The Nature of Repose Slopes in Cohesionless Materials

A Paper by Bobby O. Hardin and Brian T. Hippley that Reports on an Independent Study
By Brian T. Hippley

INTRODUCTION

The motivation for research on repose slopes was to find a simple, accurate test to

measure the angle of shearing resistance ’ for cohesionless materials at or near zero

confinement. Taylor (1948) says, “the angle of repose is the friction angle under a

pressure of practically zero...” Conventional slope stability theory indicates that the

critical slope angle for infinite dry or completely submerged slopes is equal to ’.

However, due to boundary effects, the surface profiles for finite repose slopes produced

by laboratory tests are not always linear. Thus, it is necessary to examine the entire slope

profile, and measure the angle of repose for a portion of the profile sufficiently far from

boundaries.

The term angle of repose has generally been applied to loose cohesionless

materials. However, it is possible to design laboratory tests wherein the void ratio of

repose slopes can be controlled. One purpose of this paper is to study the effect of void

ratio on repose slope profiles. The authors seek to answer the question: can the effects of

void ratio on angle of shearing resistance for cohesionless materials under zero

confinement be determined by repose slope testing?

Terzaghi (1943) defines the angle of repose with reference to the sliding caused

by digging into a bed of dry or completely immersed sand. He says, “the sliding material

does not come to rest until the angle of inclination of the slope becomes equal to a certain

angle known as the angle of repose.” Lambe and Whitman (1968) define the angle of
repose with reference to an embankment formed by end dumping from a truck or the side

slopes of a pile formed by dropping granular material through a chute. They say, “the

inclination of the slope once dumping has ceased – the maximum slope at which the

material is stable – is called the angle of repose.”

Actually two repose slope angles can be defined in the laboratory by “tipping” or

“tilt box” experiments. In these experiments a specimen of granular material with

originally horizontal surface is tilted gradually, causing the surface slope to increase until

sliding is initiated. The maximum stable slope just prior to the initiation of sliding is

called the “critical” repose slope with slope angle C. Initiation of sliding leads to a

shallow avalanche of material down the slope. The stable slope that exists immediately

following the avalanche is called the “rest” repose slope with slope angle R (Carrigy,

1970). The excavation described in the Terzaghi definition of angle of repose is expected

to produce “rest” repose slopes. The slope produced by dumping or discharge from a

chute is less certain, depending on the rate at which material is deposited. Relatively fast

deposition probably produces an approximate “rest” repose slope.

Void ratio within a repose slope depends on how the slope was formed. The bed

of sand described in the Terzaghi definition of angle of repose could have any initial void

ratio (dense to loose). In order to explore the effect of void ratio on repose slopes, a new

test called the “cylindrical extrusion test” is described where “rest” repose slopes are

produced in specimens having a full range of void ratios1. Most procedures used to form

repose slopes, including dumping and discharge from a chute as described by Lambe and

1
Repose slopes produced by the cylindrical extrusion tests reported herein may not have been long enough
to avoid boundary and local geometry effects. Tests using a larger cylinder are planned prior to submission
of this paper for publication.
Whitman, produce a loose state within the slope. Eight different laboratory tests

described in the literature that have been used to produce repose slopes are shown

schematically in Fig. 1.

AVALANCHE SIZE DISTRIBUTIONS

Physicists have used a special form of the test in Fig. 1a in an attempt to

exemplify so-called “self-organized criticality.” An apparatus was designed to add one

particle at a time to a conical sand pile deposited on an elevated circular plate whose

diameter determines the base diameter of the pile (Held, et al., 1990; Bak and Chen,

1991). The mass of sand in the pile was measured after the addition of each particle.

Variations in pile mass measure the extent of avalanching, if any, caused by each particle

deposited. Results for a 1.5 in. diameter sand pile are shown in Fig. 2 where mass of the

sand pile is plotted versus number of particles dropped. These results show a random

distribution of avalanche sizes for avalanches initiated by single particles. The results in

Fig. 2a for 2000 particle drops (19000 to 21000, box at A) are magnified in Fig. 2b

showing persistence of the random distribution for smaller avalanches. Further

magnification for 100 particle drops (700 to 800, box at B, Fig. 2b) is presented in Fig. 2c

showing some randomness for extremely small avalanches. The avalanching produced

by single particles added to small sand piles exhibits scaling characteristics similar to a

fractal. Sand piles with base diameters of 0.38 and 0.75 in. also showed this behavior.

A similar test of a sand pile with 3 in. base diameter by Held, et al. resulted in

nearly periodic behavior (Fig. 3a), where approximately 10,000 particles were dropped

on the sand pile between avalanches. Fig. 3b shows approximately periodic behavior
measured using a tilt box. Limestone sand was pluviated (36 in. drop) into a 6 in. cubical

box attached to a mechanical stage. During deposition the sand surface was horizontal.

The box was rotated, and as the sand surface approached the first critical slope, rotation

increments were 1/6 deg. (10 minutes). The equivalent plane surface slope angle plotted

as ordinate in Fig. 3b (apparent c) was computed from box dimensions and cumulative

weight of sand discharged by avalanches. The abscissa is box tilt angle. When an

apparent critical angle c was reached, an avalanche ensued. Fig. 3b shows the first and

largest avalanche at A (apparent C = 45.8? deg.), followed by small avalanches at B

(apparent C = 42.5? deg.) and C (apparent C = 42.9? deg.) before another large

avalanche at D (apparent C = 44.7? deg.). The apparent rest slope angles R for

avalanches at A, B, C and D were 39.5?, 41.5?, 42.2? and 40.3? deg., respectively.

GEOMETRY OF SLOPE PROFILES

Finite repose slope profiles are often curved near boundaries as shown by the

profile in Fig. 4 measured for lead shot in a tilt box test. In addition there are local slope

fluctuations along the profile surface. Extreme local surface slopes (nearly vertical) are

shown in Fig. 5a for dense?, crushed limestone sand with elongated angular particles and

in Fig. 5b for elongated particles of rice tested by Frette (1996). Thus, angle of repose

should be measured for a sufficiently long portion of the profile to represent global

geometry, sufficiently far from boundaries to be approximately plane.

Local slope fluctuations form “niches” in the surface that affect slope formation

as particles are deposited. One result is segregation of graded materials. Larger particles

are expected to roll farther down the slope before encountering a niche of sufficient size
to stop rolling (Baumann, et al., 1995). Small avalanches on sufficiently long slopes may

be arrested by particles finding niches before rolling off the slope (Held, et al., 1990).

Thus, avalanches of a given size may be “local” for long slopes (i.e., arrested and

undetected as in Fig. 3) while the same size avalanche will be “global” for short slopes

(unarrested as in Fig. 2).

TESTING APPARATUS

Two types of testing apparatus, the tilt box and cylindrical extrusion apparatus,

were designed to study repose slope profiles. The tilt box produces average plane strain

deformation while the cylindrical extrusion test produces an axisymmetric conical slope

surface.

Tilt Box Test

A photograph of the tilt box apparatus is shown in Fig. 6a. A 152 mm. (6 in.)

cubical box was attached to a rotational mechanical stage. The angular resolution of the

stage (with vernier) was 5 seconds. Beginning with a horizontal specimen surface, the

stage was adjusted manually to tilt the box and specimen surface to a slope somewhat

less than critical. Continued rotation stopped after every 10 minutes of rotation or

whenever an avalanche occurred that reduced the material mass in the box. A collection

pan was positioned on an electronic balance beneath the tilting box to weigh any sand

that fell out. Grid paper glued inside the right and left box walls was used to measure

coordinates along the slope surface profile after every avalanche of sufficient size to

deposit unarrested material in the collection pan. A video camera was positioned to
continuously record material movements and surface profile during testing. This record

could be viewed in slow motion or frame-by-frame to observe the development of

avalanches.

Cylindrical Extrusion Test

The cylindrical extrusion test consists of a cylindrical tube with vertically oriented

axis and a horizontal circular plate that serves as a piston fitted to the inside of the

cylinder (Fig. 6b). A specimen was formed by depositing a layer of material to be tested

inside the cylinder above the piston. When the cylinder was moved downward relative to

the piston, the specimen was extruded to form a conical slope surface. Material falling

outside the cylinder during specimen extrusion was collected in a plastic bag surrounding

the cylinder. Repose slope angles were measured at four evenly spaced locations around

the perimeter of specimens by aligning either angled cardboard cutouts or a hinged

cardboard arm parallel to the portion of the surface assumed to define the repose slope.

Cutouts were cut for each whole degree slope angle. Extrusion tests were also recorded

by video camera.

MATERIALS TESTED

Tilt box and cylindrical extrusion tests were conducted with specimens of Ottawa

sand, a quartz pool filter sand, crushed limestone sand, and lead shot 2. Physical

properties of these materials are listed in Table 1. Specimens of Ottawa sand, filter sand,

and lead shot were reused for subsequent tests without noticeable particle degradation.

2
A crushed granite from Colorado was also tested, but results are not reported because the particle size
(approximately 9.5 mm.) was too large for the testing apparatus size.
Limestone sand was also reused. Some breakage of the brittle elongated particles was

suspected, but not measured. Tests were conducted under conventional laboratory

conditions. Temperature and humidity readings taken over a period of several days after

testing was completed indicated fairly constant values, approximately 72 degrees?. F. and

80 percent?, respectively.

SAMPLE PREPARATION

All loose specimens were deposited by pouring material into a 63.5 mm diameter

cylinder which was slowly raised while keeping the lower end of the cylinder in constant

contact with the continually rising material surface (Fig. 7a, tilt box; and Fig. 7b,

extrusion test). The surface of specimens was made plane by carefully striking off excess

material with a straight edge. Dense specimens were deposited by pluviation, with sand

poured into a No. 10 sieve and falling inside a vertical tube to the specimen surface (Fig.

7c, tilt box; and Fig. 7d, extrusion test). Material was deposited in six lifts using six

tubes varying in length from 76.2 cm to 88.9 cm (2.54 cm. increments) to accommodate

the varying elevation of the specimen surface. For each lift the fall distance from sieve to

specimen surface varied from approximately 91 to 89 cm. Lead shot was poured directly

into the vertical tubes with the No. 10 sieve removed.

Rigid walls of specimen containers and plane specimen surfaces cause

interruptions in particle packing, thereby increasing measured void ratios. All measured

void ratios were adjusted using the equation developed by Hardin (1989),

D10
eadjusted  emeasured  1.6
6( V / A )

(1)
where D10 = particle size for which 10 percent of the specimen is finer; V = specimen

volume (3540 cm3 for tilt box and 2499 cm3 for extrusion specimens); and A = specimen

surface area (1394 cm2 for tilt box and 1023 cm2 for extrusion specimens). Measured and

adjusted initial void ratios for all test specimens are given in Table 2 showing either two

or three repetitions for each test. The adjustment is unimportant for specimens of Ottawa

sand, filter sand, and limestone sand (D10 = 0.625 mm, e = 0.007), but is significant for

lead shot (D10 = 3.12 mm, e = 0.033 to 0.034). These loose and dense void ratios

correspond to relative densities of 0 and 100 percent, respectively.

TILT BOX TEST RESULTS

Avalanche Initiation Zones

Video recordings of tilt box tests were reviewed to locate surface zones where

avalanches were initiated. Figs. 8 and 9 show approximate locations determined by eye

of initiation zones for the first ten avalanches for loose and dense specimens of each of

the four materials tested. With the possible exception of limestone sand, initiation zones

appear to be randomly distributed over the slope surface.

Densification Prior to Avalanching

Rotation of principal stress directions relative to principal axes of fabric caused

densification of loose specimens prior to the first avalanche. Fig. 10 shows a photograph

of the surface of loose Ottawa sand just before the first avalanche. The sand originally at

the upper edge of the cube has settled below the corner of the cube. The original corner

has deformed into a smoothly curved surface. From this photograph it is estimated that
the average volumetric strain may be less than 1 percent. This would correspond to a

change in void ratio e = 0.017, making the average void ratio at initiation of the first

avalanche probably greater than 0.7, still a very loose sand3.

Measured Repose Slope Profiles

The critical repose slope profile prior to initiation of a given avalanche is

compared to the corresponding rest repose slope profile after termination of the same

avalanche in Figs. 11 through 18. In each figure comparisons are made for six selected

avalanches occurring within a given specimen at six different box tilt angles.

Comparisons are shown for loose and dense specimens of Ottawa sand, filter sand,

limestone sand, and lead shot in Figs. 11 through 18, respectively. In the upper left

corner of each figure a plot of equivalent plane surface slope angle versus box tilt angle

(similar to Fig. 3b) shows how selected profiles relate to testing history.

These profiles show local slope variations and boundary effects. However, with

the exception of lead shot, global slopes for central sections of profiles (C and R) are

approximately constant for all avalanches of a given test. Slopes defined by central

profile sections were determined by regression analysis of the measured coordinates, and

values of C and R are plotted versus box tilt angle in Fig. 19 where trend lines are

shown. The trend lines for the three sands are nearly horizontal with some slopes

positive, some negative. Assuming  to be independent of box tilt angle for the sands,

the mean values and standard deviations shown in Table 3 were computed. These data

indicate that C is 2 to 3 degrees greater than R.

3
Accurate measurements of volume change prior to avalanche are planned before submission of this paper
for publication.
 Values of C at initiation of the first avalanche, where C = box tilt angle, are not

plotted in Fig. 19 but are given in Table 3. Values of first avalanche C for sands (based

on one to three measurements) are generally 2 to 4 degrees higher than mean values of C

for succeeding avalanches. This may be expected since the first avalanche is initiated on

a material surface that is made plane by striking with a straight edge; whereas, each

succeeding critical slope surface is formed by the preceding avalanche. A surface formed

by an avalanche has local irregularities that may be expected to trigger the next avalanche

at a lower global slope than for a surface without irregularities. Thus, C for avalanches

succeeding the first may be smaller than’. On the other hand, the change in fabric near

the plane surface produced by striking with a straight edge, may cause C for the first

avalanche to be greater than ’.

In Figs. 20 through 27 rest profile coordinates for all avalanches of a given test

are plotted on the same figure in order to compare global slopes of central profile

sections.

CYLINDRICAL EXTRUSION TEST RESULTS

Cylindrical extrusion tests were conducted on loose and dense specimens of

Ottawa sand, filter sand, limestone sand, and lead shot. As a specimen was extruded

from the test cylinder, successive avalanches continually formed new rest repose slope

surfaces. The extrusion process was stopped four to eight times to observe slope profiles

and measure slope angles for portions of slopes considered to best represent global

angles.
 Measured angles are plotted in Fig. 28 for loose and dense specimens of each of

the four materials. Each diamond symbol represents a measured value of R. Mean

values of R measured using the tilt box test are shown by the large rectangular symbols

(values in Table 3 rounded to the nearest degree). The larger extrusion test angles for a

given material in Fig. 28 are believed to result from measurements on slope lengths that

were too short to define global angles. Tests using a larger extrusion cylinder are

required to check this hypothesis, but for now, the smallest values are assumed to be

more representative of global slopes. Slope lengths required to measure global angles are

expected to increase with particle angularity and particle size. In Fig. 28 the largest

dispersions of measured angles occur for limestone with very angular particles and the

large diameter particles of lead shot; whereas, angles measured for Ottawa sand

approximate values measured in the tilt box test. Based on these results values of R for

the two tests may prove to be about equal.

EXTREMELY LOW-PRESSURE TRIAXIAL COMPRESSION TESTS

Ponce and Bell (1971) and Fukushima and Tatsuoka (1984) have reported triaxial

compression tests conducted at extremely low pressures. The lowest values of effective

minor principal stress ’3 reported for these tests are in the range 1 to 3 kPa (Fig. 29).

The tests by Ponce and Bell (Fig. 29a) show the angle of shearing resistance ’ increasing

rapidly for ’3 < 400 kPa; whereas, the tests by Fukushima and Tatsuoka (Fig. 29b)

indicate that ’ is nearly constant for ’3 < 400 kPa.

Ponce and Bell measured strengths for a rounded to subrounded sand, and

Fukushima and Tatsuoka tested angular sand. These results are compared to critical
angles of repose C for Ottawa sand and filter sand, respectively. Values of C for first

avalanche and mean C for succeeding avalanches able 3) are plotted in Fig. 29, where

dashed lines show possible extensions of triaxial test data that are consistent with

measured angles of repose. The results for dense specimens support the conclusion by

Fukushima and Tatsuoka that ’ is approximately constant for ’3 < 400 kPa. Values of

C for loose specimens are consistent with an increase in ’ of about 4 degrees below 400

kPa. The high values of ’ measured by Ponce and Bell at extremely low pressures are

probably erroneous.

CONCLUSIONS

1. Repose slope profiles produced by laboratory tests are often curved near

boundaries and exhibit local slope variations. Hence, proper measurement of repose

slope angles requires measurement of global slopes sufficiently far from boundaries.

2. Critical repose slope angles rather than rest repose slope angles are expected to

relate directly to ’. The tilt box test can measure C. However, in order to determine ’

from tests like the cylindrical extrusion test that measure R but do not measure C, it is

necessary to first establish the relationship between C and R. The tilt box tests reported

herein indicate that C is 2 to 3 degrees greater than R.

3. Repose slope testing will be most useful as a strength test when it can be used

to measure variation of strength with density. Repose slope void ratio can be controlled

for both tilt box and cylindrical extrusion tests.

values from cylindrical extrusion tests indicate that R for plane and axisymmetric slopes

may be about equal.

5. It is possible to measure C by depositing material in small amounts on large

conical piles, but not possible to control the void ratio. This test may be used to

determine ’ for loose materials only.

6. Values of C reported herein have been compared to values of ’ measured by

triaxial tests under extremely low pressures. The data for dense materials are consistent

with the conclusion that ’ for dense materials is nearly constant for ’3 < 400 kPa, and

that ’ for loose materials increases by about 4 degrees over the pressure range ’3 < 400

kPa.

7. The high values of ’ at extremely low pressures reported by Ponce and Bell

(1971) are probably erroneous.

REFERENCES

Bak, P. and Chen, K. (1991). “Self-organized criticality.” Scientific American, 264(1),