Engineering Geology 229 (2017) 73–84

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Three-dimensional stability analysis of slopes in hard soil/soft rock with MARK

tensile strength cut-off
Dowon Park, Radoslaw L. Michalowski⁎
Department of Civil and Environmental Eng., University of Michigan, 2028 G.G. Brown Bldg, Ann Arbor, MI 48109-2125, United States

A R T I C L E I N F O A B S T R A C T

Keywords: Many collapse analyses of slopes in soils or soft rocks use the classical Mohr-Coulomb yield function to define the
Landslides strength of geomaterials. In the presence of bonded particles and grains, this function predicts uniaxial tensile
Slope stability strength and even greater isotropic tensile strength. Testing for material properties, however, is typically carried
Strength envelope out in the compressive regime; the tensile strength is then burdened by uncertainties, as it is a result of extra-
Tension cut-off
polation of test results into the tensile regime. A three-dimensional limit analysis of slopes is presented with the
Plasticity analysis
3D analysis
geomaterial described by a yield surface with tensile strength cut-off. The multiplicity of admissible collapse
mechanisms is enriched, as the tension cut-off allows construction of mechanisms that include rupture modes.
Stability factors for slopes with tensile strength cut-off are reduced compared to those based on the classical
Mohr-Coulomb strength envelope, with the largest drop for steep slopes subjected to seepage. The stability factor
for a 70-degree slope subjected to seepage can be reduced by as much as 69% when tensile strength cut-off is
considered.

1. Introduction (cylindrical) and log-spiral failure surfaces (Fellenius, 1927; Taylor,

Stability of both natural and anthropogenic slopes is a paramount
problem in both geological and geotechnical engineering, with well-
established literature. Most analyses include two-dimensional me-
chanisms of collapse with the soils' ability to resist yielding described
by a linear strength envelope. Both of these assumptions are relaxed in
this paper: three-dimensional mechanisms of slope collapse are con-
sidered and the soil strength is described by a nonlinear strength en-
velope, truncated in the tensile regime.
Two-dimensional analyses of slope stability are preferred by many,
because of their relative simplicity and conservative outcome.
However, most slope failures have three-dimensional features, sup-
porting the motivation for the development of 3D analyses. In cases
where the extent of the failure mass is restricted, for instance, by ad-
jacent rock formation, a three-dimensional analysis may be called for.
This is certainly the case in excavation slopes; three-dimensional ana-
lyses are also needed when back-calculating soil strength properties
from known instabilities, where plane analyses may return estimates of
strength that are not conservative.
Early observations of slope failures and their analyses all included
two-dimensional mechanisms. Observations of failure in clay embank-
ments led Collin (1846) to believe that the shape of the failure surface is
a cycloid. Early analyses of stability of slopes were based on circular
1937; Drucker and Prager, 1952). Early 3D analyses included undrained
failures (Baligh and Azzouz, 1975; Gens et al., 1988) as the rotational
mechanisms in incompressible materials are easier to achieve: any
surface of revolution forms an admissible failure surface. The analyses
in dilative materials relied on the extension of “slice” techniques to
“column” techniques (Hovland, 1977; Hungr, 1987). A rather inter-
esting approach to rotational failures was presented by Leshchinsky
et al. (1985), who applied a variational approach to finding the critical
failure surface. The kinematic approach of limit analysis with transla-
tional collapse mechanisms was applied in 3D analysis of slopes by
Drescher (1983) and Michalowski (1989), and it was extended later to
rotational collapse (de Buhan and Garnier, 1998; Michalowski and
Drescher, 2009; Michalowski, 2010). The failure surface found by
Leshchinsky et al. (1985) as a result of variational search (see also
Zhang et al., 2016) appears to be a special case of what was postulated
in limit analysis by Michalowski and Drescher (2009). The finite ele-
ment approach to stability analysis brings the advantage of no need to
predetermine the geometry characteristics of the failure surface
(Griffiths and Marquez, 2007; Liu et al., 2017), but it comes with the
issue of non-convergence at the instant of failure. The limit analysis
approach using the finite element framework was found useful in ad-
dressing 3D stability (Li et al., 2010), and this approach can cope with
complex boundary shapes and soil inhomogeneity more easily than the

⁎
Corresponding author.
E-mail address: rlmich@umich.edu (R.L. Michalowski).

http://dx.doi.org/10.1016/j.enggeo.2017.09.018
Received 6 July 2017; Received in revised form 11 September 2017; Accepted 18 September 2017
Available online 20 September 2017
0013-7952/ © 2017 Elsevier B.V. All rights reserved.
D. Park, R.L. Michalowski
Fig. 1. Head scarp of the Twin Sisters Trail shallow landslide, Rocky Mountain National
Park, Colorado (September 2013).
‘analytical’ approach of limit analysis can.
This paper focuses on 3D analysis of slope stability in hard soils and
soft rocks with tensile strength cut-off. The motivation for the devel-
opment of an analysis with tension cut-off stems from the uncertainties
in the tensile strength determined from extrapolation of test results in
the compressive regime. Tension cut-off introduces nonlinearity into
the strength envelope, and allows for kinematic discontinuities with a
large separation velocity relative to shear. This is consistent with ob-
servations of some landslides, which have very steep failure surfaces at
the head of the slide, consistent with a mode of failure similar to top-
pling. One such example is illustrated in Fig. 1.
This shallow slide occurred in September of 2013 in a mountainous
region, and intersected the trail leading to the Twin Sisters peaks in
Rocky Mountain National Park, Colorado. The area was wooded, but
the slide occurred beneath the tree roots, yet did not reach the bed rock;
it was associated with substantial seepage, and the slide was approxi-
mately parallel to the surface. The width of the slide at its head was
about 50 m, with an about-vertical failure surface, two-to-four meters
high. To form a mechanism of collapse consistent with material trans-
lation parallel to the slope, the deformation at the steep head region of
the slide likely involved a large separation component, driving the mass
away from the material at rest, consistent with the presence of tension
cut-off in the material's yield condition. Such deformation at the head of
the slide can no longer be interpreted as the volumetric strain (or di-
latancy), but rather a rupture of the material, more often seen in top-
pling failures. This conjecture is consistent with the presence of tensile
strength cut-off in the yield function. Of course, one could also con-
struct failure mechanisms that might leave a vertical head scarp
without resorting to tension cut-off.
Not relying on the tensile strength in slope stability analyses was
suggested earlier, but it was dealt with in a somewhat indirect manner.
To avoid the presence of tensile stresses in the slope mass, a tension
crack was introduced by Spencer (1968) (also, Duncan and Wright,
2005; Utili, 2013, and Michalowski, 2013). However, this is a very
different approach from the one offered in this paper. An existing ten-
sion crack is part of the geometry of the boundary value problem,
whereas the tension cut-off is part of the material model. The latter is
likely to be more useful, and not only in limit analysis. For instance, to
introduce an existing crack into the finite element analysis, its location
needs to be defined first, whereas the method proposed here removes
the tensile stresses from the slope by eliminating tension from ad-
missible stresses defined by the strength envelope.
The investigation in this paper addresses the question whether the
presence of tensile strength in the yield condition has an influence on
74
Engineering Geology 229 (2017) 73–84
the outcome of the 3D stability analysis of slopes, in particular, whether
stability factors calculated based on the classical M-C yield condition
overestimate those calculated without tensile strength. The novelty in
this paper is in presenting a three-dimensional stability analysis with
tension cut-off, demonstrating failure modes that involve rupture of the
material, and demonstrating that steep slopes subjected to seepage are
most vulnerable to tension cut-off.
2. Tensile strength cut-off
Often misinterpreted as a feature in the failure mechanism, tension
cut-off is a material property. It is an integral part of the yield surface.
The strength of rocks is typically described with non-linear yield en-
velopes (e.g., Hoek and Brown, 1980; Hoek et al., 2002), and the tensile
strength is adjusted depending on the state of the rock. For example, for
weathered rocks, the tensile strength is often reduced or even taken as
zero. For bonded soils, a linear strength envelope in the compressive
regime is used most often, extrapolated into the tensile regime, without
consideration given to the presence or absence of true tensile strength.
Drucker and Prager (1952) suggested eliminating the tensile strength
from the limit stress envelope for soils, while Paul (1961) applied
tension cut-off to the Mohr-Coulomb yield condition to consider the
brittle behavior of rocks. This concept was used later in analyses of
brittle materials (e.g., Chen and Drucker, 1969; Michalowski, 1985;
Chen and Liu, 1990), and it is used in this paper to limit or eliminate
tension from admissible stresses.
The yield surface for bonded geomaterial (e.g., soft rock) in the
principal stress space is presented in Fig. 2(a). It is the Mohr-Coulomb
surface with an additional three mutually perpendicular planes (ABCD
being one of them) limiting the admissible stress states in the tensile
regime, as proposed by Paul (1961). Limitation on tension so conceived
is essentially the Galileo-Rankine tensile strength criterion; a smooth
surface of this kind was discussed recently by Lagioia et al. (2014). A
cross-section of the surface in Fig. 2(a) with a plane intersecting the
triaxial compression (σ1 > σ2 = σ3) and extension (σ2 = σ3 > σ1) mer-
idians is illustrated in Fig. 2(b), along with the octahedral cross-section.
Every point in the space in Fig. 2(a) can be represented with three stress
circles on the σ-τ plane in Fig. 2(c). For example, the stress state at point
K(σ1 , σ2 , σ3) is represented by the three circles shown in Fig. 2(c).
Section SP of the limit circle C3 constitutes a segment of the strength
envelope. Each point on the envelope represents components of a
traction vector on a failure surface in a collapse mechanism, whereas v
depicts the velocity discontinuity vector. As the normality flow rule is
enforced in limit analysis, points on section SP allow deformation with
large normal components relative to shear, with angle δ varying from ϕ
at point S to 90° at point P. Angle δ is not a parameter defining dila-
tancy in the Reynolds (1885) sense. The nature of the deformation
process associated with tension cut-off was not intended by Paul (1961)
to be continual strain, but fracture. Hence, more appropriately, this
deformation should be interpreted as material rupture and separation
(rather than ductile strain associated with dilatancy), and angle δ re-
ferred to as the rupture angle. Deformation with rupture angle δ in the
range of ϕ to 90° is represented in Fig. 2(a) by a fan of strain rate
vectors at point K, contained by two limiting directions marked as s and
p. It will be emphasized later in Section 3 that deformation with large
angle δ does not relax the rigor of limit analysis, and, from the plasticity
standpoint, the result is still a strict bound on the true solution.
Soils that are not bonded have neither tensile strength nor uniaxial
compressive strength, even if substantial dilatancy is derived from grain
and particle interlocking, but a bonded geomaterial can exhibit con-
siderable uniaxial compressive strength and some tensile strength.
However, as mentioned earlier, tensile strength is not a subject of
routine testing; instead, the tensile strength is an outcome of extra-
polating the test results from the compressive regime. This paper ex-
amines what the influence of reducing or eliminating tensile strength
from the strength envelope is on the outcome of stability analyses. A
D. Park, R.L. Michalowski
Fig. 2. (a) Yield surface in the principal stress space for geomaterials with tension cut-off,
(b) cross-sections with diagonal plane and octahedral plane, and (c) stress circles C1, C2,
C3 on τ-σ plane representing a single point on the yield surface in the principal stress
space.
strength envelope for a bonded geomaterial with zero tensile strength is
depicted in Fig. 3(a), whereas Fig. 3(b) shows a partial cut-off with
uniaxial tensile strength of ft. The uniaxial compressive strength is de-
noted in the figure by fc. For convenience, tensile strength is described
in this paper as fraction ξ of what the uniaxial tensile strength is in the
M-C function
2c cos ϕ
ft = ξ
1 + sin ϕ
where ϕ is the internal friction angle and c is the cohesion or the
Engineering Geology 229 (2017) 73–84
Fig. 3. Strength envelopes for bonded material (hard soil/soft rock) with tension cut-off
for drained and undrained analyses (after Michalowski, 2017): (a) zero tensile strength,
(b) partial tension cut-off, and (c) cut-off in yield condition for undrained deformation,
with the circular section representing material rupture.
intercept of the linear portion of the strength envelope with axis τ.
Coefficient ξ = 0 represents full tension cut off (zero tensile strength) as
depicted in Fig. 3(a).
Three-dimensional mechanisms of slope failure considered in this
paper involve rigid rotation with soil yielding along a curvilinear sur-
face. A two-dimensional failure mechanism for soils with tension cut-off
was considered recently in Michalowski (2017). The rate of work dis-
sipation per unit area of the failure surface can be easily found as a dot
product of the traction vector T (Fig. 3(b)) on the failure surface and the
(1)
velocity discontinuity vector v (Michalowski, 2017)
75
D. Park, R.L. Michalowski Engineering Geology 229 (2017) 73–84

1 − sin δ sin δ − sin ϕ ⎞

d = c v ⎛⎜cos ϕ + 2ξ ⎟

⎝ 1 − sin ϕ cos ϕ ⎠ (2)

where v is the magnitude of vector v. This formula holds in the analysis

no matter whether the process governed by the tension cut-off is a
plastic strain or an incipient rupture (or crack opening process), as in-
tended by Paul (1961). When the failure is governed by the traditional
linear envelope (δ → ϕ), Eq. (2) reduces to the classical form of dis-
sipation in the soil governed by the M-C yield condition and the nor-
mality flow rule

d = c v cos ϕ (3)

The uniaxial tensile strength and the rate of work dissipation in Eqs.
(1) through (3) are applicable in drained analyses (effective stress
analysis). The dissipation rate in an undrained analysis (total stress
analysis, Fig. 3(c)) becomes

d = su v (4)

where su is the undrained shear strength. In the process of rupture (or

incipient crack opening) governed by tension cut-off, the rate of work
dissipation per unit area of the failure surface (total stress analysis)
becomes

d = su v[1 − sin δ (1 − 2ξ )] (5)

where ξ determines uniaxial tensile strength,ft = 2ξsu, much like in Eq.

(1). Such rupture can occur without water transfer, preserving the
undrained nature of material deformation.

3. Limitations of the approach

The method chosen to consider stability is the kinematic approach Fig. 4. Failure mechanism for drained analyses (effective stress analysis) generated by
of limit analysis. This method is based on a theorem, which requires rotation of a circle with variable diameter about axis through point O.
convexity of the yield function and associativity (normality) of plastic
flow. The material is considered perfectly plastic, thus no distinction is
4. Rotational collapse mechanism
made between peak and residual strength. The normality flow rule is
used not because the true material behavior is governed by associa-
4.1. Problem statement
tivity, but because with this assumption, the solution can be proved a
rigorous bound on the true solution. Plastic dilatancy rates for geo-
The slopes considered are defined by their height H, maximum
materials are lower than those predicted by the normality flow rule.
width B, and uniform inclination angle β. The strength of the geoma-
However, a corollary theorem of limit analysis indicates that the ki-
terial in the slope is described with the Mohr-Coulomb yield function
nematic solution for associative material is also a rigorous bound on the
with tensile strength cut-off, and the material in the slope is uniform.
solution for a material governed by the non-associative flow rule
Given stress-free slope boundaries, find the critical value of di-
(Radenkovic, 1962).
mensionless group γH/c (stability factor) describing the limit state of
In the linear range of the strength envelope (Fig. 3(b)), dilation
the slope (γ is the unit weight of the geomaterial and c is defined in
angle ψ predicted by normality is equal to the internal friction angle,
Fig. 3(a,b)).
ψ = ϕ, which is typically larger than the dilation angle observed in
experiments. In the tension cut-off region, the normality-predicted di-
lation angle is equal to δ, which can no longer be interpreted as the 4.2. Drained failure mechanism
dilation angle in the Reynolds (1885) sense, as it can assume unrealistic
values anywhere between ϕ and 90°. The deformation modes associated The 3D rotational failure mechanism is depicted in Fig. 4. The trace
with the tension cut-off on the yield condition may be better char- of the failure surface in the figure is a cross-section of a curvilinear
acterized as rupture modes (as intended by Paul, 1961), rather than (horn-like) cone with a vertical central plane. Each radial cross-section
continual deformation. As plasticity flow rules are intended for de- of the horn-like surface is a circle of a different size, increasing with an
scription of strain rates in ductile materials, their use with the yield increase in polar coordinate θ. The magnitude of the velocity at every
function defined by material rupture can no longer predict true de- point in the moving block is
formation. In any case, limit analysis cannot be used to predict de- v = ωρ (6)
formation. However, the method is still applicable for considering in-
where ω is the angular velocity (about the axis passing through point O)
stantaneous (or incipient) stability loss (but not the progressive failure
and ρ is the radial coordinate. The direction of the velocity is de-
of slopes). Some limitations of the specific mechanisms considered in
termined by the normal to ρ. The strength of the material is governed by
this paper are in the difficulties in generalizing them to cases of non-
the yield surface in Fig. 2(a) or the strength envelope in Fig. 3(b). This
homogeneous slopes and complex geometry, which might be an im-
mechanism is based on an earlier suggestion by Michalowski and
pediment to their application in design.
Drescher (2009), with the exception that, in part of the mechanism, the

76
D. Park, R.L. Michalowski Engineering Geology 229 (2017) 73–84

soil is now governed by the tension cut-off section of the strength en-
velope. This occurs when the traction vector on the failure surface in
Fig. 4 is located on segment SP of the strength envelope, Fig. 3(b). The
trace of the failure surface on a symmetry plane (vertical central plane)
is marked CDF in Fig. 4, with portion CD governed by the flow rule
associated with the tension cut-off and the deformation along DF gov-
erned by the linear portion of the strength envelope. The analysis has
shown that section CD always approaches the surface at a steep angle.
Location of point D is not predetermined, and it will be a part of the
solution.
The reader will notice that the essential difference between this
mechanism and that suggested earlier (Michalowski and Drescher,
2009) is in the vicinity of the apex of the cone. Admissibility of the
velocity field in rigid rotation, consistent with the Mohr-Coulomb yield
condition and the normality flow rule, requires that the apex angle of
the cone is equal to 2ϕ, and the lower (PDF) and upper (PD'F′) contours
are log spirals r and r′
r (θ) = r0 e (θ − θ0) tan ϕ (7)

r ′ (θ) = r0′ e−(θ − θ0) tan ϕ (8)

where θ is the angular coordinate, and r0 is the value of r at θ = θ0
(distance OA in Fig. 4). However, with the upper portion of the me-
chanism governed by the tension cut-off, apex P moves to P′, and P′D
and P′D′ are curves that render admissible kinematics when failure is
governed by the tension cut-off. The rupture angle δ in the failure
surface portion governed by tension cut-off can increase from ϕ at point
Fig. 5. Failure mechanism in undrained material (total stress analysis), with incipient
D (or D′) to its maximum value of δm at point C (or C′). This rupture
crack opening along surface with trace CD.
angle δ does not need to be constant along P′CD (or P′C′D′), though it is
constant for given azimuth θ. To ensure a smooth surface, condition
δ = ϕ at θ = θtc will be enforced, where θtc is the angular coordinate Making use of the kinematic theorem of limit analysis, a value of
defining points D and D′. With the rupture angle described with yet dimensionless group γH/c, known as the stability factor, can be calcu-
unknown function δ(θ), the kinematic admissibility requires that trace lated for geometry of the failure mechanism given by angles θ0, θh, θm,
CD be described by function (Michalowski, 2017) θtc, and ratio r0′ r0 .
θ
r (θ) = rm e ∫θm
tan δ (θ) dθ
(9) 4.3. Undrained failure mechanism
where
The strength envelope for an undrained process of failure is shown
sin θ0 in Fig. 3(c), with the rate of work dissipation determined in Eqs. (4) and
rm = r0
sin θm (10) (5) for the linear and nonlinear portions of the strength envelope, re-
and θm defines point C, as shown in Fig. 4. In order to fully describe spectively. The failure surface now assumes the shape of a torus, con-
section P′CD, function δ(θ) needs to be defined. Previous calculations sistent with incompressible deformation, Fig. 5, terminated by a non-
for 2D analysis (Michalowski, 2017) indicated that a linear function led linear cone (DCP′C′D′), where the geomaterial comprising the slope is
to good upper bound estimates for the stability factors, i.e., nonlinear governed by tension cut-off. In that region, the material is subjected to
distributions did not significantly improve the outcome of calculations. incipient rupture, associated with cavitation in the water (no drainage).
Therefore, we adopt the linear function for the 3D analysis
4.4. Alternative mechanism of failure
δm − ϕ
δ (θ) = δm − (θ − θm )
θtc − θm (11)
An alternative mechanism is shown in Fig. 6. Ratio r0′/r0 is negative
Of four angles: θ0, θm, θtc, and δm (Fig. 4), only three are in- in this mechanism. It is similar to that suggested in Michalowski and
dependent. This is because point D is common to both log-spiral PADF Drescher (2009), but part of the failure surface is now governed by the
and curve CD. With δ(θ) defined in Eq. (11), angle δm can be derived tensile strength cut-off. The failure surface is still generated by a ro-
from the following condition tating circle of diameter increasing with an increase in polar coordinate
θtc θ, but it now revolves about a chord. Contours ADB and A′D′B′ are log-
∫ tan δ (θ) dθ spirals, but sections CD and C′D’ are associated with the tension cut-off
r0 e (θtc − θ0) tan ϕ = rm e θm (12) regime. The shape of sections ADB and CD is determined by Eqs. (7) and
′ ′ ′ (9), respectively, but A′D′B′ and C′D′ are defined respectively by
Similarly to Eq. (9), contour P C D can be described as
θ r ′ (θ) = −r0′ e (θ − θ0) tan ϕ (15)
r ′ (θ) = rm′ e ∫θm
− tan δ (θ) dθ
(13)
and
where rm′ = OC′ is found from the condition that point D′ belongs to
both curve PA′D′ and P′C′D′ θ
r ′ (θ) = −rm′ e ∫θm
tan δ (θ) dθ
(16)
e−(θtc − θ0) tan ϕ
rm′ = r0′ θtc This mechanism is mentioned here, because it does provide the
e− ∫θm tan δ (θ) dθ (14)
minimum of the stability factor for some combinations of slope para-
It can be further shown that rm′ = r0′r0 rm = r0′ sin θm sin θ0 . meters. In addition, a special case of this mechanism can be used in

77
D. Park, R.L. Michalowski
Fig. 6. Alternative mechanism generated by revolving a circle of variable diameter about
a chord passing through point O.
order to arrive at the benchmark solution attributed to Baligh and
Azzouz (1975).
4.5. Plane insert
It should be clarified that all the mechanisms discussed are plane in
the kinematic sense: all velocity vectors are parallel to one plane.
However, the geometry of the mechanisms is three-dimensional, and
the in-plane curvature of the optimized failure surface is not in-
dependent of the out-of-plane curvature. This prevents these mechan-
isms from approaching plane geometry (cylindrical failure surface) for
large widths. To provide this option, a plane insert of width b is in-
cluded in the mechanism, Fig. 7(a), with the width of the entire me-
chanism limited to B. The appropriate terms for the work dissipation
rate and the rate of work of the soil weight in the plane portion of the
mechanism were developed in Michalowski (2017). The kinematics of
the insert must match the kinematics in the 3D zones at the “end caps”,
Fig. 7(a), and this is achieved by requiring anglesθ0, θh, θm, and θtc to
be the same in both 2D and 3D portions of the mechanism.
Fig. 7. (a) Toe failure mechanism with plane insert, and (b) face failure.
78
Engineering Geology 229 (2017) 73–84
4.6. Slope face failure
When the slope is limited to a narrow space (small B/H), the plane
insert is eliminated from the mechanism in the process of solving for the
stability factor. In addition, a mechanism that does not reach the toe of
the slope may be the most critical. Such a mechanism is depicted in
Fig. 7(b). This option was included in calculations and it is reported in
the results whenever the face mechanism becomes critical.
5. Analysis
The kinematic approach of limit analysis is based on the balance of
the rate of work dissipation D and the work rate of external (gravity)
work Wγ
D = Wγ (17)
Stability factor γH/c is calculated from the balance in Eq. (17) (γ is
the soil unit weight, H is the slope height, and c is cohesion or the
intercept of the linear portion of the strength envelope). Because the
kinematic theorem of limit analysis indicates that D ≥ Wγ in any ki-
nematically admissible mechanism, Eq. (17) provides an upper bound
to stability factor γH/c. The stability factor is a combination of slope
and soil parameters when the slope is in the state of limit equilibrium
(on the verge of failure). The rate of work of the soil weight Wγ in an
incipient failure mechanism is calculated from the general integral
Wγ = ∫ γ v cos θ dV
V (18)
where V is the volume of the entire rotating soil mass and vcosθ is the
magnitude of the vertical component of the velocity vector. Details of
calculations are given in the Appendix, Eqs. (28)–(30). The entire
moving block rotates as one rigid body, and the work dissipation takes
place only at the interface between the rotating block and the soil at
rest beneath it. The rate of the work dissipation is represented by the
following integral
sin δ − sin ϕ ⎞
D= ∫ c v ⎛cos ϕ 11 −− sin
⎜
sin δ
ϕ
+ 2ξ
cos ϕ
⎟ dS
S ⎝ ⎠ (19)
where S is the area of the interface of the moving block and the soil at
rest (failure surface), and the remaining symbols are consistent with
those in Eq. (2). On the part of the failure surface where the deforma-
tion is governed by the flow rule associated with the linear portion of
the strength envelope, Eq. (19) reduces to
D= ∫ c v cos ϕ dS
S (20)
Detailed expressions for the respective integrals are shown in the
Appendix, Eqs. (34) and (35). The respective expressions for undrained
analysis are analogous, with the exception that the work dissipation
rates per unit area in Eqs. (19) and (20) need to be replaced with Eqs.
D. Park, R.L. Michalowski
(5) and (4), respectively. The work rate balance in Eq. (17) allows for
easily calculating an upper bound to stability factor γH/c.
Calculations were carried out for a given constraint of the maximum
width of the mechanism B/H and given internal friction angle ϕ. The
specific geometry of the mechanism, defined by angles θ0, θh, θm, θtc,
and ratio r0′ r0 , Fig. 4, was determined from the optimization scheme
where the least value of the stability factor was sought. In that proce-
dure, angles θ0, θh, θm, and θtc were varied in a single loop of optimi-
zation with a minimum step of 0.01°, and ratio r0′ r0 was varied with a
minimum increment of 0.001, until the minimum of the stability factor
was found. The minimum was determined when the difference between
the stability factors in two consecutive loops was less than 10− 6. The
extent of the failure surface governed by tension cut-off was not as-
sumed a priori, but was part of the solution (θtc). In all cases, the failure
surface was assumed smooth, i.e., δ = ϕ at θ = θtc. Variation of the
rupture angle along the surface governed by tension cut-off was taken
as linear in θ, Eq. (11), and the maximum value of the rupture angle, δm
at θ = θm, was calculated from the kinematic admissibility condition in
Eq. (12). It is emphasized that the material is uniform in the entire
slope, but it is governed by different sections of the yield condition, in
different parts of the slope. This is not an assumption, but the outcome
of the solution.
6. Comparison to benchmark solution
The solution for an undrained failure using a spherical cap me-
chanism suggested by Baligh and Azzouz (1975) is considered by some
a benchmark solution in 3D analysis (Hungr et al., 1989; Lam and
Fredlund, 1993; Chen et al., 2005; Griffiths and Marquez, 2007). This is
essentially a local undrained failure in an infinite slope of inclination
1:2, Fig. 8. The length scale of the problem is introduced by the size of
Fig. 8. Spherical cap failure.
79
Engineering Geology 229 (2017) 73–84
the mechanism (radius R of the spherical cap surface), unlike in most
analyses of slopes where the scale is defined by the slope height. The
length scale is included in the dimensionless group su/γR. The solution
attributed to Baligh and Azzouz, 1975 is the safety factor F = 1.402,
and it was obtained for β = 26.57°, α = 30°, and su/γR = 0.1, Fig. 8.
The reader will notice that by setting θ0 = α − β, θh = π − θ0 − 2β,
θm = θ0, and taking r0′ = − r0 in the alternative mechanism (Fig. 6),
the failure surface for undrained failure turns into a spherical surface
(ADB in Fig. 8), as that considered by Baligh and Azzouz (1975). The
stability factor, defined now as γR/su, can be calculated from the work
rate balance in Eq. (17)
γR
=N
su (21)
Factor of safety F is introduced by replacing su → su/F; conse-
quently, one can write
su
F=N
γR (22)
The dimensionless group su/γR in Eq. (22) is a combination of
parameters for an existing (safe) slope and it is equal to the reciprocal of
the stability factor in Eq. (21) only if F = 1. Solutions known from the
literature for this problem are given in Table 1 in the row for
β = 26.56°. When the geometry of the mechanism is forced to be
identical to that in Baligh and Azzouz (1975), the result calculated
using the special case of the alternative mechanism (Fig. 6) is identical
to that of Baligh and Azzouz (F = 1.402). However, Baligh and Azzouz
used an arbitrary geometry of the failure mechanism with α = 30°,
whereas the minimum of F = 1.265 is found when α = 15°. Calcula-
tions were performed based on Eqs. (17) and (22), with the work rate of
the soil weight and work dissipation rate in Eqs. (28) and (34), re-
spectively, after substituting δ → ϕ = 0 and c → su. The value of safety
factor F = 1.265 found from the kinematic approach of limit analysis is
considerably better (lower) than available results based on a “column
equilibrium” approach: 1.422 offered by Hungr et al. (1989), 1.386 by
Lam and Fredlund (1993), and 1.388 by Huang and Tsai (2000). It is
also better than the value of 1.430 from kinematic limit analysis with
finite element framework in Chen et al. (2005), and 1.39 reported by
Griffiths and Marquez (2007), who used the finite element approach.
This outcome is not surprising, as all these solutions are based on a
“fixed” geometry of the mechanism, whereas the solution given in this
paper is a result of a minimizing process with varied geometry
(Table 1). The value of 1.265 from the numerical approach is identical
to the closed-form solution given in Michalowski and Drescher (2009).
Tensile strength cut-off is introduced into the mechanism by al-
lowing the upper portion of the failure surface (CD in Fig. 8) to be
governed by rupture associated with the curvilinear portion of the
strength envelope in Fig. 3(c) and ξ = 0. Point D on the failure surface
was found from the minimization process where the minimum of γR/su
was sought. It is interesting to notice that the solution is hardly affected
by tensile strength cut-off, as shown in Table 1.
7. Numerical results
Calculated stability factors are illustrated in Fig. 9. Calculations
were carried out for width ratio B/H starting at 0.5, for the range of
slope inclinations 30° to 90°, and for full tension cut-off (ξ = 0). The
results in Fig. 9(a) are for undrained failures, but they are very con-
servative estimates as the tensile strength against rupture in saturated
soils during the undrained process is likely to be substantial. For rela-
tively gentle slopes and B/H ≥ 1, the below-toe failure mechanism
yields the minimum solution (toe failure solutions are indicated by gray
lines in that range). The remaining graphs in Fig. 9 show the calculated
stability factors for geomaterials with internal friction angle ϕ in the
range 10°–30°. Numerical values of stability factors for an internal
friction angle of 30° are given in Table 2.
D. Park, R.L. Michalowski Engineering Geology 229 (2017) 73–84

Table 1
Comparison of solutions to undrained failure problem by Baligh and Azzouz, 1975

Slope angle β Factor of safety

Baligh and Azzouz Hungr et al. Lam and Fredlund Huang and Tsai Griffiths and Marquez This study This study
(1975) (1989) (1993) (2000) Chen et al. (2007) M-Ca T-Cb
(2005)

20° – – – – – – 1.654 1.643

25° – – – – – – 1.339 1.326
26.56° 1.402 1.422c 1.386c 1.388c 1.430c 1.39c 1.265 (1.402c) 1.253
30° – – – – – – 1.131 1.118
35° – – – – – – 0.986 0.971

a
Mohr-Coulomb.
b
Tensile strength cut-off.
c
Geometry of mechanism matching that in Baligh and Azzouz (1975).

Not surprisingly, eliminating tensile strength from the yield condi-
tion leads to a reduction in the stability factor. The slopes most affected
by the tension cut-off are steep slopes with a large ratio B/H, and the
plane failure solution for vertical slopes is affected the most (about 24%
difference between the M-C and T-C solutions). The impact of tension
cut-off is only slightly reduced with an increase in the internal friction
angle. For instance, the stability factor for a 60-degree slope with me-
chanism width B/H = 5 is reduced by 4.7%, 3.8%, and 2.9% for ϕ
equal to 10°, 20°, and 30°, respectively. For a vertical slope, these
numbers are 21.0%, 20.4%, and 20.1%, respectively.
An explanation why steep and very wide slopes are affected by
tension cut-off the most can be found in the geometry of the critical
failure mechanisms. We consider the area of the failure surface gov-
erned by the tension cut-off, ATC, compared to the area of the entire
failure surface, A. The ratio of the two is plotted in Fig. 10(a) as a
function of the slope inclination, for a mechanism limited to width B/

Fig. 9. Stability factor as a function of slope inclination: (a) undrained analysis, (b) ϕ = 10°, (c) ϕ = 20°, and (d) ϕ = 30°.

80
D. Park, R.L. Michalowski Engineering Geology 229 (2017) 73–84

Table 2
Stability factor γH/c for ϕ = 30°.

B/H Yield condition β

45° 60° 75° 90°

0.5 M-C 96.34a 40.60a 25.48a 18.95a

T-C 96.14a 40.22a 25.04a 18.34a
0.6 M-C 80.35a 33.88a 21.23a 15.64a
T-C 80.23a 33.66a 20.86a 15.10a
0.8 M-C 62.12 26.47 16.60 12.09
T-C 61.84 26.13 16.14 11.11
1 M-C 54.07 23.40 14.65 10.41
T-C 53.75 23.05 14.13 9.46
1.5 M-C 45.84 20.22 12.65 8.85
T-C 45.46 19.83 12.08 7.62
2 M-C 42.61 18.94 11.84 8.21
T-C 42.19 18.52 11.22 6.91
3 M-C 39.87 17.83 11.12 7.64
T-C 39.41 17.38 10.45 6.24
5 M-C 37.97 17.05 10.61 7.23
T-C 37.48 16.56 9.88 5.78
10 M-C 36.70 16.52 10.26 6.96
T-C 36.19 16.01 9.49 5.48
∞ M-C 35.54 16.04 9.94 6.69
T-C 35.01 15.50 9.12 5.19

a
Face failure mechanism.

H = 1.0 and for a 2D analysis. Clearly, the contribution of the portion

of the failure surface that is governed by tension cut-off increases with
an increase in the slope inclination, and, for steeper slopes, this con-
tribution is much larger for plane failures. It is also interesting to notice
that the contribution of the plane insert width, b, to the entire width of
the mechanism is not a monotonic function of the slope inclination,
Fig. 10(b), but this observation may be of a lesser practical significance.

8. Seepage influence

The presence of seepage is a likely cause of many slope failures.

Finding the distribution of pore water pressure u associated with see-
page requires information about specific hydraulic conditions. In the
absence of such information, and in generic calculations focused on the
trends in seepage influence, a predetermined distribution of pore water
pressure is often used. Such a distribution was suggested by Bishop and
Morgenstern (1960) in the form: u = ruγh, where h is a vertical distance
from a point where the pressure is to be determined to the ground
surface, γ is an average bulk unit weight along h, and ru is a coefficient
loosely related to the presumed phreatic surface. For slopes that are not
submerged, ru can vary approximately in the range 0 ≤ ru ≤ 0.5. During
deformation, the pore pressure does work on the volumetric strain of
the soil, Wu, and the seepage can be considered in the analysis through
amending the work rate balance in Eq. (17) with work Wu
D = Wγ + Wu
It was shown earlier (Michalowski, 1995) that such a procedure
includes the work of the seepage forces as well as the buoyancy forces.
Because deformation in the mechanisms considered occurs only along
the failure surfaces, the respective work rate can be written as a sum of
Wu1 and Wu2
Fig. 10. (a) Area fraction of failure surface governed by tension cut-off for ϕ in 5° in-
tervals, and (b) width of plane insert as fraction of the entire width of the mechanism.
forces for steep slopes, and the results are shown for inclinations up to
70°.
With moderate seepage (ru = 0.25), the impact of tension cut-off is
relatively small for gentle slopes, but it increases significantly with an
increase in the slope inclination, Fig. 11. This impact also increases
with an increase in the mechanism width, and is the largest for the
plane mechanism (B/H = ∞). For a 40-degree slope (ϕ = 30°) and the
(23) plane mechanism, the drop in the stability factor caused by tension cut-
off is only about 3% when compared to the M-C solution, but it in-
creases to about 56% for a 70-degree slope. Not surprisingly, this dif-
ference increases when the seepage increases, with the reduction in
stability factor being 17% for a 40-degree slope and 69% for a 70-de-
gree slope when ru = 0.5.

Wu1 = ∫ u v sin δ dS, θm ≤ θ < θtc 9. Conclusions

S
Wu2 = ∫ u v sin ϕ dS, θtc ≤ θ ≤ θ h
S (24)
The detailed expressions used in calculations are presented in the
Appendix, Eqs. (36)–(38). A set of results with pore water pressure
consistent with ru = 0.25 and ru = 0.5 is illustrated in Fig. 11 and
Table 3. Coefficient ru leads to significant overestimation of seepage
Traditional slope stability analyses for soils and soft rocks use the
Mohr-Coulomb function to describe the yield state of the material. A 3D
analysis of slope stability for bonded geomaterials (hard soil/soft rock)
was presented where the tensile strength was removed (or reduced)
from admissible stress states by introducing tension cut-off in the yield
surface. The motivation behind removing (or reducing) tensile strength
stems from uncertainties regarding tensile strength. Bonded soils are

81
D. Park, R.L. Michalowski Engineering Geology 229 (2017) 73–84

the result of extrapolation of the test results into the tensile regime. The
following question was posed in the paper: does removing tensile
strength from the yield condition have any effect on the outcome of
stability analyses?
Introducing tension cut-off in the Mohr-Coulomb yield condition
leads to non-linearity of the strength envelope; the normality flow rule
can then predict deformation with a volumetric component far ex-
ceeding dilatancy observed in plastic deformation of geomaterials.
These modes can no longer be interpreted as continual deformation, but
a rupture characterized with a rupture angle, rather than a dilation
angle. Such modes are typically observed at the inception of toppling
failures. The kinematic approach of limit analysis was utilized, and the
construction of collapse mechanisms was found to be intricate, because
of varied rupture angles associated with nonlinearity in the strength
envelope owed to tension cut-off. This nonlinearity allows construction
of collapse mechanisms with large rupture angles, which enriches the
range of admissible failure mechanisms. These new mechanisms in-
dicate stability factors lower than those calculated based on the clas-
sical Mohr-Coulomb yield criterion.
Calculations indicated that the soil likely to be governed by the
tension cut-off occupies the upper portion of the slope, with the depth
of this region increasing with an increase in the slope inclination.
Consequently, the steep slopes are most vulnerable to the tensile
Fig. 11. Stability factors for slopes subjected to seepage, ϕ = 30°. strength cut-off. For instance, the drop in the stability factor of a 30-
degree slope with tension cut-off compared to the classical M-C solution
Table 3 is almost negligible, but this difference can increase to 24% for wide
Stability factor γH/c for ϕ = 30° in the presence of seepage. vertical slopes. Calculations also revealed that the impact of tension
cut-off increases with an increase in the width of the failure mechan-
ru B/H Yield condition β
isms, with the outcome of the 2D analysis affected the most. Seepage
30° 40° 50° 60° 70°
can have a detrimental effect on the stability of slopes, and this effect is
exacerbated by the presence of tensile strength cut-off. In the presence
0.25 1 M-C – 33.22 19.03 13.01 9.65 of moderate seepage (ru = 0.25), the drop in stability factor caused by
T-C – 32.57 18.22 11.90 8.09 the tension cut-off for a 70-degree slope and ϕ = 30° is 56%, and it
1.5 M-C – 27.37 16.26 11.41 8.61
T-C – 26.79 15.30 9.58 5.54
increases to 69% when ru = 0.5.
∞ M-C – 20.66 12.85 9.26 7.09 It is demonstrated in the paper that tensile strength cut-off has a
T-C – 20.11 11.72 6.34 3.12 significant impact on the outcome of stability analyses of slopes. Most
0.5 1 M-C 34.56 18.38 11.97 8.58 6.57 vulnerable to tension cut-off are steep slopes subjected to seepage.
T-C 33.97 17.19 10.69 7.26 5.07
1.5 M-C 26.53 14.69 10.12 7.61 5.95
T-C 25.39 13.10 7.58 4.45 2.47
Acknowledgements
∞ M-C 18.01 11.05 8.05 6.27 5.02
T-C 16.68 9.17 5.66 3.17 1.58
The work presented in this paper was carried out when the authors
were supported by the National Science Foundation, Grant no. CMMI-
typically tested in the compressive regime, and the tensile strength is 1537222. This support is greatly appreciated.

Appendix

Radius R of the shaded cross-section of the curvilinear cone in Fig. 4, and coordinate of the center of the circular cross-section, rc, are expressed as
r (θ) − r ′ (θ)
R (θ) =
2 (25)

r (θ) + r ′ (θ)
rc (θ) =
2 (26)
with r and r′ in Eqs. (7) and (8). Considering the geometrical relations in Fig. 4, the infinitesimal volumetric element can be written as

dV = ρ R2 − (ρ − rc )2 dρ dθ (27)
and the rate of work of the soil weight in Eq. (18) can now be written more specifically as
θh r 2 (θ )
Wγ = 2ωγ ∫θ ∫r (θ)
m 1
ρ2 cos θ R2 − (ρ − rc )2 dρ dθ
(28)
where both R and rc are functions of θ, and the integration limits are

82
D. Park, R.L. Michalowski Engineering Geology 229 (2017) 73–84

r sin θ
⎧ 0sin θ 0 , θm < θ ≤ θ E
⎪ r sin(β + θ
h h)
r1 (θ) = sin(β + θ )
, θE < θ ≤ θB (θB = θ h toe failure)
⎨
⎪ r h sin θ h , θB < θ ≤ θ h (Below‐toe only)
⎩ sin θ (29)
and
θ
⎧ rm e ∫θm tan δ (θ) dθ , θm < θ ≤ θtc
r2 (θ) =
⎨ (θ − θ0) tan ϕ
⎩ r0 e , θtc < θ ≤ θ h (30)
where
h r
⎛ sin θ h − 1 ⎞
θE = arctan ⎜ r h H
cos θ h + cot β ⎟ (31)
⎝H ⎠
and
rh = r0 e (θh − θ0) tan ϕ (32)
Infinitesimal area element dS in Eqs. (19) and (20) is defined as
dρ ρdθ dρ ρ R
dS = dl = = dρ dθ
sin α cos δ sin α cos δ R2 − (ρ − rc )2 (33)
with dl, dρ, and α illustrated in Fig. 4, and δ = ϕ in range θh ≥ θ ≥ θtc. Consequently, the rate of work dissipation in Eqs. (19) and (20) can be
written as
θh r 2 (θ )
D = 2cω ∫θ ∫r (θ)
m 1
f (θ, ρ) dρ dθ
(34)
where

⎛
f ⎜θ , ρ⎟ =
⎪ cos δ (
⎞ ⎧ ρ2 1 cos ϕ 1 − sin δ + 2ξ sin δ − sin ϕ
1 − sin ϕ cos ϕ ) R
R2 − (ρ − rc )2
, θm ≤ θ < θtc

⎜⎜ ⎟⎟ ⎨ ρ2 R
, θtc ≤ θ ≤ θ h
⎪ R2 − (ρ − rc )2
⎝ ⎠ ⎩ (35)
where R, rc and δ are all functions of θ, Eqs. (11), (25) and (26).
The presence of seepage is considered in the analysis through inclusion of the rate of work of pore water pressure on volumetric deformation, Wu
θ h r 2 (θ)
Wu = 2γ ω ru ∫∫ f (θ, ρ) dρ dθ
θm r 1 (θ) (36)
where limits of integration r1 and r2 are as in Eqs. (29) and (30), and
R
⎧ ρ2 h (θ, ρ) tan δ , θm < θ ≤ θtc
⎪ R2 − (ρ − rc )2
f (θ, ρ) = R
⎨ ρ2 h (θ, ρ) tan ϕ , θtc < θ ≤ θ h
⎪ R2 − (ρ − rc )2 (37)
⎩

⎧ ρ−
⎪
( r 0 sin θ0
sin θ ) sin θ, θm < θ ≤ θES
⎪
h (θρ) = ρ −
⎨ ( r h sin(β + θ h )
sin(β + θ ) ) (sin θ + cos θ tan β), θES < θ ≤ θBS
⎪
⎪ ρ−
⎩
( r h sin θ h
sin θ ) (sin θ + cos θ tan β), θBS < θ ≤ θ h
(38)
where θES and θBS are polar coordinates of projections of points E and B on the failure surface CDF, Fig. 4, and for toe failure θBS = θh. Angles θES and
θBS can be found from geometrical relations in Fig. 4
r2 (θES) cos θES = r1 (θE) cos θE
r2 (θBS) cos θBS = r1 (θB) cos θB , Below‐toe only (39)

References Chen, W.F., Liu, X.L., 1990. Limit Analysis in Soil Mechanics. Elsevier, Amsterdam.
Chen, J., Yin, J.-H., Lee, C.F., 2005. A three-dimensional upper-bound approach to slope
stability analysis based on RFEM. Géotechnique 55, 549–556.
Baligh, M.M., Azzouz, A.S., 1975. End effects on stability of cohesive slopes. ASCE J. Geot. Collin, A., 1846. Recherches Expérimentales sur les Glissements Spontantés des Terrains
Eng. Div. 101, 1105–1117. argileux. Carilian-Goeury, Paris (Translation by W.R. Schriever: Landslides in Clays.
Bishop, A.W., Morgenstern, N.R., 1960. Stability coefficients for earth slopes. University of Toronto Press, 1956).
Géotechnique 10, 129–150. Drescher, A., 1983. Limit plasticity approach to piping in bins. J. Appl. Mech. 50,
de Buhan, P., Garnier, D., 1998. Three dimensional bearing capacity analysis of a foun- 549–553.
dation near a slope. Soils Found. 38, 153–163. Drucker, D.C., Prager, W., 1952. Soil mechanics and plastic analysis or limit design. Q.
Chen, W.F., Drucker, D.C., 1969. Bearing capacity of concrete blocks or rock. J. Eng. Appl. Math. 10, 157–165.
Mech. Div. 95, 955–978. Duncan, J.M., Wright, S.G., 2005. Soil strength and slope stability. Wiley, Hoboken, New

83
D. Park, R.L. Michalowski
Jersey.
Fellenius, W., 1927. Erdstatische Berechnungen mit Reibungund und Kohäsion
(Adhäsion) und unter Annahme kreiszylindrischer Gleitflächen. Ernst & Sohn, Berlin
(Translation of the Swedish edition of the book (Stockholm 1926)).
Gens, A., Hutchinson, J.N., Cavounidis, S., 1988. Three-dimensional analysis of slides in
cohesive soils. Géotechnique 38, 1–23.
Griffiths, D.V., Marquez, R.M., 2007. Three-dimensional slope stability analysis by elasto-
plastic finite elements. Géotechnique 57, 537–546.
Hoek, E., Brown, E.T., 1980. Empirical strength criterion for rock masses. J. Geotech. Eng.
Div. 106, 1013–1035.
Hoek, E., Carranza-Torres, C., Corkum, B., 2002. Hoek-Brown failure criterion. In:
Hammah, R., Bawden, W., Curran, J., Telesnicki, M. (Eds.), Proceedings of the Fifth
North American Rock Mechanics Symposium (NARMS-TAC), 2002 Edition. Vol. 1.
University of Toronto Press, Toronto, pp. 267–273.
Hovland, H.J., 1977. Three dimensional slope stability analysis method. J. Geotech. Eng.
Div. 103, 971–986.
Huang, C.C., Tsai, C.C., 2000. New method for 3D and asymmetrical slope stability
analysis. J. Geotech. Geoenviron. Eng. 126, 917–927.
Hungr, O., 1987. An extension of Bishop's simplified method of slope stability analysis to
three dimensions. Géotechnique 37, 113–117.
Hungr, O., Salgado, F.M., Byrne, P.M., 1989. Evaluation of a three-dimensional method of
slope stability analysis. Can. Geotech. J. 26, 679–686.
Lagioia, R., Panteghini, A., Puzrin, A.M., 2014. The ‘I3’ generalization of the
Galileo–Rankine tension criterion. Proc. R. Soc. A 470 (2172), 20140568. http://dx.
doi.org/10.1098/rspa.2014.0568.
Lam, L., Fredlund, D.G., 1993. A general limit equilibrium model for three-dimensional
slope stability analysis. Can. Geotech. J. 30, 905–919.
Leshchinsky, D., Baker, R., Silver, M.L., 1985. Three dimensional analysis of slope sta-
bility. Int. J. Numer. Anal. Methods Geomech. 9, 199–223.
Li, A.J., Merifield, R.S., Lyamin, A.V., 2010. Three-dimensional stability charts for slopes
based on limit analysis methods. Can. Geotech. J. 47, 1316–1334.
Liu, G., Zhuang, X.Y., Cui, Z., 2017. Three-dimensional slope stability analysis using in-
dependent cover based numerical manifold and vector method. Eng. Geol. 225,
84
Engineering Geology 229 (2017) 73–84
83–95.
Michalowski, R.L., 1985. Limit analysis of quasi-static pyramidal indentation of rock. Int.
J. Rock Mech. Min. Sci. 22, 31–38.
Michalowski, R.L., 1989. Three-dimensional analysis of locally loaded slopes.
Géotechnique 39, 27–38.
Michalowski, R.L., 1995. Slope stability analysis: a kinematical approach. Géotechnique
45, 283–293.
Michalowski, R.L., 2010. Limit analysis and stability charts for 3D slope failures. J.
Geotech. Geoenviron. Eng. 136, 583–593. http://dx.doi.org/10.1061/(ASCE)GT.
1943-5606.0000251.
Michalowski, R.L., 2013. Stability assessment of slopes with cracks using limit analysis.
Can. Geotech. J. 50, 1011–1021.
Michalowski, R.L., 2017. Stability of intact slopes with tensile strength cut-off.
Géotechnique 720–727. http://dx.doi.org/10.1680/jgeot.16.P.037.
Michalowski, R.L., Drescher, A., 2009. Three-dimensional stability of slopes and ex-
cavations. Géotechnique 59, 839–850.
Paul, B., 1961. A modification of the Coulomb-Mohr theory of fracture. J. Appl. Mech. 28,
259–268.
Radenkovic, D., 1962. Théorie des charges limites extension a la mécanique des sols.
Séminaire de Plasticité. 1961. École Polytechnique, Publications Scientifiques et
Techniques du Ministère de L'Air, pp. 129–141.
Reynolds, O., 1885. On the dilatancy of media composed of rigid particles in contact.
With experimental illustrations. Philos. Mag. 20, 469–482.
Spencer, E., 1968. Effect of tension on stability of embankments. J. Soil Mech. Found. Div.
94, 1159–1173.
Taylor, D.W., 1937. Stability of earth slopes. Reprinted in: Contributions to Soil
Mechanics 1925 to 1940. Boston Soc. Civil Eng. 24, 337–386 (1937).
Utili, S., 2013. Investigation by limit analysis on the stability of slopes with cracks.
Géotechnique 63, 140–154.
Zhang, F., Leshchinsky, D., Baker, R., Gao, Y., Leshchinsky, B., 2016. Implications of
variationally derived 3D failure mechanism. Int. J. Numer. Anal. Methods Geomech.
40, 2514–2531.