TRANSPORTATION RESEARCH RECORD 1301 107

Measurement of Aggregate Shape, Surface

Area, and Roughness
RICHARD D. BARKSDALE, MICHAEL A. KEMP, WILLIAM J. SHEFFIELD,
AND JAMES L. HUBBARD

The physical characteristics of aggregates have an important in- has a significant influence on the volume of particles retained
fluence on the behavior not only of unstabilized bases but also on a specific sieve. For material retained on a given sieve
of asphalt concrete and portland cement concrete mixes. Exper- size, Lees (1) has shown that rod-shaped particles are about
imental techniques using modern data acquisition methods for
2.5 times the size of disc-shaped particles. This difference in
measuring and analyzing aggregate characteristics are described.
A Pencept Penpad digitizer accurate to 0.0015 in. together with size affects the ability of the particles to properly fill voids of
an IBM-XT microcomputer was used to collect and analyze the coarser size aggregate.
necessary data for measuring aggregate shape, surface area, and
roughness. Measured coordinates of critical points were stored
in an Autocad DXF file. The data were edited using a Basic Simple Classification Systems
program and then transferred to Lotus 1-2-3 with which they
were analyzed and presented in the form of graphs, tables, and
histograms. All data transfer and manipulation were performed The shapes of fine- and coarse-aggregate particles can be
electronically. Methods for shape and surface area measurement, divided into the following four general shape categories (1 ,2):
classification, and interpretation were reviewed. Two different (a) flaky, (b) cuboidal, (c) blade, and (d) rod. Although Brit-
techniques for measurement of aggregate surface area were com- ish Standard BS 812 (2) separates the aggregate into these
pared and demonstrated to yield reasonably similar results. A four rather broad categories, the method does not define the
special shadow technique for studying the characteristics of par- exact location of an aggregate within each category. A special,
ticles smaller than the No. 8 sieve was described. In addition, the
simple gauge is used to measure the two indices required for
accuracy of the digitizer procedure for measuring surface rough-
ness was assessed. shape classification.
ASTM (3) and the Corps of Engineers (4) also have test
methods similar to BS 812 (2) for evaluating flat and elongated
Asphalt mix designs are based on many factors including the coarse particles in aggregates to be used in concrete. These
type and amount of asphalt, air voids, aggregate character- methods use a specially designed caliper to determine particle
istics, aggregate gradation, and mineral filler. In order to shape ratios. Measurements are performed by hand to de-
investigate the effects of aggregate characteristics, which are termine if particles have certain length-to-thickness or width-
usually not fully considered, the Georgia Department of to-thickness ratios; specific particle dimensions are not mea-
Transportation (DOT) initiated, through the Georgia Insti- sured. Although simple, these methods are just classification
tute of Technology, a comprehensive research program. In schemes and do not permit determination of surface area.
the final phase of the study, the effects of these variables will Different ratios separating aggregate classes have been pro-
be evaluated on the rutting performance of Georgia DOT posed for describing aggregate particles (1,5).
asphalt mix designs. This paper describes the measurement The four broad categories defined by these methods allow
of aggregate shape, surface area, and roughness using modern a large range of particle shape characteristics within each
digitizing techniques taking advantage of a microcomputer. classification. For research purposes, these methods might
give misleading results, affecting aggregate performance. Also,
these classification tests are not suitable for measuring the
PARTICLE SHAPE shape of particles much finer than about the No. 12 sieve,
and surface area cannot be determined using the results. Clas-
Introduction sification systems that use just one aspect ratio are not suitable
for defining particle shape.
The shape of the aggregate influences the gradation curve
obtained by sieving (1). Flaky particles tend to diagonally
pass sieves having square holes. Also, the shape of the particle Generalized Classification Systems

Both fine and coarse aggregate particle shape can be deter-

R. D. Barksdale, School of Civil Engineering, Georgia Institute of mined by measuring the flatness ratio and elongation ratio
Technology, Atlanta, Ga. 30332. M.A. Kemp, Atlanta Testing and (1). The flatness ratio (p) is the ratio of the shortest length
Engineering, 11420 Johns Creek Parkway, Duluth, Ga. 30136. W. J. (c) divided by the intermediate length ( b), and the elongation
Sheffield, Law Engineering Testing Company, 396 Plasters Avenue,
Atlanta, Ga. 30324. J. L. Hubbard, Research Institute, Georgia ratio (q) is the ratio of intermediate length (b) divided by the
Institute of Technology, Atlanta, Ga. 30332. greatest length (a). By determining the actual flatness and
108 TRANSPORTATION RESEARCH RECORD 1301

elongation ratios, a continuously varying classification can be particularly considering the relatively large observed variation
developed. This approach also permits defining a shape factor in aggregate shape and dimensions. The length was digitized
F = p/q and sphericity iji. Sphericity ijJ is the ratio of surface as the longest dimension of the aggregate, and the width as
area of a sphere of the same volume as the particle divided the average dimension, in the plane of the photocopied image,
by the surface area of the particle (J). The proposed method perpendicular to the length. The coordinates (x ,y) of each
is considerably more flexible for research purposes than the point representing one end of a dimension were digitized, and
Corps or British classification schemes. The British and Corps the actual dimension was later calculated. If the original or-
classifications can be quickly obtained from the general flat- dering of length, width, and thickness was to correct, a com-
ness and elongation ratio method described by Lees. puter program later automatically reordered the dimensions
Also, the surface area and sphericity of the aggregate can correctly.
be determined using the more general shape classification Shadows were created when trying to photocopy the profile
method. The generalized shape classification concept is a of the aggregate to measure its thickness. Therefore, aggre-
method of tridimensional shape analysis where each grain is gate thickness was not digitized directly from a photocopy.
approximated by a tetrakaidecahedron (2 ,6). Three mutually Instead, vernier calipers were used to measure the average
perpendicular particle dimensions (length, width, and thick- thickness directly from the aggregate. The calipers, open to
ness) are measured and used to calculate the ratio of surface the proper width, were then laid on the digitizing pad, and
area of the particle compared to that of an equivalent sphere, the tips of the calipers, representing thickness of the aggre-
or else surface area is directly calculated. gate, were digitized. A pen-type digitizer, as opposed to one
with cross-hairs, was used, enabling digitizing of the vernier
Particle Shape Using a Digitizer caliper measurements.
This method of measuring the dimensions proved to be
For shape classification, the aggregates studied in this inves- efficient. With experience, an operator can digitize the three
tigation were divided by sieving into the following four size dimensions of 150 aggregates in approximately 30 to 45 min.
ranges: V2 to% in., Nos. 4 to 8, Nos. 8 to 120, and smaller After digitizing the three perpendicular dimensions for all
than the No. 120 sieve. These size ranges were selected by a aggregates, the data are saved as an Autocad DXF file in
panel of engineers as being appropriate. For the two larger ASCII code.
size aggregate ranges (the V2 to% in. and Nos. 4 to 8 sizes),
an aggregate sample consisted of 150 particles of each size,
with the number of particles being counted visually. In the Aggregate Smaller Than No. 8 Sieve in Size
smaller size ranges, microphotographs and special techniques
were used to measure the aggregate shape. The number of Aggregates less than the No. 8 sieve in size require the use
particles in each sample of smaller particles varied from 50 of specially prepared optical microphotographs. Similar to the
to 150, on the basis of the number of particles captured in large aggregate, aggregate length was digitized directly from
each photograph. At least three different samples were mea- the photograph as the longest dimension and the width as the
sured for each aggregate type. This approach resulted in the average dimension, in the plane of the photograph, perpen-
use of a minimum of 450 particles for each of the coarser two dicular to the length.
sizes studied and a minimum of 150 particles for each of the Because these particles are small, the height cannot be
finer two particle sizes studied; usually 250 or more particles measured directly using calipers. Therefore, a special tech-
were included. The use of more than 150 particles was desir- nique was used relating a shadow length on the photograph
able but was too expensive to achieve iu all cases fu1 lhe to particle height. As the particles were prepared fol' the
microscopic particles. microscope, uniform reference spheres were added to estab-
Aggregate shape was determined and numerous plots and lish the scale for vertical height. A thin film of metal was
tables produced without a human hand ever working with the evaporated onto the surface at an angle to the substrate on
data. The procedure developed for particle analysis is com- which the particles set to create a shadow (7). Because the
pletely automated and uses a relatively inexpensive digitizer evaporation source is a relatively long distance away, the
that automatically feeds data into an IBM-XT microcom- angles at which it strikes the particles and reference spheres
puter. are approximately equal. Therefore, by geometry, a unique
ratio exists between the shadow lengths of the reference spheres
Aggregate Greater Than No. 8 Sieve in Size and the aggregate particles and their heights. A special tech-
nique, described subsequently, was used to capture the shadow
For the aggregate greater in size than the No. 8 sieve, photo- on the photograph.
copies were made of the flattest profile of the particles. A Similar to the large aggregate, all digitized dimensions were
Savin 7350 copying machine was used to provide images of saved as an Autocad DXF file in ASCII code. A set of micro-
50 particles at a time, which were placed in a small box. The photograph data can be digitized in 25 to 50 min depending
box had a clear plastic bottom and dividers so as to give five on the number of aggregates in a sample.
rows of 10 aggregates each. The copy machine was found not
to distort the photocopied image of the aggregate. By pro-
viding a profile view of the aggregates, the length and width Manipulation of Data Using Autocad and Lotus 1-2-3
were easily digitized directly from the photocopy using a Pen-
pad digitizing tablet manufactured by Pencept, Inc. The dig- After digitization, all dimensions were stored in an Autocad
itizer had an accuracy of 0.0015 in., which was sufficient, DXF file. An Autocad DXF file contains all the formatting,
Barksdale et al. 109

scaling, size, and other information that Autocad uses when of the limited depth of field of optical techniques. If measure-
displaying and working with a drawing . A Basic program ment of the thickness of these particles is required , a unique
called DFXTRACT was used to remove all the unwanted sample preparation problem exists.
formatting information and extract only the coordinates of
the ends of lines defining the dimensions of the aggregates.
Large Fraction-Particles of Nos . 8 to 120 Sieve Size
This program then saved the data in a form that Lotus 1-2-
3, or other spreadsheets. was able to use readily.
Sample Dispersion The solution for measuring aggregate
Once the endpoints of the lines representing the dimensions
shape of small particles is to look not at the particles them-
of the aggregates were extracted and stored in the Lotus 1-
selves but to create flat silhouette representations of the par-
2-3 PRN file, the PRN file was imported into a Lotus 1-2-
ticles from which measurements can be taken. If a shadow is
3 worksheet using the Lotus 1-2-3 import command. The
added to the silhouette directly related to a particle's height,
lengths of the dimensions were then calculated using the co-
the three dimensions of length, width , and height can easily
ordinates of the end points and stored in a Lotus 1-2-3
be measured in one flat plane. The technique of vacuum
worksheet file named "WKl."
evaporation of thin metal films , such as that used in the prep-
The conversion of endpoints to lengths defining the di-
aration of samples for transmission electron microscopy (TEM) ,
mensions of the aggregate can be performed faster using Basic
was used to prepare these flat, two-dimensional representa-
as a part of the DFXTRACT program. Using the Basic pro-
tions of three-dimensional samples (7). In order to prepare
gram requires about 30 sec on an IBM - XT computer, com-
fine aggregates so that silhouettes could be obtained, alu-
pared with 3 min for the Lotus 1-2-3 ma~ro. However.' er-
minum was used instead of platinum (which is used in TEM
rors , such as adding a stray line or extra pomt, are sometimes
preparation because of its ease of evaporation) ..
made using the digitizer and Autocad. The Lotus 1-2-3
First, a glass microscope slide was cleaned with so~p and
worksheet approach allows examination of the data; in most
water to ensure good adherence of the evaporated film. A
cases the error can be corrected even after general processing
good dispersion of the sample particles was placed on t.his
of the data has been finished . A Basic program would prob-
slide. Care was exercised to ensure that the particle spacmg
ably blow up or give useless results in the same situation.
was sufficient to allow for a shadow between the particles,
Typical results demonstrating how the resulting shape mea-
and that the dispersion was representative of the true size
surement data can be readily presented using a spreadsheet
distribution. Obtaining good sample dispersion is perhaps the
are shown in Figures 1 and 2.
hardest but most important part of the sample preparatior.
A wide variety of dispersion techniques can be used depending
on the nature of the particulate material being studied.
Three-Dimensional Measurements of Very Fine
Aggregate Samples For the particles used in this study, the dispersion was ~re­
pared in the following manner. Each sample was .placed m a
The fine aggregate samples studied (smaller than the No. 8 plastic bag . The sample was then mixed .by shakmg t.he bag
sieve in size) have a broad size range that requires the use back and forth while turning it (8) . Shakmg was earned out
for a sufficiently long period of time to thoroughly mix the
both of low- and of high-magnification techniques , which can-
sample. A number of small subsamples were taken from dif-
not be accomplished using only one instrument. As a result,
ferent areas of the bag and mixed to further ensure a represen-
aggregate varying in size from the No . 8 to the No. 120 sieve
tative sample. Because the mica consisted of relatively. large
was treated differently than aggregate smaller than the 120
flakes, an antistatic spray was not required to prevent st1ckmg
sieve. The larger fraction particle size (Nos. 8 to 120 sieve
of these particles to the sides of the bag. A number of cleaned
size) is large enough to present difficulties in direct optical
glass slides were placed on a flat surface and the extracted
measurements and especially in macrophotography because
sample allowed to drop onto the slides from a height of about

....... .. --..
1 ft. This technique was performed in an area that had no air
1.0
movement. A small quantity of uniform glass spheres was also
.1
._ Disc
• I dropped onto the slides. The size of the spheres was later
• I cut.! determined by measuring their diameters on the photograph
.0
0.8 and calculating sizes knowing the scale of the photograph .
"'
II
<:I'
•
.s:
~
c:
0.6

•
-. ••
~
-
• ii •
I
"·
I •
-
II
One of the slides that visually appeared to have the best
dispersion was selected for further processing .

.s:
~
0. 4 - Evaporation of Aluminum The slide having the best di~­
Cll
§ Blad•
persion of particles was placed in a vaccum evaporat10n umt
~ 0.2 '""" in which two filaments had been set up for evaporation of
aluminum. One filament was located directly above the slide
II 0
while the other was placed off to the side at an angle of about
0.2 0.4 0.6 0.8 1.0 1.2
30° to the slide surface. The unit was evacuated to a pressure
Flatness Rat;o, p = c I b of at least 10- 4 mm of mercury and the aluminum evaporated .
FIGURE I Typical shape classification scatter diagram The proper amount of aluminum evaporated was determined
for a selected Georgia specimen. experimentally to give the best contrast both for shadow and
110 TRANSPORTA TJON RESEARCH RECORD 1301

::2

-=
,-..

cu
....'cu"'
28

24
Q.
'-'
20
~
'....cu"'
= '---··--- - - - - - - - -..·--
.... 16
='"'
'"' 12
....0
0
.... 8
'cu"'
=
& 4
cu
....
~
0
0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3 3.3

Shape Factor
FIGURE 2 Typical shape factor histogram for a selected Georgia aggregate.

silhouette. For a single particle, two areas are present on the ously described. The paricles were left in place on the slide
slide that may be coated by only one layer of aluminum, the because depth of field is not a problem in the SEM. The
shadow area and an area opposite the shadow if the particle micrographs were taken using the backscatter signal, which
is not square with the surface. is sensitive to elemental differences. Good contrast was ob-
The slide is removed from the coating unit, and the particles tained using this technique between the particle, shadow, and
are then removed from the slide by blowing them off with background. Uniform glass or latex spheres were included in
air. If the fine particles resist removal by blowing, the slide the dispersion for shadow and thickness determinations.
is placed in a beaker of water containing a small amount of
wetting agent and then treated in an ultrasonic bath for a few
seconds. Estimation of Particle Thickness

After following the previously given procedures for sample

Fine Fraction Smaller Than No. 120 Sieve Size preparation, all of the particle information is now represented
in the single plane of the slide, which can be photographed
Particle size measurements of the fine fraction less than the at any magnification or viewed on a projection screen where
No. 120 sieve in size were made from micrographs taken using direct measurements can be made. Figure 3 shows the shad-
the scanning electron microscope (SEM). The particles were owed silhouette of a single particle; A is the particle length,
dispersed on a plastic substrate and the preparation coated B the width, and C the shadow length. The shadow length of
with carbon by evaporation to prevent charging the SEM. the spheres can be used to convert shadow length of the
The dispersion was then shadowed with aluminum as previ- particles to thickness using the formula

FIGURE 3 Optical presentation of coated and shadowed slide with

particle removed.
 Barksdale el al. 111

(1) where

where S = surface area,

PL = average number of particle intersections per circle,
T = particle thickness V0 = volume of the sample, and
ShP = particle shadow length, N = total number of particles in the sample.
rap = sphere radius, and
Shap = sphere shadow length. The quantitative stereology approach makes no geometric
For low shadowing angles, the simpler formula assumptions concerning aggregate shape (9). This method is
statistically exact provided a sufficient number of measure-
(2) ments are performed . However, the sample must be statis-
tically representative of the aggregate, and a sufficient number
can be used as a close approximation. of circles must be drawn on the cut faces. The best results are
obtained if the particles are randomly positioned in the con-
SURFACE AREA tainer, with the distribution being homogeneous. If a random
distribution does not exist, more sampling planes cutting through
Introduction the block of aggregates at different angles are required, or
more sampling circles must be drawn on the cut faces or both.
The surface area of the aggregate for a given quantity of Even if the aggregates are not randomly oriented, the correct
asphalt has a significant effect on the asphalt film thickness surface area can be obtained if a sufficient number of circles
and as a result can influence mix performance. Surface area and sections are used together with a sufficiently large number
can be determined by a number of methods including (a) the of particles.
tridimensional approximation described by Aschenbrenner (6) ,
(b) quantilative stereology (9), (c) surfac coatings including
wax and paint (1,5), ( d) air and mercury permenbility (JO), Sample Preparation
and (e) gas adsorption. The tridimensional method described
by Aschenbrenner (6) has been previously summarized. In The procedure used for the quantitative stereology method
addition to this approach, the quantitative stereology and the is considerably more labor intensive than the computer method
gas (usually nitrogen) adsorption methods probably offer the used to obtain aggregate shape and surface area. Approxi-
best techniques for determining surface area. The gas ad- mately 500 aggregates from the 112- to %-in. sieve size were
sorption method, however, indirectly measures the external counted for each sample from each of the seven selected
surface area of the particle and also any pore greater in size quarries for which surface area was measured using this tech-
than about 4 A. This method requires severa l ideal assump- nique. Of these 500 aggregates , 100 were digitized in this study
tions to calculate surface area using thermodynamic into the computer for analysis by the Aschenbrenner method
principles . to compare results. After digitizing was complete, these 100
aggregates were combined with the remaining 400 particles
Quantitative Stereology and later placed in a cylinder.
A two-part epoxy glue, which was quick drying and strong,
Fundamentals was used to bind the aggregates together. Plastic cylinders
5.25 in. high and 3 in. in diameter were used as molds. This
An interesting method for measuring surface area of aggre- size is convenient to work with and handle, and provides a
gates is by using quantitative stereology (9). Quantitative ste- sufficient volume to produce a representative sample of the
reology is a direct measurement method and consists of pre- size of aggregate studied.
paring a random sample of N aggregates placed in a container After mixing, a small amount of epoxy was poured into the
of known volume. The aggregates are encased in a cementing bottom of the mold . Several aggregates were then dropped
agent such as an expoxy to form a solid block . The solid block into the mold. The mold was tapped for several minutes with
is then sawed into several random pieces with the cuts oriented a metal rod to move the aggregates into a dense packing and
in different directions. A number of circles of radius R are to drive any air bubbles present to the surface. When most
inscribed on each saw cut surface, and the number (P) of of the air bubbles were out, more epoxy and more aggregates
times each circle intersects an aggregate boundary is counted. were added and the tapping was repeated. This preparation
Now let PL equal cycle was continued until all the aggregates were placed in
the mold. Extra epoxy was also added to top off the mold
(3) and to act as a handle to hold the sample when it was cut.
The mold was then placed in a warm location and allowed to
where R is the radius of th~ circle and P is the number of harden for approximately 24 hr.
intersections . Next , calculate the average value of PL (i.e., The mold was stripped away from the sample after hard-
PL) for all the circles drawn on all sections. The average ening. The sample was then labeled with a permanent marker.
surface area S of the particles inside the block of aggregate Measurements were then taken of the height of the aggregate-
is then equal to epoxy specimen; the total height of the epoxy cylinder was
not measured because the volume of actual aggregate is used
(4) in the formulas for calculating surface area. Next, the lower
112 TRANSPORTATION RESEARCH RECORD 1301

portion of the epoxy-aggregate sample from the bottom up This definition, which is slightly different than that used for
was cut into disks approximately Y2 to 1 in. thick. The re- flat surfaces, was developed because using the line of best fit
maining cylindrical-shaped sample was split down the center, appears to contribute to the reduction of error caused by the
forming two long, semicircular sections. One side of each of curvature of an aggregate. Coupling this definition with eval-
the three disks and one of the flat semicircular sides was uating small sections of the particle, the problems caused by
photocopied. The data were taken from the photocopies and curvature are minimized.
reduced to preserve the integrity of the original samples.

Measurements Methods of Measuring Surface Roughness

Five circles were drawn on each cross section that was photo- Most work in measuring microtexture has involved the rough-
copied. The long flat side of the semicircular section had 12 ness of flat metal surfaces. Techniques for measuring surface
circles drawn on it. The number of intersections each circle roughness of aggregates include the following (11,12):
made with the edges of aggregates was recorded. This large
number of circles, 27 in all, was used to achieve a representa- 1. Stylus. A pen stylus is drawn over the aggregate surface.
tive sample of the aggregates. The number of intersections Optical, mechanical , or electronic magnification is usually
per circle was then averaged and entered into Equations 3 used to enhance the profile and process the results.
and 4 to calculate the surface area. The epoxy cylinder radius, 2. Cut Section. The cut profile surface can be measured of
volume of epoxy cylinder containing aggregate, and the total an aggregate embedded in an epoxy. The block of epoxy and
number of aggregates in the sample are also required. Either aggregate is cut, polished, and photographed at the desired
three or five aggregate-filled specimens were studied from level of magnification such as 15 x to 125 x . The surface
each quarry. profile is then directly measured by automatic measuring
techniques.
3. Casting. A casting of the surface is made. The magnified
Comparison of Results image of the casting is then examined to determine the profile.
4. Oblique Lighting. Illuminating the surface by oblique
Table 1 presents the results of the quantitative stereology lighting produces a shadow. A projection microscope is used
method for evaluating surface area with the one described by to observe the shadow.
Aschenbrenner. For the stereology technique, the average
standard deviation of the aggregate from the seven granite Stylus-type equipment, which appears at first to be ideal,
quarries included in this portion of the study is 0.030 in. 2 , is made to measure surface roughness along a flat surface;
which is 4.3 percent of the average measured value of 0. 700 deviation from this plane con c:rnse measurement errors and
in. 2 per aggregate. For the Aschenbrenner approach, the av- even instrument damage. Also, a stylus-type instrument can-
erage standard deviation is 0.050 in. 2 , which is 6.7 percent of not follow indentations less than the radius of the stylus and
the average measured value of 0.749 in. 2 per aggregate. The cannot measure roughness where overhangs occur. Flat sur-
percent differences in average results vary for individual quar- faces on an aggregate particle where measurement is possible
ries from -10.7 to +9.2 percent. The algebraic average dif- are often limited.
ference in surface area between the two methods for the seven
quarries is 2.2 percent. These results appear to indicate that
the Aschenbrenner model is probably sufficiently accurate for
Roughness Measurement
at least most purposes, particularly considering its simplicity.
The cut section method, previously described, was used to
measure surface roughness. Data were collected automatically
SURFACE ROUGHNESS MEASUREMENT
with the same Pencept Penpad and IBM-XT computer that
was used to measure aggregate shape and surface area.
Definition of Surface Roughness

Quantifying surface roughness is not easy, particularly for

Specimen Preparation
aggregates that have curved surfaces. Further, the value of
surface roughness depends on the magnification at which
A representative, random sample of 30 aggregate particles
roughness is examined. Numerous definitions of surface
was taken from each source. The aggregate sample was then
roughness have been proposed (11-13). For this study, the
placed in a small plastic cylinder 6 in. high and 1 Y2 in. in
definition developed for surface roughness (R) is as follows:
diameter. A two-part epoxy was used to bind the aggregate
together within the cylinders. Magnolia Plastics Epoxy Com-
(5)
pound 2014 and Curing agent 346 were chosen because of
their ability to hold the aggregate particles in place while
where
cutting, their good polishing characteristics, and their ability
Lr = true length of the segment of surface being analyzed, to harden within 24 hr.
and Thirty particles 3/s to 1/2 in. in size were dropped one at a
LP = length of the line of best fit for the segment of surface. time into the cylinder, which was one-half full of epoxy. This
TABLE 1 COMPARISON OF SURFACE AREA BY QUANTITATIVE STEREOLOGY AND COMPUTER SURFACE AREA ANALYSIS
FOR SELECTED QUARRIES

SA by (l) SA by l
Aggregate Stereology Std. Computer( ) Std.
Quarry Sample Type 2
(in. ) Mean Deviation 2
(in. ) Mean Deviation

Dixie Sand CAl Alluvial 0.636

Chatt., TN CA2 Alluvial 0.641 0.636 0.005 0.580 0.580
CA3 Alluvial 0.632

Florida Rock EAl-1 Granite o. 767 0.752

Mt. View, GA EAl-2 Granite o. 715
EA2 Granite 0.843 0.816 0.042 0.733 0.738 0.015
EA3-l Granite 0.837 0.747
EA3-2 Granite 0. 745

Florida Rock GAl-1 Granite 0. 713 0.891

Tyrone, GA GAl-2 Granite 0.833
GA2 Granite 0.801 0.767 0.048 0.841 0.850 0.025
GA3-l Granite 0.788 0.830
GA3-2 Granite 0.853

GA. Marble !Al Granite 0.809 0.823

Buford, GA IA2 Granite 0.827 0.821 0.010 0.747 0. 759 0.059
IA3 Granite 0.827 0.707

GA. Marble JAl Granite 0.677 0.700

Cumming, GA JA2 Granite 0.737 0.733 0.054 0.935 0. 762 0.152
JA3 Granite 0.784 0.651

Vulcan Materials RAl Granite 0.815 0.763

Kennesaw, GA RA2 Granite 0. 774 0.813 0.038 0.825 0.789 0.032
RA3 Granite 0.849 0.780

Vulcan Materials UAl Granite 0.759 o. 770

UA2 Granite 0. 781 0. 774 0.013 0.742 0. 763 0.019
UA3 Granite 0.782 0. 777

Note 1: The surface area (SA) is given for one aggregate.

114 TRA NSPO RTATION R ESEAR CH RECO RD / JOI

technique was found to allow settlement of 1.he particles to

the bottom minimizing the number of air bubbles trapped
during particle placement in the cylinder. No tapping of the
cylinder was needed because the samples were small. After
curing for 24 hr in a warm location, two to three rock saw
cuts across the diameter were performed on each cylinder,
giving three or four cut aggregate surfaces suitable for mea-
suring roughness on.

Surface Polishing

Numbers 120, 300, and 600 polishing grits were used to obtain
a smooth aggregate surface and sharp contrast between the
1/16 IN.
aggregate surface profile and epoxy. The No. 120 coarse grit
was used to take out most of the unevenness caused by the FIGURE 4 Photograph of polished aggregate at
saw cut; at the same time it placed small grooves in the.sample. surface after 20 x magnification.
The No. 300 grit was used to polish out the grooves placed
by the coarse No. 120 grit. Finally, the No. 600 grit polished length of the digitized surface, and calculated the line of best
out any remaining tiny marks or grooves to provide a smooth, fit of the data. The macro routine also adjusted the length
finished surface. The sample was polished a minimum of 5 for magnification, calculated the surface roughness, and then
min with each grit. The samples were washed between grit created a graph of the real surface and the line of best fit or
changes to prevent any contamination of the finer grit with projected surface. Both the graph and the worksheet were
the coarser ones. saved on disk.

Surface Photography Calibration of Digitization Procedure

A photograph of the aggregate surface gives the surface pro- Several calibrations were performed to find any errors, prob-
file in a form suitable for digitizing. A scale was also photo- lems, or limitations of the overall digitization methodology
graphed to accurately quantify the level of magnification used. used to evaluate surface roughness. Calibrations were ob-
For the purposes of this study, a magnification of approxi- tained from a simple comparison of measured surface rough-
mately 20 x was selected for the photomicrograph as being ness with the calculated surface roughness of surfaces having
suitable for defining the surface roughness characteristics simple, easily defined shapes. The first surface used consisted
(Figure 4). The photographs were later blown up 50 percent of two semicircles connected together as shown in Figure 5.
using a photocopier. This procedure gave a 30 x magnifi- For all calibrations, points on the surface were digitized at
cation of the surface and resulted in significant savings on distances on the photographs varying from 0.01 to 0.1 in .
printing costs compared with those for blowing the negative Figure 5 shows that an optimum spacing of digitized points
up to 30 x during printing. of about 0.05 in . exists , which gives the minimum error for a
The use of other magnification levels of the surface would surface consisting of two semicircles. A closer spacing of dig-
be expected to yield different values of surface roughness. itization points, which intuitively would be thought to be more
Wright, for example, suggested using 125 x, which perhaps
is too much magnification to evaluate surface roughness of
the gross surface. The appropriate value of magnification to
use certainly deserves further study. ..f"
Kodak PX-125 black-and-white film was used for the
photographs. Three aggregate particles from each quarry were
evaluated for surface roughness. Pictures were taken of two
> II 1.8
i:z::
different locations on the surface of each of the three particles, .;;
resulting in six photographs per quarry. In determining sur- "'c
Q,I

.: 1.7
face roughness, each photograph was broken into three smaller tlQ

segments to minimize the curvature effects of the aggregates.

=
Q
i:z::
Q,I
This procedure resulted in 18 values of surface roughness from <J

each quarry. ~ I.II

=
V)

Digitization 1 .s-+--~--1.._~--+--...--1-----.--+--r----''r
0 o.oi 0.04 0.06 0.08 1.0
A similar scheme of manipulating the data as used for shape Increment to Digitize (inches)
analysis was also used for roughness. A macro routine within FIGURE 5 Calibration of surface roughness digitization
Lotus 1-2-3 arranged the digitized points, calculated the true using a sine wave surface.
Barksdale et al. 115

accurate, was actually found to be less desirable. The loss in Results

accuracy was apparently caused by small levels of shaking of
the hand (referred to as hand vibrations). A surface with saw- As presented in Table 2 and found by Wright (11), surface
tooth shape was also used for calibration. The optimum dig- roughness varies greatly both over the surface of a single
itization spacing was found to be 0.04 in., which was close to particle and from one particle to another for the same quarry.
that found for the circular surface. Therefore, only general trends of surface roughness should
The calibration studies indicated that a digitization incre- be considered and as many measurements as practical should
ment of 0.05 in. yielded good results . An average correction be performed. Observed variations in surface roughness were
factor of + 2.0 percent was used to correct calculated surface as follows: 1.16 to 1.26 for 15 granite gneiss quarries; 1.13 to
roughness to increase the accuracy on the basis of the cali- 1.15 for 3 limestone quarries; 1.16 for an injection quartz;
bration studies . and 1.13 for an alluvial gravel.
Reproducibility of roughness measurements on aggregate
surfaces obtained by a single experienced operator was found
to be good. In comparing the results from three digitizer CONCLUSIONS
operations (only one having a high level of experience), the
standard deviation of roughness was found to be 0.023 for The use of modern data acquisition procedures, which include
three quarries in a supplementary study. This study indicated a relatively low-cost digitizer and microcomputer, makes pos-
that the operator should become experienced using the sible the accurate and rapid acquisition of large quantities of
digitizer on reference surfaces such as the one shown in data. These devices were used together with Autocad and
Figure 5. Lotus 1-2-3 spreadsheet, to acquire and process large quan-

TABLE 2 ROUGHNESS DATA ILLUSTRATING VARIABILITY FOR A STREAM DEPOSIT-

DIGITIZATION INCREMENT OF 0.05 in., AGGREGATE% TO 'h in.

ROUGHNESS CORRECTED AVG. RGH AVG. RGH AVG. RGH

SAMPLE PER SAMPLE ROUGHNESS PER GROUP PER AGG. PER QUARRY

CA1051 1.16 1.18

2 1.10 1.12 1.14
3 1.09 1.11 1.12
CA2051 1.09 1.11
2 1.07 1.09 1.10
3 1.07 1.09
CA30 51 1.07 1.09
2 1.12 1.14 1.12
3 1.09 1.11 1.15 1.13
CA4051 1. 23 1. 25
2 1.08 1.10 1.18
3 1.16 1.18
CA5051 1.10 1.12
2 1.10 1.12 1.11
3 1.06 1.08 1.13
CA6051 1.20 1.22
2 1.10 1.12 1.16
3 1.10 1.12
-------·- ---------------------------·------------·------·-·- -------·---
Mean 1.13 Standard Deviation 0.50

Roughness per Sample - True Length/Projected Length

Corrected,.Roughness - (1.0199) +Roughness per Sample
Avg. Rgh. per Group - Average roughness of s a mples from same picture
Avg. Rgh. per Agg. - Average roughness of 2 groups (pictures) taken from
same aggregate
Avg. Rgh. per Quarry - Average roughness of 3 agg. from each quarry sample,
A or B
116 TRANSPORTATION RESEA RCH RECORD 1301

tities of data without ever touching the data after digitization. Corps of Engineers, U.S. Army Waterways Experiment Station,
The use of a spreadshseet makes possible easy interpretation Vicksburg, Mis .
5. B. Mather. Significance of Tests and Propcrti · of or1cretc :md
and presentation of the data. Sample preparation and data Concrete-Making Materials. ASTM TP 169-A. Shape, S111f11ce
acquisition have been described for shape, surface area, and Texture, and Coatings, 1966, pp . 415- 431.
roughness of aggregates. These techniques can, however, also 6. B. C. Aschenbrenner. A New Method for Expressing Particle
be applied to many other materials applications. Sphericity. Journal of Sedimentary Petrology , Vol. 26, No. l,
1957, pp. IS - 3L.
7. D. Kay. Tecl111iq11es for Electron Microscopy. 2nd ed ., F. A.
Davis o. , Philadelphia, Pa., 1965, pp. 136- )44.
ACKNOWLEDGMENTS 8. J. L. Hubbard. Microscopy and Image Analysis. Metal /-/a11d-
book , 9th ed., Vol. 7, American Society (or Metals, Metals Park,
The results of this study were primarily sponsord by the Geor- Ohio, 1984, pp. 225-230. .
9. E . E. Underwood. Quantitative Sterology Ch. 2. Add1son-
gia DOT in cooperation with the FHWA under the HPR
Wesley, Rcadfog , Mass., 1974.
Program. The fine technical support given by Pete Malphurs, 10. P. C. Cannan . Determination of the Specific Surface of Powders.
State Materials Engineer, and others of the Georgia DOT's Transactions, Institute of Chemical Engineers, Vol. 57, 193 ,
Office of Materials and Research is gratefully acknowledged. p. 225.
11. P. J . F. Wright. A Method of Measuring the Surface Texwre of
Aggregate . Magazine of Concrete Research, Vol. 7, No. 21 , L995,
pp.151-160.
REFERENCES 12. E. E. Underwood and K. Banerji. Quantitative Fractog_raphy.
Metals Handbook 9th ed. , Vol. 12, Metals Park, Ohio. pp. t99 -
1. G. Lees. The Measurement of Particle Shape and Its Influence 210.
in Engineering Materials. Journal, British Granite and Whinstone 13. J. I . McCool. Relating Profile Instrument Measurements to Func-
Federation, Vol. 4, No. 2, 1964, pp. 1-22. tional Performance of Rough Surfaces. Journal of Tribology, Vol.
2. British Standards Institute. Flakiness Index Test. BS 812. United 109, April 1987, p. 264-270.
Kingdom, 19 4.
3. tandard Test Method for Flat or Elongated Particles in Coarse
Aggregate. ASTM D-4791-89. An111wl /Jook of ASTM Standards,
Vol. 04.03 , ASTM, Philade lphia, Pa ., 1990.
4. A Method of Test for Flat and Elongated Particles in Coarse Publication of this paper sponsored by Committee on Mineral Aggre-
Aggregates. CRD-C119-53. Handbook for Concrete and Cement, gates.