Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

www.elsevier.com/locate/soildyn

Stone columns as liquefaction countermeasure in non-plastic silty soils

K. Adaliera,*, A. Elgamalb, J. Menesesb, J.I. Baezc
a
Department of Civil and Environmental Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, USA
b
Department of Structural Engineering, University of California at San Diego, La Jolla, CA 92093, USA
c
Hayward Baker, Inc., Santa Paula, CA 93060, USA
Accepted 12 June 2003

Abstract
In many cases densification with vibro-stone columns cannot be obtained in non-plastic silty soils. Shear stress re-distribution concepts
[1] have been previously proposed as means to assess stone columns as a liquefaction countermeasure in such non-plastic silty soils. In this
study, centrifuge testing is conducted to assess the performance of this liquefaction countermeasure. Attention is focused on exploring the
overall site stiffening effects due to the stone column placement rather than the drainage effects. The response of a saturated silt stratum is
analyzed under base dynamic excitation conditions. In a series of four separate model tests, this stratum is studied first without, then with
stone columns, as a free-field situation, and with a surface foundation surcharge. The underlying mechanism and effectiveness of the stone
columns are discussed based on the recorded dynamic responses. Effect of the installed columns on excess pore pressures and deformations is
analyzed and compared. The test results demonstrate that stone columns can be an effective technique in the remediation of liquefaction
induced settlement of non-plastic silty deposits particularly under shallow foundations, or vertical effective stresses larger than about 45 kPa
(1000 psf) in free field conditions.
q 2003 Elsevier Ltd. All rights reserved.
Keywords: Earthquake; Liquefaction countermeasures; Stone column; Centrifuge modeling; Silty soil; Footings

1. Introduction the stiffer columns). When dealing with non-plastic silty

soils, only the third benefit can be expected primarily to
A comprehensive literature review by the authors on mitigate liquefaction [1]. Even with the vibro-flotation
stone columns (gravel drains) revealed that there is a great installation method, densification due to vibrations in silts
need for better understanding of stone column liquefaction can be impractical to achieve. However, several cases,
hazard mitigation mechanisms, particularly when con- published [2] and unpublished, are reported in the industry
structed in silty soils. In many cases densification with as being successful when the stone columns are combined
vibro-stone columns cannot be obtained in non-plastic silty with pre-installation of wick drains at close centers. The pre-
soils. Shear stress re-distribution concepts [1] have been installed wick drains enhance the global permeability of the
previously proposed as means to assess stone columns as a silt to be treated. Subsequent installation with vibro-stone
liquefaction countermeasure in such non-plastic silty soils. columns then benefits from the increased permeability,
In order to address this issue, a centrifuge experimental allowing improved volume change (densification) potential.
program of dynamic centrifuge model experiments was It should be noted that the main purpose of the wick drain
conducted. This paper reports the results of this experimen- installation is to assist the stone columns and not to
tal study. counteract seismic induced liquefaction. For the overall
The possible benefits of stone columns include densifica- system of stone columns and wick drains, due to the very
tion of surrounding non-cohesive soil, dissipation of excess low silt permeability, the drainage benefit is limited in
pore water pressure (EPWP) and re-distribution of earth- effectiveness, especially during strong shaking situations.
quake-induced or pre-existing stress (due to introduction of This paper focuses on the assessment of the stiffening effect
of stone columns and its impact on the response of improved
* Corresponding author. Tel.: þ 1-518-276-8043; fax: þ1-518-276-4833. silty ground. Additional possible stiffening due to soil
E-mail address: adalik@rpi.edu (K. Adalier). densification/compaction during the installation process
0267-7261/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0267-7261(03)00070-8
572 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

was not addressed during the testing phase. Using the prototype scale, obtained from the actual model units
centrifuge at Rensselaer Polytechnic Institute (RPI), Troy, following basic scaling relations [4]. This means that all
NY, USA (http://www.rpi.edu/~dobryr/centrifuge), two linear dimensions, including measured deformations, as well
benchmark model tests were performed first to document as the time in the model, were multiplied by N (where N is
the dynamic response characteristics of a silty stratum with level of gravitational acceleration), and the actual model
and without a surface foundation surcharge. Under the same shaking acceleration and frequency divided by N (N ¼ 50
shaking conditions, the responses of these models, reme- and 63 for Models 1– 2 and Models 3 –4, respectively).
diated with stone columns, was studied and compared to the The first test explored the response of a 7.8 m thick
benchmark models. (prototype scale), pure silt saturated stratum (Model 1,
Fig. 1). Relative density of the silt was about 57%. It was
estimated that centrifuge spin-up reduced this relative
2. Centrifuge testing program density by 3% (this reduction was practically the same for
the test with stone columns). In the second test, a total of 45,
Centrifuge physical modeling has been recognized as an 1.27 m diameter columns were placed (2.5 m center-to-
effective tool in studying prototype behavior of various center) vertically in the model laminar container at
earth structure-soil systems. The main principle in centri- predetermined positions (Model 2, Fig. 2), giving an area
fuge modeling is that a 1=N scale model subjected to a replacement ratio ðAr Þ of 20%. Note that the area replacement
gravitational acceleration of Ng (g is acceleration of gravity) ratio is defined as the area of the stone column to the tributary
experiences the same stress as the prototype. Thus, stress – area per stone column [10]. A main goal of Model 1 and 2
strain relationships at all similar points in the model and tests was to explore the impact of stone columns on lateral
prototype will be equivalent (if the same soil is employed), site stiffness, and the resulting response implications.
and the behavior of the model will mimic the behavior of the In the third test, the response of a 10 m thick saturated silt
prototype. Consequently, with the help of scaling laws [3,4] stratum with a rigid footing surcharge (rigid steel rectangu-
measurements in centrifuge tests under closely controlled lar block applying an average vertical contact pressure of
conditions can be related directly to an equivalent full-scale 144 kPa) was studied (Model 3, Fig. 3). This surcharge
prototype. The centrifuge modeling technique allows soil simulated approximately the vertical pressure transmitted
liquefaction tests to be performed at a conveniently reduced by a 10 –15 story reinforced concrete building. The fourth
scale, and provides data applicable to full-scale problems. test investigated the response of the same system employed
Centrifuge modeling data sets play a major role in in Model 3 but with the inclusion of 36, 1.6 m diameter,
verification and refinement of liquefaction countermeasures stone columns (Model 4, Fig. 4) at pre-determined positions
[5 –7]. Such data also provides a basis for calibration of (2.55 m center-to-center). This configuration provided an
design and computational modeling procedures [8,9]. area replacement ratio of 30% within the instrumented zone
Herein, in an attempt to verify and quantify the possible below the footing. Ground water table was at the soil surface
liquefaction mitigation mechanisms due to deployment of in all tests.
stone columns, a centrifuge testing program consisting of Models 1 and 2 were tested at a 50g gravitational
four separate model tests was conducted. Unless otherwise acceleration field, whereas Models 3 and 4 were tested at
indicated, all dimensions reported in this paper are in 63g: The soil container used in the Model 1 and 2 tests was

Fig. 1. Configuration of Model 1.

 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584 573

Fig. 2. Cross-sectional and plan view of Model 2 configuration containing 45 stone columns with global area of replacement at 20%.

a rectangular, flexible-wall laminar box [11]. This container order to investigate its liquefaction response characteristics
allows relative slip between laminates in order to simulate [12]. EPWP build-up as well as expected levels of
one-dimensional (1D) shear response conditions. The model liquefaction-induced deformations were explored for differ-
dimensions (inner) of this box were 0.254 m wide, 0.458 m ent levels of applied cyclic shear strains [12]. At a Relative
long and 0.254 m high. Models 3 and 4 were constructed in Density ðDr Þ of 60%, the internal friction angle of this silt is
a rigid-wall container. The model dimensions (inner) of this estimated as 258 [12].
container were 0.252 m wide, 0.53 m long and 0.203 m The material representing the stone columns was Nevada
high. Model response was measured by miniature transdu- No. 120 sand. Extensive data including the monotonic and
cers, including horizontal accelerometers (303A03, PCB cyclic response characteristics of this soil has been
Piezotronics), pore pressure transducers (PPT, PDCR81, documented by Arulmoli et al. [13] through various unit
Druck, Inc.), and linear variable differential transformers element tests (this information is archived on the WEB at:
(LVDT MHR500, Schaevitz Engineering). http://geoinfo.usc.edu/gees/velacs/). At elevated gravita-
tional levels (i.e. 50g and 63g for the first and the second
series of tests, respectively) this fine sand simulated coarse
3. Model ground construction sand rather than gravel [14]. However, since attention was
focused on the stiffening effect, this issue was of no concern.
A 100% silt size particulate material (Sil-Co-Sil 120) In addition, the Sil-Co-Sil silt and the Nevada sand were
was employed to construct the ground layer. This material mainly chosen due to availability of laboratory testing
was subjected earlier to a series of cyclic laboratory tests in results documenting their engineering properties. Nevada
574 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

Fig. 3. Cross-sectional and plan view of Model 3 configuration.

Fig. 4. Cross-sectional and plan view of Model 4 configuration containing 36 stone columns with area of replacement at 30%.
 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584 575

120 sand is mainly a uniform, subrounded, clean sand with a about 0.002 m due to centrifuge spin-up were uniform,
mean grain size of 0.15 mm. Water permeability of the sand and were considered tolerable. In addition, placement of
tested conventionally in the laboratory at 1g (at Dr of 65%) instrumentation proved to be reliable, and the PPT indicated
is 5.2 £ 1025 m/s (roughly 600 times more permeable than satisfactory saturation.
Sil-Co-Sil 120 Silt). In view of the scaling laws applicable
to centrifuge experiments, the prototype permeability of the
sand at 50g and 63g were 2.6 £ 1023 and 3.3 £ 1023 m/s, 4. Construction of model stone columns
respectively. Internal friction angle of this sand is estimated
as 378 at Dr of 65% [13]. A procedure was needed for constructing the stone
The employed Sil-Co-Sil 120 silt material is available in columns under conditions of high quality control. Con-
the form of a dry powder. Initially, attempts were made to struction of these columns within an already built silt layer
build the soil layer from a silt-water slurry. This slurry was was deemed to be unreliable for our intended testing
formed by mixing the powder with an appropriate volume of purposes. Initially attempts were made to construct slightly
water in order to attain a uniform saturated mixture. Within cemented model stone columns within cylindrical cardboard
practical ranges of water content (e.g. 45 –60%), a number tubes (compacted to the desired relative density). This
of slurry mixtures were tested for uniformity and consist- procedure would allow for construction of virtually uniform
ency. The resulting saturated silt was found to be too sticky columns that may be placed within the model with a high
and difficult to handle. Using this saturated silt mixture, four level of precision. However, the breakdown mechanism of
trial centrifuge tests were conducted in a rigid soil container this cementation process during dynamic excitation was
to check for degree of saturation and accuracy of virtually impossible to predict, and therefore introduced an
instrumentation placement. As a result, this slurry technique undesirable element of uncertainty.
was discarded due to the following difficulties: (1) accurate An alternative ‘temporary’ cementation procedure was
placement of instruments (accelerometers and PPTs) was investigated next. After a series of calibration trials, the
virtually impossible in terms of location as well as stone column material (Nevada No. 120 sand) was
orientation for directional recording of acceleration, (2) cemented temporarily during construction with a sugar –
uniform samples were difficult to build at the desired water water solution. The sugar content in this solution was 4 – 5 %
content of about 32% and (3) PPT readings were unreliable of the water by weight. The sand-sugar water mixture was
indicating partial soil saturation and/or improper PPT poured and compacted within a cylindrical cardboard
saturation. casing. The casings were then placed in an oven for drying
Thereafter, attempts were made to build an instrumented (at 508C for a duration of one day). After drying, these sand
dry silt layer and undergo a phase of saturation under an columns became brick-like, and the cardboard casings were
applied vacuum (commonly used procedure for saturation removed. The sand columns had a void ratio of about 0.62–
of sand models). The silt powder was rained from a funnel at 0.64 (Dr ¼ 65 –70%), although in the field it is possible to
a sustained height of 0.15 m. Pluviation was interrupted achieve relative densities as high as 90% with crushed stone.
periodically to place instruments in the soil. Mild compac- Lower relative densities were desired in this experimental
tion (by hand tamping using a square wooden block) program to have a Gr -ratio of shear moduli between stone
followed deposition at intervals of 0.02 m in layer thickness. column (Nevada sand) and silt of 5 –6 (Gr ¼ GSC =GS ; where
This compaction decreased the dry soil void ratio in order to GSC is stone column shear modulus, and GS is soil shear
reduce consolidation-settlements during the subsequent modulus). This ratio is a critical parameter for stress
saturation process. The soil models were percolated with concentration or stiffening effects due to introduction of a
carbon dioxide to help dissolve the air in the void space, in stone column system [1].
order to facilitate full saturation by water (which was done During this initial testing phase (Models 1 and 2), it was
under a vacuum of 90 kPa). The water used as pore fluid was desired to entirely focus on the stiffening effect of stone
de-ionized and de-aired. Saturation of the layer under columns (i.e. completely neutralize the column draining
vacuum was performed slowly over a period of 36 –48 h capacity). Consequently, each 0.0254 m-diameter sugar-
(this time frame was established after several trials). cemented column was encased in a 0.0003 m thick latex
Saturation was performed while the model container was membrane. At the column base, the membrane was sealed
secured on the shaker platform. This construction procedure with silicon glue and end-cap circular plastic plates of
was time consuming, but was found to alleviate the 0.005 m thickness and 0.0254 m diameter. This sealing
difficulties associated with slurry construction. For detailed process prevented any water within the silt from moving
descriptions of typical model preparation techniques, into the constructed stone columns. The column/membrane
instrumentation, and testing procedures, the reader is top-end remained open, and allowed for saturation of the
referred to Adalier [14]. Three trial centrifuge laminar-box columns and for solubility of the sugar (nullifying the
experiments were conducted to verify the satisfactory cementation effect).
performance of this model construction method and the For Model 4, the stone columns were not encased with a
installed instrumentation. Consolidation settlements of latex membrane, more accurately simulating actual field
576 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

conditions. In all cases, due to the very low silt permeability, pre-shake locations. In the following, all test results are
the stone columns did not increase overall drainage or presented and discussed in prototype units, unless stated
decrease the EPWP build-up rate during the shaking phase otherwise. Dynamic responses of Model 1 (model ground
in any appreciable way. Therefore, for all tests any change without surcharge), Model 2 (model ground with 45 stone
in the behavior of the remediated ground (relative to the columns and without surcharge), Model 3 (model ground
unremediated ground) is primarily a result of the stiffening with surcharge), and Model 4 (model ground with 36 stone
effect of the stone columns. columns and with surcharge) will be discussed, analyzed,
The sugar-cemented columns were placed with high and compared based on recorded accelerations, pore water
precision within the model container before construction of pressures, and settlements. Due to space limitations, only
the surrounding dry silt layer. Thereafter, upon saturation selected-representative response data is presented below.
under vacuum, the sugar dissolved and all cementation
effects were nullified. The sugar content is minimal, and is 6.1. Model 1 (uniform ground without surcharge)
not expected to have any effect on the properties of the
saturated sand columns (stone columns). Fig. 5 displays the recorded accelerations, EPWP and the
soil surface vertical settlements measured during the first
shaking event of the benchmark Model 1. PPT P2 did not
5. Input motions function during the test and its reading is not shown here.
Severe reduction of soil strength and stiffness due to
After saturation was complete, the transducers were liquefaction was apparent in the acceleration records. The
connected to the data acquisition system, manual surface measured soil accelerations virtually disappeared after
measurements were taken, and the model was spun to a 50g about 6, 9, and 10 cycles of input shaking, at the depths of
(Models 1 and 2) or a 63g (Models 3 and 4) gravitational field. 1.3, 2.3, and 4.3 m. Close to the base at a1, significant
One-dimensional horizontal shaking (vertical accelerations attenuation was observed after 15 cycles of input excitation.
being less than 10% of the horizontal) was finally imparted All accelerometer records start with cycles that are similar
along the model long axis using an electro-hydraulic shaker in shape and magnitude to the input motion, followed by an
[15]. Models 1 and 2 were shaken two times in succession attenuation-spiky behavior (gradual EPWP-induced
with a 20-cycle harmonic base input motion of increasing strength degradation, with stress-path excursions along the
amplitude, and 1.8 Hz prototype frequency. In the first and phase transformation line). Eventually, full attenuation of
second shaking events, peak cyclic excitation was at a level motions occurs (shear strength is lost due to liquefaction). It
of about 0:30g (prototype scale). Ample time was allowed can be seen (Fig. 5) that the effects of liquefaction progress
between each phase of shaking in order to allow and monitor from the surface downwards, eventually affecting the entire
full EPWP dissipation. Since no significant difference in silt stratum.
response was observed, only responses to the first shaking All EPWP records (Fig. 5) exhibit the same tendency;
event are reported below. after an initial rise, a peak is achieved, and then the
Models 3 and 4 were shaken three times in succession EPWP remains constant until the end of the shaking (and
with a uniform harmonic base input motion with a for a long time thereafter). The maximum EPWP equals
predominant 1 Hz prototype frequency. In the first shaking the initial overburden vertical effective stresses, s0vo (i.e.
event of 10 cycles, peak cyclic excitation was at a level of ru ¼ 1:0; ru ¼ EPWP=s0vo ) at each location (i.e. the entire
about 0:08g in prototype scale, herein referred to as Shake1. silt stratum was liquefied). A short time after ru ¼ 1:0;
In the second shaking event of 30 cycles, peak acceleration the accelerations start vanishing. The liquefaction front
was about 0:18g; referred to as Shake2. Finally, the third (i.e. ru ¼ 1:0) propagates top to bottom reaching the soil
shaking event of 30 cycles was imparted at a peak level of base during the 12th cycle of input shaking.
about 0:20g; referred to as Shake3. Ample time was allowed The maximum ultimate settlement of the ground surface
between each phase of shaking in order to allow and monitor recorded by the LVDT L1 and L2 was about 0.1 m (about
full EPWP dissipation. Since the difference in response of 0.035 m of this during shaking). The corresponding final
Models 3 and 4 to Shake2 and Shake3 was not significant, permanent volumetric strain over the total silt layer
only responses to Shake1 and Shake2 are reported herein. thickness was 1.3%. The ring displacements (L3, L4, and
L5; not shown here) showed negligible permanent drift (as
expected in this level ground model).
6. Test results and analyses
6.2. Model 2 (ground with 45 stone columns
During all tests, data were recorded at a sampling rate of and without surcharge)
5000 samples/second (real time) with a filter cutoff
frequency of 600 Hz. After testing, each model was Fig. 6 displays the recorded accelerations, EPWP and
carefully inspected to measure the final locations of all the soil surface vertical settlements for the first shaking
transducers, which were found to be very close to their event of Model 2. Unfortunately, accelerometers a4 and
K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584 577

Fig. 5. Measured Model 1 response.

Fig. 6. Measured Model 2 (with 45 stone columns) response.

578 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

a6 did not function properly, and therefore their readings locations, both the softening induced initial amplification
are not shown herein. The two accelerometers near the and the subsequent severe attenuation phases are signifi-
ground surface, a9 (silty soil) and a10 (stone column), cantly delayed in Model 2 relative to Model 1. This can be
showed that initial liquefaction took place at both attributed to the reinforcing-stiffening effect of the installed
locations (it is noted that this is a very low confinement stone columns. In general, throughout shaking, Model 2
region with inherently low liquefaction resistance). behaved in a stiffer manner than Model 1.
However, the record at a10 shows a more spiky response, The EPWP traces measured in the Models 1 and 2 at
and overall a stiffer response than at a9. The stiffer stone corresponding locations in the top half of the models,
columns exhibit a more pronounced cyclic-mobility showed fairly close similarity (with Model 2 EPWP build-
behavior, compared to the surrounding silty soil. This up being slightly slower than that in Model 1). However, in
cyclic mobility behavior explains why the accelerations the bottom half of the models, EPWP build-up in Model 2
vanish at a later time than those in the silt soil. The was considerably slower than that observed in Model 1.
accelerations at a7 (depth ¼ 2.15 m) also vanished Therefore, the difference in the rate of EPWP build-up
before the end of shaking, indicating that liquefaction between Models 1 and 2 soil was more pronounced at depth.
eventually occurred at this depth. Accelerations at a3 and Fig. 7 depicts EPWP evolution throughout the soil deposit
a5 (silty soil mid-depth) were somewhat attenuated measured in Models 1 and 2. As seen in this figure, the rate
towards the end of the shaking event but never vanished. of EPWP build-up is considerably higher in Model 1,
Time histories of accelerations at a1 and a2 near the compared to Model 2 at corresponding locations. The entire
bottom of the silty soil are characterized by pronounced silt soil stratum completely liquefied at the end of 12th base
spikes of large amplitudes and strong overall response. input cycle in the Model 1, whereas even at the end of
At this depth, liquefaction of the silty soil is completely
prevented by the stone columns.
The PPT records P6, P7, P3, and P5 indicate that
ru ¼ 1:0 (initial liquefaction) condition was achieved at
about 12– 16 cycles of shaking (number increasing with
depth of PPT). However, at P4, located at about the
same depth as P3 but inside the stone column, the
maximum value of ru attained during the shaking is less
than 0.5, followed by a relatively quick reduction of
EPWP. The transducers P1 and P2 located near the
bottom of the silty soil exhibited maximum ru values of
around 0.8, and liquefaction was effectively prevented as
compared with Model 1 at the same depth.
The ultimate settlement of the ground surface recorded
by the LVDT L1 and L2 was about 0.07 m (about 0.03 m of
which during shaking). The corresponding final permanent
volumetric strain over the total silt layer thickness was
0.9%. These values are less than those recorded for the
benchmark Model 1, but the difference is not significant. It
should be noted that the effectiveness of stone columns in
reducing settlements cannot be appreciated here due to the
absence of a surcharge on the surface. The ring displace-
ments (L3, L4, and L5; not shown here) were more-or-less
uniform harmonic in nature with peaks ranging from 1.5 to
2.2 cm (decreasing with depth), and negligible permanent
drift (as expected in a level-ground model). These values
were about 20% less than the ones measured in Model 1,
indicative of the Model 2 stiffer composite-material
behavior.

6.3. Comparison of test results from Models 1 and 2

Despite both models having attained high EPWP, their

dynamic behavior was noticeably different. The decay of
accelerations (i.e. loss of strength) in Model 1 was Fig. 7. Evolution of EPWP in Models 1 and 2 (note t ¼ time after start of
significantly more rapid than in Model 2. At corresponding shaking).
 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584 579

shaking (i.e. 20th cycle), only the top-half of the silt soil the stone columns to provide significant stress re-distri-
stratum liquefied in Model 2. These EPWP records are bution and mitigate the liquefaction of the loose silt. In
consistent with the recorded accelerations, which exhibit practice, this confinement could be obtained with the weight
much stronger response in the bottom half of Model 2 of the structure. Models 3 and 4 test this hypothesis.
compared to Model 1. Even at the top half of the silt stratum,
it took about two to three times more shaking cycles for the 6.4. Model 3 (model ground with surcharge)
silt to show significant strength degradation in the Model 2
compared to Model 1. Therefore, although liquefaction was 6.4.1. Shake1
not averted by the installed stone columns (in the upper half This shaking event simulated a moderate level of
of the silt stratum) under the strong base input motion earthquake excitation. Fig. 8 displays the recorded accel-
applied during these tests, the composite ground had a erations, EPWP and the soil surface and foundation vertical
considerably higher liquefaction resistance than the uniform settlements. Accelerometers a4, a5, and a7 did not function
silt ground (i.e. the soil treated by stone columns required a during both Models 3 and 4 tests, hence their results are
higher number of cycles to liquefy). excluded. As a reference, initial effective vertical stresses
Based on reinforcement concepts proposed by Baez [1], a (combined stress by the overburden soil and the foundation)
system of stone columns and silt between them with are included in the EPWP records of Fig. 8, based on the
parameters such as the one tested in Model 2 (Gr ¼ 6; pre- theory of Elasticity (with the aid of Boussinesq charts,
treatment liquefaction Factor of Safety-FSpre ¼ 0.5), an area Lambe and Whitman [16]).
replacement ratio, Ar ; of 20%, would have been sufficient to During Shake1, the accelerations at most locations were
prevent the occurrence of liquefaction ðFS $ 1Þ in the silt generally similar to the imparted input signal, except that
profile. However, the above observations suggest that moderate spiky response appeared in most soil acceleration
sufficient vertical stress or confining pressure might be records within 3 –5 cycles (number increasing with depth).
required to ‘engage’ the reinforcing effect of the stone This spiky response is indicative of cyclic mobility, even at
columns as suggested by Baez [1]. The centrifuge tests of this low level of excitation (possibly also due to presence of
Models 1 and 2 indicate that such vertical effective stress the additional footing stress). At a6, a peculiar asymmetric
might need to exceed about 45 kPa (1000 psf) in order for response occurred with clear spikes during each cycle of

Fig. 8. Measured Model 3 response during Shake1.

580 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

excitation. Such asymmetric spiky acceleration response (1.2% vertical strain), about 0.06 m of which during
has been thoroughly investigated earlier [17,18], and has shaking. Contribution of the post-shaking consolidation
been associated with cyclic-mobility down-slope shear settlements to the total settlement of the free-ground-surface
deformations (in our case, away from the footing and varied from 50 to 70%, depending on the shaking event and
towards the free field). Thus, it may be inferred from the the model. However, movement of the underlying soils
response of a6 that below the foundation edge, soil has towards the surrounding free-ground-surface is noted to
deformed laterally towards the free field. Indeed, this contribute considerably to the recorded vertical defor-
deformation and asymmetric response was observed to mations close to the footing area.
prevail even more strongly (and deeper, e.g. a3, as the zone
of significant lateral deformation expanded) during the next 6.4.2. Shake2
two larger shaking events. Due to symmetry, the absence of Fig. 9 displays the recorded accelerations, EPWP and the
a static driving shear stress along the footing centerline soil surface and foundation vertical settlements for Shake2
dictated an essentially symmetric acceleration response (see of Model 3. This event imparted strong input excitation,
a2, a5, and a8 records). The accelerations recorded on top of with about 30 cycles of shaking at a peak acceleration of
the footing (a9) show significant amplification of the base about 0:18g: At deeper locations, a1 and a2, accelerations
excitation, possibly due to the presence of a relatively stiff stayed similar to the imparted base motion. Asymmetry in
underlying foundation. accelerations indicative of lateral deformations was
Under the heavy vertical stresses of the foundation- observed at a6 and a3 (both under the footing edge at 3
footing, this moderate earthquake resulted in EPWP that and 5 m depth, respectively) followed by a significant
corresponds only to a fraction of the initial vertical effective attenuation phase. At a8, 3 m beneath the foundation,
stresses (ru ranging from 0.12 to 0.45). Below the footing notable gradual attenuation of accelerations was also
centerline, a significantly less contractive soil response observed. Likewise, the footing accelerations were gradu-
prevailed within the foundation soil. In fact, the ru values ally deamplified compared to the input motion after five to
were lowest immediately below the foundation (P7) six cycles of shaking, as the foundation soils became softer
centerline. Earlier, a number of centrifuge and 1g shaking due to EPWP development.
table tests on foundations supported by sandy deposits have Compared to Shake1, a stronger (both in magnitude
revealed that EPWP was generally smaller under the and spatial extent) negative EPWP build-up tendency
foundation than in the free field [19 – 22]. This was was observed at the central foundation zones (as the
explained by the shear-induced dilative soil response during magnitude and the spatial extent of horizontal normal
deformation of the saturated soil below the footing. strains in the foundation grew with stronger base input
Moreover, the superposed footing loads might have caused excitation). However, away from these expansive zones,
a beneficial reduction of liquefaction potential as well [23]. significant positive EPWP was attained (i.e. P1, P2, P3,
At P7, EPWP initially rises, inducing material softening. and P6). The ru values at these locations varied from
Due to the large static shear stresses under the foundation, 0.33 to 0.72 (increased with depth, or distance away
the EPWP-softened foundation soil attempts to move from the footing).
sideways away from the footing centerline. The horizontal In Shake2, the footing was observed to undergo large
normal strains occurring at this location are large, dictating settlements of 0.47 m. A large portion of this settlement
significant reductions in EPWP. In turn, this causes the (about 0.42 m) developed almost linearly with time during
gradual reduction of EPWP observed in P7. As the shaking shaking. These large settlements were partially a result of
and the lateral straining of the foundation soil stops, P7 migration of underlying foundation soil towards the free
EPWP starts increasing again, fed by the surrounding far- field, where the ground surface was observed to have
field areas (also observed by Liu and Dobry [21]). negligible net vertical deformations (compaction settle-
Lateral outflow of foundation soils contributed in a major ments were largely masked by the heave). In Shake3
way to the observed footing settlements. The foundation (records not shown here for the sake of brevity), similar
settled about 0.17 m, about 0.15 m of which during shaking. behavior was also observed (the foundation settled 0.43 m
In every shaking event, both in Models 3 and 4, over 90% of whereas the adjacent soil surface settlement was negligible).
the foundation settlements occurred during shaking, with
only a small portion contributed by post-shaking soil 6.5. Model 4 (model ground with 36 stone columns
reconsolidation due to EPWP dissipation. This means that and with surcharge)
the inertial forces due to shaking played a decisive role in
the deformation of the soil-foundation system [21]. A large In general, the stone column retrofit process was found to
portion of the foundation settlements occurred due to reduce permanent settlements of the foundation by about
penetration of the heavy footing into foundation soil. As will 50– 55% during each shaking event. Figs. 10 and 11 depict
be discussed below, this foundation penetration will the model response during Shake1 and Shake2, respectively.
increase with the decrease in foundation soil strength. The In Shake1, the free-field settlement was about the same as in
ultimate settlement at the free-ground-surface was 0.12 m the Model 3, however the foundation settled only 0.07 m
 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584 581

Fig. 9. Measured Model 3 response during Shake2.

(versus 0.17 m) in Model 4. Overall, the recorded accelera- (i) The foundation settlements were reduced by about
tion and EPWP response during Shake1 (Fig. 10) was 50% (from 0.47 m in Model 3 to 0.23 m in Model 4).
similar to that of the benchmark model (Fig. 8), with the (ii) Accelerations in the silt were slightly stronger than the
following considerable differences. ones measured in Model 3 (note that a1 in the stone
column has rotated during shaking, leading to the
(i) The acceleration spikes that appeared at a1, a2, and a3 curvature depicted in its record). Overall, the relatively
in the Model 3 were not observed in Model 4. At a6, high recorded accelerations (including those of the
such spiky response was observed, but to a lesser footing), clearly indicate that the stone columns largely
extent. Absence of this cyclic mobility effect indicates preserved overall foundation stiffness (as was also
that shear strains were lower than those of Model 3. indicated by P4 with minimal residual EPWP build-up
(ii) The gradual reduction, and the negative EPWP within the column).
observed beneath the foundation (P7) in Model 3 (iii) The seismic shaking was effectively transferred
were not apparent in Model 4. This indicates that (actually with some amplification) directly from the
lateral deformations were reduced drastically by the base of the deposit to the footing (see a9 record) by the
stone column remediation (as indicated by the stiff composite ground (i.e. silt-stone column). Thus
settlement data as well). At the remaining PPT the composite soil block under the foundation
locations, Model 3 EPWP was consistently higher sustained enough of its initial stiffness to transmit
than those of Model 4. Insignificant residual EPWP and amplify the base accelerations to the footing. The
was observed at P4 located inside the stone column, overall foundation shear strength also provided resist-
which dissipated relatively rapidly after shaking. ance to the heavy foundation penetration during
shaking and much reduced the vertical settlements
During the second larger shaking event (Shake2), the (compared to uniform silt foundation case of Model 3).
stiffening effect of stone columns was more obvious (Fig. 9 (iv) Overall increased foundation stiffness during shaking
versus Fig. 11). Based on the recorded response, it may be also reduced the outward migration of soils beneath the
concluded that the stone column application in the footing, in turn reducing the negative EPWP tendency
foundation layer has led to the following effects: that was strongly observed at P7, P5, and P4 of Model
582 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

Fig. 10. Measured Model 4 (with 36 stone columns) response during Shake1.

3. In general, EPWP in Model 4 was somewhat slower, 7. Additional observations

reached lower ultimate values, and dissipated faster
than in Model 3. 1. In Models 3 and 4, an interesting observation is that
along the foundation centerline (i.e. P2, P5, and P7),
In both tests, the footing settlement developed more or EPWP showed a ‘double cycling’ effect [24], suggesting
less linearly with time, with a much lower rate for the stone that the soil at these locations was dilating in both
column case. Fig. 12 depicts the cumulative foundation directions. However, at the other transducer locations, no
settlement in Models 3 and 4. After one moderate (Shake1) such ‘double cycling’ of EPWP was observed, presum-
and two strong shaking events (Shake2 and Shake3), the ably because at these locations the soil dilates mainly in
foundation on the silt layer settled about 1.07 m, versus one direction (towards the free field), as also shown by
0.52 m when treated with stone columns. In both models, the acceleration response.
the settlement during Shake2 was approximately three times 2. It is interesting to note that the foundation zones under
larger than that of Shake1. Shake3 resulted in slightly the footing did not reach zero effective stress condition in
smaller settlements than those measured in the second any test. Actually, the foundation EPWP in any case did
shaking, as the soil has already undergone some compaction not appreciably surpass the corresponding (similar depth)
during the first two events. free-field initial effective vertical stress values. Similar
Although the improvement provided by the stiffer stone behavior was also observed by Koga and Matsuo [25] in
columns appears little in terms of EPWP build-up patterns 1g shake table tests on earth embankments and was
(within the soil between columns), the stiffer stone column attributed to the inability of the earlier liquefied free-field
elements provided higher overall foundation shear strength soil to provide lateral stress more than its initial vertical
and bearing capacity, preventing excessive settlements. effective stress to the foundation soil.
However, the presence of these stiff columns, also allowed 3. Overall, the test results suggests that vibro-concrete
for more efficient transmission (and even amplification) of columns, or soil-cement columns, as an element stiffer
the base accelerations propagated to the footing-super- than stone columns is likely to be a more viable solution
structure. in mitigating liquefaction, if: (a) we are dealing with
 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584 583

Fig. 11. Measured Model 4 (with 36 stone columns) response during Shake2.

a case of low permeability silty soils that cannot be of columnar inclusions may be beneficial in better
effectively densified by stone column installation and restraining the lateral outflow of EPWP-softened soils,
cannot be drained effectively during shaking, and (b) the which was observed to significantly contribute to the
only possible mitigation benefit is from the stiffening vertical foundation settlement.
stress concentration criterion [1]. In addition, concrete,
or soil-cement, columns being stiffer in compression,
may provide greater support for the overlying foun- 8. Summary and conclusions
dations reducing settlements compared to the stone
columns. Likewise, improved shear and bending stiffness A series of highly instrumented dynamic centrifuge
model tests were performed to evaluate the effectiveness of
stone columns applied to non-plastic silty deposits, as a
liquefaction countermeasure. The study was focused on the
possible stiffening benefit, rather than improved drainage
and densification due to stone column installation.
Settlement, acceleration, and EPWP data indicated an
overall stiffer foundation material response during shaking
in the models remediated by stone columns. Stone columns
somewhat retarded the EPWP build-up (in the soil between
columns), increased the foundation soil overall stiffness,
and significantly reduced the surcharge-footing settlements.
In the free-field situation, the stiffening effect provided by
the stone columns was only primarily effective in reducing
pore-pressures at depths below 5 m (45 kPa or 1000 psf)
approximately. In practice, this confinement could be
Fig. 12. Accumulated foundation-footing settlements after Shake1, Shake2, obtained with the weight of the structure. Near ground
and Shake3 events. surface, the installed columns were only of marginal effect
584 K. Adalier et al. / Soil Dynamics and Earthquake Engineering 23 (2003) 571–584

in reducing pore pressures. However, this issue does not immersed tube tunnel. International Journal of Physical Modelling in
substantially affect the important situations of remediation Geotechnics 2003; 3(2):23–32.
[7] Phillips R, Guo PJ, Popescu R. Physical modeling in Geotechnics-
below shallow foundations, where the deformation mech-
ICPMG’02. Rotterdam: Balkema; 2002.
anism is totally different, and the stone columns were found [8] Marcuson WF, Hadala PF, Ledbetter RH. Seismic rehabilitation of
to reduce settlements by about 50%. earth dams. J Geotech Engng, ASCE 1996;122(1):7–20.
The study reported in this paper may be viewed as a pilot [9] Finn WDL. State-of-the-art of geotechnical earthquake engineering
investigation to furnish a quantitative comparative frame- practice. Soil Dynam Earthquake Engng 2000;20:1–15.
work of stone column liquefaction countermeasure effec- [10] Baez JI, Martin GR. Advances in the design of vibro systems for the
improvement of liquefaction resistance. Symposium of ground
tiveness in silty deposits. Additional tests considering a wider
improvement, Vancouver, BC: Vancouver Geotechnical Society;
range of conditions such as different thicknesses of treated 1993.
deposits, area replacement ratio ðAr Þ; input motion par- [11] Laak VP, Taboada V, Dobry R, Elgamal AW. Earthquake centrifuge
ameters, and stone column permeability (testing columns of modeling using a laminar box. In: Ebelhar RJ, Drnevich VP, Kutter
higher permeability), column initial stiffness, diameter and BL, editors. Dynamic geotechnical testing II. ASTM STP 1231; 1994.
slenderness ratio are needed to fully clarify the earthquake p. 370–84.
[12] Walker AJ, Stewart HE. Cyclic undrained behavior of nonplastic and
response of silty deposits in the presence of stone columns. low plastic silts. Technical report, NCEER-89-0035; 1989. 220 pp.
[13] Arulmoli K, Muraleetharan KK, Hossain MM, Fruth LS. Verification
of liquefaction analysis by centrifuge studies laboratory testing
Acknowledgements program soil data. Irvine, CA: EarthTech Corporation; 1992.
[14] Adalier K. Mitigation of earthquake induced liquefaction hazards.
The experiments described here were performed using PhD Thesis. Department of Civil Engineering, Rensselaer Polytechnic
Institute, Troy, NY; 1996. 659 pp.
the Geotechnical Centrifuge at Rensselaer Polytechnic [15] Laak VP, Elgamal AW, Dobry R. Design and performance of an
Institute, Troy, NY, USA (Prof. Ricardo Dobry, Director). electro-hydraulic shaker for the RPI centrifuge. Proceedings of the
The reported study was supported by the NCEER (Highway Centrifuge’94; 1994. p. 139 –44.
Project) grants No. R91632 and 2302, and PYI NSF grant [16] Lambe TW, Whitman RV. Soil mechanics. New York: Wiley; 1969.
No. BCS-90-16880. Matching funds for this PYI award [17] Dobry R, Taboada T, Liu L. Centrifuge modeling of liquefaction
effects during earthquakes. Proceedings of the First International
were provided by Hayward Baker and Earth Mechanics, Inc.
Conference on Earthquake Geotechnical Engineering, Special, Key-
(Ignatius Po Lam, Principal). This support is gratefully note and Theme Lectures, IS-Tokyo, Japan; 1995. p. 129 –62.
acknowledged. [18] Elgamal AW, Zeghal M, Taboada V, Dobry R. Analysis of site
liquefaction and lateral spreading using centrifuge testing records.
Soils Found 1996;36(2):111–21.
References [19] Yoshimi Y, Tokimatsu K. Settlement of buildings on saturated sand
during earthquakes. Soils Found 1977;17(1):23– 38.
[20] Whitman RV, Lambe PC. Earthquake like shaking of a structure
[1] Baez JI. A design model for the reduction of soil liquefaction by
founded on saturated sand. In: Corte JF, editor. Proceedings of the
vibro-stone columns. PhD Thesis. University of Southern California;
International Conference on Geotechnical Centrifuge Modeling.
December 1995.
[2] Luehring R, Dewey R, Mejia L, Stevens M, Baez J. Liquefaction Rotterdam: A.A. Balkema; 1988. p. 529 –38.
mitigation of a silty dam foundation using vibro-stone columns and [21] Liu L, Dobry R. Seismic response of shallow foundation on
drainage wicks: a test section case history at Salmon Lake Dam. liquefiable sand. J Geotech Geoenviron Engng, ASCE 1997;123(6):
Proceedings of the Annual Conference of the State Dam Safety 557–67.
Officials, Nevada; 1998. p. 719–29. [22] Adalier K, Elgamal AW, Martin GR. Foundation liquefaction
[3] Schofield AN. Dynamic and earthquake geotechnical centrifuge countermeasures for earth embankments. J Geotech Geoenviron
modeling. Proc Int Conf Recent Adv Geotech Earthquake Engng Soil Engng, ASCE 1998;124(6):500 –17.
Dynam, Rolla, MO 1981;3:1081 –100. [23] Adalier K, Zimmie TF, Pamuk A. Seismic behavior of rubble-mound
[4] Tan TS, Scott RF. Centrifuge scaling considerations for fluid particle moles on sandy marine deposits. Proceedings of the 21st International
systems. Geotechnique 1985;35(4):461 –70. Conference on Offshore Mechanics and Arctic Engineering, Oslo,
[5] Kimura T, Takemura J, Hiro-oka A, Okamura M, Matsuda T. Norway; 2002.
Countermeasures against liquefaction of sand deposits with structure. [24] Adalier K, Elgamal AW. Seismic response of adjacent saturated dense
Proceedings of the First International Conference on Earthquake and loose sand columns. Soil Dynam Earthquake Engng 2002;22(2):
Geotechnical Engineering, Tokyo, Japan; 1995. p. 163 –84. 115 –27.
[6] Adalier K, Abdoun T, Dobry R, Phillips R, Yang D, Naesgaard [25] Koga Y, Matsuo O. Shaking table tests of embankments resting on
E. Centrifuge modeling for seismic retrofit design of George Massey liquefiable sandy ground. Soils Found 1990;30(4):162–74.