15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Proceedings of the 15th Euroseminar

on
Microscopy Applied to Building Materials
(15th EMABM)

Delft, The Netherlands

June 17-19, 2015

edited by

Oğuzhan ÇOPUROĞLU
Delft University of Technology
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Published by:
Delft University of Technology
Faculty of Civil Engineering and Geosciences
Section Materials & Environment

http://www.citg.tudelft.nl/EMABM

ISBN: 978-94-6186-480-2

Credits:
LATEX support team: Fernando F. Mendonça Filho & Xu Ma
Organization support team: Lourdes Silva de Souza, Stefan Chavez Figueiredo
& Nynke Verhulst

Printed in Delft by Sieca Repro - June 2015

This work is licensed under a Creative Commons

Attribution-NonCommercial-NoDerivatives 4.0 International License.
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Sponsors
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Committees

Local Organizing Committee

Oğuzhan Çopuroğlu, chair (TU Delft)

Erik Schlangen (TU Delft)
Iris Batterham-Markus (TU Delft)
Siska Valcke (TNO)

International Scientific Committee

Matthias Böhm (VDZ)
Ole Mejlhede Jensen (Technical University of Denmark)
Karen Scrivener (EPFL)
Per Hagelia (NPRA)
Peter Laugesen (PELCON)
Timo Nijland (TNO)
Maarten Broekmans (Norwegian Geological Survey)
Paul Stutzman (NIST)
Larry Sutter (Michigan Technological University)
Karl Peterson (University of Toronto)
Jan Elsen (KU Leuven)
Ulla H. Jakobsen (Danish Technological Institute)
Isabel Fernandes (ULISBOA)
Herbert Poellmann (University of Halle)
Ted Sibbick (W.R. Grace)
Oğuzhan Çopuroğlu (TU Delft)

International Board of Reviewers

Hong Wong (Imperial College London)
Nick Winter (WHD)
Per Hagelia (NPRA)
Vincent Thiéry (Ecole des Mines de Douai)
Andreas Leemann (EMPA)
Oğuzhan Çopuroğlu (TU Delft)
Özge Andiç-Çakır (Ege University)
Isabel Fernandes (ULISBOA)
Matthias Böhm (VDZ)
Siska Valcke (TNO)
Karl Peterson (University of Toronto)
Maarten Broekmans (NGO)
Timo G. Nijland (TNO)
Erik Schlangen (TU Delft)
José Pacheco (TNO)
Jan Elsen (KU Leuven)
Ole Mejlhede Jensen (DTU)
Alison Crumbie (Egis)

1st EMABM Photomicrography Competition Jury

Maarten Broekmans (NGO)
Paul Stutzman (NIST)
Oğuzhan Çopuroğlu (TU Delft)
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Foreword
Dear Authors, Readers,

It is my pleasure to present you the proceedings book of the 15th Euroseminar on Microscopy Applied
to Building Materials (EMABM) which was held between June, 17-19 2015 in Delft, The Netherlands.
EMABM is one of the most important meeting series which brings academicians, students, engineers,
geoscientists and industry professionals together in order to share the latest developments and results
in the field of building materials characterization. EMABM covers all traditional and modern materials
characterization techniques while special emphasis is given on microscopy.
The Euroseminar community celebrates its 15th biennial meeting in the city of Delft, The Netherlands.
The local organizing committee is pleased to host the Euroseminar for the second time in Delft, after the
successful event in 1999. Furthermore, microscopy is Delft’s historic heritage as it is one of the cradle
cities of microscopy and the home town of Anthony van Leeuwenhoek; the pioneer of microscopy in the 17th
century.
The technical program and the proceedings comprise 46 peer-reviewed papers which were grouped
in 8 sessions on various aspects of building materials microscopy and characterization. Herewith, I
acknowledge the members of the scientific committee and the international board of reviewers for their
effort and contributions to ensure high level of scientific quality for the event.
One of the highlights of the 15th EMABM is the 1st EMABM Photomicrography Competition. Participants
were invited to submit their best optical and/or electron photomicrographs to the competition. Our aim
with this competition is to promote the importance of high quality photomicrography in the field of building
materials characterization. The winner is chosen by the votes of the delegates during the Euroseminar.
On behalf of the 15th EMABM local organization committee I would like to acknowledge our sponsors
BAS Research and Technology, Phenom World and Pelcon APs for their financial support and technical
contributions. A special thanks go to Paul Stutzman (NIST) for giving the opening lecture on quantitative
microscopy, to Peter Laugesen (Pelcon APs) for the workshop on concrete petrography, and to Isabel
Fernandes (ULISBOA) and Maarten Broekmans (NGO) for the workshop on alkali-reactive aggregate
characterization. Finally, all authors and delegates are gratefully acknowledged for being part of the 15th
EMABM.
I am confident that the works presented in this proceedings book will serve the building materials
microscopy community by generating inspirations for further research, as well as by becoming a useful
resource for students and professionals who would like to gain deeper understanding of microscopy related
techniques applied to building materials.

Oğuzhan ÇOPUROĞLU
15th EMABM Chair

Replica of van Leeuwenhoek

microscope (photo by J. Rouwkema)
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

E
M
Euroseminars
on Microscopy

B
A
Applied to
Building Materials

M EMABM

Previous Editions of the Euroseminars

on Microscopy Applied to Building Materials

1st EMABM 1987, Tåstrup, Denmark

2nd EMABM 1989, Wildegg, Switzerland
3rd EMABM 1991, Barcelona, Spain
4th EMABM, 1993, Visby, Sweden
5th EMABM 1995, Leuven, Belgium
6th EMABM 1997, Reykjavík, Iceland
7th EMABM 1999, Delft, the Netherlands
8th EMABM 2001, Athens, Greece
9th EMABM 2003, Trondheim, Norway
10th EMABM 2005, Paisley, United Kingdom
11th EMABM 2007, Lisbon, Portugal
12th EMABM 2009, Dortmund, Germany
13th EMABM 2011, Lubljana, Slovenia
14th EMABM 2013, Helsingør, Denmark
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

TABLE of CONTENTS
based on the conference presentations schedule

DAY ONE

Session One: Clinker and SCM characterization

Chair: O. Çopuroğlu

Quantitative microscopy applied to cementitious materials (opening keynote lecture) 13

P. Stutzman

Benefits of microscopy for raw material preparation and clinker processing 25

P. de Toit & L. Caragacean

Evaluation of portland cement clinker with optical microscopy - case studies III 33
M. Böhm & K. Lipus

A new quantification method based on SEM-EDX to assess fly ash chemical

composition and its reactivity in composite cements 39
P. T. Durdziński , C. F. Dunant, M. Ben Haha & K. L. Scrivener

Utilization of methods of optical microscopy in terms of evaluation of

cement materials (poster presentation) 45
T. Staněk & P. Sulovský

Session Two: Natural rocks and aggregates

Chair: K. Peterson

Fissuring pathology of partly metamorphosed limestone 55

V. Bams, D. Nicaise

Petrography of roofing slate for quality assessment 61

V. Cardenes, V. Cnudde, & J.P. Cnudde

Some aspects of petrography of burnt colliery spoil 71

V. Thiery, B. Guy

Nanolime deposition in maastricht limestone: back-migration or accumulation

at the absorption surface? 77
G. Borsoi, R. v. Hees, B. Lubelli, R. Veiga & A.S. Santos

3d characterization of structure and composition of sandstone: a nanometer to

centimeter-scale approach using high-end imaging techniques 87
W. De Boever, L. V. Hoorebeke, A. Diaz, M. Holler & V. Cnudde
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Session Three: Microscopy and mechanics

Chair: F. A. Veer

Investigation of coarse aggregate surface treatments to enhance concrete strength 97

A. Dehghan, K. Peterson, P. Zacarias & M. Smith

ESEM-BSE coupled with rapid nano-scratching for micro-physicochemical analysis

of marine exposed concrete 109
D. Palin , A. Thijssen, V. Wiktor, H.M. Jonkers, & E. Schlangen

Looking at glass from a different angle: new insights into fracture patterns through
transmitted light microscopy 113
O. v.d. Velde, O. Çopuroğlu & F.A. Veer

Using nano-indentation and microscopy to obtain mechanical properties 121

M. Lukovic, E. Schlangen, B. Savija, G. Ye, & O. Çopuroğlu

DAY TWO

Session Four: Historic materials / Self-healing materials

Chair: J. Elsen

Material characteristics of prefabricated concrete elements from a late 19th century

church in lower Austria 131
F. Pintér & C. Gosselin

Microscopic study on lombard mannerist stuccoes in comparison with

Vasari’s lives of the artists 139
R. Bugini & L. Folli

Hydraulicity in ancient mortars: its origin and alteration phenomena

under the microscope 147
J. Weber, A. Baragona, F. Pintér, C. Gosselin

Morphology of self-healing products in NHL and lime-pozzolan mortars 157

M. Amenta, I. Karatasios, N. Maravelaki & V. Kilikoglou

Assessment of the functionality of bacteria-based repair system for concrete

through ESEM analysis 165
V. Wiktor & H.M. Jonkers

Investigation of self healing phenomenon by calcite precipitation using reactive

transport modelling and microscopic observation 171
B. Hilloulin, F. Grondin, A. Soive & A. Loukili

Session Five: Alkali-aggregate reactions

Chair: M.A.T.M. Broekmans

Microscopic investigation of building materials affected by alkali-silica reaction 183

D. Werner, A. Gardei, S. Simon & B. Meng
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Microscopic and chemical characterisation of ASR induced by quartz-rich aggregates 191

Š. Šachlová, A. Kuchařová, Z. Pertold, & R. Přikryl

Deterioration of Santa Maria island airport concrete pavement 199

S. Medeiros, I. Fernandes & J. C. Nunes

Alkali reactive aggregates: the importance of representative samples from the

quarries at the absorption surface? 207
I. Fernandes, M.A. Ribieiro, H. Couto, & A.S. Silva

Evaluation of potential alkali reactivity of volcanic rocks from

northern Aegean region of Turkey 215
I. Fernandes & Ö. Andiç-Çakır

ASR prevention – effect of lithium on the reaction products formed in aggregates 227
A. Leemann

Alkali-silica reaction: petrographic characterization and performance tests of

Portuguese granitic aggregates (poster presentation) 235
V. Ramos, I. Fernandes, F. Noronha, A. Santos Silva, & D. Soares

Session Six: Hydration of Portland cement

Chair: M. Böhm

Adsorbance of superplasticizers on concrete additives: microstructural

in-situ experiments 249
A. Wetzel & J. Arend

The influence of sodium salts on C-S-H 255

B. Mota, T. Matschei & K. Scrivener

Dealing with uncertainty in material characterization of concrete by education 259

M. R. de Rooij & F. F. de Mendoça Filho

Fluorescence laser scanning confocal microscopy for real-time imaging of

early cement hydration 269
M.H.Yio, H.S.Wong & N.R. Buenfeld

Composition and morphology of C-S-H in C3 S pastes hydrated by water/isopropanol

mixes using electron microscopy 279
E. Boehm-Courjault, J.E. Rossen & K. Scrivener

The interaction of pH, pore solution composition and solid phase composition of
carbonated blast furnace slag cement activated with aqueous Na-MFP 287
J. Kempl & O. Çopuroğlu

DAY THREE

Session Seven: Recycled materials / Mortars & tiles

Chair: P. Hagelia
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

The ITZ in concrete with natural and recycled aggregates: study of microstructures
based on image and SEM analysis 299
G. Bonifazi, G. Capobianco, S. Serranti, M. Eggimann, E. Wagner, F. Di Maio, & S. Lotfi

Characterization of the interface between aggregate and cement paste of mortar

based on recycled aggregate by scanning electron microscopy 309
T. Le, G. Le Saout, E. Garcia-Diaz & S. Remond

Assessment of the contaminants level in recycled aggregates and alternative new

technologies for contaminants recognition and removal 319
S. Lotfi, F. Di Maio, H. Xia, S. Serranti, R. Palmieri & G. Bonifazi

Microstructure development of a drying tile mortar containing

methylhydroxy-ethylcellulose (MHEC) 333
A.P.A. Faiyas, S.J.F. Erich, T.G. Nijland, H.P. Huinink, & O.C.G. Adan

Application of acoustic microscopy technique for the assessment of the efficacy

of water repellent treatment in fired bricks 343
M. Stefanidou, G. Karagiannis, G. Apostolodis & K. Matziaris

An experimental study on the recovery of the hardened cement from crushed

end of life concrete (poster presentation) 353
S. Lotfi & P. Rem

Session Eight: Concrete microscopy / Characterization of ageing effects

Chair: E. Schlangen

New concrete types for aggressive environment; microscopic results from

a research project 363
U.H. Jakobsen & M. Kaasgaard

Visby 1993 re-visited: Understanding the effective W/C ratio, as influenced

by aggregates and air voids 373
P. Laugesen

SEM supervised image analysis contribution on understanding sulfate attack

on blended cement mortars 379
S. Meulenyzer, E. Brouard, D. Lapillonne & R. Barbarulo

Spalling of sprayed perlite concrete caused by popcorn calcite deposition and

internally derived sodium sulfate under influence of water leakage, frost action
and dynamic loads 387
P. Hagelia

Monitoring internal sulphate reactions by x-ray tomography 401

A. Darquennes, Y. Wang, F. Benboudjema & G. Nahas

Petrographic assessment of particulate materials: some statistical considerations 409

M.A.T.M. Broekmans & I. Fernandes

Semi- and full quantitative EDS microanalysis of chlorine in reinforced mortars

subjected to chloride ingress and carbonation (poster presentation) 417
F. F. de Mendonça Filho, J. Pacheco & O. Çopuroğlu
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Clinker and SCM characterization

15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Quantitative microscopy applied to

cementitious materials
Paul Stutzman
Engineering Laboratory, National Institute of Standards and Technology, USA
⇤ stutz@nist.gov

Abstract

Microscopy has played an essential part in developing our knowledge of cementitious materials, including portland
cement clinker phase composition and is routinely used to monitor kiln operating conditions to insure cement quality.
Our developing understanding of the phase composition of clinker coincides from the early days of the polarized light
microscope, used to resolve conflicting theories on the phase constituents of cement clinker. The phase abundance,
distribution and texture of cement clinker reflect the combination of proportioning, grinding, and homogenization of
the raw materials, and the firing and cooling history of the clinkering process. The ability to visualize, record and
quantify phase compositional and textural attributes of the clinker allowed cement chemists a view into this process,
to develop a better understanding of clinker production, and to be able to identify problems in the preparation and
firing of the raw materials for improved production of the clinker. Today, we use polished sections of clinker, reflected
light microscopy, and quantitative methods that range from linear traverse, point-count analysis, and full-field image
processing and analysis. The development of certified clinker reference materials have facilitated the development of
standard test methods for clinker microscopy and X-ray powder diffraction, which have provided a means to assess the
precision of these methodologies. The application of the scanning electron microscope (SEM) allows analysis of the
fine-grained, multi-phase particles of hydraulic cements and pozzolans, expanding our capabilities in mineralogical and
textural characterization. Cement phase mineralogy and textural characteristics, captured through SEM imaging, and
particle shape characteristics quantified through X-ray tomography has allowed the generation of 3-dimensional virtual
cement particles that retain the phase and textural attributes, providing realistic inputs to cement hydration models.

Keywords: Cement, Fly Ash, Image Analysis, Microscopy, Slag

I. Introduction

Portland cement was beginning to replace dimen-

sion stone as a principal construction material by
the late 19th century because of its relative ease of
use, its ability to be cast at the construction site, and
its ability to harden when placed under water. The
first use of hydraulic cement is attributed to an En-
glish engineer, John Smeaton, for the reconstruction
of the Eddystone Lighthouse at Cornwall England
in 1759 (Bogue, 1955). Up to that time, slaked lime
mortars were commonly used but their low dura-
bility and lack of integrity under water precluded
their use for many applications. Smeaton found,
Figure 1: An early thin section micrograph (1917) of
contrary to the conventional wisdom, that softer, im-
belite crystals bounded by dark-appearing tricalcium
pure limestone produced superior cement. He also
aluminate (90x, crossed Nichols) (Bates and Klein, 1917)
utilized natural pozzolanic tuff, which reacted with
the alkaline cement pore solution in his concrete
mixtures. The term portland cement was used by
Joseph Aspdin, who patented a high-temperature how they affected the performance of cement. The
cement manufacturing process, naming the product popular conception of cement and concrete manu-
after Portland limestone, a popular dimension stone facture at the time was that the firing process ex-
used in construction at that time (Bogue, 1955). pels air and water from the limestone, and mixing
While use of portland cement increased, little with water reconstitutes it after placement (Bogue,
was known about its compositional attributes and 1955). The introduction of the light microscope pro-

13
 Stutzman
vided new insight into cement and the hydration
products in hardened cement concrete. Henri Le
Chatelier made one of the first applications of the
petrographic microscope in clinker analysis in the
late 1880’s where descriptions of the constituents
were made along with sketches of the crystals from
thin sections, documenting the optical properties
of the main phases and identifying tricalcium sili-
cate as the principal constituent (Le Chatelier, 1905).
Subsequent work by A.E. Törnebohm and Bates
and Klein (Figure 1) provided a more detailed de-
scription and naming of the phases and textures of
clinker (Insley and Fréchette, 1955).
Microscopic examination is a direct means of
analysis, in contrast to the indirect Bogue calcula-
tions that transform a bulk chemical analysis into
phase estimates. While quantitative microscopy is
an ideal technique for clinker analysis in cement
manufacturing, it is not widely used today due
to the time involved in a proper analysis. It does
remain very useful as a rapid screening tool to eval-
uate clinker production through semi-quantitative
assessment of microstructural features and their
association with potential processing conditions.
The petrographic microscope was instrumental in
resolving early cement soundness problems by es-
tablishing a link with mineral constituents such as
free lime and periclase (Campbell, 1999). Quanti-
tative microscopy of clinkers, combined with the
long-time study of many cements’ performance
provided the basis for phase-related limits in the
developing codes and specifications (Brown, 1948).
Being able to characterize the mineral and textural
characteristics of cementitious materials remains
a challenge, but presents opportunities to assess
production quality and develop performance pre-
dictive models.
Quantitative microscopy began with Delesse
(1866) recognizing the relationship between the area
fraction and volume fraction. The first quantita-
tive procedures involved patiently sketching the
microstructure on paper, then tracing the grain
outlines onto a metal foil, and finally cutting and
weighing the mineral constituents represented by
the foil traces. The Glagolev-Chayes method, re-
ferred to as the point count method, is perhaps
the most widely used technique in quantitative
mineralogical analysis when using a microscope
(Chayes 1956, Campbell and Galehouse, 1991). This
methodology was adopted in ASTM C1356, "Stan-
dard Test Method for Quantitative Determination
of Phases in Portland Cement Clinker by Micro-
scopical Point-Count Procedure" (ASTM C1356).
More recently, image analysis systems have been
used to process and analyze images to quantify
microstructural features, bringing quantitative mi-
croscopy back to its origin when the microstruc-
ture was sketched, traced, cut and weighed. The
14
challenge is in getting the machine to "see" what
the microscopist sees, particularly when the differ-
ences between constituents are based upon varia-
tions in color or texture. In principal, the mineral
constituents are being digitally traced and extracted
from the bulk microstructure by a set of operations
termed image processing. Image analysis, or mea-
surement, replaces the weighing operation by pixel
counting. The potential for automation of data col-
lection and processing promises to provide for a
much greater sampling of any microstructure and
establishing estimates of microstructural features
not easily accomplished in the past.
Both light and electron microscopes have been
invaluable in developing Standard Reference Mate-
rials for cement clinker and to capture the composi-
tional and textural aspects of hydraulic cements, fly
ash, and slag. In addition to materials characteriza-
tion, these data are used in three dimensional mi-
crostructure development models of the hydration
process. The methods for specimen preparation,
microscopy, image processing and analysis will be
described using an example from the development
of the SRM 2686a clinker.
II. Specimen preparation for
microscopy
A number of sources of systematic and random
error are present in any quantitative analyses, influ-
enced by sampling, sample preparation, and spec-
imen polishing, and operator experience in point
counting and phase identification. Obtaining a rep-
resentative sample can be a challenge in a cement
Figure 2: Reflected light microscope image of a clinker
surface after 600-grit silicon carbide grinding (1) ex-
hibits no discernible microstructure due to the rough
surface. Increased polishing time (2,3,4) using 9 or 6
µm diamond paste removes grinding damage pitting,
revealing the underlying microstructural features.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: SRM 2686a exhibits a heterogeneous silicate distribution and a fine-grained, differentiated matrix.

plant given the throughput of a few hundred tons
per hour, or more. Often the objective of the study
is to specifically analyze materials that are unusual,
like fragments of material dissimilar to the clinker.
Proper specimen preparation methods facilitate ex-
amination and interpretation of microstructural fea-
tures. Improper preparation methods may obscure
features, and even create artifacts that may be eas-
ily misinterpreted. Reflected light microscopy and
scanning electron microscopy using backscattered
electron and X-ray imaging require a highly pol-
ished surface. The polished surface has two distinct
advantages to a fracture surface: 1) clear definition
of the constituent phases, and 2) a planar surface
amenable for quantitative analysis. Using a repre-
sentative sample, specimens are potted in an epoxy
resin to permeate the material’s pore system and
to encapsulate the particles. Epoxy impregnation
of the pore system serves two purposes: A) it fills
the voids and encapsulates the particles, creating
a solid that is better able to resist plucking and
spalling during the polishing process, and B) it en-
hances contrast between the pores and cementitious
material. With relatively high permeability materi-
als or powders such as clinker or portland cement,
an epoxy of low viscosity is suitable. The selec-
tion of epoxy depends upon the materials and the
means for analysis. For polished sections, the ideal
epoxy will exhibit good capillarity, wet and bond
to pore walls and edges, fill voids, and will not
leave any residue on the specimen surface that will
adversely affect etching (for light microscopy) or
backscattered electron contrast, will not soften (pro-
motes particle plucking) from exposure to cleaning
agents such as ethanol or acetone and, for SEM
analyses, is relatively beam-stable. For clinker, a

Table 1: Common phases in ordinary portland cement clinker, blast furnace slag, and fly ash, ordered per material with
decreasing brightness in SEM-BE imaging, with average atomic number Z̄ and backscattered electron coefficient h
based upon ideal compositions.

Phase Z̄ h Phase Z̄ h
ferrite 16.65 0.186 Slag
free lime 16.58 0.188 merwinite 13.71 0.157
alite 15.06 0.172 average slag gehlenite 13.36 0.153
belite 14.56 0.166 melilite 13.11 0.150
arcanite 14.41 0.165 åkermanite 12.80 0.147
aluminate-cubic 14.34 0.164 12.25 0.105
aluminate-orth. 13.87 0.159 Fly Ash
aphthitalite 13.69 0.159 quartz
anhydrite 13.42 0.154 mullite 10.80 0.125
bassanite 13.03 0.149 hematite 10.69 0.124
gypsum 12.12 0.138 magnetite 20.59 0.223
thenardite 10.77 0.125 21.02 0.227
periclase 10.41 0.121

15
 Stutzman
Figure 4: SRM 2686a BE image (upper) with 1) free lime,
2) ferrite, 3) alite, 4) belite, 5) aluminate, 6) periclase, 7)
void, and 8) alkali sulfate and (lower) X-ray spectrum
for aluminate indicates the presence of O, Na, Mg, Al,
Si, K, Ca, and Fe by their characteristic X-ray lines.
medium-viscosity epoxy 1, 2 , is used to promote
rapid infiltration of the accessible void spaces. Vac-
uum assist pulls the air out of the less accessible
void spaces and forces the epoxy into these voids.
While complete permeation may not be possible
in many cases as a result of occluded voids, an
additional step of back filling the open voids and
a second epoxy cure after sectioning may prove
useful. A higher-viscosity epoxy is used for ce-
ment, fly ash, and slag powders. This epoxy cures
harder and is a better match to the material when
polishing. The specimens are then cut or ground
to expose a fresh cross section of particles, lapped
to smooth the surface, and then polished using
1 Certain commercial materials and equipment are identified
in order to adequately specify experimental procedures. In no
case does such identification imply recommendation or endorse-
ment by the National Institute of Standards and Technology, nor
does it imply that the items are necessarily the best available for
the purpose.
2 L.R. White, hard grade (ultra-low viscosity), Epotek 301
(medium viscosity), and Epotek 353ND (powder).
16
a series of successively finer grades of diamond
paste. This polishing stage may be subdivided
into a coarse polish where the grinding damage
is removed to expose a blemish-free cross section,
and a fine-polishing stage that removes the fine
scratches that obscure the details of the microstruc-
ture. Polishing removes the damage imparted by
the grinding operations. Using a sequence of suc-
cessively finer polishing pastes composed of fine
diamond particles (ranging from 15 µm down to
0.25 µm) and a lap wheel covered with a low-relief
polishing cloth, an undisturbed microstructure is
exposed and the fine scratches of the polishing op-
eration are removed. The coarse polishing stage
eliminates the grinding damage zone, exposing
a relatively undamaged cross section, and the fi-
nal polishing removes residual scratches from the
coarse polishing stage, ideally leaving a damaged
zone that is essentially invisible for the imaging
method applied. Figure 2 illustrates the increased
clarity of a clinker microstructure as the grinding
damage is progressively removed with initial pol-
ishing stages using a 1200 grit silicon carbide paper
or 9 µm diamond paste. Only when this stage be-
gins to cut the topographically high portions off
the specimen are the cross sections of the crystals
seen. Continued polishing produces a specimen
with a minor grinding damage, evidenced by the
elongated, irregular-edged dark voids within the
clinker crystals. Additional coarse polishing time
spent removes most of the grinding damage. Final
polishing stages of 3 µm, 1 µm, and 0.25 µm pastes
remove fine scratches from the 6 µm polish, fur-
ther improving constituent definition. A thorough
cleaning is necessary to provide a surface that will
allow a uniform etch or carbon coating (for SEM
examination). The final polish must meet a number
of criteria to ensure one achieves the best possible
surface for analysis:
• Few scratches on the specimen surface,
• Sharp particle and pore perimeter edges (over-
polishing will tend to round corners),
• Well-defined phases and crystal boundaries,
• Minimal surface relief,
• No etching due to the polishing process,
• Epoxy completely fills all voids,
• No polishing media residue trapped within
voids or on the surface, and
• The surface is cleaned using ethanol or iso-
propyl, followed by an acetone rinse.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 5: Backscattered electron (BE) and X-ray image
set for SRM 2686a that will be used to segment the
sample into the constituent phases.
III. Microscopy in the development of
a standard reference material clinker
The Standard Reference Material (SRM) clinkers
are used for developing and testing methods of
quantitative phase analysis (Stutzman and Leigh,
2002, Stutzman et al., 2008). These clinkers were
selected as representative of the range of North
American clinker production with respect to phase
abundance, crystal size, and crystal distribution.
The reference values represent consensus means
and uncertainties based upon two independent an-
alytical methods: (1) quantitative XRD, (2) light
microscope point counts, and image analysis of
scanning electron microscope images.
Clinker 2686a is intermediate in crystal size rel-
ative to the other SRM clinkers, and exhibits het-
erogeneous phase distribution and little free lime
(Figure 3). Alite occurs as subhedral to anhedral
crystals approximately 30 µm in size. Belite occurs
in large clusters with an approximate crystal size of
20 µm. The matrix is differentiated with a medium-
to fine-grained lath-like ferrite, and fine-grained
aluminate filling the inter-lath voids. Equant peri-
clase crystals up to 15 µm are common throughout
the microstructure and the alkali sulfate phases aph-
thitolite and arcanite are disseminated throughout
the microstructure.
The scanning electron microscope provides sets
of images that are suitable for processing and anal-
ysis. The uniformity of the backscattered electron
and X-ray images makes it possible to perform
image processing (feature extraction) and analysis
(measurements) for quantitative microscopy. SEM
analysis is perhaps the only microscopic means to
characterize fine-grained, multi-phase particles like
cement, fly ash, and slag as grinding destroys the
crystal arrangements that are so useful for phase
identification. Light microscopy is of limited use
because of the fine particle size and the potential
of etching to partially or completely dissolve the
finer-sized particles. The multiple SEM phase and
chemical imaging modes help overcome these limi-
tations for qualitative and quantitative microscopy
(Bullard, et al. 2011, Stutzman et al. 2008, Stutzman
2007, Bentz and Stutzman 1994, Stutzman 1994, and
Scrivener, 1987).
SEM backscattered electron (BE) and X-ray (XR)
images are captured simultaneously and registered
(same field of view), presenting different perspec-
tives (phase and chemistry) that will be used in
conjunction with fabric characteristics (framework,
matrix, or dispersed phase) to identify and quan-
tify the constituent phases. For the BE image, local
brightness is proportional to the individual phase
average atomic number (Z̄). The backscatter coeffi-
cient h is a measure of the backscattered electron
fraction and, following Goldstein et al. (2003), is
estimated using the mass fractions and h values
for each constituent. Table 1 lists phases that are
found in clinker, cement, and pozzolanic mineral
additions in descending order of their backscat-
tered coefficient and gray intensity. The contrast
between alite (Z̄=15.06) and belite (Z̄=14.56) is rel-
atively strong and their distinction is clear, while
that between belite and cubic aluminate (Z̄=14.34) is
generally too weak to distinguish these constituents.
Figure 6: Selection of typical regions for each phase
establishes the training set for classification. Field width
= 750 µm.
17
 Stutzman
Figure 7: The addition of the Al and Mg images to the BE image (left) and S (right) illustrates the phase distinction
provided by the addition of the X-ray image data. Field width = 250 µm.
These values are only estimates, but are useful in
ranking relative brightness for identification. The
actual backscattered coefficient of a phase will de-
pend upon chemical substitution and the image col-
lection dwell time. Longer image collection times
may improve the distinction between phases.
Typical SEM operating conditions for cementi-
tious materials are between 10 kV and 12 kV accel-
erating voltage, about 3 nA probe current, which is
adjusted to maximize count rates while keeping a
reasonable X-ray dead time, a 5 min. per frame scan
rate (1024 x 768 pixels) to minimize image noise,
and a total scan time of about 1 h. The magnifi-
cation is adjusted to retain a 0.75 µm/pixel spatial
resolution. Changes in the accelerating voltage af-
fect both the size of the interaction volume, and
as such, the spatial resolution, and it affects the
X-ray spectral response. Practically, compromises
are necessary and the 10kV to 12 kV range rep-
resents values that provides a reasonable spatial
resolution, improved X-ray response for the lighter
elements (Na, Mg, Al), and yet is sufficient to ex-
cite the heavier elements (Fe). The X-ray spectrum
consists of the characteristic lines for each element
present, represented by peaks on the spectrum, and
a background of white radiation. X-ray microanaly-
sis may be used to identify and quantify chemical
composition of phases and can be used to generate
images of element spatial distribution. XR imaging
is necessary for distinguishing between phases that
have the same BE coefficient yet are composition-
ally distinct and for identification of phases that are
not easily detected in the BE image; periclase and
some alkali sulfates will appear dark, like voids.
Combining the XR and BE images allows a degree
of phase discrimination that is not available from
any single image (Table 2).
An example of the backscattered electron image
and X-ray spectrum for aluminate from SRM 2686a
is provided in Figure 4. While the SEM images lack
18
the color information of the light microscope im-
ages, the view is not unfamiliar to the microscopist,
and is perhaps easier to view and understand due
to the uniformity of appearance for each phase,
the flexibility with the wide range of magnifica-
tion, and the ancillary information provided by
X-ray microanalysis. As with the light microscope,
identification is made by a combination of phase
morphology, association with other constituents
(framework grain or matrix phase), BE brightness,
and chemistry. For this clinker field of view, there
is abundant alite, minor belite, a differentiated ma-
trix with abundant ferrite and little aluminate, dis-
persed periclase, a cluster of free lime, and some
alkali sulfate along void and fracture walls.
IV. Scanning electron microscopy
analysis
While the traditional point count analysis is readily
accomplished using a SEM, image processing and
analysis using the full image field is common and is
amenable to automation in both data collection and
analysis, essentially replicating the early days of
microscopy employing sketches of the fields of view
to quantify phase fractions (Insley and Fréchette,
1955).
A BE and X-ray image set from SRM clinker 2686a
is presented in Figure 5, with the backscattered elec-
tron image labeled BE, and the X-ray images labeled
by their respective elements. Image analysis will
use a subset of these images to reduce redundancies
that can confound the mineral phase segmentation
process. An example of redundancy is between the
BE and Fe image. Ferrite is the second brightest
(Table 1) phase and exhibits a unique grey level.
The Fe image duplicates this with an image with
inherently greater noise. The use of the Fe image
for cements will not add any new information, and
may confound the analysis due to the additional
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 8: Original BE Image (left) and segmented, color-coded, indexed image (right). Field width = 250 µm.

noise. to highlight the locations of the gypsum addition

Principal X-ray images for clinker and cements
include Ca, Si, Al, Mg, Fe, K, Na, and O. A visual
assessment is usually all that is needed to identify
significant images to extract the set of phases. An
example may be seen with aluminate and belite
having distinct chemical compositions, yet exhibit
a similar backscattered electron coefficient. How-
ever, belite contains appreciable Si while aluminate
contains Al, so use of one or the other of the X-
ray images will serve to distinguish these phases.
Similarly, the calcium sulfate addition (for cement)
and the alkali sulfates are usually difficult to see
given the high brightness and contrast of the BE
image. The X-ray images Ca, S, K, and Na may be
used to aid in their distinction. The mineral phase
identification process may be visualized through a
process whereby the BE and XR images are merged
into a red-green-blue composite image. A useful
combination for cement color composites is BE =
red, Mg = green and Al = blue to see the belite
– aluminate distinction, limestone, and the contri-
bution of Mg in identification of periclase. In this
case, the addition of the aluminum image makes
the aluminate appears purple compared to the dark
red of the belite, allowing their distinction. Another
useful combination is BE= red, K=green and S=blue
to the cement, and alkali sulfates within and along
cement grains.
Processing the image set serves to enhance the
distinction between the constituent phases by re-
ducing noise and unwanted signals. This may be
accomplished, as needed, by use of a median filter
to reduce noise yet retain edge definition, followed
by background subtraction to eliminate noise intro-
duced by the continuous background of the X-ray
spectrum. If difficulties are encountered in subse-
quent processing, the median filtered image may be
useful in achieving a more successful segmentation.
All these operations may be interactively applied
using most image analysis codes, and are illustrated
here using ImageJ 3 . Traditional image processing
methods use Boolean logic and mathematical op-
erators to threshold phases individually using one
or more images from a set. These operations create
binary images for each phase that are subsequently
merged into a composite image indexed by phase.
While this approach works well, it is tedious since
constituents are segmented individually and then
merged into a final image. Difficulties in reconcil-
ing the composite image are encountered with this
3 ImageJ processing and analysis code: National Institutes of
Health, http://rsbweb.nih.gov/ij/index.html

Table 2: Criteria for Phase Identification Using SEM BE and XR imaging

Phase Backscattered Electron, X-Ray, and textural characteristics

Free lime High BE, strong Ca, rounded grains
Ferrite High BE, high Fe, prismatic matrix phase
Aluminate Intermediate BE, matrix phase, high Al, low Mg
Belite Intermediate BE, rounded grains, low Al
Periclase Low BE, equant to dendritic habit, may occur anywhere, high Mg
Alkali sulfate Intermediate BE, along grain and void boundaries, high S, high K, medium Na
Gypsum Low BE, high S, high Ca, low K
Quartz Intermediate BE, high Si
Slag Intermediate BE, angular grains, high Si, Mg, and Al

19
 Stutzman

Figure 9: The box plot provides a graphical comparison of the XRD and microscopy results through assessment of the
alignment or mis-alignment of median values and differences in interquartile ranges.

process as areas of overlapping phase assignments
and holes from incomplete segmentations will need
to be resolved. In certain circumstances, however, it
can be a rapid and useful means to isolate specific
features for quantitative analysis. Multi-spectral
processing developed for analyzing hyperspectral
remote sensing data provides an alternative method
for image processing. (Van Niekerk 2003 and Lydon
2005) applied this approach for geological studies of
rocks (Landgrebe, 2003, Landgrebe and Biehl, 2011).
Combinations of BE and XR images of cementi-
tious materials are displayed to identify the con-
stituent phases, to establish a user-defined training
set of image regions typical of each phase (includ-
ing voids), and to classify each pixel into the phase
group to which it most likely belongs. The suit-
ability of the classifier and the operator-designated
training set in segmentation may be assessed by
a visual evaluation of the resulting segmented im-
age and by evaluation of the accuracy in which the
training set was properly classified. Multi-spectral
processing begins with reading the subset of BE

Figure 10: Segmentation of a polished section of hydraulic cement, following the same techniques used for clinkers,
resulting in a segmented, indexed image. Field width = 500 µm.

20
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 3: Classifier performance table indicates the accuracy of reclassifying the training set and area fractions for each
class (phase) for a single image field.
Classification of Training Fields (Resubstitution Method)
TRAINING CLASS PERFORMANCE (Resubstitution Method)
Project Reference
Class Number 1 2 3 4 5 6 7 8
Class Accuracy
Number Samples alite belite aluminate ferrite periclase als.sulf. freelime void
Name (%)
alite 1 100.0 1362 0 0 0 0 0 0 0 0
belite 2 100.0 746 0 746 0 0 0 0 0 0
aluminate 3 98.0 252 0 2 247 1 0 0 0 2
ferrite 4 96.1 233 0 0 0 224 0 0 9 0
periclase 5 100.0 644 0 0 0 0 644 0 0 0
alkali sulfate 6 99.9 1536 0 1 0 0 0 1534 0 1
free lime 7 100.0 1450 0 0 0 0 0 0 848 0
void 8 100.0 848 0 0 0 0 0 0 0 1450
TOTAL 7071

Reliability Accuracy (%) 100 99.6 100 99.6 100 100 98.9 99.8
CLASS DISTRIBUTION FOR SELECTED AREA
Class Number Samples Percent
1 alite 116,731 60.8
2 belite 12,084 6.3
3 aluminate 4,294 2.2
4 ferrite 11,639 6.1
5 periclase 6,670 3.5
6 alkali sulfate 6,181 3.2
7 free lime 7,492 3.9
8 void 26,909 14.0
Total 192,000 100.0

and XR images into a suitable image processing
software application. Different combinations are
displayed for phase identification and establishing
the user-defined training set (Figure 6). By merging
individual images into a red-green-blue channel
pseudo-colored image, the constituent phases may
be highlighted and identified (Figure 7). Regions
of each phase (including void space) are defined
with a training set, with at least 100 pixels each
to define each phase. The next step is to select a
classifier and group like pixels and review the per-
formance of that classifier on the training set. The
classifier extrapolates the training set characteristics
for each phase for the entire image to complete the
classification.
V. Image processing
Reviewing the performance of the classifier on the
designated training field pixels provides an initial
sense of the success of the segmentation (Table 3)
where a successful classification has high reference
accuracies for each phase. The miss-classifications
in the matrix to the right provides clues to the na-
ture of any miss-classifications. The addition of a
larger number of training pixels often will reduce
the classification error through an improved def-
inition of the range of class attributes. The best
assessment of the segmentation will be a compar-
ison of the segmented image to the original BE
image. Figure 8 shows the original BE and a color-
coded, segmented image. The resulting image is
typically displays each phase uniquely identified
by a color and an index value. Since clinker is a
heterogeneous product, replicate data on different
arbitrarily selected fields of view are required to
allow an estimate of the uncertainty (Table 4).
VI. Comparison to quantitative X-ray
powder diffraction measurements
Box plots provide a useful comparison between the
independent data sets from SEM (n = 20) and XRD
(n = 33) for clinker composition. The width of each
box is proportional to sample size, the interquartile
range ("middle half") of the data are represented by
the body of the box, the median value is marked by
the X within the box, the extremes are represented
by the ends of the straight lines projecting out of the
box, and circles outside the extremes of each box
represent outliers. Recognizing that each method
has a set of unique biases, results for each method

Table 4: Phase abundance results from SEM / image analyses on n=20 specimens, expressed as mass fractions.

alite belite aluminate ferrite periclase alkali sulfate

Average (SEM) 64.2 19.7 2.2 10.1 3.2 0.7
1-s 0.5 0.7 0.1 0.2 0.1 0.1
Certificate 63.35 18.68 2.46 10.76 3.40 0.87
Uc (95% coverage) 1.29 1.42 0.67 1.44 0.40 0.27

21
 Stutzman
Figure 11: Virtual cement particle created by combin-
ing the results from analysis of SEM image data and
X-ray computed tomography. Reconstruction of an ac-
tual cement particle is used in three-dimensional cement
hydration modeling.
(assessed by different operators) appear quite close.
The microscopy appears to slightly over-estimate
the silicates relative to the XRD. Certified values (Ta-
ble 4) represent a consensus between these unique
methods by quantifying and including in the over-
all estimate of uncertainty the systematic biases
of the individual methods by using the variation
across the methods’ results.
VII. Application to cements, fly ash,
and slag
The SEM provides the means to quantitatively de-
scribe fine-grained multi-phase powders in a way
not possible using light microscopy. Imaging ce-
ments, fly ash, slag and other supplementary ce-
mentitious materials poses a more complicated
Figure 12: SEM BEI (left) and X-Ray image of SRM 2691 fly ash composed of combined images of Ca (red), Si (green),
Al (blue), and Mg (gray) illustrates its chemical and compositional complexity. Field Width = 300µm.
22
problem with material preparation, image inter-
pretation and analysis. Additional phases, such as
calcium sulfates, and pozzolanic additions, such
as limestone, fly ash, or slag, must be considered.
Figure 10 shows the original BE image of a pol-
ished cement section and the resulting segmented
image. From the segmented, indexed image, mea-
surements of area fraction, surface perimeter frac-
tion, and spatial distribution may be made for the
constituent phases (Bentz, et al. 1999, Bullard et
al., 2011). Taking these data and combining them
with X-ray computed tomography images of real
cement particles has enabled the generation of vir-
tual cement particles (Figure 11) with the phase and
textural characteristics of actual industrial cements,
which has been invaluable in the development of
virtual cement hydration models (Bullard et al.,
2011). SEM BE and XR images of fly ash and slag
are shown in Figures 12 and 13. Fly ash is a com-
plex mixture predominantly of glassy phases and
some crystalline constituents (Chancy et al. 2010).
If sufficiently coarse, the mineral constituents can
be imaged and measured. The most common min-
eral, quartz, may be seen in Figure 11 as regions
of high Si, intermediate BE, and lacking signal in
any of the other X-ray images. The slag in Figure
12 is almost all glass, however subtle differences in
the X-ray images reflect two slightly different glass
compositions.
VIII. Summary
Microscopy has played a significant role in devel-
oping our understanding of cementitious materials
compositions and their effects on cement and con-
crete performance. It continues to play an impor-
tant role in the evaluation of cement clinker kiln
operations and as a quantitative tool for assessing
phase compositions. Point-counting for quantita-
tive phase abundance is a mature method, yet is
one of the few direct methods to determine clinker
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 13: SEM BEI (top) and X-Ray image of slag
composed of combined images of Ca (red), Si (green), and
Al (blue) illustrates its relative uniformity, but subtle
chemical differences. Field Width = 300µm.
phase compositions. The application of the scan-
ning electron microscope with X-ray microanalysis
complements light microscopy by not only provid-
ing analyses of clinker, but also of the more difficult
fine-grained powders of portland cement and poz-
zolans like fly ash, and slag. While these images
may also be point-counted, image processing and
analysis provides a means for full-field quantita-
tive measurements on area fraction and surface
perimeter fraction, and spatial distribution for the
constituent phases. These data, coupled with X-ray
computed tomography, are providing the means
to generate three-dimensional particles that cap-
ture the characteristics of phase abundance and
texture and are invaluable in the development of
three-dimensional computer simulation models of
cement hydration.
References
ASTM C1356 (2011) Standard Test Method for Quantitative Determi-
nation of Phases in Portland Cement Clinker by Microscopical Point-
Count Procedure American Society for Testing and Materials West Con-
shocken, PA, Annual Book of ASTM Standards, (4.01): Cement
Bates, PH and Klein AA (1917) Properties of the calcium silicates and
calcium aluminates occurring in normal portland cement. Technological
Papers of the Bureau of Standards, 78:1-38
Bentz, DP, Stutzman PE, Haecker CJ, Remond S (1999) SEM/X-ray
imaging of cement-based materials. Procs 7th Euroseminar on Microscopy
Applied to Building Materials (EMABM): 457-466.
Bentz DP, Stutzman PE (1994) SEM analysis and computer modeling
of hydration of Portland cement particles. Petrography of Cementitious
Materials, American Society for Testing and Materials, ASTM STP 1215:60-
73.
Bogue RH (1955) The Chemistry of Portland Cement. 2nd ed., Rein-
hold, New York
Brown LS (1948) Microscopical Study of Clinkers. In Long-time study
of cement performance in concrete, PCA Bull 26: 877-933
Bullard JW, Lothenbach B, Stutzman PE, Snyder KA, (2011) Coupling
thermodynamics and digital image models to simulate hydration and mi-
crostructure development of portland cement pastes. Jnl Mat Res 26:609-
626
Campbell DH (1999) Microscopical Examination and Interpretation
of Portland Cement and Clinker, 2nd ed., Portland Cement Association,
Skokie/IL
Campbell DH, Galehouse JS (1991) Quantitative clinker microscopy
with the light microscope. Cem Concr & Aggr 13/2:94-96
Chancy RT, Stutzman P, Juenger MCG, Fowler DW (2010) Compre-
hensive phase characterization of a class F fly ash, Cem & Conc Res.
40:146-156
Chayes F (1956) Petrographic Modal Analysis. An Elementary Statis-
tical Appraisal. Wiley, New York
Delesse M.A. (1866): Procédé mecanique pour determiner la compo-
sition des roches. Annales des Mines 13, 4th series. pp. 379-388.
Goldstein J, Newbury DE, Joy DC, Lyman CE, Echlin P, Lifshin E,
Sawyer L, Michael JR, (2003) Scanning Electron Microscopy and X-Ray
Microanalysis. A Text for Biologists, Materials Scientists, and Geologists,
3rd edition. Springer Verlag, Berlin
Insley H, Fréchette V (1955) Microscopy of Ceramics and Cements,
Chapter 5, Special techniques, pp. 177 – 207, Academic Press, New York
Landgrebe D, Biehl L (2011) An Introduction and Reference for
Multispec,” https://engineering.purdue.edu/ biehl/MultiSpec/, last ac-
cessed March 4, 2015.
Landgrebe D (2003) Signal Theory Methods in Multispectral Remote
Sensing. Wiley Interscience, New York
Le Chatelier H (1905) Experimental researches on the constitution of
hydraulic mortars. English translation by JL Mack, McGraw, New York
Lydon JW (2005) The measurement of the modal mineralogy of rocks
from SEM imagery: the use of Multispec c and ImageJ freeware. Geol
Surv Canada Open File 4941:1-37
Scrivener KL (1987) The microstructure of anhydrous cement and its
effect on hydration. Mat Res Soc, Symp Proc 85:39-46
Stutzman P (2007) Multi-spectral SEM imaging of cementitious mate-
rials. Proc. 29th Int Conf Cement Micr, Québec City, Canada
Stutzman PE (1994) Scanning electron microscopy imaging of hy-
draulic cement microstructure. Cem Concr Comp 26/8:957-966
Stutzman PE, Leigh S (2002) Phase composition analysis of the NIST
reference clinkers by optical microscopy and X-ray powder diffraction.
NIST Technical Note 1441:1-44
Stutzman PE, Lespinasse G, Leigh S (2008) Compositional analysis
and certification of NIST reference material 2686a. NIST Tech Note 1602:1-
49
Van Niekerk D, (2003) Modal analysis and phase iden-
tification in meteorite thin sections using freeware for
PC. Annual Lunar and Planetary Sci Conf, League City/TX
http://www.lpi.usra.edu/meetings/lpsc2003/pdf/2015.pdf last
accessed March 4, 2015.
23
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

24
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Benefits of microscopy for raw material

preparation and clinker processing
Pieter du Toit⇤ and Lilia Caragacean
Lafarge IPC
⇤ pieter.dutoit@lafarge.com

Abstract

This paper describes the use of microscopic techinques to improve the quality of clinker through process interventions
during the manufacturing process. Microscopy is a powerful technique as it can follow the production process from
the quarry, through the milling and finally the clinker formation stages. It is essentially a historical record of the link
between the raw materials entering the kiln, the actual process conditions in the kiln and the final product, the clinker.
For the optimum clinkering process to take place, both chemical and mineralogical parameters need to be satisfied.
Burnability is the term commonly used to describe this measurement and calculates the amount of theoretical free lime
remaining in the clinker. It is here where the microscope is used with great effect to determine the characteristics of
the raw meal. Transmitted light microscopy is used mainly for diagnostic purposes on the raw materials and the kiln
feed although can be used to examine the clinker, but this is usually limited to specialist applications. Reflected light
microscopy is used mainly for clinker examination, which is carried out on an epoxy impregnated sample of clinker
treated with various stains and etches to enhance the clinker phases (alite, belite, aluminate, ferrite and periclase). The
microscopic evaluation of these phases (size, distribution and morphology) indicates the process conditions in the kiln.
Changes are made in the preparation of the raw meal (chemical and mineralogical) and process conditions (flame shape,
kiln speed/feed, temperature profile) to optimize the clinker formation.

Keywords: microscopy, raw materials, clinker microstructures, process conditions

I. Introduction

Microscopy of clinker can be traced back as far as

1887 with the work done by Le Chatelier (Camp-
bell, 1999), based on earlier examination of rocks
by the English geologist H.C. Sorby. The naming
of the clinker mineral phases can be attributed to
Törnebohm in 1897. Today both transmitted and re-
flected light are used to explain the features of pow-
der mounts and polished sections of raw mateials
and clinker. Figure 1 shows a typical transmit-
ted/reflected light microscope which is used for
the examination of both powder mounts and pol-
ished sections.

II. Transmitted light microscopy

The transmitted light microscope is used mainly Figure 1: A typical microscope used for both transmitted
for the examination of raw materials used in the and reflected light examination.
preparation of the raw meal and kiln feed. Materi-
als examined include thin-sections of rocks (petro-
graphic examination), powder mounts of the raw
mill product or kiln feed and powder mounts of
clinker.

25
 du Toit & Caragacean
Figure 2: Transmitted light image of a limestone contain-
ing calcite fossil shells.
Petrographic examination – transmitted light
A petrographic analysis of the rock types used in
clinker manufacture (e.g. limestone) will reveal the
suitability or not of these rock types for clinker
prodution. The major component for clinker man-
ufacture is the limestone which occurs in several
different forms. Pure limestone is best suited to
the process and typically contains more than 85%
CaCO3 . Figure 2 shows a transmitted light image
(Campbell, 1999) of a thin-section of limestone used
in the manufacturing process. It is possible to use
‘impure limestone’ for clinker production but then
the following mineral phases need to be taken in
account when designing a raw mix for the cement
plant.
Pyrite is found in many limestones and will form
SO2 once in contact with the hot gasses and lead to
emissions from the stack. Authigenic quartz crys-
tals which are found as rims around calcite grains
can lead to difficult grinding in the raw mills and
result in quartz grains which cannot be ground fine
enough and therefore influence the chemical reac-
tions in the kiln. Dolomite contains MgO which
forms periclase in the clinker and may cause expan-
sion in the concrete after several years. Argillaceous
limestones contain varying amounts of clay and silt
which need to be carefully controlled if used as
constituents in the process – the amount of alu-
mina and iron need to be balanced in the raw mix
composition.
Flint and chert (impure crypto-crystalline silica)
as shown in Figure 3 (Campbell, 1999), are found
as concretionary minerals mainly in chalk deposits
and need to be separated before the chalk can be
used in the manufacturing process.
Powder mount examination – transmitted light
The transmitted light microscope is also used to
examine powders. The main application is the ex-
26
Figure 3: Thin-section showing chert particle in raw
meal.
amination of the raw mill product (raw meal) and
in particular specific size fractions of limestone and
quartz.
The production of raw meal must conform to
both chemical and physical constraints: chemical
properties with respect to the accurate proportion-
ing of raw materials to give the desired phase com-
position in the clinker (C3 S, C2 S, C3 A & C4 AF); and
physical properties with respect to the fineness so
that the minerals have sufficient time to transform
in the kiln burning process.
Thermodynamically, there is a limited amount of
time for these minerals to react to form the clinker
phases and if they are particularly large, the clinker
minerals will be unable to form in the correct pro-
portions. A large amount of the limestone particles
(CaCO3 ) greater than 125µm will lead to Free CaO
in the clinker resulting in irregular setting time
of cement. On the other hand, a large amount of
quartz particles greater than 45µm will result in
excess C2 S (belite) and insufficient C3 S (alite), low-
ering the clinker reactivity. Figure 4 (Campbell,
1999) shows particles of quartz >45µm in the raw
meal which will adversely affect the formation of
clinker phases during the burning process.
Combinability formulas
Results from the examination of powder mounts of
raw meal are used to determined the combinabil-
ity of the meal and has been the subject of many
researchers (Miller, 1981; Campbell, 1999; Theisen,
1992; and Hills, 2002, 2004), who have developed
methods and formulae to evaluate the ‘theoretical
combinability’. The combinability of the raw mix
is determined by both chemical and mineralogi-
cal factors, principally the Lime Saturation Factor
(LSF) and Silica Ratio (SR), and the coarse particles
of lime, quartz and other insoluble minerals. The
formula of Theisen (1992) is presented below. The
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Clinker powder mount examination – transmit-

ted light

In the last instance, transmitted light can also be

used to examine a powder mount of the clinker. The
preparation, examination and set-up of the micro-
scope are different from the previous two methods.
Powders are immersed in a refractive index liquid
and the microscope is equipped with a Senarmont
compensator and rotating analyzer. The clinker
powder sample is also subjected to various chemi-
cal treatments to enhance certain phases e.g. treat-
ment with KOH-sugar or Salicylic acid/Methanol
solutions.
The purpose of transmitted light examination of
Figure 4: Powder mount of +45µm acid insoluble residue clinker is to determine optical parameters for the
showing coarse quartz particle. interpretation of burning conditions according to
the Ono Method (1995). Interpretation of the pro-
cess conditions - heating rate (from the alite size),
work done by Fundal (1999) shows the relevance of cooling rate (belite colour), maximum temperature
the maximum particle size for calcite and quartz to (alite birefringence) and burning time (belite size)
the burnability function. can be inferred. This technique is usually limited
to specialist applications and carried out by expert
microscopists. Figure 5 Left image shows the trans-
FCaO (1400 C) = 0.343 ⇥ ( LSF
mitted light photo of an alite crystal of the typical
93) + 2.74 ⇥ (SR 2.3) + 0.1 ⇥
size and orientation to measure the birefringence
(C+125µm + 0.83 ⇥ ( Q+45µm + 0.39 ⇥
and in the Right image a belite crystal showing a
( Aq+45µm ) (1)
light amber colour (typical of slow cooling).

Briefly the raw meal sample is subjected to chem-
ical dissolution in 20% HCl and the amount of
CaCO3 and insoluble residue (siliceous and other)
is calculated in the 125µm and 45µm size fraction
respectively. LSF and SR are determined by tradi-
tional XRF analysis. The results are entered into
the formula and the Free CaO contribution of each
component is calculated. Table 1 gives an example
of typical results from the XRF analysis, sieving and
chemical dissolution. Table 2 shows the potential
Free CaO contribution from each component.
It is clearly demonstrated that the higher amount
of coarse quartz (7.98% >45µm) in Sample A con-
tributes to a higher potential Free CaO (6,63%) in
the combinability formula. The industrial burning
of this raw mix (Sample A) will lead to a higher
fuel consumption and poorer clinker quality.
III. Reflected light for clinker
microscopy
The reflected light mode is used for the examina-
tion of polished surfaces of clinker and is the most
well established method for examining clinker. Mi-
croscopy can be used as a tool to look both into the
past and the present - to reflect what has happened
and detect reasons for poor clinker quality and thus
poor cement strength, or it can be used on-line to
improve the process conditions in real time. It is
essentially a historical record of the link between
the raw materials entering the kiln, the actual pro-
cess conditions in the kiln and the final product,
the clinker.
For on-line control, clinker samples are usually
taken every 2-4 hours depending on the prepara-

Table 2: Potential Free CaO from each component.

Table 1: Results of XRF, sieving and chemical dissolu-
tion. Sample A Sample B
LSF -0.83 0.91
Sample A Sample B SR 0.74 0.30
LSF 90.57 95.65 C+125µm , (%) 0.27 0.28
SR 2.57 2.41 Qtz+45µm , (%) 6.63 4.06
CaCO3 >125µm, (%) 2,65 2.83 Other +45µm, (%) 1.36 1.20
Qtz >45µm, (%) 7.98 4.88 Total FCaO , (%) 8.16 6.75
Other >µm, (%) 3.49 3.08 The Free CaO contribution from each component is written as e.g. LSF.

27
 du Toit & Caragacean
Figure 5: Transmitted light images of alite (left) and belite (right) prepared for optical measurement.
tion techniques. Quick set epoxy resins and high
speed polishing equipment is required to get results
timeously.
For other investigations, a more meticulous rou-
tine can be used to prepare highly polished surfaces
to examine finer details within the clinker. Standard
metallurgical equipment is used to polish the sur-
face of the clinker impregnated mount. Polishing
media include fine diamond or alumina particles
in a suspension. Care must be taken not to use any
water based lubricants as the clinker phases will
start to react.
Once an epoxy mount has been prepared, a num-
ber of stains or etches are used to enhance the
clinker phases to facilitate the microscopic exami-
nation. The clinker phases can be separated into
two groups – the silicate phases C3 S and C2 S, and
the matrix or interstitial phases C3 A and C4 AF –
each with specific chemical solutions for staining or
etching. Additional minor phases which may need
to be identified are periclase (MgO), Free CaO and
alkali-sulphates. For the etching media, care must
be taken to use analytical grade solutions as impuri-
ties may introduce artifacts and misinterpretations
can be made. A small eye-dropper or pipette is
used to cover the surface of the clinker with the so-
Figure 6: Left and Right images showing reflected light micrographs of good quality clinker.
28
lution for the required time and then the iso-propyl
alcohol is used to wash the etching solution from
the surface. A hair-dryer is used to dry the surface
before microscopic examination.
For the silicate group, the most common etching
medium is Nital, a solution of 1,5 ml nitric acid
(HNO3 ) in 100 ml iso-propyl alcohol. Nital reacts
quickly (6-10 seconds) with alite and belite. Alite
normally turns blue to green and belite blue to
brown. Other etching media for the silicates in-
clude ammonium chloride, and hydrofluoric acid
vapour. The latter is hazardous and contact with the
HF liquid or vapour must be avoided. The etched
sample needs to stand in a fume cupboard so that
excess HF vapour leaves the surface, otherwise the
microscope lenses can be damaged.
For the aluminate (C3 A) and Free CaO, a potas-
sium hydroxide (0,1 molar aqueous) solution is the
most common. After 30 seconds, aluminate turns
blue-brown and Free CaO brown. Warm distilled
water (40 C for 5-10 seconds) is a good stain to
show Free CaO (multi-coloured) without affecting
the other phases. The ferrite phase (C4 AF) is not
affected by any of the etching media and remains
brightly reflective (white to off-white).
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 7: Left image of alite crystals and Right image of belite crystals.
IV. Interpretation of clinker
micro-structures
The microscopic evaluation of the clinker phases
(size, distribution and morphology) indicates the
preparation and raw meal and the process con-
ditions in the kiln. Changes are made to the
preparation (chemical and mineralogical as de-
scribed earlier) and to the process (flame shape,
kiln speed/feed, temperature profile) to optimize
the clinker formation. The micrographs in the fol-
lowing section show features which include well
prepared raw meal and stable process conditions –
and therefore a good clinker phase composition and
structure. A good clinker microstructure is char-
acterized by well formed hexagonal alite crystals
with clean edges (35-45µm in size), well-rounded
belite showing smooth (not ragged) edges (15-25µm
in size), and a very fine cryto-crystalline matrix of
aluminate and ferrite.
Further micrographs show some instances where
either preparation of raw meal or kiln process condi-
tions have not produced clinker of optimum quality.
The microstructures seen in the polished specimens
give an indication of which changes are necessary
Figure 8: Left image shows typical dendritic belite crystals and Right image the large alite crystals found in oil-well
clinkers.
to bring the clinker quality back to the required
level.
There is more than one type of clinker and not
all have the same raw meal chemistry, mineralogy,
mill-preparation and process conditions. For ex-
ample, cement for oil-well applications has specific
requirements - usually these have a higher alite
content and lower aluminate content. Process con-
ditions also have to change to ensure a different
phase composition with respect to size of crystals.
Typical clinker microstucture
The micrographs below show clinker microstruc-
tures characteristic of well prepared raw meal and
stable process conditions. Figure 6 Left image
shows well formed (euhedral) alite crystals, with
rounded belite crystals in the left upper corner. Fig-
ure 6 Right image shows large euhedral alite and
smaller rounded belite in the matrix.
Clinker microstructure – alite and belite
The micrographs show typical shape and structure
of alite and belite. Alite crystals are angular hexag-
onal with sharp boundaries and belite are rounded
29
 du Toit & Caragacean
Figure 9: Left image shows periclase crystals and Right image a Free CaO cluster.
and show a variety of lamellae. Figure 7 Left im-
age shows alite crystals and Figure 7 Right image
shows belite crystals.
Clinker microstructure – oil well clinker
The manufacture of oil well clinker demands spe-
cial raw meal preparation, different chemistry as
well as different process conditions in the kiln. The
objective is to make the clinker less reactive, there-
fore the alite crystals are much larger than normal
and the belite should be ‘broken-up’ in the matrix.
Overall this raw meal is coarser than normal and
requires higher burning temperatures in the kiln to
combine the raw meal to the correct clinker phase
morphology. Figure 8 Left and Right shows images
of the typical microstructure of an oil well clinker.
Clinker microstructure – Periclase and calcite par-
ticles
When the amount of MgO is high (>3%) in the raw
meal and there is slow cooling at the end of the kiln,
periclase may form as small euhedral crystals in
the clinker. Similarly, if there are coarse limestone
particles, then residual Free CaO may found in the
clinker. Figure 9 Left image shows periclase crys-
Figure 10: Left image shows belite crystals with ragged edges and the Right image shows large belite nests.
30
tals and Right image shows Free CaO clusters. Both
these phases may be detrimental to the clinker per-
formance. Periclase (MgO) may hydrate to brucite
(Mg(OH )2 ) with an expansive force which causes
cracking in concrete at later ages. Free CaO may
cause plaster popping when it hydrates after addi-
tion of water the cement.
Clinker microstructure – Slow cooling and belite
nesting
The material clinkering temperature is around
1450 C in the hottest part of the kiln and after that
the clinker moves under the flame and drops into
the cooler for further cooling and transport to the
storage silo. If the flame shape is not set-up cor-
rectly, a ’slow’ cooling zone in the kiln (behind the
flame) may result in poor belite formation which is
not as reactive in the cement product. A change to
a short and sharp flame shape would change the
position of the cooling zone in the kiln and improve
the belite shape and reactivity. If the raw meal
contains a large amount of coarse siliceous parti-
cles, then belite nesting will result. The siliceous
particles cannot combine with the CaO to form the
required amount of alite, resulting in excessive be-
lite occurring as nests and clusters in the clinker.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 11: Left image shows development of large aluminate crystals and the Right image shows belite fringes on the
alite crystals.

Belite clusters are known to be hard to grind in the around alite crystals.
cement mill and lead to higher energy consumption.
Both these feature are easily seen in the clinker. Fig- Clinker microstructure – Impact of alternative fu-
ure 10 Left image shows belite crystals with ragged els
edges and the Right image shows large belite nests.
There is an ever increasing demand to use alterna-
tive fuels in the manufacturing process to replace
Clinker microstructure – Reducing conditions
standard fuels viz. coal, petcoke oil and gas. These
The clinker is normally produced in an oxidizing fuels can be liquid or solid and include the fol-
atmosphere in the kiln. However in certain cir- lowing; whole tyres, tyre chips, rice husks, paint
cumstances there may be a reducing atmosphere residues, oil sands, sewage pellets, plastics and
and this has an influence on the formation of the many more. These fuels can be fired at both the
P. du Toit and L. Caragacean
clinker phases. Reducing conditions may result back end (pre-calciner) and at the main burner.
from poor fuel preparation (coarse particles), miss- They often contain high amounts of ash consist-
aligned flame (impinging on the clinker bed) or ing of SiO2 , Al2 O3 , K2 O and Na2 O which has to be
lack of oxygen at the burner, and manifests
Clinker microstructure – Impact of alternative fuels itself taken into account in the raw mix design. These ox-
in changes in the aluminate, ferrite, belite and alite.
There is an ever increasing demand to use alternative ides also
fuelsimpact
in the the clinker phase
manufacturing formation
process to replaceand in
Largestandard
dark coloured aluminate
fuels viz. coal, petcoke crystals and
oil and gas. over-
These fuels canparticular thesolid
be liquid or belite
and and aluminate
include phases.
the following; A high
whole
tyres,oftyre
growths chips,on
belite rice husks,
the alitepaint residues,
crystals areoil sands, sewage
typical pellets,
alkali levelplastics
in theand many
fuel canmore.
lead toThese
the fuels can
formation of
be fired at both
microstructures the Extreme
seen. back end (pre-calciner)
conditions may and atre- the main burner.
alkali They often
aluminate whichcontain high
affects theamounts of ash
initial rheological
consisting of SiO2, Al2O3, K2O and Na2O which has to be taken into account in the raw mix design. These
sults in the Fe2 O3 being reduced to Fe and seen as behavior in cement. Alkali can also stabilize the be-
oxides also impact the clinker phase formation and in particular the belite and aluminate phases. A high alkali
brightlevel
spots in the clinker. The impact of reducing
in the fuel can lead to the formation of alkali aluminate lite (alkali belite)the
which affects which prevents
initial its further
rheological behaviorreaction
in
conditions on cement rheology can be quite severe. to alite. Figure 12 Left image shows
cement. Alkali can also stabilize the belite (alkali belite) which prevents its further reaction to alite. Figure long ‘bladed’
12
Figure 11 image
Left Left image
shows shows large aluminate
long ‘bladed’ crystals
alkali aluminate crystals andalkali
the aluminate
Right imagecrystals and the development
shows extensive Right imageofshows
alkali-belite.
and Right image (arrowed) shows the belite fringes extensive development of alkali-belite.

Figure 12 Left image shows development of alkali crystals and the Right image shows a large cluster
Figure 12: Left image shows development of alkali crystals
of alkali and thebelits.
stabilized Right image shows a large cluster of alkali stabilized
belites.
Conclusions
This paper demonstrates the use of both transmitted and reflected light microscopy to identify salient
feature in powder mounts and polished specimens. This information is used firstly to improve the material 31
preparation of the raw meal with respect to chemistry, mineralogy and fineness. Secondly, the microstructures
seen in the clinker indicate which process parameters need to be adjusted to optimize the clinker phase formation
 du Toit & Caragacean

V. Conclusion
This paper demonstrates the use of both transmit-
ted and reflected light microscopy to identify salient
feature in powder mounts and polished specimens.
This information is used firstly to improve the ma-
terial preparation of the raw meal with respect to
chemistry, mineralogy and fineness. Secondly, the
microstructures seen in the clinker indicate which
process parameters need to be adjusted to optimize
the clinker phase formation and overall quality of
the clinker.

References
Campbell, D.H.(1999): Microscopical Examination and Interpretation of
Portland Cement and Clinker, Portland Cement Association, Skokie, Illi-
nois, 2nd Edition, 201p.
Törnebohm, H.E (1897): The Petrography of Portland Cement,
Tonidustrie Zeitung, Vol 21, pp. 1148-1150.
Miller, F.M. (1981): “Microscopy as an Aid in Evaluation of Mix Burn-
ability and Clinker Formation”. Proceedings of the Third International Con-
ference on Cement Microscopy, International Cement Microscopic Associa-
tion, Houston, Texas, pp. 181-192.
Theisen, K. (1992): “The influence of Raw Mix Burnability on the
Resulting Cement Clinker”. Proceedings of the 14th ICMA Conference on
Cement Microscopy, California, USA, pp74-78.
Hills, L.M. (2002): Burning the mix. International Cement Review,
September, pp. 79-84.
Hills, L.M. (2004): The value of microscopy. International Cement Re-
view, April, pp. 110-114.
Fundal, E. (1979): The Burnability of Cement raw Mixes, F.L.Smidth,
Review-22, F.L. Smidth Laboratories,Copenhagen,Denmark, 6pp.
Ono, Y. (1995): Ono’s Method. Fundamental Microscopy of Portland Ce-
ment Clinker. Chichibu Onoda Cement Corporation, 229p.

32
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Evaluation of portland cement clinker

with optical microscopy - case studies III
Matthias Böhm⇤ , Klaus Lipus
VDZ gGmbH, Research Institute of the Cement Industry, Düsseldorf, Germany.
⇤ Matthias.Boehm@vdz-online.de

Abstract

Clinker microscopy is a powerful tool for the evaluation of clinker and cement properties. Microstructural inves-
tigations yield important information on phase distribution and the conditions of the phase formation. The correct
understanding of clinker microstructure is crucial for an accurate evaluation of raw material, fuel or process parame-
ters. Two case studies on clinker samples are presented, in which individual granules with unusual microstructure
characteristics were observed.
In the first case study, the clinker contained a granule the centre of which mainly consisted of belite crystals in direct
contact with free lime crystals. Additionally several alite crystals that had partially decomposed to belite and free lime
from the rim inwards were observed in the granule centre. EDX measurements, element mapping and the interpretation
of the microstructure showed that an accumulation of barium, probably introduced into the kiln as baryte, was the cause
of the observed local phenomena.
In the second case study, granules with domains containing elemental iron were observed. Additionally these domains
contained alite, high amounts of C3 A, no belite and no C4 AF. The iron was intergrown with free lime. Estimations
based upon the surface area ratio of the phases in the intergrowths show that C4 AF is probably the compound from
which the intergrowths formed after reduction of ferric iron to elemental iron.

Keywords: portland cement clinker, microstructure, case study, burning conditions

I. Introduction Selected results of two microscopic investigations

The use of alternative fuels and raw materials (AFR)
for the production of Portland cement clinker has
gained more and more importance over the last
decades and this trend continues. These materials
contribute materially to the formation of the clinker
phases and help to save fossil fuels and natural raw
materials. The use of AFR can influence the clinker
properties. Many of the effects can be observed in
the clinker microstructure.
Therefore clinker microscopy is also gaining im-
portance as an analytical method, withstanding the
trend to automatable, quantitative methods like X-
ray fluorescence (XRF) or X-ray diffraction (XRD).
Microscopy can provide information on the phase
distribution and the conditions of the phase forma-
tion, which are important for the evaluation not
only of the effects of AFR, but also of fossil fuels
and natural raw materials.
The correct understanding of the clinker mi-
crostructure is crucial for an accurate evaluation
of raw material, fuel or process parameters. This
requires extensive experience from the microscopist.
Case studies on rare or previously undescribed fea-
tures help microscopists to broaden their experience
backgound.
on clinker samples are presented here, performed
for cement plants for the evaluation of burning
conditions, especially for the confirmation of the
absence or presence of signs for reducing burning
conditions. In the respective clinker samples un-
usual microstructural characteristics were observed
and interpreted, partly with additional information
from scanning electron microscopy.
II. Sample preparation and analysis
For the examination of the clinker samples with
optical microscopy, representative subsamples
with a grain size of 2 - 4 mm were obtained by
crushing the clinker sample in a jaw crusher and
sieving the crushed material. The subsamples
were embedded in epoxy resin under vacuum.
After curing, polished sections of the embedded
samples were produced. The polished sections
were etched with a 10% KOH solution as well as
an alcoholic dimethyl ammonium citrate (DAC)
solution for several seconds, respectively, and
then investigated with an optical microscope
(Zeiss Axioplan) under reflected light. The
etching procedure enables the distinction of the
different clinker phases (alite/C3 S/Ca3 SiO5 ;

33
 Böhm & Lipus
belite/C2 S/Ca2 SiO4 ; C3 A/Ca3 Al2 O6 ;
brownmillerite/C4 AF/Ca2 (Al,Fe)O5 ; free
lime/CaO) under the microscope. While the
brownmillerite (C4 AF) is recognizable due to
its strong reflectivity without etching, the other
three main clinker phases look very similar
under reflected light. The KOH solution causes a
discolouration of the C3 A from a light grey to a
darker grey or brown. The DAC solution etches
the surface of alite crystals which produces an
apparent sharp dark line around the crystals. A
colour change of alite from light grey to a darker
grey or brown is common. Belite is slightly etched
structurally and slightly changes its colour from
light grey to a darker grey.
Additionally to light microscopy, the polished
section from case study 1 was analysed with a
scanning electron microscope (Philips ESEM XL30
FEG) with the possibility for energy dispersive X-
ray (EDX) analysis. EDX analyses were carried out
in the form of spot analyses and in the form of
element mappings.
III. Results of case study 1
In this case study a clinker sample was analysed to
estimate the effects of different fuels on the clinker
properties. In general the clinker sample was well
burned and did not show unusual phases or mi-
crostructural features.
However, one single clinker granule in the pol-
ished section consisted mainly of belite, free lime
and ground mass (C3 A, C4 AF) with a low amount
of alite crystals. The alite crystals were surrounded
by a symplectite of belite (belite I in Figures 1 and
2) and fine grained free lime crystals (free lime I in
Figures 1 and 2). The free lime crystals were often
oriented towards the alite crystals. The symplec-
tites formed a layer of up to 20 µm thickness. Many
symplectites did not contain cores of alite. Beside
these symplectites with and without alite cores, the
granule consisted of a mixture of coarse grained
belite (belite II in Figures 1 and 2) and free lime
crystals (free lime II in Figures 1 and 2).
In usual Portland cement clinker the direct con-
tact of belite and free lime occurs only as a result of
high concentrations of phosphorous (e.g. Puntke,
Schneider, 2005), of alite decomposition due to re-
ducing burning conditions (VDZ, 1965; Böhm, 2011)
or in poorly burned material. Poor burning leads to
crystal sizes below 10 µm and high porosity (VDZ,
1965; Campbell, 1999; Böhm, Pierkes, 2009). Both
features were not found here. Additionally large
alite crystals (Figure 1) prove that the material was
exposed to conditions sufficient for alite formation.
Phosphorous can stabilise belite, preventing the for-
mation of alite and leading to a mixture of coarse
grained belite and free lime (Puntke, Schneider,
34
2005; Böhm, Pierkes, 2009). However, symplectites
of belite and free lime as they were observed here
(Figure 1) usually do not occur in connection with
phosphorus.
The symplectites of belite and free lime surround-
ing alite crystals and the orientation of the elon-
gated free lime crystals pointing towards the alite
crystals indicate the (partial) breakdown of alite.
The symplectites with and without alite cores prob-
ably formed as pseudomorphs after alite. The most
common cause for the breakdown of alite in mod-
ern Portland cement clinker is the occurrence of
local reducing conditions in the kiln feed, caused
by smouldering particles of AFR. This can lead to
symplectites of belite and free lime as in Figure 1.
However, reducing conditions do not prevent the
formation of alite as indicated by the coarse grained
mixture of belite and free lime crystals around the
symplectites.
To clarify the cause for the formation and subse-
quent partial breakdown of alite while at the same
time the formation of alite was prevented in the
vicinity, the clinker granule was analysed using
SEM and EDX analyses. The measurements re-
vealed unusually high concentrations of barium in
the clinker granule (Table 1). Alite contained less
than 2 mass % of barium. The belite crystals in the
symplectites (belite I) contained about 7 mass %
of barium, whereas the belite crystals mixed with
coarse grained free lime (belite II) contained about
10 mass % of barium. The barium distribution be-
tween alite and the two populations of belite is also
illustrated in the element maps in Figure 2.
The different barium contents and the microstruc-
tural features lead to the following interpretation.
The high barium content in the belite II-crystals
prevented the formation of alite and led to a mi-
crostructure comparable to that caused by high
concentrations of phosphorous. The concentrations
of barium in the remaining alite crystals was not
high enough to prevent the formation of alite at
sintering temperatures or the breakdown of alite to
belite and free lime during cooling. However, in
some alite crystals the barium concentration was
low enough to allow the formation of alite at sin-
tering conditions, but high enough to destabilise
the crystal structure, leading to its disintegration
during cooling and the formation of symplectites
of belite I and free lime. This process either led to
the breakdown of complete alite crystals or only of
the outer rims, indicating zonation in the original
alite crystals with increasing barium contents from
core to rim.
It remains unclear if the inhomogeneous distribu-
tion of barium is the consequence of the inhomoge-
neous distribution of the element in its source. An-
other plausible explanation would be that the alite
crystals selectively incorporated lower amounts
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 1: Reflected light micrograph of clinker; alite directly surrounded by a symplectite of free lime crystals often
oriented towards alite (free lime I) and belite crystals (belite I), itself surrounded by a mixture of coarse round free
lime crystals (free lime II) and belite (belite II); outer rim of symplectite marked with dashed red line.

of barium at the beginning of their formation at
lower temperatures and increased the incorporated
amount in outer zones formed later and therefore
at higher temperatures. The observation that zones
containing higher barium concentrations disinte-
grated during cooling indicates that the amount of
barium, which can be integrated into the alite crys-
tal structure, increases with increasing temperature.
The granule described here was the only one in
the polished section showing microstructural fea-
tures influenced by barium. The source for barium
was therefore an exceptional compound in the raw
materials or fuels used. The most probable source is
a crystal of baryte (BaSO4), which can be found in
the limestone formations used in the cement plant
in which the clinker was produced.
The average amount of barium in cements is 280
mg/kg. It replaces Ca in all clinker phases ex-
cept C4 AF (Bhatty, 1995). It can decrease the clink-
erisation temperature, improve the mineralogical
composition and increase cement strength (Bhatty,
1995). However, most studies worked with smaller
concentrations of barium than found in the clinker
granule described here (e.g. Bhatty, 2006).
IV. Results of case study 2
Also in this case study a clinker sample was anal-
ysed to estimate the effects of different fuels on the
clinker properties. The clinker sample was well
burned and mostly did not show unusual phases
or microstructural features.
However, in some granule fragments in the pol-
ished section alite crystals partially decomposed
to belite and free lime and/or C4 AF along crys-
tallographic preferred orientations. In some gran-
ule fragments alite crystals partially decomposed
to symplectites of belite and free lime. Both mi-
crostructural features indicate reducing burning
conditions (VDZ, 1965; Böhm, Pierkes, 2009; Böhm,

Table 1: EDX measurements of the composition of alite, belite I (crystals close to alite and in contact with fine grained
free lime), and belite II (crystals remote from alite and in contact with coarse grained free lime).

alite belite I belite II

Oxide
(3 measurements) (3 measurements) (7 measurements)
CaO 66.4 ± 0.2 59.0 ± 2.1 55.4 ± 1.5
SiO2 26.4 ± 0.2 28.8 ± 2.7 30.3 ± 1.2
BaO 1.3 ± 0.7 6.8 ± 0.2 9.8 ± 1.6
Al2 O3 2.5 ± 0.1 2.5 ± 0.5 2.0 ± 0.4
Fe2 O3 0.8 ± 0.1 1.1 ± 0.1 1.0 ± 0.4
MgO 1.6 ± 0.4 0.7 ± 0.5 0.4 ± 0.2
K2 O 0.3 ± 0.1 0.7 ± 0.1 0.7 ± 0.1
Na2 O 0.7 ± 0.5 0.4 ± 0.1 0.3 ± 0.1

35
 Böhm & Lipus

Figure 2: Reflected light (top left) and SEM micrograph (top right) of clinker; elemental maps (Ba mid left, Ca mid
right, Si bottom left, Al bottom right) of alite directly surrounded symplectitic arrangement of free lime and belite,
itself surrounded by a mixture of coarse free lime and belite; width of each image 235 µm.

2011). The decomposition of alite is caused by the
incorporation of Fe2+ ions in the crystal structure
of alite. The ions form under reducing conditions
in the kiln and they destabilise the crystal structure
of alite (e.g. Sylla, 1981).
Additionally several clinker granule fragments
contained elemental iron particles forming sym-
plectites with free lime crystals (Figure 3). These
particles also contained alite, partially decomposed
to fine grained symplectites of belite and free lime,
as well as belite and C3 A, while C4 AF was not pre-
served. In these clinker particles belite and C3 A
were hard to discern with the etching procedure
described above (Figure 3).
Iron particles can form under strongly reducing
conditions, but they are usually not intergrown
with free lime (e.g. Pierkes, Böhm, 2009). The re-
curring symplectitic structure of iron and free lime
indicates an equally recurring precursor phase con-
taining CaO as well as Fe. The symplectites were
xenomorph and seem to form, together with the
areas consisting of C3 A and belite, a groundmass
in which alite crystals are embedded.
Probably C4 AF was the precursor phase for the
symplectites. It must have formed before the ma-
terial was exposed to strongly reducing conditions,
which led to the conversion of ferric iron in the
phase to elemental iron. The aluminium content of
C4 AF was bound in the form of C3 A, whereas the
excess CaO formed free lime intergrowing with the
newly formed elemental iron. Estimations based
on chemical composition, molar masses and densi-
ties result in a CaO/Fe-volume ratio of 0.85, which
roughly coincide with the area ratio of free lime and

36
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Reflected light micrograph of clinker; symplectites of free lime and elemental iron together with alite crystals
in a ground mass free of C4 AF.

iron particles in the symplectitic structures (Figure

3).
The concerned clinker granules must have been
exposed to the reducing conditions after passing
the sintering zone, since relatively large C4 AF crys-
tals seem to have been the precursor for the free
lime-iron symplectites. However, the temperatures
must have been high enough to allow for the re-
crystallization of the ground mass.

References
Bhatty, J.I. (1995): “Role of Minor Elements in Cement Manufacture
and Use.” PCA Research and Development Bulletin RD109T. Portland Ce-
ment Association (PCA), Skokie, Ill.
Bhatty, J.I. (2006): “Effect of Minor Elements on Clinker and Cement
Performance: A Laboratory Analysis.” PCA Research and Development Bul-
letin RD130. Portland Cement Association (PCA), Skokie, Ill.
Böhm, M., Pierkes, R. (2009): “Evaluation of Portland cement clinker
with optical microscopy - Case Studies.” Proceedings of the 12th Eurosem-
inar on Microscopy Applied to Building Materials, Dortmund, Germany,
September 15 - 19.
Böhm, M. (2011): “Evaluation of Portland cement clinker with opti-
cal microscopy - Case Studies II.” Proceedings of the 13th Euroseminar on
Microscopy Applied to Building Materials, Ljubljana, Slovenia, June 14 - 18.
Campbell, D.-H. (1999): “Microscopical examination and interpretation
of Portland cement and clinker.” 2. Edition. Portland Cement Association
(PCA), Skokie, Ill.
Puntke, S, Schneider, M. (2005): “Effects of phosphate on clinker min-
eralogy and cement properties.” Cement International. Vol. 6, No. 5, 80-93
VDZ Verein Deutscher Zementwerke (1965): “Mikroskopie des Ze-
mentklinkers.” Beton Verlag, Düsseldorf.

37
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

38
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

A new quantification method based on

SEM-EDX to assess fly ash chemical
composition and its reactivity in
composite cements
P. T. Durdziński⇤,a , C. F. Dunanta , M. Ben Hahab , K. L. Scrivenera
a Laboratory of Construction Materials, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland.
b. HeidelbergCement Technology Center GmbH (HeidelbergCement AG), Leimen, Germany.
⇤ pawel.durdzinski@gmail.com

Abstract

Calcareous fly ashes are high-potential reactive residues for composite cements. However, accurate qualification and
use in concrete are hindered by their inherent heterogeneity and large variability. Current characterization techniques
often fail to deliver satisfactory detail, in particular on the dominant amorphous fraction of the fly ash. Furthermore,
a lack of appropriate tools to quantify the extent of fly ash reaction in cement leaves many aspects of their reactivity
poorly understood. We developed a new approach to characterize fly ash using electron microscopy. EDX element
composition of millions of points is plotted in a ternary frequency plot. A quick visual analysis informs the operator of
the number, approximate chemical composition and relative proportions of populations that constitute the fly ash. For
quantitative characterization of the raw fly ash, these populations are grouped as predominantly: silicate, calcium-silicate,
aluminosilicate and calcium-rich aluminosilicate. The technique was further developed to quantify the consumption of
these fly ash populations in Portland-fly ash cement during hydration. Two calcareous fly ashes of similar bulk chemical
composition by XRF actually contained very different amounts of the populations, which react at different rates. Fly
ash 1 contains around 90 vol.-% of moderately-reactive aluminosilicates and calcium-silicates. Overall reaction of fly
ash 1 reached 60% at 90 days. Fly ash 2 reacted faster, reaching 60% reaction at 28 days as it contains 55 vol.-% of
highly-reactive calcium-rich aluminosilicates. Further reaction was slower as 20 vol.-% of fly ash 2 were non-reactive
silica-rich phases.

Keywords: composite cement, fly ash, characterization, reactivity, EDX.

I. Introduction but a collection of compositions (Pietersen, 1993).

Calcareous fly ashes derived from coal combustion
are high-potential reactive residues for composite
cements (Papadakis, 2000). However, these fly ashes
are inherently heterogeneous and variable and it
is difficult to predict and to control the properties
of cements containing them. Characterization and
qualification of these ashes is not easy and in con-
sequence many are discarded out of hand. Indeed,
only about 30 wt.-% of globally available calcareous
fly ashes are used for cement and concrete.
Calcareous fly ashes are composed of crystalline
and amorphous phases, among which the amor-
phous phases usually dominate. Indeed, the amor-
phous fraction in fly ashes may represent up to
around 90 wt.-% of this material. The crystalline
phases can be identified by XRD diffraction and
quantified with Rietveld refinement. The amor-
phous fraction is the main contributor to reactive
components. Unfortunately it is not homogeneous
Most of the currently available characterization tech-
niques fail to distinguish between the various amor-
phous compositions in a fly ash. As a result, bulk
chemical information is often the only data read-
ily available for characterization. It is frequently
reported in studies on fly ashes and even used in
standards as a basis to classify the ashes (ASTM
2012; EN 2011). However, any kind of bulk descrip-
tion will always lack the necessary detail to describe
a complex multi-phase material like fly ash.
Electron microscopy EDS can be used to investi-
gate chemical composition of glassy particles in fly
ash. The EDS data can be collected in point mode
(operator chosen or a grid of points) or as map-
ping. As the number of points in the point mode
is around one to several hundred, the large hetero-
geneity of fly ash composition is likely to cause a
significant scatter in the results and thus hinder
any detailed analysis. With the advent of fast EDS
detectors it is now possible to collect a spectrum

39
 Durdziński et al.

a quick and intuitive determination of the number

Figure 1: Particle size distributions by laser diffraction.
of glass populations in a fly ash and the ranges
of their chemical composition. These ranges are
then used as selection criteria for image analysis
and quantification of the populations in the raw
ash. Once the raw ash is analyzed, this approach
can be used to track the anhydrous populations
in a hydrating Portland-fly ash paste. Quantities
reported over hydration times serve to analyze the
reaction progress of each of the populations indi-
vidually. More details on the described procedure
can be found in (Durdziński et al., 2015).

for every pixel of the image frame within an hour

(mapping). The spectra are represented in a set of
intensity maps, one per element. Such maps can be
analyzed with MultiSpec c , a multispectral image
analysis software, as presented in (Chancey et al.,
2010) or similar. Such software uses an algorithm
to identify and quantify groups of mutually exclu-
sive compositions. This approach can treat large
amount of data but the analysis may be difficult
and may require a skilled operator. The main draw-
back of using an algorithm is that the operator must
specify the number of groups or let the algorithm
do it. In both cases the control over the analysis is
partially lost. Further, if EDS X-ray count is used
as an input and not the more time-consuming cal-
culated element amounts in atomic-% or mass-%,
the chemical composition of the segmented groups
can only be analyzed semi-quantitatively. This is
because X-ray counts are not a direct measure of
element content.
In this study we quantify the spectra in EDS
maps using profile fitting and PhiRhoZ matrix cor-
rections to obtain maps that contain directly the
element content in atomic-%. These results are
plotted in a ternary plot with Al, Si and Ca - the
dominant elements - in the apexes. A statistically
reliable analysis requires millions of data points
and this amount quickly saturates a plot of single
data points. Our solution is to re-plot the data using
a ternary frequency plot. In such plot the dominant
compositions appear as color blobs, which allows
II. Materials
Two calcareous fly ashes (FA1 and FA2) and Port-
land cement (PC) were used in this study. Their
bulk chemical composition was determined by XRF
(Table 1) and their phase composition by XRD-
Rietveld refinement (Table 2). These materials were
ground to 4200 cm2 /g Blaine fineness and the re-
sulting particle size distributions were measured
by laser diffraction using Malvern MasterSizer S as
shown in Figure 1.
III. Methods
Sample preparation
For characterization of raw fly ashes, powder fly
ash samples were compacted by pressing into pel-
lets and embedded in epoxy resin. To study fly
ash hydration we prepared blends of 55 wt.-% PC
and 45 wt.-% fly ash (PC-FA1 and PC-FA2). These
blends were mixed with water at a ratio 0.4 with a
laboratory mixer at 1600 rpm for 2 minutes. The
pastes were cast in cylindrical PP containers, sealed
and cured at 20 C. After 1, 7, 28, 56, 90 and 365
days discs were cut from the cylinder and stored in
isopropanol for 7 days to stop hydration by solvent
exchange. The discs were then air-dried and stored
for another 7 days under vacuum in a desiccator
with silica gel to remove the isopropanol. Part of
the disc was hand ground in a mortar and analyzed

Table 2: Phase composition by XRD-Rietveld with rutile

Table 1: XRF bulk chemical analysis [wt.-%]. external standard [wt.-%].

PC FA1 FA2 PC FA1 FA2

Al2 O3 5.5 19.8 18.2 C3 S 66.3 - -
SiO2 20.7 42.3 33.6 C2 S 8.3 2.5 0.5
CaO 65.7 20.7 26.5 C4 AF 8.7 2 1.5
Na2 O 0.2 0.3 1.9 C3 A 7.7 1 4
K2 O 0.4 1.5 0.4 Quartz 0.5 1.3 3.5
MgO 1.5 2.2 6.4 Gypsum 2.6 - -
Fe2 O3 2.7 8.2 6.4 Anhydrite 3 1.8 2
SO3 2.6 1.4 2.2 Free lime - 1.7 0.6

40
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 2: BSE image of a polished section of an epoxy impregnated raw FA2 (left) and a visual representation of the
distribution of Al - red, Si - green and Ca - blue, as measured by EDS (right).

by XRD-Rietveld. The remainder was prepared
for electron microscopy by hand polishing and em-
bedded in epoxy resin. The resin mounts were
gradually polished down to 14 µm. The samples
were washed in isopropanol in an ultrasound bath
and kept in a vacuum desiccator until microscopy
measurements.
Microscope setup, data collection and analysis
A FEI Quanta 200 microscope with Bruker XFlash
4030 EDS detector was operated at 15 kV acceler-
ating voltage and 12.5 mm working distance. At
the frame size of 252 µm ⇥ 189 µm and 1024 ⇥
768 image resolution, the pixel size was 0.246 µm
⇥ 0.246 µm. To ensure high quality of spectra in
EDS mapping we chose a large spot size so the X-
ray count rate would reach around 100 000 counts
per second and a dwell time of 256 µs to keep the
scanning speed moderate. This is a compromise
between poorer statistics at high scanning speeds
and larger sample damage at lower speeds. An au-
tomated procedure collected BSE images and EDS
maps of eight frames per sample, which was found
enough for the subsequent analysis and required
around 4.5 h of collection time.
The amounts of Al, Si, Ca, Na, K, Mg, Fe, Ti
and P in atomic-% were obtained from EDS spec-
tra using profile fit with carefully collected stan-
dards and PhiRhoZ matrix corrections. These
calculations were carried out automatically by
the Esprit software and required around 1 h per
frame. The results were treated in MATLAB soft-
ware. To account for the X-ray excitation vol-
ume the raw data was smoothed with a Ham-
ming window. The size of the window matched
the size of the excitation volume modelled using
Monte Carlo simulation in Casino v2.41 software
(www.gel.usherbrooke.ca/casino/index.html) for a
bulk composition of FA2 and the parameters the
microscope was operated at. Background pixels
were removed by thresholding of an associated BSE
image, for which the threshold value was deter-

Figure 3: An Al-Si-Ca ternary plot (A) and a ternary frequency plot (B) of the FA2 composition in atomic-%. Many
of the points appearing in (A) are actually due to noise and do not appear in (B). The fraction > 95 atomic-% Si
contains a high intensity peak and was truncated from B to better visualize other phases. White contours show the
arbitrary boundaries between the groups which are used in this study.

41
 Durdziński et al.

Figure 4: Illustration of the segmentation criteria of the four fly ash populations. Selected areas are hexagons centered
around the points of highest intensity found in the studied fly ashes. The exception is group 2, whose area was
extended towards Al. Using hexagons for selection leaves some areas unselected.

mined by the operator from BSE image gray level
histogram. The remaining data was smoothed us-
ing a Hamming window and plotted in an Al-Si-Ca
ternary diagram in atomic-%. A ternary plot of
individual data points cannot be interpreted as it is
saturated with millions of points (Figure 3A). Our
solution is to re-plot the data in a ternary frequency
plot (Figure 3B). For this we divide the area of the
plot into 10 000 equal triangles (bins), thus corre-
sponding to 1 atomic-% on each of the three axes,
and we represent the number of points that fall in
each bin on a linear color scale. In a frequency plot
the dominant compositions appear as bright blobs
while irrelevant and noise pixels are not displayed.
Visual analysis of the ternary frequency plot is
carried out to identify compositionally different
populations and determine their ranges of chemical
composition. These ranges are then used as segmen-
tation criteria for image analysis. The amount of a
given population is calculated as an area fraction
of all pixels belonging to this population per image
frame, which according to fundamental stereology
equals the volume fraction of this population in the
material.
IV. Results and discussion
Characterization of a raw fly ash
BSE image of raw FA2 reveals a complex microstruc-
ture and a multitude of phases distinguished by
their grey levels (Figure 2). The EDS data for Al, Si
and Ca is displayed with color attributed to each of
these elements: red – Al, green – Si, blue – Ca (Fig-
ure 2). One can observe compositionally different
populations and that some particles are composed
of multiple phases. The vast heterogeneity of fly ash
microstructure is likely a result of different minerals
contained in coal particles; as their combustion and

Figure 5: Quantitative comparison of populations present in FA1 and FA2.

42
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
cooling occurs locally the melt does not undergo
homogenization. Calcareous fly ashes may contain
large amounts of Ca, Mg, Na and K, which can
cause phase separation in glass (Pietersen, 1993; Vo-
gel, 1977) and further contribute to the complexity
and heterogeneity of fly ash microstructure.
FA2 was composed of several populations readily
distinguishable in the Al-Si-Ca ternary frequency
plot (Figure 3). The predominantly silicate- and
aluminosilicate- populations are represented by
high and narrow peaks, which suggest well de-
fined chemical composition and large abundance.
The Ca-rich alumino-silicate group has a diffuse
shape of nonetheless high intensity. This phase or
group of phases seems to be the most abundant
in the fly ash but its chemical composition spans
a large section of the ternary graph. Intermediate
phases such as calcium-silicates with moderate Al
and alumino-silicates with moderate Ca are well
defined but present in lower amounts. Phases ap-
pearing at the 100 % Ca apex typically include free
lime (CaO), calcium sulfates and calcium carbonate
(calcite). According to XRD Rietveld analysis, FA2
contains around 4 wt.-% C3A, which appears in the
plot as a blob below the calcium-rich aluminosili-
cate group. The quantity of non-Ca-Al-Si phases is
below 3 wt.-%, as determined by XRD-Rietveld (Ta-
ble 2); they do not appear in the Ca-Al-Si diagram.
To analyze the specific populations in more detail
one must isolate them using selection criteria: lim-
its on Al, Si and Ca were sufficient to separate the
groups but in different systems any of the quanti-
fied elements and more than three elements may be
used. Wide limits provide the necessary margin for
often broad chemical compositions and reduce the
measurement error due to data scatter and poten-
tially low content of the populations. We propose
four main populations, roughly corresponding to
(1) mainly silicate, (2) calcium-silicate with low to
moderate aluminum, (3) alumino-silicate with low
to moderate calcium and (4) calcium-rich alumino-
silicate (Figure 4).
The presented selection criteria can be used to
compare different fly ashes on a quantitative basis
(Figure 5). FA1 is mainly composed of aluminosil-
icates with low to moderate calcium content (70
vol.-%) and calcium-silicates with low to moderate
aluminum (15 vol.-%). In FA2 these two phases
account together for only around 20 vol.-%. FA2
contains however a significant amount of around 55
vol.-% of potentially very reactive calcium-rich alu-
minosilicate phases, which in FA1 are only present
in a small quantity, below 10 vol.-%. Nevertheless,
almost one fifth of FA2 volume are highly silicate
phases, which are expected to not react at all.
Reactivity of fly ash populations in cement paste
The reaction of fly ash is a sum of reactions of dif-
ferent contributing populations and it is crucial to
track the reaction of each of them separately. Each
fly ash population can be extracted from hydrated
cement paste using a procedure extended from that
for a raw fly ash. Here the difference is that one
must filter out hydrate phases of compositions over-
lapping with those of fly ash populations. Hydrates
contain water and because neither oxygen, nor hy-
drogen are quantified by EDS, the total amount of
the quantified elements is lower for hydrates than
for anhydrous phases. On this basis hydrates can
be removed from further analysis. The previously
defined criteria were used to assign pixels to fly ash
populations. Volume fraction of each population
was calculated from segmented images. The values
at time zero were calculated from the proportions of
Portland clinker, calcareous fly ash and water in the
mix and their densities. Overlap of Ca-rich alumi-
nosilicates with anhydrous C3 A and C4 AF phases
from Portland clinker was taken into account.
The unreacted volume fractions for each of the
fly ash populations in hydrated Portland – fly ash
pastes from 1 to 365 days are shown in Figure 6. The
two investigated fly ashes contained significantly
different proportions of the identified populations.
These populations contribute differently to the re-
action of fly ash. Silica-rich phases are either inert
or only slightly reactive, which is an important ob-
servation for FA2, containing around 20 vol.-% of
these phases. In FA1 silica-rich phases were a mi-
nor component. The Al-silicate and the Ca-silicate
phases reacted at a moderate rate of around 50 %
reacted at 90 days. These phases dominated in FA1
but accounted for only around 25 vol.-% of FA2.
The calcium-rich aluminosilicates were the most
reactive, exceeding 80 % degree of reaction at 28
days (FA2) and 56 days (FA1) of hydration and con-
tributed the most to the reaction of FA2. Indeed,
Ca-rich glasses tend to be more reactive than those
with moderate and low Ca contents (Aughenbaugh
et al., 2013).
Large amounts of highly reactive Ca-rich
alumino-silicate glass can have an important im-
pact on the hydration products of PC-FA mixes,
leading to formation of significant amounts of AFm
and AFt phases. As these phases can interact with
Cl- and SO24 ions from aggressive environments
they may affect the durability of the binder. Large
quantities of readily available alumina from the
fast reacting Ca-rich alumino-silicate glass could
potentially react with carbonates from limestone
(Matschei, Lothenbach, and Glasser, 2007) and im-
prove its reactivity in composite cement, which may
subsequently allow larger clinker replacement.
43
 Durdziński et al.
Figure 6: The time decrease of the amount of anhydrous fly ash populations in PC-FA1 (left) and PC-FA2 (right).
Clinker phases from the fly ashes are excluded.
V. Conclusions
This study presents a novel technique to identify, vi-
sualize and quantify populations of different chem-
ical composition in a complex material, here fly ash.
The technique was extended to track these popu-
lations in hydrated Portland-fly ash pastes. The
results confirmed that different populations con-
tribute differently to the reaction of fly ash, which
strengthens the view that fly ashes should not be
judged by their bulk chemical composition only.
Understanding of fly ash composition, reactivity
and effect on concrete properties is a challenge for
cement research. The method presented here pro-
vides a tool that could facilitate investigations and
may lead to an improved classification and use of
calcareous fly ashes in concrete.
Acknowledgements
The authors would like to acknowledge Heidel-
bergCement for the financial support, John Rossen
for helpful advice on microscope setup and Ruben
Snellings for comments and discussions.
44
References
ASTM Standard C618: Standard Specification for Coal Fly Ash and
Raw or Calcined Natural Pozzolan for Use in Concrete.
Aughenbaugh, K.L., Chancey R.T., Stutzman P., Juenger M.C., and
Fowler D.W. (2013): “An Examination of the Reactivity of Fly Ash in
Cementitious Pore Solutions.” Materials and Structures 46 (5): 869–80.
Chancey, R.T., Stutzman P., Juenger M.C., and Fowler D.W. (2010):
“Comprehensive Phase Characterization of Crystalline and Amorphous
Phases of a Class F Fly Ash.” Cement and Concrete Research 40: 146–56.
Durdziński, P.T., Dunant C.F., Ben Haha M., and Scrivener K.L. (2015):
“A New Quantification Method Based on SEM-EDX to Assess Fly Ash
Composition and Study the Reaction of Its Individual Components in
Hydrating Cement Paste” Cement and Concrete Research (in press).
EN 197-1:2011 - Cement - Part 1: Composition, Specifications and
Conformity Criteria for Common Cement.
Matschei, T., Lothenbach B., and Glasser F.P. (2007): “The Role of
Calcium Carbonate in Cement Hydration.” Cement and Concrete Research
37 (4): 551–58.
Papadakis, V.G. (2000): “Effect of Fly Ash on Portland Cement Sys-
tems: Part II. High-Calcium Fly Ash.” Cement and Concrete Research 30
(10): 1647–54.
Pietersen, H.S. (1993): “Reactivity of Fly Ash and Slag in Cement.” TU
Delft.
Vogel, W. (1977): “Phase Separation in Glass.” Journal of Non-
Crystalline Solids 25 (1-3): 170–214.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Utilization of methods of optical

microscopy in terms of evaluation of
cement materials
T. Staněk⇤,a , P. Sulovskýb
a. Research Institute for Building Materials, Hněvkovského 65, 617 00 Brno, Czech Republic
b. Department of Geology, Faculty of Science, Palacký University, tř. 17. listopadu 12, 771 49 Olomouc, Czech Republic
⇤ stanek@vustah.cz

Abstract

The report describes possibilities of utilization of optical microscopy in cement industry. It deals with historical
importance and with advantages and disadvantages of actual possibilities. Methods that utilize transmitted or reflected
light are described in this paper as well as the method of high temperature microphotometry and possibilities of utilization
of optical microscopy in a cement factory.

Keywords: Portland clinker, optical microscopy, microstructure

I. Introduction with representativeness of observed objects occurs

Optical microscopy played a significant role in the
development of mineralogy of cement clinker al-
ready in the pioneer research of LeChatelier and
Törnebohm at the end of 19th century. However,
the importance of optical microscopy currently de-
creases due to a great development of measuring
and analytical techniques. Compared to other meth-
ods, the uniqueness of this method lies firstly in
the possibility of qualitative and quantitative anal-
ysis of clinker phases and secondly in the ability
to study the structure of clinker, which reflects the
processes of the clinker formation (Maki, 2004).
The optical microscope enables direct observa-
tion of structure and phase composition of Port-
land clinker. The phase composition of clinker is
the result of two main production operations – the
preparation of raw meal and burning – and clinker
microstructure is a result of these operations. It
often shows also the properties of the source of
raw meal at the locality. The composition and the
structure of clinker also significantly influences the
properties of cement; therefore clinker is an inter-
mediate product with crucial importance in both
determination of these properties and analysis of
preceding processes.
Microscopy is irreplaceable during laboratory re-
search on clinker. In the production process, micro-
scopic analysis is not necessary, but provides useful
information besides common chemical and XRD
methods. If the situation requires, other methods
with higher resolution should be used additionally,
such as electron microscopy. However, a problem
due to the scale used.
During microscopic evaluation, human factor as
well as image definition has to be taken into account.
Overcoming of weaknesses of optical microscopy
lies in development of methods for better image
processing and particularly in exact recognition of
morphology of every clinker mineral.
II. The process of Portland cement
clinker formation
Optical microscopy played and still plays a signifi-
cant role in the research on mechanism and kinetics
of the formation of Portland clinker and in the re-
search on the influence of different factors on the
formation and properties of Portland clinker, e.g.
the temperature programming, i.e. the temperature
of preheating, the temperature of firing and also
the chemical and mineralogical composition of the
raw meal.
A significant step in the process of clinker forma-
tion is the liquid phase creation. In the beginning of
the sintering process, before the creation of liquid
begins, decomposition processes of the raw meal oc-
cur (such as dehydratation and dehydroxylation of
clay minerals and micas, dissociation of carbonates)
and also reactions in solid phase develop, which
can be monitored by DTA. In current production
systems, these processes run firstly in dispersed
state in the heat exchanger (or in calciner), then in
loose state in rotary kilns; these processes are over-
all highly endothermic. Belite is created already

45
 Staněk & Sulovský
during the dissociation of calcite via reactions in
solid phase (Chromý, 1976).
The clinker liquid is created in the beginning of
the sintering process. Before this happens, local eu-
tectics are created in areas with higher Fe content,
which is the beginning of aggregation of loose ma-
terial. Gradual creation of continuous clinker liquid
is at the beginning accompanied by quick exother-
mal generation of belite (Chromý, 1982; Kondo and
Choi, 1968; Christensen et al., 1978; Staněk et al.,
2004), which is soon finished, and the end of liquid
creation is strongly endothermic.
Before the end of belite formation according to
equation 2C + S = C2 S, a thin layer of alite is cre-
ated around aggregates of free lime, the layer being
called “the limiting layer”. The last, slowest and
therefore critical reaction of alite formation runs
according to equation C + C2 S = C3 S. Free lime
dissolves in the liquid on the surface of the aggre-
gate, then it is transported by diffusion through the
liquid phase of alite layer; alite crystallizes on the
outer surface of this layer immediately after the dis-
solving belite. The concentration difference of CaO
in liquid between the surface of free lime aggregate
and the C3 S – C2 S border is stable and very small
(Johansen, 1979). That is why the speed of this
reaction is very low and progressively decreases
with the degree of conversion (Christensen et al.,
1978; Chromý and Hrabě, 1982). Removing of the
last percent of free lime requires almost doubling
of isothermal burning time, which is an important
finding for the energetic economy of production.
Along with the creation of liquid at the begin-
ning of the sintering zone aggregation of originally
loose material occurs. The intensity of granulation
decreases with the increasing conversion by the
reaction C + C2 S = C3 S, i.e. with increasing alite
content in clinker. Granules with excess of belite
are plastic with higher coverage of their surface
by liquid (due to closer packing of spherical belite
grains, which do not intergrow), whereas the later
intergrowing angular alite crystals build a solid
structure of grains with lower coverage of the sur-
face by liquid. Further aggregation of particles runs
particularly by sticking of the clinker powder on
the surface of granules or by aggregation of smaller
particles (Weber, 1989).
III. Relationship between
mineralogical and chemical
composition of clinker
The mineralogical composition of clinker deter-
mines the properties of cement. There were two
basic methods to evaluate phase composition of
clinker in the not far-off history (Aldrige, 1975).
The methods have nearly the same precision but
46
the results significantly differ in accuracy. The meth-
ods are microscopic point counting method and a
method of calculation of composition from chemi-
cal analysis.
The main equation for the calculation of the so-
called potential phase composition of clinker was
derived by Bogue. The calculated phase composi-
tion of clinker considerably differs from the real
one, sometimes by 10% or more. The reason for
the difference between the calculated and the real
phase composition of fabricated clinker are devia-
tions from the assumed chemical composition of
clinker phases, which depend on the content of
minor oxides in the raw meal, on kinetics of re-
actions and the cooling rate, i.e. above all on the
source of the raw materials and the production fa-
cilities. The possibility of finding a universal valid
method for calculation of phase composition of in-
dustrial clinker from chemical analysis is highly
improbable. Such method would have to take into
account the interrelationships in the creation of
solid solutions of clinker phases in conditions of
non-equilibrium crystallization, the differences in
the raw meal source localization and in production
facilities.
At present, phase composition can be assessed
by another method: X-ray diffraction analysis. X-
ray diffractometers with quick detectors (PSD) and
Rietveld optimalisation (Schmidt and Kern, 2001)
may provide relatively precise information about
the phase composition of the clinker, but cannot
give information about the structure of the clinker.
However, another source (Pritula et al., 2003) men-
tions, that Rietveld analysis for Portland clinker
does not give sufficiently precise and accurate re-
sults yet.
IV. Optical microscopy methods
The polarization microscope is optimal for the re-
search on cement raw materials and cement prod-
ucts because of its universality. With the micro-
scope, powder samples and thin sections can be
observed in transmitted light, or polished-sections
in reflected light.
Transmitted light
Mineralogical composition of powder samples of
raw meal, kiln dusts, cements, slags, ashes, gypsum
etc. can be observed in transmitted light in petro-
logical microscope. The microscopic specimen is
prepared by mounting the sample on microscope
slide, then a drop of suitable immersion liquid is
added and the sample is covered by a cover slip.
The phase composition is identifiable by character-
istic optical properties of minerals such as color,
cleavage, shape, index of refraction, birefringence,
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
pleochroism, isotropy and anisotropy, extinction
angle and other special properties.
Furthermore, it is possible to prepare thin sec-
tions, when the observed sample is cut by a dia-
mond blade saw, than it is ground and glued to a
microscope slide with appropriate suitable liquid.
After glueing it is cut again by the diamond saw
to get only a few tenths of millimeter in thin wafer,
which is subsequently ground to thickness of about
0.03 mm. After that a cover slip is stuck on the
surface with the same medium. This preparation is
appropriate to determine the mineralogical compo-
sition and structure of clinker and rocks destined
for preparation of raw meals. The preparation may
be used also for powder samples, but it has to begin
with preparation of casts from the powder mixed
with appropriate mounting material, which can be
cut and ground.
Reflected light
In the cement industry, observation in reflected
light is used more often, because the preparation
of the specimen is simpler and the main interme-
diate product – Portland clinker – and possibly
also cement can be observed. It is not necessary
for the observer to have a special mineralogical-
petrographic knowledge, a training by an expert is
sufficient.
The polished section is prepared as a cast from
representative clinker chippings (or cement) in an
appropriate matrix. The surface is ground and at
the end polished with diamond paste.
During the reaction of clinker silicates and free
lime with fumes of acetic acid, a layer of reaction
product is formed on the surface of the polished sec-
tion. The thickness of the layer is a function of the
CaO : SiO2 ratio. Due to the interference of reflected
light these minerals get distinctly different colours
(Chromý, 1974a). Practical realization consist of a
short exposition (around 1 s) of the polished section
placed on piece of textile tissue, which is tightened
over the neck of a bottle with acetic acid. Beyond
this time the bottle has to be covered by a glass
plate. Alite crystals are pale blue to blue-green,
belite grains are brown and free lime is bright yel-
low, orange or red (see Figure 1). Interstitial mass
cannot be colored but if it is well crystalized, its
components can be distinguished, C3 A creates gray
and C4 AF white areas in the interstitial mass (see
Figure 2).
Compared to other methods, the advantage of
optical microscopy is the possibility to observe also
the structure of clinker, which tells us more about
processes during its formation. The size of alite and
belite crystals can be measured, their shape and the
inclusions in alite crystals can be observed, their
zonation, the creation of lamellas in belite, which
indicates polymorphous composition. Due to this
method, the degree of burning and therefore the de-
gree of crystallization of silicates can be evaluated,
which can influence the initial strength of cement.
From the observed structure, the properties of raw
meal can be estimated, particularly its reactivity.
It is also possible to identify aggregates of belite,
formed as residues after big grains of SiO2 , or to
reveal excessive amounts of SO3 or MgO in the
clinker.
High temperature microphotometry
High temperature microphotometry is a special op-
tical method, which among other things allows to
identify polymorphous composition of clinker min-
erals alite and belite. In common industrial clinkers,
alite is present mainly in two monoclinic modifica-
tions M1 a M3 or in their mixtures. This informa-
tion is important because clinker with alite modi-
fication M1 can have higher compressive strength
(up to by 10%) than the clinker with M3 modifica-
tion (Staněk and Sulovský, 2002). In belite, high-
temperature modifications (a, a’) can be identified
that have higher hydraulic activity compared to b
modification.
This method combines optical microscope, mi-
crophotometer and a heating stage for automatic
registration of the course of thermal changes in
minerals as well as in inhomogenous materials up
to 1430 C (Chromý, 1974b, Staněk and Mátl, 2003).
The starting modification of clinker minerals alite
and belite can be detected from the mineral bire-
fringence changes with temperature. This requires
quality double-sided polished section of the mea-
sured sample. A slice of material is cut from sam-
ple, mounted on a slide, polished, than the slice
is flipped and re-mounted on a new slide, and
polished on the exposed surface. Polished section
is unstuck by evaporating the resin at high tem-
perature in a fume hood. The polished section is
then placed on a platinum support plate with a
hole that enables the light passage into the heat-
ing stage of the microscope. The intensity of light
(transmitted through optically anisotropic crystal
between crossed nicols) and a relative retardation
of the beam can be utilized for determination of
birefringence using a microphoyometer. However,
the method is sufficiently precise only when small
beam delays are measured and it is comparable
with other methods only if the interference colours
are of the first order. Therefore, the polished section
needs to be as thin as possible.
For the determination of the modification of
alite, which exhibits low birefringence, the suffi-
cient thickness of the thin section is 10 – 15 µm and
the maximum heating temperature 1100 C. Belite
has higher value of birefringence so the thickness
47
 cement. From the observed structure, the properties of raw meal can be estimated, particularly its reactivity. It is
also possible to identify aggregates of belite, formed as residues after big grains of SiO2, or to reveal excessive
amounts of SO3 or MgO in the clinker. Staněk & Sulovský

Figure 1: The structure of clinker in polished section colored by acetic acid fumes. Alite crystals are angular and blue, belite
Figure 1: The structure of clinker in polished section colored by acetic acid fumes. Alite crystals are angular and blue,
grains are oval and brown (in the central part on the right side); grains of free lime are also oval, but orange (in the upper left
belite grains areinterstitial
corner), oval andmass
brown (in the
creates lightcentral part on
small areas, thepores
open rightareside);
blackgrains of free
and closed limeareare
pores also
gray oval,
due but orange (in
to epoxide.
the upper left corner), interstitial mass creates light small areas, open pores are black and closed pores are gray due to
epoxide.

must be 5 – 10 µm; belite has to be heated up to
1430 C.
V. Quantitative phase analysis of
clinker
Clinker granules usually differ in the degree of
burning, therefore, polished sections representing
average phase composition of the clinker must be
used. The whole volume of the clinker sample
is crushed to pass through a 1 mm sieve. The
crushing process has to be repeated and be as mild
as possible not to influence the results by different
grindability of alite and belite (Chromý, 1978). The
0.045 – 1 mm fraction of the crushed clinker sample
is quartered to get the volume that is necessary
to prepare polished section. Crushed clinker is
mixed with mounting liquid (e.g. epoxy resin) in
cylindrical mold made of aluminium film. The
hardened pellet is cut along the cylinder axis to get
a rectangular area of the polished section. The cut
area is firstly ground with SiC papers of different
roughnesses, then polished with diamond paste on
a textile cloth.
Quantitative phase composition in vol. % is de-
termined by a microscopic point counting method
in polished section dyed by fumes of acetic acid.
Volume percentages of clinker phases are propor-
tional to the number of points, which are evenly
distributed. The task is to identify the phase un-
der the crosshair of ocular and presses appropriate
button on the semiautomatic device which registers
the phase and moves the preparation to a position
at given distance.
Precision of the determination depends on the
number of registered points and on the content of
the phase. Presumed deviation of determination
3
can be calculated as standard deviation Sc from
binominal distribution:
r
p · (100 p)
Sc = (1)
n
where p is content of determined phase and n is
total number of registered points in the area.
The influence of human factor has to be taken
into account, i.e. errors in identification of the phase
and errors in registration of points may occur. The
result can be also influenced by the choice of the
material the microscopic specimen consists of. The
main error originates in sampling and the error of
polished section analysis is reflected in the Sc value.
If 2000 points are registered, then the resulting stan-
dard deviation for alite (in 60 – 80 vol. %) is Sc =
1.5 – 1.8 %, for belite and the interstitial mass (in
5 – 20 vol. %) Sc = 0.6 – 1.0 % and for free lime
(around 1 vol. %) Sc = 0.2 %.
Specific weights of clinker minerals (in g·cm 3 )
can be used to recalculate the results from volume
percentage to weight percentage: C3 S – 3.15; C2 S
– 3.28; C3 A – 3.03; C4 AF – 3.77; free lime – 3.35;
periclase – 3.65 (g·cm 3 ).

48
 T. Staněk and P. Sulovský
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure2:2:Detail
Figure Detail of
of clinker
clinkerstructure.
structure.Interstitial mass mass
Interstitial can becan
distinguished here, C3Ahere,
be distinguished creates
C3gray areas and
A creates grayC4areas
AF white
andareas
C4 AF
in interstitial mass.
white areas in interstitial mass.

High temperature microphotometry

VI. Quantitative
High temperaturephase analysisisof
microphotometry a special VII. The amount
optical method, which among of amorphous
other things allowsphase to
identify polymorphous cement
composition of clinker minerals alite and belite. inInslag
commondetermination
industrial clinkers, alite is
present mainly in two monoclinic modifications M1 a M3 or in their mixtures. This information is important
because clinker with alite modification M1 can have higher This compressive
methodstrength (up to by transmitted
uses polarized 10 %) than the clinker
light from
Quantitative phase composition
with M3 modification (Staněk andofSulovský,
cements in vol- In belite, high-temperature modifications (α, α´) can be
2002). petrological microscope. Approximately 100 g of
ume percentage
identified canhigher
that have be determined by microscopic
hydraulic activity compared to βslagmodification.
is dried and milled in a vibration mill. Af-
point counting Thismethod
method too. The phase
combines composition
optical microscope, microphotometer andofa 0.063
heating stagemmfor isautomatic
terwards, fraction – 0.090 extracted
ofregistration
used clinker as well
of the courseasofthethermal
content of each
changes in com-
minerals as well as in inhomogenous materials up to 1430 °C
for examination in polarized transmitted light of
(Chromý,
ponent 1974b, StaněkEpoxy
is determined. and Mátl, 2003). section
polished The starting
is modification of clinker minerals alite and belite can be
microscope. This fraction is placed on a microscope
prepared from the whole grain size range of the ce- temperature. This requires quality double-sided polished
detected from the mineral birefringence changes with
section
ment of theThereby,
sample. measuredthe sample.
method A slice material is cut slide,
is notofinfluenced
so that the
from sample, grains
mounted onare distributed
a slide, polished,evenly and
than the
each can be evaluated separately. Diluted Canadian
byslice is flipped
errors and re-mounted
that come on a new
from different slide, and polished
grindability of on the exposed surface. Polished section is unstuck
by evaporating the resin at high temperature in a fume hood. balsam is then added
The polished sectionand the placed
is then slide isoncovered with
a platinum
the clinker (alite has better grindability than belite
a cover slip.
support plate with a hole that enables the light passage into the heating stage of the microscope. The intensity of
and overburning
light (transmittedcauses
throughworse grindability
optically anisotropicthan un- between crossed nicols) and a relative retardation of the
crystal Slag grains are observed under crossed nicols
derburning). Also errors from different grindability
beam can be utilized for determination of birefringence using using gypsum
a microphoyometer. However, the method is
plate. Amorphous phase is isotropic
ofsufficiently
clinker and components
precise only when aresmall
eliminated (slag has
beam delays are measured and it is comparable with other methods only if
and pink-colored in this configuration of micro-
worse grindability,
the interference gypsum
colours are ofand calcite
the first have
order. better the polished section needs to be as thin as possible.
Therefore, scope. Crystalline grains show different colours.
grindabilityFor thantheclinker). All these
determination errors
of the originate of alite, which exhibits low birefringence, the sufficient
modification
Opaque phase (e.g. coke) is counted to the crys-
inthickness
preparation of theofthin
thesection
microscopic
is 10 – specimen
15 μm and onlythe maximum heating temperature 1100 °C. Belite has higher
talline fraction.
valuea of
from birefringence
specific fraction sowhich
the thickness must be in
is eliminated 5 –this
10 μm; belite has to be heated up to 1430 °C.
Quantitative evaluation can be done by register-
case.
Quantitative phase analysis of clinker ing crystalline and amorphous particles in each
The best Clinker
way to granules
identify usually
the components
differ in thein degree
ce- field
of of view
burning, of microscope
therefore, polishedorsections
by microscopic point
representing
ment is to prepare a polished thin section
average phase composition of the clinker must be used. Thewithout counting methodofasthe
whole volume described above.is crushed to
clinker sample
a pass
coverthrough
slip. Then
a 1 mm thesieve.
sameThe areacrushing
of the thin section
process has to be repeated and be as mild as possible not to influence
can
thebe observed
results in transmitted
by different grindabilityand reflected
of alite light(Chromý, 1978). The 0.045 – 1 mm fraction of the crushed
and belite
ofclinker
a polarization
sample ismicroscope.
quartered to get the volume that is necessary VIII.to prepare
Optical microscopy
polished as process
section. Crushed clinker is
control
mixed with mounting liquid (e.g. epoxy resin) in cylindrical mold made of aluminium film. The hardened pellet tool
Tabulated values of specific weights of clinker
is cut along the cylinder axis to get a rectangular area of the polished section. The cut area is firstly ground with
minerals and compounds are used to recalculate Optical microscopy
SiC papers of different roughnesses, then polished with diamond paste on a textileiscloth.
used in some cement plants
volume percentages to weight percentages. as an additional method. It can give information
Differential thermal analysis (thermogravimetry) about properties and quality of clinker. A big ad-
can be added to microscopic analysis to determine vantage of optical microscopy is the possibility of 4di-
particularly the content of gypsum (CaSO4 ·2H2 O) rect and complex view on key intermediate product
and calcite (CaCO3 ) more precisely. in cement production and also the low investment

49
aggregates shows that the reaction is still in progress. When the reaction of free lime is finished, recrystallization
of silicates takes place and the clinker structure is homogenized. After slight overburning belite clusters are still
visible; they gradually decrease in size and vanish,Staněk & Sulovský
and belite gets dispersed in the prevailing alite (Fig. 3) until
homogenous distribution after strong overburning emerges.

Figure 3:
Figure 3: The
Thestructure
structureof of
overburned clinke.
over-burned Gradual
Figure. vanishing
Gradual of belite aggregates
vanishing and homogenization
of belite aggregates of belite grainsof belite
and homogenization
among alite crystals is apparent.
grains among alite crystals is apparent.
The same happens with free lime aggregates in oversaturated clinker after the end of belite reaction. Furthermore
it is possible to assess structure features like grain size, crystal shape, defects and inclusions, which are closely
and operational
related to nucleationexpenditures. The of
and crystallization disadvantage
alite from liquid. tion. Furthermore
The main effect on itthese
is possible
featurestohasassess structure
content of
minor oxides,sample
is difficult reactivity of raw meal
preparation and (which
humanincludes
error also its grain-size
features like graindistribution
size, crystalandshape,
mineralogical-
defects and
petrographic
during the character
evaluation of of
raw materials), images.
microscopic and the course
Sam- andinclusions,
temperaturewhichof preheating,
are closely burning andtocooling.
related nucleation
According to orientation of lamellae in belite,
pling and processing of observed sample is very the modifications of this phase can be determined.
and crystallization of alite from liquid. Size and The
shapemain
ofimportant,
belite grains and their clusters enable to assess
like with other methods, so that suffi- the properties of silicate component in the raw meal,
effect on these features has content of minor oxides, its
mineralogical nature and relative size of its grains.
cient representativeness and explanatory power is reactivity of raw meal (which includes also its grain-
guaranteed. Method of operative microscopy is be- size distribution and mineralogical-petrographic
Conclusions
ing constantly developed (Chromý, 1992; Chromý, character of raw materials), and the course and
Optical microscopy provides many of possibile uses in research and in production of Portland clinker
2004). Using modern semiautomatic devices can
and cement. It enables the study of starting raw materials as well temperature of preheating,
as of endproducts of the cementburning andIncooling.
industry. the
also facilitate
research sample
practice, preparation.
the whole More
potential of progressis in
this method According
used, while to orientation
in industrial productionof it lamellae
is usually in belite, the
restricted
tothis field liesand
qualitative in the employmentphase
semiquantitative of image analysis
analysis of Portland modifications
clinker. A big of advantage
this phaseof can be method
this determined.
is the Size
for phaseofcomposition
possibility and structure
direct observation evaluation
of materials, which canof contribute
and shape of belite
to solution grains
of the and their
relationship clusters
between enable
their
clinker. to assess the properties of silicate component in the
Modern operating microscopy allows estimating raw meal, its mineralogical nature and relative size
of content of clinker phases and fast determina- of its grains.
6
tion of free lime content with precision similar to
other methods used in practice. Besides modal
analysis, also the structure of burned clinker can IX. Conclusions
be assessed, which is impossible by other methods.
Clinker structure can provide valuable information, Optical microscopy provides many of possible uses
for example about the degree of burning. Typi- in research and in production of Portland clinker
cally, structure with belite and free lime aggregates and cement. It enables the study of starting raw
shows that the reaction is still in progress. When materials as well as of end products of the cement
the reaction of free lime is finished, recrystallization industry. In the research practice, the whole po-
of silicates takes place and the clinker structure is tential of this method is used, while in industrial
homogenized. After slight overburning belite clus- production it is usually restricted to qualitative and
ters are still visible; they gradually decrease in size semiquantitative phase analysis of Portland clinker.
and vanish, and belite gets dispersed in the prevail- A big advantage of this method is the possibility
ing alite (Figure 3) until homogenous distribution of direct observation of materials, which can con-
after strong overburning emerges. tribute to solution of the relationship between their
The same happens with free lime aggregates in composition, properties and genesis. This advan-
oversaturated clinker after the end of belite reac- tage can at the same time a disadvantage, as the

50
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
observer and the evaluator are the same subject.
The greater and more important utilization of op-
tical microscopy in cement-making practice could
be enhanced by implementation of image analysis,
which so far meets many pitfalls.
Acknowledgement
This research was done within the project No.
P104/12/1494 financed by the Czech Science Foun-
dation.
References
Aldrige, L. P. (1975): “International Cement Analysis Study.” Part 2.
“Microscopic Analysis of Clinkers.“ Dept. of Sci. and Ind. Res., Report
No. C. D. 2200, New Zeeland.
Christensen, N. H., Jepsen, O. L., Johansen, V. (1978): “Rate of alite
formation in clinker sandviches." Cement and Concrete Research. Vol. 8,
693 – 702.
Chromý, S. (1974a): “Anfärben des freien CaO und Silikate in an-
schliffen von Portlandklinker.“ Z-K-G 27, 79 – 84.
Chromý, S. (1974b): “High-temperature microphotometry and mi-
crodilatometry. “ Silikáty. Vol. 18, 105 – 123.
Chromý, S. (1976): “Mechanism of white clinker formation. “ Pro-
ceedings of the VI. JCCC, Moscow, Vol. III, 268 – 271.
Chromý, S. (1978): “Accuracy and precision of microscopic qualitative
phase analysis of Portland clinker.“ (in Czech). Silikáty. Vol. 22, (3), 215
- 226.
Chromý, S. (1982): “Prozess der Portlandzementklinkerentstehung.“
“Reaktivität und Brennbarkeit von Zement Rohstoffen.“ Teil I. Z-K-G 35,
204 – 210.
Chromý, S., Hrabě, Z. (1982): “Prozess der Portlandzementklinker-
entstehung.“ “Reaktivität und Brennbarkeit von Zement Rohstoffen.“
Teil II. Z-K-G 35, 368 – 373.
Chromý, S. (1992): “Kontrolle der Qualität von Portlandzementk-
linker durch Mikroskopie in Betrieb.“ Z-K-G 45, 338 – 542.
Chromý, S. (2004): “Operating microscopy – precondition of the
economy and quality of the cement production.“ (in Czech) Sborník XI.
Silichem, Brno, 40 – 44.
Johansen, V. (1979): “Anwendung des Gleichgewichtdiagrams beim
Klinkerbildungvorgang industriell herstelter Klinker.“ Z-K-G 32, 176 –
181.
Kondo, R., Choi, S. (1968): “Mechanism and kinetics of portland ce-
ment formation for and example of the solid state reaction in the pres-
ence of a liquid phase.“ Proceedings of the V. ISCC, Tokyo, Part I, 163 –
171.
Maki, I. (2004): “Formation and microscopic textures of Portland
clinker minerals.“ Ayaragi Minami-machi Shimonoseki, Japan.
Pritula, O., Smrčok, L., Baumgartner, B. (2003): “On reproducibil-
ity of Rietveld analysis of reference Portland cement clinkers.“ Powder
Diffraction Journal 18, 16 – 22.
Schmidt, R., Kern, A. (2001): “Quantitative XRD Phase Analysis.“
World Cement 8, February.
Staněk, T., Sulovský, P. (2002): “The influence of the alite polymor-
phism on the strength of the Portland cement.“ Cement and Concrete
Research 32, 1169 – 1175.
Staněk, T. (2002): “Influence of the minor oxides of SO3 and MgO
on the properties of the Portland clinker.“ (in Czech) Proceedings of the
CEMENT 2002, M-1 – M-9, Vysoké Tatry – Slovensko .
Staněk, T., Mátl, M. (2003): “Observation of the clinker phases
polymorphism by the high-temperature microphotometry method.“ (in
Czech) Proceeding of the XVI. Conference on Thermal Analysis and
Calorimetry TERMANAL 2003, SL-16, Stará Lesná, Vysoké Tatry –
Slovensko.
Staněk, T., Chromý, S., Mátl, L. (2004): “Kinetics of belite clinker
formation.“ (in Czech) Proceedings of the XI. SILICHEM, 174 – 178, Brno.
Weber, M. (1989): “Burning stabilization of the Portland cement
clinker in the rotary kiln.“ (in Czech) Candidate Dissertation, Brno Uni-
versity of Technology, Czech Republic.
51
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

52
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Natural rocks and aggregates

53
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

54
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Fissuring pathology of partly

metamorphosed limestone
Veerle Bams⇤ , Dominique Nicaise
Belgian Building Research Institute
⇤ Veerle.Bams@bbri.be

Abstract

Although the southern part of Belgium has several good quality limestones and sandstones, the market is flooded
with imported exotic natural stones. In contrast to our large experience with the native stones, the characteristics of
the imported natural stone are only known by the laboratory assessments. We lack the knowledge of their behavior in
our climate and our building applications within one or several decades. Over the last years we have been receiving a
significant number of complaints concerning cracking of bluish gray colored compact limestones.
Microscopic analyses show that these compact limestones, in most cases, resemble in fact more a marble than
a limestone. Signs of recrystallization and secondary dolomitization of the calcareous matrix are pointed out by
petrographic analysis. Even though the matrix is that of a marble, the fossils of the original limestone are still
recognizable. The observed recrystallization due to metamorphosis creates less adhesion between the calcite and dolomite
crystals. Other observed features such as pyrite and stylolithes can also be linked to cracking or flaking of the stone
tile. These microscopic observations proof a weakening of the material which makes it more susceptible to frost action,
thermal expansion or mechanical shock.

Keywords: limestone, cracking

I. Introduction

Although the southern part of Belgium has several

good quality limestones and sandstones, the market
is flooded with imported exotic natural stones. In
contrast to our large experience with the native
stones, the characteristics of the imported natural
stone are only known by the laboratory assessments.
We lack the knowledge of their behavior in our
climate and our building applications within one
or several decades. Over the last years we have
been receiving a significant number of complaints
concerning cracking of bluish gray colored compact
limestones.
Microscopic analyses show that these compact Figure 1: Crack in a Vietnamese limestone pavement.
limestones, in most cases, resemble in fact more a
marble than a limestone. Signs of recrystallization
and secondary dolomitization of the calcareous ma- Belgian and Asian limestones: a general compar-
trix are pointed out by petrographic analysis. Even ison
though the matrix is that of a marble, the fossils of
the original limestone are still recognizable. The ob- The Belgian Bluestone is a special facies belonging
served recrystallization due to metamorphosis cre- to the Carboniferous, Dinatian (Tournaisian- about
ates less adhesion between the calcite and dolomite 355 million years) and mined in four regions in the
crystals. Other observed features such as pyrite and southern part of Belgium. The bluestone originates
stylolithes can also be linked to cracking or flaking always from the same geological layers. The petrog-
of the stone tile. These microscopic observations raphy defines the Belgian Bluestone as a compact
proof a weakening of the material which makes it limestone of sedimentary origin and is the result
more susceptible to frost action, thermal expansion of the clustering of innumerable crinoïde- or sea
or mechanical shock. lily residues that are held together with varying

55
 Bams & Nicaise

Figure 2: Crack in a Chinese limestone pavement. Figure 3: Different geological type of Vietnamese lime-
stone

amounts micrite and sparite. The blue color of the

stone can be attributed to the presence of very finely
dispersed carbon. (de Barquin, F., 2001)
The appearance of the Vietnamese limestone is
similar to the Belgian one as the constituent com-
ponents are largely identical. Also in terms of age,
there is a similarity. They are also of Carboniferous
age (Perm - 350-270 million years). The limestone is
mined in the area of Than Hoa (de Barquin, F. 2006)
and comes from dozens of different quarries where
it is exploited from many geological layers. This
result in the existence of a large diversity in the
nature of the limestone which can be a biomicrite
(bioclastic packstone) and can vary up to almost
a marble, called a metalimestone. The limestones Figure 4: Discolouration of Chinese limestone
which are supplied to Europe tend to be a mixture
of stones from different quarries which can lead to
a large difference in color between the delivered to explain the pathology of cracking.
stone elements, ranging from dark blue to light
gray. II. Methodology
Chinese blue coloured limestone differs from the
two other natural stones, as well in composition as For optical investigation of the limestones, thin sec-
in age. It concerns a oolitic grainstone (oosparite) of tions were made. The preparation and finishing
the Middle Cambrian (500 million years). Besides of the thin sections and polished sections was per-
(partly) sparitised oolites, the limestone contains a formed according to NBN EN 12407. The thin sec-
certain amount of dolomite and ferriferous miner- tions have a thickness of 25 to 30 µm and a surface
als, which give this blue coloured limestone a far of 3 ⇥ 5 cm. The samples were impregnated under
different aspect as the stones mentioned above. The vacuum with a fluorescent epoxy resin allowing
oolitic limestone is won in the hills of the Shang- studying the porosity and cracks under reflected
dong province, in the Northeast of China. The fluorescent light. The mineralogy of the stone sam-
extraction is done by dozens of small family en- ples was examined under transmitted light (for the
treprises who excavate, horizontally and vertically, non-opaque minerals).
those hills in small quarries. Various geological lay- Additional characterization of the mineralogy
ers are exploited and collected in a small number was carried out with powder X-ray diffraction
of processing entreprises. (XRD) (Bruker AXS D8-advance diffractometer).
The geological nature of the supplied and fixed The samples were crushed and pulverized. The
Asiatic limestones is rather variable due to way of obtained powder was pressed into a special holder
mining by uncoordinated small family businesses. and placed in the diffractometer. The rotating
The last years an increase of plaints concerning (speed of 30.000 rpm) sample is irradiated with
breaking and fissuring of these limestones is noted. CuKa-radiation (40 kV and 40 mA). The analysis
Microscopic examination was conducted in order was executed with a divergence slit of 1 and covers

56
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

an 2q-area of 10 to 75 with a step size of 0,02 (1s

per step).
To investigate the composition and form of the
opaque minerals an Environmental Scanning Elec-
tron Microscope (SEM) (Quanta 200 FEG) fitted
with an EDS-system (EDAX - Apollo 40 Silicon
drift detector and Genesis XM-4i system) was used.
The ESEM analysis is carried out with a BSE detec-
tor. The specific used conditions are mentioned on
each individual picture.

III. Petrographic analysis

For the uncovering of a possible cause of the crack
formation in the Asian compact limestones, only
the petrographic analysis of these two stones will
be discussed.
Figure 5: Petrographic analysis by transmitted light
and plane polars - overview of the composition and mi-
crostructure of a metalimestone (Vietnamese).

Vietnamese limestone: composition

Generally, this limestone consists mainly of fossils
and allochems in a micritic to sparitic matrix. De-
pending on the degree of metamorphism the con-
stituent parts can whether or not be identified and
is the calcitic sediment converted to microsparite or
sparite. The allochems that can be recognized, as it
is not a metalimestone, are crinoidic residues, shell
fragments, pellets and sporadic coral fragments and
foraminifera. Dispersed in the matrix, dolomite and
opaque minerals can be found. These minerals can
also be observed in the stylolites or in the vicinity of
these features. In thin section, the dolomite crystals
Figure 6: Petrographic analysis by transmitted light
can be identified based on their typical diamond
and plane polars - overview of the composition and mi-
shape.
crostructure of a biomicrite (Vietnamese) with mortar
A XRD-analyse (and Rietveld evaluation method)
beneath.
showed that about 30% of the limestone consists
of dolomite and the remaining 70% is calcite. It
can be assumed that there is a certain amount of
anhedral dolomite crystals present in the matrix
that can not be recognized in a thin section. This
analyis was performed on one of the test samples
(not representative amount for the rock formation).
On basis of an ESEM analysis, it was found that
the opaque minerals present in the matrix (less
than 1%) are pyrite crystals. Individual crystals
show a cubic shape, but they are often present as
spheric microcrystals that not exceed 1 micrometer
in size. Several of these microcrystals form small
agglomerates, in this case it is called framboidal
pyrite.

Vietnamese limestone: microstructure

In some Vietnamese limestones, the original micritic
allochems are converted to (micro)sparite. This
Figure 7: ESEM image of framboidal pyrite.
eliminates the distinct delineation of the particles
and they become unrecognizable in a thin section.
Furthermore in the matrix, calcite crystals with a

57
 Bams & Nicaise

Figure 8: EDS analysis of the framboidal pyrite in Figure

7.
Figure 10: Petrographic analysis by transmitted light
and plane polars – shell fragment in a recrystallized and
dolomitized matrix.

Figure 9: Petrographic analysis by transmitted light and

plane polars – residues of pellets and foraminifera within
a sparitic and dolomitic matrix. Figure 11: Petrographic analysis by transmitted light
and plane polars – overview of the composition and mi-
crostructure of a Chinese limestone.
size up to 50 micrometer can be observed. The
presence of the (micro)sparitization and the relative
large size of the crystals show that a recrystalliza-
tion has occurred and that the calcitic matrix is not
be noted in the middle but these crystals can also
the sparite as in a sedimentary rock.
be present according to the micritic layers of the
The limestone generally has a low porosity, and
oolites. Sometimes the nucleus is a shell fragment
two types can be defined. Firstly, there exists an
around which a (elliptical) oolite has formed. The
intragranular porosity which is formed by the cleav-
oolites are held together by a sparite cement and
age planes in the calcite crystals, and secondly there
microsparite. In some specimens the sparite crys-
is an intergranular porosity present, which is repre-
tals can be up to 250 micrometers and show cal-
sented by space around the calcite crystals.
citic cleavage planes. Dispersed throughout the
In addition, the rock exhibit the presence of fine
matrix opaque minerals can be found but they are
calcitic veins and continuous stylolites that may or
also observed in areas with a high concentration of
may not be open.
dolomite crystals.

Chinese limestone: composition An XRD analysis was performed on a powdered

This natural stone consists mainly of circular or
elliptical oolites with a diameter ranging from 50
micrometers to 2.5 milimeters. They occur with
a cylindrical layer build-up or without this layer
build-up as a result of recrystallization. In the cen-
ter of the oolites often rhombic dolomite crystals
are present. Correspondingly opaque minerals can
stone sample (not representative amount for the
rock formation) in order to be able to estimate the
dolomite content. The analysis indicates that there
is about 71% calcite and 29% dolomite present.
An ESEM analysis identified the in the thin sec-
tion observered opaque mineral as both cubic and
framboïdal pyrite crystals.

58
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 12: Petrographic analysis by transmitted light
and plane polars – dolomite concentration in a sparitic
matrix with large crystals.
Figure 13: ESEM image of framboidal and cubic pyrite
Figure 14: EDS analysis of the pyrite in Figure 13.
Figure 15: Petrographic analysis by reflected, fluorescent
light – porosity around the sparite crystals and in zones
with dolomite concentration
Figure 16: Petrographic analysis by transmitted light
and plane polars – fissure along the sparite crystals.
Chinese limestone: microstructure
Throughout the natural stone fine calcitic veins are
observed. These veins contain more different crys-
tal sizes than the matrix. The present stylolites
are rather short stylolites which are oriented in the
same direction than the continuous styloliets ob-
served in the Vietnamese limestone. Within these
solution surfaces, accumulations of opaque miner-
als are present. Both veins and stylolites can occur
as closed and as open features.
The porosity of the limestone is low and mainly
intergranular. The open veins and stylolites disre-
garded, the porosity can consist of small voids in
the cement and open spaces along the sparite crys-
tals. Furthermore, the oolitic nuclei are relatively
porous by their dolomitisation.
From the analysis of the composition and the mi-
crostructure can be concluded that the geological
nature of the Asian stones can vary from a sedi-
mentary limestone in which still micrite is present
up to a metalimestone which almost can be con-
59
 Bams & Nicaise
sidered as a marble. The Vietnamese limestone is
situated between a dolomite bearing biopelsparite
and a dolomite bearing metalimestone. The Chi-
nese limestone on the other hand, departs from a
dolomite bearing oösparite.
IV. Fissuring pathology
As with all the limestones, the veins and stylolites
are the weak zones of a stone element along which
cracking may occur. The petrographic analysis of
the Asian limestones have revealed a complemen-
tary weakening of the rock. The observed rela-
tively large calcite crystals in the matrix and the
conversion of the micritic allochems to microsparite,
indicate recrystallization of the rock through meta-
morphosis. As a result, there is a certain porosity
present along the sparite crystals. In addition, a sec-
ondary dolomitisation slightly increased the level
of intergranular porosity near the concentrations of
dolomite.
60
The created porousness can be seen as a network
of micro-cracks that makes the rock more brittle.
When these stone elements are applicated as pave-
ment, they will be susceptible to frost action, ther-
mal expansion or mechanical shock. This will lead
to cracking, breaking or flaking. (Nicaise, D., 2010)
V. Conclusion
Not only physical and mechanical testing is im-
portant to gain knowledge of the behavior of a for-
eign natural stone, but microscopic analysis pointed
out that cracking of compact limestones is mostly
caused by a combination of its mineralogical nature
and the unfavorable choice of the stone application.
References
Dreesen, R., Nielsen, P. and Lagrou, D. (2007): “The staining of blue
stone limestones petrographically unraveled” Materials Characterization.
vol. 58, 1070–1081.
de Barquin, F. (2001): “TV 220 - Belgische blauwe hardsteen of "Petit
granit" uit het Tournaisiaanse geologische tijdperk”. BBRI publication
Nicaise, D. (2010): “Deux pathologies spécifiques à la pierre naturelle”
BBRI-dossier 04.17
de Barquin, F., Nicaise, D. and Bams, V. (2006): “TV 228 – Natu-
ursteen” BBRI publications
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Petrography of roofing slate for quality

assessment
V. Cardenes⇤ , V. Cnudde, J.P. Cnudde
Department of Geology and Soil Science-UGCT, Ghent University, Krijgslaan 281, S8, 9000 Ghent, Belgium.
⇤ victor.cardenes@ugent.be

Abstract

The commercial term roofing slates refers to slate and stone for discontinuous roofing and external cladding, according
to the European Norm EN-12326. This group comprises stones such as slate s.s., shale, phyllite, and carbonate slate, but
even schist, cinerite, sandstone, meta-arkose, schistose quartzite, quartzitic flagstone, gneiss and mica schist. However,
the main lithologies used for roofing are the first four.
All these rocks have a well-developed and layered structure in common, allowing them to be split into large, thin
and regular tiles. There are several places in the world where roofing stone could be quarried, but nowadays the main
exploitations are located in Spain, China and Brazil. The difference in nature (petrology and characteristics) of the
roofing stones from different parts of the world plays a major role in the final use, although builders often think that all
stones are equally suited for the same purpose.
Generally speaking, almost all of the World’s roofing slates are mainly composed of quartz, mica and chlorite, together
with some accessory minerals such as iron sulfides, iron oxides and carbonates. The weathering of these two minerals
creates characteristic pathologies, oxidation and gypsification, respectively. Some iron sulfides alteration causes oxidation,
which can be seen as brown to red stain marks over the surface of the tile. Oxidation is far the main cause of complaints
and rejections, even when is just a aesthetical issue. On the other hand, gypsification can lead to the tile destruction,
but it has less impact on the complaints.
Since weathering is related to the mineralogical composition and so indirectly to the origin of the roofing slate,
identifying and quantifying that mineralogy becomes critical to secure a proper use of the slate. In this sense, optical
examination with Polarized Optical Light (POL) using both transmitted and reflected light and X-ray Diffraction are
powerful methods, especially when combined with other techniques, such as Micro Computed X-ray Tomography (MCT).
POL and XRD give information about the type of the rock, its internal structure and mineral components, while MCT
gives information about the occurrence, size and distribution of the metallic minerals, specially the iron sulfides.
This works revises the main types of roofing slates found nowadays in the market, together with their main
characteristics found by POL, XRD and MCT.

Keywords: Roofing slate, petrography, MCT, weathering, quality

I. Introduction $/ton. However, the price for each country depends

The term roofing slate comprises the natural rocks
used in construction which have in common the
capacity to be split into plane, thin and regular tiles
that are used for roofing. Then, these rocks must
have a very defined cleavage (Walsh, 2003), that can
be of sedimentary or metamorphic origin. There
are many rocks with these characteristics (Cárdenes
et al., 2014), but only a few can be mined as roof-
ing slates. The main exploitations of roofing slates
are found in Spain, China and Brazil. In terms of
world’s production of natural stone, roofing slate is
the third stone in volume (Figure 1). Roofing slate
industry is then the economical engine for several
regions in the world, with a volume of exporta-
tions for the year 2013 of 1,128.1 million tons for
a value of 690.8 million $ (UNSTATS, 2014). The
average price of the slate is currently around 612
greatly on factors like the quality and availability
of the slate. For Spain, the main producer, the
average price is 710 $/ton. Spanish slate is well
consolidated in the market, being sold mainly in
Europe. New producing countries, like China (394
$/ton) and Brazil (472 $/ton) have lower prices,
since their product is not completely accepted by
the market, maybe due to its heterogeneity (Cár-
denes et al. 2014). This fact is sometimes due to
a higher hardness combined with a difficult for-
matting with supplementary breakings and losses
during roof coverings. Also the poor quality of the
first imported Chinese slates, with some very high
carbonate content or sometimes with a high poros-
ity combined with their rapid weathering made that
their products are technically less competitive. On
the other hand, countries with a long tradition of
slate mining, like the United Kingdom (745 $/ton)

61
 prices, since their product is not completely accepted by the market, maybe due to its heterogeneity (Cárdenes et
al. 2014). This fact is sometimes due to a higher hardness combined with a difficult formatting with
supplementary breakings and losses during roof coverings. Also the poor quality of the first imported Chinese
slates, with some very high carbonate content Cardenes et al.
or sometimes with a high porosity combined with their rapid
weathering made that their products are technically less competitive. On the other hand, countries with a long
tradition of slate mining, like the United Kingdom (745 $/ton) and Germany (1.551 $/ton), still maintain some
remaining quarries that provide slate for restorations of architectural heritage and special buildings.

Figure 1. Left: volume of world’s exportations of natural stone slabs and tiles, 2013. Right: Evolution of the slate production
by countries for the period 1997-2013. Source: United Nations Statistical Database, UNSTATS.
Figure 1: Left: volume of world’s exportations of natural stone slabs and tiles, 2013. Right: Evolution of the slate
Generally speaking, lithologies correspond to siliciclastic stones with different degrees of
production by countriesmetamorphism.
for the period
Despite 1997-2013. Source:
of the different types of rocks, United Nations
main minerals are alwaysStatistical Database,
quartz, mica and chlorites UNSTATS.
(Ingham, 2005; Ward et al., 2003), with different amounts of accessory (tourmaline, chloritoid, rutile, zircon,
monazite, organic matter, among others) and secondary minerals (mostly iron sulfides and carbonates). In terms
of chemical composition, SiO2 always comprises more than 50% of the total weight of the sample, followed by
and Germany (1.551 $/ton), still maintain some re- conditioned by the amount of organic matter that
Al2O3 and Fe2O3 (Cárdenes et al., 2014). Color of roofing slates greatly vary depending on the type of stone. For
maining quarries that provide slate for restorations can be found. This organic matter is present under
most of them, the color ranges between light and dark grey with a slight blue hue, but it is possible to find green,
red and purple slates. Green color is mainly due to the existence of clinochlore, while red and purple colors are
of architectural heritage and special buildings. the form of graphite and can range up to 2% in
due to the presence of Fe3+. Color essentially reflects both mineral content and chemistry. The darkness of the
some slates (Cárdenes et al., 2012a).
slate is mainly conditioned by the amount of organic matter that can be found. This organic matter is present
Generally speaking, lithologies correspond to
under the form of graphite and can range up to 2% in some slates (Cárdenes et al., 2012a).

siliciclastic stones with different

Weathering of roofingdegrees
slates of meta-
morphism. Despite of the The different
response to weathering of rocks,
types of II. depends
the different roofing slates Weathering
mainly on their modal of roofing slates
mineral
main minerals are always quartz, mica and chlo-
composition. Iron sulfides and carbonates are known (Cárdenes et al., 2012b) to be the most damaging groups of

rites (Ingham, 2005; Ward et al., 2003), with differ- The response to weathering of the different roofing
ent amounts of accessory (tourmaline, chloritoid, slates depends mainly on their modal mineral com-
rutile, zircon, monazite, organic matter, among oth- position. Iron sulfides and carbonates are known
ers) and secondary minerals (mostly iron sulfides (Cárdenes et al., 2012b) to be the most damaging
and carbonates). In terms of chemical composition, groups of minerals for the slate integrity. The most
SiO2 always comprises more than 50% of the to- abundant species of iron sulfides on roofing slates
tal weight of the sample, followed by Al2 O3 and are pyrite and pyrrhotite, and for the carbonate
minerals for the slate integrity. The most abundant species of iron sulfides on roofing slates are pyrite and
Fe2 O3 (Cárdenes et al., 2014). Color of roofing slates ankerite, calcite and siderite (Gómez-Fernández et
pyrrhotite, and for the carbonate ankerite, calcite and siderite (Gómez-Fernández et al., 2009).
greatly vary depending on the type of stone. For al., 2009).
most of them, the color rangesoxidation
Iron sulfides between maylight
produce and red stains Iron
that run sulfides
along theoxidation may produce
slate tile’s surface, red stains
changing the
dark grey withoriginal aspect
a slight of the
blue hue, slate,
butand usually
it is possibleaffecting the slaterun
that tilesalong
beneath.theThisslateoften causes
tile’s an important
surface, changing the
to find green,aesthetical
red andproblem,
purpleeven when the integrity of the cover is not usually involved. Slates with iron sulfides are
slates. Green color original aspect of the slate, and usually affecting
then considered as second or even third choice. Gypsification may cause also change in color. The occurrence of
is mainly due carbonates
to the existence of clinochlore, while the slate tiles beneath. This often causes an impor-
in roofing slates usually lead to gypsification, a process by which the carbonate content reacts with
red and purple colors are due to the presence
the atmospheric sulfates, forming gypsum. Since gypsum of tanthas aesthetical
a higher volume problem, evenits when
than carbonate, growth inthe
theintegrity
3 +
Fe . Color essentially
slate matrix reflects both mineral
causes swellings that can leadcontent of the cover
to tile disintegration. However,iseconomic
not usually involved.
losses due Slates with
to gypsification
and chemistry.areThemuchdarkness
lower than of thosethedueslate is mainly
to oxidation. iron sulfides
From an economical point ofare then
view, considered
the impact as second
of oxidation is more or even
important than gypsification (Figure 2).

Figure 2: Pie graph:

Figure 2.Estimation of theofcost
Pie graph: Estimation ofofthe
the cost twomain
the two main pathologies
pathologies in roofing in roofing
slates slates
(data from (data
customer’s from customer’s
complaints
analysis of a leading company of the sector). Modified from Cárdenes et al 2012b. Images: Left; oxidation stain running
complaints analysis of a leading company of the sector). Modified from Cárdenes et al 2012b. Images: Left; oxidation
along a roof. Right; development of gypsification (white) on the edge of slate tiles, together with some oxidation (red).
stain running along a roof. Right; development of gypsification (white) on the edge of slate tiles, together with some
oxidation (red). Oxidation potential has become then the problem for slate quality, and is probably the main concern for
slate customers. It can be measured directly by using laboratory weathering tests, or indirectly through the
petrographic characterization of the slate itself. According to the slate norm EN 12326 (AEN/CTN_22, 2011),
62 the oxidation potential is estimated by accelerated weathering with the thermal cycling (TC) test, while content
in carbonates is determined by chemical analysis and can be estimated with the sulfur dioxide exposure test. TC
test is based on hydration and de-hydration of the iron sulfides. For the oxidation potential, the accelerated

15th Euroseminar on Microscopy Applied to
Rutile has an AC close to pyrite, but its proportion in the bulk mineralogy of the sample is insignificant (belo
0.01%), so the error induced in the total quantification of iron sulfides can be assumed. On the other han
monazite has a higher AC than pyrite. In roofing slates, this mineral is usually regular in size (around 200 µm
Building Materials • 17-19 June 2015 • Delft, The Netherlands
with a characteristic rounded shape, so it can be easily identified during the POL analysis, and thus
occurrence can be taken into account when doing the MCT analysis. Then, it is always very important to have
good petrographic description of the sample before undertaking the MCT analysis.

third choice. Gypsification may cause also change

in color. The occurrence of carbonates in roofing
slates usually lead to gypsification, a process by
which the carbonate content reacts with the atmo-
spheric sulfates, forming gypsum. Since gypsum
has a higher volume than carbonate, its growth in
the slate matrix causes swellings that can lead to
tile disintegration. However, economic losses due
to gypsification are much lower than those due to
oxidation. From an economical point of view, the
impact of oxidation is more important than gypsifi-
cation (Figure 2).
Oxidation potential has become then the problem
for slate quality, and is probably the main concern
for slate customers. It can be measured directly Figure 3. Attenuation Coefficient of several minerals depending of the X-ray energy.
by using laboratory weathering tests, or indirectly Figure 3: Attenuation Coefficient of several minerals
Iron sulfides can clearly be determined in the slate matrix by means of MCT (Vavro et al., 2014), wi
through the petrographic characterization of thethe additional depending of the X-ray energy.
advantage that this method is not completely destructive, as the others. MCT itself does n
slate itself. According to the slate norm EN 12326 modify any property of the slate, but the sample preparation destroys a certain amount of the original slate b
drilling the cylinders that will be used for the scans. MCT analysis of the iron sulfides of the slates c
(AEN/CTN_22, 2011), the oxidation potential isaccuratelytoshow correctly interpret the results. The disadvantage
the size, shape and distribution of them of the slates’ matrix. However, MCT uses sma
estimated by accelerated weathering with the ther-cylindrical samples of several millimeters of diameter. Also, the information about chemical composition of t
objects analyzed can only be deduced bydone
is that it is always comparing onamongsmallthem. samples,
For a cylindricalso theof 10 mm diamet
sample
mal cycling (TC) test, while content in carbonatesand 10 mm selection of the samples and thus the knowledge ofresolution increase
length, the expected resolution is close to 6 µm voxel. For smaller sample sizes,
is determined by chemical analysis and can be esti-Size, shape and mineralogy are closely linked. According to the Aggregational Model of framboid formatio
(Wilkin the
et al.,material
1997), pyrite is necessary
is formed for some
by aggregation reliable
and pyritization of FeSresults.
compounds. Then, pyrrhoti
mated with the sulfur dioxide exposure test. TC testconstitutes a precursor of pyrite. The framboids have sizes between 5 and 150 µm. Processes described in t
is based on hydration and de-hydration of the ironAggregational Model are held during the sin- and post- depositional stages of the sediment. Later processes, li
Analysis
hydrothermalism, techniques
can alter and even add new foriron iron
sulfides sulfides inthen
to the slate, but roofing
the morphologies are mu
sulfides. For the oxidation potential, the accelerateddifferent. Resulting morphologies are then mainly cubic. The oxidations on slate always generates fro
slates
weathering consists in 20 cycles of immersing therelatively big iron sulfides that can be seen with the naked eye. Small iron sulfides usually lead to the smoo
test pieces in water at (23±5) C for 6 h and thenchanges in color that occasionally occur in some slates (Cárdenes, Mateos, et al., 2012). This slight change
color is One
known of as the twoamong
“tanning” techniques
the miners, and included
is generallyinconsidered
EN 12326 to in the trad
unproblematic
dry them in a ventilated oven at (110±5) C for 17Regarding characterize roofing slate is the petrographic analy- This fact
to oxidation potential, pyrite cubes present lower risk than amorphous pyrrhotite.
h. The report must contain all signs of oxidation,corroborated by miners, which distinguish “silver pyrite” (pyrite s.s.) from “iron pyrite” (pyrrhotite) based o
sis with Polarized Optical Light (POL) microscopy
differences on luster and color. Other problematic iron sulfide is marcasite, which commonly has a bit yellowi
staining, changes of colour of metallic inclusions,with a greenish
in both hue. transmitted and reflected light (also called
swelling, splitting, flaking or exfoliation and the ap- ore microscopy). This technique can give precise
propriate code (T1, T2 or T3). The whole test takes information about the structure, mineralogy and
20 working days. However, experienced quarrymen texture of any building stone (Ingham, 2013), and it
can classify, with a reasonable degree of success, is considered to be the main characterization tech-
the oxidation potential of a given slate. One pecu- nique, not just for slate but for stones in general.
liar method used by quarrymen to determine the The other characterization technique included is
oxidability of the slate consists in urinating several X-ray diffraction (XRD). Other techniques that can
times on the surface. The saline urine readily oxi- provide useful information when combined with
dizes iron sulfides present. In saline environments, the petrological analysis are Scanning Electronic
like the coast, the oxidation is greatly accelerated Microscopy (SEM) together with Energy Dispersive
by saline mist, but the effect of this saline mist is X-ray analysis (EDX), and X-ray Computed Micro-
not considered in the TC test. tomography (MCT). Each technique has its own
Weathering tests are an accurate way to deter- advantages and limitations (Table 1), and should be
mine the oxidation potential, but have some dis- used as a complementary technique.
advantages. These tests take at least one month An accurate characterization of iron sulfides is
to complete, and more important, the weathering then a task for several techniques. POL and XRD
test conditions not always are the same as the final analysis can together estimate the abundance of
conditions for the slate. Thus, a slate with a good iron sulfides, but the data obtained must be taken
result in thermal cycles may resist less long as nor- carefully into account, since both techniques are
mally expected from the test. The results of the acid much dependent on the skills of the technician. In
test or a combination of TC and SO2 -exposure are transmitted POL it is rather common to confuse
necessary. Weathering tests may be misinterpreted organic matter with iron sulfides, since both are
easily (Cárdenes et al. 2012c, Gomez-Fernandez et opaque components. Features such as shape and
al. 2012), but have the advantage of the possibility size can give important clues about what is actually
to check a large number of large surfaces. On the present in the thin section. However, the best esti-
other hand, characterization of the slate is more mation of the abundance of iron sulfides and their
accurate and fast, but needs to be done by a techni- distinction is obtained from Ore Microscopy. XRD
cian with a good knowledge of the material in order is not accurate for minerals with abundance below

63
 Cardenes et al.

Table 1: Characterization techniques used for roofing slates.

Technique Advantages Disadvantages Iron sulfides Resolution

POL
Transmitted light Determination of Some very small Distinction be- Up to 5 µm
Microscopy structure and min- minerals (<3 µm) tween main species
eralogy and distinc- are difficult to char- (pyrite, pyrrhotite)
tion between non acterize
opaque minerals
POL
Reflected light Ore Determination of Some very small Distinction be- Up to 5 µm
Microcopy structure and min- minerals (<3 µm) tween main
eralogy of opaque are difficult to char- opaque species
minerals acterize (pyrite, marca-
site, pyrrhotite,
titano-magnetite,
ilmenite, magnetite,
hematite)
XRD Mineralogical Minerals below 5% Estimation of the Up to 5% mineral
composition of the in abundance diffi- abundance abundance
whole sample cult to characterize
SEM+EDX Close visualization General overview Exact characteriza- <1 µm
of single minerals of the sample is tion of the mineral-
together with their problematic, diffi- ogy
chemical composi- cult to calculate av-
tion erage abundances
MCT 3D information of Depending on their Determination of Up to 1-5 µm of
the structure and composition, some the abundance, voxel size
the distribution of minerals can be dif- shape and distribu-
the metallic miner- ficult to differenti- tion
als ate

⇠5vol%, depending sample/specimen preparation,
instrument settings, and data post-processing by
an experienced operator. SEM+EDX is a good com-
plement of these two techniques, but it should not
be the main technique (Gómez-Fernández et al.,
2009). The information given by these techniques
can be improved by MCT, that incorporates the 3D
analysis, taking into account its limitations in res-
olution. Generally speaking, MCT discriminates
the elements of a sample by their difference in den-
sity and atomic number. These two characteristics
are described by the Attenuation Coefficient (AC)
(Cnudde et al., 2013). In roofing slates, the main
minerals (quartz, mica and chlorites) have different
AC values than the iron sulfides (mainly pyrite), so
it is possible to clearly distinguish them from the
rock matrix (Figure 3). However, for some types
of roofing slates, like phyllites, mica is composed
mainly by biotite, which have an AC close to the
one for iron sulfides. Also, common detrital min-
erals like monazite and rutile, have AC that may
lead to confusion. Rutile has an AC close to pyrite,
but its proportion in the bulk mineralogy of the
sample is insignificant (below 0.01%), so the error
induced in the total quantification of iron sulfides
can be assumed. On the other hand, monazite has
a higher AC than pyrite. In roofing slates, this min-
eral is usually regular in size (around 200 µm) with
a characteristic rounded shape, so it can be easily
identified during the POL analysis, and thus its oc-
currence can be taken into account when doing the
MCT analysis. Then, it is always very important to
have a good petrographic description of the sample
before undertaking the MCT analysis.
Iron sulfides can clearly be determined in the
slate matrix by means of MCT (Vavro et al., 2014),
with the additional advantage that this method is
not completely destructive, as the others. MCT it-
self does not modify any property of the slate, but
the sample preparation destroys a certain amount
of the original slate by drilling the cylinders that
will be used for the scans. MCT analysis of the
iron sulfides of the slates can accurately show the
size, shape and distribution of them of the slates’
matrix. However, MCT uses small cylindrical sam-
ples of several millimeters of diameter. Also, the
information about chemical composition of the ob-
jects analyzed can only be deduced by comparing
among them. For a cylindrical sample of 10 mm di-
ameter and 10 mm length, the expected resolution
is close to 6 µm voxel. For smaller sample sizes,
resolution increases. Size, shape and mineralogy
are closely linked. According to the Aggregational
Model of framboid formation (Wilkin et al., 1997),
pyrite is formed by aggregation and pyritization of
FeS compounds. Then, pyrrhotite constitutes a pre-
cursor of pyrite. The framboids have sizes between
5 and 150 µm. Processes described in the Aggre-
gational Model are held during the sin- and post-
depositional stages of the sediment. Later processes,
like hydrothermalism, can alter and even add new
iron sulfides to the slate, but then the morphologies

64
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 2: Macroscopic description of the slates studied in this work.

Type Sample Origin Stratigraphy Description

Anchimetamorphic BRA Minas Gerais, Bambui Group, Green slate, just in

slate Brazil Ediacaran the limit of meta-
morphism, with no
visible iron sulfides
inclusions

Slate s.s. VDR Villar del Rey, Gevora Unit, Sil- Dark slate, with
Spain urian cubes of pyrite (2-
30 mm)

Carbonate slate ITL Liguria, Italy Val Lavagna Forma- Dark-grey slate
tion, Cretaceous with no visible iron
sulfides inclusions

Phyllite VXE Lugo, Spain Candana Group, Greenish phyllite,

Cambrian coarse grain, no vis-
ible iron sulfides in-
clusions

are much different. Resulting morphologies are III. Methodology

then mainly cubic. The
This workoxidations
presents a POL, on slate
XRD and always
MCT study of the iron sulfides abundance and shape of the four
generates frommost common lithologies
relatively big iron used as roofing slates:
sulfides Samples
shales, slates
that can wereslates
s.s., carbonate selected among the roofing slate col-
and phyllites.
be seen with the naked eye. Small iron sulfides lection deposited at the Department of Geology and
Methodology
usually lead to the smooth changes in color that oc- Soil Science (Ghent University) of Professor Dr. J.P.
casionally occur in somewere
Samples slates (Cárdenes,
selected among theMateos, Cnudde.
roofing slate collection The selection
deposited of theofsamples
at the Department Geology and was made in
et al., 2012).SoilThis
Science (Ghent
slight University)
change in of Professor
color Dr. J.P. Cnudde.
is known order to cover
The selection of the rangewas
the samples of made
rocks in used
order tofor roofing
cover the range of rocks used for roofing worldwide: anchimetamorphic worldwide: slateanchimetamorphic
(BRA), slate s.s. (VDR), slate
carbonate
(BRA), slate
as “tanning” among the miners, and is generally
slate (ITL) and phyllite (VXE). Thus, the samples can be arranged according to the increase of the metamorphic
considereddegree
unproblematic in the trade. Regarding s.s. (VDR), carbonate slate (ITL) and phyllite (VXE).
(Figure 4). The samples were then petrographically characterized according to EN 12326, by thin section
to oxidation potential,
examination andpyrite cubes The
XRD analysis. present lowerexamination
thin section
Thus,included
the samples can be arranged according to the
the calculation of the Mica Stacking
risk than amorphous
Index (MSI), pyrrhotite.
which measures Thisthefact is corrob-
thickness, increase of the metamorphic
abundance and shape of the microscopic mica degree
planes (Figure
at a 4). The
orated by miners, which
magnification distinguish
of 500x. This index is “silver
related pyrite” samples were then petrographically
with the bending strength of the slates (Cárdenes et al., 2010), but is characterized
(pyrite s.s.) only
from valid for slates
“iron pyrite”s.s., since in other lithologies
(pyrrhotite) based on according
it is difficult to EN to12326,
or just impossible by the
distinguish thin
micasection
planes. examina-
POL examination were done at Ghent University (Belgium)tion in anand
optical microscope Zeiss Axioskop
XRD analysis. The thin section A.1, while examina-
differences on luster and color. Other problematic
XRD analysis were done at the Scientific Services of Oviedo University (Spain) following the powder method
iron sulfidewithis marcasite, which commonly has a bit tion included the calculation of the Mica Stacking
a Philips diffractometer PW1830, Cu cathode with a wavelength Kα =1.54051Å. The angular scan was
yellowish with a greenish hue. Index (MSI), which measures the thickness, abun-
recorded from 3º to 65º 2θ with a digital register Philips PW 1710. Obtained data were processed with the
Xpowder software (v. 2004.04.44). For the MCT scans, HECTOR dance (Masschaele
and shape of 2013)
et al., the microscopic
from the Centre for mica planes
This work presents a POL, XRD
X-ray Tomography (Ghent University, and MCT study at ) was used. One cylinder of 5 mm radius and a is related
a magnification of 500 ⇥ . This index
of the iron sulfides abundance and shape of the four
height of 10 mm was drilled out from each sample, and then scanned with the bending strength
at a voxel size of 5.25 of
µm,thewithslates
a tube (Cárdenes
most common voltagelithologies
of 130KeV and used
a tubeas roofing
power of 10W.slates: et al.,were
Resulting images 2010), but is only
reconstructed valid for
using Octopus andslates
analyzeds.s., since in
shales, slates s.s., carbonate slates and phyllites.
with the in-house software Octopus Analysis v.1.1 other lithologies it is difficult or just impossible to
(Vlassenbroeck et al. 2006, Brabant et al. 2011). Image
rendering was performed with the software Avizo Fire, v.8.1.

Figure 4. Top panel: metamorphic facies for pelitic rocks. Bottom panel: macro images of the sample materials investigated
Figure 4: Top panel: metamorphic facies for
here, arranged according pelitic
to their rocks.degree.
metamorphic Bottom panel:
Field of view ofmacro images
each image 10 cm. of the sample materials
investigated here, arranged according to their metamorphic degree. Field of view of each image 10 cm.
Results

POL analysis shows a set of rocks with two clear features, a similar composition for the main minerals
except for the sample ITL (Table 2), and a well-developed cleavage. Besides this, there are differences inherent 65
to the nature of the rock itself, but according to EN 12326, all these rocks can be considered as roofing slates. In
all of the rocks, iron sulfides could be observed, especially in VDR and VXE. Regarding to XRD, the
 Q Mica Chl Cte Cte Py FeOx OM Chlt
BRA 32.3% 57.8% 9.6% + + ++ ++
ITL 11.9% 25.4% 62.7% + +
VDR 19.9% 64.6% 15.4% Cardenes et al. +++ + ++
VXE 28.6% 58.8% 12.6% +

Figure 5.
Figure 5: Optical Optical micrographs
micrographs of theof studied
the studiedsamples.
samples. The estimated
The size of the
estimated sizeironofsulfides
the ironin BRA is ratherinsmall
sulfides BRA (5-20
is rather small
µm) and homogeneous, same as for ITL. On the other hand, VDR presents well developed clusters of pyrite, while VXE have
(5-20 µm) and homogeneous,
Petrographic description same as for ITL. On the other hand, VDR presents well developed
iron sulfides (20-30 µm) uniformly distributed. All images taken in plane polarized light.
clusters of pyrite,
while VXE have iron sulfides (20-30 µm) uniformly distributed. All images taken in plane polarized light.
BRA: Lepidoblastic texture, with grain sizes between 10 and 50 µm. There is a clear alignment following
sedimentary planes (S0). These planes are noticed due to iron oxides accumulations parallel to them (Fig. 5).
distinguish the mica planes. POL examination were a tube power of 10W. Resulting images were re-
Main minerals: Quartz, chlorite, mica and opaque minerals (iron sulfides and organic matter, sizes 5 – 50 µm)
done at Ghent University (Belgium) in an optical constructed using Octopus and analyzed with the
homogeneously distributed all over the slate.
microscope Zeiss Axioskop A.1, while XRD analy- in-house software Octopus Analysis v.1.1 (Vlassen-
sisITL:
were done at the Scientific Services of Oviedo broeck et al. 2006, Brabant et al. 2011). Image
Lepidoblastic texture, grain sizes between 10 and 50 µm, with an evident slaty cleavage (S1) deformation.
University (Spain) following the powder method rendering was performed with the software Avizo
Matrix is composed by carbonate and some quartz grains, surrounded by mica levels well defined (Fig. 5). Main
with a Philips diffractometer PW1830, Cu cathode Fire, v.8.1.
minerals: Carbonate, mica and quartz. No evidence of iron sulfides or other opaque minerals.
with a wavelength Ka =1.54051Å. The angular scan
was recorded
Texturefrom 3 to 65 2q withwith a digital
grainsregis-
VDR: porfiro-lepidoblastic, of quartz and chlorite (50-150 IV.µm)Results
standing out from the mica
ter Philips PW 1710. Obtained data were processed
matrix. Both chlorite and quartz present elongation according to S1 (Fig. 5). Mica are distributed in dense and
with the Xpowder software (v. 2004.04.44). For the
continuous levels. Main minerals: quartz, chlorite and mica,POL analysis
accessory showsiron
minerals: a set of rocks
sulfides with two The
and chloritoid. clear
MCT scans, HECTOR (Masschaele et al., 2013) from
iron sulfides are forming large cubes (up to 5 mm length) that in some parts are being transform into goethite. min-
features, a similar composition for the main
the Centre for X-ray Tomography (Ghent Univer- erals except for the sample ITL (Table 2), and a
sity, www.ugct.ugent.be) was used. One cylinder
VXE: Porfiro-lepidoblastic texture, composed by alternating levels of quartzcleavage.
well-developed Besides(clinochlore)
and green chlorite this, there ofare
of 5 mm radius and a height of 10 mm was drilled differences inherent to the
homogeneous sizes (5-50 µm) Some grains of chlorite have a bigger size that the rest nature
(Fig. 5),ofup thetorock
100 itself,
µm.
out from each sample, and then scanned at a voxel but These
according to minerals
EN 12326,
Main minerals: quartz, chlorite, mica and opaque minerals. opaque areall theseiron
mainly rocks can be
sulfides,
size of 5.25 µm, with a tube voltage of 130KeV and considered as roofing slates. In all of the rocks,
homogeneously distributed all over the slate, with sizes between 10-40 µm.

100

75
Relative intensity

50

25 VXE
VDR
ITL
0 BRA
5 10 15 20 25 30 35 40 45 50 55 60 65
2Theta Angle
Figure 6. Stack with the diffractograms of the four samples. Despite the different nature of the slates, diffractograms are
Figure 6: Stack with the diffractograms of thefor
rather similar, except four
ITLsamples.
(in blue),Despite
due to itsthe different
high nature
carbonate of the slates, diffractograms
content.
are rather similar, except for ITL (in blue), due to its high carbonate content.
The MCT analysis (Fig. 7) shows the abundance and size distribution of the iron sulfides (Fig. 8) with a
voxel size higher than 6 µm. The sample that presents the highest content in iron sulfides is VDR. This slate is
66
known to have a high oxidation potential, produced by large cubes of pyrite. It has the largest size distribution of
pyrite, with sizes ranging up to 1.8 mm. The average volume for the pyrite inclusions is 1340 µm3, and the total
3
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

2 mm 2 mm
VDR

2 mm 2 mm
BRA
Figure 7. Up: Left images are reconstructed slices of the slate volume, while right images correspond to the renderings of
Figure 7: Up: Left images are reconstructed slices
each slate of the
sample slate volume,
showing while right images correspond to the renderings
all the components.
of each slate sample showing all the components.

BRA
iron sulfides could be observed, especially in VDR
and VXE. Regarding to XRD, the composition is
rather similar for all the rocks but for ITL, the car-
bonate slate. In this sample, carbonate is the main
component (Table 3). This type of slate is com-
piled in EN 12326 under the definition "carbonate
slate", for slates with more than 20% volume of
carbonate. Mineral determination was then calcu-
lated by POL and XRD. The main minerals and
the structure were clearly defined with these two
VDR
techniques, but the determination of minor min-
erals was not possible. Iron sulfides content was
only able to be determined by naked-eye estimation
on POL (Figure 5), together with an estimation of
the mineralogical species (pyrite and pyrrhotite).
The XRD (Figure 6) gave no information about the
iron sulfides, even when their presence may cause
weathering problems.
ITL
Petrographic description
BRA: Lepidoblastic texture, with grain sizes be-
tween 10 and 50 µm. There is a clear alignment
following sedimentary planes (So ). These planes
are noticed due to iron oxides accumulations par-
allel to them (Figure 5). Main minerals: Quartz,
chlorite, mica and opaque minerals (iron sulfides
and organic matter, sizes 5 – 50 µm) homogeneously
distributed all over the slate.
VXE
ITL: Lepidoblastic texture, grain sizes between
10 and 50 µm, with an evident slaty cleavage (S1 )
deformation. Matrix is composed by carbonate and
some quartz grains, surrounded by mica levels well
defined (Figure 5). Main minerals: Carbonate, mica
and quartz. No evidence of iron sulfides or other
opaque minerals.
VDR: Texture porfiro-lepidoblastic, with grains of

Table 3: XRD Figure

and POL mineralogical
8. Iron quantification
sulfides size population of the
distribution by mineralogy. The(blue)
number of objects XRDand
assessment applies to main minerals
volume (red).
only, minor and accessory minerals the abundance was visually estimated by POL (+ <1vol%, ++ 1-2vol%, +++
>2vol%). Q: quartz, Chl: chlorite, Cte: carbonate, Py: pyrite, FeOx: Iron oxides, OM: organic matter, Chlt: chloritoid.

Main minerals Secondary minerals Accessory minerals

Sample
Q Mica Chl Cte Cte Py FeOx OM Chlt
BRA 32.30% 57.80% 9.60% + + ++ ++
ITL 11.90% 25.40% 62.70% + +
VDR 19.90% 64.60% 15.40% +++ + ++
VXE 28.60% 58.80% 12.60% +

67
 2 mm
Cardenes
Figure 7. Up: Left images are reconstructed slices of the
slate volume,
each slate sample showing all the components.
BRA
VDR
Figure 8. Iron sulfides size population distribution by number of objects (blue) and volume (red).
Figure 8: Iron sulfides size population distribution by number of objects (blue) and volume (red).
quartz and chlorite (50-150 µm) standing out from
the mica matrix. Both chlorite and quartz present
elongation according to S1 (Figure 5). Mica are
distributed in dense and continuous levels. Main
minerals: quartz, chlorite and mica, accessory min-
erals: iron sulfides and chloritoid. The iron sulfides
are forming large cubes (up to 5 mm length) that
in some parts are being transform into goethite.
VXE: Porfiro-lepidoblastic texture, composed by
alternating levels of quartz and green chlorite
(clinochlore) of homogeneous sizes (5-50 µm) Some
grains of chlorite have a bigger size that the rest
(Figure 5), up to 100 µm. Main minerals: quartz,
chlorite, mica and opaque minerals. These opaque
minerals are mainly iron sulfides, homogeneously
distributed all over the slate, with sizes between
10-40 µm.
The MCT analysis (Figure 7) shows the abun-
dance and size distribution of the iron sulfides (Fig-
ure 8) with a voxel size higher than 6 µm. The
sample that presents the highest content in iron
sulfides is VDR. This slate is known to have a high
oxidation potential, produced by large cubes of
pyrite. It has the largest size distribution of pyrite,
with sizes ranging up to 1.8 mm. The average vol-
ume for the pyrite inclusions is 1340 µm3 , and the
total content is 0.71%. The second slate in iron sul-
fides content is BRA, with an average volume of 400
µm3 and a total content of 0.05%. For this Brazilian
slate, oxidation does not represent a real problem,.
Nevertheless, in some buildings it has developed a
color change, or tanning, like in the new stadium
of Léon, Spain. Finally, ITL and VXE have little and
scarce iron sulfides (240 µm3 , 0.01% and 160 µm3 ,
0.04% average volumes and total content of iron
68
2 mm
BRA
et al. while right images correspond to the renderings of
ITL
VXE
sulfides, respectively). These slates have never pre-
sented documented oxidation troubles at all, even
when ITL has suffered some complaints regarding
troubles caused by their carbonate content.
V. Conclusions
One of the main factors for quality on roofing slates
is the content of iron sulfides. Techniques to deter-
mine the oxidation potential of iron sulfides can be
distinguished in weathering performance tests and
detailed petrographic characterization. Weathering
tests are standardized, but they generally apply
accelerated conditions to reduce experiment time,
but which not necessarily reflect realistic use condi-
tions. MCT analysis of four different roofing slate
varieties were able to characterize two different pop-
ulations of iron sulfides, micro (between 5 and 210
µm size) and macro (generally visible with naked
eye). The micro iron sulfides can be associated with
slight and homogeneous color changes in the en-
tire slate tile surface, while the macro iron sulfides
constitutes potential oxidation spots. From a com-
mercial quality point of view, macro iron sulfides
are a potential source of oxidation, while micro iron
sulfides do not represent a risk. Then, the size and
distribution of the iron sulfides present in a roofing
slate is closely related with the oxidation potential.
Petrological characterization techniques gathered
in the roofing slate norm EN 12326, XRD and POL,
are very limited in terms of deep characterizing the
iron sulfides features. Besides, the measured iron
sulfides contents in the samples ranges from 0.71%
till 0.01%, far away from the minimum mineral con-
tent needed for the XRD determination. Only with
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
MCT the real distribution of the iron sulfides can be
accurately described, so it is suggested to perform
at least a MCT scan in the initial characterization
of a rock as roofing slate. On the other hand, MCT
uses small cylindrical samples of several millime-
ters of diameter, so it is necessary to determine first
if the results are representative. Also, no direct
chemical information is provided. The use of dif-
ferent techniques combined (POL, SEM and MCT,
together with weathering tests from EN 12326) can
accurately determine the distribution, size and oxi-
dation potential of iron sulfides on roofing slates.
References
AEN/CTN_22. (2011). EN 12326. Slate and stone products for discontin-
uous roofing and cladding - Parts 1 (Product Specification) and 2 (Methods of
test).
Brabant, L., Vlassenbroeck, J., De Witte, Y., Cnudde, V., Boone, M.
N., Dewanckele, J., & Van Hoorebeke, L. (2011). "Three-Dimensional
Analysis of High-Resolution X-Ray Computed Tomography Data with
Morpho+." Microscopy and Microanalysis, 17(2), 252-263.
Cárdenes, V., Prieto, B., Sanmartín, P., Ferrer, P., Rubio, A., & Monter-
roso, C. (2012a). "The influence of chemical-mineralogical composition
on the color and brightness of Iberian roofing slates." Journal of Materials
in Civil Engineering, 24(4), 460-467.
Cárdenes, V., Mateos, F., Rubio Ordoñez, A., & Monterroso, C.
(2012b). "Standard tests for the characterization of roofing slate patholo-
gies." Materiales de Construcción, 62, 251-268.
Cárdenes, V., Monterroso, C., Rubio, A., Mateos, F. J., & Calleja, L.
(2012c). "Effect of freeze-thaw cycles on the bending strength of roofing
slate tiles." Engineering Geology, 129-130, 91-97.
Cárdenes, V., Rubio-Ordoñez, A., López-Munguira, A., de la Horra,
R., Monterroso, C., Paradelo, R., & Calleja, L. (2010). "Mineralogy and
modulus of rupture of roofing slate: Applications in the prospection and
quarrying of slate deposits." Engineering Geology, 114, 191-197.
Cárdenes, V., Rubio-Ordoñez, A., Wichert, J., Cnudde, J. P., &
Cnudde, V. (2014). "Petrography of roofing slates." Earth-Sciences Reviews,
138, 435-453.
Cnudde, V., & Boone, M. N. (2013). "High-resolution X-ray computed
tomography in geosciences: A review of the current technology and ap-
plications." Earth-Science Reviews, 123, 1-17.
Gómez-Fernández, F., Castaño, M. A., Bauluz, B., & Ward, C. (2009).
"Optical microscope and SEM evaluation of roofing slate fissility and
durability." Materiales de Construcción, 59(296), 93-106.
Gómez-Fernández, F., Castañón, A. M., & Ward, C. R. (2012). "Anal-
ysis of the methodology of the petrographic examination test (European
Standard EN 12326-2) and the relation between petrography and modu-
lus of rupture for Spanish roofing slates." Engineering Geology, 141–142(0),
114-123.
Ingham, JP (2005): Characterisation of roofing slate by optical mi-
croscopy. In: Hughes, JJ, Leslie, AB, and Walsh, JA (editors): Proceed-
ings of the 10th Euroseminar on Microscopy Applied to Building Materials,
June 21-25, Paisley, Scotland: pp18.
Ingham, J. (2010). Geomaterials under the microscope. London: Manson
Publishing.
Masschaele, B., Dierick, M., Van Loo, D., Boone, M. N., Brabant, L.,
Pauwels, E., Cnudde, V., & Van Hoorebeke, L. (2013). "HECTOR: A
240kV micro-CT setup optimized for research." 11th International Confer-
ence on X-Ray Microscopy (Xrm2012), 463.
UNSTATS. (2014). http : //unstats.un.org.Available : http :
//unstats.un.org.
Vlassenbroeck, J., Dierick, M., Masschaele, B., Cnudde, V., Van Hoore-
beke, L., & Jacobs, P. (2007). "Software tools for quantification of X-ray
microtomography." Nuclear Instruments & Methods in Physics Research Sec-
tion a-Accelerators Spectrometers Detectors and Associated Equipment, 580(1),
442-445.
Vavro, M., Soucek, K., Danek, T., & Stas, L. (2014). "Selected non-
destructive methods suitable for evaluation of roofing slate." Advanced
Materials Research, 923, 63-70.
Walsh, JA (2003): The relationship between the cleavage properties
of natural roofing slate and the thickness of the finished material. In:
Broekmans, MATM, Jensen, V, and Brattli, B (editors): Proceedings of the
9th Euroseminar on Microscopy Applied to Building Materials, Trondheim,
Norway, cd-rom: pp16.
Ward, C., & Gómez-Fernandez, F. (2003). "Quantitative mineralogical
analysis of spanish roofing slates using the Rielveld method and X-ray
powder diffraction data." European Journal of Mineralogy, 15, 1051-1062.
Wilkin, R. T., & Barnes, H. L. (1997). "Formation processes of fram-
boidal pyrite." Geochimica et Cosmochimica Acta, 61(2), 323-339.
69
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

70
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Some aspects of petrography of burnt

colliery spoil
Vincent Thiéry⇤,a , Bernard Guyb
a Dpt. Génie Civil et Environnemental, École des Mines de Douai, Douai, France.
b SPIN/Géosciences, École des Mines de Saint-Étienne, Saint-Étienne, France.
⇤ vincent.thiery@mines-douai.fr

Abstract

Burnt colliery spoil is a typical waste material from former coal mining regions (e.g. North of France, Tchelyabinsk
coal basin in Russia, Silesia in Poland. . . ) where mining residues are stored in heaps. When such coal tips are caught
on fire, the materials undergo burning of organic residues, baking and sometimes melting and recrystallization. The
resulting product, which is generally red but also yellowish, orange or violet, finds various uses as a building material:
roadmaking, backfill, decorative aggregate for private domestic roads, sand for tennis court; it has also been used in
the past to design synthetic aggregates (Gutt and Nixon, 1979). Based on a set of samples from various localities in
France (Guy et al., 2001; Masalehdani et al., 2009; Masalehdani and Paquette, 2013), we discuss the petrography and
mineralogy of burnt colliery spoil. The highly heterogeneous initial leftover material from the coal mining (sandstones,
coal dust, pelites, limestone. . . ) as well as the occasional use of tips as garbage repositories is at the origin of a high
diversity of textures within the materials. The high temperatures (800-1000 C, sometimes more) reached during the
burning of the heaps leaves no remains of organic matter and are associated with a mineralogical evolution (clay
dehydration, oxidation of sulfides. . . ) and most often with the preservation of the initial rock texture. The local melting
of those rocks lead to the formation of glassy products characterized by a high amount of amorphous content. All those
mineralogical evolutions, presenting some similarities with the manufacture of bricks, contribute to the hardness of the
material and to its characteristics because of its sintering. Finally, secondary minerals (gypsum, zeolites) fill fractures
(Masalehdani et al., 2009).

Keywords: coal mining waste, petrography, annealing, baking, aggregate

I. Introduction Thermoscientific Ultradry EDX detector.

Coal mining is probably the best known type of

mining within the general knowledge, associating
the black of coal with the dirtiness of dust and the
huge coal tips (or coal heaps) close to the mining
shaft (Figure 1). Colliery spoil, also known as mine-
stone (Skarżyńska, 1995), is the typical leftover from
coal extraction. The extracted amount of unused
material can be of one ton of colliery spoil for each
ton of coal (Wiggering, 1993).
Typically, residues from the coal treatment are a
waste (La Nauze and Duffy, 1985) which may be
interesting to re-use. Thus, a thorough characteri-
zation of what is found in a heap may yield some
interesting prospects for potential applications.
II. Materials and methods
Thin sections were produced according to standards
procedures (e.g. Humphries, 1992) and studied
using a Leica DMR XP optical microscope. SEM-
EDX has been carried out on a Hitachi S-4300SE/N
SEM working in high vacuum mode, coupled to a
Spoil heaps characteristics
In order to understand the complex, unusual and
highly variable mineralogy which results from the
burning of coal heaps, it is necessary to present
coal tips characteristics. A coal tip is a heap col-
lecting all the leftovers from the coal exploitation:
sterile rocks (sandstones, arkoses, shales. . . : Figure
2) which may contain organic matter and various
by-products (old metallic elements such as railways,
demolition wastes, fly ashes. . . ). Historically, the
first heaps were flat shaped and did not reach im-
portant heights because the leftovers were simply
thrown by man from wagons. The separation pro-
cesses were not very efficient and thus old heaps
are still rich in coal (Limacher, 1963) and are thus
more likely to be affected by burning. Technical pro-
gresses led to the elevation of typical cone-shaped
heaps using conveyor belt systems. Moreover, im-
provements in the washing of coal led to an im-
poverishment of the coal dust rate in coal heaps.

71
 Thiéry & Guy

Figure 1: Typical conic coal tips from the North of France (Loos-en-Gohelle). The two tips at the background, known
as the “twin tips” are among the highest in Europe and culminate at ca. 130 meters above the surrounding plain. At
the foreground lies a quarry corresponding to the emplacement of two former tips; both unburnt (black) and burnt
(red, at the foreground of the heap) colliery spoil is exploited.

Colliery spoil production can account for half of
the production of a mine (Wiggering, 1993). For
example, in Britain, at the dawn of the eighties,
there were 3000 million tons of colliery spoil avail-
able, with a production rate of 56 million tons/year
(Sherwood, 1979).
Spoil heaps combustion
Apart from the provoked combustion of coal heaps
in order to transform them into valorizable mate-
rials (see below); the combustion of a coal heap
occurs generally accidentally. Lightning strikes, a
nearby forest fire are frequent cases but there is also
the case of electrical short-circuit in underground
cables (Nichol and Tovey, 1998).
Spontaneous combustion of carbon-rich residues,
and thermal input by the exothermic oxidation of
pyrite are the inner phenomenon. Moreover, the
poorly compacted nature of a coal heap will facili-
tate air penetration in its deeper parts. Moldiness
on wood fragments, tissue residues. . . can also
account for an increase in temperature (Limacher,
1963). It has been estimated in the Nord-Pas-de-
Calais basin that one heap out of three ignited,
which, in the early eighties, meant 74 burning heaps
(Ghouzi, 1982). In any case, ignition of coal heap
fires is a complex phenomenon, described and dis-
cussed by Misz-Kennan & Fabiańska (2011).

Petrography

In the inner parts of a burning coal heap, temper-

Figure 2: Synthetic log presenting rocks accompanying
coal in a deposit, which will typically be disposed as
waste after coal exploitation (after Skarżyńska, 1995).
ature can exceed 1000 C and even reach 1300 C
(Grapes, 2006; Guy et al., 2001; Sokol et al., 2002),
at an almost ambient pressure. At such tempera-
tures, mineralogical transformations are likely to
occur. This phenomenon is known as pyrometamor-
phism, more specifically “anthropogenic pyrometa-
morphism” (Grapes, 2006). In the Russian literature
as well as in eastern Europe it is common to name
this as “combustion metamorphism” (Sokol, 2005;
Žáček et al., 2014).
The terminology of baked rocks and products
formed during the combustion of coal tips is quite
confusing and sometimes contradictory. Red, un-
melted baked rocks are termed “clinker” or “glassy
clinker” when they are melted but they do not
display evidence of flow (Grapes et al., 2009). “Par-
alavas” are melted vesicular rocks and can show
some evidences of flow (Grapes et al., 2011).
Thus, due to both an extreme variability within
the initial products, the nature (melted or not) and
the heat and atmosphere variability within the burn-
ing heap, extremely contrasted products can be
found in burnt colliery spoil (Figure 3).

72
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Unmelted, baked and annealed rocks material which vary within meters, and also on the
type of gas resulting from the burning of the heap.
In the simplest cases, the thermal transformations
Paralava’s aspect is typically very close to that
affecting the rocks from the heap will consist on
of a natural lava, hence its name (Fermor, 1918).
a “baking” during which the initial texture (shale,
Various types can occur but they are typically vesic-
sandstone) is preserved. This can be very spectacu-
ular with a ropy aspect. In the samples from the
lar in the cases where fossils are preserved. Many
present study, the porosity can be very high, ca.
similarities can be found between the baking of
half of the material consists of vesicles (Figure 5a).
argillites or shales and brick production (Grapes,
The microtexture reveals that a large part of the
2006). Typically, baked shales, argillites or sand-
material is glassy; in this matrix, some minerals can
stones, will be macroscopically similar to what they
be observed. Skeletal, slightly green spinel (Figure
were before baking: bedding is preserved (Figure
5b, c) is associated with elongated (up to 150-200
4 a, b). Investigations under optical microscope
micrometers long) mullite needles (Figure 5c, d and
using moderate magnifications are in good agree-
Figure 6a).
ment with this macroscopic observation. However,
The glass composition, revealed by EDX analysis,
under higher magnifications, the matrix appears
is rich is Si and Al with slight amounts of K, Ca,
very close to that of bricks (Ingham, 2011) and fired
Na, Mg and Fe (Figure 6b). Exception made of Fe,
clays (Figure 4c). Mica flakes are affected by the
this composition is quite close to that of a felspdar.
heating and coexist with a diffusive glassy matrix
EDX analysis on spinel crystal (Figure 6c) ca. 10
in which iron oxides are abundant (Figure 4d).
micrometers across reveals that this is an Al-Fe-Mg
spinel, i.e. hercynite- spinel s.s.
Melted rocks: paralavas

Melted rocks from burnt spoil heaps are termed
“paralavas”, sometimes they are also referred to
as “parabasalts” when their chemistry is close to
that of a basalt (Sharygin et al., 1999). They are
very complex rocks, their composition, and thus
their mineralogy, is dependent on the initial type of
III. Use of colliery spoil
Colliery spoil, available in huge amounts in coal
mining areas, has been for a long time considered
as a waste to reuse (La Nauze and Duffy, 1985). This
concern the global material contained in the coal

Figure 3: Macroscopic samples from various localities. a) baked sandstone, Rieulay, North of France; b) baked argillite,
same provenance; c) preserved fossil leaves on a baked shale, Germignies, North of France; d) porous paralava with
flow texture, Rieulay, North of France; e) ropy paralava, la Ricamarie, center of France; f) extremely vacuolar paralava,
same provenance.

73
 Thiéry & Guy

Figure 4: Baked argillite, exact location unknown, North of France. a) general view (scan of a thin section); b)
microscopic view plane polarized light, showing the preserved bedding; c) microtexture consisting mainly of quartz
grains in a more or less fused fire clay matrix, crossed polarized light; d) detail of the matrix showing indistinct grain
boundaries between quartz and fired clay. Poorly preserved clayey flakes are visible and show weak birefringence. The
glassy matrix contains abundant iron oxides.

tips, both burnt and unburnt. This all-in material
has served to infill former mining works (galleries,
pits. . . ), sometimes along with lime and fly ash
(Ghataora and Jarvis, 1996; Karfakis et al., 1996).
Prospects for a more general use of colliery spoil,
regardless of its burnt or unburnt nature, include
embankments, foundations (Ferreras Cadierno et
al., 2014), and, mixed with Portland cement, only
materials for which minor mechanical strength is
needed seems to be achievable (Okagbue and Ochu-
lor, 2007).
However, the special nature of burnt colliery spoil
has also attracted civil engineers because of its hard-
ness. Thus, its potential use as an aggregate in
concrete have been investigated (Runguphan and
Guthrie, 2009). Those authors, using a substitution
of natural aggregates with 60 % of burnt colliery
spoil, have achieved a concrete showing a 39 MPa
of unconfined compressive strength at 56 days.
Some prospects may be found in the very special

Figure 5: Some examples of the microtexture from a paralava, la Ricamarie, France. a) general view (scan of a thin
section) illustrating the high porosity of the rock; b: dendritic spinel (Spl) crystals in a glassy (Gl) matrix, plane
polarized light; c) elongated mullite (Mul) crystals, confirmed by EDX analysis (plane polarized light); d) elongated
mullite crystals (crossed polarized light).

74
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 6: EDX spectra for a) mullite, b) glass and c) spinel from the glassy paralava.
case of paralavas. Those ones being more or less
amorphous and rich in Si and Al may potentially
be interesting candidates for geopolymerization or
any kind of alkali activation.
Finally, it is worth mentioning a very famous
use of crushed burnt colliery spoil: it is indeed
used on tennis courts (the famous red sand), the
Rolland-Garros court having been supplied by a
former quarry in the North of France (Robaszynski
and Guyétant, 2010).
References
Fermor, L.L. 1918. Preliminary note on the burning of coal seams at
the outcrop. Trans. Min. Geol. Metall. Inst. India, Vol. 12 No., 50-63.
Ferreras Cadierno, J., Guerra Romero, M.I., Valdés, A.J., Morán del
Pozo, J.M., García González, J., Rodríguez Robles, D., Vallepuga Es-
pinosa, J. 2014. Characterization of Colliery Spoils in León: Potential
Uses in Rural Infrastructures. Geotechnical and Geological Engineering,
Vol. 32 No. 2, 439-452.
Ghataora, G.S., Jarvis, S.T. 1996. Use of colliery spoil for infilling
mine workings. Waste Management, Vol. 16 No. 1-3, 189-193.
Ghouzi, D. 1982. The case of the French Nord/Pas-de-Calais coalfield.
Minerals and the Environment, Vol. 4 No. 2-3, 67-74.
Grapes, R. 2006. Anthropogenic Pyrometamorphism. Pyrometamor-
phism, pp. 191-218. Springer.
Grapes, R., Korzhova, S., Sokol, E., Seryotkin, Y. 2011. Paragenesis
of unusual Fe-cordierite (sekaninaite)-bearing paralava and clinker from
the Kuznetsk coal basin, Siberia, Russia. Contributions to Mineralogy
and Petrology, Vol. 162 No. 2, 253-273.
Grapes, R., Zhang, K., Peng, Z.-l. 2009. Paralava and clinker products
of coal combustion, Yellow River, Shanxi Province, China. Lithos, Vol.
113 No. 3-4, 831-843.
Gutt, W., Nixon, P.J. 1979. Use of waste materials in the construction
industry. Matériaux et Construction, Vol. 12 No. 4, 255-306.
Guy, B., Garcia, D., Le Coze, J., Theisse, V. 2001. Estimation de la
température atteinte dans le crassier Saint-Pierre de la Ricamarie (Loire)
à partir de l’observation de textures de solidification/oxydation dans
un élément de chaîne métallique prélevé sur le site. Les cahiers de la
Rotonde - Publications de l’Université de Saint-Étienne, Vol. No., 61-69.
Humphries, D.W. 1992. The preparation of thin sections of rocks,
minerals, and ceramics. Oxford University Press. 83 pp
Ingham, J.P. 2011. 9 - Bricks, terracotta, and other ceramics. Geoma-
terials Under the Microscope, pp. 163-170. Manson Publishing Ltd.
Karfakis, M.G., Bowman, C.H., Topuz, E. 1996. Characterization of
coal-mine refuse as backfilling material. Geotechnical & Geological En-
gineering, Vol. 14 No. 2, 129-150.
La Nauze, R.D., Duffy, G.J. 1985. Coal rejects - a wasted resource?
Environmental Geochemistry and Health, Vol. 7 No. 2, 69-79.
Limacher, D. 1963. A propos de la formation de minéraux secondaires
lors de la combustion des charbons. Annales de la Société Géologique
du Nord, Vol. 83 No., 287-288.
Masalehdani, M.N.N., Mees, F., Dubois, M., Coquinot, Y., Potdevin,
J.-L., Fialin, M., Blanc-Valleron, M.-M. 2009. Condensate minerals from
a burning coal-waste heap in Avion, Northern France. The Canadian
Mineralogist, Vol. 47 No., 573-591.
Masalehdani, M.N.N., Paquette, Y. 2013. Chapter 8 - Coal Fires of
Northern and Massif Central, France. Coal and Peat Fires: A Global
Perspective, pp. 121-136. Elsevier, Boston.
Misz-Kennan, M., Fabiańska, M.J. 2011. Application of organic petrol-
ogy and geochemistry to coal waste studies. International Journal of
Coal Geology, Vol. 88 No. 1, 1-23.
Nichol, D., Tovey, N.P. 1998. Remediation and monitoring of a burn-
ing coal refuse bank affecting the Southsea Looproad at Brymbo, North
Wales. Engineering Geology, Vol. 50 No. 3-4, 309-318.
Okagbue, C.O., Ochulor, O.H. 2007. The potential of cement-
stabilized coal-reject as a construction material. Bulletin of Engineering
Geology and the Environment, Vol. 66 No. 2, 143-151.
Robaszynski, F., Guyétant, G. 2010. Des roches aux paysages dans le
Nord-Pas-de-Calais : Richesse de notre patrimoine géologique. 152 pp
Runguphan, T., Guthrie, P.M. 2009. Investigation on the use of burnt
colliery spoil as aggregate in low to normal strength concrete. Excellence
in Concrete Construction through Innovation, pp. 559-566. Crc Press-
Taylor & Francis Group, Boca Raton.
75
 Thiéry & Guy
Sharygin, V.V., Sokol, E.V., Nigmatulina, E.N., Lepezin, G.G., Kalugin,
V.M., Frenkel, A.E. 1999. Mineralogy and petrography of technogenic
parabasalts from the Chelyabinsk brown-coal basin. Russian Geology
and Geophysics, Vol. 40 No. 6, 879-899.
Sherwood, P.T. 1979. A review of the use of mineral wastes and by-
products in road construction. Minerals and the Environment, Vol. 1 No.
4, 127-130.
Skarżyńska, K.M. 1995. Reuse of coal mining wastes in civil engineer-
ing - Part 1: Properties of minestone. Waste Management, Vol. 15 No. 1,
3-42.
Sokol, E. 2005. Combustion metamorphism (in Russian). Publishing
house of the Siberian branch of Russian Academy of Science, Novosi-
birsk. 250 pp.
76
Sokol, E.V., Sharygin, V., Kalugin, V.M., Volkova, N.I., Nigmatulina,
E.N. 2002. Fayalite and kirschsteinite solid solutions in melts from
burned spoil-heaps, South Urals, Russia. European Journal of Miner-
alogy, Vol. 14 No., 795-807.
Wiggering, H. 1993. Sulfide oxidation - an environmental problem
within colliery spoil dumps. Environmental Geology, Vol. 22 No. 2,
99-105.
Žáček, V., Skála, R., Dvořák, Z. 2014. Combustion Metamorphism
in the Most Basin, Czech Republic. In G. Stracher, A. Prakash, and E.V.
Sokol, Eds). Coal and Peat Fires: A Global Perspective. Volume 3: Case
Studies – Coal Fires, pp. 161-202.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Nanolime deposition in Maastricht

limestone: back-migration or
accumulation at the absorption surface?
G. Borsoia , R. van Heesb , B. Lubellib , R. Veigac , A. Santos Silvac
a Heritage & Architecture Section, Faculty of Architecture, Delft University of Technology, Delft, The Netherlands
b Division of Structural Reliability, TNO, Delft, The Netherlands
c Building Division and Material Division, National Laboratory of Civil Engineering - LNEC, Lisbon, Portugal
⇤ G.Borsoi@tudelft.nl

Abstract

The development of nanomaterials is growing exponentially in the last decade. New discoveries and their applications
have taken place in different fields, such as electronics, chemistry, biology and physics. Some innovative nanoproducts
have become available also for the conservation of cultural heritage (e.g. nano-Ca(OH)2 , nano-SiO2 , nano-TiO2 , CNTs,
etc.), showing interesting potentialities and advantages. However, there are also drawbacks.
Extensive literature is available on nanolimes, i.e. colloidal alcoholic dispersions of Ca(OH)2 nanoparticles, which
are promising materials for the consolidation of lime-based plasters and calcareous stone, often present in ancient
masonry structures. Nanolimes have several advantages such as high stability, fast carbonation rate and good chemical
and physical compatibility with calcareous stone and lime-based plasters. However, in comparison to other existing
consolidant products, like TEOS-based (Tetraethylorthosilicate), the penetration depth of nanolime is often limited,
which may consequently result in a low effectiveness.
The research presented in this paper aimed at understanding the penetration and deposition process of nanolime when
applied in Maastricht limestone, a soft, highly porous substrate, widely used in the Dutch and Belgian provinces of
Limburg. This research focused on the nanolime transport process to clarify whether nanolime is able to penetrate and
deposit in depth in the material or if it simply accumulates near the surface. In order to investigate the distribution
and particle size of nanolime deposited in the stone, next to simple phenolphthalein tests, optical microscopy, AFM and
SEM-EDS were used. On the basis of the obtained results, the transport and deposition of a commercial nanolime in
Maastricht limestone has been elucidated. The comprehension of the nanolime transport mechanism (penetration and
distribution) within the treated material is crucial for improving the effectiveness of this consolidation treatment.

Keywords: Nanolime, Consolidation products, Transport, Penetration depth, Limestone

I. Introduction application of a consolidant, which can recover the

Calcareous stone (e.g. limestone) and lime-based
finishes (e.g. renders, plasters and mural paint-
ing) are often subjected to several degradation pro-
cesses (e.g. salt crystallization, frost action, biolog-
ical growth); these factors that can lead to decay
patterns i.e. powdering, sanding, scaling or chalk-
ing (Figure 1a,b).
Interventions on these kinds of historical sub-
strates anyway differ significantly; different theo-
retical approaches and practical recommendations
are found in the European context. A restoration,
meant recover the loss of cohesion of materials, can
include pre-consolidation (to avoid material loss in
the following phases), cleaning (e.g. by poultice
methods or laser ablation) and desalination (e.g.
with poultice methods).
These processes are generally followed by the
lost cohesion; consolidation can easily stabilize su-
perficial decay forms as powdering, sanding and
chalking; pathologies as exfoliation, scaling or flak-
ing can generally not be treated.
Consolidants are usually applied to stone, ren-
ders or plasters by brush, spray or by immersion
and penetrate into the treated material by capillary
action (Doehne 2010). Anyway the application of
a consolidant is a physically irreversible treatment
and is therefore a critical treatment that has to be
properly studied and well conceived. Consolida-
tion is in fact a risky operation and it is not difficult
to find situations where consolidation has left a
negative imprint and serious drawbacks (Delgado
2010).
A consolidation treatment should guarantee three
fundamental requirements: effectiveness (i.e. im-
provement of the mechanical strength) compatibil-

77
 Borsoi et al.

Figure 1: (a) Loss of cohesion, powdering and traces of past interventions are visible in this masonry of Maastricht
Limestone (Keverborg castle, Kessel, The Netherlands); b) detachment, loss of cohesion and material in coloured
lime-based mortars applied on historical stone masonry (Alcantâra Convent, Lisbon, Portugal)

ity (with the treated substrate) and long term dura-
bility (resistance to different damage mechanisms)
(Hansen, 2003; Toniolo, 2010; Van Hees, 2014). The
concept of compatibility of an intervention has nowa-
days replaced that of reversibility (Teutonico, 1997;
Van Hees, 2014). Products used for conservation
issues should be compatible, from the mechanical,
chemical physical and aesthetical point of view,
with the substrate on which they are applied.
Inorganic lime-based consolidants are in theory
more suitable for limestone than the organic ones
(as TEOS-based consolidants), because of their bet-
ter compatibility [Doehne 2010]. The historically
most known inorganic lime-based consolidant prod-
uct is limewater, a Ca(OH)2 aqueous solution. Lime-
water, however, has a poor penetration depth and
effectiveness, because of the low solubility of cal-
cium in water.
Alternatives to limewater have been explored
over the last decade, as the use of alcoholic dis-
persions of calcium hydroxide nanoparticles, com-
monly known as nanolimes (Giorgi, 2000; Dei, 2006).
Nanolimes are colloidal alcoholic dispersions with
high stability and high lime concentration, facts
that in theory would improve the consolidating ac-
tion (Chelazzi, 2013; Ziegenbalg, 2010). Nanolimes
are usually synthesized under controlled conditions
by homogeneous phase reaction (Salvadori, 2001;
Dei, 2006) or solvothermical reaction (Favaro, 2008;
Ziegenbalg, 2010; Poggi, 2012). Calcium hydroxide
nanoparticles have a size between 50 and 600 nm.
This lime-based colloidal dispersion has been ap-
plied in conservation for the pre-consolidation and
for the recovery of the superficial cohesion (Am-
brosi, 2001; Dei, 2006; Poggi, 2012; Borsoi, 2012;
Baglioni, 2013; Chelazzi, 2013; Rodriguez-Navarro,
2013, Carretti, 2014). However, when mass con-
solidation is required, as in the case of render or
stone, the potential of nanolimes is still questioned
(Campbell, 2010; Hull, 2012). One of the reasons
of the limited effectiveness of these products is the
accumulation of nanolime at or just beneath the
drying surface (often with the formation of a white
patina) (Ghaffari, 2012; Costa, 2011).

Figure 2: Specimen set up and relative sampling spots; TA: absorption and drying perferomed on the same side; TB:
absorption on one side (then sealed), drying on the opposite side.

78
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
In order to improve the consolidation effect in
depth, the transport mechanism and the deposition
of nanolimes within porous materials, as stone or
renders, should be better understood. Nanolime
transport mechanism has not been systematically
studied yet. Does nanolime accumulate at the sur-
face during absorption or is it able to penetrate in
depth in the material, but it migrates back to the
surface during the drying process? What is the role
of nanolime stability and nanoparticles aggregation
phenomena in nanolime deposition in a treated
material?
This research aimed at clarifying the transport
and deposition mechanisms of nanolime in porous
materials. The transport of a commercial nanolime
product was studied on Maastricht limestone, a
high porosity yellowish limestone. The absorption
and drying behaviour of nanolime was measured
and compared to that of ethanol (the dispersant
of this nanolime commercial product) and water
(section 3.2). Next to simple phenolphthalein tests
on cross sections (section 3.3), microscopy obser-
vations by AFM (section 3.1), optical microscopy
(3.4) and SEM-EDS (3.5) were performed; the pur-
pose is to investigate nanolime penetration depth
and deposition within the material as result of the
transport mechanism.
II. Materials and methods
Substrate
The Maastricht limestone is an almost pure cal-
careous building material (⇡95% CaCO3 ) (Nijland,
2006), quarried in the Belgian and Dutch provinces
of Limburg (Dubelaar, 2006). It has high poros-
ity (45-50%) and unimodal pore distribution (35-40
µm). Those characteristics, combined with a weak
cementation, lead to some unbelief with respect
to its use as a major good-quality building stone
(Dreesen, 2004).
A large number of historical buildings have in-
deed been built almost exclusively with this mate-
rial and have remained rather unaffected after cen-
turies of environmental exposure (Dreesen, 2004;
Van Hees, 2009). Its durability is possibly linked to
the dissolution and re-precipitation of carbonates
within the limestone pores network, when exposed
to atmospheric agents (Dusar, 2003). This process
was traditionally enhanced by the application of
limewater (kalkwei), which forms a sort of durable
protective layer (Dreesen, 2004). Despite its good
durability, this material may in some cases show
decay in the form of loss of cohesion of the surface
(Figure 1a) (Nijland, 2006).
Nanolime
The commercial nanolime CaloSil E25 (IBZ, Ger-
many), reported in the text just as E25, has been
tested. This colloidal dispersion has a calcium
hydroxide nanoparticles concentration of 25 g/l
and is dispersed in ethanol. E25 is obtained by
solvothermical reaction of metallic calcium in water
or alcohols (e.g. ethanol, propanol, isopropanol).
Nanoparticles have hexagonal plate-like shape and
size ranges between 50 and 250 nm (Ziegenbalg,
2010). Before use, E25 was placed in an ultrasonic
bath (60 Hz) for 10 minutes, to minimize nanoparti-
cle aggregation phenomena. Ethanol (p.a. >99.5%,
by Sigma Aldrich) and distilled water (conductivity
< 2 µs/cm) were used for comparison with the E25.
Absorption and drying tests
Maastricht limestone core specimens (4 cm height,
4 cm diameter), drilled from sound blocks, were
used for the capillary absorption and drying mea-
surements. The capillary absorption and drying
were measured for E25 as well as for ethanol and
water.
The specimens were sealed on the lateral side
with parafilm in order to prevent evaporation on
these sides. The capillary absorption was measured
by partially immersing the specimens in a petri dish,
filled with nanolime, ethanol or water. The weight
of the core specimens was measured at regular in-
tervals during the absorption process till saturation
was reached, and the height of the wetting front
was monitored and photographically recorded. Wa-
ter, ethanol and nanolimes absorption were carried
out in sequence on the same specimens, in order to
minimize the effect of the stone variability.
The drying behavior of specimens saturated with
nanolime, ethanol and water was evaluated by mea-
suring the weight loss over time. Both absorp-
tion and drying measurements were performed
in threefold and under controlled conditions (50%
RH, T=20 C, air speed < 0,1 m/s).
Additionally, in the case of nanolime, in order to
clarify if any nanolime accumulation at the absorp-
tion surface occurs during absorption or drying,
the absorption and drying were repeated according
to two different test procedures, as shown in Fig-
ure 2. In one case (specimen TA), the drying was
performed on the same side of the absorption (as
it normally occurs in the practice of conservation);
in the other case (specimen TB) the drying was
performed through the surface opposite to the ab-
sorption surface (which was sealed after specimen
saturation with nanolime).
79
 Borsoi et al.
Phenolphthalein test
The penetration of the lime nanoparticles immedi-
ately after absorption was assessed by phenolph-
thalein test. After complete saturation, the cross
section of the specimen was nebulized with a 1%
phenolphthalein solution in ethanol and water. Phe-
nolphthalein is a pH indicator that turns purple
when pH conditions are higher than 9,8. In this
case a purple color of the substrate, treated with
phenolphthalein, indicates the presence of nanolime
(Ca(OH)2 , pHE25 > 11). By comparing the results of
the phenolphthalein test with the photographical
observations of the wetting front, conclusion can be
drawn on possible phase separation between dis-
persant (in this case ethanol) and nanolime particles
during absorption.
The phenolphthalein test was performed on the
specimens saturated with nanolime also during the
drying process (at 1, 4, 8, 24 and 48h), in order to
understand nanolime migration within the material.
Drying tests were performed both in atmospheric
and in a nitrogen-rich atmosphere, in order to check
the relevance of nanolime carbonation during the
drying period. A N2 Boy nitrogen gas-supplying
device, by Taitec, was used to create a nitrogen-rich
atmosphere within a glass desiccator, where treated
specimens were stored.
Microstructural observations
AFM
A TT-AFM device was used to characterize E25
samples; the aim of this analysis was to verify the
morphology of the nanoparticles and possible ag-
gregation phenomena, that can affect the transport
of nanolime. Nanolime was diluted in its pure
ethanol and after several trials a dilution of 1:100 (in
volume, that is a concentration of 0.25 g/l of nano-
Ca(OH)2 in ethanol) was considered as suitable for
AFM analysis. An amount of 5?l was dropped with
a micropipette on the freshly cleaved mica substrate
(sample holder). Images were aquired in vibration
using silicon tips (Brucker model NCHV) and both
topographic and phase images were processed. The
equipment allows a measurement of the surface
topography with a high precision up to nanometric
scale.
Optical Microscopy
The drying surface and the cross section of the
specimens treated with nanolime were observed
with a stereomicroscope Zeiss Stemi SV 11. Images
were recorded with a Zeiss AxioCam MRc5 digital
microscopy camera. The AxioVision 4.8 software
and its interactive measurement tools were used to
record and analyse the specimens.
80
SEM-EDS
The cross-section of the specimens treated with
nanolime were studied by SEM-EDS; the equipment
(Nova NanoSEM 650, by FEI) is coupled with a low
vacuum solid-state detector BSED (GAD) which al-
lows high resolution imaging (up to 1,4 nm) and
a high flexibility of the working conditions (1 to
30kV). Samples were collected, at the end of the
drying, in the most representative spots (top, mid-
dle, bottom) of both specimens types (TA and TB),
as seen in Figure 2.
III. Results
Nanolime characterization by AFM
The observation of E25 by AFM clarifies the particle
size of nanolime and aggregation phenomena in
the dried phase. Figure 3a shows a heterogeneous
distribution of nanolime and a high polidispersivity
of the nanoparticle aggregates (with different size
and shape); the size of these clusters range from few
nanometers to more than 10 µm. Figure 3b and 3c
show that clusters are formed by nano (30-100 nm)
and sub-micrometric(up to 400 nm) lime particles;
the hexagonal plate-like shape, typically observed
for these type of nanoparticle (Salvadori, 2001), is
possibly deformed by dragging effect originated by
the weak absorption of the lime nanoparticle on the
mica sample holder (Ambrosi, 2001), and /or to the
fraction exerter by the scanning of the AFM tip.
Lime nanoparticles, when the alcoholic solvent
has completely evaporated, tend then to form clus-
ters; these aggregation phenomena are an impor-
tant factor that can affect the deposition of nanolime
clusters within the treated material.
When considering the liquid phase (nanoparticle
dispersed in the alcoholic solvent), diffuse electrical
double layer repulsive interactions would prevent
nanoparticle aggregation, at least in diluted dis-
persions (Rodriguez-Navarro, 2005). Furthermore
the low dielectric constant of a low-polarity sol-
vent as ethanol can guarantee a stable dispersion.
However lime nanoparticles tend spontaneously
to aggregate and attractive Van der Waals forces
will keep particles attached once agglomeration has
taken place. As argued by Chekli (2013), a particle
size concentration dependence at nanoparticle con-
centrations above 0,05 mg/L has to be considered;
when nanoparticle concentration is relevant, like in
our case (25g/l), the distance between nanoparticles
in the sample is reduced; this presumably increases
the chance of collision between particles and hence,
their aggregation.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 3: a) Topographical view (50x50 µm2 ) by AFM of nanolime clusters in the dried phase with different shape
and size; zooom on specific spots: a) topography and b) AFM image of two clusters of E25, formed by the aggregation
of lime nanoparticle.
Absorption and drying kinetics
Figure 4a shows the capillary absorption curve of
water, ethanol and E25. All these liquids saturate
the sample in less than 5 minutes, due to the coarse
porosity of Maastricht limestone.
When comparing the behaviour of the different
liquids, water absorption is faster than ethanol and
E25 absorption. Comparing pure ethanol and E25,
a slower absorption is measured for E25. This is
due to lime nanoparticles (25g/l), which modify the
density (and therefore viscosity) and possibly affect
the surface tension of the alcoholic solvent; this af-
fects the capillary transport. Duan (2011) points out
that nanoparticle agglomeration, that easily takes
place as seen in the previous section, can affect
the viscosity of the dispersion and therefore the
absorption of the dispersion.
When considering the drying behavior (Fig-
ure 4b), the evaporation rate of water is much
slower (up to 7 days for complete drying) than
that of ethanol (complete drying in 48-72h) or nano-
lime (up to 72h). The faster evaporation of ethanol
can be easily explained by its lower boiling point
(78,37 C) compared to that of water (⇡100 C).
Moreover the superficial tension of water (3 times
higher than that of ethanol) can explain the higher
water retention within the Maastricht pore network.
In the drying curves two stages can be observed.
In the 1st stage of drying, called the constant drying
period, the drying front is at the surface and the
drying rate is constant (Hall, 2012). This first phase
ends after 24h in the case of the alcoholic solution
(EtOH and E25), while for water it is approximately
at 96h. In the 2nd stage of drying, called the falling
drying rate period and identified by the change in
the slope of the drying curve, the liquid content can
no longer support the demands of the evaporation
flux. In this stage transport occurs in the vapor
phase; thus lime nanoparticles cannot physically
take part in the process and are deposited in the
material before the end of the 1st stage. It is thus
important to observe the deposition of nanolime at
24h, which corresponds to the end of the 1st stage.
By comparing the drying curves of ethanol and
nanolime, a slight delay in the drying of the mate-
rial is observed in the case of E25. Nanolime does
not hinder significantly the drying kinetic but a
partial occlusion of the pores, caused by the forma-
tion of a nanolime layer during drying, could have
delayed the evaporation of the solvent.
Phenolphthalein test
Phenolphthalein test was performed as described in
section 2.4, immediately after complete saturation
of the specimens with nanolime at standard condi-
tions in air (50% RH, 20oC) and in a nitrogen-rich
atmosphere. Nanolime has penetrated the whole
depth of the specimen (40 mm), as shown in figure
5a. No accumulation of nanolime at the absorption
surface is macroscopically visible immediately after
specimen imbibition.
By analysing the specimens stored in a nitrogen-
rich atmosphere, it is observed that nanolime is
stable and homogeneously distributed within the
cross section during the first 8 h of drying (Figure
5b,c). At 24 h, a remarkable amount of nanolime
is visible nearby the drying surface, whereas just
traces are detected in the outer 10 mm beneath the
81
 Borsoi et al.

Figure 4: a) Volume comparison of the capillarity wetting curve of water, ethanol and E25; b) Capillarity drying curve
in volume of H2 O, EtOH and E25; the red circles indicate the critical liquid content and the end of the first step of
drying.

drying surface (Figure 5d). Further microscopical
details are presented in the next section.
Microstructural observations by optical mi-
croscopy
In order to clarify nanolime transport and depo-
sition within limestone pores network, specimen
type TA and TB (described in section 2.3) were
characterized by optical microscopy.
In Figure 6 stereomicroscope microphotographs
of the cross sections of specimens TA and TB, ob-
served after complete drying (⇡72h), are shown.
In both cases nanolime accumulates at 0,5-0,7 mm
beneath the drying surface. No deposition of nano-
lime can be identified at the absorption surface of
TB.
Observations on specimen TB clarify that nano-
lime penetrates in the material without any accumu-
lation at the absortion surface during the absorption
phase. No phase separation between the particles
and the solvent is observed during the absorption
process in this stone.
At the same time, traces of nanolime are observed
in depth in both specimens, as visible in Figure 7a
and 7b. Nanoparticle tends to aggregate. Reduced
amount of nanolime are indeed visible in the entire
section of both specimens.
The specimens analyzed in section 3.3 (phenolph-
thalein test) were observed more in detail by optical
microcopy (Figure 8a,b); no white deposition was
macroscopically identified on the dying surface,
but nanolime formed a thin layer beneath the dry-
ing surface (at 0,5 mm in depth, Figure 8a,b). The
solvent, by evaporating, flows in the direction of
the drying surface and nanoparticles partially mi-
grate with the solvent, depositing just beneath the
drying surface. This accumulation of nanolime is
remarkable at the complete saturation of the solvent
(48-72h); this evidence can confirm that nanolime
is gradually moving back to the surface with its
solvent during drying.
Although most of the nanolime is accumulated
just beneath the drying surface, small and heteroge-
neously distributed amounts of nanolime particles
are also observed in depth (20 to 40 mm).

Figure 5: Pictures of the phenolphthalein test performed on the cross sections of specimens treated with E25 (and
observed a) immediately after saturation (in air at 50% RH, T=20 C), with complete saturation of the specimen
(arrow); specimens were then stored in a N2 chamber and observed at b) 4h, c) 8h and d) 24h during drying (each
cross section has been obtained by breaking the specimen at the given time and drying surface is on the top of the
specimen); a layer of nanolime is visible at 0,5 mm beneath the drying surface (arrows).

82
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 6: Microphotographs of the limestone cross section in proximity of the drying surface of a) TA and b) TB; in
both cases nanolime is accumulated in a thin layer near the drying surface.

Microstructural observations by SEM-EDS
Microstructural observations and microanalyses
were performed by SEM-EDS in order to further
study the deposition of nanolime within limestone.
The cross sections of specimens TA and TB were
observed. In both cases, samples were collected
from the middle of the specimen, and at the absorp-
tion/drying surfaces (Figure 2).
The SEM observations of specimen TA show a
significant deposition of nanolime at 0,5 mm be-
neath the drying surface (Figure 9a, 9b); the rel-
ative EDS spectra ( 9c) highlights the presence of
large amounts of carbonated nanolime (CaCO3 ).
Nanoparticles have mainly hexagonal plate-like
shape and dimension between 30 to 100 nm. Due
to random Brownian motion and attractive short-
range van der Waals forces, nanoparticles have
mainly agglomerate in clusters (Figure 9b). In fact,
similar strong aggregation phenomena are reported
in literature for particle size of 1-100 nm, resulting
in clusters of typical size 0,1-10 µm (Gaurav, 2009).
Significant deposition of nanoparticles is visible
up to 10 mm in depth from the absorption/drying
surface; the nanolime particles have deposited fill-
ing voids and pores. Pores with diameter of 1 to 5
µm are easily filled with the nanoconsolidant.
Next to the accumulation of nanoparticles clus-
ters near the drying surface, local depositions of
isolated nanolime particles are visible almost every-
where in the observed specimen section. Deeper
in the stone, at 20 mm from the absorption/drying
surface, only sporadic deposits of nanolime par-
ticles are observed, heterogeneously distributed
(Figure 9d, 9e). At 30 to 40 mm in depth from
the drying surface, the presence of nanoparticles
(Figure 9f, 9g) is very limited.
When considering the particle size of the nano-
lime deposits in the cross section, it can be observed
that the size of the nanoparticle is smaller (30 to 100
nm) at the drying surface than in depth in the mate-
rial (150 to 300 nm). This suggests that smaller lime
nanoparticle are back-transported more easily to
the drying surface, meanwhile bigger nanoparticles
(70 to 300 nm) deposit preferentially in depth.
The SEM-EDS analysis of specimen TB shows
significant nanolime deposition at 0,5 mm beneath
the drying surface (Figures 10a, 10b). Nanolime
is visible as well at 10 mm in depth; nanolime de-
position is decreasing gradually with depth, with
local clusters visible up to 20-25 mm (Figures 10c,
10d). At 35 to 40 mm in depth, in proximity of
the absorption surface, some sporadic accumula-
tions of nanoparticles are locally visible (Figure 10e).
Nanoparticles have in some cases a tetragonal mor-
phology (Figure 10f), which is typically attributed
to aragonite, a thermodynamically unstable poly-
morph of calcium carbonate.

Figure 7: Microphotographs of the limestone cross section: a) traces of nanolime at 20 mm in depth in specimen TA; b)
reduced amount of nanolime at 25 mm in depth in specimen TB.

83
 Borsoi et al.

Figure 8: Microphotographs of specimens treated with E25, stored in a nitrogen-rich atmosphere and observed at a)
24h and b) after complete drying (48 to 72h) in a N2 rich atmosphere. A layer of nanolime is visible at 0,5 mm in
depth in the drying surface.

These observations demonstrate that the accumu- IV. Conclusion

lation of nanoparticles, observed at the absorption
surface in specimen TA, is not due to reduced pen-
etration and accumulation of the nanoparticles at
the absorption surface during absorption, but to mi-
gration together with the solvent during the drying
phase. These results definitely confirm the obser-
vations carried out with phenolphthalein, showing
migration of the nanoparticles during drying.
As observed in TA, also in the case of TB the
size of nanoparticle is smaller (30 to 60 nm) at the
drying surface and considerably larger (70 to 300
nm) in depth.
This paper clarifies the transport of nanolime within
Maastricht limestone, a highly porous calcareous
material. The results show that a colloidal dis-
persion of lime nanoparticles in ethanol (CaloSil
E25) can easily penetrate in depth in this material
without any accumulation at the absorption surface
during the absorption phase. Nano and submicron
particles (up to 400 nm) can widely fit in the lime-
stone porosity (mean pore size of 35 ?µ). However it
appears that accumulation just underneath the sur-
face occurs during drying, when the solvent moves
back to the drying surface, depositing an important
part of the nanolime particles.
It was observed that the size of the nanoparticles

Figure 9: SEM microphotographs of specimen TA at different depths in the cross section (from the drying surface);
analysis of sample At at a) 0,5 mm and b) 1,7 mm and c) its relative EDS spectra; analysis of sample Am at d) 20 mm
and e) 22 mm; analysis of sample Ab at f) 30 mm and g) 35 mm in depth.

84
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 10: SEM microphotographs of specimen TB at different depths in the cross section (from the drying surface);
analysis of sample Bt at a) 0,5 mm and b) 1 mm, with a remarkable accumulation of nanolime; analysis of sample Bm
at d) 20 mm and e) 22 mm, with sporadic amounts of nanolime; analysis of sample Bb at f) 32 mm and g) 35 mm,
where some lime nanoparticle clusters are visible.
determine the location of their deposition in the
stone: smaller lime nanoparticle (30 to 60 nm) much
easily migrate back during drying and precipitate
in clusters in the proximity of the drying surface,
whereas some of the bigger nanoparticles (70 to 400
nm) deposit in depth,
Because of the fast absorption of this coarse
porous stone, phase separation between solvent
and nanoparticles does hardly take place during
absorption; only a small amount of nanoparticles,
generally those of large size, is deposited in depth
in the material. Aggregation phenomena, which
take place in the dispersion mostly due to Brownian
motions, favour the formation of clusters; Van der
Waals forces keep nanoparticles attached. Anyway
these weak bonding phenomena do not guarantee
a significant deposition of nanolime in depth in the
material.
Besides, the dense nanolime layer (accumulated
beneath the drying surface) limits the consolidation
effect in depth and the penetration of successive
consolidant applications.
These results suggest that in order to improve
nanolime precipitation in depth, the transport and
deposition mechanism of nanolime should be mod-
ified.
A more homogeneous nanoparticles distribution
in depth might be obtained by modifying the prop-
erties of the solvents. A solvent with slower drying
rate and higher dielectric constant might be used
to favor nanoparticle/solvent phase separation and
induce nanoparticle deposition in depth.
Next to solvent modification, the optimization of
improved application strategies might help favor-
ing nanolime precipitation in depth. Research is
ongoing to verify these possibilities.
Acknowledgement
Special thanks go to prof. Jorge Caldeira (FCT-UNL,
Lisbon) for support in AFM analysis and to Willem
Duvalois and Timo Nijland for SEM-EDS analysis
References
Ambrosi, M., Dei, L., Giorgi, R., Neto, C., & Baglioni, P. (2001):
" Colloidal particles of Ca(OH)2 : properties and applications to restora-
tion of frescoes." Langmuir, 17(14): 4251-4255.
Baglioni, P., Chelazzi, D., Giorgi, R., Poggi, G. (2013): " Colloid and
Materials Science for the Conservation of Cultural Heritage: Cleaning,
Consolidation, and Deacidification." Langmuir, 29: 5110?5122.
Borsoi, G., Tavares, M., Veiga, R., Santos Silva, A. (2012):
" Microstructural Characterization of Consolidant Products for Histori-
cal Renders: An Innovative Nanostructured Lime Dispersion and a More
Traditional Ethyl Silicate Limewater Solution." Microscopy & Microanaly-
sis, 18:1181-1189.
Costa, D., Delgado Rodrigues, J., (2012): " Consolidation of a porous
limestone with nanolime." In proceeding of 12th Int. Congress on Deteri-
oration and Conservation of Stone, Columbia University, New York, 22-26
October 2012.
85
 Borsoi et al.
Campbell, A., Hamilton, A., Stratford, T., Modestou, S., Ioannou, I.
(2011): " Calcium Hydroxide Nanoparticles for Limestone Conservation:
Imbibition and Adhesion." In Symposium Adhesive and Consolidants for
Conservation: Research and Applications, ICC, Ottawa, Canada, 17-21 Octo-
ber 2011.
Chekli, L., Phuntsho, S., Kandasamy, J., Shon, H. (2013): " Assessing
the aggregation behaviour of iron oxide nanoparticles under relevant en-
vironmental conditions using a multi-method approach." In Conference
Nanotech 2013, Washington DC, Volume: 1.
Doehne, E., Price, C.A. (2010): " Stone Conservation. An Overview
of Current Research, Research in Conservation." 2nd edition, The Getty
Conservation Institute, Los Angeles.
Dreesen, R., Dusar, M. (2004): " Historical building stones in the
province of Limburg (NE Belgium): role of petrography in provenance
and durability assessment." Materials Characterization, 53: 273-287.
Duan, F., Kwek, D., Crivoi, A. (2011): " Viscosity affected by nanopar-
ticle aggregation in Al2 O3 -water nanofluids." Nanoscale Research Letters,
2011, 6:248.
Dubelaar, C.W., Dusar, M., Dreesen, R., Felder, W.M., Nijland, T.G.
(2006): " Maastricht limestone: a regional significant building stone in
Belgium and the Netherlands. Extremely weak, yet time-resistant." in
Heritage, Weathering and Conservation Conference - HWC 2006, Madrid, 21-
24 June 2006.
EN 13775 (2008): " Natural stone test methods - Determination of wa-
ter absorption at atmospheric pressure." European Standard, European
committee for Standardization (CEN).
Favaro, M., Tomasin, P., Ossola, F. and Vigato, P.A. (2008): " A novel
approach to consolidation of historical limestone: the calcium alkoxides."
Applied Organometallic Chemistry, 22(12): 698-704.
Gaurav, P. (2009): " Understanding nanoparticle aggregation." PhD
thesis, Iowa State University.
Ghaffari, E., Koberle, T., Weber, J. (2012): " Methods of polarising
microscopy and SEM to assess the performance of nano-lime consoli-
dants in porous solids." In 12th International Congress on the Deterioration
and Conservation of Stone, Columbia University, New York, 22-26 October
2012.
Giorgi, R.; Dei, L.; Baglioni, P. (2000): " A new method for consoli-
dating wall paintings based on dispersions of lime in alcohol." Studies in
Conservation, 45: 154-161.
86
Hall, C., Hoff, W.D. (2012): " Water transport in brick, stone and con-
crete." Spon Press (Taylor & Francis Group), New York.
Hansen, E., Doehne, E., Fidler, J., Larson, J., Martin, B., Matteini, M.,
Rodrigues-Navarro, C., Sebastian Pardo, E., Price, P., de Tagle, A., Teu-
tonico, J. M., Weiss, N. (2003) " A review of selected inorganic consoli-
dants and protective treatment for porous calcareous materials." Reviews
in Conservation, 4: 13-25.
Nijland, T.G., Dubelaar, C.W., Tolboom, H.J., Van Hees, R.P.J. (2006):
" Building stones from a muddy delta: Native natural stone from the
Netherlands." In Heritage, Weathering and Conservation Conference- HWC
2006, Madrid, 21-24 June 2006.
Rodriguez-Navarro, C., Ruiz-Agudo, E., Ortega-Huertas, M., Hansen,
E. (2005): " Nanostructure and irreversible colloidal behavior of Ca(OH)
2 : implications in cultural heritage conservation." Langmuir, 21(24),
10948-57.
Rodriguez-Navarro (2013): " Alcohol Dispersions of Calcium Hydrox-
ide Nanoparticles for Stone Conservation." Langmuir, 29: 11457?11470.
Salvadori, B., Dei, L. (2001): " Synthesis of CaOH2 nanoparticles from
diols. " Langmuir, 12: 2371-2374.
Teutonico, J. M., Charola, A. E., De Witte, E., Grassegger, G., Koestler,
R. J., Laurenzi Tabasso, M., Sasse, H. R., Snethlage, R. (1997): " Group re-
port: How can we ensure the responsible and effective use of treatments
(cleaning, consolidation, pro- tection)?" In Saving Our Architectural Her-
itage: The Conservation of Historic Stone Structures; Dahlem Workshop on
Saving Our Architectural Heritage, The Conservation of Historic Stone Struc-
tures, Berlin, 3-8 March 1996, ed. N. S. Baer and R. Snethlage, 293-313.
Toniolo, L; Paradisi, A; Goidanich, S; Pennati, G. (2010): " Mechanical
behavior of lime based mortars after surface consolidation". Construction
and Building Materials, 25 (4): 1553-1559. Van Hees, R., Nijland, T. (2009):
" Assessment of the state of conservation of a Middle Neolithic flint mine
in Maastricht limestone". Heron, Vol. 54, No. 4.
Van Hees, R.P.J., Lubelli, B., Nijland, T., Bernardi, A. (2014):
" Compatibility and performance criteria for nano-lime consolidants". In
proceedings of 9th International Symposium on the Conservation of Monu-
ments in the Mediterranean Basin - Monubasin 2014 , Ankara, Turkey, 3-5
June.
Ziegenbalg, G., Brummer, K., Pianski, J. (2010): " Nano-Lime - a
new material for the consolidation and conservation of historic mortars".
In 2nd Historic Mortars Conference HMC10 and RILEM TC 203-RHM final
workhop, Prague, 22-24 September.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

3D characterization of structure and

composition of sandstone: a nanometer to
centimeter-scale approach using high-end
imaging techniques
Wesley De Boever⇤,a , Luc Van Hoorebekeb , Ana Diazc
Mirko Hollerc , Veerle Cnuddea
a Dept. of Geology and Soil Science, Ghent University, Ghent, Belgium. – PProGRes / UGCT –
b Dept. of Physics and Astronomy, Ghent University, Ghent, Belgium. – UGCT –
c Paul Scherrer Institute – SLS/cSAXS – Villigen, Switzerland.
⇤ wesley.deboever@ugent.be

Abstract

Most building materials, including natural stone and concrete, are complex, heterogeneous materials, with feature
sizes varying from a few nanometers to several centimeters, inside one single sample. For a full characterization of these
materials, the full scale-range needs to be analyzed, since all scales are important for the behavior of these materials.
Laboratory based high-resolution X-ray CT (HRXCT, µ-CT) has become a mature and frequently used tool for the
3D characterization of many types of materials, including building materials (Cnudde et al., 2011; Cnudde et al.,
2009; Cnudde and Boone, 2013; Lenoir et al., 2004; Rozenbaum, 2011). Despite its succes, there are some improtant
shortfalls to the technique. Although sub-micron resolutions can be achieved at high-end systems, these systems
are still far from standard, and some important sample-size trade-offs have to be made to achieve these resolutions.
Especially in heterogeneous (geological) building materials, this leads to questions concerning representativity of samples.
Furthermore, the acquired chemical information is usually limited to a relative value in terms of density and atomic
number. We propose a workflow that combines multiple techniques to characterize structure and chemical composition of
the Vosges sandstone, a heterogeneous, clay-rich sandstone consisting out of a mixture of quartz, feldspars and feldspars.
µ-CT on samples of different sizes is combined with scanning electron microscopy and energy dispersive spectroscopy,
to gain additional information on both structure and chemistry of the stone. By using 2D and 3D image registration
techniques, a pseudo-3D look into the distribution of different mineral phases is obtained, using a representative sample
size. Next, focused ion beam nanotomography (FIB-nt; Holzer et al., 2004; Holzer and Cantoni, 2011; Keller et al.,
2011) is used to investigate the clay mineral fraction, since these features cannot be resolved using µ-CT by the lack
of resolution. The most important drawback of FIB-nt is its destructive nature, making it unsuitable for the analysis
of dynamic processes such as water uptake, which are important in building materials. Therefore, a novel technique,
ptychographic tomography (Diaz et al., 2012; Dierolf et al., 2010; Holler et al., 2014; Trtik et al., 2013), is demonstrated.
This technique, possible at synchrotron facilities, enables 3D imaging of materials at resolutions up to 16 nm, in a
non-destructive way. Ptychographic tomography opens the door for the dynamic imaging of processes at the nanoscale

Keywords: SEM, micro-tomography, data fusion, ptychography, FIB-nt

I. Introduction graphical analysis) or later through post-processing

Analysis techniques based on imaging have always
been important methods for the study of (geologi-
cal) building materials. The most well-known and
mature techniques to do this are optical microscopy
(both in reflected or transmitted light mode and
in fluorescent light mode) and scanning electron
microscopy (SEM). These microscopic techniques
provide 2D images, in which information on both
the structure and composition of materials can be
found, be it directly during the analysis (e.g. petro-
the images (e.g. thresholding the porosity from
SEM images; Marcelino et al., 2007; Prakongkep et
al., 2010). When used together with energy disper-
sive spectroscopy (EDS), SEM analysis can provide
quantitative compositional data of the analyzed
samples. During the last two decades, high resolu-
tion X-ray tomography (µ-CT) has become a very
valuable technique for the acquisition of 3D images
of building materials; as a bonus, this technique
images the internal structure of these materials in
a non-destructive ways, allowing the study of dy-

87
 each method, and to complement for the weaknesses of others. A workflow was used to go
the lowest resolutions, up to adevery small sub-sample, that could be imaged at extremely hig
Boever et al.
1shows a schematic representation of that workflow.

namic processes, fast and slow ones, on the same

sample (Cnudde and Boone, 2013; Rozenbaum,
2011; Wildenschild and Sheppard, 2013). Further-
more, CT is capable of imaging a big range of sam-
ple sizes, going from mm-scale microplugs to full-
size, 10 cm diameter drill cores. Using this method,
information on connectivity of the pore space, stack-
ing of the grains, dissolution/precipitation pro-
cesses, drying rate, etc. can be analyzed, all of
which are highly important in building stone and
concrete properties. Despite its obvious advantages,
there are some fields in which µ-CT has to face
defeat against the previously mentioned, more tra-
ditional research techniques. In standard laboratory
µ-CT setups, it is not possible to acquire exact chem-
ical information about the sample, since the images
acquired with CT are 3D distribution of the lin-
ear attenuation coefficient of the samples interior,
Figurenumber
which depends on the atomic 1: Workflow for multi-scale analysis of Vosges sandstone sample. A large, centim
and density
of the sample. Since mineralogy and chemical com- Figure 1: Workflow for multi-scale analysis of Vosges
is analyzed using µ-CT, sides are imagedsandstone using SEM and EDS, a millimeter-scale microplug is used fo
sample. A large, centimeter-scale base sample
positon have a big influence on various processes
subsamplediseven more for FIB-nt
analyzed using µ-CT, and
sides areptychographic
imaged using SEM tomography.
(Marszałek et al., 2014; Rivas et al., 2003) in the
pore space of materials, qualitative and quantita- and EDS, a millimeter-scale microplug is used for high-
3
The original
tive mineralogical information sample
is a definitive in thisresolution
must study wasCT is asubsampled
14 x 14evenx 10more
mm cuboid
for FIB-nt andsample of the V
in orderfurther ptychographic tomography.
research steps, the sides of this sample were polished flat up to 1 µm. A first analy
to fully understand pore-scale processes
inside building
scan of materials.
the entire specimen, at the highest resolution possible for this sample size. T
A second drawback is the limit in resolution,
performed at the HECTOR setup of the
which is currently at about 400 nm for high-end
UGCT
als. The stone (Masschaele et stage
represents the last al., 2013), allowing for a spa
of the fluvi-
under 9 (although
µm. Thehigher
source was operated atile conditions during the Bundsandstein (Gall and
at a voltage of 150 keV, and 2000 projections with a
laboratory setups resolutions up
Grauvogel-Stamm, 2005; Shear et al., 2009). This
to 200 nmseconds were taken. Next, the polished
can be obtained at synchrotron facilities, sides
particular of thiswas
sandstone sample
chosen were coated with carbon, to
for its structural
or by using a system equipped with X-ray optics).
and chemical properties, rather than its importance
More importantly, to reach this resolution, samples
as building material. The stone’s microstructure is
have to be very small (<1 mm in diameter), due
composed out of a mix of macropores (>20 µm) and
to the cone-beam configuration of laboratory se-
micropores (< 1 µm), and there is enough variation
tups. Since evolution in systems, reconstruction
in chemical composition to justify a more detailed
algorithms and analysis software is extremely fast,
research.
µ-CT has become a mature and reliable technique
for many applications, especially if the features of
interest are not too tiny. III. Methods
As mentioned, all imaging techniques have their
advantages and disadvantages. The goal of this The Vosges sandstone was characterized using dif-
study is to combine different techniques, and over- ferent methods, to make full use of the strengths of
come the disadvantages by exploiting the strengths. each method, and to complement for the weak-
nesses of others. A workflow was used to go
from a big sample at the lowest resolutions, up
to a very small sub-sample, that could be imaged
II. Material: grès a meules / Vosges at extremely high resolutions. Figure 1 shows a
sandstone schematic representation of that workflow.
The original sample in this study was a 14 ⇥ 14 ⇥
The grès a meules (Meules sandstone), often re- 10 mm3 cuboid sample of the Vosges sandstone. For
ferred to as ‘Vosges sandstone’ is a sandstone from further research steps, the sides of this sample were
the Vosges mountain range in north-eastern France. polished flat up to 1 µm. A first analysis step was a
It makes up the lower region of the grès a Voltzia µ-CT scan of the entire specimen, at the highest res-
formation, and therefore is corresponds to the more olution possible for this sample size. This µ-CT scan
general Buntsandstein formation, deposited in the was performed at the HECTOR setup of the UGCT
lower Triassic. The Meules sandstone is a fine- (Masschaele et al., 2013), allowing for a spatial reso-
grained red to pink sandstone, containing about lution of just under 9 µm. The source was operated
75 % quartz and 20-30 % feldspars and clay miner- at a voltage of 150 keV, and 2000 projections with an

88
extremely small. In this study, several 50 µm diameter samples were imaged at a resolution of ± 100 nm, and
two 25 µm diameter samples could be measured at a resolution of about 45 to 60 nm spatial resolution.
th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Ptychographic 15
experiments were carried out at the cSAXS beam line of the Paul Scherrer Institute in Villigen,
Switzerland (www.psi.ch).

Figure 2: Typical setup for ptychographic tomography. The narrow (collimated) X-ray beam can either be
Figure 2: Typical setup for ptychographic
obtained tomography.
by a pinhole, The narrow
or by using (collimated)
a Fresnel zone plate.X-ray beam can either be obtained by
a pinhole, or by using a Fresnel zone plate.
Results and discussion
exposure time of 2 seconds were taken. Next, the porosity of the 3D volumes were analyzed using the
X-ray computed tomography
polished sides of this sample were coated with car- inhouse developed Morpho+ software package (Bra-
bon, to prepare it for SEM analysis. Using a JEOL bant et al., 2011; wwww.octopusanalysis.eu). To ac-
For easemicroscope,
5310-LV of reading,backscattered
results of both tomography
electron images scans (high
quire and of
images lower resolutions)
the tiniest featuresare put sandstone,
in the together,
despite theoffact
the that they surfaces
sample’s do not follow each other
were taken. Of 5 in the chronology
cuboid of the
FIB-nt was used.used
In workflow. The SEM
this procedure, reconstructed
images of
slices of the µ-CT
sides, scan images
mosaic of the entire ‘large’ sample
were composed out ofclearly
± 300indicated thatare
a region a resolution of 8.9each
taken, and after µmimage,
was not enough
a thin layer
to resolveimages
all features
with of the Vosges
a pixel sandstone
size of 1.08 clearly.that
µm, meaning Although onesample
of the can have anofidea
is cut ofthe
using theFIB.
overall
This structure
way, a 3D
a total
and porosity, of 1500 images
individual grains was
and acquired. Thenot
pores could sides
beofsegmented
image ofand a very small region
separated on this(usually
sample.no However,
more than
the sample were analyzed using EDS, to determine 50 ⇥ 50 ⇥ 50 µm3 ) can be imaged, at the same reso-
the mineral content of the stone. Due to time con- lution acquired in standard SEM images (De Winter
straints, the analyzed surface was limited to the top et al., 2009; Holzer et al., 2004). FIB-nt was per-
5 mm of each side, and the pixel size was reduced formed at the department of Information Technol-
to 2 µm. 28 EDS maps (for a total area of 14 ⇥ 5 ogy of Ghent University (www.fib.intec.ugent.be) 3
mm2 ) were acquired and stitched together for each using a FEI Nova Nanolab 600 Dual-beam system.
surface, creating a total EDS analysis time of over The pixel size of the individual slices was 16 nm,
40 hours. EDS and SEM images were compared and 200 slices were imaged at a z-interval of 50
to each other by performing 2D image registration, nm. This resulted in a 17 ⇥ 17 ⇥ 10 µm3 volume,
carried out using the Avizo R software package. Af- recorded over a time period of about 3 hours.
ter these 2D analysis methods, a 1 mm diameter
sub-sample was drilled out of one of the surfaces
of the stone, which was imaged using µ-CT at sub- IV. Results and discussion
micron resolution. Using UGCT’s MEDUSA setup,
Finally, one of the first geological applications of
the sample was scanned at a voltage of 90 keV; 2800
a relatively young technique, called ptychographic
projections with an exposure of 1.5 seconds were
tomography was used. Ptychographic tomography
acquired. Both CT scans were reconstructed using
(Figure 2) is based on small angle scattering of X-
Octopus Reconstruction (Vlassenbroeck et al., 2007;
rays, and combines coherent diffractive imaging
www.octopusreconstruction.eu) .The structure and
(CDI) with tomographic reconstruction techniques

Figure 3: Left: Detail of BSE image – Center: Detail of the same area EDS map – Right: Simplified representation
of EDS map. (Legend: Blue = Quartz, Green = Feldspars and Clays, Red = Iron oxides – images acquired at 100 x
magnification)

89
 de Boever et al.
Figure 4: Registration of 2D images to a 3D volume. Left: SEM images are registered and oriented to the low-
resolution CT scan. Right: EDS maps for Iron registered in the same way. The two different planes provide a
pseudo-3D look at the orientation of the iron laminations.
(Dierolf et al., 2010; Trtik et al., 2013). In CDI, the
sample is hit by a narrow pencil X-ray beam, and
unlike attenuation tomography, where only X-ray
intensity is measured, the detector records maps of
both the absorption and the diffraction patterns of
X-rays. The diffraction pattern of the X-ray beam is
measured at hundreds of (overlapping) positions in
the sample for one particular position, allowing for
a high-resolution 2D ptychographic reconstruction.
In ptychographic tomography, this procedure is re-
peated for a large number of angular intervals, and
using dedicated algorithms, a 3D image is acquired.
Ptychographic tomography has reached 3D reso-
lutions of up to 16 nm (Holler et al., 2014), but as
in most high-resolution techniques, the trade-off is
that sample sizes are extremely small. In this study,
several 50 µm diameter samples were imaged at a
resolution of ± 100 nm, and two 25 µm diameter
samples could be measured at a resolution of about
45 to 60 nm spatial resolution. Ptychographic exper-
iments were carried out at the cSAXS beam line of
the Paul Scherrer Institute in Villigen, Switzerland
(www.psi.ch).
X-ray computed tomography
For ease of reading, results of both tomography
scans (high and lower resolutions) are put together,
despite the fact that they do not follow each other
in the chronology of the used workflow. The recon-
structed slices of the µ-CT scan of the entire ‘large’
sample clearly indicated that a resolution of 8.9 µm
was not enough to resolve all features of the Vosges
sandstone clearly. Although one can have an idea of
the overall structure and porosity, individual grains
and pores could not be segmented and separated
on this sample. However, regions or laminations
with higher or lower porosity, or zones with higher
90
concentrations of dense inclusions could clearly be
identified at this scale.
To be able to do quantitative porosity and grain
analysis, higher resolution was definitely needed.
The reconstructed volume of the 1 mm microplug
showed enough detail to allow a clear separation
between porosity and grains. Grains could be sep-
arated into dense minerals and quartz, clays and
feldspars. Between the latter three mineral classes,
some contrast was visible with the naked eye; how-
ever, grey value segmentation was not possible be-
tween these types. Elongated mica clay minerals,
showing their typical cleavage, could clearly be
identified on these high-resolution images. Poros-
ity analysis of this dataset resulted in a porosity of
16.50 %, with only 0.2 % closed porosity. This points
out that most connections between individual pores
were captured at this resolution and valuable con-
clusions about the pore space could be made for
this sample.
V. Scanning electron microscopy and
energy dispersive spectroscopy
The BSE images acquired gave contrast between
quartz and feldspars, something that was absent in
the low-resolution CT scan. Sadly, this contrast was
not big enough to measure their relative content
using image analysis software directly on the BSE
images. The very heavy, bright minerals were in
very large contrast to the rest of the mineral con-
tent, but rather scarce in a random 2D plane. Point
spectra taught us that the largest of these miner-
als, which already showed up in the µ-CT scans,
contained titanium and are probably titanite (TiO2).
The acquired EDS maps revealed more detailed
information on the distribution of different min-
eral phases; three distributions were measured:
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

potassium, silicon and iron. The grains with
the highest relative amount of silicon could be
identified as quartz (SiO2 ), potassium-rich grains
are (KAlSi3O8) or clay minerals (e.g. muscovite:
KAl2 (AlSi3 O10 )(F,OH)2 ). Places where there was
iron pointed to the presence of iron oxides, which
give the Vosges sandstone its color. The images
indicated a rather uniform distribution of quartz,
feldspars and clay minerals (from most to least
abundant); however, the iron was mainly present
in a distinct layer, running through two out of five
imaged surfaces. It was clear that most of the iron
was not present in big grains, but present in the
cement that binds the grains together. The distinct,
iron-rich layer that were observed in the SEM im-
ages were not clearly visible in the CT-volume, nor
were they easy to see in the BSE images. These
results show that even BSE images do not always
give a full idea about the number of minerals inside
a sample, and that EDS analysis is useful, even if
there are no different phases at first sight.
Data fusion: CT-SEM-EDS
The real strength of the proposed method lies in
the combination of 3D information at various res-
olutions, with good structural and chemical 2D
information. 2D-to-2D image registration between
SEM and EDS images provides grain-per-grain min-
eralogical information, even when contrast in BSE
images is too low (Figure 3). Some SEM machines
are equipped to automatically overlay correspond-
ing chemical information onto the SEM images,
but since landmark based registration was used,
the method is also suitable for merging with other
imaging techniques (e.g. light microscopy, XRF
mapping).
Besides 2D-to-2D image registration methods,
useful to merge 2D imaging techniques to each
other, 2D-to-3D registration of this 2D information
to a 3D image adds an extra dimension to planar
image data. Due to various factors, this 2D-to-3D
registration is not always straightforward and man-
ual or semi-automatic registration is often the way
to go. Problems that can arise are for instance the
fact that SEM images have a depth of view, often
an advantage, but a limiting factor in the auto-
matic registration of SEM images to 2D slices from
an image stack. Also detector position (possibly
creating a skew in SEM images), unknown orienta-
tion of images, etc. can provide difficulties in im-
age registration. Using a semi-automatic landmark
based registration technique, in which correspond-
ing points in the reference and target image were
selected, SEM/EDX images were first rotated into
the same Z plane of the target surface. Next, im-
ages only had to be translated in X and Y direction,
together with a limited degree of warping to correct
for distortion in the images For this final tweaking
step, the bookstein image warp in algorithm Avizo
Fire R was used (Bookstein, 1986).
As a result of this 2D to 3D image registration, a
pseudo-three-dimensional look at the distribution
of different mineral phases is provided. This means
that not only are the iron-rich laminations observed,
they can now be oriented in space (Figure 4), and
their orientation can be compared to that of other
features inside the stone.
A next and final type of image registration is the
registration of the reconstructed microplug CT scan
to the larger CT scan of the base sample (3D-to-3D
registration). This is an important step, since ge-
ological materials like the Vosges sandstone, but
also other construction materials such as concrete
and bricks, are heterogeneous in nature. Therefore,
it is important that one can pinpoint the location
of a sub-sample, especially if this is very small (in
this case, < 1 mm), since the properties derived
from this one microplug might not be represen-
tative for the entire sample. As in the 2D-to-3D
registration step, semi-automatic landmark registra-
tion was performed to create the alignment between
the small and large 3D volume. Figure 5 shows the
importance of this last image registration step. The
small microplug is indicated in red inside the larger
sample, which has been partially made transparent.
The sample contains areas where more or less high-
density minerals are present, and porosity is not
the same throughout the entire sample. Indication
of the location of the first subsample can help to set
a strategy for further subsampling, or to determine
which zones in the sample will have the same (flow)
parameters as the microplug.
Figure 5: Location of the microplug (Ø = 1 mm) in the
full sample (width = 14 mm). Full sample has been made
partially transparent to show variance in dense mineral
distribution.

91
 de Boever et al.
Figure 6: Left: BSE image of clay-rich area inside the Vosges sandstone; Right: FIB-nt slice of a similar area inside the
stone. Different structures can clearly be adsorbed, but not automatically quantified due to low contrast.
VI. Focused ion beam
nanotomography
Using FIB-nt, particular zones of interest could be
selected and analyzed at very high resolution in
3D. Figure 6 shows a BSE image of an area consist-
ing out of fine clay minerals, which could not be
resolved using µ-CT. Using the FIB and electron
beam, a 17 ⇥ 17 ⇥ 10 µm3 volume was imaged, at a
resolution of 16 nm. Due to the non-conductivity of
the material, a lot of charging artifacts were present,
and contrast was low. The images gave a nice in-
dication and view to the microstructure of the clay
minerals; however, it was not possible to do any
type of quantitative image analysis on them. Fur-
ther experiments with serial sectioning, combined
with backscattered electron imaging will be per-
formed in the near future. This will reduce charg-
ing of the sample, and additionally create a higher
contrast between pores and grains.
VII. Ptychographic tomography
Several samples were analyzed and a first exper-
imental sequence imaged 50 µm samples with a
voxel size of 34 nm, and a corresponding spatial
resolution of 89 to 120 nm in 3D. During a later
beam time, two 25 µm sized samples were mea-
sured using the setup described in Holler et al.
(2014), achieving a spatial resolution of 45 nm (voxel
size 17 nm) and 60 nm (voxel size 21 nm) in 3D. The
resulting images were very sharp and show excel-
lent contrast between pores and grains. The images
provide quantitative data about the electron density
at any given point in the sample, meaning that dif-
ferent mineral phases were clearly distinguishable
in the sample.
Figure 7 shows a reconstructed slice of a 25 µm
small fragment of mica minerals, at a spatial resolu-
tion of 45 nm. The results show that ptychographic
tomography is a very promising technique in the
92
geosciences, providing unique and unseen images
of a material’s microstructure. It can however never
be used as a single technique, since representative-
ness of the extremely small samples will never be
enough to characterize an entire material.
VIII. Conclusion
Neither µ-CT, SEM, EDS, FIB-nt and ptychographic
tomography can prove themselves as a perfect, all-
area characterization technique for geological build-
ing materials. However, a combination of some or
all of these methods strengthens them and gives
more complete results on porosity, chemical com-
position and grain sizes inside the material. Of the
used methods, µ-CT is probably the most universal
and flexible technique to measure porosities and
grain sizes, across different scales, especially if dif-
ferent samples and subsamples are analyzed. The
strength of having three-dimensional results, the
fact that the technique has matured a lot over the
past decades, and the rapid progress of computing
power for analysis enables excellent and quantita-
tive data using CT. However, for very small features
and for additional chemical information, data fu-
sion with other techniques is a necessity. For the
first case, relatively new techniques as FIB-nt and
ptychographic tomography can be used. Although
these methods provide very high-resolution images
about a material’s microstructure, they can only
analyze very small volumes. Since these analyzed
volumes are usually smaller than (part of) the grains
and pores of the materials of interest, they can never
be used as the only analysis technique for geolog-
ical building materials. The destructive nature of
FIB-nt will sometimes create issues, since dynamic
processes cannot be analyzed, a problem that is
not present in ptychographic imaging. Chemical
information can be derived from EDS analysis, com-
bined with backscattered electron imaging. EDS has
the advantage of being quite fast, so relatively large
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 7: Reconstructed slice through a mica mineral sample, derived from the Vosges sandstone (resolution = 45 nm;
voxel size = 17 nm)
surfaces can be analyzed, if images are stitched
together.
Characterization of rocks, concrete, mortars and
bricks using image-based techniques will always
remain a multi-scale problem. Thus, several sam-
ples of different sizes need to be analyzed using
different methods. But, if sample selection and
methodology are done correctly, these image-based
techniques can provide fantastic data, in which po-
sition and orientation of certain characteristics and
parameters can be perfectly located.
Acknowledgement
The Special Research Fund of Ghent University
provided the scholarship of drs. Wesley De Boever.
XRE is acknowledged for developing hardware
at the UGCT. We would like to thank ir. Denis
Van Loo for the help during high-resolution µ-CT
experiments.
We thank the entire team of UGCT / PProgres for
help and advice during experiments and analysis.
We acknowledge the Paul Scherrer Institut, Villi-
gen, Switzerland for provision of synchrotron radi-
ation beamtime at beamline cSAXS of the SLS and
would like to thank Ana Diaz and Mirko Holler for
assistance.
The research leading to these results has re-
ceived funding from the European Community’s
Seventh Framework Programme (FP7/2007-2013)
under grant agreement no. 312284
References
Bookstein, F.L., 1986. Size and shape spaces for landmark data in two
dimensions. Stat. Sci. 1, 181–242.
Brabant, L., Vlassenbroeck, J., De Witte, Y., Cnudde, V., Boone, M.N.,
Dewanckele, J., Van Hoorebeke, L., 2011. Three-Dimensional Analysis
of High-Resolution X-Ray Computed Tomography Data with Morpho+.
Microsc. Microanal. 17, 252–263. doi:10.1017/s1431927610094389
Cnudde, V., Boone, M.N., 2013. High-resolution X-ray computed to-
mography in geosciences: A review of the current technology and appli-
cations. Earth-Science Rev. 123, 1–17. doi:10.1016/j.earscirev.2013.04.003
Cnudde, V., Cwirzen, A., Masschaele, B., Jacobs, P.J.S., 2009. Poros-
ity and microstructure characterization of building stones and concretes.
Eng. Geol. 103, 76–83. doi:DOI 10.1016/j.enggeo.2008.06.014
Cnudde, V., Dewanckele, J., Boone, M.M., de Kock, T., Brabant, L.,
Dusar, M., de Ceukelaire, M., de Clercq, H., Hayen, R., Jacobs, P., 2011.
High-Resolution X-Ray CT for 3D Petrography of Ferruginous Sandstone
for an Investigation of Building Stone Decay. Microsc. Res. Tech. 74,
1006–1017. doi:10.1002/jemt.20987
De Winter, D. a M., Schneijdenberg, C.T.W.M., Lebbink, M.N., Lich,
B., Verkleij, a J., Drury, M.R., Humbel, B.M., 2009. Tomography of
insulating biological and geological materials using focused ion beam
(FIB) sectioning and low-kV BSE imaging. J. Microsc. 233, 372–83.
doi:10.1111/j.1365-2818.2009.03139.x
Diaz, A., Trtik, P., Guizar-Sicairos, M., Menzel, A., Thibault, P., Bunk,
O., 2012. Quantitative x-ray phase nanotomography. Phys. Rev. B 85,
020104. doi:10.1103/PhysRevB.85.020104
Dierolf, M., Menzel, A., Thibault, P., Schneider, P., Kewish, C.M.,
Wepf, R., Bunk, O., Pfeiffer, F., 2010. Ptychographic X-ray computed to-
mography at the nanoscale. Nature 467, 436–9. doi:10.1038/nature09419
Gall, J.-C., Grauvogel-Stamm, L., 2005. The early Middle Tri-
assic “Grès à Voltzia” Formation of eastern France: a model
of environmental refugium. Comptes Rendus Palevol 4, 637–652.
doi:10.1016/j.crpv.2005.04.007
Holler, M., Diaz, a, Guizar-Sicairos, M., Karvinen, P., Färm, E., Härkö-
nen, E., Ritala, M., Menzel, a, Raabe, J., Bunk, O., 2014. X-ray ptycho-
graphic computed tomography at 16 nm isotropic 3D resolution. Sci.
Rep. 4, 3857. doi:10.1038/srep03857
Holzer, L., Cantoni, M., 2011. REVIEW OF FIB TOMOGRAPHY, in:
UTKE (Ed.), Nanofabrication Using Focused Ion and Electron Beams. pp.
410–435.
Holzer, L., Indutnyi, F., Gasser, P.H., Münch, B., Wegmann, M., 2004.
Three-dimensional analysis of porous BaTiO3 ceramics using FIB nanoto-
mography. J. Microsc. 216, 84–95. doi:10.1111/j.0022-2720.2004.01397.x
Keller, L.M., Holzer, L., Wepf, R., Gasser, P., Münch, B., Marschall,
P., 2011. On the application of focused ion beam nanotomography in
characterizing the 3D pore space geometry of Opalinus clay. Phys. Chem.
Earth, Parts A/B/C 36, 1539–1544. doi:10.1016/j.pce.2011.07.010
93
 de Boever et al.
Lenoir, N., Marello, S., Viggiani, G., Besuelle, P., Desrues, J., Di
Michiel, M., 2004. X-ray micro tomography characterization of strain
localization upon deviatoric loading of saturated fine-grained stiff soils,
X-Ray Ct for Geomaterials Soils, Concrete, Rocks.
Marcelino, V., Cnudde, V., Vansteelandt, S., Caro, F., 2007. An evalu-
ation of 2D-image analysis techniques for measuring soil microporosity.
Eur. J. Soil Sci. 58, 133–140. doi:DOI 10.1111/j.1365-2389.2006.00819.x
Marszałek, M., Alexandrowicz, Z., Rzepa, G., 2014. Composition
of weathering crusts on sandstones from natural outcrops and architec-
tonic elements in an urban environment. Environ. Sci. Pollut. Res. Int.
doi:10.1007/s11356-014-3312-y
Masschaele, B., Dierick, M., Loo, D. Van, Boone, M.N., Brabant, L.,
Pauwels, E., Cnudde, V., Hoorebeke, L. Van, 2013. HECTOR: A 240kV
micro-CT setup optimized for research. J. Phys. Conf. Ser. 463, 012012.
doi:10.1088/1742-6596/463/1/012012
Prakongkep, N., Suddhiprakarn, A., Kheoruenromne, I., Gilkes, R.J.,
2010. SEM image analysis for characterization of sand grains in Thai
paddy soils. Geoderma 156, 20–31. doi:10.1016/j.geoderma.2010.01.003
Rivas, T., Prieto, B., Silva, B., Birginie, J.M., 2003. Weathering of
granitic rocks by chlorides: effect of the nature of the solution on
weathering morphology. Earth Surf. Process. Landforms 28, 425–436.
doi:10.1002/esp.492
94
Rozenbaum, O., 2011. 3-D characterization of weathered building
limestones by high resolution synchrotron X-ray microtomography. Sci.
Total Environ. 409, 1959–66. doi:10.1016/j.scitotenv.2011.02.014
Shear, W. a., Selden, P. a., Gall, J.-C., 2009. Millipedes from the Grès
à Voltzia, Triassic of France, with comments on Mesozoic millipedes
(Diplopoda: Helminthomorpha: Eugnatha). Int. J. Myriap. 2, 1–13.
doi:10.1163/187525409X462395
Trtik, P., Diaz, A., Guizar-Sicairos, M., Menzel, A., Bunk, O.,
2013. Density mapping of hardened cement paste using ptychographic
X-ray computed tomography. Cem. Concr. Compos. 36, 71–77.
doi:10.1016/j.cemconcomp.2012.06.001
Vlassenbroeck, J., Dierick, M., Masschaele, B., Cnudde, V., Hoore-
beke, L., Jacobs, P., 2007. Software tools for quantification of X-ray
microtomography. Nucl. Instruments Methods Phys. Res. Sect.
a-Accelerators Spectrometers Detect. Assoc. Equip. 580, 442–445.
doi:10.1016/j.nima.2007.05.073
Wildenschild, D., Sheppard, A.P., 2013. X-ray imaging and anal-
ysis techniques for quantifying pore-scale structure and processes in
subsurface porous medium systems. Adv. Water Resour. 51, 217–246.
doi:http://dx.doi.org/10.1016/j.advwatres.2012.07.018
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Microscopy and mechanics

95
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

96
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Investigation of coarse aggregate surface

treatments to enhance concrete strength
Alireza Dehghan⇤,a , Karl Petersona , Philip Zacariasb , Michael Smithb
a Univ. of Toronto, Dept. of Civil Eng.
b St Marys / Canada Building Materials
⇤ alireza.dehghan@mail.utoronto.ca

Abstract

In southern Ontario, the premium coarse aggregate sources for concrete construction are quarried from the Niagara
Escarpment Amabel and Lockport dolostone formations. As access to this resource becomes more and more restricted,
alternative limestone formations are being utilized by the construction industry, but they exhibit inconsistent performance
in terms of concrete compressive strength as compared to traditional sources. Several aggregate treatments, including
mechanical abrasion and surface coatings were explored in this research, but without any marked improvement in
strength. Observations of fracture patterns and aggregate porosity in split tensile cylinders suggest that the presence of
dispersed natural porous aggregate particles in concrete may play a role in the superior strength of concrete made with
Amabel dolostone.

Keywords: : surface texture, interfacial transition zone (ITZ), light weight aggregate (LWA)

I. Introduction surface improved strength. On the other hand, the

even dispersal of water-saturated lightweight aggre-
gates (LWA) has been shown to reduce the overall
The relationship between aggregate surface texture
volume of the ITZ within concrete (Bentz and Sny-
and the compressive strength of concrete has long
der 1999, Bentz 2009). Densification of the ITZ
been a topic of interest in concrete research. Ka-
has also been achieved by two-stage mixing strate-
plan (1959) reported a correlation between surface
gies where water is added incrementally (Tamimi
texture and strength; the rougher the aggregate
1994, Li et al. 2012). The use of aggregate coat-
surface, the higher the concrete strength. Subse-
ings to promote the formation of dense hydration
quent research comparing smooth river gravel to
products along the ITZ is another popular research
crushed river gravel, or where rough aggregates
theme. Nano-scale alumina and silica coatings, as
were artificially made smooth by the application of
well as calcined limestone, pozzolanic, and port-
various adherent coatings, in some cases found that
land cement based coatings have all been tried with
rough surface textures did correlate with higher
varying degrees of success (Muñoz et al. 2010, San-
strength (Nepper-Christensen and Nielsen 1969,
filippo et al. 2010, Yang et al. 1992, Popovics 1987,
Nallathambi et al. 1984, Giacco and Zerbino 1998),
Li et al. 2009, Zhihui et al. 2013). The premise
and in other cases found and in other cases found
that coarse aggregate surface texture is related to
that surface texture did not have a significant in-
compressive strength was explored in this research
fluence (Darwin and Slate 1970, Al-Rawi and Al-
for two different carbonate aggregate sources avail-
Murshidi 1978). The improvement of bond charac-
able in southern Ontario: the Amabel dolostone, a
teristics within the interfacial transition zone (ITZ)
durable rock that caps the weather-resistant Nia-
has long been a focal point of efforts to increase
gara Escarpment for over 400 km between Hamilton
strength in concrete, and is generally achieved by
and Manitoulin Island, and the Gull River lime-
more efficient particle packing near aggregate par-
stone, a stratigraphically older rock that outcrops
ticles through the addition of silica fume or poly-
east of the escarpment (Figure 1). Certainly, many
mer dispersions (Jae-Ho and Robertson 1998, Ben-
other aggregate parameters can influence concrete
tur and Alexander 2000). Other ITZ enhancement
strength, but this research was limited to an inves-
strategies include locally lowering the water to ce-
tigation into alterations of the aggregate surface
mentitious ratio (w/cm) within the ITZ by mixing
texture by abrasion or by the application of spe-
concrete using dried aggregate; in the case of imper-
cial coatings. In this research, a simple qualitative
meable aggregates, Ping and Beaudoin (1992) found
distinction was made between the macroscopically
that the elimination of water layers at the aggregate

97
 Dehghan et al.
Figure 1: Map of a portion of southern Ontario with
quarry locations (Ontario Geologic Survey 2011).
smooth surfaces of coarse aggregate produced from
Gull River limestone, and the rough surfaces of
the Amabel dolostone (Figure 2). No effort was
made to directly quantify the surface texture of the
aggregates either before or after the surface treat-
ments, but the quantification of aggregate surface
texture is a common research theme, especially in
the fields of asphalt concrete and pavement friction
(Wang et al. 2013, Luce et al 2007). Quantitative
treatments of surface texture are less common in
concrete strength research, although there has been
considerable effort recently in the field of fresh con-
crete rheology (Erdoğan et al. 2008).
The geographic distribution and availability of
high quality coarse aggregates for concrete do not
always coincide with areas of high demand. For
example, in Ontario, the rapidly growing Greater
Toronto Area (GTA) is blessed with close proximity
to sources of high quality Amabel dolostone, but
the existing quarries are approaching exhaustion.
At the same time, there is great public resistance
to the opening of new quarries, or to the expan-
sion of existing quarries in the Niagara Escarpment
(D’Alesio and Howlett 2012, The Hamilton Spec-
tator 2012, Dundas 2013, Van Dongen 2014). The
prospects for new sources of Amabel dolostone are
not promising, so the construction industry is ex-
ploring alternative aggregate sources. The GTA is
not alone in this respect. Aggregate shortages are
projected in many North American regions, as sum-
marized in recent geological survey reports (Langer
2011, Clinkenbeard 2012).
For the GTA, the Gull River limestone represents
a potential replacement for Amabel dolostone, but
preliminary mixtures indicate that concrete pro-
duced with Gull River limestone does not con-
98
sistently achieve the same compressive strengths
as control concrete produced with Amabel dolo-
stone. For example, Figure 3a compares compres-
sive strengths from a series of parallel laboratory
trial concrete mixtures exploring various cement
contents using Amabel dolostone and Gull River
limestone (St Marys/CBM & Davroc Consulting
Engineers, 2011).
Of course, for a concrete producer, there are many
strategies to increase strength, from changes in ag-
gregate proportion or gradation, to the incorpora-
tion of silica fume or water-reducing admixtures —
to name a few, but many of these strategies have
limitations. For example, blended cement contain-
ing silica fume is not available at all ready mix
concrete plants, and could raise the cost. Both silica
fume and water reducers will decrease the amount
of bleed water, and when used during hot or dry
weather, may require special care when it comes to
proper concrete curing. The motivation behind this
investigation was to allow the concrete producer
to potentially switch to a surface-treated aggregate
while maintaining their same basic mixture designs.
II. Experimental
Combined vacuum/high-pressure fluorescent
epoxy impregnation
Two representative 355 kg/m3 cement content
concrete split tensile cylinders, one with Ama-
bel dolostone aggregate, and another with Gull
River limestone were retained from the same labo-
ratory trials as shown in Figure 3b. A novel com-
bined vacuum/high-pressure process (Laugesen,
2012) was used to impregnate the cracks with a
fluorescent-dyed (0.8 wt.% solvent yellow 43) low-
viscosity (manufacturer listed 100 – 200 cP @ 100
rpm 23 C) epoxy to allow the qualitative obser-
vation of fracture patterns. The impregnation was
performed as follows:
1. Each cylinder was wrapped circumferentially
with acrylic packing tape in a few locations to
hold the halves together. Sharp corners were
also covered with tape.
2. Each cylinder was slid carefully into a ⇠ 0.5m
long segment of 0.1 mm wall-thickness low
density polyethylene (LDPE) 115mm dia. tube
with a simple overhand knot securely tied at
the bottom (Figure 4a). Tape was also wrapped
around the exterior of the bagged cylinder to
prevent bulging during subsequent filling with
epoxy. Excess tube length was folded back,
leaving the top open for easy access to the
cylinder.
3. Each bagged cylinder was individually placed
in the vacuum chamber (Figure 4b). A 6 mm
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 2: Microscope images of typical aggregate particle surfaces from Milton Amabel dolostone (a), Buckhorn Gull
River limestone (b), Orillia Gull River limestone (c), magnified 10.

dia. 1 mm wall-thickness LDPE tubing was fed inlet.

through a cable gland in the vacuum chamber
lid to connect an epoxy reservoir to the vacuum
chamber. The tubing was pinched closed at a
location between the reservoir and the chamber,
and the chamber evacuated by a rotary vane
vacuum pump. The chamber was maintained
at a pressure of 10 KPa for a period of 2 h.
4. ⇠ 350 ml of freshly mixed fluid epoxy was in-
troduced from the filled reservoir into the open
end of the bagged cylinder by releasing the
clamped segment of tubing. Atmospheric pres-
sure pushed the epoxy from the reservoir into
the vacuum chamber, filling the bagged cylin-
der. The clamp was closed before the epoxy
level in the reservoir dropped below the tubing
5. After introducing the epoxy, the vacuum cham-
ber was vented and the bagged cylinder re-
moved. A sufficient volume of epoxy was in-
troduced during the previous step to maintain
a small head (⇠ 1 cm) of epoxy above the top
of the cylinder after venting (a small drop in
head typically occurs upon venting as epoxy
is pushed into the cracks and pores of the con-
crete). A second overhand knot was securely
tied at the open end of the LDPE tube, making
sure to first squeeze out excess air from the
bagged cylinder. The excess tube was trimmed
away and discarded, and a paper towel used
to remove any remaining epoxy near the knot.

Figure 3: 28 d compressive (a) and split-tensile (b) strengths for concrete with Amabel dolostone and Gull River
limestone at various cement contents.

99
 Dehghan et al.

Figure 4: Steps in combined vacuum/high-pressure epoxy impregnation, knot in LDPE tube (a), vacuum chamber
apparatus (b), pressure chamber apparatus (c).

6. The knotted and bagged cylinder was placed
in a cold ap water filled pressure chamber,
and hydrostatically pressurized to 25 MPa us-
ing a pneumatic-driven liquid pump (Figure
4c). Cold water was used to help dissipate the
heat generated during epoxy polymerization.
The epoxy was allowed to cure in the pressure
chamber overnight.
Steps 4.-6. were performed within a 20 min pe-
riod, so as to not exceed the pot-life of the fluid
epoxy (after which the viscosity starts to increase).
After curing, each cylinder was removed from the
pressure chamber, cut into slabs with a water-
cooled diamond saw, and polished smooth with
a water-cooled hand held diamond pad polisher.
The polished slabs were photographed both under
normal room light conditions, and as illuminated
with ultraviolet light.
dolomite with abundant fossils (Vos 1969). Two
sources of Gull River limestone were used, one from
a quarry near Orillia, Ontario, and another from
a quarry near Buckhorn, Ontario (Figure 1). The
Orillia source consists of a light grey to grey-brown,
micritic to very fine-grained limestone (Armstrong
and Anastas 1993). The Buckhorn source consists
of a pale grey and brown sublithographic to litho-
graphic limestone with occasional shaly seams (Car-
son 1980). All of the coarse aggregates in this study
had a maximum nominal size of 20 mm, and met
CSA A23.1 Group 1 grading requirements. The
physical properties of the aggregate sources are
summarized in Table 1.
III. Surface treatments
Silicon carbide abrasion (SiC-abrasion)

Abraded aggregates were produced in 10 kg

Materials
The source of Amabel limestone used in this re-
search originated from a quarry near Milton, On-
tario (Figure 1). The Milton source consists of a
light grey, porous, medium to coarsely crystalline
batches by adding 3 kg of 1 mm (20 grit) SiC and 3
kg of water to a 100 L capacity concrete drum mixer
and tumbled for 3 h. After tumbling, the aggregate
was retained and rinsed over a 2 mm sieve. The SiC-
abrasion process resulted in a mass loss of ⇠ 6 %

Figure 5: Gull River limestone retained on the 4.75 mm sieve before abrasion (a), after SiC-abrasion (b), and after
qtz-abrasion (c), magnified 2.

100
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 6: Gull River limestone gradation before and after SiC-abrasion (a), qtz-abrasion (b), paste-coating (c), and
powder-coating (d).

Figure 7: SEM/BSE images of untreated Gull River limestone surface (a,b), and after nano-coating (c,d), magnified 50
and 500.

101
 Dehghan et al.

Figure 8: Normal light and corresponding ultra-violet light (inverted) images of fluorescent epoxy impregnated
polished slabs prepared from split-tensile concrete cylinders made with Amabel dolostone coarse aggregate.

for the aggregates. Figs. 5a and 5b compare images
of Buckhorn coarse aggregate particles as-received
and after SiC-abrasion. Figure 6a compares the gra-
dation of the Buckhorn coarse aggregate before and
after SiC-abrasion. SiC-abrasion had little effect on
the gradation, with minor loss of materials passing
the 4.75 mm sieve. The edges of aggregate particles
were rounded after SiC-abrasion, and the aggregate
surfaces had a frosted appearance (Figure 5b).
Natural quartz sand abrasion (qtz-abrasion)
A quartz-rich natural sand from Wakefield, Québec
was used to abrade aggregate in the same manner
as the SiC-abrasion method. The sand originated
from fluvially sorted glacial sediments generated by
the Laurentide ice sheet, and met CSA A23.1 FA1
grading requirements. The qtz-abrasion process
resulted in a mass loss of ⇠ 3.5 %. Figures 5a and
5c compare images of Buckhorn coarse aggregate
particles as-received and after qtz-abrasion. Figure
6b compares the gradation of Buckhorn coarse ag-
gregate before and after qtz-abrasion. Qtz-abrasion
had little effect on the gradation, with minor loss
of materials passing the 4.75 mm sieve. As with
the SiC-abrasion treatment, the edges of aggregate
particles were rounded, and the aggregate surfaces
had a frosted appearance (Figure 5c).

Table 1: Physical properties of aggregate.

Bulk specific gravity, Absorption (%), Flat and, Loss by wash (%),
Formation Location
CSA A23.2-12A CSA A23.2-12A elongated CSA A23.2-5A

Amabel dolostone Milton 2.7 1.32 - -

Buckhorn 2.72 0.44 5.1 0.8
Gull River limestone
Orillia 2.68 0.27 - 1

102
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 9: Normal light and corresponding ultra-violet light (inverted) images of fluorescent epoxy impregnated
polished slabs prepared from split-tensile concrete cylinders made with Gull River limestone coarse aggregate.

Figure 10: Epifluorescent mode (inverted) images of fluorescent epoxy impregnated polished thin sections prepared
from split-tensile concrete cylinders showing typical fractures through Amabel dolostone (a) and Gull River limestone
(b) coarse aggregate particles.

103
 Dehghan et al.

Figure 11: 7 and 28 d compressive strengths for non-air entrained mortar cubes made with Milton Amabel dolostone
(a), Buckhorn Gull River limestone (b), and Orillia Gull River limestone (c).

Colloidal silica coating (nano-coating)
For both the mortar and concrete experiments,
rinsed and dried aggregates were dipped in a col-
loidal silica suspension prepared from tetraethyl
orthosilicate in dilute nitric acid, and then dried
under ambient room conditions as outlined by Chu
et al. (1997) and as refined by Muñoz et al. (2010)
and Sanfilippo et al. (2010). Figure 7 shows a scan-
ning electron microscope back-scattered electron
(SEM-BSE) image of an Orillia coarse aggregate sur-
face both before and after nano-coating. A sieve
analysis was not performed after nano-coating, as
the thickness of the coating would not affect the
gradation in a detectable manner.
Portland cement paste coating (paste-coating)
Paste-coating was conducted in 10 kg batches by
mixing 1 kg of GUL cement and 1 kg of water
in a 100 L capacity concrete drum mixer for 30 s,
followed by the gradual addition of the aggregate
over a 1 min period. After an additional minute
of mixing, the mixer was stopped, and allowed
to rest for 3 min, followed by another 2 min of
mixing. Next, the mixture was emptied over a 2 mm
sieve, allowing excess paste to pass through. The
aggregates then were dispersed in a one-particle
thick layer on a sheet of polyethylene and cured
under ambient room conditions for 5 d before use
in the concrete mixtures. Figure 6c compares the
gradation of the Buckhorn coarse aggregate before
and after paste-coating. Paste-coating had little
effect on the gradation.
Portland cement powder coating (powder-
coating)
The powder-coating procedure was adopted from
the common practice of coating aggregate with hy-
drated lime for use in asphalt concrete (Bayomy
1992, Little et al. 2006). The aggregates were first
soaked in water for 24 h, drained, and then dis-
tributed in a one-particle thick layer over a sheet of
polyethylene and air-dried until achieving a slightly
higher than SSD condition with a moisture content
of sim 4 %. Next, the aggregates in 10 kg batches
were placed in the drum mixer and mixed while
gradually adding 0.5 kg of GUL cement until the

Figure 12: 7 and 28 d compressive strengths for concrete made with untreated Milton Amabel dolostone control, and
made with untreated, SiC-abraded, qtz-abraded, nano-coated, paste-coated, and powder-coated Buckhorn Gull River
limestone.

104
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
aggregates were coated and the excess moisture
completely absorbed by the cement. The coated
aggregates were cured under ambient room condi-
tions for 5 d before use in the concrete mixtures.
Figure 6d compares the gradation of the Buckhorn
limestone coarse aggregate before and after powder-
coating. Powder-coating had little effect on the
gradation.
Mortar mixtures
In addition to the concrete experiments, parallel
mortar experiments were conducted. For all mix-
tures the cement content was fixed at 617 kg/m3 ,
with a w/cm of 0.46. Oven dried aggregates were
used in all mixes. SiC-abrasion and nano-coating
treatments were performed on all three aggregate
sources. In all the mortar mixtures, the crushed
coarse aggregate consisted of 40 % of the total ag-
gregate mass, and the fine aggregate was ASTM
C-778 graded sand from Ottawa, Illinois.
Concrete mixtures
For the concrete mixtures, all five surface treat-
ments were applied to the Buckhorn aggregate,
with the untreated Milton aggregate as a control.
For all mixtures the cement content was fixed at
362.5 kg/m3 , with a w/cm of 0.45. The fine ag-
gregate for all mixtures originated from the same
source as used in the qtz-abrasion treatments, with
a bulk specific gravity of 2.58 and an absorption
of 0.55 %. For all the concrete mixture designs,
the total aggregate volume was kept constant. The
same cement was used as in the mortar mixtures,
along with an air-entraining admixture at a con-
stant dosage of 50 ml/100 kg cement. The target air
content was 6 %. The fresh concrete properties are
summarized in Table 2. Moisture contents of the
untreated and abraded aggregate stockpiles were
monitored and the material pre-weighed according
to the mixture design. The aggregates were then
soaked overnight in water. On the mix day, the
soaked aggregates were drained and re-weighed to
determine excess water assuming complete satura-
tion, and then used for mixing. The nano-coated,
paste-coated, and powder-coated aggregates were
not soaked overnight, but used directly in the con-
crete mixtures. The moisture contents were 0.55 %,
2.69 %, and 0.40 % respectively. The fine aggregate
moisture stockpile moisture content was monitored,
fell within a range of 3.50 – 3.80 %, and was used
directly in the concrete mixtures. A 60 L capacity
pan mixer was used to mix the concrete.
IV. Results
Fracture patterns of split-tensile samples
Figures 8 and 9 compare macro-fracture patterns
from split-tensile tested cylinders made with Am-
abel dolostone to cylinders made with Gull River
limestone. In the Gull River limestone concrete,
well-defined fractures frequently deviate from the
primary crack, and follow along the boundaries of
coarse aggregate particles. Branching from the pri-
mary crack is observed to a lesser extent in the Ama-
bel dolostone concrete. High-porosity coarse aggre-
gate particles are present in the Amabel dolostone
concrete, but absent from the Gull River limestone
concrete. Figure 10 compares typical micro-fracture
patterns within the coarse aggregate particles split
by the primary tensile crack.
Mortar strength
In the mortar trials all three aggregate sources ex-
perienced avg. strength increases of ⇠ 15% with
nano-coated aggregate, and avg. strength increases
of ⇠ 12% with SiC-abraded aggregate as compared
to the untreated aggregate (Figure 11). The nano-
coating and SiC-abrasion treatments also brought
the strength of the Gull River limestone mortars
to similar levels as the untreated control Amabel
dolostone mortar at both 7 and 28 d.
Concrete strength
The Gull River limestone surface treatments did not
provide any improvement in compressive strength
for the concrete mixtures (Figure 12). In most cases,
with the exception of the powder-coating, the com-
pressive strengths were even lower than that of the
original untreated Gull River limestone concrete.
V. Discussion
The surface treatments tried here enhanced com-
pressive strength in the mortar experiments, but
did not enhance compressive strength in the con-
crete experiments. One possible explanation for
the disparity could be a difference in the available
aggregate surface area between the mortar and the
concrete. If the treatments did improve the ITZ,
the available aggregate surface area might posi-
tively correlate with strength. Assuming spherical
mono-sized aggregates with a dia. of 3.56 mm (the
mid-point between the 4.75 and 2.36 mm sieves)
the air-entrained mortars would have a coarse ag-
gregate surface area of about 300 m2 per m3 of
mortar. The same calculation for concrete based on
the sieve mid-points from the gradations shown in
Figure 6 yields a coarse aggregate surface area of
about 1100 m2 per m3 . From these estimates, the
105
 Dehghan et al.
Table 2: Fresh concrete properties.
Temp.( C),
Mixture ID
CSA A23.2-17C CSA A23.2-5C
Milton untreated 28.3
untreated 28.3
SiC-abrasion 28.1
qtz-abrasion 27.7
Buckhorn
nano-coating 27.7
paste-coating 28.4
powder-coating 28.7
proposed surface area hypothesis does not seem
to explain the contrasting performance of mortar
and concrete. One important difference between
the mortar and concrete experiments for the abra-
sion treated aggregate was the moisture condition:
for the mortars the aggregate was mixed dry, and
for the concrete the aggregate was pre-saturated
and coated with surface water. Ping and Beaudoin
(1992) reported that the presence of surface wa-
ter layers on impermeable aggregates can degrade
strength. Gull River limestone is not completely im-
permeable, but it does have a very low absorption
as compared to the Amabel dolostone (Table 1). An-
other possibility is the rounding of particles from
the abrasion treatments (Figure 5) led to a reduction
in strength as compared to the more angular un-
treated particles (Kaplan 1959). An explanation for
the low strength of the nano-coated, paste-coated,
and powder-coated treatments might be found in
the more aggressive mixing environment of the pan
mixer as opposed to the mortar mixer. Mixing en-
ergy was not quantified in this research. However,
based on the relatively smooth flow of the mortar
between the bowl and the paddle, and the noisy
scraping of the concrete between the pan and mixer
blades, it is not unreasonable to speculate that the
applied coatings may have been compromised in
some way for the concrete mixes. But, if this were
the case, it does not explain why the concrete with
coated aggregates would be weaker than concrete
with untreated aggregate.
Although a satisfactory explanation for the con-
trast in performance between the mortar and con-
crete experiments has not been identified, the flu-
orescent epoxy impregnated split tensile cylinders
do provide an additional clue as to the superior
performance of the Amabel dolostone. In Figure
8, porous aggregates are dispersed throughout the
Amabel dolostone concrete, but not in the Gull
River limestone concrete (Figure 9). The presence
of these naturally porous aggregate particles is per-
haps analogous to concrete containing saturated
LWA. Concrete with LWA has generally demon-
strated increased strength relative to concrete with
106
Slump(mm), Unit wt.(kg/m3 ), Air content (%),
CSA A23.2-6C CSA A23.2-4C
65 2353.3 4.5
55 2344.7 4.2
70 2326.1 4.3
65 2319.8 4.5
85 2304.1 4.5
70 2292 4.1
55 2318.4 4.1
traditional coarse aggregates (Kayali 2008). The
improved performance is attributed to enhanced
internal curing, thereby eliminating or reducing the
ITZ porosity, as well as the penetration of hydration
products into the highly textured LWA surface (Er-
dem et al. 2012). Another interesting observation
involved frequent branching of the primary tensile
crack in coarse aggregate particles from the Gull
River limestone. Branching of the primary tensile
crack in coarse aggregate particles from the Amabel
dolostone was seldom observed (Figure 10).
VI. Conclusion
Experiments to alter the surface characteristics of
Gull River limestone aggregate to enhance compres-
sive strength showed promise in mortar. Compres-
sive strengths for surface-treated Gull River lime-
stone mortars were equivalent to untreated Amabel
dolostone mortars. However, when the experiments
were scaled up to concrete the surface-treatments
failed to perform. A comparison of sectioned and
polished slabs prepared from split tensile cylinders
from Amabel dolostone and Gull River limestone
concrete showed a higher frequency of cracking
along the aggregate boundaries of Gull River aggre-
gate particles, as well a higher frequency of within-
aggregate branching of the primary tensile crack.
In addition, porous aggregate particles were dis-
persed throughout the Amabel dolostone concrete,
and were not present in the Gull River limestone
concrete. These naturally porous Amabel dolostone
particles perhaps perform the same role as LWA
particles in the improvement of the ITZ, and con-
tribute to the superior compressive strength perfor-
mance of concrete made with Amabel dolostone.
The exploration of concrete strength enhancement
strategies involving the blending of Gull River lime-
stone aggregate with LWA is recommended.
Acknowledgement
This research was supported by NSERC Engage
Grant EG #459132-13 thanks to the kind efforts of
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Daman Panesar. The assistance of student workers
Gianluca Barbieri and Oleksiy Chernoloz is greatly
appreciated, as well as Iarley Sampaio Libos and
Leonardo De Carvalho Souza Santa Rita, both of
whom were sponsored by the Ciência sem Fron-
teiras CNPq Student Research Placement Opportu-
nity. The assistance from the staff of the St Marys
Leaside Lab was invaluable throughout, and the
improved vacuum apparatus used in this research
would not have been possible without Nicola Tisato
and Giovanni Buzzeo.
References
Al-Rawi, R.S., and Al-Murshidi, K. (1978): “Effects of maximum size
and surface texture of aggregate in accelerated testing of concrete.” Ce-
ment and Concrete Research, 8: 201-210.
Armstrong D.K., and Anastas A.S. (1993): “Palaeozoic Geology of
the Orillia Area, Southern Ontario.” Ontario Geological Survey Open
File Map 222, Geological Series, Sudbury.
Bayomy, F. (1992): “Development and analysis of cement-coated ag-
gregates for asphalt mixtures.” In Effects of aggregates and mineral
fillers on asphalt mixture performance ASTM STP 1147. Edited by
Meininger R.C. American Society for Testing and Materials, Philadel-
phia, pp. 19-34.
Bentur A., and Alexander M.G., (2000): “A review of the work of
the RILEM TC 159-ETC: Engineering of the interfacial transition zone in
cementitious composites.” Materials and Structures, 33: 82-87.
Bentz, D.P., and Snyder, K.A. (1999): “Protected paste volume in con-
crete: Extension to internal curing using saturated lightweight fine ag-
gregate.” Cement and Concrete Research, 29(11): 1863-1867.
Bentz, D.P. (2009): “Influence of internal curing using lightweight
aggregates on interfacial transition zone percolation and chloride ingress
in mortars.” Cement & Concrete Composites, 31: 285-289.
Carson, D.M. (1980): “Paleozoic Geology Burleigh Falls - Peterbor-
ough Area District of Southern Ontario.” Ontario Geologic Survey Map
P. 2337, Geological Series. Sudbury.
Chu L., Tejedor M.I., and Anderson M.A. (1997): “Particulate sol-gel
route for microporous silica gels.” Microporous Materials, 8: 207-213.
D’Aliesio, R., and Howlett, K. (2012) November 21: Critics celebrate
surprise end of mega quarry north of Toronto, The Globe and Mail
[online]. http://www.theglobeandmail.com/news/national/critics-
celebrate-surprise-end-of-mega-quarry-north-of-toronto/article5520026/
[cited 31 October 2014].
Darwin, D., and Slate, F.O. (1970): “Effect of paste-aggregate bond
strength on behavior of concrete.” Journal of Materials, 5: 86-98.
Dundas, D. (2013) March 9: Carlisle quarry proposal
is dead, The Hamilton Spectator [online]. Available from
http://www.thespec.com/news-story/2525048-carlisle-quarry-
proposal-is-dead/ [cited 31 October 2014].
Erdem, S., Dawson, A.R., and Thom, N.H. (2012): “Impact load-
induced micro-strucutral damage and micro-structure associated me-
chanical response of concrete made with different surface roughness and
porosity aggregates.” Cement and Concrete Research, 42: 291-305.
Erdoğan, S.T., Martys, N.S., Ferraris C.F., and Fowler D.W. (2008):
“Influence of the shape and roughness of inclusions on the rheological
properties of a cementitious suspension.” Cement & Concrete Compos-
ites, 30: 393-402.
Giaccio G., and Zerbino R. (1998): “Failure mechanism of concrete –
combined effects of coarse aggregates and strength level.” Advances in
Cement Based Materials, 7: 41-48.
The Hamilton Spectator. (2012) October 11: Nelson quarry
expansion rejected; mayor delighted [online]. Available at:
http://www.thespec.com/news-story/2258792-nelson-quarry-
expansion-rejected-mayor-delighted/ [cited 31 October 2014].
Jae-Ho, K., and Roberston, R.E. (1998): “Effects of polyvinyl alcohol
on aggregate-paste bond strength and the interfacial transition zone.”
Advanced Cement Based Materials, 8: 66-76.
Kaplan, M.F. (1959): “Flexural and compressive strength of concrete
as affected by the properties of coarse aggregates.” Journal of the Amer-
ican Concrete Institute, 30 (11): 1193-1208.
Kayali O. (2008): “Fly ash lightweight aggregates in high performance
concrete.” Construction and Building Materials, 22(12): 2393-2399.
Langer W. (2011): “Aggregate resource availability in the coterminous
United States, including suggestions for addressing shortages, quality,
and environmental concerns.” Open-File Report 2011-1119, U.S. Geolog-
ical Survey, Reston, Virginia.
Laugesen P. (2012): Personal communication, Pelcon Materials & Test-
ing ApS,.
Li J., Xiao H., and Zhou Y. (2009): “Influence of coating recycled
aggregate surface with pozzolanic powder on properties of recycled ag-
gregate concrete.” Construction and Building Materials, 23: 1287-1291.
Li W., Xiao J., Sun Z., Shah S.P. (2012): “Interfacial transistion zones in
recycled concrete with different mixing approaches.” Construction and
Building Materials, 35: 1045-1055.
Little D., Epps J., Sebaaly P.E. (2006): “The benefits of hydrated lime
in hot mix asphalt.” National Lime Association. Arlington Virginia.
Luce, A., Mahmoud E., Masad, E., and Chowdhury, A. (2007): “Rela-
tionship of aggregate microtexture to asphalt pavement skid resistance.”
Journal of Testing and Evaluation, 35(6): 1-11.
Muñoz, J.F., Meininger, R.C. and Youtcheff, J. (2010): “New possibili-
ties and future pathways of nanoporous thin film technology to improve
concrete performance.” Transportation Research Record: Journal of the
Transportation Research Board, 2142: 34-41.
Nallathambi P., Karihaloo B.L., Heaton B.S. (1984): “Effect of speci-
men and craqcksizes, water/cement ratio and coarse aggregate texture
upon fracture toughness of concrete.” Magazine of Concrete Research,
36(129): 227-236.
Nepper-Christensen, P., and Nielsen, T.P.H. (1969): “Modal determi-
nation of effect of bond between coarse aggregate and mortar on com-
pressive strength of concrete.” Journal of the American Concrete Insti-
tute, 66( 1): 69-72.
Ontario Geological Survey (2011): 1:250 000 scale bedrock geology of
Ontario; Ontario Geological Survey, Miscellaneous Release—Data 126-
Revision 1. ISBN 978-1-4435-5705-4 [zip file]
Ping, X., and Beaudoin, J.J. (1992): “Effects of transition zone mi-
crostructure on bond strength of aggregate-portland cement paste inter-
faces.” Cement and Concrete Research, 22: 23-26.
Popovics, S. (1987): “Attempts to improve the bond between cement
paste and aggregate.” Materials and Structures, 20: 32-28.
Sanfilippo J., Muñoz J.F., Meininger R.C., Tejedor M.I., Anderson M.A.
and Cramer S. M. (2010): “Nanotechnology to manipulate the aggregate-
cement paste bond effects on mortar performance.” Transportation Re-
search Record: Journal of the Transportation Research Board, 2142: 29-
33.
St Marys/CBM & Davroc Consulting Engineers (2011) Buckhorn
Phase II Report.
Tamimi A.K. (1994): “The effects of a new mixing technique on the
properties of the cement paste-aggregate interface.” Cement and Con-
crete Research, 24(7): 1299-1304.
Van Dongen, M. (2014) January 2: Sprawling Flamborough
quarry plans to grow, The Hamilton Spectator [online]. Available at:
http://www.thespec.com/news-story/4296115-sprawling-flamborough-
quarry-plans-to-grow/ [cited 31 October 2014].
Vos, M.A. (1969): “Stone Resources of the Niagara Escarpment.” On-
tario Department of Mines, Industrial Mineral Report 31, Sudbury.
Wang, L., Sun, W., Tutumluer, E., and Druta, C. (2013): “Evaluation of
aggregate imaging techniques for quantification of morphological char-
acteristics.” Transportation Research Record: Journal of the Transporta-
tion Research Board, 2335: 39-49.
Yang, S., Zhongzi, X., Ping, X., and Mingshu, T. (1992): “New method
of enhancing cement-aggregate interfaces, ideal aggregate and its effects
on interfacial microstructures.” Cement and Concrete Research, 22: 612-
620.
Zhihui Z., Shoude W., Lingchao L. and Chenchen G. (2013): “Eval-
uation of pre-coated recycled aggregate for concrete and mortar.” Con-
struction and Building Materials, 43: 191–196.
107
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

108
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

ESEM-BSE coupled with rapid

nano-scratching for
micro-physicochemical analysis of marine
exposed concrete
Damian Palin⇤,a ; Arjan Thijssena , Virginie Wiktora,b
Henk M. Jonkersa , and Erik Schlangena
a Delft University of Technology, Faculty of Civil Engineering & Geosciences,
Section of Materials and Environment, Stevinweg 1, 2628 CN Delft, The Netherlands.
b Bartels Building Solutions, Linie 524, 7325 DZ Apeldoorn, The Netherlands.
⇤ d.palin@tudelft.nl

Abstract

Ordinary Portland cement (OPC) mortar specimens submerged in sea-water were analysed through environmental
scanning electron microscopy (ESEM) in back scattered electron (BSE) mode and nano-scratching. Results from
both sets of analysis show the presence of distinct phases associated with aragonite, brucite and cement paste. Phases
associated with porosity and aggregates were also distinguishable through the BSE analysis and less defined in the
nano-scratch data. This study indicates the powerful nature of coupling BSE image analysis with nano-scratching to
obtain information on the quality of concrete. Work is underway to improve the method in order to apply it for better
understand on the micro-physicochemical properties of marine exposed concrete.

Keywords: marine environment, concrete, ESEM-BSE, nano-scratches, micro-physicochemical.

I. Introduction ing of nano- and micro-mechanical material prop-

Concrete quality must derive from its micro-
chemical and-physical properties. Concrete may un-
dergo degradation and healing processes which can
affect its quality and subsequent durability. Evalua-
tion and interpretation of the relationship between
micro-chemical and -mechanical analysis should
then provide key insights for better assessment of
concrete durability. Since its use in the early1980s
(Scrivener and Pratt, 1983), electron microscopy in
backscattered electron (BSE) mode has shown great
potential for assessing cementitious materials. BSE
analysis produces a grey level image whose grey
level values are based on the mean atomic num-
ber of the target providing chemical information
on the sample being analyzed. Scrivener and Pratt
(Scrivener and Pratt, 1983; Scrivener et al. 1986;
Scrivener, 2004) in their pioneering work revealed
local variations in the microstructure of cementi-
tious materials allowing the identification and quan-
tification of distinct phases. A major draw back of
the technique is, however, its inability to access ma-
terials mechanical properties. Recent advances in
nanoindentation have allowed greater understand-
erties (Sanchez and Sobolev, 2010). Specimens pre-
pared for nanoindentation must be smooth and flat
making them ideally suited for BSE analysis. Re-
cent combination of nanoindentation with scanning
electron microscopy (SEM) in BSE mode has pro-
vided new insights into the micro-physicochemical
properties of cementitious materials (Chen et al.,
2010). Nanoindentation by virtue analyses me-
chanical properties at a point meaning that a large
number of indents are necessary when analysing
heterogeneous materials. An adaptation of nanoin-
dentation is the nano-scratch technique whereby
the indenter is scratched across a sample surface
providing information on its scratch hardness. The
continuous nature of the nano-scratch technique
provides far greater information over time, and
also property gradients and discontinuities across
interphases, than its nanoindentation counterpart.
This work is part of a larger study looking at the
development of autonomous self-healing concrete
for application in the marine environment. To the
authors knowledge this is the first paper to combine
BSE image analysis with nano-scratching for micro-
physicochemical material analysis of cementitious

109
 Palin et al.

Table 1: Mix-design for mortar specimens. Sample preparation and micro-physicochemical

analysis
Amount [kg/m3 ]
Constituent CEM I 42.5 N Cubes submerged for 140 days were removed from
Cement 507 the sea-water, air-dried, impregnated with epoxy
Water 253 and polished sections prepared. In order to make
Water cement ratio 0.5 the thickness of the specimens comparable and
Sand fraction [mm]: level, a thin section machine was used for cutting
1-2 608 and grinding the specimens. After grinding, each
0.5-1 426 sample was polished on a lapping table with 6 µm
0.25-0.5 167 (1 hour), 3 µm (1 hour), 1 µm (1 hour) and 0.25
0.125-0.25 319 µm (4 hours) diamond spray (DP-Spray P, Struers,
Copenhagen, Denmark). In between each spray the
samples were submerged in ethanol and ultrasoni-
materials, illustrating the potential of the combined cated (5 min) to remove debris. Following polishing
technique for analyzing marine exposed concrete. samples were kept in a desiccator until testing.
Scratch locations were identified before scratch-
ing and digital images made using ESEM (Philips
XL 30 ESEM, Eindhoven, Netherlands) in backscat-
II. Experimental approach tered electron (BSE) mode. The instrument was
operated at 20 kV accelerating voltage and at a
Specimen preparation working distance of 10 mm between the final con-
denser lens and the specimen. The spot size was 5.0
Mortar cubes were cast from OPC (CEM I 42.5 N, µm and the magnification of the images was 500⇥.
ENCI, the Netherlands) in accordance with EN Agilent Nano Indenter G200 (Agilent Technolo-
1015-11 (EN, 1999). The applied mortar mix de- gies, Oak Ridge, Tennessee, USA) with a diamond
sign is shown in Table 1. 24 hours after casting Berkovich tip was then used to produce the nano-
the mortar specimens were carefully removed from scratches. A quartz standard was indented before
their moulds, tightly sealed in polyethylene plas- and after each test series to ensure accuracy. A
tic bags and kept at room temperature for a total series of scratches were made at a wear load of
curing period of 28 days. 1 mN and velocity of 25 µm.s 1 . These scratches
were 25 µm apart, perpendicular to the specimen
surface and ran for a length of 200 µm. BSE images
Autogenous healing incubation conditions were taken of the scratched specimens. Compari-
son between the initial and subsequent BSE images
Plastic buckets were prepared containing 4 l of could then be made. MATLAB R2013b (The Math-
synthetic sea-water (20±2 C). The synthetic sea- Works, Inc., Natick, Massachusetts, USA) was used
water, to be called sea-water for the remainder of to isolate the scratch line pixels in the unscratched
the text, was produced from technical grade chemi- images. Figure 1 (C and D) shows two scratch lines
cals (Sigma-Aldrich), the composition of which is plotted in the form of two histograms. Each plot
shown in Table 2. Three cubes were submerged in is the average grey level values of 6 pixels approxi-
each bucket. Water was changed every two weeks mating the scratch width.
to mimic in situ conditions and prevent ion deple-
tion. The buckets were kept open to the atmosphere
during the experiment to allow for gas diffusion III. Results and discussion
across the water-air interface.
Figure 1 (A) and (B) show BSE micrographs of
selected areas of the mortar specimen after submer-
Table 2: Synthetic sea-water composition based after the sion in sea-water. Detailed considerations on the
major constituents of sea-water (Stumm and Morgan, effects of submerging concrete in sea-water have
1996). been given elsewhere (Palin et al., 2015, Glasser et
al., 2008; Conjeaud, 1980). Based on this work it is
Compound Amount [g/l] known that concrete tends to leach ions increasing
NaHCO3 0.19 permeability, making the concrete susceptible to
CaCl2 · 2H2 O 1.47 further attack including erosion due to wave action.
MgCl2 · 6H2 O 10.57 Reaction of these ions with sea-water constituents
Na2 SO4 · 10H2 O 9.02 can also form an aragonite-brucite mineral layer on
KCl 0.75 its surface, which may reduce permeability, con-
NaCl 24.08 tributing specific properties to concrete. Evidence

110
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 1: (A) and (B) BSE micrographs of the scratch locations; (C) and (D) the histograms grey level distributions
along the scratch lines; and (E) and (F) their corresponding scratch depth profiles.
of these phenomena can be seen in the micrograph
with the light grey area on the left hand side rep-
resenting aragonite and the darker grey to its right
representing brucite. This double layer is formed
on top of the mortar, which can be seen as the area
on the right hand side of the micrograph. Running
horizontally through the micrograph are two white
lines, which graphically represent the scratch paths.
The histograms below the micrographs (Figure 1
(C and D)) represent the grey scale levels along
the scratch lines. Due to compositional and hence
grey level variations along the lines various phases
can be distinguished. The relatively flat sections
towards the left of the histograms represent the ho-
mogenous phases of aragonite and brucite. As we
move towards the right hand side of the two his-
tograms more variance can be seen in the grey level
values representing variance in the mortars chemi-
cal composition, darker levels representing porosity
and lighter levels anhydrous phases (Zhang et al.,
2011). Figure 1 (E) and (F) show two nano-scratch
profiles. Scratch depth variations are based on the
micromechanical properties of the phases. We can
see from the nano-scratch test that both aragonite
and brucite have similar scratch profiles, indicat-
ing similar hardnesses in line with the literature
(Santhanam, 2013). This is interesting as the two
phases have quite different grey scale values and
hence chemical properties. However like the BSE
histograms, the relative flatness of the scratch pro-
files represents the homogeneity of the two min-
eral phases. Once the nano-indenter leaves the
aragonite-brucite double layer and begins scratch-
ing the mortar, the scratch profile becomes more
varied representing the heterogenic nature of the
material. Shallower depths represent harder materi-
als such as aggregates and deeper depths porosity.
Nano-scratch analysis provides information on the
nano- and miro-mechanical properties of cementi-
tious materials such as their scratch and mar resis-
tance. Interesting properties when trying to better
assess the performance of cementitious materials,
particularly those in the marine environment.
The objective this research was to see if BSE mi-
crographs could be coupled with nano-scratching
to assess the micro-physicochemical properties of
concrete exposed to the marine environment. Dis-
tinct phases can be seen in both sets of analysis.
Both techniques showed aragonite and brucite to
have relatively flat profiles representative of their
homogeneity, while the mortar profiles having more
variance representative of its heterogeneity. We also
saw correlation between some of the peaks and
dips of the two techniques within the mortar rep-
111
 Palin et al.
resenting aggregates and pores. Although good
correlation can be seen between the two methods,
differences do remain. These differences may be
based on fundamental differences between the two
techniques, but may also be explained by differ-
ences in the micro interaction volumes of the two
techniques. Any difference in the interaction vol-
umes would mean that the mechanical information
provided by the nano-indenter and the chemical
information provided by ESEM-BSE would not be
directly comparable (Chen et al., 2010). Further
investigation is required to assess whether the in-
teraction volumes of the two techniques are closely
comparable in location and size.
IV. Conclusion
BSE imaging and nano-scratching techniques have
been combined for micro-physicochemical infor-
mation on cementitious materials. Data generated
from the two techniques correlated well, with both
able to distinguish between phases associated with
aragonite, brucite and cement paste. Though fun-
damentally different, improved correlation between
the techniques may be achieved by better matching
the micro interaction volumes. Work is underway
to assess and better match the interaction volumes
of the two techniques. On a general level, this study
illustrates the great potential in coupling BSE mi-
croanalysis and nano-scratching for advancing our
understanding on the relationships between the
chemistry and mechanics of cementitious materials,
particularly those in the marine environment.
112
V. Acknowledgements
The authors wish to thank Ravi Patel of the Uni-
versity Gent for his input regarding the MATLAB
analysis, while the research leading to these re-
sults has been funded through the European Union
Seventh Framework Programme (FP7/2007-2013)
under grant agreement no 290308 – SheMat.
References
Chen, J.J., Sorelli, L., Vandamme, M., Ulm, F.-J., Chanvillard, G. (2010):
“A Coupled Nanoindentation/SEM-EDS Study on Low Water/Cement
Ratio Portland Cement Paste: Evidence for CSH/Ca(OH)2 Nanocompos-
ites” Journal of American Ceramic Society. Vol. 5, 5, pp. 1484-1493.
Conjeaud, M. (1980): “Mechanism of sea water attack on cement mor-
tar”, ACI Special Publication, 65
Glasser, F.P., Marchand, J., Samson, E. (2008): “Durability of concrete
— degradation phenomena involving detrimental chemical reactions”,
Cement and Concrete Research. 38 226-246.
Palin, D., Wiktor, V., Jonkers, H.M. (2015): “Autogenous healing of
marine exposed concrete: Characterization and quantification through
visual crack closure”, Cement and Concrete Research, 73, pp. 17-24.
Sanchez, F., Sobolev, K. (2010): “Nanotechnology in concrete - A re-
view”, Construction and Building Materials. 24 pp. 2060-2071.
Santhanam, M., (2013): “Magnesium Attack of Cementitious Materi-
als in Marine Environments, in: Performance of Cement-based Materials in
Aggressive Aqueous Environments”, Springer, pp. 75-90.
Scrivener, K.L., (2004): “Backscattered electron imaging of cementi-
tious microstructures: understanding and quantification”, Cement and
Concrete Composites. 26 pp. 935-945.
Scrivener, K.L., Patel, H., Pratt, P., Parrott L. (1986): “Analysis of
phases in cement paste using backscattered electron images, methanol
adsorption and thermogravimetric analysis, in: MRS Proceedings”, Cam-
bridge Univ Press, pp. 67.
Scrivener, K.L., Pratt, P. (1983): “Characterisation of Portland cement
hydration by electron optical techniques, in: MRS Proceedings”, Cam-
bridge Univ Press, pp. 351.
Stumm, W., Morgan, J., (1996): “Aquatic chemistry, chemical equili-
bra and rates in natural waters”, Env Sci Technol.
Zhang, X.Y., Gallucci. E., Scrivener, K. (2011): “Prognosis of Al-
kali Aggregate Reaction with SEM, Advanced Materials Research.” 194 pp.
1012-1016.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Looking at glass from a different angle:

new insights into fracture patterns
through transmitted light microscopy
Olga van der Velde⇤ , Oğuzhan Çopuroǧlu, Fred A. Veer
Delft University of Technology, Delft, The Netherlands
⇤ olgavdvelde@gmail.com

Abstract

This paper shows the benefit of using transmitted light microscopy together with a Z-scanning software in frac-
tographical analyses of glass. The strength of glass is largely dependent on processes that happen at the microscale.
In this research, 52 plates were fractured in a biaxial tensile test. These were divided into five categories according
to their fracture pattern. 6 plates were examined with a polarised light microscope and photographed with the Z-axis
scanning function. This revealed fracture markings that are barely visible with the naked eye and overlooked when only
performing a microscopic analysis of the fracture surface. This led to the conclusion that transmitted light microscopy
on glass’ fracture pattern is a valuable addition in glass fractography. It gives the researcher an overview of all fracture
markings and flaws in one image. This can be used as a guide to find the fracture origin and it gives new information
on the crack propagation and local failure processes.

Keywords: glass, microscopy, fractography

I. Introduction

In the last decades there has been an increase of

the use of glass as a building construction material.
Next to window panes, glass can also be used to
build columns, walls and beams. However, a true
glass revolution is impeded by its uncertainty in
strength. Strength tests performed on glass can
yield values as high as 150 MPa and as low as
20 MPa (Veer, 2007). Mostly the lowest value is
assumed to be the strength of glass in structural Figure 1: Flaws categorised by location (Molnár et al.
design. This is a waste of material when the true 2012)
strength of the glass is actually much higher. There-
fore, more research in the strength of glass is desir-
able. Glass researchers generally agree that these In this research 52 plates were tested in a biaxial
lower strengths are caused by flaws that weaken tensile test. Traditionally, the microscopy part of a
the glass (Bourhis, 2014, Quinn, 2007, Veer, 2007, fractographical analysis is only concerned with the
Wurm, 2007). The types of flaws that occur in glass fracture surface of the broken pieces. This paper
can be categorised by their location, see Figure 1. explores the addition of observing the fracture
This paper follows this classification (Molnár et al. pattern underneath the microscope. In other words,
2012): the angle of observation is turned 90 degrees.

• Surface, such as scratches from mishandling or

pits from weathering
• Edge, such as grinding or polishing damage
• Volume, such as air bubbles or inclusion
formed in the glass during its manufacturig
process.
Glass fractography
Quinn (Quinn, 2007) stated that the fractographical
analysis of a specimen is just as important as the
fracture stress found in the strength test. He argued
that a fracture caused by one type of flaw cannot

113
 van der Velde et al.
Figure 2: Wallner lines (Shinkai, 1994)
Figure 3: Regions marking the origin of fracture (Quinn,
2007)
be statistically compared to one caused by another
type. He promoted fractographical research to find
the origin and fully understand the strength. Such
an analysis of glass usually starts with a study
of the fracture pattern. Either the specimens are
taped to hold the fragments together or the shards
are reassembled after failure. From the pattern,
the origin crack can be determined by finding the
crack that started branching first. The analysis
finishes with collecting the fragments at that crack
and examining the fracture surface. The flaw that
caused the failure can be found by looking for
certain marks. Wallner lines at this surface are
oriented away from the fracture origin, see Figure 2.
Following them back will lead to the flaw of origin
(Shinkai, 1994). Around this flaw special marks
can be found that prove it is the origin, see Figure
3 (Quinn, 2007). Lastly, twist hackle is sometimes
reported, see Figure 4. Twist hackle shows the local
direction of crack propagation.
This is the ideal situation, but sometimes the
fracture mechanism is more complex. Therefore,
an additional procedure is suggested in this re-
search: examining the fracture pattern underneath
the microscope before collecting the fragments. The
expectation is this will give more insight into the
crack propagation and failure process.
114
Figure 4: Twist hackle (Quinn, 2007)
Figure 5: Ring on ring test values
II. Experimental
The strength values of the glass in this research
were determined by a Ring-on-Ring (RoR) test. It
was chosen because the fracture stress is not in-
fluenced by the flaw orientation and because it in-
troduces a large area in the glass under constant
bending stress. This stress can be calculated by the
formula below (Morrel, 1998). Also see Figure 5.
3P a2 b2 a
sb = {(1 n) + (1 + n)ln } (1)
2ph2 2R2 b
The dimensions used are a = 45 mm, b = 20
mm, R = 141 mm, h = 3.9 mm, n = 0.22. P is
the load at failure. The dimensions are based on
recommended ratios (Fessler and Fricker, 1984,
Morrel, 1998). 52 soda lime silica glass plates
were RoR tested in an Instron 8874 loading device.
The specimen dimensions are shown in Figure
6. The cruciform shape was chosen to be able to
introduce compressive stresses in a later stage
of the total research. The specimens were taped
with transparent foil, so that the fracture pattern
could be studied after failure. Friction is known
to be an issue in RoR tests. Various precautions
were tried to minimise this. A full description
and evaluation of them can be found in this
research’s corresponding thesis (van der Velde, in
preparation).
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 6: Dimensions of the glass specimen
After the RoR test, six of the fractured plates
were examined by a Leica DM2500P petrographic
microscope equipped with polarization accessories
and semi-apochromat (fluorite) objectives. Optical
photomicrographs of the glass specimens were
acquired with a Leica DFC310FX digital camera
at 1392 ⇥ 1040 uninterpolated resolution. In
order to overcome the limited depth of field of the
compound microscope, images were acquired
through Leica LAS Live Z-Builder software which
enabled dynamic emergence of the single in-focus
image. The resulting image has a significantly large
depth-of-field which renders the microstructural
features of a glass specimen on a two dimensional
final image. The advantage of this technique is that
it makes all flaws and fracture markings that exist
throughout the depth of the glass visible in a single
field of view.
Two light settings were used:
• Transmitted light: polarised light enters the
specimen from underneath illuminating all
flaws inside the glass,
• Reflected light: the specimen is lighted from
above, mainly illuminating the surface.
Unless mentioned otherwise, the transmitted light
setting was used. All specimens were observed
with their tensile side up.
III. Results
As is the case in most strength tests on glass, the
results were quite variable, with strength values
Figure 7: At high failure stress the pattern can include
circumferential cracking at the loading ring (Quinn,
2007)
ranging from 42 MPa to 143 MPa. What was no-
table is that there was a difference in fracture pat-
tern among the results. Some plates had the ex-
pected pattern: fracture origin inside the loading
ring (category I) or at or just outside the loading
ring (category R). Some of these were accompanied
by a circumferential cracking pattern at the load-
ing ring, as is reported in literature (?), see Figure
7. However some also showed such a pattern at
the support ring, which is not mentioned in liter-
ature (noted by an additional S). This might have
two explanations: either the cracking pattern is cre-
ated by a punching shear mechanism caused by
friction between the glass and rings or the pattern
is created by shock waves brought along by un-
stable crack growth interacting with the branching
cracks. Fractographical analysis might conclude
this. Lastly, a few plates had a line-shaped frac-
ture origin which is an indicator for a line-shaped
flaw, such as a scratch on the surface (noted by an
additional L). The plates that were examined with
the microscope are depicted in the pictures on the
next pages. The white circles mark the contact of
the loading and support ring. The red arrows mark
the likely origin crack. TT3, TT40 and TT10 were
chosen because they had the smallest failure stress:
41.79 MPa, 56.16 MPa and 57.59 MPa, respectively.
This is thought to be the case due to a line defect.
Microscopical examination might prove this. The
other three plates form a good representation of the
fracture pattern categories.
Fractographical analysis
An analysis of each plate is made.
TT3: At this magnification of the likely origin noth-
ing much is seen. However looking at the left bot-
tom, a lot of fracture marks are seen, see Figure 8c
and 8d. Quinn states that the greater the failure
stress at fracture, the more the stored energy and
the richer the fracture marks (Quinn, 2007). There-
fore it is concluded that on this line the actual origin
point lies. Apparently the branching of the cracks
was interpreted wrongly. In Figure 8e several origin
like marks are observed.
115
 van der Velde et al.

a ) Fracture pattern, category IL b ) Likely origin point c ) magnification

d ) magnification e ) magnification f ) magnification

Figure 8: Fractographical analysis of TT3

a ) Fracture pattern, category IL b ) Likely origin point c ) magnification

d ) magnification e ) magnification

Figure 9: Fractographical analysis of TT10

116
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

a ) Fracture pattern, category IL b ) Likely origin point c ) magnification

d ) magnification e ) magnification

Figure 10: Fractographical analysis of TT40

a ) Fracture pattern, category I b ) Likely origin point c ) magnification

d ) magnification e ) magnification

Figure 11: Fractographical analysis of TT6

117
 van der Velde et al.

a ) Fracture pattern, category IS b ) Likely origin point c ) magnification

d ) magnification e ) magnification f ) magnification

Figure 12: Fractographical analysis of TT8

a ) Fracture pattern, category RS b ) Likely origin point c ) magnification

d ) magnification e ) magnification f ) magnification

Figure 13: Fractographical analysis of TT16

118
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Which is the actual first one cannot be said from
this image, but it does show multiple cracking
processes followed up on one another. The last
figure zooms in on one of them.
TT10: This specimen gave quite a straightforward
analysis. At the first crack a clear shell-shaped
mark can be seen in Figure 9c. In Figure 9d stable
crack growth is observed. Interestingly at the arrow
a point where unstable crack growth commences
is marked. Figure 9e zooms in at a magnification
at the centre of shell shape where the cracking
process started. Analysis of the fracture surface
will have to conclude whether it is the actual point
of origin.
TT40: In Figure 10c, made with reflected light
instead of transmitted light, the crack branching
is clearly seen which suggests this is the origin
crack. In transmitted light in Figure 10d no starting
point can be marked, but it could be blocked from
view by the shadows. Looking at Figure 10e in
transmitted light something interesting is going
on. Twist hackle marks are observed at the sides,
both with their propagation direction (marked by
arrows) towards the origin crack. This is special
because mostly crack propagation runs away from
the origin point. Whether this is a coincidence or
the origin point is actually located somewhere else
cannot be concluded from this analysis alone.
TT6: The location of the fracture origin is clearly
visible in this specimen. However, under the
microscope so much light was coming through a
gap at the crack of interest that the image was too
bright. Therefore Figure 11c was taken in crossed
polarised light. Despite giving a spectacular
picture, in this light it is harder to distinguish
the fracture marks, so it is less suitable to draw
conclusions about the fracture pattern. On a
sidenote, crossed polarised light can be useful
in determining the composition of a material’s
components.
TT8: In this specimen the transmitted light was
blocked at the crack of interest. Therefore Figure
12d was taken with reflected light only. Especially
to the left, crack growth marks are observed. The
origin of fracture might be found there. Figure 12e
is taken to the right from the origin crack (the red
dot can be used as a reference). Here twist hackle
is captured that gives an indication of the ongoing
crack propagation
TT16: The origin crack seemed to have been
found but under the microscope hardly any
fracture marks were observed, nor in Figure 13d.
Looking back at Figure 13c a lot more marks are
observed to the left, which are shown in Figure
13e. Interestingly, stable crack growth is seen in the
form of beach marks but they seem to be taken
over by unstable crack growth in the form of hackle.
At the centre of this hackle, next to the crack more
stable crack growth marks were captured in Figure
13f. The fracture may have originated there instead.
IV. Discussion
Firstly it should be stressed that these are all 2D
images of features that are located at different
depths in the glass. While interpreting them this
should be borne in mind to make sure the right
conclusions are drawn. This can be aided by
making photographs of just the glass surface before
building the 3D images, so that it can be checked
whether a certain feature is at the surface or inside
the glass.
It is the opinion of the authors that this technique
works best in combination with fracture surface
analysis, but definitely is a valuable addition. For
TT3 and TT16 it seems the wrong origin crack was
concluded from the visual fracture pattern analysis.
In a fracture surface analysis this would have been
discovered by following the Wallner lines at the
fracture surface to the right origin point. However,
by looking at the fracture pattern by transmitted
light microscopy this can be seen in one glance
and immediately gives a direction at which point
to look in the fracture surface analysis. Therefore
it is recommended to use these Z-axis scanned
images as an overview to contemplate during the
fracture surface analysis. For example for TT10
a very clear spot is marked where the fracture
origin can probably be found. When having found
the fracture origin at the fracture surface the
researcher then can use the Z-axis scanned image
to gain an understanding of the rest of the cracking
process. This process cannot be deduced in one
view by examining only the fracture surface. And
interpreting the fracture pattern by eye can lead to
wrong assumptions. This is nicely seen in the case
of TT8 where twist hackle marks the crack prop-
agation at several location in the glass in one image.
The authors recognise that this technique on its
own does not give the final conclusion or even
the desired images as was the case for TT40 and
TT6. It turns out it depends on the position of the
shards in the glass whether enough or too much
light is let through and a good image can be made.
Although it might be possible to improve this by
exploring all settings of the polarised microscope.
These results represent a first try out of this tech-
nique. Further research of more plates including
fracture surface analysis and significant strength
data is expected to show the value of transmitted
light microscopy even better.
119
 van der Velde et al.
V. Conclusion
A transmitted light microscopy analysis combined
with Z-axis scanning software gives the glass
researcher an overview of all fracture markings
throughout the depth of the glass plate in one
image. In combination with the fracture surface
analysis this image can be used as a map which
guides the researcher in their search of the fracture
origin. More importantly, it reveals fracture
markings that give new insights into the crack
propagation and the local failure process that are
overlooked in traditional glass fractography.
Based on the findings of this study, it can be con-
cluded that transmitted light microscopy, together
with Z-axis scanning software is a valuable tool for
glass fractography and its use needs to be further
explored.
120
Acknowledgements
Many thanks to Rob Nijsse for his contribution on
the overall research, to Ger Nagtegaal and Kees van
Beek for their assistance on the strength tests and
to Kees Baardolf for his technical contributions.
References
Bourhis, E.L. Glass: Mechanics and Technology. Wiley, 2014. ISBN
9783527679423.
Fessler, H. and Fricker, D.C. . A theoretical analysis of the ring-
on-ring loading disk test. Journal of the American Ceramic Society,
67(9):582–588, 1984.
Molnár, G., Molnár, L.M., and Bojtár, I. Preparing a comprehensive
analysis of the mechanical classification of structural glass. Materials
Engineering-Materiálové inžinierstvo (MEMI), 19(2):71–81, 2012.
Morrel, R. Biaxial flexural strength testing of ceramic materials. A
National Measurement Good Practice Guide no. 12, 1998.
Quinn, G.D. . Fractography of ceramics and glasses. NBS spe-
cial publication. NIST, 2007. URL http://books.google.nl/books?id=
h8qftgAACAAJ.
Shinkai, N. . The fracture and fractography of flat glass. In R.C. Bradt
and R.E. Tressler, editors, Fractography of Glass, pages 253-297. Springer
US, 1994. ISBN 978-1-4899-1327-2. doi. 10.1007/978-1-4899-1325-8-8.
van der Velde, O. Finding the strength of glass: a mechanical and
microscopical research of the processes that determine glass’ biaxial
strength for structural purposes. Master thesis, in preparation.
Veer, F.A. The strength of glass, a nontransparent value. HERON, 52
(1/2):87, 2007. J. Wurm. Glass structures: design and construction of
self-supporting skins. Walter de Gruyter, 2007.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Using nano-indentation and microscopy

to obtain mechanical properties
Mladena Luković, Erik Schlangen⇤ , Branko Šavija,
Guang Ye, Oğuzhan Çopuroğlu
Delft University of Technology, Faculty of Civil Engineering and Geosciences, Stevinweg 1, 2628 CN, Delft, The Netherlands
⇤ erik.schlangen@tudelft.nl

Abstract

Simulation of mechanical behaviour of heterogeneous materials is only possible if the local properties of the components
are known. In recent years nano-indentation is being applied on different levels to obtain local mechanical properties. The
aim of this paper is to explore various ways to obtain these properties by combining nano-indentation and microscopy.
The method to prepare specimens and perform the testing is explained and the way the properties are used in the
modelling is discussed.

Keywords: nano-indentation, hardness, stiffness, lattice model, multi-scale modelling

I. Introduction hydrated cement and different hydration products)

are required. For the simulation of the concrete in
Figure 1c the properties of the aggregates and the
Multi-scale modelling is a popular technique to get
mortar are needed as well as the interface between
insight into mechanical properties and to explain
them. These local properties are not easily mea-
damage and deterioration mechanisms in cement
sured. Various attempts can be found in literature
based materials. With a model on a certain scale
to obtain them from direct tests on for instance an
the (non-linear) properties of a material on that
aggregate attached to a cement paste (Ağar-Özbek
scale can be predicted. These properties can subse-
et al, 2013) or trying to simulate them with a lower
quently be used as input parameters for the local
scale model (Bentz et al. 1995). Furthermore it
properties on a next scale as sketched in Figure 1.
should be noted that these local material properties,
Here the microstructure of a cement paste is sim-
such as strength and Young’s modulus, are most
ulated using the HYMOSTRUC 3D model (Figure
probably not single values, but the local behaviour
1a) according to Ye (2003). This microstructure can
is actually a non-linear stress strain relation. At the
then be used to study the mechanical properties of
atomic scale the local mechanical behaviour might
a cement-paste cube at a certain degree of hydra-
be brittle, but at all higher scales some non-linearity
tion (Qian et al. 2011). For the simulation of the
has to be implemented in the multi-scale modelling
fracture behaviour various types of models can be
concept. This non-linearity is different at each scale,
used, such as discrete element type or lattice type
which makes it rather complex. The general pro-
models (Schlangen, 1993). The outcome of such a
cedure to simulate mechanical response is to start
simulation on the scale of cement paste is a non-
with a brittle behaviour for fracture at the lowest
linear stress-strain relation which can be used as
scale of observation (see e.g. Qian, 2012) and then
input for the simulation of mechanical properties of
use non-linear behaviour at all higher scales to keep
mortar as depicted in Figure 1b. The outcome of a
the error as small as possible.
simulation on the mortar scale is again the input for
the concrete scale in which the aggregate particles Another way to obtain local mechanical prop-
in the material, see Figure 1c, should be modelled erties that can be used as input for simulations
(Qian, 2012). is nano-indentation. This method is initially de-
To be able to perform simulations of a heteroge- veloped for homogeneous materials and especially
neous material on a certain scale the local proper- thin films to obtain local mechanical properties such
ties of the components of the material have to be as hardness and stiffness. However it is also used
known. For instance, for simulating the mechanical to get local mechanical properties of heterogeneous
behaviour of the cement-paste shown in Figure 1a, materials by indenting in the different phases of the
the mechanical properties of the single phases (un- material. First applications for cement-based mate-

121
in figure 1b.. The outcome of a simulation on the mortar scale is again the input for
the concrete scale in which the aggregate particles in the material, see figure 1c,
Luković et al.
should be modelled (Qian, 2012).

Figure 1: a) Generated cement-paste microstructure, b) multi-scale modelling scheme and c) generated concrete sample
with irregular aggregates.

rials can be found in Constantinides and Ulm (2004)
Rev. 11/2014
or Hughes and Trtik (2004). Using these proper-
ties as input for modelling mechanical properties
of cement paste material is shown in Bernard et al.
(2003) and it is applied for simulating degradation
mechanisms by Çopuroğlu and Schlangen (2008).
The difficulty with nano-indentation in hetero-
geneous materials is that it is impossible to indent
in a single phase and the materials are generally
3D, which means the material below the indent
might be of a different phase. The results obtained
from the indentation will therefore always be a com-
bined property of various phases. Furthermore the
properties from nano-indentation cannot be directly
implemented into a model. Local properties of an
element in a model might be dependent on the scale
of the element or other parameters that are specific
for the model that is used.
The aim of the study presented in this paper
to give more insight in the use of different nano-
indentation techniques for cement-based materials
at different scales and the way the outcome from
the nano-indentation measurements can be used in
models.
In this paper three different ways of applying
nano-indentation to obtain mechanical properties
are discussed. All methods are supported by nu-
merical simulations with a lattice model. The first
method is used to obtain local properties in a re-
pair system of repair material (cement paste) on a
mortar. Main focus is on the properties of the zone
around the interface and whether the local mechan-
ical properties can be linked to the failure of the in-
terface. The second method is a 3D nano-indention
on cement paste samples to study the effect of the
sub-surface microstructure on the measured prop-
erties. Finally the use of nano-indentation to obtain
more global properties of cement paste is presented.
II. Experimental procedure and
results
Specimen preparation
The specimens used in the three methods of nano-
indentations are all prepared in the same way. The
specimens were prepared for a study on the perfor-
mance of repair systems (Luković, in preparation).
The substrate was 5 year old mortar casted with
Ordinary Portland Cement (OPC) CEM I 42,5 N.
Smaller mortar specimens were cut with a diamond
saw from bigger samples and fixed in the centre of

Figure 2: a) Mortar-cement
Figure 2: a) paste cylindrical sample
Mortar-cement pastetocylindrical
study local sample
mechanical
to properties
study localandmechanical
b) zoom-in on interface
between mortar and cement paste.
properties and b) zoom-in on interface between mortar and cement paste.

122
2D nano-indentation
shown by the black line in figure 3b. More results and comparison of experimental
findings and simulated
15th Euroseminar results
on Microscopy can
Applied be found
to Building in Lukovic
Materials • 17-19 June et al• (2014).
2015 Delft, The Netherlands

Figure 3: a) Nano-indentation measurements of hardness projected on top of ESEM

Figure 3: a) Nano-indentation measurements of hardness projected on top of ESEM image of microstructure in
image
the ofregion
interface microstructure
and b) simulatedin theofinterface
result region
tensile test with and b)
implemented simulated
test
local result
properties from of tensile
nano-indentation
with implemented local properties from nano-indentation measurements.
measurements.

the mould where new material was later cast (Fig- 2D nano-indentation
ure 2). Before casting it, the old mortar was placed
3D nano-indentation
for 24h in a chamber with 80% relative humidity, Agilent Nano Indenter G200 with a diamond
Nano-indentation
in order to enable hygralgives localand
equilibrium mechanical
uniform properties
Berkovich for used
tip was the for
surface of a material.
nano-indentation tests For
heterogeneous
moisture profile in materials
the substrate.this
Thealways
repair ma- results
on in
the an errorinsince
samples the
order to phasetheoflocal
measure theme-
material
terial (cement paste) was a standard grade OPC chanical properties around the interface between
below the surface might (and will) be different in most cases. The effect of evaluating
CEM I 42,5 N paste with 0.4 water-to-cement ratio. mortar and repair material. More details on these
2D instead
Cylindrical of 3D
specimens quantities
(diameter, 24 mm,is examined
height 39 tests by a combination
are described in Luković etof
al. experimental
(2014). A series and
mm) were cast.
numerical The specimens
research were cured
(Lukovic for2015).
et al., 28 ofIn thiswere
indents study, nano-indentation
performed on a tightly spaced testing
grid, was
days in sealed conditions at room temperature of with spacing of 7 µm in the direction perpendicular
applied on successive slices in cement paste. The same specimens and preparation
20 C. After 28 days, the sample was cut with a to the interface, and 14 µm in the direction parallel
method
diamond saw as into
explained
slices withabove were
a thickness of 8used.
mm. First,
to the one slice
interface. Thewas prepared,
indentation tested
depth was in nano-
700 nm.
In order to stop hydration, solvent exchange by iso-
indentation and checked in ESEM (Environmental Scanning Electron Microscope).
The Continuous Stiffness Method (CSM) developed
propanol was used. The sample was submerged by Oliver and Pharr (2004) is used. This method
Then,
five timesthe
andtop
takenlayer
out for(around
a period of16
oneμm)
minuteis removed, the sample polished, and the same
consists of superimposing a small oscillation on the
area
in ordertested
to enableagain. A procedure,
faster water-solvent exchange.ensuring
Af- thatloading
primary the signal
sameandarea is tested
analysing in every
the response
terwards, it was slice,
consecutive placed for
is 72 hours in isopropanol
developed. of the system by means
After nano-indentation of a frequency-specific
testing of consecutive am-layers,
and subsequently taken out, and solvent was re- plifier. As a consequence, it enables a continuous
all individual
moved 2D microstructures
by evaporation at ambient conditions.andIna reconstructed 3D microstructure (obtained by
measure of contact stiffness as a function of indenta-
reconstruction
order to make thicknessof individual
of the specimens2D evenslices)
and were
tion depthsubjected
and not justto simulated
at the direct
point of initial unload-tension
uniform,
testing. a thin sectioning machine was used for ing. Therefore, hardness and elastic modulus are
cutting and grinding of the specimen. An object obtained as a continuous function of surface pene-
glass was glued on the sample and a block was tration. The average E-modulus and hardness were
cut with a diamond saw (Figure 3). Afterwards, determined in the displacement range between 500
all samples were ground in the thin section ma- and 650 nm depths. The indentation areas were
chine in several passes. After grinding, each of the taken randomly within the interface between the
samples was polished with 6 µm (5 min), 3 µm (5 old mortar and new cement paste.
min), 1 µm (5 min) and 0.25 µm (25 min) diamond
Rev. 11/2014 After nano-indentation testing, samples were ex-
paste on a lapping table. After every step, an opti-
amined using environmental scanning microscope
cal microscope was used to check the effectiveness
(ESEM), in backscattered electron (BSE) mode. The
of grinding and polishing. Finally, the samples
instrument was operated at 20 kV accelerating volt-
were soaked into an ultrasonic bath to remove any
age and at 10 mm working distance between the
residue from the surface. Samples were kept in a
final condenser lens and the specimen. The spot
desiccator under vacuum until testing.
size was 5.0 µm and the magnification was 500⇥.
The vapour pressure inside the chamber was set to
1 Torr. All imaging was performed at low vacuum
in order to reduce the probability of cracking in

123
 Luković et al.

Step 4 a b

Step 1 a b

Step 5 a b

Rev. 11/2014

Step 2 a b

Step 6 a b

Step 3 a b

Step 7 a b
Figure 4: a) ESEM-images (100x100 µm) of tested locations and b) measured
hardness values with nano-indenter.

In figure 4a, only part of microstructure where the indentation testing was performed
Step 4 a b with corresponding hardness properties (figure 4b) are presented. From ESEM image
in Step 1, it seems that in the middle of the tested specimen only CH and CSH are
present.
Figure 4: a) ESEM-images (100x100 µm) of tested However,
locations and b)nano-indentation
measured hardnessresults values
indicatewith
muchnano-indenter.
higher hardness properties
than those corresponding only to CH and CSH phase. From ESEM image in Step 2, it
can be observed that exactly under this first layer, there are two unhydrated cement
particles, which contributed to locally higher micromechanical properties that are
the specimens. The nano-indentation hardness is in the
observed Step simulation is shownare
1. Similar observations by made
the black
also inline
someinofFigure
the further steps.
plotted in Figure 3a on top of the microstructure 3b. More results and comparison
Nano-indentation results at some locations, therefore, can be of experimen-
explained only if the
that is obtained using ESEM. It can be seenfollowing
that step talisfindings
analysed. and simulated
It should be kept results
in mind can be found in is, after
that nano-indentation
the
Stepvery
5 a heterogeneous structure
b all, volume Luković
of the interface testing, and results
et al. cannot be always directly related to the phases
(2014).
observed on the surface. This holds especially for highly heterogeneous materials
on microscale contributes to high variations and a
such as cement paste.
not so clear tendency around the interface. As Also,
can be seen from Figure 4, cement particles are very irregular and randomly
3D nano-indentation
these local micromechanical interface properties do
distributed. Their alignment, distribution and mutual interaction would highly
Rev. 11/2014 not provide any information about the interface
influence both on microstructure development
Nano-indentation gives localand on fracture properties.
mechanical properties
performance and its strength. Therefore, they need for the surface of a material. For heterogeneous
to be linked together in a structure where all the materials this always results in an error since the
single local tests will be assembled. This will en- phase of the material below the surface might (and
able assessment of the overall performance of the will) be different in most cases. The effect of eval-
interface zone and estimation of the fracture prop- uating 2D instead of 3D quantities is examined
erties at the micro-scale. This is done by combining by a combination of experimental and numerical
the experimental results with numerical simula- research (Luković et al., 2015). In this study, nano-
tions. Measured local micromechanical properties indentation testing was applied on successive slices
are used as input for numerical modelling with the in cement paste. The same specimens and prepara-
lattice model which enables estimation of the global
Rev. 11/2014
tion method as explained above were used. First,
micromechanical response (properties) of the tested one slice was prepared, tested in nano-indentation
microstructure. An example of this is shown in and checked in ESEM (Environmental Scanning
Figure 3b, where a 2D sample is loaded in tension Electron Microscope). Then, the top layer (around
(vertically). The local properties implemented in 16 µm) is removed, the sample polished, and the
the model correspond with the local hardness and same area tested again. A procedure, ensuring that
stiffness as measured in a sample shown in Figure the same area is tested in every consecutive slice, is
3a. The final crack pattern that is the outcome of developed. After nano-indentation testing of con-

124
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 5: Simulated tensile test on 3D cube with loading in x and y direction and local properties from nano-indentation
Figure 5: Simulated tensile test on 3D cube with loading in x and y direction and
measurements.
local properties from nano-indentation measurements.
Figure 5 shows an example of a simulation with the lattice model of a small cube
secutive layers, all individual 2D microstructures
loaded in tension. The properties use asume testing, and results cannot be always directly
input are obtained from the 3D nano-
and a reconstructed 3D microstructure (obtained related to the phases observed on the surface. This
indentation. The cube is loaded in 2 directions and the resulting load-displacement
by reconstruction of individual 2D slices) were sub- holds especially for highly heterogeneous materials
curves and crack patterns are
jected to simulated direct tension testing.
compared. The scatter in load-displacement curves is
such as cement paste.
small when the 3D input data are used and a 3D simulation is performed. This is,
In Figure 4a, only part of microstructure As can be seen from Figure 4, cement parti-
however, different when only where
2D input data from different slices are used as
the indentation testing was performed with This cles are very irregular and randomly distributed.
corre-results
discussed in Lukovic et.al 2015. in bigger scatter.
Their alignment, distribution and mutual interac-
sponding hardness properties (Figure 4b) are pre-
sented. From ESEM image in Step 1, it seems that tion would highly influence both on microstructure
Nano-indentation of small cubes development and on fracture properties.
in the middle of the tested specimen only CH and
As discussed before, nano-indentation in a heterogeneous material as cement paste
CSH are present. However, nano-indentation re- Figure 5 shows an example of a simulation with
never
sults indicate much results
higherinhardness
properties of a than
properties single phase. Properties
the lattice model that
of a are
smallfound
cube by nano-
loaded in ten-
indentation can be implemented
those corresponding only to CH and CSH phase. into models, yet an error will be made. When
sion. The properties use as input are obtained using from
From ESEM upscaling
image in methods it isbetherefore
Step 2, it can observed that necessarytheto 3Dcheck and validate The
nano-indentation. the effect
cube isofloaded
the in
choice
exactly under thisoffirst
yourlayer,
inputthere
values
are fortwoyour
unhy- model2 on every scale.
directions Thisresulting
and the should be done with
load-displacement
experiments
drated cement particles,andwhichnot contributed
by modelling because
to lo- thenand
curves thecrack
errors mightare
patterns only pile up.The
compared. In scat-
soaking
cally higherthe specimen
(Schlangen et al.for
micromechanical a short
2015) a new
properties time
that in
ob-acetone.
method
are ister Thetodicing
proposed sawcurves
study the
in load-displacement machine
failure isofsmallwas
100 µmsetthe
when in
served in cement
Step 1. paste
Similar cubes
observationsunder
are indentation
made also loading.
3D input The
data
such a way that the saw was cutting a few micrometres into the glass plate in order to aresame
used basic
and a specimens
3D simulation asis per-
in some ofdescribed
the further steps.were
Nano-indentation
used. In orderre- formed. This is, however, different when equalonly
to 2D
make
sults atsure
some that theabove
locations, cement
therefore, paste
can be cubes
explained
to make
have the
input
the thickness
required
data from
of the specimens
dimensions
different andarethat
slices usedalso the
as dis-
100 µm and even and uniform, a thin sectioning machine was used for cutting and
glue
only ifbetween
thegrinding
followingcement
ofstep paste and
theisspecimen.
analysed. glass
stepbeinwas
It should
Next sliced.
thecussed
process isThe
in Luković final cubes
et.alsmall
to create 2015. This
cubes are
resultsallinthe
from at the
bigger
dimension
kept in 100ofµm
mind that100x100x100
thin slice. For±this
nano-indentation 3 µm
is, after (Figure
all,
process vol- 6b). dicing saw for semiconductor wafers
scatter.
a diamond
was used. By running this precise saw with a 400 µm wide blade in two directions
over the thin section small cubes as shown in Figure 6a were obtained. To prevent
chipping of the edges of the small cement paste cubes while cutting, a thin layer of
nail-polish was applied on the surface of the thin section, which was later removed by

Rev. 11/2014

Figure 6: a) Small cement paste cubes on glass plate b) Single cube with dimensions c) Diamond Berkovich tip.
Figure 6: a) Small cement paste cubes on glass plate b) Single cube with dimensions
c) Diamond Berkovich tip.
125

The small cubical cement paste samples on the glass plate were placed in an
 Luković et al.

Figure 7: Three stages (a, b, c) in the nano-indentation loading process of small cubes observed in ESEM and
Figure 7: Three stages (a, b, c) in the nano-indentation loading process of small cubes
simulation result with the lattice model (d).
observed in ESEM and simulation result with the lattice model (d).
Nano-indentation of small cubes saw machine was set in such a way that the saw
For the modelling a 3D lattice network model is used. As starting point for the
was cutting a few micrometres into the glass plate
As discussed before, nano-indentation in a hetero-
simulations a 3 dimensional hydrated cementinpaste order specimen
to make sureisthat
constructed
the cement using the
paste cubes
geneous material as cement
HYMOSTRUC3D model.pasteMore
never results
detailsin on the
have construction of theseand
the required dimensions 100
thatμm
also cubes
the glue
properties of a single phase. Properties that are
and by
found thenano-indentation
implementation can of local material between
be implemented
cementobtained
properties, paste andfrom
glass was
localsliced.
nano-The
final cubes are all at the dimension of 100x100x100
indentation
into models, yettesting, can bebemade.
an error will found in us-
When Qian et al. (2011). The result of one of the
± 3 µm (Figure 6b).
ing upscaling methods
simulations it is therefore
is presented necessary
in Figure 7d astodeformed and cracked specimen.
check and validate the effect of the choice of your The small cubical cement paste samples on the
input values for your model on every scale. This glass plate were placed in an Agilent G200 nano-
should be done with experiments and not by mod- indenter and were tested with a Berkovich tip (Fig-
DISCUSSION AND CONCLUSIONS
elling because then the errors might only pile up. ure 6c). A displacement controlled test was per-
In (Schlangen et al. 2015), a new method is pro- formed by using the CSM (Continuous Stiffness
posed to study Method). The loading procedure is discussed in
In this paperthea failure of 100 µmofcement
combination paste
experimental techniques using nano-indentation and a
more detail in Poelma et al (2014) in which car-
cubes under indentation loading. The same basic
modelling technique using 3D discrete lattice bonmodel
nano tubesimulations
bundles areisloadedapplied to study
to failure with
specimens as described above were used. In order
the failure mechanism in cement paste
to make the thickness of the specimens equal to specimens
a at micro-scale.
nano-indenter. Different loading depths were ap-
Asµm
100 input
and for
eventhe
andsimulations,
uniform, a thinmechanical
sectioning properties for the local lattice elements are
plied. This made it possible to obtain crack patterns
needed.was
machine These cancutting
used for be obtained fromof local
and grinding the at various depths as shown
nano-indentation in Figure 7.in
measurements Thesingle
typical
specimen. failure mechanism obtained is crushing of the mate-
phases ofNext the step in the processcomposite.
heterogeneous is to create However, these properties cannot be used
rial under the tip and three main cracks running to
small cubes from the 100 µm thin slice. For this
directly, because they depend on
process a diamond dicing saw for semiconductor
the size of the elements
the sides of the cubes,used in from
starting the the
model and
three edges
should
wafers wasbe transferred
used. By runningtothis
model-parameters.
precise saw with Furthermore
of the Berkovichittip. should
Indenting be the
noted that the
tip further into
a results obtained
400 µm wide blade infrom nano-indentation
two directions the specimen results in complete
over the thin on the surface of a heterogeneous cement crushing of the
sample as can be seen in the ESEM image in Figure
paste small
section always includes
cubes as shownan inerror,
Figurebecause
6a were the result is influenced by the phase below
obtained. To prevent chipping of the edges of the 6c.
the surface.
small cement paste cubes while cutting, a thin layer For the modelling a 3D lattice network model is
In this paper different techniques and
of nail-polish was applied on the surface of the thin
applications of nano-indentation are presented
used. As starting point for the simulations a 3 di-
and discussed.
section, which wasFirst 2D nano-indentation
later removed by soaking the is used to determine
mensional local properties
hydrated cement paste specimen around
is con-
specimen for a short time in acetone. The dicing structed using the HYMOSTRUC3D model. More

126
Rev. 11/2014
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
details on the construction of these 100 µm cubes
and the implementation of local material proper-
ties, obtained from local nano-indentation testing,
can be found in Qian et al. (2011). The result of
one of the simulations is presented in Figure 7d as
deformed and cracked specimen.
III. Discussion and Conclusions
In this paper a combination of experimental tech-
niques using nano-indentation and a modelling
technique using 3D discrete lattice model simula-
tions is applied to study the failure mechanism in
cement paste specimens at micro-scale.
As input for the simulations, mechanical proper-
ties for the local lattice elements are needed. These
can be obtained from local nano-indentation mea-
surements in single phases of the heterogeneous
composite. However, these properties cannot be
used directly, because they depend on the size
of the elements used in the model and should be
transferred to model-parameters. Furthermore it
should be noted that the results obtained from nano-
indentation on the surface of a heterogeneous ce-
ment paste always includes an error, because the
result is influenced by the phase below the surface.
In this paper different techniques and applica-
tions of nano-indentation are presented and dis-
cussed. First 2D nano-indentation is used to deter-
mine local properties around an interface between
mortar and repair material. This method is valu-
able when studying for instance the influence of
microstructure formation in a repair system and
the resulting local mechanical properties on the per-
formance and crack formation in such a system.
A 3D nano-indentation method is shown which
is useful to evaluate the influence and scatter of
using 2D-data for studying 3D mechanisms. Fi-
nally a method is presented for making small scale
cement-paste specimens and fail them with nano-
indentation. This method is useful to validate ex-
perimentally simulation tools that are used in a
multi-scale framework.
Acknowledgement
Financial support from the Dutch Technology Foun-
dation (STW) for the project 10981-"Durable Re-
pair and Radical Protection of Concrete Structures
in View of Sustainable Construction" is greatly ac-
knowledged.
References
Ağar-Özbek, A. Weerheijm, J., Schlangen, E., van Breugel, K. Inves-
tigating porous concrete with improved strength: Testing at different
scales. Construction and Building Materials, 41 (2013) 480-490.
Bentz D, Schlangen E, Garboczi E. Computer simulation of interfa-
cial zone microstructure and its effect on the properties of cement-based
composites. Materials science of concrete IV. 1995:155-200.
Bernard O, Ulm FJ, Lemarchand E. A multiscale micromechanics-
hydration model for the early-age elastic properties of cement-based ma-
terials. Cement and Concrete Research. 2003;33(9):1293-309.
Constantinides, G. and Ulm, F.J. The effect of two types of CSH on
the elasticity of cement-based materials: Results from nanoindentation
and micromechanical modeling. Cement and Concrete Research, 2004.
34(1): p. 67-80.
Çopuroğlu O, Schlangen E. Modeling of frost salt scaling. Cement
and Concrete Research. 2008;38(1):27-39.
Hughes JJ, Trtik P. Micro-mechanical properties of cement paste
measured by depth-sensing nanoindentation: a preliminary correla-
tion of physical properties with phase type. Materials characterization.
2004;53(2):223-31.
Luković, M. Influence of interface and SHCC material properties on
the performance of the repair system, PhD thesis, Delft University of
Technology, in preparation
Luković, M., Šavija, B., Dong, H., Schlangen, E., Ye, G.. Micromechan-
ical study of the interface properties in concrete repair systems. Journal
of Advanced Concrete Technology, 2014; 12: 320-339.
Luković, M., Schlangen, E., Ye, G. Combined experimental and nu-
merical study of fracture behaviour of cement paste at the microlevel.
Cement & Concrete Research, 2015; 73:123–135.
Oliver WC, Pharr GM. Measurement of hardness and elastic modu-
lus by instrumented indentation: Advances in understanding and refine-
ments to methodology. Journal of Materials Research. 2004;19(01):3-20.
Poelma, R.H., Morana, B., Vollebregt, S., Schlangen, E., van Zeijl,
H.W., Fan, X. and Zhang, G. Tailoring the Mechanical Properties of
High-Aspect-Ratio Carbon Nanotube Arrays using Amorphous Silicon
Carbide Coatings, Advanced Functional Materials, 2014; 24, 36: 5737-
5744.
Qian, Z., Ye, G., Schlangen, E., Van Breugel, K. 3D lattice fracture
model: Application to cement paste at microscale (2011) Key Engineer-
ing Materials, 2011;452-453: 65-68.
Qian Z. Multiscale Modeling of Fractrue Processes in Cementitious
Material, PhD thesis. Delft, The Netherlands, PhD thesis, Delft Univer-
sity of Technology; 2012.
Schlangen E. Experimental and numerical analysis of fracture pro-
cesses in concrete, PhD thesis, Delft University of Technology, Delft, The
Netherlands; 1993.
Schlangen, E. Luković, M. Šavija, B., Çopuroğlu, O. Nano-indentation
testing and modelling of cement paste. Proceedings Concreep-10, Vi-
enna, Austria, 2015.
Ye, G. Experimental Study and Numerical Simulation of the Develop-
ment of the Microstructure and Permeability of Cementitious Materials,
PhD-thesis, Delft University of Technology, the Netherlands, 2003.
127
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

128
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Historic materials
f 0 F

Self-healing materials

129
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

130
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Material characteristics of prefabricated

concrete elements from a late 19th century
church in lower Austria
Farkas Pintér⇤,a and Christophe Gosselinb
a Scientific Laboratory, Federal Office for the Protection of Monuments, Arsenal 15/4, A-1030 Vienna, Austria
b Geotest SA, En Budron E7, CH-1052 Le Mont-sur-Lausanne, Switzerland
⇤ farkas.pinter@bda.at

Abstract

The late 19th century has preserved several important historic objects in the former Austro-Hungarian Empire made
of Portland cement concrete. One of the most interesting examples of early concrete objects in Austria is the Parish
church of Weissenbach an der Triesting near Vienna. The church was constructed between 1891 and 1893 using a novel
double wall system: the inner wall was built from brick and the whole façade was covered with different prefabricated
PC concrete elements.
The present study introduces the material characteristics of some of the original concrete elements at the church.
Thin and polished sections of the samples were analyzed in the optical and scanning electron microscope in order to
determine the properties of the cement and aggregates. Façade elements, such as ledges, show a double-layered structure:
a dense outer layer is followed by a porous tamped concrete layer to lighten the element. Composition and grains size
of unhydrated clinker phases and the presence of solid fuel remnants suggest the use of a shaft kiln, inhomogeneous
raw meal, firing conditions and slow cooling rates. Differences in the microporosity between the two layers refer to
different w/c-ratios during the production of materials. SEM-EDX measurements and element ratio plots highlight the
composition and change in the binder due to the interaction of cement hydrate phases, aggregates, and atmospheric
pollutants.
Additionally, dolomite aggregates exhibit the phenomenon of dedolomitization, manifesting in the transformation
of CaMg(CO3 )2 grains into calcite and brucite. The phenomenon, however, is not connected with the formation of
alkali-silica gel, thus it is harmless and cannot be classified as deleterious ACR.

Keywords: Portland cement, historic concrete, clinker phases, optical microscopy

I. Introduction Austro-Hungarian Empire and some exclusively

Portland cement was first produced in 1856 as nat-
ural cement in Austria. The last decades of the
19th century, however, were dominated by the use
of Roman cements as hydraulic binders (Zehetner,
2010). At the turn of the 20th century this trend
rapidly changed and the success story of Portland
cement has begun in the Austro-Hungarian Empire.
Some of the most interesting applications of Port-
land cement in Eastern Austria were connected to
the talented Austrian engineer Adolf Freiherr von
Pittel (1838-1900). In the 1870s he established sev-
eral cement plants near Vienna, where hydraulic
lime and Roman cement were produced, but he
promptly completed his array of products with
prefabricated casts and artificial stones made from
Portland cement (Heimel et al, 2010). Pittel’s suc-
cess resulted not only in the good quality hydraulic
binders he produced, but also in the widely dis-
tributed prefabricated products (Figure 1) in the
managed patents in the field of concrete technology
(Heimel et al., 2010).
One of Pittel’s main creations was the Parish
church in Weissenbach an der Triesting (Lower Aus-
tria, Figure 2), which he donated to his community.
It was constructed between 1891 and 1893 using a
double wall system: the inner wall was built from
brick and the whole façade was covered with pre-
fabricated PC concrete elements (Heimel et al, 2010).
The present study is a part of an international coop-
eration aimed at studying the 19th century concrete
buildings within the territory of the former Austro-
Hungarian Empire.
Due to the abundance of large unhydrated ce-
ment grains and non-carbonated cement paste in
the samples, the paper focuses on the chemical-
mineralogical characterization of residual clinker
phases and hydration products, attempting to deter-
mine the original composition, characteristics, and
production techniques of the cement and concrete.

131
 Pintér & Gosselin

II. Sampling and analytical methods

Two samples, one from a slightly damaged ledge
(WB-1) at the western façade (Figure 2) and an-
other concrete fragment from the socle (WB-2) on
the southern side, were taken for analyses. Thin
and plane sections prepared from the samples were
etched with Nital (1.5 ml of HNO3 in 100 ml of
isopropyl alcohol; Campbell 1999) for 7 to 9 sec-
onds and analyzed in the optical microscope (Zeiss
AXIOScope A1) using reflected and transmitted po-
larized light. Depending on the reactivity of the
cement phases, alite normally turned blue to tan; be-
lite was brown to blue and both phases showed de-
tails of internal structure (Campbell, 1999). Phases
of the interstitial matrix (C3 A and ferrite) were light
gray and white in color. Non-etched polished sec-
tions were coated with carbon and analyzed in the
scanning electron microscope (SEM, Zeiss EVO MA
15), coupled with an energy dispersive X-ray spec-
trometer (EDS, Oxford DryCool, 15kV, 300 pA). A
number of standardized point analyses were used
to characterize the chemical composition of the
binders.

Figure 1: Examples from Pittel´s catalog (1890s) containing III. 2:Results

Figure The western and discussion
façade of the Parish church of
Figure
precast 1: Examples
concrete from(courtesy
elements Pittel’s catalog
of the (1890s)
Countrycontain-
Museum, Weissenbach a.d. Triesting
ing precasta.d.
Weissenbach concrete elements (courtesy of the Country
Triesting) General features of the concrete samples
Museum, Weissenbach a.d. Triesting).
Based on macroscopical observations on broken sur-
Sampling and analytical methods
faces (Figure 3), samples exhibited compact struc-
Two samples, one from a slightly damaged ledge (WB-1) at the western façade (Figure 2) and another
ture; material loss due to weathering could not be
concrete fragment from the socle (WB-2) on the southern side, were
observed. taken
By the usefor
of analyses. Thin and
phenolphtalein plane sections
on fresh
prepared from the samples were etched with Nital (1.5fracture ml of HNO 3 in 100 ml of isopropyl alcohol; Campbell
surfaces an intensive purple discoloration
1999) for 7 to 9 seconds and analyzed in the optical (Figure microscope (Zeiss
3) at ca. AXIOScope
0.5 to 1 cm from theA1)surface
using indi-
reflected and
transmitted polarized light. Depending on the reactivitycated of thethecement phases, alite normally turned
preserved alkalinity of the binder. This blue to tan;
belite was brown to blue and both phases showed details of internalsuggests
phenomenon structure (Campbell,
adequately low1999). Phases of the
capillarity,
interstitial matrix (C3A and ferrite) were light gray and probably
white in color.
due toNon-etched
a relatively polished
low w/c sections were coated
ratio of the
with carbon and analyzed in the scanning electron microscope fresh concrete(SEM, Zeiss
(Idorn EVO MA1983).
& Thaulow, 15), coupled
On the with an
energy dispersive X-ray spectrometer (EDS, Oxford DryCool, contrary 15kV,
to this,300
the more
pA). porous tamped
A number concrete
of standardized point
layer (Figure
analyses were used to characterize the chemical composition of the binders. 3) in the sample WB-1 did not show
any discoloration with effect from the phenolph-
talein test, which suggests higher capillary porosity
and the influence of carbonation and/or leaching
on the cement paste, due to contact with air and
humidity along e.g. fitting gaps on the façade.

Optical microscopy
Confirming the phenolphtalein test, the high bire-
fringence of the matrix in the sub-surface zones
corresponds to carbonated binder portions. Esti-
Figure 3: Alkaline binder portions highlighted by using
mated water to cement ratios being 0.4-0.45 (outer
phenolphtalein test on the broken surface of the sample.
concrete, Figure 4, WB-1) and ⇠0.50 for the sam-
Below the discoloration the cement paste of the tamped
ng Figure 2: The western façade of the Parish church of ple WB-2 (Figure 5). The sample WB-1 contained
concrete containing large aggregates was fully carbonated
m, Figure 2: Thea.d.
Weissenbach western façade of the Parish church of
Triesting many irregularly
WB-1 shaped compaction pores (Figure
(precast ledge)
Weissenbach a.d. Triesting. 4), while the sample WB-2 had very few entrapped
air voids (1 to 2 vol.%, Figure 5). The cement paste
of both samples were free of any cracks.
ed ledge (WB-1) at the western façade (Figure 2) and another
132were taken for analyses. Thin and plane sections
uthern side,
(1.5 ml of HNO3 in 100 ml of isopropyl alcohol; Campbell
 with carbon and analyzed in the scanning electron microscope (SEM, Zeiss EV
15th Euroseminar energy dispersive
on Microscopy AppliedX-ray spectrometer
to Building (EDS,
Materials • 17-19 June Oxford DryCool,
2015 • Delft, 15kV, 300 pA). A
The Netherlands
analyses were used to characterize the chemical composition of the binders.

Results and discussion

Figure 3: Alkaline binder p
General features of the concrete samples phenolphtalein test on the
Based on macroscopical observations on broken surfaces (Figure 3), samples exhibited Below the discoloration th
compact
structure; material loss due to weathering could not be observed. By the use of phenolphtaleinconcrete containing large a
on fresh fracture
surfaces an intensive purple discoloration (Figure 3) at ca. 0.5 to 1 cm from the surface indicated
WB-1the (precast
preservedledge)
alkalinity of the binder. This phenomenon suggests adequately low capillarity, probably due to a relatively low
Figure 3: Alkaline binder portions highlighted by using phenolphtalein test on the broken surface of the sample. Below
w/c ratio of the fresh concrete (Idorn & Thaulow, 1983). On the contrary to this, the more porous tamped
the discoloration the cement paste of the tamped concrete containing large aggregates was fully carbonated WB-1
concrete layer (Figure 3) in the sample WB-1 did not show any discoloration with effect from the phenolphtalein
(precast ledge).
test, which suggests higher capillary porosity and the influence of carbonation and/or leaching on the cement
paste, due to contact with air and humidity along e.g. fitting gaps on the façade.
The most characteristic feature of both samples
is a large amounts of very coarse residual cement
particles, exhibiting a maximum of 0.9 and an av-
erage diameter of 0.2 to 0.3 mm (Figures 6 to 11),
typical for historic PC mortars and concretes (Idorn
& Thaulow, 1983; Blezard, 1984, Weber et al, 2012).
A significant portion of cement grains have re-
mained unhydrated (Figures 6 to 11). Traces of
hydration could mainly be observed on the mar-
gins of the residual phases. Due to the higher w/c
ratio the inner concrete layer of the sample WB-1
les exhibits higher degree of hydration.
bservations on broken surfaces (Figure 3), samples exhibited compact The aggregate contained in the mortars at grain
ing could not be observed. By the use of phenolphtalein on fresh fracture sizes of 0.1 to 3 mm (WB-1) and 0.1 to 4 (WB-2)
ion (Figure 3) at ca.Figure
0.5 to4:1High
cm amounts
from theofsurface indicated
compaction the low
pores and preserved
to is exclusively
Figure of crushed
5: The compact dolomite;
cement paste exibits the estimated
slightly
non suggests adequately low
Figure 4:w/c
moderate capillarity,
High
ratioamountsprobably due to a relatively
of compactioncement
of the non-carbonated pores paste inlow
and low to
the inhomogenous capillary porosity and contains few rounded
binder to aggregate ratio being 1 to 2.5-3 (WB-1)air
n & Thaulow, 1983). On WB-1
sample
moderatethe w/c
contrary
(thin to the
section,
ratio of this,
UV the more porous
microscopy)
non-carbonated tamped
cement paste voids (WB-2, thin section, UV microscopy)
and 1 to 4 (WB-2) with a definetely higher amount
WB-1 did not show any discoloration
in the sample WB-1 with effect
(thin fromUV
section, themicroscopy).
phenolphtalein
porosity and the influence of fine fraction (0 to 0.05 mm) in the sample WB-2,
Optical of carbonation and/or leaching on the cement
microscopy
dity along e.g. fitting gaps on the which
façade. the phenolphtalein test, the high birefringence probably also had an impact on the higher
Confirming of the matrix in the sub-surface zones
corresponds to carbonated binder portions. Estimated water water demand
to cement and being
ratios thus higher
0.4-0.45capillary porosity
(outer concrete,
of the mortar (cf. Figures
Figure 4, WB-1) and ~0.50 for the sample WB-2 (Figure 5). The sample WB-1 contained many irregularly 4 and 5).
shaped compaction pores (Figure 4), while the sample WB-2 had very few entrapped air voids (1 to 2 vol.%,
Figure 5). The cement paste of both samples were free of anyCharacteristics
cracks. of cement phases
Residual cement grains that entirely or partly sur-
vived the hydration and later carbonation were
used to determine the composition and properties
of the original cement. In terms of composition and
grain size of single crystals, cement grains exhibited
very inhomogeneous features. Interpretation of the
observations was based on Insley (1936), Gille et al
(1965), Idorn & Thaulow (1983), Blezard (1984), and
Campbell (1999).

es and low to Figure 5: The compact cement paste exibits slightly Alite
ement paste in the Figure 5: The compact
inhomogenous capillary cement
porositypaste exibits slightly
and contains inho-air
few rounded
y) mogenous
voids (WB-2,capillary porosity
thin section, and contains few rounded
UV microscopy) Residual cement grains contain large amount of
air voids (WB-2, thin section, UV microscopy). unhydrated euhedral to subhedral, partly distorted
C3 S (Figures 7, 8 and 10). The size of single crystals
ein test, the high birefringence
Figure 6: Coarseofunhydrated
the matrix in nest
belite the (center
sub-surface zones
left) and Figure 7: Large non-hydraulic cement grain made up of platy
tions. Estimated water to residual
another cement cement
ratios grain
being(center
0.4-0.45
right).(outer
WB-1,concrete,
thin gehlenite crystals (center left, grayish blue) and a partly
hydrated residual cement grain conatining small belite and 133
ample WB-2 (Figure section, plane-polarized
5). The sample WB-1 light contained many irregularly
euhedral C3S crystals (center right). WB-2, thin section,
while the sample WB-2 had very few entrapped air voids (1 to 2 vol.%, cross-polarized light
 mostly contains small residual belite nuclei (e.g. Figure 8). Several crystals sh
decomposition into belite (Figure 10). This latter observation indicates slow
Pintér & Gosselin
sizes
The most characteristic feature of both samples is aoflarge
alite amounts
can also of
be very
explained
coarsebyresidual
an inhomogenous
cement raw meal and/or inh
s on broken particles,
surfaces exhibiting
(Figure 3),a maximum
samples exhibited compact kiln.
of 0.9 and an average diameter of 0.2 to 0.3 mm (Figures 6 to 11), typical for
not be observed. By the
historic PC use of phenolphtalein
mortars on fresh
and concretes (Idorn & fracture
Thaulow, 1983; Blezard, 1984, Weber et al, 2012).
e 3) at ca. Figure
0.5 to 4:1 High
cm fromamounts
A the ofsurface
significant
compaction pores and
indicated
portion of the low
cement
to
preserved
grains
Figure 5: The compact cement paste exibits slightly
have remained unhydrated (Figures 6 to 11). Traces of
moderate
ests adequatelyhydrationw/c ratio
low capillarity, of the non-carbonated
probably cement
due to a on paste
relatively in the
low of the inhomogenous capillary porosity and contains few rounded air
sample WB-1 (thincould mainly
section, UVbemicroscopy)
observed the margins residual
(WB-2,phases. Due UVto the higher w/c ratio the inner
low, 1983). On the contrary
concrete layer of to thethis, the WB-1
sample more exhibits tampeddegreevoids
porous higher of hydration.
thin section, microscopy)
d not show any discoloration with effectcontained
The aggregate from the inphenolphtalein
the mortars at grain sizes of 0.1 to 3 mm (WB-1) and 0.1 to 4 (WB-2) is
Optical microscopy
nd the influence of carbonation
exclusively and/or leachingtheonestimated
the cement
Confirming the phenolphtalein test, the binder
of crushed dolomite; to aggregate of
high birefringence ratio
thebeing
matrix1 to
in 2.5-3 (WB-1) and
the sub-surface 1 to 4
zones
e.g. fitting gaps on
(WB-2) the façade.
with a definetely higher amount of fine fraction (0 to 0.05 mm) in the sample WB-2, which probably
corresponds to carbonated binder portions. Estimated water to cement ratios being 0.4-0.45 (outer concrete,
also 4,
Figure hadWB-1)
an impact and on the higher
~0.50 for thewater
sampledemand
WB-2and thus higher
(Figure 5). Thecapillary
sampleporosity
WB-1 of the mortar
contained (cf. irregularly
many Figures 4 and
5). compaction pores (Figure 4), while the sample WB-2 had very few entrapped air voids (1 to 2 vol.%,
shaped
Figure 5). The cement paste of both samples were free of any cracks.
Characteristics of cement phases
Residual cement grains that entirely or partly survived the hydration and later carbonation were used
to determine the composition and properties of the original cement. In terms of composition and grain size of
single crystals, cement grains exhibited very inhomogeneous features. Interpretation of the observations was
based on Insley (1936), Gille et al (1965), Idorn & Thaulow (1983), Blezard (1984), and Campbell (1999).

Alite Figure 8: Large residual cement grain containing angular alite Figure 9: Large resid
Residual cement grains contain large amount of (tan)Figure and8:rounded
unhydrated Large residual cement
beliteto(blue)
euhedral grain
crystals.
subhedral, containing
Interstitial
partly Cangu-
phases
distorted are
3S
(brown) and C3A as
(Figures 7, 8 and 10). The size of single crystals varies lar made up of
alite (tan)
between aluminate
5 and and (grey)
80rounded and
μm (average ferrite (white).
belite (blue) WB-1,
20 to crystals.
50 μm). Intersti-
Alite
mostly contains small residual belite nuclei (e.g. Figure 8).reflected
tial phases
Several light,
areNital
crystals madeetching
show up of aluminate
corroded surfaces(grey) and ferrite
and marginal
decomposition into belite (Figure 10). This latter observation indicates
(white). WB-1, slow rates oflight,
reflected cooling,
Nitalwhile the different
etching.
sizes of alite can also be explained by an inhomogenous raw Belite
meal and/or inhomogenous firing conditions in the
kiln.5: The compact cement paste exibits slightly Belite shows round to angular, often atypical or distorted shapes.
to Figure
e in the inhomogenous capillary porosity and contains few rounded air range of 5 to 100 μm, though single crystals around 20-30 μm are frequent (F
voids (WB-2, thin section, UV microscopy) nests of coarse, spherical C2S with very little interstitial material suggest co
quartz in the raw feed.
Figure 6: Coarse unhydrated belite nest (center left) and Figure 7: Large non-hydraulic
Segmented cementsections
fingerlike grain made up of platy
(Figure 11) were observed both in thin
gehlenite
slow crystals and
cooling (center left, grayish
eventually to blue) and aconditions.
reducing partly Identical features were re
Figure 6: Coarse
the high birefringence
another residual unhydrated
of the
cement matrix inbelite
grain the nest
(center right). (center left)
sub-surface
WB-1, zones
thin and
section, plane-polarized light hydrated
and residual (1984)
Blezard cement grain
on a conatining
historic small
PC belite and cement) from 1850, burned
(Aspdin´s
imated water to cement ratios being 0.4-0.45 (outer concrete,
another residual cement grain (center right). WB-1, thin euhedral C3S crystals (center right). WB-2, thin section,
B-2 (Figuresection,
5). Theplane-polarized
sample WB-1light. contained many irregularly droplets (Figure 11) could frequently be observed as inclusions within C3A,
cross-polarized light
sample WB-2 had very few entrapped air voids (1 to 2 vol.%, silicates crystallized from the flux (Gille et al., 1965). The lamellar structure
e free of any cracks. be observed in transmitted light or after the structural etching in reflected

Figure 8: Large residual cement grain containing angular alite Figure 9: Large residual cement grain containing Type II belite
(tan) and rounded belite (blue) crystals. Interstitial phases are Figure
(brown) 9:
andLarge residual
C3A as cement
flux. WB-1, grainlight,
reflected containing Type
Nital etching
made up of aluminate (grey) and ferrite (white). WB-1, II belite (brown) and C3 A as flux. WB-1, reflected light,
reflected light, Nital etching Nital etching.
Belite
Belite shows round to angular, often atypical or distorted shapes. Its crystal size varies within a wide
varies between 5 and 80 µm (average 20 to 50 µm).
range of 5 to 100 μm, though single crystals around 20-30 μm are frequent (Figures 7 and 9). Up to 0.6 mm large
Alite mostly contains small residual belite nuclei
nests of coarse, spherical C2S with very little interstitial material suggest concentration and inhomogenities of
quartz in thenon-hydraulic
raw feed. cement grain made up of platy (e.g. Figure 8). Several crystals show corroded
Figure 7: Large
d Segmented fingerlike sections surfaces and marginal decomposition into belite
n Figure
gehlenite 7:crystals
Large non-hydraulic
(center left, cement
grayish and(Figure
blue)grain partly11)
a made up were
of observed both in thin and polished sections, referring to
slow
hydrated
platy cooling and eventually
residualcrystals
gehlenite cement grain to reducing
conatining
(center left, smallconditions.
grayish belite (Figure 10). This latter observation indicates
and aIdentical features were reported by Idorn & Thaulow (1983)
blue)and slow
and hydrated
euhedral
partly Blezard (1984)
C3S crystals oncement
(center
residual a historic
right). WB-2,
grain PC (Aspdin´s
thin rates 1850,
cement) from
section,small
conatining of cooling,
burnedwhilein a the
shaftdifferent
kiln. Verysizes of alite
small can
belite
droplets (Figure
cross-polarized light 11) could frequently be observed as inclusions also bewithin C3A, due
explained by to
anslow cooling whereraw
inhomogenous excess
meal
belite and euhedral C3 S crystals (center right). WB-2,
silicates crystallized from the flux (Gille et al., 1965). The and/or lamellar inhomogenous
structure of the larger 2S crystals can
firingCconditions easily
in the kiln.
thinbesection,
observed cross-polarized
in transmitted light.
light or after the structural etching in reflected light. The predominant types are

Belite

Belite shows round to angular, often atypical or

distorted shapes. Its crystal size varies within a
wide range of 5 to 100 µm, though single crystals
around 20-30 µm are frequent (Figures 7 and 9). Up
to 0.6 mm large nests of coarse, spherical C2 S with
very little interstitial material suggest concentration
and inhomogenities of quartz in the raw feed.

134
 th
15 with
Euroseminar on Microscopy Applied lamellae
to Building Materials • 17-19 June
crystals two or more sets of intersecting (Type I, Insley, 1936), as2015
well•asDelft, The Netherlands
irregular grains with one
set of parallel lamellae (Type II, Insley, 1936, Figure 9).

Figure 10: Distorted alite crystals (brown) showing Figure

decomposition into belite (small blue crystals) on their extensi
margin in coarse grained flux phases. WB-2, reflected light, (white)
Nital etching
crystals with two or more sets of intersecting lamellae (Type I, Insley, 1936), as well as irregular grains with one
set of parallel lamellae (Type II, Insley, 1936, Figure 9).

Figure 10: Distorted alite crystals (brown) showing Figure 11: Belite (brown) showing segmented fingerlike
Figure 10: : Distorted
decomposition alite crystals
into belite (small (brown)
blue crystals) showing
on their extensions into the a coarse crystalline ferrite-rich matrix
decomposition
margin in coarse into belite
grained flux(small
phases.blue crystals)
WB-2, onlight,
reflected their (white). WB-1, reflected light, Nital etching
Nital etching
margin in coarse grained flux phases. WB-2, reflected
light, Nital etching.
rsecting lamellae (Type I, Insley, 1936), as well as irregular grains with one
ey, 1936, Figure 9).

Figure 12: Large residue of solid fuel (right). WB-1, Figure

Figure 10: Distorted alite crystals (brown) showing Figure
reflected12: Large
light, dark residue
field of solid fuel (right). WB-1, anisotro
Figure 11: Belite (brown) showing segmented fingerlike
decomposition into belite (small blue crystals) on their reflected light, dark field.
extensions into the a coarse crystalline ferrite-rich matrix
margin in coarse grained flux phases. WB-2, reflected light, Aluminate andreflected
(white). WB-1, ferrite light, Nital etching
Nital etching Flux phases such as C3A and Ca-Al-ferrite form
(Figures 8, 9 and 11). These features are also consequence of a s
Figure 12: Large residue of solid fuel (right). WB-1, Figure 13: The same grains as in the Figure 12 containing
reflected light, dark field Remnants
anisotropic of solid
graphite fuel and
crystals. miscellaneous
Cross-polarized phases
reflected light
Black, shard-like (Figure 12), in transmitted light opa
) showing Aluminate and ferrite diameter, can be classified as solid fuel remnants. Observatio
Figure 11: Belite (brown) showing segmented fingerlike
stals) on their Figure 11:intoFlux
Belite phases such as C A and Ca-Al-ferrite form very
anisotropic coarse,
nature large and well-developed
of phases, crystals of graphite
suggesting the presence
extensions the (brown) showing segmented
a coarse crystalline 3 fingerlike
ferrite-rich matrix
B-2, reflected light, (Figures 8, 9 and 11). These features are also consequence of a slow cooling
Apart rate.
from the PC cement clinker phases described abo
extensions
(white). intoreflected
WB-1, the a coarse
light,crystalline
Nital etchingferrite-rich matrix
(white). WB-1, reflected light, Nital etching. components indicating inhomogeneous firing conditions in th
Remnants of solid fuel and miscellaneous phases gehlenite (C2AS) exhibiting platy shaped crystals of first order g
Black, shard-like (Figure 12), in transmitted light opaque inclusions, in average of 50 to 500 μm of
diameter, can be classified as solid fuel remnants. Observations in reflectedofcross
Dedolomitization polarized light proved the
aggregates
anisotropic nature of phases, suggesting the presence of graphite in the organic
Small matter of
amounts (Figure 13).
the crushed dolomite sand used
Apart from the PC cement clinker phases describedconcrete above, the samplesThe
elements. contain small products
reaction quantitiesare of other
mainly restrice
Segmented
components fingerlike
indicating sections (Figure
inhomogeneous firing11)conditions
were in the kiln. Thealongmost cracks
frequently
decomposition was observed
also observedphase(Figure
is 14). E
observed(C
gehlenite bothAS) inexhibiting
thin andplaty polished
shaped sections,
crystals refer-
of first order gray interference color (cf. Figure 7).
Figure 12: Large residue of solid fuel (right). WB-1,
2 brucite (Mg(OH) ) and calcite (Figure 14)
Figure 13: The same2 grains as in the Figure 12 containing in the zones of in
ring to slow cooling and eventually to reducing con- Figure 13:graphite
The same
reflected light, dark field expansion
anisotropic because no grains
cracks
crystals. as in the Figure
originated
Cross-polarized from the
reflected12light
con-
dedolomitized
ditions. Identicaloffeatures
Dedolomitization aggregates were reported by Idorn taining anisotropic graphite crystals. Cross-polarized
The so-called deleterious alkali-carbonate-reaction (ACR) is ma
& Thaulow Small
Aluminate (1983)
andamountsand Blezard
ferrite of the crushed (1984) on a historic
dolomite sand used as occurs
reflected
(ASR) aggregates
light. betweenshows thededolomitization
alkali-rich pore in both and micr
solution
PC (Aspdin’s
concrete cement)
elements.
Flux The from
phases such1850,
reaction Cburned
as products
3A andare
inCa-Al-ferrite
amainly
shaft restriced to thecoarse,
form very outer large
rim ofandsome of the grains,
well-developed but
crystals
kiln. Very
decomposition small alongbelite
cracksdroplets
was (Figure
also 11)
observed
(Figures 8, 9 and 11). These features are also consequence could
(Figure 14).ofEvidence of dedolomitizations
a slow cooling rate. are small rims of
brucite
frequently(Mg(OH) 2) and calcite
be observed (Figure 14)
as inclusions withinin the zones of interaction. The process itself has not produced
C3 A,
expansion
due because
to slowof
Remnants cooling no
solid fuelcracks
where originated from
excess silicates
and miscellaneous the dedolomitized
crystal-
phases aggregate into the cement paste were observed.
The so-called
lized from Black, deleterious
the flux alkali-carbonate-reaction
(Gille et (Figure
shard-like al., 1965).12),The (ACR) is
lamellar light opaque
in transmitted made up of two reactions;
inclusions, the alkali-silica-reaction
in average of 50 to 500 μm of
ght). WB-1, Figure
(ASR) 13: The
occurs
diameter,ofcan
structure same
thebe grains
between as
the
classified
larger in the Figure
alkali-rich 12
as solid fuel
C2 S crystals porecontaining
solution
can remnants. and micro- or cryptocrystalline
easily be Observations in reflected cross polarized SiO 2 as impurity
light provedin thethe
anisotropic graphite
anisotropic naturecrystals.
of Cross-polarized
phases, suggesting reflected
the light of graphite in the organic matter (Figure 13).
presence
observed in transmitted light or after the structural
etching in Apartreflectedfromlight.
the PCThe cement clinker phases
predominant types described above, the samples contain small quantities of other
components indicating inhomogeneous firing conditions in the kiln. The most frequently observed phase is
A and Ca-Al-ferrite areform crystals with two
very coarse,
gehlenite largeorand
(C2AS) exhibiting
more sets of intersecting
well-developed
platy shaped crystals crystals
of first order gray interference color (cf. Figure 7).
are also consequencelamellae
of a slow(Type
cooling I, Insley,
rate. 1936), as well as irregular
grains with one set of parallel lamellae (Type II,
Dedolomitization of aggregates
eous phases Insley, 1936, Figure
Small 9).
amounts of the crushed dolomite sand used as aggregates shows dedolomitization in both
2), in transmitted light concrete
opaque inclusions, in average
elements. The reactionofproducts
50 to 500areμm of restriced to the outer rim of some of the grains, but
mainly
fuel remnants. Observations in reflected
decomposition alongcross polarized
cracks was also light proved(Figure
observed the 14). Evidence of dedolomitizations are small rims135
of
ing the presence of graphite in the
brucite organic2)matter
(Mg(OH) (Figure(Figure
and calcite 13). 14) in the zones of interaction. The process itself has not produced
linker phases describedexpansion
above, thebecause
samplesnocontain
crackssmall quantities
originated from of
theother
dedolomitized aggregate into the cement paste were observed.
 Pintér & Gosselin

Aluminate and ferrite

Flux phases such as C3 A and Ca-Al-ferrite form
very coarse, large and well-developed crystals (Fig-
ures 8, 9 and 11). These features are also conse- dolomite stone and the above described dedolomitization (Katayama, 2010;
quence of a slow cooling rate. indicates the reaction between the high alkali pore solution and the CaMg
(Katayama, 2010) has shown that expansive forces do not happen upon
Remnants of solid fuel and miscellaneous phases formation of cracks, which reach the cement paste, are observable, the da
Since neither microcrystalline SiO2 in the dolomite aggregates, nor alkali-s
Black, shard-like (Figure 12), in transmitted light vicinty of the dedolomitzed areas, the observed marginal decomposition of th
opaque inclusions, in average of 50 to 500 µm of di- ACR.
ameter, can be classified as solid fuel remnants. Ob-
servations in reflected cross polarized light proved
the anisotropic nature of phases, suggesting the
presence of graphite in the organic matter (Figure
13).
Apart from the PC cement clinker phases de-
scribed above, the samples contain small quantities
of other components indicating inhomogeneous fir-
ing conditions in the kiln. The most frequently
observed phase is gehlenite (C2 AS) exhibiting platy
shaped crystals of first order gray interference color
(cf. Figure 7).
dolomite stone and the above described dedolomitization (Katayama, 2010; Grattan-Bellew et al, 2010), which
Dedolomitization
indicates the reaction of aggregates
between the high alkali pore solution and the CaMg(CO3)2. Former extensive research
(Katayama, 2010) has shown that expansive forces do not happen
Figure upon dedolomitization.
14: Marginal dedolomitization ofIfdolomit
expansion and
aggregates Figure 15: Fine gra
Small amounts of the crushed dolomite sand used
formation of cracks, which reach the cement paste, are observable, Figure 14: the
into brucite Marginal
and damagededolomitization
calcite. is always
WB-1, of dolomit
connected
SEM-BSE aggre-
to ASR. AFm phases (prob
asSince
aggregates
neither shows dedolomitization
microcrystalline SiO2 in the indolomite
both con-aggregates, nor alkali-silica gel haveWB-1,
been SEM-BSE.
detected in the
gates into brucite and calcite. 2, SEM-BSE
crete elements.
vicinty The reaction
of the dedolomitzed products
areas, are mainly
the observed marginal decomposition of the aggregates cannot be classified as
restricted
ACR. to the outer rim of some of the grains, but Chemical compositon of the binders
decomposition along cracks was also observed (Fig- The elemental composition of hydration products, measured by E
ure 14). Evidence of dedolomitizations are small test, on the non-carbonanated zone of the matrix, has been expressed in atom
rims of brucite (Mg(OH )2 ) and calcite (Figure 14) Si/Ca vs. Al/Ca plots (Figs. 1a and 2a) of both samples show th
in the zones of interaction. The process itself has AFt (ettringite) and AFm phases. Furthermore, the Si/Ca ratio of the inner
not produced expansion because no cracks origi- sample WB-2 (Fig 2b; yellow ellipse), higher values than those of the theo
nated from the dedolomitized aggregate into the reference for pure synthesized C1.7SH. This difference may be explained
and/or by the influence of the clay fraction in the raw materials and su
cement paste were observed. The so-called deleteri-
portlandite confirms the alkaline environment in the binder
ous alkali-carbonate-reaction (ACR) is made up of
Values plotted in the Al/Ca vs. S/Ca plots (Figs. 1b and 2b) confir
two reactions; the alkali-silica-reaction (ASR) occurs sample WB-1 (Fig. 1b) the main AFm phase is monosulphate, coexisting pro
between the alkali-rich pore solution and micro- or of monocarbonate, as suggested by the dots marked by the red ellipse. In
cryptocrystalline SiO2 as impurity in the dolomite completely replaced by monocarboaluminate, as characterized by the values
stone and the above described dedolomitization (S/Ca). The presence of monocarbonate can be explained by the high a
(Katayama, 2010; Grattan-Bellew et al, 2010), which (dolomite) filler, which has partly reacted during hydration with the aluminat
indicates the reaction between the high alkali pore There is a considerable amount of sulphate-bearing phases in
solution and the CaMg ( CO 3 ) 2 . Former extensive presence of primary Ca-sulphate in the cement. Historical data (Riepl,
Figure 14: Marginal dedolomitization of dolomit aggregates Figure 15: Fine grained dolomite aggregates, portlandite and
research (Katayama, 2010)SEM-BSE
has shown that expan- Association of Austrian Cement Plants allowed a maximum gypsum con
into brucite and calcite. WB-1, Figure 15: Fine
AFm phases grained
(probably dolomite aggregates,
monocarboaluminate). portlandite
WB-
sive forces do not happen upon dedolomitization. interpreted
2, SEM-BSE as a deliberately admixed additive. However, the use of a solid
and AFm phases (probably monocarboaluminate). WB-2,
If expansion and formation of cracks, which reach phases (Michaelis, 1869) and its effect on the clinker composition (e.g. large
SEM-BSE.
during firing) as well as on hydrate phases cannot be excluded.
Chemical
the cementcompositon
paste, areofobservable,
the binders the damage is
The elemental composition of hydration products, measured The by effect
EDS, of excess sulphate
according is suggested by the atomic ratio p
to the phenlophtalin
always connected to ASR. Since neither microcrys-
test, onSiO
talline the non-carbonanated zone of the matrix, has been expressed in atomic ratio plots Ca, Si, Al and S. which could also be dete
monosulphate to produce secondary ettringite,
2 in the dolomite aggregates, nor alkali-
Si/Ca vs. Al/Ca plots (Figs. 1a and 2a) of both sometimes samples show in the
theform of reactivated
presence C3fine
of different A around unhdrated residual cemen
grained
silica gel have been detected in the vicinity of the
AFt (ettringite) and AFm phases. Furthermore, the Si/Ca forces ratio ofcannot be seen.
the inner C-S-H indicates, especially in the
dedolomitzed
sample WB-2 areas,(Fig 2b;theyellow
observed marginal
ellipse), higher decom-
values than those of the theoretical Si/Ca ratio (0.58) given as a
position of the aggregates cannot
reference for pure synthesized C1.7SH. This be classified as may be explained either by partly decalcified C-S-H
difference
ACR.
and/or by the influence of the clay fraction in the raw materials and subsequent clinker. The presence of
portlandite confirms the alkaline environment in the binder
Chemical composition
Values plottedof in the binders
the Al/Ca vs. S/Ca plots (Figs. 1b and 2b) confirm the nature of AFm phases. In the
sample WB-1 (Fig. 1b) the main AFm phase is monosulphate, coexisting probably also with some minor amount
The elemental composition
of monocarbonate, as suggested of hydration
by the dotsproducts,
marked by the red ellipse. In the sample WB-2 monosulphate is
measured
completelybyreplaced
EDS, according to the phenolphthalein
by monocarboaluminate, as characterized by the values 0.15 to 0.4 (Al/Ca) and 0.02 to 0.05
(S/Ca). The presence of monocarbonate can be explained by the high amounts of fine grained carbonate
(dolomite) filler, which has partly reacted during hydration with the aluminate phases.
136
There is a considerable amount of sulphate-bearing phases in both samples, which suggests the
presence of primary Ca-sulphate in the cement. Historical data (Riepl, 2008) show that since 1889 the
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
test, on the non-carbonated zone of the matrix, has
been expressed in atomic ratio plots Ca, Si, Al and
S.
Si/Ca vs. Al/Ca plots (Figures 16a and 17a) of
both samples show the presence of different fine
grained AFt (ettringite) and AFm phases. Further-
more, the Si/Ca ratio of the inner C-S-H indicates,
especially in the sample WB-2 (Figure 17b; yellow
ellipse), higher values than those of the theoretical
Si/Ca ratio (0.58) given as a reference for pure syn-
thesized C1.7 SH. This difference may be explained
either by partly decalcified C S H and/or by the
influence of the clay fraction in the raw materials
and subsequent clinker. The presence of portlandite
confirms the alkaline environment in the binder.
Values plotted in the Al/Ca vs. S/Ca plots (Fig-
ures 16b and 17b) confirm the nature of AFm phases.
In the sample WB-1 (Figure 16b) the main AFm
phase is monosulphate, coexisting probably also
with some minor amount of monocarbonate, as
suggested by the dots marked by the red ellipse.
In the sample WB-2 monosulphate is completely
replaced by monocarboaluminate, as characterized
by the values 0.15 to 0.4 (Al/Ca) and 0.02 to 0.05
(S/Ca). The presence of monocarbonate can be
explained by the high amounts of fine grained car-
bonate (dolomite) filler, which has partly reacted
during hydration with the aluminate phases.
There is a considerable amount of sulphate-
bearing phases in both samples, which suggests
the presence of primary Ca-sulphate in the cement.
Historical data (Riepl, 2008) show that since 1889
the Association of Austrian Cement Plants allowed
a maximum gypsum content of 2% in PCs, which
can be interpreted as a deliberately admixed addi-
tive. However, the use of a solid fuel containing
natural SO3 -bearing phases (Michaelis, 1869) and
its effect on the clinker composition (e.g. larger
amount of alkali-sulphates formed during firing) as
well as on hydrate phases cannot be excluded.
The effect of excess sulphate is suggested by
the atomic ratio plots. External SO3 has reacted
with monosulphate to produce secondary ettringite,
which could also be detected as secondary pore
fillings or sometimes in the form of reactivated C3 A
around unhydrated residual cement. A crack pat-
tern due to expansive forces cannot be seen.
IV. Conclusions
Using reflected and transmitted polarized light mi-
croscopy, completed with SEM-EDS measurements
on polished and thin sections, the composition
and material characteristics of two historic, pre-
fabricated PC concrete elements from the Parish
church in Weissenbach an der Triesting, dating back
to 1893, were carried out. The low to moderate
w/c ratio and the use of historical grinding tech-
niques have induced the preservation of several
unhydrated residual cement grains in the samples.
These residual grains contained all four main PC
phases; their relative amount, crystal size and shape
suggested, however, different burning and cooling
conditions and raw feed composition in compar-
ison with modern PC clinkers. Features such as
inhomogeneous crystal sizes of alite exhibiting fre-
quently decomposition into belite on its margin,
fingerlike sections of belite and coarse crystalline
flux phases suggest slow cooling rates and eventu-
ally also reducing conditions during the calcination
process. Due to the coexistence of the different be-
lite types, uneven kiln temperatures and cooling
rates, typical for shaft kilns, can be assumed. Beside
the aforementioned observations, for modern PC
atypical clinker phases containing C2 AS and prob-
ably also other low temperature, non- or slightly
hydraulic Ca-(aluminate)-silicates also suggest in-
homogeneous heat distribution during firing. Small
fragments of graphite-bearing inclusions indicate
the use of a solid fuel.
Crushed dolomite sand used as aggregate in both
concrete elements exhibited partly dedolomitiza-
tion on marginal zones and along cracks. The reac-
tion, however, is not connected with the formation
of ASR, thus it is harmless and cannot be defined
as alkali-carbonate reaction.
Binder portions with preserved alkalinity en-
abled to determine the nature of hydration products
based on the chemical composition of the cement
paste. Although the formation of secondary et-
tringite from monosulphate due to excess SO42
rather suggests the use of Ca-sulphate as setting re-
tarder admixed to the cement, the effect of sulphur-
rich fuel producing alkali-sulphates in the clinker
phases during firing cannot be excluded either. One
of the samples clearly indicated the effect of the
fine-grained carbonate aggregate portion on the hy-
dration products: monosulphate was replaced by
monocarboaluminate. In the other sample mono-
sulphate coexisted with some minor amount of
monocarbonate.
Based on the extensive investigations of the sam-
ples it can be concluded, that the cement used for
producing the concrete elements was burned in a
shaft kiln above the sintering temperature, but still
under inhomogeneous heat distribution, with prob-
ably long residence time, slow cooling rates and
inhomogeneous raw feed. Chemical data of the ce-
ment paste suggests the use of artificially admixed
Ca-sulphate as a setting retarder.
The factors such as the w/c-ratio, grading curve
of the aggregates and the partial reaction of the
sand particles with clinker phases during the hydra-
tion have played an important role in a remarkable
performance of the concrete cast elements during
the last 120 years.
137
 Pintér & Gosselin

Figure 16a: Si/Ca vs. Al/Ca plot of the X-ray data (WB-1)Figure 16b: Al/Ca vs. S/Ca plot of the X-ray data (WB-1)
Figure 16: a) Si/Ca vs. Al/Ca plot of the X-ray data (WB-1), b)Al/Ca vs. S/Ca plot of the X-ray data (WB-1).
Figure 16a: Si/Ca vs. Al/Ca plot of the X-ray data (WB-1) Figure 16b: Al/Ca vs. S/Ca plot of the X-ray data (WB-1)

Figure 17a: Si/Ca vs. Al/Ca plot of the X-ray data (WB-2) Figure 17b: Al/Ca vs. S/Ca plot of the X-ray data (WB-2)
Figure 17a: Si/Ca vs. Al/Ca plot of the X-ray data (WB-2) Figure 17b: Al/Ca vs. S/Ca plot of the X-ray data (WB-2)
Figure 17: a) Si/Ca vs. Al/Ca plot of the X-ray data (WB-2), b)Al/Ca vs. S/Ca plot of the X-ray data (WB-2).
Conclusions
Using reflected and transmitted polarized light microscopy, completed with SEM-EDS measurements
Conclusions
on polished and thin sections, the composition and material
Acknowledgement Katayama,characteristics of twoalkali-carbonate
T. (2010): “The so-called historic, prefabricated
reaction (ACR) PC
Using reflected and transmitted polarized light microscopy, completed with SEM-EDS measurements
concrete elements from the Parish
on polished and thin sections, the composition and material church in Weissenbach an der
– Its mineralogical Triesting,
geochemicaldating back
details, with to
special1893,
referencewere
to ASR”.carried
Cementcharacteristics
and Concrete Research, ofVol.
two historic,
40, pp. 643-6751.prefabricated PC
The
concrete authors
out. The elements would
low to moderate like
from thew/c to thanks
Parish to
ratiochurch Mr.
and theinuse Helmut of historical
Weissenbach grinding
Michaelis,
an W. techniques
der Triesting, (1869): „Die
dating have
backinduced
hydraulischento Mörtel
1893, the preservation
insbesondere
were der
carried of
Heimel
several (Weissenbach
unhydrated an
residual der Triesting)
cement
out. The low to moderate w/c ratio and the use of historicalBautechniker,for
grains support-
in the samples. These
Portland-Cement
grinding techniques residual
in chemischgrains contained
- technischer Beziehungall für four main
Fabrikanten, PC
Ingenieure undhave induced
Chemiker“. thevonpreservation
- Verlag Quandt & Händel, of
ing the unhydrated
phases;
several sampling
their relative and providing
amount,
residual cement valuable
crystal
grainssize informa-
in andsamples.
the shapeLeipzigsuggested,
TheseS. 315residual
p. however, different burning
grains contained all four main and cooling
PC
conditions
tion about
phases; their and raw feed
therelative
history of composition
the object.
amount, crystalinsize comparison
and shape with modern PC
Riepl, F. (2008):
suggested, clinkers.
“Die
however, Features
wirtschaftliche
different such as and
und technologische
burning inhomogeneous
Entwicklung
cooling
der Zementindustrie unter besonderer Berücksichtigung der Verdienste
crystal sizes
conditions andofraw alite feedexhibiting
composition frequently decomposition
in comparison voninto
with modern Hans belite on itsUnpublished
PC clinkers.
Hauenschild”. margin, fingerlike
Featuresdoctoral
such thesis, sections Wien,
as inhomogeneous
Universität of belite
and coarse crystalline
crystal sizes of aliteReferences flux phases suggest
exhibiting frequently decomposition intoslow cooling rates
140 p. and eventually also reducing conditions during the
belite on its margin, fingerlike sections of belite
Weber, J., Bayer, K., Pintér, F. (2012): 19th century ’novel’ building ma-
calcination
and process. Due
coarse crystalline flux tophasesthe coexistence
suggest slowofcooling the different
rates and
terials: belite types,
eventually
examples uneven
also
of various kiln temperatures
reducing
historic mortars conditions and cooling
during
under the microscope”. the
In:
calcination process. Due to the coexistence of the different belite types, uneven kiln temperatures and PC
rates, typical
Blezard, R.G. for
(1984): shaft
“A kilns,
discussion of can
the be
Paper assumed.
“Examination Beside
of 136- the
J. aforementioned
Válek, J. J. Hughes, observations,
C.J.W.P. Groot (Eds.), for
Historicmodern
mortars- atypical
Characteriza-
cooling
Years-Old Portland Cement tion, Assessment and Repair, RILEM 7, Springer, pp. 89-103.
clinker
rates, phases
typical for shaftConcrete,”
containing kilns, C
Cement and Concrete Research, Vol. 14, pp. 154-156.
by AS
can G.M. Idorn and
2 be and assumed. N. Thaulow”.
probably Beside alsothe other low K.temperature,
aforementioned
Zehetner, (2010):observations,
“Zementwerke non-infor or modern
slightlyPC
Niederösterreich“. hydraulic
G. Lind- Ca-
In:atypical
(aluminate)-silicates
clinker
Blezard,phases containing
R.G. (2004): also suggest
“The historyC2AS inhomogeneous
and probably
of calcareous cements”.also In. other low temperature, non- or slightly hydraulic graphite-
heat distribution
ner (Ed.), during
Denkmalpflege infiring. Small
Niederösterreich, fragments
Band 43: Beton,ofpp. Ca-
201-
(Hewlett, P.C. ed.) Lea’s chemistry of cement and concrete. 4th Edition, 21.
bearing inclusions indicate
(aluminate)-silicates also suggestthe use of a solid fuel.heat distribution during firing. Small fragments of graphite-
inhomogeneous
Elsevier, pp. 1-21.
bearing
Campbell,inclusions
Crushed
D.H. (1999):indicate
dolomite the use
“Microscopical sand ofused
a solid
Examination andfuel.
as aggregate
Interpreta- in both concrete elements exhibited partly dedolomitization
on
tion marginal
of Portland Crushed
zonesand
cement dolomite
and alongsand
Clinker”. usedThe
cracks.
2nd Edition, as aggregate
reaction,
Portland Cement As- in both concrete
however, is not elements
connectedexhibited with thepartly dedolomitization
formation of ASR, thus it
sociation, USA, 201p.
on marginal
is harmless
Gille, zones
and
F., Dreizler, I., and
cannot
Grade,along
be cracks.
K., defined
Krämer, H.,asThe reaction,
alkali-carbonate
Woermann, however,
E. (1965): reaction.is not connected with the formation of ASR, thus it
is harmlessdes
“Mikroskopie and
Binder cannot be defined
portions
Zementklinkers, as Association
with preserved
Bilderatlas“. alkali-carbonate
of the Ger-reaction.
alkalinity enabled to determine the nature of hydration products based
man Cement Industry, Beton-Verlag, Düsseldorf, West Germany, 75 p.
onHeimel, Binder
the chemical portions
composition with preserved
of the alkalinity
cement
C., Heimel, H., Schicht, P. (2010): “Frühe Betonfertigteile in
enabled
paste. to determine
Although the the nature ofofhydration
formation secondary products basedfrom
ettringite
on the chemical
monosulphate
Niederösterreich duecomposition
– Freiherr to
vonexcess
Pittel undSO of the
2-
die Kirche
4
cement
rather paste.
suggests
von Weissenbach Although
the use of the formation
Ca-sulphate as of secondary
setting retarderettringite
admixed from to the
an der Triesting“. In:
monosulphate
cement, the dueG. Lindner
effect to
of excess
sulphur-rich SO42- rather
(Ed.), Denkmalpflege
fuel
in Niederösterre-
suggests the
producing use of Ca-sulphate
alkali-sulphates in the as setting
clinker phasesretarder
during admixed
firing to the be
cannot
ich, Band 43: Beton, pp. 42-45.
cement,
excluded the
Idorn, G.M., effect
either. One
Thaulow, ofN.sulphur-rich
of the “Examination
(1983): samples fuel clearly
producing alkali-sulphates
indicated
of 136- Years-Old the effect in of the
theclinker
fine-grainedphasescarbonate
during firing cannot portion
aggregate be
excluded
Portland Cement
on either. One
Concrete”. of the
Cement samples
and Concreteclearly
Research, indicated
Vol. 13, pp. the effect of the fine-grained carbonate aggregate portion
the hydration products: monosulphate was replaced by monocarboaluminate. In the other sample
739-743.
on the H.hydration
monosulphate
Insley, (1936):coexisted products:
“Structural with monosulphate
some ofminor
Characteristics was replaced
amount
Some Constituents by monocarboaluminate. In the other sample
of monocarbonate.
monosulphate
of Portland Cementcoexisted with
Clinker,” Journal some
of Research minor amount
of the National Bureau ofofmonocarbonate.
Standards, Vol. Based
17, ResearchonPaper
the RP917,
extensive investigations
Washington, D.C., pp. 353-361. of the samples it can be concluded, that the cement used for
Based onMitchell
the extensive investigations of the samples it can be concluded, that the cement used for
producing
Grattan-Bellew theP.E.,
concrete elements
L.D., Margeson wasJ., burned in a shaft kiln above the sintering temperature, but still under
Min, D. (2010):
producing
“Is alkali-carbonatethe concrete
reaction is elements
just a variantwas burned in
of alkali-silica a shaft kiln above the sintering temperature, but still under
reaction
ACR=ASR?”. Cement and Concrete Research, Vol. 40, pp. 556-562.

138
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Microscopic study on Lombard Mannerist

stuccoes in comparison with Vasari’s Lives
of the Artists
Roberto Bugini⇤ , Luisa Folli
CNR-ICVBC, via Roberto Cozzi 53, Milano
⇤ bugini@icvbc.cnr.it

Abstract

Different samples of stucco-work from some rooms of a Mannerist italian building (Palazzo Te, Mantua - built in
1525-35 by Giulio Romano) were analysed using two microscopic methods (optical microscopy on thin section, scanning
electron microscopy) in order to investigate the mineralogical composition of the binder and the aggregate of the finish
coat. Two kinds of stucco were detected: the first one contains dolomitic lime as binder and quartz and silicate crystals
as aggregate; the second one contains calcium lime as binder and calcite crystals or limestone clasts as aggregate.
The stucco composition and the presence of two different combinations in two stucco-works of the same room were
compared with some excerpts of Vasari’s Lives of the Artists (1568) concerning Giulio Romano, Francesco Primaticco
and Giovanni da Udine.

Keywords: stucco, aggregate, Italian Mannerism, Mantua

I. Introduction II. Mantua and Palazzo Te

Stucco is a building material easily shaped and

largely used for architectural decoration and sculp-
ture (Davey, 1961). The employ started in Roman
times for the bas-relief decoration on the walls
and vaults (Robertson, 1964); the technique was
brought back by Giovanni da Udine (early 16th cen-
tury)(Murray, 1967), then it was improved in the
Baroque and Rococo decorations of Northern Italy
(Wittkover, 1963), Austria and Bavaria (Hempel,
1965).
The typical stucco is made of two superimposed
coats: the lower one, or render coat, is made of
lime and fluvial sand, the upper one, or finish coat,
is made of lime and marble powder. As other ar-
tificially made materials, the stucco features may
change according to the location, to the craftsman
or to the period of setting, particularly whit regard
to the aggregate employed in the finish coat (Cav-
allo et al., 2005; Rampazzi et al., 2012).
An outstanding example of stucco-work of the
Mannerist period is the decoration of Palazzo Te
in Mantua, built and decorated on a plan made
by Giulio Romano. The comparison between the
analytical data and some excerpts of Vasari’s Lives
of the Artists regarding the decoration of Palazzo
Te, should be useful to discriminate the work of
different craftsmen in this building.
Mantua is a Lombard town lying in the lower Pada-
nia plain near the river Po and it was the capital of
the Gonzaga’s Duchy from 1328 to 1708. Palazzo Te
was built outside Mantua by Giulio Romano in the
years 1525/35 for the leisure of the Duke Federigo.
Giulio Romano, born in Rome (1499), was a follower
of Raphael and worked at the Gonzaga’s court from
1524 until his death (Mantua 1546). Palazzo Te has
only a ground floor and mezzanine with a Doric
order, all in brick with rough-cast facing (Figures 1-
2); Giorgio Vasari described it as a quadrangular
building with an open courtyard in the centre (ex-
cerpt A). As declared by Burckhardt: "Palazzo Te
contains the most complete group surviving intact
of many decorated rooms of differing size, height,
architectural design and purpose, the decoration
largely as an accompaniment to cycles of pictures;
these presented in very varied forms and arrange-
ments, in oil as well as fresco. Colour and decora-
tive motives determined according to the scale and
function of the room; alternation of painting and
white or polychrome stucco" (Burckhardt, 1987).
Three decorated rooms were chosen in order to
analyse different kinds of stucco-works.

139
 Bugini & Folli
Figure 1: Palazzo Te, looking from south.
III. Methods
The scientific study is pointed out on the stucco fin-
ish coat (2-3 mm thick) in order to identify the com-
ponents. The study employed some microscopic
methods as: optical microscopy on thin section
cut normally to the surface; scanning electron mi-
croscopy on polished section and X-ray diffraction
on powders.
IV. Sampling
a) Camera dei Venti (#7 in the plan, Figure 2): a
moulded cornice with Harpies at the top of the
wall; a mythological personage below the corbels
of the vault.
b) Camera delle Aquile (#8 in the plan, Figure
2): two putti, part of a band surrounding some
quadrangular panels in the vault; a winged figure
(Harpy) above the corbels of the vault.
c) Camera degli Stucchi (#10 in the plan, Figure
2): two parallel friezes with soldiers below the vault,
the Hercules Statue in the lunette.
Figure 2: Plan (digits refer to the rooms studied).
140
Figure 3: Harpy of the moulded cornice.
V. Results
a) Camera dei Venti
a1) Leg of a Harpy of the moulded cornice (Fig-
ures 3- 4).
Binder: dolomitic lime, round cavities. Aggre-
gate: high proportion; well sorted angular quartz
(crystals and polycrystalline grains, size range 0.1-
0.6 mm), together with plagioclase crystals (maxi-
mum size 0.3 mm), muscovite laminae (maximum
size 0.2 mm) and limestone clasts (maximum size
0.3 mm).
a2) Hairs on a shoulder of a mythological person-
age (Figures 5- 6).
Binder: dolomitic lime, round cavities. Aggre-
gate: medium-high proportion; moderately well
sorted angular quartz (crystals and polycrystalline
grains, size range 0.1-0.6) mm) together with gneiss
clasts (maximum size 0.8 mm), plagioclase crystals
(maximum size 0.4 mm), muscovite laminae (maxi-
mum size 0.4 mm) and limestone clasts (maximum
size 0.4 mm).
Figure 4: Thin section, cross polarized light.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 5: Hair of a mythological personage.
b) Camera delle Aquile
b1) Arm of a putto (Figures 7- 8).
Binder: dolomitic lime, irregular shaped cavities.
Aggregate: medium-high proportion; poorly sorted
angular quartz (crystals and polycrystalline grains,
size range 0.1-0.8 mm) together with gneiss clasts
(maximum size 0.8 mm), plagioclase crystals (maxi-
mum size 0.6 mm), muscovite laminae (maximum
size 0.2 mm) and limestone clasts (maximum size
0.6 mm).
b2) Leg of a putto (Figures 9- 10).
Binder: dolomitic lime, round cavities. Aggre-
gate: high proportion; moderately well sorted angu-
lar quartz (crystals and polycrystalline grains, size
range 0.1-0.8 mm) together with gneiss clasts (max-
imum size 0.8 mm), plagioclase crystals (maximum
size 0.6 mm), muscovite laminae (maximum size
0.3 mm) and limestone clasts (maximum size 0.3
mm).
Figure 6: Thin section, cross polarized light.
Figure 7: Arm of a putto.
b3) Wing of a Harpy (Figures 11- 12)
Binder: dolomitic lime, round cavities. Aggre-
gate: medium-high proportion; poorly sorted an-
gular quartz (crystals and polycrystalline grains,
size range 0.1-0.8 mm) together with gneiss clasts
(maximum size 1.2 mm), plagioclase crystals (size
0.4 mm), muscovite laminae (maximum size 1.0
mm) and limestone clasts (maximum size 1.0 mm).
Presence of straw.
c) Camera degli Stucchi
c1) Upper frieze: arm of a soldier (Figures 13- 14).
Binder: dolomitic lime, round cavities. Aggre-
gate: high proportion; moderately well sorted angu-
lar quartz (crystals and polycrystalline grains, size
range 0.1-1.0 mm) together with gneiss clasts (max-
imum size 0.8 mm), plagioclase crystals (maximum
size 0.2 mm), muscovite laminae (maximum size
0.2 mm) and limestone clasts (maximum size 0.4
mm). Presence of straw.
Binder: calcium lime, round cavities. Aggregate:
medium-high proportion; poorly sorted angular cal-
cite (crystals, size range 0-1-0.6 mm; polycrystalline
Figure 8: Thin section, cross polarized light.
141
 Bugini & Folli
Figure 9: Leg of a putto.
grains, maximum size 1.0 mm); sub-angular quartz
grains (maximum size 0.4 mm).
Binder: calcium lime and gypsum, cracks and
irregular shaped cavities. Aggregate: low propor-
tion; moderately sorted sub-rounded limestone and
dolomite clasts (size range 0.1-0.6 mm) together
with calcite crystals (maximum size 0.3 mm); sub-
angular quartz grains (maximum size 1.0 mm).
Figure 10: Thin section, cross polarized light.
142
Figure 11: Harpy.
c2) Lower frieze: back of a soldier (Figures 15-
16).
c3) Hercules statue: tail of the lion skin (Fig-
ures 17- 18).
VI. Discussion
The main feature of the stucco finish coat set in
some rooms of Palazzo Te, is the use of an aggre-
gate mainly made of quartz: this kind of aggregate
is present both in moulded and in modelled ar-
tifacts. The use of quartz aggregate represents a
noticeable divergence in comparison with the tra-
ditional aggregate composition reported in ancient
stucco recipes. This divergence is also present in
comparison with several stucco finish coats set in
various periods of Lombard architecture and ana-
lyzed by microscopic methodologies (Bugini et al.,
1998; Folli-Bugini, 2001; Pracchi-Bugini, 2001).
Calcite aggregate
The stucco composition, reported by Vasari (ex-
cerpts B,C) and by many other Ancient Authors
Figure 12: Thin section, cross polarized light.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 13: Soldier of the upper frieze.
(Latin as Vitruvius and Plinius, Italian as Alberti
and Milizia, French as D’Aviler and Rondelet)
points out the presence of an aggregate made of
crushed calcite called marble powder.
In Palazzo Te, only two stucco-works of Camera
degli Stucchi (one of the friezes and a statue) con-
tain a calcium carbonate aggregate: calcite crystals
or limestone clasts (Bugini-Folli, 2011).
In the first case, calcite crystals are of two sorts:
bladed monocrystalline grains (maximum size 0.6
mm) with sharp corners at oblique angles (75 )
according to the rhombohedral cleavage and with-
out twinning; equant subrounded polycrystalline
grains (maximum size 1.0 mm) with polysintethic
twinning. The origin of this kind of aggregate is a
carbonate raw material (calcite marble, calcite vein)
artificially crushed according to the the Vitruvius
description of the marble for use in stucco: <In
places where marble is not found , the broken bits
of marble (...) which marble-workers throw down
as they works, may be crushed and ground and
used in stucco after being sifted> (Vitruvius).
In the second case, limestone clasts features (size
range 0.1-0.6 mm, equant shape, subrounded cor-
Figure 14: Thin section, cross polarized light.
Figure 15: Soldier of the lower frieze.
ners, good sorting) are compatible with the features
of river sand deposits pertaining to a basin contain-
ing limestone outcrops: river basins on carbonate
rocks are spreaded all around the Lombard Prealps.
Quartz aggregate
The whole stucco-work of both Camera dei Venti
and Camera delle Aquile and one of the friezes of
Camera degli Stucchi were made using an aggre-
gate mainly made of quartz (single crystals and
polycrystalline grains) with a low percentage of
clasts made of gneiss or limestone. Quartz clasts
are of two sorts: opaque, rod-shaped angular poly-
crystalline grains (maximum size 1.0 mm) with
undulose extinction; transparent uneven angular
single crystals (maximum size 0.6 mm) with unit
extinction (Bugini-Folli 2011).
On the basis of a microscopic comparison within
grain morphology of different kind of sands, the
quartz crystals of the stucco-work of Palazzo Te are
similar to the quartz crystals (conchoidal fractures
and irregular edges) produced by the fine crushing,
using a hammer, of a quartz pebble, pebbles of this
composition are diffused in loose deposits of the
Figure 16: Thin section, cross polarized light.
143
 Bugini & Folli
alpine rivers of Lombardy. The use of quartz aggre-
gate is neglected by Ancient Authors, but it is well
documented in many plasters made by the Romans
in Mediolanum-Milan (Bugini et al., 2013). The scant
variation observed in the aggregate composition,
normally occur in a artificially made material as
stucco: in the case of Palazzo Te, a minor fraction
of sand was probably added to the crushed quartz
fraction.
Binder composition
The binder composition also seems to modify ac-
cording to the change of aggregate composition:
the use quartz aggregate is associated with the use
of dolomitic lime; the use of calcite or limestone
aggregate is associated with the use of calcium lime.
Craftsmen
The use of two kinds of aggregate (quartz-silicate
or carbonate) in some rooms of Palazzo Te can be
a proof of the work of two craftsmen using hetero-
geneous working method. Moreover the presence
of three aggregates in the stucco-work of the Cam-
era degli Stucchi, carries several hypotheses: the
work of two craftsmen to make different parts of the
stucco decoration; the work of a craftsman using
different kinds of aggregate to make stucco accord-
ing to the typology (bas-relief, statue); a subsequent
remaking of some parts of the decoration with a
change of the raw material source.
The presence of different craftsmen during the
decoration of a building is well documented in
Palazzo Te, according to Vasari: many craftsmen
worked to decorate the rooms under the leadership
of Giulio Romano (excerpt D). In particular, two
artists are indicated concerning the decoration of
Camera degli Stucchi: Francesco Primaticcio and
Giovan Battista Mantovano (excerpt D). Primatic-
cio (Bologna 1504 - Paris 1570) is one of the few
artists, together with Titian and Sansovino, still liv-
ing at the time of the print (1568) of Vasari’s book;
G.B. Scultori called G.B. Mantovano (Mantua 1503
- 1575) is recorded as disciple of Giulio Romano
and it is also mentioned in " Lives of Marc’Antonio
Bolognese and other engravers of print" and in
" Lives of Benvenuto Garofalo and Girolamo da
Carpi painters". It is still difficult to attribute the
stucco-work to each craftsman because of the ab-
sence of comparative scientific analyses. The read-
ing of Vasari’s biographies nevertheless introduces
a lack of certainty: in a first quotation (excerpt D)
the author ascribes the double frieze of Camera
degli Stucchi to the work of Primaticcio and Man-
tovano, in a second one (excerpt E) Vasari ascribes
the same frieze to the sole work of Primaticcio.
144
VII. Conclusion
The finish coat of Mannerist stucco-works of three
rooms of Palazzo Te in Mantua shows different fea-
tures: angular quartz crystals and silicate clasts as
aggregate together with a dolomitic lime as binder
(cornices and figures of Camera dei Venti and Cam-
era delle Aquile, upper frieze of Camera degli Stuc-
chi); calcite crystals as aggregate and calcium lime
as binder (lower frieze fo Camera degli Stucchi);
limestone clasts as aggregate and calcium lime (and
also gypsum) as binder (Hercules statue of Camera
degli stucchi).
Raw materials come from outside and they are
not available in the town surroundings: Mantua,
lying in the center of the Northern Italian alluvial
plain (Padania), is far from stone outcrops and from
stone quarries.
The use of quartz as aggregate, instead the cal-
cite, is not reported by Ancient Authors, but it is
documented in plaster finish coats of some Roman
sites of Lombardy.
The use of different aggregate in two similar
stucco-works of the same room (Camera degli
Stucchi) may be explained by the presence of two
craftsmen working at the same time in the room.
This detail is reported by Vasari’s Lives of the Artists
in some excerpts related to the erection and the
decoration of Palazzo Te in Mantua: in particular
Vasari points out the names of F. Primaticcio and
G.B. Mantovano.
Excerpts from Giorgio Vasari’s Lives of the
artists [Devere 1912-15, Maclehose 1907]
Excerpt A - Life of Giulio Romano painter [vol.
VI]:
<Giulio set his hand to the work [Palazzo Te](...).
And since the place [Mantua] has no living rock,
and no quarries from which to excavate material
for hewn and carved stones, (...) he made use
of brick and baked stone, which he afterwards
worked over with stucco; and with this material
he made columns, bases, capitals, cornices, doors,
windows and other things, all with most beautiful
proportions. And he executed the decorations of
the vaults in a new and fantastic manner, with
very handsome compartments, and with richly
adorned recesses. (...) The form of the palace is as
follows: the building is quadrangular, and it has
in the centre an open courtyard (...), two others
[entrances] lead into various apartments, and these
are all adorned with stucco-works and paintings.>
Excerpt B - Introduction chap. IV
<(...) instead of sand the pounded marble is
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 17: Tail of the Hercules’ lion skin.

taken and is sifted finely and kneaded with the
lime.>
Excerpt C - Life of Giovanni da Udine painter
[vol. VIII]:
<(. . . ) it may be necessary to mix with white
lime of travertine, in place of pozzolana, some
substance white in color (. . . ) chips of travertine
(. . . ). But finally, having caused chips of the whitest
marble that could be found to be pounded and
reduced to a fine powder, and then sifted, he mixed
it with white lime of travertine, and discovered
that thus he had secceeded without any doubt in
making the stucco of ancients, with all that he had
desired therein.>
Excerpt D - Life of Giulio Romano painter [vol.
VI]:
<(...) from that room of Psyche one passes into
another [or Camera degli Stucchi] full of doubles
friezes with figures in low-relief, executed in stucco
after the designs of Giulio by Francesco Primaticcio
of Bologna, then a young man, and by Giovan
Battista Mantovano, in which friezes are all the
soldiers that are on Trajan’s column at Rome,
wrought in a beautiful manner.>
Excerpt E - Description of the works of Francesco
Primaticcio (...)[vol. IX]:
<[Primaticcio] going afterwards to Mantua,
where at that time Giulio Romano was working
at Palazzo Te. There, (...) he learned very well (...)
to work in stucco (...). This may be seen in a great
chamber [Camera degli Stucchi], round which he
made two friezes of stucco, one above the other,
with a great abundance of figures that represent the
ancient Roman soldiery (...).>

Figure 18: Thin section, cross polarized light.

145
 Bugini & Folli
Acknowledgement
Photo credits: Fratelli Alinari (Fig. 1); Touring Club
Italiano (Fig. 2); Bugini and Folli (Figs. 3-18).
References
Alberti, L.B. (1989): " De re adificatoria", italian translation by G. Or-
landi. Il Polifilo, Milano (First published 1485; book 6, chap. 9).
Bugini, R., Della Torre, S., Portosa, C., Pracchi, V., Sansonetti, A.,
and Tajana, C. (1998): " La conservazione del Palazzo Natta di Como:
un progetto in cantiere". Proceedings of 14th Convegno Scienza e Beni
Culturali, Bressanone, Italy, June 30 - July 3, 271-281.
Bugini, R., and Folli, L. (2011): " Monomineralic aggregates in mor-
tars: examples from Lombardy". Proceedings of the 10th International
Congress Applied Mineralogy, Trondheim, August 1-5, 85-92.
Bugini, R., Folli, L., and Biondelli D. (2013): " Grain morphology of
aggregates in Roman plasters". Proc. of 14th Euroseminar Microscopy
Applied Building Materials, Helsingor, June 10 - 14, 25-28.
Burckhardt, J. (1987): " The architecture of the Italian Renaissance",
english translation by J. Palmes. Penguin Books, Harmondsworth (First
published 1867; book 2, chap. 7).
D’Aviler, A.C. (1755): " Dictionnaire d’Architecture civile et hy-
draulique". Paris (s.v. Stuc).
Cavallo, G., Derighetti, L., and Napoli, S. (2005): " The emigration of
prealpine lake masters: a study of a 17th-century mantelpiece attributed
to the Carlone family of Rovio (Ticino-Switzerland)". Archaeometry. Vol.
47, No. 2, 425-439.
Davey, N. (1961): " A history of building materials". London.
Folli, L., and Bugini, R. (2001): " Le ricette degli stucchi in Italia set-
tentrionale dal XV al XX secolo". Proceedings of 17th Convegno Scienza
e Beni Culturali, Bressanone, Italy, July 10 - 13, 207-218.
146
Hempel, E. (1965): " Baroque art and Architecture in Central Europe".
London.
Milizia, F. (1785): " Principi di architettura civile". Bassano (Part 3,
book 3, chap. 3).
Murray, L. (1967): " The high Renaissance and Mannerism". Oxford.
Plinius (1962): " Naturalis Historia", english translation by D.E. Eich-
holz. Harward University Press (book 36, chap. 53-55 and 59).
Pracchi, V., and Bugini, R. (2001): " Gli stucchi di G.B. Barberini nella
Sacrestia del Collegio dei Gesuiti a Mantova". Proceedings of 17th Con-
vegno Scienza e Beni Culturali, Bressanone, Italy, July 10 - 13, 141-150.
Rampazzi, L., Rizzo, B., Colombo, C., Conti, C., Realini, M., Bar-
tolucci, U., Colombini, M.P., Spiriti, A., and Facchin, L. (2012): " The
stucco technique of the Magistri Com’acini: the case study of St Maria
dei Ghirli in Campione (Como, Italy)". Archaeometry. Vol. 54, No. 5,
926-939.
Robertson, D.S. (1964): " A handbook of Greek and Roman architec-
ture". Cambridge.
Rondelet, J. (1803): " Traité théorique et pratique de l’art de bâtir".
Paris (Tome 1, book 2, s. 3, art. 12).
Vasari, G. (1568): " Le vite de’ piu eccellenti pittori, scultori, architet-
tori". Giunti, Firenze. English edition: " Lives of the most eminent
painters, sculptors and architects, newly translated by Gaston Duc De-
vere in ten volumes". Ph. Lee Warner Publisher to The Medici Soci-
ety, London, 1912-15". English edition of the " Introduction": " Vasari
on technique being the Introduction to the three arts of design, architec-
ture, sculpture and painting, prefixed to the Lives of the most excellent
painters, sculptors and architects by Giorgio Vasari painter and archi-
tect of Arezzo". Louisa Maclehose translation, J. M. Dent & Company,
London 1907".
Vitruvius (1914): " De Architectura", english translation by M.H. Mor-
gan. Harward University Press (book 7, chap. 6).
Wittkover, R. (1963): " Art and architecture in Italy (1600-1750)". Lon-
don.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Hydraulicity in ancient mortars: its origin

and alteration phenomena under the
microscope
Johannes Webera , Anthony Baragonaa , Farkas Pintérb , Christophe Gosselinc
a Institute of Arts and Technology/ Conservation Sciences, University of the Applied Arts Vienna, Austria
b Scientific Laboratory, Federal Office for the Protection of Monuments, Vienna, Austria
c Geotest SA, Lausanne, Switzerland
⇤ johannes.weber@uni-ak.ac.at

Abstract

Historical hydraulic mortars can comprise different reactive phases involved in the early- and mid-age development of
binding properties such as pozzolana, brick powder or other reactive silica and alumina. However, the specific nature of
the reactive phases in such binders is rarely clearly identified, particularly when different degradation or weathering
patterns, frequently related to the carbonation of the binder, interact with dissolution and precipitation reactions. The
question of which binder and aggregate components have contributed to the hydraulicity of a given historic mortar, a key
to a better understanding of ancient technologies, is frequently not investigated in the case of mortars which underwent
the above mentioned alteration processes. Microscopy provides a powerful analytical technique to determine the type of
binders and differentiate between primary and alteration phenomena.
By the example of three groups of binders, namely Roman opus caementitium/cocciopesto mortars from Ephesus
(Asia Minor) and Vindobona (Roman Vienna), a 16th C. Ottoman horasan masonry mortar from Budapest, and 19th C.
highly hydraulic lime filling mortars from Switzerland, the present article discusses two phenomena frequently observed
when dealing with ancient hydraulic mortars: firstly as to the source(s) of hydraulicity of the mortar and secondly, that
after a long period of time exposed to moist environments, the binder tends to present an inhomogeneous composition of
an impure silica gel separated from calcium carbonate, accompanied by leaching and precipitation of binder constituents.
These two phenomena are illustrated by polarized light microscopy (PLM) and scanning electron microscopy combined
with energy-dispersive x-ray spectroscopy (SEM-EDX) and discussed in light of both conservation and understanding
of ancient structures as well as how it can be applied to the current issues of alkali-aggregate reaction (AAR) and
leaching in modern concretes.

Keywords: historical mortar, hydraulicity, carbonation, leaching, AAR

I. Introduction / Methodology two phenomena is not only useful in the histori-

PLM and SEM-EDX are powerful tools when study-
ing historical mortar samples, especially when
used in conjunction on the same polished thin sec-
tion. This is particularly true when observing the
binder and binder-aggregate interaction. While
PLM is a quick and convenient first step to visu-
alizing reactive areas, characterizing the reactions
and phases involved requires additional instrumen-
tal techniques such as SEM-EDX. This research
presents two of these observations made from com-
bining these techniques as applied to various types
of mortar. One focuses on the potential of siliceous
aggregates generally considered inert to contribute
to the overall hydraulicity of a mortar; the other
on the alteration of highly hydraulic binders, espe-
cially evident in applications where the mortar was
used in highly humid conditions. Studying these
cal or archeological context, but should further the
understanding of such issues as alkali-aggregate
reaction (AAR) in contemporary structures.
The issue of reactive aggregate was observed
across a wide variety of sample types, all of which
additionally show signs of binder alteration linked
to carbonation and leaching. How these phenom-
ena could be linked will be discussed in the con-
clusion. Presented here are a number of mortar
samples from ancient Roman sites, an Ottoman-era
bath from Budapest and a highly hydraulic mortar
from the turn of the 20th century. Cocciopesto mor-
tars (such as those from the Roman era discussed
here) have been well researched in terms of the
interaction at the interface between ceramic frag-
ments and the lime binder; what is presented here
is the observation that the binder itself is highly hy-
draulic throughout, evidence that either the binder

147
 Weber et al.
Figure 1: 90-day old experimental cocciopesto mortar;
beside the larger ceramic fragments, a large amount of
fines are visible as red particles. Incident light, white
background.
has an additional source of hydraulic material, or
that a much larger amount of the aggregate has
reacted than previously observed. Similar high hy-
draulicity has been observed in mortars with little
or no ceramic aggregate and no obvious primary
silica source from the binder itself. The example of
the early 20th century restoration mortar shows that
these phenomena are observable in much younger
mortar, potentially forming a bridge between ob-
servations on ancient mortar and issues seen in
construction today.
Most of the micrographs presented and discussed
in this study were obtained from polished thin sec-
tions produced after vacuum impregnating mortar
samples with blue-dyed epoxy resin. Once a struc-
ture was observed by polarizing light microscopy
(PLM) in the transmitted and incident modes, care
was taken to study the same structure in the scan-
ning electron microscope (SEM) at an acceleration
voltage of 20 kV. In order to avoid irreversible im-
Figure 2: Sample as Figure 1; (a) Interior of brick fragment rich in mica and fine quartz. (b) Binder paste near brick
fragment containing the same mica and quartz as fines embedded in a hydrated binder rich in silica.
148
pairment of the section by coating, most of the sam-
ples were analyzed uncoated under low-vacuum
conditions of the SEM. Since the EDX results were
supposed to be less accurate at low vacuum, quan-
titative data of analysis are to be handled with
care. This does not apply to the data plotted in
the graph of Figure 17, which were measured on
carbon-coated samples at high vacuum at 15 kV.
II. Historical samples discussed
Late Republican/ Early Imperial Roman
mortar from Ephesus, Turkey and Sisak,
Croatia
One of the largest cities in the Eastern Mediter-
ranean in ancient times, Ephesus was the capital
of the province of Asia during the Roman Empire.
Designed for approx. 25,000 spectators, the Roman
theater was constructed and enlarged in several
steps between the 1st and the 4th century A.D. The
samples presented in this study come from various
construction phases and parts of the building and
comprise both opus caementitium and masonry
mortars. The composition of these mortars did not
show any significant difference in dependence to
phase or mode of application. Additionally ana-
lyzed is a masonry mortar from Sisak (Siscia), Croa-
tia, an important center of Southern Pannonia in
Roman times. All mortars faced similar conditions
of exposure in the interior of ancient walls which
were later on buried until their excavation.
Early Imperial Roman mortar from Vi-
enna, Austria
The Vienna Basin was controlled by the Roman Em-
pire from roughly the turn of the first century until
the early 5th century A.D. as part of the province of
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Ancient cocciopesto mortar from Vindobona. (a) Observed by PLM at incident light in the dark field mode,
the brick fragment seems unreacted.(b) The bright field appearance of the same fragment evidences its high degree
of dissolution.(c) The same fragment by SEM-BSE confirms that a significant amount of the ceramic fragment has
undergone reaction; while quartz and mica have remained in the otherwise hydrated core part of the fragment, only
quartz is present in the surrounding hydrate matrix where mica has apparently reacted away. The binder is altered to
a silica-rich gel with popcorn like tiny crystals of calcium carbonate. Patchy compaction, very evident in the PLM
micrographs by bright colors, is due to precipitation of a Mg-rich carbonate phase.

Pannonia. A legionary fortress called Vindobona,
surrounded by a civilian settlement, was built at
the current site of the innermost district of Vienna.
To supply this city with fresh water, an aqueduct
was built leading from the nearby hills. The sample
discussed is of the cocciopesto lining of this aque-
duct and includes a thick layer of accreted calcium
carbonate from centuries of use. In the course of
studying this mortar, experimental mortar of simi-
lar composition was made for comparison. Mortar
was prepared in the proportion of 1 part lime to 3
parts aggregate (based on Vitruvius), with a portion
of the aggregate being ceramic, and the other por-
tion being siliceous (based on observed historical
mortar compositions).
16th C. Ottoman mortar from Budapest,
Hungary
Between 1526 and 1686, part of the Hungarian King-
dom was occupied by the Ottoman Empire. During
this time, many buildings were constructed, includ-
ing thermal baths. One of the largest Ottoman baths
in Budapest, the Császár (Emperor) Bath, was built
in 1574 by Sokollu Mustafa and has been rebuilt
several times in the following centuries. During
renovation-restoration activities of the bath, several
samples have been taken in order to support art
historical and scientific research (Pintér et al, 2011).
The sample presented in this study is a lime-brick
horasan bedding mortar of a water pipeline orig-
inating from the western part of the central bath
building.
19th C. Restoration mortar of a Swiss Ro-
man site
Aventicum, the capital of the Helvetians, was built
in the early first century A.D. It grew considerably
during the reign of Emperor Vespasian, who in 72
AD gave it the rank of colony. It was during this
period that the imposing city wall (5.5 km) was
built. The East Gate of this wall dominated the
city and indicated the general direction of a transit
route through several roads delimitated by adjacent
walls. Almost completely razed to the ground, the
gate and the walls were partly rebuilt during the
restoration campaigns of the East Gate and the For-
tification Wall (1910-1940), under the direction of
the archaeologist Louis Bosset. The original Roman
wall core was conserved and protected by a new
wall over two meters high. The present sample has
been collected from the filling joints of internal lime-
stone rubbles of the wall n 8, the external masonry
being composed of limestone joined ashlar.

Cocciopesto/horasan mortars
Figure 4: Apparently calcium-rich area within brick The ability of ceramic fragments or powder to im-
(framed); Vindobona Aqueduct. Transmitted light, dark prove the strength of a mortar, generically known
field. as cocciopesto (called horasan in the case of Ot-

149
 Weber et al.

Figure 5: (a) SEM-BSE of the framed area of Figure 4: All dark grey areas are supposed to be hydrated. (b) Framed
area from Figure 5a: C2 S with hydration rim.

toman mortars), is well known as a result of being
recommended by Vitruvius. A number of recent pa-
pers have contributed to the understanding of the
mechanisms involved, confirming the good mechan-
ical performance and durability of these mortars
[e.g. (Baronia & Binda, 1996); (Baronio et al, 1996);
(Binda et al, 1999); (Böke et al, 2006); (Moropoulou
et al, 2002); (Silva et al, 2007); (Nezerka et al, 2015)].
The strength enhancing reaction at the ceramic-lime
interface is easily observable through microscopy.
It is the contention of this paper that it can be
assumed that the same reaction is taking place on
the ceramic fines within the binder matrix, indeed
to greater extent due to their greater surface area in
contact with lime. Over time, these smaller particles
would mainly become invisible, leaving only ves-
tiges such as iron oxide, quartz or mica as the clay
minerals have reacted (Figures 1,2), interspersed
within a highly altered binder of silica gel and com-
pacted calcium carbonate (discussed below). As is
well known and stated by numerous authors, the
main reactive portion of a brick consists of dehy-
droxilated clay (Cultrone at al., 2001). When these
residuals are found within the matrix, there is a ten-
dency to oversee them or assume that they are part
of the additional aggregate or the original binder,
perhaps leading to an inaccurate assessment of the
binder-aggregate ratio. These unreacted trace min-
erals leave a fingerprint of former ceramic aggre-
gates, such as reddish areas visible in PLM (Figures
3 and 4).
However, there exist other potentially reactive
constituents including calcium silicates formed dur-
ing the firing of lime-rich bricks as well as other
reactive silicate minerals included in the temper
that may contribute to the overall hydraulicity of a
mortar (Figures 4-6). The presence of partially hy-
drated silicates completely enclosed within a brick
fragment lends credence to the idea that such phe-
nomena would likely occur as well in the matrix.
These cases exhibit the potential for additional ob-
servable hydraulicity in cocciopesto mortar beyond

Figure 6: SEM-BSE of the interior of a brick fragment showing reactive albite with silica-rich hydration rim, Ephesus
Theater.

150
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 7: Ancient highly hydraulic masonry mortar with almost no ceramic aggregate from Ephesus Theater. Most of
the aggregate is metamorphic, with frequent fragments of albite gneiss.(a) Highly altered plagioclase crystal, PLM,
transmitted plane polarized light.(b) Incident light, bright field mode.The brown (a)/bright (b) areas in a and b are
binder compacted by calcium carbonate precipitated in the course of carbonation.(c) SEM-BSE shows a thick rim of
alteration with hydration to a silica-rich gel, preferentially along cleavage planes. EDX shows albite content of about
75 % in the remaining plagioclase.

that which can be explained by the pozzolanic reac- identifiable by a similar reaction ring that can be
tion between lime and ceramic fragments. found around a pozzolan (Figures 7-11). As is
the case with ceramic aggregate, over time only
the larger aggregates would be visible, while the
Other highly hydraulic ancient mortars finer portion could be easily overseen or would be
completely reacted away. While not commonly dis-
There exist instances of ancient mortars in which
cussed in literature on historical mortars, a similar
neither brick powder nor volcanic pozzolans are
phenomenon is discussed in works on alkali-silicate
present in sufficient quantity to explain the observ-
reaction (ASR), a type of alkali-aggregate reaction
able high hydraulicity, nor infer that they were in-
in OPC based concrete. The conditions known to fa-
tentionally added for this purpose. One possibil-
vor ASR are lattice defects and/or low crystallinity
ity is that a hydraulic lime was used, however in
in SiO2 , high humidity over prolonged periods of
this case generally remnants of the lime source can
time, sufficiently high pH and alkali content. At
be found, for example marly lime lumps or un-
least some of these conditions can be present in
hydrated calcium silicates. The other possibility
historical mortars, especially when used in appli-
rarely discussed is that of reactive aggregates. Un-
cations that expose them to high humidity such as
der certain conditions, plagioclase, feldspar, chert
masonry filling mortars - e.g. opus caementitium,
and potentially even quartz can react, releasing
hypocaust floors etc. Other observations support-
silicate into the binder and contributing to the over-
ing the idea that the conditions were in favor of
all hydraulicity (Chou & Wollast, 1985; Hodgkin-
the hydration of otherwise stable silicate minerals
son & Hughes, 1999; Broekmans, 2004). These are

Figure 8: A similarly ancient highly hydraulic masonry mortar with almost no ceramic aggregate from Ephesus
Theater. Again, the aggregate is of metamorphic origin, with frequent fragments of albite gneiss.(a) Aggregate fragment
of plagioclase (plag) and quartz (qz); plagioclase appears slightly corroded at margins and along cleavage; incident
light, bright field mode.(b) SEM-BSE of the same fragment. While plagioclase shows a thin layer of a silica-rich gel,
quartz appears largely unreacted.

151
 Weber et al.

Figure 9: Ancient highly hydraulic masonry mortar from Siscia. Much of the aggregate in this mortar is chert and
siltstone with microcrystalline quartz. Aggregate fragment of chert (center) with rim hydrated to silica gel. Also the
margin of the big siltstone partly visible on the right appears hydrated. The surrounding binder matrix is very rich in
silica, with Ca:Si = 1:2. Calcium carbonate is enriched along the cracks. (a) PLM, plane polarized light. (b) Crossed
polarized light. (c) Incident light, bright field. (d) SEM-BSE.

are the eventual presence of dolomitic aggregates
undergoing alteration to brucite and calcium car-
bonate due to dedolomitization (Figures 12 and 13),
a process described e.g. by Katayama (2010) and
Mittermayr et al. (2011), indicating elevated pH.
An explanation for alkaline conditions that would
favor dissolution of feldspar over a long period of
time are a slow rate of carbonation due to limited
exposure to the atmosphere in the case of the men-
tioned types of application. The condition of high
humidity over prolonged periods of time is also
evidenced by strong alteration of the binder matrix.
It is likely that under such conditions even quartz,
especially the metamorphic quartzes generally ob-
served in all of the mortars presented here, may
be subject to dissolution at low rates. (Broekmans,
2004).

Figure 10: Altered alkali feldspar grain with residues of the original mineral exhibiting birefringence and surrounded
by porous reaction products. The original shape of the mineral has entirely preserved. The altered matrix is composed
up of an impure silica gel and finely dispersed CaCO3 . Ottoman mortar, thin section. (a) PLM, crossed polars. (b)
SEM-BSE.

152
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 11: Ancient highly hydraulic opus caementitium mortar with just occasional ceramic additions from Ephesus
Theater. Partly corroded polycrystalline quartz aggregate (center) next to a marginally altered dolomite aggregate
(right margin).(a) PLM, plane polarized light.(b) Crossed polars.(c) SEM-BSE.

Alteration of highly hydraulic binder
The carbonation of hydrates constituting the matrix
of hydraulic binders is a well-known fact attested
to by a large amount of studies such as Goto et al.
(1995). Much knowledge comes from research on
concrete and modern cementitious materials; how-
ever, it is naturally applicable to hydraulic binders
in historical mortar. Binder alteration linked to car-
bonation is reflected by a number of interesting
phenomena observable in the microstructure, in
particular when using appropriate modes of illumi-
nation under an optical microscope, in particular
dark-field mode. The areas thus discovered can
then be chemically analyzed with EDX, but it is im-
portant to remember that because of inhomogeneity,
it is not possible to access the quantitative composi-
tion of the binder in this way. The process of carbon-
ation is system-wide: not only is calcium hydroxide
converted to calcium carbonate, but calcium silicate
hydrates are likewise converted to calcium carbon-
ate and a silica-rich gel. In cement mortars, the
depletion in calcium, which forms compact rims
of carbonate around the clinker, drastically alters
the composition of calcium silicates in unhydrated
clinker residues [e.g. Shtepenko et al. (2006), Gos-
selin et al. (2012)]. When carbonation is accompa-
nied or followed by the circulation of water through
the pore system of the binder, calcium carbonate
can get dissolved and re-precipitated in specific
places. When these conditions are met, this may
lead either to "popcorn" like crystals of calcium car-
bonate growing within a silica-rich matrix (Figures
14a and b), often combined with re-crystallization
along the margin of pores and cracks, or to the
development of diffuse binder areas compacted by
submicroscopic carbonate crystals. This effect is a
clear evidence for degraded C-S-H due to strong
leaching caused by percolating water. Additionally
this can also be observed in more modern materi-
als such as an early 20th Century highly hydraulic
lime mortar (Figure 15). In this example a belite
clinker nodule has separated into silicate and car-
bonate rich areas. The described phenomena are
well-known for modern concretes. It is important
to reiterate that as in the question of additional
reactive aggregates, this phenomenon is particu-
larly evident in building applications that featured
a large amount of moisture or humidity.
From published XRD data we know that this
secondary calcium carbonate, intermixed with de-
calcified C-S-H, frequently occurs as aragonite or
vaterite (Elsen et al, 2012). In either case is the
remaining binder, a chemically impure silica gel,
strongly depleted of calcium. Though the described
inhomogeneity makes it impossible to assess the
original binder composition by EDX or other meth-

Figure 12: Ancient highly hydraulic opus caementitium mortar from Ephesus Theater. Dolomite aggregate with rim
of conversion to a Mg-phase, presumably brucite. The surrounding binder is highly altered by carbonation.(a) PLM,
transmitted light, crossed polars.(b) Incident light, bright field.(c) SEM-BSE.

153
 Weber et al.

Figure 13: Ancient highly hydraulic opus caementitium mortar from Ephesus Theater. Rim of a dolomite aggregate
converted to brucite and calcium carbonate (bright areas). The surrounding binder is an impure silica gel; SEM-BSE.

ods related to microscopy (Elsen et al, 2012), it
is nevertheless possible to recognize its hydraulic
nature and then search for potential sources of hy-
draulicity (Figures 16 and 17). It is worth mention-
ing in this context that the described process of
alteration is usually resulting in increased strength
and durability of the historical mortars affected.
Observations of the microstructure by microscopy
and SEM generally support this fact: in contrast to
the destructive effect of gel formation by AAR in
modern concrete, silica gel formed in the course
of carbonation of ancient hydraulic binders seems
to form a fairly porous structure, chemically stable
and not subject to volume expansion. The pre-
cipitation of carbonate on the other hand, though
creating inhomogeneous patterns of compaction
as described above, is yielding additional strength
since it is likely to crystallize in weak and porous
zones of the mortar, thus contributing to the so-
called self-healing effect, a term frequently cited
as advantageous of lime mortars (e.g. Borges et
al., 2014), even if their inhomogeneity is enhanced
which can cause the formation of weaker binder
depleted areas In this way, the adhesion between
binder and aggregates is usually strengthened due
to preferential carbonate crystallization in fissures
and voids between these two systems. In the case
of porous aggregates such as brick fragments, the
precipitation of calcium carbonate affects also the
marginal zones of the aggregate, binding it tightly
to the matrix (Nežerka et al, 2015). The latter effect
yields a zoning of the brick often visible even by
the naked eye, and can be easily misinterpreted as
a reaction rim of hydrates. Both phenomena, how-
ever, are likely to improve the mortar strength in a
similar way.
III. Conclusion
The wide variety of mortar samples discussed here
share two common traits: that they have additional,
unexpected observable hydraulicity not attributable
to silicate phase in the raw feed (e.g. hydraulic
lime) and that this hydraulic binder has drastically
altered over time, as manifest by a silica-rich gel
with clustered carbonate areas. These observations

Figure 14: Horasan mortar from the Emperor‘s Bath, Budapest. Both PLM (a) and SEM-BSE (b) clearly show CaCO3
"popcorn". EDX confirms that the binder is composed of this and impure and amorphous silica gel. White cracks are
filled with re-crystallized CaCO3 . Small red fragments are remnants of reacted ceramic particles added as a pozzolan.

154
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 15: (a) Alteration of belite clinker remnant used to repair the Swiss Roman site of Aventicum(b) Image showing
silica gel interspersed within matrix near clinker. SEM-BSE.

were not the initial goal of analysis of any of the
samples; rather the similarities were discovered in-
dependently. Hydraulicity in each of the historical
samples was enhanced by reactive aggregates, tem-
pered and untempered, ceramic or natural. Despite
these similarities, and the similarities of the silica
gel and carbonate binder separation phenomenon,
one must keep in mind that the rate of these reac-
tions are not necessarily the same across all samples.
One thing that all ancient mortars had in common
was high exposure to humidity and probably a high
pH over prolonged periods of time.
As a matter of fact, both phenomena were ob-
served in the same mortar samples, because the
extensive binder alteration of the described type
can only occur in hydraulic mortar. Frequently
found in ancient renders and plasters, the effect
of binder dissolution and precipitation in air lime
mortars is known to result in the re-distribution of
the carbonate binder where, however, no dissolu-
tion of silicate aggregate is observed. This points
to the necessity of a prolonged alkalinity in the ser-
vice life of a mortar, a condition enhanced by high
humidity as is the case for filling mortar and water
pipes, water buildings, etc.
It is evident that feldspars would form alkalis
upon their dissolution, thus adding a source of high
pH at least in their immediate surroundings. In this
way, a self-sustaining cycle of feldspar dissolution
could be assumed. It would continue to operate un-
til the alkalis get washed out and leave the system,
which would allow carbonation. Whether or not
this proposed mechanism can have played a role
for the whole bulk of the mortar binder is under
question at this stage of research.
This work represents a preliminary set of ob-
servations regarding important chemical processes
and how they apply to the durability of ancient

Figure 16: Aventicum mortar, morphology of the cement grains (original quartz core and C2 S rim) showing inner and
outer hydration products discriminated by a hollow shell.

155
 Weber et al.
Figure 17: Composition of the inner and outer hydration
products in the Ottoman and Aventicum mortars (cf.
Figures 10a, b and 16).
mortar. The lessons that can be learned from this
study and how they can be applied to the issues
of conservation as well as to contemporary build-
ings require further clarification. Microscopy is an
important starting point that should be followed
by the use of other analytical techniques such as
Raman microscopy. Additional studies on exper-
imental mortars to establish e.g. the potential of
feldspar dissolution to create sufficient alkaline con-
ditions would include analyses of pore solutions as
well as of alkali and feldspar contents, respectively.
Acknowledgement
Thanks are due to the following colleagues for pro-
viding archaeological samples: M. Kronberger from
the Wien Museum (Vindobona mortars), T. Lolic
from the Ministry of Culture/Conservation Depart-
ment in Zagreb (Siscia mortars), G. Styhler-Aydin
from the Vienna University of Technology (Ephesus
Theater mortars), J. Lászay from the former State
Center for Restoration and Conservation of Historic
Monuments, Budapest (Emperor’s bath sample),
N. Terrapon from the Laboratory of Conservation-
Restoration of the Roman Museum in Avenches
(Aventicum samples).
156
References
Baronio G. and Binda, L. (1996): " Study of the pozzolanicity of some
bricks and clays" Construction and Building Materials. Vol. 11, No.1, 41-46.
Baronio, G., Binda, L., and Lombardini, N.: (1997): " The role of brick
pebbles and dust in conglomerates based on hydrated lime and crushed
bricks."Construction and Building Materials. Vol. 11(1), 33-40.
Borges, C., Silva, A.S. and Veiga,R. (2014): "Durability of ancient lime
mortars in humid environment." Construction and Building Materials. 66,
606-620.
Böke, H., Akkurt, S., Ipekoglu, B. and Ugurlu, E. (2006):
" Characteristics of brick used as aggregate in historic brick-lime mortars
and plasters". Cement and Concrete Research. Vol. 36, 1115-1122.
Broekmans, M. A.T.M (2004): " Structural properties of quartz and
their potential role for ASR". Materials Charecterization 53,129-140.
Chou, L. and Wollast, R. (1985): " Steady state kinetics and dissolution
mechanisms of albite". American Journal of Sciene, 285, 963-993.
Cultrone, G., Rodriguez-Navarro, C., Sebastian, E., Cazalla, O. and
De La Torre, M. J. (2001): " Carbonate and silicate phase reactions during
ceramic firing". European Journal of Mineralogy. 13(3), 621-634.
Elsen, J., Van Balen, K. and Mertens G. (2012): "Hydraulicity in his-
toric lime mortars: a review". Historic Mortars. Springer Netherlands,
125-139.
Gosselin, C., Girardet, F. and Feldman, S. B. (2012): "Compatibility of
Roman cement mortars with gypsum stones and anhydrite mortars: The
example of Valere Castle (Sion, Switzerland)". Proc. 12th International
Congress on the Deterioration and Conservation of Stone, New York, 22-25
October 2012 (in press).
Goto, S., Suenaga, K., Kado, T. and Fukuhara, M. (1995): " Calcium
Silicate Carbonation Products". Journal of the American Ceramic Society, 78,
2867-2872
Hodgkinson, E. and Hughes, C. (1999): " The mineralogy and geo-
chemistry of cement/rock reactions: high-resolution studies of experi-
mental and analogue materials". Geological Society, London, Special Publi-
cations. 157.1, 195-211.
Katayama, T. (2010): " The so-called alkali-carbonate reaction (ACR) -
Its mineralogical and geochemical details, with special reference to ASR".
Cement and Concrete Research. 40(4). 643-675.
Mittermayr, F., Klammer, D., Köhler, S., Leis, A., Höllen, D. and Diet-
zel, M. (2011): " Dissolution of Dolomite in alkaline cementitious media".
Proceedings 13th ICCC, Madrid, 3-8 July 2011, 278: 1-6.
Moropoulou, A., Cakmak, A. S., Biscontin, G., Bakolas, A. and Zen-
dri, E. (2002): " Advanced Byzantine cement based composites resisting
earthquake stresses: the crushed brick-lime mortars of Justinian’s Hagia
Sophia". Construction and Building Materials.Vol. 16, 543-552.
Nežerka, V., Nìmeéek, J., Slížková, Z. and Tesárek, P. (2015):
" Investigation of crushed brick-matrix interface in lime-based ancient
mortar by microscopy and nanoindentation". Cement and Concrete Com-
posites. 55, 122-128.
Pintér, F., Weber, J., Bajnóczi, B. and Tóth, M. (2011): " Brick-Lime Mor-
tars and Plasters of a Sixteenth Century Ottoman Bath from Budapest,
Hungary". Turbanti-Memmi, I. (Ed.): Proceedings of the 37th International
Symposium on Archaeometry, 12th-16th May 2008, Siena, Italy. Springer
Berlin Heidelberg, 293-298.
Shtepenko, O., Hills, C., Brough, A. and Thomas, M. (2006): " The
effect of carbon dioxide on ?-dicalcium silicate and Portland cement".
Chemical Engineering Journal. 118(1), 107-118.
Silva, J., de Brito, J. and Veiga, R. (2009): " Incorporation of fine ce-
ramics in mortars". Construction and Building Materials. 23(1), 556-564.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Morphology of self-healing products in

NHL and lime-pozzolan mortars
Maria Amenta⇤,a , Ioannis Karatasiosa , Pagona Maravelakib , Vassilis Kilikogloua
a Institute of Nanoscience and Nanotechnology, N.C.S.R. Demokritos, Athens, Greece.
b Laboratory of Analytical and Environmental Chemistry, Technical University of Crete, Chania, Greece.
⇤ m.amenta@inn.demokritos.gr

Abstract

Mortars are known for their ability to heal their defects in an autogenic way. This phenomenon is expressed by the
filling of micro-cracks by secondary products, restoring or enhancing the material’s performance. The reactions that
take place inside the cracks can be either carbonation or hydration, producing secondary calcite or hydraulic phases
respectively. Overall, the "healing" potential of a cementitious material is largely governed by the binder composition,
as well as the morphology, chemistry and mineralogy of the healing products. As it is an autogenic phenomenon, the
main interest lies in a deeper understanding of its mechanism, along with the parameters that may promote or hinder it.
In the present study the autogenic self healing phenomenon was studied in lime-pozzolan and natural hydraulic
(NHL) mortars, considering aspects of binder composition, curing time before initial cracking, duration and conditions
of the healing period. In order to parameterize the phenomenon, prismatic specimens of both compositions were cracked
after 15 and 30 days of curing and were then examined after another 30 or 90 days of curing at different environmental
conditions(45, 75, 95 RH %) or after continuous wet-dry cycles. The identification of the secondary formed products was
carried out on thin-sections and fractured surfaces, in Petrographic and Scanning Electron Microscope equipped with
Energy Dispersive X-rays Analyser (SEM/EDAX), based on their optical properties, micro-morphology and chemical
composition of the new phases. Additionally, thermo-gravimetric analysis (DTA/TG) was performed in all specimens in
order to quantitatively assess the available unreacted components in the binder at all ages.
The main products formed during healing consist of calcite, aragonite and various C-S-H/ C-A-H phases. Depending
on the parameters mentioned above, there is wide diversity in the intensity, typology and topography of the secondary
phases inside the cracks. The main differences discussed, were observed between specimens cracked at very early age (7
days) and those damaged after 30 days of curing. These differences were correlated with the available each-time amount
of lime, determined by thermo-gravimetric analysis, which found to be related to the formation of carbonate or hydraulic
phases. Regarding the humidity conditions, it was observed that secondary phases may also be formed even at the lower
humidity levels, while prolonged healing period doesn’t necessarily lead to higher healing efficiency.

Keywords: self-healing, NHL, lime-pozzolan mortars, petrography, electron microscopy

I. Introduction hydration products, which derive either from the

Lime-based mortars are extremely susceptible to
crack generation since their first days of applica-
tion. These cracks, even not macroscopically visible,
could lead to further degradation and decreased
durability, as micro-cracks make the structure more
vulnerable to both mechanical stress and environ-
mental actions. Therefore, the prolonged durability
of historic lime-based mortars applied centuries
ago, is often attributed to their self-healing capac-
ity (Nijland, 2007). This phenomenon is autogenic,
which means that these materials have the ability
to heal micro-defects in their microstructure by fill-
ing the cracks with their own generic materials. A
simplified description of the phenomenon involves
carbonation and hydration mechanisms (Edvarsen,
1999); cracks are filled with calcium carbonate or
carbonation of calcium hydroxide or by further re-
action of non-hydrated phases inside the mortar
matrix respectively. The potential of self-healing is
a favorable function of modern building materials,
since they could greatly enhance the service-life
and durability of constructions and thus, reduce
the need for intervention or restoration of struc-
tures (Breugel, 2012). Although, in order to pro-
duce engineered self-healing materials, a deeper
understanding of the phenomenon is imperative.
For instance, the presence of water, which is a
crucial parameter of the self-healing phenomenon
(Hearn, 1998) (Yang, 2009), varies in the natural
environment of structures. It could be present in
excess e.g. in underwater structures, or at different
RH levels as atmospheric humidity or through cap-
illary actions. An additional parameter that should

157
 Amenta et al.

Table 1: Parameters tested in LP and NHL mortars.

Group Code Binder Cracking Age (d) Healing/ curing conditions Healing/ curing period (d)
NHL73045 NHL 7 45%RH 30
NHL73075 NHL 7 75%RH 30
NHL73095 NHL 7 95%RH 30
NHL730cy NHL 7 Wet-dry cycles 30
NHL79045 NHL 7 45%RH 90
NHL79075 NHL 7 75%RH 90
NHL79095 NHL 7 95%RH 90
NHL790cy NHL 7 Wet-dry cycles 90
NHL303045 NHL 30 45%RH 30
NHL303075 NHL 30 75%RH 30
NHL303095 NHL 30 95%RH 30
NHL3030cy NHL 30 Wet-dry cycles 30
NHL309045 NHL 30 45%RH 90
NHL309075 NHL 30 75%RH 90
NHL309095 NHL 30 95%RH 90
NHL3090cy NHL 30 Wet-dry cycles 90
LP153045 LP 15 45%RH 30
LP153075 LP 15 75%RH 30
LP153095 LP 15 95%RH 30
LP1530cy LP 15 Wet-dry cycles 30
LP159045 LP 15 45%RH 90
LP159075 LP 15 75%RH 90
LP159095 LP 15 95%RH 90
LP1590cy LP 15 Wet-dry cycles 90
LP303045 LP 30 45%RH 30
LP303075 LP 30 75%RH 30
LP303095 LP 30 95%RH 30
LP3030cy LP 30 Wet-dry cycles 30
LP309045 LP 30 45%RH 90
LP309075 LP 30 75%RH 90
LP309095 LP 30 95%RH 90
LP3090cy LP 30 Wet-dry cycles 90

not be neglected is time, especially in lime-based
mortars where non-reacted components may be
present for months or years after the initial applica-
tion of the mortar. Today, it is not yet entirely un-
derstood whether this phenomenon can take place
only in the first stage of a mortar life (Schlangen,
2006) when there still are unreacted components,
or whether it is possible to be the result of sec-
ondary dissolution and re-precipitation of hydra-
tion or carbonation products. Another question is
related to the age of the specimen when cracking
occurs (Hearn, 1997). As Ca-bearing phases exhibit
a range of solubility values (depending always on
the pH and temperature of the pore solution), it is
expected that self-healing ability of a mixture will
also depend on the stage of carbonation and/or
hydration progress in the binder matrix. During
carbonation, portlandite is transformed into calcite
which is a much less soluble mineral. In hydraulic
binders the produced C-S-H and C-A-H phases
are also susceptible to dissolution but, as they are
less soluble than portlandite, the dissociation of Ca
begins after the dissolution of the latter (Adenot,
1991). With respect to the theoretical assumption
that chemical reactions take place inside the crack
space described earlier, an important parameter to
be studied is the availability of the chemical species
involved. More specifically the concentration of cal-
cium ions in the pore solution should be a decisive
factor since the carbonate content is not a limiting
factor (Edvardsen, 1999). Water could contain a
small concentration of calcium ions but in order
saturation to be achieved the main source of cal-
cium would be the calcium-baring components of
the binder. In the study of Huang and Ye (Huang,
2014) was found that the mineralogy of the reac-
tion products of self-healing induced with saturated
Ca(OH )2 solution is similar to that caused by dis-
tilled water, but the efficiency was improved. This
suggests that calcium ions play the most significant
role on the evolution of the autogenic phenomenon.
Therefore, the characterization of the healing
products is of prime importance as it constitutes the
first step towards the comprehension of the auto-
genous phenomenon. Calcium carbonate has been
identified as the main secondary product filling
the crack by several studies as well as the presence
of portlandite (Lauer, 1956) (Lubelli, 2011). At the
same time, hydration has also been reported as a
self-healing mechanism with ettringite (Jacobsen,
1995). It is clear that the morphology of secondary
products has not been sufficiently studied and cor-

158
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 1: Sample preparation for SEM and polarizing light microscope. One specimen was split in two to reveal the
crack surface (top). Epoxy impregnated specimens were cut vertically to expose a cross-section of the crack (bottom).
related with factors affecting it.
Aiming to contribute on the systematic attempt
to parameterize the autogenic phenomenon of lime-
based mortars, this study, mainly focuses on the
microscopic examination for assessing the morphol-
ogy of the secondary formed self-healing prod-
ucts. To this end, lime-pozzolan and natural hy-
draulic lime mixtures were deliberately damaged
and cured under varying climatic conditions in
order to identify the secondary products that are
formed under different curing conditions.
II. Materials and methods
The two different mortars mixes prepared and stud-
ied (NHL and lime-pozzolan) are considered typi-
cal traditional materials, similar to those used from
the Roman period to the end of the last century.
Nowadays, they are mostly used in restoration
interventions because of their compatibility with
the original historic materials. The raw materi-
als used for this study were a commercially avail-
able lime powder (KRINOS SA) mixed with a nat-
Figure 2: Normalized Ca(OH )2 amount (wt%) in NHL
and LP mixtures, determined by DTA/TG.
ural pozzolan of volcanic origin (Dalkafoukis) in
a 1:1 ratio (LP-mixtures) and NHL5 by St.Astier
(NHL-mixtures). In both mixes siliceous standard
sand (EN196-1) was added as aggregate with a ce-
ment/aggregate ratio of 1:3 w/w.
Prismatic specimens (15⇥15⇥80 mm) of each
mortar mixture (n=48) were cast and cured in con-
trolled conditions (95% RH, 20 C). In order to
study the healing mechanisms, the specimens were
damaged by controlled cracking, under three point
bending. A continuous load (100 µm/min) was ap-
plied to all specimens, allowing the deformation of
specimens up to 50 µm after crack initiation. The
characteristics of the cracks produced, was deter-
mined by means of optical microscopy. The spec-
imens were then divided in 16 groups of 3. In
each group one parameter was examined (age of
the specimens at the time of the crack formation,
conditions of healing and healing period), as listed
in Table 1. The composition of all mixtures was
assessed after varying curing periods with thermal
analysis (DTA/TG) by a Perkin Elmer Pyris G2.
Figure 3: Chemical analysis of the secondary products
performed by EDAX revealed the two main groups
formed in the examined specimens.
159
 Amenta et al.

Figure 4: Typical patterns of calcium carbonate formations of varying size and morphology. The elemental composition
of the products was identified by EDAX.

The amount of portlandite (wt%) was monitored 45±55%RH). These specimens were examined af-
and correlated with the age of the specimens, when ter 10 and 30 cycles (30 and 90 days respectively).
cracking was introduced. Furthermore, three humidity levels were selected
in order to understand the effect of relative humid-
Healing Conditions ity on the self-healing mechanism. Lower humidity
(45%RH) could refer in an indoor environment or in
Cracked specimens were cured in four different en-
an outdoor environment during the summer when
vironments, in order to simulate varying outdoor
long and relatively dry periods take place. Higher
conditions. One group was subjected to wet-dry cy-
humidity levels are met often in exterior masonry
cles, which refer to an outdoor environment where
in areas not directly subjected to rainfall but could
masonry is subjected to rainy days followed by a
also refer to indoor masonry that is not heated.
period of drying. The specimens were immersing
in water for 24 hours and letting them afterwards to
dry for two days in laboratory conditions (25±5 C,

Figure 5: "Needle-like" hydration products, developing mainly inside cracks.

160
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 6: L1490cy-micro crack filled completely with hydration products after thirty wet-dry cycles (left). LP3030cy
hydration products formed after ten wet-dry cycles (right).

Sample preparation precipitates or reaction products in great detail. Ad-

At the end of each healing period the speci-
mens were used for the examination of the sec-
ondary products formed inside the cracks, in a
FEI Quanta-Inspect Scanning Electron Microscope
coupled with Energy-dispersive X-ray spectroscopy
(SEM/EDAX). Microscopy is one of the most widely
used techniques for the characterization of cementi-
tious materials and its proven to be a useful tool in
the assessment of both damage degree and healing
capacity of these materials (Çopuroğlu, 2013). Vari-
ous microscopy techniques provide insight concern-
ing different aspects of the problem. The examina-
tion of thin sections at the petrographic microscope
is the main technique for the characterization of
primary and secondary reaction products (Lubelli,
2012), as well as for the quantification of the sec-
ondary phases inside the crack. Scanning Electron
microscopy provides the option of examining frac-
tured surfaces of samples due to its ability to obtain
large depth of field thus making it possible to ex-
amine the crack surface and characterize secondary
ditionally, when SEM is combined with EDAX it
may be proved useful in the characterization of dis-
tinct points of a polished section by quantitative
chemical analysis. The use of the above techniques,
especially when they are combined, constitutes an
important asset in the study of the mechanism of
self-healing.
One specimen was used for the characterization
of the secondary products morphology by exami-
nation of the crack surface (Figure 1). In order to
obtain supplementary information regarding the
topography of the secondary products inside the
cracks, a cross-section of the crack was examined.
For this reason two specimens were impregnated
under vacuum in epoxy resin to produce thin and
polished sections of the crack plane (Figure 1).
III. Results and discussion
The hardening of NHL and lime-pozzolan mortars
is basically controlled by two main mechanisms,

Figure 7: "Bee-hive" formations, commonly found in NHL and LP (left) mortars. These gel-like products are in all
cases encountered in flat plains, e.g. in the area that was between binder and aggregate. Higher magnification reveals
its gel-like morphology (right).

161
 Amenta et al.

M. Amenta et al.

Concerning the effect of humidity level, in samples cured at 45%RH (regardless of their composition,
age of cracking and healing period), no secondary products were observed. In the specimens cured at 75%RH,
the secondary products were scarce in NHL mortars, whereas LP mortars revealed an abundance of both
crystalline and gel-like secondary phases (Fig.8 left). This was also confirmed during the examination of the
polished surfaces (Fig.8 right). The above findings highlight the critical role of water and aqueous solutions in
the hydration of hydraulic binders and the development of the healing mechanism. The presence of water is
crucial for the development of carbonation and hydration mechanisms (Cizer, 2012) (Karatasios, 2012).
Optimum conditions for the carbonation reaction are met in high humidity levels, whereas when pore space is
completely
Figure dry or fully saturated,
8: Microphotographs carbonation
of LP303075. is inhibited
Hydration (Moorehead,
products 1986).
formed along a smaller perpendicular crack (left) and a
micro crack completely filled by calcite (right).

hydration and carbonation. The concentration of which seem to be richer in aluminum. In contrast,
portlandite in the binder can be used as an indica- secondary products in LP specimens exhibited a
tor of the progression of these mechanisms, as in very narrow distribution in regard to Al/Si ratio,
both mixtures it is the main reactant. Portlandite whereas a clear separation as to Ca/Si ratio (crystal
was determined quantitatively in samples cured at vs gel-like products).
95% RH (Figure 2) by thermal analysis (DTA/TG). The occurrence of different healing products
The Ca(OH )2 content in both mixtures is reduced varies within the samples examined, but there are
exponentially with curing time, but even after 60 some repetitive patterns than can be recognized.
days in high humidity environment there is still a More specifically, calcium carbonate is present in
considerable amount detected. all samples. Calcite is formed on the outer part
The observation of the crack surfaces at the SEM of the crack near the surface, similar to the results
Figure 8: Microphotographs of LP303075. Hydration products formed along a smaller perpendicular crack (left) and a
revealed a variety of secondary products
micro crack completely filled by calcite (right). formed of other studies (Sisomphon, 2012), and is easily
during each healing condition. The secondary prod- recognizable as it crystallizes on the edge of the
In LP specimens cured at
ucts were classified into two main groups, namely higher humidity levels (95%RH)
specimen, the formation
creating a sealingoflayer.
secondary calcite of
The presence
dominates, and that is consistent in the specimens cured for one and
calcite for three
deeper into months
the under
crack is these conditions.
observed mainly in
crystal-like
However,and LP gel-like
specimensproducts (Huang,
after wet-dry cycles 2013).
exhibited a different mechanism. Calcite was the main product
Chemical examination
while C-S-H performed
phases were at SEM/EDAX
scarce, while LPcycles
after 20 additional specimens,
the onlywhereas
secondary inproduct
NHL specimens,
witnessed was calcite
confirmed
calcite. two main
Finally, in thegroups
cases of(Figure subjectedwas
3): crystal-
NHL specimens rarely observed.
to wet-dry cycles and Thethosecarbonation
cured at 95%products
RH, the dis-
like products
prolonged are polymorphs
healing of in
time resulted calcium carbonate,
the enrichment playsurface
of the crack a large diversity
with in terms of both crystal shape
gel-like products.
whereas gel-like Theproducts
effect of cracking at early-age
were consisted was studiedand
of differ- size (Figure
in specimens cured4).for 15 (lime-pozzolan mixtures)
and 7 daysphases
ent hydraulic (NHL mixtures)
(C-S-H andbefore cracking,
C-A-H). where both mixtures
Gel-like Gel-likeareproducts
characterized by an5-7)
(Figures excess
were amount of in
observed
portlandite (Fig.1), compared with those cured for 30 days. The wet-dry cycles on NHL specimens produced a
products show a higher dispersion as to the Al/Si both LP and NHL mortars, mainly as needle-like
dense network of “beehive-like” formations. After 90 days of curing, precipitated calcite was identified, while
ratio,hydraulic
which suggests that there is a wide variety
bee-hive formations are still dominant. The advance formations (Figure
of calcite 5), whilewas
precipitation their density
also andindistri-
witnessed
in thepolished
stoichiometry amongstatthe
sections examined the hydraulic phases. (Fig.9).
polarizing microscope butionSimilarly,
frequency vary between
early-age cracking in different specimens.
LP mortars led
Secondary phases produced
to the enhanced formation ofincalcite
NHL which
specimens dis-
was also Gel-like
observed products
in polished and thinwere commonly
sections examinedfound in small
at the SEM
andhigher
played under polarizing
diversitymicroscope
in Al/Sirespectively.
ratio, whereasEDAXa analysis of the
cracks secondary
where theyproducts
form a formed
dense inside
network the crack
(Figure 6
is presented
sub-group in Figure 10.
was identified After 30gel-like
amongst cycles calcite precipitation
products, left),was
butmore
alsoadvanced,
in lowerwhich suggests
density an ongoing
on planar surfaces,
mechanism (Fig.11).

Figure 9: Secondary formed calcite in NHL mortars after ten (left-NHL730cy) and thirty wet-dry cycles (right-NHL790cy).
Figure 9: Secondary formed calcite
Similar to in NHL
LP mixtures, the mortars after
precipitation ten (left-NHL730cy)
of calcite and cycles.
advances with additional thirty wet-dry cycles (right-
NHL790cy). Similar to LP mixtures, the precipitation of calcite advances with additional cycles.

162 7
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 10: Secondary formed calcite in LP mortars completely filling a micro crack (left). Chemical characterization of
secondary calcite was also determined by EDAX analysis (right).

mostly at the interface between binder and aggre-
gates (Figure 6 right). Another common feature
of gel-like products is the “beehive-like” pattern,
which is formed mainly on planar surfaces (Figure
7).
Concerning the effect of humidity level, in sam-
ples cured at 45%RH (regardless of their compo-
sition, age of cracking and healing period), no
secondary products were observed. In the spec-
imens cured at 75%RH, the secondary products
were scarce in NHL mortars, whereas LP mortars
revealed an abundance of both crystalline and gel-
like secondary phases (Figure 8 left). This was
also confirmed during the examination of the pol-
ished surfaces (Figure 8 right). The above findings
highlight the critical role of water and aqueous so-
lutions in the hydration of hydraulic binders and
the development of the healing mechanism. The
presence of water is crucial for the development
of carbonation and hydration mechanisms (Çizer,
2012) (Karatasios, 2012). Optimum conditions for
the carbonation reaction are met in high humidity
levels, whereas when pore space is completely dry
or fully saturated, carbonation is inhibited (Moore-
head, 1986).
In LPFigure
specimens cured
10: Secondary at higher
formed calcite inhumidity
LP mortars completely calcite
els (95% RH) the formation secondary calcite wascalcite
of secondary
dominates, and that is consistent in the specimens
cured for one and for three months under these
conditions. However, LP specimens after wet-dry
cycles exhibited a different mechanism. Calcite was
the main product while C-S-H phases were scarce,
while after 20 additional cycles the only secondary
product witnessed was calcite. Finally, in the cases
M. Amenta et al.
of NHL specimens subjected to wet-dry cycles and
those cured at 95% RH, the prolonged healing time
resulted in the enrichment of the crack surface with
gel-like products.
The effect of cracking at early-age was stud-
ied in specimens cured for 15 (lime-pozzolan mix-
tures) and 7 days (NHL mixtures) before cracking,
where both mixtures are characterized by an ex-
cess amount of portlandite (Figure 1), compared
with those cured for 30 days. The wet-dry cycles
on NHL specimens produced a dense network of
"beehive-like" formations. After 90 days of curing,
precipitated calcite was identified, while hydraulic
bee-hive formations are still dominant. The ad-
vance of calcite precipitation was also witnessed
in polished sections examined at the polarizing mi-
croscope (Figure 9). Similarly, early-age cracking
in LP mortars led to the enhanced formation of
lev-
which
filling a micro was
crack also
(left). observed
Chemical in polished
characterization of and
also determined
thin by sections
EDAX analysis (right). at the SEM and under po-
examined

Figure 11: Secondary formed calcite in LP mortars after ten (left-LP1430) and thirty wet-dry cycles (right-LP1490). It was
Figure 11: Secondary formed calcitethat
observed in the
LPprecipitation
mortars after ten (left-LP1430)
of calcite and thirty
advances with additional wet-dry cycles (right-LP1490).
cycles.
It was observed that the precipitation of calcite advances with additional cycles.

163
Conclusions
Although complete healing of the main crack induced by three point bending, was not observed in any
mixture, smaller side-cracks were completely filled with secondary products under specific each-time conditions.
 Amenta et al.
larizing microscope respectively. EDAX analysis
of the secondary products formed inside the crack
is presented in Figure 10. After 30 cycles calcite
precipitation was more advanced, which suggests
an ongoing mechanism (Figure 11).
IV. Conclusions
Although complete healing of the main crack in-
duced by three point bending, was not observed in
any mixture, smaller side-cracks were completely
filled with secondary products under specific each-
time conditions.
The study of crack surfaces under SEM revealed
the presence of carbonates and a large variety of
hydration products, depended on the initial com-
position of the binder. The hydration products pro-
duced in both NHL and LP mortars were mostly
occupying smaller cracks, especially in the interface
between binder and aggregates, while carbonation
products were observed in larger cracks.
At the humidity level of 75%, the healing prod-
ucts were abundant in LP specimens, whereas in
NHL were hardly spotted. Lower humidity con-
ditions (45% RH) found to be inadequate for the
evolution of the autogenous healing phenomenon
to take place. Nevertheless, under these conditions
a high percentage of unreacted components remain
active in the mortar matrix, allowing a potential fu-
ture initiation of the self-healing, when the proper
conditions are reached.
The effect of cracking age was a crucial parameter
to the healing efficiency, since specimens cracked
at early age displayed higher percentage of crack-
healing than those cracked after one month of cur-
ing.
Finally, curing time was found to be of prime
importance for early-age cracked specimens under-
wet dry cycles, where additional curing time was
found to have a substantial effect on the percentage
of the cracks filled by precipitated calcite (both
in LP and NHL mortars). In contrast, specimens
cracked at later age showed little or none evolution
of the phenomenon after the first ten cycles.
164
References
Adenot, F., and M. Buil. (1992): “Modelling of the Corrosion of the
Cement Paste by Deionized Water.” Cement and Concrete Research, Special
Double Issue Proceedings of Symposium D of the E-MRS Fall Meeting
1991, Vol. 22, No. 2–3, 489–96.
Breugel K. van. (2012): “Self-healing material concepts as
solution for aging infrastructure”. 37th Conference on our world
in Concrete &Structures, Singapore. Accessed April 8, 2015. http :
//www.cipremier.com/e107 f iles/downloads/Papers/100/37/100037009.pd f
Çizer, Ö., Rodriguez-Navarro,C., Ruiz-Agudo, E., Elsen, J., Gemert, D.
V. and Balen, K. V. (2012): “Phase and Morphology Evolution of Calcium
Carbonate Precipitated by Carbonation of Hydrated Lime”. Journal of
Materials Science. Vol. 47, No.16, 6151-6165.
Çopuroğlu, O., Schlangen, E., Nishiwaki, T., Tittelboom, K.V., Snoeck,
D., Belie, N.D., Rooij, M.R. de, (2013): “Experimental Techniques Used to
Verify Healing”, in: Rooij, M. de, Tittelboom, K.V., Belie, N.D., Schlangen,
E. (Eds.), Self-Healing Phenomena in Cement-Based Materials, RILEM State-
of-the-Art Reports. Springer Netherlands, pp. 19–63.
Edvardsen, C. (1999): “Water Permeability and Autogenous Healing
of Cracks in Concrete.” ACI Materials Journal Vol. 96, No. 4, 448–54.
Hearn, N., Morley, C.T., (1997): “Self-sealing property of con-
crete—Experimental evidence”. Materials and Structures Vol. 30, 404–411.
Hearn, N. (1998): “Self-Sealing, Autogenous Healing and Continued
Hydration: What Is the Difference?” Materials and Structures Vol.31, No.
8, 563–67.
Huang, H., and Ye G.. (2015) “Self-Healing of Cracks in Cement
Paste Affected by Additional Ca2+ Ions in the Healing Agent.” Journal of
Intelligent Material Systems and Structures, vol. 26, No. 3, 309-320
Huang H., Ye G., and Damidot D. (2013): “Characterization and quan-
tification of self-healing behaviors of microcracks due to further hydra-
tion in cement paste”. Cement and Concrete Research Vol. 52, 71–81.
Jacobsen, S., Marchand, J., Hornain, H.. (1995) “SEM Observations
of the Microstructure of Frost Deteriorated and Self-Healed Concretes.”
Cement and Concrete Research Vol. 25, No. 8
Karatasios, I., Amenta, M., Tziotziou, M., Kilikoglou, V., (2012): “The
Effect of Relative Humidity on the Performance of Lime-Pozzolan Mor-
tars”, in: Valek, J., Hughes, J.J., Groot, C.J.W.P. (Eds.), Historic Mortars,
RILEM Bookseries. Springer Netherlands, 309–318.
Lauer and Slate. (1956): “Autogenous Healing of Cement Paste.” ACI
Journal Proceedings. Vol. 52, No. 6
Lubelli, B., Nijland, T.G., van Hees, R.P.J. (2011): “Self-healing of
lime based mortars: microscopy observations of case studies”. Heron
Vol. 56(1/2), 75–91.
Lubelli, B., Nijland, T. G. , van Hees R. P. J. (2012): “Simulation of
Self-Healing of Dolomitic Lime Mortar,”. Materiali In Tehnologije Vol.46,
No. 3, 291-296.
Moorehead, D.R., (1986). “Cementation by the carbonation of hy-
drated lime”. Cement and Concrete Research Vol.16, 700–708.
Nijland, T.G., Larbi, J.A., van Hees, R.P.J., Lubelli, B., de Rooij, M.
(2007): “Self-healing phenomena in concretes and masonry mortars: a
microscopic study.” In: Proceedings of the First International Conference on
Self-healing Materials, Noordwijk aan Zee,The Netherlands, p. 31
Schlangen, E., ter Heide, N. and van Breugel, K., (2006): "Crack heal-
ing of early age cracks in concrete." In: M.S. Konsta-Gdoutos, Eds. Mea-
suring, monitoring and modeling concrete properties. Netherlands: Springer
Sisomphon, K., Çopuroğlu, O. and Koenders, E.A.B., (2012). "Self-
healing of surface cracks in mortars with expansive additive and crys-
talline additive." Cement and Concrete Composites, Vol. 34, No.4, 566-574.
Yang, Y., Lepech, M.D., Yang, E.H., and Li, V.C. (2009): "Autogenous
healing of engineered cementitious composites under wet-dry cycles."
Cement and Concrete Research, Vol. 39, No. 5, 382-390.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Assessment of the functionality of

bacteria-based repair system for concrete
through ESEM analysis
V. Wiktor⇤,a,b , H.M. Jonkersa
a Delft University of Technology, Faculty of Civil Engineering and Geosciences, Section of Materials and Environment,
Delft, The Netherlands.
b Bartels Building Solutions, Apeldoorn, The Netherlands
⇤ v.a.c.wiktor@gmail.com

Abstract

Biodeposition, a method by which calcium carbonate (CaCO3 ) precipitation is induced by bacteria, has been proposed
as an interesting approach to protect building materials. The liquid-based system presented in this paper aims at the
sealing of cracks and decrease of the porosity due to the production of a calcium-based biomineral. In this system, a
silicate-based compound, which has a composition similar to concrete, is associated with Microbial Induced Precipitation
(MIP), without involving urea hydrolysis pathway. Instead, the metabolic conversion of organic salts through bacterial
respiration is used for MIP. The novelty of such system is to combine advantages of both, traditional repair system for
concrete (fast reacting, and short term efficiency), and bio-based methods (more sustainable, slow process, and long-term
efficiency).
This paper presents how environmental scanning electron microscope (ESEM) has been used in the development of
the bacteria-based repair system. It is a good evaluation technique to assess the functionality of the bacteria-based repair
system and to evidence the involvement of bacteria in the mineral production. The functionality of the bacteria-based
repair system has first been assessed in the repair system alone, and then after injection into porous concrete.
These results bring more insight regarding the formation of the biomineral. Though the bacteria are active after 1
day, it requires longer time to indeed observe mineral formation. Several days are needed to bacteria to actively convert
the feed and produce substantial amount of CO2 , leading to favorable chemical environment for calcium carbonate
precipitation. The results also showed how FTIR and ESEM analysis are complementary. Vaterite and calcite have been
identified thank to FTIR while ESEM observation highlighted the unique features of the biomineral in size, shape and
texture.

Keywords: bacteria-based system, repair, concrete, biomineral, bacteria imprints

I. Introduction form a gel when mixed together. This gel is strong

The development of bacteria-based systems for the
protection of concrete structures has gained lot of
attention over the past few years. These systems,
engineered for self-healing concrete or concrete re-
pair (Dhami et al., 2012), are based on Microbial
Induced Precipitation (MIP), a method by which
calcium carbonate precipitation is induced by bac-
teria.
The system presented in this study is a liquid-
based system for concrete repair aiming at the seal-
ing of the cracks and decrease of the porosity due
to the production of calcium-based biomineral. In
this system, a silicate-based compound, which has a
composition similar to concrete, is associated with
MIP using metabolic conversion of organic salts
through bacterial respiration. The system is com-
posed of two solutions, named A and B, which
enough to allow a rapid sealing of the crack and
is also providing a suitable environment for the
bacteria to precipitate calcium carbonate (CaCO3 ).
By the time the gel becomes too weak, a substan-
tial amount of CaCO3 has been precipitated to seal
the crack. And while assessing the crack sealing
efficiency of the system can easily be done through
water permeability test, the investigation and proof
of the bacterial involvement in the mineralization
process is not an easy task.
Indeed, the basic principle behind the proposed
system is that organics (feed for bacteria) are effi-
ciently converted by the bacteria into CO2 . Due to
the alkaline environment in the system, the CO2 is
in turn converted into carbonate ions CO23 , which
precipitate with the calcium present in the feed to
form CaCO3 (biomineral). Therefore, the perfor-
mance of the proposed system is directly linked

165
 Wiktor & Jonkers
Figure 1: Details of the FTIR spectra from the powder obtained after filtration of the bacteria-based repair solution at
1, 2, 4, 6 and 8 weeks. (a) C-O asymmetric stretching vibration (n3), (b) C-O out of plan bending vibration (n2).
to the efficiency of microbial induced precipitation
and so to the biomineral production. However,
after the crack has been successfully sealed, it is
very delicate to determine whether the bacteria fa-
cilitated the CaCO3 formation or if it is simply a
result of physico-chemical conditions in the crack
micro-environment as for instance it can also result
from natural carbonation of concrete. In that area,
environmental scanning electron microscopy is a
precious tool as it may enable the observation of
the biomineral which is characterized by a unique
feature: the presence of bacteria imprints.
This paper presents how environmental scanning
electron microscope (ESEM) has been used in the
development of the bacteria-based repair system. It
is a good evaluation technique to assess the func-
tionality of the bacteria-based repair system and to
evidence the involvement of bacteria in the mineral
production. The functionality of the bacteria-based
repair system has first been assessed in the repair
system alone, and then after injection into porous
concrete.
II. Materials and methods
Bacteria-based repair system
The bacteria-based repair system is a liquid-based
system which transports a bio-based agent into
concrete. The bio-based agent is composed of con-
crete compatible bacteria and feed which produces
calcite-based minerals resulting in decreased poros-
ity and sealing of the crack. The bacteria are from
the genus Bacillus. They are added as endospores,
dormant bacteria cells with characteristic compact
166
round shape, typically in the size range of 0.8-1µm.
When the environmental conditions are favorable
(presence of water, nutrients and oxygen) these en-
dospores germinate and grow into vegetative bacte-
rial cells. Vegetative cells of Bacillus are rod-shapes
and micrometer size (Buczynski and Chafetz, 1993).
The preparation and composition of the two so-
lutions forming the repair system, solution A and
solution B, is as described by Wiktor and Jonkers
(2012). Briefly, it is composed of:
(i) Solution A – Sodium-silicate (alkaline buffer),
Sodium-gluconate (Carbon source for bacteria
growth), alkaliphilic bacteria.
(ii)Solution B – Calcium-nitrate (nitrate source
for denitrification when O2 is depleted and calcium
for CaCO3 precipitation), alkaliphilic bacteria.
Biomineral formation in the bacteria-based sys-
tem
Previous results (Wiktor et al., 2013) have shown
that the bacteria are active within 24h after the mix-
ing of the two solutions and that efficient sealing
of a crack is obtain after 6 weeks. However very
little is known on the kinetic of the biomineral for-
mation over these 6 weeks and whether the crack
sealing is indeed due to MIP. The goal here is to
follow in time the formation of the biomineral in
the repair system. For this purpose, solution A
and B are mixed together in small flasks and left
in the lab at room temperature for 1, 2, 4, 6, and
8 weeks. At each time, the precipitate and gel
are separated by filtration on a sintered-glass fil-
ter (pore size 10-16µm), washed with demi-water,
dried at 35 C, grinded, and further analysed with
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 2: ESEM pictures of the precipitate obtained after filtration from the bacteria-based repair solution at (a) 1
week, (b) 2 weeks, (c) 4 weeks, (d) 6 weeks and (e) 8 weeks. (f) detail of picture (e). The arrows indicate bacteria
imprints.

Fourier-Transformed Infra-Red spectroscopy (FTIR) Biomineral formation in concrete-based material

for mineral identification and analysed with ESEM
(Philips XL30 Series) equipped with an Energy Dis-
persive X-ray element analyzing system (EDS) to
possibly observe bacteria imprints. The samples are
not coated prior ESEM observation.
FTIR spectra were collected on a Perkin–Elmer
Spectrum 100 Series spectrometer equipped with
universal Attenuated Total Reflexion (ATR) unit.
The spectra were recorded in the range of 4000–600
cm 1 with 2 cm 1 resolution, and 32 scans were
collected each time. The ATR analyses require very
small amount of sample (<5 mg), and furthermore
no preparation or dilution of the sample is needed.
The FTIR was first calibrated for background signal
scanning, and then the experimental sample scan-
ning was conducted. The spectra were normalized
in order to compare them.
In order to assess the functionality of the bacteria-
based repair system in the concrete environment
the repair system has been injected into porous
concrete. By using porous concrete, large volume
of repair solution can be injected in the material
and therefore results in many spots within one
specimen where the biomineral can be formed. The
injection of the bacteria-based repair solution and
preparation of the porous concrete is as described
by Sangadji and Schlangen (2012).
Three weeks after the injection, the specimens
are dried, and fully impregnated with low viscosity
epoxy mixed with 1% by weight of hudson yel-
low pigment prior to sawing along the longitudinal
cross-section. After epoxy impregnation of the new
exposed surface, polished sections were then pre-
pared from each sample, by grinding with water

167
 Wiktor & Jonkers

Figure 3: Observation of polished section of porous concrete specimen prepared 3 weeks after injection of the bacteria-
based repair system: (a) and (b) ESEM pictures, (c) and (d) Elemental mapping corresponding to (a) showing the
silicon and calcium distribution respectively. Arrows indicate bacteria imprints. Agg.=aggregate, c.p=cement paste,
b=biomineral.

the specimen surface with paper grade P320, P500,
P800, P1200 for 10 min each and P4000 for 20 min.
The polished sections were observed with ESEM
and elemental mapping was acquired with the
ESEM/EDS. The polished sections are not coated
prior ESEM observation.
III. Results and discussion
Biomineral formation in the bacteria-based re-
pair system solution
The FTIR spectra (Figure 1) are indicative of the
presence of calcium carbonate. The FTIR spectra
of crystalline calcium carbonate polymorphs (cal-
cite, aragonite and vaterite) have been extensively
reported in the literature (Ghosh, 2001). Because of
their different crystal structure, they can be discrim-
inated using FTIR, a different spectrum is observed
for each of the structural forms. The occurrence of
the major vibrational bands specific to calcite are
observed: the C-O asymmetric stretching vibration
(n3), and the C-O out of plan bending vibration
(n2) of carbonate centered at 1420 and 865cm 1 re-
spectively. The C-O planar bending vibration (n4)
centered at 712 cm 1 , which is also the least intense
of the three, has not been distinctly observed as
it was probably overlapped by the peak present
at 700 cm 1 and therefore not shown on figure 1.
Peaks at 835 and 872 cm 1 are also observed at 1, 2
and 4 weeks. They decrease in time while the peak
centered at 865 cm 1 is increasing. These peaks
likely correspond to the C-O out of plan bending
vibration (n2) of vaterite (Sato and Matsuda, 1968).
After the FTIR analysis, the powder has been
observed with the ESEM and the corresponding
pictures are presented on figure 2. A change in the
morphology of the particles is noted as the time
increases. Indeed, at 1 week (Figure 2a), the parti-
cles are arranged in small flaks with no distinctive
structure neither well define edges. The presence
of silicon and calcium as indicated by EDS analy-
sis, and the presence the Si-O stretching vibration
centered at 1000cm 1 on the FTIR spectra (data
not shown) suggest that the powder obtained after
1 week is mainly composed by a form of calcium
silicate. This calcium silicate results from the imme-
diate reaction between the silicates present in the
solution A and the calcium in solution B.
At 2 weeks (Figure 2b) globular morphology
(calcium-based) with small rod-like imprints are
observed. These globular shapes become bigger
in size and more structured as the time increases
(Figure 2c-e) suggesting crystallization. Taking into

168
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
account that vaterite has been detected with FTIR
at 1, 2 and 4 weeks and calcite has been detected
at 4, 6 and 8 weeks, we can assume that the glob-
ular morphologies are calcium carbonate, mainly
in the form of vaterite in the first weeks turning
into calcite from week 4. The globular morphol-
ogy observed in the first weeks are characteristic of
vaterite. However, at week 4, this morphology is
very different from the typical rhombohedral mor-
phology usually observed for calcite. This can be
explained by the presence of organics in the sys-
tem which are known to have a profound influence
on the crystal shape and size. Interestingly, the
rod-like imprints are randomly distributed over the
crystal surface and their occurrence also increase
with time. The shape and size of these imprints are
in good agreement with those of bacteria suggest-
ing them to be bacteria imprints rather than default
in the crystal structure. This is a strong indication
of the involvement of the bacteria in the calcium
carbonate formation. Indeed, vegetative cells are
characterized by rod-shaped structure whereas en-
dospores (the form in which bacteria have been
added) are smaller compact round cells (Buczynski
and Chafetz, 1993). Moreover, a closer look at the
globular particle surface (Figure 2f) shows that this
morphology indeed seems to be the results of the
assembly of small calcite crystals grown around the
bacteria cells.
Biomineral formation in concrete-based material
After showing that the bacteria-based repair system
is functional in solution, it is of prime importance
to assess its functionality and behavior in concrete-
based material. The Figure 3 show a polished sec-
tion of a cross section of porous concrete specimen
3 weeks after it has been impregnated with the
bacteria-based repair system.
It can be seen from Figure 3a and b that the bond-
ing between the epoxy and the specimen is not
good as some epoxy seems to have scaled off from
the surface. Considering that during the prepa-
ration of the polished sections the grinding was
performed with water and that each specimen has
been in contact with the grinding paper and water
for 60 min, it can be concluded that the food dis-
solved resulting in holes what then appears as bad
bonding between epoxy and the matrix. However,
this “de-bonding” can serve as indicator of the pres-
ence of food, and help to locate CaCO3 formed due
conversion of food by bacteria.
Moreover, though morphological information is
lost due to the polishing, the de-bonding around
round particles (Figure 3b) reveals a globular mor-
phology and bacteria imprints, similarly to the pre-
vious ESEM observation suggesting that it is actu-
ally the biomineral. The silicon and calcium map-
ping (Figure 3 c and d) of the specimen indicate that
these globules are calcium-based and are primar-
ily formed in the vicinity of silicon based material
possibly cement paste and silicates from the repair
system where the pH is alkaline. Therefore con-
sidering that the bacteria added to the system are
alkaliphilic (grow at high pH) these observations
support the assumption that the round calcium-
based particles are probably calcium carbonate re-
sulting from MIP.
IV. Conclusion
These results bring more insight regarding the for-
mation of the biomineral. Though the bacteria are
active after 1 day, it requires longer time to indeed
observe mineral formation. Several days are needed
to bacteria to actively convert the feed and pro-
duce substantial amount of CO2 , leading to favor-
able chemical environment for calcium carbonate
precipitation. The results also showed how FTIR
and ESEM analysis are complementary. Vaterite
and calcite have been identified thanks to FTIR
while ESEM observation highlighted the unique
features of the biomineral in size, shape and tex-
ture. The presence of bacteria imprints gave strong
indication on the bacterial involvement in the min-
eral formation underlying the functionality of the
bacteria-based repair system in solution as well as
in cement-based material.
Acknowledgement
The authors would like to thank Arjan Thijssen
for his help with the ESEM observations. The fi-
nancial support from Agentschap NL (IOP grant
SHM012020) for this work is gratefully acknowl-
edged.
References
Buczynski, C., and Chafetz, H.S., (1993): “Habit of bacterially induced
precipitation of calcium carbonate”. In Rezak R. and Lavoie D. Eds, Car-
bonate microfabrics. Springer, 105-116.
Dhami, N.K., Reddy, S.M., and Mukherjee, A., (2012): “Biofilm and
Microbial applications in biomineralized concrete”. In: Seto J, ed. Ad-
vanced Topics in Biomineralization. Rijeka: InTech, 137-64.
Ghosh, S.N., (2001): “Infrared spectroscopy” in Ramachandran, V.S.,
and Beaudoin, J.J Eds, Handbook of analytical techniques in concrete science
and technology - Principles, techniques and applications. Noyes publication,
William Andrew publishing, 174-204.
Sangadji, S., and Schlangen, E., (2012): “Self-healing of Concrete
Structures - Novel approach using porous network concrete.” Journal of
Advanced Concrete Technology. Vol 10, 185-194.
Sato, M. and Matsuda, S., (1968): Structure of vaterite and infrared
spectra. Zeitschrift fur kristallographie, S405-410, 129.
Wiktor, V., and Jonkers, H.M., (2012): “Application of bacteria-based
repair system to damaged concrete structures”. Proceedings of the 2nd
International Workshop on Structural Life Management of Underground Struc-
tures, Daejeon, Republic of South Korea, October 18-20, 31-34.
Wiktor, V., Sangadji, S., Jonkers, H.M., and Schlangen, E., (2013):
“Potential of bacteria-based repair solution as healing agent for porous
network concrete”. Proceedings of the 4th International conference on Self-
healing materials, Ghent, Belgium, June 16-20, 592-593.
169
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

170
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Investigation of self healing phenomenon

by calcite precipitation using reactive
transport modelling and microscopic
behavior.
B. Hilloulin⇤,a,b , F. Grondina , A. Soivea,c , A. Loukilia
a LUNAM Université, Institut de Recherche en Génie Civil et Mécanique (GeM), UMR-CNRS 6183,
Ecole Centrale de Nantes, 1 rue de la Noë, 44321 Nantes, France.
b Magnel Laboratory for Concrete Research, Department of Structural Engineering, Faculty of Engineering, Ghent
University, Technologiepark Zwijnaarde 904, B-9052 Ghent, Belgium.
c Centre d’Etudes et d’Expertise sur les Risques, l’Environnement, la Mobilité et l’Aménagement (Cerema),
9 rue René Viviani - BP 46223, 44262 - Nantes Cedex 2, France.
⇤ benoit.hilloulin@ec-nantes.fr

Abstract

Cracks facilitate the ingress of aggressive and harmful substances into concrete. They lead to substantial inspection
and repairing works which considerably increase the cost of some constructions. Self- healing of cracks could be a
mean to significantly improve the durability of concrete structures and reduce reparation costs. Concrete’s intrinsic
ability to heal, called autogenic healing, has been reported for many years. This natural process is being improved and
supplemented for some years by promising engineered additions such as mineral additions, capsules containing healing
agents, minerals producing bacteria, or fibres limiting the crack width. However, a deeper understanding of the natural
phenomenon could help to design innovative healing solutions based on cementitious materials themselves and / or
supplement the actual construction codes concerning their maximum allowable cracks widths for structures.
In this study, self-healing phenomenon in concrete by calcite formation is studied in order to provide information
about the sealing process which leads to an impermeability increase. Experimental works, based on water immersion,
are conducted on pre-cracked specimens. A focus is put on the influence of initial curing and environmental healing
conditions on the kinetics and the potential of the phenomenon (maximum healable crack width). A reactive transport
model is proposed to add explanations for some observed features of the phenomenon and determine some of its influencing
parameters.
Experimentally, initial curing conditions have an important impact on the self-healing kinetics and potential as air
cured specimens do not reduced healing properties. A new method consisting of immersing these mortar specimens in
CO2 enriched water is proposed to heal cracks that cannot heal in tap water.
Numerically, some experimental observations are explained such as the end of the healing process which is probably
due to the local decrease of porosity nearby the crack, leading to a very slow ionic transport from the matrix.

Keywords: self-healing; reactive transport modelling; cementitious materials; environmental conditions, 3D microscope.

I. Introduction solutions could help the design of structures with

Cracks in concrete generate important inspection
and repairing costs, especially for public services
and governments. Some of them can seriously af-
fect the durability and the stability of the structures
because of ingress of harmful substances causing
corrosion or leakage. Thus, self-healing concrete
could be a mean to make important savings by
decreasing direct costs due to monitoring, reabilita-
tion or reparation, and indirect costs inherent with
repairing works such as trafic jams (van Breugel,
2009; De Rooij et al., 2013). Moreover, self- healing
long term durability properties for future nuclear
waste disposal for example.
Self-healing can occur naturally, without any par-
ticular additive, under favourable conditions with
the continuous or alternate presence of water by lo-
cal restart of hydration (Granger et al., 2007; Huang
et al., 2013; Hilloulin et al., 2014a) or formation of
healing products by reaction with the environment
(mainly calcium carbonate from calcium in the con-
crete and carbonate ions contained in water) (Clear,
1985; Edvardsen, 1999). Some studies have reported
the predominance of the precipitation phenomenon

171
 Hilloulin et al.
in common concrete with a water-to-cement (w/c)
ratio around 0.4 - 0.5 (Edvardsen, 1999; Ranaivo-
manana et al., 2013). Natural healing, called ‘au-
togenic’ healing, has been observed to be limited
to cracks with a maximal width of around 100 –
200 µm (Clear, 1985; Edvardsen, 1999; Aldea et al.,
2000; Reinhardt and Jooss, 2003). Due to this actual
limitation, novel engineered techniques have been
developed to maximize healing, such as embedded
capsules/vascular system (Dry and McMillan, 1996;
Li et al., 1998; Mihashi et al., 2000; Van Tittelboom
et al., 2011; Hilloulin et al., 2015), or calcium car-
bonate embedded bacteria production (Wiktor and
Jonkers, 2011; Wang et al., 2014).
Autogenic healing by further hydration has been
studied by several researchers because of expected
mechanical regains due to the creation of new cal-
cium silicate hydrates (CSH) (Jacobsen and Sell-
evold, 1996; Ter Heide, 2005; Granger et al., 2007;
Ter Heide and Schlangen, 2007). A global recovery
of stiffness can be achieved by further hydration
but mechanical regain seem very limited when cal-
cite is formed into the crack (van Tittelboom et
al., 2012). However, the kinetics and the limits of
the phenomenon are not totally understood. More
experimental tests are needed to determine the in-
trinsic characteristics of the healing products, with
the ultimate goal to propose predictive models. Self-
healing observation and monitoring, usually per-
formed by optical microscope on the crack surface,
needs to be supplemented by tomography (Fukuda
et al., 2013) or non- destructive techniques (In et
al., 2013; Hilloulin et al., 2014b), in order to de-
scribe and quantify the filling of cracks inside the
specimens.
Up to now, only few models were developed
to describe self-healing of concrete. Some mod-
els were proposed to determine the amount of un-
hydrated cement particles in concrete specimens
considering w/c ratio and cement fineness which
underlies the self-healing potential (He et al., 2007b;
He et al., 2007a), or to calculate the amount of heal-
ing product due to further hydration considering
two crack modes (Lv and Chen, 2012; Lv and Chen,
2013). A model simulating further hydration by
water release by a capsule in a crack using wa-
ter transport theory, ion diffusion theory and ther-
modynamics theory has been recently developed
(Huang and Ye, 2012; Huang et al., 2014), while
characterization of the healing products relatively
low mechanical properties have been assessed at a
mesoscopic scale (Hilloulin et al., 2013; Hilloulin
et al., 2014a). However, these models do not invest
which parameters mainly influence the creation of
calcite at the concrete – water interface which is the
mostly observed phenomenon, and cannot assess
the self-healing extent, e.g. speed and potential.
In this study, a reactive transport model has been
172
developed to simulate autogenic healing by calcium
carbonate precipitation which seems to be the pre-
dominant phenomenon in most of the cases (Van
Tittelboom et al., 2012), with the aim to explain the
influence of self-healing conditions and qualify the
self-healing potential (i.e. maximum crack width
closed) of a given composition regarding its cur-
ing conditions. The qualitative numerical results
from the model are compared with experimental
observations aiming at studying the influence of the
environmental and curing conditions on the phe-
nomenon. New environmental healing conditions
are proposed to increase and accelerate natural heal-
ing efficiency by calcium carbonate precipitation.
Moreover, the use of 3D microscope to observe the
healing product has been employed in order to ob-
tain informations on the volume it occupy. This
observation gives indications on the reliability of
classic 2D microscope use.
II. Reactive transport model
A 1-D reactive transport model has been developed
in TOUGHREACT (Xu et al., 2012) to study the
formation of mineral precipitations, especially cal-
cite by reaction of carbonate ions from water in the
crack and calcium from the cementitious matrix
(from interstitial water, or due to degradation of
portlandite and/or CSH). A mortar sample with
a length of 5 mm with an initial porosity of 20 %
discretised into 100 µm cells (and 50 µm cells at the
surface) along the X direction, was considered in
contact with 500-µm wide water filled zone at one
side (representing the crack). Water, originally at
pH = 7, was discretised into 100 µm cells in order to
have access to the eventual quantity of precipitates
formed. The boundary water cell which is not in
contact with the sample had a higher simulation
volume, e.g. several litres, in order to represent
outside water far from the sample. This hypothesis
was considered in order to not quickly saturate the
whole water volume with calcium ions and calcite
and induced approximately constant ion concentra-
tions and therefore a constant pH of around 7 in
this cell. Mortar was initially assumed saturated
in order to focus on ionic diffusion and chemical
reaction, so that water transport had no influence
in the initial model.
Also, a 2D version of the model was developed
to study the influence of water flux parallel to the
crack on the calcite precipitation. The 2D model
was discretized over the Y axis, parallel direction
to the crack, into 3 4-mm long cells in order to get
an eventual influence of the water flux on the ionic
concentrations and pH. Three water flux speeds
were simulated: 0.3 µm/s, 3 µm/s and 30 µm/s.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Table 1: Ion species, solids and chemical reactions considered in the simulation.
Primary species Minerals
Al3+ , Ca2+ , H+ , H2 O Calcite, Portlandite, Ettringite,
H4 SiO4 ( aq) , HCO3 , K+ Monosulfoaluminate, CSH_1.6
Na+ Mg2+ , SO24
Ionic diffusion is based on the Fick’s law :
Fj = D ⇥ rCj (1)
where Fj represents the ionic flux of the j species,
D the ionic diffusion which is assumed to be the
same for all the ionic species (equal to 10 9 m2 /s;
which is close to Ca2+ and HCO3 diffusion coeffi-
cient (Thiery, 2005), and Cj the ionic concentration
of the species. Ion species present in interstitial
mortar water were considerably reduced to only
H+ to calibrate pH and Ca2+ determined during
the first iteration closed to 20 mmol/l for uncar-
bonated mortar, in order to obtain a reasonable
running time, likely leading to a small overestima-
tion of healing speed due to the shortage of alkali
ions K+ and Na+ . Laboratory experiments have
shown that modest reductions in porosity from
chemical precipitation can cause large reductions
in permeability in rocks (Vaughan, 1987), and in
the case of self-healing, to a probable stop of the
phenomenon due to the formation of a calcite layer
decreasing the speed of an initially slow diffusion
controlled process (Edvardsen, 1999) (compared
with the first surface-controlled process). Recently,
(Ducasse-Laeyrusse, 2014) experimented that heal-
ing of carbonated mortars was slower than the one
of uncarbonated specimens. He attributed this find-
ing to the difficulty for ions from the matrix to cross
the small calcite layer in the crack due to carbon-
ation. Thus, we decided to compare simulations
considering the influence of the porosity variation
Figure 1: Influence of the healing conditions on the mean calcite layer thickness created during the first 3 days of
healing.
Precipitation/Dissolution Equilibrium constants
equations (log K, 25 C)
CaCO3 + H+ ! Ca2+ + HCO3 1.85
Ca(OH)2 + 2H+ ! Ca2+ + 2H2 O 22.81
on flow or not. The influence of porosity changes
on flow is taken into consideration by a decrease of
the effective ion diffusivity D according to the ini-
tial diffusivity D0 , the tortuosity t and the porosity
f:
D = t ⇥ f ⇥ D0 (2)
Mineral precipitates considered in this study are
calcite (CaCO3 ) and portlandite (Ca(OH)2 ) with
equilibrium constant from Thermoddem (Blanc et
al., 2012). The chemical reactions are described
thanks to a set of mass balance equations (equa-
tion 3, zero superscript indicates initial values) and
equilibrium equations (equation 4):
(c j + c0j ) + Sn(cm + c0m ) = 0 (3)
’ cn ⇥ gn
SI = log10 =0 (4)
K
Where c j and cm represent the concentrations of,
respectively, the primary species (ions) and miner-
als, n the stoichiometric coefficients, g the Debye-
Hückel activity coefficients, SI the saturation index
and K the equilibrium constants of the considered
reactions. Chemical reactions and equilibrium con-
stants are summarized in Table 1.
In order to model the effect of curing condi-
tions on self-healing potential, simulations were
conducted using lower portlandite volume fraction
in the first cells of mortar in contact with water to
reproduce a consequence of a drying phenomenon
173
 Hilloulin et al.
Figure 2: Evolution of crack widths over time for a specimen C1-H2: estimation of the maximum healable crack width
and assessment of healing variability.
(portlandite volume fraction of 0.1 instead of 0.2).
Other simulations were led considering a higher
porosity in this zone (20% instead of 10%).
Various parameters can be studied to represent
the healing conditions (calcium hydroxide satura-
tion (Huang et al., 2013), wet/dry cycles (Yang et
al., 2009). In this study, we focused on the concen-
tration of carbonate ions in water and on eventual
water fluxes.
III. Experimental program
Material and methods
Mortar specimens (7⇥7⇥28 cm3 ) with a w/c of
0.35 were mixed. Their composition is detailed in
Table 2. After 24 h curing in sealed conditions at
20 C, specimens were demoulded. Further curing
conditions were then applied: specimen from series
C1 were cured at 20 C, 80% of relative humidity
(RH) during 36 days (referred as ‘dry curing’), speci-
men from series C2 were cured in sealed conditions
at 20 C during 18 days and specimen from series
C3 were cured in calcium hydroxide saturated wa-
ter during 18 days at 20 C. Specimens were then
cracked by mean of a crack mouth opening dis-
placement (CMOD) controlled three-point bending
test until the formation of a single crack with a
targeted maximum width of around 100 µm for 5
series, or 300 µm for the 2 other series (large cracks
later called ‘L’).
After cracking, a first set of microscopic pic-
Table 2: Mortar formulation.
Cement (CEM II 42,5) [kg/m3 ] Sand 0/2 [kg/m3 ]
740 1250
174
tures from different specific locations all around
the cracks were taken. Thus, initial crack width
(winitial ) was determined in order to constitute ini-
tial measurement for crack width monitoring. Spec-
imens were then subjected to one of the following
healing conditions:
• immersion in tap water (H1),
• immersion in tap water at equilibrium with air
with a partial CO2 pressure of 4 % and 65 %
RH, referred as ‘CO2 enriched water’ (H2 ),
Curing and healing conditions for all the speci-
mens are detailed in Table 3 with the CMOD after
unloading.
Because the thickness of the healing product layer
is limited by the initial crack width which thus influ-
ences the comparison between two specimens that
do not have exactly the same initial crack width, a
fair comparison can be made using healing percent-
ages as follows:
winitial wt
HPt,% = ⇥ 100 (5)
winitial
where wt represents the crack width at a given
healing time. Spatial variability of crack filling
fraction was quantified by extracting the maximum
and minimum crack width which can be healed
over different locations. Thus, different local crack
widths have been monitored.
In order to get informations on the spatial mor-
phology and distribution of the healing product,
two specimens were observed by 3D microscope
Water [kg/m3 ] Superplasticizer [kg/m3 ] W/C
260 1.2 0.35
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 3: Curing conditions, healing conditions and cracking details for the different specimens.

Series name Curing conditions Age at cracking (d) CMOD (µm) Healing conditions
C1-H1 80% RH 36 90 Tap water
C1-H2 80% RH 36 90 CO2 enriched water
C2-H1 Sealed 18 85 Tap water
C2-H2 Sealed 18 85 CO2 enriched water
C2-H2-L Sealed 18 360 CO2 enriched water
C3-H1 Water immersed 18 95 Tap water
C3-H2-L Water immersed 18 300 CO2 enriched water

afterwards: one with an unhealed crack (specimen Results on the influence of healing conditions
kept out of water) and one with an almost complete
healing with only small zones that were supposed As illustrated in Figure 4, specimens which were
empty with classic 2D optical microscope. dry cured (C1) did almost not heal while the spec-
imen cured in the same condition immersed in
carbonated water almost completely healed within
some days, with the formation of a white cubic
precipitate on the entire mortar surface. However,
Results on the influence of initial curing condi- dry cured specimens (C1) show an increasing trend
tions in healing product thickness while healing product
thickness is almost constant after 50 h for the other
specimens. One should also observe on Figure 4 b)
Initial curing conditions have an important impact
that crystals extend relatively far from the crack tra-
on self-healing potential. ‘Dry curing’ at 80 % RH
ducing the occurrence of the carbonation reaction
limits considerably self-healing phenomenon as al-
at a lower speed far from the crack.
most no precipitation occurred during the water
Immersion in carbonated water increases the
immersion stage (Figure 1). Observations were
speed of healing and also leads to a higher max-
prolonged during two weeks and no precipitation
imum healable crack width as illustrated by the
was observed while immersed in tap water. Sealed
Figure 5a which compares the evolution of the thick-
curing and water curing preserve the self-healing
ness of the calcite layer for specimen C1-H1 and
potential, explaining why this curing conditions
C1-H2. From Figure 5b representing the healing
have been extensively studied in previous research
percentage for these two specimens after 48 h, one
works. Thus, structures which are not immersed
can see that all healing percentage of specimen C1-
are not likely to exhibit a natural self-healing ability
H2 is higher than the one of C1-H1 after 48h. It is
and novels engineered techniques might be devel-
worth to note that the healing potential is almost
oped. Healing capabilities of specimens cured in
constant for the different crack widths monitored
carbonated water (H2) are higher than the ones of
in this study. However, the healing potential should
specimens cured in tap water (H1). This faster heal-
ing phenomenon could be attributed to the higher
amount of carbonate ions.
For specimens presenting a healing ability, the
evolution of crack width over time can be moni-
tored as illustrated in Figure 2. For a given loca-
tion, the crack width gradually decrease over time.
Affine or linear regression coefficients, depending if
there are some healed cracks or not, give an indica-
tion of the global healing trend and can be used as a
criterium to compare healing after different periods.
The coefficient of determination gradually decrease
with time which indicates the increasing of healing
variability as the crack heals at some locations and
does not heal at other locations as illustrated by
Figure 3. This increasing spatial variability could
be explained by the creation of preferential seed- Figure 3: Microscopic observation of the healing spa-
ing zones for the healing products which could be tial variability: specimen C1-H2 after 48h healing with
related to higher ionic concentrations. visible healing product bridges.

175
 Hilloulin et al.
Figure 4: Microscopic monitoring of the crack width
(µm) during healing for specimen C1-H1 subjected to
dry curing after 48 h immersion in tap water (a) or
C1-H2 in CO2 -enriched water leading to calcite crystals
precipitation (b).
gradually decrease for larger cracks with an initial
widths around 300 µm – 500 µm because of the
lower probability of such cracks to heal naturally.
Figure 5: Microscopic monitoring results for specimens
C1-H1 and C1-H2 after 48h healing: (a) self-healing
product thickness, (b) healing percentage.
176
Figure 6: 3D reconstruction of photonic images of a
healed specimen (a) and an unhealed specimen with a
crack width of around 90 µm (b). Void (blank) zones are
due non convergence in the investigated range of ± 200
µm from the surface.
Observations using 3D microscope
The use of 3D microscope helps the observation of
the healing product inside the crack as illustrated
in Figure 6a. On the other hand, it seems difficult to
get information on the crack lips geometry as the in-
vestigated range is smaller than the effective depth
of the crack (Figure 6b). Thus, a clear difference
can be made between a healed and an unhealed
specimen using 3D microscope.
3D microscopy reveals that the healing product
forms a continued layer inside the crack but with a
variable thickness (Figure 7). At most of the loca-
tions, the healing product exits from the crack and
can be easily detected. However, at some locations
where it cannot be observed using naked eyes or
2D microscope, the healing product is generally
present inside the crack at a depth of some dozens
of microns.
The microscopic observations reveal a lot of in-
formation on the healing process but they are lim-
ited to a qualification of the healing zone and to a
steady-state measurement. The modelling could be
add to these observations in order to quantify the
components into the healing zone and to follow the
healing kinetic.
IV. Modelling results
Influence of modelling parameters
CSH degradation is not well reproduced by the
thermodynamic database: CSH with a C/S ratio of
1.7 is directly degraded to amorphous silica which
leads to a considerable increase of porosity. The
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

calcite formation calculated by the model is almost

the same for a mortar with a portlandite content
of 20%, and a mortar with a volume fraction of
both 10% CSH and portlandite. The model repro-
duces the pH drop after complete leaching. pH
thus gradually decreases from a value close to 13
inside the mortar until the boundary value of 7.
The maximum volume fraction of calcite is also
produced in a zone where the pH is between 9
and 10. Considering the healing phenomenon is
rather fast, especially in CO2 enriched water, we
will only present lately modelling results where
CSH degradation is not taken into account.
Consideration of the influence of porosity on flow
leads to a localisation of calcite production as il-
lustrated in Figure 8 comparing calculated solid
profiles after 15 days of immersion in CO2 enriched
water. In some cases, this phenomenon participates
to the nullification of self-healing speed after some
days due to porosity drop (until volume fraction
Figure 8: Calcite, portlandite and porosity volume frac-
close to zero) as illustrated in Figure 9 where the
tions profiles after 15 days of immersion in CO2 -enriched
calcite thickness is calculated from the volume frac-
water (a) without taking into account the effect of poros-
tion and the cell thickness. Thus, the local porosity
ity on flow, (b) taking into consideration the effect of
decrease can explain the end of the healing phe-
porosity on flow. Negative depth corresponds to water.
nomenon while its slowdown is principally due to
the diffusion controlled process of calcium leaching
through the calcite layer and carbonated zone. Concrete cover porosity has a smaller effect and
mainly influences the localisation of the calcite
formed. Calcite seems to precipitate preferentially
Influence of curing conditions
in the higher porosity zone than in the water. Com-
From the parameters studied to represent the effect bined together, the lower portlandite quantity and
of curing conditions on self-healing potential, the higher porosity in the concrete cover lead to a signif-
portlandite content around the concrete surface has icantly reduced self-healing potential as observed
the most significant effect. Calcite amount formed in the experiments for C1 series.
near the surface is lower than for the original speci-
men. The maximum amount of calcite in the crack, Influence of healing conditions
i.e. self-healing potential is also affected (Figure
10). Therefore, the better the hydration at the con- Water flux into the crack decreases the amount of
crete surface, the higher the portlandite amount calcite formed in early hours, and even stop healing
and self-healing potential. by dissolution of calcite after some days for a flux

Figure 7: Surface topology of a healed specimen. Yellow and red colors denote that the healing product exits outside
the crack.

177
 Hilloulin et al.
Figure 9: Influence of the porosity drop inside the mortar at 0.5mm depth on the healing process, e.g. calcite content
in the water in contact with the mortar.
of 30 µm/s compared to a flux of 0.3 µm/s, as
illustrated in Figure 11. Thus, although this model
does not deal with the mechanical resistance of the
healing product against the water flux and does not
takes into consideration some seeding effects likely
responsible for the observed spatial variability, the
model demonstrates that water flux can accelerate
the dissolution of the healing product by water.
This could be a reason of the observed relatively
small healing ability of cracks exposed to water
pressure in water pipes or reservoirs (Meichsner
and Stelzner, 1989).
Carbonate concentration in water increases the
global speed of creation of calcium carbonate in
the water or the concrete cover and also increases
the self-healing potential. This result consists well
with the experimental observation of a better self-
healing potential when mortars are immersed in
water with a higher carbonate ions content. How-
ever, it appears in the model that, for lower carbon-
ate concentrations, calcite is preferentially formed
in the water rather than in the mortar. Then, the
correlation between numerical self-healing poten-
tial and experimental healing capacity might be
calculated averaging the numerical results around
Figure 10: Evolution of the calcite thickness formed in the crack for the two initial portlandite contents.
178
the first cells of water in contact with the mortar
and the first cells of the mortar.
V. Conclusions and perspectives
In this study, kinetics and potential of self-healing
were studied. Experimentally, it has been shown
that initial curing conditions have an important im-
pact on the self-healing kinetics and potential. A
new method consisting of immersing mortar speci-
mens in CO2 enriched water was proposed to heal
cracks that cannot heal in tap water, which could be
useful to quickly heal cracks of structures cured in
low humidity environments. A reactive transport
model was developed to simulate autogenic healing
by calcium carbonate precipitation, with the aim
to explain the influence of self-healing conditions
and qualify the self-healing potential. Experimental
observations have thus been explained, e.g. stop
of the phenomenon due to porosity decrease and
re-dissolution of healing product due to water flow
into the crack, putting self-healing phenomenon
at a cross-road between carbonation, leaching and
hydration studies. Although these first results seem
promising, more research is needed to experimen-
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 11: Influence of the water flux parallel to the crack on the calcite development.
tally describe the self-healing product and the as-
sociated concrete cover composition in order to
input more representative material parameters in
the model (porosity description, local hydration
product quantities) because of the nature of the
natural self-healing phenomenon, at the cross-road
between leaching and carbonation. Moreover, it
could be interesting to study the interest of wet-dry
cycles on self- healing.
Finally, the authors hope that, with the help of
reactive transport models, easy environmental con-
ditions for fast self-healing can be designed.
Acknowledgements
The authors gratefully acknowledge the financial
support provided for this study by GIS LIRGeC
and France’s Loire Valley Regional Council. Special
thanks are due to Michel Grasset and Laurent Gail-
let (MAST, IFSTTAR Nantes) for their collaboration
to perform 3D microscopic measurements.
References
Aldea, C., Song, W., Popovics, J., and Shah, S. (2000): “Extent of heal-
ing of cracked normal strength concrete”. Journal of Materials in Civil
Engineering. Vol. 12, No 1, 92–96.
Blanc, P., Lassin, A., Piantone, P., Azaroual, M., Jacquemet, N., Fabbri,
A., and Gaucher, E. (2012): “Thermoddem: A geochemical database fo-
cused on low temperature water/rock interactions and waste materials”.
Applied Geochemistry. Vol. 27, No 10, 2107–2116.
Clear, C. A. (1985): “The effect of autogenous healing upon leakage of
water through cracks in concrete”. Technical report, Cement and Concrete
Association.
de Rooij, M., Van Tittelboom, K., De Belie, N., and Schlangen (Eds),
E. (2013): “Self-Healing Phenomena in Cement-Based Materials”. Springer
Netherlands.
Dry, C. and McMillan, W. (1996): “Three-part methylmethacrylate
adhesive system as an internal delivery system for smart responsive con-
crete”. Smart Materials and Structures. Vol. 5, No 3, 297.
Ducasse-Lapeyrusse, J. (2014): “Étude du potentiel d’autocicatrisation et
de biocicatrisation de matériaux cimentaires fissurés” (in French). PhD thesis,
Ecole Nationale Supérieure des Mines de Douai (France) and Université
de Sherbrooke (Canada).
Edvardsen, C. (1999): “Water permeability and autogenous healing
of cracks in concrete”. ACI Materials Journal. Vol. 96, No 4, 448–454.
Fukuda, D., Nara, Y., Hayashi, D., Ogawa, H., and Kaneko, K. (2013):
“Influence of fracture width on sealability in high-strength and ultra-low-
permeability concrete in seawater”. Materials. Vol. 6, No 7, 2578–2594.
Granger, S., Loukili, A., Pijaudier-Cabot, G., and Chanvillard, G.
(2007): “Experimental characterization of the self-healing of cracks in
an ultra high performance cementitious material: Mechanical tests and
acoustic emission analysis”. Cement and Concrete Research. Vol. 37, No 4,
519 – 527.
He, H., Guo, Z.-Q., Stroeven, P., Hu, J., and Stroeven, M. (2007a):
“Computer simulation study of concretes self-healing capacity due to
unhydrated cement nuclei in interfacial transition zones”. In Proceedings
of the First International Conference on Self Healing Materials. Noordwijk
aan Zee, The Netherlands,18-20 April 2007.
He, H., Guo, Z.-Q., Stroeven, P., Stroeven, M., and Sluys, L. (2007b):
“Self-healing capacity of concrete - computer simulation study of unhy-
drated cement structure”. Image Analysis & Stereology. Vol. 26, 137–143.
Hilloulin, B., Grondin, F., Matallah, M., and Loukili, A. (2013): “Nu-
merical modelling of autogenous healing and recovery of mechanical
properties in ultra-high performance concrete”. In Proceedings of the
Fourth International Conference on Self-Healing Materials. 16-20 June, Ghent,
414–417.
Hilloulin, B., Grondin, F., Matallah, M., and Loukili, A. (2014a):
“Modelling of autogenous healing in ultra high performance concrete”.
Cement and Concrete Research. Vol. 61, 64 – 70.
Hilloulin, B., Zhang, Y., Abraham, O., Loukili, A., Grondin, F., Du-
rand, O., and Tournat, V. (2014b): “Small crack detection in cementitious
materials using nonlinear coda wave modulation”. NDT & E Interna-
tional. Vol. 68, 98–104.
Hilloulin, B., Van Tittelboom, K., Gruyaert, E., De Belie, N., and Louk-
ili, A. (2015): “Design of polymeric capsules for self-healing concrete”.
Cement and Concrete Composites. Vol. 55, 298–307.
Huang, H. and Ye, G. (2012): “Simulation of self-healing by further
hydration in cementitious materials”. Cement and Concrete Composites.
Vol. 34, No 4, 460–467.
Huang, H., Ye, G., and Damidot, D. (2013): “Characterization and
quantification of self-healing behaviors of microcracks due to further hy-
dration in cement paste”. Cement and Concrete Research. Vol. 52, 71–81.
Huang, H., Ye, G., and Damidot, D. (2014): “Effect of blast furnace
slag on self-healing of microcracks in cementitious materials”. Cement
and Concrete Research. Vol. 60, 68–82.
In, C.-W., Holland, R. B., Kim, J.-Y., Kurtis, K. E., Kahn, L. F., and Ja-
cobs, L. J. (2013): “Monitoring and evaluation of self-healing in concrete
using diffuse ultrasound”. NDT & E International. Vol. 57, 36–44.
Jacobsen, S. and Sellevold, E. J. (1996): “Self healing of high strength
concrete after deterioration by freeze/thaw”. Cement and Concrete Re-
search. Vol. 26, No 1, 55–62.
Li, V. C., Lim, Y. M., and Chan, Y.-W. (1998): “Feasibility study of a
passive smart self-healing cementitious composite”. Composites Part B:
Engineering. Vol. 29, No 6, 819–827.
Lv, Z. and Chen, H. (2012): “Modeling of self-healing efficiency for
cracks due to unhydrated cement nuclei in hardened cement paste”. Pro-
cedia Engineering. Vol. 27, 281–290.
Lv, Z. and Chen, H. (2013): “Self-healing efficiency of unhydrated
cement nuclei for dome-like crack mode in cementitious materials”. Ma-
terials and Structures. Vol. 46, 1881-1892.
Meichsner, H. and Stelzner, S. (1989): “Ãœber die selbstdichtung von
risen”. Betontechnik. Vol. 10, 82–83.
Mihashi, H., Kaneko, Y., Nishiwaki, T., and Otsuka, K. (2000): “Fun-
damental study on development of intelligent concrete characterized by
self-healing capability for strength”. Transactions of the Japan Concrete In-
stitute. Vol. 22, 441–450.
Ranaivomanana, H., Verdier, J., Sellier, A., and Bourbon, X. (2013):
“Sealing process induced by carbonation of localized cracks in cementi-
tious materials”. Cement and Concrete Composites. Vol. 37, 37–46.
179
 Hilloulin et al.
Reinhardt, H.-W. and Jooss, M. (2003): “Permeability and self-healing
of cracked concrete as a function of temperature and crack width”. Ce-
ment and Concrete Research. Vol. 33, No 7, 981–985.
Ter Heide, N. (2005): “Crack healing in hydrating concrete”. Master’s
thesis, Delft University of Technology.
Ter Heide, N. and Schlangen, E. (2007): “Self healing of early age
cracks in concrete”. In Proceedings of the First international conference on
Self Healing Materials, Noordwijk aan Zee, The Netherlands, 18-20 April
2007.
Thiery, M. (2005) : “Modélisation de la carbonatation atmosphérique des
matériaux cimentaires Prise en compte des effets cinétiques et des modifications
microstructurales et hydriques”. PhD thesis, Ecole Nationale des Ponts et
Chaussées.
van Breugel, K. (2009): “Self-healing concepts in civil engineering for
sustainable solutions: Potential and constraints”. In Proceedings of the
Second International Conference on Self-Healing Materials, Chicago, USA, 28
june – 1 july.
Van Tittelboom, K., Belie, N. D., Van Loo, D., and Jacobs, P. (2011):
“Self-healing efficiency of cementitious materials containing tubular cap-
sules filled with healing agent”. Cement and Concrete Composites. Vol. 33,
No 4, 497–505.
180
Van Tittelboom, K., Gruyaert, E., Rahier, H., and De Belie, N. (2012):
“Influence of mix composition on the extent of autogenous crack healing
by continued hydration or calcium carbonate formation”. Construction
and Building Materials. Vol. 37, 349–359.
Vaughan, P. (1987): “Analysis of permeability reduction during flow
of heated, aqueous fluid through westerly granite”. Coupled Processes
Associated with Nuclear Waste Repositories. Vol. 22, No 4, 529–539.
Wang, J., Soens, H., Verstraete, W., and Belie, N. D. (2014): “Self-
healing concrete by use of microencapsulated bacterial spores”. Cement
and Concrete Research. Vol. 56, 139–152.
Wiktor, V. and Jonkers, H. M. (2011): “Quantification of crack-healing
in novel bacteria-based self-healing concrete”. Cement and Concrete Com-
posites. Vol. 33, No 7, 763–770.
Xu, T., Spycher, N., Sonnenthal, E., Zheng, L., and Pruess, K. (2012):
“Toughreact user’s guide: A simulation program for non-isothermal multiphase
reactive transport in variably saturated geologic media, version 2.0”. Technical
report, Lawrence Berkeley Laboratory.
Yang, Y., Lepech, M. D., Yang, E.-H., and Li, V. C. (2009): “Auto-
genous healing of engineered cementitious composites under wet-dry
cycles”. Cement and Concrete Research. Vol. 39, No 5, 382–390.
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Alkali-aggregate reactions

181
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

182
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Microscopic investigation of building

materials affected by alkali-silica reaction
Daniel Werner⇤ , André Gardei, Sebastian Simon and Birgit Meng
BAM Federal Institute for Materials Research and Testing, Division 7.1 ’Building Materials’, Berlin, Germany
⇤ Daniel.Werner@bam.de

Abstract

The alkali-silica reaction (ASR) is a harmful process, which can occur in concrete. As product of this chemical reaction
a swellable alkali-silica gel can exert pressure leading to cracking of the material. The damages induced by ASR can
differ in their magnitude, and to assign certainly that ASR is the main damage mechanism a microscopic investigation
is necessary. Therefore this article presents a study on concrete affected by ASR with the purpose of assessing the degree
of damage and evaluating various alkali-sensitive aggregates.
For texture analysis thin section petrography is the major tool. Optical microscopy offers the best way to analyse
ASR-affected concrete samples for damage classification by visualising the gel and the crack formation. The chemical
investigation of the gel enables a precise determination of the involved amounts of the major oxides SiO2 , CaO, Na2 O
and K2 O. The proportions of these components of the gel are decisive with regard to its ability to swell. For this analysis
scanning electron microscopy coupled with energy dispersive X-ray spectroscopy was utilised. Furthermore micro X-ray
fluorescence analysis showed elemental distributions, especially the depletion or enrichment of elements involved in
ASR.

Keywords: alkali-silica reaction; concrete; damage analysis; materials characterisation; preparation of thin sections

I. Introduction ponents) or external sources (de-icing salt or sea

Concrete is the most applied building material in
modern times. It is present in all kinds of struc-
tures and no other material seems to replace con-
crete as the most selected building material in the
near future. The prime advantages are the high
performance, the easy production and a facile pro-
cessability. The sustainability and durability are
important requirements to concrete. Even if con-
crete is significantly less susceptible against factors
promoting a damage, compared to other materials,
there are still mechanisms affecting it. In particular,
high-ways and hydraulic structures made of con-
crete worldwide show cases of alkali-silica reaction
(ASR) damages.
This article refers to experience at BAM Federal
Institute for Materials Research and Testing for sev-
eral decades to ASR research and damage assess-
ment. ASR is a severe damage mechanism that can
occur in concrete, when certain conditions related
to composition and exposition arise in a critical
combination.
The chemical reaction is based on the alkalis and
the silica in the concrete and was first identified by
Stanton in highway structures in California (Stan-
ton, 1940). Actually three components are needed
to initiate ASR: alkali-sensitive siliceous aggregates,
alkalis coming from internal (cement or other com-
water) and water. If all of these components are suf-
ficiently present an alkali-silica gel could be formed
(Hobbs, 1988). This gel itself needs additional space
when it is formed. Furthermore it is able to swell
in the presence of humidity, both processes gener-
ating an expansion pressure inside the concrete. As
a consequence of intensive expansion, cracks can
occur, which significantly lowers the strength of
the material. Parameters like amount and chemical
composition of the gel as well as the pattern and
width of cracks can help to classify the ASR damage
(Swamy, 1992).
The objective of this article is to demonstrate the
microscopic work on ASR affected concrete sam-
ples, carried out to assess the damaging process
or to evaluate various alkali-sensitive aggregates.
It will explain the approach of ASR investigation
from the macroscale to microscale, starting briefly
with the condition assessment, over treatment and
investigations of samples in the laboratory, to the
use of different microscopical and analytical tech-
niques. In the main part the characteristic features
for identifying ASR in concrete are presented by
using microscopic images. Especially the emphasis
lies on the role of the aggregate. Different kinds of
aggregates and varying damage patterns are shown
and explained. The formation and resulting char-
acteristics of the gel are discussed more detailed.

183
 Werner et al.
Figure 1: ASR affected concrete construction showing extensive cracks and the escape of the alkali-silica gel. As the gel
reacts with the atmosphere, a characteristic white carbonated efflorescence remains on the surface of the construction.
Especially the amount and properties of the gel are
known to deliver important information relating
to the damage process. Ongoing research is aimed
to reveal interrelationships between the chemical
composition of the gels and its damaging potential.
Examples for these investigations, carried out with
scanning electron microscopy coupled with energy
dispersive X-ray (SEM/EDX) spectroscopy and mi-
cro X-ray fluorescence (µ-XRF), are presented in the
last chapter.
II. Methodology
A common investigation due to ASR begins with
a broad assessment of the affected structure and
ends with a specific sample being studied micro-
scopically at high magnification. Structures suffer-
ing from ASR damages typically show extensive
macroscale cracks, even on the surface. An escape
of the alkali-silica gel out of the material is typical,
but it reacts quickly with the carbon dioxide from
the atmosphere. This causes a carbonated efflo-
rescence, which deposits as a characteristic white
pattern on the surface of the concrete (Figure 1).
In addition, expansion of single aggregates located
near the surface can result in pop-outs, which rep-
resent conical fragments leaving a sinkhole in the
concrete.
The distinct presence of cracks at the surface sim-
plifies the consideration, where to take samples out
of the concrete construction for further laboratory
tests. If an estimation of the risk for the further
damage development is needed, the storage of con-
crete samples under ASR provoking conditions (e.g.
slightly elevated temperature, increased alkali and
water supply) for a specific time is a frequently
184
used tool (Chatterji, 2005). All these accelerated
ASR tests are conceptualised to evaluate mainly
the ASR sensitivity of aggregates for classification
before using it in concrete production. As a re-
sult of ongoing ASR, the concrete sample expands,
which is observed by measuring length changes.
The length change is used to classify if the sample
is either sensitive due to ASR or not. Examples
for often used tests are the accelerated mortar bar
method ASTM C 1260 or the concrete prism test
ASTM C 1293. In Germany a special guideline for
the classification of aggregates related to ASR by
the ’German Committee for Reinforced Concrete’
specifies the slightly modified way of testing (DAf-
Stb, 2013).
The following detailed description of ASR in this
article is based on microscopic work. For this pur-
pose the combination of optical microscopy and
SEM/EDX offers a comprehensive method to ac-
quire the following essential information. Three
major aspects can be taken into consideration. First
of all the petrographic analysis offers the best way
to describe the present aggregates and moreover to
classify the ASR sensitivity of these aggregates. The
second focus lies on the visualisation of cracks and
the alkali-silica gel, even for microscale formations.
Finally the quantification of the chemical compo-
sition can be obtained by measuring the gel. All
aspects together contain important information for
the precise identification of ASR and a potential
risk for further damage.
To gain representative results, a special thin sec-
tion preparation is necessary. Normally ASR dam-
aged samples show a weak cohesion. Therefore the
preparation process is based on the solidification
of the sample by using a specific resin. In addition
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 2: ASR damaged concrete of a hydraulic structure. a) Macroscopic overview of a thin section in which the
opal in the centre shows strong ASR induced cracks spreading in different directions through the hardened cement
paste and the aggregates. b) Enlarged PPL-Image of the same sample showing massive alkali-silica gel accumulating
in a pore. c) Enlarged PPL-Image of the same sample showing alkali-silica gel filled cracks spreading from the big
aggregate at the right side into the hardened cement paste and through other aggregates.

this prevents the sample from further deteriora-
tion during the preparation steps. After drying at
moderate temperatures and a vacuum-pressure im-
pregnation the resin penetrates the whole sample
and after curing it obtains the sample material in a
fixed state. In this way the alkali-silica gel, which is
very sensitive to leaching and drying, is conserved
in the resin with as small as possible alteration. A
fluorescent coloured pigment is added to the resin,
which offers a unique way to discern those finer
resin-filled areas in detail by its colour and by irra-
diating with UV light. In consequence of the size
of the aggregates of the concrete, special dimen-
sioned thin sections in the order of 60 x 90 mm are
commonly produced to include a sufficiently large
area to analyse. Water is an involved factor of ASR.
The whole preparation process has to be carried out
completely anhydrous. The replacement of water
with petroleum prevents changes of the reaction
products and the leaching of the gel. For this pub-
lication optical microscopic images were acquired
by using the polarising microscope Zeiss Axioskop
40 and the stereo microscope Olympus SZX16 both
using transmitted illumination in plane-polarised
light (PPL) and cross-polarised light (XPL). In addi-
tion, both microscopes are equipped with UV light.
The chemical composition of the gel can be deter-
mined by using the SEM Leo Gemini 1530 VP and
the µ-XRF EDAX Eagle III. Those allow fast simulta-
neous EDX-multi-elemental analysis for mappings
and point analysis.
III. Asr induced damage in concrete
The presence of an alkali-silica gel is not always
resulting in a deterioration of the concrete. Basi-
cally it is decisive how much sensitive aggregate is
contained in the concrete and how it is distributed.
Furthermore, the reactivity of the used aggregates
is important and how much gel can possibly be
formed. The concrete could also offer enough pores
for the gel to spread out. This will inhibit or delay
an expansion pressure leading to cracking. The
chemical composition can vary leading to the gen-
eration of swellable or non-swellable gels. How-
ever, with the incorporation of water by the gel
it increases the probability of expansion and will
likely damage the concrete microstructure. The ab-
sorption of water is reversible, which results in a
recurring mechanism of wetting and drying. In-
deed the absence of gel cannot exclude ASR as the
main trigger for the damage. For example precip-

185
 Werner et al.

Figure 3: PPL-images of different ASR affected aggregates: a) Microcrystalline quartz shows a uniform reaction rim
due to dissolution of the silica at the whole surface. b) Gneiss shows crack progression following the boundaries of the
distorted quartz minerals.

itations could leach out all of the damaging gel
remaining just the unfilled cracks in the building
material. Therefore the microscale view is even
more important to possibly see a few tiny remnants
of the alkali-silica gel.
The main ASR induced damages in concrete are
attributed to cracks eminating from aggregates.
These can occur through all components of the
concrete, through the hardened cement paste or
through the aggregates as well as along the inter-
face of them. Those typical patterns are mainly
influenced by the mineralogy of the aggregates. It
can take many years until the ASR damages can
be observed on a concrete construction (Blight and
Alexander, 2011). Structures close to sea are sub-
jected to additional alkali ingress of the salt water.
Also de-icing salts used for highways supply al-
kalis to the concrete. An ASR damaged sample
of a hydraulic structure is indicated in Figure 2a.
This macroscopic overview of a thin section shows
the widespread deterioration due to cracks, but the
alkali-silica gel is not clear identifiable. To certify
that ASR is the trigger of the cracks, an enlarged
microscopic view is necessary. Figure 2b and Fig-
ure 2c refer to the gel accumulation inside pores
and cracks by using optical microscopy.
Role of the aggregate
In all cases the ASR damage depends on the pres-
ence of alkali-sensitive siliceous aggregates, which
is in turn connected to the mineralogical structure
of the aggregate. The crystallinity of an aggregate
is of essential importance (Dron and Brivot, 1992).
ASR on highly crystalline aggregates made primar-
ily of silica just takes place at the exterior surface
of the aggregate. This process is slow and dissolves
little amounts of silicate ions. Therefore those aggre-
gates (e.g. quartz) are approximately inert (Broek-
mans, 2004). In comparison to that, amorphous or
poorly crystalline forms of silica (e.g. opal or glass)
are prone to attack by hydroxyl and alkali ions into
its interior (Dent Glasser and Kataoka, 1981). Espe-
cially for porous aggregates the hydroxyl ions can
penetrate easily, which increases the ASR intensity
compared to denser aggregates. In general aggre-
gates consisting of micro and crypto crystalline
quartz varieties afford more surfaces for ASR and
hence show higher sensitivity.
Besides different mineralogical structures, var-

Figure 4: PPL-images of the cracking patterns of different aggregates: a) Aggregates, containing fine grained quartz,
show cracks through the whole particle due to ASR. b) Siliceous rock shows an ASR induced crack penetrating straight
through the highly crystalline quartz veins.

186
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 5: Microscopic images of a completely with alkali-silica gel filled pore. Amorphous and crystalline parts are
visible. a) PPL-image. b) XPL-image.

ious aggregates differ in the appearance of their
damages (Ferraris, 1995). The dissolution and crack
initiating of an aggregate depends on the position
of the most weak silica parts. For an aggregate built
up just by amorphous SiO2 the entire particle can
dissolve and will show a uniform reaction zone on
the whole surface (Figure 3a). In comparison to that
a polymineral aggregate (e.g. granite or greywacke)
can consist of pure silica (mainly quartz) with dif-
fering crystal structures and other minerals like
mica, feldspar or pyroxene. This specifies the path
of cracks inside polymineral aggregates along the
most sensitive silica phases. As indicated in Fig-
ure 3b, one gneiss aggregate, containing thoroughly
distorted quartz due to tectonic distortions (often
called stress quartz), shows a crack accumulating
and branching out at these areas. Aggregates con-
sisting of fine grained quartz (e.g. flint or chert)
typically show a uniform texture of interlocking
grains. These kinds of aggregates show single and
almost linear cracks (Figure 4a). Siliceous rocks, as
shown in Figure 4b, may contain a lot of different
sized and oriented quartz veins. In this example
the ASR induced crack penetrates the aggregate
regardless of the orientation of those veins. That
is because the quartz veins represent no distorted
silica phases and thus don’t affect the propagation
of the crack.
Texture characteristics of the alkali-silica
gel
Under PPL the amorphous alkali-silica gel appears
mainly transparent (white) and often shows a char-
acteristic mud-cracked structure because of shrink-
age effects due to desiccation (Figure 5a). This
could either originate during the life time of the
concrete construction or caused from the sample
preparation. The gel is isotropic, showing full ex-
tinction under XPL (Figure 5b). Usually it fills out
the available space, already existing cracks and the
pores. When there is no further space for the gel to
spread out, a pressure on the surrounding material
is applied and with increasing pressure cracking
will probably be enhanced. Hence the air pore con-
tent in concrete is an important parameter, as it

Figure 6: a) PPL-image of an ASR induced crack through a quartzitic rock fragment showing alkali-silica gel
accumulation at the endings of the crack. b) PPL-image of another quartzitic aggregate showing alkali-silica gel
deposition completely along the crack walls.

187
 Werner et al.

Table 1: General chemical composition of alkali-silica gels and certain ratios to develop a swelling pressure in wt%.

SiO2 CaO Na2 O K2 O

General composition(Helmuth et al., 1992) 40 - 83 1 - 48 0 - 23 0 - 24
Necessary composition for swelling(Kawamura and Iwahori, 2004) Both alkalis not more than 25
Necessary composition for swelling(Mansfeld, 2008) 5 - 30

offers more space for the gel.
The alkali-silica gel in cracks or air voids is not al-
ways a pure gel, since crystallisation can take place.
Areas of much gel often show a mix of amorphous
and crystalline parts, where typical interference
colours from white to brownish can be visualised
using XPL (Figure 5b).
As already mentioned, ASR induced cracks can
be filled entirely (Fig 2c) or only partially with gel
or even no gel can be present. In Figure 6 two
quartzitic rock grains show different gel accumula-
tions. In the first case (Figure 6a) it is accumulated
just at both ends of the crack and in between it
is unfilled. This remaining ’unfilled’ area is filled
just with the blue resin based on the sample prepa-
ration. The other example (Figure 6b) shows an
aggregate in which the walls of the crack are con-
sistently filled with gel. The narrow unfilled range
is probably a result due the drying by the sample
preparation process.
rameters (e.g. source of alkalis, aggregates and
water), the sensitive aggregate for the gel formation
and the age of the gel. The proportion of these
ingredients affect the physical parameters (e.g. the
viscosity) and of course the microstructure. The
amount of alkalis and calcium is prior responsible
for the swelling capacity and thus for the potential
damage (Tab. 1). Above an alkali content of 25
wt% no great expansions are expected (Kawamura
and Iwahori, 2004). Additionally when the gel is
high in calcium no swelling occurs (Ferraris, 1995).
To evoke the swelling ability, the amount of CaO
has to be between 5 and 30 wt% (Mansfeld, 2008).
Without CaO no swelling pressure will be build-up
and too much CaO results in the formation of non
swellable C-S-H phases.
For a comprehensive investigation of ASR af-
fected structures the analysis of the chemical com-
position of the reaction products is advisable to
affirm if alkali-silica gel is present.

IV. The chemistry of alkali-silica gel
The alkali-silica gel mainly consists of SiO2 , CaO,
Na2 O, K2 O and H2 O. The amounts of all of these
oxides can considerably vary, as indicated in Ta-
ble 1. This depends primarily on the initial pa-
Analysing alkali-silica gel via SEM/EDX
SEM/EDX is the commonly used tool to detect
the quantitative chemical composition of an alkali-
silica gel. Two samples were chosen for the SEM-
investigation, which are shown by PPL- and XPL-

Figure 7: Amplified views of the two quartzitic aggregates as indicated in Figure 6 using PPL (a, c) and XPL (b, d)

188
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 8: a) BSE-image of the same quartzitic rock grain showing in Figure 7 (left). b) BSE image of the highlighted
area showing the position of EDX-analysis. c) EDX result of the chemical composition of the measurement spot. d) - f)
Same procedure using the quartzitic aggregate of Figure 7 (right) with focus on the layered structure of the alkali-silica
gel.

images in Figure 7. In these figures the presence
of amorphous and crystalline reaction products are
both clearly visible.
Figure 7 (left) shows a crack perpendicular to the
contact zone between the aggregate and the hard-
ened cement paste. Remarkably just the area close
to the contact zone contains gel, which reveals with
higher magnification a typical mud-cracked struc-
ture. Obviously this very small reaction zone and
low amounts of gel was sufficient to form the whole
crack. This can be explained by the dense structure
of the quartzitic grain. Figure 7 (right) shows an
ASR affected crack inside another quartzitic rock
grain. The gel in this example fills the entire crack
and shows a characteristic layered structure with
mud-cracks and a brownish colour. In both cases
gel is present next to crystalline reaction products.
By using SEM the Back-Scattered-Electron (BSE)
images contain information about the distribution
of different elements in the sample (Figure 8). Upon
a higher magnification the gel gets even better vi-
sualised (Figure 8b and 8e) and offers a very good

Figure 9: a) PPL-image of an ASR damaged opal sandstone. b) µ-XRF acquired RGB-image showing the element
distribution of potassium (red), calcium (green) and silicon (blue).

189
 Werner et al.
setting of the measurement points (spotsize of 5
µm resp. 2 µm). The results of the EDX analysis
are shown in Figure 8c and 8f. In both cases the
EDX-spectra show a characteristic alkali-silica gel
pattern with high silicon, calcium and potassium
peaks. Minor amounts of sodium are present in
both EDX-spectra. The carbon content in both sam-
ples is attributed to the resin used for the sample
preparation.
Analysing alkali-silica gel with µ-XRF
The use of µ-XRF offers another opportunity to gain
the chemical composition of ASR affected structures
and identifying the alkali-silica gel. Besides mea-
suring single spots, the elemental mapping of a
specific surface area is a very useful tool to anal-
yse for instance an entire aggregate. An elemental
mapping of an ASR affected region can offer infor-
mation about chemical changes in the vicinity of
the alkali-silica gel formation, e.g. a depletion of
calcium inside the hardened cement paste or silicon
inside the aggregate as well as an enrichment of
alkalis inside the crack.
One ASR affected sample is shown in Figure 9.
The aggregate shows heavy damages with several
cracks (Figure 9a). The µ-XRF-image, indicated in
Figure 9b, illustrates the elemental distribution of
this area (spotsize of 50 µm). By overlaying of three
elements RGB-images can be acquired, showing
potassium in red, calcium in green and silicon in
blue. This exemplifies the depletion or enrichment
of these three elements in detail. The concrete in
Figure 9 shows cracks with distinct potassium en-
richments at the contact zone between aggregate
and hardened cement paste. Especially at the crack
walls potassium accumulations are visible. The fur-
ther inside of the aggregate, the cracks contain less
of the alkali-silica gel, which is explicable with the
presence of the blue resin caused by the sample
preparation indicated in Figure 9a.
V. Conclusion and perspectives
This article shows that the combination of different
microscopic devices offers a suitable methodology
to evaluate the material properties and to deter-
mine the damage potential due to ASR. In the last
decades a huge number of studies were published
worldwide concerning ASR, but there are still ques-
tions e.g. about the exact reaction mechanism and
related to testing the ASR damaging potential of
aggregates and also concrete in the most reliable
and fast way.
The division ’Building Materials’ of the BAM
Federal Institute for Materials Research and Testing
carries on intensive work with ASR. In many cases
materials from potentially ASR affected buildings
190
are examined and evaluated. Beyond that ongoing
ASR research work is reflected to different aspects.
One finished PhD thesis considered the natural
pozzolana ’Rhenish trass’ and its effect on ASR in
concrete (Bürgisser, 2011). Additional topics among
others are the sampling and laboratory investiga-
tions of motorway sections with different ASR dam-
age classes, the effect of hydrophobizing of concrete
surfaces to reduce the ASR damage. Furthermore
the range of topics includes investigations of ASR
in structural concrete elements under simultaneous
dynamical loading and external alkali supply.
The PhD project of the main author of this arti-
cle is focused on the grain size dependence of the
damaging ASR potential. The aim is to study the re-
lationship between the aggregate properties (grain
sizes and geometry) of different aggregate types
and the ASR mechanism based on mineralogical
and microstructural characterisation of the reaction
products. In particular, differences of the amount
and the chemical composition of the reaction prod-
ucts, based on the chosen grain geometry, are the
major concern.
References
ASTM C 1260-07 (2007): " Standard Test Method for Potential Alkali
Reactivity of Aggregates (Mortar-Bar Method)." American Society for
Testing and Materials, 5 pages
ASTM C 1293-08b (2009): " Standard Test Method for Determination
of Length Change of Concrete Due to Alkali-Silica Reaction." American
Society for Testing and Materials, 7 pages
Blight, G.E.., Alexander, M.G. (2011): " Alkali-Aggregate Reaction and
Structural Damage to Concrete - Engineering Assessment, Repair and
Management." CRC Press, Taylor & Francis Group, London, UK, 234
pages
Broekmans, M.A.T.M. (2004): " Structural properties of quartz and
their potential role for ASR." Materials Characterization, Vol. 53, pp. 129
- 140
Bürgisser, P.G. (2011): " Der Einfluss von puzzolanischen Zusatzstof-
fen vulkanischen Ursprungs auf die Alkali-Kieselsäure-Reaktion (AKR)
im Beton." Freie Universität Berlin, 103 pages
Chatterji, S. (2005): " Chemistry of alkali-silica reaction and testing of
aggregates." Cement and Concrete Composites, Vol. 27, pp. 788 - 795
Deutscher Ausschuss für Stahlbeton (2013): " Vorbeugende Maßnah-
men gegen schädigende Alkalireaktion im Beton." German Committee
for Reinforced Concrete, 43 pages
Dent Glasser, L.S., Kataoka, N. (1981): " The Chemistry of Alkali-
Aggregate Reaction." Cement and Concrete Research, Vol. 11, pp. 1 -
9
Dron, R., Brivot, F. (1992): " Thermodynamic and kinetic approach
to the alkali-silica reaction. Part 1: Concepts." Cement and Concrete
Research, Vol. 42, pp. 941 - 948
Ferraris, C. F. (1995): " Alkali-Silica Reaction and High-Performance
Concrete." NISTIR 5742, pp. 1 - 20
Helmuth, R., Stark, D., Diamond, S., Moranville-Regourd, M. (1993):
" Alkali-Silica-Reactivity: an overview of Research." Strategic Highway
Research Program, 108 pages
Hobbs, D.W. (1988): " Alkali-silica reaction in concrete." Thomas
Telford, London, 183 pages
Kawamura, M., Iwahori, K. (2004): " ASR gel composition and expan-
sive pressure in mortars under restraint." Cement and Concrete Research,
Vol. 26, pp. 47 - 56
Mansfeld, T. (2008): " Das Quellverhalten von Alkalisilikatgelen unter
Beachtung ihrer Zusammensetzung." Bauhaus-Universität Weimar, 145
pages
Stanton, T.E. (1940): " Expansion of concrete through reaction be-
tween cement and aggregate." Proceedings American Society of Civil
Engineers, pp. 1781 - 1811
Swamy, R.N. (1992): " The alkali-silica reaction in concrete." Blackie
and Son Ltd, Glasgow and London, Van Nostrand Reinhold, New York,
336 pages
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Microscopic and chemical characterisation

of ASR induced by quartz-rich aggregates
Š. Šachlová⇤ , A. Kuchařová, Z. Pertold, R. Přikryl
Institute of Geochemistry, Mineralogy and Mineral Resources, Fac. of Science, Charles University in Prague, Prague, Czech Republic
⇤ s.sachlova@seznam.cz

Abstract

Various quartz-rich aggregates (pegmatite quartz, quartzite, quartz meta-greywacke, and chert) were subjected to an
accelerated mortar bar test (AMBT, following the ASTM C1260 standard) and a chemical test (CT, following the ASTM
C289 standard) with the aim to evaluate their potential for the alkali-silica reaction (ASR). After testing, the accelerated
mortar bars (AMBs) were analysed by scanning electron microscopy combined with energy dispersive spectrometer
(SEM-EDS).
Only two samples (quartz metagreywacke, and pegmatite quartz) were evaluated as being deleterious, according to
the standard classification of the CT; the other aggregates were classified as innocuous. The AMBT indicated different
results. Quartz metagreywacke, chert, and quartzite exceeded the 0.1% expansion limit after the 14-day test period.
Only pegmatite quartz indicated a lower value.
SEM-EDS analysis of the AMBs confirmed the presence of alkali-silica gels (ASGs) and ASR-microcracks. The most
extensive deterioration was found in those AMBs containing chert and quartz metagreywacke aggregates, in which the
microcracks formed an extensive network. Massive deposits of ASG were found as filling of pore voids, microcracks, and
aggregate/cement paste boundaries. The high ASR potential of chert and quartz metagreywacke was attributed to the
small grain size and the poor crystallinity. ASR microcracks, with no connection to grain boundaries, were the most
common feature observed in both quartzite and pegmatite quartz. These microcracks and microholes, without connection
to the grain boundaries, were several mum long, generally randomly oriented. Microscopic examination confirmed the
high (resp. middle) ASR potential of quartz metagreywacke and chert (resp. quartzite) well correlating with expansion
values of the AMBs. Only the pegmatite quartz seemed to be safe from ASR. No correlation was found with the results
of CT, which under-evaluated the ASR potential of two of the aggregates investigated.

Keywords: alkali-silica reaction, accelerated mortar bar test, electron microscopy

I. Introduction demand for laboratory equipment, resultant low

Petrographic examination of aggregate source rocks
by microscopic techniques, in combination with the
accelerated mortar bar test (AMBT) and/or con-
crete prism test (CPT), are the most widely used
methods to assess the potential of aggregates to
alkali-silica reaction (ASR) (e.g., Lindgård et al.,
2010). Such a methodical approach enables quan-
tification of the ASR potential of aggregates, based
on the expansion values of accelerated mortar bars
(AMBs) (resp. concrete prisms); also to identify
deleterious components of the aggregates, based on
their mineral composition and microstructure (e.g.,
Hewlett, 2003). Another group of methods is based
on the leaching of a specific grain size fraction of
aggregates in alkaline solution, and on the analy-
sis of the amount of Si4+ leached into the alkaline
solution combined with the reduction of alkalinity
of the solution (e.g., chemical test, CT, following
the ASTM C289 standard). The main advantage of
the CT is regarded to be: the short test period, low
cost of the method, and almost no influence of the
technician. Although some inaccuracies of CT have
been pointed out in recent studies (e.g., Rolim et al.,
2012), CT has been successfully used as a method
to indicate the amount of SiO2 soluble in alkaline
solutions (e.g., Hünger et al., 2012; Gao et al., 2013)
In the Czech Republic, the standard method for
ASR testing of aggregates (TP 137) requires all three
method types: petrographic analysis of aggregates,
AMBT (modified according to the ASTM C1260
standard), CPT (modified according to the RILEM
AAR2 standard), and CT (modified according to
the ASTM C289 standard). Recent research has
pointed out some inaccuracies of the CT when CT
was applied on volcanic rocks (Št’astná et al., 2013).
Five of seventeen investigated volcanic and/or vol-
canosedimentary rocks were classified as innocu-
ous, although they exhibited AMBs expansion val-
ues higher than 0.1% after 14 days. It was con-
cluded that the CT results were affected by weath-
ering and alteration, affecting the mineral compo-

191
 Šachlová et al.

Figure 1: Classification of aggregates according to the CT. Q0B - quartzite; Q3 - chert; Q7C - quartz meta-greywacke;
Q9A - pegmatite quartz. The limits between aggregates considered deleterious, potentially deleterious, and innocuous
were adopted from the ASTM C289 standard (indicated by the grey dashed line).

sition of the aggregates (Št’astná et al., 2013). The II. Aggregates

weathering caused increasing amounts of clay min-
erals, which increased reduction of the alkalinity
(e.g., Wakizaka, 2000). The current paper has as its
aim to evaluate the ASR potential of aggregates, em-
ploying all the methods required by TP 137 (2003).
The testing was performed on quartz-rich aggre-
gates exhibiting almost no signs of weathering or
alteration. With the aim to visualize and document
the origin of ASR, the standard AMBT was fulfilled
by microscopic analysis of AMBs after 1, 5, 9, and
14 days of testing.
An experimental study was performed using a set
of quartz-rich rocks exhibiting different microstruc-
tural and quartz deformation characteristics (Table
1). Two of the selected aggregates (quartzite and
quartz metagreywacke) have been used in the con-
struction industry as fillers for both concrete and
asphalt mixtures as well as in gravel basements (Ka-
menolomy ČR s.r.o., 2013, Persson, personal com-
munication). Another two samples were sampled
from a natural outcrop (chert) and in a quarry aban-
doned (pegmatite quartz, Lännstyrelssen, Väster-
norrland, 2015). From each of the samples, pieces
ranging from several cm to several dm were sam-

Figure 2: Expansion curves of AMBs. The grey line indicates the area of non-reactive samples (below the line) and the
area of reactive samples (above the line). Q0B - quartzite; Q3 - chert; Q7C - quartz meta-greywacke; Q9A - pegmatite
quartz.

192
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 1: Geologic and macroscopic descriptions of selected samples. Abbreviations: BCM - Baltoscandian continental
margin, BM - Bohemian Massif, CO - Caledonian Orogen, JN - Jämtlandian Nappes, MZ - Moldanubian Zone, PP -
Palaeoproterozoic, P - Proterozoic, S - sample, SO - Svecokarelian Orogen, TBU - Teplá-Barrandian Unit. Separation
of pegmatite quartz (Sample No. Q9A) was done mechanically, according to the macroscopic characteristics of other
minerals. Classification according to: (a) Gillespie and Styles (1999), Robertson (1999); (b) Passchier and Trouw
(2005). Mineral composition: (c) - mostly K-feldspar and plagioclase plus a minor content of mica; (d) - mostly
cryptocrystalline quartz and amorphous SiO2 plus a minor content (< 1 vol.%) of clay and mica minerals (complete
data on mineral compositions are detailed in Št’astná et al. 2015; and Kuchařová et al., submitted).

S. No. Q9A Q0B Q7C Q3

Locality Edsele Těchobuz Klövsjö Litohlavy
Geological unit, SO, PP BM, MZ CO, BCM, JN BM, TBU, P
age
Macroscopic char- white-to-light- grey colour, quasi- light white-to- dark grey with
acterisation yellow, quasi- isotropic (compact), yellow, quasi- red veins, veined,
isotropic (compact), undistinguishable isotropic (compact), undistinguishable
undistinguishable grain boundaries undistinguishable grain boundaries
grain boundaries grain boundaries
Petrographic pegmatite quartz quartzite quartz meta- chert
classificationa greywacke
Quartz content (vol. 100 71 74 35
%)
SiO2 -rich matrix - 2 26d 65 d

content (vol. %)
Other minerals (vol. - 27c / 0 - -
%)
Microstructure b the whole grains homeoblastic, inter- heteroblastic, heteroblastic,
impossible to be lobate grain bound- quartz grains par- quartz grains pref-
captured in polar- aries, undulose ex- tially float in the erentially float
izing microscope, tinction and origin matrix, partially in the matrix,
intruded by mi- of quartz subgrains exhibit interlobate partially exhibit
crocracks healed boundaries, weak interlobate bound-
by fine-grained undulose extinc- aries.
quartz, undulose tion, diagenetic
extinction alteration
Mean / max. grain > 20 / - 0.09 / 0.46 0.07 / 0.68 0.01 / 0.53
size (mm)
Index of grain size - 0.024 0.016 0.005
homogeneity

pled in total amounts from 30 - 50 kg.
III. Experimental methods
CT
A CT was performed, following the ASTM C289
standard (2007). Prior to performance of the test,
each of the samples were crushed to obtain an ag-
gregate of the desired granulometry. The aggre-
gates (25 g of each) were stored in Teflon bottles.
After adding 25 ml of 1M NaOH, the bottles were
tempered at 80 C for 24 hours. After rapid cooling,
the solution was separated. The reduction of alka-
linity was analyzed based on the titration procedure
described in the standard.
The amount of Si4+ dissolved into the solu-
tion was analyzed employing the ICP-OES method
in the Laboratories of the Geological Institute
(Charles University in Prague, O. Šebek opera-
tor). A ThermoScientific iCAP 6500 radially-viewed,
inductively-coupled plasma optical emission spec-
trometer (ICP OES, ThermoScientific, Cambridge,
UK) equipped with an ASX-520 auto sampler was
used for Si concentration measurements, using the
instrument parameters recommended by the man-
ufacturer. The ICP OES instrument was calibrated
using external calibration standard solutions. The
calibration standard solutions used in this study
were diluted from commercial stock solutions (1000
mg·L 1 , Analytika Praha, Czech Republic). Matrix
matched calibration was used for solution analy-
sis. The analytical protocol consisted of analysis
of an instrumental blank (2% (v/v) HNO3 ), a cal-

193
 Šachlová et al.
ibration solution, a procedural blank, samples, a
calibration check, and a quality control solution
(Certipur R , Merck, Germany; and Aristar R and
BDH Laboratory Supplies, UK) diluted to a concen-
tration of 5 mg L 1 . The differences between the
calibration and control solutions did not exceed 5%
relative. All the calibration and control solutions
were freshly prepared before the measurements,
and all the samples were measured immediately.
AMBT
An AMBT was performed, following the ASTM
C1260-07 standard (2007). Prior to performance
of the AMBT, each of the sampled materials were
crushed to obtain an aggregate of the desired gran-
ulometry. The mortar was then mixed using aggre-
gate (0.125/5 mm fraction), CEM I 42.5 R cement,
and distilled water in the ratio of 2.25/1/0.47 (ag-
gregate/cement/water). Four different AMBs were
prepared from each sample. After 24 hours of hard-
ening, plus 24 hours of tempering, the AMBs were
placed in 1M NaOH solution at 80 C. Expansion of
the AMBs were measured on 3 AMBs for fourteen
days through the test period. Every 1st , 5th , 9th ,
and 14th day of the testing, 1/4 of one AMB from
each sample was cut and used for the preparation
of polished sections.
Scanning electron microscopy combined with en-
ergy dispersive spectrometer
The polished sections (25⇥25⇥4 mm) were pre-
pared from AMBs after 1, 5, 9, and 14 days of
AMBT, and then subjected to microscopic analy-
sis employing a scanning electron microscope with
an energy dispersive spectrometer (SEM-EDS, con-
ducted at the Laboratory of Electron Microscopy
and Microanalysis, Charles University in Prague,
M. Racek operator). The AMB sections were coated
in a carbon atmosphere. The measurements were
performed on a Tescan Vega instrument with an
energy-dispersive analytical system (Oxford Instru-
ments LINK ISIS 300) under the following condi-
tions: 0.8 nA (for BSE, EDS); 120 s counting time;
and accelerating voltage of 15 kV. For the standard
quantitative calibration, a 53 Mineral Standard Set
#02753-AB (SPI Supplies) was used.
IV. Results
CT
Standard classification (according to ASTM C289)
indicated two deleterious aggregates (pegmatite
quartz, and quartz metagreywacke) and two innocu-
ous aggregates (quartzite, and chert). Both values
(reduction in alkalinity, and dissolved silica) of the
chert sample were found very close to the borders
194
between the area of the innocuous aggregates (resp.
deleterious, and potentially deleterious) (Figure 1).
Surprising results were found when comparing the
values of dissolved silica, which increased in the fol-
lowing order: quartzite < chert < pegmatite quartz
< quartz metagreywacke.
AMBT
ASR potential, according to the AMBT (ASTM
C1260), is designed for rapid assessments between
non-reactive (aggregates indicating < 0.100% expan-
sion of AMBs) and reactive (aggregates indicating
> 0.100% expansion of AMBs) types of aggregates.
Three of the investigated aggregates exceeded the
limit of 0.100% after the fourteen day test period
(quartzite, quartz metagreywacke, and chert). Only
one aggregate (pegmatite quartz) indicated a lower
value (Figure 2).
Microscopic analysis of AMBs
AMBs were subjected to microscopic analysis (em-
ploying SEM/EDS) after 1, 5, 9, and 14 days of
testing. The analysis confirmed a microstructure
typical for mortar specimens containing crushed
aggregates. Aggregates (mostly sharp edged) were
covered by cement paste. Pore voids (typically cir-
cular, with a 1.5 mm maximum diameter) were ran-
domly distributed throughout the AMB. Variable
lengths of the immersion in 1M NaOH were mainly
observed in the microstructure of the cement paste
and on the extension of ASR products.
ASR was documented according to the network
of microcracks penetrating the aggregates (Figure
3a), cement paste, aggregate/cement paste bound-
aries (Figure 3b), and according to the alkali-silica
gels (ASGs, Table 2). Microcracks in the cement
paste, pore voids, and aggregate boundaries were
50 mm wide, maximum. Microcracks penetrat-
ing aggregates varied in their lengths and widths,
mostly corresponding to the type of aggregate.
ASGs accumulated in the pore voids and micro-
cracks (Figure 3c).
V. Discussion
Time dependence of ASR
The initial stages of ASR were observed accord-
ing to the shape of the expansion curves after 1-2
days in AMB containing chert and quartz meta-
greywacke, and after 5-6 days in AMB contain-
ing quartzite and pegmatite quartz. This obser-
vation had a close correspondence to the results
of the microscopic analysis of AMB. After 1 day,
ASGs accumulated only in pore voids (Figure 4d).
Their presence in microcracks were rare. Microc-
racks penetrated mainly the cement paste and ag-
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 2: Selected EDS analysis of ASGs (in wt. %).

AMB No. Analysis No. Na2 O MgO Al2 O3 SiO2 K2 O CaO FeO Total
Q9AT Sp23 5.18 1.27 71.31 0.77 9.77 88.3
Q0BT Sp9 3.2 0.16 49.8 0.32 11.73 65.21
Q0BT Sp10 6.24 0.42 56.69 0.41 13.09 0.18 77.02
Q0BT Sp11 7.76 0.73 0.4 66.42 0.85 14.55 0.25 90.97
Q7CT Sp2 5.96 0.62 0.1 59.68 0.33 14.13 80.81
Q7CT Sp3 4.95 0.56 0.17 61.67 0.26 14.99 82.62
Q7CT Sp4 5.57 0.67 0.07 62.24 0.86 17.72 87.12
Q7CT Sp5 4.83 0.5 61.69 0.31 12.93 80.26
Q7CT Sp6 8.68 0.88 0.22 55.96 1.06 16.5 83.3
Q3T Sp12 1.35 1.41 0.81 49.69 0.55 14.63 0.25 68.69
Q3T Sp13 5.02 0.9 0.39 54.18 0.81 17.45 0.23 78.97
Q3T Sp15 3.95 0.8 0.07 45.64 1.16 15.76 67.37
Q3T Sp16 4.88 0.28 49.62 1.18 17.34 73.31

gregate/cement paste boundaries (Figure 4a, 4g).
Several microcracks penetrated aggregates. This
initial stage of ASR was observed, preferentially, in
AMB containing chert and quartz metagreywacke.
After 5 days of immersion in alkaline solution,
AMB containing chert and quartz metagreywacke
already exhibited extensive accumulations of ASGs,
as well as the dissolution of areas rich in cryptocrys-
talline and amorphous SiO2 (Figure 4a, 4h). On the
contrary, AMB containing quartzite and pegmatite
quartz still only exhibited the initial stage of ASR
(Figure 4b; microcracks penetrating cement paste
and aggregate/cement paste boundaries, rare accu-
mulations of ASGs).
After 14 days, if the immersion was in alkaline
solution, the most extensive network of microcracks
was found in AMBs containing chert and quartz
metagreywacke aggregates (Figure 4f, 4i) propagat-
ing through cement paste, aggregates, and pore
voids. The microcracks followed grain-to-grain
boundaries on the aggregate particles. Parts of the
aggregates were completely leached out, typically
in matrix-rich areas. Massive deposits of ASG were
found filling pore voids (Figure 4d), microcracks
(Figure 4f), and aggregate/cement paste bound-
aries. Less extensive deterioration was observed in
quartzite (Figure 3b). ASR microcracks penetrating
aggregates were rare. Microcracks and microholes
(several µm long, and generally randomly oriented)
were observed penetrating aggregate particles un-
connected to grain boundaries. ASGs (mixed with
cement paste) mostly filled pore voids (Figure 4c).
Factors affecting ASR potential of aggregates
The investigated samples exhibited different ASR
potentials. The highest ASR potential was indicated
by those aggregates containing a cryptocrystalline
matrix in significant volumes (25-65 vol.%, Figure
5b, 5c). The dissolution of the matrix originated on
grain boundaries, propagating through the entire
volume of the matrix. In contrast, parts of a chert
and metagreywacke sample, composed mainly of
well-crystallized quartz (up to 200 µm), exhibited
ASR mostly located along the grain boundaries.
The high ASR potential of chert and quartz meta-

Figure 3: ASR microcracks penetrating aggregates (part a), aggregate / cement paste boundaries (parts a-c), ASGs
(ASG) filling microcracks and pore voids (part c), and SiO2 dissolution (part c). AMB containing quartz meta-
greywacke (part a), quartzite (part b), and chert (part c) after 1 (part b) and 9 (parts a and c) days of immersion in
alkaline solution. SEM/EDS, BSE microphotographs.

195
 Šachlová et al.

Figure 4: ASR after 1 day (parts a, d, g), and 5 (parts b, e, h), and 9 (parts c, f, i) days of immersion in alkaline
solution in AMB containing pegmatite quartz (parts a-c), quartz metagreywacke (parts d-f), and chert (parts g-i). The
degree of SiO2 dissolution was observed on chert and quartz metagreywacke particles reflecting approximately the
length of immersion in the alkaline solution (parts e-f, resp. g-i). ASG-CP pore void partially filled with ASG mixed
with cement paste. SEM/EDS, BSE microphotographs.

greywacke was attributed to the very small grain VI. Conclusion

size and poor crystallinity of the matrix (e.g., Bektaş
et al., 2007).
Variable ASR potential of quartzite and pegmatite
quartz was attributed to quartz deformation and
recrystallization (Figure 5a, 5d). Undulose extinc-
tion and formation of quartz subgrains indicated
increased dislocation density in the quartz crystal
lattice, which can increase its ASR potential (e.g.,
Wigum, 1995; Wenk et al., 2008; Locatti et al., 2010).
The role of K-feldspar and plagioclase was more
complicated, forming up to 28 vol.% of one of the
aggregates tested. The possible dissolution of al-
kalis from the feldspars, and their possible role in
ASR, has been discussed previously (e.g., Lu et al.,
2006). Almost no ASR was observed in AMBs con-
taining pegmatite quartz. ASGs were only found
in a small portion of pore voids (Figure 4c) and
microcracks. ASR microcracks propagated mostly
along aggregate particle boundaries (Figure 4b, 4c).
Quartz-rich aggregates were tested for their poten-
tial to ASR employing AMBT, CT, and based on
petrographic analysis of the aggregates. Addition-
ally, AMBs were subjected to microscopic analysis
with the aim to visualize the ASR products.
The CT results indicated quartzite and chert sam-
ple to be innocuous; and pegmatite quartz and
quartz metagreywacke to be deleterious, due to
ASR. On the contrary, AMBT results indicated
quartzite, chert, and quartz metagreywacke to be
deleterious (exceeding the limit of 0.100% after four-
teen days); and only pegmatite quartz to be non-
reactive. Microscopic analysis of AMB showed the
initial stage of ASR in all AMB after 1-5 days of
immersion in alkaline solution. After 14 days of
immersion, the degree of ASR increased rapidly in
AMB containing quartz metagreywacke and chert.
The ASR produced an extensive network of microc-
racks penetrating cement paste, aggregates, and ag-

196
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 5: Microfabric of the quartz-rich rocks studied: quartzite (sample No. Q0B, a); chert (sample No. Q3, b); quartz
meta-greywacke (sample No. Q7C, c); pegmatite quartz (sample No. Q9A, d), as seen in a polarizing microscopy
(crossed nicols).
gregate/cement paste boundaries. Extensive signs
of SiO2 dissolution were found in chert and meta-
greywacke particles, mainly composed of a cryp-
tocrystalline matrix. Poor crystallinity and very
small grain size facilitated dissolution of these ar-
eas. Less extensive (resp. almost no) deterioration
in AMB had contained quartzite (resp. pegmatite
quartz). ASR microcracks propagated along aggre-
gate/cement paste boundaries. Aggregate particles
penetrated by microcracks were rare in these AMB,
comparing to those containing chert and quartz
metagreywacke.
Petrographic analysis of the aggregates con-
firmed cryptocrystalline quartz in significant
amounts in chert and quartz metagreywacke, which
was well correlating with their high ASR poten-
tial. The quartzite sample indicated the presence
of strained quartz; typically with undulose extinc-
tion and the origin of quartz subgrains, indicat-
ing deformation and recrystallization under middle
temperature regimes, which can increase the ASR
potential of the aggregate comparing to the non-
deformed quartz. Significant contents of K-feldspar
and plagioclase probably released alkalis and in-
creased the ASR potential of the aggregates. The
low (but detectable) ASR potential of pegmatite
quartz was explained by weak undulose extinction
and the origin of quartz subgrains.
The current study found a good correlation be-
tween the AMBT results and the microscopic anal-
ysis of AMB. On the other hand, no correlation
was found between AMBT and CT. The CT under-
evaluated the ASR potential of quartzite and chert;
and over-evaluated the ASR potential of pegmatite
quartz. The misleading results of the CT remained
unexplained. For the follow-on study, it is proposed
to correlate the results with a long term concrete
prism test, with the aim to obtain results corre-
sponding more closely with ASR origin in real con-
crete.
Acknowledgement
This study was financially supported from Research
Project P104/12/0915 provided by the Czech Sci-
ence Foundation. Special thanks belongs to Prof.
Karel Miskovsky from Envix Nord AB for his as-
sistance with the selection of Swedish samples. We
also thank O. Šebek for the analysis of alkaline
solutions employing ICP-OES.
References
ASTM C289 “Standard test method for potential alkali-silica reactiv-
ity of aggregates (simple chemical method)”. ASTM C-289. Philadelphia,
2007.
ASTM C1260-07 “Test method for potential alkali reactivity of aggre-
gates (Mortar-bar method)”. ASTM C1260, West Conshohocken, 04.02,
2011.
Bektas, F., Topal, T., Goncuoglu, M.C., Turanli, L. (2007): “Evaluation
of the alkali reactivity of cherts from Turkey”. Construction and Building
Materials, Vol. 22, 1183-1190.
Gao, X.X., Cyr, M., Multon, S., Sellier, A. (2013): “A comparison of
methods for chemical assessment of reactive silica in concrete aggregates
by selective dissolution.” Cement and Concrete Composites, Vol. 37, 82-94.
Gillespie, M.R., Styles, M.T. (1999): “BGS Rock Classification Scheme,
Volume 1, Classification of Igneous Rocks.” British Geological Survey; RR
99-06, 54 pp.
Hewlett, P.C. “Lea’s Chemistry of Cement and Concrete” Butterworth-
Heinemann; 4th ed., 2003.
197
 Šachlová et al.
Hünger, K.-J., Bachmann, R., Scholz, Y., Hübert, C., Mo, X. (2012):
“An accelerated chemical test method for ASR carried out directly at ag-
gregate grains.” In: Drimalas, T., Ideker, J.H., Fournier, B. (Eds.) Proceed-
ings of the 14th International Conference on Alkali-Aggregate Reaction, Austin,
USA, 10 pp.
Kamenolomy ČR s.r.o. http : //www.mineral
cesko.com/cz/provozovny lomy/morava
jih/contactid2 4/kamenolom techobuz.aspx (rev. 26.9.2013).
Kuchařová, A., Götze, J., Šachlová, Š., Pertold, Z., Přikryl, R. (submit-
ted): “Microscopic characterisation of crystalline and amorphous vari-
eties of SiO2 exhibiting different ASR potential.” Proceedings of the 37th
Annual ICMA Conference, Seattle, Washington, USA, May 3 – 7, 2015, 15
pp.
Lännstyrelssen Västernorrland.
http://www.lansstyrelsen.se/vasternorrland/Sv/djur-och-
natur/friluftsliv /geologivagen/solleftea-kommun/edsele-
gruva/Pages/default.aspx (rev. 10.2.2015).
Lindgård, J., Nixon, P.J., Borchers, I., Schouenborg, B., Wigum, B.J.,
Haugen, M., Åkesson, U. (2010): “The EU “PARTNER” Project — Euro-
pean standard tests to prevent alkali reactions in aggregates: final results
and recommendations.” Cement and Concrete Research, Vol. 40, No. 4, 611-
635.
Locati, F., Marfil, S., Baldo, E. (2010): “Effect of ductile deformation
of quartz-bearing rocks on the alkali-silica reaction.” Engineering Geology,
Vol. 116, Nos. 1-2, 117-128.
Lu, D., Zhou, X., Xu, Z., Lan, X.,Tang, M., Fournier B. (2006): “Evalu-
ation of laboratory test method for determining the potential alkali con-
tribution from aggregate and the ASR safety of the Three-Gorges dam
concrete.” Cement and Concrete Research, Vol. 36, No. 6, 1157-1165.
Passchier, C.W., Trouw, R.A.J. “Microtectonics” Springer Science &
Business Media, 2nd ed., 2005.
198
RILEM Recommended Test Method AAR-2 (2000): "Detection of po-
tential alkali-reactivity of aggregates — the ultra-accelerated mortar-bar
test." Materials and Structures, Vol. 33, No. 5, 283-289.
Road and Motorway Directorate of the Czech Republic. “TP 137 Elim-
ination of alkaline reaction of aggregates in concrete structures.” Road
and Motorway Directorate of the Czech Republic. Road and Motorway
Directorate of the Czech Republic technical regulations (in Czech), 2003.
Robertson, S. (1999): “BGS Rock Classification Scheme, Volume 2,
Classification of Metamorphic Rocks.” British Geological Survey, RR 99-02,
26 pp.
Rolim, P., Gleize, P., Hasparyk, N., Monteiro, P. (2012): “Chemical
method and the variations in the test for siliceous rocks.” In: Drimalas,
T., Ideker, J.H., Fournier, B. (Eds.) Proceedings of the 14th International
Conference on Alkali-Aggregate Reaction, Austin, USA, 10 pp.
Št’astná, A., Nekvasilová, Z., Přikryl, R., Šachlová, Š. (2013): “Micro-
scopic examination of volcanic rocks indicating ASR potential.” The 3rd
International Conference on Sustainable Construction Materials and Technolo-
gies, Kyoto, Japan, August 19th–21st 2013, 10 pp.
Št’astná, A., Šachlová, Š., Pertold, Z., Přikryl, R. (2015): “Factors af-
fecting alkali-reactivity of quartz-rich metamorphic rocks: Qualitative vs.
quantitative microscopy.” Engineering Geology, Vol. 187, 1-9.
Wakizaka, Y. (2000): “Alkali–silica reactivity of Japanese rocks.” Engi-
neering Geology, Vol. 56, Nos. 1–2, 211-221.
Wenk, H.R., Monteiro, P.J.M., Shomglin, K. (2008): “Relationship be-
tween aggregate microstructure and mortar expansion. A case study of
deformed granitic rocks from the Santa Rosa mylonite zone.” Journal of
Materials Science Vol. 43, No. 4, 1278-1285.
Wigum, B.J. (1995): “Examination of microstructural features of Nor-
wegian cataclastic rocks and their use for predicting alkali-reactivity in
concrete.” Engineering Geology, Vol. 40, Nos. 3-4, 195-214.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Deterioration of Santa Maria island

airport concrete pavement
Sara Medeiros⇤,a , Isabel Fernandesb , João Carlos Nunesa
a Universidade dos Açores, Departamento de Geociências.
b Centro de Geologia da Universidade do Porto/DGAOT-FCUP .
⇤ saramedeiros@uac.pt

Abstract

The alkali silica reaction (ASR) has been found in several concrete structures in which volcanic rocks were used as
aggregates. ASR caused by volcanic rocks has been studied in various countries among them Argentina, Iceland, Japan,
New Zealand and Turkey. The reaction occurs between alkali (Na+ and K+ ) and hydroxyl (OH ) ions in the cement
and certain reactive forms of silica in the aggregates. The reactivity of volcanic rocks is usually associated with the
presence of Si-rich volcanic glass (SiO2 >62-65%), altered minerals and the occurrence of some free silica forms.
Santa Maria is an island located in the southern part of the Azores archipelago in the North Atlantic Ocean. It is
formed by volcanic rocks of basic composition with intercalated sedimentary rocks. It is the oldest island of Azores with
an age of 8.12 Ma.
In order to characterize the reactivity of Azores volcanic aggregates a research project has been implemented through
the ReAVA project (Characterization of Potential Reactivity of the Volcanic Aggregates from the Azores Archipelago:
Implications on the Durability of Concrete Structures). This project comprises mainly the petrographic assessment of
the aggregates and concrete petrography of some concrete core samples. In Santa Maria Island a program was carried
out to produce thin sections from: (1) rock samples of two quarries of the island and (2) concrete core samples from the
local airport pavement. One of the quarries was exploited for the local airport construction.
Both rock samples showed significant alteration that was confirmed by the petrographic examination and by
geochemical analyses. The concrete thin sections showed that the sand fraction is mostly composed of volcanic glass
and that the cement paste has almost completely been replaced by a fibrous alkali-silica gel, as confirmed by scanning
electron microscopy.

Keywords: concrete, volcanic aggregates, alkali-silica-reactions, Azores

I. Introduction archipelago with an age of 8.12 Ma (Abdel-Monem,

Concrete is a widely used building material. It
can be applied in structures of different complex-
ity. The aggregates represent the major component
(about 3/4 in volume) playing an important role
in the quality of concrete and consequently in its
durability.
The Azores archipelago (Figure 1) is located in
the North Atlantic Ocean and consists of nine vol-
canic Islands. The aggregates used in Azores are es-
sentialy volcanic rocks of basic nature (e.g. basalts).
Volcanic rocks have been used widely for concrete
purposes in countries such as Argentina, Iceland,
Japan, New Zealand and Turkey. In these countries
several concrete structures appeared to be affected
by ASR due to the use of volcanic rocks as aggre-
gates. In Portugal, there is only one case reported
regarding the performance of volcanic aggregates
which is the study of the pavement of Santa Maria
Island airport (Braga Reis et al., 1996).
Santa Maria Island is the oldest island of the
1975). The study of Santa Maria Island aggregates is
integrated in a research project called ReAVA (Char-
acterization of Potential Reactivity of the Volcanic
Aggregates from the Azores Archipelago: Impli-
cations on the Durability of Concrete Structures).
This project includes: (1) petrographic study of the
aggregates, (2) assessment of their performance in
laboratory expansion tests and (3) site inspection
and core sampling of existing large structures in
the Azores.
This paper presents the results regarding the as-
sessment of the potential reactivity of Santa Maria
Island aggregates which included the petrography
examination of the aggregates and of the concrete
cores obtained from the pavement of the airport of
Santa Maria Island.
II. Materials
The rock samples were collected from two quarries
of Santa Maria Island. One of the quarries (Facho)

199
 Medeiros et al.
Figure 1: Azores archipelago.
is the main source of aggregates of the island. This
quarry is located inside a submarine eruptive centre
which emitted basaltic lava flows and pyroclasts.
The area of the quarry consists of: (1) a volcanic
neck 150 m high, (2) several submarine lava flows
with fractured and altered zones, and (3) interca-
lated levels of reddish hialoclastites (Figure 2a). The
other quarry (Brejo) is no longer active but in the
mid 1940s it was heavily exploited for the construc-
tion of the airport. The quarry consists of a thick
wall of a subaerial basaltic lava flow (Figure 2b).
The two quarries belong to the Facho-Pico Alto
complex of Pliocene age. This complex is formed by
submarine and subaerial lava flows. The rock sam-
ples were named SMA-SM1 and SMA-SM2 for the
first and second quarry, respectively. Four concrete
cores were drilled in Santa Maria airport pavement
in order to characterize the aggregates that were
used in the concrete and to identify manifestations
of ASR such as rims around the aggregate particles,
cracks and possible products of ASR.
III. Methods
The methods used included: (1) geochemical analy-
ses of both rocks (bulk rock chemical analysis) and
Figure 2: Quarries: a) Facho quarry; b) Brejo quarry.
200
(2) petrographic examination.
The geochemical analyses were made by the Act-
labs Laboratory, in Canada. The major oxides
were calculated according to the method of fusion-
inductively coupled plasma (FUS-ICP: Thermo Jar-
rell – Ash ENVIRO II ICP). The results were plotted
on a TAS diagram to classify the rock samples.
To prepare thin-sections, billets of the rock hand
samples and of concrete were cut and mounted
with araldite on a standard sized carrier glass. The
specimen were impregnated with epoxy resin to
reinforce the sample material, seal off porosity
and reduce preparation artefacts, then cured up
to ⇠40 C until set. Thin-sections of 30µm thickness
were lapped by hand using silicon carbide pow-
der, and finally polished with 0.25 µm diamond
grit. For the SEM/EDS analyses the thin-sections
were sputter-coated with carbon under vacuum in a
VG MICROTECH E6700/T800 instrument.The pet-
rographic study of the rock and concrete samples
was performed under an optical polarizing micro-
scope Nikon Eclipse E 400 POL, both in plane po-
larized light (PPL) and cross polarized light (XPL),
with automatic camera AXION cam MRC. The mi-
crophotographs were obtained by Axion Vision 3.1.
software. A scanning electron microscope (SEM)
FEI QUANTA 400 FEG ESEM/EDAX PAGASUS
X4M equipped with energy dispersive spectrometer
(EDS) was used to obtain images in backscattered
electron detector (BSED) mode and to determine
the qualitative composition of the reaction products.
The operation conditions were 15 kV and collection
time of 50 ses.
Site inspection of the airport pavement was per-
formed according to BCA (1992). Manifestations
of deterioration were registered and sites marked
for coring. The places selected were located on the
runaway pavement where cracks were visible. The
extracted cores were 100 mm in diameter and about
400 mm in length, crossing all the pavement slab
thickness.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Aggregate petrography: a) Sample SMA-SM1 with iddingsitized olivines and carbonate minerals filling the
crack (crossed polarized light - XPL); b) sample SMA-SM2 with thick iddingsitized olivines (plane-polarized light -
PPL).

IV. Results However, they allowed the detection of anhedral

Aggregates
The two rock samples were classified as basanites
according to the TAS diagram. The bulk geochemi-
cal analyses of both samples show a high value of
loss on ignition (LOI = 2.15% SMA-SM1 and LOI =
3.15% SMA-SM2) which suggests there is alteration
of these rocks. LOI is widely used as a measure of
volatile content and alteration of rocks. Chemically
the more highly altered basalts are characterized
by relatively high LOI. The two basanites are very
similar in mineral composition and in textural fea-
tures. They are composed of phenocrysts of olivine,
clinopyroxene and rarely plagioclase (Figure 3a and
3b). The matrix is formed mostly by pyroxenes, pla-
gioclase and opaque minerals with much less of
olivines. The olivines show: (1) thick iddingzitized
rims, (2) sometimes appear with skeletal forms and
(3) inclusions of spinel, oxides and some volcanic
glass. The clinopyroxenes show mostly hourglass
structures or zoning. In sample SMA-SM1 carbon-
ate minerals and zeolites were identified. The car-
bonate minerals are filling some of the cracks in
the rock (Figure 3a) and the zeolites occur in the
groundmass as irregular masses with very weak
birefringence.
The SEM/EDS analyses were merely qualitative.
intergranular crystals in several areas of sample
SMA-SM1 (Medeiros et al., 2012). The composition
and the habit of the crystals suggested they might
correspond to zeolites (Figure 4a and 4b).
Concrete
The concrete drilling was carried out only in places
where the concrete showed the most pronounced
cracking. A total of four concrete cores were taken
in different places of the airport pavement, one in
the middle of the runaway track and the other three
on opposite sides of the runaway pavement. The
cores that were sampled along to longer longitu-
dinal crack showed a white deposit covering the
crack walls when the concrete portions split. In
these broken surfaces, also the coarser aggregates
particles are coated by several white rims (Figure
5).
The concrete petrography showed that the aggre-
gates are crushed basalt. The sand fraction contains
mostly olivine, pyroxene, plagioclase crystals and
rock fragments in which volcanic glass is the main
component. In thin-sections the volcanic glass ap-
pears light-brown, yellowish or dark brown (Figure
6) in plane polarized light and always isotropic in
cross polarized light. In some particles, there are
clay minerals forming rims of short acicular yellow-

Figure 4: Aggregate petrography: a) SEM (BSED) image of fibrous zeolite (15.0 mm working distance); b) EDS
spectrum of the zeolite.

201
 Medeiros et al.
Figure 5: Concrete: a) core of 100 mm in diameter. It was drilled from the site where a long longitudinal crack was
observed. The coarse aggregates were rimmed by a whitish material. The crack surface is also covered by a whitish
powder; b) pavement of Santa Maria airport where the largest crack in the runaway pavement is shown.
ish crystals, with very low birefringence, around
and partially replacing the phenocrysts (Figure 7).
The interfaces between the aggregate particles and
the cement paste are frequently debonded. Others
are lined with a fibrous colourless to brownish ma-
terial with low interferences colours (grey to white)
which might correspond also to zeolite. The crys-
tals of this mineral develop perpendicular to the
interfaces. In other interfaces, the cracks are filled
with carbonate minerals (Figure 8).
The SEM/EDS analysis indicated that the glass
is composed mainly of Si and Al with low content
in Ca, Mg, Na and K (Figure 9). In some locations
volcanic glass is heavily altered to clay minerals
forming yellow rims. Several microcracks were
observed crossing either aggregates particles and
the cement paste.
The concrete petrography analysis under opti-
cal microscope also allowed the characterization of
the cement paste. It was verified that the cement
paste has been replaced/recrystallized forming an
agglomerate of fibrous minerals with radial texture
(Figure 10). The crystals formed were very light-
coloured to whitish in plane polarised light and
almost isotropic under cross polarised light.
Within the cement paste, also agglomerations of
fibrous minerals were observed, presumably replac-
ing particles of the sand fraction. Intense cracking
is observed. The SEM/EDS analysis of the cement
paste indicated a composition of mainly Si with
much lower contents of Al, Ca, Na and residual
Mg (Figure 11a and 11b), quite different from the
characteristic composition of a cement paste. Some
of the cracks, specially the ones that cross the ag-
gregate particles, were also filled with the acicular
crystals, as described above. The habit and the
qualitative composition obtained by SEM/EDS sug-
gested that the crystals might correspond to zeolites
(Si-Al-Ca) (Figure 11c).
202
V. Discussion
The reactivity of basaltic rocks is normaly related
to the presence of volcanic glass, altered minerals,
SiO2 content of the bulk rock (Korkanç and Tugrul,
2005) and pseudomorphs of quartz such as cristo-
balite and trydimite (Diamond, 2000). Regarding
alteration, volcanic glass tends to devitrify with
time and one of the alteration products are clay
minerals. Batic et al. (1994) studied basalts contain-
ing volcanic glass and clay minerals and concluded
that montmorilonite (or smectite) disseminated in
the groundmass was the principal material involved
in the ASR. Also Marfil et al. (1998) found out that
the expansion of basalts was due to the occurrence
of volcanic glass and its alteration products from
devitrification, such as argillaceous minerals and
poorly crystallized silica.
It is important to mention that andesitic glass
is considered not deleterious in general but it be-
comes deleterious with the increasing silica content
(> 62-65%) in the course of the crystallization of the
magma (e.g. dacitic and rhyolitic glass) (Katayama
et al., 1989). According to Katayama (2010) even
basalts can contain alkali reactive rhyolitic glass.
Many authors (Katayama et al., 1989; Korkanç and
Tuğrul, 2005; Katayama, 2010) associated dacitic
and rhyolitic glass with ASR.
The petrography of the Santa Maria rock sam-
ples and the SEM/EDS study showed that there
were no deleterious species such as cristobalite and
trydimite, opal or chalcedony. On the other hand
the concrete petrography showed that there is vol-
canic glass presented both in the coarse aggregates
and in the sand fraction. Although the analysis is
qualitative, the SEM/EDS study showed that the
volcanic glass has a composition of high silica con-
tent, which is the dominant component.
The study performed in 1990 by the National
Laboratory for Civil Engineering (LNEC) on the
pavement of Santa Maria Island airport revealed
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

that the deterioration of the concrete was due to a
complex process. This process involved expansive
reactions associated with montmorillonite and zeo-
lites (Braga Reis et al., 1996). According to Batic et
al. (1994) similar results were obtained in basaltic
rocks from Argentina in a study in which the NBRI
test method was used to identify the behaviour
of volcanic rocks. The petrographic examination
confirmed that the products of the reaction origi-
nated in the test were zeolites as the main material.
In the present study also zeolites seem to be the
most abundant product for the reaction affecting
the airport concrete pavement. According to the
conclusions in Batic et al. (1994) the minerals of al-
teration of the rocks studied might have contributed
to the reaction.

Figure 6: Concrete petrography: aggregate particles and cement paste (cp) are shown. The aggregate particles show a
high content of volcanic glass (vg) (PPL); a) light brown to yellowish; b) dark brown. In the interfaces there is a rim of
fibrous zeolite (ze).

Figure 7: Clay minerals partially replacing a crystal of pyroxene: a) PPL and b) XPL.

Figure 8: Concrete petrography: aggregate particles (agg) and cement paste (cp) are shown. The aggregate particles
show: a) a high content of brownish volcanic glass (PPL), b) isotropic (XPL). The crack observed between the cement
paste and the aggregate particle is lined with carbonate minerals (cc).

203
 Medeiros et al.

Figure 9: Volcanic glass in an aggregate particle observed in the concrete thin sections: a) volcanic glass under optical
microscope (PPL); b) SEM (BSED) image of the glass and crystallites (10.0 mm working distance); c) EDS spectrum
of the volcanic glass (point Z3).

VI. Conclusion
In order to evaluate the potential alkali reactivity of
the aggregates of Santa Maria Island, a petrography
study was carried out.
The petrographic examination of both basanites
showed that there were no forms of free silica. Both
rock samples had similar mineral composition and
textural features. The presence of olivine with thick
iddingzitized rims, the occurrence of clay minerals
and the high values of LOI indicated that the rocks
have some degree of alteration. The alteration of
these rocks could also be observed in situ in the
reddish coloration of some areas of both quarries.
The petrography of concrete samples identified
volcanic glass both in the coarse and in the sand
fraction. The SEM/EDS analysis indicated that
the yellowish volcanic glass seemed to have, on a
qualitative basis, a high content in Si.
Under the optical microscope the cement paste
showed abundant cracks with a typical pattern
known as map-cracking and abundant needle-
like/acicular crystals. This pattern could be as-
sociated to ASR with the formation of zeolite-like
crystals. The EDS analysis confirmed that the ce-
ment paste is composed mainly by Si with lower
contents of Al, Ca, Na and residual Mg. This com-
position is similar to the alkali silica gel described
in the literature of ASR and quite different from the
usual composition of the cement paste.
A number of aggregate particles are rimmed by
a whitish material that also replaces the smallest
particles of the sand fraction. According to the
SEM/EDS analysis the white material has a com-
position of Si-Al-Ca which might correspond to
zeolite.
Acknowledgement
The authors would like to acknowledge the fi-
nancial support of Azorean Regional Govern-
ment through the Regional Fund for Science
(FRC - Fundo Regional para a Ciência) for
the project ReAVA (ref. M5.2.2/I/05/2011).
The analyses by SEM/EDS were performed
at CEMUP which equipment was funded
by the projects REEQ/1062/CTM/2005 and
REDE/1512/RME/2005 of Portuguese Foundation
for Science and Technology (FCT). The authors also
acknowledge the contribution of airports of Por-
tugal (ANA – aeroportos de Portugal), specially
Santa Maria airport for allowing the access to the
pavement of the airport.

Figure 10: Cement paste replaced by a crystalline product composed of fibrous crystals showing a radial texture,
mostly isotropic: a) PPL and b) XPL.

204
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 11: Concrete petrography: a) SEM image of cracked cement paste and of fibrous material (10.0 mm working
distance); b) EDS spectrum of the cracked cement paste (point Z2) and c) EDS spectrum of the fibrous material (point
Z1).

References Katayama, T., John, S.T. & Futawaga, T. (1989): The petrographic
comparison of rocks from Japan and New Zealand Potential reactivity
related to interstitial glass and silica minerals. Proceedings of the 8th In-
Abdel-Monem, A.A., Fernandez, L.A., Boone G.A., (1975). K-Ar ages
ternational Conference on Alkali-Aggregate Reaction in Concrete (K. Okada, S.
from the Eastern Azores group (Santa Maria, São Miguel and Formigas
Nishibayashi and M. Kawamura (Eds)), Japan, pp. 537-542.
islands). Lithos 8; 247-254.
Korkanç, M. & Tuğrul, A. (2005): Evaluation of Selected Basalts from
Batic, O., Maiza, P. & Sota, J.(1994):Alkali-silica reactions in basaltic
the point of alkali-silica reactivity. Cement and Concrete Research, vol 35,
rocks. NBRI method, Cement and Concrete Research, vol 24/7, pp. 1317-
No 3, pp. 505-512.
1326.
Marfil, S.A., Maiza, P.J., Bengochea, A.L., Sota, J.D., Batic, O.R. (1998).
BCA, 1992, “The diagnosis of alkali-silica reaction.Report of a work-
Relationships between SiO2 , Al2 O3 , Fe2 O3 , CaO, K2 O, and expansion
ing party”, British Cement Association Publication 45.042, 44p.
in determination of the alkali reactivity of basaltic rocks. Cement and
Braga Reis, M.O., Silva, H.S., and Silva, A.S. (1996): Ocorrência de
Concrete Research (28): 189-196.
Reações Álcalis-Inerte em Portugal. Estudos de Casos. Atas do encon-
Medeiros S., Katayama T., Zanon V., Fernandes I., Silva A.S., Nunes
tro Nacional Betão Estrutural 1996, LNEC, Lisboa, 2.37-2.48. (in Por-
J.C., Miranda V., Ramos V. & Soares D. (2012): Assessment of the po-
tuguese).
tential alkali-reactivity of volcanic aggregates from Azores islands. Pro-
Diamond, S. (2000): Chemistry and other characteristics of ASR gels.
ceedings of the 12th International Conference on Alkali-Aggregate Reaction in
Proceedings of the 11th International Conference on Alkali–Aggregate Reaction,
Concrete (K. Okada, S. Nishibayashi and M. Kawamura (Eds)), USA, 10p.
Québec, Canada (Bérubé MA, Fournier B and Durand B (eds)). CRIB,
Université Laval, Québec, Canada, pp 31–40.
Katayama, T. (2010): Diagnosis of alkali-aggregate reaction – po-
larizing microscopy and SEM/EDS analysis. Proceedings of the 6th In-
ternational Conference on Concrete under Severe Conditions (CONSEC’10),
Merida, Mexico. CRC Press, United Kingdom, vol. 1, pp 19-34.

205
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

206
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Alkali reactive aggregates: the importance

of representative samples from the
quarries
Isabel Fernandesa,⇤ , Maria dos Anjos Ribeirob , Helena Coutob , António S. Silvac
a Department of Geology, Faculty of Sciences, University of Lisbon/ICT, Institute of Earth Sciences, University of Porto, Portugal
b Department of Geosciences, Environment and Spatial Planning/ICT, Institute of Earth Sciences, University of Porto, Porto, Portugal
c LNEC, Materials Department, National Laboratory for Civil Engineering, Lisbon, Portugal
⇤ mifernandes@fc.ul.pt

Abstract

Some Portuguese granitic rocks have shown to be alkali-reactive when used in concrete. In order to understand the
processes and components involved in the reaction, three quarries were selected from more than 50 granitic samples
analysed in the scope of a research project. These quarries were mainly formed by medium to coarse-grained granitic
rocks, which showed expansive behaviour when tested in concrete according to RILEM AAR-4.1 test-method. The
petrographic analysis evidenced that besides the coarse quartz crystals with straight boundaries there were portions
of granitic rocks in each batch of aggregate containing microcrystalline quartz and/or strained quartz. The optical
microscopy analysis was complemented by the use of cathodoluminescence. The features presumably related to the
potential reactivity of these rocks are presented and the usefulness of complementary methods discussed.

Keywords: granitic aggregates, petrographic characterization, representativity

I. Introduction grains and a larger surface in contact with the pore

Alkali-silica reaction (ASR) is a deleterious inter-
nal chemical reaction which occurs in mortar and
concrete involving certain forms of silica, namely
opal, chalcedony, trydimite, cristobalite and Si-rich
volcanic glass present in the aggregates. Opal and
chalcedony are present in the so-called normally-
reactive rocks (Lindgård et al., 2010). Concrete man-
ufactured with aggregates containing these forms of
silica (e.g. chert, some types of sandstone, volcanic
rocks) exhibit cracks 5 to 20 years after the construc-
tion. However, there are other aggregates, desig-
nated as slowly-reactive (e.g. quartzite, cataclasite,
gneiss) with which concrete structures show mani-
festations of ASR at more than 15 to 20 years after
construction. These may contain microcrystalline,
cryptocrystalline quartz or quartz with variable de-
grees of deformation. For these rocks the reactivity
depends mainly on the textural and microstructural
characteristics rather than on the mineralogical com-
position and petrographic nomenclature (Wigum,
1995; Locati et al., 2010).
A lot of research has been developed regarding
the reactivity of rocks showing features of deforma-
tion such as subgraining, bulging, strain lamellae,
ribbons and stretched quartz, usually correspond-
ing to an increase of the specific surface of the
solution in concrete. In literature there are exam-
ples of the correlation between deformation fea-
tures and expansion test results such as in Kerrick
& Hooton (1992), Monteiro et al. (2001), Wenk et al.
(2008) and Locati et al. (2010). Locati et al. (2010)
refer that a mylonitised rock, affected by ductile
and brittle-ductile deformation with development
of cataclasis and pseudotachylytic veins, showed
an increase of about 97% in reactivity with respect
to the mylonitised sample, due to formation of sub-
grains, grain size reduction and the presence of
glassy material. On the other hand, Velasco-Torres
et al. (2010) studied granitic rocks reported to origi-
nate ASR in two concrete dams and concluded that
micro-cracks work as facilitators for the access of
the pore solution to the interior of the crystals of
quartz, explaining the reactivity of some granitic
rocks containing no microcrystalline quartz.
More than 50 granitic aggregates have been stud-
ied in the scope of the research project IMPROVE –
Improvement of performance of aggregates in the
inhibition of alkali-aggregate reactions in concrete,
with the aim to evaluate the potential reactivity
of different aggregate types and to define the best
test method for the identification of the Portuguese
aggregates reactivity. Bulk rock petrography was
performed, as well as accelerated mortar-bar and

207
 Fernandes et al.
two different concrete prism tests. It was concluded
that there was a better correlation between the re-
sults obtained by petrographic aggregate examina-
tion and the RILEM AAR-4.1 concrete prism test
(Ramos et al., 2013; Santos Silva et al., 2014).
Therefore, after the conclusion of the RILEM
AAR-4.1 concrete expansion test, thin sections from
two of the three tested concrete prisms were pro-
duced. The petrographic analysis showed that dif-
ferent features were observed in different particles
of the aggregate within the same thin section of
each concrete prism, revealing the heterogeneity of
the quarries. The study of the prisms highlighted
the importance of having representative sampling
of the quarries and also that the petrographic study
is a needful tool in the study of aggregates regard-
ing ASR. Besides, the microscopic manifestations of
ASR such as cracks and gel were signaled in order
to understand which crystals and features were in
the origin of ASR. In complement to the optical
concrete petrographic characterization, cathodolu-
minescence (CL) was applied to the same thin sec-
tions. CL studies revealed the boundaries of the
crystals and the opened intergranular spaces be-
tween crystals which facilitate the percolation of
the interstitial fluids of the concrete. This method
showed also the proximity of feldspar crystals to
the sites where alkali-silica gel was identified.
This work aims at debating the importance of
the detailed lithological and structural study of the
quarries for the assessment of aggregates regarding
ASR. It also presents the contribution of CL for the
definition of the features of the rocks which seem
to be in the origin of ASR.
II. Materials and methods
The granitic aggregates were studied according
with the Portuguese Specification LNEC E 461
(2007) and RILEM AAR-0 (2015). Samples were
collected from the stock piles in each quarry. Due
to the need of having a total of about 70 kg of
samples to perform the laboratory expansion tests,
samples were picked up from piles corresponding
to the different grain size fractions available in the
quarries and which are used in concrete manufac-
ture. For petrography, larger blocks of the rocks
were collected.
In the case of GR-C, also the mapping of the
quarry was carried out in the scope of another re-
search project. In consequence of the geological
mapping, it was verified that there were veins of a
fine-grained granitic rock cutting the main granitic
mass. The fine-grained granite was also sampled as
rock block for the petrographic analysis. It is worth
mentioning that the fine-grained granite was ex-
ploited with the medium-grained dominant granite
and it was disseminated in the stock piles used for
208
concrete production and sampled for the laboratory
tests.
The petrographic analysis was the first method
to be used in the assessment of aggregates regard-
ing ASR, followed by laboratory expansion tests
on mortar bars and concrete prisms. For the ac-
celerated concrete prism method (RILEM AAR-4.1,
2015) three 75⇥75⇥250 mm concrete prisms were
prepared with fine and coarse aggregate of the same
granite and a CEM I type with an alkali content of
0.9% Na2 Oeq . The prisms were placed and sealed in
a metallic container over water, which were stored
in a climatic chamber generating constant temper-
ature of 60±2 C and relative humidity as close as
possible to 100%. Periodic length measurements
were made during twenty weeks. The limit con-
sidered of potential reactivity was of 0.03% at 20
weeks following the conclusion in Lindgård et al.
(2010).
Two sets of thin sections were produced. The
first set of thin sections was obtained from the sam-
ples of rock as received from the quarries. The
blocks were cut in three orthogonal directions to
produce the thin sections so that any anisotropy
that the rock might show could be observed. The
second set was composed of thin sections of the
concrete prisms of GR-A and GR-B. At the end of
the expansion tests, the prisms were cut with a dia-
mond saw in slices, impregnated with epoxy resin
to reinforce the sample material, seal off porosity
and reduce preparation artefacts, and lapped by
hand using silicon carbide powder to produce pol-
ished thin sections of 25⇥43⇥0.030 mm, using the
same methods as for the rock thin sections. Miner-
alogical and textural characteristics of the samples
were obtained by the analysis of these thin sections,
which were studied using a Nikon Eclipse E400POL
petrographic microscope in plane polarized light
(PPL) and crossed polarized light (XPL). Photomi-
crographs were obtained using a Zeiss MRc5 cam-
era and Axiovision software. CL studies were car-
ried out on the polished thin sections using a hot
cathode CL microscope HC3-LM from the Centre
of Geology of Faculty of Sciences of University of
Porto. The system was operated at 14 kV and a fila-
ment current of 0.18 mA. The samples were coated
with gold using Cressington 108Auto equipment
to prevent any build-up of electrical charge dur-
ing the CL analysis. Luminescence images were
acquired during the CL analysis with an adapted
digital video-camera (KAPPA PS 40C-285 (DX) with
dual stage Peltier cooling) and an acquisition time
between 5.51 and 8.02 sec.
III. Results
The quarries of granitic aggregates are located in
northern and central part of Portugal. In Table 1
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Table 1: Petrographic features of the aggregate samples as received from the quarries and results of the RILEM
AAR-4.1 expansion test.
Composition
Quartz, K-feldspar,(microcline),
plagioclase, biotite, muscovite, with
GR-A
minor sillimanite,apatite, titanite,
zircon and opaque minerals
Quartz, K-feldspar,(microcline and
orthoclase), plagioclase, biotite,
GR-B
muscovite, with minor apatite, zircon,
rutile, opaque minerals, chlorite, sillimanite and titanite
Sample 1
Quartz, K-feldspar (microcline and orthoclase),
plagioclase, biotite, muscovite,
GR-C
with minor chlorite, rutile,
apatite and zircon.
Sample 2
Quartz, K-feldspar, plagioclase,
muscovite and minor
biotite, opaque minerals,
zircon and chlorite.
a short description of the selected samples is pre-
sented, based on the thin sections produced from
the bulk rock samples. The results of the expansion
tests at 20 weeks performed according with RILEM
AAR-4.1 are also presented. The three granitic rocks
selected were all classified as potentially reactive
by the petrography and the concrete prism test per-
formed according to RILEM AAR-4.1 (2015). One
of the samples (GR-C) was also reported as reactive
in field structures.
GR-A is composed of medium grained two-mica
granite containing subhedral crystals of variable di-
mensions. The main volume of rock shows quartz
crystals with straight to slightly convex boundaries
(Figure 1). However, locally the quartz crystals
show sutured boundaries and some subgraining
in the borders of the larger crystals. K-feldspar
is very perthitic, plagioclase crystals are altered,
mainly in the cores. Myrmekites are abundant.
There are frequent symplectic fringes in the mus-
covite crystals which are usually smaller than the
plates of biotite. The larger crystals of quartz and
feldspar are strongly cracked. GR-B corresponds
Figure 1: In GR-A straight to sutured boundaries dominate. Undulatory extinction is observed in the larger crystals
of quartz; microcracks are abundant (NX).
Texture Expansion %
Heterogranular. Quartz
with straight to sutured boundaries. 0.09
Undulatory extinction.
Heterogranular. Fine to,medium-grained.
Quartz with straight to slightly irregular boundaries. 0.05
Undulatory extinction.
Sample 1
Heterogranular. Quartz crystals
with straight boundaries
0.04
and slight undulatory extinction.
Note: this aggregate
There is microcrystalline quartz
proved to be reactive in concrete
as inclusions and as myrmekites.
structures
Sample 2
Fine-grained.
Sutured boundaries of the quartz,
which is dominantly
microcrystalline.
to fine to medium-grained granite. Quartz shows
undulatory extinction but the crystals have straight
boundaries, locally slightly sutured. There is rare
microcrystalline quartz as inclusions in the crys-
tals of feldspar (Figure 2). K-feldspar is usually
altered and forms the larger crystals. Plagioclase
is mainly altered in the core of the crystals and
exhibits sericitization. The crystals of muscovite
are larger than the biotite crystals, which are more
elongated. Biotite is locally chloritised. GR-C is
composed of two facies which were observed at
the scale of the quarry. In the quarry the medium-
grained granite, with no manifestations of deforma-
tion, is dominant (Figure 3a and 3b). Locally, the
rock showed stronger alteration and higher content
of biotite. Quartz presents straight boundaries. The
larger crystals are of K-feldspar, usually showing
albitization and moderate alteration. Plagioclase
crystals are sericitized, mainly in the core. There
is another generation of smaller crystals of plagio-
clase with defined twinning and without signs of
alteration. Biotite and muscovite crystals occur as-
sociated to each other. The crystals of biotite are
209
 Fernandes et al.

Fernandes et al.
Figure 2: In GR-B the crystals of quartz show straight boundaries with no features of deformation. The smaller
crystals of quartz occur as inclusions in K-feldspar. The crystal boundaries are lined with iron oxides (NX).

smaller and slimmer than the muscovite. They contain inclusions of muscovite disposed in rims
are locally chloritised containing needles of rutile. within the quartz. In this granite quartz is mainly
Muscovite exhibits symplectitic fringes. Microcrys- microcrystalline (< 100 µm) both in the matrix and
talline quartz occurs as inclusions in feldspar crys- as inclusions in the larger crystals of K-feldspar.
tals and in abundant myrmekites. The rock shows The study of the concrete thin sections of GR-
some opened cracks. The fine-grained granite of A and GR-B allowed the observation of different
GR-C is composed of quartz, K-feldspar, plagioclase important aspects:
and muscovite, with rare biotite (Figure 3c and 3d).
The boundaries of the crystals are concave-convex, • The variability of the aggregates present in
commonly sutured. Feldspar crystals are slightly each sample;
altered. There are euhedral crystals of quartz which
Figure 2. In GR-B the crystals of quartz show straight boundaries • The identification
with no features ofof the manifestations
deformation. The smallerof ASR;
crystals of quartz occur as inclusions in K-feldspar. The crystal boundaries are lined with iron oxides (NX).

a) b)

c) d)
Figure 3. GR-C corresponds dominantly to medium-grained granite, in which quartz crystals are developed and
Figure 3:
showGR-C corresponds
no signs dominantly
of deformation to in
(a and b); medium-grained granite,granite
the vein of fine-grained in which quartz crystals
microcrystalline quartzare developed
is common, and show
both
no signsinof
thedeformation
matrix and as(a
inclusions
and b); ininthe
thefeldspars
vein of (cfine-grained
and d) (NX).granite microcrystalline quartz is common, both in the
matrix and as inclusions in the feldspars (c and d) (NX).

210
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 4: Deformation features in GR-A: intense subgraining and bended cleavage planes of the biotite; myrmekites
are abundant (NX).

• The connection between ASR signs and the
different features of the aggregate particles.
In the thin section of the concrete prism produced
with GR-A particles, besides the medium-grained
above described, fractions of deformed granite were
identified. The mineral composition is similar to
the one already described but there are different
textural features, namely abundant subgraining in
alignments along the boundaries of the larger crys-
tals. Microcrystalline (<100 µm) to cryptocrystalline
(<10 µm) quartz is quite common. Tiny inclusions
of quartz and myrmekites are abundant (Figure
4). Deformation is also visible in the bent cleavage
planes of the mica crystals.
The concrete prism thin section of GR-B con-
tains mainly particles of the granitic rock described
above. However, there are particles of coarse ag-
gregate with quite different characteristics. It cor-
responds to a tectonite of granitic composition in
which there are manifestations of strong deforma-
tion and moderate alteration, with plagioclase crys-
tals almost completely sericitized. There is pre-
ferred orientation and alignment of the minerals,
the larger crystals of quartz are cracked and show
strong undulatory extinction, mica crystals exhibit
flame-like planes of cleavage and strain lamellae
(Figure 5). The main characteristic in what concerns
alkali-reactivity is the subgraining and abundant
cryptocrystalline quartz along alignments that may
correspond to shear zones. Both deformed and not-
deformed aggregate particles are cracked, and the
boundaries of all the crystals are lined with iron
oxides.
Regarding the manifestations of ASR, in the thin
sections of GR-A and GR-B, rims of gel around
some coarse (> 4 mm) aggregate particles were
identified. Cracks occur cutting the aggregate par-
ticles. These cracks develop perpendicular to the
interface cement paste-aggregate particle and they
extend to the cement paste, containing scarce alkali-
silica gel. The cracks dominantly develop along
the intergranular spaces of the crystals of quartz.
The gel observed is colorless to brownish, without
a defined internal structure. Colorless gel forms
the rim in the interfaces between the cement paste
and the aggregate particles while the brownish gel
is observed filling the cracks that cut the aggregate
particles, close to the boundaries in contact with

Figure 5: Strongly strained granite in GR-B, showing bending of the cleavage planes of mica crystals. There is intense
subgraining of quartz, usually in alignments in the boundaries of the larger crystals (NX).

211
 Fernandes et al.
Fernandes et al.

a) b)
Figure 6. Manifestations of ASR in the concrete prisms. a) (GR-A) Rim of colorless gel around a coarse aggregate
Figure 6: Manifestations
particle; of ASR
b) (GR-B) cracks within
gelthe concrete
cutting prisms.particle
the aggregate a) (GR-A) Rim of to
and extending colorless gelpaste
the cement around
(N//).a coarse aggregate
particle; b) (GR-B) cracks with gel cutting the aggregate particle and extending to the cement paste (N//).

The sites where the gel and cracks occur were marked with an ink pen for observation using CL.
For each
the cement site,(Figure
paste microphotographs
6). were obtained under plane light (N//),
structures with crossed
graniticpolarsaggregates,
(NX) and CL.Portuguese
This
observation facilitated the identification
The sites where the gel and cracks occur were of the minerals and structures closely related to the
standard LNEC E 461 (2007) immediately classifies ASR signs. It
was concluded that:
marked with an ink pen for observation using CL. these rocks as Class II (potentially reactive) irre-
- Many rims of gel occur in the vicinity of feldspar and mica crystals (Figure 7);
For each site, microphotographs
- were obtained un-
There are usually cracks and some dissolution spective of of
thethe resultsparticles
aggregate of the petrographic
in the vicinity of analysis.
der plane light (N//), crossed polars (NX) and CL. Wigum
gel. Plagioclase crystals are cracked and altered to sericite (Figure 8). (1995) refers to inclusions, myrmekites and
This observation facilitated
Features the identification
like dissolution of theare emphasized
and alteration deformation lamellae
in the observation as possibly
under CL.involved
Also, thein ASR
minerals openand structures
boundaries andclosely related to the
open intergranular ASRcan bebut
spaces concluded
seen, enhancingthat the the
detailstotal grain
of the boundary area
petrographic
signs. Itstudy.
was concluded that: of quartz and the mean grain size are the quantita-
tive parameters related to the expansion of different
• Many4. rimsDiscussionof gel occur in the vicinity of
rock types. In fact, deformation processes result
feldspar and Themica threecrystals
samples(Figure
presented
7); in this study were selected because they were classified as
potentially reactive both by petrography and/or by thein grain AAR-4.1
RILEM size reductionconcreteand subgrain
prism expansion development
test.
• There
One are usually
of the graniticcracks andshowed
rocks also manifestations of ASR when used in a concrete structure. Although et al.,
some dissolution and seem to create also myrmekites (Menegon
of the
theprevious
aggregate petrographic
particles analysis performed
in the vicinity of on 2006).
the rock
gel. For low
as received fromreactive
the quarry rocks,
was the unstable quartz
not conclusive
regarding potential reactivity, further
Plagioclase crystals are cracked and altered to petrographic in
study myrmekites
proved to be and
crucialtheir
to probably
understand low crystallinity
the results
obtained in the
sericite (Figure 8). laboratory tests. might play a role. Monteiro et al. (2001) also stud-
The occurrence of ASR with slow reactive rocks has been reported in other papers
ied rocks of similar composition but differing (Shayan et in
al., 2008;
Features Fernandes etand
like dissolution al., alteration
2004; Fernandes,
are em-2014)the and is usually explained by the presence of
deformation state and grain size distribution
phasized microcrystalline
in the observationand/or strained
underquartz. However,
CL. Also, the the cause of reactivity of granitic rocks is not clear yet
and therefore, due to the occurrence of ASR in concrete and concluded
structures with that there
granitic is a correlation
aggregates, Portuguese between
open boundaries and open intergranular spaces can
standard LNEC E 461 (2007) immediately classifies these rocks as Class II (potentially reactive) with
expansion and the development of foliation
be seen, enhancing
irrespective of the
the details
results ofofthe
thepetrographic
petrographic analysis.reduction
Wigum (1995) of grain size.
refers to The authors
inclusions, also stated that
myrmekites
study. and deformation lamellae as possibly involved in ASR but the concluded
preferredthat orientation of biotite
the total grain boundaryandareathe of
foliation
quartz and the mean grain size are the quantitative parameters providerelated
a pathwayto theforexpansion of different
the percolation ofrock
interstitial
types. In fact,
IV. deformation
Discussion processes result in grain sizeconcrete
reductionfluids.
and subgrain development
The preferred and seem of
orientation to biotite
create also myrmekites (Menegon et al., 2006). For lowwas reactive rocks, the unstable quartz
later mentioned by Wenk et al. (2008) who in myrmekites
and their probably low crystallinity might play a role. Monteiro et al. (2001) also studied rocks of similar
The three samples presented in this study were also referred the increase of dislocation
composition but differing in the deformation state and grain size distribution and concluded that there density
is a in
selectedcorrelation
because between
they were classified as potentially quartz with deformation. The
expansion and the development of foliation with reduction of grain size. The authors authors concluded
reactivealso
both by petrography
stated that the preferredand/or by theofRILEM
orientation thatfoliation
biotite and the there is a linear
provide relationship
a pathway between the tex-
for the percolation
AAR-4.1 of concrete
interstitialprism
concreteexpansion
fluids. Thetest. One oforientation
preferred the ture
of of biotite
biotite wasand laterreactivity.
mentionedŠachlová
by Wenk(2013) et al. found
granitic rocks also showed manifestations of ASR that decreased grain size and elongated grain shape
when used in a concrete structure. Although the increase the specific surface and the ASR and Mari-
previous petrographic analysis performed on the noni & Broekmans (2013), by studying sedimentary
rock as received from the quarry was not conclusive rocks, concluded that large microstrain and small
regarding potential reactivity, further petrographic crystallite size values imply high alkali-reactivity
study proved to be crucial to understand the results potential, confirming the previous observation of
obtained in the laboratory tests. McNally et al. (2004) that domain size alone is not
The occurrence of ASR with slow reactive rocks enough to reliably predict expansion behaviour.
has been reported in other papers (Shayan et al.,
2008; Fernandes et al., 2004; Fernandes, 2014) and In the samples studied, the reduction in the size
is usually explained by the presence of microcrys- of the crystals of quartz was evident due to defor-
talline and/or strained quartz. However, the cause mation (GR-A and GR-B) and to the occurrence
of reactivity of granitic rocks is not clear yet and of fine-grained granite. The presence of microcrys-
therefore, due to the occurrence of ASR in concrete talline quartz (in GR-C) and of micro- to cryptocrys-

212
 summarized above. In fact, GR-A is the sample which presents stronger deformation manifestations in
more abundant particles, while in GR-B the cataclasite was just detected in the concrete prism sample.
15thRegarding
Euroseminar on no
GR-C, Microscopy Applied
cryptocrystalline to Building
quartz Materials
was identified 17-19
but •just June 2015
unstrained • Delft, The Netherlands
microcrystalline quartz in
the fine-grained facies.

gel
ms

pl

a) b)

gel

Kfs
qz

pl
c) d)
Figure 7. GR-A. Above - Rim of alkali-silica gel in the vicinity of a crystal of plagioclase (pl) showing intense
Figure 7:cracking
GR-A. and
Above
some- Rim of alkali-silica
alteration. The darkergel in thebetween
mineral vicinitytheof feldspar
a crystaland
of the
plagioclase
rim of gel(pl) showing intense
is muscovite (ms). a) cracking
and somePhotomicrograph
alteration. Theunder
darker mineral between
transmitted light (N//);the
b)feldspar and
CL photo. the rim
Below - Rimof gel is muscovite
of gel (ms). composed
around a particle a) Photomicrograph
of
quartz (qz), plagioclase (pl) and K-feldspar (Kfs). c) Photomicrograph transmitted light (N//).
under transmitted light (N//); b) CL photo. Below - Rim of gel around a particle composed of quartz (qz), plagioclased) CL photo.
Magnification of 5x.
(pl) and K-feldspar (Kfs). c) Photomicrograph under transmitted light (N//). d) CL photo. Magnification of 5x.

talline quartz in alignments occurs in some particles quartz in the fine-grained facies.
of the aggregates present in the stock piles prepared
All the samples showed microcracks crossing the
to be used in concrete. The relative content of these
larger crystals of the main minerals (quartz and
components is related to the values of expansion ob-
feldspar) which might have contributed to facilitate
tained (Table 1) as stated by other authors, as sum-
the access of the interstitial fluids to the inner of the
marized above. In fact, GR-A is the sample which
particles, in agreement with the findings by Velasco-
presents stronger deformation manifestations in
Torres et al. (2010). On the other hand, it has been
more abundant particles, while in GR-B the catacl-
confirmed that, at least in some particles, gel oc-
asite was just detected in the concrete prism sam-
curs close to the plagioclase or K-feldspar crystals,
ple. Regarding GR-C, no cryptocrystalline quartz
Fernandes etraising
was identified but just unstrained microcrystalline al. again the question about the possible con-
tribution of the feldspars to the ASR. According

qz
pl

gel

a) b)

Figure 8:Figure
GR-B.8. Rim
GR-B.ofRim
alkali-silica gel ingelthein vicinity
of alkali-silica of aofcrystal
the vicinity ofofplagioclase
a crystal plagioclase(pl)
(pl) and quartz(qz).
and quartz (qz).The
The rock shows
rock
shows intense cracking and some alteration. a) Photomicrograph under transmitted light (N//);
intense cracking and some alteration. a) Photomicrograph under transmitted light (N//); b) CL photo. Magnification b) CL photo.
Magnification of 5x.
of 5x.

All the samples showed microcracks crossing the larger crystals of the main minerals (quartz
and feldspar) which might have contributed to facilitate the access of the interstitial fluids to the inner of 213
the particles, in agreement with the findings by Velasco-Torres et al. (2010).
On the other hand, it has been confirmed that, at least in some particles, gel occurs close to the
 Fernandes et al.
to Velasco-Torres et al. (2010) this effect can be en-
hanced in granite containing myrmekite or perthite
structures, like the ones studied. The application of
CL in complement of other petrographic methods,
namely the optical microscope and the scanning
electron microscope, has been investigated by a
number of authors (e.g. Götze, 2009; Št’astná et al,
2012) and it is known to be a crucial methodology
in the identification of minerals and observation of
alteration and cracking. In the present study, CL al-
lowed the observation of the microstructure closely
related to the occurrence of alkali-silica gel, namely
the open intergranular spaces and the dissolution
around the cracks in the feldspar crystals.
V. Conclusion
The main conclusion of the work here presented is
that ASR is not only related with the presence of
microcrystalline and strained quartz but it seems
to be facilitated by the presence of microcracks.
The possible contribution of the feldspars and/or
muscovite can also be inferred by the petrographic
study of the concrete prisms. In the three quar-
ries, medium-grained granite is the dominant rock.
However, the production of concrete prisms thin
sections from GR-A and GR-B alerted to the fact
that different features can be found within the same
quarry. About 70 kg were sent to the laboratory
taken from the piles which are directly used for
the manufacture of the concrete, containing unde-
formed and deformed rocks. For the third quarry,
a geologist was involved in the geological mapping
and therefore the different facies were identified
at the scale of the quarry, allowing that a correct
representative sampling was made. In this case,
the potentially reactive rock occurs is limited areas.
During the quarrying operation, this fine-grained
rock is mixed with the innocuous rock in an un-
known content, generating ASR. The heterogeneity
of the quarries can give, or not, potential reactive
aggregates in consequence of the relative content
of the reactive fraction. Therefore, the main note
is that it is important to establish in the standards
the need of the involvement of geologists in the
mapping of quarries in special when there are sig-
nificant zones of deformed rock or fine-grained
facies. The performance of concrete petrography
allowed the identification of manifestations of ASR,
after the expansion tests have classified these ag-
gregates as Class II (potentially reactive to alkalis).
The use of CL helps in the definition of the features
of the particles likely to originate ASR.
Acknowledgement
The authors wish to acknowledge the Foundation
for Science and Technology (FCT), under project
214
IMPROVE PTDC/ECM/115486/2009, and LNEC
for the financial support. The authors affiliated in
the Center of Geology of University of Porto are
thankful to FCT for the funding in the scope of the
Strategic Project PEST-OE/CTE/UI0039/2014.
References
Fernandes, I., Noronha, F., Teles, M. (2004). Microscopic analysis of
alkali-aggregate reaction products in a 50-year-old concrete, Materials
Characterization, Vol. 53, Issues 2-4: 295-306.
Fernandes, I. (2014). Role of granitic aggregates in the deterioration of
a concrete dam. Bulletin of Engineering Geology and the Environment.
DOI: 10.1007/s10064-014-0590-z.
Götze J. (2009). Application of Nomarski DIC and cathodolumines-
cence (CL) microscopy to building materials, Materials Characterization,
vol. 60, pp 594-602.
Lindgård, J., Nixon, P.J., Borchers, I., Schouenborg, B., Wigum, B.J.,
Haugen, M., Åkesson, U. (2010). The EU "PARTNER" Project – European
standard tests to prevent alkali reactions in aggregates: final results and
recommendations. Cement and Concrete Research (40): 611-635.
LNEC E 461 (2007). Concrete. Methodologies to prevent internal
expansive reactions (Betões. Metodologias para prevenir reacções expan-
sivas internas). Laboratório Nacional de Engenharia Civil, Lisboa (in
Portuguese): 6 pp.
Locati, F., Marfil, S.A., Baldo, E. (2010). Effect of ductile deformation
of quartz-bearing rocks on the alkali-silica reaction. Engineering Geol-
ogy (116): 117-128.
Marinoni, N., Broekmans, B. (2012). Microstructure of selected ag-
gregate quartz by XRD, and a critical review of the crystallinity index.
Cement and Concrete Research (54): 215-225.
McNally, C., Richardson, M.G., Carr, A.J., Strogen, P. (2006). Domain
size as a parameter for studying the potential alkali–silica reactivity of
chert-bearing aggregates. Magazine of Concrete Research (56): 201–209
Menegon, L., Pennacchioni, G., Stunitz, H., 2006. Nucleation and
growth of myrmekite during ductile shear deformation in metagranites.
Journal of Metamorphic Geology (24): 553–568.
Monteiro, P.J.M., Shomglin, K., Wenk, H.R., Hasparyk, N.P. (2001).
Effect of aggregate deformation on alkali-silica reaction. ACI Materials
Journal, Technical Paper, March-April: 179-183.
Ramos, V. (2013). Characterization of the potential reactivity to al-
kalis of Portuguese aggregates for concrete. PhD Thesis on Geology,
University of Porto: 417 pp.
RILEM AAR-0 (2015). Outline guide to the use of RILEM Methods
in assessments of aggregates for AAR potential. In: 2013 Special Issue
RILEM TC219-ACS. Materials & Structures (46): (in press).
RILEM AAR-4.1 (2015). Detection of potential alkali-reactivity – 60o C
method for aggregate combinations using concrete prisms. In: Special
Issue RILEM TC219-ACS. Materials & Structures (46): (in press).
Šachlová, S. (2013). Microstructure parameters affecting alkali-silica
reactivity of aggregates. Construction and Building Materials (49): 604-
610.
Santos Silva, A., Soares, D., Fernandes, I., Custódio, J., Bettencourt
Ribeiro, A. (2014). Prevenção das reações álcalis-agregado (RAA) no
concreto – melhoria do monitoramento da reatividade aos álcalis de
agregados. In: Véras Ribeiro, D. (Coord.). Proceedings do 1o Encon-
tro Luso-Brasileiro de Degradação de Estruturas em Concreto Armado
(DEGRADA), Bahia, Brasil (in Portuguese): 293-306.
Shayan, A., Xu, A., Morris, H. (2008). Comparative study of the con-
crete prism test (CPT 60o C, 100% RH) and other accelerated tests. In:
Broekmans, M.A.T.M. and Wigum, B.J. (Eds). Proceedings of the 13th In-
ternational Conference on Alkali-Aggregate Reaction, Trondheim, Nor-
way: 391-400.
Št’astná A., Šachlová Š., Pertold Z., Přikryl R. & Leichmann J. (2012).
Cathodoluminescence microscopy and petrographic image analysis of
aggregates in concrete pavements affected by alkali-silica reaction, Mate-
rials Characterization, Springer, vol. 65, pp 115-125.
Velasco-Torres, A., Aejos, P., Soriano, J. (2010). Comparative study of
the alkali-silica reaction (ASR) in granitic aggregates. Estudios Geologi-
cos (66/1): 105–114.
Wenk, H.R., Monteiro, P.J., Shomglin, K. (2008). Relationship between
aggregate microstructure and mortar expansion. A case study of de-
formed granitic rocks from the Santa Rosa mylonite zone. Journal of
Materials Science (43): 1278–1285.
Wigum, B.J. (1995). Examination of microstructural features of Nor-
wegian cataclastic rocks and their use for predicting alkali-reactivity in
concrete. Engineering Geology (40): 195-214.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Evaluation of potential alkali reactivity of

volcanic rocks from northern Aegean
region of Turkey
Isabel Fernandesa,⇤ , Özge Andİç-Çakirb
a Department of Geology of University of Lisbon/Institute of Earth Sciences, Portugal,
b Civil Engineering Department, Ege University, İzmir, Turkey
⇤ mifernandes@fc.ul.pt

Abstract

Volcanic rocks have shown to be potentially reactive in concrete structures and/or in expansion laboratory tests in
several countries and regions of the world. The main forms of silica which can occur in volcanic rocks are silica-rich
volcanic glass, opal, tridymite and cristobalite.
An interesting case was observed during a long-term research study carried out in the laboratories of Ege University,
Izmir. The expansion results of the concrete prisms of the basalt samples provided from an aggregate source have recently
showed completely different expansion results compared to the expansion results of the aggregate that was brought to
the laboratory from the same region, Aliağa, 9 years before.
Bulk chemical analyses were carried out and polished thin sections of samples of both batches were produced in order
to perform the petrographic assessment of the aggregate aiming to understand the reasons for the different performance in
laboratory. Visual analysis of the micrographs was performed by Matlab. The observation under optical microscope was
conducted by scanning electron microscope with energy dispersive X-ray analysis and by electron probe micro-analysis.
It was concluded that the samples present similar texture and mineral composition but different degrees of alteration.
The potential reactivity was attributed to the abundant silica-rich volcanic glass in the groundmass.

Keywords: volcanic rocks, petrographic characterization, visual analysis, scanning electron microscopy, energy dispersive
x-ray analysis

I. Introduction Wakizaka, 2000; Koga et al., 2011), as was also con-

Alkali-silica reactions (ASR) are deleterious reac-
tions in concrete which occur when simultaneously
three components are available: alkalis, usually
present in the cement paste; moisture conditions
and reactive forms of silica as components of the
aggregates. Some forms of silica are universally ac-
cepted as being potentially reactive, namely amor-
phous and poor crystalline silica: opal, Si-rich vol-
canic glass, tridymite, cristobalite and chalcedony.
Volcanic rocks have been found to be alkali reac-
tive in a number of regions and countries such as
Argentina, Australia, Brazil, Czech Republic, China,
Iceland, Italy, Japan, Mexico, New Zealand and
Turkey (e.g. Gudmundsson & Asgeirsson, 1975;
Shayan & Quick, 1988; Katayama et al., 1989, 1996;
Deng et al., 1996; Marfil et al., 1998; Munhoz et al.,
2008; Román et al., 2008; Vola et al., 2011). How-
ever, different components have been pointed out
as possible cause of reactivity of volcanic rocks.
It is believed that dacitic and rhyolitic volcanic
glass (SiO2 > 65 wt%) can be alkali-reactive (e.g.
Katayama et al., 1989; Katayama, 1997, 2010, 2012;
firmed by Shayan & Quick (1988) for an Australian
basalt with a glassy matrix containing 67wt% SiO2 .
Previous studies on Turkish rocks showed that re-
activity of andesites with intermediate to acidic
composition was due to the presence of volcanic
glass, alteration minerals and SiO2 content (>50%)
(Korkanç & Tuğrul, 2004, 2005).
Volcanic glass tends to alter to zeolites, clay min-
erals and silica minerals (Coombs, 1961) which,
according to some researchers, might be involved
in alkali-silica reaction. For example, Batic et al.
(1994) studied basaltic aggregates from Argentina
and concluded that smectite-type clays were in the
origin of ASR while Marfil et al. (1998) found out
that the expansion of basalts was due to the occur-
rence of volcanic glass and its alteration products
from devitrification, such as argillaceous minerals
and poorly crystallized silica. Opal, a reactive form
of silica, was identified by Katayama & Kaneshige
(1986) in Japanese moderately altered volcanic rocks
belonging to the smectite zone. In a recent work
Tiecher et al. (2012) identified a residue from the
sudden cooling of volcanic magmas (mesostasis)

215
 Fernandes & Andiç-Çakır

Figure 1: Geological map of Bozdivlit mountain and its surrounding.

which had different crypto-microcrystalline silica
phases, including quartz.
In the course of a long-term research project
developed by the laboratories of Ege University,
Izmir in the scope of the assessment of the poten-
tial reactivity of Turkish aggregates an interesting
case was observed. The alkali silica reaction expan-
sion results of the andesite batch provided recently
showed completely different expansion results com-
pared to the expansion results of the same quarry
aggregate that was brought to the laboratory from
the same region, Aliağa, 9 years before. In conse-
quence of these uneven results, the need to clarify
the performance of the different batches was raised.
In the present work various methods have been
used in order to detect the possible differences be-
tween the two batches of aggregates. The results of
the bulk chemical analysis, the mineralogical and
textural characteristics, as well as analyses of the
volcanic glass that forms the groundmass of the
rock are presented.
II. 2. Materials and methods
The two batches studied in the scope of the present
work are 9 years aside. They are named "old sam-
ple" (OS) and "new sample" (NS).
The aggregate quarry from which the old and the
new samples were provided is located in the skirts
of Bozdivlit Mountain, Çaltılıdere, Aliağa (Figure
1). According to the reports of Turkish General
Directorate of Mineral Research and Exploration
(MTA), the rocks located in the region are named
as Bozdivlit Andesitic Basalts (Eşder et al., 1991).
Upper Miocene-aged rocks are also referred to as
basalts, basaltoids and basaltic andesites by other
researchers (e.g. Çopuroğlu et al., 2009).
During the site investigation to the quarry, it

Figure 2: Quarry section where the aggregates were collected, afore: the new sample (NS), behind: the old sample
(OS).

216
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 1: ASR expansion tests conducted on mortar and concrete specimens containing the andesite from Aliağa region.

Old sample* New sample Expansion limit at the specified test period
Ultra accelerated mortar bar test expansion (%, 16 days_ 0.60 0.50 0.10
Concrete prism test expansion (%, 52 weeks) alkalinity:,%1.25 Na2Oeqv 0.16 0.02 0.04
Concrete prism test expansion (%, 52 weeks) alkalinity: constant 0.05 0.0012 —
*Andiç-Çakır (2007)

was observed that the sample taken nine years ago
(OS) was situated in the andesitic upper part of
the quarry where the alteration was comparatively
higher than the part of the quarry that is being run
by recently (NS). The location where the NS was
taken is characterized by columnar joints as shown
in Figure 2, in a compacted less altered rock mass.
The quarry shows heterogeneous characteristic and
the rocks exhibit different behaviour depending on
the vertical location from where they are exploited.
The study performed followed the recommenda-
tions in RILEM AAR-0 (2003) as explained in the
following paragraphs. The methodology was:
• Petrographic analysis of rock samples (AAR-1,
2003)
• Bulk rock chemical analysis
• Ultra accelerated mortar bar test expansion
(AAR-2, 2003)
• Concrete prism test expansion (AAR-3, 2003) in
two modalities: with constant alkalinity; with
boosted alkalis to 1.25% Na2 Oeqv .
• Concrete petrography of the mortar and con-
crete samples after the laboratory expansion
tests were over.
The samples for the chemical and petrographic
analyses consisted of the gravel fraction of the man-
ufactured aggregate from the quarry. From each
sample (OS and NS) three portions composed of
several particles were selected.
Geochemical analyses of major elements were
performed on six pieces of aggregate at Acti-
vation Laboratories (ACTLABS), in Canada, by
inductively-coupled plasma – atomic emission spec-
trometry (ICP-AES).
The petrographic analysis of rock and the con-
crete petrography were carried out under petro-
graphic microscope and by scanning electron micro-
scope (SEM) with energy dispersive spectrometry
(EDS). Polished thin sections were prepared with
25 ⇥ 43 ⇥ 0.030 mm dimensions and were studied
with a polarizing microscope Nikon E 400 POL.
The objective was the identification of potentially
reactive forms of silica in the aggregate and of man-
ifestations of ASR in the concrete prism specimen.
SEM-EDS (High Resolution Schottky Environmen-
tal Scanning Electron Microscope: Quanta 400 FEG
ESEM/EDAX Genesis X4M: 15 kV and 10.0 mm
working distance) was used to obtain the qualita-
tive micro-analysis of the volcanic glass composi-
tion for both samples and to confirm the occurrence
of ASR products. In addition, the thin sections were
analysed by electron probe micro-analyzer (EPMA)
JEOL JXA-8200 in order to obtain quantitative com-
positions of the volcanic glass. For SEM/EDS and
EPMA analyses the polished thin sections were
coated with carbon.
In order to determine the ASR expansions by the
AAR-3 method, three concrete prisms of 285 mm
length and 75⇥75 mm cross-section were cast. Af-
ter the specimens were kept at 20 ± 2 C in a moist
environment (>90% RH) for 24 h, they were de-
molded. Then, initial comparator readings (lo ) were
obtained in a room maintained at 20 ± 2 C. For
storage, standard containers capable of containing
three specimens were used. Specimens were trans-
ferred to two standard sealable containers which

Table 2: Chemical analysis of bulk rock for the two samples.

Old Sample (OS) New Sample (NS)

SiO2 59.84 60.27
Al2 O3 15.95 16.14
Fe2 O3 (T) 5.90 5.47
MnO 0.10 0.10
MgO 3.89 3.06
CaO 6.39 6.18
Na2 O 2.86 2.63
K2 O 2.53 2.45
TiO2 0.64 0.64
P2 O5 0.18 0.18
LOI 2.02 2.96
Total 100.31 100.06

217
 Fernandes & Andiç-Çakır

Table 3: Petrographic characteristics of the samples.

Old sample (OS) New sample (NS)

Plagioclase, pyroxene (clinopyroxene and orthopyroxene), opaque mineral (magnetite),
Phenocrysts Abundant glass inclusions in the phenocrysts,
Xenoliths of olivine
Groundmass Volcanic glass, microlites of plagioclase, augite, hypersthene and opaque minerals
Alteration products Pyroxene often altered to,smectite Xenoliths of olivine replaced,by carbonate
Volcanic glass content is
The phenocrysts are larger and more abundant,
Main differences larger, Microlites of plagioclase in
Volcanic glass darker and compact
the groundmass are less abundant

were filled with water to a depth of 20-30 mm and
attention was paid to avoid the contact of the prisms
with water and also with each other. Throughout
the AAR-3 tests containers were stored in a walk-in
chamber maintained at 38 C. Before any subse-
quent length (li) measurement for both tests, the
specimens were cooled in the sealed containers for
24 h in a room at 20 ± 2 C and each measurement
was made in a room maintained at 20 ± 2 C.
III. Results
Concrete Prism Expansion tests
As above referred the aim of the present work is to
explain the different results of laboratory tests per-
formed with two different batches of the aggregates
collected from the same quarry. Table 1 shows the
expansion data of the samples provided at differ-
ent periods, 9 years apart, named as old sample
(OS) and new sample (NS), respectively. The rec-
ommended limit for AAR-3 was proposed in an
international large-scale EU-funded project (PART-
NER) (Lindgård et al., 2010). It can be seen that
the ultra-accelerated mortar bar expansion results
of the NS are above the limit showing the reactivity
potential of the aggregate while its 52 weeks AAR-3
expansion is considerably low, compared to the ex-
pansions of the OS and the limit value. In addition,
increase in total alkalinity of concrete samples has
an important effect on total expansions.
A recent study conducted on NS aggregate has
revealed that using a high amount of NS powder
and cement dosage as in the case of self-compacted
concrete mixture leads to expansion levels above
the proposed limit (Yüksel et al., 2014)
Chemical analysis of bulk rock
The contents of major elements (as oxides) were
obtained together with loss on ignition (LOI) (Table
2). The projection of the results of 6 fractions of rock
(3 of the OS and 3 of the NS) were plotted on a TAS
diagram which indicates both samples correspond
to andesites (Figure 3). The SiO2 content of the NS
is slightly higher than that of the OS. The values for
LOI indicate that the samples show some alteration,
higher for the NS.

Figure 3: TAS diagram in which the compositions of 6 samples were plotted (1, 2 and 3 – OS; 4, 5 and 6 – NS)
showing that all the pieces of rock analysed correspond to andesite.

218
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 4: Photomicrographs of the OS: a) Phenocryst with abundant glass inclusions (brown); b) agglomerates of
crystals of augite with opaque minerals and brown volcanic glass; c) dense volcanic glass between phenocrysts. In all
fields the volcanic glass is abundant and dense. All images under PPL.

Mineralogical Evaluation of Old Sample vs New
Sample
The samples analysed (3 portions of the OS and 3
portions of the NS) correspond to porphyritic an-
desite with a very fine glassy matrix. It is worth
mentioning the study developed by Çopuroğlu et
al. (2009) in which a discussion is presented about
the best designation for this rock (basalt, andesite
or quartz tholeiite). Based on the chemical analy-
ses above and on information in Katayama (2013),
the designation here adopted is andesite. The
rock contains phenocrysts of plagioclase and py-
roxene (augite and hypersthene). Opaque minerals
(magnetite) occur in association with the pyroxene.
The plagioclase phenocrysts show strong concentric
compositional zoning and some present rims of re-
action in contact with the groundmass. The crystals
are dominantly euhedral to subhedral and occur
as fresh crystals with slight alteration along cracks
and cleavages. The phenocrysts of pyroxene are
mostly subhedral and fresh. However, in some of
the particles of the OS the phenocrysts are totally or
partially altered to clay minerals. Both plagioclase
(Figure 4) and pyroxene contain abundant Fernandes
glass& Andiç-Çakır
inclusions, sometimes arranged in zones. There
are also xenoliths of olivine (Çopuroğlu et al., 2009;
Katayama, 2013) mostly replaced by carbonate min-
erals (with Ca, Fe, Mg) which might correspond to
calcite, siderite and magnesite.
The groundmass is hyalopilitic and contains mi-
crolites of plagioclase, augite and hypersthene. Vol-
canic glass (brown) is the main component of both
samples (Figures 4 and 5). In some particles of the
rock samples the groundmass has a cloudy brown
colour. In other particles, volcanic glass is limpid
and denser/darker around the phenocrysts. Matrix
is always more abundant than the phenocrysts con-
tent. Under cross polarized light (XPL), the volcanic
glass is isotropic. As already referred in Çopuroğlu
et al. (2009) no signs of devitrification were found.
The main differences between the OS and the
NS are the size and density of the phenocrysts.
Phenocrysts of plagioclase and pyroxene are usually
larger in the NS than in the OS. The density of the
phenocrysts is lower in the OS which means that
the relative content of matrix in comparison to the
phenocrysts is larger in the OS than in the NS. In
addition to the matrix being more abundant in the
OS, the volcanic glass seems to be more compact,
with less tiny laths of plagioclase (Figures 4 and 5).
Some of the fractions of both samples contain
manifestations of alteration, with replacement of
phenocrysts of pyroxene by clay minerals (smectite).

a b c

Figure
Figure5:5:Photomicrographs of the
Photomicrographs of the NS:
NS:a)a)phenocryst
phenocrystof of plagioclase
plagioclase showing
showing reaction
reaction rimglass
rim and and inclusions
glass inclusions (PPL);
(PPL); b)
b) phenocrysts
phenocrysts of pyroxene
of pyroxene with with
opaqueopaque minerals,
minerals, volcanic
volcanic glass (brown)
glass (brown) and (yellowish)
and smectite smectite (yellowish) (PPL); crystal
(PPL); c) euhedral c) euhedral
of
zoned plagioclase
crystal and anhedral
of zoned plagioclase andcrystals of augite
anhedral (XPL).
crystals The groundmass
of augite (XPL). The shows abundant shows
groundmass brown volcanic
abundant glass,
brownisotropic
volcanic
glass, isotropic under XPL. under XPL.

219
 Figure 5: Photomicrographs of the NS: a) phenocryst of plagioclase showing reaction rim and glass inclusions (PPL); b)
phenocrysts of pyroxene with opaque minerals, volcanic glass (brown) and smectite (yellowish) (PPL); c) euhedral crystal of
Fernandes & Andiç-Çakır
zoned plagioclase and anhedral crystals of augite (XPL). The groundmass shows abundant brown volcanic glass, isotropic
under XPL.

a b c

Figure 6: Alteration of the phenocrysts which are replaced by: a and b) clay minerals (smectite) in OS (PPL); c) carbonate
Figure 6: Alteration of the phenocrysts which minerals
are replaced by: a and b) clay minerals (smectite) in OS (PPL); c)
in NS (PPL).
carbonate minerals in NS (PPL).

Contrary to the chemical analysis, the manifesta- K, Na, Ca and Mg. The main component is always
tions of alteration of the phenocrysts are more fre- Si, followed by Al. From the qualitative analysis
quently observed in the OS (Figure 6). The olivine based on the spectra, the volcanic glass of the NS
xenoliths are usually replaced by carbonate miner- seems to be richer in Si, in relation to the content
als but also cryptocrystalline quartz was observed. of Al, than the OS.
An example is presented in Figure 7 where it can In addition to this study, EPMA were performed.
be observed that the mineral is intensely altered On Table 4 the results obtained are presented, each
and silica forms nodules in the centre of the altered column corresponding to the mean value of 4 anal-
crystal. yses. It can be confirmed that the composition of
The main petrographic characteristics of the sam- the volcanic glass has much higher silica content
ples are summarized on Table 3. than the bulk rock. The values of SiO2 vary be-
tween 69.23 and 71.43% for the OS and from 71.23
Figure 7: Xenocryst of olivine replaced by carbonate minerals and cryptocrystalline quartz (NS) (PPL and XPL).
Composition of the volcanic glass to 72.59% for the NS which correspond to dacitic to
Tablethin
3 – Petrographic rhyolitic
characteristics of composition.
the samples. The second most abundant
During the study of the polished sections un-
oxide is Al O
2 3 , followed by K2(NS)
O, which content
der optical microscope different Old sample were
locations (OS) se- New sample
Phenocrysts is always higher than the
Plagioclase, pyroxene (clynopyroxene and orthopyroxene), opaque mineral Na O and FeO. It was
2 (magnetite)
lected and marked for analyses by SEM/EDS and
Abundant glassalso verified
inclusions thatphenocrysts
in the the volcanic glass was easily burnt
EPMA. Xenoliths
by the of olivine beam, suggesting the presence of
electron
The different locations corresponded
Groundmass Volcanic glass,to:
microlites of plagioclase, augite, hypersthene and opaque minerals
volatiles.
Alteration products Pyroxene often altered
• Vesicles containing forms of silica - these show to smectite Xenoliths of olivine replaced by carbonate
Main differences Volcanic glass content is larger The phenocrysts are larger and more abundant
a colourless mineral under PPL and first order Image
Microlites of plagioclase in the groundmass are analysis Volcanic glass darker and compact
grey under XPL with radial texture; less abundant
In order to understand if the amount of the glassy
• Volcanic glass in the groundmass. matrix itself might be the reason for the different
In Figure 8 examples of the vesicles as well as the expansion behaviour of the samples provided at
analyses of the volcanic glass can be observed. It different periods from the same quarry, image anal-
could be confirmed that the filling of the vesicles is ysis was performed. Visual analysis using Matlab
pure silica. program was carried out on the photomicrographs
The analyses by EDS indicate that the volcanic of the same magnification taken from the OS and
glass is composed mainly of Si and Al with minor from the NS. In the beginning, the three layer (RGB)

Figure 7: Xenocryst of olivine replaced by carbonate minerals and cryptocrystalline quartz (NS) (PPL and XPL).

220
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

OS

NS

Figure 8: Vesicles in the OS (above) and of the NS (below) filled with a colourless mineral which was confirmed by
SEM/EDS to be pure silica. Microphotograph (PPL), SEM image (BSE – backscattered electrons) and EDS spectrum
both of the silica and of the glass in the matrix.

221
 Fernandes & Andiç-Çakır

Table 4: Analyses by EPMA of the volcanic glass present in the groundmass.

Old Sample (OS) New Sample (NS)

Matrix 1 Matrix 2 Matrix 3 Matrix 4 Matrix 1 Matrix 2 Matrix 3 Matrix 4
SiO2 70.73 71.43 69.23 70.97 72.14 71.92 72.59 71.23
Al2 O3 11.65 11.87 11.77 11.62 11.63 11.79 12.03 12.11
FeO 2.48 2.30 2.63 2.45 1.45 1.31 1.44 1.72
MnO 0.05 0.02 0.04 0.03 0.03 0.01 0.03 0.02
MgO 0.24 0.16 0.21 0.20 0.09 0.08 0.08 0.20
CaO 0.91 0.78 0.88 0.93 0.49 0.48 0.56 0.98
Na2 O 1.13 1.08 0.95 1.02 1.16 1.24 1.23 1.19
K2 O 6.05 6.06 5.71 5.73 6.73 6.48 6.58 6.02
TiO2 0.89 0.90 0.87 0.71 0.74 0.66 0.67 0.67
NiO 0.04 0.02 0.01 0.02 0.02 0.08 0.02 0.04
Cr2 O3 0.05 0.03 0.07 0.04 0.07 0.04 0.07 0.06
BaO 0.35 0.29 0.39 0.26 0.09 0.11 0.17 0.22
Cl 0.16 0.15 0.14 0.12 0.19 0.10 0.19 0.21
F 0.14 0.08 0.03 0.12 0.06 0.09 0.08 0.08
H2 O 5.16 4.91 7.13 5.88 5.17 5.60 4.33 5.33
Total 100.08 100.08 100.06 100.10 100.06 99.99 100.07 100.08

coloured image was taken into consideration, and
the difference between the pixel values of the R
layer and the B layer was obtained.
In this new image, the pixel values over a chosen
threshold value (seen in white) denoted the ma-
trix and the pixel values under a chosen threshold
value (seen in black) denoted the crystals and voids
(Figure 9). Thus, it was possible to differentiate
the matrix from the rest of the micrograph. The
threshold values were chosen by trial and error. In
the last part of the visual analysis, the matrix area
with respect to the total area in consideration was
calculated for the OS and the NS, respectively. The
matrix area for the old and the new sample was
found as 38.68% and 38.86%, respectively. As a
result, the matrix area values of old and new sam-
ples are so close to each other that it cannot be
concluded that the amount of matrix phase is the
key to reactivity.
Concrete petrography
After the conclusion of the concrete prism tests,
slices were cut and prepared for thin sections pro-
duction. The study of the concrete thin sections al-

OS

NS

Figure 9: Left, polarised microscopy image (PPL); right, visual analysis output.

222
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 10: Manifestations of ASR: cracks and gel are observed in both samples (left, OS; right, NS) in PPL.

lowed the observation of the features generated by
ASR such as cracks and formation of scarce alkali-
silica gel, which were found in both samples (OS
and NS) (Figure 10), although these manifestations
were not very exuberant.
In addition to the identification of ASR signs it
could be observed that some of the aggregate parti-
cles in the concrete samples showed slight different
characteristics in relation to the thin sections of
the manufactured aggregate particles. This was
especially visible for the OS in which altered an-
desite particles were more abundant. The most
altered particles showed stronger manifestations
of reaction, namely rims of fibrous crystals in the
interfaces between the aggregate particles and the
cement paste. These rims were dark yellow un-
der PPL and had very low birefringence under
XPL, with the acicular/fibrous crystals in radial
disposition. Under higher magnification they corre-
sponded to rosette-like agglomerates (Figure 11).
IV. Discussion
The role played by the volcanic glass and by the
minerals of alteration on the development of ASR
is not yet totally understood.
The study developed started with the findings on
laboratory tests with contradictory results with sam-
ples from the very same quarry. The first approach
was related to the need to determine the bulk chem-
ical composition of the two different batches of
aggregates in order to find out if reactive samples
were oversaturated in silica. However, the results
obtained (59.84% for the OS and 60.27% for the
NS) were not different enough to sustain the dif-
ferent behaviour based on the content of silica of
both samples. The chemical analysis of the OS was
very similar to the results already published in Çop-
uroğlu et al. (2009). The composition of both the
OS and the NS was of the same order for all the ox-
ides. Both samples have SiO2 content >50wt.% and
could therefore be classified as potentially reactive
to alkalis.
The second step was the performance of petro-
graphic analysis in order to identify the main dif-
ferences between the OS and the NS. This study
revealed that in both cases the matrix is more abun-
dant than the phenocrysts and that the main com-
ponent of the matrix is volcanic glass. There is,
however a slight difference in that the content of
the matrix seems to be higher for the OS than for
the NS, which contains larger and more numerous
phenocrysts. However, this could not be confirmed
by image analysis.

a b c

Figure 11: Fibrous crystals observed in rims in the boundaries of the aggregate particles of the OS. a) Photomicrograph
under PPL; b and c) SEM images (BSE).

223
 Fernandes & Andiç-Çakır
It is well known that the crystallization of the
magma generates a residual melt which is richer in
silica and, therefore, in volcanic rocks the volcanic
glass will have a higher content in silica then the
bulk composition of the rock. In this context, and
according to the results of other researchers already
referred above, SEM/EDS analyses were carried out
as well as EPMA on the volcanic glass that occurs in
the matrix of different fractions of the two batches
(OS and NS). The results obtained are encouraging
as they allow the classification of the volcanic glass
as having dacitic to rhyolitic composition and being
potentially reactive to alkalis. Again, the results are
far from explaining the different performance of
the OS and the NS observed in the laboratory tests.
Following the conclusion of the laboratory tests,
the thin sections produced for concrete petrogra-
phy brought some interesting information. Even at
the scale of the thin section, it is possible to notice
that there is some variability in the quarry, mainly
in what concerns the OS. In fact, a number of al-
tered particles were observed showing the presence
of clay minerals which were only discrete in the
thin sections of the manufactured aggregate. The
concrete petrography analysis led to the identifica-
tion of cracks, some partially filled with alkali-silica
gel. These cracks develop dominantly parallel to
the interfaces of the aggregate particles with the
cement paste, but also perpendicular, extending to
the cement paste and connecting to other particles
of aggregate.
There are rims of fibrous minerals which are quite
different from the manifestations of ASR found
in other types of rocks but similar to what has
been observed with volcanic rocks (Medeiros et al.,
2014) and also to the ones already published in
Çopuroğlu et al. (2009).
In addition, in a recent study developed by Yük-
sel et al. (2014) it was verified that the increase in
the amount of filler-sized NS and higher cement
dosages (470 kg/m3 instead of 440 kg/m3 ) led to
higher AAR-3 expansions. This situation is related
with the composition and amount of siliceous ma-
terials, alkalinity and permeability of concrete mix-
ture. Thus, it can be concluded that reactivity of
aggregate seems to be dependent on various factors
in addition to type and amount of siliceous phases
present.
V. Conclusions
The results obtained in the present study do not
totally explain the origin of the potential reactivity
of andesites to alkalis but it might contribute with
some clarification.
From all the tests carried out, it could be con-
cluded that both samples’ compositions indicate
potential reactivity, due to the SiO2 rich volcanic
224
glass. Volcanic glass is the main component of these
andesites but is more abundant in the OS as this
sample shows lower density in phenocrysts and
small size of the crystals. Also, the volcanic glass
is denser in the OS and forms darker rims around
some of the phenocrysts. Regarding alteration, the
OS shows stronger alteration than the NS with clay
minerals replacing some phenocrysts of pyroxene
and plagioclase. In consequence of the results ob-
tained, it is believed that the degree of alteration of
the rocks might have influenced the development
of the ASR in the OS. Based on previous experience,
NS aggregate has the potential to show reactivity if
certain conditions are met.
The concrete petrography of the concrete prisms
specimen suggests that there is variability of the
features of the rocks, namely related to different de-
grees of alteration and that the most altered pieces
of rock were the ones exhibiting more evident man-
ifestations of ASR. The origin of the rims in the
interfaces, showing features different from ASR in
other types of rocks, also deserves some attention
and further study.
Acknowledgements
The authors are thankful to Fundação para a Ciên-
cia e Tecnologia (FCT) for the financial support
the Strategic Project Pest-OE/CTE/UI0039/2014
and Scientific and Technical Research Coun-
cil of Turkey (TUBITAK) under grant no.
110M569. The analyses by SEM/EDS were
performed at CEMUP which equipment was
funded by the projects REEQ/1062/CTM/2005 and
REDE/1512/RME/2005 of Portuguese Foundation
for Science and Technology (FCT). The authors
would like to thank to Prof. Gökçen Bombar for
their help in image analysis of samples.
References
Andiç-Çakır, Ö. (2007). Alkali Agrega Reaktivitesinin Tespitinde Kul-
lanılan Deney Metotlarının İncelenmesi. PhD Thesis, Ege Üniversity, 270
pp (in Turkish).
AAR-0 (2003). RILEM TC 191-ARP: Alkali-reactivity and preven-
tion—assessment, specification and diagnosis of alkali-reactivity, RILEM
recommended test method AAR-0: Detection of alkali-reactivity poten-
tial in concrete—outline guide to the use of RILEM methods in assess-
ments of aggregates for potential alkali-reactivity, Materials and Structures
(36): 472–479.
AAR-1 (2003). RILEM TC 191-ARP: Alkali-reactivity and pre-
vention—assessment, specification and diagnosis of alkali-reactivity,
RILEM recommended test method AAR-1: detection of potential alkali-
reactivity of aggregates—petrographic method, Materials and Structures
(36): 480–496.
AAR-2, AAR-3 (2000). RILEM TC 106-AAR, Alkali aggregate re-
action. A. TC 106-2—Detection of potential alkali-reactivity of aggre-
gates—the ultra-accelerated mortar-bar test B. TC 106-3—Detection of
potential alkali-reactivity of aggregates—method for aggregate combina-
tions using concrete prisms, Materials and Structures (33): 283–293.
Batic, O., Maiza, P., Sota, J. (1994). Alkali-silica reactions in basaltic
rocks. NBRI method, Cement and Concrete Research, (24/7), pp. 1317-1326
Coombs, D.S. (1961). Some recent work in the lower grades of meta-
morphism. Australian Journal of Science (24): 203-215.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Çopuroğlu, O., Andiç-Çakır, Ö., Broekmans, M.A.T.M., Kühnel, R.
(2009). Mineralogy, Geochemistry and Expansion Testing of an Alkali-
Reactive basalt from Western Anatolia, Turkey, Materials Characteriza-
tion - special issue for 11th Euroseminar on Microscopy Applied to Build-
ing Materials (EMABM), 60, 756-766. DOI: 10.1016/j.matchar.2008.09.011.
Deng, M., Xu, Z., Lan, X., Han, S., Tang, M. (1996). Microstructures
of some alkali-silica reactive aggregates in China. Cement and Concrete
Research (26): 663-668.
Gudmundsson, G., Asgeirsson, H. (1975). Some investigations on
alkali-aggregate reaction. Cement and Concrete Research (5): 211-220.
Katayama, T. (1997). Petrography of alkali-aggregate reactions in con-
crete – reactive minerals and reaction products. In: Nishibayashi, S and
Kawamura, M (eds). Proceedings of the East Asia Alkali-Aggregate Reaction
Seminar, Supplementary papers, Tottori, Japan: A45-A59.
Katayama, T. (2010). Diagnosis of alkali-aggregate reaction – polariz-
ing microscopy and SEMEDS analysis. Proceedings of the 6th International
Conference on Concrete under Severe Conditions (CONSEC’10), Merida, Mex-
ico (1): 19-34.
Katayama, T. (2012). Petrographic study of alkali-aggregate reactions
in concrete. PhD thesis, Department of Earth and Planetary Science,
Graduate School of Science, The University of Tokyo, Japan: 168 pp.
Katayama,T. (2013). Petrographic examination of andesite aggregate
from Turkey. Unpublished report.
Katayama, T., Helgason, T.S., Olafsson, H. (1996). Petrography and
alkali-reactivity of some volcanic aggregates from Iceland. In: Shayan,
A (ed). Proceedings of the 10th International Conference on Alkali-Aggregate
Reaction, Melbourne, Australia: 377-384.
Katayama, T., Kaneshige, Y. (1986). Diagenetic changes in potential
alkali aggregate reactivity of volcanic rocks in Japan-A geological inter-
pretation, In: Grattan-Bellew, PE (editor): Proceedings of the 7th Interna-
tional Conference on Alkali-Aggregate Reaction in Concrete (ICAAR), Ottawa,
Canada: 489-495.
Katayama, T., St John, D.A., Futagawa, T. (1989). The petrographic
comparison of rocks from Japan and New Zealand-Potential reactivity re-
lated to interstitial glass and silica minerals. In: Okada, K, Nishibayashi,
S and Kawamura, M (editors). Proceedings of the 8th International Confer-
ence on Alkali-Aggregate Reaction in Concrete (ICAAR), Kyoto, Japan: 537-
542.
Koga, H., Hyakutake, T., Watanabe, H., Sakamoto, T. (2011). Alkali-
silica reaction in Japan. Dams and reservoirs under changing challenges. In:
Schleiss, AJ, Boes, RM (eds). Taylor and Francis Group, London, UK:
163-170.
Korkanç, M., Tugrul, A. (2004). Evaluation of selected basalts from
Nigde, Turkey, as source of concrete aggregate. Engineering Geology
(75/3-4): 291-307.
Korkanç, M., Tuğrul, A. (2005). Evaluation of selected basalts from
the point of alkali-silica reactivity. Cement and Concrete Research (35): 505-
512.
Lindgård J., Nixon P.J., Borchers I., Schouenborg B., Wigum B.J., Hau-
gen M. and Åkesson U. (2010) The EU “PARTNER” Project – European
standard tests to prevent alkali reactions in aggregates: Final results and
recommendations. Cement and Concrete Research, (40): 611-635.
Marfil, S.A., Maiza, P.J., Bengochea, A.L., Sota, J.D., Batic, O.R. (1998).
Relationships between SiO2, Al2O3, Fe2O3, CaO, K2O, and expansion
in determination of the alkali reactivity of basaltic rocks. Cement and
Concrete Research (28): 189-196.
Medeiros, S., Fernandes, I., Nunes, J. (2014). Identification of alkali-
silica reactions in Santa Maria Island airport. International Multidisci-
plinary Scientific GeoConferences (SGEM), Vol. II, paper 62, Bulgaria: 487-
494.
Munhoz, F.A.C., Kihara, Y., Cincotto, M.A. (2008). Effect of mineral
admixtures on to the mitigation of alkali-silica reaction in concrete. In:
Broekmans, MATM and Wigum, BJ (eds). Proceedings of the 13th Interna-
tional Conference Alkali-Aggregate Reaction, Trondheim, Norway: 591-599.
Román, H.J.M., Camacho, R.E.R., Afif, R.U., Martínez, G., Rodriguez,
D.R. (2008). Evaluation of alkali-aggregate reaction in aggregates from
igneous rocks. In: Broekmans, MATM and Wigum, BJ (eds). Proceedings
of the 13th International Conference on Alkali-Aggregate Reaction, Trondheim,
Norway: 310-319.
Shayan, A., Quick, G.W. (1988). An alkali-reactive basalt from
Queensland, Australia. The International Journal of Cement Composites and
Lightweight Concrete (10): 209-214.
Tiecher, F., Molin, D.C.C.D., Gomes, M.E.B., Hasparyk, N.P., Mon-
teiro, P.J.M. (2012). Influence of mesostasis in volcanic rocks on the alkali-
aggregate reaction. Cement and Concrete Composites (34): 1130-1140.
Vola, G., Berra, M., Rondena, E. (2011). Petrographic Quantitative
Analysis of ASR susceptible aggregates for concrete. Proceedings of the
13th Euroseminar on Microscopy Applied to Building Materials (EMABM),
Ljubljana, Slovenia.
Wakizaka, Y. (2000). Alkal-silica reactivity of Japanese rocks. Engi-
neering Geology (56): 211-221.
Yüksel, C., Mardani-Aghabaglou, A., Beglarigale, A., Yazıcı, H., Ram-
yar, K., Andiç-Çakır, Ö (2014). Influence of water/powder ratio and
powder type on alkali-silica reactivity and transport properties of self-
consolidating concrete. Materials and Structures (accepted for publication,
published online: 18 December 2014).
225
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

226
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

ASR Prevention - Effect of lithium on the

reaction products formed in aggregates
Andreas Leemann
Swiss Federal Laboratories for Materials Science and Technology, Dübendorf, Switzerland
andreas.leemann@empa.ch

Abstract

Important progress has been made in understanding the mechanisms enabling LiNO3 to limit the extent of alkali-silica
reaction (ASR) as reported in recent publications. However, there are still some gaps in the understanding of the
ASR-suppressing mechanisms. They concern the influence of aggregate characteristics on the effectivness of LiNO3 to
supress ASR. In this study concrete and mortar are used to address this specific question. Concrete and mortar were
investigated by SEM combined with EDX. The effectiveness of LiNO3 to suppress ASR is dependent on the specific
surface area of the reactive aggregate. Whereas only minor amounts of reaction products limited to the surface are
formed on the dense artificial aggregate, a substantial amount of reaction products are formed in a porous artificial
aggregate with high specific surface area.

Keywords: Alkali-silica reaction, prevention, concrete, lithium

I. Introduction (Fournier et al., 2003; Folliard et al., 2006; Tremblay

In the last years the understanding about the mech-
anisms enabling LiNO3 to limit the extent of alkali-
silica reaction (ASR) in mortar and concrete has
improved (Tremblay et al., 2010; Feng et al., 2010a;
Feng et al., 2010b; Leemann et al., 2014). As shown
on mortar samples, LiNO3 considerably alters the
composition and the porosity of the ASR prod-
uct leading to a dense phase limiting the diffu-
sion of ions to reactive minerals (Leemann et al.,
2014). Additionally, it is possible that the lithium-
containing reaction product is less expansive than
the lithium-free reaction product. In model systems
LiNO3 addition changes the structure of the reac-
tion product from one dominated by Q3 (layers of
SiO2 -tetrahedra) to one dominated by Q2 (chains
of SiO2 -tetrahedra) (Mitchell et al., 2004; Leemann
et al., 2014). A structure dominated by Q3 seems
to be better suited to adsorb alkalis compensating
the negative charges of these layers. Such a higher
binding capability of alkalis can lead to higher bind-
ing capability of water. As the alkalis form effective
hydrogen bonding networks with water molecules
and are able to stabilize large amounts of water
in the reaction product (Kirkpatrick et al., 2005),
a reaction product dominated by Q3 may have a
higher swelling potential.
The current study extends the investigation made
on mortar (Leemann et al., 2014) on concrete and
focuses on one pending question about the effect
of LiNO3 on ASR: it is still unclear why LiNO3 is
effective with some aggregates but not with others
et al., 2007). Two concrete mixtures with natural ag-
gregates were produced: one as a reference without
additions and two with LiNO3 addition. Mortars
with corresponding compositions were produced.
A very dense SiO2 -cylinder and a highly porous
SiO2 -filter were embedded as artificial aggregates
in the mortars to investigate the effect of specific
surface area on the ASR-supressing effect of LiNO3 .
The composition of the reaction products in con-
crete and mortar was analyzed with scanning elec-
tron microscopy (SEM) and energy-dispersive X-ray
spectroscopy (EDX).
II. Materials and methods
Materials
An ordinary Portland cement (CEM I 52.5 R, Table
1) was used to produce concrete and mortar mix-
tures. The mix design of the concrete and mortar is
shown in Table 2. Gneiss aggregates with a grain
size of 0/22 mm (0/4 mm: 35 mass-%, 4/8 mm:
19 mass-%, 8/16 mm: 21 mass-%, 16/22 mm: 25
mass-%) were used for the concrete and for the
mortar a grain size of 0/4 mm. Three prisms per
mixture with a dimension of 70x70x282 mm3 were
cast with the concrete. The mortar was filled into a
cylindrical container with a diameter of 54 mm and
a height of 45 mm. Two artificial aggregates were
embedded in the mortar, a SiO2 -cylinder (diameter:
10 mm, length: 25 mm) and SiO2 -filter disc (diame-
ter: 20 mm, height: 4 mm). Both consist of highly

227
 Leemann

Table 1: Composition of cement in mass-%.

Material SiO2 Al2 O3 Fe2 O3 Cr2 O3 MnO TiO2 P2 O5 CaO MgO K2 O Na2 O SO3 L.O.I.
CEM I 52.5 R 20.16 5.29 3.25 0.01 0.06 0.29 0.22 62.71 1.65 0.90 <0.03 3.82 1.34

pure SiO2 -glass (SuprasilR). The SiO2 -cylinder is III. Results

dense material without measureable porosity. The
SiO2 -filter is highly porous (total porosity of 35-40
volume-%) with a pore size of 10-16 µm accord-
ing to the producer. The samples for SEM analysis
were dried in an oven at 50 C for three days, epoxy
impregnated, cut to obtain cross-sections of the
mortar/aggregate interface, polished and carbon
coated. Due to the high brittleness of the artificial
glass aggregates, small particles continuously splin-
tered during polishing leading to a relatively high
surface roughness within these aggregates.
Methods
The potential alkali reactivity of the concrete was
measured according to AFNOR P18-454 at 60 C
and 100% relative humidity. The length of the con-
crete prisms was measured every 4 weeks for 40
weeks. After 24 weeks one prism of concrete Ref
and Li 0.74 was removed for analysis with SEM.
Concrete Li 0.37 was not analyzed. The mortar
samples were investigated at the same age. SEM
images were acquired using an ESEM-FEG XL30 in
the high vacuum mode (2.0-6.0 ⇥ 10 6 Torr) with
an accelerating voltage of 12 kV and a beam current
of 260-280 mA in the backscattered electron mode
(BSE). The chemical analysis was conducted with
energy dispersive X-ray spectroscopy (EDX). An
EDAX 194 UTWdetector, a Philips digital controller,
and Genesis Spectrum Software (Version 4.6.1) with
ZAF corrections were used. The reaction prod-
ucts were analyzed in six to seven aggregates per
concrete mixture, resulting in a total of about 200
analyzed points per concrete. Approximately 150
points were analyzed in each mortar per artificial
SiO2 -aggregate.
Concrete
Compared to concrete Ref expansion is reduced to
nearly 50% by adding LiNO3 in the higher dosage
((Li/(Na+K) = 0.74) in concrete Li-0.76, although
the resulting value is still at the level of the limit
value distinguishing between non-reactive and reac-
tive concrete mixtures according to Swiss standard
MB SIA 2042 (2012; Figure 1A). The lower dosage
in concrete Li-0.37 is not effective in reducing ex-
pansion. Concrete Li 0.37 and Li 0.74 show a mass
increase (Figure 1B), while concrete Ref stabilizes
at its initial weight at the age of 16 weeks after an
initial mass loss.
The morphology of the reaction products in con-
crete Ref shows a layered texture with some mi-
crocracking as a result of drying due to sample
preparation (Figure 2A). The reaction product in
concrete Li 0.74 is significantly more homogenous
and has no layered texture (Figure 2B). Differences
are present as well in the chemical composition of
the reaction products of the different concrete mix-
tures. The reaction product in concrete Ref is in the
usual range of ASR products present in aggregates
(Leemann and Merz, 2013). The chemical compo-
sition of the reaction products in concrete Li 0.74
is entirely different (Figure 3). The majority of the
analyzed point shows a lower alkali concentration
and significantly lower calcium concentration in
relation to silicon compared to concrete Ref. These
data are in agreement with Kawamura and Fuwa
(2003) and Leemann et al. (2014).
Mortar
The SiO2 -cylinder embedded in the mortars shows
a rim of reaction products in mortar Ref (Figure 4A).
Additionally, some channel-like dissolution features
filled with reaction products penetrate into the SiO2 -

Table 2: Mix design of the concrete and mortar (Mk = Metakaolin)

Concrete concrete Ref concrete Li 0.37 concrete Li 0.74 mortar Ref mortar Li 0.74
Cement [kg/m3 ] 400 400 400 600 600
Additions [kg/m3 ] - LiNO3 : 2.8 LiNO3 : 5.6 - LiNO3 : 8.1
Aggregates [kg/m3 ] 1776 1770 1770 1318 1297
Water [kg/m3 ] 192 192 192 300 300
w/b 0.48 0.48 0.48 0.50 0.50

228
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 1: Expansion (A) and mass change (B) of concrete Ref, Li 0.37 and Li 0.74 versus time. The dotted line in A
shows the allowed limit value for 20 and 48 weeks according MB SIA 2042 (2012).

cylinder. The reaction product shows a Ca/Si-ratio
in the range of 0.15-0.20 and a (Na+K)/Si-ratio in
the range of 0.3-0.4 (Figure 5A). The SiO2 -cylinder
in mortar Li exhibits only a very thin (⇠ 1 µm)
layer of reaction product identifiable by the in-
creased backscattering contrast compared to the
SiO2 -cylinder and the cement paste (Figure 4B). Be-
cause of its dimensions the compositions of the
thin layer in mortar Li cannot be determined re-
liably (Figure 5B): the interaction volume of the
electron beam can be expected to extend from the
reaction product into the SiO2 -cylinder and the ce-
ment paste.
After 6 weeks the SiO2 -filter in mortar Ref shows
an ingressing front of reaction products into a depth
of 0.5-1.0 mm. After 24 weeks reaction products
are precipitated in the pores of the filter through
the entire cross section of 4 millimeters. They are
more abundant in a distance to the interface with
the cement paste of 0-1 mm than in the midsection
of the filter. The same distribution of the reaction
products is present in the filter embedded in mortar
Li 0.76 (Figure 6). However, the amount of reaction
products is lower assessed qualitatively compared
to mortar Ref. In a distance to the interface with the
cement paste of about 50 micrometers the reaction
products in all mortars contain calcium leading to a
molar Ca/Si-ratio between 0.05-0.50 after 24 weeks.
In greater depth the Ca/Si-ratio varies between 0.01
and 0.05. The reaction products mainly contain
potassium as alkalis reflecting the composition of
the cement used. The molar (Na+K)/Si-ratio of
the reaction products does not vary systematically
across the filter and is in the range of 0.1-0.3 (see
Figure 5B). There are no systematic differences in
the composition of the reaction products in the
different mortars as determined with EDX.
IV. Discussion
The addition of LiNO3 in a Li/(Na+K)-ratio of 0.74
clearly leads to a significant reduction of expan-
sion. Moreover, morphology and composition of
the reaction products are changed the same way
as already observed in mortars with LiNO3 addi-
tion (Leemann et al., 2014). The reaction product is

Figure 2: ASR products in concrete Ref (A) and concrete Li 0.74 (B) in veins located in gneiss aggregates.

229
 Leemann

Figure 3: Atomic ratio of (Na+K)/Si versus Ca/Si in the reaction products of concrete Ref and Li 0.74.

dense and texture-less (Figure 3B). Its alkali and cal-
cium contents are considerably decreased (Figure 4).
As postulated in Leemann et al. (2014) the effective-
ness of LiNO3 is caused by the formation of a dense
product protecting reactive minerals from further
dissolution and by the formation of a reaction prod-
uct with lower expansion potential. In order to have
this impact on ASR-induced expansion lithium has
to be present in a sufficient concentration in the
pore solution of the concrete. Obviously, the addi-
tion in the lower dosage (Li/(Na+K) = 0.37) is not
effective at all with the aggregate used. As shown
in several studies the required LiNO3 addition de-
pends on the aggregate used (Fournier et al., 2003;
Folliard et al., 2006; Tremblay et al., 2007).
The artificial aggregates in the mortars clearly
show that the surface area of an aggregate is influ-
encing the effectiveness of LiNO3 to suppress ASR.
While the dense SiO2 -cylinder shows only a thin
layer of reaction product on its surface, there is a
substantial amount of reaction product present in
the SiO2 -filter. The occurrence of this situation has
to be related to the change of pore solution compo-
sition with time. In general, lithium concentration
in the pore solution decreases with time (Tremblay
et al., 2008; Leemann et al., 2014). As the surface of
the SiO2 -cylinder is in contact with mortar pore so-
lution from the time of sample production, it reacts
at a time where lithium concentration in the pore
solution is still high. This first dense reaction layer
is capable of protecting the SiO2 -cylinder from any
further reaction even after lithium concentration in

Figure 4: A. SiO2 -cylinder in mortar Ref with reaction rim at the surface and channel-like dissolution features, B.
SiO2 -cylinder in mortar Li with thin reaction rim at the surface (please note the difference in magnification compared
to A)

230
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 5: Composition of the reaction products of the embedded SiO2 -cylinder (A, including the transition between
cement paste and reaction product) and the SiO2 -filter (B, from the interface towards greater depth) in mortars Ref
and Li 0.76 at the age of 24 weeks.
the pore solution has decreased. As such, the postu-
lation that the dense reaction product formed in the
presence of lithium protects reactive minerals from
further reaction (Leemann et al., 2014) seems to be
correct. In the case of the SiO2 -filter, the distribution
of the reaction products in the filters and their com-
position indicate the formation of a reaction front
within the filters. As the difference in the distribu-
tion of the reaction products after 6 and 24 weeks
shows (Figure 6), the formation of reaction products
reaches greater depth after a longer reaction time.
Additionally, calcium concentration decreases sig-
nificantly in a depth > 50-100 µm (Figure 6). This
clearly indicates that the high surface area of the
SiO2 -filter makes it likely that the reaction between
pore solution and SiO2 is proceeding over weeks
even at a stage where the lithium concentration in
the pore solution has decreased considerably. As a
consequence, the lithium available is not sufficient
anymore to form a protective layer as observed on
the surface of the SiO2 -cylinder and an expansive
reaction product is formed. This effect of the spe-
cific surface area is clearly related to change of pore
solution composition with time. There might be an
additional effect in aggregates of high specific sur-
face area. Alkalis are continuously removed from
the ingressing pore solution due to the reaction
with SiO2 . As potassium and sodium are adsorbed
slower by reaction products than lithium (Tremblay
et al., 2008; Leemann et al., 2014), they may lead to
a reaction in parts of the filter not (yet) reached by
lithium. However, this is purely hypothetical. Such
a mechanism would likely lead to the formation of
a lithium front that penetrates less far into the filter
than the sodium and potassium front. ToF-SIMS
measurements should allow detecting such a front
as the method is able to analyze lithium distribu-
tion with high spatial resolution (Leemann et al.,
2014; Bernard and Leemann, 2015). Specific surface
area has a significant influence on the effectiveness
of LiNO3 to suppress ASR as demonstrated with
the SiO2 -cylinder and the SiO2 -filter. As such it can
explain, why the effectiveness of LiNO3 and the re-
quired dosage to suppress ASR is dependent on the
type of aggregate (Fournier et al., 2003; Folliard et
al., 2006; Tremblay et al., 2007). The specific surface
area and the dissolution kinetics of reactive SiO2 in
natural aggregates can be expected to vary consid-
erably and is surely difficult to assess. Therefore,
the required dosage of LiNO3 is best determined
with expansion tests.
V. Summary and conclusions
In this study the effect of LiNO3 was analyzed in
concrete with natural aggregates and in mortar with
two artificial SiO2 -aggregates of different specific
surface area. Based on the results obtained in con-
crete and mortar the following conclusions can be
drawn:
• The reaction products in concrete Li-0.74 are
dense and texture-less contrasting to the lay-
ered texture and higher porosity of the reaction
product in concrete Ref. Moreover, the sodium,
potassium and calcium content in the reaction
products of concrete Li-0.74 are significantly
reduced. This indicates that the decrease in
expansion can be attributed to a protection of
reactive minerals from further reaction by the
dense reaction product and likely to a reaction
product with less expansion potential.
• The dense SiO2 -cylinder in mortar Li-0.74
shows no reaction products except of a thin
and dense surface layer (⇠ 1 µm wide). This
surface layer protects the SiO2 -cylinder from
further reaction. This seems to verify the hy-
231
 Leemann

Figure 6: Backscattered electron image (BSE) and element mappings of calcium (Ca), potassium (K) and silicon (Si) at
the interface of the SiO2 -filter to cement paste in mortar Li 0.76 after 6 (row on left) and 24 weeks (row on right).
Reaction products correspond to the areas with a lower silicon signal compared to the unreacted SiO2 of the filter and
to the areas showing a potassium signal.

232
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
pothesis made based on the results of con-
crete Li-0.74 on the protective effect of lithium-
containing reaction products.
• The SiO2 -filter in mortar Li-0.74 contains a sub-
stantial amount of reaction products even if it
is lower than in the one in mortar Ref. The
occurrence of these reaction products in the
SiO2 -filter clearly indicates that the difference
in effectiveness of LiNO3 to prevent the forma-
tion of reaction products is dependent on the
specific surface area of aggregates. The surface
layer on the SiO2 -cylinder likely formed soon
after sample production when the lithium con-
tent in the pore solution was still high. In the
SiO2 -filter a reaction fronts ingresses with time.
This limits the effectiveness of lithium to pre-
vent ASR. At the time the reaction front reaches
the inner parts of the SiO2 -filter, lithium con-
centration in the mortar pore solution has de-
creased to a level where it is only partly ef-
fective. Moreover, the lithium content of the
ingressing solution may continuously decrease
by the ongoing reaction with SiO2 . Sodium and
potassium may penetrate further than lithium
and lead to the formation of reaction products
in the filter not (yet) reached by lithium.
• LiNO3 is able to decrease ASR-induced con-
crete expansion when it is added in a sufficient
amount that depends on the specific surface
area of reactive SiO2 in the aggregates.
• It is essential to investigate the lithium distri-
bution in the two artificial SiO2 -aggregates to
verify the effect of specific surface area on the
effectiveness of LiNO3 . Such an investigation
could be made with ToF-SIMS on impregnated
and polished samples of the mortars as the
method permits lithium analysis with high spa-
tial resolution (Leemann et al., 2014; Bernard
and Leemann, 2015).
Acknowledgement
The author would like to thank K. Burkhard, D.
Käppeli and B. Ingold for sample production and
preparation.
References
Bernard, L., Leemann, A. (2015): "Assessing the potential of ToF-
SIMS as a complementary approach to investigate cement-based mate-
rials - Applications related to alkali-silica reaction." Cement and Concrete
Research. Vol. 68, 156-165.
Feng, X., Thomas, M.D.A., Bremner, T.W., Folliard, K.J., Fournier, B.
(2010a) New observations on the mechanism of lithium nitrate against
alkali silica reaction (ASR), Cement and Concrete Research. Vol. 40, 94-101.
Feng, X., Thomas, M.D.A., Bremner, T.W., Folliard, K.J., Fournier, B.
(2010b) Summary of research on the effect of LiNO3 on alkali-silica reac-
tion in new concrete, Cement and Concrete Research. Vol 40, 636-642.
Folliard, K. J., Thomas, M. D. A., Fournier, B., Kurtis, K. E., Ideker, J.
H. (2006): "Interim recommendations for the use of lithium to mitigate or
prevent alkali-silica reaction ASR." Publication No. FHWA-HRT-06-073.
Fournier, B., Stokes, D., Ferro, A. (2003): "Comparative field and
laboratory investigations on the use of supplementary cementing ma-
terials SCMs and lithium-based admixtures to control expansion due
to alkali-silica reaction ASR in concrete." Proceedings of the 6th CAN-
MET/ACI International Conference on Durability of Concrete, Thessa-
loniki, Greece, 823-851.
Kawamura, M., Fuwa, H. (2003): "Effects of lithium salts on ASR gel
composition and expansion of mortars." Cement and Concrete Research.
Vol. 33, 913-919.
Kirkpatrick, R.J., Kalinichev, A.G., Hou, X., Struble, L. (2005): "Exper-
imental and molecular dynamics modeling studies of interlayer swelling:
water incorporation in kanemite and ASR gel." Materials and Structures.
Vol. 38, 449-458.
Leemann, A., Lötscher, L., Bernard, L., Le Saout, G., Lothenbach, B.,
Espinosa-Marzal R.M. (2014): " Mitigation of ASR by the use of LiNO3 -
Characterization of the reaction products. " Cement and Concrete Research.
Vol. 59, 73-86.
Leemann, A., Merz, C. (2013): "An attempt to validate the ultra-
accelerated microbar and the concrete performance test with the degree
of AAR-induced damage observed in concrete structures." Cement and
Concrete Research. Vol. 49, 29-37.
MB SIA 2042 (2012):Vorbeugung von Schäden durch die Alkali-
Aggregat-Reaktion (AAR) bei Betonbauten. Schweizerischer Ingenieur-
und Architektenverein, Zürich.
Mitchell, L.D., Beaudoin, J.J., Grattan-Bellew, P. (2004): "The effects
of lithium hydroxide solution on alkali silica reaction gels created with
opal. Cement and Concrete Research. Vol. 34, 641-649.
Tremblay, C., Bérubé, M.A., Fournier, B., Thomas, M.D.A., Folliard,
K.J., Nkinamubanzi, P.C. (2008): "Use of the accelerated mortar bar test
to evaluate the effectiveness of LiNO3 against alkali-silica reaction - part
1: pore solution chemistry and influence of various parameters." Journal
of ASTM International. Vol. 5, 1-17.
Tremblay, C., Bérubé, M.A., Fournier, B., Thomas, M.D.A., Folliard,
K.J. (2007): "Effectiveness of lithium-based products in concrete made
with Canadian natural aggregates susceptible to alkali-silica reactivity."
ACI Materials Journal. Vol. 104, 195-205.
Tremblay, C., Bérubé, M.A., Fournier, B., Thomas, M.D.A., Folliard,
K.J. (2010): Experimental investigation of the mechanisms by which
LiNO3 is effective against ASR, Cement and Concrete Research. Vol. 40,
583-597.
233
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

234
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Alkali-silica reaction: Petrographic

characterization and performance tests of
Portuguese aggregates
Violeta Ramosa,⇤ , Isabel Fernandesa , Fernando Noronhaa
António Santos Silvab & Dora Soaresb
a Department of Geosciences, Environment and Spatial Planning/Centro de Geologia, University of Porto, Porto, Portugal
b Materials Department, National Laboratory for Civil Engineering, Lisboa, Portugal
⇤ violetaramos@fc.up.pt

Abstract

Portuguese granitic rocks are commonly used as aggregates for concrete, especially in the North and Center of the
country. Although this type of rock is regarded as slow/late reactive or even non-reactive to alkalis, numerous reports
indicate its deleterious behaviour in several major structures. The urge to avoid this problem led to the development
of the Portuguese project IMPROVE (Improvement of performance of aggregates in the inhibition of alkali-aggregate
reactions in concrete). Under the scope of this investigation, Portuguese aggregates of different lithologies have been
studied, including the granitic ones presented in this paper. The used methods included petrographic characterization
and performance tests: mortar-bar and concrete expansion tests. The results indicate that the content of microcrystalline
quartz correlates better with the results of the concrete prism expansion tests than with the ones of the mortar-bar
expansion test. In this paper, evidence is given to the type of quartz of the granitic aggregates which proved to be reactive
both in expansion tests and/or in concrete structures.

Keywords: ASR; granitic aggregates; petrographic characterization; expansion tests

I. Introduction ubility due to a large surface area. The presence

Alkali-aggregate reaction, which includes the alkali-
silica reaction (ASR) and the alkali-carbonate re-
action, is a chemical reaction in either mortar or
concrete involving the hydroxyl (OH ) ions associ-
ated with the alkalis from Portland cement or other
sources, with certain mineral phases that may be
present in the aggregates (RILEM AAR-6.1, 2013)
leading to the premature distress and loss in ser-
viceability of the affected concrete structures.
The importance of textural features of aggregates
in the development of ASR started to grab the atten-
tion of researchers around the 1950s (Brown, 1955).
In the following two decades, several studies were
published trying to establish a correlation between
the values of the undulatory extinction of quartz
crystals and the deformation degree of the crys-
tal’s lattice (DeHills and Corvallán, 1964; Gogte,
1973). This parameter was later on used in the in-
vestigation of the potential reactivity of aggregates
(e.g. Dolar-Mantuani, 1981). Afterwards, it was
verified that quartz grains presenting high undu-
latory extinction angles were usually associated to
the presence of microcrystalline quartz (< 100 µm)
(Grattan-Bellew, 1992), which showed a high sol-
of microcrystalline quartz then became the most
important parameter in the evaluation of the po-
tential reactivity to alkalis of quartz-bearing rocks.
Tiecher et al. (2012), after investigating deformed
quartz-rich aggregates, concluded that rocks mainly
composed by quartz crystals with a well-defined
undulatory extinction and deformation bands are
highly prone to the development of ASR. Shayan
(1993) had already attributed the occurrence of ASR
to the presence of microcrystalline quartz, stating
that the reaction occurs at the grains’ boundaries
in zones that contain the grained quartz and with
a large surface area. In 1995, Wigum proposed
a new approach for the quantification of the av-
erage surface area of quartz crystals in cataclastic
rocks which allowed to estimate the quantity of
microcrystalline quartz present in the rock. The
presence of myrmekites, defined as a vermicular
intergrowth of quartz and sodic plagioclase formed
by replacement of K-feldspar, typically in deformed
granitic rocks (Vernon, 2004), was also considered
by Wigum (1995) as a preferential site for ASR.
Several studies also showed that there is a posi-
tive correlation between the amount of microcrys-
talline quartz, resulting from the conversion of a

235
 Ramos et al.

Table 1: Codification, lithology and age of the studied aimed not only for predicting the behaviour of Por-
granitic rocks according to Dias et al. (2010). tuguese granitic rocks, but also for establishing the
(syn-D3 granites 312-321 Ma, late-D3 305-312 Ma; late expansion test that best suits the field performance
to post-D3 300Ma) of this slow reactive aggregates.

Aggregate Lithology Age

GR1 granite late-D3
GR2 granite syn-D3
GR3 granite late- to post-D3
GR7 granite late- to post-D3
GR17 granite late-D3
GR18 granite syn- to late-D3
GR19 granite syn-D3
GR20 granite syn- to late-D3
mineral grain which previously showed undula-
tory extinction into an agglomeration of smaller
new grains in response to deformation, and mor-
tar and/or concrete expansion tests (e.g. Grattan-
Bellew, 1992; Kerrick and Hooton, 1992; Thomson
et al., 1994; Wenk et al., 2008).
In spite of microcrystalline quartz being pointed
out as the major phase responsible for ASR in slow
reactive aggregates, other minerals such as micas
and feldspars have been suggested to be also to be
prone to this deleterious reaction (e.g. Yan et al.,
2004; Hagelia and Fernandes, 2012).
Though granitic aggregates are usually pointed
out as being innocuous or unlikely reactive to al-
kalis (Concrete Society, 2010), the literature on the
subject shows that this type of rocks can behave,
in some parts of the world, as reactive (RILEM
AAR-1, 2003). In Portugal, granitic rocks are the
second most commonly used material in the con-
struction of major concrete structures and several
ASR cases involving this kind of aggregates have
been reported and investigated (e.g. Fernandes et
al., 2004; Santos Silva, 2005). The behaviour of such
material towards ASR is complex and the current
methodologies of assessment have not proven to-
tally satisfactory. The study presented in this paper,
and which is included in the Project IMPROVE
(Improvement of performance of aggregates in the
inhibition of alkali-aggregate reactions in concrete),
II. Materials and Methods
Under the scope of the Portuguese project IM-
PROVE, eight granitic aggregates from active quar-
ries from northern and central Portugal were in-
vestigated. All the investigated granitic rocks out-
crop in the Central Iberian Zone and are Hercynian
in age (Dias et al. 2010). These are designated
according to the nomenclature established by the
aforementioned project (Table 1).
The aggregates were assessed by petrographic
characterization followed by expansion tests,
namely, the accelerated mortar-bar test at 80 C and
the concrete prism tests at 60 C and 38 C. A total
of about 100 kg of crushed aggregate of different
grain sizes was collected from each quarry. Hand
samples were selected in order to produce thin-
sections with the following dimensions: 25*35*0.03
mm
Petrographic characterization
The thin-sections were studied with an optical po-
larizing microscope Nikon Eclipse E 400 POL, cou-
pled with an automatic PELCON point-counter and
with a photographic camera Zeiss Axio Cam, avail-
able at "Centro de Geologia da Universidade do
Porto" (CGUP). The evaluation of the reactivity fol-
lowed the methodology described in RILEM AAR-
1.1 (2013) and in LNEC E 461 (2007), which de-
fine similar reactivity classes (Class I - non-reactive;
Class II - potentially reactive; Class III - reactive). It
was based in the quantification of microcrystalline
quartz (< 100 µm) plus myrmekites, and in the mea-
surement of the undulatory extinction angle. This
last parameter was evaluated according to the pro-
cedure described in DeHills and Corvállan (1964)
and with the classification in the Portuguese speci-
fication LNEC E 415 (1993).

Table 2: Testing periods and criteria used for the expansion tests for considering an aggregate as potentially reactive to
alkalis.

Test method Limits/Criteria

ASTM C 1260 > 0.10 % at 14 days > 0.10 % at 21 days
RILEM AAR-0 2013 Shayan 2007
RILEM AAR-4.1 > 0.02 % at 12 weeks > 0.03 % at 15 weeks (± 0.010 % > 0.03 % at 20 weeks (±
- uncertainty band) 0.010 % - uncertainty band)
LNEC E 461 2007 RILEM AAR-0 2013 Lindgard et al. 2010
RILEM AAR-3 > 0.10 % at 1 year (0.05 % - 0.10 % ambiguous) > 0.05 % at 1 year
RILEM AAR-0 2013 LNEC E 461 2007

236
 Ramos et al.
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

200 µm 200 µm
A B

200 µm C 100 µm
D

50 µm E 200 µm F

200 µm 100 µm
G H

Figure 1: Microphotographs of the investigated aggregates. Photos A, B, D, E, F, G and H in cross-polarized light (XPL).
Figure 1: Microphotographs of thelight
Photo E in plane polarized investigated
(PPL). A) GR1 aggregates. Photos
– Mycrocrystallinne A, B, quartz
“globular” D, E,in F, G and Hwith
a plagioclase in polysynthetic
cross-polarized light
(XPL). Photo E in plane polarized light (PPL). A) GR1 - Mycrocrystallinne " globular" quartz in
and Carlsbad twinning; B) GR2 – Chess-board subgrain pattern in a quartz crystal with irregular boundaries; C) aGR3
plagioclase
– with
Microcrack filled with oxides crossing the essential minerals of aggregate GR3; E) GR7 – Myrmekites in plagioclase; F)
polysyntheticGR18
and–Carlsbad twinning;
Chessboard subgrain patternB) GR2
in an - Chess-board
elongated subgrain
quartz crystal (on pattern in quartz
the top). Recrystallized a quartz
occurscrystal with irregular
in the interface
boundaries; C) GR3 of the-larger
Microcrack filled– Deformation
grains; G) GR19 with oxides crossing
lamellae the
in quartz; H) essential
GR20 – Quartz minerals of aggregate
with serrated boundaries andGR3; E) GR7 -
intercrystalline microcracks.
Myrmekites in plagioclase; F) GR18 - Chessboard subgrain pattern in an elongated quartz crystal (on the top).
Recrystallized quartz occurs in the interface of the larger grains; G) GR19 - Deformation lamellae in quartz; H) GR20
- Quartz with serrated boundaries and intercrystalline microcracks.
5

Expansion tests imens. The cement had an alkali content ranging

The aggregates were crushed and/or sieved and
weighed according to the requirements of ASTM C
1260 (2007) mortar-bar test, RILEM AAR-4.1 (2013)
and RILEM AAR-3 (2013) concrete prism tests. All
expansion tests were carried out at " Laboratório
Nacional de Engenharia Civil" (LNEC). A CEM I
42.5 R cement from Alhandra (Portugal) was used
for the manufacture of mortar and concrete spec-
from 0.86 to 0.89 Na2 Oeq. Table 2 displays the var-
ious testing periods and expansion criteria used
for classifying aggregates as potentially reactive to
alkalis.
Post-mortem concrete petrography
In order to confirm the occurrence of ASR in the
concrete prisms and mortar-bars tested in labora-

237
 Ramos et al.

Table 3: Results of the petrographic characterization of the studied granitic aggregates.

Aggregate Brief petrographic description Undulatory extinction angle ( ) Potentially reactive silica
of quartz crystals (median) and forms (< 100 µm) (vol%)
respective class and reactivity class
GR1 Medium-grained with porphyritic ten- 16.0 - Moderate Mycrocrystalline quartz;
dency granite. Microcracks are frequent, myrmekites - 2.4 - Class II
especially in quartz and plagioclase crys-
tals.
GR2 Medium to coarse-grained porphyritic 20.0 - Moderate Tectonized mycrocrystalline
granite with strong deformation evi- quartz; myrmekites - 21.7 -
dences. Class II
GR3 Medium to coarse-grained porphyritic 18.0 - Moderate Mycrocrystalline quartz;
granite. The more well-developed crys- myrmekites - 1.0 - Class I
tals are crossed by abundant microc-
racks.
GR7 Medium to coarse-grained porphyritic 14.0 - Weak Myrmekites - 0.1 - Class I
granite.
GR17 Fine to medium-grained granite. 15.0 - Moderate Mycrocrystalline quartz;
myrmekites - 1.5 - Class I
GR18 Medium-grained porphyritic granite 23.0 - Moderate Tectonized mycrocrystalline
with deformation evidences. quartz; myrmekites - 4.8 -
Class II
GR19 Medium-grained granite. 20.0 - Moderate Mycrocrystalline quartz;
myrmekites - 1.4 - Class I
GR20 Fine to medium-grained granite. Microc- 19.0 - Moderate Mycrocrystalline quartz;
racks are abundant. myrmekites - 1.2 - Class I
Classes of undulatory extinction angle: Weak – [0 - 14 ]; Moderate – [15 - 24 ]; Strong – [25 - 35 ]

tory conditions by petrographic methods, selected
samples were prepared for the production of pol-
ished thin-sections. Their observation under the
optical microscope aimed at the identification of
the manifestations of ASR in the concrete and mor-
tar, namely: cracks in the aggregates; cracks in the
cement paste or in the interfaces aggregate-binder;
presence of alkali-silica gel in the cracks; presence
of alkali-silica gel in the voids; and gel replacing
fine aggregate particles.
SEM-EDS (High Resolution Schottky Environ-
mental Scanning Electron Microscope: Quanta 400
FEG ESEM/EDAX Genesis X4M: 15 kV and 10.0
mm working distance) was used to confirm the oc-
currence and to identify the composition of ASR
products in the polished thin-sections previously
studied by petrographic methods. The SEM-EDS
examination was performed along X-ray microanal-
ysis and electron backscattered diffraction analysis
at "Centro de Materiais da Universidade do Porto"
(CEMUP). Samples were coated with carbon by a
vacuum evaporator equipment JEOL JEE - 4X. All
the images presented in this paper were made in
backscattering mode.
III. Results
Petrographic characterization
The results of the petrographic characterization of
the aggregates are summarized in Table 3. Accord-
ing to the determined percentages of microcrys-
talline quartz of each aggregate, only aggregates
GR1, GR2 and GR18 present values over 2.0 vol.%,
being classified as both by RILEM AAR-1.1 (2013)
and LNEC E 461 (2007) as Class II (potentially re-
active). All the other aggregates are classified as
Class I (non-reactive) by the aforementioned recom-
mendation/specification. Figure 1 illustrates the
evidence of potentially reactive silica forms as well
of undulatory extinction in the investigated aggre-
gates. Sometimes, microcrystalline quartz presents
as tectonized quartz.
Expansion tests
The performance of the granitic aggregates in the
expansion tests is displayed in Figure 2.
Figure 2A shows that all aggregates used in the
accelerated mortar-bar test at 80 C expansions
lower than 0.10 % at 14 days. According to the
criteria in ASTM C 1260 (2007), these results in-
dicate a non-reactive behaviour. The test period
was extended till 28 days, according to what is sug-
Table 4: List of concrete prims with granitic aggregates
of which thin-sections were prepared and studied.
Aggregate AAR-3 AAR-4.1
GR1 X X
GR2 X X
GR3 - X

238
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Ramos et al.

0.30 0.100
GR1 GR2 GR3 GR7

GR17 GR18 GR19 GR20 0.080

Expansion (%)
0.20

Expansion (%)
0.060

0.040
0.10

0.020

0.00 0.000
0 7 14 21 28 0 4 8 12 16 20
Time (weeks)
Time (days) GR1 GR2 GR3 GR7 GR17 GR18 GR19 GR20

Santos Silva and Braga Reis (2000)

A B
RILEM AAR-0 (2013) and Lindgård et al. (2010)

RILEM AAR-0 (2013)

LNEC E 461 (2007)

0.10
GR1 GR2 GR3 GR7
0.09 GR17 GR18 GR19 GR20
0.08
Expansion (%)

0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0 28 56 84 112 140 168 196 224 252 280 308 336 364
Time (days) C
RILEM AAR-0 (2013)

Figure 2: Performance of the granitic aggregates in the expansion tests

Figure 2: Performance of the granitic aggregates
(A – ASTM C 1260; Bin– the expansion
RILEM AAR-4.1; tests.
and C –(A - ASTM
RILEM C 1260; B - RILEM AAR-4.1;
AAR-3).
and C - RILEM AAR-3)
Aggregate GR1
The thin-section produced from the RILEM AAR-3 concrete prism just contained alkali-silica gel
filling one void, confirmed under SEM. The thin-section of RILEM AAR-4.1 exhibits more abundant
gested in manifestations
Santos Silvaofand ASRBraga
such as:Reis (2000),
gel lining and and, Allthin
filling voids; other aggregates
layers present
of gel in cracks from final expansions
debonding between lower
after this period,
some aggregate particles and the cement paste. Figure 3 presents the photomicrographs of voids as wellnon-reactive.
it is confirmed that all aggregates than 0.05 % which classifies them as as the
are classified
imagesas obtained
non-reactive.
by SEM However, aggregates
and the qualitative Amongst those aggregates, GR7 is the only one
analyses by EDS.
GR2, GR18 and GR20 The composition
present theof the gel in the
highest interfaces with
expan- aggregate
whose particlesseems
expansion is slightly
to different from that
have levelled in
off.
the voids. Gel is mainly composed of Si and Ca with very low contents of K and Na.
sions (0.10 %, 0.10 % and 0.08 %, respectively). It
must be noted that all mortars are still showing
Aggregate GR2 Post-mortem concrete petrography
increasing expansions at the of
This sample end of this
granite gavetest.
the highest expansion results in what concerns the expansion tests. The
concrete prisms both from RILEM AAR-3 and RILEM AAR-4.14 are
In Table thethose
list that exhibit theselected
of samples most severe
for post-
In whatmanifestations
concerns the concrete
of ASR from allprism testsamples.
the tested at 60 The thin-sections analysis showed there is gel mainly in the
C (Figurecracks
2B), and
only aggregate
voids that occurGR7
in theisinterfaces
classified as the aggregate particles and the cement paste. However, in is pre-
between mortem analysis by concrete petrography
the according
non-reactive thin-section of
to the
theRILEM AAR-3
criteria in LNECprism Ethere sented.
461are also cracks crossing the aggregate particles. These cracks
are wider close to the interface with the cement paste and are thinner inside of the particles. The alkali-silica gel
(2007) andidentified
(Lindgard et al., 2010). Yet, according
is compact, non-crystalline and exhibits several shrinkage cracks.
to the criterion inFigureRILEM AAR-0
4 presents (2013),
images besides
and the Aggregate
qualitative analysis of gelGR1
in the border of the aggregate particles
aggregatefrom
GR7, an aggregate
RILEM AAR-3 GR1prism.
is also
Thisconsidered
dense gel is disposed in layers lining the void and crack. The colour is
slightly darker
as non-reactive. The(brownish)
absence thanof acommonly
plateaufound
at 20in voidsThe thin-section
and it produced
presents abundant shrinkagefrom theBesides
cracks. RILEM theAAR-3
procedures applied during the production of the thin-sections, concrete prismtypical
the vacuum just contained alkali-silica
of SEM observations gel filling
can also
weeks indicates that thecracks.
create shrinkage levelling off of
As already the there
stated, expan-
is some variation in the content of Ca in theSEM.
analyses performed
one void, confirmed under The thin-section
sion lines has not the
although been gel reached
exhibits a for anytexture.
similar of the tested
of RILEM AAR-4.1 exhibits more abundant man-
aggregates. The results obtained for the prism tested according to RILEM AAR-4.1 are presented in Figure 4. The
gel is compact and forms in layers in the aggregate particle ifestations
boundaries. of ASR such as: gel lining and filling
As for the concrete prism test at 38 C (Fig- voids; thin layers of gel in cracks from debonding
ure 2C), and according to the criteria in RILEM between some aggregate particles and the cement
AAR-0 (2013), results under 0.05 % at 1 year of paste. Figure 3 presents the photomicrographs of
testing, indicate a non-reactive behaviour, whilst voids as well as the images obtained by SEM and
results over 0.10 % indicate reactive aggregates. For the qualitative analyses by EDS.
values between 0.05 % and 0.10 %, there is still no The composition of the gel in the interfaces with
agreement towards the interpretation of the aggre- aggregate particles is slightly different from6that in
gate’s behaviour. As so, the criterion established the voids. Gel is mainly composed of Si and Ca
in LNEC E 461 (2007) was followed. According with very low contents of K and Na.
to this, and based on the Portuguese experience,
an expansion above 0.05 % at 1 year is enough to
Aggregate GR2
regard an aggregate as potentially reactive. This
is the case for aggregate GR2, which presents a fi- This sample of granite gave the highest expan-
nal expansion of 0.06 %, though still non-stabilized. sion results in what concerns the expansion tests.

239
 Ramos et al.
The concrete prisms both from RILEM AAR-3 and
RILEM AAR-4.1 are those that exhibit the most
severe manifestations of ASR from all the tested
samples. The thin-sections analysis showed there
is gel mainly in the cracks and voids that occur in
the interfaces between the aggregate particles and
the cement paste. However, in the thin-section of
the RILEM AAR-3 prism there are also cracks cross-
ing the aggregate particles. These cracks are wider
close to the interface with the cement paste and are
thinner inside of the particles. The alkali-silica gel
identified is compact, non-crystalline and exhibits
several shrinkage cracks.
Figure 4 presents images and the qualitative anal-
ysis of gel in the border of the aggregate particles
from an RILEM AAR-3 prism. This dense gel is
disposed in layers lining the void and crack. The
colour is slightly darker (brownish) than commonly
found in voids and it presents abundant shrinkage
cracks. Besides the procedures applied during the
production of the thin-sections, the vacuum typi-
cal of SEM observations can also create shrinkage
cracks. As already stated, there is some variation
in the content of Ca in the analyses performed al-
though the gel exhibits a similar texture.
The results obtained for the prism tested accord-
ing to RILEM AAR-4.1 are presented in Figure 4.
The gel is compact and forms in layers in the aggre-
gate particle boundaries.
Aggregate GR3
The thin-sections produced from the concrete
prisms tested under the RILEM AAR-4.1 condi-
tions show very discrete manifestations of ASR. Gel
occurs in slim cracks in the boundaries of the ag-
gregate particles and in rare cracks crossing the
aggregate particles. The examination performed
with SEM-EDS confirms that the gel is composed
mainly of Si with high content of Ca and very low
content of K and Na. However, some Al was also
detected.
A crack crossing an aggregate particle was also
analysed in order to identify the material inside the
crack. It shows a particulate texture not common in
gel. The EDS analysis indicates a composition close
to that of gel but with very high content of Ca and
scarce Al (Figure 5).
IV. Discussion
The obtained results for each method are summa-
rized in Table 5.
According to the petrographic characterization,
only aggregates GR1, GR2 and GR18 present per-
centages of microcrystalline quartz above 2.0 vol.%,
which classify them as Class II according to RILEM
AAR-1.1 (2013) and LNEC E 461 (2007) (Ramos,
240
2013). Aggregate GR2 is clearly distinguishable
for having the higher volume of microcrystalline
quartz, which presents itself mainly as tectonized
quartz, followed by aggregates GR18 and GR1.
Though all the other granitic aggregates are re-
garded as non-reactive (Class I), the Portuguese
experience, in which LNEC E 461 (2007) is based,
establishes the use of concrete expansion tests as
compulsory to complement the petrographic char-
acterization.
As demonstrated in Figure 6, there is a positive
but moderate correlation between the percentage
of microcrystalline quartz and the undulatory ex-
tinction angle (correlation coefficient r = 0.41). For
example, aggregate GR2 exhibits by far the highest
content in microcrystalline quartz, but it is aggre-
gate GR18 that shows the quartz grains with the
highest undulatory extinction angles. However,
both aggregates are classified as Class II by petro-
graphic examination. Also, aggregate GR19, which
is rated as Class I, presents a very similar value
of extinction angle to aggregate GR2, but a much
lower content of microcrystalline quartz in compar-
ison to the same aggregate. Furthermore, the aggre-
gates presenting the lowest angles of undulatory
extinction seem to establish a better correlation with
the percentages of microcrystalline quartz, plotting
close to the regression line.
The above is in accordance with West (1991, 1994),
who pointed out undulatory extinction angle as a
possible indicator of potential alkali-reactivity of
rocks, rather than a diagnostic feature.
The comparison between the results of the pet-
rographic examination and the ASTM C 1260 con-
firmed that the accelerated mortar-bar test was not
able to support the results of the petrographic ex-
amination in what concerns aggregates GR1, GR2
and GR18, classifying all the granitic aggregates as
non-reactive. Though this method is widely used
for the assessment of aggregates for concrete pur-
poses, it is still not well recognized that its limits
correspond to the reactivity of some of the slow
reactive aggregates, leading to their incorrect classi-
fication as false non-reactive (Shayan, 2007), which
is surely the case of aggregate GR2 and GR18. This
method has not been successful when performed
on Portuguese granitic aggregates and, therefore, it
has been considered as inappropriate to evaluate
the alkali-reactivity of this type of slow reactive
rocks (Santos Silva, 2005). The explanation for the
results obtained might remain in the findings re-
ported by Lu et al. (2006a,b), in which it is stated
that the crushing of rocks to obtain very fine aggre-
gates can destroy the original microstructure and
thus underestimate the alkali reactivity of the rocks
in accelerated mortar-bar tests. In fact, the influence
of the size of the aggregate particles used in labora-
tory tests has been discussed and results published
 Ramos et al.
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

A1 A2 A3

B1 B2
Ramos et al. B3

A1 A2 A3

Figure 3: Post-mortem petrography of concrete prisms with aggregate GR1 tested according with RILEM AAR-
4. (A) Voids filled by gel observed in the thin-section from the concrete prism RILEM AAR-4: 1) Under optical microscope
Figure 3: Post-mortem petrography of concrete prisms with aggregate GR1 tested according with RILEM AAR-4. (A)
(PPL); 2) Image obtained by SEM; 3) Spectrum by EDS. (B) Gel in the interface of an aggregate particle and the cement
Voids filled by gel observed paste: in the thin-section
1) Under from
optical microscope the 2)concrete
(PPL); prismbyRILEM
Images obtained AAR-4:by1)EDS.
SEM; 3) Spectrum Under optical microscope
(PPL); 2) Image obtained by SEM; 3) Spectrum by EDS. (B) Gel in the interface of an aggregate particle and the
cement paste: 1) Under optical microscope (PPL); B2
2) Images obtained by SEM; 3) Spectrum by EDS.
B1
B3
A1 A2 A3

(e.g. Lu et al., 2006a,b; Multon et al., 2010). non-reactive (Class I); although classified as Class II
by the petrographic examination, aggregate GR1 is
From the comparison between the results of rated as potentially reactive by the criteria in LNEC
RILEM AAR-4.1 and the ones from the petrographic E 461 (2007) and Lindgard et al. (2010); although
examination it can be said that: all the criteria used classified as Class I by the petrographic examina-
for RILEM AAR-4.1 (twelve weeks, fifteen weeks tion, aggregates GR3, GR17, GR19 and GR20 are
and twenty weeks) Figure confirm aggregate
3: Post-mortem GR2 as
petrography of poten-
concrete prismsclassified
with aggregateasGR1
potentially reactive
tested according by RILEM
with RILEM AAR- AAR-4.1
tially reactive
4. (A)in agreement
Voids with the
filled by gel observed results
in the of from
thin-section the the concrete
by prism RILEM AAR-4: 1) Under optical microscope
all interpretation criteria.
(PPL); 2) Image obtained by SEM; 3) Spectrum by EDS. (B) Gel in B2the interface of an aggregate particle and the cement
petrographic examination (ClassB1 II); there is also
paste: 1) Under optical microscope (PPL); 2) Images obtained by SEM; 3) Spectrum by EDS. B3
consensus in the classification of aggregate GR7 as For a better interpretation of the aforementioned

A1 A2 A3

Figure 4: Post-mortem petrography of concrete prisms with aggregate GR2. Gel in the interface of an aggregate
particle and the cement paste, in the thin-section from the concrete prism tested according with RILEM AAR-3 (A): 1) Under
optical microscope (PPL); 2) Image obtained by SEM; 3) Spectrum by EDS. (B) Gel in a void and filling a crack and void in
the interface of cement paste with a granitic particle, in the thin-section from the concrete prism test RILEM AAR-4.1: 1)
Under optical microscope (PPL); 2) Image obtained by SEM; 3) Spectrum by EDS.

B1 B2
B3

Figure 4: Post-mortem petrography of concrete prisms with aggregate GR2. Gel in the interface of an aggregate
particle and the cement paste, in the thin-section from the concrete prism tested according with RILEM AAR-3 (A): 1) Under
Figure 4: Post-mortem petrography of concrete prisms with aggregate GR2. Gel in the interface of an aggregate
optical microscope (PPL); 2) Image obtained by SEM; 3) Spectrum by EDS. (B) Gel in a void and filling a crack and void in
particle and the cement paste, in the
the interface of cement paste withthin-section from
a granitic particle, thethin-section
in the concretefrom prism testedprism
the concrete according with
test RILEM RILEM
AAR-4.1: 1) AAR-3 (A):
Under optical microscope (PPL); 2) Image obtained by SEM; 3) Spectrum by
1) Under optical microscope (PPL); 2) Image obtained by SEM; 3) Spectrum by EDS. (B) Gel in a void and filling a EDS.

crack and void in the interface of cement paste with a granitic particle, in the thin-section from the concrete prism test
RILEM AAR-4.1: 1) Under optical microscope (PPL); 2) Image obtained by SEM; 3) Spectrum by EDS.

7 241
 Ramos et al.

Table 5: Summary of the results achieved by the different methods applied on the assessment of the studied aggregates.

Method Interpretation criteria GR1 GR2 GR3 GR7 GR17 GR18 GR19 GR20
RILEM AAR-1.1 (2013) Class II Class II Class I Class I Class I Class II Class I Class I
Petrographic character- LNEC E 461 (2007) Class II Class II Class I Class I Class I Class II Class I Class I
ization
Microcrystalline quartz 2.4 21.7 1.0 0.1 1.5 4.8 1.4 1.2
(vol.%)
Class of undulatory ex- Mod. Mod. Mod. Weak Mod. Mod. Mod. Mod.
tinction
ASTM C 1260 RILEM AAR-0 (2013) NR NR NR NR NR NR NR NR
Santos Silva and Braga NR NR NR NR NR NR NR NR
Reis (2000)
LNEC E 461 (2007) PR PR PR NR PR PR PR PR
RILEMAAR-4.1 RILEM AAR-0 (2013) NR PR PR NR PR PR PR PR
Lindgard et al. (2010) PR PR PR NR PR PR PR PR
RILEM AAR-3 RILEM AAR-0 (2013) NR PR NR NR NR NR NR NR
NR - non-reactive; PR - potentially reactive; Class of undulatory extinction angle: Weak - [0 - 14 ]; Moderate (Mod.) - [15 - 24 ]; Strong - [25 - 35 ]

described, it is important to keep in mind that
granitic aggregates are regarded as slow reactive ag-
gregates in terms of ASR development. Therefore,
it will more adequate to admit that for RILEM AAR-
41 the criterion in Lindgard et al. (2010), which uses
a test period of 20 weeks. In fact, this is one of the
criteria that classifies a larger number of aggregates
as potentially reactive, though some reactivity val-
ues are comprised in the uncertainty band 0.03 %
±0.010 % suggested in Lindgard et al. (2010) and
RILEM AAR-0 (2013). However, although the crite-
rion in LNEC E 461 (2007) uses the lowest expan-
sion limit and test period, the interpretation of the
results in RILEM AAR-4.1 shows the effectiveness
Ramos et al.
graphic characterization is not, by itself, sufficient
for classifying the reactivity of granitic aggregates,
seems to be justified since the potentially reactive
behaviours observed in RILEM AAR-4.4 were not
identified by petrographic characterization. It also
cannot be ignored that micas and feldspars can par-
ticipate in ASR (Gillot and Rogers, 1994; Bérubé
et al., 2002; Constantiner and Diamond, 2003; Lee-
man and Holzer, 2005; Lu et al., 2006c; Hagelia
and Fernandes, 2012) releasing alkalis to the pore
solution. These minerals are quite abundant in
the investigated granitic aggregates, and although
some leaching can occur during the concrete prism
test (Lindgard et al., 2012), accordingly to the litera-

of this criterion in detecting all aggregates regarded ture the alkalis released by some minerals should
as potentially reactive by the criterion in Lindgard be sufficient to promote ASR when silica reactive
et al. (2010), which uses a test period of 20 weeks. constituents are present. Table 6 refers to the consti-
TheAggregate
potentialGR3reactivity detected in RILEM AAR-4.1 tutive minerals of the investigated aggregates which
for aggregates TheGR3
thin-sections
and GR20 produced from the
is, possibly, concrete prisms
justified tested under
are regarded the RILEM
by LNEC E 415AAR-4.1 conditions
(1993) and LNEC E
show very discrete manifestations of ASR. Gel occurs in slim cracks in the boundaries of the aggregate particles
by the abundant presence of microcracks in these 461 (2007) as possible alkali suppliers.
and rare cracks crossing the aggregate particles. The examination performed with SEM-EDS confirms that the
aggregates. Microcracks
gel is composed mainly ofare
Sipointed
with highout by Velasco-
content of Ca and veryAccording
low contenttoofLindgard
K and Na.etHowever,
al. (2012), RILEM
some Al wasAAR-
Torres et al.
also detected. (2008) as facilitators of the access of 3 is considered the most adequate test to simulate
interstitial fluids to crossing
A crack the silicaanpresent in aggregates.
aggregate particle was also the
analysed in orderoftoslow-reactive
behaviour identify the material inside the
aggregates. Also,
The assumption
crack. It shows ain LNEC Etexture
particulate 461 (2007), that petro-
not common in gel. The EDS analysis indicates a composition
this concrete prism method was pointed out close to that
as not
of gel but with very high content of Ca and scarce Al (Figure 5).

A B C

Figure 5: Post-mortem petrography of concrete prisms with aggregate GR3. Gel and material from the cement
Figure 5: inside
paste Post-mortem petrography
a crack crossing of concrete
a granitic particle inprisms with aggregate
the thin-section from the GR3. Gel
concrete andtested
prism material from with
according the cement
RILEM paste
AAR-4.1. A) Under optical microscope (PPL); B) Image obtained by SEM; C) Spectrum by EDS.
inside a crack crossing a granitic particle in the thin-section from the concrete prism tested according with RILEM
AAR-4.1. A) Under optical microscope (PPL); B) Image obtained by SEM; C) Spectrum by EDS.
Discussion
The obtained results for each method are summarized in Table 5.
242
Table 5: Summary of the results achieved by the different methods applied on the assessment of the studied aggregates.
Method Interpretation criteria GR1 GR2 GR3 GR7 GR17 GR18 GR19 GR20
 same aggregate. Furthermore, the aggregates presenting the lowest angles of undulatory extinction seem to
establish a better correlation with the percentages of microcrystalline quartz, plotting close to the regression line.
The above
15th is in accordance
Euroseminar withApplied
on Microscopy West (1991, 1994),
to Building who pointed
Materials out undulatory
• 17-19 June 2015 • Delft,extinction angle as a possible
The Netherlands
indicator of potential alkali-reactivity of rocks, rather than a diagnostic feature.

LNEC E 461 (2007)

LNEC E 415 (1993)

Figure 6: Comparison between the percentages of microcrystalline quartz and the undulatory extinction angles of the quartz
Figure 6: Comparison between the percentages of microcrystalline quartz and the undulatory extinction angles of the
crystals.
quartz crystals.
The comparison between the results of the petrographic examination and the ASTM C 1260
confirmed
being thatatthe
conclusive theaccelerated
age of one mortar-bar
year and a longertest was notwith
abletheto support
work ofthe results ofetthe
Lindgard al.petrographic
(2012), which examination
in period
test what concerns
may be needed aggregatesfor some GR1, GR2
slow and GR18,
reactive classifying
supports all theAAR-4.1
that RILEM graniticis aggregates
more effective as than
non-reactive.
Though this
aggregates method expansion
to produce is widely used over for
the the assessment
critical RILEMof aggregates
AAR-3 in the for assessment
concrete purposes, it is still not well
of slow-reactive
recognized that its limits correspond
limit in the conditions of this test (Shayan, 2007; to the reactivity of some of the slow reactive
aggregates, even though this last method uses aggregates, leadinga to their
incorrect classification as false
Shayan et al., 2008; RILEM AAR-0, 2013). This non-reactive (Shayan,
lower testing temperature (38 C), simulating the GR2 and
2007), which is surely the case of aggregate
GR18. This method has not been
assumption was taken from the continuing positive successful when performed on Portuguese
real conditions to which granitic aggregates
structures and, therefore, it
are exposed.
has of
trend been
the considered
expansion curves as inappropriate
at the end to of evaluate
the test. the Such
alkali-reactivity
behaviour is, ofalso,
this type of slowinreactive
recognized rocks (Santos
the studied
Silva,AAR-0
RILEM 2005).(2013)
The explanation
suggests the for the resultsof obtained
continuation might which
aggregates, remainstart in totheexpand
findings reported
earlier in RILEM by Lu et al.
the(2006a,b),
test untilinexpansion
which it isceasesstatedor that
it the
has crushing
become of rocks AAR-4.1to obtain
ratherverythanfine aggregates
in RILEM AAR-3.can destroy the original
microstructure
clear whether or and not thethuscriteria
under-estimate the alkali reactivity of the rocks in accelerated mortar-bar tests. In fact,
will be exceeded.
From in all laboratory
the studied tests sampleshasbybeenpost-mortem
discussedpet-
In the present study, this phenomenon occurs particles
the influence of the size of the aggregate for used and results
published (e.g. Lu et al., 2006a,b; Multon et al., rography
2010). (GR1, GR2 and GR3), the one that showed
all aggregates, with the exception of aggregate GR7
whose expansion From has the levelled
comparison off. between
For aggregatethe resultsstronger
of RILEM manifestations
AAR-4.1 and of ASRthe was
onesaggregate
from theGR2, petrographic
examination also the aggregate which presented higher values
GR2 there is no itdoubt
can be thatsaid
it is that: all thereactive,
potentially criteria used for RILEM AAR-4.1 (twelve weeks, fifteen weeks and
twenty ofreactive
expansion in the concrete prism tests.of From a
but for allweeks)
the otherconfirm aggregate
aggregates GR2 asaspotentially
regarded non- in agreement with the results the petrographic
examination (Class generic point of view, it is verified that there are
reactive after one year,II);anthere
extendedis also consensus
test period isin the classification of aggregate GR7 as non-reactive (Class I);
although classified as Class II by the behaviours rare cracks aggregate
petrographic examination, inside theGR1 aggregate
is ratedparticles and the
as potentially reactive by
surely necessary for confirming their
cracks in the cement paste are scarce, very thin and
the criteria
(except GR7). inTheLNEC
obtained E 461 (2007)
results are and Lindgård et al. (2010); although classified as Class I by the petrographic
in agreement
have no infilling material. Gel is found filling some
examination, aggregates GR3, GR17, GR19 and GR20 are classified as potentially reactive by RILEM AAR-4.1
voids partially or totally. More often, gel occurs as
by all interpretation criteria.
Table 6: Minerals zoned lining of the cracks in the interfaces. This lay-
Forina better
the studied granitic aggregates
interpretation that
of the aforementioned described, it is important to keep in mind that granitic
canaggregates
contribute with alkalis to ASR according to LNEC ering shows colourless gel and brownish gel, with
are regarded as slow-reactive aggregates in terms of ASR development. Therefore, it will more
E 415 (1993)toand LNEC E 461 (2007). AAR-41 the criterion fibrous character in some locations.
adequate admit that for RILEM in Lindgård et al. (2010), which uses a test period of 20
weeks. In fact, this is one of the criteria that classifies a larger
Figure number
7 displays theofcomparison
aggregatesbetween
as potentially
the ex- reactive,
though some
Aggregate reactivity
Minerals that can values arewith
contribute comprised
alkalis in the uncertainty
pansion values band of 0.03% 0.010%aggregates
the investigated suggestedininthe Lindgård et
al. (2010) and
GR1 RILEM
orthoclase; AAR-0
microcline; (2013).
biotite; However,
muscovite; al- although
different theexpansion
criterion tests
in LNEC E 461
according (2007)
to the uses the lowest
respective
bite; oligoclase
expansion limit and test period, the interpretation ofinterpretation the results in criteria
RILEMversus AAR-4.1 the shows
contentthe of effectiveness
micro- of
GR2 microcline; biotite; muscovite; oligoclase
this
GR3 criterion in detecting all aggregates
orthoclase; microcline; biotite; muscovite; regarded as potentially
crystalline reactive
quartz by
and thethecriterion
undulatoryin Lindgård et
extinction al. (2010),
which usesoligoclase
a test period of 20 weeks. The potential reactivity angle. Thedetected in RILEMare
best correlations AAR-4.1
achievedfor aggregates GR3
between
GR7
and GR20 microcline;
is, possibly, biotite; muscovite;
justified by albite; oligo-
the abundant presence of microcracks
the percentages in these aggregates.
of microcrystalline quartz and Microcracks
the are
clase
GR17 orthoclase; microcline; biotite; muscovite;
concrete expansion tests, namely with RILEM AAR-
oligoclase 4.1, when the criterion in LNEC E 461 (2007) is used.
GR18 orthoclase; microcline; biotite; muscovite; al- As for the comparison with the undulatory extinc-
bite; oligoclase tion angle, good correlations are found between
GR19 orthoclase; microcline; biotite; muscovite; al- 9
bite; oligoclase
this parameter and the accelerated mortar-bar test,
GR20 orthoclase; microcline; biotite; muscovite; al- especially at 28 days. However, as already referred,
bite; oligoclase this test is not effective in the assessment of the
potential reactivity of granitic aggregates.

243
 Ramos et al.

Ramos et al.

found between this parameter and the accelerated mortar-bar test, especially at 28 days. However, as already
referred, this test is not effective in the assessment of the potential reactivity of granitic aggregates.

A B

RILEM AAR-0 (2013) RILEM AAR-0 (2013)

LNEC E 461 (2007) LNEC E 415 (1993)

C D

Santos Silva and Braga Reis (2000) Santos Silva and Braga Reis (2000)
LNEC E 461 (2007) LNEC E 415 (1993)

E F

LNEC E 461 (2007) LNEC E 461 (2007)

LNEC E 461 (2007) LNEC E 415 (1993)

G
H
RILEM AAR-0 (2013) RILEM AAR-0 (2013)
LNEC E 461 (2007) LNEC E 415 (1993)

Figure 7: Correlations between the expansion tests and respective interpretation criteria, the microcrystalline quartz content
Figure 7:
andCorrelations
the undulatorybetween
extinctionthe expansion
angle tests and
of quartz crystals. A) respective
ASTM C 1260 interpretation criteria, the microcrystalline
(14 days) vs Microcrystalline quartz; B) ASTM Cquartz
content1260
and(14
thesays) vs Undulatory
undulatory extinction
extinction angle;
angle C) ASTM
of quartz C 1260 (28
crystals. A) days)
ASTM vs Microcrystalline
C 1260 (14 days) quartz; D) ASTM C 1260 (28
vs Microcrystalline quartz;
days) vs Undulatory extinction angle; E) RILEM AAR-4.1 (12 weeks) vs Microcrystalline quartz; F) RILEM AAR-4.1 (12
B) ASTMweeks) vs Undulatory extinction angle; G) RILEM AAR-4.1 (15 weeks) vs Microcrystalline quartz; H) RILEM AAR-4.1 (15 D)
C 1260 (14 days) vs Undulatory extinction angle; C) ASTM C 1260 (28 days) vs Microcrystalline quartz;
ASTM C 1260 (28 days) vs Undulatory extinction weeks) vs angle; E) RILEM
Undulatory extinctionAAR-4.1
angle. (12 weeks) vs Microcrystalline quartz;
F) RILEM AAR-4.1 (12 weeks) vs Undulatory extinction angle; G) RILEM AAR-4.1 (15 weeks) vs Microcrystalline
quartz; H) RILEM AAR-4.1 (15 weeks) vs Undulatory extinction angle

11

244
 Ramos et al.
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

I J
Lindgård et al. (2010) Lindgård et al. (2010)
LNEC E 461 (2007) LNEC E 415 (1993)

K L
RILEM AAR-0 (2013) RILEM AAR-0 (2013)
LNEC E 461 (2007) LNEC E 415 (1993)
Figure 7 (cont.): Correlations between the accelerated expansion tests and respective interpretation criteria, the
microcrystalline
Figure quartz content and
7: (cont.) Correlations the undulatory
between extinctionexpansion
the accelerated angle of quartz
testscrystals. I) RILEM interpretation
and respective AAR-4.1 (20 weeks) vs the
criteria,
Microcrystalline quartz; J) RILEM AAR-4.1 (20 weeks) vs Undulatory extinction angle; K) RILEM AAR-3 vs
microcrystalline quartz content and the undulatory extinction angle of quartz crystals. I) RILEM AAR-4.1 (20 weeks)
Microcrystalline quartz; L) RILEM AAR-3 vs Undulatory extinction angle.
vs Microcrystalline quartz; J) RILEM AAR-4.1 (20 weeks) vs Undulatory extinction angle; K) RILEM AAR-3 vs
Microcrystalline
Conclusions quartz; L) RILEM AAR-3 vs Undulatory extinction angle.
From the comparison between the results of the petrographic characterization and of the different
expansion tests of the investigated granitic aggregates it is possible to conclude the following:
- theV.undulatory
Conclusion
extinction angle should only be usedreactive than indicator
as a possible the ASTM C 1260. reactivity
for potential The use of
and not as a diagnostic tool; the interpretation criteria in LNEC E 461 is
From the comparison between theofresults
- besides the quantification of the quartz, suggested
microcrystalline the presenceas ofbeing
microcracks
the moremustadequate
be considered
one for
petrographic characterization
in the evaluation of the aggregates;and of the different this kind of aggregates;
- thereofisthe
expansion tests theinvestigated
possibility thatgranitic
other minerals,
aggre- other than quartz, can take part in ASR (e.g. micas and
feldspars);
gates it is possible to conclude the following:
- the possibility of other minerals to enhance ASR due to alkalis
• the contentcontribution should also be quartz
of microcrystalline regarded;corre-
- ASTM C 1260 must never
• the undulatory extinction angle should only be used solely in the assessment of slow-reactive aggregates; the
lates better with the results of AAR-4.1 than
modification
be used as of athe testing indicator
possible period and/or of the reactivity limit of this test to adequate values must be
for potential
investigated in future research; with the other expansion tests. However, fur-
reactivity and not as a diagnostic tool;
- RILEM AAR-4.1 is the test that detects a higherther studies
number involving
of granitic granitic
aggregates as rocks with a
potentially
reactive thanthe
thequantification
ASTM C 1260.of The widerindiversity
use of the interpretation criteria LNEC E of 461microcrystalline quartz
is suggested as being thecon-
• besides microcrystalline
more adequate one for this kind of aggregates; tent must be carried out in order to investigate
quartz, the presence of microcracks must be
- the content of microcrystalline quartz correlates better whether thisresults
with the feature of has a direct
AAR-4.1 thaninfluence
with the on
considered
other expansion in the
tests.evaluation
However,offurther
the aggregates;
studies involving the granitic rocks with a ofwider diversity of
potential reactivity this type of rocks.
microcrystalline quartz content must be carried out in order to investigate whether this feature has a direct
• there is the possibility that other minerals,
influence on the potential reactivity of this type of rocks.
other than quartz, can take part in ASR (e.g.
micas and feldspars);
Acknowledgements
Acknowledgement
The authors are thankful to Fundação para a Ciência e Tecnologia (FCT) for the financial support
• the possibility of other minerals to enhance
concerning the Project IMPROVE (Ref: PTDC/ECM/115486/2009). The authors affiliated to CGUP are thankful
to ASR duethe
FCT for tofunding
alkalis ofcontribution
the Strategicshould
Projectalso be
Pest-OE/CTE/UI0039/2014.
The authors are Thethankful
author V. Ramos is gratefulpara
to Fundação to a
FCTregarded;
for the financial support of her PhD studies (Ref: SFRH/BD/47893/2008). The authors are
Ciencia e Tecnologia (FCT) for the financial also indebted to
CEMUP for SEM-EDS analysis. support concerning the Project IMPROVE (Ref:
• ASTM C 1260 must never be used solely in
the assessment of slow-reactive aggregates; the PTDC/ECM/115486/2009). The authors affiliated
modification of the testing period and/or of the to CGUP are thankful to FCT for the funding of
reactivity limit of this test to adequate values the Strategic Project Pest-OE/CTE/UI0039/2014.
must be investigated in future research; The author V. Ramos is grateful to FCT for
12
the financial support of her PhD studies (Ref:
• RILEM AAR-4.1 is the test that detects a higher SFRH/BD/47893/2008). The authors are also in-
number of granitic aggregates as potentially debted to CEMUP for SEM-EDS analysis.

245
 Ramos et al.

References Multon, S., Cyr, M., Sellier, A., Diederich, P. and Petit, L. (2010).
" Effects of aggregate size and alkali content on ASR expansion." Ce-
ment and Concrete Research. Vol. 40, 508-516. Ramos, V. (2013).
ASTM C 1260 (2007). "Standard test method for potential alkali reac-
" Characterization of the potential reactivity to alkalis of Portuguese ag-
tivity of aggregates (mortar-bar method)." The American Society for Testing
gregates for concrete". PhD thesis, Faculdade de Ciencias da Universi-
and Materials, Philadelphia, USA, 4 p.
dade do Porto and Universidade de Aveiro, 417 p.
Bérubé, M.A., Duchsene, J., Dorion, J.F., Rivest, M. (2002).
RILEM AAR-0 (2013). " Outline guide to the use of RILEM Meth-
" Laboratory assessment of alkali contribution by aggregates to concrete
ods in assessments of aggregates for AAR potential". In: Special Issue
and application to concrete structures affected by alkali-silica reactivity."
RILEM TC219-ACS, Materials & Structures. Vol. 46, in press.
Cement and Concrete Research. Vol. 32: 1215-1227.
RILEM AAR-1 (2003). " Detection of potential alkali-reactivity of ag-
Brown, L.S. (1955). " Some observations on the mechanism of alkali-
gregates - petrographic method. TC 191-ARP, Alkali-reactivity and pre-
aggregate reactions." ASTM Bulletin. Vol. 205, 40 p.
vention - assessment, specification and diagnosis of alkali-reactivity." Pre-
Concrete Society (2010). " Concrete petrography: an introductory
pared by Sims, I. and Nixon, P.. Materials & Structures. Vol. 36: 472-479.
guide for the non-specialist." Report of a Concrete Society Working Party.
RILEM AAR-1.1 (2013). " Detection of potential alkali-reactivity, Part
Technical Report 71, 101 p.
1: petrographic examination method for aggregates." Working Group
Constantiner, D., Diamond, S. (2003). " Alkali release from feldspars
Draft, August, 2013.
into pore solutions". Cement and Concrete Research. Vol. 33: 549-554.
RILEM AAR-3 (2013). " Detection of potential alkali-reactivity - 38 C
DeHills, S.M. and Corvalán, J (1964). " Undulatory extinction in
method for aggregate combinations using concrete prisms". In: Special
quartz grains of some Chilean granitic rocks of different ages." Geological
Issue RILEM TC219-ACS, Materials & Structures. Vol. 46, in press.
Society of America Bulletin. Vol. 75, 363-366.
RILEM AAR-4.1 (2013). " Detection of potential alkali-reactivity -
Dias, G., Noronha, F., Almeida, A., Simoes, P.P., Martins, H.C.B. and
60µm method for aggregate combinations using concrete prisms." In: Spe-
Ferreira, N. (2010). " Geocronologia e petrogénese do plutonismo tardi-
cial Issue RILEM TC219-ACS, Materials & Structures. Vol. 46, in press.
Varisco (NW de Portugal): síntese e interferências sobre os processos de
RILEM AAR-6.1 (2013). " Guide to diagnosis and appraisal of AAR
acreçao e reciclagem crustal na Zona Centro-Ibérica." In: Neiva, J.M.C.,
damage to concrete structures, Part 1: diagnosis." Godart, B., Rooij, M.
Ribeiro, A., Mendes Victor, L., Noronha, F. and Magalhaes Ramalho,
and Wood, J.G.M. (eds). RILEM State-of-the-Art Reports, Springer: 87 p.
M. (eds.). Ciencias Geológicas - Ensino e Investigação e sua História, Geolo-
Santos Silva, A (2005). " Degradação do betão por reacçoes álcalis-
gia Clássica, Publicaçao Comemorativa do Ano Internacional do Planeta
silica. Utilização de cinzas volantes e metacaulino para a sua prevenção."
Terra, Associação Portuguesa de Geólogos, Sociedade Geológica de Por-
PhD thesis, Laboratório Nacional de Engenharia Civil e Escola Superior
tugal. Vol. 1, Chap. II - Petrologia e Geoquímica (in Portuguese), 143-
de Engenharia da Universidade do Minho, 340 p.
160.
Santos Silva, A. and Braga Reis, M.O. (2000). " Avaliação da reac-
Dolar-Mantuani, L.M.M. (1983). " Handbook of concrete aggregates:
tividade aos álcalis dos agregados para betão." Encontro Nacional de Betão
a petrographic and technological evaluation (Building Materials Science
Estrutural. Faculdade de Engenharia da Universidade do Porto, Portugal,
Series)." Park Ridge, Noyes Publications, 345 p.
23-32.
Fernandes, I., Noronha, F. and Teles, M. (2004). " Microscopic analysis
Shayan, A. (1993). " Alkali-reactivity of deformed granitic rocks: a
of alkali-aggregate reaction products in a 50-year-old concrete." Materials
case study". Cement and Concrete Research. Vol. 23,1229-1236.
Characterization. Vol. 53: 295-306.
Shayan, A. (2007). " Field evidence for inability of ASTM C 1260 limits
Gillott, J. E. and Rogers, C. A. (1994). " Alkali - aggregate reaction and
to detect slowly reactive Australian aggregates." Australian Journal of Civil
internal release of alkalis." Magazine of Concrete Research . Vol. 46: 99-112.
Engineering. Vol. 3, 13-26.
Gogte, B.S. (1973). " An evaluation of some common Indian rocks
Shayan, A., Xu, A. and Morris, H. (2008). " Comparative study of the
with special reference to alkali-aggregate reactions." Engineering Geology.
concrete prism test (CPT 60 C, 100% RH) and other accelerated tests".
Vol. 7, 135-153.
In: Broekmans, M.A.T.M. and Wigum, B.J. (eds.). Proceedings of the 13th
Grattan-Bellew, P.E. (1992). " Microcrystalline quartz, undulatory ex-
International Conference on Alkali-Aggregate Reaction, Trondheim, Norway,
tinction & the alkali-silica reaction." In: Poole, A.B. (ed.). 9th International
391-400.
Conference on Alkali-Aggregate Reaction in Concrete, London, UK, 383-394.
Thomson, M.L., Grattan-Bellew, P.E. and White, J.C. (1994).
Hagelia, P. and Fernandes, I. (2012). " On the AAR susceptibility of
" Application of microscopic and XDR techniques to investigate alkali-
granitic and quartzitic aggregates in view of petrographic characteristics
silica reactivity potential of rocks and minerals." In: Gouda, G.R., Nis-
and accelerated testing." 14th Conference on Alkali-Aggregate Reaction in
peros, A., Bayles, J. (eds.). Proceedings of the 16th International Conference on
Concrete, Austin, USA, 10 p.
Cement Microscopy, International Cement Microscopy Association, Texas,
Kerrick, D. and Hooton, R. (1992). " ASR of concrete aggregate quar-
USA, 19 p.
ried from a fault zone: results and petrographic interpretation of accel-
Tiecher, F., Rolim, P.H., Hasparyk, N.P., Molin, D.C.C.D., Gomes,
erated mortar bar tests." Cement and Concrete Research. Vol. 22, 949-960.
M.E.B. and Glieze, P. (2012). " Reactivity study of Brazilian aggregates
Leemann, A., Holzer, L. (2005). " Alkali-aggregate reaction - identifying
through silica dissolution analysis". 14th Conference on Alkali-Aggregate
reactive silicates in complex aggregates by ESEM observation of dissolu-
Reaction in Concrete, Austin, USA, 10 p.
tion features." Cement and Concrete Composites. Vol. 27: 796-801.
Velasco-Torres, A., Alaejos, P. and Soriano, J. (2010). " Comparative
Lindgard, J., Nixon, P.J., Borchers, I., Schouenborg, B., Wigum, B.J.,
study of the alkali-silica reaction (ASR) in granitic aggregates". Estudios
Haugen, M. and Åkesson, U. (2010). " The EU" PARTNER" Project - Euro-
Geológicos. Vol. 66, 105-114.
pean standard tests to prevent alkali reactions in aggregates: final results
Vernon, R.H. (2004). " A practical guide to rock microstructure." Cam-
and recommendations". Cement and Concrete Research. Vol. 40: 611-635.
bridge University Press, 594 p.
Lindgard, J., Andiç-Çakir, O., Fernandes, I., Thomas, M.D.A. and
Wenk, H.-R., Monteiro, P.J.M. and Shomglin, K. (2008). " Relationship
Ronning, T.F. (2012). " Alkali-silica reactions (ASR): literature review on
between aggregate microstructure and mortar expansion. A case study
parameters influencing laboratory performance testing". Cement and Con-
of deformed granitic rocks from Santa Rosa mylonite zone." Journal of
crete Research. Vol. 42, 223-243.
Materials and Science. Vol. 43, 1278-1285. West, G. (1991) " A note on
LNEC E 415 (1993). " Inertes para argamassas e betões - determinação
undulatory extinction of quartz in granite." Engineering Geology. Vol. 24,
da reactividade potencial com os álcalis. Análise petrográfica." Especifi-
159-165.
caçao LNEC, Lisboa, Portugal, 6 p.
West, G. (1994). " Undulatory extinction of quartz in some British
LNEC E 461 (2007). " Betões. Metodologias para prevenir reacções
granites in relation to age and potential reactivity." Quarterly Journal of
expansivas internas". Especificaçao LNEC. Lisboa, Portugal, 6 p.
Engineering Geology. Vol. 27, 69-74.
Lu, D., Fournier, B. and Grattan-Bellew, P.E. (2006a). " Effect of aggre-
Wigum, B.J. (1995). " Examination of microstructural features of Nor-
gate particles size on determining alkali-silica reactivity by accelerated
wegian cataclastic rocks and their use for predicting alkali-reactivity in
tests." Journal of ASTM International. Vol. 3, 11 p.
concrete." Engineering Geology. Vol. 40: 195-214.
Lu, D., Fournier, B. and Grattan-Bellew, P.E. (2006b). " Evaluation of
Yan, L., Lee, C.F. and Pei-Xing, F. (2004). " Alkali-silica reaction (ASR)
accelerated test methods for determining alkali-silica reactivity of con-
characteristics of concrete made from granite aggregates." In: Tang, M.
crete aggregates." Cement and Concrete Composites. Vol. 258, 546-554.
and Deng, M. (eds.). Proceedings of the 12th Conference on Alkali-Aggregate
Lu, D., Zhou, X., Xu, Z., Lan, X., Tang, M., Fournier, B. (2006c).
Reaction in Concrete, Beijing, China, 369-376.
" Evaluation of laboratory test method for determining the potential al-
kali contribution from aggregate and the ASR safety of the Three-Gorges
dam concrete." Cement and Concrete Research. Vol. 36: 1157-1165

246
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Hydration of Portland cement

247
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

248
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Adsorbance of superplasticizers on concrete

additives: microstructural in-situ experiments
Alexander Wetzel⇤ , Johannes Arend
Department of Structural Materials and Construction Chemistry, University of Kassel, Kassel, Germany
⇤ alexander.wetzel@uni-kassel.de

Abstract

Superplasticizers are used in a wide range for concrete or mortar to enhance the workability and much more to
reduce the water/binder ratio while keeping the workability. Today, mainly polycarboxylate ethers (PCE) are used as
superplasticizers in concrete. The number and amount of additives for concrete has increased in recent years. The main
reasons for using these supplementary cementitious materials are saving CO2 -intensive Portland cement and to increase
the performance of the concrete. The effectivity of superplasticizers on certain additive or fine filler grains is therefore
of major interest. The distribution of PCE molecules within a hardened cementitious matrix can be measured after
PCE is stained by a fluorescing dye via UV-microscopy or laser scanning microscopy (Wetzel et al. 2015). Using such
FITC-stained PCE molecules in a suspension of fine binder and inert particles the absorbance was analyzed. For these
analyses a special flow-through cell was build, which enables the in-situ measurements via UV-microscopy as well as
via laser scanning microscope. Compared to existing methods measuring the adsorbance of admixtures on inorganic
particle surfaces, this novel method gives the possibility for in-situ observations.

Keywords: polycarboxylate ether, cement, supplementary cementitious materials, FITC, UV-microscopy

I. Introduction which are only some nanometer big in maximum in

To achieve a specific workability of mortars and
concretes polycarboxylate ether based superplasti-
cizers (PCE) are added as admixtures. These super-
plasticizers adsorb on particle surfaces and effect
dispersion by a combination of electrostatic and
steric repulsion of the particles the superplasticiz-
ers are adsorbed on. Beneath improvement of the
workability these superplasticizers enable a low wa-
ter/binder ratio and therefore improve the strength
and durability. Due to this, the use of superplasti-
cizers in high performance concretes is necessary.
PCE’s are characterized by an adsorbing backbone
and side chains made up by hydrophilic polyethy-
lene oxides (Plank et al. 2008). The adsorbance
of PCEs on particle surface depends on the charge
of the particle surface or more precisely on the
zeta potential of the particles. The zeta potential in
turn is influenced by early hydration in case of ce-
mentitious systems (Plank, Hirsch 2007). A proofed
method analyzing the affinity of absorbance of poly-
mers on mineralic particle surfaces is TOC (total
organic carbon). This method, which often is cou-
pled to zeta potential measurements, is an ex-situ
measurement, whereas the method used in these
investigations displays a novel in-situ method.
The localization and much more the quantifica-
tion of organic admixtures within building mate-
rials is difficult due to the low size of admixtures,
case of PCEs. Illustrations of PCE-comb-polyers ad-
hering on inorganic particles are drawn oversized
many times (Figure 1).
II. Methods
Materials
Polycarboxylate ethers are usually used in combi-
nation with a low amount of defoamers, because
beneath plasticizing these PCEs tend to stabilize
air voids. In these investigations superplasticizers
without defoamers were used and measurement
artefacts caused by defoamers can be neglected.
The inorganic particles used in this investiga-
tions are (a) Portland cement clinker, (b) ground
granulated blast furnace slag (GGBFS), (c) quartz
powder (d) lime stone powder and (e) corundum.
The corundum was chosen as a particle type with
exemplarily low zeta potential in contrast to the
other particles with rather high zeta potential.
To locate the superplasticizer via UV-microscopy
it is necessary to bind a fluorescent dye, in this case
fluorescein iso-thiocyanate (FITC), chemically to
the PCE-molecule. FITC binds effectively to amino-
groups, which are usually not represented in PCEs.
It is kind to introduce these functional groups by
a linker molecule e.g. ethylene diamine. This way,
FITC can be bond to the superplasticizer easily.

249
 which often is coupled to zeta potential measurements, is an ex-situ measurement, whereas the method used in
these investigations displays a novel in-situ method.
The localization and much more the quantification of organic admixtures within building materials is
Wetzel & Arend
difficult due to the low size of admixtures, which are only some nanometer big in maximum in case of PCEs.
Illustrations of PCE-comb-polyers adhering on inorganic particles are drawn oversized many times (Fig. 1).

A. Wetzel and J. Arend

Methods

Materials
Figure 1: Figure
Mostly1: the
Mostadsorbance of superplasticizers
time the adsorbance are displayed
of superplasticizers on sketches,
are displayed where
on sketches, the the
where polymers
polymersarearedrawn
drawnmuch much too
too big. big.
Polycarboxylate ethers are usually used in combination with a low amount of defoamers, because
beneath plasticizing these PCEs tend to stabilize air voids. In these investigations superplasticizers without
defoamers were used and measurement artefacts caused by defoamers can be neglected.
Therefore, The the solid PCE-polymer
inorganic wasinseparated
particles used Therefore,
this investigations are (a) a suspension
Portland cement of inorganic
clinker, (b) particles
groundin
out of the stock solution (MC-Bauchemie) by evap- water or isopropanol,
granulated blast furnace slag (GGBFS), (c) quartz powder (d) lime stone powder and (e) corundum. The depending on the reactivity
1
oration.
corundum After
was all water
chosen as awas removed
particle by vacuum-
type with exemplarily low with water,
zeta wasindisintegrated
potential contrast to theby ultrasound.
other particles with Sub-
rather high
drying, the zeta
PCEpotential.
was dissolved in toluene under sequently, a drop of that suspension was placed on
stirring and Tomoderate
locate the superplasticizer
heating (60 C).viaAfter UV-microscopy
that, a it is necessary
glass substratetoandbinddried
a fluorescent dye, in
in a cabinet this case
dryer. After
fluorescein iso-thiocyanate (FITC), chemically to the PCE-molecule. FITC binds effectively to amino-groups,
ethylene diamine was added as well as the cata- drying the particles are fixed to the substrate. A
which are usually not represented in PCEs. It is kind to introduce these functional groups by a linker molecule
lyst
e.g.p-toluene sulfonic This
ethylene diamine. acid.way,
In the lastcan
FITC step,
be small
bond to thesuspension of stained
superplasticizer easily.PCE (1%), with
Therefore, a pH-value
the solid PCE-
amounts
polymer wasof FITC are added
separated to stock
out of the the solution
solution with of 12 gained
(MC-Bauchemie) by correction
by evaporation. After with Ca(OH)
all water 2 , is then
was removed
by vacuum-drying,
additional reaction the PCEThe
time. was stained
dissolvedsuperplas-
in toluene under stirring
added on and
the moderate
particles heating
and the(60 °C). After
surface that,
is covered
ethylene
ticizer was diamine
obtainedwas after
addedthe
as removal
well as the ofcatalyst
toluenep-toluene
by asulfonic acid.immediately.
cover glass In the last step, small amounts of
FITC are added to the solution with additional reaction time. The stained superplasticizer was obtained after the
and dialytic purification. To be able to control the
removal of toluene and dialytic purification. To be able to control Due to the theconcentration
pH-sensitivity of the FITC-PCE
of FITC in the
the intensity
concentration
experiment, theofdialysed
the FITC-PCE in the
solution was experiment,
vaporized to obtain a resoluble solid Wetzel et al. (2015).
the dialysed solution was vaporized to obtain a of fluorescence is reduced with a pH-value of 12,
Experimental setup
resoluble solid (Wetzel et al., 2015). but due to the fixed pH-value of every experiment
the differences of particle fluorescence intensity is
In a first approach the adsorbance of stained PCE on inorganic
caused by differentparticles fixed on
adsorbance of athe
glass substrate
stained PCEs
Experimental setup
were measured via UV-microscopy in reflection mode. Therefore, on the aparticles
suspension of inorganic
surfaces. Theparticles
resulting Ca2+ -ion
in water or
Inisopropanol, depending
a first approach on the reactivity
the absorbance withPCE
of stained water,
on was concentration
disintegrated byafterultra-sonic.
correctingSubsequently,
the pH-value a drop
to 12ofby
that suspension
inorganic wasfixed
particles placedonona aglass
glasssubstrate
substrate and
weredried in a cabinetisdryer.
Ca(OH) aboutAfter drying the
0.01 mol/l. Theparticles are fixed
UV-measurement
2
to the substrate. A suspension of stained
measured via UV-microscopy in reflection mode. PCE (1%), with a pH-value of 12 gained by correction with
is conducted about 1 minute after contact of 2PCE- Ca(OH) , is
then added on the particles and the surface is covered by a cover glass immediately.

Figure 2: Experimental setup in UV-Microscope

Figure 2: Experimental setup in UV-Microscope.

Due to the pH-sensitivity of FITC the intensity of fluorescence is reduced with a pH-value of 12, but
due
250 to the fixed pH-value of every experiment the differences of particle fluorescence intensity is caused by
different adsorbance of the stained PCEs on the particles surfaces. The resulting Ca2+-ion concentration after
correcting the pH-value to 12 by Ca(OH)2 is about 0.01 mol/l. The UV-measurement is conducted about 1
 quartz powder an increase could be observed (Table 1). The positive charged Ca -ions adsorbe on the negative
charged surface of the quartz particles and change the zeta potential into positive direction. The more positive
zeta15potential of the on
th Euroseminar surface than enhances
Microscopy Appliedthe adsorbance
to Building of PCE• on
Materials theJune
17-19 surface.
2015 It• Delft,
could The
be observed that an
Netherlands
increase of Ca2+-ions increased the fluorescence intensity and therefore the adsorbance of PCE with a maximum
at the concentration of 0.02 mol/l.

Figure 3: Distribution of stained PCE´s on (a) cement grain, (b) GGBFS-grains, (c) Carbonate-grains and (d)
quartz-powder-grains.
Figure 3: Distribution Microscopic
of stained pictures
PCE’s on (transmitted light)grain,
(a) cement are overlain by the adsorbance(c)
(b) GGBFS-grains, of Carbonate-grains
PCE (increasing intensity
and (d)
quartz-powder-grains. Microscopic pictures from blue-green-yellow-red)
(transmitted light) are overlain by the absorbance of PCE (increasing
intensity from blue-green-yellow-red).
Table 1: Intensity of fluorescence normed to maximum of grey scale (100%)
Substrate Intensity per pixel Concentration of Ca2+
[%] [mol/l]
suspension with the particles.
Ground For Blast
Granulated the excitation the1.72 III. Results 0.01and discussion
filter 470/40 and for the emission
Furnace Slag the filter 525/50
was used. The highest magnification
Portland Cement Clinker (161⇥) on1.83 The intensity differences 0.01 at constant pH-value of 12
LimeMStone
that microscope (Leica 205 Powder
FA) was chosen. The1.76 show the highest intensities 0.01 for GGBFS and Port-
Quartz Powder
exposure time were set on 20 seconds for every 1.27 0.01
land cement clinker, followed by lime stone and the
Portland cement clinker 1.82 0.11
measurement. The pictures were
Quartz Powder analysed using 1.74 lowest intensity shows the quartz powder (Figure 3
0.11
ImageJ in order to gain an average pixel/intensity and Table 1). Due to zeta potential measurements
value for the grains. Therefore the background was of these powders this results were expected, be-
subtracted first by rolling ball method. Then the cause quartz shows a negative zeta potential, thus
picture was converted to 8bit-image, followed by adsorbing of the negative charged backbones of the
setting a threshold to identify the particle outline. PCEs is reduced on the particle surfaces, which
Then the particles were analysed. The result is the show a more positive zeta potential. While en-
average pixel intensity of the particle area. The in- hancing of the Ca2+ -ion-concentration by adding
tensity given in grey values were then normalized Calcium formate with a concentration of 0.1 mol/l
to percent with the maximum of 255 as the grey shows no difference in case of the Portland cement,
scale equals 100%. In a next step, single particles on quartz powder an increase could be observed
were analysed by SEM first and on the same par- (Table 1). The positive charged Ca2+ -ions adsorbe
ticles the FITC-stained PCE were added to locate on the negative charged surface of the quartz par-
A. Wetzel and J. Arend
the fluorescence via UV-microscopy and therefore ticles and change the zeta potential into positive
to locate the adsorbance of PCE-molecules. direction. The more positive zeta potential of the

Figure
Figure 4: Cement
4: Cement particle
particle inin(a)(a)secondary
secondary electron
electron mode
mode(SEM),
(SEM),(b)(b)
Distribution of aluminium
Distribution and silicium
of aluminium gained by
and silicium gained
element mapping (EDX) and (c) Stained PCE overlay on Secondary Electron
by element mapping (EDX) and (c) Stained PCE overlay on Secondary Electron picture of SEM. picture of SEM

In order to find a correlation between clinker phases and adsorbance of PCE a combined SEM-UV-
microscope test series was conducted. Particles fixed on a substrate were analysed via SEM gaining secondary 251
electron pictures and element maps (EDX). The secondary electron picture gives information about the
topography of the cement particle and the element map gives information about the clinker phase distribution
especially the differentiation between alite/ belite and aluminate/ferrite. Subsequently, the stained PCE-solution
 concentration on the particles surfaces increases continuously reaching a maximum after about 75 seconds in
case of the aluminium oxide. The intensity of the solution increases as well, but then decreases after about 75
seconds. This might be interpreted by a diffusion controlled
Wetzel enrichment on particle surface accompanied by a
& Arend
decrease within the solution. Additional tests will be done to check these effects i) on particle with different
mineralogical composition and ii) with different starting concentration of the PCE-solution.

2+2+ -concentration
FigureFigure 5: Time
5: Time seriesseries corundum
corundum with
with stainedPCE-solution
stained PCE-solution (1%,
(1%,pH=12,
pH=12,CaCa-concentration of 0.02
ofmol/l)
0.02 mol/l).

surface than enhances the adsorbance of PCE on
the surface. It could be observed that an increase of
Ca2+ -ions increased the fluorescence intensity and
therefore the adsorbance of PCE with a maximum
at the concentration of 0.02 mol/l.
In order to find a correlation between clinker
phases and absorbance of PCE a combined SEM-
UV-microscope test series was conducted. Particles
fixed on a substrate were analyzed via SEM gain-
ing secondary electron pictures and element maps
(EDX). The secondary electron picture gives infor-
mation about the topography of the cement particle
and the element map gives information about the
clinker phase distribution especially the differenti-
ation between alite/ belite and aluminate/ferrite.
Subsequently, the stained PCE-solution is added on
the particle and the distribution is estimated via
UV-microscopy.
It was assumed to find a correlation of
enrichment of PCE and the distribution of
aluminate/ferrite-phases due to early formed et-
tringite which promotes the adsorbance of PCE.
Such a correlation could not be found (Figure 4b),
for which further improvement of the experimental
setup might be necessary. However, an enrichment
on kinks and edges were not found as well.
The concentration of superplasticizers in solution
must be chosen low in order to keep an intermedi-
ate pH-value. In case of higher concentrations the
pH-value had to be corrected by a high concentra-
tion of hydroxides. This has been done by KOH
and Ca(OH)2 . In both cases the measured adsor-
bance decreased. An explanation for that might
be the complexation and counter-ion condensation
of PCE with the cations of the alkali hydroxides
(Plank, 2009).
In order to proof the concept of measurement,
an additional test series was conducted measuring
corundum powder with a grain size in the range of
the particles measured before. Corundum showed
a strong adsorbance and therefore a much shorter
exposure time was necessary. Therefore, it was
possible to gain a time series in which the develop-
ment of adsorbance within the first minutes could
be analyzed in-situ. The concentration on the par-
ticles surfaces increases continuously reaching a
maximum after about 75 seconds in case of the
aluminium oxide. The intensity of the solution in-

Table 1: Intensity of fluorescence normed to maximum of grey scale (100%).

Intensity per pixel Concentration of Ca2+

Substrate
[%] [mol/l]
Ground Granulated Blast Furnace Slag 1.72 0.01
Portland Cement Clinker 1.83 0.01
Lime Stone Powder 1.76 0.01
Quartz Powder 1.27 0.01
Portland cement clinker 1.82 0.11
Quartz Powder 1.74 0.11

252
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
creases as well, but then decreases after about 75
seconds. This might be interpreted by a diffusion
controlled enrichment on particle surface accompa-
nied by a decrease within the solution. Additional
tests will be done to check these effects i) on par-
ticle with different mineralogical composition and
ii) with different starting concentration of the PCE-
solution.
IV. Conclusion
A new approach measuring the adsorbance of su-
perplasticizers on the surfaces of inorganic particles
were developed using FITC-stained polycarboxylate
ether in a flow cell. The distribution of the poly-
mers were measured in-situ via UV-microscopy. An
enhanced adsorbance can be observed for Portland
cement and GGBFS, followed by limestone. Quartz
powder shows the lowest adsorbance of PCE which
might be explained by the mainly negative zeta po-
tential. The adsorbance of PCE could be increased
in case of quartz particles by increasing the Ca2+ -
ion concentration. A correlation between clinker
phase distribution and PCE-distribution by a com-
bined SEM-UV-microscope approach could not be
proofed yet. Therefore, the experimental setup need
to be reconsidered and maybe improved.
Generally, it could be shown that the adsorbance
of PCE on mineralic surface can be measured as a
function of electrolyte concentration in-situ. Using
corundum as test reference a short exposure-time
was necessary and therefor beneath the effect of
cation-concentration a temporal evolution of adsor-
bance could be measured.
Acknowledgement
We thank MC-Bauchemie for supporting us by pro-
viding the superplasticizers used in these investiga-
tions. Christin Koch is gratefully acknowledged for
acquisition of data during her bachelor thesis.
References
Plank, J.; Pöllmann, K.; Zouaoui, N.; Andres, P. R.; Schaefer, C. (2008):
Synthesis and performance of methacrylic ester based polycarboxylate
superplasticizers possessing hydroxy terminated poly(ethylene glycol)
side chains. In Cement and Concrete Research 38 (10), pp. 1210–1216.
Plank, Johann; Hirsch, Christian (2007): Impact of zeta potential of
early cement hydration phases on superplasticizer adsorption. In Ce-
ment and Concrete Research 37 (4), pp. 537–542.
Plank, J.; Sachsenhauser, B. (2009): Experimental determination
of the effective anionic charge density of polycarboxylate superplas-
ticizers in cement pore solution. In CCR 39 (1), pp. 1–5. DOI:
10.1016/j.cemconres.2008.09.001.
Wetzel, A.; Glotzbach, Ch.; Maryamh, K.; Middendorf, B. (2015): Mi-
crostructural investigations on the skinning of ultra-high performance
concrete. In Cement and Concrete Composites 57, pp. 27–33.
253
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

254
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

The influence of sodium salts on C-S-H

Berta Mota⇤,a , Thomas Matscheib , Karen Scrivenera
a Laboratory of Construction Materials, Ecole Polytechnique Fédérale de Lausanne
b Innovation, Holcim Technology Ltd., thomas.matschei@holcim.ch
⇤ berta.motagasso@epfl.ch

Abstract

This work aims to clarify the impact of alkali and sulfate on the morphology of C-S-H, which is the main hydration
product of Portland cement hydration, responsible for the development of strength. The morphology, as well as the
chemical composition, of this phase is dependent on the surrounding solution composition. The presence of alkali salts
appears to impact the morphology of C-S-H at early ages which at the same time seems to influence its ability to fill the
pore space at later ages. An ordinary white Portland based cement system is compared to systems with the additions of
NaOH and Na2 SO4 , two Na-salts which accelerate the hydration of Portland cement at early ages. In the plain white
cement, C-S-H adopts a needle type divergent morphology which grows outwards from the surface of cement grains.
The presence of NaOH seems to lead to a C-S-H morphology which is more parallel to the surface, without growing
outwards. On the contrary, Na2 SO4 leads to a C-S-H morphology comparable to the one in the plain system, thus
sulfate ions seem to correct the morphology of C-S-H or avoid the presence of a more planar C-S-H morphology adopted
in the presence of NaOH.

Keywords: C-S-H, hydration, microstructure, alkalis, sulfate

I. Introduction Methods

It is generally accepted that alkalis accelerate the hy-
dration of Portland cement (at least initially), which
usually leads to a higher strength development at
early ages. However, alkalis may have a detrimental
impact on strength at later ages (Jawed and Skalny,
1978). At present there is no clear explanation of
the impact of alkalis to Portland cement. However,
this has not been systematically confirmed and find-
ings in the literature are often contradictory (Kumar
et al., 2012; Škvára, 2007; Way and Shayan, 1989).
In this study, we combine the information from
Secondary Electron (SE) images and Backscattered
Electron (BSE) images to analyse respectively the
morphology of C-S-H at early ages (during the ac-
celeration period) and its ability to fill the pore
space at later ages.
II. Experimental
Materials
The chemical composition (analysed by X-ray Flu-
orescence) of the white cement used for the exper-
iments is: 24.40 wt.% SiO2 , 2.12 wt.% Al2 O3 , 0.32
wt.% Fe2 O3 , 68.13 wt.% CaO, 0.56 wt.% MgO and
1.94 wt.% SO3 . Two alkali solutions were prepared:
1.45M NaOH and 0.725M Na2 SO4 were dissolved in
water prior to being mixed with cement to ensure a
homogeneous distribution.
40g of cement were mixed with de-ionized water
(DI-water) or the appropriate aqueous solution of
NaOH or Na2 SO4 to achieve a water to cement
(w/c) ratio of 0.4. Care was taken to minimize
exposure to the air to limit carbonation during the
mixing and casting process, and the pastes were
stored in sealed conditions at 20 C.
The heat release was recorded by isothermal
calorimetery. The ambient temperature was con-
trolled at 20 C. 10 grams of paste were placed in
glass vessel and then into the calorimeter.
The surface of the grains was observed at dif-
ferent times during hydration, with a FEI XLF-30
SFEG-SEM in SE mode. The overall development
of microstructure was observed on polished sec-
tions with a FEI quanta 200 SEM microscope in BSE
mode. The operating conditions are shown on the
micrographs or indicated in the captions.
For all the microscopy samples the hydration
was stopped by solvent exchange technique with
isopropanol. For the SEM-SFEG specimens (frac-
tured surfaces), 0.5g of the paste was taken out
from the sealed sample at the required times and
immediately stirred (retention filter f > 5 µm) with
isopropanol in a funnel for 3 minutes to remove the
water. A small amount of the dried powder was
dispersed on an adhesive carbon tab. To ensure
electrical conductivity in the SEM, a conductive
carbon layer of 10-15 nm was deposited on top

255
 Mota et al.

for 2 days. It was followed by impregnation of a

piece of the slice (a square of ⇠1.5cm side and ⇠
3mm height) in a low-viscosity epoxy resin, polish-
ing with a diamond powder down to 1µm. They
were further stored in dessicator for 2 more days
and they were coated with a Ballec CED 030 car-
bon coater. EDX measurements were done with a
Brucker AXS XFlash dectector for Energy dispersive
X-ray analyses.

Figure 1: Heat release of each sample over the first 24
hours. Full dots in the nucleation and growth period
indicate the time at which hydration was stopped to
study the mid-way of this period.
with a Cressington carbon coater. To ensure a good
representativity, a minimum of 10 grains were ana-
lyzed. For the BSE specimens (polished samples),
a slice (f t 2.5cm, s 3mm height) was cut at 7
and 28 days from the rest of the paste sample and
the hydration was stopped by immersion in iso-
propanol for 6 days and further stored in dessicator
III. Results
Figure 1 shows the evolution of the heat release
from isothermal calorimetry as a function of the
hydration time (the first 24h). It is observed that
both Na-salts lead to a steeper slope during the
acceleration period, an earlier occurrence of the
silicate peak and a shorter induction period. Figure
1 shows the times at which hydration of cement
was stopped to analyse the morphology of C-S-H
in SE mode. For each system, the samples were
studied at the middle of the nucleation and growth
period with a similar degree of hydration in order
to follow the morphology of C-S-H during its early
growth.
Not only is the kinetics changing with the ad-
dition of alkali salts but also the morphology of
C-S-H. SEM morphology observations with SE in
Figure 2 show the following:
• In the reference cement system (Figure 2a)

Figure 2: SE micrographs show the surfaces of cement grains covered with C-S-H at the middle of the acceleration
period (full dots in the calorimetry curve): a) plain white cement – 6h30, b) with NaOH – 2h30 and c) with Na2 SO4 –
4h30. SEM operating conditions: 1 kV, working distance 2 mm, spot size 3, magnification 35,000⇥.

256
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 3: BSE micrographs at later ages (28 days): a) plain white cement, b) with NaOH and c) with Na2 SO4 . SEM
operating conditions: 15 kV, working distance 12.5 mm, spot size 4, current 0.12 nA, magnification 10,000⇥. The
images were selected to be representative of the whole matrix.
the C-S-H has a tendency to grow outwards
from the surface with a needle-like morphol-
ogy. Those needles can grow in a divergent
morphology or sometimes they appear to grow
in parallel. Somecrystals of CH can also be
observed in Figure 2a.
• The addition of NaOH (Figure 2b) leads to a
more planar, foil-like C-S-H. It is not growing
outwards generally but closer to the anhydrous
cement grain surface.
• The addition of Na2 SO4 (Figure 2c) does not
show the same behaviour as NaOH. C-S-H
leads to a more divergent morphology (similar
to the plain system). The same divergent effect
was checked to take place when adding of gyp-
sum to the system with NaOH. This suggests
a competition between the effect of sodium
(probably earlier in time) and sulfate ions, thus
their impact on the kinetics for C-S-H growth
when combined needs to be further studied.
2D microstructure at later ages studied by SEM
using BSE images and polished sections of later
ages in Figure 3 show that the Na2 SO4 (Figure 3c)
leads to a similar distribution of hydrates than the
plain one (Figure 3a): C-S-H is the main phase that
fills the space (inner and outer product) including
some elongated crystals of CH. This confirms that
a divergent morphology of C-S-H during the accel-
eration period will allow it to grow outwards from
the surface to better fill the porosity. On the con-
trary, in the case of NaOH (Figure 3b), BSE images
reveal limited growth of outer C-S-H, being difficult
to identify it. This can be linked to a more planar,
foil-like C-S-H observed during the nucleation and
growth period which is not able to grow outwards
and, consequently, is not filling the space as it does
in the other two systems. The small plates of CH
observed in Figure 3b appear to be the main phase
filling the space.
Figure 4 shows the EDX-SEM analysis (atomic
ratios) of the samples at 7 days of hydration. The
divergent morphology of C-S-H observed in the
presence of sulfates could result from a higher sul-
fate sorption on C-S-H product. Care was taken
to avoid EDX analyses of C-S-H near other phases
like CH crystals or anhydrous grains. The EDX
analysis were done at a magnification of 5,000⇥
where it is possible to identify which are the areas
of interest to be analyzed. C-S-H can include sulfate
which co-adsorbs with calcium ions, therefore a cor-
rected value for calcium is used according to (Ca –
S) which assumes that every sulfate ion adsorbed
on C-S-H has done so with one calcium ion.
IV. Conclusion
In this work, we observe that the addition of alkali
salts accelerates the hydration of cement during the
first hours and at the same time influence the 3D
morphology of C-S-H. In the presence of NaOH it is
not possible to identify a clear needle structure in C-
S-H but more planar as it is not growing outwards
257
 Mota et al.

Figure 4: Al/(Ca-S) ratio versus S/(Ca-S) ratio resulting from the EDX analysis of the C-S-H product at 7 days of
hydration. Operating conditions: spot size 5.5, 15kV, working distance 12.5 mm, current 0.8 nA.

from the surface. On the contrary, Na2 SO4 results
in a similar 3D morphology of C-S-H compared to
the plain system and the addition of sulfate seems
to be a clear determining factor for a divergent
needled-structure as suggested by EDX analysis .
The morphology of C-S-H during the acceleration
period seems to impact its ability to fill the empty
space at later ages. In the plain cement system,
SEM-BSE images allow the observation of a matrix
of hydrates where C-S-H product is easy to identify
as it mostly fills all the space. On the contrary, the
addition of NaOH clearly leads to a different matrix
distribution in which the space does not appear to
be filled by C-S-H product.
This study demonstrates the influence of the
chemical composition of the pore solution on the
3D morphology that
C-S-H adopts at early ages of hydration. It is also
possible to see that changes induced on the C-S-H
at very early ages may have a remarkable impact
on the development of the final matrix of hydrates.
At the same time, it may have an impact on the
compressive strength, as well as e.g. porosity. This
needs to be further studied.
References
Jawed, I., and Skalny, J. (1978): “Alkalies in Cement: A Review. II.
Effects of Alkalies on Hydration and Performance of Portland Cement.”
Cement and Concrete Research. Vol. 8, No. 1, 37–52.
Kumar, A., Sant, G., Patapy, C., Gianocca, C., and Scrivener, K. (2012).
“The Influence of Sodium and Potassium Hydroxide on Alite Hydration:
Experiments and Simulations.” Cement and Concrete Research. Vol. 42, No.
11, 1513–23.
Škvára, F. (2007). “Alkali Activated Materials or Geopolymers?”. Ce-
ramics - Silikáty. Vol. 51, No. 3, 173–77.
Way, S.J., and Shayan, A. (1989). “Early Hydration of a Portland Ce-
ment in Water and Sodium Hydroxide Solutions: Composition of Solu-
tions and Nature of Solid Phases.” Cement and Concrete Research. Vol. 19,
No. 5, 759–69.

258
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Dealing with uncertainty in material

characterization of concrete by education
Mario R. de Rooij⇤ and Fernando F. de Mendonça Filho
Building Materials, TNO, Delft, The Netherlands
⇤ mario.derooij@tno.nl

Abstract

In this article a trip is taken to characterize concrete through its hardened cement paste based on its microstructure.
Optical microscopy is coupled to advanced image analysis, as well as statistical data analysis to characterize heterogeneous
concrete material. The article takes a closer look at the determination of the w/c ratio through the analysis of capillary
porosity by using image analysis. It is shown that accurate analysis does not shown identical results in the w/c ratio
even for standard reference samples with known w/c ratio. In the article it is shown that this is not because the
techniques we use to determine the w/c ratios are not accurate enough, but much more because of the local variation in
the microstructure. The attention shifts with this article from the analysis technique to what we actually observe in
the microstructure. Hopefully this will change our thinking about the distribution in local microstructure variation
and will help to set off more research towards the capturing and modelling of these local microstructure variation into
consequences at a more structural level.

Keywords: water-cement ratio, microscopy, thin section, image analysis, microstructure distribution

I. Introduction be a way forward in determining material perfor-

Some things never seem to change and yet they do.
It seems like we are making concrete already more
than 100 years in much the same way. We require a
certain amount of rocks (both fine and coarse), ce-
mentitious material and water to make our trusted,
well known concrete material. To be sure we can
rely on the quality of our concrete materials the
standards regularly prescribe e.g. minimum ce-
ment content, maximum water-cement ratios, and
more if deemed necessary, see (European Standard-
ization Committee, 2009).
However, times are changing. Slowly but surely
our way of making concrete changes from a pre-
scriptive way of defining and regulating our con-
crete to a more performance based approach (Eu-
ropean Economic Community, 2011). Driven by
the increased attention for sustainability, allowing
a wider range of material components to be part of
our concrete, thus fulfilling the wish to reuse and
recycle, making concrete by prescribing its compo-
nents and the recipe is not enough anymore. A
need is emerging to determine and satisfy perfor-
mances of the concrete material which goes beyond
the 28 day strength alone.
In this article a trip is taken to characterize mate-
rial based on its microstructure. Optical microscopy
is coupled to advanced image analysis, as well as
statistical data analysis to characterize heteroge-
neous concrete material. It is thought that this can
mance in an ever growing world of different ways
and components to make concrete.
Before the material characterization is described,
a start is made at the structural level; what does
the structural engineer need and how does he deal
with uncertainties in available data? With this illus-
tration in mind a change in scale and representative
volume is made to go down to the microstructure
level and its heterogeneity. How accurate can we
determine this through the proposed methods, and
how can we deal with the uncertainties found in
this process?
II. The structural design approach
Just to serve as an example reference is made to
the North Boulevard Bridge project in the city of
Baton Rouge, Louisiana (McLellan, 2009). Here
Louisiana’s first high performance concrete bridge
was built, open to traffic in 2006. The bridge aes-
thetics were provided by the smooth surfaces of the
graceful and slender precast, prestressed concrete
U-beams, and the uniquely sculptured concrete
arch-shaped piers, see Figure 1.
To design the box girder, the structural analysis
and calculation required a minimum strength of
69 MPa (10,000 psi) at 56 days as one of the per-
formance requirements for the concrete. The term
minimum already indicates there could be a varia-

259
 de Rooij & Mendonça Filho

Figure 1: The North Boulevard Bridge in Louisiana made from high strength concrete

tion in the strength of the produced beams. In other
words, standing below the bridge, looking up at the
box girders, there is certainly one that is the weak-
est, but which one would this be? If a coring sample
would be necessary from the box girders to check
the strength, which beam should be taken and how
many samples would be needed to indicate if the
performance has been met? Keep this example in
mind when we come back to a similar choice in
sampling for microstructure characterization.
With the performance requirement set, the next
step is to design a concrete mix that can fulfil the
requirement. In concrete design it was the research
work of Abrams (Abrams, 1918) that showed a
simple relation between the concrete strength and
its water-cement ratio. Abrams’ law as it became
known, effectively states that for workable concrete,
the lower the water-cement ratio, the higher the
strength of the concrete. Abrams reported his re-
sults through Figure 2, using a water-cement ratio
by volume. As nowadays the water-cement ratio is
defined by mass, the graph has been replotted in
Figure 3, considering that the cement density used
by Abrams was 94 lb. per cubic ft. (Abrams 1918).
The mathematical expression Abrams came up
with to aid the material design was a marvelous
simplification of all the various components and
influences on the measured compressive strength.
However, often it is forgotten that it was just an
aid in the design process. Taking a closer look at
Figure 3, it shows two things. First of all, the w/c
ratio should be at the very low end of the scale,
probably towards 0.20. Second, even the data of
Abrams did not provide a clear strength value for a
specific w/c ratio. It is more of a range.
In concrete technology, the possible variation in
compressive strength is so well known, that it is
sometimes forgotten or overlooked when consider-
ing materials properties of concrete not related to
strength. Hence, just as a reminder a quick recapit-
ulation. Take e.g. a ready-mix truck of concrete. It
is one batch, all mixed in the same procedure. In
practice you do not get much more homogeneous
concrete than this. Now use this batch of concrete
completely to produce standard cubes or cylinders,
which are stored under standard conditions and all
tested after 28 days for compressive strength.
The results from a testing point of view could

Figure 2: Original plot from Abrams (Abrams, 1918), using water-cement ratio by volume

260
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Replotted Abrams graph, using water-cement ratio by mass

be considered strange, because the compressive
strength values are not all the same (see Figure 4)
even though all the samples are produced from
the same batch and hence are from a composition
(recipe) point of view all identical. Nevertheless
we have come to accept this variation as part of the
inherent heterogeneity of concrete. In fact, being
able to capture the distribution of the results in
a Gauss curve, we have learnt to deal with the
material variation at a structural design level. It
is exactly the reason why the needed structural
strength as a performance requirement is specified
as a minimum strength or sometimes characteristic
strength.
For the North Boulevard Bridge project the high
performance mix proportions used came to be with
a w/c ratio of 0.25. A little over 10 liter of high-
range water reducer was necessary per m3 to main-
tain a workability of the concrete within the spec-
ified range. The compressive strength from the
concrete produced from this high performance mix
was followed and measured based on test cylin-
ders for 23 castings. After 1 day the compressive
strength was on average already 47 MPa with a
range of 28 - 60 MPa. After 28 days the numbers
had increased to an average value of 94 MPa with
a range of 68 - 109 MPa. Hence, the minimum
strength requirement of 69 MPa after 56 days was
not a problem.
It is interesting to note that the variation in the
results reduces in the relative sense when going
from 1 day test results (spread around 30%) to 28
day test results (spread around 20%). Nevertheless,
going back to the original design where the struc-
tural engineer had designed the entire structure
with a compressive strength of 69 MPa, it is fair to
state that the vast majority of the bridge material
has a strength well above the design strength.
III. Connection to the material
microstructure
From the previous section it is clear that the con-
crete used is not a homogeneous material, but has
a distribution, at least for the property of strength.

Figure 4: Example of strength distribution of one batch of concrete (histogram)

261
 de Rooij & Mendonça Filho
Figure 5: Schematic of optical microscope in transmitted light mode to examine a thin section specimen. A specimen
with more porous microstructure shows brighter than a denser microstructure due to the difference in the amount of
epoxy present in the microstructure.
This is not a problem for structural engineers be-
cause they define their performance requirements
in terms of a minimum or characteristic strength.
One could ask if this heterogeneity is also present
for other material properties of concrete. If so, it
could perhaps help us to understand why certain
parts of the structure degrade faster than other
parts. It could help us in the understanding why
the ingress of substances like e.g. chloride does not
have a uniform frontline like in our models, but
has instead a much more variable front. In order
to see if it is possible to detect such microstructural
difference a closer look is taken at the concrete
microstructure.
Staying with the parameter strength, it was the re-
search work of Abrams that linked the compressive
strength of workable concrete to the water-cement
ratio. Even though for normal concrete mix design,
the concrete with w/c = 0.25 is outside the workable
concrete range specified by Abrams, addition of a
high-range water-reducing admixture created the
right direction for designing a high performance
concrete. Hence, even in this particular situation,
the w/c ratio introduced by Abrams was not that
far off. It helped in the design direction.
Following this importance of the w/c ratio,
would it then be possible to also use the w/c ra-
tio to characterize the material microstructure of
concrete? In earlier times this first connection was
made by Idorn (Idorn, 1967) and later by Thaulow
and co-workers (Thaulow et al., 1982). The reason-
ing behind this was relatively simple. If a material
has a certain property like strength, the value of that
property should come from the material itself; in
other words, its microstructure. A material with a
262
weak microstructure would result in a material with
a low compressive strength. Also, if the material
microstructure would show strong characteristics,
then the material as a whole would probably show
strong.
The next step in the reasoning may have caused
some misconception in the past. Historically, the
reasoning is as follows. If the microstructure is
responsible for the overall performance of the ma-
terial, and if the microstructure is determined by
its components then knowing the amount of com-
ponents, like knowing the amount of water and
the amount of cement (the water-cement ratio) cre-
ates a direct relation to properties like strength.
Hence, reasoning shows that it is perfectly logical
that Abrams found that water-cement ratio plays
such an important role in the determination of the
compressive strength of concrete.
Two mistakes were made in this reasoning. First
the microstructure of a material is not only deter-
mined by its components (its mix design), but also
by the process in which it is made (e.g. temperature,
mixing, placing, curing, etc.). Second, the water-
cement ratio had obtained too strong a position in
the determination of the compressive strength. It
had gotten to the point that the water-cement ratio
determined the strength. However, already in the
work of Abrams it is shown that one unique water-
cement ratio does not result in one unique value of
the compressive strength.
Nevertheless, led by the Nordic countries in Eu-
rope a method was developed and turned into a
NORDTEST standard (NT Build 361 (Nordtest 1991-
02), 1991), based on the analysis of microstructure
using thin sections and optical microscopy. The
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
method has been used extensively over the last 20
years to determine w/c ratio of hardened concrete.
According to Round Robin tests performed in Den-
mark, the expected accuracy of the method is ±0.02
when the analysis is performed by an experienced
microscopist (Jakobsen et al., 1995).
A drawback of the method is that the analysis
include the personal judgment of the microscopist,
who estimates as best he can, in comparison with
reference samples, the w/c ratio of the microstruc-
ture under investigation. This aspect together with
the observation that the determined water-cement
ratio came often back as a range rather than a fixed
number may well have resulted in the very limited
use of the technique for w/c ratio determination in
other countries around the world.
It is here that the story also may have reached its
own tail. As the w/c ratio has obtained a high com-
mon practice for the compressive strength, deter-
mining the w/c ratio back through the microstruc-
ture of a concrete expects again a high precision
in the determination of the w/c ratio, as they are
considered to be linked directly together. As the
determination of the w/c ratio does not provide
the accurate number that is expected, there must
be something wrong with the method, forgetting in
the process that one w/c ratio does not provide one
strength, but a range of strengths, which is reflected
in itself by a range of microstructures. Hopefully
this article will bring back some nuance and under-
standing in this relationship between w/c ratio and
strength.
IV. Representative volume
When a microscopy technique is involved, at some
point the discussion need to touch upon the subject
of representative volume. Through the very basic
of the procedure of using a microscope one zooms
in on the details. But at what point has the magni-
fication become too large and are the details that
are being looked at not representative any more for
the whole structure? This dilemma is addressed
through the topic of representative volume.
A material microstructure can always be consid-
ered as a repeating system. As long as the smaller
volume through repetition can recreate the original
larger microstructure, the representative volume is
still there and can be used to say something about
the complete microstructure.
For concrete under a microscope this system of
repeated representative volumes could break down
rather quickly when concrete is made with larger
aggregates in the order of 32 mm or above; hav-
ing a thin section with an area of 50 x 30 mm is
probably not representing the entire microstructure
anymore. When as a rule of thumb at least three
times the largest aggregate is being used to obtain
a representative volume, the thin section sample
size should be in the order of 100 x 100 mm. Even
though at TNO it is possible to produce such large
thin sections, it is expensive and not very practical.
Ordinarily the microstructure of importance is
the hardened cement paste, as well as the distri-
bution of the smaller filler materials in interaction
with the hardened cement paste. As this article fo-
cuses on the water-cement ratio, the representative
volume of the level of the largest aggregate is not
needed. However, at the level of the hardened ce-
ment paste the microstructure within a thin section
of 50 x 30 mm is generally well within the necessary
representative volume.
Next step is to determine what at that micro-
scopic level is the required representative volume.
This turns out to be a bit of a chicken and the egg
story. When the microstructure is very homoge-
neous, larger magnifications are possible before the
image in the microscope is not representative for
the generalized microstructure anymore. At that
time two choices can be made.
First and most easy choice is to zoom out; go to
lower magnifications until the representative vol-
ume is restored. However, this approach loses de-
tails observed at higher magnifications. Therefore,
the second approach is also used a lot. This consists
of collecting more image information through mul-
tiple images. Ordinarily this was done by moving
the sample around and try to mentally summarize
the different field of views. Nowadays, the more
chosen approach especially when quantifying pa-
rameters is to take multiple images and average
through image analysis software and procedures.
In this article the road of taking multiple images
has been followed.
V. w/c ratio determination
At TNO w/c ratio determination is performed rou-
tinely by our microscopists in the way described
by NT Build 361 (NT Build 361 (Nordtest 1991-02),
1991). However, as a side research line over the past
years procedures have been developed to determine
the w/c ratio of concrete through the use of image
analysis techniques.
On this path we have not been the first to try
so. For example Jakobsen et al. (Jakobsen et al.,
1995) and Elsen et al. (Elsen et al., 1995) have tried
and reported on these techniques already in the
nineties of last century. A combination of camera
resolution, computer possibilities and image analy-
sis software may have resulted in a resting state of
the developments possible in this area.
TNO gained renew interest in the topic around
2008. We have reported on our findings regularly
in the conference series of EMABM with articles in
2009 (Einarsson et al., 2009) and in 2011 (de Rooij
263
 de Rooij & Mendonça Filho
Figure 6: Same location of a concrete thin section observed under two different light modes: (A) normal light mode;
(B) ultra violet light mode.
et al., 2011). This article follows up on the line
reported in these two articles. For details on the
technique please see the mentioned articles. Fur-
thermore, publications to describe the entire pro-
cedure are in preparation for journal publication.
Here only a short highlight of the used principles
are given.
The backbone of the NT Build standard is pre-
sented in Figure 5. A concrete sample is prepared
into a thin section, which is thin enough to look
through. Normally thin sections are 25 - 30 µm
thick, sometimes thinner. In the process an epoxy is
used to stabilize the porous and brittle microstruc-
ture of concrete at these fragile scales. The epoxy
enters the sample through a vacuum impregnation
process causing it to settle at the dried pores of the
capillary pore system. The epoxy contains a yellow
dye that reacts in the ultra violet light range by
exciting light. With the right filters this transforms
an ordinary thin section to change from a normal
light image as shown in Figure 6A, to an ultra violet
light image as shown in Figure 6B.
An experienced microscopist can compare the
general brightness of the UV-image to a set of refer-
ence samples with known w/c ratios to determine
the w/c ratio of an unknown sample. The tech-
nique is based on the relation that a sample with
a higher w/c ratio has a higher capillary porosity,
which shows up as a brighter image due to the
higher amount of epoxy being present.
Over the last years TNO has worked consistently
to improve this technique from an experienced mi-
croscopist observation to a computer interpreted
value with a calculated number rather than the
approximate estimation of the experienced micro-
scopist. In other words, we would like to take the
honest but subjective interpretation of the individ-
ual as much as possible out of the equation. The
current article is one step further on this path.
264
VI. Should two images provide the
same answer?
In previous publications the technique how to de-
termine a value for a water-cement ratio using com-
puter analysis has already been described. It is
possible. What remained is the question if two im-
ages taken from the same thin section should give
the same w/c ratio result? That topic is analyzed
in more detail here.
To obtain a reasonable answer on this question
naturally the microstructure to be analyzed should
be relatively homogeneous. Hence the investiga-
tion was not performed on damaged samples or on
concrete with placement or degradation problems.
Instead the analysis were performed on the refer-
ence concrete samples TNO has for its Portland
cement with a known water-cement ratio. As these
thin sections are made from samples with a known
water-cement ratio, the first assumption is that the
analysis of two spots of the hardened cement paste
in the same thin section should provide the same
water-cement ratio.
Unfortunately, regardless of how accurate and
precise we tried to perform our analysis, almost
never the results were exactly the same. Naturally,
as the thin section was only containing one mi-
crostructure, for long we concluded that something
must be wrong with our method, as we did not
obtain the answer we were expecting. That is until
we reached the point that no other conclusion was
possible but the obvious one: it was not the method
itself, but the local variation in the microstructure
that we picked up.
Similar to the truck mixer of concrete not produc-
ing compressive strength results with all the same
value, the production of reference samples did not
result in an equal microstructure at all the same lo-
cations. But if the microstructure is not everywhere
the same, how would we know which microstruc-
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
ture would represent the actual true water-cement
ratio?
Again we followed the all too familiar path also
used in compressive strength measurements: we
need more images to be able to average towards the
most probable water-cement ratio.
VII. Analysis of multiple images
More images, especially in the digital age is not a
problem. However, it does make a difference how
to analyze these images. If we have established in
the previous section that it is very unlikely that two
images give exactly the same result upon analysis,
then the images are different. For reasoning pur-
poses, let’s assume that we have taken four images.
To quantify these relatively to each other we have: a
dark image, a medium-dark image, a medium-light
image and a light image. When all four images
are analyzed we will know the exact distribution
of the results. However, what would be the result
if only two images are analyzed? Does this then
depend completely random on which images we
took first? It could be that the averaged values are
on the darker side when by chance only the dark
and medium-dark image are taken. It could also be
that the average results are on the light side when
again by chance the light and medium-light image
are analyzed only.
To solve this problem the research and analy-
sis were set up as follows. From a thin section
with a known water-cement ratio a total of 15 lo-
cations of the microstructure were photographed
and analyzed individually. Next, the individual re-
sults were (figuratively speaking) placed in a basket.
From this basket, using a computer script, 50,000
times sets of images were drawn and analyzed for
mean and standard deviation. The drawing of the
samples was done in such a way that once an image
was analyzed in a set, it could not be redrawn and
reappear again in the same set during the same
drawing sequence. Thus plots were made for an-
alyzing sets of 3 images, sets of 4 images, all the
way up to sets of 15 images. This was done for the
range of reference samples of Portland cement with
different water-cement ratios.
In Figure 7, as an example, the results are shown
for a w/c ratio of 0.65. The results are plotted in
series with different amount of images selected to
be analyzed. So the series of 3 images is the result
of 50,000 times drawing sets of 3 images out of the
total of 15 images. From this graph it becomes clear
that more images indeed do narrow down the pre-
cision of the methodology. Please also note with
what accuracy the water-cement ratio can be ana-
lyzed using this technique; the axis representing the
w/c ratio is drawn with marks of 0.01 difference.
In Figure 8 the results of the different w/c sam-
ples are plotted in one graph, using only three
images to analyze the w/c ratio each time. As the
plot shows, it is quite difficult based on such a low
number of analyzes to determine the w/c ratio of
the sample under investigation. There are large
overlaps in the possible water-cement ratios. How-
ever, turning to Figure 9, it becomes clear that an
increase in the number of images provides a much
more accurate reliability on the w/c ratio of the
sample under investigation.
Figure 9 also shows another interesting result.
Even though the preparation of the reference sam-
ples has been done in laboratory conditions with
the utmost precaution, the variation in the final
microstructure is not everywhere identical. From
the distribution in the results it becomes clear that
the sample with w/c ratio of 0.61 is slightly more
heterogeneous than the other samples presented
in the graph. This is an observation that can only
be quantified in numbers through the image anal-
ysis technique used here. It shows that even our
preciously made reference samples show different
local variation in their microstructure.
VIII. How many images should we
analyze?
The easy answer to the question how many samples
should be analyzed is: more is better. However, that
is only in relation to collecting data. Equally impor-
tant is the realization what is needed, or requested
from the results of all the data analyses. If the re-
quest is that the outcome of the analysis should be
an exact w/c ratio without any distribution, than
the quest to determine this has failed before it has
started. The microstructure of concrete is hetero-
geneous and not identical at each and every place.
Because there is variation it is not possible to come
up with an answer without any deviation from the
mean. It is simply not present in the material.
Hence, the question should be: what is being
done with the result of the analysis? And how ac-
curate should this result be to be useful input for
further processing. Now the answer has a relation
with the next step in the process. That introduces a
dependency, which could lead to a more demand-
ing accuracy in one case (and hence possibly more
images to be analyzed), while in another case the
answer could leave room for more uncertainty re-
sulting in a lower number of images to be analyzed.
Obviously in this process two other parameters
play a very important factor: more images usually
mean a higher price that needs to be paid. The
other factor is the heterogeneity of the concrete it-
self. A more homogeneous microstructure requires
less images to be analyzed to come to a certain
265
 de Rooij & Mendonça Filho

Figure 7: Histogram plot of 50,000 times selecting sets of 3, 5, 7, 9, 11, 13 or 15 images (see legend) out of a set of 15
images taken from a thin section with an original w/c ratio of 0.65 to determine the water-cement ratio.

precision than a very heterogeneous microstructure numbers of pictures taken, and average the light in-
with large differences in the local microstructure. tensities of these images. This then is being used to
obtain a mean w/c ratio. In the process of obtaining
To give some guidance in this process, the follow-
an average light intensity also a standard deviation
ing comparison has been made. In the procedure
is obtained. Taking the mean brightness ± once
described so far, random picks have been made out
or twice the standard deviation and using this to
of a collection of 15 images per thin section sample.
determine the corresponding w/c ratio provides a
This provides us with a w/c ratio calculated from
different way of obtaining w/c ratio distribution.
the average brightness value of the cement paste
These lines are marked ’All’ in Figure 10. As can
from the randomly picked set of specified number
be seen, using this latter method to determine the
of images. By doing this 50,000 times a distribution
w/c ratio distribution gives a much wider range of
in the w/c ratio is obtained. Using the standard
results.
deviation of this distribution, for instance the mean
± once the standard deviation (68% of all values) Hence, to really characterize the distribution in
or mean ± twice the standard deviation (95% of the w/c ratio, or rather in the heterogeneity of the
all values) can be plotted in a graph to give an in- microstructure, the first method is strongly advised.
dication on the accuracy. This has been done in
Figure 10 with the graphs labelled ’Pick’. The accu-
IX. The usefulness of these analysis
racy with which the w/c ratio can be determined
is very precise.
What can we do with all these analysis? First of all
Another option is to take just 3, 4, 5, or any of the it provides us with a way to say something about

Figure 8: Distribution of the w/c ratios for the samples mentioned in the legend based on analyzing only 3 images

266
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 9: Distribution of the w/c ratios for the samples mentioned in the legend based on analyzing 9 images per thin
section of a specific w/c ratio mentioned in the legend

the local variation in capillary porosity. Hence, all
properties that are influenced by this local variation
of capillary porosity can now be studied at a mi-
crostructure level in more detail through the use
of optical microscopy. Obviously it would be nice
to be able to link the variation in strength to the
variation in local microstructure. However, much
more interesting would be the characterization of
the local microstructure in terms of transport prop-
erties. Can we relate the local variation to values
and distribution of transport parameters so we can
start to understand and model the ingress patterns
that we see with for example Rapid Chloride Mi-
gration or carbonation, see e.g. Figure 11. We know
these ingress patterns are not straight lines into the
material as our current models assume. However, is
it possible with this new local variability to predict
the local variation in ingress that we see at a higher
scale. If so, then we are making progress if only
by understanding that our concrete material is so
much more than just a recipe with emphasis on the
water-cement ratio. A possible way to do so could
be by using the width of the w/c ratio distribution,
or rather the variation in the capillary porosity as
a measure for the variation in the ingress rate of
substances. This would open up possibilities to
calculate ingress based on probability distributions
rather than as a pure deterministic diffusion alone.
X. Conclusion
This article has taken a closer look at the determina-
tion of the w/c ratio through the analysis of capil-
lary porosity by using image analysis. It has shown
that accurate analysis does not shown identical re-
sults in the w/c ratio even for standard reference
samples with known w/c ratio. This is not because
the techniques we use to determine the w/c ratios

Figure 10: Plot to indicate the possible accuracy for determining w/c ratio. The lines marked ’All’ are based on just
averaging over the collected images. The lines marked ’Pick’ use a random picking of the indicated number of pictures
out of a set of 15 images for 50,000 times.

267
 de Rooij & Mendonça Filho
Figure 11: Ingress pattern of chlorides into a concrete
sample after a Rapid Chloride Migration test
are not accurate enough, but much more because
of the local variation in the microstructure.
For the first time the attention has shifted through
this article from the analysis technique we use to
what we actually observe in the microstructure.
Hopefully this change in thought about the dis-
tribution in local microstructure variation will help
to set off more research towards the capturing and
modelling of these microstructure variation into
consequences at a more structural level.
268
References
Abrams, D. (1918). Design of Concrete Mixtures. Chicago: Bulletin No.
1, Structural Materials Research Laboratory.
Einarsson, G., Valcke, S., Nijland, T., Copuroglu, O., & Larbi, J. (2009).
An alternative approach to estimate the w/c ratio of hardened concrete
using image analysis. Proceedings of the 12th Euroseminar on Microscopy
Applied to Building Materials. Dortmund, Germany.
Elsen, J., Lens, N., Aarre, T., Quenard, D., & Smolej, V. (1995). Deter-
mination of the w/c ratio of hardened cement paste and concrete sam-
ples on thin sections using automated image analysis techniques. Cement
and Concrete Research, 25(4), 827-834.
European Economic Community. (2011). EU regulation No 305/2011
of the European Parlement and the Council of 9 march 2011 laying down
harmonised conditions for the marketing of construction products and
repealing Council Directive 89/106/EEC. Official Journal of the European
Union, L88/5.
European Standardization Committee. (2009). Survey of national re-
quirements used in conjunction with EN 206-1:2000. CEN Technical Report
15868.
Idorn, G. (1967). Durability of concrete structures in Denmark - a study
of field behavior and microscopic features. Copenhagen, Denmark: Technical
University of Denmark.
Jakobsen, U., Johansen, V., & Thaulow, N. (1995). Estimating the
capillary porosity of cement paste by fluorescence microscopy and image
analysis. Materials Research Society Symposium Proceedings, 370, 227-236.
McLellan, R. (2009, March/April). Louisiana’s first 10,000 psi box
girder bridge with U-beams. HPC Bridge View(54).
NT Build 361 (Nordtest 1991-02). (1991). Concrete, hardened: water-
cement ratio. Nordtest method.
de Rooij, M., Valcke, S., & Suitela, W. (2011). An update on using
image analysis to determine w/c ratio of concrete. Proceedings of the
13th Euroseminar on Microscopy Applied to Building Materials. Ljubljana,
Slovenia.
Thaulow, N., Jensen, A., Chatterji, S., Christensen, P., & Gudmunds-
son, H. (1982). Estimation of the compressive strength of concrete sam-
ples by means of fluorescence microscopy. Nordisk Betong(2-4), 51-52.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Fluorescence laser scanning confocal

microscopy for real-time imaging of early
cement hydration
Marcus H. N. Yio⇤ , Hong S. Wong and Nick R. Buenfeld
Concrete Durability Group, Department of Civil and Environmental Engineering, Imperial College London, SW72AZ
⇤ marcus.yio11@ic.ac.uk

Abstract

This is the first investigation into the feasibility of fluorescence laser scanning confocal microscopy (LSCM) for
real-time imaging of hydrating cementitious materials. Fluorescence LSCM allows continuous imaging of a wet
sample under ambient conditions. Hence, it is potentially suitable for studying cement hydration and microstructural
development of cementitious materials which are very sensitive to moisture and temperature. Four cement pastes
containing fluorescein and supplementary cementitious materials were continuously observed with LSCM from 15 min
after mixing up to 7 days. Results clearly show the development of amorphous and crystalline phases in originally
water-filled pores. Formation of ’Hadley’ grains (hollow-shell particles) was also observed. Image analysis was performed
to quantify the growth of hydration products as a function of time and results were compared with simulation data
obtained from the Virtual Cement and Concrete Testing Laboratory. The strengths and limitations of fluorescence LSCM
for real-time imaging of cement hydration are discussed.

Keywords: Cement hydration, laser scanning confocal microscopy, concrete, real-time imaging, microstructure

I. Introduction with keeping the sample at the required relative

t When raw cement comes into contact with water, it
undergoes flocculation, agglomeration, dissolution
and eventually forms a range of hydration products.
Within 24 hours, the cement-water mixture evolves
into an interconnected multiphase solid that gradu-
ally develops strength. The processes become more
complicated when supplementary cementitious ma-
terials (SCM) or other additives are involved. The
hydration of cement governs not only the early age
properties of concrete such as workability and set-
ting, but also the development of microstructure
which influences the mechanical properties and
long-term durability of concrete.
There are several methods to monitor the early
hydration of cement. Examples include isothermal
calorimetry which measures the heat of hydration
(Lagier & Kurtis, 2007), X-ray diffraction (Scrivener
et al., 2004), nuclear magnetic resonance (Johansson
et al., 1999), and Fourier transform infrared spec-
troscopy (Ylmén et al., 2009) which follow the chem-
ical transformations of constituent phases at early
age. In contrast, microscopy techniques enable
morphological study of the evolving microstruc-
ture. Environmental scanning electron microscopy
is able to image hardened cement paste without
the need for drying. However, continuous imag-
ing at early ages is challenging due to difficulties
humidity and temperature. Moreover, observation
is made in the secondary electron mode and is qual-
itative. A relatively new method called WETSEM
enables fresh cement paste to be encapsulated in
a sealed capsule and viewed continuously in the
backscattered electron mode (Gallucci & Scrivener,
2007). However, problems with wall effects and
growth of hydration products on the surface mem-
brane of the capsule limit the extent of observation.
Synchrotron transmission soft X-ray microscopy of-
fers high-resolution imaging of hydration products
forming on cement grains over time, but this tech-
nique is limited to dilute specimens with a high
water/cement ratio of 5 and very thin specimens
that are less than 10 µm thick (Juenger, 2005).
Fluorescence microscopy techniques are well-
established and widely used in life sciences. Quanti-
tative data can be derived based on the fluorescence
emitted by fluorophores. One of the most com-
monly used fluorescence microscopy techniques is
laser scanning confocal microscopy (LSCM). LSCM
is able to perform real-time imaging at very high
spatial and temporal resolutions under ambient
conditions without requiring sample drying. More-
over, it has the ability to perform three-dimensional
imaging by means of non-intrusive optical section-
ing. Examples of applications in life sciences in-

269
 Yio et al.

Table 1: Mix proportions.

Mix ID Binder (g), Fluorescein Water w/b SCM/CEM I SCM/binder

CEM I LF FA GGBS (g) (g)
CEM I 25.00 - - - 0.0063 12.5 0.5 - -
CEM I + 17% LF 20.75 4.25 0.0063 12.5 0.5 0.20 0.170
CEM I + 23% FA 19.25 5.75 0.0063 12.5 0.5 0.30 0.230
CEM I + 37.5% GGBS 15.63 9.375 0.0063 12.5 0.5 0.60 0.375

clude dynamic imaging of cell division (Rump et
al., 2011) and host-pathogen interactions in tissues
(Coombes & Robey, 2010). In cement and concrete
research, fluorescence LSCM has been used by the
authors’ group for static 3D imaging of capillary
pore structure, ’Hadley’ grains and microcracks in
dried and epoxy impregnated concrete (Head &
Buenfeld, 2006, Head et al., 2006). Recently, a new
method that combines fluorescence LSCM with se-
rial sectioning was proposed to enhance its ability
to image large volumes of hardened concrete at
sub-micron resolution (Yio et al., In press). Never-
theless, fluorescence LSCM has never been applied
to study hydrating cementitious materials.
In this paper, we present an approach to continu-
ously image the early age microstructural evolution
in four different cementitious systems using fluo-
rescence LSCM. Fluorescein was incorporated into
the samples to act as tracer. Quantification of the
growth rate of hydration products was performed
by means of image analysis and results were com-
pared with simulations. The advantages and limita-
tions of the method are identified and discussed.
II. Experimental
excited using a 488 nm argon laser at a very low in-
tensity of 1% to ensure that variation in the emitted
fluorescence was detected. The sensitivity of the
photomultiplier tube (PMT) was set at a constant
625 V. It was found that a concentration of about
0.5 wt.% gave the maximum fluorescence intensity
(Figure 1). Higher concentrations led to reduced in-
tensities due to self-quenching of the fluorophores.
However, when a higher laser intensity of 15% (i.e.
the default setting for fluorescein) was used, all
concentrations > 0.01 wt.% were able to give the
maximum fluorescence intensity with no detectable
variation.
The effect of fluorescein on the hydration of ce-
ment at 20 C was also investigated using an isother-
mal conduction calorimeter (Calmetrix I-Cal 4000).
CEM I pastes at a water/cement ratio of 0.5 pre-
pared using aqueous solutions of four different
fluorescein concentrations (0, 0.1, 0.5 and 1.0 wt. %)
were tested. Results show that fluorescein concen-
trations  0.1 wt.% had no influence on the heat of
hydration (see Results for further details). Based on
the findings above, a fluorescein concentration of
0.05 wt.% was used for the batching water through-
out this study

Materials Sample preparation

Four different paste samples with a water/binder Pastes were prepared by hand-mixing cement with
(w/b) ratio of 0.5 were prepared using ordinary fluorescein-dyed water for 2 min in a plastic cup.
Portland cement (CEM I), fly ash (FA), ground gran- In the case of blended systems, SCM/fillers were
ulated blastfurnace slag (GGBS), limestone filler pre-mixed with CEM I for 30 s before addition of
(LF) and fluorescein-dyed tap water. Mix propor-
tions for all samples are shown in Table 1. The Table 2: Composition of fly ash and slag.
Portland cement used complies with BS EN 197 1.
The limestone filler had 70 wt.% CaCO3 and a
particle size distribution of > 70 wt.% passing 63 Chemical analysis (wt.%) FA GGBS
µm. The composition of the fly ash and slag used SiO2 72.2 36.5
are shown in Table 2. Al2 O3 24.3 11.6
Water-soluble fluorescein with a maximum ab- Fe2 O3 0.4 1.4
sorption wavelength and a maximum emission CaO 0.1 40.8
wavelength of 490 nm and 520 nm respectively was MgO 0.1 7.5
used. In order to establish the optimum concentra- SO3 0.1 2.1
tion of fluorescein, aqueous solutions containing Na2 Oequivalent 0.3 0.5
0.0005 to 5 wt.% fluorescein were prepared and
mixed with a magnetic stirrer for 24 h and left at Physical analysis
room temperature until the fluorophores dissolved Laser granulometry d50 (µm) 7 8
completely. Droplets of the solutions were then

270
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 1: Plot showing fluorescence intensity versus
concentration of fluorescein (excited using an argon laser
of 488 nm at 1% intensity).
water. A small amount of the freshly mixed paste
(⇠0.06 g) was promptly transferred onto a glass
slide and mounted with a coverslip. The coverslip
was gently pressed down on the slide to flatten the
paste to approximately 500 µm thick. A small quan-
tity of fluorescein water (⇠0.1g) was then carefully
placed around the cement paste using a pipette to
fill the empty space between the coverslip and glass
slide. This was to prevent the paste from drying
and to provide curing water. In order to minimise
evaporation, a generous amount of nail polish was
applied to seal the edges of the sample assembly.
Figure 2 shows a schematic representation of the
sample. Two samples containing just fluorescein
water (no cement paste) were also prepared and
weighed after 30 min, 24 h, 4 d and 7 d to check the
effectiveness of the seal. Results show that the total
mass loss at 7 d was 18.85% of the water originally-
fed. By experience, paste can remain saturated for
at least 4 weeks. All weighing procedures described
above were performed using an analytical balance
with a precision of 0.0001g.
LSCM
A Leica TCS SP5 laser scanning confocal microscopy
was used for imaging. An argon laser line of 488 nm
was used to excite fluorescein and the beam splitter
was configured to range from 500 nm to 600 nm to
allow emitted fluorescence to be collected by the
PMT. A 20⇥ (NA 0.5) dry objective with a spatial
XY resolution of 0.39 µm was used. Images were
zoomed at 3.5⇥ and digitised to 1024 ⇥ 1024 to give
a pixel spacing of 0.216 µm. The Nyquist theorem
states that the smallest feature that can be resolved
is either 2.3⇥ the pixel size or the spatial resolu-
tion of the objective, whichever is larger (Pawley,
1995). Based on this criterion, the smallest feature
that could be observed here would be around 0.5
µm. The laser intensity was kept constant at 15%
whilst the PMT was adjusted correspondingly to
prevent oversaturation of pixels. The pinhole size
was maintained at 1 Airy Unit and two-times line
averaging was applied to reduce noise.
Real-time imaging was performed starting from
15 min after mixing at 1 h intervals up to the age of
24 h, then every 12 h up to 48 h, and finally every
24 h up to 168 h (7 d). A total of ⇠ 990 images
were captured per sample. At each time step, a
3⇥3 mosaic of 3D image stacks (9 stacks in total)
with a z-step size of 0.84 µm was captured from the
surface to the deepest possible plane in the sam-
ple. The total scanned area was around 600 ⇥ 600
µm2 . Following image acquisition, an image slice
at ⇠ 7 µm beneath the sample surface was selected
from each 3D stack for image analysis. This was
to minimise any possible ’wall effects’ (see Results
and Discussion sections). A pseudo flat-field filter
was applied to the images to remove uneven bright-
ness. Subsequently, the selected mosaic images
were stitched together based on phase correlation
to form a larger image. The reconstructed images
were then aligned, contrast enhanced by means of
histogram stretching to compensate for signal loss
due to photo bleaching1 and finally median filtered
to be ready for analysis.
III. Results
Isothermal calorimetry
Results from isothermal calorimetry up to 48 h of
hydration show that the heat flow curves of CEM I
pastes containing fluorescein-dyed water were simi-
lar to that of the reference sample which contained
CEM I and pure water (Figure 3). No shoulder was
observed on any of the curves for samples contain-
ing fluorescein. These observations suggest that flu-
orescein does not significantly alter the cement hy-
dration mechanism, within the concentration range
studied. The data for sample with 0.1 wt.% fluores-
cein overlapped that of the control throughout the
induction, acceleration and retardation stages. This
confirms that fluorescein concentrations of up to 0.1
wt.% have no influence on hydration. However, a
slight delay in the acceleration period was observed
at higher fluorescein concentrations. This effect was
particularly apparent for the sample containing 1.0
wt.% fluorescein.
1 Photo bleaching is a photochemical process where fluo-
rophores gradually lose their ability to fluoresce after being
exposed to prolonged laser irradiation.
271
 Yio et al.
Figure 2: Schematic of sample cross section (not to scale).
Observations
Figure 4 shows a series of z-stack images cover-
ing an imaging depth of 0.8 to 16.8 µm below the
surface of the CEM I + 23% FA sample after 15
min of hydration. The images were cropped, con-
verted to greyscale and inverted to highlight the
features of interest. Note that dark areas represent
fluorescein water-filled spaces and the bright areas
correspond to solid phases. The images show that
the microstructure of the first several microns below
the surface is highly porous due to accumulation
of water from bleeding (and possibly curing wa-
ter) beneath the coverslip. Small particles, some of
which were tips of larger particles from within the
sample, were also observed. As the imaging depth
increased, the size of the particles became larger
and their shape became more defined. However, a
gradual loss of signal was observed as the image
plane moved further (> 7 µm) into the sample. This
was because the laser became increasingly blocked
Figure 3: Heat flow and cumulative heat of CEM I pastes
containing various concentrations of fluorescein.
272
by overlying particles and hence was unable to ex-
cite fluorescein from deeper planes. As a result,
the feature resolution degraded with depth. As
mentioned in Methodology, images at ⇠ 7 µm be-
neath the sample surface with no discernible signal
degradation were used for image analysis.
Figure 5 shows examples of areal matching im-
ages from each mix captured at ⇠ 7 µm beneath
the sample surface at 15 min, 12, 24 and 168 h. It is
clear from the figure that all images at 15 min show
particles with rather distinct shapes and boundaries.
Fly ash particles can be easily distinguished based
on their spherical shape. However, it is relatively
more difficult to distinguish slag and limestone
filler particles from cement particles.
It can be observed that cement particles gradually
lose their distinct boundary over time (see particle
marked ’C’ in the CEM I mix in Figure 5(a) for ex-
ample). This is due to dissolution of cement which
releases ions into the mix water to form hydration
products in water-filled spaces and topochemical
reaction that results in the growth of hydration
products on the surfaces of cement particles. In con-
trast, SCM and limestone fillers maintained their
shape throughout the imaging period (see particles
marked ’L’, ’F’ and ’S’ in Figure 5(b), (c) and (d))
due to their low reactivity at early ages. The ap-
pearance of hydration products in the water-filled
spaces is indicated by arrows in Figure 5. It can
be observed that the size and contrast of the pore
spaces gradually decreases over time as they be-
come filled with hydration products that resemble
a network of amorphous phases. It is interesting
to note that some large pores that were initially
occupied by the mix water remained porous at 168
h, indicating that such pores would not be com-
pletely filled with hydration products. Comparing
only the images at 168 h, the CEM I mix appeared
markedly denser than the blended mixes. This is
in line with current understanding of the effect of
SCMs on early age hydration. For example, re-
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 4: Z-stack images at 0.8 to 16.8 µm below the surface of the CEM I + 23% FA sample after 15 min of hydration.
’F’ represents fly ash.

sults obtained from image analysis and the Virtual
Cement and Concrete Testing Laboratory (VCCTL)
(Bullard et al., 2009) simulations show that the CEM
I mix has the highest cumulative volume fraction
of hydration products (see Image analysis).
The development of ’Hadley’ grains i.e. hollow
shell particles was also detectable. Figure 6 presents
a sequence of images showing the evolution of a
shell around a cement particle (marked ’H’) found
in the CEM I mix. At 15 min, the original bound-
ary of the cement particle is distinct, but after 12 h,
the cement particles become increasingly engulfed
with diffuse hydration products. At around 24 h,
a distinct shell structure can be seen forming pref-
erentially on the top of the grain, connecting with
the outer hydration products. Perhaps, the hollow
shell formed much earlier but was too thin (Gal-
lucci et al., 2010) to be resolved by LSCM. Within
the shell, needle-like hydration products that con-
nect the outer products and the anhydrous core
were observed. The hollow shell increased in size
with time, while the anhydrous core diminished.

Figure 5: Imaging results of (a) CEM I, (b) CEM I + 17% LF, (c) CEM I + 23% FA and (d) CEM I + 37.5% GGBS
at 15 min, 12h, 24 h and 168 h. ’C’, ’L’, ’F’ and ’S’ represent cement, limestone filler, fly ash and slag, respectively.

273
 Yio et al.

Figure 6: Development of a ’Hadley’ grain particle marked ’H’ found in the CEM I system.

Ultimately, the shell could become entirely hollow.
Indeed, Hadley et al. (2000) and Head et al. (2006)
observed completely hollow-shell grains in 28-day-
old cement paste and concrete samples. The hydra-
tion products that appear to have an amorphous
morphology are most likely the C-S-H phase with
structures beyond the resolution of LSCM. How-
ever, large plate-like crystals in the size of tens of
µm that are uniform and well-defined were also
observed (see arrows in Figure 7). Some of these
crystals exhibit fluorescence on their edges. These
are portlandite crystals that are preferentially ori-
ented with their C-axis perpendicular to the cov-
erslip interface. Such features are similar to those
reported in studies of the interfacial transition zone
(ITZ) between cement paste and flat rock surfaces
(Yuan & Odler, 1987). The time to the appearance
of the first plate-like crystal was generally between
2 to 3 h for all systems. Similar observations were
made by Gallucci and Scrivener (2007) using the
WETSEM capsules to image the hydration of CEM
I with SEM-BSE. However, in the present study, the
growth of portlandite did not block or limit obser-
vation because imaging was performed subsurface.
This enabled observations to be made for much
longer than what is possible with the WETSEM
capsules.
bright pixels than image It 1 due to the growth of
hydration products. Therefore, the result of sub-
traction can be regarded as the growth of outer
hydration products between time t and t 1. How-
ever, inner hydration products that form within the
original boundary of cementitious particles are ne-
glected. The subtracted images are then segmented
by thresholding the transition point near the higher
end of the histogram to measure the growth of hy-
dration products. This transition point corresponds
arbitrarily to where the change in the cumulative
curve is ⇠ 0.1%. Noise is removed from segmented
images by means of a morphological closing oper-
ation. Artefacts which may result from the move-
ment of imaging plane in the z-axis when mosaic
imaging is performed are reduced by discarding
pixel groups containing < 10 pixels and with a cir-
cularity > 0.75; these criteria were found to yield
satisfactory results based on visual inspection. The
image analysis procedures described above are illus-
trated in Figure 8(a-e). We stress that the results are

Image analysis
Ideally, image segmentation would be performed
directly on the reconstructed images to separate
pore spaces from solids. However, this is rela-
tively challenging because the grey histogram spans
across the entire scale range with no distinct peaks
that can be used to guide thresholding. For pre-
liminary quantification, image subtraction was car-
ried out on a pixel-by-pixel basis between consecu-
tive area-matching images at every time step using
equation (1) to quantify the growth of hydration
products:

DI = It It 1 (1)
Where It is the image captured at time t = Figure 7: Observation of plate-like portlandite crystals
1, 2, 3 . . . 168 h. Image It will always have more (marked by arrows) in the CEM I system at 18 h.

274
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 8: Demonstration of the image analysis procedure on a large crystal found in the CEM I + 17% LF system: (a)
reconstructed image at 6 h, I6 , (b) reconstructed image at 7 h, I7 , (c) result of the subtraction I7 - I6 , (d) segmented
area of (c) and (e) cumulative segmented areas at 6, 12, 24 and 168 h.

preliminary; research is on-going to develop a more
accurate approach to quantify the microstructure.
The cumulative segmented area representing the
total area fraction of hydration products is plot-
ted as a function of time up to 24 h in Figure 9(a)
and up to 168 h in Figure 9(b). During the first
12 h, the trends for all four samples were similar
although a slight acceleration was observed with
those containing 17% limestone fillers and 37.5%
slag. Limestone fillers are known to accelerate the
hydration of cement because they provide nucle-
ation sites for hydration products (Lothenbach et
al., 2008). The slight acceleration observed after
4 h seems to be in agreement with the findings
of Lothenbach et al. (2008). However, slag was
not expected to accelerate hydration at early ages
and therefore further work is needed to clarify if
this observation is a real effect or due to imper-
fections in the image analysis approach adopted
here. After 12 h, the cumulative segmented areas
for blended samples were much lower than that of
the control. The disparity increased with increasing
SCM/filler content. This was due to the reduced
cement content and lower reactivity of the SCMs.
It is also well-established that the pozzolanic reac-
tion of SCMs occurs at much greater ages whereas
limestone fillers are not pozzolanic.
It is instructive to compare the results with mod-
elling predictions. Here, the VCCTL model (Bullard
et al., 2009) was used to simulate the hydration
of the four systems investigated under saturated
condition at 20 C. The volume fractions of outer
hydration products are plotted against time in Fig-
ure 9 (c) and a good qualitative agreement with
our measurements (Figure 9(b)) can be seen. In line
with our observations, the volume fraction of outer
hydration products decreased with increasing SCM
content. However, the retardation effects caused
by SCM were apparent from the start of hydration.
A very slight accelerating effect was also observed
with the system containing 17% limestone filler but
this occurred at a much later time, i.e. between 10
and 19 h. The observed differences could be due to
the fact that our analyses were carried out near the
surface region where the microstructure is affected
by bleeding as discussed in the previous section.
Clearly, this needs to be investigated further.
IV. Discussion
The experimental results obtained show that flu-
orescence LSCM has great potential for real-time
imaging of the hydration process and microstruc-
ture development in cementitious materials. The
main advantage of fluorescence LSCM lies in its
ability to perform continuous imaging of a sample
under wet conditions. Using the sample prepara-
tion procedure described above, imaging of a fresh
cement paste with a realistic w/b ratio of 0.5 can
start as early as 15 min and can last more than
7 d until the sample begins to dry as a result of
self-desiccation.
LSCM is also able to perform three-dimensional
imaging by means of optical sectioning to obtain

275
 Yio et al.
Figure 9: Cumulative segmented area of outer hydration products obtained from image analysis: (a) up to 24 h; (b) up
to 168 h, (c) simulation results from the VCCTL.
subsurface information. This allows imaging of in-
ternal microstructural away from the surface zone.
Measurement of mass of the samples before and
after a routine scanning procedure showed no re-
duction in mass, thus confirming that no drying
occurred due to laser irradiation. Laser exposure is
also not expected to generate heat that can cause
localised acceleration of cement hydration since
the applied intensity is low and is widely used for
imaging living cells and tissues.
However, the imaging depth of LSCM is lim-
ited. As a result, wall effects cannot be completely
eliminated. Another limitation is that only the
fluorescein-filled spaces can be detected. Despite
these limitations, LSCM has advantages over the
WETSEM capsule which permits only surface imag-
ing within a limited time of  13 h (Gallucci &
Scrivener, 2007). Moreover, by means of appro-
priate image processing methods such as contrast
limited adaptive histogram equalisation (CLAHE)
(Zuiderveld, 1994), it may be possible to select im-
ages from deeper planes for analysis. Work to de-
velop this aspect is on-going.
276
V. Conclusion
This study has demonstrated the potential of flu-
orescence LSCM for real-time imaging of hydra-
tion and microstructure development of cementi-
tious materials. It is shown that, unlike other tech-
niques, fluorescence LSCM is capable of performing
three-dimensional continuous imaging of hydrat-
ing cementitious materials at sub-micron resolution
under ambient conditions. This aspect is particu-
larly important for studying these materials which
are moisture and temperature dependent. Four
cementitious systems containing various amounts
of SCM/fillers were observed with LSCM from 15
min after mixing up to 7 days of hydration. Various
interesting observations such as the appearance of
amorphous hydration products in pores, growth of
portlandite crystals and formation of ’Hadley’ grain
particles were made. Results from image analysis
show that the volume fraction of outer hydration
products decreased with increasing SCM/filler con-
tent and this was consistent with simulation data
from the VCCTL.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Acknowledgement
M.H.N. Yio would like to acknowledge the Dixon
scholarship awarded by the Department of Civil
and Environmental Engineering, Imperial College
London. We would also like to acknowledge Cal-
metrix Inc. for the use of their calorimeter.
References
Bullard, J., Stutzman, P., Belloc, L. O., Garboczi, E. & Bentz, D. (2009)
Virtual cement and concrete testing laboratory for quality testing and
sustainability of concrete. ACI Special Publication, 266.
Coombes, J. L. & Robey, E. A. (2010) Dynamic imaging of host-
pathogen interactions in vivo. Nature Reviews Immunology 10 353-364.
Gallucci, E., Mathur, P. & Scrivener, K. (2010) Microstructural devel-
opment of early age hydration shells around cement grains. Cement and
Concrete Research, 40, 4-13.
Gallucci, E. & Scrivener, K. (2007) In situ dynamic SEM imaging of
hydration of cement using WETSEM technology. Advances in Applied
Ceramics, 106, 319-326.
Hadley, D. W., Dolch, W. L. & Diamond, S. (2000) On the occurrence
of hollow-shell hydration grains in hydrated cement paste. Cement and
Concrete Research, 30, 1-6.
Head, M. K. & Buenfeld, N. R. (2006) Confocal imaging of porosity
in hardened concrete. Cement and Concrete Research, 36, 896-911.
Head, M. K., Wong, H. S. & Buenfeld, N. R. (2006) Characterisation
of ’Hadley’ grains by confocal microscopy. Cement and Concrete Research,
36, 1483-1489.
Johansson, K., Larsson, C., Antzutkin, O. N., Forsling, W., Kota,
H. R. & Ronin, V. (1999) Kinetics of the hydration reactions in the ce-
ment paste with mechanochemically modified cement 29Si magic-angle-
spinning NMR study. Cement and Concrete Research, 29, 1575-1581.
Juenger, M. C. G., Monteiro, P. J. M., Gartner, E. M. & Denbeaux, G. P.
(2005) A soft X-ray microscope investigation into the effects of calcium
chloride on tricalcium silicate hydration. Cement and Concrete Research,
35, 19-25.
Lagier, F. & Kurtis, K. E. (2007) Influence of Portland cement composi-
tion on early age reactions with metakaolin. Cement and Concrete Research,
37, 1411-1417.
Lothenbach, B., Le Saout, G., Gallucci, E. & Scrivener, K. (2008) In-
fluence of limestone on the hydration of Portland cements. Cement and
Concrete Research, 38, 848-860.
Pawley, J. B. (1995) Handbook of biological confocal microscopy, Plenum
Press, New York.
Rump, A., Scholz, T., Thiel, C., Hartmann, F. K., Uta, P., Hinrichs, M.
H., Taft, M. H. & Tsiavaliaris, G. (2011) Myosin-1C associates with micro-
tubules and stabilizes the mitotic spindle during cell division. Journal of
cell science, 124, 2521-2528.
Scrivener, K. L., FÃijllmann, T., Gallucci, E., Walenta, G. & Bermejo, E.
(2004) Quantitative study of Portland cement hydration by X-ray diffrac-
tion/Rietveld analysis and independent methods. Cement and Concrete
Research, 34, 1541-1547.
Yio, M. H. N., Mac, M. J., Wong, H. S. & Buenfeld, N. R. (In press) 3D
imaging of cement-based materials at sub-micron resolution by combin-
ing laser scanning confocal microscopy with serial sectioning. Journal of
Microscopy.
Ylmén, R., Jäglid, U., Steenari, B.-M. & Panas, I. (2009) Early hydra-
tion and setting of Portland cement monitored by IR, SEM and Vicat
techniques. Cement and Concrete Research, 39, 433-439.
Yuan, C. Z. & Odler, I. (1987) The interfacial zone between marble
and tricalcium silicate paste. Cement and Concrete Research, 17, 784-792.
Zuiderveld, K. (1994) Contrast limited adaptive histogram equaliza-
tion. In: Graphics gems IV (ed. S. H. Paul). Academic Press Professional,
Inc., San Diego, United States.
277
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

278
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Composition and morphology of C-S-H in

C3S pastes hydrated by water/isopropanol
mixes using electron microscopy
E. Boehm-Courjault⇤ , J.E. Rossen and K. Scrivener
Laboratory of Construction Materials, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
⇤ emmanuelle.boehm@epfl.ch

Abstract

The present paper aims at better understanding the mechanisms controlling the kinetics of cement hydration and of
the limiting factors of the characteristic periods observed by calorimetry, slightly simplifying the problem by studying
the hydration of pure C3 S instead of real cement. During hydration, C-S-H is formed. It forms the major part of the
binding phase in the cement paste. In C-S-H, the ratio between Ca and Si is not fixed and can vary according to the
experimental conditions and especially the composition of the solution it is in equilibrium with.
Four C3 S pastes were produced by mixing C3 S powder with pure water or with mixes of water and isopropanol
(with isopropanol contents of 7.5, 17.5 and 25 vol%) with a constant liquid/solid ratio of 0.5. Hydration was stopped
by freeze-drying at different characteristic periods of the calorimetry curves: during the induction period, at the main
hydration peak and in the deceleration period at a degree of hydration of 35 %.
Different techniques were used for characterising the pastes at each hydration stage: thermo- gravimetric analysis
(TGA), scanning electron microscopy (SEM) and X-ray energy-dispersive spectroscopy (EDX) in both a SEM and
a transmission electron microscope (TEM). For SEM-EDX analyses, polished sections of the pastes were produced,
whereas for TEM-EDX analyses, samples were thinned down to TEM lamellae by focused ion beam (FIB).
In this paper, SEM and TEM images are presented, showing the fibrillar morphology of C-S-H at the different
hydration stages for each paste. The Ca/Si ratio measured by TGA, SEM-EDX and TEM-EDX are presented as well.
The results are in good agreement with the generally observed values.
The effect of isopropanol on hydration is studied, considering the morphology and the composition of C-S-H. It is
shown that it has no influence on the hydration mechanism but only on the kinetics of the reaction.

Keywords: C3 S; C-S-H; scanning electron microscopy; transmission electron microscopy

I. Introduction C3 S pastes, keeping a liquid/C3 S ratio of 0.5. Four

Tricalcium silicate Ca3 SiO5 or C3 S is the major
phase of Portland cement (PC) clinker. Most of
it is impure, incorporating ions such as aluminium,
magnesium or zinc and called alite. Its hydration
controls the hydration of cement at early age. The
reaction of C3 S with water results in the forma-
tion of both calcium hydroxide (CH or Portlandite)
and a calcium silicate hydrate phase with variable
composition (C-S-H), according to the following
equation:
C3 S + (3 x n) H2 O ! Cx S Hn + (3
Isopropanol is an inert solvent and it is known
for its weak impact on the microstructure of C-S-
H. It is commonly used to stop the hydration of
cement phases (Zhang and Sherer, 2011; Makar and
Sato, 2013).
In this study, isopropanol was chosen for replac-
ing part of the water mixed with C3 S to obtain
different compositions of isopropanol/water mixes
were chosen: 0; 7.5; 17.5 and 25 vol% of isopropanol.
The sample hydrated with 0% of isopropanol is
used as a reference.
By hydrating C3 S with these different mixtures,
the effect of isopropanol on the hydration kinectics
of C3 S is studied. The morphology and composition
of C-S-H at different degrees of hydration are also
investigated.
As it can be seen in reaction (1), C-S-H does
not exhibit a stoichiometric composition. Studies
regarding the composition of this phase have given
x )CH (1) results for mean Ca/Si ratio ranging from 1.2 to 2.3
(Richardson, 1999) in C3 S or PC pastes.
Different characterisation methods can be used
to determine this Ca/Si ratio, such as thermo-
gravimetric analysis (TGA), X-ray diffraction (XRD)
and X-ray energy dispersive spectroscopy (EDX). In
this study, three of them were used: TGA, EDX in a
scanning electron microscope (SEM-EDX) and EDX

279
 Boehm-Courjault et al.
Figure 1: Calorimetry curves of the 4 pastes (a) and related DoH vs. time (b).
in a transmission electron microscope (TEM-EDX),
in order to compare the results at different degrees
of hydration.
II. Materials and methods
C3 S synthesis
Pure C3 S was synthesised by mixing three parts of
calcium carbonate (CaCO3 ) with one part of quartz
powder (SiO2 ). Both raw materials are of high pu-
rity, both supplied by Merck. The mixture was
homogenized 24 hours in water and then dried at
100 C for another 24h hours. The dried material
was crushed and pressed into pellets. The latter
were calcined at 1650 C for 8 hours, with a heating
ramp of 200 C per hour ensuring a complete de-
carbonation of the raw materials. They were then
quenched in air and ground for 30 seconds in a disc
mill. Finally, the resulting material was tumbled
in a Turbula blender for 8 hours, and then sieved
to keep only the fraction smaller than 25 microns.
Batches of around 350 g of C3 S can be obtained by
this technique.
XRD was carried out to ensure that C3 S is the
main phase which was formed as expected, and
Figure 2: Calorimetry curves of the 7.5 % isopropanol-
sample and of a C3S paste with a w/C of 0.4625.
280
that other phases (remaining CaCO3 or SiO2 , free
lime. . . ) are only traces. The C3 S batch used for the
experiments described in this paper contained less
than 0.5 wt% of impurities.
The synthesised C3 S is then always kept in a
desiccator in order to avoid absorption of water
and CO2 from air.
Isothermal calorimetry and preparation of the
specimens
As reaction (1) is exothermic, its kinetics can be
monitored by isothermal calorimetry. A TAM Air
isothermal calorimeter from Thermometric is used.
Figure 1a shows the calorimetry curves of the 4
systems (0; 75; 175; 250) with a liquid/C3 S ratio of
0.5.
At first, C3 S is dissolved very rapidly (some min-
utes). Then there is a sudden slowdown in the
reaction, followed by a period of low chemical ac-
tivity. It is called the induction period and it is very
short for our C3 S batch (some minutes) because the
particles are small (< 25 µm). Thirdly, the main
hydration peak occurs, due to the precipitation of
both C-S-H and CH: this is the acceleration peak.
Then there is a second deceleration, with a long
period of low chemical activity (some days). It is
called the deceleration period.
The degree of hydration (DoH) of the reaction
(1) vs. time can be calculated from the calorimetry
curve, knowing the theoretical enthalpy of hydra-
tion of C3 S (517 J/g, (Taylor, 1997)). Figure 1b
shows the calculated DoH of the 4 pastes vs. time.
Table 1: Hydration times for the 4 pastes, depending on
the chosen period of the calorimetry curve.
% of isopropanol 0 7.5 17.5 25
Induction period 1h 1.17 h 1.33 h 1.75 h
Acceleration peak 6.5 h 7.5 h 9.5 h 16 h
DoH = 35 % 14 h 30 h 66 h 144 h
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Table 2: Hydration times for the 4 pastes, depending on the chosen period of the calorimetry curve.
% of isopropanol 0 7.5
Induction period 0-1h
Acceleration peak 0-6h30
DoH = 35 % 0-14h
It is observed that the overall reaction is slowed
down as the isopropanol replacement increases. It
is caused by the amount of available water for C3 S
to hydrate, but this is not the only explanation.
The deceleration is more pronounced than for C3 S
hydrated with the same amount of water, without
isopropanol. For example, for 7.5 % of isopropanol,
the effective water to C3 S ratio (w/C) is 0.4625.
Figure 2 compares the calorimetry curves of the 7.5
%-sample and a C3 S paste with a w/C of 0.4625.
In presence of isopropanol the induction period
is shifted in time, the main hydration peak occurs
later and its height is smaller. It suggests that there
is another factor influencing the kinetics of the hy-
dration when isopropanol is used. Mixing water
and isopropanol generates electrostatic interactions
in the liquid and results in a decrease of the "reac-
tivity" of water. C-S-H is formed but the kinetics of
this reaction is slowered. It is not currently under-
stood why this occurs.
Hydration was stopped at different times by
freeze-drying; this method is chosen rather than sol-
vent exchange with isopropanol in order to avoid
introducing isopropanol in the system as the hydra-
tion mixes already contain some. At the desired
time, the paste was immersed into liquid nitrogen
(77 K) for 10 to 15 minutes. After freezing, it was
evacuated in a freeze-dryer at low temperature (223
K) and pressure (6 Pa), for at least 24 hours. It was
then stored in a desiccator for at least 2 days before
doing the first measurements.
Hydration times were defined using the calorime-
try curves for each content of isopropanol. Hydra-
tion was stopped within the induction period, at
the acceleration peak and in the deceleration period
at a constant DoH of 35 %. The hydration times
are summarised in Table 1, while the labels of the
corresponding samples are given in Table 2.
Characterisation techniques
Thermo-gravimetric analysis (TGA)
The amounts of C-S-H and CH formed at a given
hydration time were measured by TGA on a Met-
tler Toledo TGA/SDTA 851e device. 50 mg of each
dried sample were introduced in an alumina cru-
cible. It was then heated at 900 C with a 10 C/min
ramp under a 30 ml/min nitrogen flux. The weight
17.5 25
75-1h10 175-1h20 250-1h45
75-7h30 175-9h30 250-16h
75-30h 175-66h 250-144h
of the sample is measured very accurately and con-
tinuously.
Combining this data with the DoH determined
by calorimetry permits the calculation of the Ca/Si
ratio in the C-S-H gel.
Scanning electron microscopy (SEM) and X-ray
energy-dispersive spectroscopy (EDX)
Two different imaging modes were used in the SEM,
with two different devices. The Secondary Electrons
(SE) mode was useful to determine the morphol-
ogy of the C-S-H. This was done with a FEI Sirion
SFEG microscope equipped with an in-lens ultra-
high resolution detector (it will be referred to as
HR-SEM technique in the text), using low acceler-
ating voltages (1-2 kV). A very small quantity of
powder was dispersed on a carbon tape. A car-
bon layer of approximately 15 nm was deposited
on top with a Cressington carbon coater to ensure
electrical conductivity in the SEM.
Backscattered electrons (BSE) were also used to
characterise the microstructure of polished sections.
A FEI Quanta 200 microscope was used, with an
accelerating voltage of 15 kV. Samples are embed-
ded in an epoxy resin and polished until 1-micron
diamond grade. They are coated with a Ballec CED
030 carbon coater. This analysis can only be done
for samples with a rather high DoH (> 35 %), as
for lower ones, the samples are not hard enough to
be polished.
X-ray energy-dispersive spectroscopy (SEM-EDX)
was also carried out in the FEI Quanta 200 (with a
Bruker XFlash EDX detector) on polished sections,
in order to measure the chemical content of Ca and
Si in the C-S-H of the 4 pastes with DoH = 35 %.
Transmission electron microscopy (TEM)
A lot of techniques can be used in a TEM, because
of the wide range of detectable signals. The re-
sults described in this paper are recorded with the
scanning transmission electron microscopy (STEM)
mode: the electron probe is focused into a small
probe and scanned over the sample. The generated
signal is detected at any point of the specimen. Two
different detectors were used: bright field (BF) and
high angle annular dark field (HAADF) ones. X-
ray energy-dispersive spectroscopy (TEM-EDX) was
281
 thinning by FIB.
In each sample, 2 or 3 windows of 5*10 µm2 (called “TEM lamellae”) were thinned down by FIB to
about 100-150 nm of thickness to achieve theBoehm-Courjault et al. needed for TEM analyses. Fig. 4 shows an
electron transparency
example of a window thinned by FIB for the sample 175-66h.

Cu
half-ring
sample

200 µm FIB sample-holder

Figure 3:Fig.
Sample geometry
3: Sample usedused
geometry for FIB thinning.
for FIB thinning.

also carried out in this mode. For all the TEM anal-
yses, a FEI Tecnai Osiris microscope was used, oper-
ating at 80 kV and at low current, these parameters
aiming at not degrading the sample and preserv-
ing the C-S-H microstructure. The EDX detector,
provided by Bruker, is a Nano XFlash detector.
Preparation of the TEM samples by focused ion
30 µm
beam (FIB)
Fig. 4: SEM images of a TEM lamella of about 5*10 µm thinned down by FIB for the sample 175-66h, showing
the electron transparency of the main part of the lamella, variableinwithin each
Pastes (crushed as powders) were first mixed with
a G2 hard resin. This mixture was centrifuged at
15000 rpm during 15 minutes in order to have a
good density of particles and only a small part of
resin. After polymerisation of the resin at 80 C
for 12 hours, a slice of about 700 microns thick
and 4 mm diameter was cut. This slice was cut as
a semi-disc and was then polished on both sides
until reaching a thickness of 30 to 40 microns. It
was then glued on a copper half-ring of 3 mm of
outer diameter and 1.5 mm of inner diameter, to
ensure mechanical resistance while handling the
sample. Figure 3 shows the geometry of the sample
before thinning by FIB.
In each sample, 2 or 3 windows of 5*10 µm2
(called "TEM lamellae") were thinned down by FIB
to about 100-150 nm of thickness to achieve the
electron transparency needed for TEM analyses.
Figure 4 shows an example of a window thinned
by FIB for the sample 175-66h.
III. Results and Discussion
resin
Morphology
C-S-H of CC-S-H
3S at different stages
Figure 5 shows images of the surface of C3 S grains
slightly covered by C-S-H for the 4 isopropanol re-
placements. They were taken during the induction
period, as the formation of C-S-H had just started,
which is why the surface of the grains is not com-
pletely covered. The morphology of C-S-H appears
to be fibrillar,
2 µm as it was already shown by (Bazzoni,
2014). No difference in morphology between the
2 samples can be seen. The amount of C-S-H fibrils
is veryespecially
C-S-H and sample,
resin areas. depending on
the examined C3 S grain.
The length of the fibrils was measured for each
sample: 25 to 30 measurements were taken on dif-
ferent grains. The precision of measurements is
4
affected by the angle between the C-S-H fibril and
the surface of the grain, considering if the fibril
is in the plane of the image or not. But as the
used method is the same for the 4 samples, the val-
ues can easily be compared. The obtained values,
summarised in Table 3 are similar for the 4 pastes,
showing that C-S-H seems to grow the same way
despite the presence of isopropanol. The morphol-
ogy of C-S-H cannot explain the kinetics difference
for C3 S hydrated with or without isopropanol.
Figure 6 shows images of the surface of C3 S
grains covered by C-S-H taken at the maximum
heat flow for the 4 isopropanol replacements. The
morphology of C-S-H is still fibrillar and like for
the induction period it is similar for the 4 pastes.
The surface of all C3 S grains is completely covered,

Figure 4: SEM images of a TEM lamella of about 5*10 µm2 thinned down by FIB for the sample 175-66h, showing
the electron transparency of the main part of the lamella, especially in C-S-H and resin areas.

282
 Fig. 5 shows images of the surface of C3S grains slightly covered by C-S-H for the 4 isopropanol
replacements. They were taken during the induction period, as the formation of C-S-H had just started, which is
why the surface of the grains is not completely covered. The morphology of C-S-H appears to be fibrillar, as it
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
was already shown by (Bazzoni, 2014). No difference in morphology between the samples can be seen. The
amount of C-S-H fibrils is very variable within each sample, depending on the examined C3S grain.

Surface of
C3S grain

C-S-H
fibrils

(a) (b)

(c) (d)

Fig. 5: HR-SEM images of the samples stopped in the induction period: (a) 0-1h; (b) 75-1h10; (c) 175-1h20; (d) 250-1h45.
Figure 5: HR-SEM images of the samples stopped in the induction period: (a) 0-1h; (b) 75-1h10; (c) 175-1h20; (d)
250-1h45. The length of the fibrils was measured for each sample: 25 to 30 measurements were taken on
different grains. The precision of measurements is affected by the angle between the C-S-H fibril and the surface
of the grain, considering if the fibril is in the plane of the image or not. But as the used method is the same for
which is consistent with
the 4 samples, the the
valuestheory of be
can easily (Bazzoni,
compared. Thenm, which
obtained is consistent
values, summarised inwith
Tablethe values
3 are similar given
for the in
4 pastes,toshowing
2014). According that C-S-H
this author, the seems to grow
surface of CtheS sameTable
way despite
3. OntheFig
presence
7c of isopropanol.
and 7d, The morphology
impingement of the
3
of C-S-H cannot explain the kinetics difference for C3S hydrated with or without isopropanol.
grains is progressively covered by hydrates during fibrils growing from different C3 S grains and not
the acceleration period and
Tableit3:isAverage
completely
length ofcovered inmeasured
the C-S-H fibrils the same direction
in pastes within the can be period.
induction seen. It seems that
at the moment of the maximum Sampleheat flow. Such a 0-1h
label the densification
75-1h10 of 175-1h20
the C-S-H had just started, but
250-1h45
phenomenon is also Average
observed with
length a high
of the fibrilscontent
(nm) 242there
± 69 is still
214 ±some
76 space
228 ± 65between 233 C-S-H
± 69 and C3 S
of isopropanol, even if it is delayed in time. grains. The formation of inner-product had not
Figure 7 shows STEM images of the 0-6h30 sam- started yet, which is still in accordance with the
ple, in BF and HAADF modes. C-S-H fibrils can be theory of (Bazzoni, 2014).
observed. The length of the fibrils is about 150-200 STEM images of the 175-66h and the 250-144h

Figure 6: HR-SEM images of the samples stopped at the acceleration peak: (a) 0-6h30; (b) 75-7h30; (c) 175-9h30; (d)
250-16h.

283
 Boehm-Courjault et al.

Table 3: Average length of the C-S-H fibrils measured in pastes within the induction period.

Sample label 0-1h 75-1h10 175-1h20 250-1h45

Average length of the fibrils (nm) 242±69 214±76 228±65 233±69

samples are shown in Figure 8. Fibrils of C-S-H can
clearly be seen. Their length is about 200 nm, which
is also consistent with the values given in Table 3.
In images (c) and (d), two different morphologies
can be distinguished: near the C3 S grains, C-S-H
appears to be denser than at the surface. Accord-
ing to the theory of Bazzoni, 2014), as images were
taken in the deceleration period, C-S-H fibrils can-
not form anymore as the surface of C3 S grains was
already completely covered. Dense inner-product
had thus started to form.
Ca/Si ratios obtained by the different
characterisation techniques
The Ca/Si ratios obtained for different samples
and with different characterisation techniques are
summarised in Table 4. The values determined by
TGA are all in the same range, around 1.7 - 1.8.
For the samples stopped within the deceleration
period (0-14h, 75-30h,175-66h, 250-144h), the results
obtained by TGA and SEM-EDX are comparable,
E. Boehm-Courjault et al.
considering the precision of the techniques. The
values measured by SEM-EDX are slightly higher,
probably due to the fact that the analyses were car-
ried out on a polished surface in supposedly C-S-H
areas without knowing if there is also C-S-H be-
low (the interaction volume at 15 kV is approx. 2-3
µm). Thus Ca and Si contents may be under- or
overestimated if what is really measured is the com-
position of a mixture of C-S-H with C3 S and/or CH,
which could explain the high standard deviation
on SEM-EDX measurements.
This intermixing of phases could normally be
avoided by using TEM, as the precise area of the
EDX measurement can be much more controlled
(due to the sample thickness). Nevertheless the
Ca/Si ratios obtained by TEM-EDX for the 175-
66h and 250-144h samples are even higher than the
SEM-EDX ones. At the moment, the authors have
no explanation about this fact. Further tests are still
in progress.
Finally it has to be highlighted that the pres-
ence of isopropanol seems to have no effect on the
composition
of C-S-H. The DoH seems to have no
influence on the Ca/Si ratio either.

resin

C3S
C-S-H

C3S

C3S

(a) (b)

(c) (d)

Fig. 7: STEM images of the 0-6h30 sample: (a) and (c) in BF mode; (b) and (d) in HAADF mode.
Figure 7: STEM images of the 0-6h30 sample: (a) and (c) in BF mode; (b) and (d) in HAADF mode.
Ca/Si ratios obtained by the different characterisation techniques

The Ca/Si ratios obtained for different samples and with different characterisation techniques are
284
summarised in Table 4. The values determined by TGA are all in the same range, around 1.7 - 1.8.
For the samples stopped within the deceleration period (0-14h, 75-30h, 175-66h, 250-144h), the
results obtained by TGA and SEM-EDX are comparable, considering the precision of the techniques. The values
 15th Euroseminar on Microscopy Applied to Building
E. Boehm-Courjault Materials • 17-19 June 2015 • Delft, The Netherlands
et al.

C3S
CH
C3S

CH

resin

C-S-H C-S-H
(inner-product)

C3S
C3S
CH
CH
C-S-H
(outer-product;
fibrils)
(a) (b)

C3 S

C3S

C3 S
C3S

(c) (d)

Fig. 8: STEM images (BF mode) of: (a) and (b) the 175-66h sample; (c) and (d) the 250-144h sample.
Figure 8: STEM images (BF mode) of: (a) and (b) the 175-66h sample; (c) and (d) the 250-144h sample.
Table 4: Comparison of the values of the Ca/Si ratios measured by different techniques for different samples.
Sample label DoH (%) Ca/Si (TGA) Ca/Si (SEM-EDX) Ca/Si (TEM-EDX)
0-6h30 14 n.m.
tionn.m. ± 0.05 to grow until the surface
IV. Conclusion
75-7h30 11 1.70 ± 0.10
period.
n.m.
They 1.63
continue
n.m.
175-9h30 7 1.79 ± 0.22 of then.m. grains is completely
n.m. covered, which corre-
In this paper, the250-16h
morphology and7 composition
1.82 ± 0.24 of spondsn.m. to the maximumn.m. heat flow of the calorime-
0-14h
C-S-H were investigated 1.70C± 0.10
35 pastes of
in four 1.78 ± 0.38 n.m.
3 S hy- try curve. Then a deceleration of the kinetics of the
75-30h 35 1.66 ± 0.10 2.02 ± 0.23 n.m.
drated with different
175-66h mixtures
35 of water and iso-
1.75 ± 0.10 hydration
1.97 ± 0.29 occurs,2.14
corresponding
± 0.08 to the formation
propanol. They were observed
250-144h 35 at characteristic
1.70 ± 0.10 the± 0.35
of1.84 1.99 ±This
inner-product. 0.17 mechanism seems not to
n.m.: not measured
times of the hydration: within the induction pe- be changed by the presence of isopropanol. Only
riod, at the maximum heat flow and at a degree of the time scale is different.
hydration of 35 % (deceleration period). It has been
shown that:
The measured Ca/Si ratios are consistent with
The presence of isopropanol changes the kinetics
the generally observed values.
of the hydration: it is slowed down, but the fibrillar
morphology and the composition of the hydrates
are not changed. The length of the fibrils is about The Ca/Si ratios obtained by SEM-EDX are
200 nm at every stage of the hydration. higher than the ones calculated from the TGA tech-
8
The following hydration mechanism is observed, nique. The results measured by TEM-EDX seem
as it has already been shown by (Bazzoni, 2014): to be even higher, but this has to be confirmed by
fibrils start to grow on C3 S grains within the induc- further investigations.

Table 4: Comparison of the values of the Ca/Si ratios measured by different techniques for different samples.

Sample label DoH (%) Ca/Si (TGA) Ca/Si (SEM-EDX) Ca/Si (TEM-EDX)
0-6h30 14 n.m. n.m. 1.63 0.05
75-7h30 11 1.70 0.10 n.m. n.m.
175-9h30 7 1.79 0.22 n.m. n.m.
250-16h 7 1.82 0.24 n.m. n.m.
0-14h 35 1.70 0.10 1.78 0.38 n.m.
75-30h 35 1.66 0.10 2.02 0.23 n.m.
175-66h 35 1.75 0.10 1.97 0.29 2.14 0.08
250-144h 35 1.70 0.10 1.84 0.35 1.99 0.17
n.m.: not measured

285
 Boehm-Courjault et al.
Acknowledgement
The authors gratefully acknowledge the Centre of
Microscopy of EPFL (Centre Interdisciplinaire de
Microscopie Electronique, CIME) for providing fa-
cilities and support for FIB and TEM, especially Dr
Marco Cantoni and Mrs Fabienne Bobard.
286
References
Bazzoni, 2014: EPFL Thesis n 6296, available online on:
http://infoscience.epfl.ch/record/200217/files/EPFL_TH6296.pdf
Makar and Sato, 2011: Makar J.M. and Sato T., Materials and Structures
(2013) 46, 1-12.
Richardson, 1999: Richardson, I.G., Cement and Concrete Research
(1999) 38, 137-158.
Taylor, 1997: H.F.W. Taylor, Cement chemistry, Thomas Belford Pub-
lishing (1997).
Zhang and Sherer, 2013: Zhang J. and Sherer G.W., Cement and Con-
crete Research (2011) 41, 1024-1036.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

The interaction of pH, pore solution

composition and solid phase composition
of carbonated blast furnace slag cement
paste activated with aqueous sodium
monofluorophosphate
Josepha Kempl⇤ , Oğuzhan Çopuroğlu
Delft University of Technology, Materials & Environment (CiTG), Delft, The Netherlands
⇤ j.kempl@tudelft.nl

Abstract

Blast Furnace Slag (BFS) is a waste product of industrial steel production and a common additive in the cement
industry in Northern European countries. However, cementitious materials made from slag-rich cement, particularly
CEM III /B, are very susceptible to carbonation. Recent investigations have shown that the surface application of aqueous
sodium monofluorophosphate (Na-MFP) as pre- and post-carbonation treatment can improve the surface durability of
cementitious materials with a high BFS content. Significant improvements have been observed in the micro-mechanical
characteristics of concrete surface and frost salt scaling resistance.
On the basis of previous studies we are investigating self-healing of blast furnace slag cement (BFSC) products
treated with the inorganic self-healing agent Na-MFP from a mineralogical point of view. In this study we combine
results of pore solution pH analyses and main element composition under the influence of Na-MFP with the presence of
crystalline phases found in the solid aliquot of the samples. Pore solutions were investigated by inductively coupled
optical emission spectrometry (ICP-OES). Solid-material investigation was performed by X-ray powder diffractometry,
including Rietveld quantitative phase analyses.
Our results show that the element concentration and the pH of the paste pore solutions have direct influence on the
formation of crystalline and amorphous phases forming in the solid sample aliquot during carbonation and self-healing
by Na-MFP. In this work we focus especially on the influence of sulfur in solution and the formation of ettringite. In
addition we discuss, why the formation of the crystalline phosphate apatite does not occur in cementitious products after
Na-MFP treatment.

Keywords: blast furnace slag cement, sodium monofluororphosphate, X-ray diffractometry, amorphous calcium phos-
phates

I. Introduction tion to keep BFSC equally attractive for the building

Sodium monofluorophosphate, also known as Na-
MFP (chemically Na2 PO3 F) has previously been
used as an inorganic corrosion inhibitor for steel re-
inforced concrete structures (e.g. Alonso et al. 1996).
Over the past decade, Na-MFP has gained interest
as an inorganic self-healing agent on cementitious
products rich in blast furnace slag.
BFSC is an important product of the cement in-
dustry in Northern European countries. In the
Netherlands it holds a market share of about 60%
and besides many technical advantages it is an
environmentally friendly product. However, its
carbonation rate is a huge drawback compared to
ordinary Portland cement (OPC) performances and
requires a large-scale industrial and feasible solu-
industry. Taylor (1997) explained the mechanisms
of the poor carbonation resistance by shrinkage of
the cement matrix, which occurs during the car-
bonation of a hydrated BFSC product. Due to the
loss of chemically-bound water during carbonation,
the matrix shrinkage induces micro-cracking. Pro-
gressing attack through environmental impact and
seasonal temperature variations lead to large-scale
surface damage in BFSC products and influences
their life-time durability significantly.
In order to avoid this specific damage, recent
experimental studies revealed the recovering effect
of Na-MFP on the microstructure of carbonated
BFSC pastes with respect to their frost-salt scaling
durability (Çopuroğlu et al. 2006; Sisomphon et al.
2009; 2010a, 2010b). On the basis of preliminary

287
 Kempl & Çopuroğlu
results, further research about the compatibility of
Na-MFP for the self-healing materials concept is
developed at present and adjusted to the specific
slag-bearing cement type CEM III /B.
In this experimental study we are investigating
the influence of cementitious pore solutions to the
formation of crystalline and amorphous phases.
Blast furnace slag cement pastes, particularly made
of CEM III /B, were impregnated with a 45 wt%
aqueous solution of Na-MFP under vacuum condi-
tions. Pore solutions were extracted with a high-
pressure apparatus previously described by Barney-
back & Diamond (1981) or Byfors et al. (1986). Sub-
sequently pore solutions and paste powders were
prepared for OES-ICP and X-ray powder diffraction
(XRPD).
II. Experiments
Cement paste specimens were prepared from CEM
III /B (42.5N HSR LH) with a BFS content of 67
wt%; and for reference purposes from CEM I (32.5
R). Both cement types were produced by the Dutch
cement manufacturer ENCI and processed accord-
ing to previous experimental studies (Sisomphon
et al. 2009, 2010a, 2010b). The water-cement-ratio
(W/C) was chosen to be 0.45. For paste preparation
common tap water was used for mixing in a com-
mercial Hobart mixer with one minute low-speed
and one minute high-speed mixing. Specimens
were cast in commercial polyethylene jars. A de-
tailed working scheme is shown in Figure 1. All
pastes were cured in the fog room at 100% relative
humidity (RH) for 28 days after demolding from
the polyethylene jars after 24 hours. Subsequently
the samples were sawn in slices of 1 cm thickness in
order to ensure complete chemical reaction during
the different cycles of chemical treatment.
For carbonation sample aliquots were exposed
to a 3% CO2 and a 75% RH atmosphere. For the
Na-MFP treatment solutions of Na-MFP were pre-
pared from tap water with 45 wt% Na2 FPO3 . The
paste samples were vacuum impregnated with the
respective solution either before or after carbona-
tion. Chemical reaction times of 28 days were given
to the specimens after Na-MFP treatment or carbon-
ation, respectively. Between the treatments samples
were stored in an oxygen free atmosphere at room
temperature conditions in a 55% RH desiccator.
For every chemical treatment a reference sample
was stored in the desiccator during the whole study.
This results in reference sample materials of an
ordinarily hydrated CEM III /B paste, a carbonated
CEM III /B paste and a CEM III /B paste vacuum
impregnated with the Na-MFP self-healing agent.
The hydrated CEM I paste specimen is added to
the present dataset for analytical comparison of the
X-ray refinements. A sample list with specific steps
288
of chemical treatment is given in Table 1.
Pore solutions of all samples were gained by the
application of a high-pressure apparatus and a pres-
sure load of up to 500 tons for each paste specimen.
The TU Delft CiTG block tester (MacBen-type) with
an oil-hydraulic end-load capacity of 5000 kN was
used to insert a steel-cylinder assembly with an
inner diameter of 34 mm and a height of 115 mm
exactly fitting the size of the paste cylinders pre-
pared in the polyethylene jars. The steel cylinder
assembly was build in-house from the noble-steel
type 34CrNiMo6 with a maximum tensile strength
of 1400 N/mm2 to avoid high-pressure phase tran-
sitions of the steel during exposure of the material
to high pressures. The experimental set up was
adapted from Barneyback & Diamond (1981).
From all paste samples a powder aliquot of five
gram was taken and finely ground under ethanol in
an agate mortar down to a finish of 5 µm. Samples
were prepared as dry powders for X-ray powder
diffraction on an aluminum sample holder.
III. Analyses
In this study the pH was analyzed with a “827 pH
lab” pH meter of the Metronohm AG, Switzerland.
The pH meter was calibrated with Metronohm pH
buffer solutions of pH 4 and 9 at 25 C prior to the
analyses.
The main element concentrations of the pore so-
lutions were analyzed with a VARIAN Vista 720
ICP-OES. The instrument was calibrated with a 1%
v/v HNO3 blank solution and differently concen-
trated solutions of the commercial multi-element
standard for main elements, trace elements and
rare earth elements (111355 Merck IV). All samples,
the standard and the procedural blanks were ana-
lyzed in 1% v/v HNO3 . The detection limit of the
analyzed elements was below or equal to 200 ppb.
XRPD was performed on a Philips PW 1830/40
powder diffractometer in the micromechanical labo-
ratory (Microlab) of Delft University of Technology
(TUD). The machine is operated with an acceler-
ating voltage of 40 kV and an X-ray beam current
of 30 mA. The X-ray source is a Cu-tube working
with characteristic CuKa wavelength of 0.15418 nm.
The machine is equipped with a Ni-filter. Analyses
were performed in air on hydrated, carbonated and
on Na-MFP treated and finely ground paste pow-
ders with a stepsize of 0.02 for a 2q angle range
between 5 and 70 . Every step was analysed for
10 seconds.
A total number of 3099 data points per sam-
ple could be used for quantitative phase analy-
ses performed by the Rietveld-refinement method
with the GSAS software package. For the Rietveld-
refinement an eight-term Chebeyshev polynomial
model was used. Structure models for the identi-
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 1: Schematic of working sequence and experimental treatments on cement pastes of CEM I and CEM III /B.

feid phases were taken from the Crystallography
Open Database and are listed with their database
number and reference in Table 2. For the iden-
tification of the C(A)SH phases, several different
C(A)SH polymorphs (hillebrandite, afwillite, 14-Å-
tobermorite and hydrogarnet) were used for the
refinement. The polymorphs vary in their Ca:Si
ratio or their water concentration, or both. C(A)SH
polymorphs were primarily chosen on the basis of
their peak occurrence. Due to their partly poor
crystallinity and often low intensity, and due to
their overlapping diffraction bands in the XRD pat-
terns the C(A)SH phases were secondarily chosen
after their Ca:Si ratio and their water concentrations
as Ca and Si are to different amounts available
for C(A)SH formation in each cement type. The
refined phase specific parameters of the distinct
crystalline and major phases were the lattice pa-
rameters, the peak profile parameters and the scale
factors. Due to the high amount of amorphous
phases in the BFS pastes, but also to the high band
of bremstrahlung in the diffractograms between 5
and 20 2q (Figure 3) the occupancy factors of the
ions in the solid solution series were not refined. In
order to model the peak shape and the variation
of the peak shift for distinct and major criystalline
phases the pseudo-voigt function was chosen. The
pseudo-voigt Lorentzian-Gaussian mixing parame-
ter was refined and peak assymetry correction was
also performed, again only for distinct and ma-
jor crystalline phases. Finally, spherical harmonic
preferential orientation corrections were applied
for cylindric and platy mineral phases (portlandite,
ettringite, C3 S and partially C2 S). To avoid exces-
sive deviations, constraints were put on the lattice
parameters and the profile fitting parameters with
a 2% allowance for the deviation from literature
values for the lattice parameters.
IV. Results and Discussion
Analyses of pH and main elements
The pH of all cement paste pore solutions is listed
in Table 3. As expected, the hydrated cement pastes
have a very basic character with a pH of 13.1 for
CEM I and 13.0 for CEM III /B. The carbonated
CEM III /B paste pore solution has a pH of 8.8.
These results are in good agreement with the com-
monly analyzed pH values of ⇠13 for hydrated ce-
ment pastes and 8-9 for carbonated cement pastes
(e.g. Taylor, 1997). If an aqueous solution of 45
wt% Na-MFP is added to the CEM III /B paste via
vacuum impregnation, the pH of the pore solution
is increasing to a pH of 13.1 (Table 3). In the frame-
work of corrosion inhibition, especially Na has been

Table 1: Sample list showing sample ID, chemical treatment and sample age of each specimen.

sample ID treatment sample age [d]

CEM I-01 hydrated CEM I 30
CEM III-01 hydrated CEM III /B 30
CEM III-02 hydrated & carbonated CEM III/B 58
CEM III-06 hydrated & 45% Na-MFP impregnated CEM III /B 58
CEM III-08 hydrated, carbonated & 45% Na-MFP impregnated CEM III /B 86

289
 Kempl & Çopuroğlu

Table 2: Phase names, mineralogical formulas, abbreviations, cement chemistry notation (CCN) and the crystal-
lographical open database (COD) reference number as well as the literature references of the phases used during
Rietveld-refinements.

phase formula abbrev. CCN COD no. reference

cement paste phases
portlandite Ca(OH)2 po CH 1008780 Busing & Levy (1957)
alite Ca3 SiO5 C3 S 9016125 De la Torre et al. (2008)
belite Ca2 SiO4 C2 S 9014595 Udagawa et al. (1980)
ettringite Ca6 Al(SO3 )4 (OH)12 ·26H2 O et C6 AS̄3 H32 9015084 G.-Neunhoffer & Neubauer (2006)

melilite (Ca,Na)2 (Al,Mg,Fe2+ )[(Al,Si)SiO7 ] me 9000055 Smith (1953)

aragonite arg 9000226 De Villiers (1971)

calcite CaCO3 cc CC̄ 9000095 Graaf (1961)
vaterite vat 9007475 Wang & Becker (2009)

C-(A)-S-H phases
afwillite Ca3 [SiO3 OH]2 ·2H2 O C3 SH5 9007428 Megaw (1952)
hillebrandite Ca6 [Si3 O9 ](OH)6 C6 S 3 H 9001698 Dai & Post (1995)
hydrogarnet Ca3 Al2 [SiO4 )3 x (OH)4x ] C3 AH6 1007235 Cohen-Addat et al. (1964)
tobermorite (14Å) Ca5 [Si6 (O,OH)18 ]·5H2 O C5 S6 H28 9013974 Bonaaccorsi & Merlino (2005)

shown to be responsible for the increase of the pH S4+ .

(e.g. Jin et al. 1991). The CEM III /B paste that
was first carbonated and subsequently impregnated
with the Na-MFP solution has a pH of 11.7. This
shows that the treatment of a slag-rich carbonated
cement paste with a 45 wt% solution of Na-MFP
has the potential to recover the pH of a hydrated
CEM III /B with up to 90% in a chemical reaction
time of 28 days.
The results of the main element analyses are also
listed in Table 3. In order to illustrate the percent-
age of main element concentrations in the slag-rich
cement pastes, the relative main element concen-
trations are shown in pie diagrams in Figure 2. In
the pore solution of a hydrated CEM I (CEM I-01)
paste the alkali elements Na+ and K+ make up the
highest concentration with 44.6wt.% and 52.2wt.%,
respectively and stabilize the very basic pH, while
Mg2+ , Ca2+ , S4+ and the trace elements (Li, Ba, Sr,
Fe, Al and Si) comprise a total of 3.2%. The high
amount of alkali elements has previously also been
observed by Song & Jennings (1999) and Brouwers
& van Eijk (2003). A similar result was analysed
in the hydrated CEM III /B (CEM III-01) paste in
which Na+ makes up about 48.5% and K+ makes
up 46.3% of the main elements in the pore solu-
tion. The remaining 5.2% consist of Mg2+ , Ca2+ ,
S4+ and the trace elements. In CEM I pastes as well
as in CEM III /B pastes the high pH is controlled
by the high concentrations of alkali elements (Song
& Jennings, 1999; Brouwers & van Eijk, 2003). The
increase of S4+ in the pore solution of a slag-rich
cement paste compared to the pore solution of a
hydrated CEM I paste can be explained by the pres-
ence of BFS containing an additional amount of
The main element concentrations in a slag-rich
cement paste change drastically after carbonation
(CEM III-02); especially the concentration of S4+ is
increasing remarkably with up to 53.6% in the pore
solution. Also Ca2+ (28.2%) and Mg2+ (13.1%) are
increasing in the pore solution of a carbonated CEM
III /B paste, while Na+ , K+ and the trace elements
can be summed up to 5.1%. If a hydrated CEM III
/B paste is impregnated with a 45 wt% solution of
Na-MFP (CEM III-06), an increase of Na+ (65.1%)
must be obvious. Relative to Na+ , K+ makes up
about 23.7%, while Mg2+ , Ca2+ , S4+ and the trace
elements result in a total percentage of 11.3% and
seem to be getting activated by Na-MFP. Again, ex-
pecially S4+ is remarkably increasing in the pore
solution. Finally, the Na+ concentration in a carbon-
ated CEM III /B paste impregnated with Na-MFP
solution (CEM III-08) increases up to 57.1%. While
K+ , Mg2+ , Ca2+ and the trace elements make up
2.4%, S4+ is again present as a major element with
up to 39.8%.
The relatively high concentrations of the main
elements in the two CEM III /B cement pastes (Ta-
ble 3) treated with Na-MFP solutions are mainly
caused by the Al3+ and Fe3+/2+ , likely being acti-
vated with the addition of Na-MFP.
X-ray powder diffraction and quantitative phase
analyses by Rietveld-refinement
Prior to the quantitative phase analyses occurring
crystalline phases were identified. The x-ray diffrac-
tograms of all samples are shown in Figure 3. The
diffraction pattern of hydrated CEM I paste was

290
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 2: Main element concentrations of slag-rich cement paste pore solutions in percent; upper left: hydrated CEM
III /B; upper right: hydrated and carbonated CEM III /B; lower left: Na-MFP impregnated CEM III /B and lower
right: carbonated and Na-MFP impregnated CEM III /B.

primarily added for reference purposes. Distinct
peaks of crystalline phases such as portlandite, et-
tringite and alite could clearly be identified and
were used for further preliminary peak identifica-
tion in the slag-rich cement paste diffractograms.
Less distinct peaks of belite could also be found,
but were partly overlapped by alite (Taylor, 1997),
which can form solid solutions with Mg2+ , Al3 +,
P4+ , Fe2+ and Si4+ (Nishi et al. 1984). Due to
their poor crystallinity, but also to their complex
solid solution formation with alkali elements and
the occurrence of numerous polymorphs and vary-
ing water concentrations, the C(A)SH phases could
not clearly be identified. The primarily amorphous
character of these phases also causes a relatively
large background and a diffuse diffraction band
between 25 and 40 2q. However the refinement
was performed with afwillite and 14-Å tobermorite,
of which afwillite was found to be the major CSH-
phase, while tobermorite was quantified with an
amount of less than 2 wt%, which is below the ac-
tual detection limit of the method. The results of
the quantitative analyses are listed in Table 4.
X-ray diffraction patterns of slag-rich cement
pastes generally show a diffuse, asymmetric band
from the amorphous blast furnace slag ranging
from 20 to 48 2q and peaking at about 31 2q
(Regourd, 1986). In addition the diffuse band of the
poorly cristalline to amorphous CSH phases is un-
derlying the amorphous BFS band in the diffraction
pattern in Figure 3. This even increases the back-
ground of the diffraction patterns for the slag-rich
cement pastes.
However in a hydrated CEM III /B (CEM III-01)

Table 3: pH values and main element concentrations of the pore solutions gained from high-pressure pore solution
expression. The trace elements are a total of Li, Ba, Sr, Fe, Al, and Si.

sample ID Elements dissolved in pore solution [mg/l]

pH Na K Mg Ca S trace elements
CEM I-01 13.1 2277.1 2662.9 0.5 110.9 15.6 35.9
CEM III-01 13.0 1630.9 1557.1 0.5 83 54.4 39.8
CEM III-02 8.8 229.3 93.8 877.9 1881.3 3582.6 18.7
CEM III-06 13.1 5069.3 1841.7 4.36 74.43 576.81 220.9
CEM III-08 11.7 24124.3 397.5 12.23 177.3 17100.7 430.4

291
 Kempl & Çopuroğlu

paste distinct peaks of ettringite and portlandite
could clearly be identified. Wider peaks of alite
(31 -33 2q), and a diffuse mixture of portlandite
and ettringite bands (27 -29 2q) could be observed
and were used for the quantitative phase analyses.
The amount of portlandite quantified in CEM III /B
is approximately three times lower than the amount
of portlandite in the comparable CEM I paste (Ta-
ble 4). This is in good agreement with the starting
composition of the cement types: blast furnace slag
cement contains about 67 wt% BFS and therefore
roughly one-third less clinker. Consequently, only
about one-third of CaO can undergo hydration dur-
ing paste manufacturing and curing in CEM III /B.
The analyses of the slag-rich cement paste resulted
in a relatively high concentration of alite (71.3 wt%).
Though a higher amount of alite could be expected
in slag-rich cement pastes, due to the higher con-
centration of Si in the BFS, this value is unusally
high. It might be caused during the refinement
as a consequence of overlapping peaks with belite
or the large peak width on top of the amorphous
CSH fraction of the sample. The latter has likely
happened since it is visible in the relatively low
amount of quantified crystalline CSH phases (Table
4). CEM III /B has a much larger chemical variabil-
ity due to the addition of BFS, which generally has
a melilitic composition. As a result the formation of
portlandite solid solition e.g. with the isostructural
brucite (mineralogically Mg(OH)2 ) and additional
dipositive ions would be expected (Taylor, 1997 and
references therein). In addition, the poorly crys-
talline CSH phases can substitute a couple of ions.
In a gel or amorphous phase these ions can even
multiple-positively be charged. In general this in-
cludes minor amounts of Al3+ , Fe3+/2+ , Al3+ , Si4+ ,
K+ , Na+ , but it can also comprise ions such as P4+
(Taylor, 1997). The non-stoichiometric amorphous
amount of phases existing in the samples surely
tamper with the phase quantification.
With the addition of aqueous Na-MFP to a hy-
drated slag-rich cement paste (CEM III-06) the et-
tringite peaks show a slightly higher intensity com-
pared to the hydrated CEM III /B paste. The quan-
titative analyses resulted in 15 (±2) wt% ettrinigite
compared to the hydrated slag-rich cement paste.
The amount of portlandite in CEM III-06 is given
with 9 wt%, which scatters within the common
standard deviation of ±2 wt% for each clearly crys-
talline quantified phase. Compared to the hydrated
slag cement paste the amount of portlandite is not
significantly varying. C(A)SH phases chosen for

Figure 3: X-ray diffrcation patterns of the cement pastes investigated in this study; from top to bottom CEM I-01,
CEM III-01, CEM III-02, CEM III-06 and CEM III-08. Some crystalline phases, such as portlandite, ettrinigite, alite,
vaterite, calcite and aragonite show distinct peak positions, whereas poorly cristalline phases, amorphous phases or
phases with a lower quantity cause a rather high background or broad peaks, respectively.(see Table 2 for mineral
abbreviations).

292
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 4: Results of the quantitative phase analyses by Rietveld-refinement. The standard deviation for the amount of
distinct crystalline phases is ±2 wt% in a refinement with about 100% of a crystalline sample.

sample ID ! CEMI-01 CEM III-01 CEM III-02 CEM III-06 CEM III-08
sample treatment ! hyd hyd hyd, cc hyd, Na-MFP imp hyd, cc, Na-MFP imp
Statistical parameters of the Rietveld refinement
RP 0.023 0.032 0.03 0.025 0.024
RWP 0.039 0.043 0.042 0.043 0.041
DWD 0.414 0.507 0.657 0.7 0.593
c2 5.696 7.948 4.389 5.478 4.798

Results of the quantitative phase analyses of the crystalline phases in wt%

Portlandite 32 10 3 9 7
Alite 23 71 22 18 18
Belite 3 6 7
Ettringite 13 13 7 15 15
Melilite 2
Aragonite 18
Calcite 19 13
Vaterite 25 10
crystalline C-(A)-S-H 29 2 51 36

the refinement were hillebrandite (17 wt%), 14-Å impregnated with aqueous Na-MFP. The diffraction
tobermorite (26 wt%) and hydrogarnet (8 wt%). pattern shows a recovery of portlandite from 3 (±2)
These phases were chosen due to there smaller wt% in the carboanted cement paste to 7 (±2) wt%
Ca:Si ratio in comparison to afwillite and due to in the sample that was first carbonated and subse-
their better compatibility in the course of the re- quently impregnated with Na-MFP. It also shows
finement. Again the C(A)SH phases can substitute a decent re-occurence of the ettringite peaks, quan-
numerous positively charged ions in the form of a tified with about 15 (±2) wt% of the crystalline
non-stoichiometric gel. fraction of the samples and a re-occurrence of the
diffuse diffraction band of the CSH phases (36 (±2)
After carbonation of a hydrated CEM III /B paste
wt%) after the recovery of the pH up to 11.7 (Table
(CEM III-02), the distinct peaks of portlandite dis-
3). The dominant CSH phase chosen for the re-
appear almost completely (Figure 3). The amount
finement was 14-Å tobermorite. The clinker crystal
of portlandite quantified in this sample is 3 (±2)
alite seems to be stable during the whole carbon-
wt% (Table 4). Also the intensity of the ettringite
ation and Na-MFP impregnation process within
peaks is clearly decreased after carbonation. The
the investigated sample age, while belite could not
quantified crystalline amount in this sample is 7
be identified in this sample. Instead a decent to-
(±2) wt%. Instead of distinct ettringite and port-
tal percentage of carbonates of 23 (±2) wt% is still
landite peaks, large amounts of vaterite, aragonite
present.
and calcite were identified (Table 4) amongst the
crystalline fraction of the sample. This is in good
agreement with the observations by Brocken & Nij- Rietveld-refinement and its challenges to cement
land (2004), who assume the formation of vaterite product analyses
after the interaction of carbon dioxide with ettrin-
The Rietveld-refinement method generally allows
gite during carbonation. Crystalline CSH phases
quantitative phase anlayses of crystalline materials.
have not been detected and the diffuse band of CSH
Thanks to the development of high-performance
phases seems to be decreased in the diffraction pat-
X-ray powder diffractometers and detectors as well
tern of CEM III-02. Greenberg & Chang (1965)
as stable-functioning software, improved sample
found CSH phases to be instable at pH values be-
preparation techniques and analyses, the Rietveld
low 9.5. This finding would be in good agreement
refinement has become a common tool in the ce-
with the absence of CSH phases in the diffraction
ment, cement clinker and cement product charac-
pattern and the analysed pH of 8.8 (Table 3). Alite
terization. However, a Rietveld refinement has not
and belite do not seem to be significantly influenced
become a trivial analysis. Specific issues of quanti-
by the rather short time period of hydration and
fying hydrated cement phases are the occurrence
carbonation given to the sample in this study.
of numerous poorly crystalline or even amorphous
CEM III-08 was first carbonated and subsequently phases such as CSH phases. Some of them are

293
 Kempl & Çopuroğlu
only present in minor amounts of less than 2 wt%;
which is just the detection limit for a crystalline
phase in a good quality refinement. Additionally
the occurrence of overlapping diffraction peaks, as
well as the common paragenesis of solid solution se-
ries and polymorphs of occurring hydrated cement
phases as well as the occurrence of other amor-
phous supplementary cementitious materials such
as BFS, influence the statistical quality parameters
of a refinement on cementitious composites.
In the refinement presented in this study the
DWD and c2 of each analyses did not reach the
necessary values to speak about a 90% certainty for
phase quantification (Table 4). This is mainly due to
the amounts of amorphous BFS and the formation
of poorly crystalline C(A)SH phases, increasing the
background of the analyses. In addition the Philips
pw 1830/40 is a relatively old machine showing the
large band of bremsstrahlung between 5 and 20
2q in the diffractograms (Figure 3). After the analy-
ses of a 98% crystalline quartz standard, the same
band has been observed and could be excluded to
be purely caused by the presence of amorphous
phases.
Due to the available equipment and filters we
consciously disclaimed the refinement of the occu-
pancy factors of the ions in the occurring solution
series and the subtraction of the amorphous phases
as suggested by Snellings et al. (2014a; 2014b). In
addition, the diffraction patterns show numerous
spurious peaks originating from the tungsten of the
old X-ray tube that also influences the statistical
parameters of the quantitative phases analyses. In
this context it should be mentioned that our results
“only” show the crystalline phases included in the
analyses and that the sums are normalized up to
the phase contents of 100% as usually performed
by the refinements. That means that only relative
proportions of the crystalline phases are obtained,
while the amorphous fraction of the samples should
better be identified by e.g. electron-microscopy and
image analysis.
The concentration of magnesium in pore solution
and the presence of apatite
The acidic character and composition of a cementi-
tious paste pore solution has a strong influence on
the stabilization or destabilization of crystalline and
amorphous phases. In this context hydration and
carbonation processes of numerous cement types
have intensively been studied (e.g. Hong & Glasser,
2002; Bullard et al. 2011).
Yet, several studies have controversially dis-
cussed the formation of apatite (Çopuroğlu et al.
2006; Zemskov S.V. et al. 2012; Söylev & Richard-
son, 2008; Douche-Portanguen et al. 2005; Farcas
et al. 2002; Bastidas et al. 2010; Tadic et al. 2002),
294
based on results of experimental work that have
shown that the crystalline calcium phosphate with
the mineral formula Ca5 (PO4 )3 (OH/F/Cl) can form
from portlandite and Na-MFP in aqueous solutions
of different pH values (Farcas et al. 2006, Basti-
das et al. 2010). Yet, the formation of apatite af-
ter Na-MFP treatment of BFSC products has never
been observed (Sisomphon et al. 2011; Douche-
Portanguen et al. 2005) and could also not be jus-
tified by thermodynamic calculations (Bastidas et
al. 2010). Although the occurrence of apatite-like
structures was earlier detected in calcium alumi-
nate cements by Pöllmann (2012), neither hydroxyl-
nor fluoroapatite could ever be identified in cemen-
titious products treated with Na-MFP (Alonso et
al. 1996; Çopuroğlu et al. 2006; Söylev & Richard-
son, 2008; Douche-Portanguen et al. 2005). In-
stead, the interaction of hardened cement paste
with Na-MFP in solution resulted in the forma-
tion of portlandite, carbonates and a paragenisis
with clinker minerals (Douche-Portanguen et al.
2005) and non-stoichiometric amorphous calcium
phosphates (Çopuroğlu et al. 2006) or traces of
FePO4 ·H2 O (Alonso et al. 1996; Söylev & Richard-
son, 2008) in the environment of steel reinforce-
ments.
Boskey and Posner (1974) experimentally investi-
gated the conversion of Amorphous Calcium Phos-
phate (ACP) to hydroxyl-apatite in the presence of
magnesium and found very low concentrations of
magnesium already influencing the amorphization
of apatite. The concentrations of Mg2+ presented
in Table 3 are sufficient to prevent the formation of
crystalline apatite in cementitious products. The re-
sults of the X-ray powder diffraction analyses also
confirm that no crystalline phosphate mineral is
formed during Na-MFP treatment of portlandite-
bearing cementitious products – at least not within
the age and time range of sample investigation.
Sulfur in pore solution, carbonation and the sta-
bility of ettringite
The results of the main element analyses (Table
3) show that S4+ in a carbonated slag-rich cement
paste pore solution is increasing to an amount of 3.5
g/l compared to the non-carbonated slag-cement
paste. Correlating this result with the result of the
crystalline phase analyses (Table 4) we observe the
amount of the crystalline sulfur-bearing mineral et-
tringite, as formed during cement hydration, to be
decreasing by progressing carbonation. That way
S4+ can be released to the pore solution. Brocken
& Nijland (2004) assume that during carbonation
vaterite and possibly other carbonates are forming
from ettringite and carbon dioxide, but do not dis-
cuss to where sulfur could partition. Though S4+
can be bound to amorphous hydration products
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
(C(A)SH phases) in cement pastes, the amount of
S4+ in the pore solution of the carbonated slag-
cement paste is significantly high. This again can
be explained by the pH of a carbonated paste (in
this case 8.8) and the fact that only at pH values
higher than 9.5 the C(A)SH phases are stable. The
high sulfur concentration in the pore solution of a
carbonated CEM III /B paste is a result of ettringite
dissolution.
Brocken & Nijland (2004) show an SEM photomi-
crograph, illustrating the formation of ettringite
needles that preferably occur at the interface be-
tween pore paste surface and paste matrix. If et-
tringite breaks down during carbonation, sulfur is
increasing especially in the pore solutions.
V. Conclusion
Through the application of aqueous Na-MFP as a
self-healing agent on CEM III /B this study shows
that Na-MFP has the potential to recover the pH
of a carbonated cement paste up to 90% with an
Na-MFP concentration of 45 wt% in the healing
agent.
Parallel to our solid-phase analytical results we
show that during carbonation of a slag-rich cement
paste phase transitions from portlandite, ettrin-
gite and C(A)SH occur to form carbonates such
as vaterite, aragonite and calcites, while the clinker
phases do not show a large reactivity in the time
period of sample investigation.
The influence of Na-MFP on a carbonated cement
paste is stabilizing secondary portlandite, ettringite
and C(A)SH, possibly due to recovery of the basic
character of the pore solution. Diffraction peaks
of the crystalline phosphate mineral apatite were
not observed within the time period of sample in-
vestigation and are also not expected to occur at
any later sample age. The latter assumption is un-
derpinned by the existence of several studies on
thermodynamic calculations, the high pH destabi-
lizing apatite and by the existence of Mg2+ in the
pore solution.
Future work by polarized light and electron mi-
croscopy will give more insight in the composition
of the Amorphous Calcium Phosphates (ACP’s)
that are forming in paragenesis with C(A)SH
phases in slag-rich carbonated cement paste treated
with the inorganic self-healing agent Na-MFP.
Acknowledgement
We would like to express our gratitude to the Dutch
Enterprise Agency (Agentschap.NL) who kindly
supported this work financially through the IOP
Self Healing Materials (SHM) Program.
References
Alonso C., Andrade C., Argiz C., Malric B. (1996) “Na2PO3F as in-
hibitor of corroding reinforcement in carbonated concrete” Cement and
Concrete Research; 26: 405-415.
Barneyback R.S., Diamond S. (1981) “Expression and Analyses of
Pore solutions from hardened Cement Pastes and Mortars” Cement and
Concrete Research; 11: 279-285.
Bastidas D.M., La Iglesia V.M., Criado M., Fajardo S., La Iglesia A.,
Bastidas J.M. (2010) “A prediction study of hydroxyapatite entrapment
ability in concrete” Constr Build Mater; 24: 2646-2649.
Bonaccorsi E., Merlin S. (2005) The crystal tobermorite 14 A (plom-
bierite) phase locality: Ballycraigy, Northern Ireland” Journal of the
American Ceramic Society; 88, 505-512.
Boskey A.L., Posner A.S. (1974) “Magnesium stabilization of amor-
phous calcium phosphate: A kinetic study” Material Research Bulletin;
9: 907-916.
Brocken H., Nijland T.G. (2004) “White efflorescence on brick ma-
sonry and concrete masonry blocks, with special emphasis on sulfate
efflorescence on concrete blocks” Construction and Building Materials;
18, 315-323.
Brouwers H.J.H., van Eijk R.J. (2003) “Alkali Concentrations of Pore
Solutions in hydrating OPC” Cement and Concrete Research; 33: 191-
196.
Bullard J.W., Jennings H.M., Livingston R.A., Nonat A., Scherer G.W.,
Schweitzer JS, Scrivener K.L., Thomas J.J. (2011) “Mechanisms of Cement
Hydration” Cement and Concrete Composites; 41: 1208-1223.
Busing W.R., Levy H. (1957) “A Neutron diffraction study of calcium
hydroxide” Journal of Chemical Physics; 26, 563-568.
Byfors K, Hansson CM, Tritthart J. (1986) “Pore Solution Expression
as a Method to determine the Influence of Mineral Additives on Chloride
Binding” Cement Concrete Res; 16: 760-770.
Cohen-Addad C., Ducros P., Durif A., Bertaut E.F., Delapalme A.
(1964) “Determination de la position des atomes d’hydrogene dans
I’hydrogrinat Al2O3, (CaO)3 (H2O)6 par resonance magnetique nucle-
aire et diffractionique” Journal de Physique; 25, 478-483.
Çopuroğlu O., Fraaij A.L.A., Bijen JMJM. (2006) “Effect of Sodium
monofluorophosphate treatment on microstructure and frost salt scal-
ing durability of slag cement paste” Cement and Concrete Research; 36:
1475-1482.
Dai Y.S., Post J.E. (1995): “Crystal structure of hillebrandite: A natural
analogue of calcium silicate hydrate (CSH) phases in Portland cement”
American Mineralogist; 80: 841-844.
De la Torre A.G., De Vera R.N., Cuberos A.J.M., Aranda M.A.g. (2008)
“Crystal Structure of low magnesium-content alite: Application to Ri-
etveld quantitative phase analysis” Cement and Concrete Research; 38,
1261-1269.
De Villiers (1971) “Crystal structures of aragonite, strontianite, and
witherite” American Mineralogist; 56, 758-767.
Douche-Portanguen A., Prince W., Malric B., Arliguie G. (2005)
“Study of interactions between sodium monofluorophosphate and hard-
ened cement paste and their consequences on concrete transfer proper-
ties” Cement and Concrete Research; 35: 1714-1723.
Farcas F., Chaussadent T., Fiaud C., Mabille I. (2002) “Determination
of the sodium monofluorophosphate in a hardened cement paste by ion
chromatography” Anal. Chim. Acta 472: 37-43.
Goetz-Neunhoeffer F., Neubauer J. (2006) “Refined ettringite
(Ca6Al2(SO4)3(OH)12*26H2O) structure for quantitative X-ray diffrac-
tion analysis” Powder Diffraction; 21, 4-11.
Graf D.L. (1961) “Crystallographic tables for the rhombohedral car-
bonates” American Mineralogist; 46, 1283-1316.
Hong S-Y., Glasser F.P. (2002) “Alkali sorption by CSH and CASH
gels Part II. Role of Alumina” Cement and Concrete Composites; 32:
1101-1111.
Jin S.X., Sagoe-Crentsil K.K., Glassert F.P. (1991) “Characteristics of
corrosion inhibition admixtures in OPC paste with chloride additions.
Part I: Chemistry and electrochemistry.” Magazine of Concrete Research;
43: 205-213.
Megaw H.D. (1952): “ The structure of afwillite,
Ca3[SiO3OH]2•2H2O Locality: Scawt Hill, Northern Ireland”
Acta Crystallographica; 5: 477-491.
Nishi F., Takeuchi Y., Maki I. (1984) “The tricalcium silicate Ca3OSiO4:
The monoclinic superstructure” Zeitschrift für Kristallographie; 172, 297.
Pöllmann H. (2012) “Calcium Aluminate Cements – Raw Materials,
Differences, Hydration and Properties” Reviews in Mineralogy and Geo-
chemistry; 74: 1-82.
Regourd M. (1986) “Caracterisiques et activation des produits
d’addition” Proc. 8th ICCC; 1, 199. Rio de Janeiro, Brasil
Schellings R., Salze A., Scrievener K.L. (2014a) “Use of X-ray diffrac-
tion to quantify amorphous supplementary cementicious materials in
anhydrous and hydrated blended cements” Cement and Concrete Re-
search; 64: 89-98
Schellings R., Salze A., Scrievener K.L. (2014b) “The existence of amor-
phous phase in Portland cements: Physical factors affecting the Rietveld
quantitative phase analyses” Cement and Concrete Research; 59: 139-
146.
295
 Kempl & Çopuroğlu
Sisomphon K, Çopuroğlu O, Fraaij A. (2009): “Transport Properties
and Frost Salt Scaling Resistance of Carbonated Blast-Furnace Slag Mor-
tars after Sodium Monofluorophosphate Treatment” 2nd International
RILEM Workshop on Concrete Durability and Service Life Planning;
Haifa, Israel. 9: 138-145.
Sisomphon K., Çopuroğlu O., Fraaij A. (2010a): “Properties of carbon-
ated blast furnace slag mortars after Na2FPO3 treatment” Science Asia;
36: 223-230.
Sisomphon K, Çopuroğlu O, Fraaij A. (2010b) “Development of Blast
Furnace Slag Mixtures against Frost Salt Attack” Cement and Concrete
Research; 32: 630-638.
Sisomphon K., Çopuroğlu O., Fraaij A. (2011) “Application of en-
capsulated lightweight aggregate impregnated with sodium monofluo-
rophosphate as a self-healing agent in blast-furnace slag mortar” Heron;
56: 13-32.
Smith V. (1953) “Reexamination of the crystal structure of melilite”
American Mineralogist; 38, 643-661.
Song S., Jennings H.M. (1999) “Pore solution of alkali-activated
ground granulated blast-furnace slag” Cement and Concrete Research;
29:159-170.
296
Söylev T.A., Richardson M.G. (2008) “Corrosion inhibitors for steel in
concrete: State-of-the-art report” Const Build Mater; 22:609-622.
Tadic D., Peters F., Epple M. (2002) “Continous synthesis of amor-
phous carbonated apatites” Biomaterials; 23: 2553-2559.
Taylor H.F.W. (1997) Cement Chemistry, 2nd ed., Thomas Telford,
London.
Udagawa S., Urabe K., Natsume M., Yano T. (1980) “Refinement of the
crystal structure of gamma-Ca2SiO4” Cement and Concrete Research; 10,
139-144.
Wang J., Becker U. (2009) “Structure and carbonate orientation of va-
terite (CaCO3)” American Mineralogist; 94, 380-386.
Zemskov S.V., Ahmad B., Çopuroğlu O., Vermolen F.J. (2012) “Model-
ing of a self-healing process in blast furnace slag cement exposed to accel-
erated carbonation.” IC-MSQUARE: International Conference on Mathe-
matical Modelling in Physical Sciences; Journal of Physics: Conference
Science; 410: 1-4.
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Recycled materials
f 0 F

Mortars & tiles

297
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

298
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

The ITZ in concrete with natural and

recycled aggregates: Study of
microstructures based on image and SEM
analysis
G. Bonifazia , G. Capobiancoa , S. Serrantia,⇤ , M. Eggimannb
E. Wagnerc , F. Di Maiod , S. Lotfid
a Department of Chemical Engineering Materials & Environment, Sapienza University of Rome, Rome, Italy
b Holcim Technology Ltd., Innovation, 5113 Holderbank, Switzerland
c Heidelberg Cement AG, Germany
d Sec. of Materials and Environment, Fac. of Civil Engineering and Geosciences, Delft University of Tech., The Netherlands
⇤ silvia.serranti@uniroma1.it

Abstract

Aim of this work was to investigate the microstructure of the Interfacial Transition Zone (ITZ) between cement paste
and aggregate in concrete utilizing Scanning Electron Microscope (SEM) in order to identify possible effects on the ITZ
related to different recipes and production parameters. SEM is an important tool to carry out concrete, cement and
aggregate microstructure characterization. SEM imaging facilitates identification of elements in concrete. Thanks to the
higher spatial resolution, and the analytical capabilities, it allows to perform a materials-imaging-related analysis. Many
information can thus be extracted from these data, such as composition, phase abundance and distribution. In particular,
during the last years, backscattered electron imaging of polished surfaces became an important method for the study of
cement and concrete microstructures. The technique has many advantages, as the possibility to visualize representative
cross-sections over a wide range of magnifications and to correlate image contrast with materials composition (i.e. atomic
number). Comparing the information acquired starting from SEM imaging and X-ray microanalysis, the complex
microstructure of concrete can be analyzed and images with sub-micrometer definition acquired and processed, obtaining
compositional, morphological and morphometric information allowing to understand and evaluate the mechanism
affecting the interaction of the aggregates with the cement paste. A common image representative of the ITZ in
concrete involves approximately a domain of about 30 µm surrounding each aggregate and showing, theoretically, an
increase of the porosity at the aggregate interface. In order to study this transition zone, 12 different types of concrete
mixtures were analyzed, characterized by different combination of cement and virgin/recycled aggregates. The influence
of these aggregates (i.e. virgin and recycled) on the resulting concrete microstructure was studied. The significant
morphological and overall textural characteristics (i.e. area with higher porosity around aggregate, micro-cracks, size of
the crystals in the ITZ) were detected by SEM analysis. Image analysis and SEM investigation, obtained from different
concrete samples, provided very interesting results due to the different porosity and the possible interactions between the
aggregates and the cement paste.

Keywords: Recycled aggregates, concrete, Interfacial Transition Zone (ITZ), SEM, image analysis

I. Introduction concrete mix design, the operative conditions to re-

The possibility to reutilize aggregates, resulting
from recycling of End-Of-Life (EOL) concrete, repre-
sents a challenge in the sector. In these last years,
more and more recycled coarse aggregates have
been proven to be commercially and technically
sound for both non-structural and structural appli-
cations.
Recycled Aggregate Concrete (RAC) behavior and
characteristics are strongly conditioned and influ-
enced by the "source" waste aggregates, the adopted
alize the mix, and the intrinsic RAC attributes (i.e.
mineralogical composition, surface status, morpho-
logical and morphometrical properties). A preemi-
nent role in final concrete characteristics is played
by the cohesive interface properties resulting from
RAC utilization, properties strongly linked to Inter-
facial Transition Zone (ITZ) characteristics, old mor-
tar quality and the old mortar content of the origi-
nal concrete (Xia et al., 2013).
The ITZ, that is the region between aggregate
and cement paste, is usually characterized by a

299
 Bonifazi et al.

Figure 1: Polished sections of the investigated samples, as resulting after specimen preparation, finalized to perform
SEM analyses. The details about the adopted samples labelling is reported in Table 1.

different microstructure than the bulk paste. ITZ
microstructure characterization is usually carried
out analyzing Backscattered Electron Images (BEI) ob-
tained by Scanning Electron Microscopy (SEM) based
investigations. Following this approach ITZ higher
porosity and lower content of un-hydrated cement
grains, in respect of the bulk paste, is usually de-
tected. These characteristics change according to
aggregate surface distance up to reach those of the
bulk paste. This distance is usually defined as ITZ
thickness (Elsharief et al., 2003). Water-cement ratio
(w/c) has been considered to have only small effect
on the ITZ microstructure (Scrivener, 1999).
Mortar and concrete porosity in the ITZ is higher
compared to the bulk of the hardened paste. Fresh
concrete subjected to compaction shows, around the
areas of the granules, a greater size of the aggregate
and a water film (of hundreds of microns) collecting
bleeding water. As result a higher water/cement
ratio occurs in the ITZ in respect of the cement
matrix, this fact determines, as a consequence, a
higher porosity in the ITZ.
The formation of the first-crystalline germs
of calcium hydroxide (Ca(OH)2 ) and ettringite
(3CaO.Al2 O3 .3CaSO4 .32H2 O), occurring through-
out the cementitious matrix (i.e. cement silicates
and aluminates hydration), is followed by a greater
crystal growth of these products in the transition
area and a corresponding production of higher
porosity due to the "local" existence of higher water-

Table 1: Concrete samples labelling according to utilized aggregates and cement based recipes C1 and C2 (see Table 2).

Sample Description ADR feed rate Autogenous mill

[t/h] tilt angle [degrees]
C1R-REF Concrete produced with natural roundish aggregates - -
C1-D15-100 15 5
C1-C50-100 50 5
Concrete produced with 100% recycled aggregates adopting recipe C1
C1-A100-100 100 7
C1-B100-100 100 5
C1-ADR-100 Concrete produced with 100% recycled aggregates without preliminary 100 -
autogenous milling
C1C-REF Concrete produced with natural crushed aggregates - -
C2-REF Concrete produced with natural roundish aggregates - -
C2-D15-100 15 5
C2-C50-100 50 5
Concrete produced with 100% recycled aggregates adopting recipe C2
C2-A100-100 100 7
C2-B100-100 100 5

300
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
solid ratio. This latter effect can be attributed to
the "wall effect" leading to a relatively low con-
centration of large cement particles in the ITZ pro-
ducing, as a consequence, an increased porosity.
During cement hydration the pores in the ITZ are
partially filled, according to a redistribution of cal-
cium hydroxide and calcium-silicate-hydrates (C-
S-H) (Elsharief et al., 2003). The ITZ of individual
aggregates can be interconnected depending on
their width and the content of aggregates. Such
connectivity, or percolation, is considered as an
important characteristic that can influence and/or
condition the permeability and transport properties
of mortar and concrete (Leemann et al., 2006). Ex-
tra pore space, originally associated with the wall
effect in ordinary concrete, is thus mostly filled up
by deposits of calcium hydroxide and by deposits
of (C-S-H) passing through solution, leading to the
retention of only a relatively low residual extra av-
erage pore content within the ITZ (Diamond and
Huang, 2001).
Historically, recycled aggregates properties were
studied through the evaluation of mix workability
and "final product" compressive strength. The long-
term properties evaluation/comparison of normal
and recycled based aggregates concretes, studying
shrinkage and creep phenomena, was also carried
out. However, few studies have been conducted on
exploring the microstructure of recycled aggregate
concrete and interactions with cement paste (Poon
et al., 2004).
ITZ characteristics can be qualitatively revealed
via SEM performing a BSE analysis. A BSE image,
in fact, depends on the mean atomic number of
the concrete constituents. The grey scale of a BSE
image is thus linked to the composition and textu-
ral attributes of the investigated regions (Hussina
and Poleb, 2011). Following this strategy porosity
and individuals (i.e. cement particles and/or crys-
tals) are clearly evidenced (i.e. epoxy impregnated
pores are seen as black). Furthermore the possi-
bility to perform a digital processing of the image
could allow to quantify several topological and mor-
phometrical attributes ITZ related. Combining the
information derived from the BSE images and those
obtained by EDX analysis (i.e. maps of elements)
on the same ITZ areas, it is thus possible to acquire
a huge amount of information allowing to evalu-
ate the interaction existing between aggregates and
cement paste, in respect of different recipes and
aggregates origin: virgin or recycled.
The results reported in this paper are re-
lated to some ITZ studies performed inside
a project financed by the European Union
(EU): Collaborative Project n.265189, C2CA: "Ad-
vanced Technologies for the Production of Cement
and Clean Aggregates from Construction and Demo-
lition Waste" (http://www.c2ca.eu/activities/the-
c2ca-project/) whose main target is to compare
the interaction of cement/aggregate between nat-
ural and recycled aggregates. The C2CA project,
in fact, aims to develop a cost-effective approach
for recycling EOL concrete high-volume streams
into prime-grade aggregates and cement. To reach
this goal an innovative low-cost combined com-
minution/classification, called Advanced Dry Recov-
ery (ADR) was applied to EOL concrete in order to
produce secondary cleaned aggregates, removing
at the same time the presence of fines and light con-
taminants, in size class interval ranging between 1
and 4 mm. The utilization of the ADR approach
allows to break the bonds existing between by mois-
ture and fine particles and to realize a materials
classification independent from their moisture con-
tent. After breaking up the material into a jet, the
fine particles are separated from the coarse particles.
ADR separation produces as results the aggregates
concentrations into a coarse aggregate product and
the production of fine fraction constituted by ce-
ment paste and contaminants (i.e. wood, plastics
and foams) (Lotfi et al., 2014).
II. Materials and methods
The study was performed with reference to 12 dif-
ferent types of concrete mixtures (Figure 1) realized
utilizing different aggregates and cement. The se-
lected aggregates were constituted by virgin ones
and as resulting from recycling according to dif-
ferent sequential autogenous and ADR based com-
minution strategies (Table 1). EOL concrete was
preliminary milled by autogenous milling, then the
-16 cm fraction was fed to ADR. Two series of con-
crete, identified as C1 and C2 were produced by
Holcim and HeidelbergCement AG, respectively,
utilizing different cements and mix designs. Their
characteristics are outlined in Table 2.
All the analyses have been carried out by a SEM
equipped with an Energy Dispersive X-ray Spec-
troscopy (EDS) analytical unit.
Cast samples were cut in slices of about 1 cm
thickness. From each slice a prismatic sample of
about 2x1x1cm (Figure 2) was realized and utilized
for polished samples preparation (Figure 3). To
reach this goal, concrete "prismatic specimens" were
immersed in ethanol and impregnated with epoxy
resin. Samples were then cut by diamond-bladed
saw and polished, utilizing a series of abrasive pa-
pers (i.e. 240, 600, 1000 and 2000 grit) for about
2 min each, with white spirit. Finally, a diamond
paste of 3 µm on a low-relief polishing cloth was
utilized for a further surface specimen polishing.
Polishing was manually carried out for about 2 min.
Samples were examined by a S2500 Hitachi SEM
Microscope equipped with Kevex 8000 for X-ray
microanalysis. Polished samples were examined
301
 Bonifazi et al.
Table 2: Composition of C1 and C2 concrete mixtures.
Recipes Cement [kg/m3 ] w/c ratio Water [kg/m3 ]
C1 300 CEM III 0.60 180
C2 340 CEM I 52.5R 0.45 153
both under secondary electronic imaging and BEI.
Detection was performed with a low acceleration
voltage of 20 kV. Six magnifications, i.e. 50x, 100x,
150x, 200x and 400x were adopted. Images were
"randomly" acquired around aggregates in order
to statistically guarantee the representative of the
analyses.
SEM analyses have been carried out in order to
identify and quantitatively assess, the most signif-
icant morphological, morphometrical and overall
textural characteristics of the ITZ, that is: i) pres-
ence of regions of porosity around aggregates, ii)
presence of micro-cracks, iii) presence of multi-layer
oriented crystals around the aggregates and iv) size
of the crystals detected in the ITZ. Micro-cracks
around the ITZ have been observed at low mag-
nification (50x). ITZ morphological and porosity
characteristics have been identified adopting an in-
termediate magnification (from 200x to 300x). Crys-
tal presence (i.e. ettringite or hydroxide of calcium)
and size have been detected at high magnification
(from 400x to 2000x). X-ray mapping was systemat-
ically utilized (400x) to identify the possible pres-
ence of the ITZ in respect of cementiceous matrix.
III. Results
The C1C-REF sample (Figure 4a), constituted by vir-
gin crushed aggregates, shows a variable porosity
around greater aggregates. The average dimension
of the pore, detectable in the ITZ. is about 30 µm.
The effect of water bleeding is quite evident (Fig-
ure 4b). A clear separation between aggregates and
cement matrix exists. Oriented multi-layer crystals
structures can be also identified in the areas char-
acterized by a higher porosity (i.e. water bleeding
Figure 2: Example of quartering performed with reference to C2 C50 100 sample (2a) and a slice samples (about
2x1x1cm) utilized to prepare the polished sections analyzed by SEM (2b).
302
Plasticizer Air entrainer Consistency Class Aggregate grading
none none S3 DIN A-B 16 mm
For slump none S3 DIN A-B 16 mm
effect) around bigger aggregates (Figure 4c).
The C1R-REF sample (Figure 5a), composed with
roundish aggregates, shows an ITZ characterized
by a lower porosity in comparison of that detected
in C1C-REF sample. A complete detachment of
the aggregate from cement matrix can be observed
(Figure 5c). The average thickness of ITZ is about
40-50 µm. X-ray mapping shows a higher calcium
concentration around the biggest aggregates and a
lower, as found in C1C-REF sample, silicate content
(Figure 5b). These phenomena can be associated to
the presence of calcium hydroxide along the ITZ.
The C2-REF matrix is well linked with the aggre-
gates. Micro-cracks at low magnification (Figure 6a)
are practically undetectable. ITZ average area is
below 5 µm in this sample (Figure 6b) changes of
matrix cement structure and characteristics around
aggregates are not detected. Micro-cracks can be
observed only at higher magnification (Figure 6c).
Micro-cracks, for their characteristics, can be associ-
ated to cement shrinkage and not to point of weak-
ness around aggregate. Making a comparison with
the results obtained for C1R-REF it results clear as
C2-REF samples present better ITZ characteristics
than C1R-REF. This difference can be mainly at-
tributed to the morphological and morphometrical
characteristics of the aggregates.
Cement matrix of C2-B100-100 sample is well
linked with the aggregates. The presence of micro-
cracks is not significantly detectable (Figure 7a)
both at low and high magnification. Around the
aggregates no significant variation in the cement
matrix is observed. The presence of oriented ag-
gregates, due to bleeding effect, is also evidenced
(Figure 7b). The ITZ is very small with an average
dimension ranging between 6 and 8 µm (Figure 7c).
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Samples preparation sequence. Vacuum system and impregnated specimens (3a) , diamond-bladed saw
(3b),polishing treatment (3c) and diamond pastes utilized to perfom hand made surface finishing (3d).

Figure 4: Backscattered Electron Image (BEI) based acquisition of C1C-REF concrete sample at 50x (4a), 200x (4b)
and 400x (4c) magnification, respectively.

Figure 5: BEI based image acquisition of concrete C1R-REF sample at 50x (5a) magnification, maps of element of Si,
O , Ca (5b), and BEI image at 400x (5c) magnification.

303
 Bonifazi et al.
Figure 6: BEI based image acquisition of concrete C2-REF sample at 50x (6a), 200x (6b) and 400x (6c) magnification,
respectively.
Micro-fractures are almost absent. ITZ character-
istics are in good agreement with the operative con-
ditions adopted for the EOL concrete comminution
and the adopted low water/cement (W/C) ratio in
concrete mix design.
The results obtained for C2-B100-100 sample also
fit with those of C2-D15-100 sample. Based on their
analysis, in fact, it is clear as EOL concrete feed
rate to ADR is not so important with respect to
the milling actions carried out and the adopted
concrete mix design.
C1-B100-100 sample shows an ITZ extremely
complex, with a high porosity and a large pres-
ence of micro-fractures. Furthermore the BEI shows
as a marked difference between matrix and cement
and as the ITZ, surrounding the aggregates, shows
the presence of oriented crystal aggregates due to
water bleeding effects. The visualized transition
zone is characterized by a width around 65 µm (Fig-
ure 8a). The analyses carried out by EDS produce
an elements mapping confirming the probable pres-
ence, in the ITZ, of calcium hydroxide (Figure 8b).
The pores of larger dimension are characterized by
the presence of clearly visible great size crystals of
ettringite and calcium hydroxide (Figure 8c).
C1-C50-100 sample shows an area around greater
Figure 7: C2-B100-100 concrete sample. BEI image fields acquired at different magnification: 50x (7a), 200x (7b)
and 400x (7c) magnification.
304
aggregates, characterized by high porosity. In this
area (ITZ) the presence of micro-cracks is evident
(Figure 9a). Comparing the elements map (Fig-
ure 9b), resulting from EDS analysis, is possible to
detect the presence all around the aggregates of Ca.
The thickness of the area showing the presence of
Ca is of about 30 µm). BEI determinations show
the presence of an ITZ of 50 µm average thickness.
At higher magnification (400x), it is possible to de-
tect, inside the ITZ, the presence of dark regions,
representative of a marked porosity. The presence
of micro-aggregates, ranging between 5 and 20 µm,
is noticeable and the texture that they originates
can be clearly attributed to water bleeding effects
(Figure 9c).
C2-C50-100 sample shows the presence of old ce-
ment paste attached to the aggregates. Analyses
show that the old cement is characterized by a dif-
ferent matrix, embedding small calcium carbonate
individuals and/or aggregates. The presence of
micro-cracks is low (Figure 10a). Apparently, any
marked discontinuity seems to exists, between the
old and new concrete (Figure 10b). The dimension
of ITZ is not relevant, being the detected thickness
always below 10 µm (Figure 10c).
These results are quite interesting, evidencing
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 8: C1-B100-100 concrete sample. BEI image field acquired at 200x (8a). Si, O and Ca, X-ray mapping (8b).
BEI image field acquired at 400x (8c).
as ITZ characteristics are not particularly influ-
enced by the utilization of recycled aggregates, on
condition that correct operative conditions to recy-
cled aggregates and proper concrete manufacturing
recipes are adopted.
C1-D15-100 sample is characterized by a high
porosity, clearly identified by the presence of darker
color regions around the aggregates, mainly around
the big ones (Figure 11a). The presence of micro-
cracks, detectable inside the ITZ (Figure 11b), is
also noticeable. ITZ thickness ranges in average, be-
tween 30-40 µm and below. ITZ texture, pictorially
evidenced by grey tone variations, is clearly due to
a marked bleeding water effect (Figure 11c).
C2-D15-100 sample does not show a high poros-
ity around the ITZ. The presence of the old cement
matrix is quite evident (Figure 12a). Any marked
discontinuity linked to the old (i.e. presence of indi-
viduals, constituted by cement and aggregates, be-
longing to the utilized EOL concrete ) and new con-
crete (Figure 12b) seems not to be present, porosity
is low and material composition around aggregates
does not present significant variations (Figure 12c).
Making a comparison with the corresponding
sample (C1-D15-100) obtained with a different mix
Figure 9: C1-C50-100 concrete sample. BEI image field acquired at 50x (9a). Si, O and Ca, X-ray mapping (9b).
BEI image field acquired at 400x (9c).
design, it seems quite clear the role that these pa-
rameters play in respect of the final attributes of
concrete, as well as ITZ characteristics.
Analyzing the C1-A100-100 sample (Figure 13a),
the presence of micro-cracks inside the cementitious
matrix is clear (Figure 13a). The presence of old
cement elements, around the bigger aggregates, is
also detected as well as the presence of micro-cracks
all around the ITZ of new concrete and inside the old
one (Figure 13b). At higher magnification (400x),
darker areas, representative of porosities and/or
physical discontinuities inside the sample, can be
detected. The ITZ average thickness is estimated to
be 30 µm (Figure 13c).
The C2-A100-100 sample does not show the pres-
ence of a significant amount of micro-cracks inside
the cementitious matrix (Figure 14a). The maps of
element confirm the presence of calcium hydroxide
in the ITZ (Figure 14b). ITZ average thickness, de-
tected at higher magnification (400x), is about 12
µm (Figure 14c). Also for the detected textural char-
acteristics of samples C1-A100-100 and C2-A100-
100 the considerations already done for samples
C1-D15-100 and C2-D15-100 are valid; these are the
influence of water water/cement (w/c) ratio and
305
 Bonifazi et al.

Figure 10: C2-C50-100 concrete sample. BEI image field acquired at 50x (10a), 200x (10b) and 400x (10c) magnifi-
cation, respectively.

Figure 11: C1-D15-100 concrete sample. BEI image field acquired at 50x (11a), 200x (11b) and 400x (11c) magnifi-
cation, respectively.

Figure 12: C2-D15-100 concrete sample. BEI image field acquired at 50x (12a), 200x (12b) and 400x (12c) magnifi-
cation, respectively..

306
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 13: C1-A100-100 concrete sample. Secondary electron image acquisition of concrete at 50x (13a). BEI image
field acquired at 200x (13b) and 400x (13c) magnification, respectively.

Figure 14: C2-A100-100 concrete sample. BEI image field acquired at 50x (14a), X-ray mapping of Si, O and Ca
elements (14b). BEI image field acquired at 400x (14c).

Figure 15: C1-ADR-100 concrete sample. BEI image field acquired at 50x (15a), X-ray mapping of Si, O and Ca
elements (15b). BEI image field acquired at 400x (15c).

concrete mix design. nification (Figure 15a). The map of the elements
referred to calcium, silicon and oxygen confirms
Making a comparison between the two pairs of
the probable presence of calcium hydroxide in the
C1-A100-100, C2-A100-100 and C1-D15-100, C2-
transition zone between matrix and aggregate (Fig-
D15-100 samples, it is also quite clear the influence
ure 15b). ITZ also shows a marked porosity, its
that autogenous mill tilt angle set-up and feed rate
thickness varies between 50 and 70 µm and the
play on ITZ.
presence of oriented crystals, determined by wa-
C1-ADR-100 is characterized, in respect of the ter bleeding effect (Figure 15c), can also be clearly
other samples, by the marked presence of micro- detected.
cracks all around the greater aggregates. Further-
more the ITZ can be identified also at lower mag-

307
 Bonifazi et al.
IV. Conclusion
The micro-structure and the attributes of the ITZ
in different concrete sample, realized with natural
and recycled aggregates of different characteristics,
according to different comminution strategies (i.e.
adoption of autogenous milling and different ADR
feeding conditions) and adopting different concrete
mix designs was carried out.
According to the achieved results, it is possible
to affirm as concrete samples produced by natural
roundish aggregates (i.e. C2-REF samples) and
those produced with recycled aggregates (i.e. C2-
C50-100), obtained after ADR processing with a
medium feed rate (i.e. 50 t/h), adopting the same
concrete C2 mix design (i.e. low water/cement
ratio), seems to be characterized by the presence
of an ITZ with an average thickness of about 2
µm. This result is in good agreement with the
operative conditions adopted to produce aggregates
(i.e. lower values of ADR feed produce a higher
degree of liberation of aggregates form the matrix)
and concrete mix design (C1 and C2).
Samples C2-D15-100 and C2-B100-100 show an
ITZ characterized by an average thickness ranging
between 2-5 µm and 5-12 µm, respectively. These
results are clearly influenced by ADR feeding set up.
For sample C2-A100-100 the values of ITZ increase
in agreement with the different autogenous mill
operative conditions, in respect of those adopted for
samples C2-D15-100 and C2-B100-100, that is higher
tilt angle and, as a consequence, lower residence
time of the material in the mill before to be fed to
ADR.
Comparing ITZ characteristics detected for the
concrete obtained with natural roundish (i.e. C1R-
REF) and crushed (i.e. C1C-REF) aggregates, adopt-
ing the same C1 mix design (i.e. high water/cement
ratio), it is possible to see as the first sample shows
lower ITZ thickness values (about 50 µm) than the
second one (about 60 µm). Such behavior can prob-
ably due to the different aggregates morphological
and morphometrical attributes. On the contrary,
comparing the ITZ concrete samples characteris-
tics as resulting from the utilization of recycled ag-
gregates not preliminary milled, but directly feed
to ADR (C1-ADR-100), the results, independently
from recipes, show the highest ITZ thickness, the
highest porosity and the larger presence of micro-
cracks.
The study of the ITZ for all the investigated sam-
ples shows that:
• autogenous milling-ADR based combined com-
minution influence aggregates characteristics
(i.e. aggregate degree of liberation in respect
of cementitious matrix,
• the water/cement (w/c) ratio in mix concrete
308
is influenced by the aggregates degree of liber-
ation and, as a consequence, the ITZ character-
istics will change,
• a clear correlation can be established among
recycled aggregates production strategies, se-
lected concrete manufacturing mix designs,
ITZ characteristics, concrete mechanical behav-
ior and durability.
Further studies will be addressed to validate the
obtained results, analyzing more concrete samples
produced with the same mix designs and to assess
ITZ attributes adopting imaging based strategies.
Acknowledgement
This research is funded by the European Commis-
sion in the framework of the FP7 Collaborative
project "Advanced Technologies for the Production of
Cement and Clean Aggregates from Construction and
Demolition Waste (C2CA)". Grant Agreement No.
265189.
References
Elsharief A., Menashi D. Cohen M.D. and Jan Olek J. (2003): "Influence
of aggregate size, water cement ratio and age on the microstructure of
the interfacial transition zone". Cement and Concrete Research, vol. 33,
1837-1849.
Hussina A., Pooleb C., (2011): " Petrography evidence of the inter-
facial transition zone (ITZ) in the normal strength concrete containing
granitic and limestone aggregates". Construction and Building Materials,
Vol. 25, Issue 5, 2298-2303.
Lotfi S., Dejab J., Rem J.P., Mróz R., van Roekel E. and van der Stelt
H. (2014): " Mechanical recycling of EOL concrete into high-grade aggre-
gates". Resources, Conservation and Recycling, vol. 87, 117-125.
Scrivener K.L. (1999): " Characterization of the ITZ and its quantifi-
cation bytest method", in: M.G. Alexander, G. Arlliguie, G. Ballivy, A.
Bentur, J. Merchand (Eds.), Engineering and Transport Properties of the
Interfacial Transition Zone in Cementitious Composites, RILEM Report,
vol. 20, 3-14.
Xia J., Li W. and Corr D.J. (2013): " Shah S.P Effects of interfacial tran-
sition zones on the stress-strain behavior of modeled recycled aggregate
concrete". Cement and Concrete Research Vol. 52, 82-99.
Leemann A., Münch B., Gasser P. and Holzer L.(2006): " Influence
of compaction on the iterfacial transition zone and the permeability of
concrete" Cement and Concrete Research, vol. 36, 1425-1433.
Diamond S. and Huang J. (2001): " The ITZ in concrete - a different
view based on image analysis and SEM observations". Cement and Con-
crete Composites,vol. 23, Issues: 2-3, 179-188.
Poon C.S, Shui Z.H and Lam L. (2004): " Effect of microstructure of
ITZ on compressive strength of concrete prepared with recycled aggre-
gates" Construction and Building Materials vol. 18, Issue 6, 461-468.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Characterization of the interface between

aggregate and cement paste of mortar
based on recycled aggregate by scanning
electron microscopy
T. Le⇤,a,b , G. Le Saouta , É. Garcia-Diaza , S. Remondb
a C2MA, École des mines d’Alès, 30100 Alès, France.
b LGCgE-GCE, École des Mines de Douai, 59508 Douai, France.
⇤ thang.le@mines-ales.fr

Abstract

The demolition of concrete structures generates large amounts of wastes that could be recycled into concrete as
aggregates. The presence of porous adherent paste in the recycled concrete aggregates (RCA) leads to a greater water
absorption that must be taken into account in the formulation of concrete. Depending on the saturation state of RCA,
the microstructure of the Interfacial Transition Zone (ITZ) between old paste adhering to RCA and new cement paste
differs and could influence the mechanical properties of concrete containing RCA. In this work, we characterize the
ITZ between the old and the new cement paste for two moisture conditions of aggregate (dry and over-saturated). Two
mortars, one based on natural aggregates (NA) and the other on RCA, were manufactured with the same volume of
dry aggregates and the same effective water to cement ratio. The ITZ after 2 and 28 days of hydration is characterized
by scanning electron microscopy. From the image analysis, the proportions of anhydrous phase, hydrated phase and
porosity are computed as a function of distance to interface. The porosity in the ITZ of mortars containing RCA is larger
than that of mortars made with natural aggregates. After two days of hydration, the average porosity in the ITZ of
mortar with over-saturated RCA is slightly higher than that with dry RCA. But the saturation state of aggregate has a
significant influence on the distribution of porosity in the ITZ. However, after 2 and 28 days of curing, the compressive
strengths of mortars containing dry and over-saturated RCA are close and the different distribution of porosity in the
two ITZ do not influence significantly the mechanical properties of mortars.

Keywords: recycled concrete aggregate, microstructure, interfacial transition zone

I. Introduction between aggregates and cement paste and the qual-

The construction industry is a large consumer of
natural resources. In the same time, it gener-
ates large amounts of construction and demolition
wastes that eventually end up in landfills without
any form of recovery. According to the European
Union directive 2008/98/EC, one aim is to increase
recycling of construction and demolition waste to
at least 70 % by the year 2020 (Blengini et al. 2012).
The recycled concrete aggregates (RCA) have the
potential to replace natural resources however it is
necessary to assess the effect of recycled aggregates
on the final concrete.
RCA are composed of a mixture of natural aggre-
gates roughly coated with hardened cement paste.
The presence of cement paste which is very porous
leads to a greater absorption of water that must be
taken into account in the formulation of the con-
crete. In concrete, the amount of water available
for cement hydration, the quality of the bond zone
ity of the RCA strongly influence its mechanical
properties.
The quality of this bond zone depends on the
presence or not between the aggregate and the bulk
cement paste of a more porous cement paste zone
than the bulk paste called interfacial transition zone
(ITZ). Several authors (Scrivener et al. 1996; Ollivier
et al. 1995; Scrivener et al. 2004) have studied in
details the microstructure of the ITZ and its forming
mechanisms.
The ITZ is constituted by a cement paste with a
microstructure gradient between the bulk cement
paste and the aggregate: the porosity and the port-
landite content increases in the ITZ from the bulk
cement paste to the aggregate surface. This mi-
crostructure gradient is mainly the consequence of
the “wall effect” by the aggregate on the fine ce-
ment particles. Because of this “wall effect” the
initial cement content decreases and respectively
the water content increases from the bulk cement

309
 Le et al.
paste to the aggregate surface. The thickness of the
disturbed area by the “wall effect” is of several tens
of microns and corresponds to the initial thickness
of the ITZ. During the hardening and because of
a filling of the porosity by the hydrated products
the depth of the ITZ generally decreases. During
the hardening, a mechanism of transport of the
hydrated products (mainly the portlandite) occurs
from the rich to the poor cement area. Because of
the portlandite transport the formation of a “du-
plex film” rich in portlandite of few microns can
be observed at the aggregate surface. In the case of
concrete based on RCA, three different kinds of ITZ
are present: ITZ within the RCA, ITZ between the
old and new cement paste, ITZ between the new
paste and natural aggregate present in the RCA.
The difficulty to control the quality of recycled
aggregate leads to different conclusions in literature.
We can find different contradictory opinions on the
influence of absorption on quality of ITZ and on
mechanical properties.
Du et al. (2010) used the digital micro-hardness
meter to characterize the strength of ITZ and re-
ported that the micro-hardness of the RCA was
much higher than that of the ITZ in concrete based
on RCA, and the micro-hardness of the interface
between the the old cement paste of recycled ag-
gregate and the new paste was the lowest. On
the other hand, Tam et al. (2005) and Etxeberria
et al. (2006) concluded that the old mortar of the
RCA formed the weak link in the concrete based
on RCA due to the presence of many porosity and
micro-cracks within the RCA. According to Xiao
et al. (2012) and Rasheeduzzafar et al. (1984), the
failure behavior of concrete based on RCA depends
on the relative quality of old ITZ and new ITZ. In
the work of Xiao et al. (2012), the first bond crack
appeared around the weakest link, the ITZs, and
then propagated into the mortar region approxi-
mately along the loading direction. For modeled
recycled concrete aggregate and concrete based on
RCA, the microcrack location, size and shape in the
ITZs were significantly influenced by the relative
Table 1: Physical properties of the sands.
Sands
absolute particle density1
particle density on an oven dried basis1
particle density on a saturated and surface dried basis1
water absorption at 24 h, W24h
water content in wet sand, WWS
open porosity3
1: measured according to the European standard EN 1097-6,
2 calculated as equation 1, 3 calculated as (1- particle density on an oven dried basis /absolute particle density),
4 measured using the evaporometry method (Mechling et al. 2003).
310
strengths of the old mortar and new mortar. When
the strength of the new ITZ was lower than that
of the old ITZ, the first observable crack mostly
appeared at the new ITZ region. Also, when the
strength of the new ITZ was better than that of the
old ITZ, the first observable crack mostly appeared
in the old ITZ area. Contrary to the common opin-
ion, Nagataki et al. (2004) reported that the primary
parameter determining the quality of the concrete
based on RCA is not always the adhered mortar.
Xiao et al. (2012) concluded that the old mortar
of the RCA forms the weak link in RAC, which is
composed of many porosities and cracks. These
pores and cracks increase the consumption of water,
leading to less water for hydration at the new ITZ.
Zhao (2014) studied the microstructure of ITZ
in concrete based on RCA by SEM. SEM observa-
tions of recycled mortars with saturated fine RCA
demonstrated that the new ITZ can clearly be seen,
while with dried fine RCA, the new ITZ zone cannot
clearly be seen. It means the properties of new ITZ
are improved by the water absorption of fine RCA.
And a higher compressive strength with dried fine
RCA than that with saturated fine RCA was also
observed.
Another study carried out on the influence of
saturation state of different aggregate of different
absorption coefficients on the properties of mortars
has been performed by Nguyen (2013). In the case
of low porosity limestone aggregate, a little or no
difference in the porosity in the ITZ between mor-
tars based on dry and wet sand limestone all along
the distance from aggregate is observed. However,
in the case of more porous limestone aggregates,
the porosity in the ITZ is much higher for the mor-
tars based on wet sand than that of mortars based
on dry sand. In the case of large absorption co-
efficient, the higher porosity in the ITZ obtained
for saturated aggregates in comparison to dry ones
could explain the lower mechanical strengths.
As a conclusion, the quality of ITZ depends on
the surface characteristics of aggregate particles,
chemical bonding, absorption and saturation state
NS RS
(g.cm 3) 2.67 2.66
(g. 3 ) 2.63 2.25
(g. 3 ) 2.64 2.41
(%) 0.51 6.81 /11.64
(%) 10.5 20
(%) 1.5 15.4
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

of aggregate. A few studies about the saturation

state of aggregate on the properties of ITZ are car-
ried out. A few quantitative studies have been
carried out on the new ITZ in the RAC and on
the influence of saturation state of aggregate on its
properties. However, there is no quantitative study
of new ITZ in RAC.
The aim of this work is to characterize the ITZ,
particularity at a mesoscopic scale, between the old
and the new paste of cement for different ages of
curing and then to compare them to ITZ of mortar
based on natural aggregate.
Figure 1: Cumulated particle size distribution of the
sands (NS and RS).
II. Materials and methods

The RCA and natural sand are provided by the Na-
tional French Project for Research and Development
RECYBETON. For the preparation of the mortars,
the RCA were crushed and sieved to obtain the
same grain size distribution as the natural sand
(NS) (Figure 1). The recycled sand (RS) is reconsti-
tuted from the six size fractions, according to the
size of the natural sand. The physical properties of
the sands are given in Table 1. The cement used
in the study is an ordinary Portland cement OPC
(CEM II/A-L 42,5N CE CP2 NF).
Measurement of the moisture content and absorp-
tion coefficient of aggregates is essential, because if
the methods of formulation enable determination of
the optimum dosage of water in concrete, it should
take into account the water contained or absorbed
by the aggregates to determine the effective water
content. The effective water content is defined in
the case of absorbing aggregates as the difference
between the total water present in the fresh concrete
and the water absorbed by aggregates.
In order to study the effect of saturation degree
of aggregate on the microstructure and mechanical
properties, wet and dry sand (respectively labeled
WS and DS) have been used. DS was obtained
by oven drying at 105 C until constant mass. WS
was pre-saturated in a plastic box and stored in a
room at 20 C for one week. The amount of water
added to the pre-saturation is equal the amount of
absorbed water plus 10%, so the water content is
equal to W24h +10%. The water absorption of NS
was determined by standard method (0.5%). How-
ever, in the case of RS, the water absorption was
calculated as the average of the water absorption
measured by the standard method 6.8% and the one
measured by evaporometry method 11.6% (Mech-
ling et al. 2003). So the water absorption of RS use
in here is 10%.
The relative amount of water WWS in wet sand is
calculated as:
MWS MDS
WWS = 100 · (1)
MDS
where MWS and MDS are respectively the weight
of wet and dry sand. The mortars were formulated
to have effective water to cement mass ratio (w/c)
of 0.5 and an identical volume of sand of around
60% in both mortars (sand to cement ratio of 3
based on dry natural aggregate) so any water in the
aggregate has to be subtracted from the water to be
used in batching the mortar. Mix proportions are
presented in Table 2.

Table 2: Compositions of the mortars with different moisture states of sand.

Natural sand mortar Recycled sand mortar

Moisture state of sand Dry state Over-saturated state Dry state Over-saturated state
Nomination NM Dry NM Sat RM Dry RM Sat
Cement (g) 450 450 450 450
Dry sand (g) 1350 1155
Wet sand (MWS ) (g) 1491.8 1386
Water absorption coefficient (%) 0.5 0.5 10 10
Effective amount of water (Me f f )1 (g) 225 225 225 225
Absorbed water (g) 6.8 6.8 115.5 115.5
Batch water (g) 231.8 90 340.5 109.5
1 effective water to cement mass ratio of 0.5.

311
 Le et al.
40⇥40⇥160 mm3 mortar bars were cast and
stored 1 day at 20 C and 95% relative humidity.
The specimens were demoulded the following day,
then cured under water at 20 C and tested at 2 and
28 days. Compressive strengths were measured
according to the European standard EN 196-1.
For the microscopical investigations, pieces of
hydrated samples of 1⇥2⇥2 cm3 were cut from
the core of mortar bars. Two main methods are
available to stop hydration and remove the water:
freeze drying and solvent exchange. The disad-
vantage of the freeze drying method is the pres-
ence of many micro cracks in the cement matrix
(Zhang & Scherer 2011). We chose the solvent ex-
change method using isopropanol-2 as solvent and
immersion of the sample for one week. The sam-
ples were then passed to vacuum to remove solvent
and stored in a desiccator with silica gels. After-
ward, they were impregnated using a low viscosity
epoxy and polished down to 0.25 µm using dia-
mond pastes. The samples were further coated
with carbon (⇠5 nm) and examined using a Quanta
200 FEG scanning electron microscope (SEM) from
FEI coupled to an Oxford INCA X-sight energy
dispersive X-ray spectroscopy (EDX) analyzer. The
sample is introduced into the chamber of the SEM
setting a mode high vacuum to avoid disruptions
and analytical distance of 10mm. We chose a volt-
age of 15kV for optimum observation through the
analysis of secondary and backscattered electrons.
A slow scanning speed was chosen (1-2 photos /
min) to capture an image. SEM is used in imaging
mode "backscattered electron diffraction" (BSE) for
studies of polished sections. This observation mode
provides better contrast between different phases
in the cementitious material (Figure 2).
Backscattered electron imaging has been used
to study the ITZ in mortars following the lead of
Scrivener et al. (1987). Around 30 BSE images of
a field 182 ⇥ 167 µm2 that corresponds to a resolu-
tion of 0.18 µm per pixel, were acquired per sample
which make it possible to detect elements with di-
mensions 0.03 µm2 . The fractions of anhydrous
cement, hydrates and porosity were evaluated in
8 bands of 10 µm in width, around 250 µm long
(Diamond 2001; Basheer et al. 2005; Elsharief et
al. 2003). The results were plotted against distance
from the aggregate and averaged over the 30 im-
ages. The relative area of each component in each
band was calculated from the grey level histogram.
The lower threshold level of pores has been set to
zero (black pixel). Several methods are available
for determining the thresholds on the histogram of
an image (Hemavibool 2007; Scrivener et al. 2004;
Brough et al. 2000). In this study, we used inflection
points on the graph of accumulated area depend-
ing on the intensity of the image to determine the
thresholds as proposed by Wong et al. (2006).
312
A visual comparison of the compatibility between
the image and the phase contrast image is first car-
ried out to validate the selected thresholds. To facili-
tate comparison, the images are enlarged. If there is
an underestimation or overestimation of the phases,
the thresholds are shifted until the isolated phases
are compatible with the base image (Figure 3). The
Delesse principle states that the two-dimensional
areas of profiles of components are related to the
three-dimensional volumes occupied in space by
these components, assuming random distribution
and random orientation of components (Delesse
1848). The porosity deduced by this method is lim-
ited by the pixel size, thus very fine capillary and
C-S-H pores cannot be separated and are included
in the area of hydrated products (Diamond 2001).
The distance measured from the aggregate surface
on a random section overestimates the true perpen-
dicular distance in three dimensions (Scrivener et al.
2004; Chen et al. 2007). It is difficult to correct the
distances unless measurements are made only on
interfaces that are normal to the image planes as in
(Head et al. 2008). In this study, the distance from
aggregate is the distance measured on 2D sections
without corrections. Considerations with regards
to image analysis and its limitations in the study
of the ITZ have been discussed in literature and
not detailed here (Scrivener et al. 2004; Wong et
al. 2006). When the thresholds are known, seg-
mentation phase is performed using a programmed
Matlab tool.
III. Results
Mechanicals properties
The mechanical tests are carried out after 2 and 28
days of curing. The results are presented in Figure
7.
The mortars based on natural sand present larger
compressive strengths than the mortars based on
recycled sand. The strength loss is almost 20 MPa
at 28 days when replacing NS by RS. This loss of
strength can be explained by a lower quality of
transition zone, a lower strength of RS compared to
the cement matrix and/or an incorrect estimation
of water absorption coefficient.
We can also notice that the initial saturation state
of aggregates has little influence on the mechanical
properties if we work with the same effective water-
cement ratio. This result was expected in the case
of natural sand with a low porosity, and is in agree-
ment with previous studies on sands of limestone
with low porosity (Nguyen et al. 2014).
The case of RS is, at first sight, more surprising.
Indeed, it had previously been shown in the case
of porous limestone aggregate (Nguyen et al. 2014),
that the effective water calculation underestimates
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 2: Backscattered image of mortar based on dry recycled sand at 2 days. In dotted line, the limit between old
cement paste of the RCA and the new cement paste.
the water released by the aggregates. This excess
water carries with it a more porous transition zone
and leads to a decrease in the mechanical prop-
erties of mortar based on saturated aggregates in
comparison to mortar containing dried aggregates.
SEM observation and image analysis
SEM images of mortar with dry and oversaturated
sand at 2 and 28 days, are presented in Figures 5
to 8. In the case of mortar based on natural sand,
the border between natural aggregate and cement
paste can easily be found; it is a natural siliceous
aggregate so the difference is very clear.
In the case of RM, at two days of hydration, the
degree of hydration of the new cement paste is still
low, but the old paste cement is mature, so we can
easily distinguish the boundary between the two
pastes. With RM-Sat at two days of hydration, a
wide layer of portlandite is formed on the surface
of the former paste. The duplex film zone is rich
in portlandite and has been described by previous
authors (Barnes et al. 1978; Ollivier et al. 1995;
Nguyen 2013)
At 28 days, in the case of RM-Dry, we cannot
distinguish anymore the boundary between the old
and new cement paste and plot the ITZ. However,
for RM-Sat, the layer of portlandite is still present
that permits to plot the boundary between the old
and new cement paste.
Figure 9 presents the average porosity of ITZ
after two days of hydration from around 30 images.
At two days of hydration, it is found in the case of
RM-Sat a significant decrease of porosity in the first
ten microns compared to the ITZ of other mortars.
However, when we get away from the surface of
the aggregates, the porosity is higher in the case of
mortar of over-saturated recycled sand.
Comparing both ITZ of mortars made from recy-
cled sand, the average porosity of the ITZ in both
saturation states is close (22.2% and 24.1% respec-
tively for RM-Dry and RM-Sat). However the state
of saturation, in the case of RM, significantly influ-
ences the distribution of the porosity in this zone.
In the case of mortars composed of NS (low
porosity sand), the average porosity and the poros-
ity distribution of ITZ in both saturation states,
dry and over-saturated, is identical. This result is
in agreement with previous studies on calcareous
sands with low porosity (Nguyen 2013).
Figure 10 presents the percent of anhydrous ce-
ment in ITZ after two days of hydration. The wall
effect can here clearly be observed. In fact, the
cement concentration increases with the distance
from interface. This observation concerns both ag-
gregates (natural and recycled sand) and both satu-
ration states (dry or over saturated).
However, this effect is less important for the re-
cycled sand at over-saturated state. This result may
be due to the draining of water out of recycled ag-
gregate that can push the cement grains away from
the surface of the aggregates.
At 28 days, the porosity of ITZ is decreased in
comparison to mortars after 2 days. The distribu-
tion of porosity in ITZ of NM is coherent with the
literature corresponding to low absorption aggre-
gates: high porosity in the first twenty microns, but
from twenty microns the porosity is reduced and
313
 Le et al.

T.Le, G. Le Saout, E.Garcia-Diaz, S.Rémond

ITZ base image Zooming in specific zone

Phase 1: Porosity Specific zone in the porosity phase

Phase 3: Anhydrous cement Specific zone in the anhydrous cement phase

Figure 3 : Comparison of specific zones between base images and binary images of porosity and anhydrous cement phases.
Figure 3: Comparison
Results
of specific zones between base images and binary images of porosity and anhydrous cement
phases.
Mechanicals properties
The mechanical tests are carried out after 2 and 28 days of curing. The results are presented in Fig. 7.
The mortars based on natural sand present larger compressive strengths than the mortars based on
stabilizes In therecycled
RM-Sat, sand.with the “duplex
The strength film”
loss is almost 20 MPaatat 28 days
landite at the NS
when replacing aggregate
by RS. Thissurface, but can
loss of strength after the poros-
the surface aggregate, a low porosity was found in
be explained by a lower quality of transition zone, a ity is increased.
lower strength of RS compared to the cement matrix and/or
an incorrect estimation of water absorption coefficient.
the first ten microns due to the presence of port-
We can also notice that the initial saturation state ofTable 3 presents
aggregates the average
has little influence porosity of ITZ from
on the mechanical
properties if we work with the same effective water-cement ratio. This result was expected in the case of natural

Figure 4: Compressive strength of mortars after 2 and 28 days of hydration (NM-dry: mortar based on dry natural
sand; NM-Sat: mortar based on over-saturated natural sand).

314
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 5: BSE images of mortar based on dry natural sand (left: 2 days; right: 28 days).

Figure 6: BSE images of mortar based on over-saturated natural sand (left: 2 days; right: 28 days).

Figure 7: BSE images of mortar based on dry recycled sand (left: 2 days; right: 28 days).

Figure 8: BSE images of mortar based on over-saturated recycled sand (left: 2 days; right: 28 days).

315

Table 3: Average porosity of ITZ
Nomination NM-Dry NM-Dry
2d
Average porosity of the ITZ at
(0-60µm) (%) 18.1
Average porosity of the ITZ at
(10-60µm)1 (%) 17.2
1: 10-60 micron bands zone from the aggregate, the first 10 microns are not taken in account
in the calculation of the average porosity.
0 to 60 µm and the average porosity of ITZ from 10
to 60 µm from the aggregate surface. The porosity
between 10 and 60µm does not take into account
the layer of portlandite that is present at the sur-
face of the old cement paste (duplex film). It can
be seen that the porosity between 10 and 60 µm
Figure 9: Average porosity of ITZ at 2 days with distance
from the aggregate.
Figure 10: Gradient of anhydrous cement in ITZ at 2
days.
Figure 11: Average porosity of ITZ at 28 days with
distance from the aggregate.
316
Le et al.
NM-Sat NM-Sat RM-Dry RM-Sat RM-Sat
28d 2d 28d 2d 2d 28d
8.3 16.8 10.7 22.2 24.1 14.8
7.2 16 9.7 21.9 26 15.7
is much larger when recycled aggregates are satu-
rated, probably because of a release of water from
the aggregates.
IV. Conclusion
A low influence of moisture condition of recy-
cled aggregates on the mechanical properties is
observed. We observe a little difference between
the compressive strength of both mortars, NM or
RM, with dry sand or over-saturated sand.
The ITZ of RM at two days of hydration has been
easily found with a visible difference between the
new and old cement paste. However, in the case of
RM-Dry at 28 days of curing, the new ITZ has not
been found because of a good continuity between
the old and new cement paste. The new ITZ of
RM-Sat at 28 days has been found with a layer of
portlandite at the surface of old cement paste.
With the RM, the moisture conditions of aggre-
gate influence significantly the porosity distribu-
tion in ITZ. A decrease of porosity in the first ten
micrometers is observed for saturated aggregates.
However, the average porosity of ITZ and the com-
pressive strength of the two mortars, RM-Dry or
RM-Sat, are identical, suggesting that the porosity
distribution in the ITZ does not affect significantly
the mechanical properties.
Some experiments are in progress and consist
to identify the adhesive “p” parameter and the ag-
gregate strength “q” parameter of RCA and NA
according to the De Larrard approach (De Larrard
1999). These identifications could help us to have a
better understanding of the local mechanical behav-
ior of these mortars and to answer the following
question: are the lowest mechanical properties of
RCA mortar in comparison to NA mortar mainly
due to a lower strength of RCA or to a more porous
interphase because of a higher average effective
water cement ratio?
Acknowledgement
Authors would like to thank project EcoReb and Na-
tional French Project for Research and Development
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
RECYBETON for providing us all raw materials and
for its financial support. The authors would like to
acknowledge J.-M. Taulemesse for his support dur-
ing the SEM experiments and C.Buisson, A. Diaz
for their experimental support in the laboratory.
References
Barnes, B.D., Diamond, S. and Dolch, W.L. (1978). The contact zone
between portland cement paste and glass “aggregate” surfaces. Cement
and Concrete Research, 8, pp.233–244.
Basheer, L., Basheer, P.A.M. and Long, A.E., (2005). Influence of
coarse aggregate on the permeation, durability and the microstructure
characteristics of ordinary Portland cement concrete. Construction and
Building Materials, 19(9), pp.682–690.
Blengini, G.A., Garbarino, E., Solar, S., Shields, D.J., Hamor, T., Vinai,
R., and Agioutantis, Z.(2012). Life Cycle Assessment guidelines for the
sustainable production and recycling of aggregates: the Sustainable Ag-
gregates Resource Management project (SARMa). Journal of Cleaner Pro-
duction, 27, pp.177–181.
Brough, A.R. and Atkinson, A. (2000). Automated identification of
the aggregate–paste interfacial transition zone in mortars of silica sand
with Portland or alkali-activated slag cement paste. Cement and Concrete
Research, 30(6), pp.849–854.
Chen, H., Wei, S., Piet, S. and Lambertus, J.S. (2007). Overestima-
tion of the interface thickness around convex-shaped grain by sectional
analysis. Acta Materialia, 55(11), pp.3943–3949.
Delesse, A. (1848). Procédé mécanique pour déterminer la composi-
tion des roches. Annales des Mines, 13, pp.379–388.
Diamond, S. (2001). Considerations in image analysis as applied to
investigations of the ITZ in concrete. Cement and Concrete Composites,
23(2-3), pp.171–178.
Du, T., Wang, W. H., Lin, H.L.,Liu, Z.X., and Liu. J. (2010). Experi-
mental Study on Interfacial Strength of the High Performance Recycled
Aggregate Concrete. Earth and Space 2010, pp.2821–2828.
Elsharief, A., Cohen, M.D. and Olek, J. (2003). Influence of aggregate
size, water cement ratio and age on the microstructure of the interfacial
transition zone. Cement and Concrete Research, 33(11), pp.1837–1849.
Etxeberria, M., Vazquez, E. and Mari, A. (2006). Microstructure anal-
ysis of hardened recycled aggregate concrete. Magazine of Concrete Re-
search, 58(10), pp.683–690.
Head, M.K., Wong, H.S. and Buenfeld, N.R. (2008). Characterising ag-
gregate surface geometry in thin-sections of mortar and concrete. Cement
and Concrete Research, 38(10), pp.1227–1231.
Hemavibool, S. (2007). The microstructure of synthetic aggregate pro-
duced from waste materials and its influence on the properties of concrete. Uni-
versity of Leeds.
De Larrard, F. (1999). Concrete Mixture Proportioning: A Scientific Ap-
proach, Spon (E&FN).
Mechling, J.-M., Lecomte, A. and Merriaux, K. (2003). Mesure de
l’absorption d’eau des additions minérales des bétons par évaporométrie.
Materials and Structures, 36, pp.32–39.
Nagataki, S., Gokce, A., Saeki, T., and Hisada, M. (2004). Assessment
of recycling process induced damage sensitivity of recycled concrete ag-
gregates. Cement and Concrete Research, 34(6), pp.965–971.
Nguyen, T.-D. (2013). Etude de la zone d’interphase“granulat calcaires
poreux-pâte de ciment”: influence des propriétés physico-mécaniques des gran-
ulats; conséquence sur les propriétés mécaniques du mortier. Ecole Nationale
Supérieure des Mines Saint-Etienne.
Nguyen, T.-D., Le Saout, G., Devillers, P., and Garcia-Diaz, E (2014).
The effect of limestone aggregate porosity and saturation degree on the
interfacial zone. In the 2nd International Symposium on Cement-based Ma-
terials for Nuclear Wastes. Avignon, France.
Ollivier, J.P., Maso, J.C. and Bourdette, B., (1995). Interfacial Transi-
tion Zone in Concrete. Advanced Cement Based, 2(1), pp.30–38.
Rasheeduzzafar and Khan, A., (1984). Recycled Concrete A Source
for New Aggregate. Cement, Concrete and Aggregates, 6, pp.17–27.
Scrivener, K.L., Crumbie, A.K. and Laugesen, P. (2004). The Interfa-
cial Transition Zone ( ITZ ) Between Cement Paste and Aggregate. inter-
face science, (12), pp.411–421.
Scrivener, K.L. and Gartner, E.M. (1987). Microstructural Gradients
in Cement Paste Around Aggregate Particles. MRS Proceedings, 114,
pp.77–86.
Scrivener, K.L. and Pratt, P.L. (1996). Characterisation of interfacial
microstructure. In J. C. Maso, ed. Interfacial Transition Zone in Concrete. E
& FN SPON, pp. 3–18.
Tam, V.W.Y., Gao, X.F. and Tam, C.M. (2005). Microstructural analy-
sis of recycled aggregate concrete produced from two-stage mixing ap-
proach. Cement and Concrete Research, 35(6), pp.1195–1203.
Wong, H.S., Head, M.K. and Buenfeld, N.R. (2006). Pore segmenta-
tion of cement-based materials from backscattered electron images. Ce-
ment and Concrete Research, 36(6), pp.1083–1090.
Xiao, J., Wengui, L., Yuhui, F., and Xiao, H. (2012). An overview of
study on recycled aggregate concrete in China (1996–2011). Construction
and Building Materials, 31, pp.364–383.
Zhang, J. and Scherer, G.W. (2011). Comparison of methods for
arresting hydration of cement. Cement and Concrete Research, 41(10),
pp.1024–1036.
Zhao, Z. (2014). Valorisations des fines de concassage de béton pour la
fabrication de mortiers. Université Lille 1.
317
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

318
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Assessment of the contaminants level in

recycled aggregates and alternative new
technologies for contaminants recognition
and removal
Somayeh Lotfi⇤,a , Francesco Di Maioa , Han Xiaa
Silvia Serrantib , Roberta Palmierib , Giuseppe Bonifazib
a Department of Materials and Environment, Delft University of Technology, Delft, The Netherlands
b Department of Chemical Engineering Materials & Environment, Sapienza University of Rome, Italy
⇤ s.lotfi@tudelft.nl

Abstract

One of the main challenging problems associated with the use of Recycled Aggregates (RA) is the level of mixed
contaminants. For utilizing RA in high-grade applications, it is essential to monitor and minimise the content of the
pollutants. To this extent the C2CA concrete recycling process investigates a combination of smart demolition, followed
by new innovative technologies to produce high-grade secondary aggregates with low amount of contaminants. This
paper firstly reports the level of contaminants in different fractions of recycled aggregates coming from a real case
study. Results show that the wood content of 4-16 mm recycled aggregates is well within the strictest limit of the EU
standard. However, there are still large visible pieces of wood and plastic in the +16 mm RA fraction which, albeit
within the standards, does not satisfy the users. In order to solve this problem the feasibility of applying two existing
technologies (near infrared sensor sorting and wind sifting) to remove contaminants, is studied. Furthermore, two types
of online quality control sensors (hyper spectral imaging and laser induced breakdown spectroscopy) are introduced and
a summary of their recent developments towards the quality control of RA are presented.

Keywords: construction and demolition waste (C&DW), recycled aggregate, contaminants, quality control

I. Introduction tural concrete containing RA with high chloride

The efficient high-grade recycling of Construction
and Demolition Waste (CDW) is of increasing in-
terest from an environmental and economic per-
spective. From an environmental point of view, the
urgency of saving resources and reducing human-
ity’s impact on the environment is evident. The
need of increasing recycling and improving the
quality and homogeneity of recycled materials to
minimize environmental pollution and the use of
primary resources is a topical subject for European
Community (Enterprise and Industry reports of the
EC).
In order to enhance the quality of RA for high-
grade applications, the content of contaminants
such as organic materials (wood, plastic and foams),
gypsum and glass must be minimised (Vegas, et
al. 2015). Many organic substances such as wood
are unstable in concrete when submitted to dry-
ing and wetting or freezing and thawing (Hansen,
1990). Water-soluble sulphates (coming from gyp-
sum plaster) in RA are reactive and may produce
expansive reactions (Silva, et al. 2014) while struc-
content may deteriorate more rapidly due to the
corrosion of reinforcement bars (RA coming from
concrete subjected to marine may have high soluble
chloride content). Plate glass from windows has
the density similar to the stone’s and brick’s and
therefore it complicates its separation. Thus pre-
sorting of the glass is essential also because of the
alkali- silica reactions which can take place due to
non-crystalline metastable silica (Hansen, 1990).
Considering the importance of upgrading the
quality of the RA and removing the contaminants,
currently different technologies and procedures
such as smart demolition and dismantling of End-
Of-Life (EOL) buildings, automated sensor sorting
and online quality control sensors have been de-
veloped (Serranti, et al. 2012) and (Palmieri, et
al. 2014) and (Xia, et al. 2014). A novel con-
crete recycling process developed within an Eu-
ropean funded project (Lotfi, et al. 2015), aims at a
cost-effective system approach for recycling high-
volume EOL concrete streams into high-quality
aggregates and cement. The best practices and
technologies implemented are smart demolition to

319
 Lotfi et al.

produce crushed concrete with low levels of con-

taminants, followed by mechanical upgrading of
the material on-site into an aggregate product and
a cement-paste concentrate that can be processed
(off-site) into a low-CO2 input material for new
cement production. Sensor-based on-line quality
assurance allow for a proper monitoring of the out-
put. Achieving in-situ recycling of the EOL concrete
is one of the main goals of this process. Therefore,
the liberation of the cement paste as well as the
sorting and size classification of the aggregate, is
performed purely mechanically and in the moist
state, i.e. without prior drying or wet screening.
This choice reduces process complexity and avoids
problems with dust or sludge while providing eco-
nomic benefits in terms of process costs and logis-
tics. After crushing, liberation of the cement paste
is promoted by several minutes of grinding in a
small-diameter (D = 2.2 m) autogenous mill and
at the same time producing as little as possible
fine silica. Then a new low-cost classification tech-
nology, called Advanced Dry Recovery (ADR) is
applied to remove the fines and light contaminants
with an adjustable cut-point of between 1 and 4
mm for mineral particles. ADR uses kinetic energy
to break the bonds that are formed by moisture
and fine particles and is able to classify materials al-
most independently of their moisture content. After
breaking up the material into a jet, the fine parti-
cles are separated from the coarse particles. ADR
Figure 1: General layout of the C2CA technology show-
separation has the effect that the input aggregate is
ing different steps for contaminants removal. Two devel-
concentrated in two main streams: a coarse aggre-
oping sensors (HSI and LIBS) are being developed for
gate product and a fine fraction which includes the
the quality control of the products.
cement paste and contaminants (e.g. wood, plastics
and foams).
In the current study, the influence of different re- III. Smart dismantling and selective
cycling steps in C2CA process (See Figure 1) in the demolition
level of contaminants is investigated and solutions
to make clean final products are presented. The aim An EOL building may be conventionally or selec-
is to enrich our understanding of the importance of tively demolished. Although the construction and
existing recycling steps with respect to eliminating demolition industries still see the concept of the
the contaminants in recycled aggregate. selective demolition doubtful from economic point
of view, it may be more profitable than the conven-

II. End of life building (case study in

Groningen)

The case study of the C2CA project involved the de-

molition of a governmental complex in the province
of Groningen in the Netherlands and the building
of an underground garage from concrete with recy-
cled aggregate. The scope of the demolition part of
the project mainly consisted of two identical high-
rise towers (KB2 and KB6) with the blue dotted line
in Figure 2. The section plan of the towers can be
seen in Figure 3. Figure 2: Overview of the end of life buildings

320
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 4: Material composition (wt%) of KB2/KB6

Figure 3: Section plan of the EOL building tower.

tional demolition approach and the most effective

way of minimising the amount of contaminants in constituents. Samples were portioned to 5-10 mm,
CDW materials (Coelho, et al. 2011). 10-20 mm and +20 mm fractions and their contami-
In the 70’s and 80’s the Dutch construction sector nants were hand-picked according to the procedure
used asbestos in the buildings. Therefore prior to previously explained.
the dismantling and demolition of KB2 and KB6
Figure 6 shows the mass percentage of the hand-
buildings, asbestos was removed and collected in
picked contaminants from mill input and output for
the total amount of 40 tons. The further strategy
three aforesaid fractions. Considering the results
for the dismantling of the KB2 and KB6 involved
after milling, the mass of contaminants (bigger than
the detailed removing of all materials from the
5 mm) is reduced by 30%. It appears that by milling,
concrete skeleton before starting the demolition:
contaminants are broken down in smaller parts so
air-conditioners, radiators, lamps, piping systems
that the less than 5 mm fraction increases. There is a
of water and heating, electric cables, carpets, gyp-
clear effect of the milling also on the size reduction
sum plates from ceilings and walls, window glass,
of brick (compare Figures 6-A with 6-D). A similar
frames of doors and windows etc. For the demo-
trend can be seen for wood contaminants albeit in a
lition, two methods were applied: the top-down
less outstanding way. In general it is observed that
method to demolish the top 12 floors, and short-
milling has an obvious effect on the size reduction
reach method to demolish the lowest 2 floors of
of brick and a slight effect on size reduction of other
the towers. The materials composition of KB2/KB6
contaminants.
tower can be seen in Figure 4. It demonstrates
that the amount of EOL concrete was 87wt% of the
whole CDW materials.

IV. Autogenous milling

Autogenous milling of the crushed End of Life
(EOL) concrete is a mechanical method to remove
cement paste from the surface of aggregates. Dur-
ing autogenous milling, the combination of shear-
ing and compression forces, promotes selective at-
trition and delivers a better liberation of cement.
Beside liberation of cement, the acting forces could
affect the size of the contaminants. In this regard,
a batch test with around 15 tons of crushed EOL
concrete was carried out. The residence time of
the materials inside of the mill was estimated as 12
minutes. About 2 tons of sample from both mill Figure 5: Constituents of the contaminants mixed with
input and output was taken in order to analyse the concrete.

321
 Lotfi et al.
V. Screening and ADR processing
In the C2CA concrete recycling process, autoge-
nous milling of the materials is followed by a 16
mm screen and an ADR. Materials smaller than 16
mm are fed into the ADR and using a jet, the fines
(0-4 mm) are separated from the coarse particles.
The air knife installed in ADR helps to concentrate
contaminants like wood, foam and plastic in the
fines. In order to figure out the amount of wood
in the coarse ADR products (4-16 mm RA) a sink
floating test on approximately 1700 kg of RA and ac-
cording to EN 12620 (for application of the material
as coarse recycled aggregate) was carried out.
Result shows that the total amount of floating
wood in 4-16 mm RA is almost 0.117 cm3/kg which
is well within the strictest norm of EU standard
(EN 12620). ADR separation has the effect that the
aggregate is concentrated into a coarse aggregate
product and a fine fraction including the cement
paste and contaminants such as wood, plastics and
foams (see Figure 7B). ADR fines can be used as the
input of cement kilns so that the wood and plastics
contaminants are even beneficial for the process.
During the first C2CA case study, it became clear
that because of the contamination, the +16 mm frac-
tion does not have the market potential as such (see
Figure 7A). Therefore it would impair the economic
attractiveness of the recycling process being devel-
Figure 6: A, B and C show the amount of contaminants in the +5mm input of the autogenous mill. D, E and F show
the amount of contaminants in the +5mm output of the autogenous mill.
322
oped. According to the visual evidence, there are
big contaminants of wood and non-ferrous metals
in +16 mm oversize fraction. In order to satisfy
the customers’ demands and use +16 mm RA for
high-grade applications, contaminants should be
removed from this fraction. Since +16 mm RA is a
small stream (almost 30wt% a sensor sorter could
be a cost effective option to clean this fraction. In
this regard, the possibility of applying Near In-
frared (NIR) sensor sorting technology to clean +16
mm RA was examined.
NIR sensor sorting and wind sifting to remove
contaminants from +16mm RA
Advanced automated sensor-based sorting tech-
nologies use physical – chemical properties of differ-
ent materials such as density, electrical conductivity
or magnetic susceptibility, as well as surface and
material properties, such as NIR spectrum or the
visible colour (Vegas, et al. 2015). For this part of
study, NIR sensor sorting facility of TOMRA GmbH
sorting in Germany was used (Figure 8 shows the
functional principle). Input material (1) is evenly
fed onto a conveyor belt, where it is detected by
the NIR and/or VIS spectrometer (2). If the sensors
detect material to be sorted out, it commands the
control unit to blow the appropriate valves of the
ejection module at the end of the conveyor belt. The
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 7: A) +16mm RA which still contain contaminants, B) ADR fines in which light contaminants like wood and
foam are concentrated.
detected materials are separated from the material
flow by jets of compressed air. The sorted mate-
rial is divided into two or three fractions in the
separation chamber (3).
For testing the performance of the NIR sen-
sor sorting system, around 900 kg of +16 mm of
crushed concrete was delivered to TOMRA sort-
ing GmbH. During the experiment, contaminants
(wood, plastic and metal) with size of 10-20 mm
and +20 mm were added to the clean crushed con-
crete. The output of the NIR sensor sorting system
consists of an accepted portion (clean concrete) and
ejected contaminations. Figure 9 shows an example
of the input contaminants to the NIR sensor sorting
system and accepted and ejected outputs.
The process flow diagram for 20 t/h/m of the
throughput shows that by using NIR sorting, almost
88% mass% of wood is thrown out and small pieces
are left in the product. Plastic and metal are also
removed (see Table 1 and 2).
The investment cost of a NIR sorting system (in-
cluding high speed conveyor and separation cham-
Figure 8: Functional principle of TOMRA NIR sensor
sorting technology.
ber) vary depending on the width of the system
between approximately e95,000 (0.6 m width) to
roughly e230,000 (2.8 m width). Considering the
investment costs for other equipment which are
needed to operate a NIR sorting system (com-
pressor, cables, electrical equipment, conveyors
for output fractions, installation) total investment
cost would be e140,000 (0.6 m wide system) and
e320,000 (2.8 m wide system). It is estimated that
for a 2 meter width of the belt (40 t/h), with run-
ning time of 1600 hours/year the cost of processing
will be 1.2 euro/ton.
Considering the above mentioned results, the
only limitation with NIR sensor sorting system is
its current inability to remove small contaminants
which therefore remain in the accepted output. To
reach a more accurate contaminants removal, the
combination of the wind sifting technology with
NIR sorting is considered beneficial. In this regard,
a test was carried out to examine the performance of
a wind sifter to remove plastic and wood. The test
was performed at REDOX B.V. in the Netherlands
(see Figure 10). The input materials of the wind
sifter consisted of almost 500 kg of +16mm crushed
concrete mixed with a specific amount of wood and
plastic.
According to the process flow diagram (see Table
3 and Table 4), it is concluded that wind sifting
is able to remove 80% (by number) of wood and
plastic contaminants from the stream (for 35t/h
Table 1: Mass analysis of NIR input.
Materials Weight(gr) Mass,%
Clean crushed concrete 278000 99.87
Wood 230 0.08
Plastic 121 0.04
Total 278351 100
323
 Lotfi et al.
Table 2: Mass analysis of NIR output
Materials
Crushed concrete in accepted materials
Crushed concrete mixed in ejected materials
Wood in ejected materials
Plastic in ejected materials
Total
throughput ). Big particles which are heavier can-
not be removed and remain in the final products.
In Figure 11 the remained big pieces of wood in the
final product can be seen.
The investment cost of the REDOX wind sifter
is e75,000 euro. It is estimated that for a 40 t/h
throughput, with running time of 1600 hours/year
the cost of wind sifting process will be 0.1 euro/ton.
VI. Sensors for On-line quality
control
LIBS
Laser Induced Breakdown Spectroscopy (LIBS) is an
optical spectroscopic technique employing a pulsed
laser to produce a high power density light beam
(>108 W/cm2 ) to ablate tiny amounts of material
from target material surface, resulting locally in
high dissipation and accompanying breakdown of
molecules or crystalline structures into a partly ion-
ized plasma with plasma temperature and elec-
tron density typically in the order of 104 K and
1018/cm3, respectively. During the cooling-off of
the plasma, the fingerprint LIBS spectra including
element specific atomic-ionic emission lines can
be observed, whereby photon wavelengths may be
linked to specific elements, allowing identification
of the elements in the plasma which represents the
target sample material. The number of photons pro-
duced by a certain type of element may in principle
be linked to the element concentration therein. The
LIBS technique has great potential on real-time pro-
cess and in-situ quality control: it has relative sim-
Figure 9: Input contaminants and accepted and ejected output of NIR sensor sorting system.
324
Weight(gr) Mass, %
267900 96.24
10136 3.64
203 0.07
123 0.04
278362 100
ple instrumentation, fast measurement with only
optical access, needs minute sample preparation
etc. The LIBS experimental setup being developed
within C2CA is shown in (Figure 12). This set- up
consists of a 1064 nm Q-switched YAG-laser (11-25
mJ, 10 ns/pulse, 1-100 pulses/s), fibre-optics and
focusing assembly to collect the light, and a spec-
trograph with attached CCD to disperse and detect
the photonic emissions. The timing of the whole
system is facilitated by a waveform generator.
One of the quality control steps within C2CA, is
inspection of drill cores of concrete coming from
the EOL building. Drill cores usually contain infor-
mation on the type of materials used in the concrete
(cement, sand, granulate, rebar), and the possible
ingress of chlorines (in case of outside exposure) or
degree of carbonation. The possibility to inspect
drill cores and detect these properties in-situ consti-
tutes an advance on present practices, since the dis-
mantling and demolition process may be adapted
according to the information to produce the highest
quality secondary raw materials for recycling.
The laboratory LIBS set-up has been shown to
be sensitive enough to perform semi-quantitative
surface mappings which enable visualization of el-
emental distributions of various elements. Figure
13a shows the grayscale LIBS raster scanning results
on one of the drill cores (no.7-from the pillar wall
in the room 2.03 of the KB2 building) measured in
the air compared with the photograph. Starting at
the surface, the drilling core was linearly scanned.
Each measurement was made per millimetre av-
eraged over ten single laser shots. The distance
between two lines was 1 mm. Each spectrum was
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 10: Wind sifting facility in REDOX.

Figure 11: Input of wind sifter(upper images) which consists of wood, plastic and clean +16 mm crushed concrete.
Downer images demonstrate the removed wood and plastic using wind sifter and remained wood in the products.

Table 3: Mass analysis of the wind sifter input.

Input weight(gr) mass,% number of particles

clean crushed concrete 524000 99.90
wood 324 0.06 134
plastic 195 0.04 48
total 524520 100

325
 Lotfi et al.
Table 4: Mass analysis of the wind sifter output.
output
crushed concrete in accept materials
crushed concrete in eject
wood in eject
plastic in eject
total
background subtracted and was normalized to the
whole spectral integral against shot-to-shot varia-
tions. Ca (Ca 422.4 nm) was abundant in the cement
paste, while the Si (288.1 nm) and Fe (274.4 nm) in
aggregate and rebar, respectively. In particular, the
inverse relation between the Ca and the Si content
can be observed by the comparison between aggre-
gate and cement. Si, Al (Al 309.1 nm), Fe, K (K
766.3 nm), Mg (Mg 285.0 nm), Li (Li 670.5 nm), Na
(Na 589.0 nm), S (S 922.0 nm) were also present
in cement paste. Furthermore, there was higher
content of Si in the white (more like quartz) aggre-
gate than that in the dark green aggregate and vice
versa for Al and Na. C (C2 516.6 nm) was observed
in both cement paste and rebar. O (O 777.2 nm)
and H (H 656.3 nm) were mainly in aggregates,
possibly as combined H2 O. Cl (Cl 837.7 nm) was
mainly detected in the dark green aggregate and
rebar. It should be emphasized that the images
show the relative distribution of the element, and
not the absolute concentration. Comparison with
the photograph of the scanned drill core surface
proves that cement paste, coarse aggregate and re-
bar were clearly differentiated. The finer gravel
or sand could also be distinguished using higher
scanning resolution.
To facilitate the smart dismantling at earliest
stage, it was attempted to differentiate between ag-
gregate, cement paste and rebar of the drill cores us-
Figure 12: Laboratory LIBS experimental setup.
326
weight(gr) mass, % number of particles
513390 97.88
10790 2.06
209 0.04 107
130.6 0.02 39
524520 100
ing LIBS. Successful classification of different types
of aggregate, sand and cement paste, as well as re-
bar using PCA and neural network algorithms were
achieved (see Figure 13b). In addition the area ra-
tios of the different materials could be determined
(Figure 13c). Drill core concrete samples taken from
the building site before demolition at the start of the
C2CA project have been successfully categorized
according to the used cement types. As a represen-
tative of each group, drill core no. 7 (Portland) had
a higher content of Ca while no. 15 (blast furnace
slag cement, from the floor in the room 0.36 of the
KB6 building) had higher contents of Si, Al, Fe and
Mg (see Figure 14). Here, the LIBS signal intensity
for each element is normalized to the summation
over those of all elements and hence, is not pro-
portional to its elemental content. These findings
will be instrumental for smart dismantling, because
it helps to determine which parts of the concrete
structure may be demolished together and which
parts should be kept separate to obtain concrete
batches of known and consistent quality.
The potential capability of the LIBS system to
detect and characterize the waste pollutants in a
real-scale setting has been investigated in (Xia, et
al. 2014). The classification of eight waste materials
with selected types from a stream of demolition
concrete produced using their LIBS spectra com-
bined with principal component analysis (PCA) is
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 13: LIBS grid scanning 20⇥20 mm2 on drill core no. 7 in 1 mm steps. The laser spot diameter is 300 µm. (a)
2D distribution of Ca, Si, Al, Fe, K, Mg, H, Li, C, O, Na, S and Cl. The data are normalized to unity, corresponding
to white in a grey scale. (b) classification of rebar, cement paste and aggregate/sand using PCA-neural network
algorithm, corresponding to 1, 0.5 and 0 in the grey colour bar, respectively; (c) cumulative number of shots for each
material, indicating the area ratios.

presented in Figure 15. The representative materi-
als for most of the materials found in demolition
concrete were: cement (CEM I 42.5 R HS, CEM
II/B-S 52.5 N, CEM III/B 32.5 N), brick (yellow,
brown and red), gypsum block (white, blue and
red), wood (pine, ash and walnut), PVC plastic
(grey, black pipe and grey hard plate), glass (white,
green and brown), two of the most common types
of natural rock used as aggregate (sandstone and
gabbro) and steel rebar, of which only one type is
used in the Netherlands. All materials could be
successfully classified out of a random (data) mix-
ture of these materials using partial least squares
discriminant analysis (PLS-DA). As a percentage of
the tested data set, the misclassifications made up
0.28% (single-shot), 0.14% (2 fold averaged), 0.56%
(4 fold averaged) and 0% (10 fold averaged).

HSI

An Hyperspectral Imaging (HSI) approach, acting

Figure 14: Element-wise comparison of drill cores 7
and 15. Note that the y-axis adds up to unity for each
element.
in the NIR range (1000-1700 nm) was adopted as
quality control tool for the recovered products (i.e.
recycled aggregates). More in detail, a system able
to recognize concrete aggregates and unwanted
contaminants, such as brick, wood, plastic, gypsum
and foam after the ADR processing was developed,
implemented and set up.
Utilized HSI platform, realized by DV srl
(Padova, Italy), is an integrated hardware and soft-
ware architecture allowing to digitally capture and
manage spectra, as they result along a pre-defined
alignment on a surface sample properly energized
(Hyvarinen, et al. 1998) and (Geladi, et al. 2007).Hy-
perspectral data are called “hypercube” because of

327
 Lotfi et al.

Figure 15: PCA score plot of waste materials using the first three PCs. Indicated in the axis label is the PC contribution
to the total variance.

their 3D structure, characterized by two spatial and
one spectral dimensions. For each pixel of the hy-
perspectral image, a full discrete is obtained being
the wavelength bands typically an equally spaced
sequence (Geladi, et al. 2004). Spectral features
are related to the chemical-physical properties of
samples, allowing a full characterization. Spectral
adsorption mechanisms usually affecting vegeta-
tion, mineral, chemical products, etc. also affect
secondary raw materials as generated in an Ur-
ban Mining context. In this latter case alterations,
due to “life-time” related constraints of both ma-
terials and/or products, as well as, the frequent
presence of “composite”, dramatically influence de-
tected spectra characteristics, not allowing and easy
identification of specific absorption features. Fur-
thermore, man-made are a source, at the origins, of
chemical absorptions which are not readily found
in natural materials. Finally, weather alteration can
modify the original material and as a consequence
their spectra (Roberts, et al. 2004). For all these
reasons, acquired spectra interpretation is difficult,
but it is simply possible to relate most absorption
bands to C-H, O-H and N-H stretching vibrations
specific for the NIR range.
Pollutant samples (i.e. wood, plastic, gypsum,
brick and foam), provided from C2CA case study,
were representative of some contaminant materials

Figure 16: Digital image of Experimental set up 1 (a) and Experimental set up 2 (b).

328
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 17: Acquired spectra of the six different analysed materials.

collected after demolition of a building in Gronin-
gen (NL). RA samples are 4-16 mm ADR products.
Analysed hyperspectral images were acquired in
the 880-1720 nm wavelength range, with a spectral
resolution of 7 nm, for a total of 121 wavelengths.
In order to remove effects due to the background
noise, a preliminary reduction of investigated wave-
lengths was applied and spectral variables were
thus reduced from 121 to 93. The investigated wave-
length range was from 1006 nm to 1650 nm. The
image width was 320 pixels, while the number of
frames was variable according to the length of the
desired acquisition.
Two different hyperspectral image sequences
have been thus acquired in order to build the clas-
sification model and to validate it. A first and a
second series, respectively identified as Experimen-
tal set up 1 and the Experimental set up 2 (Figure
16). Twelve particles clearly identified as brick (2
particles), aggregates (3 particles), wood (1 particle),
gypsum (3 particles), foam (3 particles) and plastic
(4 particles) were arranged in lines, forming the Ex-
perimental set up 1 used to build the classification
model. Experimental set up 2 was used as valida-
tion data set: particles were arranged in six groups
according to their class membership.
Looking at the acquired raw spectra (Figure 17)
related to the Experimental set up 1, differences in
the spectral behaviour are visible. In order to high-
light them, a combination of three pre-processing
algorithms (Detrend, Standard Normal Variate and
Mean Center) is applied, the obtained spectra are
shown in Figure 18.
After pre-processing, Principal Component Anal-
ysis (PCA) was applied to the image corresponding
to the Experimental set up 1 in order to explore

Figure 18: Pre-processed spectra of the six different analysed materials.

329
 Lotfi et al.
Figure 19: PCA score plot: Experimental set-up 1 pixels projection onto the space of PC1 vs PC2.
the data. PCA projects the samples into a low
dimensional subspace, whose axes (the principal
components, PCs) point in the direction of maxi-
mal variance, compressing the data. Looking at
the distribution of samples into the PCs space, it is
thus possible to interpret differences and similari-
ties among them: the more they are grouped, the
more they have similar spectral features. Further-
more, PCA highlights the trends among samples,
giving preliminary information about distribution
of different specimen on an image (Amigo, et al.
2013). The obtained score plot is shown in Figure
19: six different clusters, corresponding to the differ-
ent materials, are recognizable onto the score plot.
Therefore the training dataset was easily created
by removing some border cluster points “different”
from all other spectra of the same category and by
setting class of the remaining pixels.
Partial Least-Square Discriminant Analysis
(PLSDA) was adopted in order to build a predictive
model able to classify the different specimen in the
image: recycled aggregates and contaminants. This
technique is a linear method based on the partial
least squares regression requiring prior data knowl-
edge: it is a supervised classification. The PLS-DA
model allows to assign to each unknown sample
(in this case, pixel) only one of the available defined
classes, according to the similarity among spectra.
Prediction maps are PLS-DA results: each class is
defined by a different colour. In Figure 20 PLS-DA
prediction map of Experimental set up 2 is shown.
A good classification is obtained and it is easy
to associate each object in the image to a specific
category, but some sporadic misclassified pixels are
visible. This phenomenon can be attributed to the
330
rough surface of the samples and the consequent
light scattering effect. Moreover, it is necessary to
take in account the presence of dust particles, due to
the “dirty” nature of the samples, that can influence
the analysis because of their own spectrum.
HSI based approach allows to develop an objec-
tive, fast and non-destructive method in order to
control the quality of ADR products. It is possible
to discriminate between recycled aggregates and
contaminants in the ADR outputs, evaluating the
pollutant content in order to monitor the entire re-
cycling process. The utilization of the proposed
approach is preferred in the secondary raw mate-
rials sector, where expensive and/or sophisticated
quality control devices cannot be practically pro-
posed, both for technical (i.e. particles of different
size, shape and composition, etc.) and economic
(i.e. environmental constraints, maintenance, etc.)
reasons.
VII. Conclusions
According to the results the following conclusions
can be drawn out:
• Smart dismantling and selective demolition
is the most important step to minimize con-
taminants in RA. Results show that although
the wood content of 4-16 mm recycled aggre-
gates is well within the strictest limit of the EU
standard, the still visible pieces of wood and
plastic in the +16 mm RA fraction reduces the
economic potential of the RA. Using selective
demolition in this study, the amount of gyp-
sum and wood in the crushed EOL concrete
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 20: Prediction map based on PLS-DA model built for the classification of foam (1), gypsum (2), wood (3),
aggregates (4), brick (5) and plastic (6) in the Experimental set up 2.
was reduced to almost 60 ppm and 2 cm3 /kg
respectively.
• Autogenous milling reduces the mass percent-
age of contaminants bigger than 5mm by 30%.
There is a clear effect of the milling on the
size reduction of brick. A similar trend can
be seen for wood contaminants but in a less
outstanding way.
• The amount of wood in 4-16mm ADR product
was measured as 0.117 cm3 /kg which is well
within the strictest norm of EU standard.
• It is revealed that the combination of two tech-
nologies (NIR sensor sorting and wind sifting)
will remove most of the contaminants from +16
mm RA.
• Two developing sensors (LIBS and HSI), should
allow on-line methods for quality control and
quality assurance of the concrete recycling
products. The concept is to avoid the need
for laboratory analysis and intermediate stor-
age and if possible quality and end- of-waste
certification at the site without human inter-
vention. Recent investigations related to the
aforesaid sensors show their high potential to-
wards achieving the mentioned goals.
Acknowledgements
This research is funded by the European Commis-
sion in the framework of the FP7 Collaborative
project “Advanced Technologies for the Production
of Cement and Clean Aggregates from Construction
and Demolition Waste (C2CA)”. Grant Agreement
No.265189.
References
Amigo J.M, Marti I, Gowen A, Hyperspectral imaging and chemomet-
rics: a perfect combination for the analysis of food structure, composi-
tion and quality, in Federico Marini (Ed.), Data Handling in Science and
Technology, Vol.28, Amsterdam: The Netherlands, 2013, pp. 343-370.
Coelho A, de Brito J. Economic analysis of conventional versus selec-
tive demolition—A case study. Resources, Conservation and Recycling.
2011;55:382-92.
Communication from the Commission to the European Parliament
and the CouncilThe raw materials initiative- meeting our critical needs
for growth and jobs in Eroupe (COM(2008) 699 final.
EN 932–1. Tests for general properties of aggregates. Part 1: Methods
for sampling.
EN 12620. Aggregates for concrete
Geladi P, Grahn H, Burger J, Multivariate images, hyperspectral imag-
ing: background and equipment, in H. Grahn and P. Geladi (Eds.), Tech-
niques and Applications of Hyperspectral Image Analysis, Jhon Wiley &
Sons, West Sussex, England, 2007, pp. 1-15.
Geladi P, Burger J, Lestander T, Hyperspectral imaging: calibration
problems and solutions, Chemometr. Intell. Lab. 72 (2004) 209–217.
Hansen TC. Recycling of demolished concrete and masonry: Taylor
& Francis; 1990.
Hyvarinen T, Herrala H, Dall’Ava A, in: Proceedings of SPIE Elec-
tronic Imaging, 3302, San Jose, CA, USA (1998).
Lotfi S, Deja J, Rem P, Mróz R, van Roekel E, van der Stelt H. Mechan-
ical recycling of EOL concrete into high-grade aggregates. Resources,
Conservation and Recycling. 2014;87:117-25.
Palmieri R, Bonifazi G, Serranti S. Automatic detection and classifica-
tion of EOL-concrete and resulting recovered products by hyperspectral
imaging. SPIE Sensing Technology+ Applications: International Society
for Optics and Photonics; 2014. p. 91060D-D-14. Roberts D. A, Herold M,
Imaging spectrometry of urban materials, in King, P. Ramsey, M.S. and
G. Swayze (Eds.), Infrared Spectroscopy in Geochemistry, Exploration
and Remote Sensing, Mineral Association of Canada, Short Course Se-
ries Volume 33, London, Ontario, 2004, pp. 155-181.
Silva RV, de Brito J, Dhir RK. Properties and composition of recycled
aggregates from construction and demolition waste suitable for concrete
production. Construction and Building Materials. 2014;65:201-17.
Serranti S, Gargiulo A, Bonifazi G. Classification of polyolefins from
building and construction waste using NIR hyperspectral imaging sys-
tem. Resources, Conservation and Recycling. 2012;61:52-8.
Vegas I, Broos K, Nielsen P, Lambertz O, Lisbona A. Upgrading the
quality of mixed recycled aggregates from construction and demolition
waste by using near-infrared sorting technology. Construction and Build-
ing Materials. 2015;75:121-8.
Xia H, Bakker MCM. Reliable classification of moving waste materials
with LIBS in concrete recycling. Talanta. 2014;120:239-47.
331
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

332
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Microstructure development of a drying

tile mortar containing
methylhydroxy-ethylcellulose (MHEC)
A.P.A. Faiyasa,b , S.J.F. Erichb,c , T.G. Nijlandc , H.P. Huininkb and O.C.G. Adanb,c
a Materials Innovation Institute, Mekelweg 2, P.O.Box 5008, 2628 CD Delft, The Netherlands
b Department of Applied Physics, Eindhoven University of Technology, P.O.Box 513, 5600 MB Eindhoven, The Netherlands.
c Netherlands Organization for Applied Scientific Research (TNO), P.O. Box 49, 2600 AA Delft, The Netherlands

Abstract

Cement based mortars are widely used as adhesive for tiles in building and construction. They have a limited
timespan during which a tile can be placed effectively in order to develop sufficient bond strength. This timespan,
usually called ’open time’, is controlled, amongst others, by adding water soluble polymers such as methyl cellulose
or modified cellulose ether like methylhydroxy-ethylcellulose (MHEC) to the mortar. MHEC influences the drying
behavior of the mortar. As such, it may be expected to influence the final microstructure of the tiling mortar. The paper
reports the results of a combined scanning electron microscopy and nuclear magnetic resonance imaging study on the
development of the microstructure of hardening mortar.

Keywords: : tiling mortar, methylhydroxy-ethylcellulose, PFM, SEM, NMR

I. Introduction of MHEC, the following samples have been studied:

Cementitous tiling mortars are widely used in the
building and construction industry to apply ce-
ramic tiles on walls etc. Besides cement, water and
aggregate, such tiling mortars usually contain ad-
ditives like methylhydroxy-ethylcellulose (MHEC).
MHEC is a water soluble cellulose ether, synthe-
sized from cellulose by substituting methyl and
ethylene oxide groups by an etherification process.
Such water soluble polymers are added to control
the so-called ’open time’, i.e. the limited timespan
during which a tile can be placed effectively in or-
der to develop sufficient bond strength. Though
several studies have been published on the effect
of MHEC on kinetics of cement hydration (Müller
2006, Pourchez 2006), water retention, transport
and microstructure (Pourchez et al. 2010, Wetzel et
al. 2011, Marliere et al. 2012), viscosity (Faiyas et al.
2015) and skin formation (Bühler et al. 2013), the
working principle and effect of MHEC on the de-
velopment of materials properties of tiling mortars
is still not properly understood. The current study
focusses on the effect of MHEC on the microstruc-
ture development of tile mortars by a combination
of optical polarizing-and-fluorecence microscopy
(PFM), scanning electron microscopy (SEM) and
nuclear magnetic resonance imaging (NMR).
In order to get a better understanding of the effect
1) The effect of the presence of MHEC on the mi-
crostructure of cement paste and mortar in closed
systems. 2) The influence of increasing MHEC con-
tent on the air void content of the hardened mi-
crostructure in closed systems, studied by optical
microscopy and NMR, and 3) The relationship be-
tween the presence of MHEC on microstructure and
drying behaviour in open and closed systems, stud-
ied by optical and scanning electron microscopy
and NMR.
II. Sample preparation
Cement paste samples for experiment 1
Cement pastes were made with CEM I 52.5R only.
Pastes were made with 0 and 0.5 wt.% MHEC. A
commercial grade MHEC (Tylose MHS 300000 P6)
provided by SE Tylose GmbH & Co.KG, Germany,
with a degree of substitution (DSME, ME=Methyl)
of 1.3 and a molar degree of substitution (MSHE,
HE=Hydroxyethyl) of 0.3 was used.
Cement pastes were prepared according to the
following procedure: cement and the MHEC pow-
der were mixed in a plastic cup for two minutes
and subsequently 34 wt. % of water was added to
the mixture. The sample was carefully stirred for 30
seconds and after one minute the sample was again

333
 Faiyas et al.
Figure 1: Plane polarized light microphotographs of cement pastes (upper row) and mortars (lower row) with 0 wt.%
MHEC (left) and 0.5 wt.% MHEC.
stirred for one minute. After five minute, stirring
was performed for 15 seconds. Finally the cement
was put in to cylindrical Teflon tube with 18 mm
diameter and 10 mm length and the Teflon Tube is
sealed to avoid any evaporation.
Mortar samples for experiment 1
The dry composition for mortar samples for experi-
ment 2 was the following: 38 wt.% CEM I 52.5R, 57
wt.% quartz sand, 5 wt.% limestone filler and 0 or
0.5 wt.% MHEC. The MHEC was similar as those
used for the cement pastes.
Mortar paste samples were prepared according
to the following procedure: the dry components
were mixed in a plastic cup for two minutes and
subsequently 34 wt.% of water was added to the
mixture. The sample was carefully stirred for 30
seconds and after one minute the sample was again
stirred for one minute. After five minute, stirring
was performed for 15 seconds. Finally, the mortar
was put in to cylindrical Teflon tube with 18 mm
diameter and 10 mm length and the Teflon Tube is
sealed to avoid any evaporation.
Mortar samples for experiments 2 and 3
Mortars for experiments 1 and 3 have the following
dry composition: 38 wt.% CEM I 52.5 R, 57 wt.%
quartz sand, 5 wt.% limestone filler. MHEC con-
centrations were varied from 0, 0.21, 0.5, 2.1 and
4.3 wt.% MHEC relative to water. A commercial
grade MHEC (Tylose MHS 6000 P6) provided by
334
SE Tylose GmbH & Co.KG, Germany was used in
experiments 1 and 3. This MHEC has a degree of
substitution (DSME, ME=Methyl) of 1.3 and a molar
degree of substitution (MSHE, HE=Hydroxyethyl)
of 0.3.
Mortars were prepared according to the follow-
ing procedure: the dry components were mixed in
a plastic cup for two minutes and subsequently 23
wt.% of water was added to the mixture. The sam-
ple was carefully stirred for 30 seconds and after
one minute the sample was again stirred for one
minute. After five minute, stirring was performed
for 15 seconds. Finally the mortar was put in to
cylindrical Teflon tube with 18 mm diameter and
10 mm length and the Teflon Tube is sealed to avoid
any evaporation or left open (part of the samples
in experiment 3).
III. Experimental details
Microscopic methods
Thin sections for optical microscopy were made
and investigated according to procedures outlined
in Nijland & Larbi (2010). Scanning electron mi-
croscopy (SEM) was performed using a FEI Nova
Nanosem 630 with solid state (GAD) and secondary
electron (SE) detectors and energy dispersive spec-
trometeres. This type of SEM allows for the use of
uncoated samples.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Nuclear Magnetic Resonance
NMR principle - Nuclei with a net magnetic mo-
ment (such as 1 H,23 Na, 35 Cl, 19 F and 6 Li etc..)
will start to precess, when placed in a magnetic
field with a resonance frequency, called Larmor
frequency ( f l ) and is given by,
g and an internal surface area S, move randomly due
fl = B0 (1)
2p
where 2pg
is the gyromagnetic ratio of the nuclei
(42.58 MHz/T for hydrogen nuclei), and B0 is the
applied magnetic field strength. To the constant
magnetic field B0 , a spatially varying magnetic field
B can be applied to introduce position dependency,
B = B0 + ( G · x ), (2)
where G is the magnetic field gradient and x
the position. This enables in-situ monitoring of
moisture content at different positions in the sample
since the resonance frequency varies as a function
of position.
The sum of all individual magnetic moment is
called the net macroscopic magnetization, and can
be manipulated by applying an oscillating magnetic
field B1 exactly at the resonance frequency of the
nuclei. The intensity of the resulting spin echo sig-
nal is proportional to the density of the magnetic
moment and thus the hydrogen or water concen-
tration at different positions. Besides the proton
density also information about the interaction of
the spins can be obtained. In principle the intensity
of the spin echo decay is given by
Figure 2: Cross polarized light microphotographs of cement pastes (upper row) and mortars (lower row) with 0 wt.%
MHEC (left) and 0.5 wt.% MHEC.
nt E
S = S0 exp( ) (3)
T2
wherein S0 is the intensity of the signal, T2 the
spin-spin relaxation time and t E the echo time.
Surface relaxation - During an NMR measurement,
water molecules inside a pore with a volume V
to Brownian motion and probe the pore structure.
This will influence the T2 relaxation time, which
contains information on the pore size distribution
of the porous material. Depending on the pore size
distribution various surface relaxation time can be
identified. In the fast diffusion limit (Brownstein
& Tarr 1979), where pores are so small that a typi-
cal molecule crosses the pore surface many times
before it relaxes in the bulk, the surface relaxation
rate is given by
1 1 S
= + rs (4)
T2,s T2,b V
where rs is the surface relativity, which is a con-
sequence of the susceptibility difference between
the water and the porous material. VS is the surface
to volume ratio of the pore, and T2,b is the bulk
relaxation time, which is negligible compared to
surface relaxation time (T2,b » T2,S ).
NMR setup - The experiments were performed
with a home built NMR setup with a static magnetic
field of 0.7 T, resulting a resonance frequency value
of 31.57 MHz and a gradient of 400 mT/m, with a
theoretical spatial resolution of 0.8 mm. Moisture
profiles were obtained by using a Hahn spin echo
sequence at different sample positions, divided by
335
 Faiyas et al.

5.5 0.24
Cement
5.0 cement +0.5 Wt % MHEC
4.5 0.20
0.5 wt % MHEC cement +0.5 wt % MHEC
4.0 0 wt % MHEC
0.16 0.5 wt % MHEC
3.5 0 wt % MHEC

S(t)
3.0
T2 (ms)

0.12
2.5

2.0
0.08
1.5

1.0 cement 0.04

0.5

0.0 0.00
0 10 20 30 40 50 0 10 20 30 40 50
t(h)
t(h)

Figure 3: T2 (left) and signal (right) evolution of cement paste with and without 0.5 wt.% MHEC in a closed system
as function of time.

the reference sample of equal volume. Spin–spin
relaxation time (T2 ) values are obtained by conven-
tional CPMG pulse sequences. The NMR set up is
designed to measure moisture profile of cylindri-
cal samples. A step motor is used to position the
sample inside NMR setup.
Mercury intrusion porosimetry
The pore size distribution of the dried samples was
measured by Mercury Intrusion Porosimetry (MIP).
The pore size distribution from MIP is depending
on the pore geometry, since MIP only measures the
connected pores.
IV. Results
Experiment 1: Effect of the presence of MHEC on
microstructure of cement paste and mortar
Figure 1 shows an overview of the microstructure
of cement paste and mortar with an without 0.5
wt.% MHEC. From the pictures, it becomes clear
that, with the same preparation method, addition
of MHEC to both cement paste and mortar results
in considerably more entrapped air and by conse-
quence a significantly more porous microstructure.
Addition also has a profound effect on the mor-
phology of portlandite crystals (aggregates), which
become larger, more developed, and in case of the
cement paste seem to exhibit some kind of regu-
lar distribution pattern (Figure 2). Figure 3 shows
NMR relaxation curves and signal curves. The sig-
nal curves shows the amount of water being present
in a free state. The signal level drop is the result of
hydration process ongoing. Clearly in both cases
the signal loss in approximately the same, indicat-
ing a similar hydration curve. The T2 curves show
a higher T2 value, which indicates larger pores,
which seems to be in line with the observation of
more air being entrapped in the mortar. The same
observations holds for Figure 4. However, the key
difference seems to be in the larger delay in hydra-

14 0.30

0 wt % MHEC 0 wt % MHEC
12 0.5 wt % MHEC 0.25 0.5 wt % MHEC

10
0.20

8
T2 (ms)

0.15
S(-)

0.10
4

2 0.05

0 0.00
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t(h) t(h)

Figure 4: T2 (left) and signal (right) evolution of mortars with and without 0.5 wt.% MHEC in a closed system as
function of time.

336
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

8 0.35
0 wt % MHEC
7 0 wt % MHEC 0.21 wt % MHEC
0.21 wt % MHEC 0.30 0.4 wt % MHEC
6 0.4 wt % MHEC 2.1 wt % MHEC
2.1 wt % MHEC 0.25 3.4 wt % MHEC
3.4 wt % MHEC
5
T2(ms)

0.20
4

S(-)
0.15
3

0.10
2

1 0.05

0 0.00
0 5 10 15 20 25 0 5 10 15 20 25
t(h) t(h)

Figure 6: NMR T2 evolution (left) and signal (right) of tile mortars with increasing MHEC content in a closed system
as a function of time.

tion (shift of T2 curve to the right in case of presence
of MHEC), and a increase and decrease signal at 3
hours is observed in the signal curves.
Experiment 2: Increasing MHEC content and air
void content of mortar
Figure 5 shows overview of the microstructure of
the tile mortars with increasing MHEC content in a
closed system. As in experiment 1 (Figure 1), addi-
tion of MHEC results in a considerable amount of
entrapped air and porous microstructure. There is
no clear relationship between MHEC content and
the amount of air voids for mortars with 0.4 wt.%
MHEC or more (Figure 5). Corresponding T2 re-
laxation time and signal evolution versus time as
determined by NMR (Figure 6) show an increase
in T2 for mortars with 0.4 wt.% MHEC or more in
the period up to 12 hours; the mortars with MHEC
show lower signal intensity than the one without
MHEC present. With increasing concentration of
MHEC the hydration is delayed, until a concentra-
tion of 2.1 wt% is reached, after which the delay
does not increase. The results of these experiments
are comparable with the results during experiment
1. Note that the viscosity of the first experiment
was much higher at similar concentrations.

Figure 5: Microphotograph with overview of the mi-

crostructure of tile mortars with increasing MHEC con-
tent (Tylose MHS 6000 P6) in a closed system: 0 wt.%
(top left), 0.21 wt.% (middle left), 0.4 wt.% (bottom left),
1.3 wt.% (top right), 2.1 wt.% (middle right) and 3.4
wt.% (bottom right).
Figure 7: Complete cross section of mortars without (left)
and with 0.5 wt.% MHEC (right) in a closed system
(plane polarized light).

337
 Faiyas et al.

Figure 8: Plane polarized light microphotographs of the microstructure of mortars with 0.5 wt.% MHEC in an open
(left) and closed (right) system.

Experiment 3: Effect of MHEC on microstructure open system in presence of MHEC can be explained
and drying behaviour in open and closed systems based on the MRI profiles made for the drying sys-
tems. Figure 12 shows the moisture distribution of
Figure 7 shows the complete cross sections of sam-
drying mortars in an open system with different
ples used for the NMR experiments in a closed
cellulose concentration as function of position. The
system. In line with the other experiments, the ad-
moisture profile of 0 wt.% MHEC mortar initially
dition of MHEC again results in more entrapped
shows shrinkage (indicated by horizontal arrow
air and a considerably more porous microstructure,
in Figure 12a) due to evaporation, i.e. the mortar
as far as macropores (air voids) are concerned. Fur-
compacts, but otherwise the water is distributed
thermore no depth dependence is observed. Figure
homogeneously (indicated by the vertical arrow)
8 shows the microstructure of mortars hardened in
till the end of the process. The moisture profile
open and closed systems. On the scale visible by
of the mortar with 0.4 wt.% MHEC also shows a
optical microscopy, no differences are visible.
homogeneous distribution of water initially; below
Contrary to optical microscopy, SEM shows dif- a moisture content of about 0.10 (m3 /m3 ), the water
ferences in microstructure in case of drying mortars distribution becomes slightly inhomogeneous. At
in an open system both with and without MHEC. higher MHEC concentrations of 1.3 (Figure 12c) and
In the zone below the drying surfaces, similar mi- 2.1 wt.% (Figure 12d), a clear drying front develops
crostructures are observed (Figure 9, top left and at the surface; simultaneous to the ingress, the wa-
right). However, within the mortar with 0.5 wt.%
MHEC, there is a clear variation of microstructure
as a function of depth, showing less cohesion in the
middle (in the zone of the receding drying front;
Figure 9 bottom left) and a more well developed
microstructure with well developed ettringite at
the bottom (Figure 9 bottom right). This is con-
firmed by pore size distributions obtained by MIP
(Figure 10). These show a slight shift to smaller
pore sizes, but still the major peak is close to 1 µm
for the middle of the sample. The bottom of the
sample shows a bimodal pore size distribution, in
which the amount pores around 1 µm is reduced
and a second peak around 0.1 µm appears (Fig-
ure 10). In Figure 11, the pore size distributions
obtained by MIP for mortars drying in open and
closed systems are compared. The pore sizes in
the mortars from closed systems are considerably
smaller than those found in the open systems, re-
gardless whether MHEC is present or not (Figure
11). When MHEC is present in the open system a Figure 9: SEM microphotographs with overview of the
shift is observed to smaller pore sizes, but the sys- microstructure in mortar without (top left) and with 0.5
tem still contains a considerable amount of larger wt% MHEC at the surface (top right), half the cross
pores. section (bottom left) and at the bottom (bottom right) in
The reason for obtaining smaller pores in the closed system.

338
9: SEM microphotographs with 15thoverview of theon
Euroseminar microstructure
Microscopy in mortar to
Applied without (top Materials
Building left) and with 0.5 wt%
• 17-19 JuneMHEC
2015 • Delft, The Netherlands
at the surface (top right), half the cross section (bottom left) and at the bottom (bottom right) in closed system.

0.016 the addition of MHEC to mortar or cement paste

unsealed 0 wt % MHEC mortar
0.014 unsealed 2.1 MHEC mortar middle has an influence on the portlandite morphology. As
unsealed 2.1 MHEC mortar bottom
Incremental pore volume (ml/g) a consequence SEM images were made. These SEM
0.012
observations on the mortar with 0.5 wt.% MHEC
0.010 (open system) show relatively large booklets of port-
0.008
landite (Figure 13) with a nanostructured surface
(Figure 14). EDS analysis confirms the presence
0.006
of an organic compound in these booklets (Figure
0.004 15), indicating coprecipitation of portlandite and
MHEC.
0.002

0.000
0.01 0.1 1 10 100
Pore diameter(nm) V. Discussion and conclusion

Figure
Figure 10: The influence 10: The
of MHEC influence
on the pore sizeofdistribution
MHEC on the pore mortar
of unsealed size as determined
The combined
by MIP.
distribution of unsealed mortar as determined by MIP.
ter content reduces homogeneously at the bottom
of the sample. This receding drying front seems
to correspond to the change in microstructure in
the vertical cross section
Pageobserved
9 of 12 by SEM. The rea-
son for that is that due to the presence of MHEC
a receding front forms, which results in a longer
presence of water at the bottom of the sample. The
top of the sample will dry first. As a result of this
increased retention of water with increasing depth
in the sample, the time of hydration will also in-
crease with depth in the sample. The longer time of
hydration at the bottom of the sample will increase
the degree of hydration and as a consequence of
the larger formation of hydration products the pore
size will decrease. As can be clearly seen from the
SEM pictures and MIP measurements.
In experiment 1 optical microscopy showed that
PFM, SEM, NMR and MIP study
shows a clear effect of the addition of MHEC on
both the porosity / pore structure of the hardened
mortar and cement paste, as well as portlandite
morphology. The amount of entrapped air and with
that macro-pores is significantly increased when
MHEC is added (Figure 1, 5, 7). At the same time,
addition of MHEC causes a shift in the average
pore size, forming a bimodal pore size distribu-
tion in which a larger amount of small pores are
formed (Figures 10, 11). This observation holds for
both open and closed systems having MHEC. In
open systems, this shift in pore size distribution is
accompanied by the development of a receding dry-
ing front at higher MHEC concentrations (Figure
12). Due to the higher viscosity of the pore fluid
it cannot maintain homogenous drying towards
the surface anymore; Consequently, in the upper
part, evaporation occurs and less water is available
for hydration, where as the part of the mortar be-
low the drying front, develops the micro- and pore

0.020
0 wt % MHEC unsealed mortar
2.1 wt % MHEC unsealed mortar
0 wt % MHEC sealed mortar
Incremental Pore Volume (ml/g)

2.1 wt % MHEC sealed mortar

0.015

0.010

0.005

0.000
0.01 0.1 1 10 100
Pore diameter (μ )

Figure 11: The influence of MHEC on the pore size distribution of sealed and unsealed mortar with and without 2.1
wt.% MHEC.

339
 Faiyas et al.

0.35
0.30
a b
0.30
0.25

0.25
0.20
)

)
0.20
0.15
0.15
Θ(

Θ(
0.10
0.10
Topp Bottom

0.05 0.05

0.00 0.00
-2 0 2 4 6 8 10 12 -2 0 2 4 6 8 10 12
X(mm) X(mm)

0.30
0.30
c d
0.25 0.25

0.20 0.20

)
)

0.15 0.15

Θ(
Θ(

0.10 0.10

0.05 0.05

0.00 0.00
-2 0 2 4 6 8 10 12 -2 0 2 4 6 8 10 12
X(mm) X(mm)

Figure 12: NMR Moisture profile of unsealed mortar samples with 0 (a), 0.4 (b), 1.3 (c) and 2.1 (d) wt.% MHEC. 0
and 0.4 wt % MHEC the profiles have been plotted every 12 minutes and 1.3 and 2.1 wt % MHEC profiles have
been plotted every 24 minutes. The top and bottom side of the samples are located on left and right side of the figure,
respectively.

structures (Figure 9, 10) containing a large number Acknowledgement

of small pores.
MHEC clearly adsorb on the portlandite surface
(Figure 14, 15), a phenomenon observed in previous
studies by non-microscopic techniques (e.g. Ou et
al. 2012). This adsorption is probably responsible
for the observed change in portlandite morphology
(Figures 2, 13).
This research was carried out under the project
number M81.6.08315 in the framework of the Re-
search Program of the Materials innovation insti-
tute (M2i). W. Duvalois assisted the SEM work. We
wish to thank Dr. C.-J. Haecker from Shin-Etsu/SE
Tylose De.
This research was carried out under the project
number M81.6.08315 in the framework of the Re-
search Program of the Materials innovation institute
(M2i). W. Duvalois assisted the SEM work. We wish
to thank Dr. C.-J. Haecker from Shin-Etsu/SE Ty-

Figure 13: SEM microphotographs showing details of the morphology of portlandite in mortar with 0.5 wt.% MHEC.

340
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

A.P.A. Faiyas et al.
clearly adsorb on the portlandite surface (Fig. 14, 15), a phenomenon observedHefte
croscopic techniques (e.g. Ou et al. 2012). This adsorption is probably responsible
Figure 14: SEM microphotographs showing detail of
in portlandite morphology (Fig. 2, 13).
portlandite in mortar with 0.5 wt.% MHEC (right pic-
ture in Figure 13).
References
Brownstein, K.R. & Tarr, C.E. (1979): "Importance of classical diffusion
in NMR studies of water in biological cells." Physical Review A Vol. 19,
2446..
Bühler, T., Zurbriggen, R., Pieles, U., Huwiler, L. and Raso, R.A.
(2013): ”Dynamics of early skin formation of tiling mortars investigated
by microscopy and diffuse reflectance infrared Fourier transformed spec-
troscopy.” Cement & Concrete Composites Vol. 37, 61-170.
Faiyas, A.P.A., Erich, S.J.F., Van Soestbergen, M., Huinink, H.P., Adan,
O.C.G. and Nijland, T.G. (2015): "How methylhydroxyethylcellulose
(MHEC) influences drying in a porous media." Chemical Engineering Sci-
ence Vol. 123, 620-628.
Marliere, C., Mabrouk, E., Lamblet, M. and Coussot, P. (2012): "How
water retention in porous media with cellulose ethers works." Cement &
Concrete Research Vol. 42, 1501-1512.
Müller, I. (2006): "Influence of cellulose ethers on the kinetics of early
Portland cement hydration." Karlsruher Mineralogische und Geochemische
in previous
Vol. 32, 110 studies
pp. by
Ou, Z.H., Ma, B.G. and Jian, S.W. (2012): "Influence of cellulose ethers
for the observed
molecular parameters on hydration kinetics of Portland cement at early
ages." Construction and Building Materials Vol. 33, 78-83.
Pourchez, J. (2006): Aspects physico-chimiques de l’interaction des éthers
de cellulose avec la matrice cimentaire. PhD thesis, École Nationale
Supérieure des Mines de Saint-Étienne, 285 pp.
Pourchez, J., Ruot, B., Debayle, J., Pourchez, P. and Grosseau, P.
(2010): "Some aspects of cellulose ethers influence on water transport
and porous structure of cement-based materials." Cement & Concrete Re-
search Vol. 40, 242-252.
Wetzel, A., Herwegh, M., Zurbriggen, R. and Winnefeld, F. (2012):
"Influence of shrinkage and water transport mechanisms on microstruc-
ture and crack formation of tile adhesive mortars." Cement & Concrete
Research Vol. 42, 39-50.

15: EDS spectrum of typical portlandite in mortar with 0.5 wt.% MHEC. As the samples have not been coated, the
Figurepreparation,
n peak is no artefact of sample 15: EDS spectrum of typical
but indicates portlandite
the presence in mor-
of an organic compound intergrown with the
tar with 0.5 wt.% MHEC. portlandite.
As the samples have not been
coated, the carbon peak is no artifact of sample prepara-
tion, but indicates the presence of an organic compound
wledgements
intergrown
This research was carried withthe
out under theproject
portlandite.
number M81.6.08315 in the framework of the Research
m of the Materials innovation institute (M2i). W. Duvalois assisted the SEM work. We wish to thank Dr.
aecker from Shin-Etsu/SE Tylose De.
lose De.This research was carried out under the
project number M81.6.08315 in the framework of
nces
the Research Program of the Materials innovation
tein, K.R. & Tarr, C.E.institute
(1979):(M2i). W. Duvalois
"Importance assisted
of classical the SEM
diffusion work.studies of water in biological
in NMR
We wish to thank
cells." Physical Review A Vol. 19, 2446.. Dr. C.-J. Haecker from Shin-
Etsu/SE
T., Zurbriggen, R., Pieles, Tylose De.This
U., Huwiler, research
L. and Raso, R.A. was carried
(2013): out of early skin formation of
''Dynamics
under theby
tiling mortars investigated project numberand
microscopy M81.6.08315 in the frame-
diffuse reflectance infrared Fourier transformed
spectroscopy.'' Cement
work & Concrete Composites
of the Research Vol. 37,
Program of61-170.
the Materials in-
A.P.A., Erich, S.J.F., Van Soestbergen,
novation M., Huinink,
institute (M2i). H.P., Adan,
W. Duvalois O.C.G.
assisted the and Nijland, T.G. (2015):
"How methylhydroxyethylcellulose (MHEC) influences
SEM work. We wish to thank Dr. C.-J. Haecker drying in a porous media." Chemical
Engineering Sciencefrom
Vol. 123, 620-628.
Shin-Etsu/SE Tylose De.This research was
e, C., Mabrouk, E., Lamblet, M. and Coussot, P. (2012): "How water retention in porous media with
carried out under the project number M81.6.08315
cellulose ethers works." Cement & Concrete Research Vol. 42, 1501-1512.
in the framework of the Research Program of the
I. (2006): "Influence of cellulose ethers on the kinetics of early Portland cement hydration." Karlsruher
Mineralogische undMaterials
Geochemischeinnovation institute
Hefte Vol. 32, 110(M2i).
pp. W. Duvalois
H., Ma, B.G. and Jian, S.W. (2012): "Influence of cellulose ethersDr.
assisted the SEM work. We wish to thank C.-J. parameters on hydration
molecular
Haecker from Shin-Etsu/SE Tylose De.
kinetics of Portland cement at early ages." Construction and Building Materials Vol. 33, 78-83.
ez, J. (2006): Aspects physico-chimiques de l’interaction des éthers de cellulose avec la matrice
cimentaire. PhD thesis, École Nationale Supérieure des Mines de Saint-Étienne, 285 pp.
ez, J., Ruot, B., Debayle, J., Pourchez, P. and Grosseau, P. (2010): "Some aspects of cellulose ethers
influence on water transport and porous structure of cement-based materials." Cement & Concrete
Research Vol. 40, 242-252.
A., Herwegh, M., Zurbriggen, R. and Winnefeld, F. (2012): "Influence of shrinkage and water transport
mechanisms on microstructure and crack formation of tile adhesive mortars." Cement & Concrete
Research Vol. 42, 39-50.
341
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

342
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Application of acoustic microscopy

technique for the assessment of the
efficacy of water repellent treatment in
fired bricks
Maria Stefanidoua,⇤ , Georgios Karagiannisb
Georgios Apostolidisc , Katia Matziarisd
a,⇤ Assistant Professor, Laboratory of Building Materials, Civil Engineering Department AUTH, Greece
b Dr. Electrical Engineer, “Ormylia” Foundation, Art Diagnosis Center, Sacred Convent of the Annunciation,
630 71 Ormylia, Chalkidiki, Greece
c PhD Candidate Electrical Engineer, “Ormylia” Foundation, Art Diagnosis Center, Sacred Convent of the Annunciation,
630 71 Ormylia, Chalkidiki, Greece
d Dr. Chemical Engineering, R&D Manager in Polychimiki G. Matziaris S.A., 570 11 Thessaloniki, Greece
⇤ e-mail: stefan@civil.auth.gr

Abstract

Even though low fired clay bricks provide energy saving to masonry structures, water lies in wait on their physical
and chemical degradation. In an effort to protect them from water and its effects on their degradation, different solutions
have been tested both commercial and nano-modified. Brick samples in cubic shapes (4⇥4⇥4cm) were immersed for 2
minutes duration in coating solutions and then left to dry for 7 days in normal conditions (20 C, R.H. 65%). The water
repellence effectiveness has been tested by means of capillarity, porosity, water absorption, static contact angle (SCA) and
micro-roughness measurements. Since documented by numerous researchers that surface roughness is strongly related
to hydrophobicity, we focused to sample’s profilometry and furthermore the penetration depth of protective coatings
has been visualized and evaluated by means of acoustic microscopy measurements. Acoustic signals on the material
interfaces have been transformed and depicted in order to eliminate unpleasant noise, attributing the impregnated
mineral surface relief. The derived results showed that all the coatings exhibit high penetrability and sufficient protection
against moisture deterioration. Eventually, it seems that acoustic microscopy is a powerful non- destructive technique
that contributes to the evaluation of water repellency treatments in mineral substrates.

Keywords: brick, water-repellent, acoustic microscopy, capillarity, porosity

I. Introduction wide variety of materials with water-proof proper-

Bricks are porous materials and they are susceptible
to deterioration. Water presence, in any status like:
humidity, rain, capillary absorption, water vapour
condensation, causes the most important patholo-
gies in constructions (Watt, 2000). As a result take
place material’s degradation and downgrading of
their initial characteristics (Foraboschi, 2014). Dif-
ferent protection measures either on the building
itself or on its environment have been applied in
order to restrict the problem. Traditional materi-
als that were used to impregnate the surface of
building materials were organic substances such as
waxes and oils (Mac Mullena, 2011) (Nunes, 2014)
while waterproof clay was reported as protective
measure in some cases (Binks, 2005). The evolution
of technology had as a result the production of a
ties (Simionescu, 2009) (Castellino, 2011) (Esteves,
2014).
Fired bricks are some of the most primeval and re-
silient building materials and have been used since
around 3000 BC (Stefanidou, 2014). Clay bricks
have an impressively high load-bearing capacity
while keeping high dimensional stability and com-
pressive strength. All the mentioned characteristics
minimize the risk of cracking, ensuring the struc-
tural integrity of buildings. Bricks absorb and store
heat throughout the day, and then slowly release
it at night, including thermal mass to buildings.
Consequently, bricks offer lower energy consump-
tion, due to reduced costs for heating and cooling.
Salt presence, discoloration, biological alteration,
degradation, swelling and dissolution are some of
the symptoms bricks can present due to water ac-

343
 Stefanidou et al.
tion (Larsen, 1990) (Morton, 1990). Many studies
have been reported in an effort to protect bricks
especially those incorporated in monuments and
historic buildings with special value in the cultural
heritage and different techniques have been used
to evaluate the effectiveness of the treatments (Ste-
fanidou, 2013) (Carrascosa, 2014). In the present
paper silicon-based water proof products have been
applied by the impregnation technique in specially
produced bricks used for conservation purposes.
The difference of these bricks in relation to indus-
try made ones, is the fact that they were produced
hand–made according to compatibility principles
based on the analysis of old bricks (Papayianni,
1998). Thus, their porosity is high the compressive
strength and specific gravity is low in relation to
industrially produced bricks. The special character-
istics that these bricks have, renders them expensive.
Their protection and the extension of their service
life are important for the sustainability and econ-
omy of the intervention works.
II. Experimental part
Materials and methods
Clay brick samples have been prepared by labo-
ratory methods according to monument restora-
tion demands from a clay matrix paste with main
salts consistency as shown in Table 1. Clay moulds
have been prepared and after natural drying all
samples have been fired at temperature lower than
1000 C, imitating old traditional practices. Finally
bricks have been cut by laser machinery in cubical
shapes with dimensions 4⇥4⇥4cm. The mineralogi-
cal composition has determined by XRD analysis as
shown in Figure 1. The colour of the brick samples
have been characterized by use of Munsel chart as
2.5YR5/6 red.
For the protective coatings synthesis has been
selected a commercial product based on triethoxy
(2,4,4-trimethyl-pentyl) silane that was diluted in
heavy naptha (mixture proportion 1:10 parts w/w).
The diluted solution has been divided into 2 parts
and the first part left as it described (coded S) and
the second (coded SN) was enriched by the addition
1% w/w of hydrophobic silicon dioxide nanopar-
ticles (14nm diameter) with BET 200±25m2 /g and
the above mentioned nano-modified coating was
combined with micro-clay addition to 1% w/w. The
added micro-clay is fine grained and it is used for
pottery. The gradation of clay was recorded by par-
ticle size analysis (Mastersizer2000 Scirroco). More-
over 90% of the sample was below 400µm, while
50% of the sample was below 65µm and finally 10%
was below 5.5µm.
The added nanoparticles have been dispersed
by ultrasound bath for 60 minutes. Brick samples
344
were immersed into solutions for 2 minutes dura-
tion. Totally five series of samples were tested and
all recorded values represent the mean average of
these measurements. Note that in each case an un-
treated sample was used for comparison reasons.
All impregnated samples were left for 7 days under
normal conditions (20 C, R.H. 65%) till solvent’s
evaporation.
In order to interpret hygric properties of the
treated low fired clay bricks in comparison to the
untreated samples porosity and water absorption
capacity of the samples have been tested accord-
ing to RILEM CPC11.3 recommendations on water
absorption under vacuum (RILEM, 1984). Further-
more capillarity absorption and the capillary index
were recorded based on EN1015-18 regulation espe-
cially useful for determination of water absorption
coefficient due to capillary action of hardened mor-
tars (BS EN 1015-18, 2002). The capillary forces are
determined by: the surface tension of the liquid,
the contact angle between the liquid and the pore
walls, and the radius of the pores. Narrow pores
attract moisture stronger (higher) than wide pores
are able to do. A viscous liquid is less quickly at-
tracted, than a thinner one. Small contact angles
indicate molecular attraction between the liquid
and the solid substrate. A liquid drop spreads over
a flat surface; the meniscus in a capillary lies above
the level of the surrounding liquid and is hollow
towards the “dry” side. In the absence of such
attraction between a solid substance and a liquid,
a drop remains on the surface in the form of a
sphere and high values of the angle are recorded.
The hydrophobicity of the treated surface has been
evaluated by Static Contact Angle (SCA) measured
by means of a goniometer Kruss DSA 100 with
the special fitted software Drop Shape Analysis.
The penetration depth was tested by our laboratory
empirical protocol based on macroscopically mea-
surements of penetration depth. The cut samples
of bricks have been cleaned by acetone and then
have been dried for 24 hours at normal climatic con-
ditions (25 C and 50% R.H.). Samples have been
immersed for 5 minutes into the measured to be
solution. For curing and development of hydropho-
bicity, samples are stored for 7 days. Afterwards
samples are wetted with water. The hydrophobic
zone does not absorb the water but remains light
and does not change appearance. The untreated,
inner area, is not water repellent, absorbs water
and becomes darker. So the hydrophobic zone can
be measured (i.e. with a ruler or vernier calipers)
or it can be documented by taking a picture. The
macroscopic results of penetration depth have been
combined by Scanning Acoustic Microscopy (SAM)
in order to visualize inside the very thin protec-
tive film mass, the micro to nanoscale structure
of the mineral clay bricks and the organosilanic
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 1: Clay matrix water soluble salts average consistency.

Water soluble, salts in clay matrix (% w. w.) Cl NO3 SO4

Average, consistency of 4 samples 0.0156 0.010 0.422

Figure 1: XRD diagram of the clay added in 1%w.w. of solution in the SN solution.

protective coating. Acoustic Microscopy is an in-

novative, non-destructive technique, by using very Zo = p · c (1)
high frequency ultrasounds. It helps to superficial
where p is the density of a medium and c is
and internal features visualization of solid samples.
the velocity of sound into this medium. Therefore
Generally fewer scientists are familiar with this tech-
practically total reflection appears into material in-
nique because it involves more complex processes
terfaces, which have different acoustic impedance,
that require greater specialization and training for
while wave reflection turns to minimal when two
test operation and results interpretation. However,
materials/media have similar acoustic impedance.
acoustic microscopy gives some advantages com-
The echoes are received by the transducer at differ-
pared to other available technologies that make it
ent time instances, i.e., time of flight (TOF), which
superior as a tool for some materials and processes,
are proportional to the distance of the interfaces
e.g. artifacts diagnosis (Briggs, 1995) (Briggs, 1996)
from the transducer. The signal that contains the
(Karagiannis, 2009) medical applications etc. The
echoes is called the A-Scan signal. Acoustic mi-
ultrasonic waves are emitted from the transducer
croscopy allows the roughness measurement of a
into the measured sample through the used cou-
micro scale surface. To find the roughness from
pling medium. Since ultrasound propagates from
acoustic measurements (A-Scans) it needs to cal-
the transducer to subject, echoes are generated by
culate the time of flight of the acoustic wave from
the interfaces of the various micro-structures. An
the time it was produced until the time the first re-
echo is the reflective component of the initially gen-
flected echo is received from the transducer (RILEM,
erated wave. The density of the reflection depends
1979) BS EN1015-18, 2002). The procedure is shown
on the difference of the acoustic impedances of
in Figure 5. This is accomplished by gating the
the engaged materials. The acoustic impedance is
A-Scans in a time interval which is related to wave
defined by:
return. The calculation of Time of Flight (TOF) re-

Table 2: Physical properties of the tested samples.

Porosity Water Absorption Capillary index

Sample
% % g/cm2 min0.5
Untreated sample 24.34 12.69 0.330
S 17.55 9.76 0.185
SN 14.57 6.88 0.004

345
 Stefanidou et al.

sults an image where its pixel represents the TOF, Table 3: Water drop angles measured on the surfaces of
i.e., a high pixel value means a deep spot, whereas, the samples.
a small value means a tip spot (Briggs, 1995). In or-
der to calculate the roughness, the two dimensional Sample Contact Angle ( )
Fourier Transform is used, Untreated sample 0
S 109.8
N1 1 N2 1 SN 137.0
X ( w1 , w2 ) = Â Â x [n1 , n2 ]e j(w1 ·n1 +w1 ·n2 ) (2)
n1 =0 n2 =0
3.75mm (Figure 4a), while SN coating reached an
where x[n1 ,n2 ] is an image of size N1 ⇥ N2 and average penetration depth of 8.66mm (Figure 4b).
X (w1 , w2 ) is the resulted Fourier transform. Espe- The results are shown in Table 4. It seems that
cially the surface roughness of the samples was the roughness was increased after treatment with S
measured using a profilometer (Mitutoyo) and R a solution but the NS solution made a considerable
values were calculated. The roughness of the sam- difference and a rather rough surface was achieved.
ples have a detrimental role in hydrophobicity as
it enhances hydrophobicity of a solid surface lead-
Acoustic Microscopy Measurements
ing to high contact angles with water (Nosonovsky,
2007). In this case R a value which is a parameter The configuration followed for the measurements
that characterizes the surface based on the vertical were:
deviations of the roughness profile from the mean
line is recorded. R a is the arithmetic average of the • Transducer: 110MHz
absolute values. In order to check results, some • Scanning Area: 1⇥1 mm
preparation tests have been conducted for compare
both results. • Scanning Step: 0.01⇥0.01 mm

• Sampling Rate: 500MHz

III. Results and discussions
• ABS Gate: 7-8.1 µsec
Capillary and porosity tests • Coupling Medium: Deionized Water
The physical characteristics recorded in the samples
The images of roughness for three samples and
are shown in Table 2. The silicone coating (S) im-
the respective 2D-Fourier are illustrated in Figure
posed a better behaviour in relation to the untreated
6.
samples both to capillary index and to water ab-
The 2D Fourier Transforms show that the rough-
sorption test. The nanosilica modified coating (SN)
ness is limited in sample S and NS comparing with
showed even better behaviour in the above tested
the untreated sample. Also, S sample show less
parameters. The capillary absorption trend indi-
roughness that NS which is in agreement with the
cates (Figure 2) that treated with S samples after a
results of the profilometer. Moreover, according to
certain time increase the quantity of the absorbed
the dominating values of Fourier Transforms, it can
water but it still remains low comparing to the un-
be said that the untreated sample has alterations
treated samples. The lowest absorption is recorded
in the whole spectrum, i.e., 50 alterations/mm,
in samples treated with SN coatings.
whereas, S appears roughness from 0 to 15 alter-
ations/mm and NS from 0 to 30 alteration/mm.
Static Contact Angle (SCA) test To point that, as it was waited, roughness is in-
dependent of the spatial dimension. The acoustic
The water drop test results are tabulated in Table 3.
transmittance in a material is related to its acoustic
The untreated sample absorbed water droplets al-
impedance. In order this material behavior to be
most immediately and no values could be reported.
recognized, it needs to examine the C-Scans (3D
This phenomenon occurs when a porous building
representation of the measurement). Longitudinal
material such as low fired clay brick comes into
sections of samples’ C-Scans are illustrated in Fig-
contact with (liquid) water, because water is sucked
ure 7.
immediately into pores by capillary forces.
The best hydrophobic behaviour was recorded by
SN coating, shown characteristic images (Figure 3). Table 4: R a values from profilometry measurements.

Sample R a (mm)
Penetration depth-roughness Untreated sample 0.041
The protocol measured penetration depth indicates S 0.058
that samples with S coating have a depth of almost SN 0.158

346
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 2: Capillary absorption curves.

Figure 3: Water drop angles measured on the surfaces of the samples.

Figure 4: Macroscopic calculation of the penetration depth

347
 Stefanidou et al.

Figure 5: Photos during measurements

It is shown that S solution allows an acoustic
wave to be transmitted, but less than the untreated
sample, whereas, NS solution does not allow the
wave to be transmitted or the transmitted power
is significantly reduced compared to the untreated
sample.
To quantify these results, the A-Scans of every
measurement will be processed. Specifically, we
want to measure the depth a wave travels into the
material. This is realized by monitoring the am-
plitude of the wave. When the time during which
the amplitude of wave’s envelop is larger than a
threshold is considered as the effective duration of

Figure 6: 3D representation of the roughness (above) and 2D Fourier images.

348
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 7: Longitudinal sections of a) untreated sample, b) of samples treated with S and c) of samples treated with SN.

the wave. Finally, for the mean effective duration
among all A-Scans of a measurement is calculated.
The threshold can be a ratio of maximum value
of the wave, in this case, one hundredth. Also, to
mention, that A-Scans are low pass filtered in order
to eliminate noise artifacts. The envelop (senv (t)) of
a signal (s(t)) is calculated as following;
s a (t) = s(t) + jHT {(t)} (3)
senv (t) = |s a (t)| (4)
where HT {.} is the Hilbert Transform.
The Mean Effective Durations (MED) of the three
samples and the relative MED according to the used

Figure 8: Measuring Effective Duration of an echo by monitoring its envelop. Illustration of the operation in an
arbitrary A-Scan.

349
 Stefanidou et al.
Table 5: Values of the MED for the three samples.
Sample MED (µsec)
Untreated sample 0.5697
S 0.4128
SN 0.158
time window are as shown in Table 5.
The resulted MED values offer indirect informa-
tion about the acoustic impedance of the materials.
The fact that the MED of reference sample is large
means that water has been absorbed in some depth
during the measurement, allowing the coupling of
acoustic wave for more time.
About S treated samples, the MED indicates less
transmittance, hence more reflectance. This means
that the wave found a surface between two of differ-
ent impedance materials. Consequently, the more
different the materials, the more the reflectance.
These two materials can be considered as the resin
absorption region and the non absorption one. The
resin has changed the mechanical properties of the
material, making it more compact, so less resistive
to acoustic waves. On the other hand, the more
porous non absorption region is more resistive to
acoustic waves. As a result, the common surface of
these two different materials reflects very strongly
the wave, not allowing it to travel in depth.
Moreover, about NS treated samples, the same
wave transition principle governs the results. In
this case, the wave is transmitted more, so the inter-
mediate surface between resin/non-resin regions is
less definite. This means that the resin in this case
made the region where was absorbed less compact
comparing with NS.
Initially the testing results starting from our em-
pirical laboratory protocol shows that the penetra-
tion depth varies by type of protective coatings and
strongly depend on its additives (nano or micro).
Furthermore profilometry is confirming that each
coating exhibits high penetrability and sufficient
protection against moisture deterioration. In addi-
tion to our findings acoustic microscopy visualizes
the interfaces in micro and nano scale between the
organosilanic coating and the crystal brick surface.
Hereby, confirmed the evaluation of water repel-
lency treatments in mineral substrates.
IV. Conclusion
The brick samples have been impregnated by use of
an organosilanic coating that furthermore this coat-
ing was enriched with the addition of 1.5% w/w
nanosilica and 1% w/w micro-clay. The achieved
physical properties comparing to the untreated sam-
ples show a significant improvement of all the brick
protection parameters. Additionally all test have
350
shown that the nano and micro addition conducted
to a deeper penetration of the coating, thus result-
ing higher hydrophobicity, endorsed by higher SCA
test results. In other words less water was absorbed
either by capillary size porous or water absorption
under vacuum. This phenomenon is related to the
roughness’s increase, which was achieved on the
surface of the bricks when they were treated with
nano and micro protrusions. The roughness in-
crease was recorded individually by profilometer
measurements and also by the acoustic microscopy
technique.
Furthermore into our study has been revealed the
practical contribution of acoustic microscopy to the
surface study of the mineral profiles. A simple pro-
filometer measurement is extrapolated from micro-
sizing analysis with the incorporation of scanning
acoustic microscopy to a full nano-scale visualiza-
tion of the mineral and organosilanic compound
interface area. As well measuring the Effective Du-
ration of an echo, could give us conclusions for
the material’s compactness, the internal structure,
micro fragments, pore mapping etc., thus giving
indirect information on the penetration depth of
the protecting coatings. Conclusively acoustic mi-
croscopy seems to be a very powerful method in
order to record the efficacy of impregnating treat-
ments on porous clay materials. The combination
of these techniques could help scientists to have
valuable information about mineral materials, in
order to upgrade their properties.
References
Watt, D., Colston, B (2000): “Investigating the effects of humidity and
salt crystallization on medieval masonry”. Building and Environment. Vol.
35, 737-749.
Foraboschi, P., Vanin A. (2014): “Experimental investigation on bricks
from historical Venetian buildings subjected to moisture and salt crystal-
lization” Engineering Failure Analysis, Vol. 45, 185-203.
Mac Mullena, J., Zhanga, Z., Rirschb, E., Nath Dhakal, H., Bennett,
N. (2011): “Brick and mortar treatment by cream emulsion for improved
water repellence and thermal insulation”. Energy and Buildings, Vol. 43,
1560-1565.
Nunes, C., Slížková, Z. (2014), “Hydrophobic lime based mortars with
linseed oil: Characterization and durability assessment”. Cement and
Concrete Research, Vol. 61-62, 28-39.
Bernard, P., Binks, J., Clint, H. and Whitby C. (2005): “Rheological
Behavior of Water-in-Oil Emulsions Stabilized by Hydrophobic Bentonite
Particles”, Langmuir, Vol. 21, No 12, 5307-5316.
Simionescu B. Olaru M. (2009): “Assesment of Siloxane-based poly-
meric matrices as water repellent coatings for stone monuments”. Euro-
pean Journal of Science and Theology, Vol. 5, No 1, 59-67.
Castellino, V., Cheng, Y.L. and Acosta E. (2011): “The hydrophobicity
of silicone-based oils and surfactants and their use in reactive microemul-
sions”. Journal of Colloid and Interface Science, Vol. 353, No 1, 196-205.
Esteves, C., Flores-Colen, I. and Veiga, R. (2014): “Performance as-
sessment of hydrophobic treatments on different substrates”, Portugal,
7th International conference on water repellent treatment and protective surface
technology for building materials, Lisbon, September 2014.
Stefanidou, M., Papayianni, I., Pachta, V. (2014)“Analysis and Charac-
terization of Roman and Byzantine fired bricks from Greece”. Materials
and Structures. DOI 10.1617/s11527-014-0306-7
Larsen, E-S, Nielsen, C.B. (1990): “Decay of bricks due to salts”. Ma-
terials and Structures. Vol.23, 16-25.
Morton J (1990): “The maintenance of Brick and Stone Masonry Struc-
tures”. Editor Sowden A.M. E & F SPON,17-21.
Stefanidou M., Matziaris K., Karagiannis G. (2013): “Hydrophobiza-
tion by Means of Nanotechnology on Greek Sandstones Used as Building
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Façades”. Geosciences, Vol. 3, 30-45.
Carrascosa, L., Facio, D. and Mosquera, M. (2014): “Producing super-
hydrophobic coatings for roof tiles”, 7th International conference on water
repellent treatment and protective surface technology for building materials, Lis-
bon, September 2014.
Papayianni, I., (1998): “Criteria and methodology for manufacturing
compatible repair mortars and bricks”. Proceedings of International Con-
ference, Compatible Materials for the protection of European Cultural Heritage,
Editor Biskotin G. et al., December, Athens Al, 179-190.
RILEM (1979): Recommendation CPC11.3 Absorption of water by im-
mersion under vacuum, Materials and Structures Vol. 17, No 101.
BS EN1015-18 (2002): Methods of test for mortar for masonry - Part
18: Determination of water absorption coefficient due to capillary action
of hardened mortar.
Briggs, A. (1995): Advances in acoustic microscopy, Vol 1. Plenum
Press.
Briggs, A. & Walter, A., (1996): Advances in acoustic microscopy Vol
2. Plenum Press.
Karagiannis, G. Vavliakis, K. Sotiropoulou, S. Damtsios, A. Alexiadis,
D. Salpistis, C. Daniilia, S. (2009): “Using Signal Processing and Semantic
Web Technologies to Analyze Byzantine Iconography”. IEEE Instrumen-
tation and measurements, Vol. 60, No 9, 1-18.
Nosonovsky, M., Bharat, B. (2007): “Hierarchical roughness makes
superhydrophobic states stable” Microelectronic Engineering. Vol. 84, No
3, 382-386.
351
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

352
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

An experimental study on the recovery of

the hardened cement from crushed end of
life concrete
Somayeh Lotfi⇤ and Peter Rem
Section of Materials and Environment, Faculty of Civil Engineering and Geosciences, Delft University of Technology
Delft, The Netherlands
⇤ S.Lotfi@tudelft.nl

Abstract

In the C2CA concrete recycling process, autogenous milling of the crushed End of Life (EOL) concrete is a mechanical
method to remove cement paste from the surface of aggregates. During autogenous milling, the combination of shearing
and compression forces, promotes selective attrition and delivers a better liberation. In order to investigate the effects
of shear and compression on the cement recovery and specify the importance of them, a new set-up is designed and
constructed. This set-up permits aforesaid forces to be determined and controlled. For experimental design, the MINITAB
16 software was used and 13 different experimental runs based on varying shear and compression forces were conducted.
After each experiment, the amount of cement recovery using XRF analysis, water absorption of the recycled aggregates
and energy consumption during the process were measured. Results show that both shear and compression forces have
influence on improving the cement recovery. With simple changes in the setting of an autogenous mill like bed height or
residence time the need for high-cost secondary crushing during concrete recycling could be eliminated.

Keywords: concrete recycling, cement recovery, recycled aggregate

I. Introduction fine fraction.

The idea of using autogenous milling instead of
the secondary crushing has a root in a fundamental
Recycling End of Life (EOL) concrete is challeng-
principle of mineral processing. In mineral process-
ing for the building sector because of the compet-
ing, one of the aims of comminution is liberation of
ing constraints of low recycling process cost and
valuable components separated from gangue. In an
high product quality. The C2CA process aims at a
autogenous mill, internally created shear and com-
cost-effective system approach for recycling high-
pression forces produce a gentle attrition among
volume EOL concrete streams into prime-grade ag-
particles. Therefore, surface layers, edges or corners
gregates and cement. One of the main technologies
from crushed EOL concrete can be removed (King,
considered within C2CA is autogenous milling. Af-
2012).
ter crushing the EOL concrete, liberation of the
In the present study, the influence of shear and
cement paste is promoted by several minutes of
compression forces on the cement recovery from
grinding in an autogenous mill while producing
EOL concrete was determined. Evidence is pre-
as little as possible fine silica (Lotfi et al.,2014) and
sented that the amount of cement recovery in the
(Lotfi et al., 2013).
crushed fine fraction EOL concrete is influenced
The conventional EOL concrete recycling circuit by both shear and compression forces. The aim of
is composed of simple size reduction and classi- this investigation is to enrich our understanding of
fication for RA production. However, it does not the importance of noted forces with respect to the
deliver a high amount of liberated aggregate and cement recovery and also enable advances in the
cement paste(Kim et al., 2012). According to some field of concrete recycling.
studies ( Florea et al., 2013) the liberation of the
cement paste in the fine fraction will enhance by
increasing the number of crushing processes. A
II. Materials and methods
secondary crusher can apply pure compression and
Parent concrete and primary crushing
bring higher cement recovery. However, using pure
compression is not always economic and beneficial. Table 1 shows the mix design of the parent concrete
It requires a substantial amount of energy and there with the strength class of C30/37. After casting
is a high possibility of breaking aggregate into the and six months curing of the parent concrete, a jaw

353
 Lotfi & Rem

Table 1: Mix proportions of the concrete per m3 Table 2: Variables and their level for experimental design

Component Wet [kg] Dry [kg] Variable Low level High level
CEM III/B 330 330 Compression: Force (KN) 1.2 KN 36 KN
Sand 0,125-0,250 mm 74.72 74.72 Shear: Duration (min) 2 min 8 min
Sand 0,250-0,500 mm 242.85 242.85
Sand 0,500-1 mm 242.85 242.85
Sand 1-2 mm 149.45 149.45
Sand 2-4 mm 93.4 93.4
Gravel 4-8 mm 373.61 373.61
Where b i represents the linear effect of xi , b ii
Gravel 8-16 mm 691.18 691.18
Water 165 represents the quadratic effect of xi , b ij represents
Super plasticizer 0.27 the interaction between xi and x j and e is the fitting
Total 2363 2198 error.
In this research, design and analysis of the central
composite experiment were carried out using the
crusher with the opening of 20 mm was used for MINITAB 16 software. The variable of compression
the primary crushing. Figure 1 shows the particle was defined as the force exerted by the cylinder,
size distribution of the crushed parent concrete. whereas the variable of shear was represented by
the duration of the back and forth motion of the
Set-up for applying shear and compres- arm. Both variables had two different levels (low
sion force and high, see Table 2). In line with Equation (1), the
total number of simulated runs was 13 experiments.
A new experimental set-up with the purpose of
applying controlled shear and compression on
the bulk crushed concrete was designed and con-
structed. The schematic of the Shear - Compression
Machine (SCM) can be seen in Figure 2. The SCM Analysis of the experimental samples
consists of a vertical cylinder for the application
the compression force. A ring-shaped container After each experimental run, the particle size dis-
is placed under the vertical cylinder which is con- tribution of crushed concrete was analysed. The
nected to an arm. An electrical engine connected to chemical composition of CEMIII/B used in the par-
the arm is applied to move the container back and ent concrete was determined using XRF (model
forth. The effective surface area of the container is PANalytical-Epsilon 3x spectrometers) (see Fig-
0.12 m2 , and for each test it can be filled out with ure 4). The XRF result of the cement shows the
approximately 22 kg of crushed parent concrete. amount of SiO2 and CaO, 27.86 mass% and 47.55
mass% respectively (See table 3). In order to cal-
Experimental design culate the cement recovery, XRF analysis for all
0-1mm and 0-0.5mm fractions was conducted. Ac-
In order to figure out the cumulative effects of two
cording to the primary weight of cement and silica
variables (shear and compression), it was decided to
(sand plus aggregate), and assuming that the main
use Response Surface Methodology (RSM). Within
source of CaO in in the parent concrete was cement
RSM, Central Composite Design (CCD) is a popu-
(compare Table 3 and 4), the total mass of CaO
larly used method (Montgomery, 2008). As shown
in the parent concrete could be calculated. XRF
in Figure 3, a two-variable CCD is composed of
analysis of the 0-1mm and 0-0.5mm fractions yields
22 =4 factorial points, extended by 2⇥2 additional
the mass% of CaO. Thus using the mass (gr) of the
axial points and 5 centre points (t0 ) (five replica-
considered fractions, their total mass(gr) of CaO
tions). In general, for a k-variable CCD, the total
can be calculated. The cement recovery is then cal-
number of simulated runs T is calculated by:
culated based on the calculated mass of CaO in the
considered fraction divided by the total mass of
T = 2k + 2k + t0 (1)
CaO in parent concrete.
Repeating runs at the centre of the design intro-
The water absorption of the 4-16 mm fractions
duces a check on variability and repeatability into
was determined to assess the amount of cement
the data, providing a means to eliminate noise in
paste reduction and the energy consumption for
the experimental results. The general form of RSM
each experimental run was calculated according
with the second-order model is expressed as:
to the torque measurement on the arm. Figure 5
shows the experimental set-up. A sensor was ap-
y = b 0 + Â b i xi + Â b ii xi2 + Â b ij xi x j + e (2) plied to determine the amount of torque and energy
i i i< j consumption during the experiments.

354
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 1: Particle size distribution of parent concrete after primary crushing

Table 3: XRF analysis of applied cement (CEMIII/B) in the parent concrete

Component MgO Al2 O3 SiO2 SO3 K2 O CaO TiO2 V2 O5 Cr2 O3

Mass% 7,85% 10,33% 27,86% 3,48% 0,47% 47,55% 0,72% 0,01% 0,00%
Component MnO Fe2 O3 CuO ZnO SrO Y 2 O3 ZrO2 SnO2 BaO
Mass% 0,17% 1,32% 0,01% 0,02% 0,10% 0,01% 0,03% 0,02% 0,04%

Table 4: XRF analysis of applied aggregate in the parent concrete

Component MgO Al2 O3 SiO2 SO3 K2 O CaO TiO2 V2 O5 Cr2 O3

Mass% 0,13% 2,79% 94,83% 0,23% 1,30% 0,21% 0,13% 0,01% 0,01%
Component MnO Fe2 O3 CuO ZnO SrO Y 2 O3 ZrO2 SnO2 BaO
Mass% 0,01% 0,17% 0,01% 0,01% 0,00% 0,00% 0,02% 0,13% 0,00%

355
 Lotfi & Rem

III. Results and discussion
Interpretation of the regression analysis
Experiments were performed according to the ex-
perimental plan and the results are given in Table 5
along with the results predicted by the model. Ta-
bles 6 and 7 show estimated regression coefficients
for cement recovery in the 0-1 mm and 0-0.5 mm
fractions, respectively. Using T-test and P-values
regression analysis was carried out. In general, the
larger the magnitude of T and the smaller the value
of P, the more significant is the coefficient term
(Montgomery, 2008). Considering Tables 6 and 7,
the effects of the linear factors on cement recovery
are highly significant (P< 0.001). A positive sign
of a coefficient represents a synergistic effect while
a negative sign shows antagonistic effect. In both
tables, it can be seen that the linear terms of com-
pression force and shear duration, the quadratic
term of force and the interaction term of force and
duration have a positive effect on cement recovery.
Those coefficients show that with an increase in
the amount of force and duration the recovery per-
centage of cement will increase. Considering the
regression coefficients, two regression equations for
cement recovery in two different fractions 0-1mm
and 0-0.5 mm result as following:
Y = 0.309341 + 0.079612X1 + 0.039063X2 +
(3)
0.014693X12 0.008136X22 + 0.014098X1 .X2
Y 0 = 0.201083 + 0.049825X10 + 0.030094X20 +
(4)
0.002352X102 0.005516X202 + 0.022357X10 .X20
Where Y is the Recovery of cement into the 0-
1mm fraction and Y0 is the recovery of cement into
the 0- 0.5mm fraction. In both tables, the value of S

Figure 2: Schematic of Shear-Compression Machine(SCM)

Figure 3: Layout of the experiments in a 2-factor Central Composite Design

356
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 4: XRF facility in the resources and recycling laboratory -TUDelft
(standard deviation) between the measured and pre-
dicted results shows that the equation adequately
represents the relation between the response and
significant variables. In particular, S is close to the
experimental error of the data. The high value of
(R2 = 97.06%, R2 = 93.12%) and (R2 (adj) = 94.96%,
R2 (adj) = 88.20%) show high correlation between
the observed and predicted values of response.
Main effect plot
A main effect is present when different levels of
factor influence the response differently. It is cre-
ated by plotting the response mean for each factor
level. A line is drawn to connect the points for
each factor and a reference line is also drawn at
the overall mean (Greenfield and Metcalfe, 2008).
When the line is not horizontal, there is a main
effect present. Different level of the factor affects
the response differently. The greater the difference
in the vertical position of the plotted points, the
greater is the magnitude of the main effect. The
main effect of the parameters Force and timing on
cement recovery from 0-1mm and 0-0.5 mm fraction
are given in Figure 6A and 6B. Reference line in
figures 6A and 6B is 0.3114 and 0.2001 respectively.
From the figures, it is observed that both timing
and force have a positive effect on the cement re-
covery. From the main effect plot, it is obvious that
Figure 5: Shear-Compression Machine
force has slightly more influence.
Normal probability plot
The normality of the data can be checked by plot-
ting the normal probability plot of the residuals.
The normal probability plot is a graphical tech-
nique for assessing whether or not a data set is
approximately normally distributed. Figures 7A
and 7B show normal probability plot of residual
values. Trends observed in those figures reveal well-
behaved residuals. Based on this plot the residuals
appear to be randomly scattered.
Interpretation of surface and contour
plots
Contour and surface plots give a better understand-
ing of the influence of variables and their interac-
tion on the response. A contour plot provides a
two-dimensional view, where all points having the
same response are connected to produce contour
lines. A surface plot provides a three- dimensional
view that may provide a clearer picture of the re-
sponse surface. Figure 8 shows the 3D or 2D plots
relationship between two variables (force and tim-
ing) and properties like cement recovery, water ab-
sorption and energy consumption. According to
the results, with increasing the amount of force and
357
 Lotfi & Rem

Table 5: Full factorial central composite design matrix of two factors, with experimental and predicted response
(cement recovery)

Run Order Force(KN) Duration (min) Recovery predicted Residual Recovery of predicted Residual
of cement Recovery of cement in Recovery of
in (0-1mm) cement in (0-0.5 mm) cement in
Based on (0-1mm) Based on (0-0.5mm)
XRF results XRF results
1 6.30 2.88 25.00% 23.60% 1,40% 16.89% 15.40% 1,50%
2 18.60 5.00 30.93% 30.90% 0,00% 20.93% 20.10% 0,80%
3 18.60 8.00 35.03% 34.00% 1,00% 23.54% 22.60% 1,00%
4 1.20 5.00 23.31% 24.40% -1,10% 14.27% 15.40% -1,10%
5 36.00 5.00 41.66% 40.40% 1,30% 26.38% 25.30% 1,10%
6 18.60 5.00 30.57% 30.90% -0,40% 19.53% 20.10% -0,60%
7 6.30 7.12 27.82% 27.70% 0,10% 17.51% 17.40% 0,10%
8 30.90 2.88 33.13% 33.40% -0,30% 20.19% 20.20% 0,00%
9 18.60 5.00 30.81% 30.90% -0,10% 20.76% 20.10% 0,70%
10 18.60 2.00 25.38% 26.20% -0,80% 15.54% 16.50% -1,00%
11 30.90 7.12 38.77% 40.40% -1,60% 25.28% 26.70% -1,40%
12 18.60 5.00 30.90% 30.90% 0,00% 19.47% 20.10% -0,60%
13 18.60 5.00 31.46% 30.90% 0,50% 19.85% 20.10% -0,30%

Table 6: Estimated regression coefficient for cement recovery into the 0-1 mm fraction

Term Coefficient Standard error coefficient T-Value P-Value

Constant =(X0 ) 0.309341 0.005281 58.577 0.000
Force(KN) = (X1 ) 0.079612 0.005904 13.484 0.000
Timing(min) ( X2 ) 0.039063 0.005904 6.616 0.000
Force(KN)*Force(KN) =(X21 ) 0.014693 0.008954 1.641 0.145
Timing(min)*Timing(min) =(X22 ) -0.008136 0.008954 -0.909 0.394
Force(KN)*Timing(min) =(X1 .X2 ) 0.014098 0.011808 1.194 0.271
S(Standard error) = 0.0118084 R2 = 97.06% R2 (adj) = 94.96%

Table 7: Estimated regression coefficient for cement recovery into the 0-0.5 mm fraction

Term Coefficient Standard error coefficient T-Value P-Value

Constant =(X’0 ) 0.201083 0.005462 36.816 0.000
Force(KN) = ( X’1 ) 0.049825 0.006107 8.159 0.000
Timing(min) = ( X’2 ) 0.030094 0.006107 4.928 0.002
Force(KN)*Force(KN) =( X’21 ) 0.002352 0.009261 0.254 0.807
Timing(min)*Timing(min) =( X’22 ) -0.005516 0.009261 -0.596 0.570
Force(KN)*Timing(min) =( X’1 . X’2 ) 0.022357 0.012213 1.831 0.110
S(Standard error) = 0.0122130 R2 = 93.12% R2 (adj) = 88.20%

358
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 6: A: Main effects plot for recovery of cement in (0-1mm), B: Main effect plot for recovery of cement in
(0-0.5mm). The duration of back and forth motion is indicated as timing

duration, the weight of 0-1 mm fraction and the
cement recovery are increased. Recovery of cement
is affected by both compression and shearing and
it is increased to more than 40% in 0-1mm fraction
(see Figures 8A an 8B). Decreasing the amount of
water absorption in coarse fraction 4-16mm, by in-
creasing the amount of force and timing, is another
evidence to prove the reduction of cement paste on
the surface of recycled aggregates (see Figures 8C
and 8D). The energy consumption raises by in-
creasing the amount of timing and force . However,
according to Figures 8E and 8F even by using the
highest amount of force and timing, the energy con-
sumption is less than 700 kJ/ton (0.19 KWh/ton).
It shows that the cost of autogenous milling could
stay in a reasonable range during concrete recycling
process.
IV. Conclusion
nous mill in a controlled way, a new set-up was
constructed. A central composite experimental de-
sign with the help of the MINITAB 16 software
for predicting the results of 13 experimental runs
was used. According to the regression analysis, the
effect of shear and compression on the cement re-
covery for both 0-1 mm and 0-0.5 mm fractions was
found to be strongly linear (P< 0.001). Comparing
the main effect plots, force (compression) is slightly
more effective than timing (shear). However, based
on the achieved results, it is possible to replace
the shear and compression with each other with
the purpose of raising the cement recovery. There-
fore, high amount of produced low-cost shear in
an autogenous mill will eliminate the need for the
expensive pure compression in a crusher. Variation
in the strength of concrete could be compensated
by simple changes in the mill feeding, the residence
time and the bed height.

In the C2CA concrete recycling process, autogenous Acknowledgement

milling of crushed end of life concrete is used to
increase the liberation of the cement paste. This
research is carried out to understand how shear
and compression, and the combined effect of them
inside of an autogenous mill, influence the cement
recovery. In order to simulate forces in an autoge-
This research is funded by the European Commis-
sion in the framework of the FP7 Collaborative
project "Advanced Technologies for the Production
of Cement and Clean Aggregates from Construction
and Demolition Waste (C2CA)". Grant Agreement

Figure 7: A: residual plots of cement recovery in (0-1 mm), B: residual plots of cement recovery in (0-0.5 mm)

359
 increased. Recovery of cement is affected by both compression and shearing and it is increased to more than
40% in 0-1mm fraction (see Figures 8A an 8B). Decreasing the amount of water absorption in coarse fraction 4-
16mm, by increasing the amount of force and timing, is another evidence to prove the reduction of cement paste
on the surface of recycled aggregates (see Figures
Lotfi8C&and
Rem8D). The energy consumption raises by increasing the
amount of timing and force . However, according to Figures 8E and 8F even by using the highest amount of
force and timing, the energy consumption is less than 700 kJ/ton (0.19 KWh/ton). It shows that the cost of
autogenous milling could stay in a reasonable range during concrete recycling process.

Contour Plot of Recovery of cement in (0-1mm)

7.8 Recov ery of
cement in
7.2 (0-1mm)
< 0.25
6.6 0.25 – 0.30
0.30 – 0.35
6.0 0.35 – 0.40
> 0.40

Timing(min)
5.4

4.8

4.2

3.6

3.0

2.4

0 5 10 15 20 25 30 35
Force(KN)

A B

Contour Plot of water absorption (4-16mm)

8.0
water
absorption
7.2
(4-16mm)
< 0.038
6.4
0.038 – 0.040
0.040 – 0.042
5.6
0.042 – 0.044
0.044 – 0.046
4.8

Timing(min)
0.046 – 0.048
> 0.048
4.0

3.2

2.4

1.6

0.8

0.0
0 5 10 15 20 25 30 35
Force(KN)

C D

E F
Figure 8: A and B: surface and contour plot for cement recovery in 0-1mm, D and E: surface and contour plot of water
absorption 4-16mm, E and F: surface and contour plot for energy consumption during milling. In all figures duration of back
Figure 8: A and B: surface and contour plot for cement recovery in 0-1mm, D and E: surface and contour plot of
and forth motion is indicated as timing

water absorption 4-16mm, E and F: surface and contour plot for energy consumption during milling. In all figures
duration of back and forth motion is indicated as timing.

No.265189.

References
Florea MVA, Brouwers HJH. Properties of various size fractions of
crushed concrete related to process conditions and re-use. Cement and
Concrete Research. 2013;52(0):11-21.
Greenfield, Tony, and Andrew Metcalfe. Design and Analyse Your
Experiment Using Minitab. Hodder Arnold, 2006.
Kim KH, Cho HC, Ahn JW. Breakage of waste concrete for liberation
using autogenous mill. Minerals Engineering. 2012;35(0):43-5.
King, R. Peter. Modeling and simulation of mineral processing sys-
tems. Elsevier, 2012.
Montgomery DCDaaoeJWS, 2008. Design and analysis of experi-
ments: John Wiley & Sons, 2008.
Lotfi S., Dejab J., Rem J.P., Mróz R., van Roekel E. and van der Stelt
H. (2014): " Mechanical recycling of EOL concrete into high-grade aggre-
gates". Resources, Conservation and Recycling, vol. 87, 117-125.
Lotfi S, Deja J, Rem P, Mróz R, van Roekel E, van der Stelt H. A Me-
chanical Process for In Situ Recycling of EOL Concrete, SB 13, Singapore,
9 - 10 September 2013.

360
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Concrete microscopy
f 0 F

Characterization of ageing effects

361
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

362
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

New concrete types for aggressive

environment; microscopic results from a
research project
Ulla H. Jakobsen, Martin Kaasgaard
Danish Technological Institute
⇤ uhj@teknologisk.dk

Abstract

This paper presents microscopic data from a research project aimed at finding new suitable concrete mixes for sewer
pipes in contact with very aggressive waste. Several mix designs were tested, where both the binder type and aggregate
composition were changed. The concrete was submersed in 1 wt-% sulphuric acid for 105 days and was examined both
before and after submersion in the acid, using polarizing microscopy and SEM-EDX.
Totally, nine different concretes were tested. The binders were respectively CEM I SR with fly ash, calcium aluminate
cement, CEM III/B slag cement and geopolymer. The CEM I SR with fly ash binder was tested with different
combinations of both siliceous and calcareous aggregate, while only siliceous aggregate was used for the other binders.
The test showed various degree of deterioration of the concrete surfaces. Typically, the paste crumbled during the
submersion in the acid leaving exposed aggregates. Depending on the rock type present and its porosity, the aggregate at
the surface was intact, fractured or recrystallized into gypsum.
Except for the concrete with geopolymer, all other binder types showed a similar change in the surface region, where
the paste transformed into gypsum. Not only the paste transformed into gypsum, but gypsum was also seen in surface
parallel cracks and in air voids. In cases where the aggregate (sand and stone) contained limestone (CaCO3 ) this was
also transformed into gypsum.
The project showed that the most suitable mix design studied in order to withstand aggressive waste water appeared
to be concrete containing CEM I SR cement with fly ash and a very fine sand consisting of porous limestone (the coarse
aggregate was normal Danish siliceous material). In this sample, a rather straight, dense zone of gypsum was formed
and almost no scaling of the surface occurred. It was further observed, the denser and more fragile the aggregate, the
more scaling of the surface occurred.
The deepest ingress of acid was in concrete with CEM I SR and fly ash, and concrete with slag cement, both with
normal siliceous aggregate in both the fine and coarse fraction.

Keywords: Optical microscopy, SEM-EDX, acid attack, aggressive environment, new mix design

I. Introduction Limestone aggregate will react with acid and

The chemically aggressive environment within con-
crete sewer pipe structures exposes the inside sur-
face of the structures to sulphuric acid attack. The
cost of maintenance and repair of concrete sewer
pipe is substantial. In Denmark alone the yearly
costs of maintenance, repair and expansion is esti-
mated to 1 billion DDK (Miljostyrelsen, 2006). The
study described in this paper is part of a Danish
research project aiming to develop concrete sewer
pipe structures, which are more resistant to acid
attack in order to achieve a prolonged service life
of minimum 75 years.
Different technologies are known to enhance the
resistance of concrete to sulphuric acid attack, in-
cluding limestone aggregate, calcium aluminate
cement, slag cement and geopolymer binder (Al-
lahverdi, 2000).
thereby function as a sacrificial material, which
neutralizes the acid so attack on the cement paste is
reduced (Chang, 2005). Hydration of calcium alumi-
nate cement results in a paste without Portlandite
and the formation of alumina hydrate, which is sta-
ble down to pH 3-4. Furthermore, the dissolution
of other calcium aluminate hydrates during acid
attack leads to formation of more alumina hydrate
which infills pores (Scrivener, 1999). Slag cement
is known to give a more dense paste microstruc-
ture and thereby the ingress of acid during acid
attack will be reduced (Bijen, 1996). Geopolymer is
a binder composed of fly ash or slag activated by al-
kaline solutions like sodium hydroxide and sodium
silicate. This forms a polymeric network, which
is known to be resistant to acid attack (Thokchom,
2009).

363
 Jakobsen & Kaasgaard

Table 1: Composition and test periods of the analysed samples.

Sample ID Mix design Microscopy before Acid Microscopy after Acid

Binder Sand Stone
RC CEM I + FA Siliceous aggregate Siliceous aggregate X X
K CEM I + FA Limestone, soft Limestone, soft X X
CA Calcium Aluminate Siliceous aggregate Siliceous aggregate X X
G Geopolymer Siliceous aggregate Siliceous aggregate X X
SL CEM III/B Siliceous aggregate Siliceous aggregate - X
NOK CEM I + FA Norwegian lime- Norwegian lime- - X
stone, high density stone, high density
2700 kg/m3 2700 kg/m3
KGRSA CEM I + FA Siliceous aggregate Limestone, soft - X
& soft limestone
KSA CEM I + FA Fine grained soft Siliceous aggregate - X
limestone (>2mm)
KST CEM I + FA Siliceous aggregate Limestone, soft - X
(>2mm)

These technologies are investigated in the study
on samples of earth moist concrete, which is the
concrete type primarily used in Denmark for pro-
duction sewer pipes.
II. Experimental
A total of nine different earth moist concrete mix
designs with different binder and aggregate types
were tested (Table 1). The concretes were designed
with equal water/cement ratio of 0.34 and equal
volume content of paste (and aggregate). The spe-
cific mix designs are presented in another article
(Juel-Hansen, 2014).
Cylindrical samples of diameter 100 mm and
height 80 mm were prepared in a compactor, where
the earth moist concrete was compacted between
two angled eccentrically rotating piston heads, us-
ing a compaction pressure of 4 bar.
After 28 days of storage immersed in water at
20 C, the samples were exposed to a 1 wt-% sul-
phuric acid solution (pH⇠1) in order to investigate
their sulphuric acid resistance based on a modified
version of ASTM C267. At selected time steps up
to 126 days of exposure, the samples were removed
from the solution, rinsed under running water and
weighed (in SSD state). Furthermore, the visual
appearance was recorded and photo documented.
At all the selected time steps, the 1 wt-% sulphuric
acid solution was replaced by fresh solution. The re-
sults from the weight measurements are presented
in another article (Juel-Hansen, 2014). In this arti-
cle, the results from microscopic evaluation of the
samples after 105 days of exposure to the sulphuric
acid solution are presented.
The concrete was analysed using optical polar-
izing microscopy, OPM, and scanning electron mi-
croscopy, SEM-EDX. One fluorescent impregnated
thin section and one polished section were pre-
pared from each test cylinder. The sections were
positioned along and including one of the cylinder
faces.
The thin sections are made by vacuum impreg-
nating a cut sample piece with an epoxy resin con-
taining a fluorescent dye. Subsequently, the impreg-
nated slices are mounted on objective glass and
ground to a thickness of 0.020 mm (20 µm). Finally,
the section is covered by a cover glass. The thin
sections are then examined in a Leica DM2500P op-
tical polarizing microscope using transmitted light,
crossed polarized light, and fluorescent light. The
filter combination used for fluorescent light is a
LWP 530 yellow blocking filter and a BG 12 excita-
tion filter (Jakobsen, 2000). The light source at the
microscope is a 100W halogen bulb. The vacuum
impregnation of the samples with epoxy causes all
voids and cavities in the samples to be filled with
fluorescent epoxy. By transmitting fluorescent light
through the thin section in the microscope, the flu-
orescent epoxy in the various porosities will emit
yellow light that makes voids, cavities and cracks
easy to identify. The fluorescent epoxy also impreg-
nates the capillary pores in the hardened cement
paste causing a dense cement paste with low water
to cement ratio to appear darker green while a more
porous cement paste with a high water to cement
ratio appears lighter green. By this, the water to
cement ratio (w/c) of the concrete can be estimated
with an accuracy of ± 0.02. A detailed description
of the sample preparation is found in (Jakobsen,
2000).
The left over piece from the thin section produc-
tion is again impregnated and then polished using
first fixed diamond disks from Struers (MD Piano)
with grid 500, 1200 and 2000 and then a Texmet
Perforirt cloth from Buehlers with 6 µ m, 3 µm and
1 µm diamond spray. The polishing is performed
using a Buehler Beta Polished with a Vector Power
Head.
SEM-EDX analysis is performed on the polished

364
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 1: Appearance of the concrete cylinder after exposure to acid.

sections using a Quanta 400 from FEI equipped
with a NNS EDX-system from Thermo. An accel-
erating voltage of 20 kV, spot size 5 and a working
distance of around 10 mm are used during analy-
sis. The data are Proza corrected. The sample are
carbon-coated and analysed in high vacuum mode.
III. Results
Microstructural appearance before acid
exposure
In order to check whether the concrete to be tested
was properly cast, the microstructural appearance
of 4 of the concretes were analysed before they
were exposed to the acid (Table 1). The concretes
all appeared relatively well-compacted with a ho-
mogeneous, dense paste. Large irregular air voids
are present. The cylinder sides appear relatively
even with a thin pasta layer covering the aggregate.
No larger cracks are present in the concrete. Mi-
nor difference are however observed, differences
with are related to the binder combination; e.g. the
concrete with geopolymer has initially a distinct
higher capillary porosity than concrete based on
CEM I and calcium aluminate cement, and appears
initially soft.
Microstructural appearance after acid ex-
posure
After the 105 days exposure to acid, the con-
cretes show different degree of deterioration (Figure
1). The concretes showing the most distinct scal-
ing/abrasion are the reference concrete (RC), the
concrete with calcium aluminate cement (CA), the
CEM III /B concrete (SL), and the concrete with
CEM I+FA and coarse limestone aggregate (KST).
An intermedia scaling/abrasion is observed in con-
crete with dense Norwegian limestone in both frac-
tions (NOK), and in concrete with limestone in
the coarse fraction and both limestone and Dan-
ish siliceous material in the sand fraction (KGRSA).
The concretes, which apparently have performed
best, are the two concretes containing soft limestone
in the sand fraction, (K) and (KSA).

Figure 2: Appearance of the cylinder face, which has been in contact to the bottom of the acid container.

365
 Jakobsen & Kaasgaard

Table 2: Summary of observations regarding the surface region. ⇤ Measured from the outer surface of exposed aggregate,
app. 2.5 mm paste has disappeared.

Sample ID Width of surface Width of gypsum With of iron-rich Carbonation band, Gypsum in air
change, mm zone, mm band, mm mm voids, mm
RC 2-3 1.5 0.5 0.8 4-5
K 1.2 0.4 0.2 0.6 2
CA 2 2 not obs. not obs. 4
G 0 0 0 0 9
SL 3-4 3-4 0.1 0.4 4-5
NOK 1.5 1.2 0.2 0.2 6
KGRSA 1-2 0.8 (2) 0.2 0.4 4-5
KSA 0.4 (1.2) 0.6 0.06 0.2 2.5
KST 4* 1.5-2.5 0.2 0.2 6

Almost no deterioration of the surface is seen in increase of S can be traced.

the concrete with geopolymer (G); but generally,
this concrete appears very soft and porous.
The face of the cylinders in contact with the bot-
tom of the acid container shows a distinct rusty-red
staining (Figure 2). The degree of staining appears
to correlate positively with the iron content of the
cement; most intense staining is seen on the calcium
aluminate concrete (CA) and least on the CEM III/B
concrete (SL). Of the CEM I concretes, the reference
(RC), without any limestone at all, show the most
staining.
All concretes, except the one with geopolymer
(G), show distinct evidence of acid attack in the
surface region (Table 2, Figures 3 and 4). From
the surface and inwards, a 0.4 to 4 mm thick zone
appears where the microstructure of the concrete
is changed. This changed zone is generally more
porous compared to the interior non-affected paste.
In the concrete with geopolymer, the influence of
the acid is only visible due to the presence of gyp-
sum in air voids and chemically by EDX a weak
Based on the mineralogy and chemical compo-
sition, the porous surface region can be divided
into 3 or 4 zones, starting with an outer S-rich zone
(Figure 3) containing gypsum, then a thin band rich
in iron and or aluminium/silicon (Figure 4), and fi-
nally a thin carbonated zone occurs before entering
non-affected paste.
Even though the paste behind the carbonated
area appears non-affected by the acid, gypsum oc-
curs in air voids to depths more than twice the
depth of the distinctly affected surface region.
Gypsum zone
In the surface region, gypsum is typically replacing
the cement paste and replacing limestone aggregate
if present and is occasionally observed in surface
parallel cracks and air voids.
In concretes without limestone aggregate (RC,
CA and SL), the paste and occasionally lime-
containing porous flint aggregate converts into gyp-

Figure 3: The surface of the concrete is enriched in S to different depth. The horizontal arrow indicates that the S-
ingress of sample KST actually is much deeper as at least 2.5 mm of the surface has scaled.

366
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 4: In front of the S-enrichment, a narrow zone rich in Al occurs. In front of the peaks, Al is depleted.

sum (Figure 4). In these concretes, surface parallel
cracks are frequent, and typically they directly tra-
verse dense aggregate such as chalcedonic flint. The
cracks are mainly filled with gypsum but de-bond
in the adhesion zone to the aggregate. Concrete
with this texture appears to be the most porous and
show the most severe crumbling and scaling at the
surface.
The concretes containing coarse limestone aggre-
gate (NOK, KGRSA and KST) are in many ways
similar to the concretes with Danish siliceous aggre-
gate (RC, CA and SL), as cracks in these concretes
also appear to penetrate the outermost positioned
aggregate, even though they are limestone. The
KST concrete containing coarse soft limestone but
siliceous sand is the concrete showing the most
severe scaling (Figure 6).
In the concretes containing soft limestone sand (K
and KSA), both paste and limestone sand is totally
converted into gypsum forming a straight relatively
dense zone, where typically only few cracks are
observed, except in the front of the gypsum zone
(Figure 7 and 8). In these concretes, debonding are
not observed within the adhesion between aggre-
gate and gypsum-filled cracks.
Figure 9 shows examples from the X-ray mapping
with distribution of S, Al and Ca. As seen, all
aggregate as well as paste in surface of concrete
KSA appears as one straight S-rich zone (turquoise)
with few S-containing cracks at the interface to
the more normal paste. The paste of the surface
region of concrete KST is also rich in S, but the

Figure 5: Typical deterioration pattern at the surface of concrete with siliceous material in both the coarse and fine
aggregate fraction. Sample RC. Exposed surface is towards the top of the image.

367
 Jakobsen & Kaasgaard
Figure 6: Similar deterioration pattern as seen in con-
crete with pure siliceous aggregate as is seen in concrete
with siliceous sand, and limestone in the coarse fraction.
Sample KST. Exposed surface is towards the top of the
image.
Figure 7: In concrete with porous limestone in both the
fine and coarse aggregate, aggregate as well as pasta
converts into a dense, uniform gypsum zone. Sample K.
Exposed surface is towards the top of the image.
highest S-concentration can be recognized in cracks.
No conversion of the siliceous aggregate is seen in
concrete KST.
Chemically, the maximum content of S found
in the zones are very alike at a maximum level
between 12-16 at.%. The binder composition does
therefore not seem to influence the max. S the
sample can carry, but it distinctly influences the
depth of S-ingress. This feature is illustrated in the
bar-diagrams in Figure 13.
368
Iron- and aluminum/silicium-rich bands
In front of the gypsum zone, a distinct concentra-
tion of iron and aluminum is observed (Figure 10).
The iron band is typically less than 0.5 mm thick,
the aluminum less than 0.3 mm, but the two "zones"
are somewhat intermixed with each other.
The iron band appears dense, with a Fe content
above the background level of the various pastes.
The highest content of about 5 at.% is found in
the sample with calcium aluminate cement, CA.
In the other samples, the Fe content is around 1
atomic% in this zone. In optical microscopy, and
even macroscopically, the band can easily be seen
by its distinct rusty red colour (Figure 10 & 11),
similar to the colour observed on the bottom of the
cylinders, when removed from the acid container
(Figure 2).
The aluminum/silicium-enriched band is diffi-
cult to observe by OPM, but fluorescent light and
SEM reveals a porous band rich in empty, very tiny
small, random oriented cracks (Figure 12). The max-
imum Al content of the zone varies depending on
binder type from 3-6 at.% in samples with Portland
and Slag cement to 18 at.% in the sample with cal-
cium aluminate cement. This feature is illustrated
in the bar-diagrams of Figure 13.
Carbonation
Just at the interface between the porous Al-band
and "non-affected" paste, a thin band, less than 0.5
mm thick, of slightly carbonated paste appears. The
band is easy recognized by OPM, but difficult to
see using only SEM.
Figure 8: In concrete with soft limestone in the sand
fraction but siliceous sand in the coarse fraction a similar
dense gypsum zone is formed at the surface. Sample KSA.
Exposed surface is towards the top of the image.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 9: X-ray mapping showing the distribution of S (turquoise), aluminium (red) and calcium (blue) in concrete
KSA to the left and sample KST to the right. Exposed surface is towards the top of the image.
Interior non-affected paste
The interface between affected and non-affected
paste is generally sharp. The paste in front of the
Figure 10: A distinct, wide, rusty red zone is seen in
front of a dense gypsum zone in concrete KSA. Exposed
surface to the left.
Figure 11: A distinct, thin, rusty red zone is seen in
front of a highly cracked surface in concrete SL. Exposed
surface to the left.
interface looks un-affected by the acid, but do show
evidence of the attack as gypsum crystals appear
in air voids to depths of max. 9 mm. No distinct
build-up of ettringite occurs at the exterior of the
un-affected paste.
IV. Discussion and Conclusion
The petrographic study illuminates the different
chemical changes and degradation of the concrete
surfaces when exposed to a strong acid.
As described, the concretes showing the deepest
ingress of S, are the concretes with slag cement (SL),
the reference concrete with Portland cement and fly
ash (RC), and the concrete with Portland cement
and fly ash, but with limestone in the coarse aggre-
gate fraction instead of siliceous aggregate (KST).
These three concrete mixes are also the concretes vi-
sually showing the most distinct scaling and crum-
bling of the cylinder surfaces. When looking at the
data from the mass change study performed on
the same samples (Juel-Hansen, 2014), the slag ce-
ment concrete does, however, have a very low mass
loss, especially much lower than the Reference (RC).
Therefore, from a mass loss point of view, the slag
cement concrete performs well, but from an acid
ingress point of view, it performs the worst of the
studied concretes. The reason for this conflicting
data is not yet completely clear.
In the microscope, the cause of the severe surface
damages seems to be related to the formation of
surface parallel cracks, infilling of gypsum, debond-
ing, cracking of aggregates and decalcification of
paste. Resultantly paste erodes, leaving exposed
damaged aggregate at the surface.
The concrete with the absolute less S-ingress is
the one where the siliceous sand material is sub-
stituted by limestone (KSA). This concrete is also
369
 Jakobsen & Kaasgaard

Figure 12: In all concrete independent on binder and aggregate type, a distinctly cracked zone is observed between the
gypsum zone and the non-affected interior paste. This zone is enriched in Al and Si. Exposed surface is towards the
top of the image.

Figure 13: Bar diagrams showing the variation in S (blue), Al (red) and Si (green) in the gypsum-rich surface region,
the interface region and in the interior paste. The content is expressed relative to Ca (at.%). Each bar represents the
average of several analysis performed in the zones.

370
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
one of the concretes showing the least scaling of the
surface area. The microanalysis show, that in this
concrete, and in concrete with similar limestone ag-
gregate, the paste (CSH) as well as the limestone is
converted into gypsum, forming a relatively dense
gypsum zone, where only minor abrasion occurs.
Furthermore, only minor surface parallel cracking
of paste and aggregate is observed. The mass con-
version of the CSH binder as well as the limestone
sand to gypsum seem to have a protecting effect
instead of a damaging effect.
So why do we see damage related to crack for-
mation in some concretes and in others full paste
conversion and less deterioration? According to
Chang et. al, the exchange of normal aggregate
with limestone aggregate neutralises the sulphuric
acid, reducing the attack on the cement paste. Their
theory is simply that the limestone is sacrificed dur-
ing the attack. This study show, the finer and the
more soft the limestone aggregate becomes the bet-
ter the effect of neutralization. So why cracking? It
appears that in the most scaled concrete aggregate
do also crack. Actually, it is mainly the lime con-
taining flint aggregate, which crack. It is speculated
whether the acid entering these typically porous
aggregate, react with calcium, forms gypsum that
resultantly crack the aggregate and the surrounding
paste. This may explain the severe scaling observed
in concrete containing dense Ca-containing rock
material in the aggregate fraction.
The analysis further show that acid does enter
the surface of all examined concretes; even the one
with geopolymer, however, in this concrete the acid
does not deteriorate the paste as the content of Ca
present in the binder is too low for gypsum to form.
The average S level of the surface region is however,
constant in all concrete independent on binder and
aggregate type. The average S/Ca level is around
1.2 at.%; probably indicating the maximum S ca-
pacity of the concrete when concerting to gypsum
(S/Ca of pure gypsum equals 1 at.%).
Acknowledgement
The authors would like to thank all the involved
project partners and in particular the Environmen-
tal Protection Agency of the Danish Ministry of the
Environment for their funding of the project.
References
Allahverdi, A. and Skvara, F. (2000): Acidic corrosion of hydrated
cement based materials. Part 2 - Kinetics of the phenomenon and math-
ematical models. Ceramics - Silikaty 44 (4) pp. 152-160.
Bijen, J. (1996): Blast furnace slag cement for durable marine struc-
tures. Stichting BetonPrisma.
Chang, Z, T., Song, X,J, Munn, R. and Marosszeky, M. (2005): Using
limestone aggregates and different cements for enhancing resistance of
concrete to sulphuric acid attack. Cement and Concrete Research 35, pp
1486-1494.
Jakobsen, U.H., Laugesen, P. and Thaulow, N. (2000): Determination
of water to cement ratio in hardened concrete by optical fluorescence
microscopy. ACI Symposium Volume: " Water-Cement ratio and other Dura-
bility Parameters: Techniques for determination". SP 191, pp. 27-42.
Juel-Hansen, L., Kaasgaard, M., Anderson, J. and Pade, C. (2014):
Development of resistant concrete pipes to be exposed in aggressive en-
vironments. Proceedings of the XXII Nordic Concrete Research Symposium
pp. 495-498.
Miljostyrelsen (2006): Aflobssystemts levetid og renovering (in Dan-
ish). Miljoprojekt 1065.
Scrivener, K.L., Cabiron, J.L. and Letourneux, R. (1999): High perfor-
mance concretes from calcium aluminate cements. Cement and Concrete
Research 29, pp 1215-1223
Thockchom, S. Ghosh, P. and Ghosh, S. (2009): Acid resistance of fly
ash based geopolymer mortars. International Journal of Recent Trends in
Engineering, Vol 1, No 6.
371
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

372
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Visby 1993 re-visited:

Understanding the effective W/C ratio, as
influenced by aggregates and air voids
Peter Laugesen
Pelcon Materials & Testing ApS, Denmark
pl@pelcon.dk

Abstract

When adding more and more (SSD: Saturated Surface Dry) sand to a mix of cement paste, one observes that it dries
out and loses workability. This was explained by Laugesen (1993) and has since proven its worth in understanding
concrete microstructure, not least the water cement ratio in relation to (capillary) porosity.
This paper recalls the findings of Laugesen (1993) presented at the 4th Euroseminar in Visby, Gotland, and draws
lines to a series of examples from real-life concrete. Furthermore, it is emphasized that general laboratory studies and
research based on cement-based materials essentially needs to take this into account.
Finally it is introduced, that the wall effect is also relevant for air void systems, trapping moisture at the rim and
depleting the general cement paste for water. Microstructure details strongly supports this hypothesis, as well as
experience from real life show that air entrained concrete have less tendency of bleeding.

Keywords: cement packing, wall effect, ITZ, ’effective w-c ratio’, air void influence on w-c ratio.

I. Influence of sand & stones on the pletion of the water content of the general cement
‘effective water cement ratio’ of paste. The more large grains (i.e. the more ‘large
hardened concrete grain surface area’ the more the cement paste will
be generally lowered in water content; hence the
bulk cement paste will be denser, having a lower
When adding more and more (SSD: Saturated Sur-
(capillary) porosity corresponding to a lower water-
face Dry) sand to a mix of cement paste, one ob-
cement ratio. All in all this redistribution of water
serves that it dries out and loses workability. This
in the cement paste of concrete and the implica-
was explained by Laugesen (1993) and has since
tions to performance and analysis is the scope of
proven its worth in understanding concrete micro
this presentation.
structure, not least the water cement ratio in rela-
tion to (capillary) porosity.
This paper recalls the findings of Laugesen (1993)
presented at the 4th Euroseminar in Visby, Gotland,
and draws lines to a series of examples from real-
life concrete. Furthermore, it is emphasized that
general laboratory studies and research based on
cement-based materials essentially needs to take
this into account.
As pointed out in (Laugesen, 1993) and (Scrivener
et al., 2004) the packing of cement grains in a mor-
tar or concrete is influenced by larger objects such Figure 1: Illustration of the “wall” effect. A flat solid
as sand grains and coarse aggregates, a fact that object placed at random in an assembly of cement grains
can be described by the ‘wall effect’ (Figure 1). In would cut through grains. As this is impossible the pack-
short the normal packing and ‘interlocking’ of ce- ing of grains is disrupted to give a zone of higher porosity
ment grains in the paste are disturbed/reduced by and smaller grains in the zone close to the “aggregate”.
a larger (sand) grain replacing an array of fresh From (Scrivener et al., 2004).
cement paste. Along the boundary the increased
space is filled by: Water! - being the only available The implied ‘halo’ of increased (capillary) poros-
easy-moving component. This, in turn, causes a de- ity around the sand grains and stones that must be

373
 Laugesen

Figure 2: Brighter ‘halos’ around sand grains in concrete mixed with coarse grained Portland cement. Microscope
mag.: x63; UVL (right-hand photo manipulated to highlight the ‘halos’). Thin section by courtesy of Dr. Yilmaz
Akkaya; ITU.

Table 1: Selected data for the Examples 1&2

Data: Suggested
1 Waterbinding, mean layer thickness, mm 0,005
Sand grain size (specific surface, a), mm 1 20
Sand grain size, r, mm 0,15
Density of sand, kg/m3 2650
Density of water, kg/m3 1000
Calculation of water binding, given the data above Volume, mm3 Weight, g
One sand grain, (4/3 ⇥p ⇥ r3 = 4/3 ⇥p ⇥ (0.153 )) 0,0141 3.75⇥10 1

Sand grain + water binding, (4/3 ⇥p ⇥ r3 = 4/3 ⇥p ⇥ (0.15+0.005)3 ) 0,0156

Water binding 0,0015 1.46⇥10 2

Water binding, weight % of SSD sand 3,9

1 This is a suggested and probably not correct value - and it is surely NOT a uniform film!!

Table 2: Example 1: Influence of water binding on ’effective W/C-ratio’ in concrete.

Cement, kg/m3 380

Free water, kg/m3 171
Water-cement ratio 0.45
Ssd sand, kg/m3 700
Water binding, kg/m3 - Applying the data from Table 1 27
’Effective’ w-c ratio of bulk cement paste 0.38

Table 3: Example 2: Design of concrete with bulk properties as 0.40 pure cement paste

Sand, kg/m3 700

Water binding, (at example density), kg/m3 27
Cement, kg/m3 380
Normal concept,of free water, kg/m3 152
Total water,content for w-c ratio calculation, kg/m3 179
Water-cement ratio, as normally calculated 0.47
’Effective’ w-c ratio 0.40

374
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 3: Same cement paste with additions of no, little and some sand, respectively. Microscope mag.: x200, PPL.
expected as a result of the above, are rarely seen
in the analysis of fluorescence impregnated thin
sections. One exception is in case of coarse grained
cement being used, re. Figure 2. This implies that
the amount of moisture being trapped along the
sand grain boundaries is related to the Blaine value
of the cement: the coarser the cement the more
moisture is trapped.
The straight forward explanation of the apparent
absence of a bright ‘halo’ may be that the trapped
water does not constitute a uniform film, but is
rather placed in the local small vacancies that would
be taken by cement grains be it not for the sand
grain. Furthermore, the relatively large thickness of
thin sections of 20-30 microns may blur the appear-
ance of an expected ‘halo’. Finally the precipitation
of portlandite at the ITZ as detailed later, to some
degree compensates the formation of ‘halos’.
All in all the water trapped locally at the sand
grain surface leads to larger local porosities here,
and reduced (capillary) porosity in the bulk cement
paste. This is easily seen in thin section in PPL due
to denser packing of cement grains in bulk paste,
re. Figure 3. Also in UVL the fluorescence thin
sections will readily reveal the denser bulk cement
paste due to reduced porosity, re. Figure 4.
The implication of the actual microstructure, ce-
ment packing and local water-cement ratio (capil-
lary porosity) is sought quantified in the calcula-
tions given in Table 1, 2 and 3.
For a typical concrete mix design, as in the Exam-
ples 1 and 2 (in Table 1, 2 and 3), the implication of
the water binding is seen to be at least of the mag-
nitude as the influence of aggregate absorption.
Please keep in mind that water binding at the
sand/aggregate surfaces is what actually happens
in all concrete mixed!
The implication to the cement paste, as studied
in e.g. fluorescence impregnated thin sections, is
that the bulk cement paste is already denser, and
the ITZ (Inter Transition Zone) is more porous. This
does influence the analysis we perform as detailed
in the following.
Due to the wall effect the bulk cement paste was
, as calculated in Table 1, actually changed from
a w-c ratio of simply calculated 0.45 to an actual
0.38! I repeat: This is what has actually already
happened in all concrete samples that we analyze!
Working with pure cement paste of 0.40 in w-c
ratio, for e.g. research purposes, one will actually
perform the experiments on a microstructure that
corresponds to a w-c of (normally calculated) 0.47
in concrete, as seen in the example in Table 2; with
other sands and mix designs these values change.
This implies that ANY research results obtained
from cement paste of w-c ratio 0.40 CANNOT be
directly related to a concrete of (normally calcu-
lated) w-c ratio of 0.40.
It must be emphasized that the applied ‘thick-
ness’ of the water binding of 0.005mm has not been
thoroughly documented for a range of concrete mix
designs. However, a series of carried out experi-
ments indicate that the magnitude may not be all
too wrong.
II. Relations with influence on the
’effective w-c ratio’
Normal concrete mix designs actually have rela-
tively small variations in sand content; furthermore,
the fineness of the sand (⇠ the ‘water demand) is
compensated by adjusting the sand content. This
implies that the actual variations in ‘effective water-
cement ratio’ are not as large as may first be antic-
ipated; which is great for the field of thin section
fluorescence analysis...
Another issue is the location of the portlandite
375
 Laugesen

Figure 4: Same cement paste with additions of little, some and much sand, respectively. Microscope mag.: x63, UVL.

being formed – due to the larger available spaces
at the sand grains and stones, the porous ITZ is
somewhat densified by the precipitation of port-
landite. Some literature erroneously describes this
as a ‘rim-of-portlandite’ being formed round the
aggregate grains.
III. Influence of air void structure
on effective water-cement ratio
Playing with air bubbles you will find that when-
ever a bubble bursts it leaves a wet mark on your
hand, shirt or on the ground – wherever it bursts.
Air voids in fresh concrete acts as a particle,
hence to move an air void it needs to ‘push away’
the cement paste, sands or stone in front of it. This
‘particle’ would have similar implications to trap-
ping water as seen in the following.
With the analogy of the sand grains in the cement
paste, it is hereby claimed that the formation of air
voids likewise traps moisture around the periphery
of the air void. This in turn further depletes the bulk
cement paste for moisture. The microstructural
features described for the general wall effect seems
also to be relevant for the ’air void ITZ’; hence
larger cement grains are not located right at the rim,
and portlandite crystals may well be concentrated
locally at the rim; re. Figure 5.
It is a well-known fact from real life, that bleeding
of a mix design of concrete can be remedied by
adding an air void system to the concrete. With the
above we may now understand why.
The influence on w-c ratio of the cement paste
has not been assessed by experiments, but from
the small calculation below, it appears to be on a
smaller scale that the sand, as may be expected.

Figure 5: Cluster of air voids in concrete; note that the rim of each air void is very complete and smooth, that the
larger cement grains are placed away from the rim (same as typically seen at the ITZ), and that some portlandite is
precipitated along the rim (right-hand photo). Microscope mag.: x100, PPL and XPLG (right).

376
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 4: Selected data for the Examples 3

Data: Suggested
2 Waterbinding, mean,layer thickness, mm 0.005
Mean air void,size, a, mm 1 30
Air void size, r, mm 0.1
Density of water, kg/m3 1000
Calculation of water binding at air voids Volume, mm3
One mean air void (4/3 ⇥p ⇥ r3 = 4/3 ⇥p ⇥ (0.13 )) 0.0042
Air void + water binding (4/3 ⇥p ⇥ r3 = 4/3 ⇥p ⇥ (0.1+0.005)3 ) 0.0048
3 Water binding 0.0007
Water binding, vol.% of air content 15.8
2 This is not the correct value but possibly a rather uniform ‘film’ is formed.
3 Apparent ‘rounding error’.

Table 5: Example 3: Influence of water binding at air voids on ’effective w-c ratio’

Cement, kg/m3 380

Free water, kg/m3 171
Water-cement ratio 0.45
Air content, vol % 5.0
Water binding at air voids, kg/m3 7.9
’Effective’ w-c ratio 0.43

However, with the outlined magnitude, this phe-
nomenon should still be taken into account.
The examples presented in Table 4 and 5 show
that by entering an air void system of 5% air with a
specific surface of 30 mm 1 , the effective w-c ratio
of the cement paste is lowered from 0.45 to 0.43.
HOWEVER: It must be noted that the here ap-
plied ‘thickness’ of the water binding at the air
voids of 0.005 mm has not yet been experimentally
documented!
IV. Concluding remarks
What we analyze at fluorescence microscopy of the
cement paste in concrete thin sections, using UVL,
is (capillary) porosity – this is an enormously strong
tool, unsurpassed by others in the allowed detail,
quantification and ease of analysis. Any correlation
with this porosity to composition and w-c ratio
needs to be based on reference samples.
It should be emphasized that we have ALWAYS
used the bulk (capillary) porosity of the cement
paste, i.e. the ‘effective’ w-c ratio in our thin section
analysis, as demonstrated in the present paper.
Looking at old concrete to evaluate the durability
and present condition, the actual chloride ingress
boundary and the carbonation fronts in general,
surely appear to be a uniform ingress front without
apparent curving neither along the coarse aggre-
gates nor the sand grains. (One very large exception
from this is naturally the many concrete structures
where the concrete was bleeding, or the microstruc-
ture is inhomogeneous, due to otherwise poor mix
design, inadequate mixing, and/or poor workman-
ship, such as e.g. over vibration.)
Hence, the capillary porosity of the bulk cement
paste is one of the main controlling parameters
in concrete durability. This feature is easily, fast
and accurately assessed by fluorescence thin section
microscopy, a tool that should be implemented in
concrete research to a much higher degree than at
present.
References
Laugesen, P.: Effective W/C Ratio of Cement Paste in Concrete, in
Jan E. Lindqvist et. Al. (eds.) Proceedings of the Fourth Euroseminar on
Microscopy Applied to Building Materials. Visby - Sweden, 1993
Scrivener, K.L, Crumbie, A.K. & Laugesen, P.: The interfacial transi-
tion zone (ITZ) between cement paste and aggregate in concrete: Special
Issue: Cementitious materials. Interface Science, October 2004, vol. 12,
no. 4, pp. 411-421(11).

377
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

378
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

SEM supervised image analysis

contribution on understanding sulfate
attack on blended cement mortars
Samuel Meulenyzer⇤ , Eric Brouard, Didier Lapillonne, Rémi Barbarulo
Lafarge Centre de Recherche (LCR), St Quentin Fallavier, FRANCE.
⇤ Samuel.Meulenyzer@lafarge.com

Abstract

The mechanisms of deterioration in concrete due to sulfate exposition are amply studied over years. This study
presents the results that tend to explain the degradation mechanisms of five different mortars, including regular OPC and
OPC-additions systems (slag and limestone filler) immerged in sulfate solutions (16g/L). In particular, we try to explain
through quantitative measurements the benefic role of certain mineral additions. We have thus associated conventional
characterization technics (such as SEM, EDX and expansion measurements) to quantitative mineralogical maps (so
called ‘mineralogical gradients’) obtained by image analysis using recent supervised classification development. In this
study we formalized through these quantitative mineralogical maps obtained by image analysis that microstructural
changes which occur when sulfate ions penetrate into mortars are visible in zones, as mentioned by several past studies.
These maps allowed us to convert images in mineralogical profiles in which portlandite, monosulfoaluminate (MSA),
gypsum, porosity and additions are precisely quantified in function of depth and time. We have especially focused our
interest on the quantity of MSA pockets and tried to relate their distribution on their ability to resist on sulfate attack.

Keywords: sulfate attack, multispectral imaging, SEM-EDX, mortars, expansion

I. Introduction ets’. Combining these ‘maps’ to more conventional

Sulfate ions from groundwater and soils may in-
duce concrete deterioration. This type of degra-
dation is generally attributed to the precipitation
of ettringite from the aluminum-bearing phases of
the hydrated cement paste (mainly calcium mono-
sulfoaluminate) and sulfate ions from the concrete
environment, leading to internal stresses and expan-
sion. Recent research (Yu, 2013) has shown that the
initial stage of expansion could be more specifically
related to the formation of microcrystals of ettrin-
gite from calcium monosulfoaluminate confined
within C-S-H. On the opposite, monosulfoalumi-
nate in the so called ‘capillary porosity’ (pockets)
would convert to ettringite in presence of sulfate
without causing stress in the material, thus playing
the role of ‘buffer’ by consuming sulfate without
causing any harm.
The development of a new image analysis tech-
nique using supervised classification allowed us to
better quantify the role of the ’buffer’ monosulfoa-
luminate. Indeed, with this technique it becomes
possible to classify objects of similar chemical com-
position of a few µm2 , on a representative area (sev-
eral mm2 ), in a reasonable time. The ’mineralogical
maps’ obtained allow to quantify for the first time
the distribution (in space and size) of these ’pock-
techniques (expansion of mortar prisms, SEM /
EDS spot analyses) helps to confirm the origin of
these ’pockets’, their fate under sulfate attack and
their role in expansion.
Besides, OPC-limestone mixes are generally con-
sidered to demonstrate poor durability in sul-
fate environments. At low temperature (typically
5 C), thaumasite precipitation (and C-S-H dissolu-
tion) can be the reason for this poor performance
(Barcelo, 2014). However it is not the case at higher
temperature, in which case poor performance of
OPC-limestone blends That has been attributed to a
‘dilution effect’ (in case OPC is simply replaced by
limestone), leading on one hand to a higher poros-
ity and higher diffusivity of the hydrated cement
paste thereby increasing the ingress of sulfate ions
(Dhir, 2004) and the other hand to lower mechanical
properties. Only a few studies have investigated the
possible mineralogical and microstructural reasons
for this poor performance (Schmidt, 2008). As for
pure OPC, Mineralogical maps combined with spot
analyses reveal that in presence of limestone filler
there are no pocket of monocarboaluminate formed
that could lead to a more important formation of
microcrystals of ettringite and to earlier expansion.
The application of the new image analysis tech-
nique described here give insights on the potential

379
 Meulenyzer et al.

Table 1: Chemical composition of materials.

LOI at 950 C SiO2 Al2 O3 Fe2 O3 CaO MgO SO3 K2 O Na2 O

% % % % % % % % %
CEM I 52,5 N 1.13 20.28 4.76 3.05 64.47 0.81 3.49 0.78 0.17
Limestone 40.74 6.43 0.36 0.16 51.4 0.32 0.02 0.04 0
slag 36.09 11.44 42.09 7.62 2.03 0.41 0.27

role of limestone in sulfate environments. Supervised classification image analysis

This study investigates mineralogical evolutions
in deterioration of three mortars based on a CEM
I, a mixture of CEM I with 30wt% slag and CEM
I with 30 wt% limestone filler immersed for two
years in a sodium sulfate solution.
Material and mixture characteristics
The CEM I cement was commercial cement con-
taining a medium amount of C3 A. The main oxide
contents, the main phase contents and other char-
acteristics of these materials feature in Tables 1 and
2.
The volumes of the cement or combination, sand
(AFNOR sand) and water were kept constant for all
mortars (EN mortar) with a cement or combination
content of 510 kg/m3 and w/cm ratio of 0.50.
The test solution contains sodium sulfate, 16 g/l
SO4 (23.66 g/l or 0.17 mol/l sodium sulphate). The
test solution was prepared with tap water. The solu-
tion volume to mortar volume ratio was fixed at 5.2.
This solution was renewed after every measurement
(8 times over 2 years).
II. Experimental procedure
Length changes
For every mortar 6 bars (20⇥20⇥160 mm) were
casted. The all specimens were first cured for 28
days in tap water in capped plastic boxes. The water
volume to mortar volume ratio was fixed at 5.2. At
the end of the curing time, 3 bars were placed into
the sodium sulphate solution while the 3 others
stayed in water. Water was never changed. Length
changes of all the specimens were measured after
28, 56, 84, 144, 196, 231, 252, 310 and 365 days of
exposure and the level of visual degradations was
quantified thanks to a "visual degradation index".
As it is detailed in (Meulenyzer, 2013), the ensemble
of a BSE-SEM image and its corresponding SEM-
EDS elemental maps forms a cube dataset, in which
each pixel represents a vector of B elements, or,
more generally, B bands. For SEM-EDS, we are
concerned with 10 or more major elements, thus
making the image a large multispectral image.
Due to the high dimensionality of a multispectral
image, we have use an advanced image classifi-
cation algorithm that has shown excellent perfor-
mance on classifying such images. This ’spectral-
spatial’ algorithm consists on exploiting both the
spectral information in each pixel, as well as the
spectral information of the neighboring pixels.
The spectral classification algorithm is based on
a Support Vector Machine (SVM) algorithm. In
the SVM approach, 2 main inputs are needed: i)
a multispectral image (e.g. the SEM-BSE image
+ SEM-EDS maps), and ii) ’training’ or ’learning
points. The training points are user-defined areas
in the multispectral image, which define the spec-
tral ’signature’ of each of the classes in the image.
The signatures of each phase are then used by the
SVM algorithm to efficiently divide the spectral
space into regions by maximizing the margin be-
tween classes. When the multispectral image of
interest is analyzed, each pixel, with its correspond-
ing spectra, is mapped and then assigned to a class.
The probability for a pixel belonging to that class is
also calculated.
The SVM algorithm is a ’pixel wise classifier’,
since it classifies each pixel using only the spectral
information at each pixel, without regard to the
properties of the neighboring pixels. This approach
has limitations in real images, where neighboring
pixels are generally coupled, or related.
This coupling is especially true for SEM-EDS
maps, where the X-ray signal is generated from

Table 2: Phase content (Rietveld analysis) and fineness.

Alite Belite Aluminate Ferrites Specific gravity SSB Glassy phase

% % % % g/cm3 cm2 /g %
CEM I 52,5 N 60.2 19.2 5.2 10.2 3.14 3561
Limestone 2.72
slag 2.91 3656 99.6

380
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 1: Expansion curve in function of time for the studied systems. Only T1 and T2 are shown but all experiments
are followed up to 720 days (2 years), corresponding to T3 in the next sections, and later.
a sub-surface volume that is larger than the size of
the pixel. To overcome this coupling effect, the SVM
classification map is ’regularized’, or refined. Here,
we use a Markov Random Field (MRF) regulariza-
tion strategy. The MRF algorithm is based on the
assumption that it is generally likely to have neigh-
boring pixels belonging to the same class. To do
this, the MRF algorithm compares the class prob-
abilities (calculated by the SVM) of the pixel of
interest and of each neighboring pixel, and assigns
the most likely class (with highest probability) for
the pixel of interest.
Sample preparation and EDS measurements
Each sample was prepared using a classical Log-
itech Pack 301r epoxy resin is used and the two
constituents of resin are mixed with 4 volumes of
resin for 1 volume of hardener during around 1
minute. Before impregnated the powder or the
paste, the epoxy was curing-vacuum during 5 min-
utes in order to remove enclosed air. Grinding steps
have been automatically done with slight down-
ward pressure applied using progressively finer
SiC powders and Buthandiol alcohol solution on a
slotted steel plate. An optical microscope is used
to control the final quality of surface. Then sam-
ples have been polished using successively finer
diamond paste in the following succession: 1 µm
and 0.25 µm during several hours. After the sam-
ple was cleaned, the specimen is then coated with
carbon to provide a conductive surface for SEM
microanalysis.
The microstructural changes were studied with
backscattered electrons and X-Ray maps in the SEM
(FEI Quanta 400 ESEM FEG with an acceleration
voltage of 15 keV). The changes of the elemental
composition were followed by energy dispersive
X-ray spectroscopy (EDS). The EDS measurements
were corrected with the ZAF correction. Oxide con-
tents were calculated from the measured elements.
III. Results and discussion
Expansion on mortars
Expansion of CEM I and CEM I + 30% slag mortar
bars were very similar over a year. Expansion of the
CEM I + 30% limestone filler mortar started earlier
and was nearly three times higher after one year
of immersion. Divergence of slopes of the CEM I
and the CEM I + limestone was significant after 150
days.
Quantitative sulfate concentration in function of
depth
In order to determine the minimal depth in which
the gradients are representative (ie until the varia-
tions of phases is no more measurable), we have per-
formed SO3 (atomic %) quantification using SEM-
EDS as described before. SO3 profiles are repre-
sented in Figure 2.
We illustrate through this Figure the methodol-
ogy used to measure SO3 concentration from the
surface to core of sample. We have decided to
choose areas (at magnification ⇥2000) in the paste,
included all phases without taking in account sand
aggregates. This allowed us to have a ’global’ con-
tent of SO3 in the paste, and only in the C-S-H. As
shown in Figure 2, SO3 concentration is stable after
3 mm from the surface. We have in consequence
chosen 3.5 mm as the maximal depth analyzed in
the next sections. It will correspond to 9 field of
view at the magnification ⇥350 (with an image size
of 1024⇥884 pixels).
381
 Meulenyzer et al.

Figure 2: SO3 concentration measured by SEM-EDS in function of depth for the studied systems.

General overviews of microstructures are given in the legend.

In order to determine a map of gradients profile Gradients measurements and representativeness

for each mineral phase, we have performed image
analysis as described in section 2.2.4 on individual
multispectral set (ie BSE + elementary mapping of
Al, Si, Ca, S, Na Mg, K, Fe, Mn, Ti). Each view in
Figures 4 and 5 is composed by a mosaic of 9⇥8
individual images. Each image (1024⇥884) was
taken at ⇥350 magnification. As a reference before
sulfate ingress (and deterioration), we show also a
28 day image for each mortar, excepted for OPC-
Slag sample which is in progress at the time this
paper was written. Colors for phases in Figure 4
All the samples have been observed considering
that’s important to be representative enough in our
conclusions. We have focused our representative-
ness study on the OPC mortar in order to make
in evidence what could be the number of fields of
views required to be stable in phase measurements.
Thus we have observed the cumulative fraction of
paste (paste is chosen as a pertinent indicator) in
function of the number of images classified. Of
course the higher number of images we integrate in

Figure 3: Cumulative fraction of paste in OPC mortar in function of the number of images taken in account in height.

382
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 4: Backscattered electron images of OPC, OPC-LF (Limstone Filler) and OPC-Slag at 28 (before immersion),
250 (T1), 350 (T2) and 720 (T3) days. The global field of view for each image is around 3.5 mm. Each image is a
composition of a mosaic composed by 9⇥8 individual images with a pixel size of 0.350µm /pixel (visible as an artifact
of reconstruction in the latest on).

cumulative fraction, the higher representative value
we have.
In Figure 3, we are trying to determine what
could be precisely the optimal size (height and
depth) of the global field of view for a selected
system in order to get a representative gradient of
mineralogical phases. We have made the hypoth-
esis that the precision of phase quantification in

Figure 5: Classified images of OPC, OPC-LF (Limstone Filler) and OPC-Slag at 28, 250, 350 and 720 days,
corresponding to previous BSE images (Figure 3). Scale bar is 0.5 mm. The code of colors are: Sand (Orange),
Porosity (light brown), Porlandite and/or calcite in OPC and OPC-Slag mortars (dark green), ’Outers’ (light green),
C3 S C2 S silicates (light and dark blue), Slag (Yellow), monosulfoaluminates grains (red), trisulfoaluminate coarser
phases (rose), limestone filler (dark brown).

383
 Meulenyzer et al.

Figure 6: Evolution of phase fraction for: 1/ monosulfoaluminate (MSA) pockets and trisulfoaluminate (TSA)
‘coarse grain’ in function of depth for OPC (at left), 2/ Monocarboaluminate (MCA) and trisulfoaluminate (TSA)
‘coarse grain’ in OPC-Limestone filler (middle) and 3/ sulfoaluminates(MSA + MCA) and gypsum ‘coarse grain’ in
OPC-Slag mortars (right)

function of depth (profiles) depends on the num- Evolution of the microstructure upon sulfate
ber of images we have in a column. The measures ingress
have been performed with 3 different zones of 9⇥8
In this section we focus our analysis of microstruc-
images. The cumulative fraction of paste includes
ture on local scale from core to surface of OPC and
the porosity (macro / capillary porosity that could
OPC-LF mortars (at T2 = 360 days). We have been
be observed in BSE images). That’s the reason why
especially interested by the evolution of zones com-
the first µm from surface haven’t been shown here
bining a high concentration of sulfates (S in SO3
is due to leaching that increase highly this value
through Sulfur X-Ray maps) and alumina (Al in
and also the resin beyond the surface. As shown in
Al2 O3 through Al x-ray maps). During the learning
consequence in Figure 3 the cumulative percentage
step, which consists of assigning a color to a phase,
of paste starts to stabilize with 20 images in a given
we have identified using EDS spectroscopy phases
height (column that represents a depth from sur-
like monosulfoaluminate, trisulfoaluminate and
face). We have decided to take in account 3 zones
monocarboaluminate (OPC limestone filler mortar)
of 9⇥8 images to get a representative gradient.
structures in zones containing high concentration
of sulfates. Two of these three phases appeared
in the form of ’pockets’ and they have been then
Meulenyzer et al. classified in all images.
We describe in Figure 7 and Figure 8 what could
3.3.1 Microstructural évolution of OPC-T2 mortar

Figure 7: Microstructural phase evolution of OPC-T2 mortar at local scale from surface (left) to core of mortar (OPC).
KK=clinker; Agg.=Aggregates; CH=portlandite.

384
 15th Euroseminar on Microscopy Applied to Building
Meulenyzer et al.
3.3.2 Microstructural évolution of OPC–Limestone filler–T2 mortar
Figure 8: Microstructural phase evolution at local scale from surface (left) to core of mortar (OPC-LF). KK=clinker; Agg.=Aggregates; CH=portlandite;
Figure 8: Microstructural phase evolution of OPC–Limestone
LF=Limestone
Filler; Mca= Monocarboaluminate
mortar (OPC-LF). KK=clinker; Agg.=Aggregates; CH=portlandite; LF=Limestone Filler; Mca= Monocarboaluminate.
be a continuous replacement of monosulfoalumi-
nate (MSA) or monoarboaluminate by ’grains’ of
trisulfoaluminate (TSA) whereas the concentration
of sulfate decreases from surface to core of sample.
Gradients measurements
We have calculated the gradient of a given phase
taking in account representativeness study de-
scribed before. Each ‘depth’ is represented by 24
images. Each point of the curves below is a mean
value of a window with a size equivalent to total
height of the 24 images. The larger of this window
is half a single image (512 pixels) and is moving by
step of 35 µm.
Figure 6 shows the evolution of phase fraction
(MSA, MCA, TSA, gypsum) for the 3 studied sys-
tems at T2 in function of depth. It is interesting in
the case of OPC to note that the MSA grains are in-
creasing with depth whereas the TSA is decreasing.
In the case of OPC-LF, we have observed a ‘dense’
zone containing TSA grains in the middle of sam-
ple whereas the quantity of MCA grains slowly
decreasing from core to surface.
The case of slag must be study in more details
since the difference between MSA and TSA was
more difficult to obtain. We only observe at this
moment a front of gypsum grains near 1 mm from
the surface.
Materials • 17-19 June 2015 • Delft, The Netherlands
filler–T2 mortar at local scale from surface (left) to core of
2
IV. Conclusion and perspectives
We have presented a methodology based on recent
developments in image analysis that allowed us
to study phase evolutions in mortars immerged in
sulfate solutions. Our present paper was focused
on OPC, OPC-LF and OPC-Slag (T2) and allowed
us:
• To evaluate qualitatively the evolution of mi-
crostructure at local scale
• To measure gradients of selected mineralogical
phases in function of depth.
We have especially observed that fraction of pock-
ets of MSA or MCA evolved from surface to core
of each sample.
More investigations are needed to attribute a pre-
cise role of these pockets. Especially we will drive
our efforts on others T1 and T3 and compare all
the gradients measured. Also EDS spectroscopy on
these pockets must be performed EDS in order to
understand more the origin of expansion.
References
Barcelo, L., Gartner, E., Barbarulo, R., Hossack, A., Ahani, R., Thomas,
M., Hooton, D., Brouard, E., Delagrave, A., Blair, B. (2014), “A modi-
fied ASTM C1012 procedure for qualifying blended cements containing
limestone and SCMs for use in sulfate-rich environments”, Cement and
Concrete Research, 09/2014; 63:75–88.
Dhir R.K., McCarthy, M.J., Zhou, S. (2004), “Role of cement content
in specifications for concrete durability: cement type influences”, Pro-
ceedings of the ICE - Structures and Buildings, Volume 157, Issue 2, 01 April
2004 , pages 113 –127.
385
 Meulenyzer et al.

Schmidt, T., Lothenbach, B., Romer, M., Scrivener, K., Rentsch, D., Yu, C., Suna, W., Scrivener, K.., “Mechanism of expansion of mor-
Figi, R. (2008), “A thermodynamic and experimental study of the condi- tars immersed in sodium sulfate solutions”, Cement and Concrete Research
tions of thaumasite formation”, Cement and Concrete Research. 38 (2008) Volume 43, January 2013, Pages 105–111.
337–349.
Meulenyzer, S., Chanussot, J., Chen, J.J., Crombez, S., “Spectral-
spatial image processing strategies for classifying multispectral SEM-
EDS X-Ray maps of cementitious materials”, 14th Euroseminar on
Microscopy Applied to Building Materials, 2013, Helsingor, Denemark;
06/2013.

386
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Spalling of sprayed perlite concrete

caused by popcorn calcite deposition and
internally derived sodium sulfate under
influence of water leakage, frost action
and dynamic loads
Per Hagelia
Norwegian Public Roads Administration, TMT, Tunnel and Concrete Division
per.hagelia@vegvesen.no

Abstract

A Norwegian road tunnel opened for traffic in 1988 was closed in 2011 due to severe spalling of light weight sprayed
concrete. Initially a 2-3 cm thick layer of steel fiber reinforced sprayed concrete with extruded perlite aggregate was
employed for fire protection of polyethylene (PE-) foam panels used for water and frost protection. Repair work in
2002 involved spraying of some new perlite concrete and a ca 40 mm thick outer layer of fire resistant sprayed concrete
with micro-polypropylene (micro-PP) fiber. Petrography, SEM and XRD in 2011 showed that the spalling zone, being
located around the interface between the two concrete layers, contained mainly thenardite. Also the adjacent micro-PP
fiber concrete was partly influenced. Mirabilite was not detected, whilst minor amounts of a thaumasite-like phase
was occasionally present. The entire layer of perlite concrete was very friable and had suffered destructive steel fiber
corrosion. This was characterised by extensive carbonation, internal sulfurised Popcorn calcite deposition (PCD) and
severe Ca-depletion of the cement paste matrix, hence resembling some features of acid attack. Spalling was particularly
well developed where water had seeped through the PE-foam plates and affected by freeze/thaw. The tunnel is located
within a Permian rhomb-porphyry lava sequence with no external sulfate source, and the concrete sand aggregate was
devoid of sulfur-bearing minerals. However, stable S isotopic signatures (d34 S) imply that perlite aggregate contributed
sulfuric acid: The S isotopes in thenardite, sulfurised carbonate and acid leached residue were lighter than common
cement gypsum sources. Variable moisture and temperature conditions, frost action and dynamic loads from the traffic
provided a favourable environment for micro cracking, ion migration, precipitation /dissolution and combined action of
thenardite and pore-ice.

Keywords: Sprayed concrete, internal sulfate attack, thenardite, carbonation, frost action

I. Introduction 2007, 2013). In general thicker layers of sprayed con-

Sprayed concrete is used in Norwegian road tun-
nels for rock mass reinforcement in conjunction
with rock bolts and as fire protection of polyethy-
lene (PE) foam panels used for water and frost
protection. Degradation of tunnel concrete is not a
general problem. However, steel reinforced sprayed
concrete for rock support has suffered some slow
degradation in the Alum shale environment, due
to Thaumasite Sulfate Attack (TSA) with associated
popcorn calcite depositiom (Hagelia et al. 2001,
2003). Also a spotwise significant sprayed concrete
deterioration caused by combined bacterial and abi-
otic attack occurs in some subsea tunnels, leading
to localised thinning where Mn and Fe-rich biofilms
are acting simultaneously with Ca-leaching, pop-
corn calcite deposition, TSA and Mg attack (Hagelia
crete are more durable than relatively thin layers
(Davik 1998, Hagelia 2011). In contrast, degrada-
tion of sprayed concrete used for fire protection
is apparently uncommon, although mechanically
induced fracturing has been observed in some tun-
nels (Norwegian Public Roads Adminsitration 1997,
Davik 1998).
This paper reports on a severe attack within
sprayed concrete used for fire protection of PE-
foam panels in a road tunnel in East Norway. This
unusual incident, involving sprayed concrete with
light weight perlite-aggregate, had caused severe
spalling and closure of the tunnel for nearly three
years.

387
 Hagelia
History of concrete degradation in the Nes road
tunnel and main technical data
The Nes road tunnel on expressway E16 in East
Norway was opened for traffic in 1988. The tunnel
is 1276 m long, serving about 11000 vehicles every
day. Steel fiber reinforced concrete and rock bolts
were used for rock mass reinforcement, whilst PE-
foam panels were installed as an inner lining for
water and frost protection. Due to the inflammable
properties the PE- panels were fire-protected by a
2-3 cm thick layer of steel fiber reinforced light-
weight sprayed concrete, containing very porous
extruded perlite and sand aggregates. After about
ten years in service some damage was discovered
within the perlite sprayed concrete at several places.
A detailed inspection was undertaken in 2002, un-
raveling need for reparations. Loose rock mass
amounting to about 500 kg was also discovered
at one location behind the water panel. The dam-
age in perlite sprayed concrete was characterised
by horizontal and vertical penetrative cracks, sev-
eral of which carried water leakages. Reparations
in 2002 involved sealing of wide cracks with PE-
foam, whilst smaller cracks with water leakages
were sealed with silicon. Thereafter the cracks were
sprayed by new steel fiber reinforced perlite con-
crete. A special adhesive was applied for making
the concrete stick to the PE-panels. In order to fur-
ther strengthen the light-weight sprayed concrete
a ca. 4 cm thick layer of concrete with micro-PP
fiber was sprayed in the lower parts of the tun-
nel profile along the entire tunnel length, reaching
about one meter above the road level. The micro-PP
fibers concrete was subsequently covered by a ca.
3 cm thick spray of fiber-free “ordinary” sprayed
concrete (Figure 1%2).
In the early morning on March 3rd , 2011 about
200 kg of sprayed concrete fell down, and one ve-
hicle was damaged. The tunnel was immediately
closed for traffic, at first for a thorough inspection.
The conclusion after inspection was that the tun-
nel should be closed on permanent basis and that
complete refurbishment would be the most feasible
solution. The tunnel was not reopened for traffic
until February 2014.
The 1988 sprayed perlite concrete (Robotic 10)
was made with Portland cement, silica fume and
steel fiber reinforcement. The cement was most
likely CEM I. Water glass was used as setting accel-
erator. However, the water/binder ratio and other
details have not been reported. In 2002 thorough
water cleaning of the entire tunnel was undertaken
prior to reparation. The new concrete mixes are pre-
sented in Table 1. According to internal documents
at NPRA the work was thoroughly controlled dur-
ing and after it was finished, showing good and
satisfactory workmanship.
388
Inspections in 2011 by NPRA and Norconsult
showed that spalling of the perlite concrete was
developed throughout most of the tunnel. It was
not possible to differentiate visually between dam-
aged and undamaged areas, because hammering
unraveled a much wider extent of spalling than was
obvious through superficial inspection. Observa-
tions clearly indicated that the spalling involved
frost action and carbonation, likely assisted by vi-
brations due to dynamic air pressure induced by
vehicles (Holm 2011). However, no laboratory anal-
ysis and microscopy were applied.
Objectives
Due to the severe implications of the concrete degra-
dation it was decided to undertake a more detailed
investigation, with the clear aim of diagnosing the
deterioration mechanism and its causes. The objec-
tives of this work were to:
• Present the main tunnel observations and sam-
pling of degraded sprayed concrete
• Characterise and diagnose the degradation
mechanisms by the aid of concrete petrography,
X-ray diffraction and Scanning electron microscop
(SEM) analysis
• Investigate the ultimate source of aggressives
involved, assisted by stable S isotope analysis
• Discuss the spalling mechanism
II. Methods in the field and
laboratory
Field methods
Samples were collected in April 2011 after the tun-
nel was closed. Spalls and smaller samples of de-
graded concrete were collected from different layers
along with degradation products located at the in-
terface between perlite concrete and outer micro
PP-fiber concrete. Spall samples were collected in
plastic bags and sealed and debris samples were
put into small glassed with tight plastic lids. All
samples were essentially dry.
Laboratory methods
The analytical work is summarised in Table 2. The
thin sections were investigated under a standard
polarising microscope and analysed by SEM. Debris
samples and chips of friable concrete were analysed
by X-ray diffraction (XRD) and Scanning Electron
Microscopy (SEM). Stable S-isotopes were analysed
in two samples.
Preparation of thin sections. Two concrete spalls,
approximately two cm thick, were cut at right angle
across layering with a diamond saw and washed
thoroughly. Standard polished thin sections with-
out fluorescent dye were prepared on 27 mm x 45
totally disintegrated concrete. The PE-foam panels were exposed at several locations (Figure 1). It was otherwise
evident that the perlite concretes had suffered extensive carbonation with severe and mainly destructive steel
15th Euroseminar
fiber corrosion all overonthe place. Also
Microscopy the adjacent
Applied parts
to Building of the• 17-19
Materials outerJune
micro-PP fiber The
2015 • Delft, concrete were in part
Netherlands
cracked and friable. The features of the concrete attack were essentially the same everywhere.

Figure 1: Left: Well developed spalling in the 1988 sprayed concrete, terminating about 1 m above the road. Apparently the
Figure 1: Left:
wider lower spallsWell developed
had also spalling
developed in the in theparts
upper 1988ofsprayed
the 2002concrete, terminating
sprayed layers. Middle:about
Severe1spalling
m above the road.
in 1988 and
Apparently
2002 sprayed the wider being
concrete, lower typically
spalls had also developed
developed within in
thethe upper
perlite parts
layer: of the 2002
Previous sprayed
spalling layers.
in 1988 Middle:
perlite concreteSevere
(upper
arrow) isin
spalling partly
1988covered
and 2002by 2002 sprayed
sprayed concrete,
concrete, beingdisplaying
typicallyextensive
developedand morethe
within recent spalling
perlite layer:with exposed
Previous PE-foam
spalling in
panel
1988 (lower
perlite arrow) Right:
concrete Bright
(upper white
arrow) is thermonatrite
partly coveredefflorescence deposited
by 2002 sprayed fromdisplaying
concrete, ground water on rhomb
extensive andporphyry rock
more recent
spalling with exposed PE-foam masspanel
(widths of photos;
(lower arrow)about
Right:3 m; 1 m &white
Bright 0.8 m,thermonatrite
respectively).efflorescence deposited from
ground water on rhomb porphyry rock mass (widths of photos; about 3 m; 1 m & 0.8 m, respectively).
Jo
in
te
d
ro
c
k diffractograms were checked ver-
mm glass slides, representing perlite concrete and 1.54178 Å. The m
as
outer sprayed concrete with micro-PP fiber. Water sus the Powder Diffraction s Files database from the
was used as coolant; hence possible presence of sol- International Centre for Diffraction Data. Mineral

of rh
PE-foam panel
uble phases such as thenardite were not preserved proportions are regarded as qualitative to semi-

omb
in the thin sections. quantitative.

porp
Polarising microscopy. The thin sections were stud- Scanning
Inner spray of steel electron microscopy (SEM). Thin sections
fibre reinforced

hyry
perlite concrete: leached with
ied under a standard polarising microscope with and concrete
thenardite adjacent debris and other friable samples were
to outer spray
objectives 5x, 10x, 20x and 50 x. Domains with analysed. This involved point analysis (1-2 µm2 )
Outer spray w/micro-PP fibre
degradation products and other features were iden- and X-ray element mapping. All samples were
tified, several of which were further analysed by mounted on carbon tape without coating. A Hi-
SEM. tachi 3600N scanning electron microscope with an
Figure 2: Cross section showing three layers of sprayed concrete: EDS 1) inner light colored
unit from Thermo 1988 perlite concrete
Electronic with internal
Corporation was
X-ray diffraction (XRD). Small samples about 0.5-2
spalling due to extensive carbonation and destructive steel fiber corrosion, followed by 2002 concrete; 2) grey sprayed
grams werewith
handpicked fromanddebris and friable used. The instrument was operated with an acceler-
concrete micro-PP fibers 3) outer dark sprayed concrete without fibers. Breath of photo is about 50 cm.
concrete under a stereomicroscope at 20-50x magni- ated voltage of 15 kV and 10 Pa vacuum. Analytical
fication Spalls
and ground in small agate mortars under precision for major elements is within 1-2 wt.% for
and debris samples were collected from each sprayed concrete layer at suitable locations within
theethanol.
middle The
partssamples were left
of the tunnel. In to dry cases
most in air the samplesspot
at am- could analysis
be brokenon thin sections
loose without and other
using even sur-In
a hammer.
bient laboratory temperature, mounted on sample faces. The analytical
general it was not easy to differentiate between the two generations of perlite concrete. No water leakage work was undertaken soonwas
holders for
available andproper
run inwater
a Siemens D 5000
sampling. after collection and preparation
Spectrome-the chemical nature of ground water load could not be
Consequently in 2011, thus avoid-
ter. All
directly diffractograms
established. However,wereat recorded from
one location 2o towhite ing
bright aging.
surface However, X-ray
efflorescence mapping
sitting on barren and twomass
rock points
was
o
70 on(Figure
reached the 2-Theta scalehoped
2). It was at 0.050 o
that per were analysed in thin section N2
second. Themight shed light on the ground water chemistry from which
its composition in January.
spectrometer
it had was setAtowhite
been precipitated. 40 kV and 40 nAwas
efflorescence using Sulfur
also collected isotopes.
from Sulfur in
the southern thenardite-rich powder
portal.
Ni-filtered CuKa radiation with wavelength (l) = (Nes 1) was obtained by homogenization of the

Table 1: Concrete mixes in the Nes road tunnel. Weights of cement, aggregate and perlite are given in units of 1 m3 of
concrete; Sand grading 0-4 mm; Cylinder strength. *) = No data; **) = SF by cement weight. 4
Perlite concrete 1988 Perlite concrete 2002 PP-fiber concrete 2002 “Ordinary” concrete
Portland cement, * Portland cement * = 450 kg Portland cement * = 465 kg Portland cement * = 465 kg
Silica fume* Silica fume = 20 kg** Silica fume = 30 kg** Silica fume = 30 kg**
Water glass Water glass Water glass Water glass
Fine sand* Sand = 550 kg Sand = 1580 kg Sand = 1577 kg
Perlite* Perlite = 1450 kg
Steel fiber Steel fiber = 1 kg Micro-PP fiber Fiber free
w/c* w/b = 0.47** w/b = 0.42** w/b = 0.42**
Strength = 13-22 MPa Strength ⇡ 35 MPa Strength = 31-37 MPa
r= 1500-1850 kg/m3 r⇡2225 kg/m3 r= 2211-2260 kg/m3

389
 possibly also governed by previous penetrative cracks in the 1988 concrete. Although the tunnel concret
essentially dry to at the time of sampling, there was clear evidence for Ca-leaching, suggesting that wate
Hagelia impact. Indeed, the fire protected PE-foam panels were installed along the
frost must have had a profound
length of the tunnel due to the presence of widespread moisture and water seepages in the tunnel in 1988
local climate is characterised by cold winters, rainy seasons and variably hot summers.
sample, followed by direct combustionSpalling was controlled
of the sam- by a weakpanels
The PE-foam very light-colored
were exposed layer,
atinseveral
most cases
loca-located in the outer pa
the perlite concrete.
ple in presence of V2 O5 . Three equal samples were The spalling
tions (Figure 1). It was otherwise evident that of
zone was frequently characterised by presence thea white powder represe
totally disintegrated concrete. The PE-foam panels were exposed at several locations (Figure 1). It was othe
prepared in tin capsules. The extensively carbon- perlite concretes had suffered extensive carbonation
evident that the perlite concretes had suffered extensive carbonation with severe and mainly destructive
ated concrete sample N1 was first crushed to a with severe and mainly destructive steel fiber cor-
fiber corrosion all over the place. Also the adjacent parts of the outer micro-PP fiber concrete were in
fine powder in an agate mortar. From this bulk
cracked and friable. The featuresrosion allconcrete
of the over the place.
attack wereAlso the adjacent
essentially the sameparts
everywhere.
material three equal samples were packaged into of the outer micro-PP fiber concrete were in part
tin capsules in the presence of V2 O5 . The remain- cracked and friable. The features of the concrete
ing material was washed in 2N HCl to release S attack were essentially the same everywhere.
bound to calcite and other soluble phases, and sub- Spalls and debris samples were collected from
sequently washed with deionised water, dried, and each sprayed concrete layer at suitable locations
three equal samples were packaged into tin cap- within the middle parts of the tunnel. In most cases
sules in the presence of V2 O5 . The sample capsules the samples could be broken loose without using
were combusted at 1700 o C and reduced to SO2 , a hammer. In general it was not easy to differenti-
which was analysed in a Nu Horizon isotope mass ate between the two generations of perlite concrete.
spectrometer along with IAEA standard NBS 127 No water leakage was available for proper water
and Elemental Microanalysis standard B2308 (quan- sampling. Consequently the chemical nature of
tification reference). Isotopic composition and S ground water load could not be directly established.
concentration were analysed in the same instru- However, at one location bright white surface efflo-
ment. The analytical precisionFigure
in d341:S Left:
is ±Well developed rescence
0.2 ‰(1 spalling in the 1988 on
sitting sprayed concrete,
barren rockterminating
mass wasabout 1 m above the road. Apparentl
reached
wider lower spalls had also developed in the upper parts of the 2002 sprayed layers. Middle: Severe spalling in 1988
s) and the isotopic values are 2002
given relative to the (Figuredeveloped
sprayed concrete, being typically 2). It was hoped
within that its
the perlite composition
layer: might
Previous spalling in 1988 perlite concrete (u
Canyon Diablo troilite standard. shed
arrow) is partly covered by 2002 lightconcrete,
sprayed on thedisplaying
groundextensive
water and chemistry
more recentfromspalling with exposed PE-fo
which
panel (lower arrow) Right: Bright whiteitthermonatrite
had been efflorescence
precipitated. A white
deposited from efflores-
ground water on rhomb porphyry
mass
cence(widths
was of photos;
also about 3from
collected m; 1 mthe& 0.8 m, respectively).
southern portal.
III. Results
Jo
in
te
Field observations and sampling d
ro
ck
m
The sprayed concrete was severely spalling in most as
s
parts of the tunnel, being developed in a spot-wise

of rh
PE-foam panel
fashion in the ceiling and tunnel walls. The sizes of

omb
spalled areas varied from a few square decimeters

porp
to more than one square meter, whilst the spall Inner spray of steel fibre reinforced

hyry
perlite concrete: leached with
thicknesses were about 0.5 to 2.5 centimeter. The thenardite adjacent to outer spray
spalling was at several places developed as ver-
Outer spray w/micro-PP fibre
tical long and narrow zones connected to wider
spalls in the lower parts of the tunnel walls. Yet
the sprayed concrete below about 1 m above the
Figure 2: Cross section showing three layers of sprayed concrete: 1) inner light colored 1988 perlite concrete with int
road level, corresponding to 2002 repair
spalling due concrete, Figure 2:and
to extensive carbonation Cross sectionsteel
destructive showing three layers
fiber corrosion, of sprayed
followed by 2002 concrete; 2) grey spray
had not developed much visible spalling
concrete(Figure 1). fibers
with micro-PP and 3) outer
concrete: dark sprayed
1) inner concrete1988
light colored without fibers.concrete
perlite Breath of photo is about 50 cm.
However, also the 2002 perlite concrete was spalling with internal spalling due to extensive carbonation and
(Figures 1 & 2). Spalls and debris samples weresteel
destructive collected from each sprayed
fiber corrosion, followedconcrete
by 2002layer
con- at suitable locations w
The vertical narrow spallsthe middle
were parts of the
sometimes re- tunnel. In most cases the samples could be broken loose without using a hamm
crete; 2) grey sprayed concrete with micro-PP fibers and
general it was not easy to differentiate between the two generations of perlite concrete. No water leakag
lated to joints in the PE-foam panels behind, and 3) outer dark sprayed concrete withoutnaturefibers. ofBreath of water load could n
available for proper water sampling. Consequently the chemical ground
were possibly also governed by previous penetra- photo is about 50 cm.
directly established. However, at one location bright white surface efflorescence sitting on barren rock mas
tive cracks in the 1988 concrete. Although
reached (Figurethe tun-
2). It was hoped that its composition might shed light on the ground water chemistry from w
nel concrete was essentially itdry to at the time
had been precipitated. of A white efflorescence was also collected from the southern portal.
sampling, there was clear evidence for Ca-leaching,
XRD
suggesting that water and frost must have had a
profound impact. Indeed, the fire protected PE- XRD analysis was performed on perlite concrete,
foam panels were installed along the entire length associated degradation products and white efflores-
of the tunnel due to the presence of widespread cence material both from barren rock (Figure 1) and
moisture and water seepages in the tunnel in 1988. the tunnel portal. The results from XRD analysis
The local climate is characterised by cold winters, are summarised in Table 3. In general, quartz and
rainy seasons and variably hot summers. feldspar represents aggregate minerals. Calcite in
Spalling was controlled by a weak very light- degraded and friable perlite concrete represents ex-
colored layer, in most cases located in the outer tensive carbonation of the cement paste. “Possible
parts of the perlite concrete. The spalling zone minors” were indicated by very minor peaks, but
was frequently characterised by presence of a white were not rigorously proven.
powder representing totally disintegrated concrete. Sodium sulfate was always present in the white

390
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Table 2: Summary of analytical work on sprayed concrete samples from the Nes road tunnel.

Sample No./characteristics/type of material Polarising microscopy XRD SEM S- isotopes

Nes-1 White powder in perlite concrete at interface w/PP concrete X X X
Nes-2 White very friable perlite concrete X X
Nes-3 Grey friable mico-PPfiber concrete X
Nes-4 White powder at interface w/PP-concrete X
Nes pr 1 Efflorescence on rhomb porphyry rock mass X
Nes pr 2 White powder at interface w/PP-concrete X
Nes pr 3 Efflorescence from crack in tunnel portal X
N-1 Perlite concrete w/corroded steel fiber X X X
N-2 Perlite concrete & outer PP-concrete X X

to light colored powder samples. Nes-1 consisted
of thenardite and a little quartz. Nes-4 contained
quartz, calcite and lesser amounts of thernardite,
with a possible very minor amount of vaterite.
Nes pr. 2 was particularly interesting in that two
forms of sodium sulfate occurred. Thenardite, be-
ing face-centered orthorhombic Na2 SO4 (form V)
and base-centered orthorhombic sodium sulfate
(form III) occurred together with quartz and a lit-
tle calcite. Portlandite and gypsum were possibly
present. Mirabilite (Na2 SO4 ·10H2 O) was not de-
tected.
Sample Nes-2 of white and very strongly de-
graded perlite concrete contained much thenardite
and also Ca-mordenite (CaAl2 Si10 O24 ·7H2 O),
which is a zeolite. XRD analysis of Nes-2 also indi-
cated presence of a silicon oxide (Si100 O200 ). This
substance is known mainly as a molecular sieve.
Its occurrence in degraded perlite concrete remains
obscure, but it seems possible that this silicon oxide
may be related extruded perlite or silica fume.
Sample Nes pr 1 of white efflorescence sitting
on the rhomb porphyry rock mass behind the
PE-foam panel (Figure 1) contained thermonatrite
(Na2 CO3 (H2 O)) with lesser amounts of quartz and
anorthite. There was no evidence of sulfur bearing
phases. Hence, it seems very unlikely that ground
water in the area is enriched in sulfate. Sample
Nes pr 3 of white efflorescence precipitated on a
penetrative crack in one of the tunnel portals (cast
concrete) consisted mainly of pure calcite, yet with
possible traces of portlandite and gypsum. This
likely represents leaching of the concrete.
Petrography - polarising microscopy
Thin section N1 of extensively carbonated sprayed
perlite concrete was characterised by a light brown-
yellowish cement paste matrix when viewed in
plane polarised light. The paste was extensively
leached, with isotropic character in crossed po-
lars. The carbonation was represented by ubiq-
uitous Popcorn calcite deposition (PCD) with in-
dividual calcite grains typically about 20-50 µm.
However, also much finer carbonation had occurred,
resembling features of ordinary carbonation. Cal-
cite was also deposited in some air voids and lo-
cally within the pore space of perlite. Portlandite
was apparently absent. Only few suspect grains of
unhydrated cement were observed within the Ca-
depleted cement paste matrix. Some undispersed
brown silica fume globules, about 50-150 µm big,
were found scattered about in the matrix. Also,
dark brown, transparent and glass-like spheres
about 20-50 µm big were found, probably repre-
senting fly ash agglomerations.
Rusty iron-rich deposits occurred, which had also
led to a brown staining of the adjacent cement paste
matrix. No steel fiber were preserved in the thin
section, however, holes in the middle of some dense
rust deposits suggested that less corroded steel had

Table 3: Summary of XRD results on degraded concrete from the Nes road tunnel. Two structural forms of Na2 SO4
were identified; thenardite (form V) and sodium sulfate * (form III).

Samples XRD minerals Possible minors

Nes-1/white powder
Nes-2/white friable fragment
Nes-4/white powder
Nes pr 1/efflorescense
on rock mass
Nes pr 2/white powder
Nes pr 3/efflorescence
from crack in tunnel portal
Thenardite > quartz
Thenardite > mordenite quartz > silicon oxide
Quartz > calcite > thenardite Vaterite?
Thermonatrite > quartz ⇡ anorthite
Thenardite > sodium sulfate* > quartz > calcite Portlandite? Gypsum?
Calcite quartz Portlandite? Gypsum?

391
 Hagelia
been lost during thin section preparation.
The perlite aggregate particles were ranging in
sizes from about 50 to 1000 µm and were extremely
porous. Suspect thin glass shards of perlite were
also observed scattered about in the cement paste
matrix, which was confirmed in SEM. The remain-
ing aggregate consisted of granitic rocks including
micro-granite, siltstone and single grains of plagio-
clase, quartz, biotite and amphibole. The aggregate
particles were rounded to angular with Dmax about
5 mm. Opaque minerals were mainly Fe-Ti-oxides.
Inspection in reflected light, using an external light
source, showed that sulfide minerals were absent.
Thin section N2 included both sprayed perlite
concrete and outer micro-PP fiber sprayed concrete.
The perlite concrete here was sharing the same
characteristics as sample N1. In contrast, the ce-
ment paste matrix in the outer sprayed concrete
with micro-PP fiber was dark brown in plane polar-
ized light, and did not appear to be much leached.
The contact region adjacent to perlite concrete was,
however, somewhat porous, with some large entrap-
ment voids (reaching 300 µm). In a small domain
next to an entrapment void apparent thaumasite
had replaced cement paste, showing typical straw
yellow color in plane polarised light and grey first
order birefringence in cross polarised light. How-
ever, the SEM work showed this to be a little more
complex.
The aggregate in outer concrete consisted of
granitic rocks, siltstone, limestone, mica-gneiss,
quartzite, and single grains of quartz, feldspars, am-
phibole and biotite. The aggregate particles were
otherwise similar in shape and grading as in the
sprayed perlite concrete. In a single micro-domain
a gel-like isotropic material was found forming a
thin rim inside an air void.
Scanning electron microscopy, micro-analysis
and X-ray mapping
The SEM analysis was performed on the thin sec-
tions and several small debris samples. Analytical
data from these two sample types must be regarded
as complimentary. Sodium sulfate is soluble and
not expected to have been preserved during thin
sections preparation, which involved water cool-
ing. Moreover, additional SEM work on samples
analysed by XRD might shed further light on their
composition, because of the possibility of mixed or
amorphous phases.
Perlite concrete in thin sections N1 and N2. Fig-
ure 3 shows a typical micro domain of perlite con-
crete in sample N2. The image is from outer perlite
concrete few mm beneath the micro-PP fiber con-
crete and may represent the 2002 spray. Severe
carbonation here had caused destructive steel fiber
corrosion. The mineral aggregate was quartz-rich
392
with no sign of sulfides. The cement paste matrix
was characterised by extensively developed PCD
and finer grained carbonates. The initial calcium
silicate hydrate (CSH) was completely degraded
due to Ca-leaching, and the matrix was very friable
and porous with variably penetrative micro-cracks.
Undispersed silica fume lumps (100-200 µm) and
suspect fly ash spheres (20-50 µm) were observed.
The results of X-ray element mapping are shown
in Figure 4. The element distribution shows that
Ca, O and lower intensities of C occupy the same
domains in the paste. This reflects the presence of
calcium carbonate, which was also Mg-substituted
especially in the upper right parts of the analysed
frame. However, the highest intensities of carbon
represent epoxy in highly porous perlite, air voids
and micro-cracks. The iron stained paste is clearly
indicated by high Fe intensity. It is noteworthy
that S was mainly occurring in the cement paste
matrix, whilst the lower intensities of Na and Cl
were scattered about without any particular relation
to phase boundaries observed in back scatter. Most
likely the latter feature reflects presence of soluble
de-icing salt being smeared out on the thin section
surface as a result of preparation. The mineral
aggregate particles were clearly indicated by the
higher intensities of Si, Al, K, Na and Ti. Perlite was
not well indicated on this map, being overshadowed
by other phases.
An important observation was that a very signif-
icant amount of the pore space of extruded perlite
had in fact collapsed. Strings and shards of perlite
glass occurred scattered about everywhere, being
well embedded within depleted cement paste ma-
trix with PCD (Figure 5). These textural relations
clearly suggest that collapse must be a primary
feature, e.g. that partial perlite collapse has taken
place already during the spraying operation.
Micro-PP fiber concrete in thin section N2. SEM
imaging confirmed that this outer concrete was
sounder than the perlite concrete. The cement
paste was often characterised by a normal look-
ing cement paste matrix without PCD and other
obvious signs of extensive moisture transport and
Ca-depletion. However, the surface region was very
porous and inhomogeneous, with occasional big
lumps of undispersed silica fume, as seen in Fig-
ure 6C. The thaumasite like material (Figure 6B),
located in a small micro domain about 200 µm
away from perlite concrete, had replaced local ce-
ment paste, hence resembling features of a localised
Thaumasite sulfate attack (TSA). However, chemical
analysis showed that this is not a typical thaum-
asite but rather a very carbonate rich thaumasite-
ettringite like phase (see details below). This min-
eral occurred in presence of a large entrapment
void, and had also formed a faint fringe deposit
inside this void. The entrapment void was joined
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Back scatter image of severely carbonated perlite concrete (thin section N2) with Fe-rich corrosion product
(bright) having stained adjacent carbonated cement paste matrix. Aggregate particles were mainly sand consisting of
quartzite with biotite (upper left), pure quartz and granitic rocks. Undispersed silica fume (SF) and suspect fly ash
(FA?) occurred. The dark areas are pores in extruded perlite and air voids, and a hole in the thin section due to a lost
corroded steel fiber (cf. Figure 4).

with precipitates of calcium carbonate, represent-
ing a local weakness zone near the perlite concrete
layer. However, grey friable micro-PP fiber concrete,
represented by sample Nes-3, clearly showed that
also parts of the outer concrete was partly severely
attacked, as shown in Figure 8.
Chemical analysis of thin sections N1 and N2.
Table 4 summarises the EDS analyses performed
on the thin sections. Most of the analysed points
and frames are shown in Figures 5 and 6. The
carbonates were represented by three forms; 1)
PCD in the cement paste matrix, 2) deposits on air
voids in paste (“Cc” in Table 4) and within perlite
and 3) very fine grained carbonation resembling
ordinary carbonation due to atmospheric CO2 . The
finer carbonation was not analysed.
The chemical composition of perlite glass was
similar to rhyolite: In terms of oxide components
(weight percent), the analysis N1-7 (Figure 7A) cor-
responds to: Na2 O = 4.08 %, K2 O = 4.56 %, Al2 O3
= 13.69 %, SiO2 = 74.63% and CaO = 3.04 %. The
apparent absence of iron and magnesium is typi-
cal of a source material, which represents highly
differentiated late stage volcanic melts. No sulfur
was detected in extruded perlite glass. N1-8 repre-
sents apparent non-degraded CSH-rich paste sitting
within a pore in a big perlite aggregate. Carbon and
sulfur were not detected. In contrast analysis N1-5
from cement paste among PCD in presence of col-
lapsed perlite, showed a very severe Ca-depletion,
which is a typical feature of PCD concrete (Sibbick
et al. 2003, Hagelia and Sibbick 2009). The de-
pleted paste was otherwise apparently carbonated
and sulfurised.
The PCD compositions and calcium carbonate de-
posits were not pure calcite. They were substituted
by a small amount of magnesium, and were also
sulfurised. Smaller amounts of alkalis, aluminium
and silicium also occur, which may be chemically
bound to the carbonates. The C/Ca+Mg –atomic
ratio in calcite is 1. However, apart from a very low
ratio of 0.33 in N1-3, the ratios in PCD and calcium
carbonate deposits varied from 0.53 to 0.72 (cf. Fig-
ures 5 and 6). The C/Ca+Mg–ratio in carbonate
analysis N2-4 (cf. Figure 6B) was 0.53, which is
very close to 0.5 in Ca(HCO3 )2 . Hence PCD and
the other carbonate deposits were apparently a mix-
ture of calcite and calcium bicarbonate, although
amorphous calcium carbonate (ACC) might not be
ruled out. Also N1-6 (rust) of a thick corrosion
product after steel fiber apparently contained an
iron-carbonate component.
The thaumasite–like phases (N2-3 and N2-5), had
an odd composition with unusually high carbonate
contents. The Si/Al ratios were about 2 indicat-
ing intergrowth of thaumasite and ettringite. The
Ca/Si+Al ratios were 1.02 and 1.4 and the Ca/S
ratios were 2.0 to 3.8 as opposed to thaumasite and
ettringite ratios, being equal to 3 for both ratios in
both minerals. The mineral resembles aspects of
Ca6 Al3 (CO3 )3 (OH)12 · 26H2 O (carbonate ettringite).
However, in the present case Si substituted for Al
and carbonate might have substituted for some of
the OH groups.
The silica fume lump (N2-2) had a surprisingly
low Si and unexpected high Mg. SiO2 in silica
fumes are usually well above 80 wt. %, whilst
the Si atomic % in Table 4 corresponds to SiO2 =

393
 Hagelia

Figure 4: X-ray map of the domain shown in Figure 3. High carbon intensities within perlite and corrosion product
represent the epoxy used in thin section preparation, whilst lower C intensities represent calcium carbonate. High
K, Na, Al and Si represent aggregate. The cement paste matrix was extensively carbonated (Ca) containing a
little S, reflecting variably Mg substituted calcium carbonate phases. Na and Cl were scattered about at fairly low
concentrations in the matrix.

22.97 wt. %. The composition of brown mainly
unreacted transparent glass sphere (N2-23) corre-
sponded quite well with Type C fly ash. CaO in the
analysed particle is 22.64 wt. %. Type C fly ash is
not a highly pozzolanic material.
Small samples. The analysed domains were cho-
sen from thick areas and single phases such that
possible influence from the carbon tape should be

Table 4: SEM – EDS analysis of thin sections N1 (carbonated perlite concrete) and N2. All samples were from
extensively degraded perlite concrete; except for adjacent micro-PP fiber concrete N2-3 through N2-5. All data in
atomic %.

N1-3 N1-4 N1-5 N1 N1-6 N1-7 N1-8 N2-1 N2-2 N2-3 N2-4 N2-5 N2-23 N2-24
PCD PCD Paste Cc Rust Perlite CSH Cc SF T-E* Cc* T-E* FA? PCD
C 23.64 20.29 29.72 20.98 14.34 0.00 0.00 21.75 21.36 30.45 19.53 27.94 16.52 22.55
O 63.37 61.63 65.86 60.80 59.26 63.01 60.27 61.39 64.03 65.35 62.30 64.50 62.23 61.93
Na 0.40 0.45 0.29 0.58 0.00 2.72 1.31 0.66 0.45 0.34 0.85 0.36 1.02 0.00
Mg 0.49 0.50 0.16 0.40 0.00 0.00 3.07 0.27 4.31 0.00 0.45 0.20 0.67 0.10
Al 0.77 0.73 0.47 0.31 1.61 5.54 4.19 0.29 0.98 0.46 1.24 0.36 2.72 0.29
Si 3.23 3.18 2.01 1.03 3.37 25.61 20.13 1.50 6.52 1.02 5.03 2.04 7.35 0.65
S 0.13 0.13 0.03 0.22 0.00 0.00 0.00 0.29 0.02 0.76 0.06 0.89 0.00 0.26
Cl 0.00 0.00 0.00 0.00 0.00 0.00 0.31 0.00 0.06 0.05 0.04 0.08 0.00 0.00
K 0.36 0.44 0.17 0.09 0.00 2.00 1.45 0.00 0.00 0.06 0.13 0.05 0.50 0.00
Ca 7.40 12.47 1.23 14.78 1.00 1.12 8.26 13.85 1.90 1.51 9.99 3.38 7.73 14.23
Fe 0.21 0.20 0.06 0.79 20.42 0.00 1.01 0.00 0.37 0.00 0.38 0.21 1.28 0.00

394
 Figure 4: X-ray map of the domain shown in Figure 3. High carbon intensities within perlite and corrosion product represent
15th Euroseminar onin Microscopy
the epoxy used Applied
thin section preparation, tolower
whilst Building Materials
C intensities 17-19carbonate.
represent•calcium June 2015High•K,Delft,
Na, Al The
and SiNetherlands
represent aggregate. The cement paste matrix was extensively carbonated (Ca) containing a little S, reflecting variably Mg
substituted calcium carbonate phases. Na and Cl were scattered about at fairly low concentrations in the matrix.

A B

4
5

Figure 5: Thin section N1 of severely leached perlite concrete (Back Scatter). A: Dark grey = depleted cement paste, Light
colored irregular = Popcorn calcite deposits (PCD); Light colored strings = broken up and collapsed perlite. The aggregate to
Figure 5: Thin section N1 of severely leached perlite concrete (Back Scatter). A: Dark grey = depleted cement paste,
the right of inset image is quartz. Width of field = 300 μm: B: Detail with analytical points, with numbers corresponding to
Light colored irregular = Popcorn calcite deposits (PCD);
Table 5. Width Light
of field colored strings = broken up and collapsed perlite.
= 75 μm.
The aggregate to the right of inset image is quartz. Width of field = 300 µm: B: Detail with analytical points, with
numbers corresponding to Table 4. Width of field = 75 µm.

kept at a minimum. In contrast to the thin sections, detected here.

sodium was detected in all analysis (Table 5). The
analysed areas and points are indicated in Figure 7,
being less accurate than the thin section analyses.
Sample Nes 1 of thenardite-rich white powder of
completely degraded perlite concrete was analysed
as bulk and in two individual grains. The bulk
analysis (Nes 1a), representing an about 0.35 mm x
0.4 mm area centered on Figure 7A, showed very
low Si, being much lower than the quartz content
by XRD would seem to indicate. Moreover, the Na
and S concentrations (7.24 and 2.60 atom %) were
lower than expected for this thenardite rich mate-
rial. Thus, the sample is not exactly the same as
the one used for XRD, despite taken from the same
small batch. However the Na/S ratio was near that
of thenardite. The bulk analysis also indicated high
contents of C and O, indicating presence of carbon-
ate, with possible interference with the carbon tape.
The point analyses of thenardite –like single crys-
tals (Nes 1b and Nes 1c) showed that ratios of Na,
S and O are compatible with thenardite. In contrast
point analysis Nes 1c had no C despite sitting next
to the former thenardite with C, suggesting that C
variation is real. Minor amounts of Si, Al and Ca
may represent very fine debris. Cl and Mg were
absent.
Sample Nes 2 representing white degraded and
very friable perlite concrete was analysed in three
points. Nes 2b is an acicular phase quite similar
to CSH, yet with much S and Na. The Na/S ratio
was higher than ordinary sodium sulfate. Nes 2c is
a point analysis in fine grained material, perhaps
reflecting the composition of more than one phase.
Yet, the Na/S ratio was very high due to low S, sug-
gesting Na is not just restricted to sodium sulfate.
Nes 2d of a single grain has a Na/S ratio being
compatible with thenardite. It should further be
noted that Nes 2 and especially point 2c contained
Cl, likely derived from de-icing salts. Mg was also
Sample Nes 3 represents a small sample
7 collected
from grey friable micro-PP fiber concrete. Nes 3a is
a point analysis of a tiny grain, which may reflect
other near by grains. The analysis is somewhat com-
parable with bulk analysis of the entire frame of Fig-
ure 7C (3b), reflecting a mixture of CSH-like phases
and sodium sulfate. However, also in this case the
high Na/S ratios suggest excess Na. Point Nes 3c
reflects an outer thin deposit of Na and S bearing
material with Na/S ratio similar to thenardite, sit-
ting on a substrate of CSH-like material. Also 3d
resembles a mixture of thenardite and CSH. The
Ca/Si ratios in friable micro-PP concrete do not
indicate extensive Ca-depletion. A little Cl and Mg
were also present. Figure 8, which encompass Fig-
ure 7D, shows that this concrete was extensively
influenced by degradation in presence of sodium
sulfate.
The general impression of the debris samples
was that sodium sulfate is omnipresent and dissem-
inated within CSH-like phases. Sodium also seems
to be in excess in comparison with thenardite.
Sulfur isotopes
Sulfur isotopes were analysed in order to put con-
straints on possible sulfur sources. The thenardite-
rich sample Nes-1 and sample N1 with extensive
carbonation by sulfurised PCD and partly collapsed
perlite within Ca-depleted cement paste matrix
were selected. The relatively heaviest sulfate was
detected in bulk analysis of N1 (d34 S = +9.19 ‰),
whilst sulfate in thenardite and sulfate in bulk
residue of N1 after acid leaching were lighter (d34 S
= +7.44 ‰ and d34 S = +5.73 ‰, respectively). The
standard NBS 127 was run as unknown, yielding
d34 S = +21.60 ‰, which is 0.50 ‰ higher than the
true value. The data allow us to calculate the iso-
topic composition of sulfur in the carbonates, on
the likely assumption that sulfur isotopes behaved

395
 The silica fume lump (N2-2) had a surprisingly low Si and unexpected high Mg. SiO2 in silica fumes
are usually well above 80 wt. %, whilst the Si atomic % in Table 4 corresponds to SiO2 = 22.97 wt. %. The
composition of brown mainly unreacted transparent glassHageliasphere (N2-23) corresponded quite well with Type C
fly ash. CaO in the analysed particle is 22.64 wt. %. Type C fly ash is not a highly pozzolanic material.

Figure 6: Back scatter images of thin sections N1 and N2. A: Close up of perlite showing analytical points on glass (N1-7)
Figure
with 6: Back
fragment scatter images
of CSH-like of thin and
paste (N1-8) sections
calciteN1 andprecipitate
(Cc) N2. A: Close
insideupperlite
of perlite showing
(width of fieldanalytical
= 350 μm).points on glass
B: Carbon-rich
(N1-7) with fragment
thaumasite-ettringite of CSH-like
phases (N2-3 andpaste (N1-8)
N2-5) and calcite
and Ca(HCO (Cc) precipitate
3)2 (N2-4) (width of inside
field =perlite
350 μm).(width of field = 350
C: Undispersed silica B:
µm).fume
(N2-2) within outer porous micro-PP fiber concrete (width of field = 780 μm). D: Extensively carbonated
Carbon-rich thaumasite-ettringite phases (N2-3 and N2-5) and Ca(HCO3 )2 (N2-4) (width of field = 350 µm). C: perlite concrete
Undispersed silica fumewith suspect
(N2-2) fly ash
within (N2-23)
outer porousand circularfiber
micro-PP PCDconcrete
(width of field =
(width of 400
fieldμm).
= 780 µm). D: Extensively
carbonated perlite concrete with suspect fly ash (N2-23) and circular PCD (width of field = 400 µm).
Small samples. The analysed domains were chosen from thick areas and single phases such that
possible influence from the carbon tape should be kept at a minimum. In contrast to the thin sections, sodium
wasasdetected in all analysis
a conservative mixture.(Table
Thus,5). The analysed
accounting for areas and points
involving are indicated
depleted paste and in perlite.
Figure 7,The being less
inter-
accurate than the thin section analyses.
the mix proportions of bulk and residue of sample mediate composition of thenardite sulfur suggests
Sample PCD
N1, sulfurised Nes 1and of thenardite-rich white powder
associated carbonate and of completely
influence fromdegraded perlitesources.
two different concrete was analysed
as possibly
bulk and other
in twosoluble
individual grains.
sulfate 34 S =
The dbulk
yielded analysis
+9.57 (Nes 1a), representing an about 0.35 mm x 0.4 mm area
centered on Figure
‰. However, in 7A,
viewshowed very low
of the results fromSi, NBS
being127,
much lower than the quartz content by XRD would seem to
indicate.
the realMoreover, the Na
values should and S 0.5
be about concentrations
‰ lighter in (7.24
all and 2.60 atom %) were lower than expected for this
thenardite
samples.rich
Thematerial.
lightest Thus, the sample
S isotopic is notin
composition exactly
the the same as the one used for XRD, despite taken from
theresidue
same small batch. However
represents refractorythe Na/Ssitting
sulfur ratio was near that of thenardite. IV.
within Discussion
The bulk analysis also indicated high
contents of C and O, indicating presence of carbonate,
less soluble material in perlite concrete, possibly with possible interference with the carbon tape. The point
analyses of thenardite –like single crystals (Nes 1b and Nes 1c) showed that ratios of Na, S and O are compatible
with thenardite. In contrast point analysis Nes 1c had no CThis paper
despite has shown
sitting next to that the exceptional
the former thenarditedegree
with C,
of spalling and steel fiber
suggesting that C variation is real. Minor amounts of Si, Al and Ca may represent very fine debris. corrosion withinCl sprayed
and Mg
were absent. concrete for fire protection in the Nes road tunnel
Sample Nes 2 representing white degraded and very was intimately associated
friable perlite with was
concrete severe Ca-depletion
analysed in three
points. Nes 2b is an acicular phase quite similar to CSH, yet with much S and Na. The Na/S ratio wasinflu-
and extensive carbonation under significant higher
than ordinary sodium sulfate. Nes 2c is a point analysis ence in offine
sodium
grainedsulfates. Theperhaps
material, presentreflecting
discussionthe
composition of more than one phase. Yet, the Na/S ratio was focuses
veryon thedue
high causes, mechanisms
to low and Na
S, suggesting the issources
not just
of aggressives, as high-lighted by
restricted to sodium sulfate. Nes 2d of a single grain has a Na/S ratio being compatible with thenardite. It should the overall con-
text of the tunnel concrete. The tunnel
further be noted that Nes 2 and especially point 2c contained Cl, likely derived from de-icing salts. Mg was also environment
detected here. here was characterised by variable moisture loads,
Sample Nes 3 represents a small sample collectedfrost fromaction
grey and heavy
friable traffic fiber
micro-PP underconcrete.
influenceNes of de-
3a is
a point analysis of a tiny grain, which may reflect other near icingbysalts, being
grains. The dependent
analysis isonsomewhat
seasonal variations.
comparable
with bulk analysis of the entire frame of Figure 7C (3b), The fact that
reflecting spallingof
a mixture in CSH-like
the perlitephases
concrete andhad led
sodium
to repair work already in 2002 and that spalling in
both new and old perlite concrete continued, hav-
gure 8: Sample Nes 3 of micro-PP fiber concrete showing significant degradation due toing
sodium sulfateweakening
caused and also in adjacent new outer
Figure 8: Sample Nes 3carbonation.
of micro-PP fiber concrete show- micro-PP concrete without perlite, seems to suggest
s
ing significant degradation due to sodium sulfate and that the degrading agents/mechanisms were acting 9
carbonation.
ur isotopes were analysed in order to put constraints on possible sulfur sources. overThea long time. This needs to be explained.
thenardite-
es-1 and sample N1 with extensive carbonation by sulfurised PCD and partly collapsed perlite
leted cement paste matrix were selected. The relatively heaviest sulfate was detected in bulk
396
(δ34S = +9.19 ‰), whilst sulfate in thenardite and sulfate in bulk residue of N1 after acid leaching
δ S = +7.44 ‰ and δ34S = +5.73 ‰, respectively). The standard NBS 127 was run as unknown,
34

= +21.60 ‰, which is 0.50 ‰ higher than the true value. The data allow us to calculate the isotopic
not indicate extensive Ca-depletion. A little Cl and Mg were also present. Figure 8, which encompass Figure 7D,
shows that this concrete was extensively influenced by degradation in presence of sodium sulfate.
15thThe
Euroseminar on Microscopy
general impression Applied
of the debristosamples
Buildingwas
Materials • 17-19 sulfate
that sodium June 2015 • Delft, The Netherlands
is omnipresent and disseminated
within CSH-like phases. Sodium also seems to be in excess in comparison with thenardite.

A B
2d
2b
2c
1b
1c

C D

3c
3a

Figure 7: BSE images of selected domains in small samples. The analytical results are presented in Table 5. A: Sample Nes 1
Figureconcrete)
(perlite 7: BSE images of selected
with analysed domains
points 1b andin1c;
small samples.
Image width =The
250analytical results Nes
μm. B: Sample are presented in Table 5.with
2 (perlite concrete) A: Sample
analysed
Nes 2b,
points 1 (perlite
2c and concrete)
2d; Imagewith
widthanalysed
= 250 μm. points 1b andNes
C: Sample 1c; 3Image widthfiber
(micro-PP = 250 µm. B:with
concrete) Sample Nes 2point
analysed (perlite concrete)3b
3a. Analysis
represents
with a bulk
analysed scan
points 2b,over the entire
2c and frame;
2d; Image Image
width width
= 250 µm.= 410 μm. D: Nes
C: Sample Nes33 (micro-PP
with point fiber
3c. Image width
concrete) = 200
with μm.
analysed
point 3a. Analysis 3b represents a bulk scan over the entire frame; Image width = 410 µm. D: Nes 3 with point 3c.
Table
Image 5: width
SEM –=EDS200 analysis.
µm. All samples were from extensively degraded perlite concrete, except for Nes-3 of micro-PP
fiber concrete. The micro domains are shown in Figure 7 (except 3d).
Nes 1a Nes 1 b Nes 1 c Nes 2b Nes 2c Nes 2d Nes 3a Nes 3b Nes 3c Nes 3d
Formation Bulk paste Na-sulf
conditions for sodium Na-sulf Single mixed black
sulfate Single
shale Paste
and subsea Bulkenvironments
paste mixed (Hageliamixed et al.
C 28.47 19.94 0.00 7.62 11.07 2001;10.98 14.29 22.66 17.23
Hagelia et al. 2003, Hagelia 2011); in 19.28
cast con-
DueOto the 61.03severe Ca-depletion
49.44 24.43of CSH 51.61involv-
46.36 crete51.4involving
59.99seawater
62.00 attack55.81 60.44
in brackish water
ing Na
capture7.24of Ca in calcite
18.40 and calcium12.22
46.06 bicarbon-
21.58 (Sibbick
17.71 et 4.46
al. 2003a1.81 8.30
& b) and concrete 4.27
in sewage
ate, Mg
formation
0.00 of gypsum 0.00 and 0.00thaumasite 0.37in pres-
0.68 0.4 0.89 0.54 1.12 0.63
systems (Fernandes et al. 2012). PCD is frequently
enceAlof sulfate
0.04 ions should
0.02 be0.23
strongly1.69 1.38
inhibited. 1.61 2.31 0.56 0.82 1.45
Si 0.13 0.11 0.42 6.74 3.76 associated
7.65 with
8.95 thaumasite
2.03 sulfate attack5.61
2.51 and sulfu-
Thus, when significant Na and sulfate is present,
S 2.60 11.97 28.59 3.84 1.52 ric acid
7.53 attack,
0.68 but may
0.20 also take place
4.09 in
1.43absence
thenardite and other sodium sulfates may form.
Cl 0.00 0.00 0.00 1.73 10.35 of sulfate,
0.23 e.g.
1.22 in through
0.36 solution
0.28 processes
0.64 with
ThisKhas also
0.00 been observed in0.00
some Austrian
0.63 tun- 0.28 ingress
0.03 of a0.72
0.00 carbonate source, such as bicarbonate
0.22 0.00 0.39
nels,Cawhere0.12mirabilite0.13formed 0.33
at a very7.26late stage
2.34 (Gaze 1.34 and 5.88
Crammond 2000; Hagelia et5.73
8.83 9.47 al. 2001;
afterFethaumasite
0.00 sulfate
0.00attack0.00(Mittermayr6.30 2012).
0.68 Sibbick
1.03 et al.0.54
2003). 0.19 0.38 regarded
PCD is generally 0.00 as a
Hence,
Ti in0.00such contexts0.00internal 0.00carbonation
0.00 is 0.00
a 0.00of carbonation
0.08 0.60
form which may0.00 0.12
lead to significant
prerequisite
F 0.35for later stage
0.00 formation
0.00 of sodium
0.00 sul-
0.00 0.00 0.00 0.00 0.00 0.00
weakening of the cement paste matrix, involving in-
fate. Water glass used as setting accelerator was
creased water permeability and steel reinforcement
likely a main source of sodium, although de-icing
corrosion. This stands in marked contrast to ordi-
salt likely also contributed. This would explain
nary surface carbonation, being mainly harmful to
some evidence of excess Na relative to sulfur. Be-
steel reinforcement.
fore turning to the discussion of possible sulfate
sources we need to explain the mechanisms causing Van der Sloot (2000) reported that Ca is much
severe Ca-depletion and internal carbonation. more leachable than Si, being about 90-95 % and
0.015 % respectively. Obviously, these figures are
manifested in concretes which have suffered Pop-
Mechanism of Ca-leaching and carbonation corn calcite deposition, e.g. where ingress of exter-
Ca-leaching was omnipresent in the perlite con- nal bicarbonate-bearing water with a neutral pH
crete with ubiquitous presence of Popcorn calcite meets the alkaline cement pore fluids, causing en-10
deposition (PCD) with additional very fine grained trapment of leached Ca in secondary calcite, leaving
carbonation. The PCD form of carbonation is com- behind a substrate mainly consisting of amorphous
monly observed in submerged concretes structures, silica and other less mobile elements.
such as tunnel sprayed concrete for rock support in The analytical data from the Nes tunnel showed

397
 Hagelia

Table 5: SEM – EDS analysis. All samples were from extensively degraded perlite concrete, except for Nes-3 of
micro-PP fiber concrete. The micro domains are shown in Figure 7 (except 3d).

Nes 1a Nes 1 b Nes 1 c Nes 2b Nes 2c Nes 2d Nes 3a Nes 3b Nes 3c Nes 3d
Bulk paste Na-sulf Na-sulf Single mixed Single Paste Bulk paste mixed mixed
C 28.47 19.94 0.00 7.62 11.07 10.98 14.29 22.66 17.23 19.28
O 61.03 49.44 24.43 51.61 46.36 51.4 59.99 62.00 55.81 60.44
Na 7.24 18.40 46.06 12.22 21.58 17.71 4.46 1.81 8.30 4.27
Mg 0.00 0.00 0.00 0.37 0.68 0.4 0.89 0.54 1.12 0.63
Al 0.04 0.02 0.23 1.69 1.38 1.61 2.31 0.56 0.82 1.45
Si 0.13 0.11 0.42 6.74 3.76 7.65 8.95 2.03 2.51 5.61
S 2.60 11.97 28.59 3.84 1.52 7.53 0.68 0.20 4.09 1.43
Cl 0.00 0.00 0.00 1.73 10.35 0.23 1.22 0.36 0.28 0.64
K 0.00 0.00 0.00 0.63 0.28 0.03 0.72 0.22 0.00 0.39
Ca 0.12 0.13 0.33 7.26 2.34 1.34 5.88 8.83 9.47 5.73
Fe 0.00 0.00 0.00 6.30 0.68 1.03 0.54 0.19 0.38 0.00
Ti 0.00 0.00 0.00 0.00 0.00 0.00 0.08 0.60 0.00 0.12
F 0.35 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

that PCD and associated calcium carbonate precipi- rographic features.

tates were not pure calcite. Instead they apparently
represent a mixture of calcium carbonate and cal-
cium bicarbonate (Ca(HCO3 )2 ) and were otherwise
sulfurised, substituted by Mg and also contained a
little K, Na, Al and Si. Standard chemical literature
states that calcium carbonate dissolves in excess of
CO2 to form calcium Ca(HCO3 )2 . It is interesting
that the XRD of small samples did not record a
lot of calcite as opposed to what was indicated by
generally elevated C and O in the SEM-data. Hence,
it seems possible that these carbonates in fact may
represent some form of amorphous calcium carbon-
ate (ACC).
In the Nes tunnel, it seems evident that bicar-
bonate was present in the ground water, due to
the occurrence of thermonatrite efflorescence on
the rock mass. However, spalling had occurred all
over the place, and was not restricted to water seep-
ing through joints in the PE-foam panels. Hence
we mainly need to invoke atmospheric CO2 as the
likely main source of carbonate, which when dis-
solving into water forms carbonate ions when pH
is much alkaline, and bicarbonate at circum-neutral
pH. Nevertheless, this cannot explain why severe
Ca-depletion was brought about. Indeed, such a
reaction is most typically associated with influence
of acidic waters, as was also suggested by the pet-
Origin of thenardite sulfate
No report suggests significant concentrations of sul-
fate in the local ground water. There is no industry
in the area, and the rock mass is completely domi-
nated by rhomb porphyry which has a feldspar-rich
mineralogy with composition similar to latite. Most
of the spalling concrete was also screened from ex-
ternal water by still intact PE-foam plates. Hence, it
must be concluded that the sulfate originated from
the sprayed perlite concrete itself.
The S isotopic data showed that thenardite sulfur
must reflect two different sources. Possible sources
are the gypsum added to the cement and perlite,
because the concrete aggregate did not contain any
sulfur bearing mineral. The heaviest sulfur was
sitting in the sulfurised carbonates (calculated to
d34 S = +9.57 ‰), whilst the lightest was sitting in
residue after acid leaching (d34 S = +5.73 ‰). For
comparison, previous analysis of fairly intact CSH
in thaumasite bearing Norwegian sprayed concrete
were ranging from d34 S = +10.6 ‰ to d34 S = +11.8
‰. These concretes were made in 1987 and 1999,
respectively, and their S isotopic compositions were
interpreted to largely reflect gypsum added to the
cements (Iden and Hagelia 2003). However, associ-
ated thaumasite derived from external sulfate had

Table 6: S isotopic signatures of sulfates in thenardite; bulk carbonated concrete and residue after acid leaching.

Sample No. Sample type d34 SCDT (‰) S (Wt. %)

Nes-1 Powder rich in thenardite with some quartz +7.44 -
N1 Bulk of extensively carbonated concrete +9.19 0.10
N1 Residue after washing with 2M HCl +5.73 0.01
NBS 127 Standard with true value d34 SCDT = 21.1‰ +21.60

398
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
d34 S ⇡ +2 ‰, which hints that the composition of
added gypsum might have been slightly heavier
than about 11-12 ‰.
It should be noted that gypsum sources used in
cement production are represented by ancient evap-
orites, having distinct age- dependant S isotopic
signatures. This depends on the isotopic variations
in the ancient seas from which gypsum evaporated.
The lightest possible S isotopic signatures of sig-
nificant gypsum deposits are represented by up-
permost Permian to lowermost Triassic evaporites,
being d34 S ⇡ +10-13 ‰(Claypool et al. 1980). Data
from Zechstein anhydrite units in West Poland have
d34 S = +9.6 ‰to +12.6 ‰(Peryt el al. 2010). On the
other hand other sources of gypsum with heavier
sulfur are possible. Unpublished data of the au-
thor from Norwegian cement shows that gypsum
with d34 S = +21.3 ‰has been used just after year
2000. However, there are reasons to believe that the
Nes tunnel was made with cements from the same
factory as our previously investigated sprayed con-
cretes, being constructed at more or less the same
time. Hence, it may be concluded that all of the S
isotopic signatures obtained from the Nes tunnel
are significantly lower than the minimum estimate
for added gypsum (d34 S ⇡ +11-12 ‰): The d34 S
compositions obtained in a) sulfurised carbonate
(PCD), b) thenardite and c) the residue after acid
leaching are about 2 ‰, 4 ‰and 6 ‰lighter than
any possible gypsum source, respectively. All com-
positions may be regarded as intermediate between
two sulfur sources. It is implicit that the heavier
end-member must be represented by added gyp-
sum, whilst the lighter end-member clearly seems
represented by sulfuric acid.
The lighter sulfate source involved was bound to
the less soluble phases in the Nes carbonated per-
lite concrete. It is therefore quite possible that this
is represented by SO2 gas initially trapped within
the pore space of extruded perlite. In the first place
SO2 must be expected to have been present in the
perlite raw material, being a common volcanic gas.
Moreover, during the production of extruded per-
lite this gas can be trapped in the pore space if oil
or coal has been used for combustion. However,
for the extruded perlite used in the Nes tunnel, no
chemical documentation is available.
The main point here is that sulfur dioxide has a
very light S isotopic composition, frequently show-
ing negative d34 S signatures (Faure 1986), and that
it reacts with water to form sulfuric acid. In view
of the observation that the pore space of perlite had
frequently collapsed during the spraying operation
(Figure 5B) this should indeed facilitate direct ac-
cess to any possible aggressive gas at the earliest
stage, likely before and during early setting of the
cement. If so, it would seem likely that cement
hydration never went well, and that portlandite
perhaps never formed. This may also explain why
sulfur was distributed evenly throughout the ce-
ment paste matrix as seen in Figure 4. This is not
at odds with the fact that the perlite glass was sul-
fur free. It is, however, not possible to put further
constraints on the amount of sulfuric acid involved,
because we do not explicitly know whether the
added gypsum sulfur was similar to Zechstein or
even heavier.
Reaction mechanisms
The fact that thenardite (type V) and sodium sulfate
(type III) occurred in the main spalling zone shows
that salt crystallisation represented an important
ingredient in the spalling process. It was argued
above that Ca-leaching assisted by sulfuric acid de-
rived from perlite, with PCD acting as a sink for
liberated calcium was a prerequisite for the forma-
tion of sodium sulfate. Yet, the sprayed concrete
used for fire protection in the Nes tunnel had clearly
been influenced by frost and also variable moisture
loads in presence of de-icing salts, depending on
seasonal variation. Moreover, this kind of tunnel
lining is sensitive to pressure waves from traffic, be-
cause the construction is pretty slim and mounted
to the rock mass by bolts. The thin concrete layers
outside foam panels has in fact been subjected to
dynamic loads, which must have resulted in micro-
cracking at an increasing rate, hence making the
concrete more and more susceptible to degradation
by combined freeze/thaw, salt crystallisation and
dissolution.
The relative humidity (RH) in Norwegian tun-
nels is generally < 60 %. For this reason it is of
no surprise that mirabilite was not detected, since
this mineral is only stable at RH > 65 % for Norwe-
gian tunnel temperatures. Until recently the conver-
sion of thenardite to mirabilite has been regarded
as most important for build-up of crystallisation
pressure and spalling (Skalny et al 2002), owing to
their great differences in molar volume (Vthenardite =
53.30 cm3 /mol versus Vmirabilite = 218.56 cm3 /mol).
However, research has shown that the effects of
thenardite have been underestimated. This is due
to more complex phase changes, along with the
significant impact from degree of supersaturation.
Indeed, the growth pressure of thenardite may be
about four times higher than that of mirabilite at
high supersaturation. Cyclic conversion tests have
demonstrated that thenardite alone can inflict con-
siderable damage to rocks of different porosities (Yu
and Oguchi 2013). Rodriguez-Navarro et al (2000)
reported that fast evaporation due to lowering of
RH conditions, leading to high supersaturation in
micropores before thenardite precipitation, results
in high crystallisation pressure of thenardite, whilst
mirabilite crystallises at lower level of supersatura-
399
 Hagelia
tion generally as efflorescence.
Due to the extensive presence of weak and
powder-like concrete it is likely that mass trans-
fer during freezing in the spalling region might be
compared to mechanisms taking place within saline
soils. Chuvilin et al. (1998) have demonstrated that
water soluble salts such as sodium sulfate have a
fundamental influence on the process of ice for-
mation and the formation of cryogenic structures.
They reported that the mass transfer of salt in freez-
ing saline soils is a result of ion transfer within
the water flux, surface conduction, diffusion and
thermodiffusion; the most important process is the
transfer of salt ions within the water flux. The max-
imum accumulation of water corresponded to the
largest accumulation of salt, hence being prone to
supersaturation of thenardite.
V. Conclusions
The Nes road tunnel was closed for nearly three
years due to extensive spalling of perlite concrete
used for fire protection. The present investiga-
tion has shown that the degradation process began
with early stage leaching of the cement paste ma-
trix, due at least in part to involvement of sulfuric
acid derived from sulfur dioxide gas in the per-
lite pore space. PCD and other carbonates formed
a sink for Ca, which consequently made way for
sodium sulfate formation with significant and re-
peated salt crystallisation/dissolution events taking
place within the main spalling zone. During win-
ter salt scaling was acting in harmony with the
freeze/thaw cycles. The effects of dynamic loads
from traffic accentuated the speed of the attack.
Acknowledgements
The author is deeply indebted to mineralogist and
analyst Hans-Jørgen Berg († 2014) of the Natural
History Museum (NHM), Oslo, for technical assis-
tance with XRD and SEM, and not the least for
his knowledge and insight, which he always shared
with us with the greatest dedication. I also acknowl-
edge technical assistance by Harald Folvik NHM
and skillful preparation of thin sections by Sala-
hadin Akhavan at the University of Oslo. Ingar
Johansen and Christian Alexander Schöpke at Insti-
tute for energy technology, Kjeller, deserve thanks
for doing the isotope analysis. Lastly the author
wishes to thank Synnøve Myren at the Norwegian
Public Roads Administration for assistance and for
digging up important information. This report was
prepared for the R&D project “Durable structures”
at NPRA.
400
References
Chuvilin, E.M., Ershov, E.D. and Naletova, N.S. (1998): “Mass trans-
fer and structure formation in freezing saline soils.” Permafrost - Pro-
ceedings of 7th International Conference, Yellowknife, Canada, Collection
Nordicana, No 55, 173-179.
Claypool, G.E., Holster, W.T., Kaplan, I.R. Sakai, H. and Zak, I. (1980):
“The age curve of sulfur and oxygen isotopes in marine sulfate and their
mutual interpretation”. Chemical Geology, 28, 199-260.
Davik, K.I. (1998): “Proper use of sprayed concrete”. Nordic Road and
Transport Research, 1998-1, 16-17.
Fernandes, I., Pericão, M., Hagelia, P. Noronha, F., Ribeiro, M.A. and
Maia, J. (2012): “ Identification of acid attack on concrete of a sewage
system”. Materials and Structures, 45-3, 337-350.
Faure, G. (1986): “Principles of Isotope Geochemistry”. (2nd Ed.).
John Wiley and Sons, 589 p.
Gaze, M.E. and Crammond, N.J. “The formation of thaumasite in a
cement, lime, sand mortar exposed to cold magnesium and potassium
sulfate solutions”. Cement and Concrete Composites, 22, 209-222.
Hagelia, P. (2007): “Sprayed concrete deterioration influenced by
saline ground waters and Mn-Fe biomineralisation in subsea tunnels”.
In: B. Jamtveit (Ed), Mechanical Effects on Reactive Systems, the 20th Kongs-
berg Seminar, 2007, p 26 (abstract).
Hagelia, P. (2011): “Deterioration Mechanisms and Durability of
Sprayed Concrete for Rock Support in Tunnels” PhD thesis, TU-Delft,
205 pp and Appendices.
Hagelia, P. (2013): “Interaction of abiotic and biochemical reactions
and their role in concrete deterioration”. Concrete, July/August, 49-51.
Hagelia, P., Sibbick, R.G., Crammond, N.J., Grønhaug, A.W. and
Larsen, C.K. (2001): “Thaumasite and subsequent secondary calcite de-
position in sprayed concrete in contact with sulfate bearing Alum Shale,
Oslo, Norway”. 8th Euroseminar on Microscopy Applied to Building Materi-
als, Athens, Greece, 131-138.
Hagelia, P., Sibbick, R.G., Crammond, N.J. and Larsen, C.K. (2003):
“Thaumasite and secondary calcite in some Norwegian concretes”. Ce-
ment and Concrete Composites, 25, 1131-1140.
Hagelia, P. and Sibbick, R.G. (2009): “Thaumasite sulfate attack, pop-
corn calcite deposition and acid attack in concrete stored at the Blind-
tarmen test site Oslo, from 1952 to 1982.” Materials Characterisation, 60,
686-699.
Stenger, H. G., Hu, K., & Simpson, D. R. (1993). Competitive adsorp-
tion of NO, SO2 and H2 O onto mordenite synthesized from perlite. Gas
separation & purification, 7(1), 19-25.
Holm, J.V. (2011): “E16 Nestunnelen – nedfall av sprøytebetong”.
Norconsult - Notat nr.1, Oppdragsnr. 5111160.
Iden, K.I. and Hagelia, P. (2003): “C, O and S isotopic signatures in
concrete which have suffered thaumasite formation and limited thauma-
site form of sulfate attack”. Cement and Concrete Composites, 25, 839-846.
Mittermayr, F. (2012): “Why thaumasite is forming in concrete struc-
tures”. PhD thesis, TU-Graz, 119 pp.
Norwegian Public Roads Administration (1997): “Proper use of
sprayed concrete in tunnels”. Parts A, B, C, D, E and Final report (In
Norwegian).
Peryt, T.M., Hałas, S. and Hryniv, S.P. (2010): “Sulphur and oxygen
isotope signatures of late Permian Zechstein anhydrites, West Poland:
seawater evolution and diagenetic constraints”. Geological Quarterly,
54(4), 387-400.
Rodriguez-Navarro, C., Doehne, E. and Sebastian, E. (2000): “How
does sodium sulfate crystallise? Implications for the decay and testing
of building materials”. Cement and Concrete Research, 30, 1527-1534.
Sibbick, R.G., Crammond, N.J. and Metcalf, D. (2003a): “The micro-
scopical characteristics of thaumasite.” Cement and Concrete Composites,
25, 831-837.
Sibbick, R.G., Fenn, D. and Crammond, N.J. (2003): “The occurrence
of thaumasite as a product of seawater attack”. Cement and Concrete Com-
posites, 25, 1050-1066.
Skalny, J.P., Marchand J., Odler I. (2002): “Sulfate Attack on Concrete”.
Modern Concrete Technology Series 10, Spon Press – Taylor and Francis
Group, London and New York
Van der Sloot, H.A. (2000) “Comparison of the characteristic leaching
behaviour of cements using standard (EN 196-1) cement mortars and an
assessment of their long-term environmental behaviour in construction
products during service life and recycling”. Cement and Concrete Research,
30-7 1079-1096.
Yu, S. and Oguchi, C.T. (2013): “Is sheer thenardite attack impotent
compared with cyclic conversion of thenardite-mirabilite mechanism in
laboratory simulation tests?” Engineering Geology, 152, 148-154.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Monitoring internal sulphate reactions by

X-ray tomography
A. Darquennesa , Y. Wanga , F. Benboudjemaa , G. Nahasb
a LMT (ENS Cachan, CNRS, Université Paris Saclay), 94235 Cachan, France
b Institut de radioprotection et de sureté nucléaire, Fontenay-aux-Roses, France
⇤ aveline.darquennes@ens-cachan.fr

Abstract

For certain cement chemical compositions and under specific temperature and humidity conditions, concrete deterio-
ration can be attributed to internal sulphate reactions. This pathology mainly affects massive structures (e.g. bridge
piers, dams, nuclear power plants) or precast concrete structures that undergo specific heat treatments. Without an
external sulphate source, sulphate reaction is characterized by delayed ettringite formation leading to swelling of the
cementitious matrix and cracking. Nevertheless, the mechanisms and the impact of this pathology on the microscopic
and structural scales are still today not thoroughly understood. A performance based approach (accelerated treatment)
was recently proposed by IFSTTAR (test method n 66) in order to determine the sensitivity of a cementitious mixture to
the delayed etttringite formation. It consists in longitudinal swelling measurements on cylindrical specimens submitted
to drying and wet cycles followed by a water curing at 20 C.
In this work, the internal sulphate reaction in a concrete specimen that has followed an accelerated treatment is
monitored by means of X-ray tomography. X-ray tomography is a non-destructive test that provides information on
the 3D mesostructural evolutions. Using a post-treatment method implemented in Matlab, the evolution of several
parameters (porosity, cracking) is quantified. The interfacial zone between aggregate and paste matrix is also studied.
Supplementary observations performed by means of optical and scanning electron microscopy confirm that the delayed
ettringite is principally formed in the large pores and in the interfacial transition zone of the cementitious matrix.
Finally, using a 3D numerical reconstruction method aggregates can be seperated from the cementitious matrix to create
a finite element mesh of the specimen at the mesoscale level that can be used for numerical simulations.

Keywords: Cracking, Delayed ettringite, Mesoscale, Porosity, X-ray tomography

I. Introduction after several years. In order to compare the sensi-

In France, the internal sulphate reaction is a rela-
tively newly studied pathology - since the nineties.
It is linked to the delayed ettringite formation with-
out an external sulphate source. Ettringite develop-
ment depends on several parameters as the cement
chemical composition, the maximal temperature
inside the concrete at early age and its humidity
content. When temperature reaches a value supe-
rior to 65 C - this is the case for massive structures
as bridge piles and precast concrete - ettringite for-
mation slows down, stops or dissolves (Divet et al,
1998). After concrete cooling and for an adequate
moisture content in the concrete porous network,
delayed ettringite can be formed inside the hard-
ened concrete. Ettringite mainly appears in the
large pores and in the aggregate/paste interface (To-
sun, 2006). Its formation can induce high swelling
pressure leading to a global matrix expansion and
to its cracking. It is important also to underline
the fact that the kinetics of this phenomenon is
very slow. Indeed, concrete swelling can appear
tivity of different concrete mixtures to the delayed
ettringite formation, IFSTTAR developed an accel-
erated method, named test method n 66 (Pavoine
et al, 2006 and 2007). It consists on applying a
thermal treatment at early age following by wet
and drying cycles. To quantify the evolution of the
delayed ettringite formation, the global expansion
and the mass variation are monitored (Pavoine et al,
2006; Escadeillas et al, 2007) as also the permeabil-
ity (Al Shamaa et al, 2014), the strength (Bouzabata
et al, 2012), the Young’s modulus (Al Shamaa et al,
2014), and the microstructure by means of Scanning
Electron Microscopy (SEM) pictures (Tosun, 2006).
However, the majority of these experimental tests
lead to the specimen destruction. In this study, we
propose to use a 3D non-destructive method, the
X-ray tomography, combined with an optical and
scanning electron microscopy (SEM) to observe and
quantify the effect of the internal suphate reaction
on the microstructure.

401
 Darquennes et al.
Figure 1: Position of the sample for the global (a) and the local (b) scanning (top view).
II. Experimental
Materials
In order to determine the sensitivity of a cementi-
tious mixture to the delayed etttringite formation, a
performance based approach (accelerated test) was
proposed by IFSTTAR (test method n 66) (Pavoine
et al, 2007). It consists in longitudinal deformation
measurements on cylindrical specimens submitted
to drying and wet cycles followed by a water curing
at 20 C. In this study, the sensitivity of a Portland
cement concrete with siliceous aggregates - insensi-
ble to alkali-aggregate reactions - is studied using
a cylindrical specimen (diameter f = 11 cm and
height h = 22 cm). Its Water/Cement (W/C) ratio is
equal to 0.48 and the concrete composition is given
in Table 1.
A thermal treatment for a period of 7 days was
applied (Al Shamaa, 2012). After mixing, the speci-
men was kept at 20 C during 2h. Then, the temper-
ature increased until 80 C (heating velocity = 2.5
C/h) and kept constant during 3 days. Finally, the
specimen was cooled until a temperature of 20 C
(cooling velocity rate is equal to -1 C/h) and then
it was demoulded. It was kept at 20 C and 91%
of Relative Humidity (R.H.) for 334 days, and then
under water. The tested specimen was 774 days old
and was characterized by a linear expansion equal
to 1.3%. After the X-ray tomography, it was sliced
in order to analyze its microstructure.
Figure 2: Sub-volume of the studied sample after the 3D
numerical reconstruction.
402
Experimental tests
The effects of the internal suphate reactions on the
concrete microstructure were monitored by X-ray
tomography. This analysis was completed by opti-
cal and scanning electronic microscopy (SEM) tests.
X-ray tomography
During the X-ray tomography test, the concrete
specimen was placed on a rotating support between
the X-ray tube and the detector (Figure 1). The
specimen was then crossed by X-ray beams and
2D radiographies around the rotation axis were ob-
tained. Finally for each specimen a 3D image was
reconstructed. The specimen dimensions control
the X-ray tomography picture resolution. Indeed, a
thinner specimen facilitates crossing, leading to a
smaller voxel size and to a better resolution. As the
specimen dimensions could not be modified, two
types of scanning were applied: a local (Figure 1a)
and a global (Figure 1b) one. To limit the length
L1 or L2 (Figure 1) and thus to improve resolution,
scanning was performed in three positions: at the
top, the middle and the bottom of the specimen.
The obtained picture resolution is equal to (approx-
imately) 60 µm and (approximately) 30 µm for the
global and local scan respectively.
A 3D image of the specimen was finally numeri-
cally reconstructed (Figure 2). In this image, grey-
levels vary as function of the material capacity to
absorb X-rays. Areas with a lower density (such
as air in cracks or in pores) are characterized by a
darker grey-level making thus feasible to identify
the porosity and the paste matrix.
Table 1: Concrete Composition.
Materials Quantity (kg/m3 )
Cement CEM I 52.5 N 400
Sand 0/0.315 130
Sand 0.315/1 130
Sand 1/4 450
Aggregate 4/8 210
Aggregate 8/12.5 570
Aggregate 12.5/20 310
Total Water 190
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Initial 2D image from the sub-volume (a) and the corresponding grey-level histogram (b).

The method applied to isolate and to quantify grey-level threshold is identified to isolate the
the void volume is briefly described hereafter: interface between aggregates and matrix by
segmentation (Figure 8a) and the aggregates
• To limit the effect of artifacts related to the
(Figure 8b).
specimen boundary, a sub-volume (Figures 2
and 3a) is isolated inside the specimen. • As Figure 8b has a lot of noise and the ag-
gregates are presented with white and black
• To determine the grey-levels threshold separat-
pixels, a further post-treatment is applied. To
ing the voids from the paste matrix (segmen-
isolate the aggregates, the continuity of the
tation), lines are drawn through the 2D image
line giving the aggregate contour is amplified
(Figure 4). This step is performed after each
and the pixels of the aggregate surface are
scan.
represented with white pixels using a dilation
• Based on the grey-level threshold separating method (Figure 9).
the porosity from the paste matrix (Figure 4),
It is important to identify the localization and the
the picture is segmented: white pixels corre-
form of aggregates as the delayed ettringite seems
spond to voids and black pixels to the paste
to form preferentially in this area (Tosun, 2006).
matrix (Figure 5a).
• The segmentation is applied on all 2D images
by means of numerical loops (using Matlab)
or on the 3D volume using a specific software
(e.g. ImageJ) (Figure 5b). These two methods
provided similar results (Table 2).

Notice however that this method cannot be di-

rectly applied to isolate aggregates. Indeed, the
grey-levels corresponding to aggregates and to the
cementitious matrix are very close (Figure 6). It
is therefore necessary to amplify the contrast be-
tween these two areas. The following procedure is
applied:

• To limit the artifacts related to the specimen

boundary, the grey-level of the entire surface
around the specimen is defined as the grey-
level corresponding to the porosity (Figure 7).
• The contrast between matrix and aggregates Figure 4: Two linear profiles containing paste matrix
on the 2D picture is amplified (Figure 7). The and porosity: line 1 (b) and line 2 (c).

403
 Darquennes et al.

Figure 5: Segmented images in 2D (a) and in 3D (b).

Microscopy analysis tomography several observations (Figure 10) can be

To identify the different areas (voids, interface be-
tween paste and aggregate, etc.) observed on the 3D
X-ray tomography images, the different specimen
slices were analyzed by means of optical and scan-
ning electron microscopy (SEM) with energy dis-
persive X-ray spectroscopy (EDS). The observations
were performed on coarse and polished surfaces.
III. Results and Discussion
The main effect related to the internal sulphate reac-
tions is the ettringite dissolution/recrystallisation.
Moreover, the delayed ettringite located on the in-
terface between aggregates and paste can generate
high pressures leading to cracking. From the X-ray
made:
• Large porosity is probably related to the ettrin-
gite dissolution.
• Darker areas characterize the interfacial area
around the aggregates.
The different areas were isolated using previously
described approach. Results are detailed hereafter
and the validation of the X-ray tomography observa-
tions is provided by means of microscopy analyses.
Porosity
Based on the post-treatment of the X-ray tomogra-
phy images, porosity (large pores, dissolution sites,

Figure 6: Raw 2D image (a) - Two linear profiles containing paste matrix and aggregates: line 1 (b) and line 2 (c)

404
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Table 2: Porosity (%) determined with a 2D method
(Matlab) and a 3D method (ImageJ).
Global scan Local scan
2D method 3D method 2D method 3D method
Top 1.7 1.0 0.8
Middle 1.5 1.2 0.4
Bottom 0.7 0.8 0.5
cracks) is defined as the ratio between the white
pixels and the total number of pixels. The results
obtained for the global and local scans are given in
Table 2. The 2D (Matlab) and 3D (ImageJ) methods
provide similar results. It appears that porosity
is lower at the specimen bottom for all the tested
methods (similar results were obtained by perme-
ability tests). The difference between the global
and local scans can be partially attributed to noise
which is more important for the local scans due to
the presence of matter between the X-ray source
and the studied volume. Moreover the total volume
is lower for the local scans.
Aggregate/paste interface
The X-ray tomography images indicated darker ar-
eas around several aggregates (Figure 11a and 11b).
The optical microscopy observations confirmed the
existence of a different matrix texture around these
aggregate (Figure 11c). Moreover, the large dark
round marks shown in Figure 11b and 11d corre-
spond to large pores covered by mineral crystals.
To determine the hydration products in these
areas, SEM analyses combined with EDS were ap-
plied. Results show clearly the presence of delayed
ettringite (Figure 12). It has a more compact mor-
phology (Figures 13 and 15) and it is particularly lo-
cated in vacuoles and large pores - probably formed
Figure 7: A 2D image with amplified contrast between aggregates and matrix.
during the dissolution stage - (Figure 14) and in the
interfacial zone (Figure 13). Moreover, the interfa-
cial zone is characterized by cracks (Figure 12). The
delayed etttringite located in these cracks tends to
occupy all the free space. All these observations
0.7
0.4
confirm the X-ray tomography observations and
0.7 the fact that the interfaces between paste and ag-
gregate (dark lines around aggregates on the X-ray
tomography images) are filled by delayed ettrin-
gite. Isolating cracking located at the interfacial
area around aggregates can therefore be used to
quantify this product. Unfortunately however in
this study the crack width is inferior to the pixel
size leading to an overestimation of the delayed
ettringite.
IV. Conclusions
The aim of this experimental campaign is to study
the consequences of internal sulphate reactions on
a concrete specimen using a 3D non-destructive
method, the X-ray tomography: cracks, voids due to
ettringite dissolution and the formation of delayed
ettringite in the aggregate/paste interfaces. This
last point is also confirmed by SEM observations.
Aggregates are also isolated from the paste matrix
and a finite element mesh of the specimen at the
mesoscale level useful for finite element calculations
can be created.
The adopted picture resolution does not how-
ever allow quantifying the aggregate/paste inter-
faces despite the use of two different level scanning
methods (global and local). A solution consists
in limiting the dimensions of the specimen. This
technique will be used in a future study on a new
sample with a smaller diameter, directly after the
thermal treatment and for several years.
The adopted picture resolution does not however
405
 Darquennes et al.

Figure 8: 2D images where matrix interfaces (a) and aggregates (b) are isolated.

Figure 9: 2D final image with isolated aggregates.

Figure 10: Coarse image (pixels size = 31.2 x 31.2 µm) from a local scan in the middle of the specimen.

406
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 11: Coarse X-ray tomography images for a global scan (a) zoom on an aggregate (b) observation of the aggregate
by optical microscopy (d) zoom on the interfacial zone (c) zoom on a large pore (3D picture - (e)).

Figure 12: Cracking in the interfacial area and delayed ettringite on a coarse surface.

Figure 13: Delayed ettringite in the interfacial zone on a coarse surface.

407
 Darquennes et al.

Acknowledgement
This work has benefited from the support of
the French "Agence Nationale de la Recherche",
through the "Investissements d’avenir" program un-
der the reference "ANR-10-EQPX-37MATMECA".
The authors would like also to thank B. Smaniotto
for his contribution.

Figure 14: Delayed ettringite in a vacuole on a coarse
surface.
Figure 15: Delayed ettringite on a coarse surface.
allow quantifying the aggregate/paste interfaces
References
Al Shamaa, M. (2012):"Etude du risque de développement d’une réac-
tion sulfatique interne et de ses conséquences dans les bétons de struc-
ture des ouvrages nucléaires." PhD Thesis, Université Paris-Est. (In
French)
Al Shamaa, M., Lavaud, S., Divet, L., Nahas,G., and Torrenti, J.-M.
(2014):"Coupling between mechanical and transfer properties and expan-
sion due to DEF in a concrete of a nuclear power plant." Nuclear Engi-
neering and Design, Vol. 266, 70-77.
Bouzabata, H., Multon, S., Sellier, A., and Houari, H. (2012): "Effect
of restraint on expansion due to delayed ettringite formation. " Cement
and Concrete Research, Vol.42, 1024-1031.
Divet, L., and Randriambololona R. (1998): "Delayed ettringite forma-
tion: the effect of temperature and basicity on the interaction of sulphate
and C-S-H Phase." Cement and Concrete Research, Vol. 28, 357-363.
Escadeillas, G., Aubert, J.-E., Segerer, M., and William, P. (2007):
"Some factors affecting delayed ettringite formation in heat-cured mor-
tars. " Cement and Concrete Research, Vol.27, 1445-1452.
Pavoine, A., Divet, L., and Fenouillet, S. (2006): " A concrete perfor-
mance test for delayed ettringite formation: Part I optimisation. " Ce-
ment and Concrete Research, Vol.36, 2138-2143.
Pavoine, A., and Divet, L. (2007): "Réactivité d’un béton vis-a-vis
d’une réaction sulfatique interne - Essai de performance : Méthode
d’essai des LPC n 66." Laboratoire Central des Ponts et Chaussées. (In
French)
Tosun, K. (2006): "Effect of SO3 content and fineness on the rate of
delayed ettringite formation in heat cured Portland cement mortars." Ce-
ment and Concrete Composites, Vol. 28, 761-772.

despite the use of two different level scanning meth-

ods (global and local). A solution consists in limit-
ing the dimensions of the specimen. This technique
will be used in a future study on a new sample
with a smaller diameter, directly after the thermal
treatment and for several years. The adopted pic-
ture resolution does not however allow quantifying
the aggregate/paste interfaces despite the use of
two different level scanning methods (global and lo-
cal). A solution consists in limiting the dimensions
of the specimen. This technique will be used in a
future study on a new sample with a smaller di-
ameter, directly after the thermal treatment and for
several years. The adopted picture resolution does
not however allow quantifying the aggregate/paste
interfaces despite the use of two different level scan-
ning methods (global and local). A solution con-
sists in limiting the dimensions of the specimen.
This technique will be used in a future study on a
new sample with a smaller diameter, directly after
the thermal treatment and for several years. The
adopted picture resolution does not however allow
quantifying the aggregate/paste interfaces despite
the use of two different level scanning methods
(global and local). A solution consists in limiting
the dimensions of the specimen. This technique
will be used in a future study on a new sample
with a smaller diameter, directly after the thermal
treatment and for several years.

408
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Petrographic assessment of particulate

materials: some statistical considerations
Maarten A.T.M. Broekmans⇤,a , Isabel Fernandesb
a Department of Mineral Resources, Geological Survey of Norway – NGU, PO Box 6315 Sluppen, N 7491 TRONDHEIM, Norway;
b Departament of Geology, Faculty of Science of the University of Lisboa, Edifício C6. Campo Grande, 1749-016 LISBOA, Portugal
⇤ maarten.broekmans@ngu.no

Abstract

Reliable assessment of particulate materials for building and construction is subject to dispute. Some common and
hardy misunderstandings coexist with confusion about sample size and representativity. Statistics determines the
minimum number of particles to be assessed to achieve a certain statistical error (analytical, instrumental, and operator
error come in addition). This paper outlines analytical procedures as common in aggregate assessment, and makes a
comparison with the metals mining industry.

Keywords: petrographic assessment, particulate materials, concrete aggregate, counting statistics

I. Introduction tion of (usually fine-grained) lithologies in coarse

In daily concrete construction and engineering rou-
tine, ‘microscopy applied to building materials’ is
largely synonymous with ‘aggregate assessment’.
For one, this reflects the enormous annual produc-
tion volume of concrete (⇠24G tonne/annum glob-
ally, or about 10km3 ), for another, the recognition
of deleterious alkali-aggregate reaction AAR as a
worldwide damage mechanism (e.g. Broekmans
2012, Fernandes & Broekmans 2013). Nowadays,
civic and infrastructural structures are normally
designed to hold at least half a century, whereas
critical and/or costly structures like coastal defense
works and power installations (e.g. hydropower,
nuclear) may be designed to last for two centuries
or even longer, in both cases anticipating for a lim-
ited amount of degradation from ‘natural decay’ as
well as ‘wear and tear’ from normal use. By con-
trast, distress from damages like e.g. unforeseen
AAR (and disregarding calamities) cannot be antici-
pated for, and strictly reduces a structure’s lifetime
expectance.
One strategy proven effective to minimize the
chance for deleterious AAR in forthcoming struc-
tures is petrographic assessment of the aggregate
material, prior to mixing and emplacement of the
concrete it is to be used in. Petrographic assessment
thus comprises the determination of relative propor-
tions in weight or volume percent of the composing
lithologies (including potentially deleterious), by
particle counting of a collected or submitted sample.
Typically, gravel fractions are assessed macroscop-
ically by manual sorting, sand fractions using a
petrographic microscope in thin section. Identifica-
particles may require preparation of additional thin
sections, and/or supplementary methods like XRD,
SEM-EDS/EPMA, or other. The analytical protocol
is elaborated in the forthcoming RILEM guideline
AAR 1.1 Petrographic Method (2015), bringing the
preceding AAR 1 (2003) guideline up-to-date. The
procedures adopted by (inter ) national standards
are very similar, e.g. ASTM C295 03 (2003), BS7943
99 (1999), CUR Recommendation 89 08 (2008) (also
see Jensen & Sibbick 2001). Regardless which petro-
graphic standard or guideline is adhered to validate
concrete aggregate, most rely primarily on litholog-
ical classification and application of standardized
nomenclature (see e.g. Broekmans et al 2009). The
forthcoming petrographic atlas in RILEM Guideline
AAR 1.2 (Fernandes et al 2015) intends to facili-
tate correct identification of potentially deleterious
rock species. Subsequently, relative proportions of
constituent lithologies are determined by particle
counting in coarse materials (followed by weigh-
ing and commonly conversion to volume percent),
in fine materials by point counting in thin section,
typically around one thousand points (expressed in
volume percent).
Determination of modal contents by parti-
cle/point counting is strenuous for the analyst, de-
manding undiluted attention for the duration of the
entire assessment, possibly while being fixed in an
ergonomically less-than-optimal posture. The type
of work itself is tedious, and arduous if done for
prolonged time. At consultancy businesses, a typi-
cal assessment including writing up observations
and tabulating data takes one half day by quali-

409
 Broekmans & Fernandes
fied/highly educated personnel (sample/specimen
preparation come in addition), and the commer-
cial market price is to a large extent determined
by the hours spent. To remain price-competitive,
possibilities to extend assessment of a single sample
to count more particles/points are strictly limited.
Consequently, the size of a sample that will be as-
sessed as a whole is in practice limited too. It is
therefore interesting to take a closer look at the re-
lationship between a one thousand particle/point
assessment versus sample size.
II. Pierre Gy’s Theory of Sampling -
ToS
The theory of sampling was devised by Pierre M.
Gy in the 1970’s. Originally in French (Gy 1975),
it was translated into English (Gy 1979), and has
since then found pervasive acceptance and appli-
cation throughout the metals mining industry, and
elsewhere (see e.g. Minkkinen 2004). The Histori-
cal Summary (Gy 1982, p3) quotes a statement by
Weed which can be traced back to the Glossary in
The Mines Handbook from (1922, p17): “Sampling.
[. . . ] Honest sampling requires good judgment and
practical experience.”. Referring to the latter aspect,
the books by Gy elaborates extensively on rigor-
ous and reliable procedures for the extraction and
preparation of representative (sub ) samples. Sam-
pling might seem a sinecure, but reducing a 400,000
ton iron ore port stockpile to a ⇠35g sample for
analysis requires both suitable hardware and rigor-
ous procedures (e.g. Petersen et al 2004), to arrive
within 0.2wt% of the quarry sample.
Far more importantly, Gy’s ToS also developed a
mathematical-statistical underpinning (which was
substantially extended in Gy 1992) for the minimum
size of samples from solid and/or particulate ma-
terials to be considered statistically representative.
Thus, for truly representative samples, variation
between different samples extracted from one and
the same bulk parent remains within predefined
limits. This intrinsically relies on repeatability of
the (extraction) procedure as well as (analytical)
reproducibility (Howarth 2000).
As an example from real life, the port sample of a
bulk batch of gold ore concentrate was found to con-
tain only 19ppm Au, one-fourth of the quarry sam-
ple at 72ppm (parts per million – microgram per
gram). By detailed assessment using XRD, QEM-
SCAN and supplementary techniques, the differ-
ence could eventually be attributed to gravity segre-
gation during long-distance rail transport (while ex-
cluding ‘salting’ of the bulk concentrate with lower
grade material). After remixing the concentrate
at the port, the original Au-content was again ‘re-
stored’ within analytical error (Lombard & Youlton
410
2014). The findings and conclusions in above (much
excerpted and abridged) example totally relied on
the undisputed representativity of the quarry sam-
ple, extracted and prepared according to the princi-
ples defined in the ToS. A large number of similar
examples and case studies on a variety of minerals
in diverse situations have been published in jour-
nals and reported at conferences, and invariably
arrive at the conclusion that a) sample representa-
tivity according to the ToS is quintessential for a
reliable result, but b) yet and despite is commonly
ignored (see e.g. the January 2004 Special Issue
titled “50 years of Pierre Gy’s Theory of Sampling
– Proceedings of the First World Conference on
Sampling and Blending etc” in the journal Chemo-
metrics & Intelligent Laboratory Systems [sic!]).
Unfortunately, ‘sample size and representativity’
as a concept only enjoys very limited popularity.
At first glance, the mathematical background of the
statistics involved uses unfamiliar notations and
jargon and probably discourages most geoscientists
save math-aficionados. In addition, the interactions
between various aspects of sampling and sample
size are intricate and not instantly clear, and mu-
tual effects are complex and occasionally counter-
intuitive. Even when the book by Gy (1982, 1998)
are well written, the extensive deliberations and
mathematical derivations still demand appreciable
time to sink in and comprehend.
As an author, Gy was well aware of the limited
popularity of the subject of statistically correct sam-
pling for analysis, a fact illustrated rather sadly in
Figure 1 (also see Tkachev 2014). Nevertheless, at
the end of Chapter 28, Gy concludes: “[. . . ] Sam-
pling and sample preparation should be on an equal
footing with assaying and given exactly the same
consideration. Sampling has become a science and
must be treated as such.” (Gy 1982, p350). Yet and
despite, statistics is generally perceived as the dis-
cipline capable of proving a given hypothesis true,
but make the opposite –skillfully phrased– sound
equally plausible: “There’s lies, damn lies, and
statistics.”, which has been a handy marketing tool
in many cases and for a long time (see e.g. Huff
1954).
III. Modal assessment of solid rock
materials
As referred to in the introduction of Howarth (1998,
p594), Delesse (1848, p379) states: “To study a rock
completely, it is not sufficient to know the differ-
ent minerals that compose it, one should also de-
termine the proportion of each of them; the so-
lution of that question presents some difficulties,
that cannot be resolved directly by comparing their
densities, that is to say if the rock encloses more
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
Figure 1: Popularity of ToS in practice: the book by Gy was published in 1998, and also purchased in 1998 according
to the librarian’s pencil markings, but never loaned for 15 years until it was deaccessed in 2013.
than two minerals.” [translation MATMB], which
is followed by a procedure for the modal analy-
sis of rocks considering their relative surface areas
(reportedly, previous attempts on modal analysis
by weight using liberated mineral grains were un-
successful). It is interesting to know that the first
rock thin section ever was created one year later
in 1849, by metallurgist/meteoriticist Sir Henry
Clifton Sorby (in: Humphries 1992). Since then,
many researchers have been contributing to modal
thin section analysis, notably Rosiwal (best known
from his abrasion hardness scale) until ⇠1900 and
Chayes and co-workers in the mid of the 20th cen-
tury (see e.g. Howarth 2004). The comprehensive
historical overview by Hykšova et al (2012) extends
into stereology as used in pathology and microbiol-
ogy, and sports many great names from science.
The graph from Van der Plas & Tobi (1965) to
judge reliability of statistical point counting data
from thin section analysis has been applied widely
(at the time of this writing, Google Scholar returns
470 citations for the paper), but has been shown to
be inaccurate to species of interest with a very low
or very high abundance (i.e. <10vol%, or >90vol%)
and getting worse towards the extremes (see the
discussion in e.g. Neilson and Brockman 1976). The
landmark paper by Howarth (1998) demonstrates
this inaccuracy through thorough mathematical-
statistical analysis, and in doing so, also disproves
the undeservedly popular allegation that “Perhaps
300 [counts] will suffice for most ordinary work”
(ipse dixit Dryden 1931, p237).
His contribution also presents to new graphs that
since their publication have been widely adopted,
also by RILEM AAR 1.1 (2015). While these dia-
grams (i.e. Figure 2AB on p596-597) do provide cor-
rect statistical errors also for low/high-abundant
species, they are a bit challenging to read. They
show that analytical error (= statistical uncertainty,
i.e. the chance in percent that the true result is
different from the one actually found) increases
towards low/high extreme abundances, and in-
creases towards lower overall point/particle counts.
Thus, analytical error is lowest for abundances in
the range 20-80vol% counted in a large popula-
tion of points/particles. The remainder of this pa-
per relies extensively on Howarth (1998) and the
graphs therein as mentioned above, which have
been adopted by RILEM AAR 1.1 (2015).
IV. Grain versus particle
In the mineral industry, sample representativity
is totally quintessential for the correct evaluation
and validation of a mineral stockpile (see example
from Lombard & Youlton 2014 elaborated above),
or indeed an entire deposit. In industrial rock pro-
cessing, raw material is comminuted down to ap-
proximately the grain size of the (metallic or non-
metallic) minerals of interest, to minimize presence
of non-recoverable composite particles and max-
imize liberation, as only fully liberated mineral
grains are useful for further processing. Imagine
a platinum ore: Pt-contents are typically on the
order of 2ppm and sometimes less, and the de-
posit can only be operated economically profitable
if most of the Pt can be recovered from the ore
(see e.g. Dill 2010). Thus, the raw ore is commin-
uted (first crushed, ultimately pulverized in up to
five successive steps) to free all platinum-minerals
from unwanted host that would impede further
concentration or refinement. The mineral process-
ing industry invests a lot of money in equipment
to obtain a powder product consisting of individ-
411
 Broekmans & Fernandes
ual, mono-mineralic grains, as well as in QA/QC
at multiple stages from raw material to final prod-
uct. Consequently, literature describing experiences
from the mineral industry regarding sample rep-
resentativity and counting statistics thus refers to
(liberated) ‘mineral grains’.
By contrast, assessment of the AAR-potential of
coarse aggregate (cf. RILEM AAR 1.1 2015) relies
on lithological identification of constituent litholo-
gies based upon mineral content, structure, and
texture/fabric. Next, relative volume proportions
are determined per lithology, essentially a modified
version of the Gazzi-Dickinson method originally
developed for provenance determination (see e.g.
Ingersoll et al 1984, and references therein). Fi-
nally, the AAR-potential of bulk aggregate material
is classified according to local experience (RILEM
AAR 1.1 2015). Thus, the entities counted in pet-
rographic aggregate assessment comprise whole
particles (whether crushed or of natural origin), as
opposed to liberated individual mineral grains as
in the mineral industry.
Even when ‘grain’ and ‘particle’ are commonly
used interchangeably, both terms are in fact not syn-
onymous. The difference is not just trivial: a par-
ticle may be composed of multiple mineral grains,
whereas the opposite makes no sense – a single min-
eral grain cannot consist of multiple particles, or
else it would simply be a ‘composite particle’. Only
if particle size approaches mineral grain size, then
‘particle’ becomes identical with ‘grain’, and the
distinction may appear redundant. However, pet-
rographic aggregate assessment is explicitly based
on particle counting independent of size fraction,
even when some of the particles assessed may be
identical with grains.
V. Routine petrographic assessment
of concrete aggregate
Procedures are described in detail in RILEM AAR
1.1 (2015). In brief, a representative sample of
coarse aggregate shall be separated into its con-
stituent lithologies by handpicking, and relative
proportions are then determined by weighing (can
be converted into volume fractions using specific
weight). Thin sections may be prepared from indi-
vidual coarse particles for identification purposes,
typically of fine-grained lithologies. Representative
sample sizes are given in Table 1 below, in which
the specified sample weights are recalculated as
numbers of granitic or arkosic cubes of the nominal
size on edge, and spheres of nominal size in diame-
ter respectively, to give an idea on the number of
particles accessible for assessment (also see Broek-
mans 2006). At 50mm and 40mm particle sizes,
samples cf. RILEM AAR 1.1 (2015) are larger and
412
hence contain more particles than cf. ASTM D75-
03, whereas from 20mm down, ASTM-compliant
samples tend to contain more particles.
Fine aggregate is not separated, but instead as-
sessed in thin section by point counting in a petro-
graphic microscope. Commonly, two thin sections
are prepared for the fraction 2-4mm, one single thin
section for the fraction 1-2mm (optional 0.063-1mm,
or alternatively one single thin section 0.063-2mm –
finest material can be analyzed by XRD after addi-
tional milling). The size of a standard petrographic
thin section is ⇠26⇥48mm (⇠1”⇥1 7/8”, fits stan-
dard holders by many instrument manufacturers),
of which ⇠24⇥44mm represents specimen surface
(here rounded to 1000mm2 ). The number of parti-
cles that fit onto one single thin section is strictly
limited by their size, and amounts to about 150 par-
ticles 2-4mm, and about 600 for 1-2mm. To ensure
that ‘all particles’ are involved in the assessment as
well as to compensate for the ‘tomato salad prob-
lem’ (term originally coined by Goldsmith 1967),
the number of counts per thin section is set so as
to exceed the number of particles contained, usu-
ally by an arbitrary factor of ⇠1.5⇥ or thereabout
– always assuring that the whole specimen area is
covered.
Regardless the total number of thin sections
and/or points counted, statistical error remains
uniquely determined by the number of particles
actually assessed (in addition to abundance of
‘species of interest’). Thus, errors can be deter-
mined for numbers of particles (not counts) and
the observed volume percentage (abundance) of a
given ‘species of interest’, using the graph in Figure
2A of Howarth (1998). The en-echelon curves define
absolute errors in volume percent, which can be
recalculated as relative error. For instance, an abso-
lute error of 1.5vol% represents a 7.5%-relative error
at 20vol% abundance, which quadruples to 30%-
relative error at 5vol% abundance. At 2vol% abun-
dance (as allowed for chert/flint in Dutch guideline
CUR-Recommendation 98-08; similar values apply
in UK and DK), 1.5vol% absolute error represents
a 75%-relative error – the true abundance is some-
where in the range 0.5-3.5vol%. Table 2 summarizes
thus calculated relative errors, for a range of parti-
cle numbers and abundances at 20-10-5vol%.
Obviously, relative error is largest (64%) for low
abundance (5vol%) at low particle count (150, one
single thin section only). In an assessment of two
thin sections 2-4mm aggregate comprising ⇠300
individual particles together, a species of interest
at 5vol% abundance achieves a 44% relative error.
Abundances of that magnitude are quite common
in polymict aggregate materials, which are again
widely applied in concrete. The only way to re-
duce (both relative and absolute) error is to include
more particles in the assessment. This is simply
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
and solely determined by the mathematical back-
ground of counting statistics for which no shortcuts
exist, hereunder ‘extensive operator experience’ as
well as ‘petrographic wizardry’ (note: Dryden 1931,
as debunked in Howarth 1998). The easiest way
to resolve this is not by counting more points in
available thin sections, but instead by preparing
and assessing additional sections (Howarth and
French 1998, p174). Table 3 tabulates the number of
standard-sized (=1000mm2 ) thin sections needed to
contain a certain number of particles 2-4mm or 1-
2mm, to achieve a predefined relative error, at three
selected abundances (20-10-5vol%). If a large error
is acceptable (as e.g. for non-critical concrete appli-
cations), then assessment of one or two thin sections
is sufficient. However, the number of thin sections
rapidly increases towards lower abundances and
towards lower error. Thus, as an example, assess-
ment of 2.0vol% chert (+flint+chalcedony+opaline
silica, cf. CUR-Recommendation 89-08 2008) with
10%-relative error requires counting 15000 particles
(extrapolated from Figure 2A in Howarth 1998),
for 2-4mm aggregate comprising one hundred thin
sections, or ‘only’ about twenty-five if for 1-2mm
aggregate. While such numbers of thin sections
are practically unfeasible and unrealistic, they can
to a large extent be circumnavigated by additional
comminution.
VI. Routine assessment of a
hypothetical noble metal ore
Many different procedures and protocols exist, but
all are essentially based on the same principles.
Consider a metallic ore containing ⇠2ppm Au
(=0.0002wt%, as determined by chemical assay),
present as isolated massive rounded grains of na-
tive gold of 50µm size hosted in a quartzitic matrix
with grain size ⇠1mm and density 2.65kg·L 1 (as
determined by reflected light ore microscopy on
selected samples). To assess this amount with 10%-
relative error (i.e. 0.0002±0.00002wt%), a represen-
tative sample must contain at least on the order of
Table 1: Minimum sample masses in kg, for particulate materials with maximum particle sizes in mm. Numbers
of cubes/spheres calculated assuming rock density of 2.65kg·L 1 (as for e.g. granite or arkose), rounded to nearest
integer.
RILEM AAR 1.1-15, Table 1
max. size min. mass number of granite...
[mm] [kg] cubes spheres
50 200 604 1 154
40 100 590 1 127
20 25 1 179 2 152
10 10 3 774 7 208
4 5 29 481 56 305
107 (=10,000,000, i.e. ten million) individual grains
of the ore, according to mathematical statistics elab-
orated by Gy (1983, 1998). With ⇠1mm grain size,
one liter of sample material contains 106 grains,
so that a sample of at least 10L (>26.5kg) may be
considered as representative. The entire sample is
comminuted to liberate all the gold grains present,
as well as to enable reduction into subsamples of a
size better suitable for assessment while preserving
representativity, i.e. maintaining total number of
particles to be assessed. Subsamples are extracted
from the parent bulk using rigorous procedures
and suitable equipment (e.g. an automated sam-
ple divider). The ultimate subsample for specimen
preparation and analysis consists of a carefully pre-
pared and weighed amount of fine powder, still
containing 107 mineral grains as determined by
statistics but now of finer size.
After specimen preparation (see e.g. Kwitko-
Ribeiro 2012), liberated mineral grains are individ-
ually identified in a specially adapted SEM-EDS
instrument (e.g. QEMSCAN – Quantitative Elec-
tron Microscope SCAN, MLA – Mineral Liberation
Analyser, now both owned by FEI), capable of ana-
lyzing up to one thousand grains per second (e.g.
Fandrich et al 2007, Grauch et al 2008). Using highly
specialized proprietary software, minerals are iden-
tified by combining signals from backscatter and
EDS detectors with image analysis and grain mor-
phology (e.g. Pirard and Sardini 2010), and modal
proportions/abundances are determined. Uniden-
tifiable grains (typically composite particles and
trace minerals) are automatically recognized and
kept separate for manual classification. Depending
on fineness of the sample powder, one standard
sized thin section may contain up to half a million
of individual grains, which at one thousand grains
per second, takes a mere five hundred seconds to
analyze.
Both the SEM instrument and the assessment
process are highly automated and require mini-
mal operator effort. Back in the 1970’s, the first
automated mineralogy instruments relied on opti-
ASTM D75-03, Table 1
maxsize min. mass number of granite...
[mm] [kg] cubes spheres
50 100 302 577
37.5 75 537 1 026
19.0 25 1 375 2 626
9.5 10 4 401 8 405
4.75 10 35 210 67 245
413
 Broekmans & Fernandes

Table 2: Examples of relative errors after assessment of 4, 2 or 1 thin section(s) for ‘species of interest’ at abundance
20-10-5vol%, for size fractions 2-4mm and 1-2mm.

total number assessed of %-relative error at abundance

thin sections particles counts 20vol% 10vol% 5vol%
size fraction 2-4mm, ⇠150 particles/⇠250 counts per section, 2.00mm grid

4 600 1 000 16 24 32
2 300 500 22 32 44
1 150 250 32 44 64

size fraction 1-2mm, ⇠600 particles/⇠900 counts per section, 0.75mm grid

4 2 400 3 600 8 12 16
2 1 200 1 800 12 16 24
1 600 900 16 24 32

cal microscopes, which were gradually replaced by operator must be ‘educated’.

SEMs with up to four EDS detectors on a single col-
umn. Thanks to ever increasing computing power
(according to Moore’s Law, doubling every eighteen
months), the time needed to process acquired data
decreases continuously, while accuracy and resolu-
tion increase (cf. the extensive review by Gu 2013;
also see Gu 2003). Alternatively, certain manufac-
turers offer software packages that turn an existing
(newer model) SEM-EDS instrument into an auto-
mated mineral analyzer. Lately, progress has been
made to extend mere ‘liberated-grain counting’ to-
wards ‘automated identification of lithologies’, to
reconstruct rock types from drill cuttings with large
variation in size and (mineral) composition (e.g.
Haberlah 2012). In addition to offering the neces-
sary speed, automated grain analysis significantly
reduces operator error, though even an automated
Nowadays, SEM-EDS instruments are available
that are the size of an office coffee machine, almost
as convenient to operate, costing roughly the same
as a research-grade optical petrographic microscope.
It is only a matter of time until ‘automated mineral
analyzer’ software applications will be ported to
such machines.
VII. Summary
Reliable assessment of particulate materials is gov-
erned by the mathematical background of statistics.
Total statistical error for a given ‘species of interest’
depends on its relative abundance in the sample,
and the overall number of particles (or grains, but
not points) counted, for which no shortcuts exist in
terms of operator experience or magic sample treat-

Table 3: Examples of minimum numbers of particles / thin sections needed, to achieve a predefined analytical error for
abundance levels 20-10-5vol%.

abundance absolute error true range min. particles min. thin sections needed*
[vol%] [vol%] [vol%] needed 2-4mm 1-2mm
predefined error 50%-relative
20 ±10.0 10.0-30.0 50 1 1
10 ±5.0 5.0-15.0 100 1 1
5 ±2.5 2.5-7.5 200 2 1

predefined error 20%-relative

20 ±4.0 16.0-24.0 350 3 1
10 ±2.0 8.0-12.0 800 6 2
5 ±1.0 4.0-6.0 1 700 12 3

predefined error 10%-relative

20 ±2.0 18.0-22.0 1 500 10 3
10 ±1.0 9.0-11.0 3 200 22 6
5 ±0.5 4.5-5.5 7 000 47 12
*: number of particles per thin section: 2-4mm: ⇠150 / 1-2mm: ⇠600
¶: extrapolated from Fig. 2A in Howarth (1998).

414
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
ment or analysis. It is essential to understand that
the ‘grains’ referred to in literature on assessment
of particulate sample materials in the mineral pro-
cessing industry, in fact comprise ‘particles’ when
applied in a concrete aggregate context. In certain
cases (e.g. ‘chert sensu lato’ in Danish/Dutch/UK
polymict aggregate), the resulting number of parti-
cles to count is inevitably very large, as determined
by the nature of the material to be assessed. How-
ever, these numbers are small compared to the num-
bers common in the metals mining industry, where
fully automated instrumentation and analysis are
the standard. Educating operators and ‘clientele
sensu lato’ that large counts are indeed required and
a worthwhile investment in structural longevity
is challenging, but necessary (Holmes and Robin-
son 2004). Equipment and software are evolving
rapidly, supported by equally rapid developments
in computing power. This offers great opportunities
for automated assessment of particulate materials
used in the building and construction industry in
the near future.
Acknowledgement
The authors would like to thank all who contributed
to the development of AAR 1.1 with ample dis-
cussion and animated review rounds, notably the
members of the RILEM team.
References
ASTM C295-03 (2003): Standard guide for petrographic examination
of aggregates for concrete. American Society for Testing and Materials,
West Conshoshocken/PA: pp8.
ASTM D75-03 (2003): Standard practice for sampling aggregates.
American Society for Testing & Materials, West Conshoshocken/PA:
pp5.
Broekmans, MATM (2006): Sample representativity: effects of size
and preparation on geochemical analysis. In: Fournier, B (editor): Marc-
André Bérubé Symposium on alkali-aggregate reactivity in concrete. 8th
CANMET/ACI International Conference on Recent Advances in Con-
crete Technology, Montréal/CA: 1-19.
Broekmans, MATM, Fernandes, I, and Nixon, P (2009): A global pet-
rographic atlas of alkali-silica reactive rock types: a brief review. In:
Middendorf, B, Just, A, Klein, D, Glaubitt, A, and Simon, J (editors): Pro-
ceedings of the 12th Euroseminar on Microscopy Applied to Building
Materials, Dortmund/DE: 39-50.
BS 7943-99 (1999): Guide to the interpretation of petrographical ex-
aminations for alkali-silica reactions. British Standard, BSI 07. British
Standard Institution, London/UK: pp15+ii.
CUR-Recommendation 89 08 (2008): Measures to prevent concrete
damage by the alkali-silica reaction. Official English translation, 2nd
revised edition. Centre for Civil Engineering Research and Codes,
Gouda/NL: pp 48.
Dill, HG (2010): The “chessboard” classification scheme of mineral
deposits: Mineralogy and geology from aluminium to zirconium. Earth
Science Reviews (100/1-4): 1-420.
Dryden, AL (1931): Accuracy in percentage representation of heavy
mineral frequencies. Proceedings of the National Academy of Sciences –
PNAS (17): 233-238.
Esbensen, KH (2004): 50 years of Pierre Gy’s “Theory of Sampling”
– WCSB1: a tribute. Chemometrics and Intelligent Laboratory Systems
(74/1): 3-6.
Fandrich, R, Gu, Y, Burrows, D, and Moeller, K (2007): Modern SEM-
based mineral liberation analysis. International Journal of Mineral Pro-
cessing (84/1): 310-320.
Fernandes, I, Anjos-Ribeiro, MA, Broekmans, MATM, Sims, I (2015):
AAR-1.2 – Detection of potential alkali-reactivity of aggregates – petro-
graphic atlas. Springer Verlag, Berlin-Heidelberg/DE: in press.
Geelhoed, B (2004): Sampling of particulate materials – a new the-
oretical approach. PhD Thesis, Technical University of Delft – TUD,
Delft/NL: pp200.
Goldsmith, PL (1967): The calculation of true particle size distribu-
tions from the sizes observed in a thin slice. British Journal of Applied
Physics (18): 813-830.
Grauch, RI, Eberl, DD, Butcher, AR, and Botha, PWSK (2008): Quanti-
tative mineralogy of fine grained sedimentary rocks: a preliminary look
at QEMSCAN. Microscopy & Microanalysis (14/2): 532-533.
Gu, Y (2003): Automated scanning electron microscope based mineral
liberation analysis: an introduction to JKMRC/FEI mineral liberation
analyser. Journal of Minerals & Materials Characterization & Engineer-
ing (2/1): 33-41.
Gu, Y (2013): Automated mineralogy – past, present and future.
Keynote lecture, 11th International Congress for Applied Mineralogy
(ICAM), 08 10 July 2014, MianYang/CN.
Gy, PM (1975): Théorie et pratique de l’échantillonnage des matières
morcelées. Editions PG, Cannes/FR: pp595.
Gy, PM (1979): Sampling of particulate materials, theory and practice.
Elsevier Scientific Publishing Company, Amsterdam/NL. Developments
in Geomathematics (4): pp450.
Gy, PM (1982): Sampling of particulate materials, theory and prac-
tice. Elsevier Scientific Publishing Company, Amsterdam/NL. Second
revised edition. Developments in Geomathematics (4): pp431.
Gy, PM (1992): Sampling of heterogeneous and dynamic material sys-
tems: theories of heterogeneity, sampling and homogenizing. Elsevier
Science Publishers, Amsterdam/NL: pp684.
Gy, PM (1998): Sampling for analytical purposes. John Wiley & Sons,
Chichester/NY: pp153.
Haberlah, D (2012): SEM-EDS-based protocol for subsurface drilling
mineral identification and petrological classification. In: Broekmans,
MATM (editor): Proceedings of the 10th International Congress for Ap-
plied Mineralogy (ICAM). Springer Verlag, Heidelberg-Berlin/DE: 265-
273.
Holmes, RJ, and Robinson, GK (2004): Codifying the principles of
sampling into mineral standards. Chemometrics & Intelligent Labora-
tory Systems (74): 231-236.
Howarth, RJ (1998): Improved estimators of uncertainty in propor-
tions, point-counting, and pass-fail test results. American Journal of
Science (298): 594-607.
Howarth, RJ, and French, WJ (1998): A statistical study of aggregate
testing data with respect to engineering judgment. In: Latham, JP (ed-
itor): Advances in aggregates and armourstone evaluation. Geological
Society London, Engineering Geology Special Publications: 169-183.
Howarth, RJ (2000): Estimation of the mean and reproducibility of ag-
gregate compositions based on constituent mass data. Quarterly Journal
of Engineering Geology & Hydrogeology (33): 247-256.
Howarth, RJ (2004): Not “just a petrographer”: the life and work of
Felix Chayes (1916-1993). Earth Sciences History (23/2): 343-364.
Huff, D (1954): How to lie with statistics. WW Norton & Company
Inc, New York: pp142.
Humphries, DW (1992): The preparation of thin sections of rocks,
minerals and ceramics. Royal Microscopical Society, Oxford Science Pub-
lications, Microscopy Handbooks (24): pp 83.
Hykšova, M, Kalousová, A, and Saxl, I (2012): Early history of ge-
ometric probability and stereology. Image Analysis & Stereology (31):
1-16.
Ingersoll, RV, Bulard, TF, Ford, RL, Grimm, JP, Pickle, JP, and Sares,
SW (1984): The effect of grain size on detrital modes: a test of the Gazzi-
Dickinson Point Counting method. Journal of Sedimentary Petrology
(54): 103-116.
Jensen, V, Sibbick, T (2001): RILEM petrographic method: practical
use and comparison with other petrographic methods in use. In: Sta-
matakis, M, Georgali, B, Fragoulis, D, and Toumbakari, EE (editors):
Proceedings of the 8th Euroseminar on Microscopy Applied to Building
Materials, Athens/GR: 579-586.
Kwitko-Ribeiro, R (2012): New sample preparation developments
to minimize mineral segregation in process mineralogy. In: Broek-
mans, MATM (editor): Proceedings of the 10th International Congress
for Applied Mineralogy (ICAM), Trondheim, Norway. Springer Verlag,
Berlin/Heidelberg/DE: 411-417.
Lombard, A, and Youlton, B (2014): Quality control: QEMSCAN from
a different angle. 21st General Meeting of the International Mineralogical
Association – IMA, Johannesburg/SA, Abstract Volume: p310.
Minkkinen, P (2004): Practical applications of sampling theory.
Chemometrics & Intelligent Laboratory Systems (74): 85-94.
Neilson, MJ, and Brockman, GF (1976): The error associated with
point counting. American Mineralogist (62): 1238-1244.
Nielsen, BK, and Thyregod, P (1984): Statistiske usikkerheds-
beregninger på analyser av grus: et samarbejdsprojekt mellem Fred-
ningsstyrelsen og IMSOR. Rapport om råstofkortlægning 2. Institute
of Mathematical Statistics and Operations Research (IMSOR), Technical
University of Denmark, Lyngby. Research Report (1984/27): pp62.
Petersen, L, Dahl, CK, and Esbensen, KH (2004): Representative mass
reduction in sampling – a critical survey of techniques and hardware.
Chemometrics & Intelligent Laboratory Systems (74): 95-114.
415
 Broekmans & Fernandes
Pirard, E, and Sardini, P (2010): Image analysis for advanced char-
acterization of industrial minerals and geomaterials. In: Christidis, GE
(editor): Advances in the characterization of industrial minerals. EMU
Notes in Mineralogy (9): 287-340.
RILEM (2003): AAR-1 – Detection of potential alkali-reactivity of ag-
gregates – petrographic method. Materials & Structures (36/7): 480-496.
RILEM (2015): AAR-1.1 – Detection of potential alkali-reactivity of
aggregates – petrographic method. Materials & Structures, Special Issue:
in press.
416
Tkachev, YA (2014): Study of mathematical statistics by geology stu-
dents: necessary minimum. (In Russian, with English abstract.) Institute
of Geology of Komi Science Center of Ural Branch Russian Academy of
Sciences. Vestnik (239): 24-29.
Van der Plas, L, and Tobi, AC (1965): A chart for judging the reliabil-
ity of point counting results. American Journal of Science (263): 87-90.
Weed, HW (1922): The mines handbook, succeeding ‘The copper
handbook’ founded by Horace J. Stevens 1900, describing the mining
companies of the world. Volume XV, supplementing volumes I-XIV. The
Mines Handbook Co, Tuckahoe/NY: pp2248.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Semi- and full quantitative EDS

microanalysis of chlorine in reinforced
mortars subjected to chloride ingress and
carbonation
Fernando F. Mendonça Filho⇤ , José Pacheco, Oğuzhan Çopuroğlu
Delft University of Technology, Materials & Environment (CiTG), Delft, The Netherlands
⇤ fernando.ni13@yahoo.com

Abstract

Energy dispersive X-ray spectrometry (EDS) is a powerful tool for research studies on building materials. Elemental
quantification in cementitious phases contained in the concrete microstructure can be performed at an excellent spatial
resolution. However, accurate compositional quantification requires a standard-based analysis, whereas currently the
most common are standardless microanalysis. Reasons behind this approach lie on the difficulty of finding appropriate
microanalysis standards in EDS studies. In standard-less analyses, elemental quantification from an EDS spectrum
provides normalized quantification, i.e. an analytical total of 100%. Most EDS detectors cannot provide reliable data for
elements with smaller atomic number than sodium, which results in the stoichiometric determination of oxygen. The
aim of this paper is raise awareness about the importance of a controlled experimental parameters and the use of mineral
standards when performing EDS analysis on cementitious materials. For this purpose, both types of investigation
(with and without standards) are performed simultaneously in order to obtain chloride profiles in six reinforced mortar
specimens subjected to chloride ingress and carbonation. Results illustrate that differences in element concentrations can
be derived by the absence of mineral standards which could lead to inaccurate interpretations of element concentrations
such as chlorine.

Keywords: EDS, chloride profiles, carbonation, reinforced mortar.

I. Introduction carbonation, causes a reduction in the pH of the

paste (Bijen, 2003; Neville, 2011).
Durability of concrete structures is a serious con-
Chloride induced reinforcement corrosion is the
cern around the globe. Reinforcement corrosion
most common cause of corrosion degradation. In
is the predominant degradation mechanism affect-
order to estimate the remaining service life of con-
ing concrete infrastructure. In normal conditions, a
crete infrastructure, assessment of chloride concen-
stable oxide layer forms at the steel surface due to
trations is usually carried out by wet chemical anal-
concrete alkalinity. This oxide layer, also known as
ysis. This method is performed in layers through-
the passive film, is capable of preventing deteriora-
out the concrete cover (chloride profiles). Then,
tion due to reinforcement corrosion. Nonetheless,
chloride concentrations are fit to a transport model
the change of local pH (carbonation) or the ingress
from which predictions on the time necessary for
of chloride ions result in the dissolution of this film
chloride concentrations to be equal to the critical
(Hansson, 1988; Bertolini et al, 2013). Chlorides
chloride content is determined. As the critical con-
dissolved in sea water or de-icing solutions pene-
tent indicates the start of attack in steel, most stan-
trate through the concrete pore network. As they
dards use it as a guidance value for service life
accumulate at the steel surface, the amount of chlo-
predictions, emphasizing the necessity of accurate
ride ions needed for destroying the passive layer is
measurements (EN 1992-1-1, 2004; fib, 2006; CUR,
known as the critical chloride content, which once
2009). Further, chlorine can usually be found in the
reached, initiates localized reinforcement corrosion
paste as a component of Friedel’s salt due to precip-
(Mehta and Monteiro, 2008; Grantham, 2011). Con-
itation or chemically bound to the C-S-H (Glass and
versely, the ingress of carbon dioxide (CO2 ) is re-
Buenfeld, 2000; Reddy et al, 2002), and carbonation
sponsible for consuming the available calcium ions
is known to change it between those two forms by
(Ca+ ) in the pore water and portlandite (Ca(OH )2 )
releasing bound chlorides (Glass and Reddy, 2000).
in a reaction resulting in the precipitation of cal-
cium carbonates (CaCO3 ). This reaction, known as Carbonation induced corrosion is common in

417
 Mendonça Filho et al.
concrete infrastructure with high water/cement ra-
tios and low cover thickness. The assessment of
the depth of carbonation is performed by splitting
concrete cores and spraying the fractured concrete
surface with phenolphtalein, or more accurately by
polarized light microscopy.. Visual inspections on
the penetration of carbonation are possible because
concrete in alkaline conditions, pH higher than 10,
will be pigmented whereas carbonated concrete, pH
lower than 9, will not. Usually, this carbonation in-
duced corrosion proceeds at a slower rate compared
to chloride induced corrosion. The rate at which
the carbonation front proceeds has been studied
in service life prediction models (DuraCrete, 2000;
LIFECON, 2004). However, a combination of chlo-
ride ingress and carbonation presents a more severe
environment for reinforcement corrosion. In fact,
this combination has been considered as the most
common case in practice (Basheer et al., 2001).
Energy Dispersive X-ray Spectroscopy (EDS) is
an increasingly available tool for element quantifi-
cation at high spatial resolutions. Compared to
WDS, EDS provides simultaneous quantification of
elements present in the cement paste (Wilson et al,
2014; Krakowiak et al, 2015). The principle of micro-
analysis through EDS is the possibility to infer infor-
mation about the concentration of elements through
the ratio K between the intensity in the number of
X-ray counts of a sample and a known standard.
As the interaction of the beam with the substance
changes for each material, corrections must be done
to take into account differences in mean atomic
number, differences in absorption of X-rays and
differences in the production of secondary X-rays
or X-ray fluorescence (Çopuroğlu and Einarsson,
2011). Also, the detection of elements with lower
atomic number than sodium is not reliable due to
their poor X-ray efficiency (Newbury, 2002). There-
fore, phases with chemically bound water should
have the amount of oxygen determined by stoi-
chiometry. A fully quantitative analysis is defined
by the consideration of matrix effect corrections
(among others, the most commons are ZAF and
FrZ ) using specific minerals for each element and
accounting for test parameters such as beam cur-
rent, accelerating voltage, type of detector, take off
angle and working distance (Goldstein et al, 2003;
Winter, 2012). The opposite, a semi-quantitative
analysis is possible because most commercial soft-
ware for microanalysis possess internal standards
for the most common elemental peaks, allowing
internal calculations to take place in disregard of
beam current and the matrix corrections mentioned.
This way of analyzing EDS spectra is considered
to be ‘point&shoot ’in this paper. This approach is
very common among concrete researchers. The rea-
son for this may lie in the complex composition and
structure of hydrated cementitious materials, but it
418
Table 1: Cement composition CEM I 52.5R
Oxide wt. %
CaO 63.98
SiO2 19.76
Al2 O3 4.93
Fe2 O3 3.16
K2 O 0.54
Na2 O 0.27
SO3 3.17
MgO 1.92
TiO2 0.28
Mn3 O4 0.11
P2 O5 0.77
Cl 0.04
Total 98.93
is also possible that many researches are conducted
without the awereness of the possibilities, since
the software is typically marketed for qualitative
analysis. Furthermore the greatest challenge to the
cement-based materials microanalysis is the lack of
proper microanalysisIdeal standards which are not
widely available as e.g. rock-forming minerals or
synthetic glasses.
In general, chloride quantification by means of
microanalysis has been published elsewhere (Mori
et al. 2006; Jensen et al. 1996; Jensen et al. 1999;
Jakobsen 2013; Anzalone et al. 2013; Pacheco et
al. 2012; Pacheco et al. 2013). Likewise, the in-
vestigation of carbonation through EDS and WDS
has been reported elsewhere (Hidalgo et al, 2008;
Groves et al, 1990; Groves et al, 1991; Lange et al,
1996; Chen et al, 2008; Liu et al, 2010; Pacheco and
Copuroglu, 2015).
This paper examines the differences between
using semi and fully quantitative microanalysis
within the framework of corrosion of reinforcement
on cementitious materials. For this purpose, mi-
croanalytical results obtained from ‘point&shoot
’settings in the EDS spectra analyses (semi) and
standarised (fully) X-ray spectra are presented. Car-
bonation is interesting in particular for this context,
since the presence of precipitated carbonates (CO3 )
could cause discrepancies in the stoichiometric cal-
culations. Reinforced mortar specimens were fabri-
cated and exposed to chloride penetration. Half of
those specimens were later exposed to accelerated
carbonation in order to study the possible influence
of a combined degradation system, i.e. the release
of bound chlorides. Attention is paid to the cal-
culations of element concentrations made by the
‘automatic ’and standard-based analyses and how
it affects the measurements and its own variability.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
II. Sample preparation
Materials
Mortar cylinders with water/cement ratio of 0.45
were cast with ordinary Portland cement (CEM
I 52,5 R). The chemical composition of the used
Portland cement is shown in Table 1. The specimens
contained standard sand (CEN-NORMSAND DIN
EN 196-1) with a maximum aggregate size of 4 mm
and tap water.
Fabrication of mortar specimens
Six mortar cylinders of 24 mm of diameter per 35
mm of height, each containing a steel bar with 70
mm in length and 8 mm in width were fabricated.
After casting, the cylinders were contained with
plastic and kept in laboratory conditions for 24
hours, then de-moulded and stored in 20 C, >90%
RH (fog) room for 28 days.
Chloride ingress and carbonation
After curing, the specimens were partially im-
mersed in a 3.5wt.% sodium chloride (NaCl) so-
lution for 60 days in laboratory conditions (22 C,
65% RH). During this period the solution was con-
stantly renewed so the concentration would be kept
constant. Then half of the specimens (Cl samples)
were prepared for EDS analysis while the remaining
specimens were subject to accelerated carbonation
for 14 days (Cl + CO2 ). The accelerated carbonation
environment was set to 3% CO2 and 75% relative
humidity. Subsequently, the specimens were pre-
pared for EDS analysis.
Specimen preparation for microanalysis
From each set of specimens two slices were sawn
with an anhydrous cutting lubricant (transversal
Table 2: Mineral standards used in DTSA-ii R Software against oxides provided by automatic settings
Semi quantitative
Element Default oxides
Na Na2 O
Mg MgO
Al Al2 O3
Si SiO2
P Not detected nor quantified
S SO3
K K2 O
Ca CaO
Ti Not detected nor quantified
Mn Not detected nor quantified Willemite
Fe Fe2 O3
Cl Cl2 O ⇤
* Note: Cl2 O is a gas at room temperature.
samples) and one was obtained by the Brazilian
splitting test (longitudinal samples). The transver-
sal samples had diameter of 24 mm with a depth of
6 mm and the longitudinal samples had a geometry
of 24 x 24 x 6 mm. From these samples, polished
sections were prepared. Surface polishing involved
grinding with #320, #500, #800 and #1200 grinding
papers; followed by 6 µm, 3 µm, 1 µm and 0.25 µm
diamond paste polishing. The specimen surfaces
were cleaned by immersing them in ultrasonic bath
with laboratory graded ethanol for 10s between
each grinding and polishing step.
EDS microanalysis
For the microanalysis and electron imaging a
Philips XL30 environmental electron microscope
equipped with EDAX energy dispersive spectrom-
eter (EDS) was used under hi-vacuum chamber
condition. EDS detector was a SUTW (sapphire)
type with a calibrated resolution of 131.2eV at 15kV
accelerating voltage. The take-off angle was 35.3
and a working distance of 10 mm. The polished
specimens to be analyzed were carbon coated in a
Leica EM CED030 carbon evaporator at a thickness
of 10 nm.
A stable beam current of approximately 0.5
nA was regularly measured by means of a pico-
ammeter and a Faraday cup. Under these con-
ditions, a stable beam current aimed at a quartz
crystal read around 2500 cps (counts per second)
at 25% dead time. The X-ray collection time was
set to 100 live-sec. per analysis. Before and after
each spot analysis, the average beam current was
measured and included into the ZAF quantification.
The approximate size of the interaction volume
was reproduced by a Monte Carlo simulation. A
square of approximately 5 µm by size is the field
in which the trajectory of electrons is estimated,
covering approximately 2 x 3 x 3 µm3 , as simulated
Fully quantitative
Mineral Composition
Albite NaAlSi3 O8
Diopside MgCaSi2 O6
Sanidine KAlSi3 O8
Quartz SiO2
Apatite Ca5 ( PO4 )3 F
Anhydrite CaSO4
Biotite K ( Mg, Fe)3 AlSi3 O10 (OH )2
Calcite CaCO3
Rutile TiO2
( Zn, Mn)2 SiO4
Hæmatite Fe2 O3
Scapolite ( Na, Ca)4 [ Al3 Si9 O2 4]Cl
419
 Mendonça Filho et al.
Figure 1: Schematic of EDS samples and analysis positions.
by CASINO R Software V. 2.48. It was chosen to
perform the measurements in C-S-H points due to
its binding capacity and abundance through the
hydrated matrix with respect to chlorides.
The analyses were carried out in three manners
as follows (shown in Figure 1):
(a) Three lines were taken in each transversal sam-
ple. Each line had EDS spectra collected at
every 500 µm from the specimen surface to-
wards the rebar;
(b) Three lines were taken in the Cl longitudinal
sample. Each line had measurements at ev-
ery 1000 µm from one extreme to the other
(crossing the rebar previous position);
(c) One line was taken in the Cl+CO2 longitudinal
sample. The line had measurements at every
500 µm from one extreme to the other (crossing
the rebar previous position).
For displaying puroses, averaging of chlorine con-
centrations of points collected at similar distances
from the reinforcement was carried out. The same
procedure was performed on the longitudinal anal-
yses.
The procedure for element quantification was as
follows: first, an EDS spectrum was collected from
the mortar sample. Then, the raw spectrum was
saved for full quantitative analysis. Full quantita-
tive analyses were processed in DTSA-ii Software
Figure 2: Atomic ratios of spectra taken for analysis.
420
using both stoichiometry and oxygen by difference
methods. Afterwards, semi quantitative analyses
were performed by the native EDS software (EDAX
Pheonix) from which normalised results were ob-
tained under automatic settings. Table 2 presents
the predetermined standards in the automatic set-
tings of the EDS software. In parallel, minerals
employed for fully quantitative analysis are pre-
sented. Apart for scapolite, the rest of minerals
were present in a commercial mineral standard
mount (ASTIMEX MINM25-53)
III. Results
Figure 2 shows Si/Ca atomic ratios plotted against
Al/Ca for all obtained spectra in order to confirm
the phases studied. Winter (2012) states that a per-
fectly pure paste of clinker and water would have
values of Si/Ca ratio for C-S-H between 0.45 and
0.55 and Al/Ca ratio of 0.04 and 0.08. As the re-
search was conducted with CEM I, it is accepted
that a small amount of fine limestone was added,
which broadens the range of results. The transver-
sal samples presented results within the expected
boundaries, represented by the red lines in Figure 3.
In contrast, longitudinal samples displayed a lower
Si/Ca ratio in a number of points, most of them
from the Cl+CO2 sample.
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Figure 3: Averaged weight percentage of chlorine in the interaction volume found with respect to transversal specimen
depth. The red vertical line represents the position of the steel bar.

Figure 3 and Figure 4 presents the concentration
of chlorine by weight percentage in the interac-
tion volume. A lower amount of chlorides can be
observed at the outer edges of the specimens, nor-
mally associated with loss by drying process. Con-
sidering the longitudinal samples, it is observed
an increasing trend from the surface to the vicin-
ity of steel, this is in opposition with the common
belief that concentrations are maximum near the
surface and decrease with increasing depth. This
could be an indication that once chloride ions reach
the reinforcement, a cumulative process start and
local concentrations surpass the ones in materials
surface. While maintaining the general behavior,
the Cl+CO2 samples seems to increase the error
between standard and standardless analysis.
profile values for analysis with standards do not
change much between Cl and Cl+ CO2 results, on
the other hand, values without standard show con-
siderable difference, which can be an indication of
error in the detection. The influence of carbonation
will be further discussed on the discussion section.
It is possible to perceive an underestimation of
chloride content from standardless measurements
with respect to the ones using standard minerals.
This difference is not constant and even presents
few exceptions where higher concentration values
are found. However, it also shows an tendency
against safety, since smaller values lead to inaccu-
rate assumptions of remaining time for service life
predictions.

For transversal samples, the trend of increasing

concentration from surface to steel is only present
in Cl samples, while the Cl+ CO2 samples present
drops in both extremes. It is also noticeable that

421
 Mendonça Filho et al.

Figure 4: Averaged weight percentage of chlorine in the interaction volume found with respect to longitudinal specimen
depth. The red vertical line represents the position of the steel bar.

IV. Discussion to this study.

Atomic ratios
Most of the spectra taken displayed a similar range
of atomic ratios, nonetheless it is noticeable that
the transversal samples possessed a higher Si/Ca
ratio. Longitudinal samples had a decreasing Si/Ca
ratio in the points collected on the interface zone.
It might seem counter intuitive that similar spec-
imens present diferent atomic ratios for the same
experiment, but the data collected from transver-
sal samples never reached the interface with the
reinforcement, while for longitudinal samples, the
focus was on that specifically. Initially the discrep-
ancies were atributed to the possible wall effect near
the steel, which causes a higher w/c ratio that leads
to increased formation of CH. In Hu et al (2014), a
higher amount of CH was one of the main reasons
to the presence of inclusions in C-S-H gel, which
diminished the Si/Ca ratio. However, the research
observed a stronger relation of this phenomenum
for smaller w/c ratios (e. g. 0.32), thus not as likely
Rather, it was reported in Koleva et al (2007)
that the presence of chloride ions, and in particular
corrosion, could change chemical composition of
hydration products, causing a shift in Si/Ca ratios
of C-S-H into values similar to the ones reported
hereby. Although the ratios on both researches
are very similar, there is no straight forward ex-
planation of how the microstructure of the gel is
influenced by the presence of chlorides to increase
the amount of calcium. Yet, this is evident in the
longitudinal samples near the interface with the
steel. Possibly, the presence of Cl- ions can attract
enough Ca+ to reproduce the conditions observed
in Hu et al (2007), creating a more propitious envi-
ronment for a calcium rich C-S-H intermixture. An
explanation for the higher number of points in Cl+
CO2 group is provided below.
Effect of carbonation
As most of the analysis presenting lower Si/Ca ra-
tio were from Cl+ CO2 group, this can be used as

422
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands
an indication that carbonation attacked portlandite
throught the specimen, providing additional free
Ca for the intermixture phases in C-S-H, the rea-
son for it to be evident only in the steel interface
zone probably is that just one of these effects is not
sufficient to change the microstructure of calcium
silicate hydrate. Nonetheless, the combination of
the effect of carbonation with the presence of chlo-
ride ions should be enough. It is interesting to
point out that this is an opposite behaviour to the
observed in Sevelsted and Skibsted (2015), where
the C-S-H studied did not contained chloride ions,
reinfocement or the overall cement products, which
allowed carbonation to be a more straight forward
process. It is also pointed out the lack in literature
concerning simultaneous effect of different degra-
dation processes in Si/Ca ratios of concrete.
One other important mechanism of carbonation
is the diminishing of the capacity of the paste in
binding the chlorides ions, since it represents a
difference in corrosion resistance to any given struc-
ture (Lin et al, 2010; Li and Shao, 2014). Since EDS
analysis take into account any chlorine in the inter-
action volume (bound or free), this process should
take place mainly in the surface of the specimens,
as drops in chloride contents caused by loss during
drying. Conversely, the drop in chlorine due car-
bonation is only observed in standardless analysis,
which will be discussed next.
Use of standards vs standardless analysis
As C-S-H is both, hydrated and porous, it presents a
challenge for elemental quantification, and the high
presence of hydrogen and oxygen intensify the mis-
calculations in a standardless analysis because the
provided analytical total of 100%. Such mistake can
be avoided by the adoption of a standard based
protocol that takes into account the factors men-
tioned previously and allows stoichiometric deter-
mination of elements with smaller atomic number
than sodium. Figures 3 and 4 show a concentration
around 5% in weitgh of interaction volume for all
measures taken using standard based analysis, but
varying chloride contents from 1% to 5% in the
standardless analysis. It is very unlikely that the
values near 1% are correct since the solution used
presented a much higher concentration and enough
time was provided to penetration. Further, since
all other analysis have a convergent result, it points
to inaccuracy of standardless analysis. The under-
esimation of chlorine is most probable due to the
choice of default peaks considering oxides that do
not properly represent C-S-H phases composition
and bulk structure.
For longitudinal samples (Figure 4), a high error
between standard-based and standradless analy-
sis is seen, and it also show a slight increase in
errors from chloride treated samples to chloride
and carbonation treated samples. The clear under-
estimation in standardless analysis (up to 90%) is
of utmost importance, since service life prediction
is greatly dependent of accurate chloride concen-
tration determination (Frederiksen, 2009; Marke-
set, 2009), which can lead to vast underrating of a
structure true state. The increase in error for the
samples subjected to carbonation is most likely due
to the inclusion of carbon and oxygen atoms, since
these elements can not be reliably measured by the
probe, therefore, appearing to present lower con-
centrations. Which is another point of concern, as
real structures are often subjected to carbonation
and chloride rich environments simultaneously and
forensic examination are usually carried regardless
of it.
For transversal samples (Figure 3), the diference
in accuracy between the use of standards or not
isn’t as evident as the one mentioned above. This
only reinforces the random nature of standardless
analysis in the sense that non controled factors (e.g.
ZAF corrections) can or not be a big influence in the
results. Comparing the data regarding carbonation,
one can notice that for analysis with standard min-
erals the values of concentration keep somewhat
constant, which is is agreement with what as previ-
ously discussed. For analysis with default software
peaks, the concentrations show a decrease through-
out the entire depth of the specimen when carbon-
ated, pointing again to a increase in error generated
by the presence of additional oxygen carbon atoms.
Although also present in the longitudinal specimen,
this phenomenum is harder to observe in them be-
cause of the already big diference between standard
based and standardless analysis.
V. Conclusion
Results were used to illustrate differences in ele-
ment concentrations that can be derived by the ab-
sence of microanalysis standards, which could lead
to inaccurate interpretations of concentration peaks
for different elements such as chlorine. By the end
of the work, it is was considered evident the impor-
tance of using standardds for microanalysis. It was
further discussed the influence of positionning of
analysis within samples. Although it seemed to not
be a major parameter regarding the acquisition of
chloride profiles, the results using default software
peaks showed major differences in concentrations
for different orientation of analysis, reinforcing the
conclusion above. It is expected that this will be
a step forward for the use of standard-based mi-
croanalysis in the future of cementitious phases
research.
Additionaly, the influence of carbonation was ob-
served. While it seemed to not be a key parameter
423
 Mendonça Filho et al.

for determination of chloride content, it played a Hu, C., Han, Y., Gao, Y., Zhang, Y., Li, Z. (2014): “Property investi-
gation of calcium–silicate–hydrate (C–S–H) gel in cementitious compos-
useful role in pointing out a weakness of standard- ites.” Materials Characterization, 95, 129-139.
less analysis. Concerning the atomic ratio, it was Jakobsen, U. H. (2013). “Microstructural surface deterioration of con-
crete exposed to seawater; results after 2 years of exposure.” 14th Eu-
concluded that in opposition to what was reported roseminar on Microscopy Applied to Building Materials.
before, it appears that might not be the presence Jensen, O. M., Coats, A. M., and Glasser, F. P. (1996). “Chloride
ingress profiles measured by electron probe micro analysis.” Cement and
of chlorides, but rather the presence of chlorides Concrete Research, 26(11), 1695–1705.
near the reinforcement that changes chemical com- Jensen, O. M., Hansen, P. F., Coats, A. M., and Glasser, F. P. (1999).
“Chloride ingress in cement paste and mortar.” Cement and Concrete
position of C-S-H decreasing the Si/Ca ratio. The Research, 29(9), 1497–1504.
effects of carbonation seem to assist this effect, but Koleva, D.A., Hu, J., Fraaij, A.L.A., van Breugel, K., de Wit, J.H.W.
(2007): “Microstructural analysis of plain and reinforced mortars under
it is still not clear. As this work is part of an on- chloride-induced deterioration.” Cement and Concrete Research, 37(4),
going research, further investigation will be carried 604-617.
Krakowiak, K., Wilson, W., James, S., Musso, S., Ulm, F. (2015): “In-
out. ference of the phase-to-mechanical property link via coupled X-ray spec-
trometry and indentation analysis: Application to cement-based materi-
als.” Cement and Concrete Research, 67, 271-285.
References Lange, L.C., Hills, C.D., Poole, A.B. (1996): “The Effect of Acceler-
ated Carbonation on the Properties of Cement-Solidified Waste Forms.”
Waste Management, 16(8), 757-763.
Anzalone, G., Peterson, K., and Sutter, L. (2013). “Concrete chloride
LIFECON (2004). Life Cycle Management of Concrete Infrastructures
concentration profile collection with an Environmental Scanning Elec-
for Improved Sustainability. TU Munchen.
tron Microscope and X-ray Energy Dispersive Spectrometry.” 35th Inter-
Li, J., Shao, W. (2014): “The effect of chloride binding on the predicted
national Conference on Cement Microscopy. Chicago, IL.
service life of RC pipe piles exposed to marine environments.” Ocean
Bertolini, L., Elsener, B., Pedeferri, P., Redaelli, E. Polder, R. (2013):
Engineering, 88, 55-62.
Corrosion of Steel in Concrete: Prevention, Diagnosis and Repair. John
Lin, G., Liu, Y., Xiang, Z. (2010): “Numerical modeling for predicting
Wiley & Sons.
service life of reinforced concrete structures exposed to chloride environ-
Basheer, L., Kropp, J., & Cleland, D. J. (2001). Assessment of the dura-
ments.” Cement & Concrete Composites, 32(8), 571-579.
bility of concrete from its permeation properties: a review. Construction
Liu, Z., Schutter, G. De, Deng, D., Yu, Z. (2010): “Micro-analysis of
and building materials, 15(2), 93-103.
the role of interfacial transition zone in “salt weathering” on concrete.”
Bijen, J. (2003): Durability of Engineering Structures: Design, repair
Construction and Building Materials, 24, 2052–2059.
and maintenance.Woodhead.
Markeset, G. (2009): “Critical chloride content and its influence on
Chen, Q., Hills, C.D., Yuan, M., Liu, H., Tyrer M. (2008): “Charac-
service life predictions.” Materials and Corrosion, 60(8), 593-596.
terization of carbonated tricalcium silicate and its sorption capacity for
Mehta, P.K., Monteiro, P.J.M. (2008) (Portuguese): Concrete: Mi-
heavy metals: A micron-scale composite adsorbent of active silicate gel
crostructure, Properties and Materials. IBRACON.
and calcite.” Journal of Hazardous Materials, 153, 775-783.
Mori, D., Yamada, K., Hosokawa, Y., and Yamamoto, M. (2006). “Ap-
Copuroglu, O. and Einarsson, G. J. (2011): “Montigny Viaduct Re-
plications of electron probe microanalyzer for measurement of cl concen-
visited: Microstructure of an ASR damaged concrete from 1959.” 33th
tration profile in concrete.” Journal of Advanced Concrete Technology,
International Conference on Cement Microscopy, pp. 1 – 13.
4(3), 369–383.
CUR, (2009) (Dutch): Durability of structural concrete with
Neville, A.M. (2011): Properties of Concrete. Pearson Education
regard to chloride induced reinforcement corrosion - Guideline
Newbury, D.E. (2002). "Barriers to Quantitative Electron Probe X-Ray
for formulating performance requirements, downloadable via
Microanalysis for Low Voltage Scanning Electron Microscopy." Journal of
http://www.curbouweninfra.nl/ CUR Gouda.
Research of the National Institute of Standards and Technology, 107(6),
DuraCrete (2005). General Guidelines for Durability Design and Re-
605 - 619.
design, The European Union – Brite EuRam III.
Pacheco, J. and Copuroglu, O. (2015): Quantitative Energy Dispersive
EN 1992-1-1 (2004) (English): Eurocode 2: Design of concrete struc-
X-ray microalanalysis of chlorine in cement paste. Journal of Materials
tures - Part 1-1: General rules and rules for buildings
in Civil Engineering (Accepted for Publication)
Fib. (2006): Fib Bulletin 34, Model code for service life design, Fib,
Pacheco, J., Çopuroğlu, O. and Polder, R.B. (2013). “Optimisation of
Lausanne.
chloride quantification in cementitious mortar using Energy-Dispersive
Frederiksen, J.M. (2009): “On the need for more precise threshold
X-ray Analysis.” 35th International Conference on Cement Microscopy.
values for chloride initiated corrosion.” Materials and Corrosion, 60(8),
Pacheco, J., Çopuroğlu, O., Savija, B., Schlangen, E., and Polder, R.
597-601.
(2012). “Assessment of critical chloride content in reinforced concrete
Glass, G. and Buenfeld, N. (2000): “The influence of chloride binding
by Energy Dispersive Spectrometry (EDS) revisited.” 3rd International
on the chloride induced corrosion risk in reinforced concrete.” Corrosion
Conference on Concrete Repair, Rehabilitation and Retrofitting.
Science, 42(2), 329–344.
Reddy, B., Glass, G., Lim, P., and Buenfeld, N. (2002): “On the cor-
Glass, G., Reddy, B., Buenfeld, N. (2000): “The participation of bound
rosion risk presented by chloride bound in concrete.” Cement and Con-
chloride in passive film breakdown on steel in concrete.” Corrosion Sci-
crete Composites, 24(1), 1–5.
ence, 42(11), 2013-2021.
Sevelsted, T.F., Skibsted, J. (2015): “Carbonation of C–S–H and
Goldstein, J. I., Newbury, D. E., Joy, D., Lyman, C., Echlin, P., Lifshin,
C–A–S–H samples studied by 1 3C, 2 7Al and 2 9Si MAS NMR spec-
E., Sawyer, L., and Michael, J. (2003): Scanning Electron Microscopy and
troscopy.” Cement and Concrete Research, 71, 56–65.
Xray Microanalysis. Springer.
Wilson, W., Krakowiak, K., Ulm, F. (2014): “Simultaneous assessment
Grantham, M.G. (2011): Concrete Repair: A practical guide. Spoon
of phase chemistry, phase abundance and bulk chemistry with statistical
Press.
electron probe micro-analyses: Application to cement clinkers.” Cement
Groves, G.W., Brough, A., Richardson, I.G., Dobson, C.M. (1991):
and Concrete Research, 55, 35-48.
“Progressive Changes in the Structure of Hardened C3S Cement Pastes
Winter, N.B. (2012): Scanning Electron Microscopy of Cement and
due to Carbonation.” Journal of American Ceramic Society, 74(11), 2891-
Concrete. WHD Microanalysis Consultants Ltd.
2896.
Groves, G.W., Rodway, D.I., Richardson, I.G. (1990): “The carbonation
of hardened cement pastes.” Advances in Cement Research, 3(11), 117-
125.
Hansson, C. M. and Sorensen, B. (1988): “The threshold concentra-
tion of chloride in concrete for the initiation of reinforcement corrosion.”
Corrosion rates of steel in concrete, 3–16.
Hidalgo, A., Domingo, C., Garcia, C., Petit, S., Andrade, C., Alonso,
C. (2008) : “Microstructural changes induced in Portland cement-based
materials due to neutral and supercritical carbonation.” Journal of Mate-
rial Science, 43, 3101–3111.

424
15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

this page is intentionally left blank

425
 15th Euroseminar on Microscopy Applied to Building Materials • 17-19 June 2015 • Delft, The Netherlands

Author Index

Adan, O.C.G. 333 Guy, B. 71 Peterson, K. 97

Amenta, M. 157 Hagelia, P. 387 Pintér, F. 131, 147
Andiç-Çakır, Ö. 215 Hees, R. van 77 Přikryl, R. 191
Apostolodis, G. 343 Hilloulin, B. 171 Ramos, V. 235
Arend, J. 249 Holler, M. 79 Rem, P. 353
Bams, V. 55 Hoorebeke, L.V. 79 Remond, S. 309
Baragona, A. 147 Huinink, H.P. 333 Ribiero, M.A. 207
Barbarulo, R. 379 Jakobsen, U.H. 363 Rooij, M.R. de 259
Ben Haha, M. 39 Jonkers, H.M. 109, 157
Rossen, J.E. 279
Benboudjema, F. 401 Kaasgaard, M. 363
Šachlová, Š. 191
Boehm-Courjault, E. 279 Karagiannis, G. 343
Santos, A.S. 77
Boever, W. de 87 Karatasios, I. 157
Šavija, B. 121
Bonifazi, G. 299, 319 Kempl, J. 287
Serranti, S. 299, 319
Borsoi, G. 77 Kilikoglou, V. 157
Brouard, E. 379 Kuchařová, A. 191 Schlangen, E. 109, 121
Böhm, M. 33 Lapillonne, D. 379 Scrivener, K. 39, 255, 279
Broekmans, M.A.T.M. 409 Laugesen, P. 373 Silva, A.S. 207, 235
Buenfeld, N.R. 269 Le, T. 309 Simon, S. 183
Bugini, R. 139 Le Saout, G. 309 Sulovský, P. 45
Capobianco, G. 299 Leemann, A. 227 Smith, M. 97
Caragacean, L. 25 Lipus, K. 33 Soares, D. 235
Cardenes, V. 61 Loukili, A. 171 Soive, A. 171
Cnudde, J.P. 61 Lotfi, S. 299, 319, 353 Staněk, T. 45
Cnudde, V. 61, 79 Luković, M. 121 Stefanidou, M. 343
Couto, H. 207 Lubelli, B. 77 Stutzman, P. 13
Çopuroğlu, O. 113, 121, 287, Maravelaki, N. 157 Thiery, V. 71
417 Matschei, T. 255 Thijssen, A. 109
Darquennes, A. 401 Matziaris, K. 343 Toit, P. de 25
Dehghan, A. 97 Medeiros, S. 199 Veer, F.A. 113
Diaz, A. 79 Mendoça Filho, F.F. de 259,
Veiga, R. 77
Di Maio, F. 299, 319 417
Velde, O. van der 113
Dunant, C.F. 39 Meng, B. 183
Wang, Y. 401
Durdziński, P.T. 39 Meulenyzer, S. 379
Wagner, E. 299
Eggimann, M. 299 Mota, B. 255
Weber, J. 147
Erich, S.J.F. 333 Nahas, G. 401
Faiyas, A.P.A. 333 Nicaise, D. 55 Werner, D. 183
Fernandes, I. 199, 207, 215, Nijland, T.G. 333 Wetzel, A. 249
235, 409 Noronha, F. 235 Wiktor, V. 109, 165
Folli, L. 139 Nunes, J.C. 199 Wong, H.S. 269
Garcia-Diaz, E. 309 Pacheco, J. 417 Xia, H. 319
Gardei, A. 183 Palin, D. 109 Ye, G. 121
Gosselin, C. 131, 147 Palmieri, R. 319 Yio, M.H. 269
Grondin, F. 171 Pertold, Z. 191 Zacarias, P. 97

426