Vous êtes sur la page 1sur 10

Cement & Concrete Composites 55 (2015) 112–121

Contents lists available at ScienceDirect

Cement & Concrete Composites


journal homepage: www.elsevier.com/locate/cemconcomp

Impact of hemp shiv on cement setting and hardening: Influence of the


extracted components from the aggregates and study of the interfaces
with the inorganic matrix
Youen Diquélou a,b, Etienne Gourlay c,1, Laurent Arnaud c,2, Bernard Kurek a,b,⇑
a
INRA, UMR614 Fractionnement des AgroRessources et Environnement, F-51100 Reims, France
b
Université de Reims Champagne-Ardenne, UMR614 Fractionnement des AgroRessources et Environnement, F-51100 Reims, France
c
Ecole Nationale des Travaux Publics de l’Etat – Université de Lyon, Département Génie Civil et Bâtiment, 3, rue Maurice Audin, F-69518 Vaulx-en-Velin Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history: The need for efficient building insulation with low environmental impact has led to a growing market for
Received 22 January 2014 hemp lightweight concretes. However, a deeper understanding of the interactions between hemp parti-
Accepted 11 September 2014 cles and binders during the setting and hardening phase is needed to meet required industrial specifica-
Available online 20 September 2014
tions. In this paper, the effects of three types of shiv and their corresponding water extracts on cement
setting are compared. Various analyses show that these extracts delay hydration, thereby negatively
Keywords: affecting the mechanical properties of the cement. A comparison of the chemical composition of shiv
Hemp
water extracts highlights the importance of plant quality and the variability of their effects on cement.
Hydration
Retardation
The development of a new methodology, combined with the use of 13C labeled plant material, made it
Mechanical properties possible to closely monitor the physiochemical phenomena occurring at the plant aggregate and cement
Interface matrix interface.
Cement Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction CO2 storage [4]. It has also been shown that HLCs offer various
interesting thermal and acoustic properties [5–9] depending on
Shiv (also called woody hemp core) refers to fragments of the their mix formulations.
central section of dried hemp stalk (Cannabis sativa) (Fig. 1a). It is However, HLCs do have some drawbacks compared to conven-
generally considered as a byproduct of long fiber and seed produc- tional hydraulic concrete. In particular, their measured mechanical
tion and until now has been used mostly as animal bedding or properties are low [9] and slow or weak setting is often observed
mulch [1,2]. While the chemical composition of shiv is comparable onsite. These problems are generally worse when HLC is produced
to that of wood, it has a much lower density due to its porous using a pure hydraulic binder like cement. HLCs are comparable to
microstructure (Fig. 1b). Because it has such low bulk density wood cement concretes but have lower densities (700–1200 kg/m3
(110 kg/m3), shiv is used as a natural aggregate to produce hemp for wood cement concrete). Various studies have reported similar
lightweight concretes (HLC) (Fig. 1c) [2]. HLCs are made with shiv, issues to those above when cement is mixed with wood [10–12] or
water and a binder, with densities ranging from 200 to 650 kg/m3 other lignocellulosic residues like wheat straw [13], oil palm [14]
depending on the mix formulation (binder content). They are used or bagasse [15]. Based on setting monitoring of cement mixed with
for various applications in building construction such as walls and different species of wood, some researchers have suggested indexes
roofing, among others [3]. to assess or predict wood/cement compatibility [16,17]. Predictive
Various types of binders are used in HLCs, such as lime, hydrau- indexes take into account several parameters that are mainly related
lic or pozzolanic binders (e.g., cement or pozzolan). These to the extractives (e.g., pH, extractives content and buffering capac-
materials have small carbon footprints thanks to plant particle ity). In fact, extractives which correspond to easily solubilized mol-
ecules (e.g., monosaccharides, fatty acids or phenolic compounds)
⇑ Corresponding author at: INRA, UMR614 Fractionnement des AgroRessources et are thought to play a major role in the setting and hardening process
Environnement, F-51100 Reims, France. Tel.: +33 3 2677 3593; fax: +33 3 2677 [18,19]. Moreover, the high alkalinity of cement can cause lignocel-
3599.
lulosic compounds to degrade and release byproducts, thereby
E-mail address: bernard.kurek@reims.inra.fr (B. Kurek).
1
Present address: CEREMA, BP9, F-67035 Strasbourg, France. impacting setting even more than simple extractable compounds
2
Present address: ENSAM, F-71250 Cluny, France. do alone [20]. The setting impairment caused by these molecules

http://dx.doi.org/10.1016/j.cemconcomp.2014.09.004
0958-9465/Ó 2014 Elsevier Ltd. All rights reserved.
Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121 113

(extracts and/or degradation byproducts) could be explained by var- binder. This study focuses solely on cement because it is a good
ious phenomena like adsorption on hydrated or non-hydrated model for hydration reactions that are responsible for early
cement particles, ionic complexation or the formation of a thin bar- strength. Additional studies currently underway on cement and
rier around cement grains by precipitation [21]. lime mixes will later provide greater insight into the full setting
To circumvent setting problems when plant aggregates are phenomenon of classic HLC binders.
mixed in cement paste, different strategies have been devised
[22]. For example, lignocellulosic residues can be pretreated by sol- 2. Materials and methods
vent extraction or a coating process. Enhanced hydration and car-
bonation through the addition of lime and CO2 injection has been 2.1. Raw materials
shown to accelerate hardening; similarly, set accelerators can also
be added to cement. Concerning HLCs, some improvements have 2.1.1. Cement
also been made. Good practices have been compiled based on The cement used in this study was Portland cement CEM I 42.5
empirical evidence [3] and systematic mechanical tests have made R-HS (Schwenk). The mineral composition of the cement was
it possible to define acceptable pairs of binder/shiv types that reach determined by X-ray diffraction. Potential phases were calculated
good performance thresholds and hence show relatively good com- using the Rietveld method (Table 1).
patibility. Finally, a projection process has been designed in order
to improve the characteristics of HLCs through increased compac-
2.1.2. Shiv
tion and better water addition control [23]. However, despite all
This study used three kinds of shiv issued from three long fiber
these advances, further progress is necessary for HLCs to meet
production chains. They differ by the cultivar, date and method of
the specifications required by industrial and final customers. For
harvest, and defibration process used.
instance, HLCs must:
– CA Shiv: Hemp grown in France. Plants were harvested at matu-
– Take into account the natural variability of hemp quality.
rity and processed via hammer mill.
– Preserve their low density and insulation capacity.
– CB Shiv: German hemp, dried onsite without cutting and har-
– Be inexpensive and directly applicable onsite.
vested via combine harvester and processed via hammer mill.
– Offer reasonably good and reliable mechanical properties.
– CC Shiv: Hemp grown in the UK. Plants were harvested before
maturity, dew-retted, and processed via hammer mill.
To achieve these objectives, the physicochemical influence of
shiv on the binder needs to be better understood. This study made
These shiv types were chemically analyzed according to the
use of two complementary approaches. First, the impact of water
techniques described in Section 2.2.1; results are shown in Table 2.
extracts on setting time, hydration rates and mechanical perfor-
Only minor differences in initial composition were observed.
mances was investigated; second, the interface between shiv and
cement paste was studied because it constitutes a key factor in
2.2. Chemical analysis
the mechanical performances of concrete.
In Europe, and more specifically in France, most HLC binders
2.2.1. Shiv chemical analysis
used by building professionals are a mix of lime and hydraulic
Prior to analysis, the shiv was extracted in a Soxhlet apparatus
with three successive solvents. The extraction time was 6 h with a
liquid to material ratio of 26.6 ml g1. The three solvents used
a b
Table 1
Parenchyma Potential phase composition as determined using the Riet-
cell veld method [24].

Woody fiber Phases % in mass

100 µm Vessel C3S 43.40


C2S 25.80
C3A 0.50
C4AF 19.80
c CaO 0.30
Gypsum 0.30
Ca(OH)2 0.00
CaCO3 2.40
SiO2 0.10
Anhydrite 3.70
Bassanite 1.60

Table 2
Chemical composition of the three types of shiv (in percent of dry weight). Ash can be
included in two component classes: soxhlet extracts and others.

Components (% of dry weight mass) CA CB CC


Cellulose 47.3 45.6 49.2
10 cm Hemicelluloses 18.3 17.8 21
Klason lignin 21.8 23.3 21.9
Fig. 1. Photograph of raw shiv particles (a); SEM micrograph showing shiv Soxhlet extracts 6 5.1 6.2
microporosity due to the organization of the different types of plant cells (b); Others 6.6 8.1 1.6
photograph of a hemp cement concrete sample (formulated for wall application) (c). Ash content 3.7 2.6 3.5
114 Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121

were: a mixture of toluene/ethanol – 95% (2:1 by v/v), followed by 2.3.3. Hydrates production monitoring
ethanol – 95%, and lastly, ultra pure water. The weight of solvent Using the mixing procedure described above, 300 g of cement
extractable components was calculated from the difference of the was mixed with 150 g of water extract (150 g of distilled water
dry matter mass before and after Soxhlet extraction. The dry mat- for the reference sample) with a W/C = 0.5. The mixture obtained
ter obtained was then analyzed. was then poured into eight plastic molds (V  6 ml) corresponding
A carbohydrate analysis was carried out as follows: 10 mg sam- to eight hydration stopping times (20 min, 1 h, 3 h, 7 h, 1 day (d),
ples were pre-hydrolyzed with 12 M H2SO4 for 2 h at 20 °C. Then, a 3 d, 7 d, 28 d). At each stopping time, the mold was frozen in liquid
full hydrolysis with 1 M H2SO4 for 1 h at 100 °C was applied. The nitrogen and then freeze dried for 48 h. Samples were then finely
released monosaccharides were then separated by high perfor- ground to be analyzed by Fourier Transform Infrared spectroscopy
mance anion-exchange chromatography (Dionex) on a PA1 col- (FT-IR) and Thermogravimetric Analysis (TGA).
umn. Triplicate analysis was completed as described by FT-IR was performed in middle infrared with the ‘‘Nicolet 6700’’
Beaugrand et al. [25]. spectrometer on 14 mm diameter pressed pellets and containing
The acid-insoluble lignin content (Klason lignin), corrected for approximately 300 mg of KBr salt and 5 mg of sample powder.
ash content, was estimated gravimetrically (in triplicate) according TGA was done with the Hi-Res Thermogravimetric TGA
to the procedure described by Effland [26] and modified by 2950 Analyzer (TA Instruments) on approximately 25 mg of sample
Monties [27]. powder. All experiments were conducted from 20 °C up to 800 °C
Ash content was estimated by incinerating 6 g of shiv in a muf- (heating rate of 10 °C/min) under dynamic nitrogen atmosphere.
fle furnace at 500 °C for 3.5 h. The ashes were then weighed (in The thermogram shows three different areas with respect to the
duplicate). chemical species responsible for the weight loss in this temperature
range.
2.2.2. Chemical analysis of water extracts Weight loss occurring between 50 °C and 380 °C is mainly due
The water extracts were prepared by mixing shiv and water to the loss of water from C–S–H hydrates, as well as from alumi-
with a shiv to water weight ratio of 0.1. After 24 h, the extracts nate hydrates. Because several types of hydrates present in an
were filtered (300 lm) and analyzed as described below. The shiv unknown ratio are responsible of this weight loss, it is impossible
was dried at 80 °C for 96 h and weighed. The weight loss was then to individually calculate the amount of each hydrate. For this rea-
calculated. son, results for this range of temperature are expressed directly as
The amount of soluble protein was determined colorimetrically the rate of water weight loss, which remains a valuable indication
in duplicate with the microscale Bradford reagent from Sigma– of the hydration rate.
Aldrich. The standard used was bovine albumin serum. The protein The amount of portlandite (Ca(OH)2) was calculated from the
concentration was based on UV absorbance at 595 nm [28]. loss of water (LH2O) occurring between 380 °C and 500 °C accord-
The amount of phenolic and lignin-like compounds was deter- ing to Eq. (1), where MW represents molecular weight:
mined directly from the liquid extracts by UV spectrometry at CaðOHÞ2 ð%Þ ¼ LH2 Oð%Þ  ðMWðCaðOHÞ2 Þ=MWðH2 OÞÞ ð1Þ
280 nm (in duplicate). The absorption coefficient of 20 for a solu-
tion of 1 g l1 of lignin was determined from previous experiments Similarly, the amount of calcium carbonate (CaCO3) was calcu-
and was used for calculation. lated from the loss of carbon dioxide (LCO2) occurring between
Ash content was estimated by incinerating 15 mg of lyophilized 550 °C and 790 °C according to Eq. (2):
extracts in a muffle furnace at 500 °C for 3.5 h. The ashes were then
CaCO3 ð%Þ ¼ LCO2 ð%Þ  ðMWðCaCO3 Þ=MWðCO2 ÞÞ ð2Þ
weighed (in duplicate).
Carbohydrate analysis was performed on liquid extracts as
described above, but without the pre-hydrolysis step (in triplicate). 2.3.4. Mechanical tests
Using the mixing procedure described above, 1500 g of cement
2.3. Procedures for analyzing cement/water extract mix was mixed with 500 g of water extract (or 500 g of distilled water
for the reference sample) with a W/C = 0.33. The mixture obtained
2.3.1. Mixing procedure was then poured into three prismatic molds (with a cross section of
Cement was poured into a 5-l mixer and water or a solution of 40  40 mm and a length of 160 mm) according to the procedure
water extractives was added. After a 1 min mixing period at low described in [30]. The test specimens were first preserved in their
speed (25 rpm), the mixer was stopped to scrape the paste from molds in a climatic room controlled at 20 °C with 50% relative
the inside walls. The paste was then mixed again at high speed humidity. After 3 days of setting, they were removed from the
(100 rpm) for 1 min 30 s. molds and kept in the same conditions until the test date (7, 14,
28, 60 or 90 days) in the climatic room.
2.3.2. Cement setting monitoring via Vicat test The flexural and compressive strengths of the specimens were
Using the mixing procedure described above, 900 g of cement measured by using a universal hydraulic servo-controlled com-
was mixed with 150 g of water extract and 150 g of distilled water pressive testing machine (INSTRON 1273) and suitable devices
(300 g of distilled water for the reference sample); the water (W) conforming to the requirements of [30].
to cement (C) weight ratio was W/C = 0.33. The mixture obtained
was then poured into three truncated cone-shaped molds 2.4. Procedure for analyzing cement/plant aggregate interface
(Height = 40 mm). Set monitoring was carried out according to
NF EN 196-3 standard [29] with a manual Vicat apparatus (no. 2.4.1. Pellet inhibition capacity calculation
23 – Maurice Perrier & Cie) consisting of a 1.13 mm diameter nee- To observe the interface between the shiv and cement paste, we
dle fixed on a 300 g moveable rod. Three penetration measure- developed a new methodology: a shiv particle (or pellets of shiv
ments per mold were taken every 30 min by recording the powder) was placed on a 20 cm square glass plate, to which a bot-
needle penetration into the paste. In this test, when the material tomless mold was attached and filled with cement paste (W/
is fluid, the needle passes through the total height of the sample C = 0.5). Once the cement paste has set, the device was turned
(40 mm), whereas when the paste sets and hardens, the depth of upside down to observe the interface between the aggregate and
needle penetration is reduced. mineral matrix through the glass (see Fig. 10).
Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121 115

Pellets were made by compressing 200 mg of plant powder extracts is relevant because the components dissolved from the shiv
(200 lm ground) at 0.64 MPa of pressure for 3 min. The pellets are directly supplied to the cement together with the hydrating reac-
obtained had a diameter of 14 mm and thickness of 1.8 mm. tant, without another solvent, which could itself interfere with the
To make a quantitative determination of the inhibition capacity setting process. This is an easy way to evaluate the net contribution
of the pellets on the cement setting, three interface imaging of the shiv extracts on modifications to the cement paste. Addition-
devices (see above) were used. Each one contained three pellets ally, the physicochemical interactions taking place at the interface
of each type of shiv (nine pellets per device). After three days of between the plant aggregate and the cement paste were analyzed.
hydration, the glass plates of the devices were scanned and the
area of the halo observed around the pellets was measured using 3.1. Impact of water extracts on cement setting and hardening
IMAGE J software [31] (Fig. 10).
3.1.1. Set monitoring of cement mixed with water extracts
2.4.2. Extractive migration monitoring Fig. 3 shows the set time measurement determined using the
Based on the same principle described in Section 2.4.1, a device Vicat needle for cement mixed with water (reference) and with
was made by placing a labeled flax pellet (made of 200 lm ground shiv water extracts. Initial and final set times of the reference sam-
flax stalk powder) at the center of a 14 mm diameter plastic tube ple (with pure water) occurred at 240 min (needle penetra-
(10 cm long) as shown in Fig. 2. Each side of the tube was then tion = 36 mm) and 400 min (needle penetration = 0.5 mm),
filled with cement paste (freshly mixed with distilled water) and respectively. It can be observed that setting was dramatically
closed with stoppers. After 20 h and 3 d, the device was frozen in delayed when cement was mixed with shiv water extract, and
liquid nitrogen and then opened to be freeze dried for 48 h. Each especially with that from CC shiv, which created a delay of more
resulting cylinder was cut into 1 mm thick slices, starting from than five hours to final setting. Most set retardation appeared to
the pellet and moving towards the outer end of the cement sample. occur during the latency state (dormant period); this could be
Slices were then ground and analyzed by TGA to determine the explained by the effect of the extracts on the dissolution of the ini-
portlandite content, as well as by mass spectrometry to measure tial phases and/or on hydrates nucleation.
the carbon isotopes content as described below. The delays were considerable, even though the shiv proportion
The pellets used here were made of flax powder enriched with (compared to cement) used to produce water extracts was signifi-
carbon-13 isotope (13C). This flax has been homogeneously 13C cantly lower (approximately one-tenth, in mass) than the propor-
labeled by growing it in an atmosphere enriched with 13C carbon tion used in a conventional HLC mix formulation [22].
dioxide, following a similar procedure as described in [32]. By Results given by the Vicat needle measurements clearly show
doing so, labeled carbonaceous extracts that migrated from the that shiv water extracts acted as a strong cement retarding agent.
pellets could be monitored by measuring the local enrichment FT-IR and TGA were then used to verify if the delay could be related
of 13C in the sample, enabling differentiation from the carbon to changes in hydrates proportion.
coming from the cement.
The 13C content of the sample was determined by full combus- 3.1.2. Impact of shiv water extracts on hydrates production
tion of 15 mg of the sample, with further analysis of the isotopes by Cement setting can be monitored using FT-IR spectra analysis
mass spectrometry, using a procedure similar to [32]. After calcu- [33]. During hydration, a gradual shift of the band centered from
lation according to the isotopic dilution principle, the data are
expressed as the mass (in g) of carbon originating from flax per 40
1 kg of sample.
Needle penetration (mm)

35
30
3. Results and discussion
25
20 Water
When samples of the cement-based HLC mixes, such as the one
15 CA
illustrated in Fig. 1c, were manipulated, the material proved to be
CB
friable with a low cohesion between shiv particles and the cement 10
CC
matrix [9]. This indicates that cement paste should have not shown 5
a normal setting and hardening. The following two sections
0
attempt to clarify the phenomena responsible for the observed par- 0 200 400 600 800
ticular physical state of the cement paste. Time (min)
The first section investigates the effect of shiv water extracts on
cement setting and mechanical properties. The effect of the water Fig. 3. Depth of Vicat needle penetration into the cement paste (made with water
or shiv water extracts) as a function of time. The height of the mold was 40 mm.
extracts from the three different shiv types are also compared and
correlated with their chemical compositions. The use of water

Stopper Space filled with Pellet


cement paste

1 cm Slice position
123456 8 10 15 30
(in mm)

Fig. 2. Schematic representation of the device designed for the extract migration monitoring. The double blue arrows indicate the analyzed slices; the numbers underneath
correspond to the distance in mm from the pellet.
116 Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121

920 to 970 cm1 can be observed, which is assigned to the asym- the amount of portlandite produced after 28 days of hydration
metrical SiO4 stretching frequency [33]. It indicates the polymeri- with reference water.
zation of silicate units SiO44 with the formation of the C–S–H Portlandite content at the beginning of setting is only linked to
phase. At the same time, there is an appearance of a band centered portlandite appearance kinetics. Nevertheless, in a second phase, it
at 3640 cm1 due to the formation of portlandite (Ca(OH)2). The is possible for portlandite to react and produce calcium carbonate;
peaks observed at 1421 cm1, 874 cm1 and 713 cm1 are linked the resulting disappearance/consumption of portlandite must
to the calcium carbonate (CaCO3) initially present in the cement therefore be taken into account. This second phenomenon offers
and to calcium carbonate produced due to the reaction of portlan- an explanation for the decreasing trend of portlandite content after
dite with atmospheric carbon dioxide. 7 days observed in the reference and CA mixes. Moreover, the cal-
As shown in Fig. 4, the amount of C–S–H hydrates was clearly cium carbonate production rate between 7 days and 28 days was
lower when cement was mixed with each of the three hemp higher in these samples (Fig. 6). It can be noted that, after 20 min
extracts, as demonstrated by the lower 970/920 cm1 height peak of hydration, calcium carbonate content was lower than the refer-
ratio. The three shiv water extracts can be ranked as follows: ence for the CC sample and even more so for the CA sample. At this
CB < CA < CC, by increasing order of inhibition capacity of C–S–H point of setting, the calcium carbonate measured in the samples
production. The results presented here were obtained after 7 days originated solely from the calcium carbonate initially present in
of hydration, but the lower C–S–H contents (and the ranking of the dry cement. The lower calcium carbonate content could be due
shiv’s effect as reported above) remained observable even after to the lower pH of shiv water extracts (pH 5.5) compared to water
28 days of hydration (data not shown). only, which could lead to the displacement of the calcium carbon-
Samples were also examined through TGA to precisely deter- ate dissociation equilibrium.
mine how much portlandite and calcium carbonate were present Under these conditions, the lower hydrates content (C–S–H and
in the cement. portlandite) after 28 days of hydration indicated an irreversible
The shiv water extracts had a strong impact on portlandite con- effect on the cement hardening process rather than a simple set-
tent when mixed with cement (Fig. 5). Again, CC extracts caused ting delay. However, it cannot be excluded that higher water evap-
the longest delay compared to the reference water, as there was oration yield occurred due to the setting delay, which could lead to
no measurable portlandite after 1 day of hydration. However, the a depletion of water for reaction, and therefore, to a lower final
final portlandite content (after 28 days) was approximately the amount of hydrates.
same for all three types of shiv, but significantly lower (40%) than
3.1.3. Impact of shiv extracts on mechanical properties
The compressive strength of the reference cement paste mixed
1.2 970 cm-1 920 cm-1 with water was 49.8 MPa after 28 days of curing (Fig. 7a), a normal
3740 cm-1 CC
value for a cement grade of 42.5. Shiv water extracts led to a
1.0 CB
significant drop in compressive strength, more pronounced for CC
Absobance (AU)

CA
0.8 Water
(about 25% after each curing period) and less marked for CA and
CB. The lower hydrates content reported in the previous section
0.6 did not appear to have much impact on the compressive strength
of CA and CB. However, for CC, the difference in the chemical com-
0.4 position of the cement paste after 28 days had a more direct effect
on the mechanical properties. It can therefore be assumed that up
0.2
to a certain delay in setting, as shown with CA and CB, there is no
0.0 major effect on the physical integrity of the material, but beyond
3430 2430 1430 430 that delay, as shown with CC, detrimental effects appear.
Wavenumber (cm-1) It can also be observed from Fig. 7b that all samples tested pres-
ent an important gain in flexural strength between 7 and 14 days
Fig. 4. FT-IR spectra of cement mixed with water and with shiv water extracts after of curing. The shiv extracts can then be ranked according to the
7 days of hydration. rate development of these properties as follows: CA > CB > CC.
However, none of the samples formulated with shiv extracts
reached the mechanical characteristics of the reference cement
20 min 1 h 3h 7h 1d 3d 7d 28 d

20 min 1 h 3h 7h 1d 3d 7d 28 d
7
25
6
20
5 Water
Water
% Ca(OH)2

% CaCO3

4 15 CA
CA
CB
3
CB 10 CC
2 CC
5
1

0 0
1 10 100 1000 10000 100000 1 10 100 1000 10000 100000
Time of hydration (min) Time of hydration (min)

Fig. 5. Change in portlandite content in cement mixed with water and with shiv Fig. 6. Change in calcium carbonate content in cement mixed with water and with
water extracts. shiv water extracts.
Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121 117

5
a
Compressive strength (MPa) 55
4.5

50 4

3.5

Extract Yield (%)


Water Others
45 3
CA Ash
2.5
40 CB Protein
2
Phenolic compounds
CC 1.5
35 Sugar
1

30 0.5
1 10 100
0
Time of hydration (days) CA CB CC

Fig. 8. Chemical composition of water extracts for the three types of shiv. The
proportion of each component is represented based on the dry weight (50 g in
500 ml) of shiv used for extraction.
7d 14 d 28 d 60 d 90 d

compounds such as lignin, teracacidin or sequirin-C have been


b 5 described as inhibitors for cement setting [15,36,37], but again,
4.5
this parameter does not seem to be a key factor in this case since
Flexural strength (MPa)

the CB shiv had a higher content of phenolic compounds than CC


Water
4 shiv but exerted a lesser influence on cement hydration.
CA After extensive quantification of the main class of the com-
3.5
pounds derived from plant cell walls, we also observed a deficit
CB
3 of 15–30% (for CC shiv) of the total mass balance of the component
CC constituting the shiv extracts. This could be partly due to the over-
2.5
or under-estimation of their absolute mass due to their low
2 amount (no more than 2% of the total shiv weight). However, this
undefined fraction could also be from a pool of various molecules
1.5 from primary and secondary plant metabolism processes, which
1 10 100
could initially be present in the sap of the hemp shiv vascular sys-
Time of hydration (days) tem, or in some contaminating tissues from the defibration process
Fig. 7. Change in compressive strength (a) and flexural strength (b) of neat cement
(i.e. leaves) or secretory structures (trichomes). Thus, while mole-
mixed with water and with shiv water extracts. cules such as terpenic compounds or fatty and resin acids [38] that
have no significant effect on cement hydration may be present
within this unknown fraction, it is also possible to have bioactive
paste (water; Fig. 7b). Moreover, all of the tested cement pastes molecules such as hormones or alkaloids whose effect on cement
were shown to have almost constant mechanical properties after has not yet been assessed, or organic acids, some of which have
14 days of curing, suggesting that shiv extracts have significant been described as accelerators (e.g., oxalic acid) [39] or retardants
irreversible effects on both short- and long-term setting as com- (e.g., malic acid) [40] of cement setting.
pared to neat cement paste. Finally, in addition to a greater quantity of unidentified compo-
Lastly, using the same basic extraction protocol, the various shiv nents, CC extracts also contained more sugar. It should be empha-
extracts had considerably different effects on cement paste setting. sized that the CC shiv came from ground-retted hemp. As a result
Although the three shiv types exhibited only slight differences in of the retting process, byproducts from microbial activity and
their initial biochemical compositions, major disparities were metabolism, as well as from cell wall degradation, could be
observed in how their water extracts affected the cement (Section extracted.
2.1.2). For these reasons, the chemical composition of water Complementary to the higher quantity of sugars within the
extractives themselves was investigated. extracts, their composition could be important (Fig. 9). In fact,
hemicelluloses constituting monomers like xylose, rhamnose or
3.1.4. Chemical analysis of shiv water extracts arabinose were found in higher proportion in CC extracts, which
Chemical analyses were carried out on water extracts of the was not the case for glucose (cellulose constituent). However, the
three shiv types to investigate which class of molecules might be specific impact of each component remains to be determined, as
responsible for the effects described above and explain the differ- well as their different combinations in the extracts.
ences observed in the three shiv types.
Fig. 8 shows the amount of total extracts based on the dry shiv 3.2. Macroscopic study of the cement/plant interface
weights as well as the proportion of each component. A large part
of the extract is ash, possibly coming from the constitutive plant The first section of this study demonstrated the detrimental
minerals or soil dust (soil residues, fertilizers) that were not effects of hemp water extracts on cement paste setting and hard-
removed during the defibration process [34]. This mineral fraction ening, including setting delay, fewer hydration products and
could be responsible for a slight delay in setting. Indeed, it has been reduced mechanical performances. These findings on hemp con-
demonstrated that zinc, lead or phosphates can impact the setting firm and complete the effects previously described for other plant
and hardening process [35]; however, there is no major difference aggregates [12]. Another key point to manage, in order to achieve
in the ash contents of the three types of shiv. Similarly, phenolic favorable final properties for HLCs, is the quality of the interface
118 Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121

0.80 1.2 970 cm-1 920 cm-1


3740 cm-1
Sample 1
0.70 1.0
Sugar monomers extract Yield (%)

glcA
Sample 2

Absorbance (AU)
0.60 0.8
galA

0.50 man 0.6

0.40 xyl 0.4

0.30 glc 0.2

0.20 gal 0.0


3430 2430 1430 430

0.10 rha+ara Wavenumber (cm-1)

Fig. 11. FT-IR spectra of cement samples taken outside the halo (1) and inside the
0.00
halo (2).
CA CB CC

Fig. 9. Sugar monomer distribution in the water extracts of the three shiv types.
Particles of shiv have an irregular shape, which does not allow
The proportion of each monomer is represented based on the dry weight of shiv
used for extraction. glcA = glucuronic acid; galA = galacturonic acid; man = man- easy quantification of the halo phenomenon. To overcome this
nose; xyl = xylose; glc = glucose; gal = galactose; rha + ara = rhamnose + arabinose. problem, cylindrical pellets were produced by compressing shiv
powder and then analyzed using the same test as for the particle
(Fig. 10b). The same kind of halo can be observed around the pellet
between the aggregate and the cement matrix. Experiments were
in the same proportions and with a similar chemistry pattern (ver-
therefore designed with shiv particles or shiv powder, under con-
ified by FT-IR – data not shown) to those observed around the
ditions similar to onsite settings in applications, to study the main
particle.
phenomena occurring near the aggregates, and which involve the
extraction of molecules from the shiv by the cement paste.
3.2.2. Effect of the type and the washing treatment of shiv on its
3.2.1. Interface observation capacity to inhibit cement hydration
The new device described in Section 2.4.1 made it possible to The test described above allowed the comparison of the hydra-
observe the interface between the shiv and the cement matrix after tion inhibition capacity of the three types of shiv by measuring the
3 days of hydration. Fig. 10a shows a brighter halo of cement area of the non-hydrated halo produced (Table 3). It can be seen
around a shiv particle with a well-defined limit. To study the that CC again presented the broadest effect; because water extracts
cement paste on both sides of this limit, samples were taken (see of CC shiv had a greater effect on cement, it can be suggested that
the arrows in Fig. 10a) and analyzed by FT-IR (Fig. 11). Unlike sam- the inhibition capacity is related to the effect of water extractives.
ple 1, the cement inside the halo of sample 2 did not present peaks To investigate this hypothesis, the hydration inhibition capacity of
corresponding to hydration products (C–S–H at 970 cm1 and pellets made from normal CC powder and pellets made from CC
Ca(OH)2 at 3640 cm1) but still presented a high peak at powder washed beforehand (4 times) have been measured
920 cm1 of the C3S phase. This proves that the cement matrix in (Table 3).
the halo corresponds to non-hydrated cement. This is a very signif- When the powder was previously washed, the pellet displayed
icant result if we consider the limited space (size of the interparticle a non-hydrated cement area of less than half the size of the halo
matrix in HLC) between hemp particles (and therefore the quantity produced by the reference (CC). Thus, water extracts were respon-
of cement) in a conventional hemp concrete (such as the one sible for the major part of the non-hydrated halo observed. How-
illustrated in Fig. 1c), which might then contain a considerable ever, a residual effect still appeared in spite of the previous
proportion of unset cement with poor binding properties. washing of the powder, which could be explained by several
hypotheses.
As a result of its surface chemistry, plant matter itself can have
a b an impact on cement setting, based on the suggestion by Sedan
et al. [41] that hemp fiber polymers could trap calcium from
cement paste. Similarly, it cannot be excluded that plant aggregate
has an effect on the distribution of the reactive water; in fact, due
to its high porosity, shiv can indeed trap this water.
Cement paste has a high alkalinity (pH > 12). Polymers
1 2 constituting the shiv cell wall can thus suffer degradation leading

Table 3
Comparison of the halo area produced after 3 days of hydration by a pellet made from
the three different types of shiv powders and water washed CC powder (mean value
of 9 samples, except water washed CC: mean value of 8).

1cm 1cm Pellet Area of halo (mm2) Standard deviation (%)


CA 172.1 17.1 (9.9)
CB 160.5 8.0 (5.0)
Fig. 10. Photograph taken after 3 days of hydration of cement paste with two kinds CC 178.6 9.2 (5.1)
of plant aggregate inclusion: one particle of shiv (arrows 1 and 2 indicate sampling Water washed CC 86.1 22.3 (25.9)
points for FT-IR analysis) (a); pellet made of shiv powder (b).
Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121 119

to the release of molecules which differ from simple water extracts 9


[42]. This hypothesis is supported by the FT-IR spectras shown in
6h
Fig. 12. When the FT-IR spectras of the pellets before and after 8
12 h
immersion in cement are compared, the disappearance of peaks
is observed. Peaks at 1740 cm1 and 1243 cm1 could be attribut- 7 1d

Proporon of C from flax (g/kg)


able either to pectins or acetyl groups of xylan [43]. Similarly, the 2d
reduction of peaks observed at 1320 cm1 and 1370 cm1 could 6 4d
reflect a decrease in lignin content. It can therefore be assumed 8d
that when plant aggregates are immersed in cement, degradation 5
of pectin or hemicelluloses (and a part of lignin) can occur, releas-
ing products which can spread in the cement matrix and have a 4

deleterious effect on cement setting.


3

3.2.3. Migration monitoring of extracts and effect on cement setting 2


The results of the previous sections showed that extracts were
involved in the non-hydrated cement halo around the plant aggre- 1
gates. As such, the diffusion of those extracts throughout the
cement paste was monitored. To do this, pellets were made with 0
a powder of 13C labeled flax. The use of flax was justified not only 0 5 10 15 20 25 30 35
because it was readily available in the laboratory, but also because Distance from pellet (mm)
its tissue organization is similar to hemp (woody core and long
Fig. 13. Migration pattern of carbonaceous extracts in the cement, followed by the
fiber peripheral). After verifying that the pellets presented the local amount of labeled 13C material as a function of the distance from the pellet
same non-hydrated halo and in a similar proportion to that and time elapsed.
obtained with hemp pellets, it was possible to determine the diffu-
sion of carbonaceous extracts through sampling at an increasing
distance from the pellet (Fig. 13). This diffusion curve of the 12
extracts was then compared to the curve of hydration rate
(Fig. 14) obtained by calculating the proportion of water loss
between 50 °C and 380 °C from the TGA thermogram, which is 10
mainly due to C–S–H dehydration.
During the early stages of hydration, the cement paste remains
8
relatively homogeneous. This leads to a normal diffusion pattern of
Hydraon rate

the extractives after 6 h, characterized by a rapid drop of carbon 6h

concentration. This migration then impacted setting: as can be 12 h


6
observed after 12 h and 1 d, hydration was slowed over 8 mm from 1d
the pellet, creating a gradient of setting in this location. Then, two 2d
situations occurred: (1) near the pellet (0–3 mm), the concentra- 4 4d
tion of extractives was high enough to completely inhibit setting, 8d
and (2) beyond 3 mm, the cement has already produced a mini-
mum setting because of the lower gradient concentration of the 2
extractives. This may diminish the sensitivity of the partly
hydrated cement to the next arriving extractives; furthermore,
the modified rheology of the medium might also reduce the diffu- 0
0 5 10 15 20 25 30 35
sion rate of those extractives. The two combined phenomena could
Distance from pellet (mm)
consequently allow the setting to continue. After 2 d of hydration,
an accumulation pattern of extractives was observed at exactly Fig. 14. Change in hydration as a function of the distance from the pellet and time
4 mm, corresponding to the limit of set and unset cement. Such elapsed.

Pectin or acetyl
1 from xylan

0.9
0.8
Absorbance (AU)

0.7
0.6 Lignin

0.5
0.4
0.3
Before
0.2 immersion
0.1
After
0 Immersion
3400 2400 1400 400
Wavenumber (cm-1)

Fig. 12. Comparison of the FT-IR spectra of shiv powder in pellet form before and after 3 days of immersion in cement.
120 Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121

accumulation of organic material could be explained by a possible between the mineral matrix and the hemp particles. The second
water flow from the unset cement towards the set cement, which, area corresponds to the location where setting was delayed,
due to its lower viscosity and/or porosity, could act as a physical whereas cement in the third area set normally. The formation
barrier to the extractives. The water flow could be due to either a and the progressive change in the chemical composition and char-
water ‘‘pump’’ effect produced by the hydration of the cement dis- acteristics of these areas depend on complex interactions between
tant enough far from the pellet, or by a potential evaporation at the the migration of extracted shiv byproducts and the hydration reac-
edges of the test device which were not perfectly sealed. tions of cement.
From this cross examination, three areas can be noted: Further research aims to study the influence of the shiv and
water extracts used in this paper on setting and hardening of other
– A first area, near the pellet (0–3 mm), where the cement kinds of binders generally used to manufacture hemp lightweight
almost never set at all, caused by excessive extract quantities concretes (air lime, hydraulic lime, etc.). Moreover, different treat-
(irreversible effect or drying effect due to considerable ments of shiv are also being studied to improve the mechanical
delay). properties of hemp lightweight concretes.
– A second area (3–8 mm), where setting was delayed in rela-
tion to the quantity of extracts present during the early stage Acknowledgements
of hydration.
– A third area, beyond 8 mm, where extracts are virtually This work was part of the 2C2E Prebat Program and received
absent, which allowed the cement to set normally. funding from the French Environment and Energy Management
Agency (ADEME) and the French National Research Agency
Altogether, our data strongly supports the hypothesis that the (ANR). The authors would like to thank Daniel Daviller (BCB,
formation of these three areas and their changes over time result France) and Ulrike Peter (Lhoist, Belgium) for the initial cement
from the combination of cement paste setting and extract migra- composition analysis. We are grateful to Sylvie Recous and Olivier
tion. In fact, these two phenomena interact: the more the extracts Delfosse (UMR FARE, Reims) who provided many valuable discus-
migrate, the more setting is delayed, and consequently, the more sions and the elementary analysis on 13C flax. We also extend our
setting is delayed, the longer the extracts are able migrate. This thanks to Younes Sadour for the work done during his Master’s
suggests an interactive competition between these simultaneous internship for the Troyes University of Technology (UTT), France.
phenomena.
References

4. Conclusions [1] Karus M, Vogt D. European hemp industry: cultivation, processing and product
lines. Euphytica 2004;140(1):7–12.
Weak settings have been regularly reported onsite or in labora- [2] Bouloc P. Le chanvre industriel: production et utilisations. Editions France
Agricole; 2006. ISBN: 2-85557-130-8.
tories during the manufacture of hemp lightweight concretes using [3] Construire en Chanvre. Règles professionnelles d’exécution, Editions SEBTP;
cement binder. To analyze and identify the mechanisms responsi- 2012. ISBN: 978-2-35917-046-7.
ble for these phenomena, three types of shiv were studied and the [4] Boutin M-P, Flamin C, Quinton S, Gosse G. Étude des caractéristiques
environnementales du chanvre par l’analyse de son cycle de vie, 2006, report
interactions between the cement matrix and shiv water extracts, as available at www.agriculture.gouv.fr (2014/09/25).
well as with shiv particles, were characterized. [5] Gle P, Gourdon E, Arnaud L. Acoustical properties of materials made of
This research clearly demonstrates that the components vegetable particles with several scales of porosity. Appl Acoust
2011;72(5):249–59.
extracted from shiv by water have deleterious effects on cement [6] Collet F, Bart M, Serres L, Miriel J. Porous structure and water vapour sorption
setting and hardening. Indeed, they act as strong retarding agents of hemp-based materials. Constr Build Mater 2008;22(6):1271–80.
and reduce the amount of hydrates (C–S–H and portlandite) [7] Cerezo V. Propriétés mécaniques, thermiques et acoustiques d’un matériau à
base de particules végétales: approche expérimentale et modélisation
formed. This considerable modification of the setting and harden- théorique, 2005. thèse de doctorat n°2005ISAL0037, INSA Lyon.
ing process is further illustrated with the loss of mechanical prop- [8] Arnaud L. Mechanical and thermal properties of hemp mortars and wools:
erties (flexural and compressive strengths) measured on both experimental and theoretical approaches. In: Third international symposium
bioresource hemp and other fibre crops. Germany: Wolfsburg; 2000.
short- and long-term curing of the cement paste containing shiv
[9] Arnaud L, Gourlay E. Experimental study of parameters influencing mechanical
extract, as compared to neat cement paste. Additionally, this study properties of hemp concretes. Constr Build Mater 2012;28(1):50–6.
highlights that small differences in the chemical composition of [10] Semple K, Evans PD. Adverse effects of heartwood on the mechanical
shiv water extracts, which depend on the shiv’s origin and process- properties of wood–wool cement boards manufactured from radiata pine
wood. Wood Fiber Sci 2000;32(1):37–43.
ing, can significantly impact the extent of the inhibitory effects [11] Wei YM, Tomita B, Hiramatsu Y, Miyatake A, Fujii T, Yoshinaga S. Hydration
described above. behavior and compressive strength of cement mixed with exploded wood fiber
A new methodology is also described in this paper which allows strand obtained by the water–vapor explosion process. J Wood Sci
2003;49(4):317–26.
the area of influence of the shiv byproducts extracted from the [12] Govin A, Peschard A, Fredon E, Guyonnet R. New insights into wood and
alkaline cement paste to be directly visualized and measured. A full cement interaction. Holzforschung 2005;59(3):330–5.
absence of setting was then shown to occur within a well-defined [13] Soroushian P, Aouadi F, Chowdhury H, Nossoni A, Sarwar G. Cement-bonded
straw board subjected to accelerated processing. Cem Concr Compos
area immediately around the shiv particle. Another major observa- 2004;26(7):797–802.
tion is that this absence of setting is linked to both water extracts [14] Hermawan D, Subiyanto B, Kawai S. Manufacture and properties of oil palm
and degradation byproducts of shiv by the alkaline cement frond cement-bonded board. J Wood Sci 2001;47(3):208–13.
[15] Bilba K, Arsene MA, Ouensanga A. Sugar cane bagasse fibre reinforced cement
medium. composites. Part I. Influence of the botanical components of bagasse on the
A second approach to explain the modification in cement set- setting of bagasse/cement composite. Cem Concr Compos 2003;25(1):91–6.
ting around the hemp aggregate was developed using 13C-labeled [16] Semple KE, Evans PD. Compatibility of some Australian acacias with Portland
cement. Holz als Roh-und Werkstoff 1998;56(1):24.
flax shiv particles. In this case, the labeled extracts were precisely
[17] Karade SR, Irle M, Maher K. Assessment of wood–cement compatibility: a new
tracked within the cement, which not only confirmed the relation approach. Holzforschung 2003;57(6):672–80.
between the presence of extract and the lack of cement setting, but [18] Pereira C, Jorge FC, Irle M, Ferreira JM. Characterizing the setting of cement
also demonstrated the existence of three well-defined areas. The when mixed with cork, blue gum, or maritime pine, grown in Portugal – I:
temperature profiles and compatibility indices. J Wood Sci 2006;52(4):311–7.
first area, composed of almost non-hydrated cement, is of [19] Hachmi MH, Moslemi AA. Correlation between wood–cement compatibility
significant interest, because it did not allow for a good interface and wood extractives. Forest Prod J 1989;39(6):55–8.
Y. Diquélou et al. / Cement & Concrete Composites 55 (2015) 112–121 121

[20] Vaickelionis G, Vaickelioniene R. Cement hydration in the presence of wood [33] Farcas Fabienne, Touze Philippe. La spectrométrie infrarouge à transformée de
extractives and pozzolan mineral additives. Ceram-Silik 2006;50(2):115–22. Fourier (IRTF): Une méthode intéressante pour la caractérisation des ciments.
[21] Jolicoeur C, Simard MA. Chemical admixture-cement interactions: Bulletin des laboratoires des ponts et chaussées 2001;230:77–88.
phenomenology and physico-chemical concepts. Cem Concr Compos [34] Bag R, Beaugrand J, Dole P, Kurek B. Viscoelastic properties of woody hemp
1998;20(2–3):87–101. core. Holzforschung 2011;65(2):239–47.
[22] Jorge FC, Pereira C, Ferreira JMF. Wood–cement composites: a review. Holz als [35] Olmo IF, Chacon E, Irabien A. Influence of lead, zinc, iron (III) and chromium
Roh-und Werkstoff 2004;62(5):370–7. (III) oxides on the setting time and strength development of Portland cement.
[23] Elfordy S, Lucas F, Tancret F, Scudeller Y, Goudet L. Mechanical and thermal Cem Concr Res 2001;31(8):1213–9.
properties of lime and hemp concrete (‘‘hempcrete’’) manufactured by a [36] Yasuda S, Iwase Y, Seguchi Y, Takemura T, Matsushita Y. Manufacture of
projection process. Constr Build Mater 2008;22(10):2116–23. wood–cement boards. 5. cement-hardening inhibitory components of sugi
[24] Rietveld H. A profile refinement method for nuclear and magnetic structures. J heartwood and behavior of catechol as a simple inhibitor model with vicinal
Appl Crystallogr 1969;2(2):65–71. phenolic hydroxyl-groups in cement paste. Mokuzai Gakkaishi
[25] Beaugrand J, Cronier D, Thiebeau P, Schreiber L, Debeire P, Chabbert B. 1992;38(1):52–8.
Structure, chemical composition, and xylanase degradation of external layers [37] Tachi M, Tange J, Nagadomi W, Suzuki Y, Terashima N, Yasuda S. Manufacture
isolated from developing wheat grain. J Agric Food Chem of wood–cement boards. 4. Cement-hardening inhibitory components of the
2004;52(23):7108–17. Malaysian fast-growing tree, acacia-mangium. Mokuzai Gakkaishi
[26] Effland MJ. Modified procedure to determine acid-insoluble lignin in wood and 1989;35(8):731–5.
pulp. Tappi 1977;60(10):143–4. [38] Miller DP, Moslemi AA. Wood–cement composites – effect of model
[27] Monties B. Dosage de la lignine insoluble en milieu acide: influence du compounds on hydration characteristics and tensile-strength. Wood Fiber
prétraitement par hydrolyse acide sur la lignine Klason de bois et de paille. Sci 1991;23(4):472–82.
Agronomie 1984;4(4):387–92. [39] Singh NK, Mishra PC, Singh VK, Narang KK. Effects of hydroxyethyl cellulose
[28] Beaugrand J, Gebruers K, Ververken C, Fierens E, Dornez E, Goddeeris BM, et al. and oxalic acid on the properties of cement. Cem Concr Res
Indirect enzyme-antibody sandwich enzyme-linked immunosorbent assay for 2003;33(9):1319–29.
quantification of TAXI and XIP type xylanase inhibitors in wheat and other [40] Rai S, Chaturvedi S, Singh NB. Examination of Portland cement paste hydrated
cereals. J Agric Food Chem 2007;55(19):7682–8. in the presence of malic acid. Cem Concr Res 2004;34(3):455–62.
[29] Méthodes d’essais des ciments – Partie 3: détermination du temps de prise et [41] Sedan D, Pagnoux C, Smith A, Chotard T. Mechanical properties of hemp fibre
de la stabilité. NF EN 196-3; 2009. reinforced cement: influence of the fibre/matrix interaction. J Eur Ceram Soc
[30] British standard, methods of testing cement – Part 1: determination of 2008;28(1):183–92.
strength. BS EN 196-1; 2005. [42] Durot N, Gaudard F, Kurek B. The unmasking of lignin structures in wheat
[31] http://rsbweb.nih.gov/ij/ (2014/09/25). straw by alkali. Phytochemistry 2003;63(5):617–23.
[32] Trinsoutrot I, Recous S, Mary B, Nicolardot B. C and N fluxes of decomposing [43] Himmelsbach DS, Khalili S, Akin DE. The use of FT-IR microspectroscopic
C-13 and N-15 Brassica napus L.: effects of residue composition and N content. mapping to study the effects of enzymatic retting of flax (Linum usitatissimum
Soil Biol Biochem 2000;32(11–12):1717–30. L) stems. J Sci Food Agric 2002;82(7):685–96.