Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Measurement of small strain shear modulus of clean and natural sands

in saturated condition using bender element test
Yuanqiang Cai a,b, Quanyang Dong a, Jun Wang b,n, Chuan Gu b, Changjie Xu a
a
Research Center of Coastal and Urban Geotechnical Engineering, Zhejiang University, Hangzhou 310058, PR China
b
College of Architecture and Civil Engineering, Wenzhou University, Wenzhou 325035, PR China

art ic l e i nf o a b s t r a c t

Article history: Bender element (BE) tests of saturated sand have increased interest to researchers currently. However,
Received 27 June 2014 the measurement of small strain modulus from BE tests shows large difference between saturated and
Received in revised form dry conditions. In this study, BE tests of a type of clean sand (Fujian sand) and two types of natural sands
18 December 2014
(Hangzhou sand and Nanjing sand) were performed. For the purposes of comparison, resonant column
Accepted 28 December 2014
(RC) test and torsional shear (TS) test were also carried out on the same specimen. The factors that
influence the determination of the travel time of shear wave in BE tests are discussed and a reliable
Keywords: method for the determination of the shear-wave velocity is obtained. It is found that the shear-wave
Small strain shear modulus velocities Vs of saturated Fujian sand (clean sand) and Hangzhou sand (natural sand) obtained from BE
Bender element test
tests are 5–10% greater than those obtained from RC and TS tests. However, the Vs of saturated Nanjing
Saturated sand
sand (natural sand) obtained from BE, RC and TS tests show good agreement with a maximum difference
Natural sand
Clean sand of about 3%. Sands with various fines contents were also tested in an attempt to explain the differences
Fines content between the two saturated natural sands. Biot's theory accounting for the dispersion of shear wave was
employed to interpret the results of BE tests. The results indicate that the fines content of natural sand
plays an important effect on the hydraulic conductivity, which affects the relative motion between soil
particles and fluid when a high frequency shear wave propagates in the specimen. Based on this, a
method for the determination of small strain shear modulus in BE test was proposed for both saturated
clean sands and natural sands.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction apparatuses such as triaxial and odometer devices so that Gmax

can be determined during these tests [12,7,3,35]. Nevertheless,
The small strain shear modulus (Gmax) of soils is a key considerable uncertainty still remains in the interpretation of BE
parameter in many geotechnical applications such as earthquake test signal. Especially, the determination of first arrival of shear
ground response analysis, design of machine foundations, and wave has always been a critical problem, and many researchers
deep excavations [1,40]. Extensive research works have been done have commented on this difficulty [11,36,29,41,24,39,13]. Largely
to study the Gmax of various soils such as sand, clay, and so on. scattered results in shear-wave velocities estimated from BE test
Traditionally, resonant column (RC) test was most widely used were observed from many literatures. A typical example is shown
because of its high reliability and accuracy ([15,10,9,26,8,20,21]). In by the results of the international parallel BE tests performed on
addition, the torsional shear (TS) test has also been used to the same Toyoura sand [38]. Many factors may have large effects
measure Gmax in the laboratory [16,17,27]. Recently, bender ele- on the signal interpretation such as frequency, interpretation
ment (BE) test, which involves direct measurement of shear waves, methods, and soil types [18,3,23,20,21,13].
has become a popular method for the measurement of Gmax in In order to obtain reliable BE test results, RC tests are usually
laboratory since Shirley and Hampton [32].The significant advan- conducted in parallel for comparison. Madshus (1985) performed
tages of BE test are (1) the strain produced by piezo-electrical both RC and BE tests during the consolidation of five different
transducer is in the order of 10
6, which can be considered as clays, and a good agreement between the Gmax of the two tests
nondestructive; (2) the methods for data interpretation are rela- was obtained. Souto et al. [33] reported that the BE and RC tests
tively simple; and (3) it can be also incorporated into other gave nearly same results for dry sands tested. However, the Gmax of
saturated sand usually showed disagreement between the BE and
RC tests. Youn et al. [41] performed BE, RC and TS tests on two
n
Corresponding author. Tel.: þ 86 577 8668 9687; fax: þ86 577 8668 9611. clean sands. The shear-wave velocity Vs obtained from the BE tests
E-mail address: [email protected] (J. Wang). was about 6–10% greater than that obtained from the RC and TS

http://dx.doi.org/10.1016/j.soildyn.2014.12.013
0267-7261/& 2015 Elsevier Ltd. All rights reserved.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
2 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎

tests for both testing sands. Gu [14] compared the Gmax of

saturated Toyoura sand obtained from BE and RC tests and the
Accelerometer Proximeter
values from BE tests were about 10–15% higher than those from RC
tests. As the behavior of saturated sand specimen becomes much Magnet
LVDT
more complex than that of dry sand specimen, only a few BE tests Coil
have been carried out on saturated sand specimens so far.
Especially, BE tests of saturated natural sand were seldom studied Driving arm
to the best of the authors' knowledge. Therefore, there is a strong Top cap
need for a deep investigation of the measurement of Gmax by BE
test for saturated sands. Bender
In this study, small strain shear modulus measurements of a element
specimen
type of clean sand and two natural sands were conducted under
saturated conditions. All the tests were carried out using an base pedestal
apparatus which allows BE, RC and TS tests to be performed on
an identical specimen, and more reliable comparisons will be
obtained. The factors that influence the interpretation of received
signals in BE tests are discussed, and a reliable method to
determine the travel time of shear wave in BE tests is obtained.
In addition, tests on sands with different fines contents were also
performed in an attempt to explain the differences in the beha-
Fig. 2. Schematic illustration of the testing equipment.
viors of shear wave between the testing sands. The test results
show that Biot's theory should be adopted to consider dispersion
of shear wave in saturated sand BE tests. Based on this, a method
for measuring small strain shear modulus in BE tests is proposed
for both clean sands and natural sands.

2. Test equipment, material and procedure

2.1. Test equipment

Fig. 1 presents a full view of the testing system. Its configura-

tion is schematically shown in Fig. 2. Fig. 3 shows a photo of the
testing equipment on the completion of sample preparation. The
system allows the BE, RC and TS tests being performed on an
identical specimen in parallel. The resonant column is of bottom-
fixed and top-free configuration (i.e., the Stokoe type), which can
offer relatively high available torque and convenient access to the
specimen for control of the effective stress as compared with the
free–free configuration. The dimensions of the element are 11 mm
in width and 1.2 mm in thickness, with a protruding depth of
2.0 mm. Unlike the conventional design, the bender element used
is able to generate not only shear waves (i.e. S-waves) but also
compression waves (i.e. P-waves). Input signals including sinusoi-
dal and square waveforms of different excitation frequencies with
a maximum voltage of 77 V can be generated. A careful calibra-
Fig. 3. Photo after sample preparation.
tion of the system is then performed. A system delay obtained by
direct contact of the element tips was also considered [36,37,13].

Fig. 1. Full view of the testing system. Fig. 4. Grain size distribution curve of three testing sands.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3

Table 2
Basic properties of Nanjing sands with different fines contents.

FC Gs D10 D50 D60 Cu Cc emax emin

0 2.66 0.109 0.173 0.193 1.771 0.892 1.142 0.625

5 2.66 0.104 0.168 0.190 1.827 0.895 1.138 0.584
10 2.66 0.075 0.163 0.185 2.467 1.144 1.131 0.535
14.5 2.66 0.048 0.158 0.180 3.750 1.695 1.124 0.508
20 2.67 0.029 0.152 0.174 6.000 4.579 1.123 0.469

Fig. 5. Grain size distribution curve of Nanjing sand with different fines contents.

Table 1
Basic properties of test clean sand and natural sands.

Sand Gs D10 D50 D60 Cu Cc emax emin

Fujian 2.64 0.263 0.542 0.612 2.327 0.935 0.848 0.519

Hangzhou 2.65 0.078 0.202 0.249 3.192 0.897 1.207 0.583
Nanjing 2.66 0.048 0.158 0.180 3.750 1.695 1.124 0.508

Fig. 6. Typical shear wave signals received by bender element for dry condition
2.2. Test materials (Fujian sand, ei ¼ 0.618, σ3 ¼ 100 kPa).

In this study, a type of clean sand and two types of natural

sands were investigated. The clean sand named Fujian sand is adjusting the input excitation voltage, which confirms the shear
widely used in China. The two types of natural sands come from modulus obtained in the “small strain” level, and a reasonable
Hangzhou and Nanjing, China, which are named Hangzhou sand comparison with shear modulus from BE tests can be obtained.
and Nanjing sand, respectively. The grain-size distributions of the
testing sands are shown in Fig. 4, and their basic soil properties are
summarized in Table 1. In addition, different fines contents of 3. Determination of travel time in BE tests
Nanjing sands were also studied. Fig. 5 shows the grain size
distribution of Nanjing sand with different fines contents. Their The shear wave produced by BE is always accompanied by the
basic soil properties are listed in Table 2. propagation of compression wave and reflected shear wave (a
signal of opposite polarity), which is known as the near-field
2.3. Test procedure effect. The presence of near-field effect affects the shape of the
received signal and it creates a difficulty in finding accurately the
Before specimen preparation, sands were first oven-dried and arrival of shear wave in the BE tests [25,22,39,13]. Jovicic et al. [18]
cooled in sealed containers to remove the potential influence of and Brignoli et al. [7] stated that the near-field effect can be
moisture on the particle surface. Then, cylindrical specimens reduced by increasing the excitation frequency. However, most of
50 mm in diameter and 100 mm in height (H/D ¼2) were prepared the BE tests reported in the literature have been done at a
using a dry tamping method. The method involves pouring the frequency less than 10 kHz.
sand using a funnel without falling height, and compacting it to For a given wave velocity, the wave length decreases continu-
desired density using a tamper in five layers [39]; then deferential ously as the frequency increases. Therefore, the near-field effect is
vacuum pressure was applied to stand the specimen. For saturated usually quantified in terms of the ratio of the wave path length (L)
specimen, CO2 was circulated into the specimen and de-aired to the wave length (λ) [24]. Sanchez-Salinero [30] indicated that
water was percolated from the bottom through the top to saturate the near-field effects would not be significant if L/λ is greater than
the specimen. Back pressure was applied and Skempton B value 2. Arulnathan et al. [3] showed that the near-field effect faded
was checked to guarantee the saturation of the tested specimen when this ratio is greater than 1. Arroyo et al. [2] found that the
(greater than 0.98). After that, isotropic effective confining pres- near-field effect would be less than 5% if L/λ is greater than 1.6. The
sures of 50, 100, 200, and 400 kPa were applied to each specimen experimental observation of Leong et al. [23] suggested a higher
step by step. The specimen was consolidated for 10 min at each limit of 3.33 for L/λ.
stress state. Then, height change for dry specimens and both To better identify the effect of frequency, BE tests with excita-
height and volume changes for saturated specimens were mea- tion frequencies ranging from 2 kHz to as high as 50 kHz have
sured to correct the test results. been performed in this study. As the Vs of sand in dry condition
Sine waves with a wide range of excitation frequencies (2– shows good agreement between BE and RC tests [33], signal
50 kHz) were used as input signals for the BE tests. TS tests with interpretation of BE test results from a dry specimen is carried
the loading frequency of 1 Hz were also performed. Both RC and TS out first. By doing so, a reliable and accurate determination of the
tests were conducted in the strain range of 0.0005–0.001% by travel time of shear wave in BE tests can be achieved.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
4 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 7. Characteristic points for travel time interpretation.

Fig. 9. Typical variation of Vs with excitation frequency using various travel time
determination methods in BE tests for dry conditions (Fujian sand, ei ¼ 0.618,
σ3 ¼ 100 kPa).

Fig. 8. Characteristic points for cross-correlation method.

Shear wave signals measured in the BE tests of Fujian sand in

the dry condition are shown in Fig. 6. The initial signal component
with negative polarity preceding the arrival of major components
represents the near-field effect. It is obvious when the excitation
frequency is small. It reduces with an increase in the excitation
frequency. When the excitation frequencies are larger than 15 kHz,
the waveforms become quite similar. In this case, a slight near-
field effect still exists but larger excitation frequency will con- Fig. 10. Comparison of Vs for different effective confining pressures and void ratios
tribute little to further reduce the near-field effect. of Fujian sand under dry condition.
Because of the uncertainty and difficulty in the determination
of travel time of shear wave, various methods working in both
time domain and frequency domain have been suggested. Among consistent results are obtained when the excitation frequency is in
them, the following three approaches are commonly used: (1) the the range of 10–20 kHz (L/λ E3–8). As the excitation frequency further
start–start (S–S) method, which uses the direct arrival of the shear increases (e.g., larger than 20 kHz), the velocities obtained by the three
wave; (2) the peak–peak (P–P) method, which uses the time methods scattered again. Therefore, increasing excitation frequency
between the first peaks of the input and received waves; and may help reduce the near-field effect, but very high excitation
(3) the cross correlation (C-C) method ([36,20,21]). The idea of the frequency contributes little to the reduction of near-field effect. The
S–S method is to find the first arrival of the shear wave by visual reason is that the near-field effect is produced by both compression
inspection of the received signal. Different characteristic points wave and reflected shear wave, and high excitation frequency
have been proposed as indicators of the first arrival of shear wave contributes only to the reduction of compression waves while it
[39]. The possible characteristic points are shown in Fig. 7, cannot reduce the reflected shear waves.
including the troughs S1 and S3, and the zero intercepts S2 and As the RC test is a more reliable and accurate method for
S4. Likewise, Fig. 8 shows the characteristic points for C-C method. measuring the small strain shear modulus, the Vs obtained from
Fig. 9 plots the shear velocity obtained by making use of the above RC test is also shown in Fig. 9 to verify the results of BE tests. It can
mentioned methods. The velocities determined using any method be seen that the Vs obtained by the P–P method with a frequency
appears to be frequency dependent. They all show a trend that of 15 kHz shows a better agreement with the RC test. Therefore,
increasing excitation frequency will lead to a small increase of velocity. the P–P method with a frequency of 15 kHz is used to determine
When the excitation frequency is smaller than 5 kHz, the velocities the shear-wave velocity in BE tests hereafter. Fig. 10 compares the
obtained by the three methods are considerably scattered. More Vs obtained from the BE, RC and TS tests, and different confining

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 5

pressures and void ratios have been considered. It can be seen that
the Vs obtained by the three tests are close to each other.

4. Results for saturated clean sand

The typical received signals obtained from BE tests of Fujian

sand under saturated conditions are shown in Fig. 11. In the
saturated conditions, the received waveforms are similar to those
obtained in the dry conditions except that they show compression
waves propagating through an effective medium formed by the
soil matrix and pore water before the arrival of shear waves. Fig. 12
shows the variation of shear-wave velocity with excitation fre-
quency. Recalling Fig. 9, it is found that the excitation frequency
imposes a similar effect on the shear-wave velocity. Fig. 13 shows Fig. 13. Variations in Vs with effective confining pressure for Fujian sand under
the variation of shear-wave velocity with confining pressure. As saturated condition.
seen, the Vs obtained from the BE tests are about 5–10% greater

than those obtained from the RC and TS tests, which is similar to

the results found in Youn et al. [41] for two saturated clean sands.
They suggested that the differences between BE and RC tests can
be explained by the dispersion of S-wave in saturated clean sand.
Generally, natural sands contain more fines than clean sands and
the properties of natural sands are very different from those of
clean sands. Nevertheless, the dispersion of S-wave in natural
saturated sands has never been studied by BE tests so far.
Furthermore, BE test results for natural sands especially saturated
natural sands were seldom reported in the literatures. Therefore,
two saturated natural sands were tested. Furthermore, Nanjing
sand with different contents of fines was also tested in this study.

5. Results for two saturated natural sands

Fig. 11. Typical shear wave signals received by bender element for saturated In this section, results of BE, RC and TS tests for two types of
condition (Fujian sand, ei ¼ 0.618, σ3 ¼ 100 kPa). natural sands, i.e., Hangzhou sand and Nanjing sand, were
reported. A series of void ratios and confining pressures have been
considered. The summary of tests conducted is listed in Table 3.
The Vs of the two natural sands is shown in Fig. 14. It can be
seen that the variation of Vs with effective confining pressure is
different for the two natural sands. Fig. 14(a) shows that the Vs of
Hangzhou sand (saturated natural sand) obtained from the BE
tests is about 5–10% greater than that obtained from the RC and TS
tests regardless of the different void ratios and confining pressure
levels. Therefore, the results for Hangzhou sand are similar to
those for the saturated Fujian sand (saturated clean sand) and the
other two clean sands in Youn et al. [41]. On the contrary, the Vs
values of Nanjing sand (saturated natural sand) obtained from BE,
RC and TS tests show good agreement between each other.
Natural sands have larger content of fines than clean sands,
which may lead to some change of the soil properties. Therefore,
sands with different fines contents have also been tested in an
attempt to explain the different results for the two saturated
natural sands. A total of 5 fines contents of Nanjing sand have been
considered, i.e., 0%, 5%, 10%, 14.5% and 20%. The tests performed are
summarized in Table 4. The grain-size distributions and basic soil
properties of Nanjing sand with different fines contents have been
shown in Fig. 4 and Table 2, respectively. Fines content (FC) is
defined as
weight of fines ð o 0:075 mmÞ
FC ¼
weight of specimen
Fig. 12. Typical variation of Vs with excitation frequency using various travel time
determination methods in BE tests for saturated conditions (Fujian sand, ei ¼ 0.618, The Vs of Nanjing sand with different fines contents is shown in
σ3 ¼100 kPa). Fig. 15. The Vs obtained from the TS test is converted from Gmax

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
6 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Table 3
Summary of test series on two types of saturated natural sands.

Sand Initial void ratio Confining pressure σ3 Void ratio Porosity Hydraulic conductivity kh Characteristic frequency fc Tortuosity factor
ei (kPa) e n (1  10
4 m/s) (kHz) α

Nanjing 0.762 50 0.761 0.432 0.0412 163.847 –

100 0.760 0.432 0.0411 164.278 –
200 0.758 0.431 0.0408 165.146 –
400 0.751 0.429 0.0398 168.239 –
Nanjing 0.615 50 0.613 0.380 0.0235 252.514 –
100 0.61 0.379 0.0232 255.004 –
200 0.608 0.378 0.0230 256.684 –
400 0.605 0.377 0.0227 259.236 –
Hangzhou 0.736 50 0.735 0.424 2.5600 2.585 1.680
100 0.732 0.423 2.5332 2.606 1.683
200 0.729 0.422 2.5065 2.628 1.686
400 0.725 0.420 2.4712 2.657 1.690
Hangzhou 0.612 50 0.61 0.379 1.5770 3.753 1.820
100 0.608 0.378 1.5635 3.778 1.822
200 0.606 0.377 1.5501 3.803 1.825
400 0.603 0.376 1.5300 3.841 1.829
   
Note: Hydraulic conductivities of sand were evaluated from kh at initial void ratio using the Kozeny–Carman equation: k1 : k2 ¼ e31 =ð1þ e1 Þ : e32 =ð1 þ e2 Þ .

6. Analysis of test results — Biot's theory on the dispersion of a

shear wave

Biot [5,6] developed the wave propagation frameworks in

porous materials saturated with viscous fluid. The frequency
equation of Biot's theory for S-wave velocity can be written as
follows in a form more suitable for a saturated particulate medium
[34,41]:




ρ
Gsk =V s ρf


mix




ρf αρf =n
jηF=ωK

¼ 0 ð1Þ

where Gsk is the shear stiffness of the granular skeleton, ρmix ¼

(1–n)ρg þ nρf is the total mass density of the saturated particulate
medium, in which ρg is the mass density of the grains (the
minerals that make up the grains); ρf is the mass density of the
fluid (fluid that fills the pores); n is the porosity of the medium; α
is a tortuosity factor; ω is the angular frequency; η is the fluid
viscosity (η ¼10
3 Pa  s is the viscosity of water at room tempera-
ture); K is the absolute hydraulic conductivity of the medium (m2);
and F is a factor which is described in detail in Youn et al. [41]. A
typical S-wave velocity dispersion curve obtained using Eq. (1) is
shown in Fig. 17.
Santamarina et al. [31] stated that when the stiffness of the
skeleton Gsk is much smaller than the stiffness of the material that
makes up the particles Gg, i.e., Gsk⪡Gg, Biot's theory can be
simplified by three characteristic values: characteristic frequency
fc, low-frequency velocities Vlfs , and high-frequency velocities Vhf
s .
The characteristic frequency fc, which separates the low and high
frequency velocities, is defined as
nη ng
Fig. 14. Variations of shear-wave velocity (Vs) with effective confining stress for fc ¼ ¼ ð2Þ
(a) natural Hangzhou sand and (b) natural Nanjing sand under saturated condition. 2πρf K 2π kh

where g ¼9.81 m/s2 is the acceleration due to gravity and kh is the

hydraulic conductivity (m/s).
using the total mass density. The Vs obtained from RC tests is
If frequencies are low (e.g., f o0.1fc), there is no relative move-
similar to that obtained from TS tests regardless of the fines
ment between solid particles and the fluid because of the viscosity
content, whereas the Vs obtained from the BE tests may differ
of the pore fluid. Therefore, the inertia of the fluid can be added to
from that obtained from RC and TS tests as the fines content
the inertia of the soils in terms of mass density, and the shear-
changes. When FC ¼0%, 5% and 10%, the Vs obtained from the BE
wave velocity of the soils is the low-frequency velocity and can be
tests is about 5–10% greater than that obtained from the RC and TS
written as
tests. However, when FC ¼14.5% and 20%, the Vs obtained from BE,
RC and TS tests show good agreement between each other. The sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
maximum difference is about 3% among all the tests. It is more lf Gsk
Vs ¼ ð3Þ
obvious in Fig. 16. ð1
nÞρg þnρf

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 7

Table 4
Summary of test series on saturated Nanjing sand with different fines contents.

FC (%) Initial void ratio ei Confining pressure σ3 (kPa) Void ratio e Porosity n Hydraulic conductivity kh Characteristic frequency fc (kHz) Tortuosity factor α
(1  10
4 m/s)

0 0.754 50 0.751 0.429 1.5800 4.240 1.666

100 0.749 0.428 1.5692 4.263 1.668
200 0.745 0.427 1.5477 4.309 1.671
400 0.739 0.425 1.5159 4.379 1.677
5 0.749 50 0.748 0.428 1.1600 5.763 1.668
100 0.747 0.428 1.1560 5.778 1.669
200 0.743 0.426 1.1402 5.840 1.673
400 0.737 0.424 1.1166 5.936 1.678
10 0.757 50 0.754 0.430 0.2160 31.088 1.663
100 0.752 0.429 0.2145 31.254 1.665
200 0.748 0.428 0.2116 31.589 1.668
400 0.742 0.426 0.2073 32.102 1.674
14.5 0.762 50 0.761 0.432 0.0412 163.847 –
100 0.760 0.432 0.0411 164.278 –
200 0.758 0.431 0.0408 165.146 –
400 0.751 0.429 0.0398 168.239 –
20 0.746 50 0.744 0.427 0.0316 210.886 –
100 0.741 0.426 0.0313 212.598 –
200 0.736 0.424 0.0307 215.496 –
400 0.731 0.422 0.0302 218.454 –
   
Note: Hydraulic conductivities of sand were evaluated from kh at initial void ratio using the Kozeny–Carman equation: k1 : k2 ¼ e31 =ð1 þ e1 Þ : e32 =ð1 þ e2 Þ .

Fig. 15. Variations of shear-wave velocity (Vs) with effective confining stress for Nanjing sand with different fines contents: (a) FC¼ 0%; (b) FC¼ 5%; (c) FC¼ 10%; and
(d) FC¼ 20%.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
8 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 19. Variations of characteristic frequency with different fines contents for
Fig. 16. Relationship between fines content and shear wave velocity under
Nanjing sand.
different confining pressures (50, 100, 200, 400 kPa) and similar void ratio
(e¼0.746–0.757).
Herein, tortuosity factor α is calculated as α ¼ (1 þn)/2n [4,41]. The
tortuosity factor of each test is also listed in Table 4.
As per Eq. (2), it is found that the hydraulic conductivity plays
an important role on the determination of characteristic frequency
fc. Therefore, a series of tests were performed using a conventional
permeameter to measure the hydraulic conductivities of Nanjing
sands with different fines contents. The specimens for perme-
ability tests were prepared in a cylindrical box using the same
preparation method for BE tests initially. Then, vertical perme-
ability tests were performed for each kind of specimen at a
constant head of 16 cm. The hydraulic conductivities kh were thus
determined. In order to obtain the kh at different void ratios, the
Kozeny–Carman equation was applied. The measured hydraulic
conductivities values are listed in Table 4 as well. Fig. 18 shows the
variation of hydraulic conductivity with fines content. It can be
seen that the hydraulic conductivity decreases with increasing
Fig. 17. Typical shear-wave velocity dispersion curve [31]. Gf is stiffness of the fluid. fines content. Especially, the hydraulic conductivity drops rapidly
when the fines content is between 5% and 14.5%. Similar observa-
tions have been found by Nagaraj et al. [28] and Kong et al. [19].
Substituting the measured hydraulic conductivities into Eq. (2),
the characteristic frequencies of sands with different fines con-
tents can be obtained, which are also listed in Table 4. Fig. 19
shows the variation of characteristic frequency with fines content.
Because the hydraulic conductivity decreases with increasing fines
content, the characteristic frequency will increase when the fines
content increases.
Under different confining pressures, the characteristic frequen-
cies vary in the range 4240–4379 Hz and 5763–5936 Hz when
FC ¼0% and 5%, respectively. As Vs is measured at a frequency near
15 kHz in BE tests, which is far greater than the characteristic
frequencies, the Vs obtained from BE tests when FC ¼0% and 5% are
high-frequency velocities. In this case, relative motion will occur
between the soil and fluid because the viscosity effect is relatively
small compared with the inertial effect. The resonance frequencies
of RC tests with different confining pressures fluctuate in a range
of 55.2–81.6 Hz and 53.6–78.2 Hz when FC ¼ 0% and 5%, respec-
tively. The loading frequency used in TS tests is 1 Hz. Hence, the
Fig. 18. Variations of hydraulic conductivity with different fines contents for
Nanjing sand. frequencies of RC and TS tests are much smaller than 0.1fc (i.e.,
576–594 Hz), indicating that the Vs obtained from RC and TS tests
is located in the low-frequency range where the soils and fluid
If f⪢fc, relative motion will occur between the solid particles move together because of the viscosity of the pore fluid. Therefore,
and fluid, and the shear-wave velocity of the soils is the high- the difference in Vs among BE, RC and TS tests is attributed to the
frequency velocity and can be written as different ranges of excitation frequency.
When FC ¼14.5% and 20%, the minimum characteristic frequen-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cies are 163.847 and 210.886 kHz, respectively. Therefore, the
hf Gsk frequencies used in RC (46.5–72.1 Hz), TS (1 Hz) and BE (15 kHz)
Vs ¼ ð4Þ
ð1
nÞρg þ nð1
1=αÞρf tests, which are all smaller than 0.1fc (16.385–21.089 kHz), are

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 9

located in the low-frequency range. In this case, the inertia of the The clean sand tested in this study is better graded than most
fluid is added to the inertia of the soils in terms of mass density. widely used clean sand (e.g., Toyoura sand, Hostun sand and
Hence, the Vs obtained from RC, TS and BE tests show good Ottawa sand), as shown in Fig. 4. Consequently, it has a lower
agreement with a maximum difference of about 3%. hydraulic conductivity and higher characteristic frequency. In
When FC ¼10%, the characteristic frequencies under different other words, most of the widely used clean sands have high
confining pressures are between 31,088 and 32,102 Hz. The fre- hydraulic conductivity and consequently low characteristic fre-
quencies in both RC (51.2–76.5 Hz) and TS (1 Hz) tests are located quencies which are usually smaller than the excitation frequency
in the low-frequency range because they are smaller than 0.1fc used in BE test. Therefore, the mass density as proposed by Biot's
(3109–3210 Hz). However, the excitation frequency in BE test theory should be used to calculate Gmax instead of the total mass
(15 kHz) is between 0.1fc and fc, making the Vs obtained from BE density. Assuming a void ratio of 0.5 for natural sands and an
tests neither high-frequency velocity nor low-frequency velocity.
Therefore, Eq. (1) should be used to calculate the Vs in this case.
Again, the difference in Vs among BE, RC and TS tests is attributed
to the different ranges of excitation frequency.
Recalling the properties of saturated natural Hangzhou sand
and natural Nanjing sand in Table 3, it can be seen that the Vs of
Hangzhou sand obtained from BE tests is a high-frequency velocity
while the Vs obtained from RC and TS is a low-frequency velocity.
However, the Vs of Nanjing sand obtained from BE, RC and TS are
all low-frequency velocities. It therefore explains the different
behavior of the two saturated natural sands in terms of Vs.
The Gmax can be calculated from Vs using Eq. (3) when the
excitation frequency is in the low-frequency region and Eq. (4) is
for the high-frequency region. The variations of Gmax obtained
from BE, RC, and TS tests with effective confining pressure are
plotted in Fig. 20. Good agreement among the results of the three
kinds of tests is observed. Fig. 21 plots the variations of Gmax with
fines content. A decreasing trend of Gmax with increasing fines Fig. 21. Relationship between fines content and shear modulus under different
content is seen for sands with a similar void ratio. confining pressures (50, 100, 200, 400 kPa) and similar void ratio (e ¼0.746–0.762).

Fig. 20. Variations of small strain shear modulus (Gmax) with effective confining stress for Nanjing sand with different fines contents: (a) FC¼ 0%; (b) FC¼ 5%; (c) FC ¼10%;
and (d) FC¼ 20%.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
10 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎

excitation frequency of 15 kHz for BE test, if kh is greater than 50979096 and 50808145, and Natural Science Foundation of
5.2  10
5 m/s (f⪢fc), relative movement will occur between solid Zhejiang Province under Grant nos. Y1090105 and Y1110751. The
particles and fluid. Therefore, Gmax should be calculated from Vs authors would also acknowledge the financial support from
using the mass density as proposed by Biot's theory. If kh is smaller Science and Technology Research Program of Zhejiang province
than 5.2  10
6 m/s (f o0.1fc), no relative movement will occur (Grant no. 2010C33182) and Science and Technology Research
between solid particles and fluid, and the Gmax should be calcu- Program of Wenzhou city (Grant no. S20100057).
lated from Vs using the total mass density.

References
7. Conclusions
[1] Andrus RD, Stokoe II KH. Liquefaction resistance of soils from shear-wave
In this study, the small strain shear modulus of saturated sand velocity. J Geotech Geoenviron Eng 2000;126(11):1015–25.
was investigated, including a type of clean sand and two types of [2] Arroyo M, Wood DM, Greening PD. Source near-field effects and pulse tests in
soil samples. Géotechnique 2003;53(3):337–45.
natural sands. For each kind of sand, wide ranges of void ratios and [3] Arulnathan R, Boulanger RW, Riemer MF. Analysis of bender element tests.
confining pressures were considered. Three kinds of tests, includ- Geotechn Test J 1998;21(2):120–31.
ing BE, RC and TS tests, were performed on the same specimen, [4] Berryman JG. Elastic wave propagation in fluid-saturated porous media.
J Acoust Soc Am 1981;69(2):416–24.
and comparisons of small strain shear modulus obtained from the [5] Biot MA. Theory of propagation of elastic waves in a fluid-saturated porous
three different tests were made. The findings achieved in this solid. I. Low frequency range. J Acoust Soc Am 1956;28(2):168–78.
study are summarized as follows: [6] Biot MA. Theory of propagation of elastic waves in a fluid-saturated porous
solid. II. Higher frequency range. J Acoust Soc Am 1956;28(2):179–91.
[7] Brignoli EGM, Gotti M, Stokoe II KH. Measurement of shear waves in
(1) The excitation frequency influences the determination of laboratory specimens by means of piezoelectric transducers. Geotech Test J
travel time of shear wave in BE tests. Increasing excitation ASTM 1996;19(4):384–97.
[8] Cascante G, Vanderkooy J, Chung W. Difference between current and voltage
frequency helps reduce the near-field effect, but very high
measurements in resonant-column testing. Can Geotech J 2003;40(4):806–20.
excitation frequency contributes little to the reduction of near- [9] Chung RM, Yokel FY, Drnevich VP. Evaluation of dynamic properties of sands
field effect. The results obtained from different methods are by resonant column testing. Geotech Test J 1984;7(2):60–9.
more consistent when the excitation frequency is between 10 [10] Drnevich VP, Hardin BO, Shippy DJ. Modulus and damping of soils by the
resonant column method. Dynamic geotechnical testing. Special technical
and 20 kHz (L/λ E 3–8). publication 654. West Conshohocken, PA: American Society for Testing and
(2) The shear-wave velocities results of Hangzhou sand (saturated Materials (ASTM); 1978. p. 91–121.
natural sand) are similar to those of Fujian sand (clean sand). [11] Dyvik R, Madshus C. Lab measurements of Gmax using bender elements. In:
Khosla V, editor. Advances in the art of testing soils under cyclic conditions.
Its shear-wave velocities obtained from BE tests are about 5– Proceedings of the ASCE conference, Detroit, MI, 24 October 1985. New York:
10% greater than those obtained from the RC and TS tests. On American Society of Civil Engineerings (ASCE); 1985. p. 186–96.
the contrary, the shear-wave velocities of Nanjing sand [12] Fam MA, Santamarina JC. Study of geoprocesses with complementary
mechanical and electromagnetic wave measure ments in an oedometer.
(another saturated natural sand) obtained from BE, RC and Geotech Test J 1995;18(3):307–14.
TS tests show good agreement between each other. [13] Gu X, Yang J, Huang M. Laboratory measurements of small strain properties of
(3) Under all different effective confining pressures, the shear-wave dry sands by bender element. Soils Found 2013;53(5):735–45.
[14] Gu XQ. Dynamic properties of granular materials at the macro and micro
velocities of Nanjing sand obtained from BE tests are about 5–
scales [Ph.D. thesis]. Hong Kong: The University of Hong Kong; 2012.
10% greater than those obtained from RC and TS tests when [15] Hardin BO, Richart Jr FE. Elastic wave velocities in granular soils. J Soil Mech
FC¼0%, 5% and 10%. However, when FC¼14.5% and 20%, the Found Div ASCE 1963;89(SM1):39–56.
[16] Isenhower WM. Torsional simple shear-resonant column properties of San
shear-wave velocities obtained from BE, RC and TS tests show
Francisco Bay mud [M.Sc. thesis]. Austin, TX: The University of Texas at Austin;
good agreement with a maximum difference of about 3%, 1979.
regardless of the effective confining pressure and void ratio. [17] Iwasaki T, Tatsuoka F, Takagi Y. Shear moduli of sands under cyclic torsional
(4) Frequency dependent Biot's theory was employed to consider shear loading. Soils Found 1978;18(1):39–56.
[18] Jovicic V, Coop MR, Simic M. Objective criteria for determining Gmax from
the dispersion of shear wave in saturated conditions. The bender element tests. Géotechnique 1996;46(2):357–62.
relative movement between solid particles and fluid does exist [19] Kong LW, Li XM, Tian HN. Effect of fines content on permeability coefficient of
as predicted. The fines content of natural sand plays an sand and its correlation with state parameters. Rock Soil Mech 2011;S2:21–6
(in Chinese).
important effect on the hydraulic conductivity, which affects [20] Kumar J, Madhusudhan BN. On determining the elastic modulus of a
the relative motion between soil particles and fluid when a cylindrical sample subjected to flexural excitation in a resonant column
high frequency shear wave propagates in the specimen. apparatus. Can Geotech J 2010;47(11):1288–98.
[21] Kumar J, Madhusudhan BN. A note on the measurement of travel times using
(5) A method to determine the small strain shear modulus Gmax in bender and extender elements. Soil Dyn Earthq Eng 2010;30(7):630–4.
BE test is proposed for both saturated clean sand and natural [22] Lee JS, Santamarina JC. Bender elements: performance and signal interpreta-
sand, which is applicable in both laboratory test and field test. tion. J Geotech Geoenviron Eng ASCE 2005;131(9):1063–70.
[23] Leong EC, Yeo SH, Rahardjo H. Measuring shear wave velocity using bender
For most of the frequently used clean sands which usually
elements. Geotech Test J 2005;28(5):488–98.
have high hydraulic conductivities, the mass density as pro- [24] Leong EC, Cahyadi J, Rahardjo H. Measuring shear and compression wave
posed by Biot's theory should be used to calculate the Gmax in velocities of soil using bender–extender elements. Can Geotech J 2009;46
(7):792–812.
BE test instead of the total mass density. For saturated natural
[25] Lings ML, Greening PD. A novel bender/extender element for soil testing.
sands, the use of the mass density as proposed by Biot's theory Géotechnique 2001;51(8):713–7.
or the total mass density for the calculation of Gmax in BE tests [26] Lo Presti DCF, Jamiolkowski M, Pallara O, Pedroni ACS. Shear modulus and
is dependent on the hydraulic conductivity kh. The boundary damping of soils. Géotechnique 1997;47(3):603–17.
[27] Mancuso C, Vassallo R, d'Onofrio A. Small strain behavior of a silty sand in
of kh is roughly between 5.2  10
5 and 5.2  10
6 m/s. controlled-suction resonant column torsional shear tests. Can Geotech J
2002;39(1):22–31.
[28] Nagaraj TS, Pandian NS, Narasimha Raju PSR. An approach for prediction of
compressibility and permeability behaviour of sand–bentonite mixes. Indian
Geotech J 1991;21(3):271–82.
Acknowledgments [29] Nakagawa K, Soga K, Mitchell JK. Pulse transmission system for measuring
wave propagation in soils. J Geotech Eng ASCE 1996;122(4):302–8.
[30] Sanchez-Salinero I. Analytical investigation of seismic methods used for
The work described in this paper is supported by Natural engineering application [Ph.D. dissertation]. Austin, TX: The University of
Science Foundation of China through Grant nos. 51025827, Texas at Austin; 1987.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i
 Y. Cai et al. / Soil Dynamics and Earthquake Engineering ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 11

[31] Santamarina JC, Klein KA, Fam MA. Soils and waves. New York: John Wiley & [37] Wang YH, Lo KF, Yan WM, Dong XB. Measurement biases in the bender
Sons Ltd.; 2001. p. 238–82. element test. J Geotech Geoenviron Eng 2007;133(5):564–74.
[32] Shirley DJ, Hampton LD. Shear-wave measurements in laboratory sediments. [38] Yamashita S, Kawaguchi T, Nakata Y, MIKAML T, Fujiwara T, Shibuya S.
J Acoust Soc Am 1978;63(2):607–13. Interpretation of international parallel test on the measurement of Gmax
[33] Souto A, Hartikainen J, Özüdǒ gru K. Measurement of dynamic parameters of using bender elements. Soils Found 2009;49(4):631–50.
road pavement materials by the bender element and resonant column tests. [39] Yang J, Gu XQ. Shear stiffness of granular material at small strains: does it
Géotechnique 1994;44(3):519–26. depend on grain size? Géotechnique 2013;63(2):165–79.
[34] Stoll RD. Experimental studies of attenuation in sediments. J Acoust Soc Am [40] Yang J, Yan XR. Site response to multi-directional earthquake loading: a
1979;66(4):1152–60. practical procedure. Soil Dyn Earthq Eng 2009;29(4):710–21.
[35] Valle-Molina C, Stokoe KH. Seismic measurements in sand specimens with [41] Youn JU, Choo YW, Kim DS. Measurement of small-strain shear modulus Gmax
varying degrees of saturation using piezoelectric transducers. Can Geotech J of dry and saturated sands by bender element, resonant column, and torsional
2012;49(6):671–85.
shear tests. Can Geotech J 2008;45(10):1426–38.
[36] Viggiani G, Atkinson JH. Interpretation of bender element tests. Géotechnique
1995;45(1):149–54.

Please cite this article as: Cai Y, et al. Measurement of small strain shear modulus of clean and natural sands in saturated condition
using bender element test. Soil Dynamics and Earthquake Engineering (2015), http://dx.doi.org/10.1016/j.soildyn.2014.12.013i