You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.


Modeling critical-state shear strength behavior of compacted silty sand via

suction-controlled triaxial testing

Article  in  Engineering Geology · December 2017

DOI: 10.1016/j.enggeo.2017.10.011


5 297

4 authors, including:

Ujwalkumar Patil Anand Puppala

University of Texas at Arlington University of Texas at Arlington


Aravind Pedarla


Some of the authors of this publication are also working on these related projects:

U.S Army Corps of Engineers - Joe Pool Dam, Texas View project (pls look for my information by searching with my last name, thanks) View project

All content following this page was uploaded by Ujwalkumar Patil on 23 October 2017.

The user has requested enhancement of the downloaded file.

This manuscript is published in Engineering Geology (2017): Vol. 231, pp. 21-33 and available online.
Engineering Geology xxx (2017) xxx-xxx

Contents lists available at ScienceDirect

Engineering Geology
journal homepage:

Modeling critical-state shear strength behavior of compacted silty sand via
suction-controlled triaxial testing
Ujwalkumar D. Patila⁠ ,⁠ ⁎⁠ , Anand J. Puppalab⁠ , Laureano R. Hoyosb⁠ , Aravind Pedarlac⁠

University of Guam, School of Engineering, UOG Station, Mangilao 96923, Guam, United States
Department of Civil Engineering, University of Texas at Arlington, Arlington, TX 76019, United States
University of Texas at Arlington, TX 76019, United States


Keywords: Most of the recently postulated unsaturated shear strength models have been calibrated only for a short vari-
Soil suction ety of soils. In addition, these models are yet to be extended and calibrated over a wider range of matric and
Suction-controlled triaxial testing total suction states. The present work focuses on further refinements of previously proposed shear strength equa-
Axis-translation tions in light of newly obtained experimental evidence of shear strength behavior of compacted silty sand at a
critical state from suction-controlled triaxial tests conducted between 0.05 MPa to 300 MPa suction range. A re-
Shear strength modeling
fined and rather simple equation comprising two independent functions, is introduced and validated, including
a thorough parametric investigation, to predict the unsaturated shear strength of compacted silty sand at a crit-
ical state for a wide range of matric and total suction states. The experimental program included a total of 21
consolidated drained (CD) triaxial tests conducted on statically-compacted specimens of silty sand under strain-
and suction-controlled conditions. Experimental results show that the angle of internal friction (ϕ′) remained

virtually constant over the entire range of induced suction states; however, the shear strength increased while
the angle of internal friction with respect to suction (ϕb⁠ ) decreased with increasing suction, with both varying
non-linearly. Finally, a gradual increase in brittleness of the test soil at peak-failure condition, as well as an in-
creasingly marked strain-softening post-failure, was observed with increasing suction.

knowledge of shear strength of soil at residual suction. Although they


1. Background and scope could predict shear strength for different soils with residual suction ψr⁠
varying between 0.1 MPa to 10 MPa, the method appears to be most
The pioneering frameworks postulated by Bishop (1959) and suitable for coarse- to medium-grained soils with relatively low residual
Fredlund and Morgenstern (1977) require expensive, time consuming suction that can be easily attained in laboratory. The procedure is lim-
experimental studies to determine the shear strength parameters and ited to soils whose increase in shear strength varies mostly between air
properties of unsaturated soils. In an effort to make the process more entry and residual suction, while remaining constant beyond residual

cost-effective, several researchers have attempted to use the soil-wa- suction, which may not necessarily be the case, as manifested in experi-
ter retention curve (SWRC) as an interpretative tool, along with the mental test results from present research.
saturated shear strength parameters, c′ and ϕ′, and thus develop rela- Nishimura and Fredlund (2003) used vapor equilibrium technique
tively simple and readily available shear strength equations or models to perform triaxial tests on silty soil at total suction ψ = 39 MPa and
for unsaturated soils (for e.g., Vanapalli et al., 1996; Fredlund et al., five different net normal stresses; however, to the authors' best knowl-
1996; Oberg and Sallfors, 1997; Khalili and Khabbaz, 1998; Vilar, 2006; edge, the results were not used to verify any of the shear strength
Houston et al., 2008; Alonso et al., 2010; Sedano and Vanapalli, 2011; prediction models available in the literature. Blatz et al. (2002) per-

Han and Vanapalli, 2016). formed triaxial tests on compacted sand-bentonite specimens that were
Rassam and Cook (2002) presented a new model by adopting a meant to be used as buffer for nuclear waste by using salt solutions
power additive function to predict shear strength of unsaturated soils of different concentrations to induce high suction between 5.0 MPa
based on the assumption of ϕb⁠ = 0 at residual suction, and with to 42.4 MPa at high net confining cell pressures (0.5, 1.0, 2.0, and

⁎ Corresponding author.
Email addresses: (U.D. Patil); (A.J. Puppala); (L.R. Hoyos); (A. Pedarla)
Received 28 September 2016; Received in revised form 10 October 2017; Accepted 11 October 2017
Available online xxx
0013-7952/ © 2017.
U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

3.0 MPa). They could best-fit three-dimensional surface that predicted

peak shear strength as a function of mean stress and suction. 2.2. Specimen preparation for SWRC and triaxial tests
More recently, Han and Vanapalli (2016) proposed a new methodol-
ogy to predict the non-linear stiffness-suction and shear strength-suction Dry test soil was hand mixed with distilled water at a water content
relationship of unsaturated soils, within the lower suction range, using of 14.2% (+ 2% of optimum), sealed in airtight zip lock bags and kept
a normalized function formulated with ‘suction times exponential de- in a 100% humidity chamber for at least 24 h to attain moisture equilib-

gree of saturation. In addition, they could account for various influenc- rium prior to static compaction. Specimens were statically compacted in
ing factors including external stress, soil structure, anisotropy, hydraulic nine equal lifts via stress-based approach by using a 50 kN load frame,
hysteresis, and testing techniques. Likewise, the effective degree of sat- at a constant rate of 1 mm/min, to a final vertical stress of 1600 kPa,

uration has been used in past to interpret and predict stiffness-suction producing homogenous specimens with overconsolidation stress history
and shear strength-suction relationships for unsaturated soils (Fredlund (Patil, 2014). The initial voids ratio for all specimen varied between
et al., 1996; Vanapalli et al., 1996; Alonso et al., 2010). 0.46 and 0.49.
Recently, a novel suction-controlled triaxial system that can accom- Likewise, Identical SWRC specimens (2 cm diam. And 1 cm ht.) were
modate both axis-translation and relative-humidity techniques has been prepared using the same stress-based approach but compacted in a sin-
implemented at the University of Texas at Arlington. The operational gle lift. Each specimen was compacted to a dry unit weight of 1.80 g/
functionality of this equipment and its capability to replicate test re- cm3⁠ (112.4 lb./ft3⁠ ) with negative pore water pressure, and hence matric

sults using both axis-translation and relative-humidity techniques have suction, between 8 and 10 kPa at water content of 14.2%.
been thoroughly verified through a short series of tests conducted over
a wide range of controlled suction states, between 0.05 and 300 MPa 2.3. Soil water retention curve (SWRC)
(Patil, 2014; Patil et al., 2015; Patil et al., 2016a, b). Test specimen
preparation, preconditioning, suction equalization, and unsaturated tri- Tempe Cell device was used to assess the soil-water retention curve
axial testing procedures via both the axis-translation and relative-hu- (SWRC) of the silty sand up to matric suction values of 500 kPa. A cus-
midity technique has been thoroughly described by Patil et al. (2016b). tom-designed relative humidity (RH) chamber, along with an in-house
The chief objectives of the present work can be summarized as fol- fabricated Plexiglas chamber, was used to obtain remainder of SWRC

lows: (1) To present a comprehensive set of experimental data from data in the higher suction range (“Residual Zone”), beyond 10,000 kPa
densely compacted (overconsolidated) silty sand using the newly imple- suction (Patil, 2014). A new specimen was prepared with identical ini-
mented triaxial system, which is suitable for testing unsaturated soils tial conditions and then used to determine water retention capacity at
well beyond residual suction state; (2) To analyze these results to test each matric suction or total suction level by drying it; thus, each point
the efficacy of previously postulated equations for unsaturated shear on the SWRC represents one test.
strength over a wide range of suction states; and (3) To refine the exist- SWRC models by Fredlund and Xing (1994) and van Genuchten
ing equations in light of new experimental evidence from pertinent soil. (1980) were used to best fit the experimental points along drying path
All the models analyzed in the present work were tested and parametri- in order to complete the characteristic curve over the entire range of
cally investigated by using experimental data at critical state failure. soil suction, from 0 to 10,000 Mpa, as shown in Fig. 1. The best fitting
Six different proposed models in their original form (i.e., Vanapalli parameters are summarized in Table 1. As has been extensively docu-

et al., 1996; Fredlund et al., 1996; Khalili and Khabbaz, 1998; Vilar, mented, an increase in matric suction causes the wetted area of contact
2006; Houston et al., 2008; and Sedano and Vanapalli, 2011) were first between soil particles to decrease and vice versa; hence, there exists a
evaluated in the matric suction range (ψm ⁠ = 0 to 1500 kPa) by plotting non-linear relationship between the soil-water retention curves and the
soil suction on natural scale, and then further extended to assess their shear strength of soil as it changes from saturated to unsaturated state
validity in the high total suction range (ψt⁠ = 20 and 300 MPa) by plot- (Vanapalli et al., 1996).
ting soil suction on logarithmic scale. Mainly, two unsaturated soil shear

strength parameters are analyzed: first, increase in cohesion intercept

(c″) i.e. suction-induced increase in shear strength with increase in suc-
tion; and second, decrease in angle of internal friction with respect to
suction (i.e. ϕb⁠ ).
In lieu of costly setups and time-consuming experimental proce-
dures, the availability of verified predictive models, with reasonable ac-
ceptability over wider suction states and soil types, could provide prac-

ticing engineers with a quick preliminary estimate of the shear strength

of unsaturated soil: a practically viable tool that is much needed in im-
plementing the said subject in actual practice (Vilar, 2006; Houston et
al., 2008).

2. Test material and variables


2.1. Soil properties

The test material used in this work consisted of 45:55 mixtures by

dry mass of fines (37% silt and 8% non-plastic clay) and coarse material
(fine sand), classifying as silty sand (SM) according to the Unified Soil
Classification System (USCS). Standard Proctor compaction tests indi-
cated a maximum dry density of 1.87 g/cm3⁠ , at an optimum water con-
tent of 12.2%.
Fig. 1. Soil-water retention curve (SWRC) along drying path for compacted silty sand test

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

Table 1 taken as a measure of openness of unsaturated soil fabric and influence

Best-fit parameters for selected SWCC functions.
the critical deviator stress the unsaturated soil can sustain at a given
van Genuchten (1980) Fredlund and Xing (1994) mean average skeleton stress.
Alonso et al. (2010) linked the relationship between effective stress
α = 0.036 α = 55 and soil microstructure in unsaturated soils via effective degree of sat-
n=1 n = 0.75 uration, by conceptually differentiating between the volume of water

m = 0.5 m = 1.9
existing in the soil for a given suction into two parts: the free water,
θr⁠ = 0.3 θs⁠ = 16.5
θs⁠ = 16.5 Ψr⁠ = 2000 kPa
partially filling the macro-pores, and the ‘immobile’ water, closely at-
tached to the clay minerals. They were successful in interpreting ex-

perimental data on the strength and stiffness changes with suction for
a variety of soils including granular and high-plasticity clays. More re-
In the present work, target values of matric suction, varying from
cently, Alonso et al. (2013) incorporated microstructural information
50 to 750 kPa, and target values of total suction, at 20,000 and
within the conceptual framework to reproduce the compression behav-
300,000 kPa, were induced on identically prepared specimens of com-
ior of compacted soils.
pacted silty sand prior to monotonic shearing.
Fredlund and Morgenstern (1977) treated net normal stress (σ − ua⁠ )
and matric suction (ua⁠ − uw ⁠ ) as two independent stress state variables
2.4. Experimental variables

in the assessment of their role in the mechanical response of unsaturated
soil. Such an approach separates the effect due to changes in net nor-
Unsaturated soil testing typically involves three essential tests. First,
mal stress from those due to change in pore-water pressure, and enables
SWRC tests that establish the relationship between water holding capac-
the independent assessment of the effect of suction and normal stress on
ity of soil with change in suction, i.e. explaining hydraulic behavior; sec-
volume change. Shear strength is hence expressed as given in Eq. (1).
ond, suction-controlled one dimensional or isotropic consolidation tests
that explain suction-controlled volumetric response; and thirdly, shear
strength tests, such as direct shear test or triaxial shear test modified
to impose and maintain suction within the specimen, that are used to where τf⁠ = shear strength at failure; c′ = effective cohesion; ϕ′ = ef-

quantify strength-deformation variation, i.e. explaining mechanical be- fective friction angle associated with net normal stress (σ − ua⁠ )f⁠ on the
havior. failure plane at failure; and ϕb⁠ = friction angle that captures the contri-
In the present work, and as previously stated, tests were conducted
bution of matric suction to shear strength. Eq. (1) can further be modi-
using a fully-automated, double-walled triaxial test equipment that ac-
fied as Eq. (2).
commodates the essential modifications for unsaturated soil testing, in-
cluding high-air-entry (HAE) ceramics in the bottom pedestal; pore-wa- (2)
ter pressure control; pore-air pressure supply via the top cap; and dif-
fused-air flushing assembly (Patil, 2014). With the operational appara- where c″ = τu⁠ s is defined as the capillary cohesion describing shearing
tus, a comprehensive series of saturated and unsaturated consolidated resistance arising from capillarity effects given by Eq. (3).

drained (CD) triaxial tests on statically compacted silty sand specimens

was performed. (3)
The axis-translation technique was used to impose and control ma-
tric suction in the range of 0 to 750 kPa. Three confining pressures were Figs. 2–4 shows all the Mohr circles obtained at critical state under
used: 100, 200 and 300 kPa. The same RH chamber used for SWRC test- net confining pressures of 100, 200 and 300 kPa, and constant matric
ing was integrated into the new triaxial testing system and a second se- suctions of 50, 250, 500 and 750 kPa, as well as high total suctions of
ries of CD tests, at higher values of total suction (20 MPa and 300 MPa), 20 and 300 MPa. Clearly, the angle of internal friction (ϕ′⁠ ), that is slope

were performed under three different confining pressures: 100, 200 and of failure line, remains constant, irrespective of the applied suction and
300 kPa. confining pressure. However, there is an upward shift in failure lines at
The convention adopted to designate specimens for matric suc- critical state, with increasing suction and at same confining pressure, re-
tion-controlled triaxial testing is “CDx⁠ -y” while specimens for total suc- sulting an increase in apparent cohesion (Figs. 2–4). The values of aver-
tion-controlled testing using relative humidity technique were desig- age apparent cohesion (c″), and the corresponding average values of the
nated as “CDRHx⁠ -y.” Here, “CD” denotes the consolidated drained test, angle of internal friction with respect to soil suction (ϕb⁠ ), as interpreted

“x” represents the net confining pressure (σ3⁠ − ua⁠ ), while “y” represents from these Figs. 2–4 using Eq. (3), are summarized in Table 2, and will
the imposed constant matric suction (ψm ⁠ ) or total suction (ψt⁠ ). Further- be used henceforth in all further analysis to relate the rate of change in
more, the terms “matric” or “total” suction essentially refer to the partic- shear strength with respect to change in matric suction.
ular “technique” that was used to impose such suction, i.e., “axis-trans- Although, not shown here (and not used in further analysis), Mohr
lation” or “relative humidity” based technique, respectively. It is also circles were also drawn at critical state under different matric suctions
worth noting that chemical analyses conducted on the silty sand soil of 50, 250, 500, and 750 kPa as well as high total suctions of 20 and
showed no presence of salts, hence the osmotic component of suction 300 MPa and under each of the constant net confining pressures 100,

can be expected to be negligible. 200, and 300 kPa. The values of apparent cohesion (c″), and the cor-
responding values of the angle of internal friction with respect to soil
3. Unsaturated test results suction (ϕb⁠ ), as interpreted from these figures using Eq. (3), are summa-
rized in Table 3. It should be noted that the values of apparent cohesion
3.1. General shear strength response for 0–300 MPa suction range and “average apparent cohesion” are not exactly same, because the for-
mer has effect of net confining pressure which is not similar while the
Nuth and Laloui (2008) presented a detail review of the histori- latter is obtained by neutralizing this effect. Nonetheless, it can be ob-
cal developments of the effective stresses with the objective of deter- served that the non-linear decrease in ϕb⁠ with increasing suction is in
mining a proper stress framework for constitutive modeling of unsatu- accordance with previous findings (i.e., Escario and Saez, 1986; Gan et
rated soils. Gallipoli et al. (2008) introduced the term capillary bond- al., 1988; and Houston et al., 2008).
ing which is uniquely related to the voids ratio of unsaturated soil to
that of saturated soil. They demonstrated that capillary bonding can be

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

Fig. 2. Critical state Mohr circles under net confining pressures, (σ3⁠ − ua⁠ ) = 100, 200, and 300 kPa: (a) ψ = 50 kPa; (b) ψ = 250 kPa.


Fig. 3. Critical state Mohr circles under net confining pressures, (σ3⁠ − ua⁠ ) = 100, 200, and 300 kPa: (a) ψ = 500 kPa; (b) ψ = 750 kPa.

Fig. 4. Critical state Mohr circles under net confining pressures, (σ3 − ua) = 100, 200, and 300 kPa: (a) ψ = 20 MPa; (b) ψ = 300 MPa.

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

Table 2
Experimentally obtained average values of apparent cohesion.

Soil confining Average Average angle of friction
suction, pressure, apparent with respect to suction,
MPa kPa cohesion, kPa deg.

(ψ) (σ3⁠ − ua⁠ ) (c″a⁠ vg) (ϕb⁠ )

0.05 100, 200 & 20 21.8


0.25 100, 200 & 37 8.4
0.50 100, 200 & 64 7.3
0.75 100, 200 & 88 6.7
20 100, 200 & 120 0.3
300 100, 200 & 150 0.03


Table 3
Experimental values of unsaturated shear strength parameters.

Soil Apparent
Net confining suction, cohesion, Angle of friction with

pressure, kPa MPa kPa respect to suction, deg.
(σ3⁠ − ua⁠ ) (ψ) (c″) (ϕb⁠ )

100 0.05 19 20.8

0.25 39 8.9
0.5 59 6.7
0.75 99 7.5
20 102 0.29
300 140 0.03
200 0.05 31 31.8
0.25 53 12

0.5 80 9
0.75 131 9.9
20 140 0.4
300 208 0.0397
300 0.05 21 22.8
0.25 50 11.3
0.5 73 8.3
0.75 96 7.3

20 162 0.46
300 180 0.03

Fig. 5(a) illustrates the nature of the increase in shear strength

of compacted silty sand with increasing suction between
50 and 300,000 kPa, and under net confining pressure of 300 kPa. The

remaining stress-strain and volume change curves along with detailed Fig. 5. (a) Stress-strain response and (b) Variation of brittleness index of SM soil from suc-
discussion can be obtained in Patil et al. (2016a, b). It is worth clarify- tion-controlled CTC tests at different matric and total suctions.

ing that most suction-controlled, consolidated-drained (CD) triaxial tests

performed in this research were extremely time-consuming, with each The rate of increase in shear strength (i.e. c″) retards as total suc-
test taking at least a month to complete, after pore-fluid equalization; tion is increased between 20 and 300 MPa and under net confining pres-
hence, critical state was identified as soon as the deviator stress reached sure of 300 kPa, as indicated in Tables 2 and 3. Beyond residual suction,
an “apparent” critical state condition (i.e., further shearing of soil sam- the specimen has adsorptive water that is mostly associated with mi-
cro-pores (Alonso et al. 2013) and hence remains ineffective in relaying

ple under virtually constant stress.

Increase in matric suction beyond air-entry suction, introduces air suction to soil grains, thereby possibly retarding the rate of increase in
in the pores and forms menisci around the solid grains that pulls the shear strength with any further increasing suction.
particle together. This causes an apparent increase in normal forces at
grain point contact resulting in an increase in effective stress. Hence, 3.2. Effect of suction on post-peak strain softening and brittleness
around 100 to 125 kPa of matric suction, there is a remarkable increase
in shear strength, irrespective of net confining pressure. However, be- The soil experiences strain softening when there is a reduction of
yond approximately 2000 kPa suction (which also happens to be resid- deviator stress from peak to critical stress during shearing. Further in-
ual suction for the test soil), this increase in shear strength becomes only sight on experimental stress-strain curves from this work strongly sug-
gradual with further increase in suction. In addition, there is an increase gests an increase in magnitude of strain-softening with increasing suc-
in shear strength with increasing net confining pressure at constant suc- tion, as shown in Fig. 5(a). Irrespective of net confinement applied, an

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

increase in suction is observed to have a more pronounced influence stress increase gradually up to ψm ⁠ = 500 kPa, with a dramatic increase
on peak shear stress than on critical shear stress. One way to quantify in both deviator stresses between ψm ⁠ = 500 to 750 kPa, and thereafter
brittleness is to assess the “brittleness index” (IB⁠ ) as defined by Bishop the increase is gradual up to ψt⁠ = 300,000 kPa, but at comparatively
(1971): much higher rate for peak deviator stress than for critical stress; clearly
manifesting increase in strain softening response with increase in suc-
(4) tion. In addition, the increase in net confinement from 100 to 300 kPa

causes the difference between two deviator stresses to gradually in-
crease at constant suction.
where, qp⁠ eak = peak deviatoric stress and qr⁠ = residual deviatoric
The analysis of test results in p′-q plane, revealed the mobilized

stress ratio at peak failure, ηp⁠ eak and at critical state, M equal to 1.60
Considering, qr⁠ as qc⁠ s = critical state deviatoric stress for experi-
ments in this paper, we can quantify IB⁠ and it varies between 0 and 1. and 1.42, respectively (Patil et al., 2016a, b). The stress ratio is use-
With decrease in value of brittleness index towards zero, the failure be- ful in identifying volumetric behavior during shearing. For instance,
havior becomes increasingly ductile. On the other hand, specimen fail- when ηp⁠ eak > M, plastic softening (Fig. 5a) and volumetric dilation oc-
ure will become increasingly brittle with increasing value of IB⁠ . For in- curs on yielding under shear (Patil et al., 2016a, b). The partially sat-
stance, IB⁠ = 1 indicates very brittle behavior. Fig. 5(b) clearly indicate urated test soil showed post-peak softening stress-strain response (Fig.
that the brittleness index increases with increasing matric/total suction, 5a) and initial compression followed by dilational volumetric response

irrespective of net confining pressure with maximum brittleness index under shear (Patil et al., 2016a, b). Such a response may be attributed
value at highest total suction applied i.e. 300 MPa. to the relatively dense or overconsolidated stress history of the spec-
Under constant net confinement, continuous shearing beyond peak imens (as explained in section 2.2). Furthermore, the observed dila-
shear stress drastically disturbs and destroys the air-water menisci im- tancy forms the basis of additional energy manifested in terms of peak
posed via axis-translation technique and formed around contact of solid strength (Schofield and Wroth, 1968; Atkinson, 2007).
grains. Hence, the contribution of suction towards strength at peak It is observed from Table 2 that there is a gradual, nonlinear (hy-
failure is often greater than at critical state; which is clearly illus- perbolic) reduction in ϕb⁠ from 50 kPa to 750 kPa of matric soil suction,
while there is a sharp reduction in ϕb⁠ over the high suction range from

trated by Fig. 6(a), (b) and (c). Both the peak and critical deviator

Fig. 6. Effect of suction on peak and critical deviator stress at different net confinements: (a) (σ3⁠ − ua⁠ ) = 100 kPa, (b) (σ3⁠ − ua⁠ ) = 200 kPa, and (c) (σ3⁠ − ua⁠ ) = 300 kPa.

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

20 MPa to 300 MPa. The shear strength variation is primarily due to the dried samples of Brazilian soils. The model was successfully used to pre-
nonlinear change in cohesion intercept with suction. dict the increase in shear strength/cohesion intercept, based on the fol-
lowing hyperbolic mathematical Eq. (5).
3.3. Effect of suction on failure mode
It is well known that, in addition to the external loading, an increase

in suction tends to apply capillary force on the solid grains at their point where c″ = c(ψ) = the cohesion intercept function of the soil suction,
of contacts, thereby inducing extra bond among them, and thus render-
ψ = ua⁠ − uw ⁠ , c′ = effective soil cohesion (at ψ = 0). Parameters a and
ing a stiffer soil structure that manifests brittleness during shearing. The
b are curve-fitting parameters obtained from tests using saturated and

growth of brittleness in specimen with increasing suction is clearly visu-
air-dried specimens. However, in the present work, similar to Reis et
alized from pictures taken at the end of the tests that indicate gradual
al. (2011), the parameters a and b were obtained from a best-fit analy-
transition in failure zones (i.e. ductile failure mode with barrel shape
sis, by plotting ψ/(c(ψ) − c′) versus ψ > 0 and fitting the experimental
or bulging at center under low to medium suctions and no barrel effect
points with a straight line given by the least square method.
under high suction) after shear failure, as illustrated in Figs. 7(a)–(d).
Although, Eq. (5) uses matric suction, attempt was made to apply
The specimen with highest value of total suction are expected to exhibit
the same equation and procedure to extend the experimental results
maximum brittle failure, which is precisely the case as visually manifest
up to total suction of 300,000 kPa by plotting soil suction on logarith-

from Fig. 7(f).
mic scale. The average value of apparent increase in shear strength
Specimens sheared in the matric suction range ψm ⁠ = 50–750 kPa
(c″) was obtained by drawing a tangent at approximately ϕ′ = 35°.
show development of multiple shear failure planes with barreling effect
to Mohr's stress circles at critical state failure at net confinements of
in the shear zone, while those sheared at total suctions ψt⁠ = 20,000 and
(σ3⁠ − ua⁠ ) = 100, 200 and 300 kPa for each suction imposed (Table 3).
300,000 kPa developed distinct single shear failure plane. Also, the brit-
The parameters a and b are captioned in Fig. 8, along with reasonably
tleness increased the amplitude of post-peak softening which is evident
good correlation between the predicted and experimental results with
from increase in difference between magnitude of deviator stress at peak
R2⁠ = 0.99.
and critical state failure, as manifested from Figs. 6(a)–(c).
Vilar (2006) used direct shear test data from Escario and Juca (1989)

on Madrid gray clay and Guadalix red clay that was extended to suctions
4. Verification of shear strength models
of about 11,000 and 8000 kPa respectively, far above their respective
residual suction value, and then compared with these predictions from
4.1. Vilar (2006) model - approach I
Eq. (5), with reasonable success. Apart from this exception, to the au-
thor's best knowledge, none of the existing models have been tested over
Vilar (2006) proposed an hyperbolic equation that considers matric
a wider range of suction (0.05 MPa to 300 MPa) as such used in this
suction contribution towards an increase in peak shear strength by an
research; thus, making the present virtually the first such attempt, espe-
increase in apparent cohesion to best fit experimental data from air-
cially with specimens prepared in an identical way (e.g., similar com-
paction method, initial water content and initial voids ratio) and vali-

dated using unsaturated triaxial test data. It is worth noting that, when
results are plotted beyond residual suction value, the shape of the curve
resembles that of the SWRC (Fig. 8).

Fig. 7. Typical specimen failure at different suction states: (a) CD300-50, (b) CD300-250, Fig. 8. Variation of cohesion intercept with matric suction (Vilar, 2006; approach I) from
(c) CD300-500, (d) CD300-750, (e) CDRH300-20 MPa, (f) CDRH300-300 MPa. average value of c″ at same suction but different confining pressures.

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

paper at critical state.

4.2. Vilar (2006) model – approach II and proposed modification

Vilar (2006) also proposed a model that use maximum measured (7)
suction and modified the parameter b as in equation below

(6) The parameters “a” and “b” are determined by using transformed lin-
ear plot of ψ∗⁠ against ψ ∗⁠ /(ϕ′ − ϕb⁠ ). Experimental values of ϕb⁠ are first

calculated. For instance, for ψ = 50 kPa, and σ3⁠ − ua⁠ = 100 kPa from
Alternatively, the predictions at critical state are made through Eqs. Fig. 2; ϕb⁠ = arctan (19/50) = 20.8°.
(2), (3), and (6) and shown in Fig. 9. For the sake of brevity, and to com- Houston et al. (2008) were able to use this model quite successfully
pensate the effect of the net confining pressure, only one plot is shown on triaxial shear strength experimental data from unsaturated CL-ML,
with average c″ value obtained by plotting Mohr's-stress circles at three SM, CL and SP soils, as well as on data from previous researchers (e.g.,
net confining pressures, i.e., (σ3⁠ − ua⁠ ) = 100, 200 and 300 kPa for each Gan et al., 1988; Escario and Juca, 1989; Oloo and Fredlund, 1996; and
suction level (Table 3). It was also cautioned that in granular soils the Thu et al., 2007). However, the data set was limited to a matric suction

model could yield conservative results. In agreement to the suggestion, range < 1500 kPa. As such, the model remains to be validated for suc-
it can be observed that predictions from model were higher as compared tion values beyond 1500 kPa.
to experimental results with R2⁠ = 0.69. Therefore, the value of “a” can Houston et al. (2008) also explained the physical significance of cor-
be modified to a = 2.5/tanϕ′ instead of the one obtained from Eq. (2). relating “b” with ϕ′ using a simple equation, b = 1/ϕ′. However, since
As seen in Fig. 9, a good agreement was obtained between the predicted they lacked the data beyond residual suction range, caution was sug-
response and measured test response with R2⁠ = 0.96 using modified gested while using the Eq. (7) for dry soils or in high suction range.
value of parameter “a” in Vilar (2006), Eq. (6). It is worth mentioning From present work, when the parameter was evaluated using maximum
here that by coincidence, the Vilar (2006) approach I and approach II suction as 750 kPa, the value of ϕ′ obtained from reciprocal of b var-

with proposed slight modification yields almost similar predictions. ied between 29 and 40°. However, when maximum suction was taken
as 300,000 kPa, equation, b = 1/ϕ′, resulted in a value of ϕ′ = 35°. Re-
markably, as can be observed from Fig. 10, irrespective of the confining
4.3. Houston et al. (2008) model and proposed modification pressure (σ3⁠ − ua⁠ = 100, 200 and 300 kPa), the value of the best fitting
parameter b, when fitted over the entire suction range (i.e. b = 0.0286)
Houston et al. (2008) proposed a hyperbolic function, as given by remained virtually constant (see Table 4), and is in agreement with 1/
Eq. (7), to predict the angle of friction with respect to suction, ϕb⁠ that b = ϕ′ = 35°., as postulated by Houston et al. (2008).
can be used along with the extended Mohr-Coulomb shear strength To extend the results in high suction range, i.e. well beyond residual
equation proposed by Fredlund and Rahardjo (1993) to further pre- suction, the soil suction was plotted on logarithmic scale to avoid ma-

dict the increase in suction-induced peak shear strength, provided that tric suction data points getting concentrated in narrow range, thus pro-
the air entry value (AEV) is known, along with the saturated effective viding a better visual representation of ϕb⁠ over the entire suction range,
stress parameters. It should be noted that analysis is performed in this as shown in Fig. 10. Good correlations are observed between predic-
tions and experimental results with R2⁠ = 0.98 over the entire suction
range of 50–300,000 kPa for net confinement of (σ3⁠ − ua⁠ ) = 300 kPa,
as shown in Fig. 10.
Fig. 11 shows the variation of gravimetric water content (w), angle

of friction with respect to suction (ϕb⁠ ), and apparent cohesion with in-
creasing suction based on model predictions as originally postulated by
Houston et al. (2008) and Vilar (2006). It can be observed that as the
suction increases, the value of ϕb⁠ and water holding capacity of soil de-
creases non-linearly. Striking reductions in both, ϕb⁠ and water holding
capacity of soil are observed beyond AEV of soil as expected for sandy

soil. However, beyond the residual suction, the further increase in suc-
tion has low impact on reduction in ϕb⁠ : an indication that soil had at-
tained its maximum strength.
In the present work, no reduction in strength was observed even
with imposed suction as high as 300,000 kPa (maximum total suction
that can be applied and maintained constant throughout testing using
relative humidity technique). For better visual comparison purposes, the

y-axis scales were adjusted so that, irrespective of their different individ-

ual range, all three plots could be accommodated within same plot size.
Again, the shape of shear strength curves resembles that of the SWRC.

4.4. Fredlund et al. (1996) model – approach I and proposed modification

Fredlund et al. (1996) and Vanapalli et al. (1996) proposed a non-

linear function for predicting the peak shear strength of unsaturated
Fig. 9. Variation of cohesion intercept with matric suction using Vilar, 2006: approach II, soil, utilizing the entire soil-water retention curve with suction be-
and approach II: modified. tween 0 and 1000,000 kPa, along with saturated shear strength para

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

content = θw ⁠ /θs⁠ ; where, θw

⁠ = volumetric water content obtained using
Fredlund and Xing (1994) equation. Third term of the Eq. (8) represents
the shear strength contribution, τu⁠ s.


Fredlund et al. (1996) compared the experimental peak shear
strength test results on glacial till from a series of multistage tests us-
ing modified direct shear test apparatus with predictions from Eq. (9).

A thorough parametric study yielded good correlations between experi-
mental and predicted peak shear strength with ĸ = 2.2.
In the present work, a new function is postulated based on previous
equations proposed by Fredlund et al. (1996), Vanapalli et al. (1996),
and Sedano and Vanapalli (2011), including critical modifications to ex-
tend their usefulness into the high suction range. Eq. (9) was used to
predict shear strength values at critical state, which were then compared
with experimental results from this work. For instance, for ψ = 50 kPa;

Θ = θw ⁠ /θs⁠ = 18.5/30.2 = 0.61; τu⁠ s = [50 ∗ (18.5/30.2)1⁠ .21 ∗ tan
(35°)] = 19.34 kPa. However, using one value of ĸ, over the entire test
suction range, did not yield good correlations, especially beyond resid-
ual suction value. Therefore, function ĸ can be modified to be com-
prised of two different values when using this model: the first value
between the AEV (i.e. ψ = 10 kPa) and up to residual suction (i.e.
ψ = 2000 kPa), and the second value beyond residual suction, as fol-

TE lows:


As shown in Fig. 12, the experimental increase in shear strength was

reasonably well predicted with R2⁠ = 0.98 for the entire test range, i.e.
0–300,000 kPa. In previous research, “peak” shear strength values were

used for comparison. In the present work, however, the increase in “crit-
ical state” shear strength is used for comparison, thus validating the
model at critical state.

4.5. Sedano and Vanapalli (2011) model – approach II and proposed


Sedano and Vanapalli (2011) introduced Eq. (11), very similar to Eq.
(9), but using degree of saturation, S instead of normalized volumetric
water content (θ), and thus were able to correlate, with reasonable ac-
Fig. 10. (a) Calibration of best-fitting parameters, (b) Variation of cohesion intercept with
matric suction (Houston et al., 2008) from CD100-xx tests. curacy, the critical state shear strength of glacial till obtained from mod-
ified ring shear apparatus using best-fit parameter, ĸ = 5.0, thus vali-
dating the model at critical state failure.

meters, c′ and ϕ′, as shown by Eq. (8). They also suggested the use
of Fredlund and Xing's (1994) equation to plot the best fitting SWRC
through experimental points.
Close observations of experimental test results indicate that the crit-
(8) ical state shear strength is fairly constant beyond 300 kPa while the tar-
get matric suction range was only up to 500 kPa; thus, the validation re-
mains to be addressed in the high suction range. In the present work, an

where, κ = fitting parameter used for obtaining best-fit between the attempt is made to extend the critical state shear failure predictions at
measured and predicted values. Θ = normalized volumetric water critical state failure, and far beyond residual suction, thereby attempt-
ing to validate this model for high suction range.

Table 4
Calibrated parameters for compacted silty sand according to Houston et al. (2008) model.

σ3⁠ –ua⁠ Suction range b ϕ′ = 1/b Suction range b ϕ′ = 1/b

100 50–750 kPa 0.0342 29.2 50 kPa–300 MPa 0.0286 35

200 0.0252 39.7 0.0286 35
300 0.0333 30 0.0286 35

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

4.6. Vanapalli et al. (1996) model - approach III

Vanapalli et al. (1996) proposed another predictive shear strength

(at peak failure) model by eliminating the use of a fitting parameter κ,
as given by Eq. (12) below:


where, θw ⁠ = volumetric water content, θw⁠ = saturated volumetric wa-
ter content, and θr⁠ = residual volumetric water content that can be es-
timated from the SWRC. Similar results were obtained while using the
degree of saturation (S) or gravimetric water content instead of volu-

metric water content in Eq. (12). Both Eq. (11) and (12) are consistent
with the stress-state variable approach and satisfy the continuum me-
chanics concept. However, it has limitations in that increase in shear
strength drops to zero at residual suction and might even become nega-
tive, i.e. decrease beyond residual suction. Vanapalli et al. (1996) tested
the equation within matric suction range of 0–500 kPa and obtained
good correlation between experimental results and predictions. Authors
Fig. 11. Variation of gravimetric water content, angle of friction with respect to suction
and increase in shear strength for test soil over 00–1000 MPa soil suction. made attempt to check the validity of this model for large suction range,

especially beyond residual suction.
Good correlations were obtained only up to ψm Present paper explores suitability of applying above equations at
⁠ = 750 kPa. How-
critical state failure. As shown in Fig. 15, good correlations were ob-
ever, much higher and unrealistic predictions were obtained in the
tained between experimental and predicted results for range of matric
high suction range. It is clearly observed from the SWRC of the test
suction up to ψ = 750 kPa. The nature of the Eq. (12) suggests its apt
soil (Fig.1) that its water retention capacity changes drastically beyond
residual suction, and hence the value of the fitting parameter ĸ should suitability for predictions of decreasing shear strength beyond residual
suction. Although it was confirmed experimentally in this research that
be adjusted to capture such dramatic change. Hence, a new value of
the shear strength continued to increase with increase in suction be-
ĸ = 1.85 is suggested for suction states far beyond residual suction.
yond residual suction, the comparative predictions showed unrealistic
Parametric investigations of the proposed equation with varying

decrease in shear strength; thus, making this model inappropriate be-

values of ĸ, beyond residual suction, clearly indicate its sensitivity,
yond residual suction for test soil under investigation.
as shown in Fig. 13. Upon using ĸ = 1.2, up to residual suction, and
ĸ = 1.85, beyond residual suction, strong correlations (R2⁠ = 0.98) were
obtained between experimental values and predictions for the entire test 4.7. Khalili and Khabbaz (1998) model and proposed modification
suction range (i.e. ψ = 0–300 MPa), as illustrated in Fig. 14. However,
the shear strength, particularly beyond residual suction drops with any Khalili and Khabbaz (1998) extended Bishop (1959) equation by in-

further increases in value of ĸ, and vice-versa. troducing an empirical constant χ as expressed in Eq. (13).

Fig. 12. Experimental and predicted increase in shear strength with soil suction using Fredlund et al., 1996 model – approach I with modification.

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

However, when applied to results from present research at critical

state, the predictions from original equation, could be improved with
slight modification to the power function used in parameter χ as follows:


The modified equation holds good correlation (R2⁠ = 0.96) with

results from experimental program from this research only up to
ψm⁠ = 750 kPa (0.75 MPa), as illustrated in Fig. 16. Beyond this level of
suction, however, the predictions are unrealistically high, thereby, lim-
iting its use below residual suction.

5. Summary and conclusions

Consolidated drained triaxial tests were conducted on identically

prepared (i.e., similar initial voids ratio, water content and compaction
method (static), unsaturated silty sand specimens over a wide suction
range between 50 and 300,000 kPa and sheared along the CTC stress
path. Shear strength of soil increased dramatically with suction between
air entry value and residual suction. On the other hand, it increased at
Fig. 13. Parametric performance of proposed equation with varying value of ĸ beyond dramatically slower rate beyond residual suction. The specimens under-
residual suction using Sedano and Vanapalli (2011) model – approach II with proposed went post-peak strain-softening that increased with increase in soil suc-
modification. tion.

Hyperbolic equation proposed by Houston et al. (2008), in its origi-
nal form was validated for first time for suction range beyond residual
suction. Slight modification was suggested to Vilar (2006) equation, and
(13) Khalili and Khabbaz (1998) equation to get better predictions that led to
validation of former equation for silty sand test soil over selected wide
suction range. Although, good correlations were obtained using modi-
where, (ua⁠ − uw⁠ )f⁠ = matric suction in the specimen at failure condition, fied Khalili and Khabbaz (1998) equation up to ψm ⁠ = 750 kPa, unreal-
(ua⁠ − uw
⁠ )b
⁠ = air entry value suction of soil. Previously available data istically high predictions were obtained in high suction range; thus, lim-
from 13 different soils indicated that the value of − 0.55 best fits the iting its use in low-medium suction (0–750 kPa) range only for the test

equation over a wider range of soil type. In addition, the air entry value soil.
is needed along with saturated strength parameters to predict the peak A new function is postulated based on previous equations proposed
shear strength using Eq. (14). by Fredlund et al. (1996), Vanapalli et al. (1996), and Sedano and
Vanapalli (2011) including critical modifications to extend their predic-
tions over high suction range. The modified equation uses two different
(14) values of best fitting parameter, ĸ; First, to improve original predictions

up to residual suction and second, beyond residual suction, thereby ex-

tending it to high suction range up to 300 MPa.

Fig. 14. Experimental and predicted suction-induced increase in shear strength up to Ψ = 300,000 kPa using Sedano and Vanapalli (2011) model – approach II with proposed modifica-

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

Fig. 15. Experimental and predicted shear strength using Vanapalli et al. (1996) model – approach III.


Fig. 16. Experimental and predicted shear strength (Khalili and Khabbaz, 1998).

Almost all the models, except the one by Sedano and Vanapalli Award No. 1039956. This support is gratefully acknowledged. Any find-
(2011), were previously validated to predict increase in peak shear ings, conclusions, or recommendations expressed in this material are
strength due to increase in suction; however, this research focusses on those of the authors and do not necessarily reflect the views of the Na-
validating them at critical state. Accurate assessments of unsaturated tional Science Foundation.
strength parameters are of extreme importance in natural slopes in fis-

sured rocks with unsaturated clayey and silty sand fills that can un- References
dergo significant shear strength changes upon wetting, or shallow fis-
sured landslides that can also be activated by wetting. Additional ex- Alonso, E.E., Pereira, J.-M., Vaunat, J., Olivella, S., 2010. A microstructurally based effec-
tive stress for unsaturated soils. Geotechnique 60 (12), 913–925.
perimental evidence of unsaturated soil shear strength on variety of Atkinson, J., 2007. The Mechanics of Soils and Foundations, Second ed. Taylor and Fran-
soils is needed in future to corroborate greater acceptance of proposed cis, 442.
semi-empirical equations in their proposed form or in modified form as Bishop, A.W., 1959. The Principle of Effective Stress. 106(39), Tecknish, Ukebland,
from this article or for their further future enhancement.

Bishop, A.W., 1971. The influence of progressive failure on the choice of stability analysis.
Géotechnique 21 (2), 168–172.
Uncited reference Escario, V., Juca, J.F.T., 1989. Strength and deformation of partly saturated soils. In: Proc.
12th Int. Conf. on Soil Mech. Found. Eng. Vol. 1. Balkema, Rio de Janeiro, pp. 43–46.
Escario, V., Saez, J., 1986. The shear strength of partly saturated soils. Geotechnique 36
Patil et al., 2015 (3), 453–456.
Fredlund, D.G., Morgenstern, N.R., 1977. Stress strain variables for unsaturated soils. In:
Proc. Amer. Soc. of Civil Eng. Vol. 103, No. GT5. pp. 447–466.
Acknowledgements Fredlund, D.G., Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils. Wiley, New
The experimental work described in this paper is part of an ongoing Fredlund, D.G., Xing, A., 1994. Equations for the soil-water characteristic curve. Can. Ge-
otech. J. 31 (4), 521–532.
research project funded by the National Science Foundation under MRI

U.D. Patil et al. Engineering Geology xxx (2017) xxx-xxx

Fredlund, D.G., Xing, A., Fredlund, M.D., Barbour, S.L., 1996. The relationship of the un- Patil, U.D., 2014. Response of Unsaturated Silty Sand Over a Wider Range of Suction States
saturated soil shear strength to the soil-water characteristic curve. Can. Geotech. J. 33 Using a Novel Double-walled Triaxial Testing System, (Ph.D. dissertation submitted to
(3), 440–448. the department of civil engineering, University of Texas at Arlington, TX).
Gallipoli, D., Gens, A., Chen, G., D'Onza, F., 2008. Modelling unsaturated soil behavior Patil, U.D., Hoyos, L.R., Puppala, A.J., 2015. Suitable shearing rate for triaxial testing of
during normal consolidation and at critical state. Comput. Geotech. 35, 825–834. intermediate soils under vapor controlled medium to high suction range. Geotech.
Gan, J.K.M., Fredlund, D.G., Rahardjo, H., 1988. Determination of shear strength parame- Special Publ. ASCE 256, 2141–2150.
ter for unsaturated soil using direct shear test. Can. Geotech. J. 25 (3), 500–510. Patil, U.D., Hoyos, L.R., Puppala, A.J., 2016. Modeling essential elasto-plastic features of
van Genuchten, M.T., 1980. A closed-form equation for predicting the hydraulic conduc- compacted silty sand via suction-controlled triaxial testing. Int. J. Geomech.22, Avail-

tivity unsaturated soils. Soil. Sci. Soc. Amer. J. 44, 892–898. able online
Han, Z., Vanapalli, S.K., 2016. Stiffness and shear strength of unsaturated soils in relation Patil, U.D., Hoyos, L.R., Puppala, A.J., 2016. Characterization of compacted silty sand us-
to soil-water characteristic curve. Geotechnique 66 (8), 627–647. ing a double-walled triaxial cell with fully automated relative-humidity control. Geot-
Houston, S.L., Perez-Garcia, N., Houston, W.N., 2008. Shear strength and shear-induced ech. Test. J. 39 (5), 742–756.

volume change behaviour of unsaturated soils from a Triaxial test program. J. Geot- Rassam, D.W., Cook, F., 2002. Predicting the shear strength envelope of unsaturated soils.
ech. Geoenviron. Eng. 134 (11), 1619–1632. Geotech. Test. J. 25 (2), 215–220.
Khalili, N., Khabbaz, M.H., 1998. A unique relationship for the determination of the shear Reis, R.M., Azevedo, R.F., Botelho, B.S., Vilar, O.M., 2011. Performance of a cubical triax-
strength of unsaturated soils. Geotechnique 48 (5), 681–687. ial apparatus for testing saturated and unsaturated soils. Geotech. Test. J. 34 (3), 1–9.
Nishimura, T., Fredlund, D.G., Leung, et al., 2003. A new triaxial apparatus for high total Schofield, A., Wroth, C.P., 1968. Critical State Soil Mechanics. McGraw-Hill, London.
suction using relative humidity control. In: 12th Asian Regional Conference on Soil Sedano, J.A., Vanapalli, S.K., 2011. The relationship between the critical state shear
Mechanics and Geotechnical Engineering, Singapore. Vol. 1, World Scientific Publish- strength of unsaturated soils and the soil-water characteristic curve. In: Alonso, Gens
ing, pp. 65–68, August 4–8. (ISSN: 9789812385598). (Eds.), Unsaturated Soils. Taylor and Francis Group, London, pp. 253–258.
Nuth, M., Laloui, L., 2008. Effective stress concept in unsaturated soils: clarification Thu, T.M., Rahardjo, H., Leong, E.C., 2007. Critical state behavior of a compacted silt spec-

and validation of a unified framework. Int. J. Numer. Anal. Methods Geomech. 32, imen. Soils Found. 47 (4), 749–755.
771–801. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., Clifton, A.W., 1996. Model for the prediction
Oberg, A., Sallfors, G., 1997. Determination of shear strength parameters of unsaturated of shear strength with respect to soil suction. Can. Geotech. J. 33, 379–392.
silts and sands based on the water retention curve. Geotech.Test. J. 20 (1), 40–48. Vilar, O.M., 2006. A simplified procedure to estimate the shear strength envelope of un-
Oloo, S.Y., Fredlund, D.G., 1996. A method for determination of ϕb⁠ for statically com- saturated soil. Can. Geotech. J. 43, 1088–1095.
pacted soils. Can. Geotech. J. 33, 272–280.



View publication stats