Soils and Foundations 2012;52(2):335–345

The Japanese Geotechnical Society

Soils and Foundations

www.sciencedirect.com
journal homepage: www.elsevier.com/locate/sandf

Properties of very soft clays: A study of thixotropic hardening and

behavior under low consolidation pressure
Sochan Seng, Hiroyuki Tanakan
Hokkaido University, Japan
Available online 29 March 2012

Abstract

In a study on the properties of very soft clays, bender element testing was used to evaluate thixotropic hardening behavior; that is, to
measure the stiffness with resting time under constant volume and water content. A laboratory vane test, which measures the undrained
shear strength of the materials, was also carried out for comparison purposes. To investigate the mechanism of the thixotropic
phenomenon, a consolidation test with very low pressure was also performed in a cell equipped with bender elements. The most
important findings from this study are as follows: (1) regardless of soil types, the effect of thixotropy was significant around the liquid
limit state and less remarkable at the lower and higher ranges; (2) the shear modulus at the liquid limit after 24 h resting is around
200 kPa; (3) the correlation between the shear modulus and the undrained shear strength of very soft clays is similar to that of cement-
treated soil proposed by Seng and Tanaka (2011); (4) the increment of the shear modulus developed in the thixotropy process appears to
be noticeably higher than that in the secondary consolidation process. It is believed that these findings are very useful to establish a new
theory for the consolidation of ground filled by very soft clays or dredged soils with extremely high water content as well as to
understand the effects of ageing on the consolidation properties of natural soils.
& 2012 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.

Keywords: Thixotropic hardening; Bender element; Vane shear; Consolidation; Shear modulus; Shear strength (IGC: D05/D06)

1. Introduction disposal areas, but it is becoming increasingly difficult to

A huge amount of dredged soils or very soft clays have
been generated from the maintenance of navigation chan-
nels or the deepening of existing port facilities to cope with
large ships. These dredged soils are normally dumped into
n
Corresponding author.
E-mail address: tanaka@eng.hokudai.ac.jp (H. Tanaka).
0038-0806 & 2012 The Japanese Geotechnical Society. Production and
hosting by Elsevier B.V. All rights reserved.
Peer review under responsibility of The Japanese Geotechnical Society
doi:10.1016/j.sandf.2012.02.010
Production and hosting by Elsevier
find such disposal site due to the environmental problems
they pose. To efficiently utilize these restricted deposit
areas, the settlement of the ground deposited by these
dredged soils with high water content needs to be precisely
estimated. In practice, the settlement caused even by self-
weight is predicted based on the effective stress. The
concept of ‘‘effective stress’’ has been widely accepted by
geotechnical engineers and it is believed that geotechnical
parameters such as strength or stiffness can be simply
correlated with effective stress. In this concept, these
parameters are independent of time, and the apparent
time dependency during consolidation can be understood
by changes in the effective stresses in the process of the
dissipation of the excess pore water pressure. On the other
hand, some behaviors opposite to the effective stress
approach are also recognized, i.e., thixotropic hardening.
The soil loses strength or stiffness by remolding, but
recovers with time. It is not unreasonable to suggest that
336 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345
this factor plays an important role in the process of
settlement. Thus, this paper will try to understand the
thixotropic hardening phenomenon of clays with high
water content.
The term ‘thixotropy’ originally came from colloid
science, as presented in Mewis (1979), Barnes (1997), and
Mewis and Wagner (2009), and as this term has come to be
used in other fields its definition has changed somewhat
according to area of interests. Simply put, in geotechnical
engineering, thixotropic hardening is, after softening by
remolding, the process of time-dependent return to a
harder state under constant water content or volume
conditions. The recovery of stiffness or strength may either
fully or partially depend on material properties. The
mechanism of soil thixotropy has been explained in terms
of both microstructure and macrostructure in geotechnical
engineering by, for example, Seed and Chan (1957),
Osipov et al. (1984), Mitchell and Soga (2005). Particu-
larly, the important role of thixotropy involving in geo-
technical engineering work was perfectly described by
Mitchell (1960), which has inspired many researchers to
pay attention to this topic. It is believed by most research-
ers that thixotropic hardening occurs due to energy
imbalances which are induced by remolding or compac-
tion. When soil is remolded, its strength or stiffness is
reduced because of the dispersing structure caused by
repulsive forces. However, the flocculated interparticles
after remolding are gradually rearranged by attractive
forces and create a structure until an equilibrium state is
reached.
Some experimental data have been reported on this field,
using the conventional destructive testing tools such as
triaxial tests, unconfined compression tests, and laboratory
vane apparatus to measure strength hardening due to
thixotropy. These methods tend not to guarantee the same
testing conditions because different samples are used and a
relatively long period of time is required for preparing the
samples for testing, i.e., it is difficult to measure properties
at a very early time, and the disturbance effect during the
insertion of the vane blade makes it difficult to obtain
reliable results, for example. To overcome these draw-
backs, a nondestructive bender element test, which allows
measurement to be taken very low strain levels, as small as
10  5 (0.001%), is introduced in this paper to study the
thixotropic behavior of very soft clays, together with a
vane shear test measuring the undrained shear strength. In
Table 1
Physical property of soil samples.
Soil types Plastic limit (%) Liquid limit (%)
Fujinomori 21.3 48.6
Kasaoka 27.5 62.0
NSF 29.0 55.0
Ariake 31.4 72.5
Hachirogata 96.5 246.0
Tokuyama 40.0 110.6
another approach, an oedometer test under very low
pressure was also carried out, equipped with the bender
element to examine the development of soil stiffness due to
consolidation. The correlation between the stiffness and
the effective stress is established, and a discussion is made
on the changes in stiffness during secondary consolidation.
Using these test results, the increment of stiffness caused
by thixotropic behavior was evaluated.
2. Sample preparation
The tests were run on very soft clays using samples
obtained from three commercial powder clays and three
natural clays. The commercial clays were Fujinomori,
Kasaoka, and NSF clays. The natural clays were Ariake,
Hachirogata, and Tokuyama clays. Their index properties
are summarized in Table 1. Large variation in the liquid
limits of these clays was noted, ranging from low (48.6%)
to very high (246.0%) values, which in effect allowed a
wide variety of materials with very different characteristics
to be investigated. Since all the clay particles were smaller
than 0.425 mm, however, it was determined that there was
no sand content in these clays.
Samples were prepared by mixing a soil with different
water contents to vary its properties. Commercial clays
prepared by the dry powder condition were mixed with a
water content lower than the target and kept for about one
day before testing to let the soil particles adjust well and to
ensure the homogenous distribution of water. Just before
testing, additional distilled water was added reaching order
to achieve the final water content and the soils were mixed
thoroughly to achieve uniformity. Natural clay samples,
which were stored under wet conditions, were prepared
immediately after remolding with the target water content.
A prepared sample was poured into a mold with various
sizes for different purposes: a plastic cylindrical mold with
a height of 10 cm and the diameter of 5 cm for the bender
element test, a gallon bucket 20 cm in diameter and with a
specimen height of 7–8 cm for the vane test, and a
consolidometer cell with a diameter of 10 cm and adjus-
table height from 4 to 15 cm for the consolidation test.
Vibration was gently applied to drive the air bubbles from
the slurry specimen, and the sample was compacted when
its water content was small and the sample was stiff. The
uniformity of the sample was not well guaranteed in the
latter case, i.e., inevitably the sample was partially
Plasticity index Particle density (g/cm3)
27.3 2.69
33.5 2.61
26.0 2.76
41.1 2.62
149.5 2.43
70.6 2.62
 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345
unsaturated. The mold with the specimen for the thixo-
tropy test was wrapped in thin plastic film, and then kept
in a high humidity box to prevent water evaporation. Since
the changes in water content in each specimen after the test
were found in the range of less than 1% compared to the
initial water content, this method was considered effective.
The water content within the specimens was distributed
such that there was a slightly difference between the top,
middle, and bottom parts. If the specimen was not well-
saturated, the lower portion had a higher water content
and vice-versa for the saturated condition. The elapsed
time for thixotropy was judged to be just after the sample
was poured into the mold.
3. Testing methods
3.1. Bender element test
A pair of bender elements is required for both the
thixotropy and consolidation tests. For the former test, a
parallel type 10 mm in length, 10 mm in width, and 0.5 mm
in thickness was used as the transmitter and a series type
with length of 13 mm, a width of 10 mm, and a thickness
of 0.5 mm was the receiver. The transmitter element was
assembled with a light acrylic cap while the receiver was
attached to a brass pedestal where the protrusion length
was approximately 7 and 9 mm, respectively. As for the
latter test, only the series type was employed and equipped
in the consolidometer cell with a 6 mm protrusion length.
It is important that the protrusion length is sufficient to
allow the bender element to penetrate into the soil sample
to generate or receive shear waves. As input signals, both
sine and rectangular waves have been alternatively used
with wide ranges of frequencies in accordance with
material stiffness, in order to attain a clear output wave-
form. Generally, high frequency waves are required for the
testing stiff soils and lower frequency waves are required
for the less stiff soils.
Methods for determining the arrival time and travel
length has been investigated by many researchers, for
example, Viggiani and Atkinson (1995), Kawaguchi et al.
(2001), and Yamashita et al. (2009). Considering those
previous researches, the ‘‘start-to-start’’ method for deter-
mining the arrival time (Dt) and ‘‘tip-to-tip’’ method for
determining the travel distances (Dd) of the shear wave
were adopted in this study. The shear wave velocity (Vs)
can be calculated from Eq. (1):
Dd
Vs ¼ ð1Þ
Dt
Shear modulus (G) can be derived from Vs, through the
shear wave propagation in an elastic body theory, as
shown below:
G ¼ rVs2 ð2Þ
where r is the density of the soil specimen.
337
It should be pointed out that measured Vs is affected by
the existence of the mold or the consolidometer cell, but
their rigidity is much larger than that of soft clays, so that
the arrival time of the shear wave should not be inter-
rupted by the existence of a mold or a consolidation cell.
For the thixotropy test, the bender elements were
embedded in the specimen for the monitoring period
without repeated insertions in order to avoid disturbance
caused by insertion.
3.2. Vane shear test
The increase in the undrained shear strength during
thixotropic hardening was also confirmed by the vane
shear test. Since conventional triaxial or unconfined tests
cannot be performed on such low strength soils as those
targeted in the present study, the vane shear test was
considered an appropriate test. Undrained shear strength
(su) is calculated as follows:
2T
su ¼ ð3Þ
pD2 ðH þ ðD=3ÞÞ
where T is the measured torque at peak, D is the vane
diameter, and H is the vane height.
The vane diameter and height used in this experiment
were 20 and 40 mm, respectively. The shear rate of the
laboratory vane apparatus was constant at 61 rotations per
minute. All tests were carried out for the same sample
created by a gallon bucket, as already mentioned. When
the predetermined time came, the vane blade was inserted
at a point away from the previous tested points not to be
influenced by the disturbance caused by the previous tests.
To attain accurate results, two different maximum capa-
cities of load cells, 1 and 20 N, were alternately used
according to the shear strength of the material.
4. Thixotropic effect measurement
4.1. Shear wave velocity and shear modulus by bender
element test
Fig. 1 shows an example of increase in the shear wave
velocity (Vs) with time, measured by the bender element
test on a specimen made from Kasaoka clay with a 60.6%
water content.
In this example, an input wave with 1 kHz of frequency
and 7 10 V of amplitude was adopted, as indicated on the
top of the figure. The waves at the lower section of the
figures are the received signals with black triangles which
define their arrival time. Since the water content of the
specimen was almost equal to the liquid limit state, the
received shear wave signals at the beginning of the
measurement were difficult to identify because of their
low amplitude and frequency. P-waves clearly appeared
since they could propagate through the liquid. As time
proceeded, the shear waves’ arrival times were faster,
indicating that the shear wave velocity increased. The
338 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345
Voltage (V)

10 4000
0 Input Signal
A1 (79.8%; 1.10)
-10 Sine 1 kHz 3500 F1 (40.9%; 0.84)
F2 (52.5%; 1.08)
K1 (62.3%; 1.00)
3000

Shear Modulus, G (kPa)

K2 (60.6%; 0.97)
1.5 hrs K3 (52.0%; 0.84)
2500 K4 (62.1%; 1.00)
2.2 hrs K5 (46.0%; 0.74)
N1 (54.0%; 0.93)
2000
2.7 hrs N2 (48.6%; 0.86)
P-waves H1 (241.3%; 0.98)
19.5 hrs 1500 T1 (134.3%; 1.21)
T2 (149.3%; 1.35)
23.7 hrs 1000

44.4 hrs 500

93.8 hrs 0
116.8 hrs 0.1 1 10 100 1000
Time (hr)
142.0 hrs
800
0 3 6 9 12 15
Time (ms)

Shear Modulus, G (kPa)

Fig. 1. Measurement of thixotropic hardening by bender element test for 600
Kasaoka clay with 60.6% of water content.

400
increase in Vs can be assumed to correspond to an increase
in stiffness, which reflects the thixotropic phenomenon.
The received shear waves became much clearer with high
amplitude and frequency after a certain time, while the 200

P-waves seemed to decay.

The same testing method was conducted on other soils
under various water content conditions, and the shear
modulus (G) derived from Vs with elapsed time is illu-
strated in Fig. 2(a). An enlarged scale for high water
content samples with low G magnitude is shown in
Fig. 2(b). The symbols of A, F, K, N, H, and T represent
Ariake, Fujinomori, Kasaoka, NSF, Hachirogata, and
Tokuyama clays, respectively, with sample number. The
number in the brackets indicates the water content (w) and
the normalized water content by the liquid limit (w/wL). It
can be seen from the figures that the G values for all
conditions increases with time even at over limit liquid
states. Also, G builds up nearly in proportion to time on a
logarithmic scale, but this magnitude clearly depends on
the type of soil and the amount of water content.
Since G increases in time, G at 24 h (G24) was used a
representative parameter for the identification of the
characteristics of various soil types and the influence of
water content. G24 is plotted in Fig. 3(a) against the
normalized w/wL, for different types of materials. A clear
correlation between the two parameters can be observed
with minor scatters. Obviously, G decreases as the normal-
ized w/wL increases, meaning that wL defines the magni-
tude of G regardless of the type of soil. In this case, the G24
value at wL appears to be around 200 kPa.
By extruding data of Hachirogata clay, because of its
strange point in the G24 and w/wL relation shown in
Fig. 3(a), the correlations of hardening G and normalized
w/wL at not only 24 h but also various times are plotted in
Fig. 3(b) to examine the effect of thixotropy corresponding
0
0.1 1 10 100 1000
Time (hr)
Fig. 2. (a) Variation in G with time at rest and (b) enlarged scale.
to a wide range of water contents. Generally, the relations
at different times seem to be parallel each other, but if
these relations are carefully examined, the ranges of G at
various observed times are somewhat different by w/wL,
i.e., the increment of G to time at around wL is the largest,
while at smaller water contents than wL, say, when w/wL is
less than 0.8, the tendency of G increasing against time
becomes smaller. To confirm this trend clearly, the incre-
mental G based on G24 is plotted against the w/wL ratio in
Fig. 3(c). Fitting curves are drawn by excluding the
three encircled data. This tendency described above can
be recognized in Fig. 3(c): the thixotropic effect on
G is prominent at around wL, and at lower and higher
water contents the thixotropic hardening becomes less
remarkable.
As pointed out by Mitchell (1960), water content has
been used as a primary parameter to evaluate the magni-
tude of soil strength gained by thixotropy, which results
from the alteration of internal soil structure between
dispersion and flocculation. The range of water content
between the plastic limit (wP) and the liquid limit (wL)
affecting the thixotropy has been discussed by many
researchers, for example Skempton and Northey (1952)
and Seed and Chan (1957). They suggested that thixotropic
 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345 339

increase in the thixotropic effect. However, Skempton and

10000
Ariake Northey showed a consistent increase in the thixotropic
Fujinomori effect only from wP until wL, then over the wL value, some
Kasaoka
Shear Modulus at 24hrs, G24 (kPa)

NSF
samples continued to increase while it dropped off in
1000 Hachirogata others. Mitchell (1960) conducted the experiment on
Tokuyama
compacted clays in the range of w from 0.36 to 0.90 of
wL and found that thixotropy had the greatest effect at
100
values of w between about 0.56 and 0.74 of wL. Therefore,
to our knowledge, an explanation of the w effect in the
range over wL has yet to be provided in the literature.
Between the ranges of w/wL from 0.75 to 1.35 in this
10 experiment, the effect of thixotropy was found to increase
from 0.75 to 1.0 of w/wL ratio, which is consistent with
Skempton and Northey (1952), but over wL, the thixo-
1 tropic effect decreases. However, the upper boundary of
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 the water content without any thixotropic effect has not
Normalized w/wL been established even at 1.35 times wL. Suthaker and Scott
(1997) carried out tests on the thixotropic strength of oil
10000 sand fine tailings which contained extremely high w of
Resting Time
5hrs
more than 8 times of wL, and proved that the thixotropic
10hrs effect was still there; nonetheless, the properties of this soil
1000 24hrs differ considerably from marine clays. In the case of
Shear Modulus, G (kPa)

48hrs Tokuyama dredged clay, the natural water content con-

120hrs
192hrs tains only 1.20 times of wL, at which point thixotropic
effect is still pronounced. From this finding, it can be
100
assumed that thixotropy plays an important role in
altering the properties of natural soil.
Fig. 4 shows the relationship between normalized G with
10 G24 and the logarithm of elapsed time for different clays.
As can be seen in the figure, tested samples may be
classified into three groups in accordance with their w/wL
ratios, i.e., less than 0.95, between 0.95 and 1.15, and
1
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 greater than 1.15. The relation at around wL shows the
Normalized w/wL highest gradient to elapsed time, followed by the greater
water content group, while the low water content samples
2.4
showed the smallest increment of G against time.
G10/G24
2.2
G48/G24
2.0 w/wL < 0.95
G120/G24 F1(40.9%; 0.84)
1.8 3.0 K3 (52.0%; 0.84)
Normalized G/G24

1.6 K5 (46.0%; 0.74)

N1 (54.0%; 0.93)
1.4 N2 (48.6%; 0.86)
2.5
1.2 0.95<w/wL < 1.15
A1 (79.8%; 1.10)
1.0
F2 (52.5%; 1.08)
Normalized G/G24

2.0
0.8 K1 (62.3%; 1.00)
K2 (60.6%; 0.97)
0.6 K4 (62.1%; 1.00)
1.5 w/wL > 1.15
0.4
T1 (134.3%; 1.21)
0.2 T2 (149.3%; 1.35)
1.0
0.0
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
Normalized w/wL
0.5
Fig. 3. (a) Relationship between G and normalized w/wL, (b) G and
normalized w/wL at different resting times, (c) Normalized G/G24 versus w/wL.
0.0
0.1 1 10 100 1000
effect at or close to wP can be very low or cannot be
Time (hr)
observed. The findings of most investigations presented in
the literature were that increasing water content caused an Fig. 4. Relationship between normalized G/G24 with time.
340 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345

4.2. Undrained shear strength by vane test steeper gradient were recognized than those in Fig. 3(a). At
wL, the magnitude of su24 varies from about 1 to 2 kPa,
Soil stiffness is found to increase due to the thixotropic with values somewhat lower than those suggested by
phenomenon as mentioned in the above section. The Wood (1990), who did not consider the thixotropic hard-
increase in the strength caused by the thixotropy was also ening effect and recommended 2 kPa of su at wL.
measured with a laboratory vane test on ten different
specimens, as depicted in Fig. 5. It is interesting to note 4.3. Comparison between shear modulus and undrained
that only some restricted conditions show a slight increase shear strength
in su with time, unlike thixotropically gained G illustrated
in Fig. 2(a) and (b). The constant su with time is observed The relationship between thixotropic hardening G and su
for some samples with a higher water content than the at various elapsed times is shown in Fig. 7 together with
liquid limit, but as previously demonstrated, G did increase that obtained from the cement-treated soil (CTS) material
even at these conditions. This interesting finding will be proposed by Seng and Tanaka (2011). They found that the
discussed in the following section in more detail. G and su relation of CTS can also be applicable to most of
The values of su measured at 24 h (su24) are plotted with the natural clays found throughout the world with su
normalized w/wL in Fig. 6, in the same manner as G24 varying from 10 to 150 kPa. It is observed in Fig. 7 that at
shown in Fig. 3(a). An almost linear correlation between very high water content corresponding to low initial
su24 in the logarithm scale and w/wL ratio was also strength, su remains constant until a certain time, unlike
identified, but more remarkable scatters and a slightly G. After G reaches a certain value, su starts to increase and
the relation of G and su seems to approach the same line as
11 CTS. Indeed, Seng and Tanaka (2011) reported that even
F1 (40.9%; 0.84)
10 F2 (52.5%; 1.08)
CTS material behaves in a similar way when the strength
of the CTS is extremely small. However, when the strength
Undrained Shear Strength, su (kPa)

K1 (62.3%; 1.00)
9 K2 (60.6%; 0.97)
K4 (62.1%; 1.00) of the CTS is greater than 1 kPa, the G and su correlation
8
K5 (46.0%; 0.74) for each sample forms a linear function, unlike very soft
7 N1 (54.0%; 0.93)
H1 (241.3%; 0.98) clays which show a monotonic increase. Both behaviors,
6 T1 (134.3%; 1.21) the constant values and slow increases of su, are quite
T2 (149.3%; 1.35)
5 interesting and might be associated with either viscosity or
4 the strain rate effect, which appears to be an important
3
factor governing soil strengths especially when material
remains soft. Further investigation is necessary, however,
2
to confirm this presumption.
1
An alternative explanation for the constant su until
0 certain times may be that it can be attributed to the order
0.1 1 10 100 1000
Time (hr) of the strain level, as presented by Seed and Chan (1957)
and Mitchell (1960). They investigated the thixotropic
Fig. 5. Variation in su with time.

10000
10 CTS (Seng and Tanaka, 2011)
Undrained Shear Strength at 24hrs, su24 (kPa)

72
F1 (40.9%; 0.84) 120
24
F2 (52.5%; 1.08) 10
0.5
K1 (62.3%; 1.00) 1000
Shear Modulus, G (kPa)

K2 (60.6%; 0.97)
144
K4 (62.1%; 1.00)
227.5 24 96
K5 (46.0%; 0.74) 168 18
N1 (54.0%; 0.93) 48 F1 (40.9%; 0.84)
4
24 F2 (52.5%; 1.08)
H1 (241.3%; 0.98) 100 20 288
5 0.5 K1 (62.3%; 1.00)
T1 (134.3%; 1.21) 72 120
1 T2 (149.3%; 1.35) 24 11 K2 (60.6%; 0.97)
168
120 2 K4 (62.1%; 1.00)
96
0.5
10 24 K5 (46.0%; 0.74)
10
0.5
N1 (54.0%; 0.93)
H1 (241.3%; 0.98)
T1 (134.3%; 1.21)
T2 (149.3%; 1.35)
1
0.1 1 10 20
0.1 Undrained Shear Strength, su (kPa)
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
Normalized w/wL Fig. 7. Relationship between G and su of very soft clays and CTS (after
Seng and Tanaka, 2011). Numbers with symbols indicate the elapsed
Fig. 6. Relationship between su and normalized w/wL. time (h).
 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345 341

strength of compacted clays using triaxial compression
tests and reported that the strength ratio measured at
smaller strains appears much greater than at large strains.
The strain level measured by the bender element test,
which was adopted in this study, is less than 0.001%, while
the strength obtained by vane test corresponds to large
strain, where the progressive failure to attain the peak
strength may destroy weak bounds created by thixotropy.
Another reason for the constant su may be the insertion of
the vane blade disturbs the samples, and therefore the
validity of the results. When a specimen is too soft, the
insertion of vane blade might destroy the particles bound-
ing created by cement or thixotropy.
5. Shear modulus measurement under low consolidation
pressure
Soil samples, namely the Kasaoka, NSF, and Hachir-
ogata clays, were mixed with different initial water con-
tents and consolidated under very low pressures as small as
2–20 kPa. Bender elements equipped inside the cell were
used to monitor the shear modulus developed during the
consolidating process, focusing mainly on G at the end of
primary consolidation (EOP) and after EOP, i.e., during
the secondary consolidation. In this paper, the time at
EOP was obtained by the Taylor method, i.e., the root
2, 4, 8, 16, and 20 kPa. Using the same input sine wave
signal with 1 kHz frequency as depicted in Fig. 1, the
arrival time of the received waves become faster as the
consolidation pressure is increased. In this case, the travel
distance decreases with settlement; however, the measured
Vs was found to increase. Interestingly, no compressional
waves appear in front of the shear waves in Fig. 1, but the
large magnitude of the reflected waves, as encircled in the
figure, was identified in accordance with the decreases in
arrival time.
Fig. 9(a) and (b) presents the relationships between the
void ratio (e) and the vertical effective stress (s0v ) for
Kasaoka and Hachirogata clays, which were prepared with
two different initial water contents (wi). As can be seen
from the figures, the specimens with higher wi (for instance,
K.C2 and H.C2) show larger void ratios at each loading
step than those with lower wi. This indicates that e cannot
be determined by s0v alone since it is influenced by wi. This
2.2
2.1 K.C1 (63.2%)
K.C2 (129.6%)
2.0
1.9
Void Ratio, e

1.8
time method.
1.7

5.1. Shear modulus measurement at EOP 1.6

1.5
Fig. 8 shows a typical example of the shear wave velocity
(Vs) measurement at EOP, for Kasaoka sample with 1.4
129.6% of initial water content, at five loading levels: 1.3

1.2
0 2 4 6 8 10 12 14 16 18 20 22 24
10
Voltage (V)

Vertical Effective Stress, 'v (kPa)

Input Signal
0
Sine 1 kHz
8.5
-10
H.C1 (319.0%)
2 kPa 8.0 H.C2 (446.0%)

7.5
4 kPa
Void Ratio, e

7.0

8 kPa 6.5

6.0
16 kPa

5.5
20 kPa
5.0
0 2 4 6 8 10 12
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (ms) Vertical Effective Stress, 'v (kPa)

Fig. 8. Shear wave velocity measurement under consolidation process for Fig. 9. Relationship between e and s0v : (a) Kasaoka clay and (b)
Kasaoka clay with 129.6% of initial water content. Hachirogata clay.
342 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345

 
tendency, which has been reported by many researchers, e
G ¼ 41; 600 0:67 s0 0:5 ðin kPaÞ
becomes prominent when s0v is very small. Nevertheless, 1þe v
the shear modulus at EOP (GEOP) of the same clay appears
ðShibata and Soelarno; 1978Þ ð5Þ
to be approximately equal for all tested soils, as illustrated
in Fig. 10(a) and (b) for Kasaoka and Hachirogata clays, G ¼ 5000e1:5 s0v 0:5 ðin kPaÞ ðShibuya and Tanaka; 1996Þ ð6Þ
respectively. It may be concluded that G is mainly
 0:2  0:6
governed by pressure, and the effect of void ratio on G 2 1þ OCR0:5
can be ignored. G ¼ 20; 000w0:8
L OCR s0v0:8
3 3
The estimation of the G value from fundamental para- ðin kPaÞ ðKawaguchi and Tanaka; 2008Þ ð7Þ
meters is quite a challenging topic in the geotechnical
0
engineering field and its formulation has been extensively where p is the mean effective stress; is the vertical s0v
proposed by many researchers. The following are examples effective stress; OCR is over consolidation ratio.
(Hardin and Black, 1969; Shibata and Soelarno, 1978; Eq. (4) proposed by Hardin and Black (1969) is well
Shibuya and Tanaka, 1996; Kawaguchi and Tanaka, known, but this formula was originally derived for prediction
2008): of clean sand and found to be overestimated for very soft
clays used in this study. As for Eq. (5), if the test results from
ð2:97eÞ2 00:5 the present study under very low consolidation pressures, i.e.,
G ¼ 3270 p ðin kPaÞ ðHardin and Black; 1969Þ under large e are applied to this formula, the calculated G
1þ e
ð4Þ becomes negative. Formulae using faction of e, including
Shibuya and Tanaka (1996), is not suitable because G is not
affected by measured e, as previously mentioned. The
3600 formula proposed by Kawaguchi and Tanaka (2008) is one
of a few equations without using the e parameter, and
3200
showed a successful applicability to both reconstituted and
Shear Modulus at EOP, GEOP (kPa)

2800 in situ soils. However, they did not prove its validity to the
low confining stress conditions dealt with in this study. To
2400 examine the applicability of their equation to G measured
2000 under low consolidation pressures, the comparison is made as
illustrated in Fig. 11. The consistent results are identified for
1600 Hachiragata and NSF clays, though they somewhat over-
estimate the value of G for Kasaoka clay.
1200
Since GEOP is the shear modulus after the dissipation of
800 the excess pore water pressure at a constant effective stress,
K.C1 (63.2%) it is attained or increased due to changes in the effective
400
K.C2 (129.6%) stress. By referring to the relations between s0v and GEOP in
0 Fig. 10(a) and the relations between G and time in
0 2 4 6 8 10 12 14 16 18 20 22 24 Fig. 2(a), the process of the thixotropic hardening can be
Vertical Effective Stress, σ'v (kPa) replotted as s0v building up with time. This relation for
Kasaoka clay is shown in Fig. 12, as an example.
2600
2400 8000
2200
Shear Modulus at EOP, GEOP (kPa)

2000 7000
G (Kawaguchi and Tanaka, 2008)

1800
6000
1600
1400 5000
1200 1
1000
4000 1:
800 3000
600
400 H.C1 (319.0%) 2000
200 H.C2 (446.0%) Kasaoka
1000 Hachirogata
0
0 2 4 6 8 10 12 14 16 18 20 22 NSF
0
Vertical Effective Stress, σ'v (kPa) 0 1000 2000 3000 4000 5000 6000 7000 8000
G (This study)
Fig. 10. Relationship between GEOP and s0v : (a) Kasaoka clay and (b)
Hachirogata clay. Fig. 11. Comparison of G with Kawaguchi and Tanaka (2008).
 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345 343

28 constant values of suction for intact clays at about

K1 (62.3%; 1.00) Suction K (49.8%; 0.80)
26 20 min over 1 h of measurements. From these experimental
K2 (60.6%; 0.97)
24 results, it might not be reasonable to conclude that the
K3 (52.0%; 0.84)
22 K4 (62.1%; 1.00) thixotropically gained stiffness is completely caused by the
20 K5 (46.0%; 0.74) increase in the suction, though the possibility that the
Effective Stress (kPa)

18 suction created by particles rearrangement may be partly

16 contributed to the increase in G due to thixotropy
14 phenomenon cannot be denied.
12
10 5.2. Shear modulus measurement during secondary
8 consolidation
6
4 The development of G during secondary consolidation
2 (GSC) is also an interesting topic, and has been discussed
0 for decades by researchers, including Anderson and
0.1 1 10 100 1000
Woods (1976), Athanasopoulos and Richart (1983),
Time (hr)
Lohani et al. (2001), and Howie et al. (2002). Most
Fig. 12. Relationship between effective stress and time. researchers showed that the value of GSC increases until
a point where it is almost linearly correlated with loga-
rithm of time. Using data from Anderson and Woods
Suthaker and Scott (1997) pointed out the importance of
(1976) and Lohani et al. (2001), the increment of GSC is
the consolidation effect caused by the self-weight of
plotted in Fig. 13. The pressures applied in their study
sediments for increasing the long-term strength of fine
varied from 35 to 450 kPa, which is a considerably larger
tailing material. Given that particle density is 2.61 g/cm3
range than those in the present study. The GSC/GEOP for
and assuming that the specimen is fully saturated, then the
Anderson and Woods’s data was derived from the square
vertical effective stress at 5 cm depth, which is equivalent
of VsSC/VsEOP ratio (where VsSC and VsEOP are the shear
to the middle point of the mold for measuring G by the
wave velocity during secondary consolidation and at the
bender element test in the present study, is 0.3 kPa for
EOP, respectively) by assuming that the soil density at
w/wL at 0.8 and 0.2 kPa for w/wL at 1.2. The order of these
EOP is equivalent to that during secondary consolidation.
self-weight consolidation pressures are far smaller than s0v
The gradient of GSC/GEOP ranged between approximately
values in Fig. 12. Therefore, although the authors do not
0.05 and 0.30 per log cycle of time, regardless of the soil
completely deny the possibility of the involvement of self-
type and initial water content. Mitchell and Soga (2005)
weight, the consolidation effect during the thixotropic
summarized the variation of DG/G1000 ratios (where DG is
hardening process on the measured G values can be
shear modulus increase in every 10-fold time increase and
neglected in this study.
G1000 is shear modulus at 1000 min) of clay and sand
As briefly explained in the introduction, there is a well-
materials from many researchers. The ratios varied from
known hypothesis that thixotropic hardening is caused by
the rearrangement of soil particles to a stable condition
(for example, Mitchell, 1960). During this process, despite 1.7
the absence of changes in the confining stress, Day (1955) Anderson and Woods (1976)
35 kPa
and Mitchell (1960) suggested that negative pore pressure 1.6
69 kPa
or suction is built up within the soil causing increase in 138 kPa
effective stress. Ideally, the meaning of suction and 1.5
Normalized GSC/GEOP

effective stress should be the same under the fully saturated

conditions. To confirm these results, a suction test using a 1.4
ceramic disc was conducted following the method pre-
sented in Horng et al. (2010). It can be seen from Fig. 12 1.3
that within 120 h of monitoring, the Kasaoka sample with
Lohani et al. (2001)
49.8% (w/wL ¼ 0.80) of water content showed a fluctuated 1.2
200 kPa
increase of suction to a maximum value about 5 kPa, 450 kPa
which is significantly lower than the 14 kPa attained by the 1.1 135 kPa
present thixotropy test for a sample with a water content 300 kPa
of 52.0%. In addition, when water content was added to 1.0
62.6% (w/wL ¼ 1.0), no change in suction was observed. 1 10 100 1000
However, the reliability of the ceramic disc method used in Normalized Time/Time at EOP
this study for longterms measurements is still required for Fig. 13. Relationship between normalized GSC/GEOP with time by EOP
further research, since Horng et al. (2010) obtained from other reported data.
344 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345

0.05 to 0.25. It can be concluded that the GSC/GEOP and
DG/G1000 ratios are somewhat similar.
The experimental results of the secondary consolidation
under low pressure in this study are shown in Figs. 14 and
15, for Hachirogata and NSF clays, respectively. Again,
the average gradients of GSC/GEOP were approximately
0.25 for Hachirogata clay and 0.15 for NSF clay. The
Hachirogata clay had a higher gradient. Kokusho (1987)
suggested that the ratios increased with increases in the
plasticity index. As seen in Table 1, the plasticity index for
Hachirogata clay is greater than that for NSF clay. It can
be therefore concluded that even at very low pressures, the
gradient during the secondary consolidation is the same as
that under the conventional stress range.
It should be noted that the even though the condition of
the secondary consolidation is under a constant effective
1.5
1.4
Normalized GSC/GEOP
stress, the volume is changing, while in the thixotropic
condition, the volume change does not take place. There-
fore, without experimental confirmation, it might be
considered reasonable to assume that the increase rate of
hardening G owing to secondary consolidation should be
greater than that of thixotropy due to the decreasing void
ratio during secondary consolidation. However, the results
from this study show that when the average incremental
gradient of normalized thixotropic G/G24 against time is
compared with that of the general tendency of GSC/GEOP,
the increment stiffness due to thixotropy is much higher
than that during the secondary consolidation. This can be
explained in terms of a rearrangement of interparticles
within the soil microstructure. From the explanation of the
thixotropic mechanism provided by Mitchell (1960) men-
tioned beforehand, Anderson and Woods (1976) proposed
the mechanism of increase in soil stiffness during second-
ary consolidation as the energy imbalance within soil
structure took place at the end of primary consolidation,
then particles rearrange with time until an equilibrium
state, resulting in a stronger bound or greater rigidity.
However, based on the results presented in this study, the
small gradient of incremental G which undergoes second-

1.3 ary consolidation might be caused by pressure which

constrains the particle from being arranged freely, unlike
under the unconfined stress condition (thixotropy). In
1.2 H.C1 (319%) addition, the applied pressures which act as induced energy
2 kPa to the soil mass from the beginning of consolidation
4 kPa process are not removed, leaving a slight imbalance in
1.1 8 kPa energy caused by the dissipated pore water pressure,
16 kPa meaning that a small amount of stiffness is left to be
20 kPa gained. Another possibility is that the continuous change
1.0 in volume during secondary consolidation may well
1 10 100 destroy the interparticle bounding, and, as a consequence,
Normalized Time/Time at EOP the rate of increase in G is noticeably low.
Fig. 14. Normalized GSC/GEOP and time by EOP for Hachirogata clay.
6. Conclusions

The experimental results of thixotropic phenomenon

1.8
measured for various clays are presented and compared
N.C2 (105.2%)
1.7 with values obtained by consolidation tests at low pres-
2 kPa
sures. The main conclusions which can be drawn are as
4 kPa
1.6 8 kPa
follows:
Normalized GSC/GEOP

16 kPa
1.5
20 kPa (1) The bender element test is a powerful tool and an
appropriate method for evaluating the thixotropic
1.4 hardening stiffness of very soft clays, since it is able
to detect even small changes in G with an extremely
1.3
small strain.
(2) G at 24 h for wL is around 200 kPa.
1.2
(3) Regardless of the type of soil, thixotropy affects the
1.1 clay most strongly at around the liquid limit state and
becomes less remarkable at lower and higher water
1.0 contents.
1 10 100 1000 (4) The correlation between G and su for very soft clays
Normalized Time/ Time at EOP
is analogous to that of cement-treated soil proposed
Fig. 15. Normalized GSC/GEOP and time by EOP for NSF clay. by Seng and Tanaka (2011). Additionally, similar
 S. Seng, H. Tanaka / Soils and Foundations 52 (2012) 335–345
behavior is recognized at very low strengths, where su
appears constant while G increases.
(5) The increment of the shear modulus under secondary
consolidation is relatively low compared with that
developed during the thixotropic process. It is sug-
gested that the difference may due to the different
constraint pressures between the secondary consolida-
tion and thixotropic condition, i.e., the free arrange-
ment of soil particles is prevented under the former
condition, while under the thixotropic condition, the
soil particles can be freely arranged. Alternatively, the
continuous changes in volume during secondary con-
solidation destroy the particle bounding.
From the present study, the importance of the thixo-
tropic hardening phenomenon for understanding soil
behavior with a high water content is clear. It is well
known that the void ratio or water content in situ is much
larger than that predicted by laboratory consolidation tests
based on the e-logp0 relation at the EOP (for example,
Tanaka et al., 2004). If we consider the secondary
consolidation effect, in situ void ratio should be smaller
than that predicted from the e-logp0 relation at the EOP.
To cope with such inconsistency, Burland (1990) has
proposed a concept of sedimentation consolidation line
(SCL), considering the cementation or fabric effect.
Further studies on thixotropy are expected to provide an
important key to understanding such a phenomenon.
Acknowledgments
This research was supported by the Program for Pro-
moting Fundamental Transport Technology Research
from the Japan Railway Construction, Transport and
Technology Agency (JRTT).
References
Anderson, D.G., Woods, R.D., 1976. Time-dependent increase in shear
modulus of clay. Journal of the Geotechnical Engineering Division,
ASCE 102 (GT5), 525–537.
Athanasopoulos, G.A., Richart, F.E., 1983. Effect of creep on shear
modulus of clays. Journal of the Geotechnical Engineering Division,
ASCE 109 (10), 1217–1232.
Barnes, H.A., 1997. Thixotropy—a review. Journal of Non-Newtonian
Fluid Mechanics 70 (1/2), 1–33.
Burland, J.B., 1990. On the compressibility and shear strength of natural
clays. Géotechnique 40 (3), 329–378.
345
Day, P.R., 1955. Effect of shear on water tension in saturated clay,
Annual Reports I and II. Western Regional Research Project, W-30.
Hardin, B.O., Black, W.L., 1969. Vibration modulus of normally
consolidated clay. ASCE 95 (SM6), 1531–1537.
Horng, V., Tanaka, H., Obara, T., 2010. Effects of sampling tube
geometry on soft clayey sample quality evaluated by nondestructive
methods. Soils and Foundations 50 (1), 93–107.
Howie, J.A., Shozen, T., Vaid, Y.P., 2002. Effect of ageing on stiffness of
very loose sand. Canadian Geotechnical Journal 39 (1), 149–156.
Kawaguchi, T., Mitachi, T., Shibuya, S., 2001. Evaluation of shear wave
travel time in laboratory bender element test. In: Proceedings of the
15th ICSMGE 1, 155–158.
Kawaguchi, T., Tanaka, H., 2008. Formulation of Gmax from reconsti-
tuted clayey soils and its application to Gmax measured in the field.
Soils and Foundations 48 (6), 821–831.
Kokusho, T., 1987. In-situ dynamic properties and their evaluations. In:
Proceedings of the 8th ARC on SMFE, Kyoto 2, 215–240.
Lohani, T.N., Imai, G., Tani, K., Shibuya, S., 2001. Gmax of fine-grained
soils at wide void ratio range, focusing on time-dependent behavior.
Soils and Foundations 41 (5), 87–102.
Mewis, J., 1979. Thixotropy—a general review. Journal of Non-New-
tonian Fluid Mechanics 6 (1), 1–20.
Mewis, J., Wagner, N.J., 2009. Thixotropy. Advances in Colloid and
Interface Science 147–148 (1), 214–227.
Mitchell, J.K., 1960. Fundamental aspects of thixotropy in soils.
Journal of the Soil Mechanics and Foundations Divisions, ASCE 86
(3), 19–52.
Mitchell, J.K., Soga, K., 2005. Fundamentals of Soil Behavior. John
Wiley & Sons, Inc., New Jersey.
Osipov, V.I., Nikolaeva, S.K., Sokolov, V.N., 1984. Microstructural
changes associated with thixotropic phenomena in clay soils. Géo-
technique 34 (2), 293–303.
Seed, H.B., Chan, C.K., 1957: Thixotropic characteristics of compacted
clays. In: Proceedings of the American Society of Civil Engineers, vol.
83 (SM4), pp. 1–35.
Seng, S., Tanaka, H., 2011. Properties of cement-treated soils during the
initial curing stages. Soils and Foundations 51 (5), 775–784.
Shibata, T., Soelarno, D.S., 1978. Stress–strain characteristics of clays
under cyclic loading. JSCE 276, 101–110.
Shibuya, S., Tanaka, H., 1996. Estimate of elastic shear modulus in
Holocene soil deposits. Soils and Foundations 36 (4), 45–55.
Skempton, A.W., Northey, R.D., 1952. The sensitivity of clays. Géotech-
nique 3 (1), 30–53.
Suthaker, N.N., Scott, J.D., 1997. Thixotropic strength measurement of
oil sand fine tailings. Canadian Geotechnical Journal 34 (6), 974–984.
Tanaka, H., Kang, M.S., Watabe, Y., 2004. Ageing effects on consolida-
tion properties –Based on the site investigation of Osaka Pleistocene
clays. Soils and Foundations 44 (6), 39–51.
Viggiani, G., Atkinson, J.H., 1995. Interpretation of bender element tests.
Géotechnique 45 (1), 149–154.
Wood, D.M., 1990. Soil Behavior and Critical State Soil Mechanics. Press
Syndicate of the University of Cambridge, USA.
Yamashita, S., Kawaguchi, T., Nakata, Y., Mikami, T., Fujiwara, T.,
Shibuya, S., 2009. Interpretation of international parallel test on the
measurement of Gmax using bender element. Soils and Foundations 49
(4), 631–650.