HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations xxx (2017) xxxxxx
www.elsevier.com/locate/sandf

Failure behavior of cement-treated soil under triaxial

tension conditions
Tsutomu Namikawa , Shota Hiyama 1, Yoshiya Ando 1, Taihei Shibata 1
Department of Civil Engineering, Shibaura Institute of Technology, Japan

Received 1 November 2016; received in revised form 2 June 2017; accepted 21 June 2017

Abstract

Drained triaxial tension tests were conducted to investigate the tensile and shear failure behaviors of cement-treated soils under eec-
tive conning pressures. In the tests, tensile force was applied on saturated cement-treated soil specimens at eective conning pressures.
The experimental result for a cement-treated soil shows that tensile failure occurred at low eective conning pressures, while shear fail-
ure occurred at high eective conning pressures. Based on experimental evidence, a failure criterion for cement-treated soil is discussed.
In this study, tensile failure is assumed to be dominated by the eective minor principal stress and the Mohr-Coulomb failure criterion is
adopted for shear failure. The observed stress states at failure lie on the failure criterion consisting of the tensile and shear failure sur-
faces, indicating that the proposed failure criterion is suitable for cement-treated soils. The experimental results provide a good under-
standing of the failure behavior of cement-treated soils that exhibit tensile and shear failure modes.
2017 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Cement-treated soil; Triaxial tension test; Tensile failure; Shear failure; Failure criterion (IGC: D6)

1. Introduction showed that bending tensile failure is induced in the

Ground improvement by cement mixing has been widely
used for structural foundations, remediation methods
against liquefaction and excavation supports in practical
projects. Recently, column and wall shaped ground
improvements have often been employed to reduce con-
struction costs. In the design of column and wall shaped
ground improvements, the tensile behavior of cement-
treated soils becomes a predominant factor when external
forces cause bending moments. Kitazume and Maruyama
(2007) carried out centrifuge model tests for cement-
treated soil columns subjected to embankment loading to
investigate the failure mode of the columns. The results
Peer review under responsibility of The Japanese Geotechnical Society.
Corresponding author.
E-mail address: namikawa@shibaura-it.ac.jp (T. Namikawa).
1
Formerly.
cement-treated columns because of the embankment
loading.
Lattice-shaped ground improvement by cement mixing
has been developed as a method for mitigation of liquefac-
tion. The performance of this method was demonstrated
during the Hyogo-ken Nanbu earthquake in 1995
(Tokimatsu et al., 1996). Namikawa et al. (2007) conducted
a nite element analysis of a lattice-shaped ground
improvement for mitigation of liquefaction. The numerical
results showed that tensile failure occurs at the corners of
the lattice-shaped ground improvement during large earth-
quakes. Khosravi et al. (2016) conducted dynamic cen-
trifuge tests to evaluate the eect of a soil-cement grid
reinforcement on the seismic response of deep soft soil pro-
les. The experimental results showed that strong shaking
events are likely to produce cracking in the soil-cement
walls. These previous studies indicate that it is important
to evaluate the tensile failure behavior of cement-treated

https://doi.org/10.1016/j.sandf.2017.08.011
0038-0806/ 2017 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
2 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx
soils in the design procedure of column and wall shaped
ground improvements.
To assess the tensile failure behavior of ground improve-
ment by cement mixing, the failure criterion of cement-
treated soils must be determined based on experimental evi-
dence under tensile loading conditions. In current design
procedures (CDIT, 2002; BCJ, 1997), the tensile strength
is used to assess the internal stability of ground improve-
ment. The tensile strength is dened as the peak tensile
stress applied to a specimen under simple tensile loading.
A splitting tension test is often used to determine the tensile
strength of cement-treated soils. However, the splitting ten-
sion test is not considered an element test because the stress
condition is not strictly dened at boundaries (Namikawa
and Koseki, 2007). A direct tension test is more appropri-
ate for investigation of the tensile behavior of cement-
treated soils. However, in the direct tension tests, the tensile
behavior is normally measured under an unconned pres-
sure condition; therefore, the failure criterion under eec-
tive conning pressures cannot be provided. The
laboratory test in which the stresses are well controlled in
three dimensions is required to reveal the failure criterion
of cement-treated soils under tensile loading conditions.
In the compression stress region, the failure criteria of
geomaterials are normally determined using triaxial com-
pression test results. Triaxial compression tests have been
conducted to investigate the shear behavior of cement-
treated soils under eective conning pressures (e.g.,
Kawasaki et al., 1981; Tatsuoka et al., 1997; Ismail et al.,
2002). However, the failure criterion at the boundary
between the tensile and shear failures cannot be determined
from a triaxial compression test in which the mean eective
stress is more than one-third of the axial stress. Therefore,
the failure criterion of cement-treated soils has never been
revealed at the boundary between tensile and shear failures.
Drained triaxial tension tests were conducted to investi-
gate the mechanical behavior of cement-treated soils at the
boundary between tensile and shear failures. Namikawa
and Mihira (2007) and Namikawa and Hiyama (2014) con-
ducted triaxial tension tests of cement-treated soils and
revealed their tension behavior under eective conning
pressures. However, those studies only provided experi-
mental results for the tensile failure behavior of cement-
treated soils. The present study provides an extension to
those studies. The main objective of this study is to inves-
tigate the failure behavior of cement-treated soils around
the region where the failure mode changes from tension
to shear. The drained triaxial tension tests were conducted
at eective conning pressures. The failure mode was deter-
mined from the stress-strain relationships and the fracture
region appearing in the specimens after the failure. The tri-
axial tension test results for the cement-treated soil with an
unconned compressive strength qu of 500 kPa show that
tensile failure occurs at low eective conning pressures,
while shear failure occurs at high eective conning
pressures.
Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
Based on the experimental evidence, a failure criterion
for cement-treated soils is discussed. In this study, tension
failure is assumed to be dominated by the eective minor
principal stress and the Mohr-Coulomb failure criterion
is adopted for shear failure. The stress states at failure
obtained from the tests lie on the failure criterion consist-
ing of tensile and shear failure surfaces, indicating that
the proposed failure criterion is suitable for cement-
treated soils. The experimental results provide a good
understanding of the failure behavior of cement-treated
soils that exhibit tensile and shear failure modes.
2. Experimental method
2.1. Specimen preparation
The cement-treated soils used in this study were pre-
pared by mixing Toyoura sand, Portland cement, distilled
water and Kaolin clay. The compositions of the cement-
treated soils used are listed in Table 1. The specimens con-
sist mainly of Toyoura sand, cement and water; the
mechanical properties of the cement-treated sands were
examined in this study. A small amount of Kaolinite clay
was added to prevent separation of the cement-water paste
from the aggregate.
Triaxial tension tests were conducted for two cement
contents. The unconned compressive strength qu required
in practical projects is generally, 3002000 kPa approxi-
mately (Nozu, 2005; Porbaha et al., 2005). The mixing pro-
portions shown in Table 1 was selected to prepare
specimens with qu of approximately 500 kPa for Case 1
and 1000 kPa for Case 2. Triaxial tension tests of the
cement-treated sand with higher strength (qu = 2000 kPa)
have already been performed by Namikawa and Mihira
(2007).
The specimens were prepared by the following proce-
dure. All the materials shown in Table 1 are mixed thor-
oughly. The mixed materials are placed into a cylindrical
mold of 140 (height)
50 (diameter) mm and compacted
with vibration. The specimen is cured under wet condition
for 5 days. Thereafter the cylindrical specimen is trimmed
to the specimen prole shown in Fig. 1. The specimen
diameter is initially 50 mm and is reduced to 45 mm for a
distance of 25 mm on either side of the center of the spec-
imen, as seen in Fig. 1. The specimen is cured in de-aired
water for 24 h to increase its saturation. In the drained tri-
axial tension tests, the specimen is consolidated for approx-
imately 1 day; hence, the total curing time is 7 days. The
Table 1
Mixing properties of cement-treated soils.
Case Water Cement Sand Kaolin
1 18.7 4.5 71.2 5.6
2 18.6 7.0 69.3 5.0
Note: weight (%).
 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx
Fig. 1. Specimen geometry.
unconned compressive strengths of the specimens result-
ing from the procedure described above are approximately
500 and 1200 kPa for Cases 1 and 2, respectively.
2.2. Test apparatus
A schematic of the triaxial tension apparatus used in this
study is shown in Fig. 2. The mechanism of applying the
tensile force on the specimen is similar to that reported
by Koseki et al. (2005) who conducted direct tension tests
of cement-treated sands. In their apparatus, a specimen is
placed between two holders. Gypsum is used as a lling
material for the gap between the specimen and the holders
Fig. 2. Schematic of testing apparatus.
Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
3
to transfer the tension force during the loading process.
The holders are attached to the top and bottom pedestals
with universal joints to avoid the bending moment that acts
on the specimen. Namikawa and Koseki (2007) numeri-
cally simulated the tensile strength of cement-treated soils
obtained by several testing methods and revealed that the
direct tension test provided reliable evaluation of the tensile
strength.
The applied tensile load was measured using an inner
load cell that is free from the eects of friction at the bear-
ing. The displacement of the loading shaft was measured
using an external displacement transducer (EDT). The dis-
placements of the specimen surface were also measured
using local displacement transducers (LDTs) (Goto et al.,
1991) so as to obtain accurate displacements without bed-
ding error. The axial strain is calculated from the displace-
ments measured by LDTs in the sections for the test results.
The volumetric change in pore water was measured using a
weight scale with a cell in which back-pressure can be
applied. The image of the apparatus taken during the triax-
ial tension test is shown in Fig. 3.
2.3. Test procedure
The specimen was surrounded by a latex membrane
when placed into the triaxial tension apparatus. After
placement, the specimen was saturated using a double vac-
uuming method (Ampadu and Tatsuoka, 1993). Moreover,
a back-pressure of 200 kPa was applied to the specimen to
achieve a fully saturated condition. Skemptons B-value
measured in the tests ranged from 0.8 to 0.95 for Case 1
and from 0.7 to 0.95 for Case 2. The B-value for a fully sat-
4 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx

0.010.1%/min. A lower strain rate than the recommended

rate was used in this study to ensure constant pore pres-
sures within the specimens during drained tensioning.
Drained triaxial compression tests were also carried out
for specimens prepared using the same method as that used
for the triaxial tension tests. Cylindrical specimens with
diameter of 50 mm and heights of 100 mm were used in
the triaxial compression tests. The experimental conditions
for the drained triaxial tension and compression tests are
summarized in Table 2.
Direct tension tests were also conducted in this study. In
the direct tension tests, the specimen was prepared in the
same manner as for the triaxial tension test, and the appa-
ratus shown in Fig. 2 was used. The latex membrane was
not used and the tension load was applied to the specimen
under an unconned condition. The tensile strengths
obtained from the direct tension tests are listed in Table 3.
The tensile strengths were approximately 50 and 200 kPa
for Cases 1 and 2, respectively.
Fig. 3. Image of triaxial tension test.

3. Experimental results
urated condition without system compliance can be calcu-
lated as 3.1. Results for Case 1
1
B 1 The triaxial tension test results for Case 1 are shown in
1 nK
Kw
s

Fig. 4. There is a slight dierence between the initial stress-

strain relationships in all the cases, indicating that the eec-
where n is the porosity, Ks is the bulk modulus of the soil
tive conning pressure r0 c does not signicantly aect the
skeleton and Kw is the bulk modulus of pore water
initial elastic modulus under the tension loading condition.
(Skempton, 1954). The value of n evaluated from the water
After the initial loading process, nonlinear behavior is
content of the specimens was approximately 0.4. The val-
observed in all the cases and the stress-strain relationship
ues of the initial elastic modulus obtained from the triaxial
varies with the eective conning pressure r0 c. At r0 c = 30 -
tension test results were approximately 1.3 GPa for Case 1
kPa (1T-1), a slight strain-hardening behavior is observed
and 3.0 GPa for Case 2. Assuming the value of Poissons
ratio of the specimens is 0.333, Ks corresponds to the elastic
modulus. The value of Kw was approximately 2 GPa. Sub- Table 2
Summary of triaxial tests.
stituting these values into Eq. (1), we obtain B-values of 0.8
for Case 1 and 0.65 for Case 2 for a fully saturated condi- Test Case r0 c (kPa) Loading
tion. Comparison between the measured and estimated B- 1T-1 1 30 Tension
values indicates that the specimens are approximately satu- 1T-2 1 100 Tension
1T-3 1 200 Tension
rated. However, the B-value obtained in the tests does not
1T-4 1 300 Tension
fulll the undrained triaxial test condition. Therefore, the 1T-5 1 400 Tension
triaxial tests were conducted under the drained condition. 1T-6 1 500 Tension
After applying the back pressure, the specimen was 1T-7 1 500 Tension
isotropically consolidated at the prescribed eective conn- 1C-1 1 50 Compression
1C-2 1 100 Compression
ing pressures for 24 h. In both cases, the eective conning
1C-3 1 200 Compression
pressure r0 c was selected to be in the range of 100500 kPa. 2T-1 2 100 Tension
In Case 1, the test was also performed at the low conning 2T-2 2 200 Tension
pressure (r0 c = 30 kPa) because the strength was low. The 2T-3 2 300 Tension
tests were performed twice at selected conning pressures 2T-4 2 400 Tension
2T-5 2 500 Tension
to ensure the repeatability of the experiment.
2C-1 2 50 Compression
After the consolidation, the drained triaxial tension test 2C-2 2 50 Compression
was carried out under displacement control at a constant 2C-3 2 100 Compression
vertical loading rate of 0.005 mm/min. The Japanese 2C-4 2 100 Compression
Geotechnical Society standard JGS2552-2015 (JGS, 2015) 2C-5 2 150 Compression
2C-6 2 200 Compression
recommends that the strain rate for direct tension tests
on rocks and articially-made geomaterials should be Note: r0 c, eective conning pressure.

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx 5

Table 3
Summary of direct tension tests.
Test Case Tf (kPa)
1DT-1 1 57
1DT-2 1 52
2DT-1 2 208
2DT-2 2 201
Note: Tf, tensile strength.
before the peak stress state, and the failure occurs suddenly
during the strain-hardening process. Hence, the strain level
at failure is very low; the absolute value of the axial strain
at failure in test 1T-1 is approximately 0.015%. A similar
stress-strain relationship before the peak stress state was
observed in the direct tension tests conducted by Koseki
et al. (2005). Thus, the failure mode of the specimen at
low r0 c for Case 1 can be identied as tensile failure.
The strain-hardening behavior before the peak stress
state becomes signicant at higher eective conning pres-
sures. The strain level at failure increases with increasing
r0 c. At r0 c = 400 kPa (1T-5) and 500 kPa (1T-6 and 1T-7),
the absolute values of the axial strain at the peak stress
are greater than 0.5%. Similar pre-peak behavior is gener-
ally observed in the triaxial compression test results. The
stress-strain relationships of the drained triaxial compres-
sion test for Case 1 are shown in Fig. 5. The stress-strain
relationships of the triaxial tension tests at r0 c = 400 kPa
(1T-5) and 500 kPa (1T-6 and 1T-7) are similar to that of
the triaxial compression tension test at r0 c = 100 kPa (1C-
1), implying that the shear failure occurs in the specimen
in the triaxial tension test at high r0 c.
In addition, Fig. 4 shows the change in the volumetric
strain in the triaxial tension test. The volumetric strain is
not signicantly generated during the tension loading pro-
cess. The volumetric strain is roughly one third of the axial
tensile strain, implying that the considerable dilatancy
behavior is not observed during the tension loading
process.
The specimens after the triaxial tension tests are shown
in Fig. 6. The fracture surfaces of the specimens at
r0 c = 30 kPa and 100 kPa are approximately perpendicular
to the applied tensile stress, indicating the occurrence of
tensile failure. Conversely, an inclined fracture surface
forms at r0 c = 200 kPa, implying that shear failure occurs
at high conning pressures. Although the failure mode can-
not be strictly determined by examining the fracture sur-
face, Fig. 6 provides experimental evidence identifying
the failure mode. The stress-strain relationships shown in
Fig. 4 and the formed fracture surface shown in Fig. 6 sug-
gest that tensile failure occurs at low r0 c, while shear failure
occurs at high r0 c under the triaxial tension conditions in
Case 1.
Fig. 6 also demonstrates that the failure regions appear
at dierent locations in each test, indicating that a weak
part is not yielded at a xed location in the specimens. Such

Fig. 4. Triaxial tension test results for Case 1: (a) stress-strain relationships for tests 1T-1, 1T-2 and 1T-3; (b) volumetric strain changes for tests 1T-1, 1T-2
and 1T-3; (c) stress-strain relationships for tests 1T-4, 1T-5, 1T-6 and 1T-7; (d) volumetric strain changes for tests 1T-4, 1T-5, 1T-6 and 1T-7.

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
6 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx

Fig. 7. Stress-strain relationships obtained from triaxial tension tests for

Case 2.

the amount of axial strain at the failure is very small.

Fig. 5. Triaxial compression test results for Case 1: (a) stress-strain
relationships; (b) volumetric strain changes.
random failure locations support the homogeneity of the
specimen prepared using the method adopted in this study.
3.2. Results for Case 2
The stress-strain relationships for Case 2 are shown in
Fig. 7. Although the maximum deviator stress increases
with increasing the eective conning pressure r0 c, no sig-
nicant dierence in shape is observed between the stress-
strain relationships. For all the values of r0 c, the stress-
strain relationship shows a slight nonlinear behavior and
The strain level at failure in Case 2 ranged from 0.01 to
0.04%. The strain level at failure is similar to that observed
for low r0 c (1T-1) in Case 1. This stress-strain relationship
implies that tensile failure occurs in all the triaxial tension
tests for Case 2.
The typical stress-strain relationships of the drained tri-
axial compression tests for Case 2 are shown in Fig. 8. Sig-
nicant nonlinear behavior is observed before the peak
stress for the compression test results. The amount of axial
strain at failure in the compression tests is much greater
than that for the tension tests. The dierence between the
stress-strain relationships of the triaxial tension and com-
pression tests indicates that the failure mode appearing in
the specimens in the triaxial tension tests diers from that
in the triaxial compression tests in Case 2.
It should be noted that the volumetric strain is evaluated
from the volumetric change in pore water measured using a
weight scale, and the accuracy of the measured volumetric
strain is reduced at low strain levels. Therefore, the volu-
metric strain change observed in the triaxial tension tests
for Case 2 is not shown in this study. In future studies, a
method for directly measuring the displacement in the lat-
eral direction should be used to accurately evaluate the vol-
umetric strain at low strain levels.

Fig. 6. Specimens at the end of the triaxial tension tests: (a) 1T-1 (r0 c = 30 kPa); (b) 1T-2 (r0 c = 100 kPa); (c) 1T-3 (r0 c = 200 kPa); (d) 1T-4 (r0 c = 300 -
kPa); (e) 1T-5 (r0 c = 400 kPa); (f) 1T-6 (r0 c = 500 kPa); (g) 1T-7 (r0 c = 500 kPa).

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx
Fig. 8. Typical stress-strain relationships obtained from triaxial compres-
sion tests for Case 2.
Fig. 9. Specimens at the end of the triaxial tension tests: (a) 2T-1
(r0 c = 100 kPa); (b) 2T-4 (r0 c = 400 kPa).
The specimens after the triaxial tension tests 2T-1
(r0 c = 100 kPa) and 2T-4 (r0 c = 400 kPa) are shown in
Fig. 9. The fracture surfaces of the specimens are approxi-
mately perpendicular to the applied tensile stress, indicat-
ing that tension failure occurs in these specimens. The
stress-strain relationships shown in Fig. 7 and the formed
fracture surface shown in Fig. 9 suggest that tensile failure
occurs in the specimens under the triaxial tension condition
for Case 2, regardless of the eective conning pressures.
Namikawa and Mihira (2007) conducted triaxial tension
tests for cement-treated soils with qu = 2000 kPa. The
results obtained in that study is shown in Fig. 10. The tri-
axial tension test results are similar to those for Case 2; the
evidence for only tensile failure is found in the stress-strain
relationships. The experimental results for the previous
study and Case 2 do not provide clear evidence for the
transitional region between tensile and shear failures. In
the result for Case 1, evidence for shear failure is found
in the triaxial tension tests at high eective conning pres-
sures. The experimental results of Case 1 clearly show the
mechanical behavior in the transitional region.
Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
7
Fig. 10. Stress-strain relationships obtained from triaxial tension tests
conducted by Namikawa and Mihira (2007).
The experimental results imply that the failure mode
appearing in the triaxial tension tests depends on the
strength of the specimen and the eective conning pres-
sure. The variation of the failure mode may be associated
with the cementation eect. Therefore, the cementation
eect on the failure mode should be considered when devel-
oping a constitutive model for cement-treated soils. Based
on the experimental results, a failure criterion, which is
one of the essential elements in a constitutive model, will
be discussed in the following sections.
4. Discussion
4.1. Failure criteria
The experimental evidence indicates that the failure
mode of the cement-treated soil specimen depends on the
eective conning pressure. This transition of the failure
mode should be considered when dening the failure
criterion for cement-treated soil. Several studies proposed
constitutive models for the mechanical behavior of
cement-treated soils (e.g., Kasama et al., 2000; Lee et al.,
2004; Horpibulsuk et al., 2010). Most of the proposed
models are based on triaxial compression test results.
Therefore, few constitutive models satisfactorily explain
the tensile failure behavior of cement-treated soils. In this
study, the criterion for the tensile and shear failures is
examined based on the experimental results.
The Mohr circles for the failure conditions are shown in
Fig. 11. These stress circles are determined from the eec-
tive principal stresses at the peak stress state for the triaxial
tension and compression test results. For some triaxial
compression tests in which the peak stress is not clearly
observed during the loading process, the stress for the fail-
ure conditions is dened as the eective stress at the axial
strain ea = 2%.
It seems that the shear strength is proportional to the
normal stress in the compression region. Therefore, the
Mohr-Coulomb failure criterion could be applied for the
failure behavior in the compression region. The
Mohr-Coulomb failure criterion is superimposed in
Fig. 11. The values of the friction angle /0 and cohesion
8 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx

Triaxial tension test Triaxial tension test

(a) Triaxial compression test
Triaxial compresson test
Tensile strength line
Mohr-Coulomb failure criterion Tensile strength line
600 Mohr-Coulomb failure criterion
2000
Shear stress (kPa)

500

Shear stress (kPa)

400
1500
300
200 1000
100
0
500
-200 0 200 400 600 800 1000
Effective normal stress (kPa) 0
-1000 -500 0 500 1000 1500 2000 2500 3000
(b) Triaxial tension test Effective normal stress (kPa)
Triaxial compression test
Tensile strength line Fig. 12. Mohr circles at failure obtained from triaxial tests conducted by
Mohr-Coulomb failure criterion Namikawa and Mihira (2007).
1000
Figs. 11 and 12 indicate that the two failure surfaces,
Shear stress (kPa)

800
i.e., the tensile and shear failure surfaces, can be adopted
600 for the failure criterion of cement-treated soils. For this cri-
400 terion, the cementation eect is expressed by Tf for tensile
failure surfaces and c0 for shear failure surfaces. Although a
200 single parameter has been introduced for the cementation
0 eect in other studies (e.g., Kasama et al., 2000;
-500 0 500 1000 1500
Horpibulsuk et al., 2010), the two parameters Tf and c0
Effective normal stress (kPa)
are adopted here to describe the cementation eect on the
Fig. 11. Mohr circles at failure: (a) Case 1; (b) Case 2. tensile and shear failure modes.
Based on the experimental results that tensile failure
occurs when the eective minor principal stress r03 reaches
c0 are determined from the triaxial compression test results
Tf, the function for the tensile failure is dened as
using a least squares method. Moreover, the tensile
strength line is superimposed in Fig. 11. The values of F t r0ij r03  T f 0 2
the tensile strength Tf are determined from the direct ten-
where r0 ij is the eective stress tensor. r0 3 can be written as
sion test results. The values of /0 , c0 , and Tf are listed in
the following general expression (Nayak and Zienkiewicz,
Table 4. Fig. 11 shows that not only the triaxial compres-
1972):
sion test results, but also some of the tension test results p  
of Case 1 follow the Mohr-Coulomb failure criterion. 2 J2 2
r03  p sin h p p0 3
Meanwhile, the triaxial tension test results for Case 2 indi- 3 3
cates that the failure occurs when the Mohr circles reach
where J2 is the second invariant of the deviator stress; h is
the tensile strength line. The Mohr circles at failure
the Lode angle, and p0 is the mean eective stress. Substitu-
obtained from the triaxial tension and compression tests
tion of Eq. (3) in Eq. (2) yields the following general
conducted by Namikawa and Mihira (2007) are shown in
expression for the tensile failure function:
Fig. 12. The Mohr circles in Fig. 12 have a similar feature
as those for Case 2; although the triaxial compression test p  
2 J2 2
results follow the Mohr-Coulomb failure criterion, the ten- F t r0ij p sin h p  p0  T f 0 4
sile strength line is vertical in this stress plane and is tan- 3 3
gent to the Mohr-Coulomb circles.
Here, the triaxial tension and compression stress conditions
are expressed at h = p/6 and p/6, respectively.
Figs. 11 and 12 indicate that the Mohr-Coulomb failure
Table 4 criterion can be adopted as the shear failure function. The
Material properties for failure criterion.
Mohr-Coulomb failure function is expressed as (Nayak
Case /0 c0 (kPa) Tf (kPa) and Zienkiewicz, 1972)
 
1 40 90 50 p 1
2 33 300 200 F s rij J 2 cosh p sinhsin/ c0 cos/0 p0 sin/0 0
0 0

Note: /0 , friction angle; c0 , cohesion; Tf, tensile strength.

3
5

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx 9

Fig. 13. Failure stress states in meridian plane: (a) Case 1; (b) Case 2.

Fig. 13 shows the tensile and shear failure functions
dened by Eqs. (4) and (5) in the meridian plane of the
stress space. The values of /0 , c0 and Tf listed in Table 4
are used for the failure functions. The stress states at failure
obtained from the tests are superimposed in Fig. 13. The
triaxial compression results lie on the failure surface
expressed by Eq. (5), indicating that the compression fail-
ure of cement-treated soils can be dened by the Mohr-
Coulomb failure criterion.
To highlight the comparison further between the failure
criterion and triaxial tension test results, Fig. 13 is enlarged
in Fig. 14. In the triaxial tension test results for Case 1, the
failure stress states at low values of p0 (1T-1) lie on the ten-
sile failure line, while the failure stress states at high values
of p0 (1T-5, 1T-6 and 1T-7) lie on the shear failure line.
These results are consistent with the failure modes identi-
ed in the stress-strain relationships and the fracture sur-
faces observed after the tests; the tensile failure mode is
observed at r0 c = 30 kPa (1T-1), while the shear failure
mode is observed at r0 c = 400 kPa (1T-5) and 500 kPa
(1T-6 and 1T-7). These results indicate that the failure cri-
terion of the cement-treated soils can be appropriately
expressed by Eqs. (4) and (5).
In almost all the triaxial tension test results for Case 2,
the failure stress states lie on tension failure surfaces. This
result is consistent with the tensile failure modes identied
by the stress-strain relationships and the fracture surfaces
observed for Case 2. At r0 c = 500 kPa (2T-5), the stress
state at failure appears to lie on the shear failure surfaces.
In this test, the stress-strain relationship and fracture sur-
face support the tensile failure mode. This contradiction
indicates that it is dicult to clearly divide the failure
modes near the boundary between tensile and shear failure.
The failure criterion and the stress-strain relationship near

Fig. 14. Failure stress states for triaxial tension tests in meridian plane: (a) Case 1; (b) Case 2.

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
10 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx

the boundary of the failure modes will be discussed in the

following section.
'1
4.2. Transition between tensile and shear failure (a)

The experimental results support the failure criterion

consisting of the tensile and shear failure functions for
cement-treated soils. A schematic of the two failure sur-
faces in the meridian plane of the stress space is shown in
Fig. 15. The failure criterion consisting of the tensile and
shear failure functions is plotted as a solid line. The mean
eective stress at the intersection of the two failure surfaces
p0 t is dened as
'3 '2
c0 cos /0  T f T f sin /0  c0 cos /0
p0t p tan h 6
1  sin /0 31  sin /0 '1
(b)
When p0 5 p0 t, tension failure occurs, and when p0 > p0 t,
shear failure occurs. p0 t varies with h. The minimum value
of p0 t is obtained for the triaxial compression condition,
h = p/6.
2c0 cos /0  3  sin /0 T f p
p0tmin ; h 7
3  3 sin /0 6
The maximum value of p0 t is obtained for the triaxial
tension condition, h = p/6.
'3 '2
4c0 cos /0  3 sin /0 T f p
p0tmax ; h 8 '1
3  3 sin /0 6
(c)
The two failure surfaces with the p plane are shown in
Fig. 16. When p0 = p0 tmin, the two failure lines intersect at
the major principal stress, and when p0 = p0 tmax, the two
failure lines intersect at the minor principal stress. When
p0 tmin < p < p0 tmax, the failure surface consists of the tensile
and shear failure surfaces. It should be noted that the eect
of the intermediate principal stress on the failure criterion
has not been examined in this study. The shape of the fail-
ure surface on the p plane should be determined consider- '2
'3

'1, '2, '3 : Effective principal stress

Fig. 16. Schematic of failure criterion in p plane: (a) p0 = p0 tmin; (b)
p0 tmin < p0 < p0 tmax; (c) p0 = p0 tmax.

ing the eect of the intermediate principal stress. Further

Fig. 15. Schematic of failure criterion in meridian plane.
experimental studies in which the intermediate principal
stress is independently controlled are required to determine
the shape of the failure surface on the p plane.
The failure criterion consisting of tensile and shear fail-
ures has been examined in this study. In this criterion, there
is a boundary between the tensile and shear failure sur-
faces. The failure criterion which describes the smooth
transition between the tensile and shear failures is dis-
cussed. The schematic of the tensile and shear failure crite-
ria for the relationship between the normal and shear
stresses is shown in Fig. 17. It can be seen that there are dif-
ferences between the friction angle /0 and eective normal

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx 11

Shear stress behavior is clearly observed before the peak stress state.
Here, the stress-strain relationship obtained from the triax-
Tensile failure Shear failure
ial tension tests is compared to that from the triaxial com-
pression tests. The normalized stress-strain relationships
Tf
' obtained from the triaxial tension and compression tests
are shown in Fig. 18. The results of the triaxial tension test
in which the shear failure is observed (1T-7) and the triaxial
compression test at low eective conning pressures (1C-1,
2C-1) are compared in Fig. 18. Here the normalized devia-
Cf Effective normal stress tor stress qn is dened as
r1  r3
Fig. 17. Schematic of tensile and shear failure lines in the relationship qn 11
between normal and shear stresses.
qf
where r1 is the major principal stress, r3 is the minor prin-
cipal stress, and qf is the deviator stress at failure dened by
stress r0 -intercept of the two failure criterion lines. Here it the Mohr-Coulomb failure criterion. Fig. 18 shows that the
is assumed that the friction angle and r0 -intercept vary in normalized stress-strain relationships obtained from the
the transitional region. The Mohr-Coulomb failure crite- triaxial tension tests agree approximately with those
rion for shear failure expressed by Eq. (5) is rewritten as obtained from the triaxial compression tests. This agree-
p p ! ment indicates that a unique normalized stress-strain rela-
2 J 3 1 tionship can be determined during the shear failure process.
F s r0ij p sin h  C f  p0 0;
2
0 cos h 
3 2 sin / 2 The normalized stress-strain relationships of the triaxial
tension tests (1T-3 and 2T-5) are shown in Fig. 19. This g-
c0
Cf 9 ure shows the normalized stress-strain relationships around
tan /0 the intersection of the tension and shear failure lines for
where Cf is the r0 -intercept of the shear failure line. Taking Cases 1 and 2 (see Fig. 14). There is reasonable agreement
the linear combination of the cosine and sine functions in between the two stress-strain relationships before the peak
the parentheses, Eq. (9) can be rewritten as stress state, implying that a unique normalized stress-strain
p s relationship can be determined around the boundary of the
0 2 J2 1 3 tensile and shear failure surfaces. The stress-strain relation-
F s rij p 2 0
sinh a  C f  p0 0;
3 4 4 sin / ships shown in Fig. 19 dier from those in the shear failure
p process shown in Fig. 18. In particular, a signicant dier-
3 ence is observed between the strain levels at the peak stress.
sin a q 10
2 0 Moreover, as previously mentioned, the strain level at the
sin / 3
peak stress is much lower in the tensile failure mode. This
Comparing Eqs. (10) and (4), we can nd that the failure variation of the stress-strain relationships suggests that a
criterion is smoothly transferred from the shear failure mechanical model for cement-treated soils is required to
function to the tensile failure function by changing /0 describe the behavior associated with the failure mode.
and Cf. In the transitional region, it is assumed that /0 The stress-strain relationships under triaxial tension
increases and Cf decreases as p0 decreases. As /0 ? p/2 (h = p/6) and compression (h = p/6) were discussed in
and Cf ? Tf, Eq. (10) corresponds to Eq. (4) dening the this study. These relationships may be aected by the inter-
tensile failure function.
A specied function for changing /0 and Cf in the tran-
1.2
sitional region is not provided in this study because Fig. 14
Normalized deviator stress

suggests that the failure criterion consisting of the tensile 1.0

and shear failure surfaces approximately expresses the fail-
ure stress state determined based on the laboratory tests in 0.8
this study. However, multiple failure modes are likely to 0.6
occur in the transitional region. Further study is required
to determine the specied function for changing /0 and 0.4 1T-7
1C-1
Cf based on more elaborate experimental results.
0.2 2C-1

4.3. Stress-strain relationship associated with failure mode 0.0

0.0 0.5 1.0 1.5
Absolute value of axial strain (%)
In the triaxial tension test results for Case 1, shear fail-
ure occurs under the high eective conning pressures. Fig. 18. Normalized stress-strain relationships for tests 1T-7, 1C-1 and
During the process of shear failure, the strain hardening 2C-1.

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
12
1.0
Normalized deviator stress
0.8
0.6
0.4
0.2
0.0
0.00 0.05
Absolute value of axial strain (%)
Fig. 19. Normalized stress-strain relationships for tests 1T-3 and 2T-5.
mediate principal stress. Further experimental studies in
which the value of h is controlled are required to verify that
a unique normalized stress-strain relationship can be
dened at various values of h.
5. Conclusions
Drained triaxial tension tests were conducted to investi-
gate the tensile and shear behaviors of cement-treated soils
under eective conning pressures. The experimental
stress-strain relationships varied with the eective conning
pressure r0 c. The absolute value of the axial strain at the
peak stress signicantly increased with increasing r0 c. For
Case 1 (qu = 500 kPa), the fracture surfaces of specimens
at low eective conning pressures were perpendicular to
the applied tensile stress, while the inclined fracture sur-
faces were formed in specimens at higher eective conning
pressures. This nding implies that under triaxial tension
loading, the tensile failure occurs at low eective conning
pressures and shear failure occurs at high eective conn-
ing pressures. However, only tensile failure was observed
for Case 2 (qu = 1200 kPa). These results indicate that the
failure mode appearing in the triaxial tension test depends
on the strength of the specimen and eective conning
pressure.
The triaxial tension test results led to a discussion on a
suitable criterion to describe the mechanical behavior of
cement-treated soils. The failure criterion consisting of ten-
sile and shear failure functions was proposed in this study.
It is assumed that tensile failure occurs when the eective
minor principal stress r0 3 reaches the tensile strength Tf,
and the Mohr-Coulomb failure criterion was adopted as
the shear failure function. The failure stress states deter-
mined based on the triaxial tests lay on the proposed failure
surfaces, indicating that the proposed failure criterion
describes reasonably the failure stress state located around
the boundary between the tensile and shear failures. The
normalized stress-strain relationship of the specimen show-
ing shear failure under the triaxial tension condition agrees
reasonably with that under the triaxial compression condi-
tion. This agreement indicates that a unique normalized
stress-strain relationship can be determined for the shear
Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011
T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx
behavior of cement-treated soils, regardless of the loading
conditions.
The results of this study have implications on the devel-
opment of constitutive models for cement-treated soils. In
particular, the failure modes observed in the triaxial ten-
sion tests provide a good understanding on the failure
1T-3 behavior of cement-treated soils around the boundary
2T-5 between the tensile and shear failure modes.
Acknowledgments
0.10 0.15
The authors acknowledge the support of Prof. Junichi
Koseki and Mr. Takeshi Sato, University of Tokyo. The
authors also acknowledge the support of the Japan Society
for the Promotion of Science (JSPS KAKENHI Grant No.
25420509). In addition, the authors thank the anonymous
reviewers for their helpful comments.
References
Ampadu, S.K., Tatsuoka, F., 1993. Eect of setting method on the
behaviour of clays in triaxial compression from saturation to
undrained shear. Soils Found. 33 (2), 1434.
Costal Development Institute of Technology (CDIT), Japan, 2002. The
Deep Mixing Method; Principle, Design and Construction. Balkema,
Lisse, The Netherlands.
Goto, S., Tatsuoka, F., Shibuya, S., Kim, Y.S., Sato, T., 1991. A simple
gauge for local small strain measurements in the laboratory. Soils
Found. 31 (1), 169180.
Horpibulsuk, S., Liu, M.D., Liyanapathirana, D.S., Suebsuk, J., 2010.
Behavior of cemented clay simulated via the theoretical framework of
the Structured Cam Clay model. Comput. Geotech. 37 (12), 19.
Ismail, M.A., Joer, H.A., Sim, W.H., Randolph, M.F., 2002. Eect of
cement type on shear behavior of cemented calcareous soil. J. Geotech.
Geoenviron. Eng., ASCE 128 (6), 520529.
Japanese Geotechnical Society (JGS), 2015. Japanese Geotechnical
Society Standard JGS2552-2015, Method for Direct Tension Test on
Rocks.
Kasama, K., Ochiai, H., Yasufuku, N., 2000. On the stress-strain
behaviour of lightly cemented clay based on an extended critical state
concept. Soils Found. 40 (5), 3747.
Kawasaki, T., Niina, A., Suzuki, Y., Saito, S., Suzuki, Y., Babasaki, R.,
1981. On the deep chemical mixing method using cement hardening
agent. Takenaka Technical Research Report, No. 26, 1342.
Khosravi, M., Boulanger, R., Tamura, S., Wilson, D., Olgun, C., Wang,
Y., 2016. Dynamic centrifuge tests of soft clay reinforced by soil-
cement grids. J. Geotech. Geoenviron. Eng., ASCE 142 (7), 04016027.
Kitazume, M., Maruyama, K., 2007. Internal stability of group column
type deep mixing improved ground under embankment loading. Soils
Found. 47 (3), 437455.
Koseki, J., Sato, T., Mihira, S., Takeya, N., Yoshizawa, M., 2005.
Comparison of tensile strength of cement treated sand by various test
methods. Deep Mixing 05, 95100.
Lee, K., Chan, D., Lam, K., 2004. Constitutive model for cement treated
clay in a critical state frame work. Soils Found. 44 (3), 6977.
Namikawa, A., Hiyama, S., 2014. Triaxial tension test of cement-treated
soil. In: Proceedings of the 15th Danube European Conference on
Geotechnical Engineering, Vienna, Austria, vol. 1, pp. 329334.
Namikawa, T., Koseki, J., 2007. Evaluation of tensile strength of cement-
treated sand based on several types of laboratory tests. Soils Found. 47
(4), 657674.
Namikawa, T., Koseki, J., Suzuki, Y., 2007. Finite element analysis of
lattice-shaped ground improvement by cement-mixing for liquefaction
mitigation. Soils Found. 47 (3), 559576.
 T. Namikawa et al. / Soils and Foundations xxx (2017) xxxxxx 13

Namikawa, T., Mihira, S., 2007. Elasto-plastic model for cement-treated
sand. Int. J. Numer. Anal. Meth. Geomech. 31 (1), 71107.
Nayak, G.C., Zienkiewicz, O.C., 1972. Convenient form of stress
invariants for plasticity. J. Struct. Divis., ASCE 98 (ST4), 949954.
Nozu, M., 2005. Regional report: Asia. In: Proceedings of the Interna-
tional Conference on Deep Mixing, Swedish Deep Stabilization
Research Centre, Stockholm, Sweden, vol. 1, R3R17.
Porbaha, A., Weatherby, D., Macnab, A., Lambrechts, J., Burke, G.,
Yang, D., Puppala, A.J., 2005. Regional report: North American
practice of deep mixing technology. In: Proceedings of the Interna-
tional Conference on Deep Mixing, Swedish Deep Stabilization
Research Centre, Stockholm, Sweden, vol. 1, R47R73.
Skempton, A.W., 1954. The pore-pressure coecients A and B. Geotech-
nique 4, 143147.
Tatsuoka, F., Uchida, K., Imai, K., Ouchi, T., Kohata, Y., 1997.
Properties of cement-treated soils in Trans-Tokyo Bay Highway
project. Ground Improve. 1, 3757.
The Building Center of Japan (BCJ), 1997. Design and Quality Control
Guideline for Building Foundation Treatment; Deep/shallow Soil
Mixing Methods with Cement Admixtures (in Japanese).
Tokimatsu, K., Mizuno, H., Kakurai, M., 1996. Building damage
associated with geotechnical problems. Special Issue Soils Found. 36,
219234.

Please cite this article in press as: Namikawa, T. et al., Failure behavior of cement-treated soil under triaxial tension conditions, Soils Found. (2017),
https://doi.org/10.1016/j.sandf.2017.08.011