The International Information Center for Geotechnical Engineers

Ground Freezing




The following sections describe the effects of AFG on the engineering properties of soils, namely hydraulic conductivity, stiffness, shear strength, and volume change capacity.  In addition, laboratory testing and classification of frozen soils is introduced per JGS and ASTM standards.




When applied to civil engineering projects for containment or control of groundwater, frozen soil is practically impermeable.  Ice fractures also have the propensity to heal themselves by refreezing.  Problems with permeability arise when freezing procedures are not performed correctly, and soil does not freeze completely as one mass, leaving “windows” of unfrozen soil which could compromise the ability of the frozen barrier to contain and control groundwater or isolate a contaminant within the soil.  Windows of unfrozen soil are often found and sized using the ultrasonic measurement method (Jessberger 1980).




The strength behavior of frozen soils, as with any other soil, is dependent on a number of factors, including soil type, temperature, confining stress, relative density, and strain rate.  Frozen soils exhibit higher strength than unfrozen soils.  In general, frozen soil strength increases as temperature decreases and confining stress increases.  


Da Re et al. 2003 performed a study on the triaxial strength characteristics of frozen Manchester Fine Sand (MFS), in which specimens were prepared over a variety of relative densities (20 - 100%), confining stresses (0.1 - 10 MPa), strain rates (3 x 10-6 - 5 x 10-4 s-1), and temperatures (-2 to -25°C).  

 The results, shown graphically in Figure 1, show two distinct regions of strain over which the frozen soil acts differently.  Small strains (less than 1% axial) result in a linear strength increase which has a slope (modulus) independent of relative density or confining stress.  The magnitude of the initial yield point (at 0.5-1% axial strain in all cases) increases with increasing strain rate and decreasing temperature.  Large strain behaviors include strain softening, exhibited by specimens prepared at low relative density and under low confining stress, to strain hardening, exhibited by specimens prepared at high relative density and high confining stress.

Strength behavior of sands

Figure 1. Strength behavior of MFS (Da Re et al. 2003)


The strain softening behavior of the MFS shown in the Da Re et al. study is explained by Kornfield and Zubeck 2013.  They state that a decrease in stress past the initial yield point is due to an increase in crushing and pressure melting of the frozen pore water.  Yang et al. 2009 and Xu et al. 2011 also showed that as confining pressure increases, shear strength reaches a peak then decreases due to ice crushing and pressure melting.  Generally, at -10°C frozen sands and frozen clays have compressive strengths of 15 MPa and 3 MPa respectively (Klein 2012).


The compressive strength of frozen clay was analyzed by Li et al. under variable temperatures, strain rates, and dry densities.  The clay was compacted  to three different dry densities and had a liquid limit of 28.8 % and plastic limit of 17.7%.  Uniaxial compressions tests were performed at different temperatures (-2 to -15°C) and different strain rates (approximately 1 x 10-6 to 6 x 10-4 s-1) at each dry density.  Results of the study showed similar strength behavior to the study performed by Da Re et al. for frozen MFS.  The compressive strength of the clay tested increased with increasing strain rate, decreasing temperature, and increasing dry density, similar to the behavior of the MFS tested in the Da Re et al. study.  Additionally, frozen clays exhibited both strain hardening and strain softening behavior after an initial yield stress was reached, which was highly dependent on the time to failure, which itself depends on strain rate.  The results of the study showed that frozen clay samples loaded at low strain rates reached a low uniaxial compressive strength (approximately 2 MPa defined at 10% strain if no failure was reached) at a longer time to failure but exhibited strain hardening behavior.  Conversely, frozen clay samples loaded at high strain rates reached a much higher uniaxial compressive strength (approximately 6 MPa at failure) but exhibited strain softening (Li et al. 2004).




 In general, frozen soils are stiffer than unfrozen soils.  Da Re et al., in their frozen soil strength study on MFS, conducted a study on Young’s Modulus.  They found that frozen MFS had a Young’s modulus of approximately 23 GPa to 30 GPa.  Because the small strain strength behavior of the frozen MFS was similar throughout the variables tested, Young’s modulus was independent of the variables tested (relative density, confining stress, strain rate, and temperature).  

Young's Modulus sands

Figure 2.  Normalized stress-strain behavior of MFS (Da Re et al. 2003)


Figure 2 from the Da Re et. al., 2003 study shows the independence of Young’s modulus of frozen sands, by normalizing the shear stress with the initial yield stress.  Figure 2 also shows the different volumetric strains due to the strain hardening or softening behavior of the frozen MFS post initial yield stress, as denoted by Type A, B, C, or D stress-strain behavior.




During a phase change from liquid to solid, water increases in volume by approximately 9%, which translates to soil heave at the ground surface (Lackner et al. 2005).  Heave due to volume expansion may damage nearby structures (tunnels, surface structures) during freezing and thawing, therefore understanding the soil properties and how they contribute to soil heave is important during AGF.  Soil that has exhibited heave will also experience settlements upon thawing, which must be considered. Soil may also observe volume changes due to creep when loaded.  


Soil heave occurs in soils where ice lenses form within voids.  The soil structure must promote transfer of water from surrounding void spaces to the freezing front of the ice lense through capillary forces.  For this reason, silty soils are particularly frost-susceptible (Widianto et al. 2009).

It is also important to note that clays may exhibit low frost susceptibility in some cases.  As the freezing front moves outward, clays exhibit heave due to volume expansion of the ice lense, however consolidation may occur ahead of the freezing front, where negative pore pressures are being generated by the movement of water into the freezing zone.  The net effect of heave and consolidation below the ice lense may be small or negligible at the surface (Han and Goodings, 2006).  Despite this, site specific soils must be tested for frost-susceptibility if frost heave is expected to be an issue to nearby structures.




With respect to frozen soils, both ASTM and JGS have some standards for laboratory testing.  However, many of these tests either apply to pavements, multiple freeze-thaw cycles, or only provide information in the direction of heat flow.  JGS 0171-2003 is a test method for predicting the frost heave of a soil.  This standard utilizes Takashi’s equation for frost heave in the direction of heat flow.  Kanie et al. 2013 has proposed the use of a three-dimensional evaluation method involving a unique laboratory apparatus and finite element method modeling.


Currently, standards exist for the determination of strength properties at constant strain (ASTM D7300-11) and creep properties (ASTM D5520-11).  Both of these tests are performed in uniaxial compression.  Standards for triaxial testing of unfrozen soil do not apply to frozen soils, and new standards are needed to generate comparable results.  There are many unstandardized laboratory and field tests currently used for frozen soils including (Oestgaard and Zubeck 2013):


- Direct Shear (Bennett and Nickling 1984, Yasufuku et al. 2003).

- Triaxial Compression (Baker et al. 1984, Arenson et al 2004).

- Uniaxial Tension (Zhu and Carbee 1987, Erckhardt 1981).

- Constant Creep (Andersland and Ladanyi 2004).

- Relaxation Test (Andersland and Ladanyi 2004).

- Thaw Consolidation (Morgenstern and Nixon 1971).

- Pressuremeter Creep (Ladanyi 1982).

- Pressuremeter Relaxation (Ladanyi 1982, Ladanyi and Melouki 1992).




The classification and description of frozen soils is currently documented by ASTM D4083-89 (reapproved in 2007).  This involves the description of both the soil phase and the ice phase of the material.  Description for the soil phase is the same as an unfrozen soil, ASTM D2488.  The frozen phase is then classified into one of two groups: N for soil with no ice visible and V for soil with significant visible ice.  These groups are subsequently broken down into subgroups described in the standard.  Figures 3 and 4 show visual representations of both the visible ice and no visible ice classifications, per the ASTM D4083-89 standard.


Figure 3.  Visible ice structured in frozen soil (ASTM D4083-89)


Visible ice is represented by the color black in Figure 3.  Visible ice may exist within the soil structure as separate pockets of ice (Vx), coatings around soil particles (Vc), irregular formations (Vr), or stratified formations (Vs).


Figure 4.  Structure of frozen soil with no visible ice (ASTM D4083-89)


As in Figure 3, ice is represented by the color black in Figure 4.  When there is no visible ice within the structure of a frozen soil, the frozen soil is classified by how well the sample is bonded by the ice.  Frozen soil with no visible ice may be poorly bonded (Nf), well-bonded with no excess ice (Nbn), or well-bonded with excess ice (Nbe).


Sayles et al. 1987 provides several recommendations for the complete description of a frozen soil.  These include unfrozen soil USCS symbol and description, frozen soil symbol and description, grain size distribution, Atterberg limits, as well as physical properties such as ice content (frozen), water content (unfrozen), unit weight, specific gravity of soil, saturation percentage, and salinity.  These parameters have a strong influence on the frozen strength and behavior of the soil.  For artificial ground freezing applications, it is recommended that the system described in Andersland and Anderson 1978 be used (Sayles et al. 1987).  Still et al. 2013 has proposed developing standardized index testing for use in frozen soil classification.


Add comment

NOTE: The symbol < is not allowed in comments. If you use it, the comment will not be published correctly.

Security code
*Please insert the above-shown characters in the field below.