Characterizing Surface Roughness and Shape of Sands

Using Digital Microscopy

Khalid A. Alshibli, M.ASCE,1 and Mustafa I. Alsaleh, A.M.ASCE2

Abstract: This paper presents detailed microscopic analyses of the surface roughness, roundness, and sphericity of sands. The surface
roughness of three silica sands was studied using the optical interferometry approach. It was statistically characterized by a set of
Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

parameters. Optical interferometry yielded very accurate measurements of surface roughness of the three sands. It was found that, as the
surface roughness increases, the friction and dilatancy angles of the sand increase. In addition, two new indices for particle roundness and
sphericity are introduced, compared with the Powers classification, and used to classify the investigated sands.
DOI: 10.1061/共ASCE兲0887-3801共2004兲18:1共36兲
CE Database subject headings: Surface roughness; Shape; Sand; Granular materials; Friction; Microscopy; Digital techniques.

Introduction sphericity may be defined as the ratio of the surface area of a

Shape, surface roughness, and gradation of particles have a sig-
nificant influence on the strength and deformation properties of
granular materials. Sands with a predominance of angular par-
ticles possess greater friction than those consisting mainly of
rounded particles. Koener 共1968兲 investigated the effects of angu-
larity, gradation, and mineralogy on the shear strength of cohe-
sionless soils and found that the angle of internal friction in-
creases with increasing angularity of particles, and decreases with
increasing effective size. Sedimentologists generally express par-
ticle shape in terms of surface texture, roundness, and sphericity.
Surface texture is used to describe the surface of particles 共e.g.,
polished, greasy, frosted, etc.兲 that are too small to affect the
overall shape. Roundness refers to those aspects of grain surface
共sharpness of corners and edges兲 that are on a larger scale than
those classed as surface texture, but that are smaller than the
overall dimensions of the grain. Sphericity is used to describe the
overall form of the particle irrespective of the sharpness of edges
and corners. It is a measure of the degree of conformity of particle
shape to that of a sphere.
Wadell 共1932兲 was the first to point out that the terms shape
and roundness are not synonymous, but rather include two geo-
metrically distinct concepts. He defined roundness as the ratio of
the average of radii of corners of the grain image and the maxi-
mum radius of the inscribed circle. There are also many other
scale definitions for the roundness 共e.g., Russell and Taylor 1937;
Pettijohn 1949; Powers 1953, 1982兲. Wadell 共1932兲 was also the
first to choose the sphere as a standard. Ideally, the property of
1
Joint Assistant Professor, Dept. of Civil and Environmental Engineer-
ing, 3507 CEBA Building, Louisiana State Univ.–Southern Univ., Baton
Rouge, LA 70803. Email: Alshibli@lsu.edu
2
Graduate Student, Dept. of Civil and Environmental Engineering,
Louisiana State Univ., Baton Rouge, LA 70803. Email:
MALSAL1@lsu.edu
Note. Discussion open until June 1, 2004. Separate discussions must
be submitted for individual papers. To extend the closing date by one
month, a written request must be filed with the ASCE Managing Editor.
The manuscript for this paper was submitted for review and possible
publication on September 10, 2002; approved on April 21, 2003. This
paper is part of the Journal of Computing in Civil Engineering, Vol. 18,
No. 1, January 1, 2004. ©ASCE, ISSN 0887-3801/2004/1-36 – 45/$18.00.
sphere of the same volume as the fragment in question to the
actual surface area of the grain.
Recently, Masad et al. 共2001兲 introduced the surface texture
index 共TI兲 in which they referred to the fast Fourier transform; a
good correlation between TI and the rutting resistance for hot-mix
asphalt was found in their study. In 1998, Grigoriev et al. studied
the surface texture at the nanoscale level and concluded that it
affects the contact behavior between two surfaces. James and
Vallejo 共1997兲 defined the roughness as the general shape and
surface irregularity; they emphasized that roughness is an impor-
tant characteristic that affects the mass behavior of the soil.
The definition of particle shape in terms of sphericity and
roundness is widely accepted. However, methods have not been
standardized because of the tedious task of making numerous
readings. Analyses of sphericity and roundness are often made
visually. Pettijohn 共1949兲 supplemented his roundness classes
with detailed descriptions so that the particle can be classified
visually. Referring to the work of Russell and Taylor 共1937兲 and
Pettijohn 共1949兲, Powers 共1953兲 introduced a new roundness
scale for sedimentary particles. He emphasized that roundness
does not depend on the particle’s shape; instead, it depends on the
sharpness of the edges. In his work, Powers introduced the sphe-
ricity terminology to describe the shape as well as the roundness
index. In 1982 Powers modified his chart to include more classes
for sphericity and he assigned index numbers to the different
roundness and sphericity classes 共Fig. 1兲. Table 1 shows the
roundness grades proposed by Russell and Taylor 共1937兲 and Pet-
tijohn 共1949兲 and Table 2 shows the grades introduced by Powers
共1953兲.
The surface roughness has been studied lately using more ac-
curate and complicated methods such as fractal geometry,
the fuzzy uncertainty texture spectrum 共Lee et al. 1998兲, struc-
tural three-dimensional 共3D兲 approaches 共Hong et al. 1999兲,
SURFASCAN 3D 共Content and Ville 1995兲, and photometric ste-
reo acquisition and gradient space domain mapping 共Smith 1999兲.
In this paper, the results of particle surface measurements
using optical interferometry are presented. The measurements
were performed on three silica sands. This paper also presents a
literature summary of particle roundness/sphericity quantification
along with introducing new roundness and sphericity indices.

36 / JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004

J. Comput. Civ. Eng., 2004, 18(1): 36-45

Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Comparison between Powers classification, I sph , and I R values

Surface Roughness Measurements specific coordination is measured based on the phase data 共i.e.,
pixel value兲 and the wavelength of the source light. Based on the
white light reflection at the surface and on the reflected wave-
Materials and Methods
length, the height of every single point 共pixel兲 on the sand particle
The surface roughness of three different sizes of silica sand was surface can be measured using the WOI in the VSI mode. The
measured using an optical surface profiler called the Wyko optical resolution of the system in the VSI mode was 3 nm; the system
interferometer 共WOI兲 manufactured by the Veeco Metrology was calibrated using known standards traceable by the National
Group 共1999兲. It is a noncontact optical profiler that uses two Institute of Standards and Technology 共NIST兲.
technologies to measure a wide range of surface heights 共rough- Three sands were used in this study to represent fine-,
ness兲. Figs. 2 and 3 show a schematic and digital photographs of medium-, and coarse-grained silica sands with different surface
the WOI, respectively. It has two operating phases: the phase roughness and shapes. The fine-grained sand is uniform silica
shifting interferometry 共PSI兲 mode allows measurements of 共quartz兲 sand with mean particle size (d 50) of 0.22 mm. It was
smooth surfaces and steps, while the vertical scanning interferom- obtained from the Ottawa Industrial Silica Company and is com-
etry 共VSI兲 mode measures rough surfaces and steps up to milli- monly denoted as F-75 Ottawa sand 共banding sand, herein labeled
meters high. In this study, the VSI mode was used since it gives as F sand兲. The medium-grained sand 共herein labeled as M sand兲
better resolution for the three sands used in the investigation. In was obtained from Unimin Corporation. It is industrial uniform
the VSI mode, a white light source is used since it works best for white quartz sand termed grade no. 30 with 0.55 mm mean par-
vertical scanning interferometry. The height of the surface at a ticle size. The third sand is crushed silica sand obtained from the
Connecticut Silica Company 共herein labeled as C sand兲 with
mean particle size of 1.6 mm.
Table 1. Roundness Grades According to Russell and Taylor and
Pettijohn Classifications 共Powers 1953兲
Russell and Taylor Pettijohn Table 2. Powers Roundness Classification 共Powers 1953兲
Arithmetic Class Geometric Classification Class intervals Geometric mean
Classification Class limit midpoint limit midpoint
Very angular 0.12–0.17 0.14
Angular 0–0.15 0.075 0.0–0.15 0.15 Angular 0.17–0.25 0.21
Subangular 0.15–0.3 0.225 0.15–0.25 0.20 Snbangular 0.25–0.35 0.30
Subrounded 0.30–0.50 0.40 0.25–0.40 0.315 Subrounded 0.35–0.49 0.41
Rounded 0.50–0.70 0.60 0.40–0.60 0.50 Rounded 0.49–0.70 0.59
Well rounded 0.70–1.0 0.85 0.60–1.0 0.80 Well rounded 0.70–1.0 0.84

JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004 / 37

J. Comput. Civ. Eng., 2004, 18(1): 36-45

Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Schematic of the Wyko optical interferometer

Samples of the three sands were randomly mounted on thin
glass slides and then gold coated. Particle gold coating is essential
in scanning electron microscopy 共SEM兲 and VSI because it
causes electron beam and light reflection for better image viewing
in both systems. A total of 120 particles for each of the sands
were scanned. They were arbitrarily selected from each slide. To
avoid errors due to the particles’ surface curvature and edges only,
the center part of each particle is used in roughness calculations
as illustrated in Fig. 4共a兲. The WOI is supplemented with com-
puter software called Vision 32 关Fig. 4共b兲兴, in which a series of
mathematical algorithms are executed for each scan to estimate
different roughness indices. The reference mean line is taken au-
tomatically, and the overall roughness is given according to the
image taken for each particle. This approach was not used in our
study since particle edges are included in the calculations, and
thus sharp edges will show up, causing a sudden drop or rise in

Fig. 3. The LSU Wyko optical interferometer

38 / JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004

J. Comput. Civ. Eng., 2004, 18(1): 36-45

Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. 共a兲 Illustrative image of sand particle captured using Wyko optical interferometer method; 共b兲 particle surface roughness measurements
acquired using Vision 32 software

Fig. 5. Flowchart of processes used to calculate particles’ surface roughness

JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004 / 39

J. Comput. Civ. Eng., 2004, 18(1): 36-45

Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Scanning electron microscopy images of the three sands at different magnification levels

the height 共represented by the pixel value兲. To avoid this problem,

after the scans were completed, an ASCII file for each image was
generated using Vision 32 software. Then, MathCad software was
used to process the generated ASCII files and to calculate the
following roughness indices for the three sands. Fig. 5 summa-
rizes the sequence of processes used to quantify the particles’
surface roughness.
• The average roughness (R a ) is the arithmetic mean of the
absolute values of the surface departure from the mean plane.
M N
1
Ra⫽ 兺兺
MN i⫽1j⫽1
兩Zij兩 (1)

where M and N⫽numbers of pixels in X and Y directions;

Z⫽surface height at a specific pixel relative to the reference
mean plane; and Z i j is defined as
Zij⫽␭Lij (2)
where ␭⫽wavelength used in the scan; and L⫽wave value for
specific coordinates at the particle surface. R a is usually used
to describe the roughness of a finished surface, so it can be
used to describe the roughness of the sand particle surface.
The main disadvantage that might be encountered here is that,
with average roughness, the effect of a single spurious, non-
typical peak or valley will be averaged out and have only a
small influence on the overall roughness. Thus this index or
average will give no information about the shape of irregulari-
ties or the surface of the particle. For granular materials and
particle-to-particle friction, R a represents, to some extent, the
overall roughness used for friction calculations.
• The root mean square roughness (R q ) is calculated as follows:

Rq⫽ 冑 1
M N i⫽1
M

兺 j⫽1
兺 Z 2i j
N

(3) Fig. 7. Typical optical microscope profiler images for the sands used
in the investigation
This index represents the standard deviation of the surface

40 / JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004

J. Comput. Civ. Eng., 2004, 18(1): 36-45

 Table 3. Statistical Summary of Roughness Parameters for F Sand
Parameter
Statistic R a (␮m) R q (␮m) R p (␮m) R v (␮m) R t (␮m) R sk R ku
Minimum 0.2620 0.3168 0.6149 0.6149 1.2298 ⫺1.1334 1.2102
Maximum 2.9116 3.1390 6.0046 6.0046 12.0092 1.0325 3.6862
Mean 0.8487 0.9815 1.9631 1.9631 3.9263 ⫺0.1539 2.0603
Median 0.7075 0.8316 1.7691 1.7691 3.5383 ⫺0.1275 2.0012
Root mean square 1.0105 1.1569 2.2466 2.2466 4.4932 0.6096 2.1047
Standard deviation 0.5510 0.6153 1.0977 1.0977 2.1953 0.5927 0.4322
Standard error 0.0538 0.0600 0.1071 0.1071 0.2142 0.0578 0.0422
Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

heights. R q has the same disadvantage as R a . The only advan- • In addition, SEM images were taken of the same sands to look
tage of R q compared to R a is due to the fact that the heights at the surface at high magnification and see the roughness of
are squared here, and thus the valleys and peaks have more the grain surface 共Fig. 6兲. To some extent, the SEM images
significance in R q . showed good uniformity in the surface roughness for each of
• The maximum profile peak height (R p ) is the distance be- the three sands.
tween the highest point of the surface and the mean surface for
the entire data set. Therefore, it represents the peak value for
the surface. Results
• The maximum profile valley depth (R v ) is the distance be- As mentioned earlier, 120 particles were randomly selected and
tween the lowest point of the surface and the mean surface for analyzed to quantify surface roughness. The ASCII files that con-
the entire data set 共i.e., it measures the depth of the valley of a tain the scan results were exported to MathCad software for fur-
given surface兲. ther analysis and calculations, including calculating the roughness
• The maximum height of the surface (R t ) is the vertical dis- indices defined in the previous subsection and generating 3D ren-
tance between the lowest and highest points on the surface derings of the particles’ surface profiles 共Fig. 7兲. Tables 3–5 give
a statistical summary of the analyses conducted on F, M, and C
Rt⫽Rp⫹Rv (4) sands, respectively. Furthermore, Fig. 8 shows the frequency dis-
• Skewness (R sk) measures the asymmetry of the surface about tribution of the average surface roughness (R a ) and the root mean
the mean plane; it is the mean-cubed roughness square roughness (R q ) for the three sands used in the investiga-
M N tion. Comparing the statistical summaries of the three sands, one
1
Rsk⫽ 兺 j⫽1
兺 Z 3i j
M NR 3q i⫽1
(5)
can see that the statistical parameters for R q are slightly higher
than those for R a for each of the three sands. Furthermore, the
mean values for R a and R q are 0.8487 and 0.9815 ␮m for the F
The advantage of this index is that, if two different surfaces sand, respectively, compared to 0.9460 and 1.0822 ␮m for the M
have the same R a and R q values, one can distinguish between sand, and 1.1169 and 1.2383 ␮m for the C sand. Fig. 8 shows that
them using the skewness. Valleys in the surface will yield the frequency distributions for R a and R q differ slightly from each
negative R sk whereas peaks will give positive ones. other for the same sand and their standard deviations are ⬃0.55–
• Kurtosis (R ku) measures the peakedness of the surface about 0.61 ␮m for the F sand, ⬃0.61–0.67 ␮m for the M sand, and
the mean plane. R ku is calculated using the following expres- ⬃0.50–0.52 ␮m for the C sand.
sion: A series of biaxial 共plane strain兲 experiments were performed
on the three sands using prismatic specimens that measure 83.3
1
M N wide⫻80.8 long⫻152.4 high. Loose 共relative density D r
Rku⫽ 兺兺
M NR 4q i⫽1 j⫽1
Z 4i j (6) ⫽39– 66%) and dense specimens (D r ⫽88– 100%) were tested
under 15 and 100 kPa confining-pressure conditions 关see Alshibli
It is mostly used for machined surfaces. R ku will register high and Sture 共2000兲 for more details about the experiments兴. Fig. 9
values when a high proportion of the surface falls within a shows the relation between surface roughness 共represented by the
narrow range of heights. mean R a value兲 and peak friction angle (␾ p ) and dilatancy angle

Table 4. Statistical Summary of Roughness Parameters for M Sand

Parameter
Statistic R a (␮m) R q (␮m) R p (␮m) R v (␮m) R t (␮m) R sk R ku
Minimum 0.3538 0.3997 0.5387 0.5387 1.0774 ⫺1.3663 1.1555
Maximum 2.8653 3.1205 5.8650 5.8650 11.7299 1.1276 2.9372
Mean 0.9460 1.0822 1.9176 1.9176 3.8351 ⫺0.2900 1.9272
Median 0.7165 0.8130 1.6201 1.6201 3.2403 ⫺0.3910 1.9149
Root mean square 1.1259 1.2685 2.1404 2.1404 4.2808 0.7315 1.9666
Standard deviation 0.6141 0.6657 0.9563 0.9563 1.9127 0.6754 0.3938
Standard error 0.0655 0.0710 0.1019 0.1019 0.2039 0.0720 0.0420

JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004 / 41

J. Comput. Civ. Eng., 2004, 18(1): 36-45

 Table 5. Statistical Summary of Roughness Parameters for C Sand
Parameter
Statistic R a (␮m) R q (␮m) R p (␮m) R v (␮m) R t (␮m) R sk R ku
Minimum 0.6515 0.7737 1.1823 1.1823 2.3646 ⫺1.2507 1.1416
Maximum 2.6958 2.9494 4.2120 4.2120 8.4239 1.0594 2.5140
Mean 1.1169 1.2383 2.0519 2.0519 4.1037 ⫺0.3373 1.6537
Median 0.9639 1.0414 1.9138 1.9138 3.8276 ⫺0.4367 1.6619
Root mean square 1.2227 1.3423 2.1663 2.1663 4.3325 0.7790 1.6805
Standard deviation 0.5033 0.5241 0.7029 0.7029 1.4058 0.7105 0.3023
Standard error 0.0768 0.0799 0.1072 0.1072 0.2144 0.1084 0.0461
Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

共␺兲. Some of the specimens did not reach the critical state, which
made it difficult to calculate the critical state friction angle; there-
fore, peak friction angles were used in the analysis. It is well
known that the frictional resistance of granular materials is
mainly attributed to interparticle sliding resistance and particle
rearrangements/interlocking 共dilatancy effects兲. Higher peak fric-
tion angles are expected when sand specimens are tested under
biaxial loading conditions compared to conventional triaxial com-
pression. Fig. 9 shows that ␾ p and ␺ increase as surface rough-
ness increases for all cases. Furthermore, dense specimens have
higher ␾ p and ␺ values compared to loose specimens tested under
the same confining pressure. ␾ p and ␺ values are also higher for
specimens tested at low confining pressure 共15 kPa兲 compared to
those with similar densities tested at high confining pressures
共100 kPa兲. Finally, it seems there is a consistency in the trend of
the rate of increase of ␾ p with surface roughness 共i.e., the slope
and shape of the curve兲 for specimens tested at the same confining
pressure value.

Fig. 8. Frequency distribution of R a and R q for the sands used in the investigation

42 / JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004

J. Comput. Civ. Eng., 2004, 18(1): 36-45

 Particle Roundness and Sphericity

The shape/roundness of granular particles can be evaluated using

different indices, such as the form factor 共FF兲, which is widely
used in the literature and is defined as
4␲A
FF⫽ (7)
P2
where A⫽projected area of the particle; and P⫽perimeter of this
area.
Masad et al. 共2001兲 proposed the following definitions to
quantify the aggregate shape characteristics.
• The sphericity parameter 共SP兲 is used in the literature to define
Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

the angularity of particles

SP⫽ 冑
3

d L2
d sd I
(8)

where d L ⫽longest particle dimension; d I ⫽intermediate par-

ticle dimension; and d s ⫽shortest particle dimension.
• The shape factor 共SF兲
ds
SF⫽ (9)
冑d L d I
Furthermore, Masad et al. 共2001兲 introduced the form index
共FI兲 and angularity index 共AI兲, which are defined as follows:
355
兩 R ␪⫹5 ⫺R ␪ 兩
FI⫽ 兺
␪⫽5 R␪
(10)

355
Fig. 9. Effect of surface roughness on friction and dilatancy angles 兩 R ␪ ⫺R EE ␪ 兩
AI⫽ 兺
␪⫽5 R EE ␪
(11)

where R⫽radius of the particle at a directional angle ␪; and R EE␪

⫽radius of an equivalent ellipse at a directional angle ␪. Masad

Fig. 10. Scion image analysis window showing particle analysis

JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004 / 43

J. Comput. Civ. Eng., 2004, 18(1): 36-45

 Table 6. Shape Factor Results for F Sand
Area ds dL P D equ
Statistical parameter 共␮m2兲 共␮m兲 共␮m兲 共␮m兲 共␮m兲 FF SP SF IR I sph
Mean 18,124.12 127.6 168.16 507.60 161.66 0.8215 0.9334 0.7703 1.0953 0.3364
Median 15,588.00 126.0 156.00 484.00 154.14 0.8222 0.9335 0.7503 1.0969 0.3411
Root mean square 20,688.74 134.6 175.04 529.29 168.57 0.8238 0.9342 0.7796 1.097 0.4032
Standard deviation 10,182.72 43.85 49.59 153.08 48.75 0.0622 0.0381 0.1224 0.0619 0.2269

Table 7. Shape Factor Results for M Sand

Area ds dL P D equ
Statistical parameter 共␮m2兲 共␮m兲 共␮m兲 共␮m兲 共␮m兲 FF SP SF IR I sph
Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.

Mean 275,247.6 454.52 753.56 2,076.52 661.31 3.5608 1.0239 0.6902 1.0878 0.6565
Median 239,205 430.00 727.00 1,973.00 628.34 0.7573 0.8999 0.6376 1.1207 0.6038
Root mean square 291,080.7 473.61 766.74 2,140.23 681.60 14.235 1.1984 0.7398 1.107 0.8645
Standard deviation 96,645.44 135.86 144.48 529.005 168.47 14.067 0.6356 0.2717 0.212 0.5741

Table 8. Shape Factor Results for C Sand

Area ds dL P D equ
Statistical parameter 共mm2兲 共mm兲 共mm兲 共mm兲 共mm兲 FF SP SF IR I sph
Mean 2.6772 1.4068 2.2556 6.482 2.064 0.7711 0.9036 0.6628 1.126 0.5933
Median 2.2600 1.3400 2.2300 6.440 2.0510 0.7887 0.9124 0.6708 1.126 0.5123
Root mean square 2.903 1.4529 2.3088 6.623 2.1092 0.7734 0.9049 0.6793 1.128 0.7030
Standard deviation 1.1455 0.3707 0.5031 1.386 0.4414 0.0618 0.0476 0.1519 0.0617 0.3847

et al. 共2001兲 proposed correlations for the rutting resistance of
asphalt using the FI, AI, and TI. The rutting resistance was mea-
sured using the Purdue wheel-tracking device 共PURWheel兲, and
strong correlation was found between these indices and the resis-
tance. Matsushima and Konagai 共2001兲 studied the grain-shape
effect on the peak strength of granular materials using the discrete
element method. In their study they concluded that the shape does
not affect the strength at the same void ratio where the surface
roughness is dominant here; they verified the result using real
sands. Brezezicki and Kasperkiewicz 共1999兲 proposed a method
for automatic characterization of the shape of coarse particles
based on an image analysis technique. Ghalib and
Hryciw 共1999兲 proposed an imaging and watershed analysis
method to determine the soil particle distribution. Yudhbir and
Abedinzadeh 共1991兲 used an image analyzer to quantify the shape
of particles. They used Eq. 共8兲 to measure the particle sphericity.
In this paper, two new indices are introduced to define the
sphericity and roundness of particles. The sphericity index (I sph)
is defined as
polygon, circle, cross hair, etc.兲 and performing the measurement
on the selected particles. Then I sph and I R values were calculated
and the results of the analysis are shown in Fig. 1, which shows
that I R ranges from 1.0 for well-rounded particles to greater than
1.5 for very angular particles, and I sph ranges from 0.0 for diskoi-
dal particles to greater than 1.0 for prismoidal particles.
Scion Image Analysis Software was also used to measure the
areas, perimeters, and the shortest and longest axes of 25 particles
of each of the three sands. High-resolution SEM images were
used to make the measurements. The equivalent circular diameter
for each particle was then calculated. The equivalent diameter
was considered to be the mean axis length as well. Due to the fact
that the SEM images were taken for the particles as is 共no flush-
ing was done兲, more than five particles were available in the
image and it was not easy to separate them. Then a free-hand
drawing tool of SIAS was used to trace the perimeter of each
particle and the measurements were made based on the features
available in this software 共Fig. 10兲. The shape/roundness factors
discussed and defined above were calculated and Tables 6 – 8 list

兺冏 冏
N statistical summaries of the analysis. Based on the classification
1 D equ共 i 兲 D equ共 i 兲 proposed by the authors (I R and I sph) the F sand mainly consists
I sph⫽ ⫺ (12)
N i⫽1 d s 共 i 兲 d L共 i 兲 of spherical well-rounded particles whereas the M and C sands
consist of subprismoidal rounded particles.
where D equ⫽equivalent particle diameter 共perimeter/␲兲; and d s
and d L are as defined before. The roundness index (I R ) is defined
as follows:
Conclusions
N
1 P act共 i 兲
I R⫽ 兺
N i⫽1 ␲ 关共 d s 共 i 兲 ⫹d L 共 i 兲 兲 /2兴
(13) Surface roughness was accurately quantified for three silica sands
using the optical interferometry approach. Roughness parameters
where P act(i) ⫽actual perimeter of the particle. Particles listed in were defined and calculated for representative samples of the
Powers’ chart 共Fig. 1兲 were scanned and converted to image for- three sands used in the study. It was found that, as the surface
mat. Scion Image Analysis Software (SIAS) has drawing tools roughness increases, the friction and dilatancy angles increase.
that enable one to manually measure the parameters in Eqs. 共12兲 The new roundness and sphericity parameters seem to accurately
and 共13兲 共Fig. 10兲 simply by choosing the proper tool 共i.e., line, classify the shape and sphericity of granular particles.

44 / JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004

J. Comput. Civ. Eng., 2004, 18(1): 36-45

 Acknowledgments
The writers gratefully acknowledge the financial support provided
by NASA George C. Marshall Space Flight Center under Contract
No. NAS8-01042. Thanks are also due to Professor Efstathios
Meletis, Department of Mechanical Engineering at Louisiana
State University, for giving access to use the WOI.
References
Alshibli, K. A., and Sture, S. 共2000兲. ‘‘Shear band formation in plane
strain experiments of sand.’’ J. Geotech. Geoenviron. Eng., 126共6兲,
Downloaded from ascelibrary.org by Purdue University Libraries on 01/17/16. Copyright ASCE. For personal use only; all rights reserved.
495–503.
Brzezicki, J. M., and Kasperkiewicz, J. 共1999兲. ‘‘Automatic image analy-
sis in evaluation of aggregate shape.’’ J. Comput. Civ. Eng., 13共2兲,
123–128.
Content, P., and Ville, J. F. 共1995兲. ‘‘Surfascan 3D—An industrial 3D
surface texture characterization instrument.’’ Int. J. Mach. Tools
Manuf., 35, 151–156.
Ghalib, A. M., and Hryciw, R. D. 共1999兲. ‘‘Soil particle size distribution
by mosaic imaging and watershed analysis.’’ J. Comput. Civ. Eng.,
13共2兲, 80– 87.
Grigoriev, A. Y. A., Chizhik, S. A., and Myshkin, N. K. 共1998兲. ‘‘Texture
classification of engineering surfaces with nanoscale roughness.’’ Int.
J. Mach. Tools Manuf., 38共5– 6兲, 719–724.
Hong, H.-K., Myung, Y.-C., and Choi, J.-S. 共1999兲. ‘‘3-D analysis of
projective textures using structural approaches.’’ Pattern Recogn., 32,
357–364.
James, P. H., and Vallejo, L. E. 共1997兲. ‘‘Fractal analysis of the roughness
and size distribution of granular materials.’’ Eng. Geol. (Amsterdam),
48, 231–244.
JOURNAL OF COMPUTING IN CIVIL ENGINEERING © ASCE / JANUARY 2004 / 45
J. Comput. Civ. Eng., 2004, 18(1): 36-45
Koener, R. M. 共1968兲. ‘‘The behavior of cohesionless materials formed
from various materials.’’ PhD dissertation, Duke University, Raleigh,
N.C.
Lee, Y.-G., Lee, J.-H., and Hsueh, Y.-C. 共1998兲. ‘‘Texture classification
using fuzzy uncertainty texture spectrum.’’ Neurocomputing, 20, 115–
122.
Masad, E., Olcott, D., White, T., and Tashman, L. 共2001兲. ‘‘Correlation of
fine aggregate imaging shape indices with asphalt mixture perfor-
mance.’’ Transportation Research Record 1757, Transportation Re-
search Board, Washington, D.C., 148 –156.
Matsushima, T., and Konagai, K. 共2001兲. ‘‘Grain-shape effect on peak
strength of granular materials.’’ Computer methods and advances in
geomechanics, C. Desai et al., eds., Balkema, Rotterdam, The Neth-
erlands.
Pettijohn, F. J. 共1949兲. Sedimentary rocks, Harper and Brothers, New
York.
Powers, M. C. 共1953兲. ‘‘A new roundness scale for sedimentary par-
ticles.’’ J. Sediment. Petrol., 23共2兲, 117–119.
Powers, M. C. 共1982兲. ‘‘Comparison charts for estimating roundness and
sphericity.’’ AGI Data Sheets, American Geological Institute, Alexan-
dria, Va.
Russell, R. D., and Taylor, R. E. 共1937兲. ‘‘Roundness and shape of Mis-
sissippi River sands.’’ J. Geol., 45, 225–267.
Smith, M. L. 共1999兲. ‘‘The analysis of surface texture using photometric
stereo acquisition and gradient space domain mapping.’’ Image Vis.
Comput., 17, 1009–1019.
Veeco Metrology Group. 共1999兲. Wyko Surface Profiler reference manual,
Version 2.2.1, Veeco Metrology Group, Tucson, Ariz.
Wadell, H. 共1932兲. ‘‘Volume shape and roundness of rock particles.’’ J.
Geol., 40, 443– 488.
Yudhbir, and Abedinzadeh, R. 共1991兲. ‘‘Quantification of particle shape
and angularity using the image analyzer.’’ Geotech. Test. J., 14共3兲,
296 –308.