Postpeak Strength of Interfaces in a

Stress-Dilatancy Framework
Matthew S. Dietz1 and Martin L. Lings2

Abstract: Laboratory sand-steel interface tests, using a range of sand sizes on a wide range of surface roughnesses, have been conducted
using a direct shear apparatus modified to enable reliable measurements of both friction and dilation. The paper looks at the minimum
interface strength after peak, termed here the postpeak strength, and assesses its dependence on roughness, density, and stress level. Its
upper limit is the large displacement direct shear friction angle, related to but not equal to the critical state friction angle. When data are
normalized by this value, they show linear dependence on the logarithm of relative roughness in the intermediate zone between smooth
and rough. Once the roughness dependence of the postpeak strength has been allowed for, dilatant interfaces are found to follow classical
stress–dilatancy relationships. It appears that there is no fundamental difference in the responses of sand-on-steel or sand-on-sand
interfaces.
DOI: 10.1061/共ASCE兲1090-0241共2006兲132:11共1474兲
CE Database subject headings: Dilatancy; Friction; Interface shear; Roughness; Sand; Steel.

Introduction and Dietz 2005兲. The model of interface behavior that emerges
unites peak strength, peak dilation, and postpeak strength using
An interface exists when an assemblage of discrete particles is in simple expressions adapted from classical stress-dilatancy theory.
contact with a solid surface. Interfaces between sand and con-
struction materials occur in a wide variety of situations, and the Background
friction mobilized between the two is often of importance to en-
gineers. If only very small relative displacements are likely to
occur, or if the maximum strength needs to be overcome, and Peak Strength
therefore quantified, then peak strength will be the parameter of Much research has been published on the peak strength of sand–
interest. If larger displacements can be expected to occur, then the steel interfaces, and the main contributions have recently been
lower postpeak strength may be the more appropriate parameter. summarized by Lings and Dietz 共2005兲. Here, a brief summary of
Common examples where the magnitude of the postpeak interface the terminology that has been used in the geotechnical interface
friction might be important include piles and retaining structures. literature is presented.
Here, postpeak interface response and its relation to peak state Uesugi and Kishida 共1986兲 introduced normalized roughness
parameters are investigated using a modified version of the direct 共Rmax / D50兲, where Rmax is the maximum roughness measured over
shear apparatus 共DSA兲. Unlike conventional DSA, the modified a gauge length L ⬇ D50, to account for the observation that, in
device provides more reliable measurements of both friction and general, a fine sand has a higher frictional resistance on a surface
dilation 共Lings and Dietz 2004兲. Seven surfaces of widely varying of given roughness than a coarse sand. Subba Rao et al. 共1998兲
roughness have been tested alongside coarse, medium, and fine introduced an alternative definition, relative roughness 共Ra / Dav兲,
sands. The full stress range available with the experimental appa- where Ra is the average roughness and Dav is the weighted aver-
ratus has been covered and densities ranging from dense to loose age particle size of the sand. Lings and Dietz 共2005兲 showed that
explored. Attention is primarily focused on those interfaces ex- both approaches are equally good at integrating peak friction data,
hibiting dilation since, for most sands, the roughness of typical and are also able to integrate peak dilation data.
rolled steel surfaces is sufficient to give a dilatant response 共Lings Paikowsky et al. 共1995兲 introduced the terms smooth, interme-
diate, and rough to describe three distinct roughness zones.
1
Research Associate, Dept. of Civil Engineering, Univ. of Bristol, Smooth interfaces mobilize low peak strengths; rough interfaces
Queen’s Building, University Walk, Bristol BS8 1TR, U.K. E-mail: mobilize peak strengths equivalent to those obtained in direct
M.Dietz@bristol.ac.uk shear; and intermediate interfaces are transitional between the
2
Senior Lecturer, Dept. of Civil Engineering, Univ. of Bristol, two. Lings and Dietz 共2005兲 extended these descriptions, showing
Queen’s Building, University Walk, Bristol BS8 1TR, U.K. E-mail: them to be equally applicable to peak dilation measurements.
Martin.Lings@bristol.ac.uk Dove and Jarrett 共2002兲 introduced the terms nondilative and
Note. Discussion open until April 1, 2007. Separate discussions must
dilative to describe interface behavior, the former applicable
be submitted for individual papers. To extend the closing date by one
when the interface is smooth, the latter when the interface is
month, a written request must be filed with the ASCE Managing Editor.
The manuscript for this paper was submitted for review and possible rough, mobilizing the full strength of the soil. Lings and Dietz
publication on November 1, 2005; approved on February 8, 2006. This 共2005兲 introduced the slightly different terms nondilatant and di-
paper is part of the Journal of Geotechnical and Geoenvironmental latant, the latter referring to interfaces that dilate by any signifi-
Engineering, Vol. 132, No. 11, November 1, 2006. ©ASCE, ISSN 1090- cant amount during shear, up to and including those that are
0241/2006/11-1474–1484/$25.00. rough.

1474 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006

 Fig. 2. WDSA used for interface testing

Fig. 1. Influence of relative roughness on strength after peak tan ␾⬘p = tan ␾ld
⬘ + tan ␺ p 共1兲
关adapted from Subba Rao et al. 共1998兲兴
where ␾⬘p⫽peak friction angle; and ␺ p⫽peak dilation angle. Other
relevant flow rules that work well for sands have been proposed
by Rowe 共1962兲 and Bolton 共1986兲, although they are formulated
Strength after Peak using plane strain parameters and, therefore, lack the directness of
Comparatively little attention has been focused on interface Taylor’s flow rule. However, they are all similar, and only diverge
strength after peak. Jardine et al. 共1993兲 investigated interface at large dilation angles 共Jewell 1989兲.
friction with a DSA and ring shear apparatus using a range of A modified version of Taylor’s flow rule, adapted for inter-
granular soils on steel surfaces with roughnesses comparable to faces, can be written as
those of industrial piles. They concluded that the strength after tan ␦⬘p = tan ␦⬘pp + tan ␰ p 共2兲
peak, referred to as the critical state interface friction angle, was
the controlling frictional parameter for piles in sand and silt. They where ␦⬘p and ␦⬘pp⫽peak and postpeak interface friction angles,
found that it was independent of density, although it appeared to respectively; and ␰ p⫽peak interface dilation angle. The validity of
be strongly dependent on stress level. It reduced with increasing this flow rule will be explored later.
D50, and was upper bounded by the critical state strength of the Dove and Jarrett 共2002兲 have suggested a simple flow rule
granular material. from tests using glass beads and sand on machined aluminum
Subba Rao et al. 共1998兲 used a DSA to measure interface interfaces, expressed in terms of the symbols adopted in this
frictional properties of sands with a range of sizes, on surfaces of paper as
stainless steel, mild steel, and ferrocement with a wide range of
␦⬘p = ␦⬘pp + ␰ p 共3兲
roughness values. All samples had a high initial relative density,
and a modest range of stresses was investigated. Critical state Lings and Dietz 共2005兲 have suggested a somewhat similar flow
friction angles 共␦⬘cv兲 were shown to vary with relative roughness, rule from tests using sand-on-steel interfaces
and to approximate to the critical state friction angle of the soil
␦⬘p = 25 ° + ␰ p 共4兲
共␾⬘cv兲 at large relative roughness values. Their data are reproduced
in Fig. 1. Both these expressions imply that interface friction angles are
Other terms have been used to describe interface strength after enhanced by the full dilation angle, akin to rock mechanics, rather
peak. Dove and Jarrett 共2002兲 refer to the strength obtained once than only by a proportion of the dilation angle, as observed for
any dilation has ceased as the steady state. Uesugi et al. 共1989兲 soil 共Bolton 1986兲.
refer to the strength achieved after multiple two-way displace-
ments as the residual strength. Here, the term postpeak strength is
used, defined as the minimum value reached after peak. It is Methodology
reached reasonably rapidly and implies nothing about a final
value at very large displacement. The term critical state implies a The winged direct shear apparatus 共WDSA兲 was used for all tests.
unique value, whereas the value depends on roughness. Also, This device retains the simplicity of the conventional apparatus,
there is clear evidence that the friction angle measured at large yet is able to generate high-quality data. The essential modifica-
displacement in direct shear 共␾ld ⬘ 兲 is not the same as the critical tion of the WDSA is that the shear force is transmitted through
state friction angle 共␾crit⬘ 兲, they are related by tan ␾ld⬘ = sin ␾crit
⬘ ball races to a pair of “wings” attached to the sides of the shear-
共Lings and Dietz 2004兲. There is, therefore, potential for confu- box, thereby preventing unwanted forces and moments from
sion in using this label. The term steady state implies that the acting, and permitting unimpeded dilation. The device was devel-
value is stable, whereas for smooth surfaces it can increase with oped by Dietz 共2000兲 and has been validated by Lings and Dietz
further displacement 共Uesugi et al. 1989兲. For all these reasons, 共2004兲.
we have chosen to adopt the label postpeak strength. The WDSA as used for interface testing is pictured in Fig. 2.
The sand is confined within an upper frame, 100 mm square in
plan. Underlying the sand is a solid steel block, the top surface of
Flow Rules
which is prepared with the surface texture to be investigated.
The earliest flow rule, formulated by Taylor 共1948兲 and specifi- Overlying the sand is the load pad onto which a vertical normal
cally for a direct shear test, can be written as load is applied. The load pad is secured within the upper frame

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006 / 1475

Table 1. Physical Properties of Test Sands
D10 D50 D60 ⬘
⌽ld
Sand Name Gs emin emax 共mm兲 共mm兲 共mm兲 Uc Shape 共deg兲a
Coarse VLB 2.651 0.506 0.802 0.64 0.78 0.81 1.27 Rounded 31.2
Medium MGS 2.657 0.494 0.806 0.34 0.44 0.45 1.32 Subrounded 31.0
Fine SFS 2.654 0.684 1.017 0.09 0.13 0.14 1.56 Subrounded 35.3
a
Values measured under 25 kPa confining stress.

prior to testing, the so-called symmetrical arrangement 共Jewell lution of the digitized points is ±0.5 nm, but background noise
1989兲, which helps reduce upper-frame rotations 共␻ in Fig. 2兲. produced vibrations of the stylus tip of around 3 nm. In general,
Prior to sample deposition, a gap of 5D50 is created beneath the roughness parameters were evaluated from unfiltered 共primary兲
upper frame, and strips of rubber edging are attached to the inter- profiles of 5 mm length employing a 0.25 ␮m sampling interval.
nal walls using silicone grease to limit sample extrusion 共Lings For the rougher surfaces, it was necessary to increase the traverse
and Dietz 2004兲. length to 70 mm to acquire sufficient profile data to produce an
accurate estimate of roughness parameters. Storage limits then
Sands required the sampling interval to be reduced to 1 ␮m. The pro-
filometer was configured to add a horizontal mean line to profile
Three different sands were used. The coarse sand was Leighton data using the least-squares method. Visual inspection revealed
Buzzard, and because the batch was new, unlike others in the that the surface profiles could be characterized solely by a rough-
laboratory, was termed virgin Leighton Buzzard 共VLB兲. The me- ness component, a consequence of the interface block preparation
dium sand, because of its golden color, was termed medium procedure. No longer-wavelength waviness was apparent.
golden sand 共MGS兲. The fine sand, because of its silver-gray A total of five profile segments were used to characterize a
color, was termed silver fine sand 共SFS兲. A number of tests were surface, running parallel to the shear displacement. Each traverse
performed to characterize the physical properties of these sands, was taken across a different portion of the surface to provide an
the results of which are displayed in Table 1. indication of the uniformity of roughness, which was generally
Samples were prepared by pluviation, using a device similar to found to be high. Average roughness values for each surface are
the multiple sieve pluviator described by Miura and Toki 共1982兲. given in Table 2, which were used to calculate relative roughness
The rate of deposition was used to control the depositional den- values 共using D50兲 for the different interface tests.
sity via the aperture size of the device’s nozzle. Loose samples
were prepared by the slow pouring method described by Miura et
al. 共1997兲. The uppermost surface of the sample was initially Test Parameters
leveled using a small vacuum device 共Stroud 1971兲, and finally
leveled by contact with a greased platen repeatedly lowered onto Three measurements are usually taken during a shear test: the
the surface of the sand to remove excess grains. horizontal displacement, vx; the vertical displacement of the load
pad, vy; and the shear load S 共Fig. 2兲. Here, vertical displacements
were recorded at the front and rear of the load pad, allowing its
Surfaces rotation, ␻, to be derived, as well as the average vertical displace-
The mild steel interface blocks were machined and ground to ment. Upward motions and clockwise rotations are here taken to
have their principal surfaces parallel to one another. Four distinct be positive. As discussed by Lings and Dietz 共2004兲, such rota-
steel roughness magnitudes were investigated, each newly pre- tions seem to be a persistent feature of the WDSA, although they
pared prior to each test using the processes described in Table 2. have an almost negligible effect on measured parameters.
Three different sand roughness magnitudes were investigated, The applied normal load, N, the self-weight of the upper frame
formed by fixing each of the test sands to an interface block as assembly, na; the self-weight of the sand deposit, ns; and the plan
described in Table 2. area of the sample, A; all need to be accounted for when calcu-
Surface topographies were digitized using a profilometer lating the average vertical normal stress, ␴⬘yy. A similar calculation
共Talysurf 112/2009兲, which has a spherical tip of radius 2 ␮m and produces the average horizontal shear stress, ␶yx, from the mea-
contact force between 0.7 and 1 mN. The nominal vertical reso- sured shear load, S

Table 2. Formative Processes and Properties of Test Surfaces

Ra
Material Name Formative process 共␮m兲
Steel POL Abrasion of ground surface using progressively finer grades of abrasive paper, before polishing with Brasso 0.147
Steel GND Grinding wheel, with direction of striations perpendicular to direction of shearing 0.356
Steel ALO Shot blasting with 120 grit aluminium oxide 共D50 = 0.2 mm兲 2.49
Steel SIC Shot blasting with 16 grit silicone carbide 共D50 = 1.8 mm兲 9.40
Sand SFS Pluviation of SFS onto uniform coating of Araldite smeared over interface block 33.7
Sand MGS Pluviation of MGS onto uniform coating of Araldite smeared over interface block 114
Sand VLB Pluviation of VLB onto uniform coating of Araldite smeared over interface block 180

1476 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006

 N + na + ns Postpeak Strength
␴⬘yy = 共5兲
A
Effect of Roughness and Particle Size
Fig. 4共a兲 plots postpeak friction angle against average roughness
S for Test Series 1. Fig. 4共b兲 presents the same data using relative
␶yx = 共6兲
A roughness 共Ra / D50兲 and normalizing ␦⬘pp by ␾ld⬘ . Values of ␾ld
⬘ are
The angles of interface friction and dilation are obtained from taken from Table 1, with the exception of the fine sand, where
there is evidence that the value measured is much too large 关as
the following equations, exactly analogous to angles 共␾⬘ and ␺兲
seen in Fig. 4共c兲 of Lings and Dietz 共2005兲兴, thought to be due to
obtained in direct shear:
the formation of oblique rupture planes. Instead, the value of
␦⬘pp from the sand–sand test 共31.6°, see Table 3兲 has been used
␶yx instead. As shown for peak data 共Lings and Dietz 2005兲, normal-
tan ␦⬘ = 共7兲 ization brings the three different trends together in the intermedi-
␴⬘yy
ate and rough zones. Data in the smooth zone remain segregated,
in fact, normalization seems to make the scatter worse.
Trend lines have been added to the intermediate and rough
dv y zones in Fig. 4共b兲. The intermediate zone trend line is defined by
tan ␰ = 共8兲
dvx 0.003ⱕ Ra / D50 ⱕ 0.1 and 0.75ⱕ ␦⬘pp / ␾ld
⬘ ⱕ 1.0. Because it is not
clear what the trend in the smooth zone is, these data are enclosed
by dashed lines.

Data
Effect of Relative Density and Stress Level
Three series of tests were carried out, each covering a wide range Fig. 5共a兲 plots postpeak friction angle against relative density for
of roughness values. The effects of particle size were explored in Test Series 2, including Tests 9–13 from Series 1. Best-fit lines
the first test series. All three sands were tested at low stress levels have been plotted through the data to see whether any trends are
共⬃25 kPa兲 and high relative densities 共⬃90% 兲 to maximize di- evident. At higher relative roughness, postpeak strength appears
lation. The second test series explored the effects of density, using to be independent of density. At lower relative roughness, post-
the medium sand at low stress levels. The third test series ex- peak strength appears to increase with density.
plored the effects of stress level using the coarse sand at high Fig. 6共a兲 plots postpeak friction angle against stress level for
relative density. All tests were carried out at a constant rate of Test Series 3, including Tests 1 and 7 from Series 1. Again, best-
shear displacement of 1.2 mm/ min. fit lines have been plotted through the data. Rough interfaces
Full details of all these tests, together with the peak and post- show a general trend of slightly decreasing postpeak strength with
peak parameters measured, are displayed in Table 3. Postpeak increasing stress level that exactly mirrors the trend of large dis-
strength is the minimum value recorded after peak, irrespective of placement direct shear friction angles, also shown on the plot,
previously noted by Lings and Dietz 共2004兲. Intermediate inter-
the displacement at which it occurs. Because it occurs at different
face data are more scattered, but show a somewhat similar trend.
displacements in different tests, these displacements have been
Smooth interfaces show, if anything, the reverse, with postpeak
included in the table for completeness.
strengths increasing very slightly with increasing stress level.
From a total of 44 interface tests performed, 14 have been
Normalized postpeak strength data from these two test series
selected to highlight important aspects of behavior within each are plotted against relative roughness in Figs. 5共b兲 and 6共b兲. In the
series. Block arrows indicate the general trends. Fig. 3共a兲 explores case of stress level, the trend line of the large displacement fric-
the influence of surface roughness on the response of coarse sand. tion angles from Fig. 6共a兲 has been used, with interpolation, to
Decreases in roughness push the curves downward, reducing peak normalize the data in Fig. 6共b兲. Reproduced in both Figs. 5共b兲 and
strengths, the magnitudes and rates of dilation, and the amounts 6共b兲 are the trend lines of Fig. 4共b兲. There is some scatter, but the
of rotation. For all tests, irrespective of roughness, a peak state trend lines seem to provide a reasonable match to both data sets.
can be identified soon after test initiation. For rough and interme-
diate surfaces, dilation and strain softening then occur until the
postpeak strength is reached. For smooth surfaces, dilation is neg-
Stress Dilatancy
ligible, and postpeak strength is reached shortly after peak. There-
after, there is an increase in resistance with increasing interface
displacement. Taylor and Bishop Plots
Fig. 3共b兲 explores the influence of depositional density on the Flow rules can be used to interpret direct shear data in two ways.
response of medium sand on the SIC surface. Decreases in den- Using Eq. 共1兲 as an example, first it can be used exactly as written
sity reduce peak strength, dilation, and rotation. Postpeak to investigate peak parameters. Second, it can be used to plot
strengths remain almost unaffected, and generally occur in the tan ␾⬘ against tan ␺ over the course of a test 共e.g., Jewell 1989兲.
latter half of the test. Fig. 3共c兲 explores the influence of stress The same approach can be followed for interfaces using Eq. 共2兲.
level on the response of coarse sand on the SIC surface. Increases If tan ␦⬘ is plotted against tan ␰ over the course of an interface
in stress level reduce peak strength, dilation, and rotation. Post- test, and if the scales are the same on both axes, the modified
peak strengths occur at a range of displacements and appear to be Taylor flow rule plots as a straight line inclined at 45°. Various
somewhat affected by stress level. such plots, referred to here as Taylor plots, are presented in Fig.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006 / 1477

Table 3. Interface Test Data
Dr ␴⬘yy ␦⬘p ␰p ⬘
␦pp ␰pp 共Vx兲pp
Number Sand Surface 共%兲 共kPa兲 共degrees兲 共degrees兲 共degrees兲 共degrees兲 共mm兲
Test Series 1
1 VLB POL 89 25.3 11.7 0.4 9.3 −0.2 1.2
2 VLB GND 94 25.2 16.7 0.2 13.4 0.0 1.1
3 VLB ALO 92 25.3 26.0 1.9 23.8 0.3 4.5
4 VLB SIC 88 25.1 33.2 9.6 26.6 0.1 11.2
5 VLB SFS 83 25.2 39.9 16.4 28.5 0.2 8.3
6 VLB MGS 89 25.2 47.5 24.8 30.9 −0.7 9.1
7 VLB VLB 86 25.2 48.1 26.6 30.8 −0.1 8.0
8 MGS POL 94 25.2 13.7 0.4 10.8 0.0 5.8
9 MGS GND 93 25.3 15.3 0.0 11.8 −0.1 1.9
10 MGS ALO 94 25.1 27.7 3.7 24.3 0.5 7.1
11 MGS SIC 93 25.2 39.0 14.1 27.7 0.3 11.6
12 MGS SFS 92 25.3 47.1 24.5 30.5 0.7 8.0
13 MGS MGS 93 25.3 49.0 26.3 31.7 0.7 8.4
14 SFS POL 99 25.3 10.9 0.4 9.3 −0.1 10.1
15 SFS GND 96 25.3 15.3 0.1 10.8 −0.1 11.3
16 SFS ALO 94 25.2 36.5 10.2 28.2 0.0 7.7
17 SFS SIC 92 25.3 42.8 15.2 30.6 0.1 5.5
18 SFS SFS 97 25.2 43.2 17.0 31.6 0.1 9.2
Test Series 2
19 MGS GND 78 25.1 13.3 0.1 9.3 0.1 9.8
20 MGS GND 69 25.2 11.5 0.1 7.1 −0.1 10.9
21 MGS GND 23 25.1 10.8 0.1 8.0 −0.4 97
22 MGS ALO 79 25.3 25.1 3.2 22.4 −0.1 9.4
23 MGS ALO 69 25.3 25.1 2.3 21.6 0.1 8.5
24 MGS ALO 24 25.3 22.6 0.0 21.3 −0.3 9.2
25 MGS SIC 75 25.3 35.8 12.7 28.4 0.6 10.7
26 MGS SIC 68 25.4 33.4 11.2 28.5 0.5 9.7
27 MGS SIC 22 25.1 29.1 1.5 28.2 0.0 64
28 MGS SFS 70 25.2 42.8 18.2 32.0 1.1 4.6
29 MGS SFS 26 25.3 33.7 4.2 32.5 1.3 5.5
30 MGS MGS 78 25.3 43.4 18.6 32.3 0.5 5.9
31 MGS MGS 62 25.4 40.6 14.6 32.0 0.6 5.4
32 MGS MGS 27 25.2 33.9 3.6 31.8 −0.5 9.9
Test Series 3
33 VLB POL 99 128.0 14.6 0.6 9.6 −0.1 3.1
34 VLB POL 93 251.1 16.1 0.9 9.9 −0.1 4.0
35 VLB ALO 93 25.2 29.1 5.2 26.1 0.0 11.7
36 VLB ALO 95 86.9 26.1 1.2 22.6 −0.1 2.8
37 VLB ALO 99 169.4 25.1 1.9 21.5 0.4 3.3
38 VLB ALO 91 251.0 26.1 2.9 23.3 0.2 3.8
39 VLB SIC 90 25.1 31.4 8.6 23.5 0.6 11.5
40 VLB SIC 99 128.6 29.2 4.5 24.7 0.6 10.0
41 VLB SIC 99 251.4 26.5 1.9 23.0 0.0 4.4
42 VLB VLB 70 86.9 45.6 22.1 30.4 −1.8 8.9
43 VLB VLB 74 169.6 45.4 21.6 29.6 −1.7 8.3
44 VLB VLB 74 251.6 44.7 21.2 29.4 −1.7 11.1
Additional Test
45 VLB Milled 91 25.1 47.5 25.2 30.6 −0.4 10.9

1478 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006

Fig. 3. Interface stress and displacement histories showing general trends from: 共a兲 decreasing roughness, using coarse sand; 共b兲 decreasing
density, using medium sand; and 共c兲 increasing confining stress, using coarse sand

7共a兲 for dense coarse sand confined under 25 kPa. Apart from steel tests given by Eq. 共4兲. There appear to be different trends
data before peak, acquired while principal axes are still rotating, displayed by sand-on-sand and sand-on-steel tests, and Lings and
data generally lie fairly close to the Taylor line. Dietz 共2005兲 speculated that sand-on-sand interfaces may follow
Bishop 共1950兲 plotted data from direct shear tests using an conventional flow rules where only part of the dilation is con-
axis equivalent to tan ␾⬘ – tan ␺ to separate the work done in over- verted into additional strength, whereas the sand-on-steel inter-
coming internal friction from that required to cause the sample to faces appear to convert the whole of the dilation into additional
dilate. This approach is essentially identical to that used by Taylor strength. This is emphasized in Fig. 8共b兲 where the same data are
共1948兲. Data from the same three interface tests are presented in presented using Taylor axes. The sand-on-sand tests appear to
Fig. 7共b兲 using axes of tan ␦⬘ – tan ␰ against horizontal displace-
follow a Taylor flow rule that is rather different from the trend of
ment. The points during each test when peak and postpeak
the sand-on-steel tests.
strengths have been reached have been marked on the plots, re-
As a first attempt at clarifying this apparent difference, an
ferred to here as Bishop plots. Generally, they show that after
peak, the interface follows a modified Taylor flow rule quite additional test result has been added to Fig. 8共b兲. A milling ma-
closely right to the end of the test. chine was used to inscribe an interface block with a 30° sym-
metrical triangular wave of 2 mm asperity height. An interface
test using dense coarse sand was carried out, and the results are
Peak State included at the bottom of Table 3. With an estimated Ra / D50 of
Peak friction and dilation data from all 44 interface tests are pre- 0.64, this interface is rougher than the roughest sand-on-sand in-
sented in Fig. 8共a兲, together with the trend line for the sand-on- terface. The data point for the sand on milled steel test sits in

Fig. 4. Effect of roughness and particle size on postpeak strength using 共a兲 standard axes; 共b兲 normalized axes

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006 / 1479

 Fig. 5. Effect of density on postpeak strength of medium sand plotted using 共a兲 relative density axis; 共b兲 relative roughness axis

among the sand-on-sand data, suggesting that there might not be a dilation reduces. As dilation disappears altogether, the slope of
fundamental difference between the two types of test after all, the line becomes very large. Once the variable postpeak strength
provided the steel surface is rough enough. This has prompted us has been accounted for, Taylor’s flow rule captures the response
to re-examine the data to try and gain a better understanding of of dilatant interfaces reasonably well.
the relationship between peak and postpeak strength.
Modified Flow Rules
From Peak to Postpeak
Bolton’s well-known plane strain 共ps兲 flow rule can be expressed
Peak and postpeak data from all tests with dense coarse sand are in terms of direct shear parameters if use is made of Davis’s
presented in Fig. 9 using Taylor axes. The postpeak dilation angle, 共1968兲 relation
␰pp, is obtained from the dilation rate coincident with ␦pp ⬘ . It is
always of small magnitude, but partly because of noise, is rarely ⬘
cos ␺ sin ␾ps
zero. Based on the evidence from Fig. 7 that data between peak ⬘ =
tan ␾ds 共9兲
⬘
1 − sin ␺ sin ␾ps
and postpeak seem to follow a Taylor flow rule quite closely, a
straight line has been drawn joining the two points. Data from A critical state friction angle of 35° 共appropriate to the sands
nondilatant tests are shown to an enlarged scale as an insert in the tested here兲 converts to a large displacement direct shear 共ds兲
figure. Rough interfaces, including the milled steel test, closely ⬘ 兲p against ␺p is then no
friction angle of about 30°. A plot of 共␾ds
follow a Taylor flow rule. Intermediate interfaces also broadly longer a perfect straight line, but a good straight-line fit is ob-
follow a Taylor flow rule, although there is increasing scatter as tained with a slope value of 0.75. Thus, Bolton’s flow rule be-
the roughness reduces. Smooth interfaces, where there is almost comes indistinguishable from
no dilation yet a significant difference between peak and postpeak
strength, show steep lines joining the two states. ⬘ 兲 p = 共␾ds
共␾ds ⬘ 兲ld + 0.75␺ p 共10兲
The gradient of the conjoined line for all 45 tests has been
plotted on a logarithmic scale in Fig. 10. Gradients are close to Conversion from plane strain to direct shear leads to a reduction
unity for highly dilatant interfaces, with increasing scatter as peak in the familiar 0.8 factor down to 0.75. A modified form of Bol-

Fig. 6. Effect of stress level on postpeak strength of coarse sand plotted using 共a兲 stress axis; 共b兲 relative roughness axis

1480 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006

Fig. 7. Tests with dense coarse sand on various surfaces displayed using 共a兲 Taylor plot; 共b兲 Bishop plot

Fig. 8. Peak interface stress-dilatancy plots using 共a兲 angles; 共b兲 tangents of angles 共Taylor axes兲

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006 / 1481

 points lie significantly below this line, lying instead close to the
Bolton flow rule. Eq. 共3兲 is, therefore, shown not to apply to
interfaces.
The same dilatant test data are presented in Fig. 11共b兲, but now
using axes of tan ␰ p and tan ␦⬘p – tan ␦⬘pp. Equation 共2兲 is superim-
posed and is seen to provide a reasonable fit to all dilatant tests,
whether sand-on-sand or sand-on-steel. The data from Dove and
Jarrett 共2002兲 almost all lie on this Taylor flow rule.
Thus, by allowing for postpeak strength that varies with
roughness, all interface tests with surfaces of sand, steel, or alu-
minum are united by flow rules that mirror classical flow rules.

Interface Stress-Dilatancy Model

For a particular sand, it is now possible to construct an approxi-
mate model for the peak strength of any dilatant interface, which
is presented in Fig. 12. For soils at a particular stress level, the
peak internal friction angle is dependent on the large displace-
ment friction angle and the density, which in turn controls dilation
and how far along the flow rule the peak state is located. For
interfaces, a similar pattern emerges, except that the postpeak
friction angle is not unique but depends on roughness. Rough
interfaces achieve the same peak state as the soil in direct shear.
Intermediate interfaces have a postpeak strength that depends on
roughness, and dilate by an amount that depends on density. The
amount of dilation also depends on roughness, with rougher sur-
faces dilating more for a given density. For soil of a given density
Fig. 9. Taylor plot showing peak and postpeak points for tests with
共medium density is portrayed兲, the peak state will follow the
dense coarse sand
chain-dotted line in Fig. 12 as the roughness varies from smooth
to rough.
It is now clear why the flow rule suggested by Lings and Dietz
ton’s equation for interfaces, expressed in terms of peak and post-
共2005兲 关Eq. 共4兲兴 appeared to show that steel interfaces were ca-
peak parameters, can then be written
pable of mobilizing the full effect of dilation. It is because of the
␦⬘p = ␦⬘pp + 0.75␰ p 共11兲 twofold effect of roughness: on postpeak strength and on dilation.
It is concluded that the model in Fig. 12 is the more fundamental
All dilatant test data 共excluding tests where peak dilation insight into the behavior of dilatant interfaces.
angles are less than 1°兲 are presented in Fig. 11共a兲 using axes of
␰ p and ␦⬘p − ␦⬘pp. Equation 共11兲 is superimposed and is seen to pro-
vide a reasonable fit to all dilatant tests, whether sand-on-sand or Discussion
sand-on-steel. The data for sand on aluminum from Dove and
Jarrett 共2002兲 have also been plotted, and it is seen that when The maximum value of postpeak interface strength has been
peak dilation is small, the points lie close to the equality line, shown to be the large displacement direct shear friction angle.
which represents Eq. 共3兲. However, with increasing dilation, the This enables the following important equation to be written

tan ␦⬘pp ⱕ tan ␾ld

⬘ = sin ␾crit
⬘ 共12兲
This confirms for sand-on-steel interfaces what has already been
suggested by O’Rourke et al. 共1990兲 for polymers.
Highly dilatant tests plot below Bolton’s flow rule in Fig. 11共a兲
and above Taylor’s flow rule in Fig. 11共b兲. This is consistent with
the finding of Lings and Dietz 共2004兲 that the peak strength of
dense coarse Leighton Buzzard sand best fits the flow rule of
Rowe 共1962兲, which lies between those of Bolton and Taylor.
The large displacement friction angles of the three sands con-
sidered in this study are broadly similar. Further work is required
to assess the influence of significantly different large displace-
ment friction angles on interface response.
It should be noted that the model presented in Fig. 12 is based
largely on behavior at low stress levels. At higher stresses, it is
anticipated that ␾ld⬘ and ␦⬘pp will be slightly lower; also, there will
be less dilation, and hence, lower peak strength, for a given
Fig. 10. Gradient of line joining peak and postpeak points on Taylor roughness and density. However, the effects of stress level ob-
plot 共all tests兲 served for ␦⬘pp in this work 关Fig. 6共a兲兴 are not as marked as those

1482 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006

 Fig. 11. Dilatant interface test data and flow rule using 共a兲 Bolton axes; 共b兲 Taylor axes

reported by Jardine et al. 共1993兲, and further work is required to ratus modified to enable reliable measurements of both friction
clarify this. and dilation. Seven surfaces of wide-ranging roughness have been
tested in conjunction with three sands, one coarse, one medium,
and one fine. A range of depositional densities and confining
Summary and Conclusions stress levels have been explored.
The focus of the paper has been the minimum strength of
The stress-dilatancy characteristics of sand-on-steel and sand-on- interfaces attained shortly after peak when any dilation has
sand interfaces have been investigated using a direct shear appa- ceased, referred to here as the postpeak strength. This terminol-
ogy is preferred to the alternatives of critical state, steady state, or
residual, because it is not unique, it can increase with further
displacement, and it may differ from values at very large dis-
placement. It is important because it is likely to be a more appro-
priate value than peak strength in many design situations.
The maximum value of postpeak interface strength is the large
displacement direct shear friction angle 共␦⬘pp ⱕ ␾ld ⬘ 兲, related to the
critical state friction angle by tan ␾ld ⬘ = sin ␾crit
⬘ . The normalized
postpeak friction angle 共␦pp ⬘ / ␾ld
⬘ 兲 shows linear behavior against
the logarithm of relative roughness 共Ra / D50兲 for dilatant inter-
faces, with values between about 0.75 and 1.0. For nondilatant
interfaces, the values can be very much smaller. The transition
between the two occurs at a relative roughness Ra / D50 ⬇ 0.003.
For dilatant interfaces, postpeak strength is primarily a func-
tion of roughness. Density, and to a lesser extent stress level,
control how much a soil dilates against a surface of particular
roughness. However, roughness not only controls postpeak
strength, it also has a significant effect on dilation. This leads to a
simplified model of interface behavior that links peak and post-
peak friction and peak dilation using a classical flow rule, where
only part of the dilation is converted to enhanced strength. When
peak strength is understood in these terms, it appears that there is
no fundamental difference between sand-on-steel and sand-on-
sand interface behavior.

Acknowledgments

The writers offer their thanks to Dr. Thomas Pearce and Alan
Speight of the Institute of Grinding Technology, University of
Fig. 12. Approximate model for peak strength of dilatant interfaces Bristol, for the use of their profilometer; also, to Mike Pope for
for a given sand technician support in the laboratory. Financial support from the

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006 / 1483

Engineering and Physical Sciences Research Council, U.K., is Dietz, M. S. 共2000兲. “Developing an holistic understanding of interface
also gratefully acknowledged. friction using sand within the direct shear apparatus.” Ph.D. thesis,
Univ. of Bristol, Bristol, U.K.
Dove, J. E., and Jarrett, J. B. 共2002兲. “Behavior of dilative sand interfaces
Notation in a geotribology framework.” J. Geotech. Geoenviron. Eng., 128共1兲,
25–37.
Jardine, R. J., Lehane, B. M., and Everton, S. J. 共1993兲. “Friction coeffi-
The following symbols are used in this paper:
cients for piles in sands and silts.” Offshore site investigation and
Dr ⫽ relative density;
foundation behavior, Kluwer, Dordrecht, The Netherlands, 661–677.
D50 ⫽ median particle size; Jewell, R. A. 共1989兲. “Direct shear tests on sand.” Geotechnique, 39共2兲,
Ra ⫽ average roughness; 309–322.
vx ⫽ horizontal displacement; Lings, M. L., and Dietz, M. S. 共2004兲. “An improved direct shear appa-
vy ⫽ vertical displacement; ratus for sand.” Geotechnique, 54共4兲, 245–256.
␴⬘yy ⫽ effective vertical normal stress; Lings, M. L., and Dietz, M. S. 共2005兲. “The peak strength of sand-steel
␶yx ⫽ horizontal shear stress; interfaces and the role of dilation.” Soils Found., 45共6兲, 1–14.
␦⬘ ⫽ interface friction angle; Miura, S., and Toki, S. 共1982兲. “A sample preparation method and its
␾⬘ ⫽ direct shear friction angle; effects on static and cyclic deformation-strength properties of sand.”
␰ ⫽ interface dilation angle; Soils Found., 22共1兲, 61–77.
␺ ⫽ direct shear dilation angle; and Miura, K., Maeda, K., and Toki, S. 共1997兲. “Method of measurement for
the angle of repose of sands.” Soils Found., 37共2兲, 89–96.
␻ ⫽ angle of rotation of upper frame.
O’Rourke, T. D., Druschel, S. J., and Netravali, A. N. 共1990兲. “Shear
Subscripts strength characteristics of sand-polymer interfaces.” J. Geotech.
crit ⫽ indicates critical state; Engrg., 116共3兲, 451–469.
ds ⫽ indicates direct shear; Paikowsky, S. G., Player, C. M., and Connors, P. J. 共1995兲. “A dual
ld ⫽ indicates large displacement; interface apparatus for testing unrestricted friction of a soil along solid
p ⫽ indicates peak; surfaces.” Geotech. Test. J., 18共2兲, 168–193.
Rowe, P. W. 共1962兲. “The stress dilatancy relation for static equilibrium
pp ⫽ indicates postpeak; and
of an assembly of particles in contact.” Proc. R. Soc. London, A269,
ps ⫽ indicates plane strain.
500–527.
Stroud, M. A. 共1971兲. “The behavior of sand at low stress levels in the
simple shear apparatus.” Ph.D. thesis, Univ. of Cambridge, Cam-
References bridge, U.K.
Subba Rao, K. S., Allam, M. M., and Robinson, R. G. 共1998兲. “Interfacial
Bishop, A. W. 共1950兲. “Discussion on Measurement of shear strength of friction between sands and solid surfaces.” Geotech. Eng., 131,
soils.” Geotechnique, 2共2兲, 113–116. 75–82.
Bolton, M. D. 共1986兲. “The strength and dilatancy of sands.” Geotech- Taylor, D. W. 共1948兲. Fundamentals of soil mechanics, Wiley, New York.
nique, 36共1兲, 65–78. Uesugi, M., and Kishida, H. 共1986兲. “Frictional resistance at yield be-
Davis, E. H. 共1968兲. “Theories of plasticity and the failure of soil tween dry sand and mild steel.” Soils Found., 26共4兲, 139–149.
masses.” Soil mechanics: Selected topics, I. K. Lee, ed., Butterworth, Uesugi, M., Kishida, H., and Tsubakihara, Y. 共1989兲. “Friction between
London. sand and steel under repeated loading.” Soils Found., 29共3兲, 127–137.

1484 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2006