Soil Dynamics and Earthquake Engineering 25 (2005) 773–783

www.elsevier.com/locate/soildyn

Centrifuge modeling of load-deformation behavior

of rocking shallow foundations
Sivapalan Gajana, Bruce L. Kuttera,*, Justin D. Phalena,
Tara C. Hutchinsonb, Geoff R. Martinc
a
Department of Civil and Environmental Engineering, University of California, Davis, CA 95616, USA
b
Department of Civil and Environmental Engineering, University of California, Irvine, CA, USA
c
Department of Civil Engineering, University of Southern California, Los Angeles, CA, USA
Accepted 11 November 2004

Abstract

Shallow foundations supporting building structures might be loaded well into their nonlinear range during intense earthquake loading. The
nonlinearity of the soil may act as an energy dissipation mechanism, potentially reducing shaking demands exerted on the building. This
nonlinearity, however, may result in permanent deformations that also cause damage to the building. Five series of tests on a large centrifuge,
including 40 models of shear wall footings, were performed to study the nonlinear load-deformation characteristics during cyclic and
earthquake loading. Footing dimensions, depth of embedment, wall weight, initial static vertical factor of safety, soil density, and soil type
(dry sand and saturated clay) were systematically varied. The moment capacity was not observed to degrade with cycling, but due to the
deformed shape of the footing–soil interface and uplift associated with large rotations, stiffness degradation was observed. Permanent
deformations beneath the footing continue to accumulate with the number of cycles of loading, though the rate of accumulation of settlement
decreases as the footing embeds itself.
q 2005 Elsevier Ltd. All rights reserved.

Keywords: Centrifuge modeling; Foundation uplift; Rocking; Shear walls; Settlement; Sliding; Rotation

1. Introduction on the building response to seismic shaking. The non-

Shallow foundations supporting building structures
might be loaded well into their nonlinear range during
intense earthquake loading. Understanding the nonlinear
behavior of shallow building foundations under large
amplitude loading is an important aspect of performance-
based design. The 1997 Federal Emergency Management
Agency NEHRP Guidelines for the seismic retrofit of
buildings [1] and the associated Applied Technology
Council document (ATC 40) [2] discuss alternative design
methods associated with the response of shear walls when
subjected to lateral earthquake induced rocking. Geotechni-
cal components of the foundation have a significant effect
* Corresponding author. Tel.: C1 530 752 8099/7929; fax: C1 530 752
7872.
E-mail address: blkutter@ucdavis.edu (B.L. Kutter).
0267-7261/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2004.11.019
linearity of the soil and the interaction between the soil and
foundation was shown to influence the building’s stiffness
and period to change [3]. The nonlinearity of the soil may
act as an energy dissipation mechanism, potentially
reducing demands exerted on the structural components of
the building. This associated nonlinearity, however, may
result in permanent deformations (settlement, rotation or
sliding) that cause damage to the building.
Many researchers have studied the nonlinear behavior of
shallow foundations and the effect of foundation rocking on
the behavior of soil–foundation system [3–8]. Foundation
rocking and yielding of soil reduce the stiffness of the soil–
structure interface, lengthen the structure’s natural period,
and hence may reduce the force demands imposed on the
structure [3]. Rocking of the footing progressively makes
the foundation soil curved with a reduction in contact area
between the footing and soil and thereby causing non-
linearity in the moment–rotation relationship [8]. With
increasing eccentricity and inclination of vertical load, the
vertical footing displacements reduce while horizontal
774 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783
and rotational displacements increase [5]. Faccioli et al. [4],
Martin and Lam [6] and Pecker and Pender [7] state that
allowing mobilization of ultimate capacity for shallow
foundations, a major change in conventional foundation
design, results in reduced ductility demands for structural
components and more accurate performance evaluation.
Although previous experimental efforts have contributed
greatly to nonlinear analysis of shallow foundations in
performance-based evaluation, many of these tests
were conducted at low confining stresses (model tests at
1 g) [3,5,8]. The confining stresses play an important role in
the nonlinear prototype behavior of soil-footing system.
This paper presents the results of tests conducted on model
foundations attached to a rigid shear wall subjected to
vertical, lateral slow cyclic and dynamic loading at 20g
centrifugal acceleration. Five series of tests, including 40
model shear wall footings, were performed to study the
effects of footing dimensions, depth of embedment, initial
static vertical factor of safety (FSV) and soil type on the
nonlinear ‘soil–foundation’ system response. The initial
static vertical factor of safety was varied by changing the
structures weight and footing dimensions. The methods of
testing, analyses and major findings are presented in the
following sections.
2. Testing program
Experiments have been conducted in a 9.1 m radius
centrifuge at the Center of Geotechnical Modeling at the
University of California, Davis at 20g centrifuge accelera-
tion. Four series of tests were conducted on dry sand
(DrZ80 and 60%) and one test series was on saturated clay
(CuZ100 kPa). All the models were tested on a soil bed
prepared in a rigid container (Fig. 1) (1.75!0.90!0.53 m,
Fig. 1. Model container and experimental setup with instrumentation for vertical push, slow cyclic lateral push and dynamic loading tests.
model scale dimensions), each of which included six to ten
footing-wall models depending on the testing series. The
preparation procedures for all the models were similar.
Sand was pluviated at required densities using air
pluviation. Nevada Sand (uniformly graded fine clean
sand with D50Z0.17 mm, CuZ1.6, emaxZ0.881 and
eminZ0.536) was used for all sand layers. The remolded
San Francisco Bay Mud (Atterberg limits: LLZ90 and
PLZ38) was consolidated on top of a dense sand layer
prior to spinning in the clay test series. The soil properties
of Nevada Sand are presented in [9,10]. All the results in
this paper are presented in prototype units unless otherwise
stated. The initial static vertical factors of safety of the
footings were varied from about 2.0 to 10.0, by changing
the weight of the structure and footing dimensions
(length, LZ2.5–4.0 m, width, BZ0.4–1.0 m, embedment,
DZ0.0–0.7 m). The depth of the sand layer inside the
container was 4.0 m, and the thickness of the consolidated
clay layer was 1.7 m in clay test series.
Fig. 1 shows the test setup and instrumentation for
vertical push, slow cyclic lateral push and dynamic tests.
The soil bed in the container was divided into stations
(about six to ten stations depending on the test series) and
tests were conducted in each station separately. Table 1
presents the details of tests conducted in every test series.
Soil strength, footing geometry, depth of embedment, initial
static vertical factor of safety (FSV) and types of loading are
given for every test in Table 1. The last column in Table 1
includes the height of push for slow cyclic lateral loading
tests. ‘Standard’ height of push corresponds to the height of
center of gravity of the structure from the base of the
footing, while ‘low’ height of push corresponds to a height
closer to the base of the footing.
In each series of tests, at least one concentric vertical
push test was conducted to estimate the bearing capacity of
 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783 775

Table 1
Test series summary

Test series Solid type Event type Static FSv Footing length (m) Footing width (m) Embed depth (m) Load height (m)
KRR01 Dry sand (60–80%) VSC 3.8 2.672 0.686 0.3 n/a
HSC 3.8 2.672 0.686 0.3 Standard
HSC 3.0 2.672 0.686 0.3 Standard
VSC 3.0 2.672 0.686 0.3 n/a
VSC 3.0 2.672 0.686 0.3 n/a
VSC 2.0 2.540 0.381 0.3 n/a
Dynamic 6.2 3.944 1.080 0.3 n/a
Dynamic 2.8 2.672 0.686 0.3 n/a
Dynamic 2.4 2.672 0.686 0.3 n/a
Dynamic 1.3 2.540 0.381 0.3 n/a
Dynamic 3.0 2.672 0.686 0.3 n/a
Dynamic 1.3 2.540 0.381 0.3 n/a
KRR02 Dry sand (DrZ60%) VSC 1.6 2.672 0.686 0 n/a
VSC 1.6 2.672 0.686 0 Standard
HSC 4.1 3.944 1.080 0 Standard
HSC 1.6 2.672 0.686 0 Standard
Dynamic 1.6 2.672 0.686 0 n/a
Dynamic 4.1 3.944 1.080 0 n/a
KRR03 Bay mud (CuZ100 kPa) VSC 2.8 2.672 0.686 0 n/a
HSC 2.8 2.672 0.686 0 Standard
HSC 2.8 2.672 0.686 0 Standard
VSC 2.8 2.672 0.686 0 n/a
Dynamic 2.8 2.672 0.686 0 n/a
Dynamic 4.8 3.944 1.080 0 n/a
SSG02 Dry sand (DrZ80%) VSC n/a 2.840 0.690 0 n/a
HSC 6.8 2.840 0.690 0 Low
HSC 6.8 2.840 0.690 0 Standard
HSC 9.6 2.840 0.690 0 Standard
HSC 3.4 2.840 0.690 0 Standard
HSC 9.6 2.840 0.690 0 Standard
HSC 3.4 2.840 0.690 0 Standard
Dynamic 6.8 2.840 0.690 0 n/a
Dynamic 5.3 2.840 0.690 0 n/a
SSG03 Dry sand (DrZ80%) VSC n/a 2.840 0.690 0.7 n/a
HSC 1.1 2.840 0.690 0 Low
HSC 8.2 2.840 0.690 0.7 Standard
HSC 8.2 2.840 0.690 0.7 Standard
HSC 4.0 2.840 0.690 0.7 Standard
HSC 11.5 2.840 0.690 0.7 Standard
Dynamic 4.0 2.840 0.690 0.7 n/a
Dynamic 6.4 2.840 0.690 0.7 n/a

HSC, horizontal slow cyclic (lateral push tests); VSC, vertical slow cyclic (vertical push tests); Dynamic, dynamic base shaking.

the soil and to back-calculate the strength of the soil bed
(either friction angle or undrained shear strength). Slow
cyclic lateral push tests were carried out by pushing the wall
with an actuator at different heights on the wall. This
produced different moment to horizontal force ratios at the
base center point of the footing. Displacements were
measured by two horizontal and two vertical linear
potentiometers attached on the wall and forces were
measured by a load-cell attached to the actuator. Dynamic
loading was applied to the models by shaking the base of the
soil container with a tapered cosine (gradually increasing)
displacement time history. Accelerations and displacements
were recorded at different positions on the footing and wall
as shown in Fig. 1. Experimental setups and testing
procedures are explained in detail in [11–18].
3. Data processing
Measured forces and displacements are used to calculate
the resultant forces and displacements at the base center
point of the footing using equilibrium equations and rigid
body translation and rotation. Results are presented in terms
of three forces and three displacements components for a
planer loading as shown in Fig. 2; vertical force (V),
horizontal force (H), moment (M), settlement (s), sliding (u)
and rotation (q). Measured forces and displacements during
cyclic tests were filtered properly and processed to calculate
the force resultants and corresponding displacements. For
lateral push tests, the additional moment caused by the self-
weight of the structure moving through a lateral displace-
ment (PKD effect) was also included.
776 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783

Fig. 2. Definition of system of forces and displacements.

In the dynamic tests, forces and displacements of the

footing were obtained by combining the measurements of
accelerometers and displacement transducers. Considering
the shear wall as a rigid body, the applied moment, M, for
dynamic tests was computed by
M Z I:a C macg hcg C mgD
where I is the structure’s moment of inertia about its center
of gravity, a is the angular acceleration of the structure, m is
the mass of the structure, acg is the horizontal acceleration at
the center of gravity, and hcg is the height of the center of
gravity from the base of the footing. The term ‘m g D’
corrects for the pKD effect, g is the centrifugal acceleration
and D is the horizontal eccentricity of the center of gravity
relative to the bottom center of the footing. The displace-
ments and rotations of the footing were obtained by
combining low-pass filtered displacement transducers
results and high-pass filtered accelerometer results.
4. Test results
4.1. Vertical push (loading and unloading)
Fig. 3 shows the results for one of the vertical push test
on a rectangular surface footing (LZ2.84 m and BZ0.65 m,
and DZ0.0 m) on dry sand (DrZ80%). The ultimate
vertical load (Vmax) was measured as 1920 kN for this test.
Results show a larger stiffness for the unloading–reloading
process, compared to virgin loading, and the unloading–
reloading lines show nonlinear behavior at very small load
levels.
4.2. Slow cyclic lateral push
(1)
Fig. 3. Vertical push test results: sand (DrZ80%), LZ2.84 m, BZ0.65 m,
DZ0.0 m.
(M/H) at the base center point of the footing was controlled
by the height of push. Most of the slow cyclic lateral push
tests have a standard height of push (hZ4.75 m), which is
close to the height of center of gravity of the structure, while
some tests were done at a low height of push (hZ1.25 m).
The height of all structures tested were 10.0 m.
Fig. 5 shows the moment–rotation, horizontal force-
sliding, settlement-rotation, and settlement-sliding relation-
ships at the base center point of the footing for one of the
lateral push test on sand (DrZ80% dry sand, LZ2.84 m,
BZ0.65 m, DZ0.0 m, hZ4.9 m, and FSVZ6.7), while
Fig. 6 shows the same results for one of the test on clay
(CuZ100 kPa, LZ2.7 m, BZ0.65 m, DZ0.0 m, hZ4.6 m,
and FSVZ3.0). The moment–rotation relationship encloses
a large area in hysteresis loops indicating a considerable
amount of energy is dissipated at the footing soil interface.
The moment–rotation plot does not show any reduction in
moment capacity with the number of cycles or with the
amplitude of rotation, but it does show a degradation of
rotational stiffness with increasing amplitude of rotation.

Slow cyclic lateral push tests were conducted on

structures with different initial static vertical factors of
safety (FSV) (varying footing dimensions, applied vertical
load, soil strength) and with height of lateral loading (h)
varied. An actuator was used to apply cyclic lateral
displacements to the structure at load height h, while a
load cell attached to the actuator measured the load on the
actuator. Fig. 4 shows the load paths followed in slow cyclic
lateral tests. The applied moment to horizontal force ratio Fig. 4. Load paths followed in slow cyclic lateral push tests.
 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783 777

Fig. 5. Slow cyclic lateral push test; sand (DrZ80%), footing lengthZ2.84 m, widthZ0.65 m, embedmentZ0.0 m, FSZ6.7, lateral load heightZ4.9 m
(forces and displacements are at the base center point of the footing).

The settlement–rotation relationship shows the accumu-
lation of permanent settlement beneath the footing. The rate
of increase in settlement per cycle of rotation decreases with
the number of cycles applied. As the footing settles down,
its depth of embedment increases, overburden stresses
increase, and vertical stiffness also increases, thus the rate of
increase of settlement reduces (Fig. 5). The settlement-
rotation plot shows the uplift behavior of the footing
associated with larger rotation amplitudes. A portion of
the footing base loses contact with the soil when the rotation
is large. The formation of a gap on one side of the
footing causes yielding of the soil on the other side of the
footing, and the yielding of the soil on the other hand
increases the uplift. Foundation rocking during high
amplitude lateral loading causes rounding of the soil
beneath the footing, and the rounding of the soil causes

Fig. 6. Slow cyclic lateral push test; clay (CuZ100 kPa), footing lengthZ2.7 m, widthZ0.65 m, embedmentZ0.0 m, FSZ3.0, lateral load heightZ4.6 m
(forces and displacements are at the base center point of the footing).
778 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783

the nonlinear moment rotation relationship and the degra-

dation of rotational stiffness due to closing of the gap.
The horizontal load-sliding plot does not show any
significant degradation in horizontal stiffness with number
of cycles or the amplitude of displacement. The footing does
not slide much until it reaches the shear capacity, but once it
reaches the shear capacity it starts sliding. The settlement-
sliding relationship shows the coupling behavior of sliding
and uplift, and the closing of the gap upon unloading. When
loaded in one direction, the footing closes its gap first and
then it slides horizontally, while a gap forms on the other
side.

4.3. Dynamic test

Fig. 7 shows one of the time histories of applied

acceleration at the base of the soil for dynamic tests. The
dynamic test results shown in Fig. 8 was carried out on the
same footing as the one presented in the slow cyclic lateral
test in Fig. 5 (LZ2.84 m, BZ0.65 m, DZ0.0 m) on sand,
but with a slight change in the weight of the wall used. This
reduced the factor of safety from 6.7 to 5.3. It should be
noted that the settlement and rotation do not start at zero for
the shaking event presented in Fig. 8, because another two
small amplitude shaking events were applied before this
event.
The dynamic moment–rotation behavior also shows a
reduction in rotational stiffness with increasing amplitude of
rotation. The footing–soil interface shows stiffer response in Fig. 8. Results of dynamic test: sand (DrZ80%), LZ2.84 m, BZ0.65 m,
moment–rotation behavior for small amplitude motions, as DZ0.0 m, FSVZ5.3.
the magnitude of shaking increases it shows softer response,
but at the end of shaking, it again shows stiffer response for Fig. 9 shows a dynamic data trend line (i.e. backbone
small amplitude of motions. The structure tilted back and curve that connects the extreme points of each cycle),
forth during shaking and ended up with a permanent rotation obtained from Fig. 8, superimposed on a slow cyclic
in one direction after shaking. The maximum amplitude of moment–rotation plot (Fig. 5). This suggests that the
cyclic rotation observed in the dynamic test (about moment–rotation behavior during slow cyclic and dynamic
0.02 rad.) was smaller than that applied in the slow cyclic tests agree very well and that slow cyclic tests may be
lateral push test (about 0.06 rad.). The settlement-rotation appropriate for simulating moment–rotation behavior in
behavior from the dynamic test does not indicate significant dynamic events.
uplift.

Fig. 9. Comparison of dynamic data trend line with slow-cyclic moment–

Fig. 7. Applied base acceleration time history for dynamic test. rotation plot: sand, DrZ80%, LZ2.84 m, BZ0.65 m, DZ0.0 m.
 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783 779

5. Rotational stiffness degradation

It is evident from the moment–rotation plots of all tests

conducted that the energy is dissipated by the foundation by
rocking. Initial cycles show significant mobilization of
moment capacity at small strains, where rotational stiffness,
kq, is high. At higher levels of strain, there is a softening of
the system and a degradation of rotational stiffness.
Although the hysteresis curves are highly nonlinear, a
consistent linear behavior can be recognized in their central
portion (Fig. 10). For the purpose of this analysis this range
was taken to be the maximum moment (MMAX, half of the
peak-to-peak moment) centered about the median moment.
Fig. 10 shows the estimation of rotational stiffness (kq) at
different levels of rotation from the moment–rotation plot of
a slow cyclic lateral push test. To quantify the amount
of rotational stiffness reduction that occurs as a result of
foundation rocking, each test was analyzed for its stiffness,
relative to a theoretical maximum stiffness, kq,MAX, at each
rotational increment. kq,MAX was estimated from proposed Fig. 11. Rotational stiffness degradation as a result of foundation rocking:
relationship for Gazetas’ elastic stiffness parameters of test results on sand for various FSV.
rectangular footings [19]
"
 # obtained at very low strains (!0.002 rad.). However, it is
GMAX 0:75 L 0:15 not likely that substantial permanent deformations will
kq;MAX Z I 3 b (2)
1 Kn B occur at small foundation rotations (even with a large
number of cycles). There seems to be no clear trend in terms
where GMAX is the initial shear modulus of the soil, obtained of the effect of vertical factor of safety on rotational stiffness
from a relationship proposed by Seed and Idriss [20], n is reduction. Both high and low FSV tests show similar
Poisson’s ratio for sand (taken to be 0.35 for all cases), I is reduction at similar strain levels. There is good agreement in
the area moment of inertia about the centroid of the footing rotational stiffness between the centrifuge and 1g tests
base normal to the direction of rocking, L is the length of the performed by Weissing [8]. Fig. 11 shows a recommended
footing, B is the width of the footing, and b is an embedment ‘mean’ stiffness reduction trend for shallow footings on
factor and is correlated with the ratio of the depth of sand for the foundation rotation range shown. Upper and
embedment to the length of the footing (D/L) [19]. lower bound lines show the G1 standard deviation. The
Fig. 11 shows kq (normalized by kq,MAX) versus rotation mean stiffness line was determined as:
for tests with varying FSV (test series KRR [16–18], SSG
kq
[11,12] and Weissing [8]). Due to the small inherent Z 3:0 !10K3 ðqK0:6 Þ (3)
kq;MAX
vibrations during the test procedure and limitations of the
data acquisition system, accurate recordings were not

6. Accumulation of permanent settlement

Test results in slow cyclic and dynamic tests (Figs. 5, 6

Fig. 10. Moment–rotation plot showing rotational stiffness degradation with
increasing rotation.
and 8) show that the footing tends to accumulate vertical
settlement as the moment and shear loads are cycled. During
horizontal slow cyclic tests, displacement was usually
applied in packets of three cycles of given amplitude,
followed by three more cycles with a larger amplitude (see
Figs. 5 and 6). The permanent displacement in a packet of
cycles of similar amplitude was divided by the number of
cycles in that packet to compute the amount of settlement
per cycle of given rotational amplitude. Fig. 12 is
constructed from selected tests (on sand) of varying FSV
and relative density, Dr. The horizontal axis represents the
rotation, q, for a given packet of cycles (slow cyclic tests)
or averaged values of similar rotational amplitudes
780 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783
Fig. 12. Permanent settlement caused by cyclic rotation: slow cyclic and dynamic test results on sand for various FSV and Dr (for height of pushZ4.0–5.0 m).
(dynamic tests). The vertical axis is the normalized vertical
settlement (normalized by the dimension of footing in
loading direction, L) per cycle of that rotation demand,
denoted UV. A family of curves may be determined based on
the specific FSV and relative density of each test.
Data points shown in Fig. 12 are chosen from slow-cyclic
and dynamic data from the KRR [16–18], SSG [11,12]
series, as well as experimental 1g data from Weissing [18]
and the TRISEE experiment by Faccioli et al. [2]. The
plotted slow cyclic lateral push test data correspond to a
moment to horizontal force ratio (M/HZ4.0–5.0 m). For
slow cyclic tests, there is a consistent trend that the
normalized settlement, Uv, decreases as FSV increases.
Also, for those tests with similar FSV but different relative
densities, Uv decreases as Dr increases. It is important to
point out, that the settlement is not only affected by vertical
factor of safety but soil state as well. The settlement in
dynamic tests is significantly larger than the settlement in
slow cyclic tests. A portion of this difference is attributed to
the free-field settlement of the ground due to the dynamic
ground shaking. Rocking of rigid shear walls during
dynamic shaking tests produced vertical acceleration
components that caused more settlement in dynamic tests
than in slow cyclic tests where the effect of inertia forces is
not present.
7. Failure envelope in (V–M–H) space
Previous studies by Cremer et al. [21] suggested that
there exists a failure envelope in normalized V–H–M
loading space (FVZV/Vmax, FHZH/Vmax and FMZM/
(Vmax.L), where Vmax is failure vertical load for pure
vertical loading). The failure envelope was assumed to
consist of parabolic sections in the V–H and V–M planes
and an elliptical section in the M–H plane
 2  2
FH FM
C K1 Z 0 (4)
aFVc ð1 K FV Þd bFVe ð1 K FV Þf
where, a, b, c, d, e, and f are constants that define the shape
of the failure envelope. The coefficients a and b define the
size of the failure envelope of elliptical shape in a FM–FH
plane whereas the coefficients c, d and e, f define the
parabolic shape of the failure envelope in FH–FV and FM–
FV planes, respectively. This failure envelope was used in a
plasticity model to simulate the behavior of strip footings
during cyclic loading on cohesive soils [21].
Houlsby and Cassidy [22] also suggested an analytical
expression for the yield surface, which also consists of
parabolic and elliptical sections in V–H–M space. They
normalized the forces by maximum past vertical load
experienced by the footing
 2  2
FH F
C M K 16Fv2 ð1 K FV Þ2 Z 0 (5)
h0 m0
where, h0 and m0 are constants that define the shape of the
yield surface. They used this yield surface to formulate a
plasticity model to simulate the behavior of surface footings
on dry cohesionless soils during monotonic loading in
V–H–M space [22].
Experimentally observed failure points are plotted with
the theoretical failure envelopes proposed by Cremer et al.
[21] and Houlsby and Cassidy [22] in Figs. 13 and 14. The
failure points were obtained from centrifuge tests conducted
on rectangular surface footings on sand for constant moment
to shear ratio (M/HZ4.9 m) loading at different vertical
loads (FSV). The data show agreement with the failure
envelopes indicating that these can be used for analytical
modeling of shallow foundations subjected to combined
moment, shear, and axial loading.
 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783
Fig. 13. Projection of failure envelope and experimental failure points on
sand in FM–FV plane for M/HZ4.9 m loading.
8. Effect of embedment on shear and moment capacities
If the footing area and vertical load remain constant as
embedment increases, the tendency for horizontal sliding
should decrease. Although the footing base friction
remains constant, additional passive pressure is mobilized
at the toe of the footing, for increasing embedment,
providing more resistance to sliding. Comparing the tests
in the SSG02 series [11] (no embedment) with those in
the SSG03 series [12] (same building weights and footing
size, with embedment, DZB), an increase of between 4
and 19% in horizontal shear capacity can be observed, as
the static vertical factor of safety increases from about 3.0
to 10.0 and the embedment is increased from DZ0 to
DZB [16].
A parametric study using basic bearing capacity analyses
was performed to assess the relationship of moment to shear
capacity ratios, considering embedments of DZ0, B, 2B and
3B, and different vertical factors of safety FSvZ2, 4, 6, 8
and 12. Dry sand, with a friction angle of fZ378 was used,
with footing dimensions matching the intermediate size
footing tested (BZ0.686 m, LZ2.84 m). The theoretical
maximum shear capacity, HMAX, and moment capacity,
MMAX were calculated as
HMAX Z Vk C 2Po k C Pp K Pa (6)
Fig. 14. Projection of failure envelope and experimental failure points on
sand in FH–FV plane for M/HZ4.9 m loading.
781
Fig. 15. Effect of embedment (D) on the ratio of horizontal force to moment
capacities for various vertical factors of safety.


LV 1 D D D
MMAX Z 1K C Pp C 2Po k K Pa (7)
2 FSV 3 3 3
where V is the applied vertical load, L is the length of the
footing, D is the depth of embedment, FSv is the vertical
factor of safety, Po, Pa and Pp are the at-rest, active and
passive earth pressures and k is the base/side shear
coefficient, taken as 0.4 in this analysis.
MMAX and HMAX both increase with embedment,
however, the relative rates of increase actually govern
footing behavior. If HMAX increases faster with embedment
than MMAX, the footing would have a greater tendency to
rotate about its base, rather than slide, as is the case for
greater depths. Fig. 15 plots the theoretical ratio HMAXL/
MMAX versus embedment for varying vertical factors of
safety, and Fig. 16 shows the percent increase in the ratio
HMAXL/MMAX as embedment increases. The initial trend for
all cases suggests that the ratio increases slightly with depth,
with higher factor of safety footings increasing more than
lower factor of safety footings. In the case of FSZ2.0, the
increase in HMAXL/MMAX is about 4% from a surface
footing to a footing with an initial embedment of 2B. For
FSZ12.0, the increase in HMAXL/MMAX is about 26% from
a surface footing to a footing with initial embedment 2B.
The theoretically predicted increase in shear to moment
capacity ratio calculated using Eqs. (6) and (7) seems to be
consistent with the general magnitude of increase observed
in the experiments. Since most design safety factors are on
the order of 2.0–4.0, it can be assumed that maintaining a
constant vertical factor of safety with increasing embedment
does not favor either sliding or rotation.
782 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783
Fig. 16. Effect of embedment on the percent increase in horizontal force to
moment capacity ratios for various vertical factors of safety.
9. Deformed shape of the footing–soil interface
Plaster was cast at the structures’ footprint after some
tests to preserve the imprint for subsequent analysis of the
shape of the soil surface beneath the footing. Measurements
of these footing imprints were taken in transects using a
computer measuring machine (CMM). These imprints
illustrate the deformed shape of the soil surface at the
foundation–soil interface, which cumulated during the
slow-cyclic lateral push tests. The soil surface beneath the
footing becomes curved as the rigid footing keeps rotating
on the deformable soil layer.
The curved shape of the soil surface at the footing–soil
interface causes partial separation of the footing from the
soil, resulting in reduced footing–soil contact area. During
moment loading, the footing begins lifting off on either side
of its contact, and progressively creates a curved soil surface
beneath the footing. As a result, the length of contact
between the footing and soil decreases as the structures’
rotation increases and as the curvature of the soil surface
becomes significant. The contact length not only reduces,
but also moves along the length of the footing as the
building rocks, causing a nonlinear bearing pressure
distribution in the soil beneath the contact length. The
changing location of the resultant of the bearing pressure
distribution with the rotation of footing dictates the
moment–rotation behavior.
To maintain stability (i.e. a factor of safety of 1.0 against
bearing failure), for a given soil-footing system, a limiting
minimum value of footing–soil contact length may be
calculated. This contact length is defined as the critical
contact length (Lc). The ultimate moment can be obtained
Fig. 17. Illustration of the geometry and contact of the rigid footing with
deformed soil surface.
when the contact length at one edge of the footing reaches
its critical contact length. Considering the free body
diagram at critical overturning illustrated in Fig. 17, and
assuming that the resultant of the bearing pressure acts at the
center of the critical contact length, for a surface footing, the
maximum moment is given by
 
V$L Lc
MMAX Z 1K (8)
2 L
In fact, Eq. (8) is similar to Eq. (7) for surface footings,
where active, passive and side soil pressures are not present.
It can be shown, using conventional bearing capacity
equations, that
Lc 1
Z (9)
L FSV
for surface footings when LcRB. Similar to the above
method, prior to ultimate moment, the moment–rotation
relationship as a function of rotation can be obtained from
the length of the footing (L) and the length and location of
the footing in contact with the soil (Lcont) as a function of
rotation. The above idea is being implemented in a single
element footing–soil contact model that simulates the
moment–rotation behavior of rigid footings on dry sand.
By keeping track of the curved soil surface and the location
of the footing contact length with the footing, moment–
rotation relationships can be obtained. The other parameters
that need to be included in the contact model are the
nonuniform bearing pressure distribution beneath the
footing–soil contact area and the location of the resultant
bearing pressure during large rotations.
10. Conclusions
Data presented in this paper show the nonlinear load–
displacement behavior of model shallow foundations resting
on moderately dense sand under high confining stresses
(20g centrifuge experiments). Models were constructed to
represent various realistic static factors of safety and
subjected to slow cyclic and seismic loading. The
moment–rotation relationships show a large amount of
work being dissipated in the foundation, indicating a great
potential for the soil beneath a footing to dissipate a large
amount of energy during dynamic loading. While this
 S. Gajan et al. / Soil Dynamics and Earthquake Engineering 25 (2005) 773–783
potential for energy dissipation in the soil can reduce the
demands on the building structure, undesired permanent
foundation deformations (rotation, sliding and settling) may
be associated with softening of the system. These permanent
deformations continue to accumulate with the number of
cycles of loading, though the rate of accumulation of
settlement decreases as the footing embeds itself.
The moment–rotation plots also show the softening of
the soil–foundation system (even before the moment
capacity is reached) with increased rotation. The slope of
the curve in the intermediate region of any moment–rotation
plot becomes less steep with larger deformations, indicating
a consistent reduction in rotational stiffness due to uplift and
separation of the base of the footing from the soil.
Normalized settlement per cycle versus amplitude of cyclic
rotation shows a consistent trend for tests in the present
study as well as for tests performed by other researchers.
The settlements tend to increase as both factor of safety and
relative density decrease and as the amplitude of rotation
increases in a consistent pattern. The consistency observed
between the experimental failure points and the failure
envelopes proposed by other researchers [21,22] in
moment-shear-axial loading space suggests that previously
developed analytical expressions for the failure envelopes
may be useful for modeling the footing–soil interface
behavior.
Experimental evidences clearly show the rounding of
soil beneath the footing as the building rocks, which is
consistent with the observed reduction in moment–rotation
stiffness associated with the uplift of the footing.
Acknowledgements
This work was supported primarily by the Pacific
Earthquake Engineering Research Center’s Program of the
National Science Foundation under Award Number EEC-
9701568 and PEER project number 2262001. Authors like
to thank Ross Boulanger and Dan Wilson for their valuable
comments and suggestions. Authors would also like to thank
the support and assistance provided by technicians Chad
Justice, Tom Kohnke and Tom Coker.
References
[1] NEHRP Commentary on the Guidelines for the Seismic Rehabilita-
tion of Buildings, FEMA publication 274; October 1997.
[2] Comartin CD, Niewiaroski RW, Freeman S, Turner FM. Seismic
evaluation and retrofit of concrete buildings: a practical overview of
the ATC-40 document. Earthquake Spectra 2000;16(1):241–62.
783
[3] Bartlett PE. Foundation rocking on a clay soil. ME Thesis. University
of Auckland, School of Engineering Report No. 154; November 1976.
[4] Faccioli E, Paolucci R, Vivero G. Investigation of seismic soil-footing
interaction by large scale cyclic tests and analytical models.
Proceedings of the Fourth International Conference on Recent
Advances in Geotechnical Earthquake Engineering and Soil
Dynamics, San Diego 2001 pp. 26–31.
[5] Georgiadis M, Butterfield R. Displacements of Footings on Sand Under
Eccentric and Inclined Loads. Can Geotechnol J 1988;25:199–212.
[6] Martin G, Lam IP. Earthquake resistant design of foundations: retrofit
of existing foundations. Proc GeoEng 2000 Conf Melbourne,
Australia 2000 pp. 19–24.
[7] Pecker A, Pender M. Earthquake resistant design of foundations: new
construction. Proc GeoEng 2000 Conf, Melbourne, Australia 2000 pp.
19–24.
[8] Wiessing PR. Foundation rocking on sand. ME Thesis. University of
Auckland, School of Engineering Report No. 203; November 1979.
[9] Arulmoly K, Muraleetharan KK, Hossain MM, Fruth LS. VELACS
laboratory testing program. Preliminary Data Rep. to National Science
Foundation, Earth Technology Corporation, Irvine, CA; 1991.
[10] Chen YR. Behavior of fine sand in triaxial, torsional and rotational
shear tests. PhD Thesis. University of California, Davis, CA; 1995.
[11] Gajan S, Phalen JD, Kutter BL. Soil–foundation–structure interaction:
shallow foundations. Centrifuge Data Report for test series SSG02,
University of California, Davis, Report No. UCD/CGMDR-03/01; 2003.
[12] Gajan S, Phalen JD, Kutter BL. Soil–foundation–structure interaction:
shallow foundations. Centrifuge Data Report for test series SSG03,
University of California, Davis, Report No. UCD/CGMDR-03/02; 2003.
[13] Kutter BL, Martin G, Hutchinson TC, Harden C, Gajan S, Phalen JD.
Status report on study of modeling of nonlinear cyclic load-
deformation behavior of shallow foundations. University of Califor-
nia, Davis, PEER Workshop; March 2003.
[14] Phalen JD. Physical Modeling of the Soil–foundation interaction of
spread footings subjected to lateral cyclic loading. MS Thesis.
University of California at Davis, School of Engineering; 2003.
[15] Rosebrook KR. Moment loading on shallow foundations: centrifuge
test data archives. MS Thesis. University of California at Davis,
School of Engineering; 2001.
[16] Rosebrook KR, Kutter BL. Soil–foundation–structure interaction:
shallow foundations. Centrifuge Data Report for test series KRR01,
University of California, Davis, Report No. UCD/CGMDR-01/09; 2001.
[17] Rosebrook KR, Kutter BL. Soil–foundation–structure interaction:
shallow foundations. Centrifuge Data Report for test series KRR02,
University of California, Davis, Report No. UCD/CGMDR-01/10; 2001.
[18] Rosebrook KR, Kutter BL. Soil–foundation–structure interaction:
shallow foundations. Centrifuge Data Report for test series KRR03,
University of California, Davis, Report No. UCD/CGMDR-01/11; 2001.
[19] Gazetas G. Foundations vibrations. In: Fang H-Y, editor. Foundation
engineering handbook. New York: van Nostrand Reinhold; 1991.
Chapter 15.
[20] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic
response analyses. Report No. EERC 70-10, Earthquake Engineering
Research Center, UC Berkeley; 1970.
[21] Cremer C, Pecker A, Davenne L. Cyclic macro-element of soil
structure interaction: material and geometrical nonlinearities. Int
J Num Anal Meth Geomech 2001;25:1257–84.
[22] Houlsby GT, Cassidy MJ. A plasticity model for the behavior of
footings on sand under combined loading. Geotechnique 2002;52(2):
117–29.