You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/273977292

Changes to particle characteristics associated with the compression of sands

Article  in  Géotechnique · June 2011


DOI: 10.1680/geot.9.P.114

CITATIONS READS

218 1,717

2 authors, including:

Fatin Altuhafi
University College London
18 PUBLICATIONS   714 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The mechanical behaviour of glacial sediments View project

All content following this page was uploaded by Fatin Altuhafi on 29 June 2015.

The user has requested enhancement of the downloaded file.


Altuhafi, F. N. & Coop, M. R. (2011). Géotechnique 61, No. 6, 459–471 [doi: 10.1680/geot.9.P.114]

Changes to particle characteristics associated with the compression


of sands
F. N. A LT U H A F I  a n d M . R . C O O P 

It is commonly accepted that the onset of particle break- Il est généralement reconnu que le début de la rupture
age in sands during compression marks the start of de particules dans le sable, au cours d’une compression,
yielding. Although particle breakage is strongly asso- marque le commencement de la déformation. Bien que la
ciated with the tensile strength of a single soil grain, rupture de particules soit fortement tributaire de la
initial density and initial sample grading have great influ- résistance à la traction d’un grain de sol unique,
ence on the probability of particle breakage. This paper la densité initiale et le calibrage initial de l’échantillon
examines the effect of initial grading and density on the influent dans une grande mesure sur la probabilité d’une
sample behaviour during one-dimensional compression rupture des particules. La présente communication se
for three sands with distinct mineralogies. It was found penche sur les effets du calibrage et de la densité initiaux
that a unique normal compression line is the outcome of sur le comportement de l’échantillon au cours d’une
a large amount of breakage in poorly graded samples compression unidimensionnelle pour trois sables aux
and that by changing the initial grading to a better minéralogies distinctes. On a relevé qu’un fort degré de
graded sample a significant reduction in particle break- rupture dans des échantillons mal calibrés donne lieu à
age is observed, until for very well-graded samples no une ligne de compression normale, et qu’en modifiant le
significant particle breakage can be measured. At this calibrage initial par un échantillon mieux calibré, on
point a difficulty in defining a unique normal compres- obtient une réduction sensible des ruptures de particules,
sion line for the sample was found and a transitional jusqu’au point où, pour des échantillons extrêmement
behaviour was identified. In addition to particle size, the bien calibrés, on ne peut mesurer aucune rupture de
paper examines the changes to some important particle particules significative. On a éprouvé alors une difficulté
characteristics such as particle shape and particle surface dans la définition d’une ligne de compression normale
roughness in an attempt to relate the soil behaviour to unique pour l’échantillon, et l’identification du comporte-
the nature of the microscopic particle damage during ment transitoire. En plus de la granulométrie, la présente
yielding. communication examine également les changements sur-
venant dans certaines caractéristiques importantes des
particules, comme la forme des particules et la rugosité
de la surface des particules, pour tenter d’établir un
rapport entre le comportement du sol et la nature de
l’endommagement de particules microscopiques lors de la
KEYWORDS: microscopy; particle crushing/crushability; sands déformation.

INTRODUCTION each particle (coordination number) and thus the force


The recognition of the role of particle breakage in defining transferred to each particle through a force chain developed
the constitutive behaviour of granular materials has imposed in the sample during loading (e.g. Muir Wood, 2008). The
major changes to the way in which the macro-scale behav- particle size distribution is of key importance in defining the
iour of these materials is interpreted (e.g. Coop & Lee, packing and hence the void ratio of the sample. The wider
1993; McDowell & Bolton, 1998; Hyodo et al., 2002). the particle size range in a single sample, the greater the
Defining the factors controlling particle breakage during packing efficiency, resulting in a higher coordination number
compression or shearing is therefore a crucial issue to estab- for larger particles and thus lowering the probability of
lish a strong foundation for a constitutive model able to breakage, as demonstrated by discrete element method
describe the deformation of sands. Particle breakage is (DEM) studies (Muir Wood, 2006, 2008).
strongly associated with the tensile strength of the particles Detailed statistical data presented by Nakata et al. (1999;
of a single soil, although there are often other factors influ- 2001) investigated the degree of crushing of individual
encing the probability of particle breakage (Bolton, 1986; particles inserted in a silica sand matrix and subjected to
Nakata et al., 1999; McDowell, 2002). A real soil matrix one-dimensional compression. This showed that catastrophic
also consists of grains with different gradings, usually with splitting of particles takes place in a uniformly graded
different tensile strengths depending on their size, with sample between the yield stress identified from a steepening
smaller grains exhibiting higher strengths than larger ones of the compression curve in the e–log v graph and the
(McDowell & Bolton, 1998; Nakata et al., 2001; McDowell, point at which the compression index reaches a maximum,
2002). The packing of soil particles in a granular assembly and that 50% of the particles had undergone major splitting
is also fundamental in defining the number of contacts for immediately after yielding. However, for a well-graded sand
there was a distinct difference in the crushing behaviour,
with an increase in the breakage of asperities and surface
Manuscript received 16 February 2009; revised manuscript accepted
15 April 2010. Published online ahead of print 17 September 2010.
grinding as the main features of the damage to the larger
Discussion on this paper closes on 1 November 2011, for further particles, whereas particle splitting was more dominant in
details see p. ii. the smaller particles.
 Geotechnics Section, Department of Civil and Environmental In this paper, data are presented which highlight particle
Engineering, Imperial College London, London. damage during the compression of sands. Changes to parti-

459
460 ALTUHAFI AND COOP
cle size through breakage are investigated for sands with The effect of the initial grading on the susceptibility to
distinct mineralogies, different initial gradings and different breakage of particles during shearing and compression has
initial densities to examine the effects of particle packing on also been noted by Altuhafi et al. (2006) in their tests on a
their behaviour. Particle shape changes either due to break- subglacial sediment from Iceland. In its natural state the soil
age or changes to surface roughness are also investigated, was very well graded and fractal as a result of the intensive
microscopy being used to provide accurate and quantitative deformation under the glacier. It was found that its behav-
data to give a better idea of the changes taking place at the iour during compression did not follow the typical pattern of
particle scale. having a unique NCL, which is the common feature during
compression of sands, but a transitional behaviour was seen,
with different compression paths for different initial densities
EVOLUTION OF PARTICLE SIZE DISTRIBUTION DUE (Altuhafi et al., 2010). When a poorly graded sample of the
TO BREAKAGE same soil was tested the behaviour reverted to a usual sand
The progress of particle breakage during the compression behaviour with significant particle breakage and a unique
and shearing of granular soils has been investigated by many NCL. This difference of behaviour for different gradings
authors. It was confirmed at an early stage that particle prompted the more systematic investigation of the effect of
breakage was associated with the compression of even the initial grading presented here.
strongest particles such as silica sands, and is usually
initiated after reaching the yield stress as identified from the
e–log  v9 plot (Coop & Lee, 1993; Hyodo et al., 2002; SOILS TESTED
McDowell, 2002). Tests on Dog’s Bay carbonate sand, by Different types of sands were tested and images of each are
Coop & Lee (1993), showed that there were unique and shown in Fig. 1. Leighton Buzzard sand (LBS) is a silica
linear relationships between relative breakage Br , as defined sand which consists of strong and regular highly spherical
by Hardin (1985), and the logarithm of mean effective stress particles. Dog’s Bay sand (DBS) is a biogenic carbonate sand
(log p9) for both the isotropic normal compression line with fragile particles, mainly composed of broken shell frag-
(NCL) and critical state line (CSL). This suggested that a ments making this sand very crushable. The third soil was a
unique relationship might exist between particle breakage basaltic sand from the Langjökull glacial sediment (GBS),
and effective stress independently of the initial density, and which was previously tested by Altuhafi et al. (2006; 2010).
for the critical states independently of the stress path taken. Each sample was subjected to high stresses in one-dimen-
DEM studies on crushable materials (Bolton et al., 2008), sional compression using two types of oedometer cells: a
showed the evolution of fragmentation of DEM agglomerates 38 mm diameter cell with a fixed ring was used in testing
during compression of a uniformly graded sample. The the carbonate sand up to 30 MPa vertical stress and a
results showed an initial phase of asperity damage, which is 20 mm diameter floating ring oedometer was used to reach
followed by a grain-splitting phase that coincides with the very high stresses up to 107 MPa. The use of a floating ring
material yielding. At high stresses, however, both modes of minimises the effect of boundary friction on the vertical
grain damage can be expected for all grain sizes. stress. The initial void ratio of the sample was calculated
McDowell & Bolton (1998) showed that the linearity of from the initial dry weight and initial dimensions of the
the NCL in granular soils is associated with the evolution of sample. The final void ratio was also monitored from the
the soil grading and they emphasised the role of the final water content for those tests which were carried out
successive fracture of the smallest particles under increasing under saturated conditions (DBS and GBS). For the tests on
macroscopic stress. Considering a compression test on a LBS, which were carried out dry to 107 MPa, the samples
uniformly graded sample, as the stress increases more fines had interlocked particles after the test which enabled meas-
are created by breakage, resulting in a more efficient pack- uring their final dimensions with a reasonable accuracy. The
ing and thus a denser sample. Since the larger particles have final void ratio could then be calculated from the final
higher coordination numbers due to the large number of weight and final dimensions of the sample. The final void
neighbouring finer particles, breakage is less likely for them ratios could then be compared with those calculated from
and so breakage is predominant in the smaller particles with the initial void ratio and the measured volumetric strain, as
lower coordination numbers. This effect was found to out- shown in Tables 1–3. Generally it was assumed that a
weigh that of increasing strength with decreasing particle difference of 0.01 in void ratio was acceptable.
size (McDowell & Bolton, 1998). Each soil was prepared for testing by dry sieving to
Particle breakage cannot, however, be a never-ending pro- separate the soil into its constituent particle size ranges.
cess and, at high stresses, a constant grading should eventually Samples of different gradings for each type of sand were
be reached (McDowell & Bolton, 1998; McDowell, 2002). then created by mixing soil from each sieve interval in the
For many granular materials, it was found that the limiting proportion required for a specific grading, ensuring that a
grading during compression is a fractal distribution of particle homogeneous mixture of all the sizes was achieved in each
sizes with a fractal dimension of around 2.5–2.6 (Sammis et sample. The initial particle size distribution was then
al., 1987; McDowell et al., 1996; McDowell & Bolton, 1998). checked using the QicPic apparatus, which will be described
This is broadly in agreement with the results presented by in the following section. The initial particle surface rough-
Coop et al. (2004) of ring shear tests on Dog’s Bay sand, ness before the tests was measured for the silica sand by
which indicated that also during shearing, at very large selecting 20 particles randomly from each sample, although
strains, as particle breakage progressed the contact stresses the particles needed to be larger than about 300 ìm to be
between the particles were eventually too low to cause any able to make an accurate measurement. The surface rough-
further breakage and a constant grading was again reached. ness measurement was made over an area of 20 3 20 ìm for
Coop et al. showed evidence that the final grading curves may each particle by using the interferometer, which will also be
be different for different stress levels and different initial described below.
gradings, but at higher stress levels there was a tendency
towards a fractal distribution with a fractal dimension of about
2.57. Changing the particle contacts by remoulding the sam- THE QICPIC APPARATUS
ple, however, can result in more breakage when further The QicPic apparatus is a laser image analysis instrument
shearing is applied to the sample. with a manufacturer’s stated capacity of measuring particle
CHANGES TO PARTICLE CHARACTERISTICS ASSOCIATED WITH THE COMPRESSION OF SANDS 461
Table 1. Initial and final void ratio values by different methods
1000 µm of calculations – tests on Dog’s Bay sand. (The first part of the
test ID symbol refers to the initial grading used in the test,
while the extension refers to the test number.)

Test ID Initial e Final e† Final e

D1-26 1.671‡ 0.419 0.417


D1-27 1.896 0.418 0.420
D1-28 1.499 0.404 0.408
D2-23 1.600 0.721 0.722
D2-24 1.267 0.689 0.692
D2-25 0.886 0.684 0.678
D3-4 2.110 0.377 0.441
D3-5 1.852 0.423 0.423
D3-6 2.308 0.440 0.440
D3-7 1.625 0.399 0.406
D3-8 2.071 0.426 0.433
(a)
D3-9 1.859 0.391 0.400
CD-16 2.202 0.448 0.455
CD-17 2.754 0.677 0.682
CD-19 2.088 0.550 0.556
1000 µm CD-20 1.885 0.423 0.424
CD-21 2.781 0.730 0.731
CD-22 2.102 0.591 0.593
D4-10 1.097 0.565 0.569
D4-11 1.143 0.624 0.631
D4-12 0.922 0.473 0.474
D4-13 0.692 0.405 0.413
D4-14 1.362 0.633 0.635
D4-15 0.849 0.427 0.433
 Void ratio calculated from final water content of retrieved sample.
† Void ratio calculated from initial void ratio and volumetric strain.
‡ Void ratio after initial loading, this sample experienced high
reduction in e at seating pressure due to its highly loose nature.

Table 2. Initial and final void ratio values by different methods


(b) of calculations – tests on silica sand. (The first part of the test
ID symbol refers to the initial grading used in the test, while
the extension refers to the test number.)

1000 µm Test ID Initial e Final e† Final e

G1-S1 0.770 0.296 0.317


G1-S2 0.622 0.280 0.290
G1-S3 0.587 0.283 0.305
G2-S1 0.490 0.246 0.252
G2-S2 0.365 0.230 0.241
G2-S3 0.395 0.242 0.246
CG-S2 0.858 0.186 –
CG-S4 0.448 0.145 –
CG-S5 0.319 0.210 0.215
CG-S6 0.226 0.170 0.189
CG-S7 0.390 0.198 0.219
OC-S4 0.260 0.218 0.220
OC-S5 0.323 0.247 0.258
OC-S6 0.449 0.256 0.275
OC-S7 0.237 0.237 0.249
(c)
 Void ratio calculated from final sample weight and final height of
Fig. 1. Sands used in this study: (a) Dog’s Bay sand (DBS); retrieved sample.
(b) Leighton Buzzard sand (LBS); (c) glacial basalt sand (GBS) † Void ratio calculated from initial void ratio and volumetric strain.

sizes between 1 ìm and 20 mm. It uses dynamic image laser and the camera detector can operate up to speeds of
analysis by examining a flow of moving particles. This 500 frames/s, allowing a very large number of particles to be
allows a large sample size to be considered, the particles to considered in a short duration. Two dispersing units were
be randomly orientated and the occurrence of overlapping used with the QicPic depending on particle size. The
particles to be reduced. The dispersing unit creates a well- GRADIS is a dry gravity feeding system which is used for
dispersed flow of particles that falls through the scanning particles with sizes between 0.05 and 20 mm and the
beam emitted by a pulsed laser. An exposure time of less LIXELL is a water circulation system which is used mainly
than 1 ns ensures that motion blur is not detectable. The for particles less than 100 ìm in size.
462 ALTUHAFI AND COOP
Table 3. Initial and final void ratio values by different methods example of Fig. 3. The surface image can also be analysed
of calculations – tests on basalt. (The first part of the test ID for absolute height measurements and roughness.
symbol refers to the initial grading used in the test, while the The vertical resolution is around 1 nm for each pixel, but
extension refers to the test number.) the vertical resolution of the image depends also on the Z
step size, which has to be chosen between a few nanometres
Test ID Initial e Final e† Final e
to 100 nm. This step size has a direct influence on the
BG1-T1 0.982 0.331 0.335
acquisition and calculation times, memory use in the perso-
BG1-T2 0.840 0.324 0.319 nal computer (PC) as well as the resolution. In the present
BG1-T3 0.770 0.311 0.300 study a step size of 15.7 nm was used, thus limiting the
BG2-T1 0.641 0.286 0.283 accuracy of the measurements made in this study to half this
BG2-T2 0.590 0.343 0.313 value, to around 8 nm.
BG2-T3 0.491 0.299 0.304 A number of statistical parameters are available in the
BCG-T1 0.663 0.281 0.273 software (Microsurf 3D). These calculations are either made
BCG-T2 0.507 0.235 0.228 on a selected surface or a cut section, which was chosen to
BCG-T3 0.433 0.194 0.207 be 20 3 20 ìm for this study. The main parameters for
 Void ratio calculated from final water content of retrieved sample. surface roughness are Sa , which is the arithmetic mean of
† Void ratio calculated from initial void ratio and volumetric strain.
the deviation from the mean height value, and Sq , which is
the square root of the arithmetic mean of squared deviation
from the mean height value (Sacerdotti et al., 2000)
X !0:5
The QicPic apparatus has an accompanying software Sq ¼ 1=mn Z 2ij (1)
known as WINDOX, which controls the settings of the ij
dispersing unit as well as storing and manipulating the X 
measurements made. Measurements for each individual par- Sa ¼ 1=mn  Z ij  (2)
ticle can be evaluated from its image, examples of which ij
are given in Fig. 2, and this enables a comprehensive analy-
sis of size and shape to be made. where m and n are the number of points in the x and y
There are several particle size definitions available within directions, and Z is the deviation at each point from the
the WINDOX software. The first is termed EQPC and is the mean height value. To avoid either particle size or shape
diameter of the equivalent circle with the same area as that influencing the values, both Sq and Sa are calculated relative
of the particle. Several Feret diameters can also be calcu- to a smoothed surface rather than a flat plane. This was
lated, but it was found that for most soils the Feret minimum done by the same software and by specifying the size of
gives the closest size to a sieve analysis and so this has been shape motifs for the designated areas, which was 5.02 ìm
used throughout the current paper. It is the minimum for the area chosen in this study.
distance between two parallel lines which touch the particle
on opposite sides in a two-dimensional image. The software
is able to calculate particle size distributions based on THE INFLUENCE OF GRADING ON THE NCL
volume or area; the former was chosen here for compatibil- The previous literature on one-dimensional compression
ity with normal soil mechanics practice of using sieved tests on uniformly graded samples of silica sand has shown
weights. that for a given void ratio samples with smaller particles
The aspect ratio (AR) in this software is the ratio between reach yield at higher stresses than those with larger particles
Feret minimum and Feret maximum diameters, while the (Nakata et al., 2001; McDowell, 2002), although McDowell
sphericity S is calculated as the ratio of the equivalent circle (2002) noted that following yield the compression index was
perimeter, PEQPC , to the real perimeter, Preal. The sphericity independent of particle size. However, the uniformity of the
value obtained from the QicPic actually equals the square sample has a much more significant effect on the compressi-
root of circularity as defined by the International Standard bility index of the NCL and the yield point of the material,
ISO/DIS (2006) (ISO, 2006; Cavarretta et al., 2009). Con- as was shown by Coop & Atkinson (1993). To demonstrate
vexity C is also available in the software and it describes the this effect more systematically, three samples of uniformly
compactness of a particle as calculated from the ratio of the graded DBS D1 with an initial mean particle size of 362 ìm
projected particle area to the gross area including any re- were one-dimensionally compressed to 30 MPa. The results
entrant sections. shown in Fig. 4(a) clearly define a unique NCL for this
grading with higher yield stresses for denser samples. Here
the yield stress is defined as the point of maximum curva-
THE INTERFEROMETER ture on the e–log  v9 plot. As shown in Figs 5 and 6(a),
The surface roughness of sands has been studied recently similar results were obtained from compression tests to
using some more accurate and complex methods such as 107 MPa, on uniformly distributed samples of the GBS and
fractal geometry, the fuzzy uncertainty texture spectrum LBS, both of which initially consisted of a single size of
(Lee et al., 1998), structural three-dimensional approaches particles between 1 and 2 mm. A unique NCL can be again
(Hong et al., 1999), SURFASCAN 3D (Content & Ville, identified for both the BG1 and G1 gradings.
1995) and photometric stereo acquisition with gradient space From the change in grading after compression for each
domain mapping (Smith, 1999). However, the optical profilo- sample for the D1 grading of DBS (Fig. 7(a)), it can be seen
metry technique is perhaps the most convenient for surface that there is more breakage for the sample with the higher
analysis. An optical view of the sample is converted to an initial void ratio. Contrary to the hypothesis of Coop & Lee
elevation map using interferogram processing. From a single (1993) the amount of breakage is therefore not unique for a
scan the height of each pixel is determined independently given stress level for a soil that has reached its NCL, but is
relative to each other with an accuracy of 1 nm in the white dependent also on the initial density of the specimen. This
light scanning mode (Fogale, 2005). The output topography might be attributed to the differences in coordination number
may be presented as an image for which the height at every in these samples. In samples with higher densities the
individual point is coded on a colour or grey scale as in the coordination number would tend to be higher, decreasing the
EQPC 1419·093 µm EQPC 1176·862 µm EQPC 552·432 µm EQPC 869·659 µm EQPC 533·738 µm
Sphericity 0·909 Sphericity 0·898 S 0·721 S 0·870 S 0·735
Aspect ratio 0·708 Aspect ratio 0·763 AR 0·358 AR 0·830 AR 0·439
Convexity 0·980 Convexity 0·971 C 0·920 C 0·940 C 0·884
Elongation 0·561 Elongation 0·560
EQPC 283·037 µm
EQPC 378·631 µm EQPC 635·509 µm
S 0·719
S 0·866 S 0·822
EQPC 1209·198 µm EQPC 1312·834 µm AR 0·316
AR 0·658 AR 0·584
Sphericity 0·830 Sphericity 0·863 C 0·903
C 0·943 C 0·931
Aspect ratio 0·526 Aspect ratio 0·645
Convexity 0·959 Convexity 0·947 EQPC 1268·014 µm EQPC 472·206 µm EQPC 362·356 µm
Elongation 0·291 Elongation 0·343 S 0·887 S 0·828 S 0·763
AR 0·885 AR 0·551 AR 0·545
C 0·960 C 0·933 C 0·878

EQPC 1218·171 µm EQPC 366·408 µm EQPC 525·138 µm


Sphericity 0·868 EQPC 1393·630 µm S 0·851 S 0·805
Aspect ratio 0·636 Sphericity 0·915 AR 0·614 AR 0·486
Convexity 0·960 Aspect ratio 0·867 EQPC 1243·367 µm C 0·929 C 0·931
Elongation 0·360 Convexity 0·973 S 0·922
Elongation 0·423 AR 0·794 EQPC 376·676 µm
C 0·983 S 0·629
EQPC 474·287 µm
AR 0·269
S 0·707
EQPC 1162·078 µm C 0·825
AR 0·809
Sphericity 0·882
EQPC 100·902 µm C 0·810
Aspect ratio 0·637 EQPC 519·721 µm
Convexity 0·980 Sphericity 0·891 EQPC 770·232 µm
Aspect ratio 0·848 S 0·790
Elongation 0·466 AR 0·555 S 0·774
Image number 7509 Convexity 0·973 AR 0·584
Elongation 0·376 C 0·925
C 0·919

(a) (b)

Fig. 2. Examples of particle shapes from the QicPic output for the LBS compressed to 30 MPa. EQPC is the diameter of the equivalent circle with the same area as
that of the particle. S, AR and C are the sphericity, aspect ratio and convexity respectively: (a) before compression; (b) after one-dimensional compression to 30 MPa
CHANGES TO PARTICLE CHARACTERISTICS ASSOCIATED WITH THE COMPRESSION OF SANDS
463
464 ALTUHAFI AND COOP
XY: 58·27 µm ⫻ 52·12 µm
X: 7·283 µm/div, Y: 7·283 µm/div
A significant reduction in the inclination of the NCL can
Z: 16·25 µm also be seen in all the less uniformly distributed samples
Z: 4·062 µm/div D2, BG2 and G2 for DBS, GBS and LBS respectively, as
can be seen in Figs 4(b), 5 and 6(a).
Changing the initial grading of DBS samples to the next
better-graded one D3 again results in a reduction of break-
age after compression, as indicated by the relatively small
Y shift in the final gradings curves (Fig. 7(b)). However, here
X
again it can be seen that more particle breakage occurs in
samples with higher initial void ratios. For this grading,
some scatter in the compression curves starts to be observed,
although they still appear to converge at higher stresses.
Again the NCL becomes flatter and the yield points are less
pronounced.
A fractal grading with fractal dimension, D of 2.57 was
then created for DBS, such that the percentage by mass
Fig. 3. Topographic image of the surface of a LBS particle as M(L , d ) of particles finer than size d is given by the
obtained by the interferometer relation
M ð L , d ÞÆd 3 D (3)
probability of breakage (McDowell & Bolton, 1998). The This grading is the final grading determined by Coop et al.
decrease in the inclination of the NCL at the highest (2004) which evolved at very large strains from shearing
pressures for the uniformly graded sample of DBS (Fig. DBS of an initial grading 300–425 ìm in the ring shear
4(a)) and LBS (Fig. 6(a)) is similar to that seen by apparatus under 650–930 kPa vertical stress. If plotted using
McDowell & Bolton (1998) and discussed by McDowell a logarithmic percentage finer scale this grading would plot
(2005), who related this phenomenon to reaching the com- as a straight line of gradient 3  D, that is 0.43. On Fig.
minution limit for particles, at which the finest particles no 7(b) it is labelled CD for ‘critical grading of Dog’s Bay
longer break but deform plastically. sand’. In this case the gradings of the samples after com-
For the LBS and GBS all the gradings analyses were pression to 30 MPa were similar to the initial one before
carried out using both the sieve and Qicpic. In Figs 8–11 a compression, indicating that no significant breakage took
number of comparisons are shown with sieve data which place during loading. The initial gradings curves for CD and
show significant differences arising from the different techni- the final ones for tests CD-16 to CD-22 overlie each other.
ques. Apart from these few examples, only the QicPic data It has previously been assumed that the NCL of a granular
are shown because, although there were differences in the material is a function of particle breakage and that every
absolute grading, the sieve data confirmed the conclusions material has a unique NCL, the location of which is
regarding the quantity of breakage that occurred. The final dependent on its grading, among other factors. This raises
grading of a compression test which was performed on an important question, which is whether soils that are
grading G1 to 30 MPa vertical stress (G1–30 MPa) is also naturally well graded, or those that have reached their
shown in Fig. 9, showing less breakage than the tests which terminal or fractal grading through compression or shearing,
reached 107 MPa as expected. Unless otherwise indicated can have a unique NCL because no further breakage is
the tests on GBS and LBS were all loaded to a vertical possible. The answer can be found in Fig. 4(d) which shows
stress of 107 MPa. the compression curves for the tests conducted using the
The effect of the initial void ratio on the amount of fractal grading CD. They form a set of parallel curves which
breakage is less pronounced for both BGS and LBS. How- do not seem to converge to a unique line even at higher
ever, small differences in the final grading can still be stresses. The position of each curve is dependent on the
distinguished, indicating that more breakage is still asso- initial density of the sample, an indication of transitional
ciated with samples with higher initial void ratios, as shown behaviour, a form of behaviour which has been seen pre-
in Figs 8(a) and 9. viously in some residual and gap-graded soils (Martins et
Packing efficiency and thus the coordination number can al., 2001; Ferreira & Bica, 2006; Shipton et al., 2006), some
also be increased by increasing the fines content of the well-graded silts (Nocilla et al., 2006) and recently on well-
sample. Changing the grading of DBS to D2 and of GBS graded glacial tills (Altuhafi et al., 2010). A particularly
and LBS to BG2 and G2, as shown in Figs 7(a), 8(b) and interesting feature is that even if the compression curves are
10 respectively, significantly reduces the particle breakage as flatter at this grading, there is still a large plastic volumetric
indicated by the final particle size distributions. Here again strain that occurs, as can be seen from the unloading curves,
the final gradings for DBS are different, with the looser and yet there has been no measurable particle breakage.
samples exhibiting more breakage than the denser ones, Since the soil is composed only of non-plastic particles, this
which is again less clear in other soils. It is particularly means that although some measurable particle breakage is
interesting to note that the final gradings curves for the commonly associated with plastic volumetric compression in
better-graded soil D2 actually lie below those of the uni- sands, there is no dependent link between the plastic strains
formly graded D1 in the case of DBS, so that as a result of and breakage. In their as-yet-unpublished work on ring shear
compression to 30 MPa the gradings curves have actually tests on DBS, Ferreira & Coop showed that if a soil that had
crossed. Similarly, it can be seen from Fig. 4(b) that the reached its terminal grading was reconstituted, additional
NCLs also cross, and that at the final stress it is the initially breakage would be seen, indicating that the cessation of
better-graded soil that has the higher void ratio and is now breakage is dependent on maintaining the fabric of the soil
less well-graded than the initially uniform soil. From the as it evolves during the test. This seems to be in contrast to
compression curves of these samples in Fig. 4(b), the yield the result here, where reconstitution of an estimated terminal
stresses are clearly higher than for the tests performed on grading was found not to result in any breakage on compres-
grading D1, with the looser sample of D2 starting to yield at sion. However, the strains involved in these compression
about 1200 kPa compared to about 800 kPa for grading D1. tests are very much less than the enormous strains of
CHANGES TO PARTICLE CHARACTERISTICS ASSOCIATED WITH THE COMPRESSION OF SANDS 465
NCL, Cc ⫽ 0·75 2·0
D1-26
D1-27 D2-23
2·0 D1-28 NCL for D1 D2-24
1·6 D2-25

1·6

Void ratio, e
1·2
Void ratio, e

1·2
0·8
0·8
0·4
Cc ⫽ 0·48
0·4
0
0 10 100 1000 10 000 100 000
Vertical stress: kPa
10 100 1000 10 000 100 000
(b)
Vertical stress: kPa
(a)

D3-4 2·0 CD-16


2·0
NCL for D1 D3-5 CD-17
D3-6 CD-19
1·6 D3-6 1·6 CD-20
D3-7 CD-21
D3-8 CD-22
Cc ⫽ 0·40 D3-9
Void ratio, e
Void ratio, e

1·2 1·2

0·8 NCL for D2 0·8

0·4 0·4

0 0

10 100 1000 10 000 100 000 10 100 1000 10 000 100 000
Vertical stress: kPa Vertical stress: kPa
(c) (d)

2·0
D4-10
D4-11
1·6 D4-12
D4-13
D4-14
Void ratio, e

1·2 D4-15

0·8

0·4

0
10 100 1000 10 000 100 000
Vertical stress: kPa
(e)

Fig. 4. Compression curves for tests on DBS: (a) grading D1; (b) grading D2; (c) grading D3; (d) grading CD; (e) grading D4

thousands of percent required even to see a small change of the samples were reconstituted from the various gradings as
grading in the ring shear tests. separated by sieving. After reconstitution, part of the sample
The final series of samples of DBS was created with a was then sieved again to check the accuracy of its grading
higher fractal dimension than the critical grading by adding before the test. No significant differences were found. In
more fines to the samples to create D4 with a fractal contrast to what was found for the DBS, the grading results
dimension at around 2.74 (Fig. 7(b)). This again flattened obtained from the QicPic for samples of both LBS and GBS
the compression curves even more and enhanced the transi- showed some divergence from those obtained from the
tional behaviour (Fig. 4(e)). Again the gradings were un- sieves and the gradings that were chosen to be fractal based
changed after compression of the samples, and are omitted on sieves are therefore not fractal when based on the QicPic
from Fig. 7 for clarity. As for the DBS a ‘critical’ or fractal data. The grading OC for the LBS was also reconstituted as
grading with a fractal dimension of 2.57 was again tested a fractal grading with a dimension of 2.73 and so was even
for GBS and LBS, and these are denoted BCG and CG in better graded. Similarly to what was found for the DBS, a
Figs 8(a) and 11 respectively. decrease in the NCL gradient and a general increase in yield
The reconstitution method for all the samples tested in stress can be observed as the grading becomes less uniform.
this study was based on sieves, as discussed earlier, since This is again associated with a significant reduction in
466 ALTUHAFI AND COOP
1·0 0·8
NCL BG1 NCL-G1 G1-S1
Basalt G1-S2
BG1-T1 G1-S3
BG1-T2 0·6 ⫹ ⫹ ⫹ G2-S1
BG1-T3 G2-S2
0·8
BG2-T1 ⫹ G2-S3

Void ratio, e
⫹ ⫹ NCL-G2
BG2-T2 ⫹ ⫹
BG2-T3 0·4 ⫹

NCL BG2 BCG-T1 ⫹
BCG-T2 ⫹
0·6 ⫹
BCG-T3
0·2
Void ratio

0·4 0
0·1 1 10 100 1000
Vertical stress: MPa
(a)
0·2

0·8

0 CG-S1
CG-S2
0·01 0·1 1 10 100 1000 0·6 CG-S4
Vertical stress: MPa CG-S5

Void ratio, e
CG-S6
Fig. 5. Compression curves for tests on GBS samples CG-S7
0·4 NCL-CG

0·2
particle breakage. Again the well-graded soil BCG for GBS
showed no tendency to converge to form a unique NCL,
indicating a clear transitional behaviour (Fig. 5). A unique 0
NCL can still be identified for the CG grading of LBS (Fig. 0·1 1 10 100 1000
6(b)), but there is a noticeable scatter in the compression Vertical stress: MPa
curves, which perhaps indicates the onset of transitional (b)
behaviour which is then evident for the OC grading (Fig.
6(c)). The very well-graded samples of GBS and LBS show
no significant particle breakage, as shown in Figs 8(a) and
11 respectively, although there is a little scatter in the data 0·4
for the CG and OC of LBS gradings.
The QicPic particle shape analyses for the tests on GBS 0·3
and LBS in Figs 8(a) and 9 showed that in both cases
Void ratio, e

particles of the uniformly graded samples BG1 and G1 were


initially highly spherical, highly convex and with aspect 0·2
ratios of around 0.8–0.9. The new particles created during OC-S4
compression of G1 for the LBS are less spherical, less 0·1 OC-S5
OC-S6
convex and with lower aspect ratios than the original parti- OC-S7
cles. Similar results were obtained for the particle shape
0
analyses for the tests on the uniformly graded samples of
GBS as shown in Fig. 8(a). The decrease in aspect ratio for 0·1 1 10 100 1000
this sand is less pronounced than for the LBS, and for the Vertical stress: MPa
(c)
very finest particles created by breakage the sphericity starts
to increase again. For the uniformly graded sample of LBS Fig. 6. One-dimensional compression curves for tests on LBS:
which was compressed to only 30 MPa, the new smaller (a) gradings G1 and G2; (b) grading CG; (c) grading OC
particles show a slightly lower sphericity and aspect ratio,
which might be an indication that these shape factors tend
to increase as the compression continues. However, in all Owing to the similarity of results for the CG and OC
cases the coarser particles remain of similar shape, even gradings of LBS, the particle shape analysis data are pre-
though many of the particles are likely to have been sented only for tests on the CG grading. The analysis in Fig.
damaged by breakage, since the gradings curves move 11 shows that the larger particles experienced no significant
significantly in the coarser fraction. The shape factor data shape changes similar to what has been found for BCG
for below 60 ìm were very scattered and appeared to be grading of GBS (Fig. 8(a)). However, even if there was very
unreliable, so these data have been omitted in the analysis little, if any, breakage evident in the gradings curve, some
presented here. small changes were observed for the shape of the fine
As for the DBS, the breakage after the compression tests fraction of the LBS. For the critical grading BCG of GBS,
on the better-graded samples BG2 and G2 was significantly however, it seems that there is no clear change in shape.
lower than for BG1 and G1 as indicated by the grading Because of the highly irregular shape of the DBS particles,
curves shown in Figs 8(b) and 10 respectively. The figures which can be clearly observed from the microscope image
also show that the finer particles of G2 after the test are in Fig. 1, the shape and surface roughness analyses were not
again less spherical with lower aspect ratio than the untested undertaken, as the data would be too scattered to be useful.
sample of LBS, but the changes for GBS are less pro- To give a simple quantitative description for the unifor-
nounced. Again for both soils the coarser particles remain mity of the sample, the relative distribution factor (RD ) has
substantially unchanged in shape. been used here, which is the ratio between d90 and d10 .
CHANGES TO PARTICLE CHARACTERISTICS ASSOCIATED WITH THE COMPRESSION OF SANDS 467
100 0·92
D1 grading

Sphericity
D1–26
80 D1–27 0·88
D1–28
Percentage finer: %

D2 BCG
60 D2–23 0·84
D2–24 0·9
D2–25

Aspect ratio
40
D1 initial
0·7
20 BG1
D2
initial 0·5
0 0·92

Sphericity
1 10 100 1000
Particle size: µm 0·88
(a)
BG1
0·84
100 100
CD BG1-untested
CD-16 BG1-T1
80

Percentage finer: %
CD-17 BG1-T2
80
CD-19 BG1-T3
Percentage finer: %

CD-20 BG1-T1, sieve


60
CD-21 D4 BCG-untested
60
CD-22 BCG-T1
D4 40 BCG-T2
D3 BCG-T3
40 CD D3-4 20
D3-5
D3-6
20 D3-7 0
D3-8
D3-9 1 10 100 1000 10 000
0 Particle size: µm
1 10 100 1000 (a)
Particle size: µm
(b) 0·9
Aspect ratio

Fig. 7. Initial and final gradings for DBS samples which were 0·8
compressed to 30 MPa: (a) gradings D1 and D2; (b) gradings 0·7
D3, CD and D4
0·6
BG2
0·5
Table 4 shows the value of RD for each initial grading of 0·92
DBS compared with both the coefficients of uniformity and 0·90
Sphericity

curvature. The RD value tends to be higher for a well-graded


sample and approaches a value of 1 for a sample which 0·88
consists only of one size of particles. The clear dependency 0·86
of the particle breakage on the uniformity of the initial BG2
grading is highlighted in Fig. 12(a), which shows the 0·84
relationship between the relative breakage defined by Hardin 100
(1985) and illustrated in Fig. 13, and RD for the initial BG2-untested
sample grading. The figure clearly illustrates that the nar- 80 BG2-T1
Percentage finer: %

rower the distribution of the particle size, the greater the BG2-T2
BG2-T3
particle breakage after compression. The curves drawn here
60
are based on the mean values of relative breakage for the
samples of the same initial grading curve. Individual values
40
for each test for DBS are also shown to highlight the
dependency of the amount of breakage on the initial void
ratio of the DBS samples. Note that this figure is based on a 20
maximum vertical stress of 30 MPa for DBS and 107 MPa
for GBS and LBS. The figure also shows the effect of 0
grading on the gradient of the NCL (Cc ) and it is interesting 1 10 100 1000 10 000
Particle size: µm
to see that for DBS, there is approximately a linear relation- (b)
ship between Cc and the logarithmic value of the initial RD .
Here, the results of tests on an intermediate grading of DBS Fig. 8. Initial and final grading curves and shape factors for
between D1 and CD, with an RD value of 100, are also tests on GBS: (a) for test on BG1 and BCG; (b) for tests on
included, but the detailed data were omitted from previous BG2
figures for brevity.
The effect of the initial void ratio or packing of the DBS
samples on the final breakage can also be derived from these the highest breakage values, whereas those tests which were
data as shown in Fig. 12(b). An approximately linear performed using the fractal grading (RD ¼ 217) exhibited
relationship can be seen between the initial void ratio and zero breakage values lying on the x-axis. The effect of the
the final breakage for each individual initial grading of this initial void ratio on breakage values for samples of GBS and
sand. The line for the more uniformly graded samples LBS were not as pronounced and were more scattered than
(RD ¼ 1.2) can be seen at the top of this figure indicating for the DBS, but small effects can be seen in Figs 8–11.
468 ALTUHAFI AND COOP
1 0·8

Aspect ratio
Convexity

0·9
0·7
0·8

0·7 0·6
0·8 0·9
Aspect ratio

Sphericity
0·7
0·8
0·6

0·5 0·7
1 100
CG-untested
0·9 CG (sieve)

Percentage finer: %
80
Sphericity

CG-S4
0·8 CG-S5
60 CG-S6
0·7 OC-untested
40 OC-S6
OC-S7
0·6
100 20 G2

80
Percentage finer: %

G1 before test 0
G1-S1 (sieve)
1 10 100 1000
60 G1-S1
Particle size: µm
G1-S3
40 G1-30 MPa
Fig. 11. Initial and final grading curves and shape factors for
tests on LBS with initial gradings CG and OC
20
G1
initial
0
1 10 100 1000
Particle size: µm

Fig. 9. Initial and final grading curves and shape factors for Table 4. The relative distribution factor and values of the
tests on LBS with initial grading G1 coefficient of uniformity Cu and coefficient of curvature Ccu
for each initial grading of DBS

Grading ID Cu Ccu RD

D1 1.24 0.88 1.2


D2 5.32 1.35 6.8
D3 11.7 1.39 19.6
0·8 D4 55.5 2.68 100
CD 72.66 2.14 216
Aspect ratio

0·7

The reason for the more distinct effect for the DBS could be
0·6 that it was possible to create a greater range of initial void
0·9
ratios in this soil.
Sphericity

CHANGES TO PARTICLE SURFACE ROUGHNESS


The particle surface roughness was measured for each
sample by measuring the Sa value on 20 3 20 ìm areas of
0·85 the surfaces of selected particles. Although the particles
100
G2-untested were randomly selected, they had to have a size of more
G2-S1 than about 300 ìm to be able to make the measurement, so
Percentage finer: %

80
G2-S2
G2-S3
the data reflect changes to particle surface roughness of the
60 G2-S1 (sieve) coarser end of the grading in each case. Owing to the high
angularity of the GBS and DBS, surface roughness analyses
40 were not possible for these sands using existing techniques.
Fig. 14 shows the particle surface roughness for the untested
20
LBS with an initial grading between 1 and 2 mm (G1),
0 along with measurements which were made after (a) one-
1 10 100 1000
dimensional compression to 30 MPa and (b) one-dimensional
Particle size: µm compression to 107 MPa. In sample G1–30 MPa care was
taken to examine unbroken particle surfaces for the rough-
Fig. 10. Initial and final grading curves and shape factors for ness evaluation. The examination of the unbroken surfaces
tests on LBS with initial grading G2 might highlight any minor damage that was occurring to the
CHANGES TO PARTICLE CHARACTERISTICS ASSOCIATED WITH THE COMPRESSION OF SANDS 469
Data from 70
0·7 D1
DBS mean values 60
Relative breakage

0·6 LBS mean values


GBS mean values 50 Before test

Frequency: %
0·5 Data from DBS individual values Mean after G1-S1 and G1-S2
40
0·4 D2
After G1–30 MPa
Data from 30
0·3
D3
0·2 20
Data from
0·1 CD 10
0 0
NCL gradient, Cc

0–49

50–99

100–149

150–199

200–249

250–299

300–349

350–399

400–449

450–499
0·2 10 100 1000
Initial RD ⫽ (D90/D10)
0·4

0·6 Surface roughness, nm, Sa


(a)
0·8

Percentage change in surface roughness: %


(a)
20 Mean value
Individual tests of OC grading
0·7 Individual tests of GC grading
D90/D10 ⫽ 1·2 Individual tests of G2 grading
0·6 D90/D10 ⫽ 6·8 Individual tests of G1 grading
D90/D10 ⫽ 19·6 0
Relative breakage

0·5 D90/D10 ⫽ 100


D90/D10 ⫽ 217
0·4
0·3 ⫺20
0·2
0·1
0 ⫺40

0·4 0·8 1·2 1·6 2·0 1 10 100 1000


RD ⫽ D90/D10
Initial void ratio
(b)
(b)

Fig. 12. (a) Changes in the amount of breakage after compres- Fig. 14. Surface roughness measurements for LBS: (a) particle
sion and the compression index with relative distribution for surface roughness before and after tests on LBS with uniform
DBS, LBS and GBS. (b) Effect of packing (initial void ratio and grading G1 after one-dimensional compression to 30 MPa or
relative distribution) on relative breakage for DBS 107 MPa; (b) percentage changes in average surface roughness
after one-dimensional compression to 107 MPa for tests on LBS
with different initial gradings

A B
100
pressed to 107 MPa and a reduction in the number of
smooth particles (with an Sa value between 0 and 50 nm)
can be seen, resulting in a slight increase in the mean value
80
Percentage passing: %

of surface roughness from 81 nm before testing to 91 nm


0.074 mm
after, although this result may reflect the effect of using the
60
different method of selecting the particle surfaces.
Surface roughness measurements were also made on the
40 G1, G2, CG and OC gradings as shown in Fig. 14, which
D Initial grading
shows the change in surface roughness after compression
20 tests to 107 MPa against the logarithm of the relative
C After test
distribution factor (RD ). The percentage change in surface
0 roughness has been calculated
0·01 0·10 1·00 10·00
Particle size: mm Percentage change of surface roughness
Total breakage, Bt ⫽ area BCDB (4)
Breakage potential, Bp ⫽ area BCAB
S a ð f Þ  S a ð iÞ
¼ 3 100
Relative breakage, Br ⫽ Bt /Bp S a ð iÞ
Fig. 13. Schematic diagram showing the breakage parameters
proposed by Hardin (1985) where Sa ( f ) is the mean value of surface roughness after
testing and Sa (i) is the mean value of surface roughness of
the untested material. The results of each individual test are
shown as well as the mean value of tests for each individual
larger particles. However, a difficulty was encountered in grading. As discussed above, there was an increase in sur-
identifying undamaged surfaces for those samples which face roughness for those tests which experienced a signifi-
were compressed to 107 MPa because of the intense degree cant amount of particle breakage (grading G1), which might
of breakage. For all tests at 107 MPa, it could not be indicate that the broken surfaces of these particles had
ensured that the surfaces examined were the original sur- higher roughness values than the original particles. In con-
faces of the particle, and some may be new broken surfaces. trast, a decrease was seen for tests conducted on G2, which
The surface roughness shows no significant change after exhibited less breakage than G1, and a similar change was
one-dimensional compression to 30 MPa, but the change observed for those samples which experienced no breakage
started to be more pronounced when the sample was com- at all (CG and OC gradings). Again this might be an
470 ALTUHAFI AND COOP
indication that with no or minimal particle breakage the that the local damage at particle contacts is not affecting the
plastic deformation of the soil mass is permitted by a overall particle shape even if it might alter the surface
rearrangement of particles which is associated with an texture (Cavarretta et al., 2010). It is also interesting to
abrasion that decreases the surface roughness for this notice that apart from the creation of these fines there was
sand. Alternatively the particle surface roughness might be no other significant change to the grading.
reduced by plastic deformation of the asperities, and
Cavarretta et al. (2010) have observed that significant plastic
strain can occur at particle contacts through the flattening of CONCLUSIONS
asperities as a process of plastic deformation rather than The results discussed in this paper clarify some important
breakage. aspects related to compression in sands. First, a unique NCL
is related to large amounts of particle breakage in poorly
graded sands. A well-graded sample will have a lower com-
PRE-YIELD BEHAVIOUR pression index and less clear yield point and exhibits less
To investigate changes to particle shape characteristics breakage in comparison to a uniformly graded one. With the
during one-dimensional compression in the pre-yield region, reduction of particle breakage for well-graded sands, a
samples were examined after a series of tests on the difficulty in defining the NCL is encountered, and a transi-
uniformly distributed LBS (G1) which was subjected to a tional behaviour evolves as the particle breakage reduces to
maximum vertical stress of 3 MPa. This value of stress was levels that are not detectable. However, it is clear that
far below the yield stress for this grading, which is around significant plastic volumetric compression can still occur in
10 MPa, as can be seen from Fig. 6(a). The tests included well-graded sands even in the absence of significant break-
one monotonically loaded sample and one cyclically loaded age. Another important conclusion is the dependency of
sample in which the sample was unloaded to 47.2 kPa and particle breakage on the initial density of the sample and the
re-loaded for ten cycles. A third test was conducted again fact that for a specific stress on the NCL the particle break-
with ten similar loading cycles, but reconstituting the sample age is not unique.
between cycles. This was done to ensure that the particle The results also show that different gradings result in
orientations and contacts were changed between each cycle. different types of deformation. For uniformly graded samples
The QicPic results for the monotonically loaded sample after many particles suffer catastrophic splitting, creating smaller
the test indicated no significant change in either the grading particles which are less spherical, less convex and with
or the shape characteristics of the sample. A similar result lower aspect ratio than the original particles. In these
was also obtained for the cyclically loaded sample as shown uniformly graded samples some increase in particle surface
in Fig. 15. However, around 0.63% of fines were produced roughness was also noted. The larger particles of a well-
during the third test in which the sample was remoulded graded sample will suffer from either asperity breakage and/
between cycles. Although the presence of the fines was an or perhaps flattening of asperities through plastic straining,
indication of damage to the sand particles at the contacts, and there will be a reduction in surface roughness, but there
the shape parameters remained unchanged, which suggests will be little, if any, particle splitting. It is probably this very
small amount of abrasion that permits the volumetric com-
pression to occur in those samples which otherwise show no
0·9 significant particle breakage.
Sphericity
The results also support the idea of critical or terminal
0·8 grading of a granular material, at which the particle size
distribution cannot evolve any more and no further breakage
Shape factor

0·7 Aspect
can be expected. This may well be a fractal grading as has
ratio been suggested previously.
0·6

ACKNOWLEDGEMENTS
0·5 The authors would like to express their gratitude to Mr
1000 1200 1400 1600 1800 2000 Antonio Borges Rodriguez and Miss Bai Xue for kindly
Particle size: µm providing some of the data for this research. The authors
would also like to thank Professor Peter Sammonds from the
Earth Sciences Department, University College London, for
100 Untested material providing the glacial basalt which was used in this study.
Cyclic-reconstituted The authors are also grateful to Dr Ignazio Cavarretta for
Cyclic his invaluable help and to Mr Alan Bolsher for manufactur-
80 Monotonic
ing the high-pressure oedometer.
Percentage finer: %

60
NOTATION
40
AR aspect ratio
Br Hardin’s relative breakage
C convexity
20 Cc inclination of the NCL in e–log  v9 graph
Ccu coefficient of curvature
Cu coefficient of uniformity
0 CSL critical state line
100 1000 10 000 D fractal dimension
Particle size: µm DBS Dog’s Bay sand
d size of particle
Fig. 15. Changes to grading and shape factor after monotonic, d10 particle size corresponding to 10% of the grading
cyclic and cyclic with reconstitution tests on LBS d90 particle size corresponding to 90% of the grading
CHANGES TO PARTICLE CHARACTERISTICS ASSOCIATED WITH THE COMPRESSION OF SANDS 471
EQPC diameter of the equivalent circle with the same area as that Hardin, B. O. (1985). Crushing of soil particles. J. Geotech. Engng
of the particle 111, No. 10, 1177–1192.
e sample void ratio Hong, H. K., Myung, Y. C. & Choi, J. S. (1999). 3-D analysis of
GBS glacial basalt sand projective tesxtures using structural approaches. Pattern Recog-
LBS Leighton Buzzard sand nition 32, No. 3, 357–364.
M percentage by mass Hyodo, M., Hyde, A. F. L., Aramaki, N. & Nakata, Y. (2002).
m, n number of points in the x and y directions Undrained montonic and cyclic shear behaviour of sand under
NCL normal compression line low and high confining stresses. Soils Found. 42, No. 3, 63–76.
PEQPC equivalent circle perimeter ISO (International Organization for Standardization) (2006). Repre-
Preal perimeter of particle two-dimensional projection sentation of results of particle size analysis. Part 1–6. Part 6:
p9 effective pressure descriptive and quantitative representation of particle shape and
RD relative distribution factor morphology. Draft International Standard ISO/DIS 9276. Geneva:
S sphericity ISO.
Sa , Sq surface roughness parameters Lee, Y. G., Lee, J. H. & Hsueh, Y. C. (1998). Texture classification
Sa ( f ) mean value of surface roughness after test using fuzzy uncertainty texture spectrum. Neurocomputing 20,
Sa (i) mean value of surface roughness of untested material No. 1, 115–122.
Z deviation at each point from the mean height value Martins, F. B., Bressani, L. A., Coop, M. R. & Bica, A. V. D.
 v9 vertical effective stress (2001). Some aspects of the compressibility behaviour of a
clayey sand. Can. Geotech. J. 38, No. 16, 1177–1186.
McDowell, G. R. (2002). On the yielding and plastic compression
REFERENCES of sand. Soils Found. 42, No. 1, 139–145.
Altuhafi, F., Baudet, B. A. & Sammonds, P. (2006). Particle break- McDowell, G. R. (2005). A physical justification for e–log  based
age in glacial sediments. In Geomechanics and geotechnics of on fractal crushing and particle kinematics. Géotechnique 55,
particulate media (eds M. Hyodo, H. Murata and Y. Nakata), No. 9, 697–698, doi: 10.1680/geot.2005.55.9.697.
pp. 21–24. London: Taylor & Francis. McDowell, G. & Bolton, M. (1998). On the micromechanics of
Altuhafi, F. N., Baudet, B. A. & Sammonds, P. (2010). The mech- crushable aggregates. Géotechnique 48, No. 5, 667–679, doi:
anics of subglacial sediment: an example of new ‘transitional’ 10.1680/geot.1998.48.5.667.
behaviour. Can. Geo. J. 47, No. 7, 775–790. McDowell, G. R., Bolton, M. D. & Robertson, D. (1996). The
Bolton, M. D. (1986). The strength and dilatancy of sands. Géo- fractal crushing of granular materials. J. Mech. Phys. Solids 44,
technique 36, No. 1, 65–78, doi: 10.1680/geot.1986.36.1.65. No. 12, 2079–2102.
Bolton, M. D., Nakata, Y. & Cheng, Y. P. (2008). Micro- and Muir Wood, D. (2006). Geomaterials with changing grading: A
macro-mechanical behaviour of DEM crushable materials. Géo- route towards modelling. In Geomechanics and geotechnics of
technique 58, No. 6, 471–480, doi: 10.1680/geot.2008.58.6.471. particulate media (eds M. Hyodo, H. Murata and Y. Nakata),
Cavarretta, I., O’Sullivan, C. & Coop, M. R. (2009). Applying 2D pp. 313–325. London: Taylor & Francis.
shape analysis techniques to regular materials with 3D particle Muir Wood, D. (2008). Critical state and soil modelling. In Defor-
geometries. Powders and grains. Proc. 6th Int. Conf. Micromech. mational characteristics of geomaterial (eds S. E. Burns, P. W.
Granular Media (eds M. Nakagawa and S. Luding), 833–836. Mayne and J. C. Santamarina), pp. 51–72. Amsterdam, The
New York: American Institute of Physics. Netherlands: IOS Press.
Cavarretta, I., Coop, M. R. & O’Sullivan, C. (2010). The influence Nakata, Y., Hyde, A. F. L., Hyodo, M. & Murata, H. (1999). A
of particle characteristics on the behaviour of coarse grained probablistic approach to sand particle crushing in the triaxial
soils. Géotechnique 60, No. 6, 413–423, doi: 10.1680/geot.2010. test. Géotechnique 49, No. 5, 567–583, doi: 10.1680/geot.1999.
60.6.413. 49.5.567.
Content, P. & Ville, J. F. (1995). Surfacscan 3D – An industrial 3D Nakata, Y., Hyde, A. F. L., Hyodo, M., Kato, Y. & Murata, H.
surface texture characterisation instrument. Int. J. Mach. Tools (2001). Microscopic particle crushing of sand subjected to high
Mf. 35, No. 2, 151–156. pressure one-dimensional compression. Soils Found. 41, No. 1,
Coop, M. R. & Atkinson, J. H. (1993). The mechanics of cemented 69–82.
carbonate sands. Géotechnique 43, No. 1, 53–67, doi: 10.1680/ Nocilla, A., Coop, M. R. & Colleselli, F. (2006). The mechanics of
geot.1993.43.1.53. an Italian silt: an example of ‘transitional’ behaviour. Géotechni-
Coop, M. R. & Lee, I. K. (1993). The behaviour of granular soils at que 56, No. 4, 261–271, doi: 10.1680/geot.2006.56.4.261.
elevated stresses. In Predictive soil mechanics (eds B. G. S. Parry Sacerdotti, F., Griffiths, B., Benati, F. & Kang, H. (2000). Measmt
and R. H. G. Parry), pp. 186–198. London: Thomas Telford. Sci. Technol. 11, No. 3, 171–177.
Coop, M. R., Sorensen, K., Freitas, T. B. & Georgoutsos, G. (2004). Sammis, C., King, G. & Biegel, R. (1987). The kinematic of gouge
Particle breakage during shearing of a carbonate sand. Géotechni- deformation. Pure Appl. Geophys. 125, No. 5, 777–812.
que 54, No. 3, 157–163, doi: 10.1680/geot.2004.54.3.157. Shipton, B. J. I., Coop, M. R. & Nocilla, A. (2006). Particle
Ferreira, P. M. & Bica, A. V. D. (2006). Problems in identifying the breakage in transitional soils. In Geomechanics and geotechnics
effects of structure and critical state in a soil with transitional of particulate media (eds M. Hyodo, H. Murata and Y. Nakata),
behaviour. Géotechnique 56, No. 7, 445–454, doi: 10.1680/geot. pp. 143–147. London: Taylor & Francis.
2006.56.7.445. Smith, M. L. (1999). The analysis of surface texture using photo-
Fogale (2005). Fogale nanotech user manual version 1.5. Nimes, metric stereo acquisition and gradient space domain mapping.
France: Fogale. Image Vis. Comput. 17, No. 14, 1009–1019.

View publication stats

You might also like