6th International Conference on Earthquake Geotechnical Engineering

1-4 November 2015

Christchurch, New Zealand

Effectiveness of ‘Gel-Push’ Sampling Technique to Retrieve

Undisturbed Sandy Specimens for Liquefaction Test

Y. Umehara 1, G. Chiaro 2, T. Kiyota 3, Y. Hosono 4, Y. Yagiura 5 and H. Chiba 6

ABSTRACT

In recent years, the advanced ‘Gel-Push’ (GP) sampling technique was developed to collect
undisturbed sandy soil samples. In this study, the quality of samples collected by GP sampler
and conventional triple-tube (TB) sampler were evaluated and compared with each other. Soil
samples were collected from two sandy soil layers (reclaimed and alluvial deposits) in
Mihama ward, Chiba City, Japan, where severe liquefaction was observed during the 2011
‘Off the Pacific Coast of Tohoku’ Earthquake. Based on the changes in void ratio (density
disturbance) and shear wave velocity (soil fabric disturbance) between field and laboratory,
the sample quality was evaluated. It was demonstrated that the GP sampler performed slightly
better than the TB sampler for the tested samples, since it is able to better minimize the soil
fabric disturbance during the sampling process. However, there was no significant difference
between the two techniques in the results of liquefaction resistance obtained by the triaxial
tests.

Introduction

To investigate liquefaction properties of sandy soils, laboratory tests on undisturbed samples

are usually performed. However, it is difficult to describe the actual soil behavior through the
laboratory tests because the tested samples can be easily disturbed during the sampling
process. Yoshimi et al. (1994) reported that samples collected by the conventional tube
sampling (TB) technique are more disturbed than frozen samples and it greatly affects the
evaluation of liquefaction resistance. However, in a practical sense, the conventional TB
method has been widely used because the freezing sampling method is very expensive.

Over the last decade, the advanced ‘Gel-Push’ (GP) sampling technique has been
progressively developed and successfully employed in New Zealand (Taylor et al., 2012),
Taiwan and Japan (Chen et al., 2014). In general, GP allows collecting samples with high
sampling quality through the use of a lubricant gel that reduces the friction between the tube
wall and soil sample. Yet, a comprehensive discussion on the qualitative assessment of GP
samples for liquefiable soils has not been made yet.

In this paper, the sample quality was evaluated by using comparisons between void ratio and
shear wave velocity directly evaluated in-situ and those measured in the laboratory on
undisturbed samples extracted by two different sampling techniques, GP sampler and

1
Graduate Student, Dept. of Civil Engineering, University of Tokyo, Tokyo, Japan, umehara1@iis.u-tokyo.ac.jp
2
Lecturer, Dept. of Civil and Natural Resources Eng., University of Canterbury, Christchurch, New Zealand,
gabriele.chiaro@canterbury.ac.nz
3
Associate Professor, Institute of Industrial Science, University of Tokyo, Tokyo, Japan, kiyota@ iis.u-
tokyo.ac.jp
4
Former Project Researcher, Institute of Industrial Science, University of Tokyo, Tokyo, Japan
5
Executive Managing Director, Kiso-Jiban Consultants Co. Ltd., Tokyo, Japan
6
Manager, Kiso-Jiban Consultants Co. Ltd., Tokyo, Japan
conventional TB sampler. For this purpose, borehole RI-density logging and P-S logging
were performed in the field and void ratio and shear wave velocity measurements were made
in the laboratory on triaxial samples. Finally, the effects of sample quality on the liquefaction
properties of soils were evaluated.

Gel-Push Sampling Technique

As aforementioned, the main purpose of this study was to evaluate the quality of GP samples
and examine the effectiveness of this sampling method compared to the conventional TB
technique. The GP sampler was introduced by Kiso-Jiban Consultants Co. Ltd. for obtaining
high quality undisturbed soil samples. As described in detail by Chen et al. (2014), the GP
sampling technique was first developed in Japan to retrieve gravel material as an alternative
to the costly ground freezing method. Then, in an attempt to obtain undisturbed sandy soil,
the GP sampler was modified to accommodate a thin wall tube inside the sampler to become
a triple-tube system. The GP sampler was aimed to allow a water-soluble polymeric lubricant
(gel) to seep into the tube wall while penetrating the tube into the soil by hydraulic pressure.
Moreover, the sampler was equipped with a cutter attached to the guiding tube to permit
smooth penetration, and a catcher fixed at the bottom of the thin wall tube to prevent the soil
sample from falling out during uplifting. Due to a very small amount employed, the
polymeric gel contaminates only a limited superficial portion of the sample. As a result, the
GP sampler can effectively reduce the wall friction, so that good quality sensitive sandy
samples can be recovered. A schematic illustration of GP sampler operation is shown in
Figure 1a. Essentially, the GP sampler is (i) lowered down a cased hole and then (ii) pushed
into the undisturbed soil. Next, (iii) the catcher is closed and, finally, (iv) the undisturbed
sample is brought to the surface.

Note that, depending on the type of soil, different GP samplers are available. In this study
GP-S sampler (Figures 1b and 1c) was used to obtain clay or clayey soil samples, while a
rotary sampler (GP-Tr) was employed to obtain soil samples.

(i) (ii) (iii) (iv)

(a) (b) (c)

Figure 1. Field instruments used in this study: (a) Illustration of GP sampler operation (from
Lee et al., 2012); (b) GP-S sampler; and (c) details of GP-S sampler cutting shoe (top) and a
sample surrounded by gel lubricant (bottom)
 Sampling Location

The samples used in this study were collected from two sandy soil layers (reclaimed and
alluvial deposits) in Mihama ward, Chiba City, Japan (Figure 2), where severe liquefaction
was observed during the 2011 ‘Off the Pacific Coast of Tohoku’ Earthquake. Mihama ward is
located in the western part of Chiba City along the coast of Tokyo Bay and consists entirely
of reclaimed ground. It was reclaimed by dredged sand and sandy silt taken from the seabed
of the nearby Tokyo Bay. Immediately after the 2011 quake, Nakai and Sekiguchi (2011)
carried out an exhaustive survey of ground damage in Mihama ward, which revealed that due
to liquefaction a huge amount of sand boiling, ground deformation, tilting and subsidence of
buildings were found in almost all zones of Mihama ward.

Figure 2. Map of liquefaction caused by 2011 quake in Mihama ward as reported by Nakai
and Sekiguchi (2011)

In-Situ Investigation

At the investigation site in Isobe area, four boreholes (Figure 2) were performed up to a depth
of 20 m below the ground surface to acquire SPT N-values, shear wave velocity (Vs)
measurements and define the soil profile (borehole No. 1), to collect undisturbed soil samples
by using conventional TB sampling (borehole No. 2) as well as by means of advanced GP
sampling (boreholes No. 3 and No. 4).

The subsoil condition at Isobe site is shown in Figure 3, including Vs measurements, N-SPT
values, density of saturated soil (ρsat) and fines content (Fc). Essentially, ground consists of a
reclaimed deposit (created by hydraulic filling of dredged marine soils) underlaying a natural
soil deposit of recent fluvial origin (Alluvial layer). The thickness of the reclaimed soil is
about 8 m.
• Fsc: reclaimed sand with fines (Vs = 80-140 m/s; N <10 and Fc = 12-24%, non-plastic
fines);
• Fc2: reclaimed clayey soil (Vs = 100 m/s; N = 0 and Fc > 70%, PI = 35-82%);
• As1: alluvial sand (Vs = 120-160 m/s; N = 10-20 and Fc > 60%, non-plastic fines);
• Ac2: alluvial clay (Vs = 110-160 m/s; N < 4 and Fc > 80%, PI = 35%);
• As2: alluvial sand (Vs = 150-200 m/s; N = 10-15 and Fc = 15%, non-plastic fines).
Based on this geotechnical data, it is expected that the two layers Fc2 and Ac2 containing
plastic fines (i.e. clay) as well as the dense sand with high fines content As1 are less prone to
liquefaction; thus, likely did not liquefy during the 2011 earthquake. On the contrary, the two
loose sands with low content of non-plastic fines Fsc and As2, are highly vulnerable against
liquefaction. For this reason, in the laboratory, a series of liquefaction triaxial tests with
dynamic measurement were conducted on specimens of Fsc and As2 soils.
Vs (m/s) N-value ρsat (g/cm3) Fc (%)
0 100 200 300 400 0 10 20 30 40 50 1 2 3 0 20 40 60 80 100
W. L. 0

Fsc
5
Fc2

As1
Depth (m)

10
Ac2

As2
15

20

Shear wave velocity SPT blow count Saturated density Fines content

Figure 3. Subsoil condition and field measurements at Isobe site

Laboratory Tests Method

The procedure of the laboratory test to evaluate the sample quality and the undrained cyclic
strength is summarized as follows. After sampling, the specimens were careful extruded from
the sample tube and trimmed to be accommodated in the triaxial apparatus (i.e. specimen size
of H =10 cm and ϕ = 5 cm). To ensure full saturation (i.e. B-value ≥ 0.96), a back pressure of
200 kPa was applied. Undrained cyclic shearing was then conducted at a frequency of 0.1 Hz
on specimens isotropically consolidated at different confining pressures representative of
field stress conditions. Note that, to ensure a better quality of test results, two parallel series
of tests were performed in the Institute of Industrial Science (IIS), University of Tokyo, and
in the geotechnical laboratory of Kiso-Jiban Consultants Co. Ltd.

Shear wave velocity measurements were made prior to cyclic loading using two equivalent
dynamic measurement devices. In the IIS, an S-wave was generated by creating a torsional
moment through a pair of actuators mounted on the top cap, while a couple of receivers
(accelerometers) glued on the membrane were used to detect the received S-wave (Figure 4a).
On the other hand, in the Kiso-Jiban Consultants geotechnical laboratory, a single actuator
placed on the top cap was used to produce S-waves and a single receiver placed underneath
the pedestal cap was used to capture the received signal (Figure 4b). In both cases, from the
analysis of the wave form, Vs was calculated by the rising-to-rising distance and the measured
travelling time (Kiyota et al., 2009, among many). The transmitted wave consisted of a
solitary sinusoidal wave having a frequency of 1 or 5 kHz. Typical received S-waves are
shown in Figure 4. The travel time was taken as the point of first zero crossing.
 1.2 1.2
Frequency = 1kHz Frequency = 5kHz

0.8 0.8
∆t
0.4 0.4

Voltage (v)
Voltage (v)
0.0 0.0

t
-0.4 -0.4

-0.8 -0.8
Lower receiver Received wave
Upper receiver Input wave
-1.2 -1.2
0 1 2 0.0 0.2 0.4 0.6 0.8
Time(ms) Time(ms)

(a) (b)

Figure 4. Sketch of dynamic measurement devices employed in this study with typical S-
wave signal: (a) in the IIS, University of Tokyo and (b) in the Kiso-Jiban geotechnical
laboratory

Sample Quality Assessment

By comparing the change in soil properties between in-situ and laboratory measurements at
the same effective stress state, sample quality was evaluated hereafter considering two
important factors:
• Change in void ratio (e), which reflects the change in density; and
• Change in shear wave velocity, which captures the change in soil fabric.

Density Disturbance

In this paper, the change in density (i.e. density disturbance) was evaluated by means of the
ratio between the void ratios evaluated in the laboratory and in the field i.e. eLab/eField. So that
for eLab/eField > 1 the sample becomes loosened during the sampling process. Alternatively, for
eLab/eField < 1 the sample densified. The sample is intact (no disturbance) for eLab/eField = 1.

Soil Fabric Disturbance

As mentioned earlier, the change in Vs would capture the change in soil fabric (i.e. soil fabric
disturbance), but actually it could include also the change in density. To have a clear
separation between the change in soil fabric and that of in density, hereafter Vs was
normalized by the square root of the void ratio function f(e):
𝑉𝑉𝑠𝑠∗ = 𝑉𝑉𝑠𝑠 /�𝑓𝑓(𝑒𝑒) (1)
where f(e) was conveniently chosen as f(e) = e-1.3, as proposed by (Jamiolkowski, 1991). The
change in soil fabric was evaluated by means of the ratio between Vs* evaluated in the
laboratory and in the field i.e. Vs*Lab/ Vs*Field. So that, for Vs*Lab/ Vs*Field > 1 the sample after
sampling has stiffer fabric than in-situ. Alternatively, for Vs*Lab/ Vs*Field < 1 the sample
weaker fabric. The sample is intact (no disturbance) for Vs*Lab/ Vs*Field = 1.

Sampling Quality Assessment

Since the sampling quality can be affected by many factors, to better judge the quality of GP
samples with respect to TB samples, in Figures 5 and 6, comparisons are only presented for
those specimens having similar index properties and retrieved at the same depth (i.e. same
stress state). Separate plots are shown for the alluvial and fill sands.
Figure 5 reports the case of alluvial sand As2 specimens retrieved at three different depths
with SPT-N values of 8-11. It appears that the GP sampler performed slightly better than the
TB. In particular, the GP sampler was able to minimize the soil structure disturbance during
the sampling process compared to the TB sampler. On the other hand, in both cases the
density disturbance was reduced.
2.0 2.0 2.0
alluvial layer alluvial layer alluvial layer
1.8 GL-15.0~16.0m 1.8 GL-16.5~17.0m 1.8 GL-17.0~18.53m
Fc<8.6% (lab sieve analysis) Fc<18.0% (lab sieve analysis) Fc<11.0% (lab sieve analysis)
1.6 N=8~9 1.6 N=9 1.6 N=10~11

1.4 1.4 1.4

Vs*Lab / Vs*Field

Vs*Lab / Vs*Field

Vs*Lab / Vs*Field
1.2 1.2 1.2

1.0 1.0 1.0

0.8 0.8 0.8

0.6 0.6 0.6

Sample quality
Sample quality Sample quality
0.4
e±10%, Vs±20%
0.4
e±10%, Vs ±20%
0.4 e±10%, Vs±20%
e±20%, Vs±40% TB e±20%, Vs ±40% TB e±20%, Vs±40% TB
0.2 GP 0.2 GP 0.2 e±30%, Vs±60% GP
e±30%, Vs±60% e±30%, Vs ±60%
0.0 0.0 0.0
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
eLab / eField eLab / eField eLab / eField

(a) (b) (c)

Figure 5. Quality assessment of undisturbed alluvial sandy samples retrieved by GP and TB

Figure 6 presents the case of filled sand Fsc specimens retrieved at the same depth with SPT-
N value of 4, but having different fines content. In general, it can be said that the TB sampler
produced both a higher soil structure disturbance and a larger density disturbance during the
sampling process compared to the GP sampler. Thus, the GP sampler performed better than
the TB sampler. Note that, due to limited number of data as well as the highly heterogeneous
properties of fill sand compared to the natural sand, additional tests are being currently
performed on fill sand specimens to confirm such findings.
2.0 2.0
reclaimed layer reclaimed layer
1.8 GL-2.5~3.0m 1.8 GL-2.5~3.0m
Fc>25.0% (lab sieve analysis) Fc<14.0% (lab sieve analysis)
1.6 N=4 1.6 N=4

1.4 1.4
Vs*Lab / Vs*Field
Vs*Lab / Vs*Field

1.2 1.2

1.0 1.0

0.8 0.8

0.6 0.6
Sample quality Sample quality
0.4 e ±10%, Vs±20% 0.4 e ±10%, Vs±20%
e ±20%, Vs±40% TB e ±20%, Vs±40% TB
0.2 e ±30%, Vs±60% GP 0.2 e ±30%, Vs±60% GP

0.0 0.0
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
eLab / eField eLab / eField

(a) (b)

Figure 6. Quality assessment of undisturbed reclaimed sandy samples retrieved by GP and

TB
 Undrained Cyclic Strength

Comparison of liquefaction resistance curves of sand specimens retrieved by both GP and TB

sampling methods for As2 and Fsc are shown in Figure 7 and 8. As well as the sample
quality assessment, comparisons are only presented for those specimens having similar index
properties and retrieved at the same depth. Note that the liquefaction resistance here was
defined as the number of cycles to cause shear strain double amplitude (εa, DA) of 5%.

As shown in Figure 7, in the case of the alluvial soil As2, it seems that the liquefaction
resistance of GP samples is a bit lower than the TB samples. However, in general, there is not
a significant difference. Note that, the value of Vs by P-S logging was slightly higher for the
TB samples than the GP samples in Figure 7c, thus some differences in liquefaction
resistance can be anticipated. From this analysis, it was found that the difference of sample
quality between the GP and TB sampling methods did not affect the evaluation of
liquefaction resistance for examined alluvial sands with SPT-N value of 8-11.

On the other hand, in the case of the reclaimed layer Fsc shown in Figure 8, data points are
rather dispersed but enclosed within a specific area, irrespective of the used sampling
technique. All samples of both sampling techniques had significant strength for liquefied
ground with SPT-N value of 4. Since there are insufficient data to assess the sample quality,
we need to perform further experiments on the Fsc samples.
0.6 0.6 0.6
alluvial layer alluvial layer TB alluvial layer
GL-15.0~16.0m GL-16.5~17.0m GP TB
GL-17.0~18.53m
0.5 Fc<8.6% (lab seive analysis) TB 0.5 Fc<18.0% (lab seive analysis) 0.5 GP
Fc<11.0% (lab seive analysis)
N=8~9 GP N=9 N=10~11
Cyclic stress ratio, σ'd/2σ'c

Cyclic stress ratio, σ'd/2σ'c

Cyclic stress ratio, σ'd/2σ'c

0.4 0.4 0.4

0.3 0.3 0.3

0.2 0.2 0.2

0.1 0.1 0.1

Axial strain DA=5% Axial strain DA=5% Axial strain DA=5%

0.1 1 10 100 1000 0.1 1 10 100 1000 0.1 1 10 100 1000

Number of cycle, Nc Number of cycle, Nc Number of cycle, Nc

(a) (b) (c)

Figure 7. Liquefaction curves of undisturbed alluvial sandy samples retrieved by GP and TB
0.6 0.6
TB reclaimed layer TB
GP GL-2.5~3.0m GP
0.5 0.5 Fc<14.0% (lab seive analysis)
Cyclic stress ratio, σ'd/2σ'c

N=4
Cyclic stress ratio, σ'd/2σ'c

0.4 0.4

0.3 0.3

> 1000
0.2 reclaimed layer 0.2
GL-2.5~3.0m
Fc>25.0% (lab seive analysis)
0.1 N=4 0.1

Axial strain DA=5% Axial strain DA=5%

0.1 1 10 100 1000 0.1 1 10 100 1000

Number of cycle, Nc Number of cycle, Nc

(a) (b)

Figure 8. Liquefaction curves of undisturbed reclaimed sandy samples retrieved by GP and

TB
 Conclusions

In this paper, the quality of samples collected by two different sampling methods,
conventional triple-tube sampling (TB) and a cutting-edge sampling technique, namely the
“Gel-Push sampling” (GP) was evaluated for samples retrieved from two sandy deposits (i.e.
reclaimed and alluvial).

By using comparisons between void ratio and shear wave velocity directly evaluated in-situ
and those measured in the laboratory, it was demonstrated that the GP sampler performed
better than the TB sampler, since it was able to better minimize the soil fabric disturbance
during the sampling process.

However, there was no significant difference between the two techniques in the results of
liquefaction resistance measured by the undrained cyclic triaxial test. It seems that the
difference of the quality between the GP and TB samples was too small to affect the
liquefaction resistance of examined sand.

References

Chen, C.C., Lee, W.F, Chen J.W and Ishihara, K. Liquefaction potential of non-plastic silty sand, Journal of
Marine Science and Technology (2014); 22 (2): 137-145.
Kiyota, T., Koseki, J., Sato, T. and Tsutsumi, Y. (2009): Effects of sample disturbance on small strain
characteristics and liquefaction properties of Holocene and Pleistocene sandy soils. Soils and Foundations, Vol.
49, No. 4, pp. 509-523.
Jamiolkowski, M.B., Leroueil, S., Lo Presti, D.C.F. Design Parameters from Theory to Practice, Proc. of the
International Conference on Geotechnical Engineering for Coastal Development - Theory and Practice on Soft
Ground (1991): Vol. 2, 877-917.
Nakai, S. and Sekiguchi, T. Damage due to liquefaction during the 2011 Tohoku earthquake. Proc. of the
International Symposium for CSMID (2011): 1-8.
Taylor, M.L., Cubrinovski, M. and Haycock, I. Application of new “Gel-push” sampling procedure to obtain
high quality laboratory tests data for advanced geotechnical analyses, Proc. of the 2012 NZSEE Conference
(2012): Paper No. 123, 1-8.
Yoshimi, Y., Tokimatsu, K. and Ohara, J. In situ liquefaction resistance of clean sands over wide density range,
Geotechniqe (1994): 44(3), 469-494.