VASCONCELOS
Characterization of microstructures
A. M, Gokhalc
ABSTRACT
Various stereological techniques for characterization of microstructures are presented.
The
methods for
the
estimation of
volume fraction, surface area, total curvature, length of lineal features,
and topological properties are described.
Emphasis is
placed on the underlying assumptions, limitations and applicability of the stereological equations.
Important derived properties are
described and their limitations and utility are discussed.
Techniques for the estimation of particle size distributions are presented.
INTRODUCTION
(1)
It must be
distribution.
(2)
valid
for
any
shape
and
size
Kanpur,
is
now
with the
71
"
2
METRIC
PROPERTIES
fl
value
Fig. 1
fl
12
placed in
the plane
fraction of
particular
Figure 2 ).
V L*
A. Volume fraction
...2
V =A~
T
... 1
10
The above
equation has been re-derived by several
other
research workers . It is not necessary to
randomize the sectioning plane with respect to the
spatial
orientation for the estimation
of volume
fraction .
The above equation is valid for any particle
shape and size distribution and it is statistically exact.
However, the area fraction must be measured ou a
T h e volume fraction can also be estimated by
representative two dimensional section. The statistical " p o i n t counting" method. This simple technique was
error involved in the estimation of volume fraction developed by T h o m s o n and Glagolev . If a number
can be kept as small as desired by increasing the -.of test points are placed at r a n d o m on a plane of
n u m b e r of observations.
polish or in the three dimensional microstructure, then
L
11
18
14
15
Transactions of The Indian Institute of Metals, Vol. 34, No. 1, Febru, ry 19S1
10
Gokhel:
Characterization oj"microsmiptvrcs
where
.73
' .
t a foil thickness
S = surface area of the phase of interest per unit
volume Equation (4) is applicable u n d e r t h e following
v
conditions :
T.L. T e s t
(2)
must be k n o w n to
estimate the
of S and t
average
T h e values
line
(1)
of volume
fraction, from a
projected
image
obtained
from a thick
distribution .
V =P
v
slice,
40
... 3
8,17
13
18,19
20
21
22
0
"
23
Let us consider the technique for estimating volume fraction from the measurements m a d e on a projected image.
The projected area fraction is usually larger than the
area fraction observed in a two dimensional section. This
difference between the projected and true areal fraction is
called " H o l m e s effect" . Calm and N u t t i n g have given
the following relation between volume fraction and the
average value of the projected area fraction, ^ A Jproj-"
21
25
\ y
7
ft
0
Test Points are the i n t e r s e c t i o n
Points of the g r i d . T o t a l number
of test Points = 9 ; number of test
Poits in ? Phase = 3 .". P = 0 33
P is an average value of P
P
S -t
v
(A )pr jA
'
Fig. 3
74
89
(2)
S = 2N
v
(3)
(4)
(5)
(6)
31
10
37
38
0
T.L.I
T.L.2
I\
.L.3
. i
Thus, the systematic point counting is the most efficient method, provided t h a t there is approximately one
test point per feature of the phase of interest.
T . L . : T e s t line ; X : I n t e r s e c t i o n s
of the
t e s t lines w i t h <p b o u n d a r i e s
N = 15/L ; U
L
B.
of N
is an a v e r a g e
yahie
S = 2N
V
Fig. 4
3 1
La =
... 6
t r i p l e point
1-L
... 7
IT
FA=14/[_
v a l u e of P
; ^ i s a n average
A
L =2P
Fig. 5
39
40
41
C.
39
40
Consider a system
lines in space.
Let L '
ted image.
4 2
4 3
.. 9
L ' i s related t o N b y equation (6).
equations (6) and (9) g i v e s !
L = 2NJt
A
be the
Thus,
combining
... 10
L = 2 P
v
Figure 5 demonstrates
... 8
an
application of this
result.
Transactions of The Indian Institute or Metals, Vol. 34, No. 1, February 19SI
76
D.
Gokhale ;
Characterization of microstruclurcs
H=(K +K ) 2
1
T h e total curvature
d\I
... 11
is
defined as follows :
dM
dS
... 12
of interface
per
M =
v
JJ
H.dS
unit volume
... 13
O M r a r u c c . n ^ r w - J M(R) n ( R ) d R - 4TT J R n ( R ) d R
v
o
.
o
... 15
where R is the size of the largest particle. The integral
on the right hand side of equation (15) is the first moment
of the particle size distribution function.
DeHoftf
has
shown that for a tubule network, M is proportional to
the length of the network per unit volume. The total
curvature is expected to be a very important microstructural parameter for the processes driven by surface
tension.
For example, the rates of grain g r o w t h
and
particle coarsening .
depend significantly on the ratio
of the total curvature to surface area. The total curvature may be an important geometrical parameter in the
kinetics of the spherodization process.
Experimental
measurements of total curvature have been carried out
only in few c a s e s . .
m
46
48
46
47
47
35
51
M = 7 7 . (T )
V
... 16
n c l
where
(T )n.==((T )+-(T )_)
A
...17
... 14
5 0
H =(hV*K )/
2
equation
for
the
dM= HdS
Fig. 6 Definition of local mean curvature.
Transactions of The Indian Institute of Metals, Vol. 34, N o . 1, February 1981
M =2n(N )
T
... 18
M t
(NAW=(N ) -(NJ_
A
... 19
Gokhale :
Characterization of microstructure
77
which truncate
Such truncated
as
1/2.
This
problem
of
n c
contributions ;
surface elements
elements ( M ) .
v
(M -)
v
~23/L
T; = J / L
T , = T;-T;
An
T,
A|K t
= T T ^
52
... 20
53
and DeHoff .
54
have shown
the
...21
volume
of
53
54
the
edge
structure.
elements
Gehl
DeHoff ,
which enclose
( N ) _ is the average
the
number
of
in
the
unit
and
that
not
V
NCT
An example of area tangent count.
Fig. 7
edge
( M ) = i { X . di
v
smooth
the
T , is an a v e r a g e value of
the
=16/l
from
M = (M ) + (M )
v
one
DeHoff *
and
curvature of edges
tangent
count
49
separately to edges
which
(Ma=^
(T )
A
22
Figure 8 ).
precisely
N o t e that
equations (16)
and
(18)
are
statistically
The
of
first points
sweep
by
of contact ) formed
sweeping
test
line
the
If the value of ( T
A rii
. oi ( N )
t
J X
n e ;
is very small or
required
The error
or ( N ) , is the
N E L
n c
and ( N ) _ .
Thus,
if ( t )
A
error involved
and
and
n e t
n e
numbers.
C N ) t are
respect
(T )j_ or ( N )
orientation
the average
possible orientations.
values
obtained
(T ) ,
A
from
0 e
all
and
In
the ( N )
A
n e t
measurement some
error
may
N>3|L
the
N;=H/L ;
N
"net
is an average value of . N .
_
A ,
(
e t
be
Transactions of The
Institute
Metals,
34, N o , 1 of
February
19S1
F i g Indian
. 8 Area
loop of
count
for Vol.
the estimation
My
65
61
48
A.
61
The total
number of sections
'
"' ^ T T ( N ) - M
v
tr
2 4
63
B.
Connectivity
. T h e choice of a parameter which describes the connectivity of a microstructural feature depends upon the
dimensionality of the feature. The connectivity of a
network can be described by its first Betti n u m b e r .
The first Betti number Pj is defined as follows :
66
1 = =
bn-r-P
... 25
where
b = number of branches in the network
62
61
N,
E.A.d
..23
C =B -K -|-N
V
... 2 6
where
C
T h e connectivity per unit volume, C , represents the magnitude of the first Betti number per unit volume. T h e
quantity C represents the number of branches per unit
volume, which may be removed from the network without
creating any new separate parts and thus it is a measure
The quantities
= *_(TAT
1v
.. 29
81
a projected
torsion, T,
53
T , is given by :
v
01
30
dl
dl
of interest in
DeHoff and G e h l
52
the
over
all
the
line
and DeHoff .
54
JA
31
08
0 =lO.d\
The quantity
out
Tv
...27
52
53
DERIVED ROPErTT.ES
0 =7rT
v
vhere x
is
...28
the
average
ormed by a sweeping
is
the
test plane
flicrostructure. T
test
number
of
tangents
be
78
per-
71
measured by
volume of
serial
A,
section-
51
so
H = J J H dS/JJ d s
. _
...32
curvature of the
surface
where H is
the local m e a n
element dS.
it follows that :
is
just
given
by
... 36
... 3 3
S<fi
v
C.
H=M,/S
The
surface
to
surface
distance
average
uninterrupted
between
the
particles
by
p a t h , a , is given by :
surface energy. G i b b s
has shown that the chemical
potential of fine particles is linearly related to
their
= 4(1-V )/S,
processes driven
70
77
60
mean curvature. In some theories of particle coarseni n g ' , the critical particle size is given by the recipro-
37
where V and
are
the
volume fraction
and
surface
particle-
matrix interface.
B.
particles in a
70
cal
71
curvature
of
the
through
the
three
dimensional
73
70
The
mean
intercept
7 of
isolated
4V
71
D.
Contiguity ratio
.. 34
where V, and S are the volume fraction and surface
area of the particulate phase. The mean tntercept
value in the three dimensional structure is the same
v
70
35
C,
Sy/(Sy)ti
38
t o t M
55
E.
18
The
average
nearest
neighbor
distance,
A,
for
1 / 3
82
74
84
:
39
75
where N is the
n u m b e r of point particles per unit
volume. Bansal and A r d e l l have derived the following
expression for the average nearest neighbor distance,
A for the randomly distributed monodispersed spherical particles :
T
88
8V
R . e v
80
Gurland
has shown that for a system of partially
contiguous phase ( say < ) in a matrix ( s a y / ? ) , the
mean intercept of a phase is given by :
y '
V v ) 1
dy
... 40
8V
where R is the radius of the impenetrable monodispersed spherical particles. Equation (40) is not valid for
large volume fractions. A r d e l l has utilized the parameter, A to predict the effect of volume fraction of
precipitates on the rate of precipitate coarsening.
K
J J Hvds
He
F.
... 41
- ^
M,
... 45
97
*dv
... 46
v
89
90
... 47
94
fj K.v.ds
7 1 v-ds
G.
averaged
93
8 8
v is the
It can be
60
82
Jeffrie's m e t h o d
of grain size determination utilizes
(A) ' as the measure of grain size. The ASTM grain
size number is also related to A . Recently, G o k h a l e
has shown that for single phase materials, a is related
to the grain edge length per unit volume.
is
vds
88
d M
dV7
v
.. 48
50
91
52
^ 2 ( M ) e
v
... 42
Lv
PARTICLE S I Z E DISTRIBUTION
H.
T = T /Ly
T
... 4 3
71
95
9G
87
0 '= 0 / L
V
I.
98
... 44
In an evolving microstructure, some derived properties related to the path of microstructural change can
be defined. These derived properties may give important information regarding the evolution process.
common
(1)
(2)
Most of the techniques, with a few notable except i o n s ' , assume that all the particles have a geometrically similar shape ( e . g . . ellipsoids of same axial
r a t i o ) and they differ only in size. The assumptions
regarding the particle shape are not generally verified
when the experimental measurements are carried out.
F o r example, if most of the sections are circular,
it is taken for granted that the particle shape is spherical. However, such a conclusion may be quite misl e a d i n g ' ' . For example, if the precipitate phase
consists of a tubule network along the grain edges of
matrix, most of the precipitate sections in the plane
of polish would be discrete and convex .
59
49
100
55
101
i a
1 / 2
102
0 ( r, R ) d R = 2 { ( R - r )
1 , a
n (R) dR
... 50
55
N
The techniques for detailed estimation of particle
size distribution ( henceforth denoted, by P S D ) require
a measurement of the distribution of some geometrical
variable ( e . g . , section sizes, linear intercepts, etc. ) on
the plane of polish. This measured distribution function
is then transformed into P S D by applying the principles
of geometrical probabilities .
3
A >
( r ) = 2 | (R r ) / n ( R ) dR
r
... 51
( r )
~ ?> n ( R ) d R
/ (R*-r)i'
v
2 r
5 2
A.
d
Consider a system of polydispersed spherical particles distributed uniformly in a matrix.
Let n ( R ) be
the particle size distribution function such that n ( R )
dR gives the number of spheres per unit volume whose
radii are between R and ( R + d P . ) . A two dimensional
section through such a microstructure consists of circles
of various radii distributed uniformly in the two dimensional matrix. Let the variable r represent the radii of
the circular particle sections. Let n ( r) be the section
size distribution ( henceforth denoted by SSD ) such
that n ( r ) dr gives the number of circles per unit area
of plane of polish whose radii are between r and ( r + d r ).
The SSD n ( r ) can be measured from the plane of
polish.
The problem is to estimate n ( R ) front n ( r ).
Let us define
n (R)
(r)
2
- 2TT
- dR
- t Hk (r R ) /
dr
... 53
108
... 54
Mj = J (R)' n (R) dR
... 55
and
10
rn
- Pt
+ 1
M,
... 56
where,
N > (r) = J n (r) dr
A
... 48
+1
.. 57
Uukhalc
where,
S3
r(o+i)orf)
r(i/2)=v^;r(i)=i
... 58
Characterization of mkfoitrwtures
follow-
expression
110
(2)
(3)
11
118
CO
Nv=2- \ - ^ - d r
... 5 9
104
105
10
...60
(2) N
V j
, N
number
of
A,
V |
spheres
... N
... N
V j
represent
Vfc
the
2 A , j A , k A , respectively.
(3) n
A i
, n
, ... n
A j
A [
A, A to 2 A , ( i - T )
to
i A ,
( k I ) A t o k A , respectively.
(4) N
T j
N . , ... N
v
, ... N ,
V j
1 0 7 - 1 0 9
"AJ , % . A , , A
);
equation :
1 1 0 - 1 1 2
113
114
115
116
1
=--
p (j, i) n
... 61
i=j
where |8 (j, i) are the Saltykov's coefficients, and these
are tabulated by Saltykov and DeHoff .
0
119
1 2 2
122
84
Gakhale
Charactization
of
microstructures
..area
distribution.
Saltykov's
method
is more flexible
119
110
*S - A.KCQj
!/CJ,i)N
...62
ratio,
Q, of the ellipsoids.
prolate ellipsoids,
axes
of
the
then n
of
polish.
used
major
The
other
notations
123
0.613
is the distribu-
A (
The N
V j
values can
Myers
has given a technique for estimation of
P S D of cubic particles. The input information required
is the measurement of altitudes of the four-sided sections
on the plane of polish. He has shown that the altitude
of a four-sided section is closely related to the edge
length of the cube, and furthermore, the four-sided sections form a fixed fraction of the section shapes observed.
Estimation of P S D of polyhedral particles is considered
by Scheil and W u r s t , Hull and H o u k
and P a l a u s .
The experimental work of White and
Van V l a c k
indicates that the pentagonal dodecahedran
is an adequate
model shape for the estimation of
cell size distribution.
120
be calculated
0.0079 N
0.0038 N
A
A
0.001 N
A ]
- 0.0003 N
0.0018 N
A l
A
j5
A j
- 0.0002 N
A j
A Q
]7
A j
...63
twelve, and N
J4
A
A
JS
0.0002 N _ . ]
where j can
j 3
j2
up to and including
13
132
133
1 5 8
127
129
the P S D
A a
131
Saltykov
to 0.613 A ,
= ~ [1.6461 N
- 0.451 N,
,
j
Dj
j
*J-i
0.1162 N
0.0415 N
0.0173 N ,
100
to estimate
... N .
and
respectively.
Recently, C r u z - O r i v e
has given a method for
estimation of size and shape distribution of ellipsoidal
particles.
He has shown that a unique solution to
130
A 2
given by Wicksell".
Johnson
, N
If the
in equation (61).
by T a l u s
A 1
lar sections
(3) N
0 3
classes.
on
(0
2 1
11
(R)
y (,/R) d R
... 64
a (R) . n (R) d R
v
148
151
103
length
distribution
of one
c is the shape factor. F o r convex particles, a (R) is onefourth of the total surface area of the particle.
F o r spherical particles, it can be shown t h a t
V 0/R) =
i
2R
1 3 6 - 1 3 8
... 65
and
o (R) = v R"
... 66
1 5 1
(/) -
equation (64)
(/))
dl
6 7
/
136
139
140
141
142
143
DeHoff and B o u s q u e t
have given the following
result for the geometrically similar triaxial ellipsoids.
1 4 4
the
DISCUSSION
1 4 6
Itoh
and
Coleman '
have considered the
intercept distribution of cubic particles. I t o h
has
derived
an analytical expression for the intercept
distribution of a cube of unit edge length. W a r r e n
has
experimentally verified I t o l r s f o r m u l a for cubic particles.
1 3 4
146
134
36
are
many unsolved
problems
in
stereology.
21.
22.
on
fractured
projections
of
investigation.
estimation
surfaces,
curved
Several
curved
surfaces
need
questions
surfaces,
a
plane
of further
the
'23.
H o w many n u m b e r of
24.
arise
of size distributions.
lot
regarding
do
the
distribution
performed
to
ment
stereology
of
estimate
when
size
is
not
complex
calculations
distribution ?
yet
are
T h e develop-
complete and
further
and
Calculations,
27.
REFERENCES
1.
2.
3.
4.
5.
6.
S. A. Saltykov. Stereometric
Metallurgizdat, Moscow, (1958).
7.
8.
Metallography,
29.
30.
2nd
ed.,
34.
37.
38.
39.
40
42.
12.
43.
13.
44.
45.
15.
46.
47.
48.
49.
50.
Differential Geometry,
of
Florida,
20.
1981
Gokliale :
-.82.
56.
57.
Mod.
84.
85.
S. Chandrasekhar, Rev.
87
87.
54.
Characterization oj micro'siruciures
Soc.
A. M. Gokhalc, to be published.
90.
Springer-Verlag,
93.
94.
and
K.
R.
Craig,
66.
Gibbs,
70.
It,
72.
75.
76.
R.
L.
77.
R.
L.
78.
E.
Orovvan,
80.
81
1,
97.
98.
99.
100.
107.
Vol.
108
109.
110.
Log-Normal
Cambridge,
Transa ions of The Indian Institute of Metals, Vol. 34, N o . 1, February 1981
88
114.
Gokhale :
Characterization of microslrnctures
117.
118.
12!.
11
134.
135.
136.
137.
138.
139.
140.
123.
124.
144.
126.
127.
12S.
129.
130.
149.
150.
(1970),
148.
Transaction of The Indian Institute of Metal, Vol, 34, No. 1, February 981
Brook,
G. B.
Dissertation.
University
of