Mechanical Properties of Frozen Soils: 5.1 Stress-Strain-Time and Strength Behavior

5
Mechanical Properties
of Frozen Soils
l
ria
ate
From the point of view of the science of materials, frozen soil is 5.1 Stress-Strain-Time and Strength Behavior
a natural particulate composite, which is composed of four dif-
ferent constituents: solid grains (mineral or organic), ice, Hydrostatic Pressure Effect on Frozen Soil Behavior
unfrozen water, and gases. The most important characteristic
dM
by which it differs from other similar materials—such as unfro- Isothermal Compression. The behavior of a frozen soil under
zen soils and the majority of artificial composites—is that an increase of hydrostatic pressure is considered to be the result
under natural conditions its matrix, which is composed mostly of combined mechanical and thermodynamic effects, the
of ice and water, changes continuously with varying tempera- former governing the stress sharing, and the latter the pressure
ture and applied stress. melting phenomena. If only mechanical effects are considered
There is only one type of ice present in soil pores (i.e., nor- and a two-phase material is assumed, the stress sharing under
hte
mal hexagonal ice of type 1h), but unfrozen water exists in two an increase of overall pressure is governed by this equation
states: strongly bound and weakly bound water. The former is (Bishop 1973):
the water film surrounding mineral particles and held to them
by high intermolecular forces that suppress freezing, even at -1
È Ê Cm ˆ ˘
very low temperatures. The rest of the pore water is weakly
Dum Í
Í ÁË C ˜¯ - 1 ˙
bound and can be frozen more easily (Anderson and Morgen- B= = 1+n
s ˙
Í ˙ (5.1-1)
rig
stern 1973). The amount of unfrozen water present in frozen Dp ÊCˆ

soil at a given temperature can be related to the specific surface
Í ÁË C ˜¯ - 1 ˙
ÍÎ s ˙˚
area of the mineral particles (Anderson and Tice 1972) and is
strongly affected by the electrolyte content (Yong and Warken- where Dp is the increase of total mean normal stress, Dum the
tin 1975). An investigation by Bourbonnais (1984) showed that resulting increase in the pore matrix stress, n soil porosity, Cm
py
in a frozen clay, unfrozen water films may exist down to a tem- compressibility of the pore matrix, Cs compressibility of the soil
perature of –110 °C. grains, and C the compressibility of the soil skeleton. When
Despite the presence of unfrozen water, when ice fills most proper values of these parameters are substituted into Eq.
of the pore space, the mechanical behavior of a frozen soil (5.1-1) (Ladanyi 1985a), it is found that, in a dense sand that is
Co
closely reflects that of the ice. The pore ice is usually of poly- either water or ice saturated, nearly all pressure is transferred to
crystalline type with a random crystal orientation. Under ordi- the pore matrix.
nary conditions, its response to deviatoric stresses is governed However, in a three-phase material such as a frozen soil con-
by the motion of dislocations and can be represented by a taining unfrozen water, the increase in pore water pressure due
power law creep equation of the Norton–Bailey type. to this mechanical effect may be quite different. One can con-
The yielding and failure of polycrystalline ice under a tri- sider the soil grains to be cemented by ice, which will result in a
axial state of stress differs from most other materials in that greatly reduced compressibility of the soil skeleton, C. For
under a high hydrostatic pressure, it first weakens and eventu- example, Wissa (1969) has found for a cement-stabilized sand a
ally melts. Conversely, when subjected to shear stresses at low value of C = 0.265 m2/GN, which leads to B = 0.575 under the
hydrostatic pressures and at ordinary freezing temperatures, it same conditions. This shows that under a sudden increase in
shows a ductile yielding at low strain rates, but it becomes hydrostatic pressure, the unfrozen water in a frozen soil may be
more and more brittle as the strain rate increases (Mellor stressed much less than that of the ice, which may have impor-
1979). tant thermodynamic consequences.
106
Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

MECHANICAL PROPERTIES OF FROZEN SOILS 107
Pressure Melting Phenomena. When the process is non-iso- structure. This consolidation time is, in fact, extremely long.
thermal, the pressure–temperature relationship for ice and For example, if one assumes that a typical value for the hydrau-
water coexisting in soil pores is considered to be properly lic conductivity of a frozen sand at –5 °C is about 10–11 cm/s,
defined by the Clausius–Clapeyron equation (Hillel 1980): and its coefficient of consolidation is 10–6 cm2/s, then, accord-
ing to Bishop and Henkel (1962), the time necessary for hydro-
dpw dpi LdT dynamic consolidation after a hydrostatic stress application for
- = (5.1-2)
rw ri T a specimen 10 cm long by 5 cm in diameter will be more than 7
months if drainage is permitted only from both ends. This time
where pi and pw denote the pressure of the ice and water, rw = may be reduced to 9 days if drainage over the entire surface of
1,000 kg/m3 is the density of water, ri = 916.8 kg/m3 is the den- the specimen is possible. In triaxial testing of frozen soils, the
sity of ice, L = 3.336 × 105 J/kg is the latent heat of water, and majority of tests reported in the literature should be classified
dT = To – T is the difference between the normal freezing tem- as unconsolidated–undrained tests, with the exception of those
l
perature of pure water, To = 273.15 K, and the actual tempera-
ria
that were consolidated before freezing (Goodman 1975).
ture, T, of the system. Substituting these values into Eq. (5.1-2)
gives
Shear Stress Effect on Frozen Soil Behavior
dpw = 1.091 dpi + 1.221 dT (5.1-3)
Sources of Strength. On the basis of findings made by many
ate
with p in MPa and T in K. More frequently, it is assumed that at investigators who studied systematically the shear behavior of
the ice-water interface, dpw = dpi = dp, which yields the freez- frozen sands (e.g., Goughnour and Andersland 1968; Ander-
ing-point depression for ice: sland and Al-Nouri 1970; Alkire and Andersland 1973; Cham-
berlain, Groves, and Perham 1972; Sayles 1973), and on the
dT
= -0.0743 K/MPa (5.1-4) basis of their own investigations, Ting (1981) and Ting, Martin,
dp
dM and Ladd (1983) concluded that the shear behavior of frozen
In an isothermal case, dT = 0, and Eq. (5.1-3) yields sands is controlled essentially by the following four physical
mechanisms: (1) pore ice strength; (2) soil strength, consisting
dpw = 1.091 dpi (5.1-5) of interparticle friction, particle interference, and dilatancy
effects; (3) increase in the effective stress due to the adhesive ice
which indicates that changes in pore-water pressure will tend to bonds resisting dilation during shear of a dense soil; and (4)
follow closely those in the pore ice, as long as there is no phase synergistic strengthening effects between the soil and ice matrix
hte
change. preventing the collapse of the soil skeleton. On the basis of

these observations, Ting, Martin, and Ladd (1983) proposed a
Consolidation. When a hydrostatic confining pressure is
schematic failure mechanism map (Fig. 5-1) which expresses
applied to a frozen granular mass, it follows from the foregoing the fact that the simultaneous presence of various mechanisms
that pressure melting will occur locally at grain-to-grain con- depends on the volume fraction of sand in the ice–sand
tacts and there will be some water migration toward lower- mixture.
rig
stress regions. Theoretically, Eq. (5.1-4) predicts that a pressure

of about 13.5 MPa is necessary to depress the freezing point of
ice by 1 °C. Consequently, under ordinary pressures, there
would be very little melting were it not for stress concentration
at ice-particle contacts, which may increase the pressures by a
py
factor of 10 to 500 (Arteau 1984). Finally, if the confining pres-

sures are very high so that the grain structure collapses, a total
pressure melting of the pore ice will occur even in dense sand,
as has been observed by Chamberlain, Groves, and Perham
Co
(1972); Simonsen, Jones, and Green (1974); and Jones and

Parameswaran (1983). The last three studies tend to support
the validity of Eq. (5.1-4) for predicting the total melting pres-
sure (e.g., dp = 100 MPa at –10 °C), but they also show that the
first sign of local melting becomes apparent at about a half to a
third of that pressure.
It has been demonstrated by several investigators (e.g., Burt
and Williams 1976; Lunardini et al. 1982; Arteau 1984) that the
hydraulic conductivity for water in frozen soils is small but not FIGURE 5-1 Failure mechanism map for unconfined compres-
zero, and that Darcy’s law is applicable to a frozen soil with sive strength of frozen Ottawa sand at –7 °C and a strain rate
pore ice if the ice pressure gradient is negligible. This conclu- of 4.4 × 10–4 s–1.
sion implies that frozen soil undergoes consolidation, which Source: Reproduced from Goughnour and Andersland 1968. Copyright 1968
eventually leads to transfer of applied stresses to the grain American Society of Civil Engineers. See also Ting, Martin, and Ladd 1983.

108 FROZEN GROUND ENGINEERING
It is clear that, in addition to soil density, the importance of 5.2 Factors Influencing Creep and Strength
any of these mechanisms in the observed strength of a frozen
soil will also depend on such factors as temperature, confining
Creep of Frozen Soil under Constant Stress
pressure, and deformation history. For example, the effect of
ice bonding will disappear after global ice melting occurs at a When a frozen soil specimen is subjected to a load, it will
high confining pressure. In addition, in fine-grained frozen respond with an instantaneous deformation and a time-depen-
soils and in saline soils at relatively high freezing temperatures, dent deformation; if the load is high enough, it will display a
in which the unfrozen water content is high, the ice bonding limiting strength. Several typical curves are shown in Fig. 5-2a.
effects may become negligible. Conversely, in ice-rich soils For step loading under uniaxial stress conditions and at con-
where the ice–soil ratio is high, (more than 1.38 in sands; stant temperature, the type of creep curve shown in Fig. 5-2b is
(Goughnour and Andersland 1968), most of the strength due common for frozen soils and a large number of other materials.
to intergranular friction and dilatancy effects vanish along with
l
Figure 5-2c shows the corresponding creep rate versus time
ria
the dilatancy-induced confinement increase (dilatancy harden- plot. The basic creep curve consists of three periods of time
ing). The latter effect can be suppressed by a sufficiently high during which the creep rate is (I) decreasing, (II) remaining
confining pressure, as has been shown by Chamberlain, Groves, essentially constant, and (III) increasing. These are often called
and Perham (1972). periods or stages of primary, secondary, and tertiary creep. For
As was noted above, at higher particle concentrations, the stresses lower than the long-term strength of frozen soil, the
ate
behavior of frozen soil is affected not only by pore ice and second period, with the minimum creep rate, and the third
unfrozen water, but also by friction, dilatancy, and the mineral period, with increasing creep rates, may not develop.
cohesion of the soil skeleton. Although the sources of strength The shape of creep curves for frozen soils is influenced not
are qualitatively well known, a proper evaluation of their only by temperature but also by the magnitude of applied
respective contributions is still considered to be a difficult task, stress, soil type, and its density. Medium- to high-density ice-
because of the impossibility of measuring directly, or inferring saturated sands and silts will exhibit the creep behavior illus-
dM
indirectly, the value and variation of intergranular stresses in a trated in Fig. 5-2b and 5-2c for a given constant temperature
frozen soil that are produced by external loading. and medium to high stress levels. For low stress levels, these
same soils will display only primary creep and will asymptoti-
Dilatancy Hardening and Softening Effects. When a two- cally approach some limiting deformation, as is shown for ice-
phase granular mass, consolidated under hydrostatic pressure, poor soils in Fig. 5-2a. In contrast, ice-rich silts and clays
is submitted to shear stresses, its initially stable structure will exhibit an abbreviated primary-creep period and a prolonged
hte
either collapse, if its density is low and/or if the confining pres- secondary-creep stage, whereas tertiary creep may never be
sure is high, or expand in the opposite case. If the pore-filling attained, as is shown by the intermediate curve in Figure 5-2a.
matrix has a low compressibility such as ice, and if overall vol- Some fine-grained soils can display primary-creep deforma-
ume changes during shear are prevented, the shear will produce tions up to strains exceeding 20%.
an increase in the matrix stresses in the first case and a decrease For ice-rich soils under moderate stress conditions, steady-
in the second case. As a result, shear under constant volume state (secondary) creep is dominant. In this case the primary
rig
will produce a decrease in intergranular stresses at low densi- stage can often be neglected and the entire creep curve consid-
ties, and an increase of these stresses at higher densities of the ered to be linear (Hult 1966; Ladanyi 1972a). Although in many
granular mass, as long as the matrix bond remains unbroken. practical frozen soil creep problems, steady-state creep is found
Because of these dilatancy-induced changes in intergranu- to be a good approximation of observed behavior (Thompson
lar stresses and the resulting softening and hardening effects and Sayles 1972), at high stress levels the specimen may appear
py
on the material behavior, these phenomena have been termed to go straight into accelerated creep without well-defined pri-
in soil and rock mechanics dilatancy softening and dilatancy mary and secondary stages, and may fail after a short period of
hardening effects. In frozen soil mechanics, although dila- time. The stress producing such a short-term failure of the spec-
tancy softening has been found to occur at high confining imen is referred to as the short-term strength of the material.
Co
pressures, even in frozen sands and silts (Chamberlain, If a series of uniaxial compression creep tests are carried out
Groves, and Perham 1972), dilatancy hardening effects have on identical samples, at the same temperature, but at various
more often been observed and discussed (Goughnour and applied loads, a set of creep curves (shown schematically in the
Andersland 1968; Andersland and Douglas 1970; Alkire and right lower quadrant of Fig. 5-3) is obtained. The information
Andersland 1973). contained in these creep curves can be represented in various
Available experimental evidence shows that in a frozen sand ways, some of which are shown in the figure (Ladanyi 1972a).
at ordinary pressures and temperatures, the dilatancy harden- For example, if the creep curves are intersected by constant
ing effect may exist only up to strains of about 1%, after which time lines and the intersection points are projected to the upper
the pore ice starts to yield in a brittle manner under combined right plot, a set of isochronous stress-strain curves is obtained,
tensile and shear stresses, and the test ceases to be “undrained.” each curve representing the state of strain in the soil after a
It is clear that the pore-ice cavitation can be prevented if much given time under load. With the exception of the instantaneous
higher confining pressures are applied (Sayles 1973; Ladanyi curve (t = 0), these curves are different from those obtained in
and Morel 1990). stress-strain tests conducted with a controlled rate of strain.

l
ria
ate
dM
FIGURE 5-3 Schematic plots of data from compression creep
tests conducted at a constant temperature and confining
pressure.
Source: Reproduced with permission from Ladanyi 1972a. Copyright 1972
hte
Canadian Geotechnical Journal.
ent plots of this kind will be obtained for any new combination
of these parameters. Because of this relatively complex rheolog-
ical behavior of frozen soils, special methods must be used for
rig
processing experimental data, in order to put the information

into an analytical form useful for engineering design.
FIGURE 5-2 Constant-stress creep test: (a) creep-curve varia-
tions; (b) basic creep-curve; (c) true strain rate versus time. Stress-Strain Behavior under Constant Strain Rate
py
Source: Reproduced with permission from Andersland, Sayles, and Ladanyi

1978. Copyright 1978 McGraw-Hill. The rheological curve in the upper left quadrant of Fig. 5-3 can
be considered as the relationship between applied stress, s, and
minimum creep rate, ė min. At the same time, the same curve can
Conversely, if steady-state (or minimum) strain rates are plot- also be considered as the relationship between peak stress, smax,
Co
ted against stress (as in the upper left quadrant of Fig. 5-3), the and applied strain rate, ė, in a strain-rate-controlled test. In
basic rheological curve of the soil for the particular test condi- other words, the same rheological curve as that derived from
tions is obtained. For frozen soil in the usual temperature stress-controlled creep tests can also be obtained from strain-
range, the curve frequently has a complex nonlinear shape. rate-controlled compression tests in an ordinary triaxial appara-
Finally, if the points of creep failure on the creep curves are tus. A set of such true stress-strain curves is shown in Fig. 5-4.
determined by some conventional method and the failure There is ample experimental evidence for ice, frozen soils,
stresses are plotted against the corresponding times to failure and high-temperature metals that a close correspondence exists
(as in the left lower quadrant of Fig. 5-3), a delayed strength between the peak stress observed at a given strain rate in a con-
curve is obtained. This shows how the strength of the frozen soil stant-strain-rate (CSR) test and the minimum strain rate in a
decreases with time from its short-term value, sst , and tends constant-stress (creep) (CSC) test. Or, as expressed by Mellor
asymptotically toward its long-term strength, slt . (1979), the ratio smax/ė in the former is approximately equal to
The creep behavior of frozen soils also depends on tempera- the ratio s/ė min in the latter. In polycrystalline ice and the two
ture and normal (or confining) pressure, and therefore differ- kinds of tests, the foregoing ratios occur at about the same

l
ria
ate
FIGURE 5-5 Schematic relationship between constant-strain-
rate compression tests, and constant-stress (creep) tests.
Source: Reproduced from Vaid and Campanella 1977. Copyright 1977 Ameri-
can Society of Civil Engineers
failure strain is affected by strain rate, temperature, and confin-

dM
ing pressure. Nevertheless, there are well-defined regions in
which this assumption holds, which makes it a convenient way
for determining creep rupture conditions.
In creep tests it is usually found that ef increases with
increasing confining pressure, and also slightly with increasing
strain rate. However, at low confining pressures and increasing
strain rates, a large drop of ef can be expected at the ductile-
hte
brittle transition. Because of increasing brittleness of response,

it can also be expected that ef will decrease with decreasing tem-
perature.
FIGURE 5-4 Typical stress-strain curves from constant-strain-rate
uniaxial compression tests on frozen medium-dense silt at –3 °C.
As for the correspondence of the complete behavior in CSR
Source: Reproduced with permission from Zhu and Carbee 1984. Copyright
and CSC tests, there is a simple practical method for directly
rig
1984 Elsevier. deducing the CSR curves from CSC data, and vice versa. For
that, as described by Vaid and Campanella (1977) in Fig. 5-5,
one must first deduce from the creep curves the corresponding
strain. This is also valid for the first peak in ice-cemented fro- ė min versus e curves, and then find the variation of stress with
zen sands. strain by intersecting the curves at a given constant strain rate.
py
Experimental support for a strain-based creep failure crite- A more general approach, based on energy considerations, was
rion can be found in the creep literature on unfrozen soils (e.g., proposed by O’Connor and Mitchell (1978).
Singh and Mitchell 1969; Campanella and Vaid 1974) and high-
temperature metals (Penny and Marriott 1971). In compres-
Ice Content Effect on Strength
Co
sion creep of frozen soils, the criterion was first mentioned by

Scott (1969) and was developed further by Ladanyi (1972a), The mechanical behavior of frozen soils depends, to a great
who used it as a basis for defining creep rupture in several later degree, on that part of the pore ice which normally binds the
studies (Ladanyi and Johnston 1973, 1974; Johnston and Lada- grains together and fills most of the pore space. It is now gener-
nyi 1972). The basic assumption in all these studies was that ally agreed that most of the ice found in soil pores is of a poly-
after long periods, crystalline type with a random crystal orientation. On the other
tf ė min = ef (5.2-1) hand, ice crystals within ice lenses are usually elongated in the
direction of heat flow during freezing (Penner 1961).
that is, the product of time to failure, tf , and the minimum The strength of ice depends on many factors; the most
creep rate was constant and equal to the total failure strain, ef . important are temperature, pressure, and strain rate, as well as
In fact, the assumption of a constant failure strain in creep, the size, structure, and orientation of grains. The strength of ice
which agrees in principle with the cumulative damage theory of increases with decreasing temperature, and its mode of failure
failure, is clearly an approximation, because it is found that the is strain-rate dependent. With varying temperature and strain

rate, its response to loading is found to vary from viscous to

brittle.
In permafrost soils, ice exists at very high homologous tem-
peratures, mostly above 90% of the fusion temperature, which
limits its deformation mechanisms to a narrow area, character-
ized by a power law type of creep, resulting mainly from the
motion of dislocations (Ashby and Frost 1975).
According to the experimental evidence shown by Gough-
nour and Andersland (1968) and Hooke, Dahlin, and Kauper
(1972), adding very small quantities of sand particles to poly-
crystalline ice has a weakening effect on the ice. But when sand
l
concentration is gradually increased, the sand particles have a
ria
strengthening effect, even if they are separated by ice. As pure
ice generally creeps faster than ice-rich frozen soil, the use of a
flow law for ice in the design of foundations in ice-rich perma- FIGURE 5-7 Average stress-strain curves for a frozen silt at
frost will lead to a safe design. This method was proposed by five different total water content ranges, at –1.67 °C.
Weaver (1979) and by Morgenstern, Roggensack, and Weaver Source: Reproduced with permission from Sayles and Carbee 1980. Copyright
ate
(1980), who checked its validity by comparing its predictions 1980 A. A. Balkema.
with long-term creep data for piles in ice and ice-rich soils.
Goughnour and Andersland (1968) studied the influence of between the ice matrix failure, occurring at small strains, and
sand concentration on strength of sand–ice mixtures at tem- shear failure of the entire mixture, which takes place at much
peratures ranging from –4 to –12 °C. When sand concentration larger strains. Their tests showed that at silt concentrations of
was increased beyond 42% by volume, the influence of inter- less than about 50%, the former dominates strength, whereas at
dM
particle friction and dilatancy became apparent, while at lower larger particle concentrations, the stress-strain curves manifest
concentrations, strengths were only a little higher than those of an increasingly strain-hardening character, which results from
pure ice. This finding, which agrees with some later test results a gradual mobilization of friction and interlocking at large
reported by Kaplar (1971), showed that at about 40% of sand strains (Fig. 5-7).
by volume, particle contact is established, giving rise to a rapid
increase in strength with increasing sand density. Clearly, as Normal Pressure Effect on Strength
hte
shown by Kaplar (1971) and T. H. W. Baker (1979) in Fig. 5-6,

the strength increases only as long as the sand remains ice satu- As noted in previous sections, at sand concentrations higher
rated. When the ice fraction tends to zero, the strength of an than about 40% by volume, the strength of frozen sand
unsaturated frozen sand decreases rapidly toward that for a dry becomes a function of the strength of both the ice cement and
sand. Sayles and Carbee (1980) considered that a better under- the soil skeleton. It has been found that these two sources of
standing of the effect of particle concentration on the behavior strength do not necessarily act simultaneously. This is because
rig
of ice-silt mixtures can be achieved if a distinction is made the ice matrix, under normal pressure and temperature condi-
tions, is much more rigid than the soil skeleton and attains its
peak strength at much lower strains. As a result, when a rela-
tively dense frozen sand is sheared in compression under a low
confining pressure, it shows two yield points: one at about 1%
py
axial strain and another at about 10% or more.

The behavior observed in triaxial compression tests on fro-
zen Ottawa sand by Chamberlain, Groves and Perham (1972)
and by Sayles (1973) is shown in Figs. 5-8 and 5-9. The behav-
Co
ior observed in compression tests at a constant strain rate and

temperature, but at different confining pressures, can be repre-
sented schematically by the Mohr plot in Fig. 5-10 (Ladanyi
1981b). The plot is composed of three failure lines: line I for ice
cement includes the region between the tensile strength of ice
and the pressure melting point; line II represents the drained
failure envelope for the sand skeleton, resulting from internal
friction and dilatancy, which is also affected by the presence of
FIGURE 5-6 Effect of total moisture content on unconfined ice; and line III is the undrained failure envelope for unfrozen
compressive strength of frozen fine sand, at –12 °C and at a sand. The complete failure behavior of frozen sand under such
strain rate of 2.2 × 10–6 s–1. conditions is then represented by the full line, which is a com-
Source: Reproduced with permission from T. H. W. Baker 1979. Copyright 1979 bination of the three basic lines, and which, together with the
Elsevier. transition zone, form four distinct regions, A to D.

FIGURE 5-8 Results of uniaxial compression tests with frozen FIGURE 5-10 Schematic representation of the entire failure
l
Ottawa sand, at –3.85 °C and at a strain rate of 5 × 10–4 s–1: envelope for frozen Ottawa sand.
ria
stress-strain curves at different confining pressures. Source: Reproduced with permission from Chamberlain, Groves, and Perham
Source: Reproduced with permission from Sayles 1973. Copyright 1973 1972. Copyright 1972 Thomas Telford. See also Ladanyi 1981b.
National Academies Press.
ate
dM
FIGURE 5-9 Results of uniaxial compression tests with frozen FIGURE 5-11 Mohr envelopes for creep strength of Ottawa
hte
Ottawa sand, at –3.85 °C and at a strain rate of 5 × 10–4 s–1: sand, at –3.85 °C and at a strain rate of 5 × 10–4 s–1.
Mohr envelopes for Ottawa sand and ice. Source: Reproduced with permission from Sayles 1973. Copyright 1973
Source: Reproduced with permission from Sayles 1973. Copyright 1973 National Academies Press.
National Academies Press.
increases to 10–3 s–1 (Sayles 1973). Because the transition

rig
In region A, which covers tension, tension-compression, and between regions B and C occurs only at confining pressures of
compression under low confining pressures, the ice cement about 55 MPa, the majority of permafrost problems will be
dominates. The stress-strain behavior is brittle in tension and limited to regions A and B.
strain softening in compression, with peak strength occurring
at strains of about 1%. In region B, the first peak is still at 1% Shape of the Failure Envelope. As noted in earlier sections,
py
strain; but the second peak, due to friction and dilatancy, dom- the shape of the failure envelope of a frozen soil tends to be
inates at strains about 10 times larger. However, at confining fairly complex, and it is expected to depend on the soil type, its
pressures of more than 50 MPa (Chamberlain, Groves, and Per- density, and ice saturation, as well as on the temperature and
ham 1972), the dilatancy is suppressed, and at still higher pres- strain rate. In addition, because intergranular stresses cannot
Co
sures it changes sign. When this happens, a large portion of be measured during a shear test, all the results can only be plot-
normal pressure is transferred to the ice during shear. In the ted in terms of total stresses. Nevertheless, if one compares the
transition zone C, ice melts partially, but because little consoli- results of strength tests obtained by various authors, there are
dation is possible, shear occurs under undrained conditions certain common conclusions.
with a limited contribution of ice matrix strength. Finally, as The shape of the failure envelope is approximately parabolic
has been explained by Chamberlain, Groves, and Perham at relatively low temperatures and high strain rates. When the
(1972), when the confining pressure becomes so high that it temperature increases and/or the strain rate decreases, the fail-
can crush the sand grains, the pore ice that is already under ure envelope shrinks and straightens, with its slope slightly
compression thaws, and shear failure occurs as in an unfrozen smaller or equal to that of the same soil when unfrozen (Fig. 5-
sand under undrained conditions. 11). At very low strain rates (or very long times under stress)
From the results of these experimental studies, it appears and/or at temperatures close to the melting point, the cohesion
that the transition pressure between regions A and B is about intercept tends to zero, and the remaining strength is then gov-
3–4 MPa, increasing up to about 7 MPa when the strain rate erned by intergranular stresses and mineral cohesion.

The ratio between values of uniaxial compressive and uniax- about four times higher strain rates than for ice. More gener-
ial tensile strengths depends strongly on the strain rate and tem- ally, when the peak strengths obtained in such tests are plotted
perature, and it varies from 1 at low strain rates up to about 5 at against the applied strain rates in a log–log plot, it is often
high strain rates, which is similar to polycrystalline ice. This is found that the resulting line is not a continuous straight line, as
because the latter strength is much less sensitive to rate and tem- is assumed by the power law creep equation of the type
perature than is the former, at least in the brittle failure range
ė = Bsn (5.2-2)
(Perkins and Ruedrich 1973; Bragg and Andersland 1982). At
ordinary temperatures a frozen soil can hardly be called brittle but its slope, defined by n = d(log ė)/d(log s), tends to be lower
in the fracture mechanics sense. But it may satisfy this definition at low rates of strain and higher at high strain rates.
at very low temperatures, such as those encountered in ground For a dense frozen sand at low temperatures, n tends to be on
freezing and in the underground storage of liquefied natural gas the order of 10 or more, and it seems to be very little affected by
l
(as will be shown further in this section). In any case, the brittle- temperatures below –5 °C. At higher temperatures, such as –2
ria
ness affects both the shape of the stress-strain curve and the °C, a break in slope at a rate of about 10–5 s–1 has been observed
amount of postpeak loss of strength (Fig. 5-12). by several investigators (Perkins and Ruedrich 1973; Parame-
swaran 1980; Bragg and Andersland 1982), as is shown in Fig. 5-
14, reducing n to 5 or even 3 at low rates. Clearly, in very ice rich
Strain Rate Effect on Strength
soils, ice governs the behavior and n = 3 closely approximates
ate
Over a range of strain rates, the behavior of a frozen soil will the results (McRoberts, Law, and Murray 1978; Weaver 1979;
probably be similar to that found by Haynes, Karalius, and Morgenstern, Roggensack, and Weaver 1980). There are also
Kalafut (1975) for a frozen silt at –9.4 °C, shown in a log–log indications that n decreases considerably with decreasing ice sat-
plot in Fig. 5-13. The observed rate sensitivity of peak strength uration and when a cyclic loading is applied to a frozen sand (Li
of the frozen silt is found to be similar to that reported for poly- and Andersland 1980), as is shown in Fig. 5-15.
crystalline ice by Hawkes and Mellor (1972), the main differ-
dM
ence being that for frozen silt the two strength lines separated at
hte
FIGURE 5-14 Compressive strength versus strain rate rela-

rig
tionship for a frozen sand at different temperatures.

FIGURE 5-12 Typical stress-strain curves for a frozen sand at Source: Reproduced with permission of P. E. Frivik from Bragg and Andersland
1982.
different strain rates and at –10 °C.
Source: Reproduced with permission of P. E. Frivik from Bragg and Andersland
py
1982.
Co
FIGURE 5-15 Comparison of creep strengths of frozen sand

FIGURE 5-13 Relationship between strength and strain rate under static and dynamic loading conditions.
for a frozen silt, at –9.4 °C. Source: Reproduced with permission from Li and Andersland 1980. Copyright
Source: Reproduced from Haynes, Karalius, and Kalafut 1975. 1980 A. A. Balkema

As for the shape of the stress-strain curve, it is generally 1974; Haynes and Karalius 1977; Haynes 1978), as is shown in
found that lower temperatures and strain rates both reduce the Fig. 5-18. Down to about –10 °C, the embrittlement effect of
failure strain. If the strain at absolute maximum strength is temperature is observed more in a frozen sand or silt than in a
considered, which may be either the first (ice–cement) or the frozen clay, which at that temperature still contains enough
second (friction) peak, the variation of failure strain with strain unfrozen water to keep it plastic.
rate will not necessarily be continuous but may show a sharp Although there have been some attempts in the past to
drop at the brittle-plastic transition. Figure 5-16 shows a typi- express the strength variation of frozen soils with temperature
cal variation of the failure strain, which is seen to be remark- by means of the theory of reaction rate processes, at present it is
ably constant in each of the two strain rate regions. A similar felt that the theory can be used only as an approximation. In
embrittlement effect with increasing strain rate was also fact, as was mentioned by Hoekstra (1969a), the theory is not
observed by Ladanyi and Paquin (1978) in a frozen sand at –6 applicable to materials in which a temperature change pro-
duces a continuous phase change. In other words, though the
l
°C. They also found an increasing loss of strength after the peak
ria
with increasing strain rates, which practically disappeared theory may be applicable to frozen soils at very low tempera-
when compression tests were performed with a confining pres- tures, it is considered better to replace it by an empirical rela-
sure of 3.55 MPa (Fig. 5-17). tionship giving the best fit to experimental results, as is shown,
for example, in Fig. 5-19a and 5-19b. One such convenient rela-
tionship will be shown in Section 5.3.
ate
Temperature Effect on Strength
Because of its direct influence on the strength of intergranular
Frozen Soil Behavior at Cryogenic Temperatures
ice and on the amount of unfrozen water in a frozen soil, tem-
perature has a marked effect on all aspects of the mechanical Although in permafrost regions the ground temperature only
behavior of frozen soils. In general, a decrease in temperature rarely descends below –40 °C, such low temperatures are cur-
results in an increase in strength of a frozen soil, but at the same rently attained in ground freezing practice if liquid nitrogen is
dM
time it increases its brittleness, which is manifested by a larger used as the refrigerant. Conversely, underground storage of liq-
drop of strength after the peak and an increase in the ratio of uefied natural gas (LNG) may bring the ground temperatures
compressive strength to tensile strength (Sayles and Haines down to –160 °C.
In the past, the results of several studies on mechanical
properties of frozen soils at cryogenic temperatures have been
published (e.g., Wolfe and Thieme 1964; Sayles 1966; Biarez,
hte
Mascarelli, and Desvarreux 1967). The following brief review

shows the results of an investigation made by Bourbonnais
(1984) and published by Bourbonnais and Ladanyi (1985a,
1985b, 1985c). The study involved a series of uniaxial compres-
sion tests with a frozen dense fine sand and a dense saturated
clay, respectively.
rig
FIGURE 5-16 Failure strain versus strain-rate relationship for a

frozen sand.
py
Source: Reproduced with permission of P. E. Frivik from Bragg and Andersland

1982; see also Ladanyi 1981b.
Co
FIGURE 5-17 Peak and post-peak strengths of a frozen sand

in unconfined and confined triaxial compression, at –6 °C. FIGURE 5-18 Average strength-versus-temperature relation-
Source: Reproduced with permission from Ladanyi and Paquin 1978. Copyright ship for a frozen silt in uniaxial compression and tension tests.
1978 National Research Council of Canada. Source: Reproduced from Haynes and Karalius 1977.

l
ria
ate
FIGURE 5-20 Low temperature uniaxial compression tests
with frozen sand: Influence of temperature on stress-strain
dM curves at a constant strain rate.
Source: Reproduced with permission from Bourbonnais and Ladanyi 1985c.
Copyright 1985 A. A. Balkema.
°C. This type of frozen sand behavior, which was also observed
earlier by Sayles (1966), can be explained by thermal micro-
cracking during cooling, due to the thermal expansion mis-
hte
match between sand grains and pore ice.
Frozen Clay. A series of uniaxial compression tests (Bourbon-

nais and Ladanyi 1985b) was conducted on an overconsoli-
dated, undisturbed Boom clay from Belgium, with a typical
rig
water content of about 25%. The tests were performed at tem-

peratures varying from –16 to –154.4 °C, and at strain rates
between 2.6 × 10–3 and 9.2 × 10–6 s–1. A set of typical stress-
strain curves from this series of tests, which are shown in Fig. 5-
22, leads to the conclusion that despite temperatures decreasing
py
FIGURE 5-19 Temperature-time dependence of uniaxial com- to about –110 °C, the frozen clay retained its plastic behavior,
pressive strength for two frozen sands. up to axial strains of more than 5%. Only below –110 °C, when
Source: Reproduced from Sayles 1968. all, even molecularly bound, water was frozen did the clay
behave in a brittle manner. As is shown in Fig. 5-21, its strength
Co
increase was exponential down to –100 °C, and became more

Frozen Sand. Specimens of a fine Le Sueur sand, prepared in a attenuated in the brittle range, below that temperature.
well-controlled manner with an average water content of 20.5% A combined plot of compression strength variation with de-
and a degree of saturation of 92%, were tested in uniaxial com- creasing temperature for these two soils, together with the strength
pression at temperatures ranging from –6.7 to –160°C, and at curve obtained for a silty sand, are shown in Fig. 5-21. The ob-
axial strain rates between 3 × 10–3 and 3 × 10–5 s–1 (Bourbon- served behavior closely resembles that published by Sayles (1966).
nais and Ladanyi 1985c). Some typical results of this study are
shown in Figs. 5-20 and 5-21. As seen in Fig. 5-20, except for
the test at –6.7 °C, all the stress-strain curves show an increas- 5.3 Analytical Representation of Creep and
ingly brittle behavior with decreasing temperature, failing at Strength Data
axial strains of about 0.5%. It is shown in Fig. 5-21 that for fro-
zen sand, strength increases sharply with decreasing tempera- The main practical purpose of creep and strength testing of fro-
ture to about –40 °C but tends to level off or peak at about –100 zen soils under different conditions (density, ice content, tem-

l
ria
ate
dM
FIGURE 5-21 Influence of temperature and soil type on uniaxial compression strength of three typical frozen
soils.
hte
Source: Bourbonnais 1984; reproduced with permission from Bourbonnais and Ladanyi 1985a. Copyright 1985 A. A. Balkema.
perature, confining pressure, etc.) is the formulation of a con- frozen soils is limited to the latter type of constitutive equa-
stitutive equation that relates together the macroscopic tions.
variables of strain rate, stress, temperature, and time (or accu-
rig
mulated strain). When frozen soil deforms, its structure

General Creep Equation
changes continuously and its current structure or state deter-
mines its further response to stress. So a properly formulated An ice-saturated frozen soil under an increase in stress con-
constitutive equation should contain, in addition to the macro- taining a substantial deviatoric component will show an
scopic variables, one or more state variables, describing the instantaneous response, both elastic and plastic, followed by
py
current state of the structure (Ashby and Duval 1985). creep, sometimes combined with consolidation. As is shown in
There are several examples of equations of this type in the Fig. 5-2b, the creep is initially of a primary type, followed by a
literature, proposed for various materials (e.g., Ashby and short steady-state (or secondary) portion, and eventually
Duval 1985, for ice; Aubertin, Gill, and Ladanyi 1991, for rock going into tertiary creep, accelerating toward failure. Because
Co
salt). However, for engineering purposes, various phenomeno- tertiary creep is considered to be affected by boundary condi-
logical constitutive equations are most often used, because of tions of the test, which may be different from those in a real
their simplicity and limited number of required experimental problem, it has been customary until now not to include it in
parameters. These equations ignore the structure and attempt design. Instead, it is usually assumed that creep failure occurs
to establish, by curve fitting, a functional relationship between just at the start of tertiary creep, or at the minimum creep rate,
strain, e, stress, s, temperature, T, and time, t : which often occurs at a reasonably constant accumulated shear
strain.
e = f (s, T, t) (5.3-1)
The complete stress-strain-time behavior of frozen soils is
One should be aware that although useful for practice, such fairly complex. For design purposes, one is mostly satisfied
equations have in fact very little physical basis; they only with a simplified picture of that behavior, which can be
describe a curve fitted to the data and are valid only over the described by some simple mathematical expressions, contain-
range of variables for which the data are available. The following a limited number of experimentally determined parame-
ing analytical treatment of creep and strength information on ters, and valid in a limited area of application.

In the latter case, for the portion of creep curves at and

beyond the inflection point (or point B in Fig. 5-2b), but before
tertiary creep, the total strain can be expressed by (Hult 1966)
e = e(i ) + e(min
c)
t (5.3-3)
where ė (c)min = de(c)/dt is the minimum (or steady-state) creep

rate and t is time. On the basis of available experimental experi-
ence with frozen soils, and following Hult (1966) and Ladanyi
(1972a), it is found that in Eq. (5.3-3) both e(i) and ė (c)min can
be expressed conveniently by power law approximations:
l
k
Ê s ˆ
ria
e(i ) = ek Á (5.3-4)
Ë sk q ˜¯
and
n
ate
Ê s ˆ
e(min
c)
= ec Á (5.3-5)
Ë sc q ˜¯
where, in Eq. (5.3-4), sk q is a temperature-dependent total

deformation modulus, corresponding to the reference strain ek,
and k £ 1 is an empirical exponent. Similarly, in Eq. (5.3-5), scq
dM
is the temperature-dependent creep modulus, corresponding to
the reference strain rate, ėc , whereas n ≥ 1 is an experimental
creep exponent. All of these experimental parameters in Eqs.
FIGURE 5-22 Low-temperature uniaxial compression tests
(5.3-4) and (5.3-5) can be determined by plotting the creep test
with frozen clay: Influence of temperature on stress-strain results in appropriate log–log plots (see Example 5.3-1).
curves at a constant strain rate.
Source: Reproduced with permission from Bourbonnais and Ladanyi 1985c. ■ EXAMPLE 5.3-1: A series of uniaxial compression creep
hte
Copyright 1985 A. A. Balkema. tests on a fine silty sand at a moisture content of 30% and at a
temperature of –1 °C resulted in a set of creep curves, similar to
those shown in Fig. 5-2b, each of them corresponding to a dif-
In frozen soil mechanics, it is usually assumed that the total ferent value of the applied constant stress s1. The curves are
strain, e, resulting from a deviatoric stress increment, is com- characterized by the intercepts e(i) at t = 0, and by their mini-
posed of an instantaneous strain, e0, and a delayed or creep mum creep rates e1,min
(c )
, as indicated in Table 5-1. It was also
rig
strain, e(c): found that the minimum creep rate (inflection point of creep
curves, as in Fig. 5-2b) occurred at a total strain, elf of about
e = e0 + e(c) (5.3-2) 0.15. On the basis of these data, determine:
The instantaneous strain, e0 (Fig. 5-2b), may contain an elastic (a) The values of parameters k and sk in the pseudo-instanta-
py
and a plastic portion, but at the usual service loads, excluding neous deformation equation [Eq. (5.3-4)].
instantaneous failure, the plastic portion may be absent. The (b) The parameters n and sc in the creep rate equation [Eq.
creep strain, in turn, is composed of both primary and second- (5.3-5)].
ary (steady-state portion) creep, although the latter may some- (c) Using these parameters, determine the values of uniaxial
Co
times be reduced to just an inflection point on the creep curve, compression strength of this frozen soil after 10 min and
preceding tertiary creep (Fig. 5-2b). 10 years, respectively.
In practice, for relatively short-term processes such as
ground freezing, the strain e0 in Eq. (5.3-2) is considered to be TABLE 5-1 Test Data for Example 5.3-1
governed by Hooke’s law, while the creep strain, e(c), is usually
Test ␴1 (kPa) ⑀(i)
1 ⑀·(c) –1
1,min, (min )
defined by an empirical primary creep formulation. For long-
1 3,131 0.100 1.895 × 10–2
term problems; such as the behavior of foundations in perma-
frost, the short-term response—including elastic, plastic, and 2 2,221 0.071 4.550 × 10–3
primary creep portions—is sometimes lumped together to 3 1,434 0.046 1.770 × 10–3
form a pseudo-instantaneous plastic strain, e(i) (Hult 1966; 4 972 0.031 4.710 × 10–4
Ladanyi 1972a). This strain is defined by the intersection on the 5 552 0.0154 3.583 × 10–5
strain axis (Fig. 5-2b), when the slope at the minimum or 6 317 0.0108 5.383 × 10–6
steady-state creep rate is extrapolated back to t = 0.

l
ria
ate
dM
hte
FIGURE 5-23 Determination of parameters in Eqs. (5.3-4) and (5.3-5).
Solution: Plot the values of e1(i) and ė 1(c) against applied these equations by their “equivalent” values (Odqvist and Hult
stresses in a log–log plot (Fig. 5-23) and draw average straight 1962), defined by
lines through the experimental points. From the slopes of the
lines, obtain k = 1 and n = 3.56. Select ek = 10–2 and ė c = 10–6 se2 = 23 Sij Sij = 3J 2¢ (5.3-6)
rig
min–1 and read from Fig. 5-23 values for sk = 315 kPa and sc =
197 kPa. Now write: ee2 = 23 eij eij = 43 I 2¢ (5.3-7)
ee2 = 23 eij eij = 43 I2¢

1
s (5.3-8)
(a) e1(i ) = 10-2 ÊÁ 1 ˆ˜ = 3.17 ¥ 10-5 s1
py
Ë 315 ¯
where Sij and eij are the deviatoric stress and strain tensors,
3.56 respectively, while J2¢ and I2¢ are the second invariants of the
s
(b) e1(c ) = 10-6 ÊÁ 1 ˆ˜ stress and strain deviator tensors. The dot above a symbol
Ë 197 ¯
Co
denotes time rate.

1
With this generalization, Eq. (5.3-5) becomes
0.28
È e1f - e1(i ) ˘ n
Ê 0.15 - 3.17 ¥ 10-5 s1f ˆ n
(c) s1 f = sc Í ˙ = 197 Á ˜ Ês ˆ
ÍÎ ec t f ˙˚ ee(cmin = ec Á e ˜ (5.3-9)
)
Ë 10-6 t f ¯
Ë sc q ¯
with stresses in kPa and time in minutes. For tf = 10 min, one
which is the well-known Norton–Bayley power law creep equa-
gets by trial and error s1f = 2,388 kPa, and for tf = 10 years =
tion, used extensively in the literature for steady-state creep for-
5.256 × 106 min, s1f = 72.5 kPa.
mulation in high-temperature metals and ice. Written in tensor
Assuming the validity of von Mises flow rule and volume form, Eq. (5.3-9) becomes
constancy for all plastic deformations, including creep strains,
the power laws of Eqs. (5.3-4) and (5.3-5), adopted for the n -1
3ec Ê se ˆ
uniaxial case, can be generalized for the triaxial state of stress eij(c ) = Sij (5.3-10)
and strain by expressing stresses, strains, and strain rates in 2sc q ÁË sc q ˜¯

It is noted that for axial symmetry, such as in triaxial compres- The first one, based on the theory of rate processes (Ander-
sion tests on cylindrical specimens, se = (s1 – s3) and ė e = ė 1, sland and A1-Nouri 1970), has the form
whereas for plane-strain conditions, assuming material incom-
pressibility, se = ( 3/2)(s1 – s3), and ė e = 2/ 3 ė 1. È Lq ˘
f1 (q) = exp Í ˙ (5.3-18)
Because of the assumed validity of the von Mises law, lead- Î 273n (273 - q) ˚
ing to the foregoing relations, the power law of Eq. (5.3-9)
becomes, for cylindrical symmetry, or, approximately, for q 273 °K,
Ê s - s3 ˆ
n È Lq ˘ (5.3-19)
f1 (q) ª exp Í
e1(c ) = ec Á 1 (5.3-11) ˙
Î (273)2 n ˚
Ë sc q ˜¯
for plane strain, where L = U/R (°K) is the ratio between the apparent activation
l
energy U (J/mol) and the universal gas constant R = 8.32 (J/
ria
n +1 n mol · K). Some reported values for L include L = 4,274 K for
Ê ˆ Ê s - s3 ˆ
e1(c ) = Á 3 ˜ ec Á 1 (5.3-12) frozen saturated Ottawa sand (Andersland and Al-Nouri 1970),
Ë 2 ¯ Ë sc q ˜¯ L = 56,000 K for frozen Sault Ste. Marie clay (Akili 1970), and L
= 10,000 K for polycrystalline ice (Gold 1970). This form of the
and for simple shear,
ate
temperature function may be valid only at lower temperatures,
n where the change of phase associated with temperature varia-
Ê t ˆ
g = e1 - e 3 = 3(n +1)/ 2 ec Á (5.3-13) tion becomes negligible.
Ë sc q ˜¯ The second form of f (q), based on Vialov (1962) and pre-
sented in Ladanyi (1972a), is a power law
These relationships make it possible to determine the creep
parameters n, b, and sc q from a series of simple laboratory tests.
dM w
Ê qˆ (5.3-20)
In the primary creep range, in turn, it is usually considered f 2 (q ) = Á 1 + ˜
that the creep strain e(c) in Eq. (5.3-2) can be expressed as a Ë qc ¯
product of independent stress, time, and temperature functions where qc is an arbitrary temperature, say 1 °C. The exponent w is
(Hult 1966): obtained by plotting log(sc q) versus log(1 + q/qc ). The slope of
e(c) = f1(s) f2(t) f3(T) (5.3-14) the straight line drawn through the experimental points gives
hte
A convenient form of the primary (or transient) creep law is D log(sc q )

w= (5.3-21)
Andrade’s empirical law: Ê qˆ
D log Á 1 + ˜
Ë qc ¯
e(c) =Asntb (5.3-15)
which can be extended to three dimensions by assuming the Finally, in certain cases and for small temperature variations, a
linear relationship between sc q and q is often observed. In that
rig
validity of von Mises flow rule. For example, Ladanyi and

Johnston (1974) wrote the law in the form case, w = 1, and one obtains a linear temperature function
n q
Ê s ˆ Ê e t ˆ
b
f3 (q) = 1 + (5.3-22)
e(ec ) =Á e ˜ Á c ˜ (5.3-16) qo
Ë sc q ¯ Ë b ¯
py
where qo is the intercept on the q-axis of the straight line pass-

In Eq. (5.3-11), n, b, and sc q are three experimentally deter- ing through the experimental data points.
mined frozen soil parameters, of which the last one, sc q, Equation (5.3-16) represents the time-hardening formula-
denotes the reference stress corresponding to an arbitrary refer- tion of primary creep. The corresponding creep rate for that
Co
ence strain rate, ė c, and to a soil temperature q (°C) = –T (°C). case is

A semigraphical method for determining values of the three
creep parameters from a series of laboratory creep tests is n 1-b
Ês ˆ Ê b ˆ
shown in Example 5.3-2. ee(c ) = ec Á e ˜ Á (5.3-23)
The effect of temperature on creep of a frozen soil can be Ë sc q ¯ Ë t ec ˜¯
included in the value of the creep modulus sc q by means of an
empirical formula (Ladanyi 1972a): ■ EXAMPLE 5.3-2: Time and creep-rate data obtained by
Sayles (1973) for frozen Ottawa sand at a constant temperature
sc q = sco f (q) (5.3-17)
are summarized in Fig. 5-24. Determine the creep parameters
where sco is the value of sc q , obtained in unconfined compres- b, n, and sc needed for the primary creep-rate equation (5.3-
sion creep tests, extrapolated back to 0 °C, as shown in Ladanyi 23); then evaluate (a) the creep rate after 24 h, and (b) the com-
(1972a). In the same reference, three different forms of the pressive strength for t = 24 h and t = 1 year, respectively, if the
temperature function f (q) were proposed. creep failure strain is 1%.

D log (Cb)
n= = tan b (5.3-30)
D log (s1 - s3 )
Finally, from any point on that line, say [(Cb)i , (s1 – s3)i], one
can determine the value of sc (for a selected value of ė c), which
is, according to Eq. (5.3-25), equal to
1
È Ê e ˆ b ˘ n
b c
sc = (s1 - s3 )i ÍÍ ÁË b ˜¯ ˙˙ (5.3-31)
ÍÎ (Cb)i ˙˚
l
Solution: From the slope of log ė 1 versus log t (Fig. 5-24),
ria
obtain (1 – b) = tan a = 0.7, and b = 0.3. Plot Cb values (ė 1(c) at
t = 1h) against (sl – s3). The resulting straight line represents
Cb = f (s1 – s3) with a slope giving n = tan b = 0.87.
For (s1 – s3)i = 1 MPa, obtain (Cb)i = 4 × 10–4 h–1. Assum-
ate
ing that ė c = 10–8 h–1 and using Eq. (5.3-31), obtain sc = 5.32
MPa. Hence
FIGURE 5-24 Determination of primary creep parameters in
Eq. (5-23) from the results of uniaxial compression creep 0.3 0.87
Ê 10-8 ˆ Ê s1 - s3 ˆ
tests with frozen Ottawa sand at –3.85 °C. e1(c ) = 0.3 Á ÁË 5.32 ˜¯ t -0.7
Ë 0.3 ˜¯
Source: Reproduced with permission from Sayles 1973. Copyright 1973
dM
National Academies Press. = 4 ¥ 10-4 (s1 - s3 )0.87 t -0.7
with stresses in MPa and time in hours.

Method: For a uniaxial compression test, Eq. (5.3-23) can be
written as (a) The strain rate after 24 h at s1 – s3 = 1 MPa will be ė 1(c) =
4 × 10–4 × 24–0.7 = 4.32 × 10–5 h–1.
e1(c ) = Cbt -(1-b ) (5.3-24)
(b) According to Eq. (5.3-37) and for a creep failure strain of
hte
where 0.01, the creep strength equation becomes
n
Ê e ˆ Ê s - s3 ˆ
b (s1 - s3 )f = 10.134t -0.345
Cb = b Á c ˜ Á 1 (5.3-25)
Ë b ¯ Ë sc ˜¯
which for t = 24 h gives 3.387 MPa, and for a year ( = 24 × 365
Note that Eq. (5.3-24) linearizes in a log ė 1(c) versus log t plot, h) a strength of 0.443 MPa.
rig
because for a constant stress, Cb = constant and If time t is eliminated from Eq. (5.3-16), one gets the strain-
hardening form of the creep rate equation,
log e1(c ) = log (Cb) - (1 - b)log t (5.3-26)
n
Ês ˆ b -(1-b )/b
e(ec ) = ec Á e ˜ (5.3-32)
py
The slope of this descending line in Fig. (5-24) gives ÈÎe(cc ) ˘˚

Ë sc q ¯
D log e(c )
1-b = = tan a (5.3-27) Although the strain-hardening formulation offers a more accu-
D log t rate representation of reality than the time-hardening formula-
Co
which permits computation of the exponent b. Conversely, for tion, the latter is nevertheless often preferred because it makes
a constant time and variable stress, it possible to obtain closed-form solutions of some simple
practical problems. In addition, as has been shown by Hult
(1966), if a time-hardening assumption is adopted, any solu-
Cb = e1(c )t 1-b = C1 (s1 - s3 )
n (5.3-28)
tion obtained for a steady-state creep law of Eq. (5.3-5) can
where readily be transformed into a transient creep form by consider-
ing that the strain rate in Eq. (5.3-5) is the result of a differenti-
Ê e ˆ
b ation of strain, not with respect to time, t, but with respect to
C1 = b Á c ˜ sc-n (5.3-29)
Ëb¯ an arbitrary time function, F(t). If F(t) = t b is selected for the
time function, it is found that the transient creep form of a
Plotting Cb (at t = constant) against (s1 – s3) in a log–log plot steady solution can be obtained by replacing ė c everywhere by
2
(Fig. 5-24) gives an ascending line, whose slope is (ė /b)b, and t by t b (Ladanyi and Johnston 1974).

Strength of Frozen Soils k

Ê s ˆ
As in unfrozen soils, the concept of failure in frozen soils eef - ek Á e ˜
Ë skq ¯
includes both rupture and excessive deformation. Depending tf = (5.3-35)
n
Ês ˆ
on soil type, temperature, strain rate, and confining pressure, ec Á e ˜
the mode of failure may vary from brittle, similar to that in a Ë sc q ¯
weak rock, through brittle-plastic, with formation of a single
For high-ice-content soils and for large time intervals, the plas-
failure plane or several slip planes, to purely plastic failure
tic strain e(i) can be neglected relative to the creep strain por-
without any visible strain discontinuities. The last type, failure
tion; hence Eq. (5.3-35) becomes
by excessive creep deformation, is typical for permafrost prob-
lems involving ground temperatures only a few degrees below eef
the melting point of ice. tf = (5.3-36)
l
n
The creep strength is defined as the stress level at which, Ês ˆ
ec Á e ˜
ria
after a finite time interval, either rupture or instability leading Ë sc q ¯
to rupture (e.g., tertiary creep) occurs in the material. In com-
pression testing of frozen soils, the creep strength is usually This also makes it possible to write the creep strength of a fro-
zen soil as a function of time to failure,
defined as the stress at which the first sign of instability occurs.
ate
In a constant-stress creep test, this condition coincides with the 1
passage from steady state to accelerated creep (point C in Fig. Ê eef ˆ n
sef ª sc q Á (5.3-37)
˜
5-2b), or simply to the inflection point on the creep curve (i.e., Ë t f ec ¯
when points B and C in Fig. 5-2b coincide). Conversely, in a
constant-strain-rate compression test (Fig. 5-4), this condition or as a function of the minimum creep rate,
corresponds to the first drop of strength after the peak of the
dM
1
stress-strain curve. È e(c ) ˘ n
Creep strength prediction consists of finding a relationship sef ª sc q Í e ,min ˙ (5.3-38)
among creep strength, se f , time to failure, t f , secondary or ÍÎ ec ˙˚
minimum creep rate, ė (c)min, failure strain, eef , and tempera-
if, for long time intervals, one defines
ture, q = –T .
If a secondary creep formulation of Eq. (5.3-3) is adopted, eef
e(ec,min
)
ª (5.3-39)
hte
one gets, from Fig. 5-2b,

tf
eef = e(i ) + t f e(min
c) (5.3-33) If the primary creep formulation of Eq. (5.3-16) is adopted,
and considering that the true instantaneous strain (eo in Fig. 5-
from which
2b) can be neglected relative to the primary creep strain, the
eef - e(i ) creep strength becomes
rig
tf = (5.3-34) b
e(min
c)
nÊ
b ˆ n
(5.3-40)
sef ª sc q e1/
ef Á ˜
Compression creep testing of frozen soils often shows that the Ë ec t ¯
amount of permanent strain at the onset of tertiary creep is Clearly, for b = 1, Eq. (5.3-40) reduces to Eq. (5.3-38).
py
approximately constant for a given temperature and type of test

(e.g., Vialov 1962; Bragg and Andersland 1982; Zhu and Carbee ■ EXAMPLE 5.3-3: Uniaxial compressive strength tests were
1984). This behavior suggests that instability occurs when the conducted by Zhu and Carbee (1984) on remolded saturated
total damage done by straining reaches a certain critical value. frozen Fairbanks silt, under various constant strain rates, tem-
Co
Although there is some experimental justification for using a peratures, and dry densities. Figure 5-4 shows the resulting
constant permanent strain as a basis for the creep failure crite- stress-strain curves for this frozen silt with a moisture content
rion in frozen soils, this criterion is convenient for design pur- of 42% and dry density of 1,200 kg/m3, at a temperature of
poses, because it limits the total strain to values acceptable for –3.0 °C, and at different strain rates. Applying Eq. (5.3-37) to
the structure. the peak strength of these curves, evaluate the creep parameters
In actual compression testing, when both constant-stress- n and sc for a given reference strain rate, write the creep
creep tests and constant-strain-rate compression test results are strength equation (5.3-37), and estimate the 10-year strength of
available, it is observed that this critical creep strain is approxi- this material, assuming the average creep failure strain to be
mately equal to the failure strain at the peak of the stress-strain about 7.5%.
curves in the latter type of tests. Solution: Inspecting Fig. 5-4, one finds the data shown in
For creep failure prediction on the basis of Eq. (5.3-34), the Table 5-2. When data from the table are plotted in a log(stress)
values of e(i) and ė (c)min should be substituted into the equa- versus log(strain rate) plot (Fig. 5-25), they can be approxi-
tion, to obtain mated by two straight lines, one between points 1 and 2, with n

l
ria
ate
FIGURE 5-25 Determination of creep parameters in Eq. (5.3-28) from the results shown in Fig. 5-4.
dM
TABLE 5-2 Data for Example 5.3-3 l k
Ê me ˆ Ê qˆ
Test Applied strain rate, ⑀·(c) Peak strength, ␴1f (MPa) sc q = wqck Á c ˜ Á 1 + ˜ (5.3-42)
–6
1
Ë l ¯ Ë qc ¯
1 1.08 × 10 1.09
–5 In practice, Eq. (5.3-16) is generally preferred to Eq. (5.3-41),
2 1.10 × 10 1.32
3 1.13 × 10–4 1.91 because the former is written in a dimensionally balanced form.
hte
4 1.12 × 10–3 3.30

Conversely, Vialov (1959) suggested that the time depen-
dence of strength of frozen soils could be represented by the
5 5.63 × 10–3 4.73
following empirical law:
so
= 12.12 and sc = 1.09 MPa at ė c = 1.08 × 10–6 s–1, and another sf = (5.3-43)
Ê f +t *ˆ
t
rig
one between test points 2 and 5, with n = 4.89 and sc = 1.32 ln Á ˜¯

MPa at ė c = 1.10 × 10–5 s–1. If the latter is extrapolated to 10 Ë t o
years = 315.36 × 106 s, corresponding to an average strain rate where so and to denote parameters that depend on soil type and
of 0.075/(315.36 × 106) = 2.378 × 10–10 s–1, one gets temperature, tf is time to failure, sf is the time-dependent
py
1
uniaxial compression strength, and
Ê 2.378 ¥ 10-10 ˆ 4.89
s1 f = 1.32 Á = 0.1467 MPa
Ë 1.10 ¥ 10-5 ˜¯ Ês ˆ
t * = t o exp Á o ˜ (5.3-44)
Ë si ¯
Co
Comparison with Vyalov’s Creep and Strength where si is the initial instantaneous strength. For long time
Equations intervals, Vyalov (1963) stated that the quantity t* may be
neglected.
To describe the primary creep of frozen soil, Vialov (1959) pro- Although the above equations have sometimes been used to
posed the following equation: approximate the time-dependent strength, they have the disad-
1 vantage of requiring a new set of empirical parameters, which is
È st l ˘ m
much less the case with Eq. (5.3-38) or (5.3-40), based on the
e=Í ˙ (5.3-41)
Î w(q + qc ) ˚
k creep theory, as was shown previously in this section. Table 5-3
shows some typical values of experimentally determined creep
It is apparent that Eq. (5.3-41) is a creep law of the Andrade parameters, b, n, w, and sco (the last for ė c = 10–5 h–1) for a vari-
type, having the same mathematical form as Eq. (5.3-15), into ety of frozen soils and for polycrystalline ice, based on the data
which it can readily be transformed by noting that b = l/m, n = published in the literature. Table 5-4 shows the main physical
1/m, and properties of the same soils.

TABLE 5-3 Values of Parameters in Eqs. (5.3-16) and (5.3-17), Obtained for Some Typical Frozen Soilsa
Frozen soil type Source b n w ␴co (MPa)b Note
Clays
Suffield clay Sayles and Haines 1974 0.33 2.38 1.20 0.17
Bat-Baioss clay Vialov 1962 0.45 2.50 0.97 0.18
Silts and loams

Hanover silt Sayles and Haines 1974 0.15 2.04 0.87 2.25
Callovian loam Vialov 1962 0.37 3.70 0.89 0.31
Ice-rich silt (undisturbed) (10 < e· < 10–4 h–1)
–8
McRoberts, Law, and Murray 1978 1.00 3.00 0.60 00.071 Long term
l
ria
Sands
Ottawa sand Sayles 1968 0.45 1.28 1.00 1.05
Manchester fine sand Sayles 1968 0.63 2.63 1.00 0.16
Karlsruhe silty sand Meissner and Eckhardt 1976 0.40 2.00 1.00 0.30
ate
Very ice-rich soil or polycrystalline ice Morgenstern, Roggensack, and 1.00 3.00 0.37 00.103 Long term
(for T £ –1 °C, 10–7 < e· < 10–2 h–1) Weaver 1980
a
ee = (se / scq)n ( e· ct / b)b [Eq. (5.3-16)]; scq = sco(1+q / 1 °C)w [Eq. (5.3-17)].
b
e· c = 10–5 h–1. dM
TABLE 5-4 Physical Properties of Soils from Table 5-3a
<0.002 mm LL PL Sr ␳d ␳
Soil (%) <0.1 mm (%) (%) (%) Gs (%) e (Mg/m3) (Mg/m3)
Suffield clay 31 0 35 20 2.69 98.5 1.045 1.312 1.79
Bat-Baioss clayb 30 10 51.2 23.6 2.73 ? 0.59 1.72 2.06–2.15
hte
Hanover silt 2 18 — — 2.74 99.5 0.923 1.42 1.86

Ice-rich siltc 20 6 33–60 25–50 ? 100 — — 1.20–1.45
Callovian loamd 2 10 ? ? 2.70 — 0.64–0.79 1.50–1.65 —
Ottawa sand (20–30 mesh) — — 2.65 100 0.587 1.67 —
rig
Manchester fine 0 10 — — 2.67 98.7 0.770 1.51 —

sand
a
LL, PL, Atterberg limits; Gs, specific gravity of soil grains; Sr , degree of saturation; e, void ratio; rd, dry density; r, bulk density.
b
w = 20–24 %.
c
w = 50–150%.
py
d
w = 39–44%.
Normal Pressure Effect on Creep and Strength on time to failure (or strain rate) and temperature, as is
Co
expressed by
The effect of normal, or confining, pressure on the creep and
strength of a frozen soil can be taken into account in several t = c(t, q) + s tan f(t, q) (5.3-45)
ways (e.g., Ladanyi 1972a, 1981b, 1983; Klein 1979). For exam- For this case, the cohesion c is related to the uniaxial compres-
ple, under triaxial test conditions, cold, ice-rich soils, contain- sion strength, sfu , by
ing too little unfrozen water to consolidate under confining
pressure, tend to behave like weak rocks, showing failure enve- s fu (t , q)
lopes of a parabolic shape, similar to those in Fig. 5-26. c (t , q) = 1 (5.3-46)
2N f2
Although these envelopes can be described reasonably well by
second-degree parabolas, it is more customary in practice to where Nf is the flow value defined by
approximate them, at least on the compression side, by a set of
straight-line Coulomb envelopes. These envelopes are defined 1 + sin f(t , q)
Nf = (5.3-47)
by Coulomb parameters c and f, both of which may depend 1 - sin f(t , q)

l
ria
FIGURE 5-26 Schematic time dependence of failure enve-
lopes for frozen soils.
Source: Reproduced with permission from Ladanyi 1972a. Copyright 1972
Canadian Geotechnical Journal.
ate
and according to Eq. (5.3-11),
1
Ê e ˆ n
(5.3-48)
s fu (t , q) = qu = sc q Á 1 ˜
Ë e ¯c
As has been shown by Ladanyi (1972a), for an ice-rich, cold (ª

dM
below –5 °C) frozen soil, where both c and f are dependent on
temperature and strain rate, a good approximation of the
observed behavior, in terms of principal stresses, can be
obtained by writing
1
Ê e ˆ n (5.3-49)
(s1 - s3 )f = Á 1 ˜ ÈÎsc q + s3 (N fc - 1)˘˚
hte
Ë ec ¯
where Nfc equals the value of Nf for f = fc , which represents
the slope angle of a Coulomb envelope at ė 1 = ė c (i.e., at the
same reference rate which determines sc q ).
rig
■ EXAMPLE 5.3-4: Figure 5-27a to 5-27c show the variation of

peak triaxial compression strength with temperature and strain
rate for a frozen dense Toyoura sand (e = 0.65) (Shibata et al.
1985). It can be shown that the experimental results can be
expressed by a general creep equation of the form of Eq. (5.3-49):
py
1
È Ê qˆ ˘ Ê e ˆ n
qmax = (s1 - s3 )f = Ísco Á 1 + ˜ + s3 (N fc - 1)˙ Á 1 ˜
Î Ë qo ¯ ˚ Ë ec ¯
(5.3-49a)
Co
with n = 16.28; sco = 5.04 MPa for ė c = 0.027%/min; qo = 8.96

°C; Nfc = 1.8114 (for fc = 16.8°), valid within the range of tem-
perature from 0 to –50 °C; strain rates between 0.027 and 2.7%/
min; and confining pressures from 0 to 9.8 MPa. Calculate the
value of qmax for T = –30 °C, s3 = 4.9 MPa, and ė 1 = 0.29%/min.
Solution:
1 FIGURE 5-27 Dependence of peak strength on temperature
Ê 0.29 ˆ È Ê
16.28
30 ˆ ˘ and confining pressure under different strain rate conditions:
qmax =Á Í5.04 ÁË 1 + ˜¯ + 4.9(0.8114)˙
Ë 0.027 ˜¯ Î 8.96 ˚ (a) ė 1 = 0.027%/min; (b) ė 1 = 0.29%/min; (c) ė 1 = 2.7%/
min.
= 29.956 MPa or 305.36 kgf/cm 2
Source: Reproduced with permission from Shibata et al. 1985. Copyright 1985
which compares well with Fig. 5-27b. A. A. Balkema.

■ EXAMPLE 5.3-5: Figure 5-28 shows an experimental rela- represents a set of straight lines with varying slope angles, fc , all
tionship between the uniaxial compressive strength and tempera- of them passing through the same point O ¢ at H = Hc (Fig. 5-
ture at different water contents for frozen Kiso River sand (Kurib- 29). An inverse form of Eq. (5.3-49) gives the corresponding
ayashi, Kawamura, and Yui 1985). It can be shown that these effect of normal pressure on creep rate:
experimental results can be empirically expressed by the equation
n
Ê s - s3 ˆ
qmax (kgf /cm2) = 0.158 w (%) (q + 15 °C) – 14.0 (5.3-49b) e1 = ec Á 1 (5.3-52)
Ë sc q ˜¯
(valid for 6 < w < 30% and 10 < q < 30 °C). Calculate the value
of sfu for T = –20 °C and w = 18.5%. with
Solution:
sc q = sc q + s3 (N fc - 1) (5.3-53)
l
qmax = 0.158 ¥ 18.5(20 + 15) - 14 = 88.3 kgf/cm2 In its integrated form, for a constant-stress difference, and for
ria
which corresponds well with Fig. 5-28, or in units of kPa, primary creep, the corresponding creep strain equation will be
n
Ê e ˆ Ê s - s3 ˆ
b
qmax = 15.5 ¥ 18.5(20 + 15) - 1, 373 = 8, 663 kPa (5.3-54)
e1 = Á c ˜ t b Á 1
Ë b ¯ Ë sc q ˜¯
In terms of t and s, Eq. (5.3-49) can be written as
ate
For frozen soils with large quantities of unfrozen water, or
t = (s + Hc )tan fc (5.3-50)
when consolidation is possible so that the confining pressure
where can be transferred to the soil skeleton, at least at failure, the
angle f may remain approximately constant, while only the
s fu cohesion will be affected by temperature and strain rate. In this
Hc = c cot fc =
dMcot fc (5.3-51)
2N 1fc/ 2 case, the shear strength can be approximated by the equation
(s1 - s3 )f = s fu (t, q) + s3 (N f - 1) (5.3-55)
or
t f = c (t, q) + s tan f (5.3-56)

hte
where f is constant, which is represented by a set of parallel

straight lines in the Mohr plot (Fig. 5-30). The corresponding
strain rate equation for such a soil will be
n
Ê s1 - N f s3 ˆ
e1 = ec Á (5.3-57)
rig
Ë sc q ˜¯
which is valid for s1 ≥ Nfs3.

In practical applications, the main difference between Eqs.
(5.3-49) and (5.3-55) is that, according to Eq. (5.3-49), when ė 1
py
Æ 0, (sl – s3)f Æ 0 and there is no true long-term strength.

This corresponds to the behavior of ice. According to Eq. (5.3-
Co
FIGURE 5-28 Relationship between uniaxial compressive FIGURE 5-29 Simplified failure envelopes for “hard frozen”
strength and temperature for a frozen sand. soils.
Source: Reproduced with permission from Kuribayashi, Kawamura, and Yui Source: Reproduced from Ladanyi 1983. Copyright 1983 American Society of
1985. Copyright 1985 A. A. Balkema. Civil Engineers.

FIGURE 5-30 Simplified failure envelopes for “plastic frozen”
l
soils.
ria
Source: Reproduced from Ladanyi 1983. Copyright 1983 American Society of
Civil Engineers.
55), when ė 1 Æ 0, the strength tends to a finite value, the long-
ate
term strength, thus
(s1 - s3 )1t = s3 (N f - 1) (5.3-58)
which is of a frictional character, as is expected in dense and

consolidated frozen soils.
dM
If in design, one would like to use a true three-dimensional
representation of creep failure in frozen soil, the validity of an
extended von Mises model must be assumed. This model is
represented by a family of cones in the principal stress space
with the same apex but with varying cone angles, similar to Fig. FIGURE 5-31 Relationship among volumetric unfrozen water
5-29. This can be done through the equation content, temperature, and salinity for a silty clay.
hte
n Source: Reproduced with permission from Patterson and Smith 1985. Copy-
Ês ˆ
ee = ec Á e ˜ (5.3-59) right 1985 Canadian Geotechnical Journal.
Ë sec ¯
where (according to Zienkiewicz and Cormeau 1974) From a large number of uniaxial compression tests on fro-
zen silts and silty clays, Nixon and Lem (1984) have established
2
rig
sec = (3cc cos fc + I1 sin fc ) (5.3-60) experimentally the dependence on salinity of the coefficient B
3 - sin fc in the steady-state creep equation
in which, cc and fc are the Coulomb parameters corresponding
e1 = Bs1n (5.3-62)
to the strain rate ė 1 = ėc , while I1 = s1 + s2 + s3 is the first
py
invariant of the principal stress tensor. The corresponding which is seen to be equivalent to Eq. (5.3-5) if
creep strength is then
ec
1 B= (5.3-63)
Ê e ˆ n
(5.3-61) snc q
sef = sec Á e ˜
Ë ec ¯
Co
where B is dependent on the salinity, S (in parts per thousand,

or ppt) and temperature q = – T (°C), as is shown in Fig. 5-32.
Salinity Effect on Frozen Soil Creep and Strength Adopting n = 3 (as suggested by Nixon and Lem 1984) and
selecting ė c = 10–4 yr–1 as a reference strain rate, one obtains for
The strength of frozen soils depends strongly on their ice con-
sc q the following empirical expressions, approximating the
tent. In saline soils, the volumetric ice content is a function of
experimental curves in Fig. 5-32:
both the pore water salinity and temperature. The influence of
salinity on the unfrozen water content of a silty clay is shown in sc q = 0.323 fc (S ) fc (q) (5.3-64)
Fig. 5-31. As for the effect of salinity on creep and strength of
frozen soils, this effect has been investigated by several authors with
in the past. The results of two such investigations involving var-
ious types of saline soils are shown below, and some of the 49.505 - S (5.3-65)
fc (S ) =
results are expressed in an empirical analytical form. 8.425 + S

with
18.78 + S (5.3-68)
fs (S ) =
5.52 + S
and
fs(q) = 3.52 + q (5.3-69)
where qu is in kPa, S in ppt, and q in °C. Equations (5.3-67) to
(5.3-69) approximate well the experimental curves in the tem-
perature range from –2 to –7 °C, and salinities from 0 to 38 ppt.
One may express the same results in the form of a creep
l
equation, such as Eq. (5.3-5), by taking n = 1.5 (Mahar and
ria
Stuckert 1985) and selecting as the reference strain rate ė c =
10–4 s–1, which gives
sc q (kPa) = 39.65 fs (S) – 225.12 (5.3-70)
ate
and is valid for strain rates below 10–5 s–1.
For frozen saline gravel, shown in Fig. 5-33c for ė c = 8.5 ×
10–4 s–1, qu can be approximated by the following expressions
(Ladanyi 1989):
qu (kPa) = fg(S) fg(q) – 169.3 (5.3-71)
dM
with
104 + S (5.3-72)
FIGURE 5-32 Relationship of the creep parameter B with
f g (S ) =
4 +S
salinity and temperature.
Sources: Reproduced with permission from Nixon and Lem 1984. Copyright and
hte
1984 Canadian Geotechnical Journal. See also Ladanyi 1989.

fg(q) = 1.77 + q (5.3-73)
For any other strain rate below 10–5 s–1, the value of qu can be
and obtained by substituting into Eq. (5.3-5), ė c = 10–4 s–1 and n = 3
(Mahar and Stuckert 1985), which gives
fc (q) = (1 + q)2 (5.3-66)
sc q (kPa) = 1.27 fg(S ) fg (q) – 82.96
rig
(5.3-74)
where sc q is the reference stress in kPa, valid for ė c = 10–4
yr–1, q = – T (°C), and S is the salinity in ppt. The curves More recently, Hivon and Sego (1995) reported a comprehen-
calculated by Eqs. (5.3-64) to (5.3-66) and shown in Fig. 5- sive study on the strength of frozen saline soils, including the
32 by dashed lines, approximate reasonably well the test combined influence of soil type, temperature, and salinity.
py
results in the temperature range from –3 to –13 °C and Their test results, based on unconfined constant strain rate tests
salinities from 0 to 35 ppt. The corresponding uniaxial on three different frozen soils, showed a predictive model for
compression strength as a function of strain rate can be strength in terms of salinity and temperature. The three soils, a
found by substituting sc q values from the equations above uniform fine sand (soil A), a silty sand (soil B), and a very fine
into Eq. (5.3-38). silty sand (soil C), were tested at three salinities (5, 10, and 30
Co
Mahar and Stuckert (1985) have studied the uniaxial com- ppt), four temperatures (–12, –10, –7, and –5 °C) and a con-
pression strengths of two saline soils: a subangular uniform stant compression strain rate of 2 × 10–6 s–1 (0.8% per hour).
sand (Monterey sand) and an Alaska gravel. The grain-size The three relationships include:
curves for these soils are shown in Fig. 5-33. All the compres-
sion tests were carried out at a strain rate of 8.5 × 10–4 s–1. The Soil A Rmax = 0.987(1 + S) –0.628 (5.3-75)
compressive strength of the two soils as functions of tempera- Soil B R10% = 0.936 – 0.0254S (5.3-76)
ture and salinity are shown in Fig. 5-33b and 5-33c. On the
basis of these results, the following empirical expressions for Soil C R10% = 1.027 – 0.0254S (5.3-77)
the uniaxial compression strength, qu , can be established
(Ladanyi 1989). For frozen saline sand, where Rmax is the peak strength ratio, R10% the ratio of strengths
at 10% strain, and S the salinity in ppt for a strain rate of 2 ×
qu = 165.2 fs(S) fs(q) – 938 (5.3-67) 10–6 s–1.

l
ria
ate
dM
hte
rig
py
Co
FIGURE 5-33 Effect of temperature and salinity on uniaxial compression strength of two granular soils: (a) grain-size distri-
bution curves; (b) test results with frozen sand; (c) test results with frozen gravel.
Source: Reproduced from Mahar and Stuckert 1985. Copyright 1985 American Society of Civil Engineers.

5.4 Frozen Soil Behavior in Uniaxial Tension TABLE 5-5 Uniaxial Tensile Strength Parameters for Various
Soils
In comparison with the published data on compression testing Dry Water
of frozen soils, the amount of data on tensile testing is much density content Temperature ␴c ␪b
more limited, but some valuable information has become avail- Soil type (Mg/m3) (%) (°C) n (MPa)
able in recent years. Quite generally, in uniaxial tension tests Silta 1.26 36.9 –5.0 10.4 0.862
the behavior of frozen soil is more brittle, but at the same time Sandy silt 1.30 34.9 –6.7 11.9 1.23
it is less sensitive to temperature and strain rate as compared to
Fine sand 1.57 24.7 –6.7 6.37 0.982
uniaxial compression tests under similar conditions (Haynes,
Karalius, and Kalafut 1975; Bragg and Andersland 1982). Gravelly sand 1.75 19.2 –6.7 5.31 0.114
An extensive series of uniaxial tensile tests on ice-saturated Gravelly sand 1.75 19.2 –1.1 3.81 0.049
Fairbanks silt (Zhu and Carbee 1984, 1987), with an average dry
l
a
After Zhu and Carbee 1987.
density of 1.26 Mg/m3 and temperature of –5 °C can be summa- For e· c = 10–5 h–1.
ria
b
rized as follows: Below and up to the ductile-brittle transition Source: After Sayles 1991.
point, which occurred at a strain rate of about 10–2 s–1, there was
a steady increase of both tensile strength and failure strain with
Based on the results of cyclic compression tests on 200-mm
increasing strain rate; beyond the transition point, the strength
ate
cubes of three different frozen soils, Tsytovich (1975) found
showed a slight decrease with increasing strain rate.
that under a pressure of 200 kPa, the variation of Young’s mod-
Comparison of uniaxial tensile and compression strengths
ulus E with temperature could be represented by the following
with strain rate showed that they remained about the same up
empirical equations (Johnston 1981):
to the plastic-brittle transition point, beyond which the com-
pression strength continued to increase while the tensile 1. For frozen sand (grain size between 0.05 and 0.25 mm, and
strength started to decrease because of the increasing brittleness total moisture content of 17–19%) at temperatures down
dM
of the soil at high strain rates. The failure strain in tension was to –10 °C,
found to be about one order of magnitude smaller than in
E = 500(1 + 4.2 q) (5.5-1)
compression, within the strain rates covered in the tests. The
initial tangent modulus, E i , of the frozen silt at –5 °C was 2. For frozen silt (grain size between 0.005 and 0.05 mm, and
observed to be relatively insensitive to strain rate, with its value water content of 26–29%) at temperatures down to –5 °C,
varying between 900 and 1,400 MPa. E = 400(1 + 3.5 q) (5.5-2)
Information on the tensile behavior of a sandy silt, a fine
hte
3. For frozen clay (with 50+% passing the 0.005-mm sieve)

sand, and a gravelly sand, at temperatures of –1.1 and –6.7 °C
and strain rates between 10–1 and 10–5 h–1, has been made avail- and a water content of 46–56% at temperatures down to –5
able by Sayles (1991). The peak strengths, plotted in a log–log °C,
plot, showed a linear increase with applied stress, indicating E = 500(1 + 0.46 q) (5.5-3)
that the data could be represented by a power law creep equa- where E is Young’s modulus in MPa and q is the number of
rig
tion such as Eq. (5.3-38) or (5.3-48). Denoting the peak tensile °C below 0 °C.
strength by st gives
When these values of E are compared with those for ice
1
Ê e ˆ n
(5.4-1)
obtained under similar conditions, it is observed that the mod-
st = sc q Á 1 ˜ ulus for ice is smaller than that of dense frozen sand and silt but
Ë ec ¯
py
is much larger than that of clay, due to the large amount of

The parameters n and sc q (MPa) for ė c = 10–5 h–1 calculated unfrozen water in the latter. Poisson’s ratio for the three frozen
from the data supplied by Sayles (1991), are given in Table 5-5. soil types was found to decrease with decreasing temperatures
until essentially all the pore water is frozen and the soil
Co
becomes rigid.
5.5 Deformability of Frozen Soils Additional triaxial test data (Zhu and Carbee 1984) for fro-
zen Fairbanks silt are summarized in Fig. 5-34a to 5-34c. These
The total deformation of a frozen soil under load is composed data (Fig. 5-34a and b) show the dependence of the initial tan-
of an instantaneous portion and a time-dependent portion, gent modulus, E50 (at 50% of peak strength) on temperature for
both of which contain a reversible and an irreversible compo- six different strain rates. The variation of E50 with strain rate
nent. The reversible component of instantaneous deformation and dry density is presented in Fig. 5-34c.
determines the response of frozen soils to very-short-term and Another set of data (Kuribayashi, Kawamura, and Yui 1985),
dynamic loading. In general, the deformation parameters, such summarizing uniaxial compression tests on frozen fine sand at
as Young’s modulus, E, and Poisson’s ratio, m, depend on soil temperatures between –10 and –30 °C, shows very good corre-
type, its temperature, and type of test. Some data available in lation between the deformation modulus E1.5 (at 1.5% uniaxial
the literature on deformability for short-term static loads and strain) and the uniaxial compression strength, qu . These results
various frozen soils will be presented. can be expressed by

l
ria
ate
dM
hte
rig
py
FIGURE 5-34 Deformation moduli for frozen silt, under various temperature, strain-rate, and density conditions: (a) initial tangent
Co
modulus versus temperature ratio for medium-density samples at various strain rates; (b) deformation modulus E50 versus tem-
perature ratio for medium-density samples at various strain rates; (c) deformation modulus E50 versus strain rates for three differ-
ent dry density ranges at –2 °C.
Source: Reproduced with permission from Zhu and Carbee 1984. Copyright 1984 Elsevier.
E1.5 ª 60qu (5.5-4) modulus Ei (at deviatoric strains of less than 0.2%) and the
uniaxial compression strength, qu :
Finally, Shibata et al. (1985) showed the results of a series of tri-
axial compression tests with a frozen sand at temperatures Ei ª 129.5qu (5.5-5)
between –2 and –50 °C, at confining pressures between 0 and
10 MPa, and at strain rates between 0.027 and 2.7%/min. From They observed that Poisson’s ratio decreased from about 0.2, at
a correlation involving a large number of test points, they temperatures close to the freezing point, down to about 0.1, at
found the following relationship between the initial elastic temperatures lower than –30 °C.

5.6 Compressibility of Frozen Soils warming are dependent on the gradual increase in permafrost
temperatures which will increase the active layer thickness
Frozen soils are usually considered to be practically incom- (annual thaw depth) and cause degradation of the permafrost.
pressible, and therefore volume-change deformations are often The gradual warming of permafrost with a concurrent decrease
neglected compared with creep deformations. Investigations in frozen soil strength leads eventually to the thawed state. The
conducted on various types of frozen soils at different freezing strength of rocks and dense gravels depends primarily on min-
temperatures show that the compressibility of frozen soils can eral bonds and internal friction with little dependence on tem-
play a significant role in some cases, especially when large areas perature. In frozen fine-grained soils, a major portion of the
carry long-terms loads (Brodskaia 1962; Tsytovich 1975). soil strength is due to ice bonding. On warming, the ice bond-
Compressibility and its time dependence in frozen soils have ing will decrease as ice is converted to water. In Arctic regions
several causes, such as the instantaneous compression of the with thaw sensitive permafrost, continued climatic warming
will result in significant settlement of structures (Esch and
l
gaseous phase; the creep of ice cement at the grain contacts;
ria
and hydrodynamic consolidation due to the expulsion under Osterkamp 1990). There is a need to estimate this potential set-
stress of unfrozen water, the amount of which varies with pres- tlement and damage to existing structures. This section will
sure (Johnston 1981; Arteau 1984). According to Brodskaia show how the decrease of frozen ground strength with temper-
(1962), the shape of the oedometer compression curve varies ature increase can be quantified in terms of a strength-sensitiv-
with the type and temperature of frozen soil. Usually, with soils ity index. This index can be used to develop design guidelines
ate
containing large amounts of unfrozen water, its shape is not for future projects (Ladanyi 1996).
very different from that obtained for the same soils when
unfrozen, at the same total water content. At lower tempera- Frozen Soil Strength Sensitivity to Temperature Change
tures, however, the effect of ice cementation becomes more
pronounced, and the behavior under compression becomes In addition to temperature, frozen soil strength depends on soil
more similar to that usually obtained for unfrozen cemented or density, ice content, and salinity. This strength also is affected
dM
preconsolidated soils (Johnston 1981). by the degree of confinement and the applied strain rate. For a
Typical data for the coefficient of total volume compressibil- general strength comparison, the uniaxial compressive strength
ity, mv , obtained in oedometer compression tests on various is the most useful. The strain rate and temperature dependence
soils at different temperatures by Brodskaia (1962) and Tsytov- (Ladanyi 1996) of frozen soil strength can be expressed by
ich (1975) are summarized in Table 5-6. qf = qf 0 ◊ f (q) (5.7-1)
with
hte
5.7 Frozen Ground Sensitivity to Climate Change qf 0 = sco · f (ė,t) (5.7-2)

Physical and mechanical properties of permafrost are tempera- The term f(q) expresses the effect of temperature on strength
ture dependent. Most engineering concerns related to climatic and creep. The most frequently used empirical form is
rig
TABLE 5-6 Coefficient of Volume Change, mv , for Several Soils
Total Unfrozen Bulk Soil Coefficient of volume change, mv (cm2/kgf × 10–4), for
moisture water content density temperature pressure intervals (kgf/cm2) of:
(kg/m3)
py
Soil type content (%) (%) (°C) 0–1 1–2 2–4 4–6 6–8
Medium-grained sand 21 0.2 1,990 –0.6 12 9 6 4 3
27 0.0 1,870 –4.2 17 13 10 7 5
27 0.2 1,860 –0.4 32 26 14 8 5
Co
Silty sand 25 5.2 1,900 –3.5 6 14 18 22 23

with massive texture 27 8.0 1,880 –0.4 24 29 26 18 14
Medium silty clay 35 12.3 1,830 –4.0 8 15 26 28 24
with massive texture 32 17.7 1,840 –0.4 36 42 37 21 14
with reticular texture 42 11.6 1,710 –3.8 5 10 18 42 32
38 16.1 — –0.4 56 59 39 24 16
with stratified texture 104 11.6 1,360 –3.6 54 54 59 44 34
92 16.1 1,430 –0.4 191 137 74 36 18
Varved clay 36 12.9 1,840 –3.6 15 22 26 23 19
34 27.0 1,870 –0.4 32 30 25 20 16
Source: Tsytovich 1975.

w
Ê qˆ (5.7-3)
f (q) = Á 1 + ˜
Ë qo¯
while f (ė,t) expresses the effect of strain rate and time on

strength and creep. Its time hardening (primary creep) form is
1 (b -1)/ n
Ê e ˆ n Ê b ˆ
f (e ,t ) = Á 1 ˜ Á ˜ (5.7-4)
Ë ec ¯ Ë ec t ¯
which, for steady-state creep (b = 1) reduces to FIGURE 5-35 Schematic frozen soil strength variation with
temperature change.
l
1
Ê e ˆ n
ria
f (e ,t ) = Á 1 ˜ (5.7-5)
Ë e ¯
c For example, if w = 1 and n = 3, a temperature increase of 1 °C,
from q2 = 2 °C to q1 = 1 °C, (or from T = –2 to T = –1 °C) will
In the above equations:
increase creep strains and creep strain rate by a factor 3.375.
q= As is shown in Fig. 5-35, when temperature increases by DTi
ate
–T = frost temperature, °C
qo = 1 °C = –Dqi , the strength will drop by Dqf i . The strength drop
t= time, s depends on the strength variation gradient (Dq/Dq)i and on the
ė 1 = uniaxial compression strain rate, s–1 temperature change Dq = – DTi . For f (q) given by Eq. (5.7-3),
ė c = reference strain rate, s–1 the gradient of strength change at qf = qf i is equal to
sco = reference stress (MPa) at ė c and when T tends to 0 °C
qf =
dM
uniaxial compression strength, MPa w -1
Ê dq f ˆ Ê df (q) ˆ Ê w ˆ Ê 1+ qˆ (5.7-10)
n= creep exponent for stress Á dq ˜ = q fo ÁË dq ˜¯ = q fo ÁË q ˜¯ ÁË q ˜¯
Ë ¯i o o
b= creep exponent for time
w£1= temperature exponent which can also be written as
Equation (5.7-1) combined with Eqs. (5.7-2) and (5.7-5) gives the
usual temperature- and strain-rate-dependence of strength. For a Ê dq f ˆ Ê f (q) ˆ
Á dq ˜ = q fo ◊ w ÁË q + q ˜¯ (5.7-11)
hte
given soil, the effect of temperature alone (at the same strain rate) Ë ¯i o
can be expressed by combining Eqs. (5.7-1) and (5.7-3),
It is proposed that the strength sensitivity index, ST , for frozen
w
Ê qˆ (5.7-6)
soil be defined by the ratio
q f = q fo Á 1 + ˜
Ë qo ¯ Dq f i
ST = (5.7-12)
rig
The typical strength variation of frozen soil with temperature, qf i

expressed by Eq. (5.7-6), is shown schematically in Figure 5-35.
Alternatively, to evaluate the effect of temperature on creep where
rate, Eqs. (5.3-17), (5.3-20), and (5.3-23) are used to derive the qfi = strength at temperature qi , and
following equations: Dqfi = its variation due to temperature change.
py
n From Eq. (5.7-1) write

For creep Ê q ˆ Ê ec t ˆ b (5.7-7)
e1 = Á Á ˜
Ë sco f (q) ˜¯ Ë b ¯ qfi = qfo · f (qi ) (5.7-13)
Co
n and, from Eq. (5.7-11)

Ê q ˆ Ê ec t ˆ b -1
For creep rate e1 = ec Á Á ˜
(5.7-8)
Ë sco f (q) ˜¯ Ë b ¯ Ê dq f ˆ Ê f (qi ) ˆ
Dq fi = Á ˜ Dqi = q fo w ÁË q + q ˜¯ (5.7-14)
where q is the applied uniaxial compression stress and ė 1 the Ë d q ¯i i o
resulting creep strain. With other terms constant, Eqs. (5.7-7)

Using Eq. (5.7-12) gives the strength sensitivity index
and (5.7-8) show that the effect of temperature change on creep
and creep rate follows a law of the form [sco · f (q)]–n. Hence,
Ê Dqi ˆ
for two temperatures q1 and q2, the corresponding creep and ST = w Á (5.7-15)
creep rate ratios will be Ë qi + qo ˜¯
-n -wn The index ST is shown to be a function of the initial permafrost

e1,q1 e1,q1 È f (q1 ) ˘ Ê q1 + qo ˆ
= =Í ˙ = (5.7-9) temperature, qi , the temperature change, Dqi , and the frozen
e1,q2 e1,q2 Î f (q2 ) ˚ ÁË q + q ¯˜
2 o soil sensitivity to temperature change expressed in terms of w.

For example, an ice-rich silt (w = 0.60) at –7 °C will, for a tem- where ė m is the minimum creep or strain rate (= ė 1,min), tm the
perature increase of 1 °C, undergo a strength loss of ST = time to attain ėm , and C a constant, usually taken equal to the
0.60[1/(7 +1)] = 7.5 %, while the same soil at a temperature of failure strain in compression. The exponent p is usually found
–3 °C, would lose ST = 0.60[1/(3 + 1)] = 15% of its original experimentally to be slightly higher than 1, in the range of 1.01
strength. to about 1.40, increasing with soil density. The effect of tem-
The above-defined strength sensitivity index, ST , can be a perature on the uniaxial (unconfined) compression strength,
useful measure for evaluating the strength loss of frozen soils in sf u , is most often expressed by an empirical relationship of the
permafrost regions where climatic warming is not expected to form
result in complete permafrost thawing. The index requires a
sfu = so · f (q) (5.8-2)
knowledge of temperature sensitivity of the strength and creep
of some typical Arctic soils. Although some such information where f(q) is given by
already exists (Table 5-3 and Section 5.8), additional laboratory
l
m
ria
and field testing of permafrost soils is needed. Eventually, com- Ê qˆ
f (q) = Á ˜ (5.8-3)
bining information on permafrost occurrence and soil types Ë qo ¯
and characteristics with the climate warming trends, it may be
possible to prepare climate warming effect maps that would or by Eq. (5.3-20)
show not only trends to active layer increase and permafrost
ate
w
thawing, but also the reduction of permafrost strength due to Ê qˆ
f (q) = Á 1 + ˜ (5.8-4)
warming. Permafrost sensitivity maps of this kind may become Ë qo ¯
a useful basis for predicting the effects of climate warming for
existing constructed facilities in the Arctic, and for developing and so is a function of strain rate. In this creep and strength
guidelines for the design of new facilities. Vyalov et al. (1993), data representation, so in Eq. (5.8-2) would, according to Eqs.
proposed establishing a permafrost sensitivity zonation in the (5.3-17) and (5.3-38), be given by
dM
Arctic, based on the average annual ground temperature of the
1
permafrost. Ê e1,min ˆ n
(5.8-5)
so = sco Á
Ë e ˜¯ c
5.8 Frozen Soil Creep and Strength: In Eqs. (5.8-2) to (5.8-4), q is the number of °C below the freez-
Generalization of Test Data ing point of water, while m and w are experimental temperature
hte
exponents, which, for a given soil, may be constant or tempera-

Experimental work reported for different frozen soil types by ture dependent. In contrast, some authors find it more conve-
various authors shows a general consensus on the generalization nient to involve temperature implicitly in the strength equation
of test results. Mellor and Cole (1982) proposed that the relation- (e.g., Orth 1986, 1988). In connection with creep tests, a very
ship between minimum strain rate and stress, observed in stress- useful stress measure was found to be the creep stress ratio,
controlled creep tests (Fig. 5-36) and strain-rate-controlled com- defined as the ratio between the applied stress and the short-
rig
pression tests was unique. On this basis, most authors consider as term peak strength, obtained at a given rapid strain rate (e.g.,
valid the relationship (equivalent to Eq. 5.3-11): Martin, Ting, and Ladd 1981; Wijeewera and Joshi 1990, 1992).
As for deformations, in addition to finding temperature-
emt mp = C (5.8-1) dependent tangent- and secant-deformation moduli, most
authors use for creep displacement calculations a primary creep
py
formulation (Eq. 5.3-16) with a time function expressed by a

power law, or alternatively by an exponential function of time,
the latter sometimes tending to cover tertiary creep (Ting 1983;
Zhu and Carbee 1987).
Co
In the following six examples, an attempt is made to present,

in a standard form, the results of a limited number of experi-
mental studies on the mechanical behavior of various frozen
soils, published since 1980. The results are presented in a stand-
ard form, giving only essential data on the soil (density, water
content, degree of saturation), and on test conditions (temper-
ature, strain rate, creep loads and confinement). Each example
summarizes the main findings relevant to their use in perma-
frost engineering and in ground freezing design.
All symbols used in the examples are defined in Appendix A.
FIGURE 5-36 Definition of tm, em, and ė m from creep tests.
Source: Reproduced with permission from Orth 1988. Copyright 1988 A. A. ■ EXAMPLE 5.8-1: From the data given below, compute the
Balkema. uniaxial compression strength sfu of frozen Fairbanks silt at a

temperature of –5 °C and a strain rate of 10–5 s–1 (Source: Zhu Findings: Ductile to brittle transition: ė 0 = 3 × 10–4 s–1.
and Carbee 1983, 1984, 1987). Peak strength: For ė m < ė 0, sfu = s0 (ė/ė 0)m, with: so = 3.8
Soil: Fairbanks silt (q/q0)0.466 (MPa), q0 = 1 °C, and m = 0.204 + 0.0066 q.
Grain size: >80% between 0.001 and 0.1 mm For ė m ≥ ė 0, sfu = s0.
Density: 1.08, 1.20, 1.40 g/cm3 Solution: Compute:
Water content: 28–43% so = 3.8(5/1)0.466 = 8.045 MPa
S r : > 95%
T: –0.5, –1.0, –2.0, –3.0, –5.0, –7.0, –10.0 °C Then
Strain rate: 6.2 × 10–2 to 1.1 × 10–6 s–1 m = 0.204 + 0.0066 × 5 = 0.237
Confinement: 0
giving
Findings: The failure criterion is defined by Eq. (5.8-1),
with: C ª em = f(rd), varying from 0.01 at low rd to 0.18 at high
l
0.237
Ê 10-5 ˆ
ria
rd . The exponent p in Eq. (5.8-1) varies from 1.01 at low rd to s fu = 8.045 Á = 3.593 MPa
1.16 at high rd . Ë 3 ¥ 10-4 ˜¯
Ductile to brittle transition occurs at: ė 0 = 1.1 × 10–3 s–1. which corresponds well to the results of measurements, pub-
Peak compression strength, Eqs. (5.8-2) and (5.8-3), with qo lished by Zhu, Zhang, and Shen (1988).
= 1 °C, and m = 0.49 + 0.0517 ln(ė 0/ė), for: 10–3 ė 0 £ ė £ ė 0 ,
ate
and m = 0.49 for ė ≥ ė 0. ■ EXAMPLE 5.8-3: From the data given below, compute the
For medium-density silt, s0 can be computed from: uniaxial compression strength of frozen Karlsruhe medium
sand at –10 °C, and a strain rate of 0.001 min–1 (Source: Orth
Ao
so = 1986, 1988).
Ê ˆ Ê e ˆ
A Soil: Karlsruhe medium sand
1 + Á o ˜ ln Á o ˜
Ë a ¯ Ë e ¯
dM Grain size: 80% between 0.2 and 2.0 mm
Where ė 0 = 1.1 × 10–3 s–1 , and Ao = 1.87 MPa; a = 3.92 MPa , Density: Dense (rd = 1.753 g/cm3)
when 10–3 ė 0 £ ė £ ė 0 , and a = 8.43 MPa, when ė ≥ ė 0. S r : Saturated
Solution: From Eq. (5.8-2): sfu = so f (q) with f (q) given by T: –2, –10, –15, –20 °C
Eq. (5.8-3). Compute: Creep load: 5, 6, 10, 11, and 12 MPa
Strain rate: 0.001–1.0%/min
Ê 1.1 ¥ 10-3 ˆ Confinement: 0
hte
m = 0.49 + 0.0517 ln Á = 0.733

Ë 10-5 ˜¯ Findings: Failure criterion: use Eq. (5.8-1), with p = 1 and
C = em = 0.024
Then
Peak strength can be expressed by the equation (Fig. 5-37):
f (q) = (5/1)0.733 = 3.253
rig
1.87
so = = 0.577 MPa
Ê 1.87 ˆ Ê 1.1 ¥ 10-3 ˆ
1+ Á ln
Ë 3.92 ˜¯ ÁË 10-5 ˜¯
giving
py
sfu = 0.577 × 3.253 = 1.876 MPa

which corresponds well to the results of measurements, pub-
lished by Zhu and Carbee (1984).
Co
■ EXAMPLE 5.8-2: Using the data given below, compute the

uniaxial compression strength sm of frozen Lanzhou medium
sand at –5 °C and at a strain rate of 10–5 s–1 (Source: Zhu,
Zhang, and Shen 1988).
Soil: Lanzhou medium sand
Grain size: > 80% between 0.05 and 1.0 mm
Density: Dense (1.80 g/cm3)
Water content: 14%
S r : 93% FIGURE 5-37 Uniaxial compression stress s1 (=sfu) versus ė m
T: –2, –5, –10, –15 °C from creep tests.
Strain rate: 5 × 10–7 to 6 × 10–4 s–1 Source: Reproduced with permission from Orth 1988. Copyright 1988 A. A.
Confinement: 0 Balkema.

Soil: Watarase River sand and Toyoura sand

È Ê e ˆ ˘
Í ln Á m ˜ ˙ Density: Variable (1.335 to 1.645 g/cm3)
Ë 1 ¯
s fu = 3.05 q0.591 Í1 + ˙ Water content: Variable
Í Ê 3817 ˆ˙ S r : Variable (15–100%)
ÍÎ ÁË 273.4 - q ˜¯ ˙˚
T: –30 °C
Loading rate: 200–300 kPa/sec
with:
Confinement: 0
sfu in MPa and ė m in %/min. Findings: Peak strength: Empirical relationships of the form:
Solution: The value of sm can be computed from the above m

Ê w ˆ Ê rd ˆ
n
s fu = qx Á
equation: Ë 100 ˜¯ ÁË rw ˜¯
l
È Ê 0.1 ˆ ˘
ria
ln Á ˜ ˙ For Watarase River sand: qx = 22.5 MPa, m = 1.06, n = 3.49.
Í Ë 1 ¯ For Toyoura sand: qx = 26.9 MPa, m = 0.852, n = 2.95.
sm = 3.05 ¥ 100.591 Í1 + ˙ = 10.0 MPa
Í Ê 3 , 817 ˆ ˙ Solution: According to the above equation,
ÍÎ ÁË 273.4 - 10 ˜¯ ˙˚
1.06 1.06
Ê w ˆ Ê w ˆ
s fu = 26.9 Á ¥ 1.53.49 = 110.74 Á
ate
which corresponds well to the data shown in Fig. 5-37. Ë 100 ˜¯ Ë 100 ˜¯
■ EXAMPLE 5.8-4: Source: Kuribayashi, Kawamura, and Yui This gives for
1985 w = 10%: sfu = 9.654 MPa
Soil: Kiso River sand
Grain size: >80% between 0.1 and 1.0 mm w = 15%: sfu = 14.82 MPa
dM
Density: Dense (compacted) w = 20%: sfu = 20.11 MPa
Water content: 6 different water contents between 3% and
30 % which corresponds well to the data shown in Fig. 5 in Enokido
S r : 4–90% and Kameta 1987.
T: –10, –18, –25, –30 °C
Strain rate: 0.1, 1.0, 6.0, 10.0%/min
Confinement: 0 Problems
hte
Findings: Failure strain: em = 5–7% at a strain rate of 0.1%/

5.1 The results of a series of uniaxial compression creep tests
min, em = 2% at a strain rate > 1 %/min.
Peak strengths, shown in Fig. 5-28, can be expressed approx- on a frozen silty sand have been expressed by a general equation
imately by Eq. (5-3.49b). [such as Eq. (5.3-3) to (5.3-5)], giving the total uniaxial strain e
Modulus of deformation at e = 1.5%: E1.5 ª 60 sm . as a function of applied stress s and time t:
rig
(For comparison with measurements, see Example 5.3-5 1.5 3

and Fig. 5-28.) Ê s ˆ Ê s ˆ
e1 = 10-2 Á 1 ˜ + 10-6 Á 1 ˜ t
Ë 0.05 ¯ Ë 0.25 ¯
■ EXAMPLE 5.8-5: Source: Shibata et al. 1985.
with s1 in MPa and t in minutes.
Soil: Toyoura sand
py
From this general equation, compute the axial strain for a

Density: Dense (e = 0.65)
similar frozen soil specimen if it is loaded up to 2.0 MPa in four
S r : Saturated
stages (i.e., 0.50, 1.00, 1.50, and 2.00 MPa), each load for a
T: –2, –10, –30, –50 °C
duration of 2 h.
Strain rate: 0.027, 0.29, 2.7%/min
Co
Confinement: 0, 50, 100 kgf/cm2 5.2 For a frozen soil, the creep equation in terms of equivalent
Findings: Approximately linear increase of qmax with q. stresses and strains is given by
Logarithmic relationship between qmax and em.
Power-law relationship between qmax and s3. ee = 3 ¥ 10-4 se2.8
The relationship qmax ∫ (s1 – s3)f = f(q,s3), can be expressed
with se in MPa and time in minutes. Using a reference strain
approximately by Eq. (5.3-49a).
rate ė c = 10–5 min–1, write the creep equation for the same soil
(For comparison with measured data, see Example 5.3-4
when it is submitted to a simple shear test (i.e., in terms of ġ
and Fig. 5-27.)
and t).
■ EXAMPLE 5.8-6: For frozen Toyoura sand, compute the 5.3 From a series of uniaxial compression creep tests, con-
values of the uniaxial compression strength sm at the density of ducted on a frozen silt at constant temperature and confining
rd = 1.5 g/cm3, and at water contents of 10, 15, and 20%, pressure, it was observed that its compressive strength varied
respectively. (Source: Enokido and Kameta 1987). with time according to the equation

(s1 - s3 )tr = 10t -f 1/ 3 Ê qˆ

w
sc q = sco Á 1 + ˜
Ë qc ¯
with stresses in MPa and time in minutes. Transform this equa-
tion into an equivalent one valid for plane strain conditions, (c) Estimate the long-term (10-year) unconfined compressive
assuming that the creep exponent b = 1 and that the failure strength for this Wedron sand at a temperature of –10 °C.
strain remains the same for the two cases. Assume that failure occurs at 4% axial strain (ef = 0.04).
5.4 Data from a series of unconfined compression tests on fro- 5.5 Uniaxial compression tests on a saturated frozen sand have
zen Wedron sand (size range of 105 to 595 mm) are summa- provided the following information: failure strain ef = 4%,
rized in Fig. 5-38. Soil data include a density of 1,700 kg/m3 creep parameter n = 12.5, temperature parameter w = 0.63,
and a water content of 20%. The sample size was 51 mm in friction angle f = 30 degrees, and proof stress (at 0 °C) sco = 7.2
MPa for a reference strain rate ė c = 10–7 s–1.
l
diameter and 127 mm in height.
ria
Determine the maximum acceptable value for the major
(a) For the –10 °C temperature, determine sc q, ėc , and n in the principal stress at failure slf with s3f = 90 kPa, T = –15 °C, and a
equation service life t = 30 days.
1 5.6 Estimate the percentage loss of strength sfu due to soil
Ê ef ˆ n
ate
s fu = sc q Á temperature increase from –5 °C to –3 °C, at a strain rate of
˜
Ë ec t f ¯ 10–5 s–1, for
(a) Fairbanks silt, described in Example 5.8-1, and
(b) Determine sco (proof stress at T = 0 °C) and the parameter
w in the equation (b) Lanzhou medium sand, described in Example 5.8-2.
dM
hte
rig
py
Co
FIGURE 5-38 Variation of compressive strength with strain rate at –6, –10, and –15 °C.

Mechanical Properties of Frozen Soils: 5.1 Stress-Strain-Time and Strength Behavior

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Mechanical Properties of Frozen Soils: 5.1 Stress-Strain-Time and Strength Behavior

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mechanical Properties of Frozen Soils: 5.1 Stress-Strain-Time and Strength Behavior

Uploaded by

Copyright:

Available Formats

5

stern 1973). The amount of unfrozen water present in frozen Dp ÊCˆ

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

change. preventing the collapse of the soil skeleton. On the basis of

stress regions. Theoretically, Eq. (5.1-4) predicts that a pressure

factor of 10 to 500 (Arteau 1984). Finally, if the confining pres-

(1972); Simonsen, Jones, and Green (1974); and Jones and

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

Canadian Geotechnical Journal.

processing experimental data, in order to put the information

Source: Reproduced with permission from Andersland, Sayles, and Ladanyi

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

failure strain is affected by strain rate, temperature, and confin-

brittle transition. Because of increasing brittleness of response,

sion creep of frozen soils, the criterion was first mentioned by

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

rate, its response to loading is found to vary from viscous to

shown by Kaplar (1971) and T. H. W. Baker (1979) in Fig. 5-6,

axial strain and another at about 10% or more.

ior observed in compression tests at a constant strain rate and

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

increases to 10–3 s–1 (Sayles 1973). Because the transition

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

FIGURE 5-14 Compressive strength versus strain rate rela-

tionship for a frozen sand at different temperatures.

FIGURE 5-15 Comparison of creep strengths of frozen sand

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

Mascarelli, and Desvarreux 1967). The following brief review

FIGURE 5-16 Failure strain versus strain-rate relationship for a

Source: Reproduced with permission of P. E. Frivik from Bragg and Andersland

FIGURE 5-17 Peak and post-peak strengths of a frozen sand

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

match between sand grains and pore ice.

Frozen Clay. A series of uniaxial compression tests (Bourbon-

water content of about 25%. The tests were performed at tem-

increase was exponential down to –100 °C, and became more

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

mulated strain). When frozen soil deforms, its structure

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

In the latter case, for the portion of creep curves at and

where ė (c)min = de(c)/dt is the minimum (or steady-state) creep

where, in Eq. (5.3-4), sk q is a temperature-dependent total

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

FIGURE 5-23 Determination of parameters in Eqs. (5.3-4) and (5.3-5).

ee2 = 23 eij eij = 43 I2¢

denotes time rate.

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

A convenient form of the primary (or transient) creep law is D log(sc q )

validity of von Mises flow rule. For example, Ladanyi and

where qo is the intercept on the q-axis of the straight line pass-

ence strain rate, ė c, and to a soil temperature q (°C) = –T (°C). case is

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com

with stresses in MPa and time in hours.

where 0.01, the creep strength equation becomes

The slope of this descending line in Fig. (5-24) gives ÈÎe(cc ) ˘˚

Copyright © 2004 John Wiley & Sons Retrieved from: www.knovel.com