JOURNAL OF MATERIALS S C I E N C E 11 ( 1 9 7 6 ) 2105--2141

Review
The physical properties of composite
materials
D. K. H A L E
Division of Chemical Standards, National Physical Laboratory, Teddington,
Middlesex, UK

In this review, the physical properties of composite materials are discussed; however,
discussion of the mechanical properties has been excluded except when necessary for the
consideration of properties such as thermal expansion or swelling and shrinkage. One of
the main aims in the review has been to show how the theoretical and experimental
information that is already available may be used (a) to design and construct composite
materials with predetermined physical properties and (b) to ensure that the physical
properties of composite materials are properly measured and properly defined.

1. Introduction For the purpose of this review it will be

The outstanding mechanical properties of fibre assumed that a composite material is a hetero-
composites and especially the unique combi- geneous mixture of two or more homogeneous
nations of low density with high strength and phases with at least one continuous solid phase or
stiffness which can be achieved with these ma- an interconnected mixture of solid phases such
terials, have led not only to extensive research on that the geometry is fixed and the material can be
the mechanical properties of composites, but also considered as a solid. The dimensions of a phase
to a highly developed technology. In comparison, are assumed to be greater than about 10nm
relatively little effort has been devoted to the (100 A). On a macroscopic scale, that is on a scale
other physical properties of composites. Properties which is large compared with the pore or particle
such as thermal and electrical conductivity, diameter or with the periodicity of the structure,
thermal expansion, swelling and shrinkage are also the material then behaves as a quasi-homogeneous
important and there is not only a need to improve solid with its own characteristic properties. The
and optimize these properties but also the possi- essential problem that has to be considered is that
bility of devising materials showing novel thermal, of how the properties of the composite depend on
electrical, magnetic or optical behaviour~ the properties of the separate phases, their volume
The physical properties, other than mechanical, fractions, dimensions and geometry. To simplify
have been discussed in a number of reviews [1-7] the discussion it will usually be assumed that there
but most of these have been concerned primarily are only two phases and that voids are absent, or if
with the properties of directionally solidified they should be present, as in a foamed material,
eutectics, or with non-structural applications then the voids can be regarded as one of the two
rather than basic properties. This review is con- phases.
cemed essentially with the relationships between The selection of a suitable plan for the dis-
the physical behaviour of composite materials and cussion of the physical properties of composites
their structure. Discussion of the mechanical presents some difficulties. The scheme adopted in
properties is excluded except where some con- this review is similar in many ways to that used by
sideration of the mechanical properties is necess- Albers [1] in discussing the physical properties o f
ary for the discussion of properties such as thermal in situ composites. Two broad headings have been
expansion or swelling and shrinkage. used to divide the physical properties into:
9 1976 Chapman and Hall Ltd. Printed in Great Britain. 2105
 (1) properties determined by the properties of The coefficient of thermal conductivity is
the constituent phases, their volume fractions and defined by Fourier's law which for an isotropic
the composite geometry, and medium may be written in the form
(2) properties depending on structural factors
J = --)k grad T (3)
such as periodicity and particle or fibre dimen-
sions. where X, the thermal conguctivity, is the pro-
These two groups of properties are considered portionality constant between the heat flux
in turn. In Section 2 we examine the way in which vector J and the temperature gradient. Other
the value, say A*, of a given property depends on transport coefficients are defined in a similar
the fixed values A1, A2 . . . for the component way as proportionality constants between fluxes
phases. T h e properties considered in Section.3 and gradients. Examples are the electrical conduc-
depend primarily on structural factors such as tivity (Ohm's law) and diffusion coefficients
periodicity, inter-fibre spacing, particle dimensions (Fick's law).
etc (cf. the importance of the inter-particle dis- With a composite material, the relations derived
tance on the mechanical properties of dispersion- for all these properties will be formally identical
strengthened metals). The properties considered and the expressions obtained for the dielectic
include the optical, magnetic and superconducting constant, for example, will be equally applicable
properties; heat transfer under conditions where to the thermal conductivity or magnetic
phase dimensions are important is also discussed. permeability.
In Section4 we consider the special case of In this section, we therefore first consider the
"product properties" to which attention has been dielectric constant of a composite and then the
drawn by van Suchtelen [8] and Albers [1]. The mathematically related properties such as the
discussion of some special properties including electrical and thermal conductivities. Next we
those associated with the Hall effect and resulting consider the thermal expansion together with the
from a combination of different properties forms swelling and shrinkage behaviour. Properties as-
the subject of the final section. sociated with time-dependent vector fields, in
The aim of this review has not been a compre- particular dielectric loss, fall outside the scope of
hensive account of the physical properties but this review.
rather a discussion of representative properties To simplify the discussion, it will be assumed
with some indication of the ways in which that the composite consists of two phases only,
"materials engineering" can be used to design and although in some cases general relationships for a
construct composite materials with predetermined property of an N-phase composite may have been
properties using the theoretical and experimental derived. It is also assumed that both phases are iso-
information that is at present available. tropic. We have to remember, however, that even
if the phases are isotropic in a particular property,
2. Properties determined by the properties the composite m a y be anisotropic in structure so
of the constituent phases, their volume that, with the composite, the property may be
fractions and the composite geometry different in different directions.
2.1. General considerations It is assumed that the composite is statistically
Many of the physical properties of materials are homogeneous, i.e. if we take small elements of the
described by a functional relation between a solen- material then these will all have the same physical
oidal vector and an irrotational vector: for properties as the whole sample provided that the
example, for a linear isotropic material we have elements are large enough compared with the
the relations periodicity or dimensions of the phase regions to
ensure that they are representative. Within an
O = eE (1)
element, however, the geometry may be regular or
where D is the electric induction, e the dielectric i~regular~ isotropic or anisotropic. We may, for
constant and E the electric field and example, have a random dispersion of spherical
particles in a continuous matrix, a regular array of
B = .I-I (2)
aligned filaments, continuous laminae, or a com-
where B is the magnetic induction, # the magnetic pletely irregular geometry with continuous or
permeability and H the magnetizing force. discontinuous phase regions (Fig. 1).
2106
 the phase geometry is available, still closer bounds
I- i
r-,

" )
can be set on the possible value o f e*. The same is
',_) I " true o f other properties. F o r a few special geo-
metries exact solutions can be obtained. Approxi-
mate solutions have also been obtained for two-
phase composites with various geometries but
b these are, in general, only applicable when the
volume fraction o f one o f the components is low.

2.2. Dielectric constants

f The earlier theoretical work on the dielectric be-
haviour o f heterogeneous systems has been de-

i
c d

Figure 1 Composite geometries: (a) random dispersion of

scribed in some detail in the critical review b y
Van Beck [10]. The following summary includes
some o f the more important relationships derived
in this earlier work together with results obtained
in more recent theoretical investigations. The di-
electric constant o f the composite is represented
spheres in a continuous matrix, (b) regular array of b y e*, the upper and lower bounds by e~+) and
aligned filaments, (c) continuous laminae, (d) irregular @ ) , and the volume fractions o f the two phases by
geometry.
V1 and V2. We first consider the effective dielec-
tric constants o f isotropic composites and then, in
The physical behaviour of a composite can be view of their technological importance for other
predicted with an exactitude which depends on properties, the relations for the dielectric con-
the precision with which the phase geometry can stants o f fibre composites.
be specified. In a fundamentally important paper
[9], Brown showed that the dielectric constant o f 2 . 2 . 1. I s o t r o p i c c o m p o s i t e s
a composite must depend not only on the Exact relations are available for very dilute disper-
dielectric constants o f the two phases and their sions o f particles in a continuous matrix. Some o f
volume fractions but also on the composite geo- the more important expressions for the dielectric
metry. A t t e m p t s to derive exact theoretical ex- constants o f suspensions containing a small volume
pressions o f general applicability for the dielectric fraction (V1) o f randomly oriented particles
constants o f mixtures o f unknown phase geometry (dielectric constant e l ) in a continuous matrix
are futile. (dielectric constant e2) are included in Table I.
The dielectric constant e* o f a mixture o f two Equation 4 is the Rayleigh mixture formula which
phases must, however, lie between certain limits was first derived for a cubical array o f spheres. If
whatever the geometry, and if some knowledge o f VI (el --e2 )/(2e2 + e l ) ~ 1 it m a y be written in the

TABLE I Formulae for the dielectric constants of isotropic two-phase composites containing randomly orientated
particles in a continuous matrix
System Dilute suspension References SCS approximation References
em = e2 em = e*

Spheres 2e~ + e I + 2V l(e~ --e~)

e*=e: 2e~ +e 1 - - V I ( % --e2) (4)[11]

3 V1e 2(e 1 - - e : ) 3 V 1 e*(q -- e 2 )

e*=% + (5) [10] e*=e 2 + (9) [13]
2e2 + e I 2e* + e l
Discs or V~ (e 1 -- e 2)(e 2 + 2e~ ) V~(e 1 -- e2)(e* + 2el )
lamellae e* = % + 3e~ (6) [12] e* = % + 3el (10) [12]

Rods, needles e* Vt (el -- e2 )(5e2 + el ) . V1(e~ -- e~ )(5e* + e 1)

orfibres =e~ + 3(e~ +e2) (7)[101 e =e2+ 3(e*+e~) (11) [10,14]

2107
simpler form given by Equation 5. This relation, rounded by its own shell of matrix with a
together with Equations 6 and 7 for dilute suspen- thickness appropriate to the volume fraction. With
sions of discs and rods, may be obtained from the this structural geometry which Hashin terms the
general expression derived by Polder and Van composite spheres assemblage, the dielectric con-
Santen [15] for the dielectric constant of a dilute stant of the composite is given exactly by
suspension of ellipsoids with a random distribution
of orientations Wl
e* = e 2 -[-
[1/(el -e2)] + (V2/3e2)"
em
e* -~- e 2 -I- V I ( e l - 6 2 ) 2 (13)
3 abe [e= + a i ( e , - e=)]
(8) For the corresponding case in which material 2
is the particulate phase and material 1 is the
where em is the effective mean value of the dielec- matrix, the dielectric constant of the composite is
tric constant of the medium around each particle given by
and the Ai are the depolarizing factors along the
ellipsoid axes a, b and c. Values for depolarizing
r -- C1 +
factors have been tabulated by Stoner [16] ; for [1/@2 - - e l ) ] + ( V l / 3 6 1 ) "
spheres the depolarizing factors are A i = 89
(14)
For very dilute suspensions, em is equal to e2
the dielectric constant of the matrix. At higher It may be noted that Equations 13 and 14 are
concentrations, the disturbing effects of neigh- mathematically equivalent to the corresponding
bouring particles on the field have to be con- Rayleigh mixture formulae given by Equation 4.
sidered. One way of doing this is to use the self- Although real composite materials may have a
consistent scheme (SCS) in which it is assumed structure approximating to the composite sphere
that each particle is surrounded by the composite assemblage, there will be uncertainty about the
material with a dielectric constant e* rather than degree of approximation involved in using
by matrix with a dielectric constant e2. Setting em Equations 13 or 14, The same is true of the SCS
equal to e* in Equation 8 then leads to Equations approximations and the other approximate re-
9, 10 and 11 in Table I. Equation 9 which was lations which have been derived for the dielectric
derived by B6ttcher [13] can be shown to be sym- constants of isotropic composites.
metrical in V1 and V2 (= 1 -- V1) and it can be Rigorous upper and lower bounds can, how-
argued that it is an appropriate approximation for ever, be set on the value for the dielectric constant
a conglomerate of particles of two different ma- of a composite material and the use of these
terials. bounds may clearly be preferable to the use of
The assumption that each particle is surrounded uncertain approximations. The earliest bounds for
by a continuum with a dielectric constant e* is, the dielectric constant of a two-phase isotropic
however, clearly an over-simplification. A more composite were those obtained by Wiener [18]
satisfactory assumption with a concentrated sus- and given by Equations 15 and 16 in Table II.
pension of spherical particles might be to suppose These bounds have also been obtained by Brown
that each particle (radius a) is surrounded by a using two minimization theories of classical
shell of matrix (outer radius b) and that this, in electrostatics [19] and, since the expressions are
turn, is surrounded by composite material. The identical with those for the longitudinal and trans-
extension of the SCS approach in this way has verse dielectric constants of a composite with
been discussed by Hashin [17] in a critical assess- alternating laminae, they are also the best possible
ment of the SCS approximation. A special bounds in the absence of any information on the
situation arises when geometry or degree of isotropy. The form of these
bounds for ea/e2 =0.1 and 0.01 is shown in
a3 V1
- (12) Fig. 2.
ba 112
Closer bounds for the effective dielectric con-
since it is then possible to construct a system in stant of an isotropic composite were derived by
which the space within the material is completely Hashin and Shtrikman [20] using variational prin-
filled by spheres of varying diameter each sur- ciples and are given by Equations 17 and 18 in
2108
TABLE I I Bounds foi the dielectric constants ofisotropie
two-phase composites I'O ~ (.--J-=O'l

Wiener [18]
I\'R- "'
el§) = qV~ + % V 2 (15) o.8 -iX NN,,
',\ \ N ' ,
1 ,. o.,-', \ \ \ ' - , ..,
e~_) = (I,, /ex ) + (V21%) (16)

0'4 ~?, x
Hashin and Shtrikman [20]
* V1 (17) o.2 I-
~(+)= ~ + [1/(~1 -~01 +(G/3q)
! ! I ] I
G (18) 0 0 ':) 0"4 0'8 0'8 I'0
e~_) = e, + [1/(% -- e, )] +(V, 13e~) VI

where e 2 > e 1

Beran [ 22 ] I'Ol~~
iX , c, =o.o,

~) = [<e)-- 3<e>+ (9(e>=/((e')2))lJ

<(e')2> ] (19) 0"8 I ~ \ x Cz
s

e~_) =
[<l) (4/9(e)Z)(e'/e}2 j] -1 (20) E-z20"6LtL XNx,,x
O13<~>=)<(J)=I~>+ i \~ ",,o )
where 0.4 \
1 r r [5 {e(O)e(r)e(s)~\risi
2 ." -: -# - 8 xx'xx
I = - -
f6~=<,,= jvsJvr~ ~ ~drd$ 0 . 2 \, "i?)
\ (3X,\ ( a \ "\,,,,

j - 1 fVsL(62<e'(r)e'(s)/e(O))~risi.. ( i ~ ' . ~
161r~<e)2 -8r -a 6s a l/ r~ s aras
0 0'2 0.4 0"6 0" 8 "O
VI
e(x) = dielectric constant at point x
e'(x) = e(x) --(e} Figure 2 Dielectric constant of composite materials
(e) = ensemble average of e (x) (assumed to (e 1/% = 0.1 and %/% = 0.01). (1) Wiener bounds
be equal to the spatial average) (equations 15 and 16), (2) Hashin and Shtfikman bounds
(d(O)e'(r)e'(s)> = ensemble average of the product of e' for arbitrary geometry (Equations 17 and 18), (3)
at some co-ordinate origin within the B6ttcher SCS approximation for spheres in a continuous
composite, e at a position r from the matrix (Equation 9).
origin and e' at position s.
F~ S = sets of three orthogonal space vectors
with a common origin. pared with the Wiener bounds and the SCS
rr, Fs = volumes of infinite spheres in vector approximation given by Equation 9 for a suspen-
spaces denoted by subscripts. sion o f spherical particles. The dielectric constants
o f all isotropic composites containing dispersed
Table II. The Hashin and Shtrikman bounds spheres in a continuous matrix should, of course,
always lie within the Wiener bounds and, since fall within the Hashin and Shtrikman bounds and
they correspond exactly with Equations 13 and 14 if the values o f el and e2 do not differ greatly
for the two composite sphere assemblages, they these bounds can show good agreement with
are the best possible bounds for the dielectric con- experimental results. A good example is shown in
stant of an isotropic two-phase material if no Fig. 3 in which the Hashin and Shtrikman bounds
structural information apart from the volume frac- are compared with Reynolds' experimental results
tions is available. If the values o f el and e2 do not [10,21] for a suspension o f glass spheres in
differ greatly the bounds are close together, but carbon tetrachloride.
when the ratio o f e2 to el is high the bounds If some information on the structural geometry
become widely separated as shown in Fig. 2 in is available, then closer bounds may be set of the
which the Hashin and Shtrikman bounds are corn- value o f the dielectric constant for an isotropic
2109
 4-6 sandstone and firebrick (Fig. 8). The Beran bounds
showed a significant improvement over the Hashin
4.2
and Shtrikman bounds but the comparison with
experimental results cannot be considered rigorous
since it was assumed that the pl~ase geometries of
3"8 the porous sandstone and firebrick were similar to
s
that of the iron-lead composites used in the
3"4 experimental measurement of the spatial corre-
lation functions.
As we have seen, the derivation of the Beran
3"0
bounds presents serious problems. Miller [27],
however, showed that for a broad class of isotropic
2"6 two-phase composites which he termed cell ma-
terials, the problem could be circumvented. For
2"2 these materials, he deduced that the functions of
0 0"2 0"4 0"6 0"8 I'0 the three-point correlation function which ap-
VI
peared in the bounds for e* were simply a number
Figure 3 Dielectric constant of composite materials: for each material with a value between 1 and 89
Hashin and Shtrikman bounds and experimental results with the number having a simple geometric signifi-
for glass spheres in carbon tetrachloride [ 10, 21 ].
cance (e.g. ~ for a spherical cell). Miller obtained
composite. To do this, statistical information is bounds for two types of material-symmetrical
needed in addition to a knowledge of the volume and asymmetrical cell materials.
fraction. Mathematically, this information can be In the symmetrical cell materials the space
introduced through n-point correlation functions within the composite is subdivided by closed sur-
which are related to the probability that n points faces into closed regions or cells so that the
thrown at random into the composite will all be in following requirements are met:
the same phase. Following the earlier work by (1) the space is completely filled by cells;
Brown [9], Beran [22] obtained bounds for the (2) the cells are distributed in such a way that
dielectric constant using a variational principle the material is statistically homogeneous and
which included the three-point correlation func- isotropic;
tion. These bounds are given by Equations 19 and (3) the dielectric constant e of a particular cell
20 in Table II. The application of statistical is statistically independent of the dielectric con-
continuum theories of this kind to the prediction stant in any other cell;
of the properties of heterogeneous materials has (4)the conditional probabilities of n points
recently been examined by Corson [ 2 3 - 2 6 ] . For being and m points not being in the same cell of a
the experimental determination of a correlation particular material, given that one point is in a cell
function, the associated probability functions have of that material, are the same for each material.
to be measured. The technique developed by Miller gives three examples of symmetrical cell
Corson involved preparing sections of the material, materials. The first example is the Poisson cell ma-
taking random photomicrographs, superimposing terial in which the space within the composite is
a matrix of points on each of the photomicro- divided into convex polyhedron-shaped cells. In
graphs as a sampling grid, and then noting in which the second example, the cells no longer have plane
phase each point fell. From the experimental or convex sides but are star-shaped instead. The
observations, the two- and three-point probability third example is a composite built up of spheres
functions can be obtained and these can then be with diameters varying in such a way that the
used to form the two- and three-point correlation space within the composite is completely filled.
functions. (The procedure used by Corson for With the asymmetrical cell materials the geo-
deriving these correlation functions is complex and metry of the cells of the two materials is dissimi-
the original papers should be consulted for lar. In this case, the space within the composite is
details.) Using the correlation functions, Corson subdivided so that the following requirements are
evaluated the Beran bounds and then compared fulfilled:
these with experimental data obtained with porous (1) the space is completely fdled by cells;
2110
T A B L E IlI Miller's bounds for the dielectric constants ofisotropic two-phase cell materials [27]
~4)(1 -- 4))(a -- 1) 2 ]
e~§ = e 2 1 + 0(c~-- 1) -- 1 + (c~-- 1){4) + 3[(1 --4))2G~ --4)2G~ ] } (21)

}4)(1 -- 4))(a -- 1) 2 ]-'

e~_) = % [~ - - 4)(~ - - I ) ] - - 1 + a + 3(~ - - I ) [ ~ ~ G2 - - (1 - - 4))2 G I ]J (22)

where 4) -- V I , 1 - ( ~ = V2, cz = e l / % and G I and G 2 are shape factors for the t w o materials (~ for spheres, 31--for discs,
and a-
6 for needles).
For asymmetrical cell materials Gx r G2
For symmetrical cell materials G~ = G 2 = G

(2) the cells are distributed in such a way that fore, be found by letting r approach zero or unity
the material is statistically homogeneous and and a approach unity and then setting the solution
isotropic; equal to the small concentration solution to solve
(3) the dielectric constant e of a particular cell for G. If an exact small concentration solution is
is statistically independent of the dielectric con- not available (e.g. with particles of irregular shape)
stant in any other cell. a G value may be obtained from experimental
An example of an asymmetrical cell material is measurements on a suspension at low concen-
a composite formed from spheres and cubes of trations.
varying size. It is assumed that the space within 9 This cell model would seem to be intrinsically
the material can be completely filled without any suitable for the representation of the behaviour of
preference as to whether a cube or sphere is in- materials with a cell-like structure such as eutectic
serted in any given location. alloys or sintered materials. The appropriate G
For these two-phase cell materials, Miller ob- values will then be determined by the geometries
tained the bounds given by Equations 21 and 22 in of the component cells. The Miller bounds can
Table III. The shape factors G1 and G2 are simple also, as described below, be applied to systems in
numbers each lying between ~ and 89 and charac- which particles are dispersed in a continuous
teristic of the cell geometries of the two materials. matrix, i.e. suspensions. With, for example, a sus-
If G1 is equal to G2 then these equations represent pension of spheres, the space between the spheres
the bounds for the symmetrical cell materials. The which is f'filed with matrix material is assumed to
bounds shown in Table III have also been derived be made up of cells of all shapes and sizes so that
by McCoy [28]. the fundamental requirements of the asymmetrical
The extreme upper bound is given by Equation cell model are satisfied. The bounds are then ob-
21 with G1 = 89 and G2 = ~ ; the extreme lower tained by the following method. At low concen-
bound by Equation 22 with G 1 = ~ and G2 = 89 trations of material 1, i.e. when q~ is small, the
For all possible values o f a, r G1 and G2, these bounds given by Equations 21 and 22 can be
bounds for the asymmetrical cell materials fall written in the form
qe
inside the Hashin and Shtrikman bounds. This is,
e(:O - 1 + r ~1--
at first sight, unexpected since the Hashin and
Shtrikman bounds actually correspond to the
62 t 3[l + ])a,]
relations f o r a real physical geometry (the com- (23)
posite spheres assemblage discussed above). This and
particular geometry cannot, however, be rep-
resented by the asymmetrical cell model. -1+r
1)
When r approaches zero or unity, and a also e 2 ot
approaches unity, the bounds for the symmetrical
cell material (Ga = G2 = G) converge and become I-I 3[l+cz--3(a--1)G1] " (24)
coincident with the exact solution for the dielec-
tric constant of a suspension containing particles If e~+) is set equal to e(*), a quadratic
of a corresponding shape at very low concen- equation in G 1 is obtained with the roots G1 =
tration. The appropriate values for G which should 89 and G1 = ~. If G 1 is set equal to ~ then
be used in these bounding equations can, there- E q u a t i o n s 23 and 24 become equivalent to
2111
Equation 5, the solution for a dilute suspension of meets the requirements of Miller's asymmetrical
spheres, so that Miller concludes that Ga = ~ has cell model. A G value of ~ is obtained for
the geometrical significance of a sphere. needles [27] by solving Equations 23, 24 and 7
The bounds for a suspension of spheres when for small perturbations (i.e. a approaches unity).
the spheres have the higher dielectric constant, are The bounds for particle suspensions obtained
then obtained from Equations 21 and 22 by using Miller's model are substantially closer than
setting Gx = ~ and G2 equal to a value in the the Hashin and Shtrikman bounds for volume frac-
range ~ to ] which gives the highest upper and the tions up to about 0.6. In practice, however, there
lowest lower bound. If the spheres have the lower may be some doubt as to when a system can be
dielectric constant then G2 is set equal to ~. The represented by the asymmetrical cell model. The
Miller bounds for the dielectric constant of a composite spheres assemblage in particular cannot
suspension of spheres in a continuous matrix be represented in this way since the dielectric
(el/e2 = 10) are shown in Fig. 4 together with the constants for systems of this type lie outside the
corresponding Hashin and Shtrikman bounds and Miller bounds.
the Rayleigh mixture formula.
For a suspension of randomly oriented discs, 2.2.2. Aligned rods, needles, or fibres
we can set G1 in Equations 23 and 24 equal to 89 I f the rods or fibres are o f u n i f o r m cross-section
and then these equations become equivalent to and aligned in one direction then the dielectric
Equation 6 so that G1 = 89 has the geometrical constant of the composite in the axial direction
significance of a disc. The bounds are then ob- (e~) is given by
tained from Equations 21 and 22 by setting G1 =
89and G2 to a value in the range -~ to ~ which gives e.~ = Vie1 q- I72e2. (2s)
the highest upper and the lowest lower bound.
Similarly, closer bounds on the dielectric This result is independent of the transverse
constant of a composite containing randomly geometry.
oriented needles can be set when the geometry For a dilute suspension of fibres with a circular
cross-section, the transverse dielectric constant is
I0'0 given by Rayleigh's formula for cylinders [10, 11]

e~ -- e 2 e 1 -- e2
- V 1 - (26)
e§ + e= el + e2

'~
(...._* 6.0

4"0-
-

Z Using a self consistent scheme in which g-- e*,

Davies [14] obtained the expression

2 7 , e§ - e=)
e} = e2 + (27)
e~ + e l
2-0 ~ ~ - (g)
which corresponds with that given by Peterson and
Hermans [29] for aligned cylinders perpendicular
I I I I
0 0"2 0"4 0'6 0"8 I'0 to the field. Numerical solutions have alSO been
VI given by Springer and Tsai [30].
For a composite cylinder assemblage similar to
Figure 4 Equations for the dielectric constant of a
the composite sphere assemblage described above,
suspension of spheres in a continuous matrix
(%/% = 10). (1) B/~ttcher's SCS approximation Hashin [5] showed that
(Equation 9). (2) Upper and lower bounds for Miller
asymmetric cell model (Equations 21 and 22). (3) Upper V1
Hashin and Shtrikman bound (Equation 17); composite e,~ = e2 + [1/(el --e2)l +(V=/2e2)" (28)
spheres assemblage with spheres of material 2 in material
1 (Equation 14). (4) Lower Hashin and Shtrikman bound This relationship is mathematically identical to
(Equation 18); composite spheres assemblage with Equation 26.
spheres of material 1 in material 2 (Equation 13);
Rayleigh mixture formula (Equation 4). (5) Rayleigh Hashin also showed that the bounds for the
mixture formula for very dilute solutions (Equation 5). transverse dielectric constant of a transversely
2112
isotropic fibre reinforced material were given by where, as before, a = el/e2 > 1. G 1 and G2 are
V1 two numbers lying within the range 88to 89which
CT(+) = e2 + characterize the fibre and matrix geometries. A G
[1/(el -- e2)] + (V2/2e2)'
value of 88corresponds to a circular shape and G =
and (29) I to the shape of parallel lamellae (two-
V2 dimensional needles). These bounds always lie
eT(_ ) = 61 -l-
[1/(e2 - el)] + ( V l / 2 e l ) " between the Hashin bounds. Beran and Silnutzer
consider the fibre geometry to be of first import-
(3o) ance since at low concentrations the term V~2G1
Since Equation 29 is identical to Equation 28 will be much larger than V] G2.
and Equation 30 with the result for a composite For circular fibres, the bounds are quite close
cylinder assemblage in which material 2 is the fibre up to volume fractions of about 0.2 when ot = 10
and material 1 the matrix, it follows that and up to about 0.1 when a = 100. In both cases
Equations 29 and 30 are the best possible bounds the bounds are close to the Hashin lower bound.
for the transverse dielectric constant of a trans- When the volume fractions are of the order of
versely isotropic fibre composite if the only infor- 0.50, however, the bounds diverge considerably.
mation available is the volume fractions and the Beran and Silnutzer concluded that the packing
dielectric constants of the two phases. geometry is extremely important and that
When additional geometrical information is changing the positioning of the fibres would have a
available, closer bounds can be set on the trans- significant effect on the value of e~..
verse dielectric constant. Thus, using a composite This analysis has been taken a stage further by
cylinder model similar to that used by Hashin and Elsayed and McCoy [34] who have derived
Rosen [31] for mechanical properties, Donea bounds which incorporate additional geometrical
[32] derived bounds for the transverse thermal information in the form of further shape factors
conductivity of a composite containing aligned and two packing parameters.
fibres of identical circular cross-section arranged
in a square or hexagonal array. 2.2.3. Fibres random/y oriented in a plane
Bounds for the transverse dielectric constant For fibres randomly oriented in a plane, Davies
of an aligned fibre composite with additional [14] obtained the SCS-type relations
information on the fibre geometry have also been
2V1 ez(el - e2)
obtained by Beran and Silnutzer [33] using a ez = C2 + (33)
method of analysis closely similar to that followed ez +el
by Beran and Miller [22, 27] in deriving bounds Vl(e I --62) (131 -1- 3fix)
for the isotropic three-dimensional composite. It ex ~ ey = e 2 +-
2(el + e l )
was, in fact, only necessary to rewrite all the
previous results using two rather than three dimen- (34)
sions. This then gave the following equations for
2.2.4. Conducting inclusions
the upper and lower bounds
For a composite system containing dispersed con-
e~'+) _ 1 [1 ducting particles el is very much greater than e2
(ele2)l/2 al/2 L + g l ( a - 1)- and approximate equations for the dielectric con-
stant can then be derived for el + oo. For example,
1 V l V2 (o( - - 1)2 ] for spherical particles, when el >> e2 and g = e2
1 + (~ - 1) (V 1 -}- 2V~ G l - - 2V] G2)J '
we have the relation

and (31) e* = e2(1 + 3 V , ) (35)

instead of Equation 5. For insulators (e.g. air
(ele2).2 vl + v2~- cavities) in a matrix with a high dielectric constant
el < e2 and for spherical inclusions we then have,
e* = e2(1 3 Vt). (36)
(c~- 1)(I/:]G2 - V~Cl) + (a/2)+(1 -- a/2)Vl
The derivation of these and similar relations has
(32) been discussed by van Beek [10].
2113
2.3. Electrical conductivity
All the expressions given in the previous section
,o /
are also valid for the electrical conductance of a
two-phase composite. It is only necessary to 30
replace el, e2, e*, e(+), e(_) etc by the conduc- O
tances O'1~ g 2 ~ 19"* , o(§
* e~_) etc. Many of the 2O
expressions derived specifically for the resistance
or conductance of a composite are indeed math- IC I I I I
ematically identical with corresponding ex- o 0.4 0.8
pressions for the dielectric constant. For example, Volume fraction Sb
Maxwell's equation [35] for the specific resistance Figure 5 Comparison of experimental values for the
of a dilute suspension of spheres is resistivity of the Cu2Sb-Sb system [167] with
theoretical curve obtained from Equation 38 (after
201 +P2 + VI(Pl --P2) Landauer [361 ).
P* = 2pl + P 2 - - 2 V l ( p l " P 2 ) p2. (37)
500
When the specific resistances are replaced by
the corresponding conductances, this becomes
mathematically identical with the Rayleigh
mixture formula for dielectrics (Equation4). 3OO
Similarly, the expression for the electrical conduc-
tivity of a two-phase composite derived by
Landauer [36] using a self-consistent scheme
approximation, IOO

O 1 -- G* 0"2 -- O* I I 1 I
V 1 - - + V 2 - - - 0 (38) 0 0.4 0'8
oi + 20* a2 + 20*
Volume f r a c t i o n Pb
corresponds with Equation 9 derived by B6ttcher
Figure 6 Comparison of experimental values for
for the dielectric system.
conductivity of the Mg=Pb-Pb system [168] with
In applying the expressions given in Section 2.2 theory. (1) Equation 17. (2) Equation 18. (3) Equation
to experimental results, it must however be 38. (After Hashin [5] ).
remembered that they involve the implicit assump-
tion that interface effects such as space charge fit the theory. The experimental values in this case
polarization are absent, i.e. it is necessary to do, however, fall inside the Hashin and Shtrikman
assume that the potentials on opposite sides of the bounds which, in this case, are not too far apart
interface but infinitely near to each other are (Fig. 6).
equal and that the current through any element of In the extreme case of composites containing
the surface is the same when measured in either conducting particles in an insulating matrix, the
medium [35]. With two solid phases, deviations dominant factor determining the conductivity at
from the theory may be expected if voids are the higher particle volume fractions would appear
present, but voids in a single solid phase can be to be the formation of infinite chains of particles
regarded as a separate non-conducting phase. in contact with one another [ 3 7 - 4 0 ] . At the
Landauer examined the application of Equation lower volume fractions these chains are unlikely to
38 to experimental data on the electrical resistance be formed and the conductivity of the composite
of binary metallic mixtures. In seven cases does not differ greatly from that of the matrix.
(Bi-Bi2Pb, Bi-Sn, Cd-Pb, Cu-Fe, Cu2Sb-Sb, The relationships between the number of particle
Mg2Pb-Pb, P b - S b ) where agreement with a mix- contacts and the degree of continuity have been
ture theory might be expected, five gave good discussed by Gurland [41 ].
agreement. A comparison of the experimental
values for the Cu2Sb-Sb system and the 2.4. Thermal conductivity
theoretical curve is shown in Fig. 5. With the As we have noted earlier, the relations given for
Mg2Pb-Pb system, however, the results failed to the dielectric constant in Section 2.2 will also

2114
apply to steady-state heat conduction and can be urethane foam). In the latter case, if the pores are
used to derive the effective thermal conductivities open there will be two continuous phases; if they
of two-phase composite materials. Many of the are closed there will be one continuous solid
relations given in Section 2.2 have indeed been phase.
rederived or reapplied in the analysis of thermal For a foamed or porous material, the thermal
conduction. The formula given by Brailsford and conductivity X is often expressed as
Major [42]
X = Xs+Xg+Xr+Xe (40)
where ks, Xg, Xr and Xe are contributions due to
conduction through the solid, conduction through
the gas, radiation and convection within the pores.
/1-Xa/X2t]
1 + Va ~,~ T ~ ] ] (39)
This description is, however, misleading since it
implies that the four processes are taking place
independently and in parallel. At normal tempera-
for the thermal conductivity of a two-phase tures, however, radiation effects will be small and,
aggregate in which particles of phase 1 (thermal if the cell diametex is less than 3 to 4 mm, con-
conductivity Xl and volume fraction Va) are vection effects will be negligible [43]. For
surrounded by a continuous phase 2 (thermal isotropic materials with small pores, cavities or
conductivity X2) corresponds to Equation 4 for interstices it can then be assumed that the thermal
the dielectric constant and to Equation 37 for the conductivity is determined by the thermal conduc-
specific resistance. These equations should how- tivities of the phases and the phase geometry. Use
ever only be applicable to dilute suspensions. For can then be made of the Hashin and Shtrikman
systems in which no special circumstances lead to bounds given by Equations 17 and 18. These
the spatial continuity of either phase and each bounds for the air/silica system are shown in Fig. 7
phase can be treated equivalently, Brailsford and in which the thermal conductivities of various
Major [42] suggested the possible application of insulating materials have been plotted against their
an SCS type equation corresponding to Equation densities. The thermal conductivity of silica
38 derived by Landauer for the electrical conduc- (1.36 W m -a K-a) does not differ very greatly
tivity. A large number of empirical or semi- from that of gypsum (1.30 W m -a K-a) and the
empirical expressions for the thermal conductivity values for diatomaceous earth, rock wool, pow-
of heterogeneous systems have also been dered gypsum and cellular gypsum all fall within
examined. As with the dielectric constant it will, the Hashin and Shtrikman bounds. It can be seen
however, only be possible to obtain a satisfactory immediately from Fig. 7 that it will be impossible
description of the thermal conductivity behaviour to make an isotropic composite from air and silica
by taking the geometry of the composite into with a thermal conductivity of less than
consideration and making proper use of the geo- 0.05 W m -a K-a unless the density of the com-
metrical information that is available. posite is less than 625 kg m -3 . It may also be
Problems of heat transfer are of considerable noted that the values for cellular gypsum in which
technological importance in situations where heat the solid phase is continuous lie towards the upper
transfer has to be encouraged, as in heat ex- bound while those for powdered gypsum lie
changers, or reduced by the use of insulation. Most towards the lower bound.
insulating materials are, indeed, essentially mix- With this type of system in which there is a
tures of a solid material and air and owe their large difference between the thermal conduc-
insulating properties to the low thermal conduc- tivities of the two components, the Hashin and
tivity of air which, in the absence of convection Shtrikman bounds are very widely separated and
and under dry conditions, has a value of more geometrical information is clearly needed to
2.45 x 10-2 W m -I K -a at 273 ~ K. (The air may, predict the thermal conductivity of the composite
of course, in some cases be replaced by another gas material. In Fig. 8 the Beran bounds calculated by
or a vacuum.) The insulating material can have a Corson [26] are compared with the experimental
fibrous or granular structure (e.g. glass wool or values obtained by Sugawara and Yoshizawa [45]
diatomaceous earth) in which case the air is the for the effective thermal conductivity of firebrick.
continuous phase or it can be cellular (e.g. a poly- The Beran bounds are closer than the Hashin and
2115
 Figure 7 Thermal conductivity of insu-
0"18 lating materials. Theoretical curves: (a)
Hashin and Shtrikman upper bound for
air-silica, (b) Hashin and Shtrikman lower
0'16 bound, (c) Davies relation for fibres ran-
domly oriented in a plane. Experimental
01 results: (1) cellular gypsum, (2) rock
0'14 wool, (3) glass wool, (4) diatomaceous
earth, (5) powdered gypsum (data from
(a) [169]).
0"12
x
01
(Win-' ,-,)
0"10

I(D 05
0"08
05
01
0"06

0"04 731~2402 02
(c)

0.02

0 I00 200 300 400 500 600 700

Density @9 m-3)

1'2 K i , , i TABLE IV Thermal conductivity of fibrous insulating

materials
Conduction through air 0.027 Wm -1 K -1
Convection 0.003Wm-1 K-1
i Radiation
Conduction through solid
0.002W m-1
0.002Wm-1
K -1
K-1
Total conductivity 0.034Wm-1 K-1

the thermal conductivity is largely determined b y

" (i)
the conductivity o f the enclosed air, but radiation
o 0'4 and convection can make significant contributions
especially when compared with that due to con-
duction through the solid. Table IV [44] gives
.~ O
details o f the contributions due to the different
,,'n mechanisms if it is assumed that these operate
I-O independently and in parallel.
Volume fraction of air. I f we neglect convection and radiation effects,
the thermal conductivity o f a composite contain-
Figure 8 Effective thermal conductivity of fire-brick
ing fibres randomly oriented in a plane can be
(after Corson [26] ). (1) Hashin and Shtrikman bounds,
(2) Beran bounds. (Experimental data from Sugawara and estimated using Equation 33. A theoretical curve
Yoshizawa [45 ] ). calculated from this equation assuming that, for
the fibres, X = I . 0 W m -1 K -1 and p = 2 6 0 0 k g
Shtrikman bounds but are still, however, widely m -a , has been included in Fig. 7. It will be seen
separated. that the experimental results for the rock wool
Fibrous insulating materials will usually be and glass wool insulation follow the same general
anisotropic but more or less isotropic in a plane. trend but the experimental values are appreciably
With glass-fibre insulation under normal conditions higher. This is presumably the result o f the effects
2116
o f radiation, convection, fibre-fibre contacts, or have a thermal expansion coefficient very much
out-of-plane fibre alignment. closer to that of tooth substance these effects are
In the absence of effects due to radiation and very much reduced [48].
convection, the thermal conductivity of a com- The effective thermal expansion coefficients for
posite is, under normal conditions, independent of a composite material are defined as the average
the particle size [46] or inter-fibre spacing strains which result from unit temperature rise in a
provided these are (a) sufficiently small for the traction-free material. In considering the response
material to be considered macroscopically hom- of composite materials to changes in temperature,
ogeneous and (b) sufficiently large compared with the basic stress-strain relations which are used in
the mean free path of the carrier (e.g. electrons or defining the elastic moduli have to be modified to
phonons) responsible for the conduction process. include the thermal expansion coefficient. Levin
The use of the relations given in Section 2.2 does, [49], in an important paper, showed that a simple
however, involve the assumption that there is no relationship between the effective expansion coef-
discontinuity or barrier at the interface which, in ficients and the effective elastic moduli could be
the presence of a heat flux, would lead to the tem- derived for two-phase materials using the thermo-
perature on one side of the interface being differ- elastic stress-strain relations. For an isotropic
ent from that on the other. With particle-filled composite with two isotropic phases the basic
materials, e.g. filled elastomers, voids may be relationship can be written in the form
expected to lead to anomalous behaviour [47].
The conditions under which phase dimensions
can become important in heat transfer are con- [(1/K1)--(1/K2)] K-~--
sidered in Section 3.4.
(41)
2.5. T h e r m a l e x p a n s i o n c o e f f i c i e n t s where ~1, c~2 are the linear thermal expansion
Thermal expansion behaviour can be important coefficients of the two phases, K1, K2 are their
when composite materials are used in conjunction bulk moduli, K* and o~* the effective bulk
with other materials and when it is necessary to modulus and effective thermal expansion coef-
match the thermal expansion coefficient of one ficient of the composite, and the bars over the
structural component with another. An important symbols indicate volume averages [50]. The re-
function of reinforcing fillers and fibres in plastics lationship represented by Equation41 has also
is the reduction and control of the thermal expan- been established by Schapery [51], Cribb [52]
sion. With dental filling materials a difference in and Steel [53].
thermal expansion between the filling material and The form of Equation 41 gives an indication
the tooth substance can lead to a marginal gap of how the effective expansion coefficient will
(Fig. 9). With the composite filling materials which deviate from the simple mixture rule, o~* = if, as a
result of the mechanical interactions represented
by the final term. If one of the phases is a fluid (or
a low shear modulus matrix) and the other consists
of rigid particles

K--7 ~-- (42)

and it follows from Equation 41 that ~* --~~, and

the mixture rule will apply [51 ].
It also follows from Equation 41 that if we
know the values of K1 and K2 and can calculate
K*, then we can also calculate ~* from ~1 and e2
for any given volume fraction Vt. If we cannot
calculate K* but can set bounds on its value then
Figure 9 Scanning electron micrograph (X 100) showing
we can also set bounds on the value ofc~*.
marginal gap produced by difference in thermal expansion
coefficients of dental filling material and tooth substance The simplest case for a statistically isotropic
(courtesy Dr W. Finger). material is that of a dilute suspension of spherical
2117
particles when we can make use of the relation equation Equation 41 we obtain the relation
derived by Hashin [54].
3K2 + 4~2 (as -- a l)K1 (3Ks + 4/~2)V1
@* ~- 0t2 - -
K* = K2 + (KI --K2) 3K1 +4pa V1 K2 (3K 1 + 4/~2) + 4(K 1 --K=)/a2 Vl

(43) (47)
for the effective linear thermal expansion coef-
for a dilute suspension in which spherical particles ficient of the composite sphere assemblage. This
with bulk modulus Ka are dispersed in a matrix relation can also be obtained from the expressions
with bulk modulus K2 and shear modulus ~2. This given by Kerner [55]. It may be noted that when
relation may be expected to apply when the V 1 = 0, o* = a2, when V1 = 1, a* = ~1 and when
volume fraction of the particles VI does not tas = 0 , ~ * =alVa +%I12.
exceed a few per cent. From Equation 41 and 43 Using the relationships
we obtain
E E
K - ~
) -3- (21v and # - 2(1+v) (48)
a2 -t (at --~s)KI(3K2 +4us)Va
K2 (3Ka + 4/a2) + (K 1 --K2) (3K2 + 4Us ) V1 where E is Young's modulus and v is Poisson's
ratio, Equation 47 can be written in the form
(44) ~* = ~2 - (as - ~1) Vl x
or, neglecting the final term in the denominator of 3(&/Es) (1 -vs)
the last term
(E1/Es) [2V,(1 --2v2)+(1 +vs)] +2(1 --2vl)Vs
0~* = (~2 + (O~1 - - ~ 2 ) K I ( 3 K s +4//2)Vl (49)
K2(3K1 + 4/.t2) which corresponds with the relations obtained by
Wang and Kwei [56] and Fahmy and Ragai [57]
(45)
from a consideration of the displacements and
For higher particle concentrations, the com- stresses in a composite sphere.
posite sphere assemblage described in Section When the composite consists of a mixture of
2.2.1 may be a more realistic model since it incor- two kinds of particle, rather than particles of one
porates an element of randomness and its use phase dispersed in a continuous matrix, the SCS
involves the implicit assumption that each com- approximation described in Section 2.2.1 may be
posite sphere is, in effect, surrounded by com- more appropriate since, in this case, each phase is
posite material instead of matrix. For the effective considered on an equal footing. The effective
bulk modulus of this system, Hashin [24] ob- thermal expansion coefficient a* will be given by
tained the exact relation Equation 41 with K* determined by the relations

K* = Ks +(K1 --K2) 3K* +4/1' V1

3K1 + 4g*
(3K2 + 4gt2)V1
K2 + (K, (5o)
--K2)3K
1- + 4~ts + 3(Ks --K1)V,"
/~* = g2 +(Pq - / 1 2 ) x
(46) 5(3K* + 4,u*)V~
(51)
9K* + 8/2*+ 6(K* + 2/a*) (/1, i/a*)
When Va is small, the last term in the denomi-
nator of the final term can be neglected and the These two expressions which were derived
relation then becomes identical with Equation 43. independently by Budiansky [58, 59] and
An expression corresponding to Equation 46 for Hill [60] will, in general, have to be solved nu-
the effective bulk modulus was obtained earlier by merically to obtain K*, but decouple when vl =
Kemer [55] but the method of derivation is diffi- v2 = 0.2 [59]. Similar expressions, which reduce
cult to understand [5]. to Equations 50 and 51, were also given by
If we substitute Equation 46 in Levin's Kerner [55]. The use of the self-consistent

2118
 method for determining thermoelastic moduli has (Equation 41) lead to the following bounds for
also been discussed by Laws [61]. 0~*;

If the geometry of the composite is not clearly

( o q --a2)K2 (3K1 + 491 )V2
defined it can be argued that it is better to use 0~_) = O~1 - -

rigorous bounds rather than an uncertain approxi-

KI(3Ks + 491) + 4(Ks --K1)gl V2
m a t i o n . Bounds on the effective thermal expan- (56)
sion coefficient of an isotropic composite Can be
obtained by using Paul's bounds [62] on the bulk
(a2 -- a a )K 1(3K2 + 4/~s) V1
modulus: 0~+) = 0~s - -
K2 (3K1 + 4#s) + 4(K1 --Ks );us VI'
(57)
1 1
/~ < ~ - / < 9 (52)
where

As we have seen, the lower bound on K* rep- ~2 -- 0~1

- - >0 and g2>gl.
resented by Equation 42 and corresponding to Ks --K1
uniform stress in the composite leads to the
When ( a s - - a a ) / ( K 2 - K 1 ) < 0 or /a2 < # 1 , the
"mixture rule", a * = a, when substituted into
bounds are reversed. When a s = a 1 or/~s =/-ta,
Equation 41. The upper bound on K* correspond-
the bounds coincide.
ing to uniform strain and given by It should be noted that Equation 57 for the
K* = K1 V1 +K2 V2 (53) upper bound a~+) is identical with Equation 47
for the effective thermal expansion coefficient of
leads with Equation 41 to the relation the composite sphere assemblage in which spheres
of material 1 are surrounded by shells of material
O~* = KlOtlVI + K2az Vs = Ks~_ 2. Similarly, Equation 56 will correspond to the
K1V1 + K2 V2 K ' relation for the effective thermal expansion coef-
(54) ficient of a composite sphere assemblage in which
spheres of material 2 are surrounded by shells of
which is the expression suggested by Turner [63] material 1. This means that, as with the bounds for
for the thermal expansion coefficient of a two- bulk modulus [5], the bounds given by Equations
phase composite. 56 and 57 are the best possible bounds for the
Closer bounds on the value of a* can, however, effective thermal expansion coefficient of an
be set by using the bounds derived by Hashin and isotropic composite of arbitrary geometry.
Shtrikman [64], Hill [65] and Walpole [66] for If, however, further statistical information on
the bulk modulus, which may be written in the the composite geometry is available closer bounds
form can be set than those given by the Hashin and
Shtrikman results. Thus, bounds for the bulk
K~_) = K1 +
modulus have been derived by Beran and
v~ Molyneux [67] and McCoy [68] using the infor-
[1/(K2 - - K , ) ] + [3V1/(3K1 + 4U,)] mation contained in the three-point correlation
function. These bounds are, unfortunately,
(55a) complex expressions involving derivatives and
K~+) = Ks + integrals of the correlation functions and, in their
original form, cannot be easily used except in
V1 special cases. Miller [69], however, showed that
[1/(KI --K2)] + [3V2/(3Ks + 4/~2)] bounds o n the bulk modulus of the two-phase cell
(55b) materials described in Section 2.2.1 could be ob-
tained in the form of relatively simple algebraic
where ~zz >/~1. When /11 >/12, the relations for functions from the Beran and Molyneux ex-
the upper and lower bounds are reversed and when pressions. These bounds depend on the values of
/am =/~z, the bounds coincide. K1, K2, /~1, gs, V1 and V2 and on the two par-
These bounds and the Levin relationship ameters Ga and Gs which describe the cell

2119
T A B L E V Axial and transverse thermal expansion coefficients for transversely isotmpic aligned fibre-gomposites
Levin [49 ]

a* = a2 + (vl( a-, - vo2)E*

q ) [(1 + v, )(E* -- V2E 2 ) = (1 + v2 ) V~E i I (60)

(o,.
a t = a2 + ~ uz,~*[(1
v. + v,)(V2u*aE 2 ---v~E*)+ V i v ~ E , ( I ' + v2) ] (61)
(1-2) a

Schapery [51 ]

(62)
V-u-}
Rosen [70]

(63)

r --- (1)] (64)

a t = 6 + [(1]K, ) -- (1]K,)] [2k t Ea* --

Van Fo Fy [71]

a~ a 2 + (a 2 - - a a ) V a - - (or2 - - a l ) ( 1 +V "(v2 --va*) (65)

where the symbols have the following significance

a*, a~ effective axial and transverse linear thermal expansion coefficients.
a I , a 2 linear thermal expansion coefficients of fibre a~d matrix.
vl, V2 iPoisson's ratio for fibre and matrix.
E*, u* Young's modulus and Poisson's ratio of composite under axial loading.
E~, E 2 Young's moduli for fibre and matrix.
K 1 , K 2 bulk moduli for fibre and matrix.
k~ transverse plane-strain bulk modulus of composite and bars over symbols indicate volume averages, e.g.

-- /'-~ V1 V ~
Ea = E t a 1V; + E : a 2V2 and kK) = K-, + K~

geometries o f the two materials. Miller showed where a = K1 ~Ks, (3 = gt/1~2, 0 = V1, 3' = la2/K2.
that the values o f Ga and G2 must lie between The bounding equations for the symmetrical cell
and 89 and that, as for the dielectric case (see material can be obtained by setting G1 = G2 in
Section 2.2.1), a value o f } corresponded to a Equations 58 and 59.
spherical shape and a value o f 89 to a plate-like Up to now, it has been assumed that b o t h the
shape. composite and the two constituent phases are
The bounds for the asymmetrical cell material isotropic. The more general case of multiphase
take the following form (after correcting some anisotropic composites with anisotropic phases has
typographical errors in the equations given in been considered by Rosen and Hashin [50] who
Miller's paper), derived bounds b o t h for the effective thermal

(K, K2),] ~ - c~] ~ 1 + ~(c~ T 1) - - ( a - 1)~ + 27(1 - $,~ + -~fi(4~ - 1) + 3(~ - 1)[a~ (1 - & - c~ ~ ]

(58)

: K~_)
(K1K2) 1/2
= a I/2 / ~ b ( I - - a )
t
( )1_1
+ ol -- 1 + ~b(a -- 1) + ~ (o~/l~y) [3(/3-- 1){3 [G~ 4~2 -- G1(1 -- ~b)2 ] - ~ qS}+ 3/3 -- 11

(59)
2120
expansion coefficients and specific heats. In the approach has been adopted by Schneider [75,76]
case of two-phase composites the bounds coincide who derived the following expressions for the
and give unique relations between the effective thermal expansion coefficients using a model in
expansion coefficients and the effective elastic which the fibres were aligned in a hexagonal array:
moduli. Rosen and Hashin also showed that these
relationships for the two phase composite could be . a2 -- a 1
% = al + (68a)
derived directly using the methods developed by (V~/(1 -- Va))E~/Ez + 1
Levin which they extended to include anisotropic
phases.
In the important case of a transversely isotropic = as - ( a s -a~) x
aligned fibre composite with two isotropic phases,
. 2(v~ --v= - - 1) 1.1 V 1
a number of expressions have been derived for the
axial and transverse thermal expansion coef- 1.1 Vl(2v~ + v2 -- 1 ) - - ( 1 + u2)
ficients. These are given in Table V. By making use
of the relationships between the elastic moduli v2 E1/E2 ]
given by Equation 48, Hill's relations between the El~E2 + (1 -- 1:1 V1)/1.1 VlJ "
fibre composite modufi [72], and with some
(68b)
effort on the algebra, it can be shown that (a) the
Levin, Schapery and Rosen expressions for the It may be noted that when vl = v2, Equations
axial expansion coefficients are equivalent, (b) the 66 and 67 reduce to
Levin and Rosen equations for the transverse
E* = E~ V1 + E2 V2. (69)
expansion coefficient are also equivalent, and (c)
these expressions for the axial and transverse On substituting this simpler relation for the axial
expansion coefficients then lead to the relation- Young's modulus in Equation 60 we obtain the
ship given in Equation 65 between a* and simple approximate expression
previously given by Van Fo Fy [71].
, alK1 V1 + a2K2 V2
From these relationships, bounds on the % = (70)
thermal expansion coefficients can be obtained K1V~ + K2 V2
from bounds on the moduli. For example, with alE1 V1 + a2E2 V2
a a* ~ (71)
Equations 60 to 62 in Table V, we can make use E1V~ + E~ Vs
of the bounds derived by Hill [72] for the axial
Young's modulus, which in most cases will provide an adequate ex-
pression for the axial thermal expansion coef-
4Vl V2(vl --v2) s ficient.
E*(+) --. Ea V1 + E2 V2 + Some of the theoretical relationships described
V~/ks + V:/k~ + 1/U,
above are compared in Fig. 10 with the exper-
(66) imental results obtained by Schneider [75] for the
4V1 Vs(va - - v 2 ) 2 axial and transverse thermal expansion coefficients
E*(_) = E 1 W 1 --~ g 2 W 2 ~- of an aligned glass-fibre composite with an epoxy-
V1/k2 + Vs/kl + 1has
resin matrix. The theoretical curves have been
(67) calculated using the following values for the
where k is the plane-strain bulk modulus given by properties of the glass fibre and epoxy-resin
matrix. Glass: Ea = 71.59 GN m -s, ua = 0.25,
al = 4.8 X 1 0 - 6 K -1 , Ka = 47.73 GN m -z,
k = K+ 89 (68)
/aI = 28.64 GN m -s , kl = 57.28 GN m -2 . Epoxy-
Alternatively, we can make use of the SCS resin: E2 = 3.14 G N m -2, vl, = 0.4, as +=
method and either use the equations in Table V 6 6 x 10 - 6 K -a, /(2 = 5 . 2 3 G N m -2, #2 =1.121
with the relations given by Hill [73] for the elastic GN m -2 , k2 = 5.60 GN m -2 . The experimental
moduli, or use the explicit formulae given by Laws results for a* lie close to the lower bound obtained
[74] for the overall thermoelastic moduli. using Hill's relations for the bounds on the moduli
Rosen [70] suggested the composite cylinder (the theoretical curve corresponding to Equation
assemblage was an appropriate model and a similar 71 has not been plotted since it is extremely close
2!21
 x ~0-6K:tI
80~-

~ 7o

~ 6o

0
u 50
g
0
~-40 --
U

2)

2O

IO

0 I I I I r
0 Oq 0-2 0-3 0,4 09 ~6 09 09 09 t-O
Volume traction of fibre (Vl)
Figure 10 Thermal expansion of aligned glass-fibre composites9 (1) Lower bound for aa, (2) upper bound for aa*, (3)
lower bound for a t, (4) upper bound for at, calculated using Equations 60 and 61 with Hill's bounds on E~. (5) SCS
,approxamation for a t , (6) Schneider s equation for a t . Experimental results [75] : oaxaal thermal expansion transverse
thermal expension coefficient9

to this lower bound), The experimental values for nate in which the layers are of various orien-
fall between the calculated bounds and lie close tations. In this theory each layer is treated as a
to the curves obtained using (a) the SCS approxi- homogeneous anisotropic material in a state of
mation for the moduli and (b) Schneider's model. plane stress. The thermal expansion coefficient
From Fig. 10 it will be seen that, at low fibre and the thermal stresses in multi-ply laminates
volume fractions, the transverse expansion coef- have been considered in some detail by Schneider
ficient of the glass fibre-epoxy resin composite [76].
will be greater than that of the matrix. This effect A balanced laminate in which the plies are
will be especially noticeable with fibres of high oriented at angles of 0 ~ 60 ~ - 6 0 ~ and 0 ~ will
modulus and low axial expansion coefficient" (e.g. show pseudoisotropic in-plane elastic behaviour
boron or carbon) in a low-modulus matrix (e.g. and the thermal expansion within the plane will
epoxy resin) [70]. also be quasi-isotropic. Halpin and Pagano [81],
The coefficient of linear thermal expansion of a however, showed that a laminate need not necess-
unidirectional laminate at an angle to the fibre arily be quasi-isotropic with respect to in-plane
direction can be obtained from purely geometrical stiffness for it to show isotropic in-plane thermal
considerations [77]. At an angle 0, the expansion expansion behaviour. Isotropic thermal expansion
coefficient a~ of a transversely isotropic laminate within the plane will be shown by any uncoupled
will be given by laminate with equal stiffnesses in two in-plane
directions.
~ = a t sin20 + aa cos20. (72)
The effective in-plane thermal expansion coef-
Once the thermal expansion properties of a ficients aT and a~ for "angle-ply" laminates in
single layer of fibre-reinforced material have been which the fibres are arranged at angles of +0 and
derived, classical laminated plate theory [78-80] --0 to the 1-direction have been evaluated both by
can be used to calculate the behaviour of a lami- H a l p i n and Pagano [81-383] and by Dow and
2122
Rosen [84, 85]. With these angle-ply laminates a the inside of the surrounding shell to compensate
"scissoring" or "lazy tongs" type of action can for the misfit due to differences in thermal expan-
occur and, with appropriate values of 0, lead to a sion of the two materials. For a rise in temperature
zero or even a negative thermal expansion coef- of AT the interfacial pressure P in an initially
ficient in one direction. Dow and Rosen found stress-free system is given by
that a ratio of filament to matrix modulus of at
least 25 to 50 was required to achieve pronounced 2(% --%) A T E i E ~
p=--
"scissoring". Their theoretical analysis indicated 2E= V=(1-- 2v1)+2E 1Vl ( 1 - 2 v = ) + E t ( l + v 2)
that the boron/epoxy system would not quite give
(73)
a zero thermal expansion in any of the 2 - D
configurations examined. However, a zero thermal This relation corresponds with that given by
expansion could be achieved theroretically with a Fahmy and Ragai [57] after correcting for an
3 - D configuration in which half the filaments in apparent error in sign in their published equation.
the comparable 2 - D configuration were rotated When V 1 + 0 , the expression reduces to that ob-
about the 1-axis into a plane at 90 ~ to the original tained by Selsing [93] for a dilute suspension.
plane. The general effects predicted both by Dow If ~2 > 0q and AT is negative, then P will be
and Rosen and by Halpin and Pagano have been positive and there will be a normal compressive
confirmed experimentally with carbon fibres in stress at the interface. Thus if a particle-
epoxy resin (86-89). With this system, negative reinforced composite is fabricated at a tempera-
coefficients of thermal expansion in one direction ture above ambient, a positive interfacial pressure
can be achieved. The thermal expansion of lami- will develop on cooling the composite to room
nated fibre composites in the thickness direction temperature if the thermal expansion of the
has been investigated by Fahrny and Ragai-Ellozy matrix is greater than that of the inclusions. On
[9O]. the other hand, if o~2 > al and the temperature of
Approximate values for the thermal expansion an initially stress-free composite is raised, a normal
coefficients of oriented short fibre composites can tensile stress will develop at the interface. This
be obtained [91] by estimating the stiffness in the may be sufficient to lead to debonding. If separ-
fibre direction from that of the corresponding ation of the two components occurs at the
continuous fibre composite and assuming that interface then the expansion of the matrix will no
Poisson's ratio for the matrix is the same as that longer be restrained by the inclusions and the
for the fibre. Laminate theory can then be used to thermal expansion of the composite will then be
obtain the thermal expansion coefficient of equal to that of the matrix [ 5 6 , 9 4 , 9 5 ] . With
randomly oriented short fibre laminates. If the fibre-reinforced materials, the internal stresses
aspect ratio is sufficiently high the properties of resulting from changes in temperature may lead to
the randomly oriented discontinuous fibre com- multiple cracking of the matrix [96]. In consider-
posite can be predicted from the behaviour of a ing the development of internal stresses during the
pseudo-isotropic laminate reinforced with fabrication of composites, other dilatational
continuous fibres [92]. effects, e.g. the shrinkage of resins on cure, must
Up to this point it has been assumed that the also be taken into account. These effects are dis-
phases in the composite remain in contact at the cussed in the following section.
interfaces and that, in particular, with two solid
phases the composite remains fully bonded. The 2.6. Swelling and shrinkage
effects of internal stresses arising from differences Volume changes in the solid components of a
in the expansion coefficients of the components composite material may arise not only as a result
have not been considered. These internal stresses of changes in temperature but also from other
can be important not only in the fabrication of the causes. Two of the more important are:
composite but also in determining its subsequent (a) a phase change or chemical reaction e.g.
behaviour. polymerization or cross-linking; and
If we take the composite sphere assemblage as a (b) the absorption of gases, liquids or vapours.
model system, the interfacial pressure can be The effects of phase changes or chemical
obtained by finding the pressure which has to be reaction during fabrication are of importance with
applied to the outside of the central sphere and a wide range of composites: excessive shrinkage of
2123
one component can result in high intemal stresses deformation of a solid induced by swelling was
which may lead to premature failure or even equivalent to that caused by a temperature change
preclude successful fabrication. Matrix shrinkage and applied this concept to the analysis of the
also has important indirect effects on the mechan- swelling behaviour of a unidirectionally reinforced
ical behaviour of fibre composites since the composite sheet using Schapery's [51] formulae
consequent internal stresses can determine the for linear thermal expansion coefficients. They
frictional forces at the fibre-matrix interface. The showed that it was possible to predict the expan-
problem of deducing the shrinkage of a composite sional strains of a nylon-reinforced elastomer in
in terms of the properties of its components is also benzene within the limits of their experimental
of importance in the prediction of shrinkage in precision. They also showed that it should be
concrete [97,98]. possible to design laminates which would contract
In principle, as indicated by Cribb [52], in one direction on swelling.
Equation 41 for the thermal expansion of a two It would appear, however, that the Halpin and
phase composite can be expected to apply equally Pagano analysis involves the assumption that the
well to the prediction of strain arising from the weight of solvent absorbed by a given volume of
shrinkage of one or both of the components as a polymer is not affected by fibre reinforcement and
result of phase changes or chemical reactions such that the expansional strains in the composite are
as polymerization or the setting of cement. It will determined solely by the elastic deformations of
be merely necessary to replace the linear thermal the components. There would seem to be some
expansion coefficients by the corresponding linear doubt as to how far this assumption is justified at
strains resulting from shrinkage. Indeed, relations any rate with composites based on polymers to
of the type given in Section 2.5 should in general, which the Flory-Rehner[100,101] theory of
be applicable not only for the prediction of the swelling is applicable. On this interpretation the
shrinkage of composites but also for the analysis condition for the equilibrium swelling of a cross-
of the resulting internal stresses. With processes linked polymer of the conventional type in which
such as resin polymerization or the setting of the ends of the chains are united tetrafunctionally
cement there is, however, a continuous change is given by
in the viscoelastic properties of the material and
ln(1 -- V:) + V2 +/2V~2 +
there will be some difficulty in deciding on
appropriate values both for the linear shrinkage (pVo/Me) (V2 u3 __ V2/2) - Ul --/-to
and the elastic moduli of the resin or the cement. RT (74)
The importance of swelling behaviour was where V2 is the volume fraction of unswollen
clearly recognized in early work [99] on cellulose polymer, # is a dimensionless parameter charac-
fibre composites since with these systems both the terizing the interaction energy of solvent and
fibre and resin matrix behaved as swelling gels polymer, p is the density, Vo is the molar volume,
which could absorb or desorb water depending on Me is the number average molecular weight of the
the environmental conditions. The consequent polymer chains between the cross-links, #1 the
swelling and shrinkage of the composites could chemical potential of the solvent in the polymer
have serious adverse effects on their performance and/2 o the chemical potential of the pure solvent.
in structural situations. With composites based on For equilibrium with the pure solvent, //1 =/ao
glass, carbon or boron fibres, swelling and and the swelling of the unconstrained polymer will
shrinkage arising through sorption or desorption of be given by
water does not have such a marked effect since
although, with glass fibres in particular, water may ln(1 - - V 2 ) + V2 + uV~ +
be absorbed at the interface, these fibres are not
( p V o / M c ) (V2 '/3 -- V2/2) = O. (75)
swelling gels and do not absorb water in the same
way as cellulose fibres. Swelling effects can, how- If the polymer is under a hydrostatic pressure
ever, still be very important in the absorption of P, the swelling ~11 be given by
water by nylon composites and in the absorption
of organic solvents by a wide range of reinforced l n ( 1 -- V2) + V2 + / I V ~ +
polymers and elastomers. eVo
Halpin and Pagano [81] considered that the ( p V o / M e ) ( v 2 ~/3 - v 2 / 2 ) - (76)
RT
2124
when the correction t e r m , - V2/2 given by Flory The dimensions and periodicity of a composite
[100] is included in Treloar's [101] equation. If structure may, however, have an important effect
swelling is constrained in one direction but can on the properties when they become comparable
take place in the other two directions, we have with, for example, the wavelength of incident
[101,102] radiation, the size of magnetic domains or the
ln(1 -- V2) + V2 + ~V~ + (pVo/Me) thickness of space charge layers at interfaces.
Interface effects will, in general, become more
(1 -- V2/2) = 0. (77) important as the dimensions are decreased.
The application of this relationship to the The effects will be especially noticeable in
swelling in toluene of unidirectionally reinforced composites in which the particle radius is less than
rubber composites has been examined by Coran some critical value and in composites with a
et al. [102]. Although the constraint due to the periodic structure in which the spacing between
reinforcement reduced the amount of solvent aligned fibres or lamellae is fixed and defined.
absorbed, the swelling in a direction perpendicular Composites with structures of the latter type can
to the reinforcement was actually higher than that be prepared by the unidirectional solidification of
in the unreinforced material. Good agreement was eutectics or eutectoids. With eutectics, the inter-
obtained between the experimental results and lamellar period X is given by X2R = a constant
those predicted by Equation 77. The effect of where R is the imposed growth rate; for eutec-
mechanical constraints on the sorption of water by toids, X4 R = a constant [107]. Another possible
cell wall material in wood was examined by Barkas method of preparation is by the unidirectional
[103] who considered that an elastic sheath sur- decomposition of non-crystalline solid single-phase
rounding the cell wall material exerted a natural materials [108].
restraint on the swelling of the wood substance. Properties dependent on phase dimensions and
The swelling of wood from the standpoint of periodicity include some of the optical, magnetic
composite theory, has also been considered by and superconducting properties. Under certain
Cave [104]. conditions, phase dimensions can also become
As with thermal expansion, differential swelling important in heat transfer through composite ma-
of matrix and reinforcement in a composite can terials.
lead to debonding at the interface. Thus, for
example, with polyester fibre composites, resin 3.1. Optical and related e f f e c t s
swelling in water at 100 ~ C was sufficient to over- Many of the optical effects which depend on struc-
come the effects of resin shrinkage during cure, tural dimensions also depend in a complex way on
producing a net tensile stress across the fibre-resin other factors, the theory can be extremely compli-
interface with the result that both untreated cated, and a detailed discussion is beyond the
carbon fibres and untreated glass fibres were scope of this review. The more important effects
rapidly debonded [105,106]. At 20 ~ C, untreated arise in light scattering and diffraction.
glass fibres were debonded but the interfacial bond For very small particles with a size very much
was retained with both the carbon fibres and less than the wavelength X of the incident light,
treated glass fibres. the intensity of scattered light is given by the
Rayleigh relationship which for unpolarized inci-
3. Properties determined by phase dent light and a single particle takes the form
dimensions and structural periodicity
In the preceding sections, the properties under Io 87r4 a 2
consideration have been dependent on the volume I - r2 X4 (1 +cos20), (78)
fractions and geometries of the phases but not on
the phase dimensions (provided these are small where I is the intensity of the scattered light at a
compared with t h e size of the specimen). The distance r from the particle, Io the intensity of the
expressions given for the effective properties do incident light, a the polarizability of the particle,
not include any dimensional parameters so that, X the wavelength, and 0 the angle between the
for example, the effective dielectric constant of a scattered and incident light. The quantity
dilute suspension of spherical particles does not Igor2/Io V where V is the volume of the particle is
depend on the particle radius. known as the Rayleigh ratio R(9o).
2125
Figure 11 Extruded polystrene-polybutadiene-polystyrene block co-polymer (a) low-angle X-ray diffraction
[ i 18 ] : (i) beam parallel to extrusion direction; (ii) beam perpendicular; (b) electron micrographs [ 119 ] of sections cut
(i) perpendicular and (ii) parallel to extrusion direction. 1.0. (Courtesy Professor A. Keller and Dr J. Dlugosz.)

For an assembly of particles we can assume, in inversely proportional to V. The light scattered by
the simplest approximation, that the particles are a given volume concentration of the particulate
very far apart and the intensity of the light scat- phage will, therefore, increase as the particle size
tered by N particles will then be given by increases. This general principle forms the basis for
the determination of the molecular weights of high
N Io 87r4a 2
IN - r2 X4 (1 + cos20). (79) polymers (and simpler molecules) in solution by
fight-scattering measurements [109] using the
The polarizability a depends on the volume V theory developed by Debye [110, 111].
of one particle so that a 2 will be proportional to The effect of particle size on the light scattering
V~. With a given volume of particles, N will be behaviour in a solid medium is very well illustrated
2126
Figure 11 Continued.

by the work of Maurer [112] on crystal scattering at angles symmetrical about the 90 ~
nucleation in titania-containing glasses. Light- position) and the degree of depolarization (the
scattering measurements were used firstly to ratio of light intensities of vertical and horizontal
detect the presence of scattering centres and study polarization scattered at 90~ Measurements on
the initial stages of crystal nucleation, and the glass before heat-treatment showed that the
secondly to estimate the average crystal size and scattering was about three times as high as that
concentration so that the crystallization process which would be expected from the refractive
could be followed and the crystal-growth rate index and the depolarization was about one-third
measured. The three important parameters which of the expected value. These results suggested that
were measured were Rayleigh's ratio, the dissym- a glass-in-glass type phase separation had occurred
metry coefficient (the ratio of the intensities of during the preparation of the titania-containing
2127
 glass. On heat-treatment at the lowest temperature diffraction patterns possessing hexagonal
(725 ~ C) there was initially a six-fold increase i'n symmetry were obtained [118] and it was shown
depolarization without any change in dissymmetry both from the X-ray and electron microscopic
and only a 30% increase in total scattering. This evidence [119] that the material consisted of
indicated that the scattering centres were polystyrene cylinders embedded in a poly-
becoming anisotropic and it was concluded that butadiene matrix with the cylinder axes along the
crystallization was taking place within the glass extrusion direction. The polystyrene rods were
particles. The growth of the crystalline centres as a 150 A in diameter and arranged in a hexagonal
result of heat-treatment was followed by lattice with a lattice parameter of 300 A. The low-
measurement of the average particle size. This was angle X-ray diffraction patterns [118] are shown
deduced from measurements o f Rayleigh's ratio in Fig. 1 la and the electron micrographs [119] in
and the volume concentration of the particulate Fig. l lb. This X-ray and electron microscope
phase using a theory developed by Maurer which evidence has been used for the interpretation of
indicated that, if the number of crystalline the mechanical properties of "single crystal"
particles remained constant, the crystal diameter samples of SBS co-polymer [120].
should vary as the sixth root of the Rayleigh Sievers [117] has discussed the diffraction
ratio. The volumetric concentration of the crystals behaviour of dielectric eutectics and described
was obtained from the integrated intensity of an results obtained in an investigation of the infra-red
X-ray diffraction line and this was then used to transmission of the NaC1-NaF eutectic. The
obtain the crystal diameter from the Rayleigh longest wavelength incident on the crystal
ratio. Crystal diameters of from 70 to 175 A satisfying the diffraction condition for rods
obtained by this method were in good agreement arranged in a hexagonal array would be expected
with those obtained from X-ray line broadening to be given by
measurements.
Xmax = x/Sn*d, (80)
Light-scattering behaviour has also been shown
to depend on the domain size in phase-separated where Xmax is the wavelength of the incident light,
borosilicate glasses [113] and on spherulite size in n* the effective refractive index and d the
films of polyethylene [114, 115]. spacing between nearest neighbours. The agreement
Periodicity in a composite structure can result between the calculated value for Xmax and the
in diffraction effects if the wavelength of the inci- experimental value for strong absorption was
dent light is of the same order as the composite within 5% tor d = 4.5/am and within 2% for d =
period X. The behaviour of the aligned lamellar 5.8/am.
Co2Si-Co eutectoid [10] is given as an example Provided the periodicity is very much less than
by Albers [1]. By preferentially etching the the wavelength of the light, a composite material
surface of a specimen in which X was equal to containing aligned elongated particles of an op-
0.5/am, a periodic relief structure of the ridge type tically isotropic material in an optically isotropic
was obtained which showed diffraction spectra matrix (with a different refractive index) will
when illuminated with white light. The diffraction exhibit birefringence (or double refraction). This is
patterns obtained by optical reflection from the a straightforward consequence of the anisotropic
polished and etched surfaces of directionally sol- dielectric properties and the effective refractive
idified eutectic alloys have been discussed by indices can be calculated from e.g. Equations 25
Clarke [116] who has shown how they may be and 26 using the relationship, n * = (e*) 1/2, be-
used tor the detection of surface symmetries and tween the effective refractive index n* and the
for the measurement of interparticle spacing. effective dielectric constant e*. For a material
The spacing between the NaF rods in an containing aligned cylinders it is then easily shown
NaC1-NaF eutectic has been obtained by [117] that
reflecting laser light from the polished surface of a #
ne = ( V l e l + V2a2) 1/2 (81)
transverse section and measuring the distance to
the first diffraction ring [117]. Similarly, in low- and
angle X-ray diffraction studies of an extruded SBS { Io2o-v1)+ ,(1 +
(polystyrene-polybutadiene-polystyrene) block = [c,(--iu v,) c,(1 - T)Jj
+
co-polymer~ it was found that sharp single-crystal
(82)
2128
and the c'6fnposite betiaves as a uniaxial material ever, fine-particle permanent magnet materials
with the optic axis along the cylinder axis. based on cobalt-rare earth alloys have been devel-
Composite systems containing aligned metallic oped which have properties superior to any
rods in a semiconductor or an insulator can also previously known [123]. The general properties
act as polarizing falters when the diameter of the of the Alnicos and other cobalt alloy permanent
rods is small compared with the wavelength. With magnets have been reviewed by Gould [124].
the unidirectionally solidified InS-NiSb eutectic In recent years considerable attention has been
the rod diameter is typically about 0.5 #m and given to the investigation of rod-type eutectics,
with a layer 70 #m in thickness the polarizability obtained by unidirectional solidification, as
at wavelengths above 12/~m is at least 0.99. permanent magnet materials since, in the absence
Electromagnetic waves with an electric vector of magnetocrystalline anisotropy, the magnetic
vibrating parallel to the axis of the rods lose particles need only be small in two dimensions
energy and transverse the composite with reduced and, as a consequence of the shape anisotropy of
amplitude. A wave with an electric vector perpen- the demagnetization energy, can be large in the
dicular to the rods will, however, be damped very third dimension. This is the concept of the
little [7]. "elongated single domain".
The behaviour in the microwave region of arti- Progress in 'this field has been reviewed by
ficial dielectrics consisting of a lattice of con- Galasso [3, 40], Albers [1] and Sahrn and Hofer
ducting elements or an array of elements of one [125]. A number of eutectic components contain-
permittivity embedded in a dielectric of a differenting magnetic rods in a non-magnetic matrix have
permittivity has been discussed by Brown [121]. been investigated including InSb - MnSb [126],
FeS-Fe [127], Bi-Co [125] and Sb-MnSb
3.2. Magnetic p r o p e r t i e s [128]. Livingston [129] in studies with the
In a ferromagnetic material, at temperatures well Au-Co eutectic found that as the diameter of the
below the Curie point, the electronic magnetic Co-rods fell from 1/~m to 0.07/1m, the coercive
moments will be aligned within small contiguous force along the rod axes increased from 30 to
regions - magnetic domains - within each of 330 Oe. In early work on the directionally solidi-
which the local magnetization is saturated. In the fied Bi-MnBi system, Noothoven van Goor and
demagnetized state, the directions of magnet- Zijlstra [130] obtained a value of 11 kOe for the
ization in the different domains will vary so that coercive force at room temperature. In more
the overall magnetization is zero. On applying a recent work [131] values as high as 24 kOe have
magnetic field, the magnetic moment can increase, been obtained with MnBi rod diameters of around
firstly by an increase in the volume of the domains 0.5 to 1.0/lm. With material containing rods of
which are favourably oriented with respect to the smaller diameter the coercive force decreased. This
field and secondly, in stronger fields, by rotation was thought to be due to an increasing lack of
of the magnetization. The high coercivity required alignment in material obtained at very high growth
in permanent magnets is achieved by suppressing rates. Although these materials exhibit a high
the possibility o f domain boundary displacement. coercive force the volume fraction of magnetic
This can be accomplished by using very fine rods and the energy product is low. As noted by
particles. A particle with a diameter less than Albers [1], the best magnetic in situ composites
about 0.1 to 0.(301/am will always be magnetized' are still probably the Alnico alloys obtained by
to saturation since the formation of a flux-closure spinodal disproportionation [132]. These contain
configuration will be energetically unfavourable a high volume fraction of ferromagnetic rods and
[122]. Magnetization reversal cannot take place in have a high energy product (11 x 106 GOe).
a sufficiently small single domain particle by Several eutectic systems with magnetic rods in a
boundary displacement. It can only take place by magnetic phase have been investigated including
rotation of the magnetization so that large fields CoSb-Co [125], Y2 Colv-Co [3], and Sin2 Co17-
(depending on the anisotropy energy and shape of Co [125]. With the latter system which contained
the particle) may be required. ferromagnetic Co-rods in a ferromagnetic
The cast or sintered alloys of iron, aluminium, Sm2Co17 matrix, Sahm and Hofer found that the
nickel and cobalt have been the most widely used coercive force increased from 50 to 500 Oe as the
materials for permanent magnets. Recently, how- rod diameter was decreased from 2/xm to 0.2/~m.
2129
 For soft magnetic materials, fine particle sizes reduced to a level at which it can be absorbed by
have to be avoided. Work on Co-Nb and the material itself without a serious rise in tem-
C o - N b - F e directionally solidified eutectics with perature which would lead to destruction of the
soft magnetic properties has been described by superconductivity. The criterion for adiabatic stab-
Coiling and Kossowsky [133,134]. ility is given by the relation [146,147].
x < (109rrSTo)l)2/4Jc (84)
3.3. Superconducting properties
The boundaries of the superconducting state are where x is the penetration depth of the magentic
defined by the critical temperature Te and, in a field (cm), S is the heat capacity (J cm -a K-1 ),
magnetic field, by the critical field He(T) at which
T0 = & / ( - d&/d73, (85)
the magnetic flux penetrates the specimen and the
superconductivity is destroyed. With type I super- and Je is the current density (A cm-2). This re-
conductors the flux penetrates abruptly at He; lation gives the approximate thickness of super-
with type II superconductors penetration occurs conductor below which flux jumping should not
gradually over a range He1 to He2. With thin films occur. Typical numerical values for the type II
in which the thickness is very much less than the superconductor NbTi are S "~ 2 • 10-a , To ~ 4
London penetration depth (typically about corresponding to a superconductor thickness or
500 A), there is no Meissner effect and the critical filament diameter of about 0.008 cm for a Je of
field H e can be very high. In a composite super- 3 x l0 s Acm -2.
conducting material the value o f H e can, therefore, An alternative approach is to slow down the
increase when the dimensions of the supercon- motion of flux by incorporating a low resistivity
ducting phase are sufficiently small. This effect is normal metal (e.g. Cu) thus allowing time for the
illustrated by the work of Watson [135,136] on heat to escape from the superconductor by
porous glass impregnated with indium. With pore thermal conduction. The corresponding dynamic
diameters of from 65 to 250 A and at a tempera- stability criterion is given, for a composite con-
ture T below T / Y e = 0.5, the critical field He was sisting of alternate layers of superconductor and
given by normal metal (e.g. Cu), by the relation [148-150]
He = (3415 +40) [1 --(T/Tc)2]/d t'~176
+- o. 14 d 2<~8T~ 1--Xs 1 (86)
(83) P Xs J~ '
where d is the pore diameter in A so that, at a where d is the thickness of the superconductor
given temperature, the critical field was approxi- (cm), To is given by Equation 85, k s is the thermal
mately inversely proportional to the pore diam- conductivity of the superconductor (W cm -1 K-1 ),
eter. Small increases in Tc have also been reported p is the resistivity of the normal metal (f2cm) and
with systems of this type [136], the greatest Xs is the volume fraction of superconducting ma-
relative effect being obtained with an A1-A1203 terial. This relation gives similar values for the
multilayer structure in which Te was 4.5 ~ K com- superconductor thickness (or filament diameter)
pared with a value of 1.2 ~ K for A1 [137]. When to those given by Equation 84. For example, with
both phases are metallic decreased phase dimen- T o ~ 4 K , J e ~ 3 x 10SAcra -2, k s ~ 1 0 - 3 W
sions are likely to lead only to decreased critical cm -1 K -1, p ~ 2 x 10-8 g2cm and X"~0.5, we
fields [138]. find that d~<0.004 cm. Wilson et al. consider,
At present, the main application of filamentary however, that adiabatic stability (which assumes
superconducting composites is in the construction no heat loss during a perturbation) provides the
of powerful electromagnets in which one of the more secure basis for achieving reliable super-
main problems is that of stability [139, 140]. conducting performance.
Instability in the shielding currents induced in the A.c. losses in filamentary superconductors also
superconductor can lead to the phenomenon depend on filament diameter and the calculation
known as flux-jumping. The conditions required to " of these losses has been discussed by London
achieve stability have been summarized and [151], Hancox [152] and Wipf[153]. Wilson
examined in a series of papers by Wilson and his et al. have summarized the basic formulae for
colleagues [141-145]. To achieve adiabatic losses in coils [141].
stability, the energy dissipated by a perturbation is Since the satisfactory performance of filamen-
2130
tary superconductors is dependent on a t'me dividual grains or crystallites. If these are of the
fdament diameter continued attention has been order of tens of microns this can lead to a de-
given to the superconducting properties of direc- pression in X(T) at low temperatures and an
tionally solidified eutectics. References to the extension of the region in which X(T) is pro-
earlier work are given in the reviews b y Galasso portional to C(T) to higher temperatures than
[3] and by Gallaso et al. [4]. Recent work has with single crystals.
included re-examinations of the superconducting There is, therefore, a size effect which may
properties of unidirectionally solidified P b - S n have to be taken into account in considering heat
eutectics [154, 155] together with a preliminary transfer in composite materials at low tempera-
investigation [155] of the Josephson effect at 4.2 tures. There may also be an interface effect since
K with different inteflamellar spacings. acoustic mismatch can limit the transmission of
phonons across the boundary and there will then
3.4. Heat transfer be a finite difference between the temperature on
In the earlier discussion of the thermal conduc- one side of the interface and that on the other
tivity of composite materials (Sections 2 4 ) it was (With metals and liquid helium this is known as
assumed that (a) interface effects were absent and the Kapitza thermal boundary resistance). A
(b) heat transfer by radiation did not occur. We theoretical treatment of this effect has been given
now consider conditions under which these two by Little [156] who showed that for a perfect
assumptions are no longer valid and the heat trans- junction with zero modulus of rigidity in both
fer properties become dependent on phase materials, the net flow of heat across the interface
dimensions. at low temperatures from material t to material 2
Heat can be carried through a solid by elec- is given by
trons, lattice waves (phonons), magnetic
excitations or electromagnetic radiation (photons). de
d t -- 5.01 x 1016pA
~ (T14 -- T~)erg see -1 ,
The thermal conductivity (X) of the solid is given
by (89)
?v = 89 2 Civili, (87)
where P is a complicated function of the densities
where i is the type of carrier, Ci is the contribution of the two materials and the ratio of the acoustic
of the carrier to the specific heat, v i is the carrier velocities, A is the area of the interface, CI is the
velocity, and li is the mean free path of the carrier. group velocity (cm sec -1) of the longitudinal
If the dimensions of the specimen are suf- phonons in the first material, and Ta and 7"2 are
ficiently small, the mean free path of a carrier will the temperatures (K) in the two materials. For
not be the same as that in the bulk material but small values of ( T I - T 2 ) / T 1 . Equation 89 can be
will be limited by the external boundaries of the written in the form
specimen. The thermal conductivity will then no
17 FA a
longer be an intrinsic property but will depend on d~ ~ 2 . 0 x 10 - - = T t A T e r g s e c -1
dt C~
the specimen dimensions.
With dielectric crystals, for example, where the (90)
phonon free paths can be long at low temperatures so that the boundary resistance is inversely pro-
the thermal conductivity is given by portional to T a. At low temperatures the effects
of this boundary resistance can become very pro-
X(i0 = 89 (88)
nounced. The addition of a fine metal powder to a
where L is of the order of the shortest linear di- dielectric would for example, be expected to in-
mension of the specimen, C is the specific heat, crease the thermal conductivity. However, at low
and v is the velocity of the lattice waves (or the temperatures, the temperature dependence of the
sound velocity). The temperature dependence of X boundary resistance can exceed that of the
will then parallel that of the specific heat due to thermal resistance of the dielectric and the thermal
lattice waves and X(T) will vary as T 3 at the conductivity of the composite will then be less
lowest temperatures. than that of the dielectric [157].
In a polycrystalline aggregate, the phonon mean These effects have been investigated by Garrett
free path can be determined by the size of the in- and Rosenberg [158] who measured the thermal
2131
N}
N~ TAB L E V I Matrix of properties (after van Suchtelen [ 8]. The table classifies the physical properties or phenomena in materials according to the kind of input
parameter (columns) and output parameter (rows).
Y X
(1) Mechanical (2) Magnetic: HIM (3) Electrical: E/P, i (4) Optical and (5) Thermal: T, (6) Chemical: (grad)
K/&l particle radiation: grad T, heat C, (grad) t~
light of particle current
flux
(1) Mechanical: elasticity (i) magnetostriction (i) electrostriction - thermal expansion osmotic pressure
K/zXl (ii) magnetoviscosity (i) Kirkendall effect
(suspensions) (iii) electroviscosity
(suspensions)
(iv) indirect: thermal
expansion
(2) Magnetic piezomagnetism x (i) superconductors photomagnetic effect {+ H} ferromagnetic dependence of Te on
[t/M (ii) galvanic material at T ~- Te C (FM)
deposition of FM
layer
(iii) direct generation
of magnetic field
(3) Electrical: (i) piezoelectricity (i){+ i~ magnetoresistance e, p (i) photoconductivity (i) thermoelectric dependence of Te on
E/P, i (ii) piezoresistivity (i.i){+ i} Hall effect {+ H} Hall effect (ii) photo-emission (ii) {+ E} ferroelectrics C (FE)
(iii) a.c. resonance (iii){+ H}PEM effect at T ~ T e
(iv) induction of (iv) ionization (i~){+ i}p(r)
voltage
(4) Optical (i) stress (i) Faraday effect (~ electro luminescence n thermoluminescence chemoluminescence
and particles: birefringence (ii) magneto-optic (ii) laser junctions fluorescence
light or (ii) tribolumines- Ke~ effect (iii) n scintillation
particle flux cence (iii) deflection of (iv) Kerr effect colour-centre
charged particles (v) absorption by activation
galvanic deposits
(vi) cold emission of
electrons
 T A B L E VI continued.
Y X
(1) Mechanical: (2) Magnetic: HiM (3) Electrical: E/P, i (4) Optical and (5) Thermal: T, (6) Chemical: (grad)
K/A l particle radiation: grad T, heat C, (grad)
light or particle current
flux

(5) Thermal: T, (i) heat of (i) adiabatic (i) dissipation in absorption ~, reaction heat
grad T, heat transition of demagnetization resistance
current pressure- (ii){+/}grad T NE (ii) Peltier effect
induced phase effect (iii){+ H}grad T
transition (iii) {+ E} magnetoresistance NE effect
(ii) piezorestivity effect + Joule heating
and Joule
heating
(6) Chemical: pressure-induced - (i) electromigration light or particle (i) Soret effect -
(grad) C, phase-transition (ii) galvanic deposition stimulated reactions (grad T)
(grad) ~ (photosensitive layers) (ii) phase transition
(iii) change of
chemical
equilibrium
Symbols
C chemical composition, concentration M magnetic polarization X thermal conductivity
E electric field n refractive index ~z chemical potential
H magnetic field P dielectric polarization p resistivity
i electric current T temperature FM ferromagnetic
K (mechanical) force e dielectric constant FE ferroelectric
2xl deformation x magnetic susceptibility NE Nernst-Ettingshausen
Bracket symbols, e.g. {+ i}, indicate that the pmarneter represented by the symbol in the brackets is essential as a second input (e.g. i and H in the Hall effect).

~o
 conductivities of epoxy resin composites at tem- With composite materials, the photons will be
peratures between 2 and 300K. The fillers in- scattered at the phase interfaces but, assuming the
cluded glass spheres and quartz, corundum and photon mean free path is independent of fre-
diamond powders. The quartz and diamond par- quency the thermal conductivity due to radiation
ticles were irregular in shape; the corundum will again be proportional to T 3 . The effects of a
particles were in the form of small platelets. Above frequency dependence of the mean free path on
about 10 to 20 K, the results with the epoxy/glass radiative heat transfer through composite materials
composites were in agreement with a modified have been considered by Klemens and
form of the Rayleigh equation derived by [162]. At low frequencies the mean free path
Meredith and Tobias [159], with the quartz and must be strongly frequency dependent and indeed
diamond powders the results were interpreted if the particles are small compared to the wave-
using the Hamilton and Crosser equation for length, the mean free path should vary as 1/p 4.
thermal conductivity [160] which includes a Long-wavelength photons can, therefore, play an
shape parameter. With all the samples the thermal important part in heat transfer through composite
conductivity above 20 K was found to be higher materials such as polycrystalline solids, densely
than that of the unfilled rasin and to increase with packed powders, foams and fibrous aggregates
increasing filler concentration. with the result that the radiative thermal conduc-
Below 20 K, marked changes in the composite tivity at moderately high temperatures may be
conductivity were observed and with some very much higher than would be expected from
samples the thermal conductivity was appreciably extrapolating down from temperatures above
lower than that of the unfilled resin. The thermal 2000 K according to a T 3 law.
conductivity in the liquid helium region was also
strongly dependent on the size of the particles and 4. Product properties
this dependence could not be explained merely in The concept of product properties was first de-
terms of a frequency-independent scattering of scribed by van Suchtelen [8] and has also been
phonons at the particle boundaries. Garrett and discussed by Albers [1]. These properties can be
Rosenberg considered that this discrepancy was distinguished from most of the material property
due to acoustic mismatch at the matrix-particle combinations considered in the previous sections
interface and showed that the low temperature since they depend on the way in which the behav-
results could be interpreted by calculating an ef- iour of one phase affects that of another.
fective thermal conductivity for the filler particles Most physical properties of materials can be de-
based on measurements of thermal contact fined in terms of an X - Y effect in which X is an
resistance. input parameter and Y a corresponding output
At high temperatures, the effects of radiative parameter. As we have noted, the behaviour of the
heat transfer may have to be considered. With a material is then described in terms of a pro-
material which is partially transparent each volume portionality tensor ~ Y / O X = A where A is a
element will absorb part of any incident radiation property such as electrical conductivity or mag-
and will also re-radiate energy. A certain amount netic susceptibility. Van Suchtelen distinguished
of energy can therefore be transmitted through two ways in which material properties can be
the material by these processes of absorption and combined in a two-phase composite material:
re-radiation in addition to that conducted by (a) The property arising from an X - Y effect in
lattice waves. When the thickness of the specimen material 1 can be combined with that from an
is much larger than the mean free path of the X - Y effect in material 2. Together these give an
photons, the apparent radiative thermal conduc- X - Y effect in the composite. These properties
tivity will be proportional to T a and inversely which were termed "sum properties" by Van
proportional to the Rosseland mean extinction Suchtelen have been discussed in the previous
coefficient. When the specimen is optically thin, sections.
the apparent radiative thermal conductivity will (b) The output from an X - Y effect in material
again be proportional to T 3 but will also be pro- 1 can form the input for a Y - Z effect in material
portional to the thickness so that the thermal 2 to give an X - Z effect in the composite material.
conductivity is no longer an intrinsic property of The corresponding properties were termed
the material [t61]. "product properties" by van Suchtelen.
2134
TABLE VII Product properties of composite materials (after van Suchtelen [8] )
X-Y-Z Property of phase I Property of phase II Product property
(Table II) X- Y X- Y X-Z
123 piezomagnetism magnetoresistance piezoresistance; phonon drag
124 piezomagnetism Faraday effect rotation of polarization by
mechanical deformation
134 piezoelectricity electroluminescence piezoluminescence
134 piezoelectricity Kerr effect rotation of polarization by
mechanical deformation
213 magnetostriction piezoelectricity magneto-electric effect
213 magnetostriction piezoresistance magnetoresistance; spin-wave
interaction
253 Nernst-Ettingshausen Seebeck effect quasi-Hall effect
effect
214 magnetostriction stress-induced birefringence
312 electrostriction piezomagnetism magnetically induced birefringence
313 electrostriction piezoresistivity ~ electromagnetic effect
343 electroluminescence photoconductivity J coupling between p and E (negative
314 electrostriction stress-induced bffefringence diff. resistance, quasi Gunn effect)
electrically induced biIefringence;
421 photomagnetic effect magnetostriction "~ light modulation
421 photoconductivity electrostriction J photostriction
434 photoconductivity electroluminescence wavelength changer (infra-red
visible etc)
443 scintillation photo conductivity radiation induced conductivity
(detectors)
444 scintillation fluorescence fluorescence radiation detectors 2-stage
fluorescence

The concept of product properties is clearly of BaTiO3-CoF%O4 eutectic and (b) the develop-
considerable potential importance in the develop- ment of an X-ray fluorescent material consisting of
ment of composite materials with novel properties. fine (--~l~m)PbC12 particles in an anthracene
The principle may be similar to that in an elec- matrix. In the latter case, the X-rays release sec-
tronic or other device but with the composite ondary electrons with high efficiency in file PbC12
material the resulting property is an intrinsic prop- and these electrons which have a range of a few
erty of the material. The X - Z effect may be /am spend most of their time in the anthracene
achieved by the use of different Y parameters and which is a good scintillator. The composite had an
the coupling mechanism can consequently take X-ray to visible light conversion efficiency ex-
different forms: electrical, optical, magnetic, ceeding that of anthracene (saturated with PbC12)
thermal, chemical and mechanical coupling are all by at least one order of magnitude in a layer
possibilities. An example of mechanical coupling thickness less than 0.05 cm.
would be a magneto-electric effect produced in a In a sense, product p/operties can be regarded
composite material by mechanical coupling be- as a special, and potentially important, aspect of
tween a magnetostrictive phase and a piezoelectric the general physical behaviour of composite ma-
phase. terials. In the general case, when considering the
The potentialities of this concept are illustrated physical behaviour of a composite material all the
by Tables VI and VII taken from the paper by van possible interactions between the phases should
Suchtelen. Table VI is essentially a 6 x 6 matrix in considered whether they be electrical, magnetic or
which the most obvious X - Y (or Y - Z ) effects mechanical. This is not usually done and in the
have been tabulated, These can then be combined consideration of dielectric properties it is normally
as indicated in Table VII to give the product prop- assumed that mechanical interactions are absent
erties listed in the table. and that piezoelectric effects do not have to be
The feasibility of this approach has been con- considered. On the other hand, in the discussion of
firmed experimentally by (a) the investigation of thermal expansion it is recognized that mechanical
magneto-electric and electromagnetic effects in a interactions must be important and these are taken
2135
into account. The concept of product properties I n S b - S b eutectics are markedly anisotropic in
is, however, of considerable importance since it electrical resistance and that the values of ~ and K
indicates the ways in which interactions arising differ considerably from those for the pure ma-
through different forms of coupling can be ex- terials. Unfortunately, the values of Z are not as
ploited in the development of materials with novel high as the values for either InSb or Sb although
properties. higher values might be obtained by a further
reduction in rod diameter. Similar studies have
5. Other special property combinations been carried out with the B i z T % - T e [3],
As we have seen, a particular property of a com- Mg-MgaTAll2 and B i - C d [164] systems but the
posite material can be varied in a controlled way figure of merit values with these materials were
by controlled changes in the composition and not sufficiently outstanding to make them com-
structure. This means that by a suitable selection petitive with single phase materials [3]. More
of components a n d structural geometry it may be recently Levinson [165] has examined the ther-
possible to achieve a combination of properties moelectric properties of an aligned CrSi2-Si
which cannot be realized with a homogeneous ma- eutectic.
terial. Thus, for example, with fibre composites we The outstanding example of the way in which a
have the possibility of very high mechanical required combination of electrical properties can
strength combined with very high electrical be achieved with a composite material is provided
conductivity. by the development at the Siemens Laboratories
The possibility of achieving a particular com- of magnetoresistive devices based on the direc-
bination of physical properties is illustrated by the tionally solidified InSb-NiSb eutectic in which
work o f Liebmann and Miller [163] on the ther- metallic NiSb rods are aligned in a InSb matrix
moelectric properties of directionally solidified (Fig. 12). This work has been summarized recently
I n S b - S b eutectics. The microstructure consisted
of antimony rods in an InSb matrix. Values for the
specific resistance (p), the thermoelectric power
(4) and the thermal conductivity (K) are given in
Table VIII which also includes values for the figure
of merit (Z) given by
42
pK

It follows from equation 91 that a reduction in

the thermal conductivity will lead to a higher value
for Z and with semi-conductors this can be
Figure12 Polished surface of InSb-NiSb eutectic.
achieved by mixed crystal formation which leads Left: surface parallel to NiSb needles. Right: surface
to statistically distributed lattice distortions [7]. perpendicular to NiSb needles [7]. (Courtesy Professor H.
It will be seen from Table VIII that the Weiss.)

T A B L E V I I I Thermoelectric properties of directionally solidified InSb -Sb eutectics [ 163 ]

Sb rod s i z e Direction of measurement p 4~ K Z X 106
(cm) (to Sb rods) (f~cm) (#V deg-1 ) (Wcrn- 1 deg- 1) (deg- 1)
4.3 X 10-4 [L 3.05 X 1 0 - 4 - - 18 0.1113 9.5
• 3.3 Xl0 -s --71 0.0822 18.6
8.5 X 10 -4 tt 3.00 x 10 -4 -- 12 0.1235 3.9
• 3.3 x10 -3 --53 0.0940 9.1
2.8 x 10 -3 II 3.15 • 10 -4 --8.2 0.1430 1.5
• 3.4 • 10 -3 --28 0.1059 2.2
Pure materials
InSb 10 -2 -- 325 0.162 65.2
Sb 4.4 X 10 .2 + 35 0.189 147.3

2136
by Weiss [7] and will only be discussed briefly in should be theoretically attainable with a short
this review. circuited material.
The origin of the large magnetoresistive effect In the InSb-NiSb eutectic the short circuits are
in these materials can be explained by first con- provided by the metallic NiSb rods (Fig. t3b)
sidering the Hall effect in an elongated plate of which have an electrical conductivity two orders
InSb traversed by a current i flowing in a direction of magnitude greater than that of the InSb matrix.
parallel to the longitudinal axis of the plate. If a With the aligned eutectic about 60% of the theor-
magnetic field B is applied in a direction perpen- etical magnetoresistive effect is obtained. The
dicular both to the plane of the plate and the more important applications of magnetoresistive
direction of the current then a Hall voltage perpen- devices based on the InSb-NiSb eutectic include
dicular to I and B is obtained. If the Hall voltage is the measurement of magnetic fields, contactless
now short circuited by wires at intervals along the variable resistances and potentiometers (Fig. 14),
plate as shown in Fig. 13a, then the Hall current and other contactless controls.
flowing through these wires produces a further
Hall voltage parallel to the primary current i and
results in an additional resistance AR which will be
proportional to the magnetic field. The current
flowing through the InSb will follow a zig-zag path
from one end of the plate to the other. With an in-
finite number of short circuits, the relative
increase in resistance will be (gBB) 2 where/l B is
the electron Hall mobility in a magnetic field B.
For intrinsic InSb, ~BB has a value of 5 in a field
of 10kG so that a 25-fold increase in resistance

iHl
U Figure 14 Contactless nominal current indicator for
electric locomotive brake (courtesy Professor H. Weiss).

i/- //i / /' II In a sense, the InSb-NiSb eutectic could be

i ~ ~ II
- - l I r I regarded as a device rather than a material.
However, as noted by Albers [1], the composite is
BS essentially a new material showing an extremely
J high magnetoresistance effect and this property is
(a) retained whgtever the shape of the specimen. The
effects arise through a special anisotropic combi-
nation of the electrical properties of InSb and
NiSb. A similar high magnetoresistive effect would
be shown by InSb itself if the electron and hole
mobilities were equally high. In fact the hole
~, II ~l I Ir iI i~ mobility in lnSb is small. The composite is, there-
HI ii II fore, essentially a new material with unique
lli i, II r,i[:i
properties.
L]lj- ~ L j _lL_l_ l LJ_ L_l_Ll.l
a J In addition to the magnetoresistive effect the
J Hall current flowing through the NiSb rods gives
(b) rise to a Peltier effect so that a temperature dif-
Figure 13 Magnetoresistive effect in directionally
ference is observed between the two surfaces of
solidified InSb-NiSb eutectic. (a) Externally short the eutectic. Unfortunately, although the figure of
circuited InSb plate. (b) Short-circuiting NiSb needles in merit is 30 times as high as that of homogeneous
InSb-NiSb eutectic. (After Weiss [71 .} InSb, the thermal conductivity of InSb is too high
2137
 3. F. S. GALASSO, d. Metals 19 (6) (1967) 17.
4. F. S. GALASSO, F. C. DOUGLAS and J. A. BATT,
]. Metals 22 (6) (1970) 40.
5. Z. HASHIN, in "Mechanics of Composite Materials"
(Proceedings of the 5th Symposium Naval Structural
Mechanics, Philadelphia, Pa., May 8 - 1 0 , 1967)
edited by F.W. WENDT, H. LIEBOWITZ,
N. PERRONE (Pergamon, 1970) pp. 201-242.
6. J, D. LIVINGSTON, Composites 4 (2) (1973) 70.
7. H. WEISS, Met. Trans. 2 (1971) 1513.
8. J. VAN SUCHTELEN, Philips Res. Repts. 27 (1972)
28.
9. W. F. BROWN, ]. Chem. Phys. 23 (1955) 1514.
10. L. K. H. VAN BEEK, "Progress in Dielectrics",
Vol. 7 (Heywood, London, 1967) pp. 69-114.
11. J. W. RAYLEIGH,Phil. Mag. 34 (1892) 481.
12. D. A. G. BRUGGEMAN, Ann. Phys. Lpz. 24 (1935).
636.
13. C. J. F. BOTTCHER Rec. Tray. Chim. Pays-Bas 64
(1945) 47.
14. W. E. A. DAVIES, Jr. Phys. D. 7 (1974) 120.
15. D. POLDER, and J. H. VAN SANTEN, Physica 12
(1946) 257.
16. E. C. STONER, Phil. Mag. 36 (1945) 803.
Figure 15 Infra-red detector with InSb-NiSb eutectic 17. Z. HASHIN, J. Comp. Mater. 2 (1968) 284.
(courtesy Professor H. Weiss). 18. O. WIENER, Abh. s~ichs. Ges. (akad.) Wiss. 32
(1912) 509.
19. W. F. BROWN, Trans. See. Rheol. 9 (1965) 357.
for devices based on this effect to have practical 20. Z. HASHIN and S. SHTRIKMAN, J. Appl. Phys. 33
value. The converse effect which is observed when (1962) 3125.
a temperature difference AT is imposed on the 21. J. A. REYNOLDS, Thesis (London 1955).
system has, however, been exploited in a 22. M. BERAN, II Nuovo Cimento 38 (1965) 771.
23. P. B. CORSON, J. Appl. Phys. (USA] 45 (1974)
room-temperature detector for the far infra-red 3159.
(Fig. 15). 24. Idem, ibid 45 (1974) 3165.
A large number of aligned eutectics involving 25. Idem, ibid 45 (1974) 3171.
III-V compounds have been examined [7] but 26. Idem, ibid 45 (1974) 3180.
27. M. N. MILLER, d. Math. Phys. 10 (1969) 1988.
none of these has shown the outstanding prop-
28. J. J. McCOY, // Nuovo Cimento 57B (1968) 139.
erties of the unidirectionally solidified InSb- 29. J. M. PETERSON and J.J. HERMANS, J. Comp.
NiSb-system. A magnetoresistive effect was Mater. 3 (1969) 338.
observed with the Cd As-NiAs eutectic [166] but 30. S. G. SPRINGER and S. W. TSAI, ibid 1 (1967) 166.
again this was not as large as that with InSb-NiSb. 31. Z. HASHIN and B.W. ROSEN, J. Appl. Mechs.
(Trans. ASME] 31 (1964) 223.
32. J. DONEA, d. Comp. Mater. 6 (1972) 260.
Acknowledgements 33. M. J. BERAN and N. R. SILNUTZER, ibid 5 (1971)
The author is grateful to the Director of the 246.
National Physical Laboratory for permission to 34. M. A. ELSAYED and J . J . McCOY, ibid 7 (1973)
publish this paper and is indebted to Professor 466.
35. J. C. MAXWELL, "A Treatise on Electricity and
A. Keller, Professor H. Weiss, Dr J. Dlugosz and
Magnetism", VoL 1. 3rd Edn. (Clarendon Press,
Dr. W. Finger for their help in providing Oxford, 1892) p. 440.
illustrations. 36. R. LANDAUER, J. Appl. Phys. 23 (1952) 779.
37. S. M. AHARONI, ibid 43 (1972) 2463.
References 38. F. BUECHE, ibid 43 (1972) 4837.
1. w. ALBERS, Proceedings of the Conference on In 39. Idem, ibid 44 (1973) 532.
Situ Composites, September 5 - 8 (1972), Lakeville, 40. Idem, or. Polymer ScL, Polymer Phys. Ed. (USA) 11
Connecticut, Vol. III (Publication NMAB 308-111, (1973) 1319.
Nat. Acad. Sci., Washington, D.C. 1973), pp. 1--19. 41. J. GURLAND, in "Quantitative Microscopy" edited
2. M. B. BEVER, P.E. DUWEZ and W. A. TILLER, by R.T. De HOFF and F. N. RHINES (McGRAW--
Mat. Sci. Eng. 6, (1970) 149. HILL, New York, 1968) pp. 278-290.

2138
42. A.D. BRAILSFORD and K. G. MAJOR, Brit. J. 78. B.W. ROSEN, in "Composite Materials", Agard
Appl. Phys. 15 (1964) 313. Lecture Series No. 55, Advisory Group for
43. R.E. SKOCHDOPOLE Chem. Eng. Prog. 57(10) Aerospace Research and Development, NATO
(1961) 55. (1972) pp. 1 - 1 to 1-22.
44. A.W. PRATT, in "Thermal Conductivity", Vol. 1 79. L. R. CALCOTE, "The analysis of laminated com-
edited by R.P. TYE (Academic Press, London, posite structures" (Van Nostrand-Reinhold, New
1969) pp. 301-405. York, 1969).
45. A. SUGAWARA and Y. YOSH1ZAWA, J'. AppL 80. S. G, LEKHNITSKII, "Theory of elasticity of an
Phys. 33 (1962) 3135. anisotropic elastic body" translated by P. FERN,
46. T. P. FIDELLE and R. S. KIRK, A.I.Ch.E.J. 17 edited by J. J. Brandstatter (Holder-Day, San
(1971) 1427. Francisco, 1963).
47. R.A. SCHAPERY and R. E. MARTIN, Mechanics 81. J. C. HALPIN and N. J. PAGANO, Technical Report
and Materials Research Center, Texas A & M A F M L - T R - 6 8 - 3 9 5 , December 1969.
University Technical Report M M 7 2 - 2 (1972). 82. Idem, Technical Report A F M L - T R - 6 9 - 3 4 1 ,
48. W. FINGER, Schweiz Monatsschr, Zahnheilkd. 84 August 1970.
(1974) 630. 83. J.C. HALPIN, J. Comp. Mater. 6 (1972) 208.
49. V. M. LEVIN, Mekhanika Tverdogo Tela 1 (1967) 84. N. F. DOW and B.W. ROSEN, NASA Report
88. CR-1324, May 1969.
50. B.W. ROSEN and Z. HASHIN, Int. J. Eng. Sci. 8 85. Idem, Israeli. Teehnol. 7 (1969) 209.
(1970) 157. 86. W. ALTHOF, Deutsche Forschungs- und Versuch-
51. R. A. SCHAPERY, J. Comp. Mater. 2 (1968) 380. sanstalt f'fir Luft- und Rauhmfahrt eV. Institut f'fir
52. J. L. CRIBB, Nature 220 (1968) 576. Flugzeugbau Report F D - 4 4 , October 1971.
53. T. R. STEEL, Int. J. Solids Structures 4 (1968) 87. A. A. FAHMY and A. N. RAGAI, J. Appl. Phys. 41
1149. (1970) 5112.
54. Z. HASHIN, Proceedings of the IUTAM Symposium 88. D. E. GEILER, 28th Ann. Tech. Conf. SPI 1973,
on Nonhomogeneity in Elasticity and Plasticity, Reinforced Plastics/Composites Institute, The So-
Warsaw, Poland (Pergamon Press, New York, 1959) ciety of Plastics Industry Inc, Section 9 - D , pp. 1-8.
p. 463. 89. W. R. GOGGIN, AppL Optics 13 (1974) 444.
55. E. H. KERNER,Proc. Phys. Soe. B. 69 (1956) 808. 90. A.A. FAHMY and A.N. RAGAI-ELLOZY, jr.
56. T. T. WANG and T. K. KWEI, J. Polymer ScL Part Comp. Mater. 8 (1974) 90.
A2 7 (1969) 889. 91. J. C. HALPIN, ibid 3 (1969) 732.
57. A. A. FAHMY and A. N. RAGAI, J. Appl. Phys. 41 92. C.C. CHAM1S, NASA Technical Note D-6696,
(1970) 5108. March 1972.
58. B. BUDtANSKY, J. Mech. Phys. Solids 13 (1965) 93. I. SELSING, J. Amer. Ceram. Soc. 44 (1961) 419.
223. 94. L. HOLLIDAY and J. ROBINSON, J. Mater. Sci. 8
59. Idem, Z Comp. Mater. 4 (1970) 286. (1973) 301.
60. R. HILL, ,L Mech. Phys. Solids 13 (1965) 213. 95. L. E. NIELSEN, J. Comp. Mater 1 (1967) 100.
61. N. LAWS, ibid 21 (1973) 9. 96. G. A. COOPER and J. M. SILLWOOD, J. Mater. ScL
62. B. PAUL, Trans. Mct. Soc. AIME 218 (1960) 36. 7 (1972) 325.
63. P. S. TURNER, J. Res. NBS 37 (1946) 239. 97. T.C. HANSEN and K. E. C. NIELSEN, Jr. Amer.
64. z. HASHIN and S. SHTRIKMAN, J. Mech. Phys. Concrete Inst. 62 (1965) 783.
Solids 11 (1963) 127. 98. G. PICKETT, ibid 2 7 ( 5 ) ( 1 9 5 6 ) P r o c e e d i n g s 52
65. R. HILL, ibid 11 (1963) 357. (1956) 581.
66. L. J. WALPOLE, ibid 14 (1966) 151. 99. F. T. BARWELL and K. W. PEPPER, Trans. Faraday
67. M. BERAN and J. MOLYNEUX, Q. Appl. Mathemat- Soc. 42B (1946) 275.
ics 24(2) (1966) 107 100. P. J. FLORY, "Principles of Polymer Chemistry"
68. J. J. McCOY, Recent Advances in Engineering Sci- (Cornell University Press, Ithaca, New York, 1953)
ence Vol. VI. Ann. Meeting Conf. Eng. Sci., Nov. p. 576.
1969, Princeton N.J. (1970) p. 235. 101. L. R.G. TRELOAR, "The Physics of Rubber
69. M. N. MILLER, J. Math. Phys. 10 (1969) 2005. Elasticity", 2nd Edn. (Clarendon Press, Oxford,
70. B. W. ROSEN, Proe. Roy. Soc. Lord. A. 319 (1970) 1958) p. 123.
79. 102. A. Y. CORAN, K. BOUSTANY and P. HAMED, J.
71. G. A. VAN FO FY Soy. Phys. Doklady 11 (1966) Appl. Poly. 15 (1971) 2471.
176;Doklady Akad, Nauk SSSR 166 (1965) 817. 103. W. W. BARKAS, Discuss. Faraday Soc. 42B (1946)
72. R. HILL, J. Mech. Phys. Solids 12 (1964) 199. 137.
73. Idem, ibm 13 (1965) 189. 104. I.D. CAVE, Wood Sci. Technol. 6 (1972) 157.
74. N. LAWS, Int. J. Eng. Sci. 12 (1974) 79. 105. R.C. WYATT and K. H. G. ASHBEE, Fibre Sci.
75. W. SCHNEIDER, Kunstoffe 61 (1971) 273. Technol. 2 (1969) 29.
76. Idem, ibid 63 (1973) 929. 106. K. H. G. ASHBEE and R.C. WYATT, Proc. Roy.
77. L. B. GRESZCZUK, AIAA Structures and Materials Soc. A. 312 (1969) 553.
Conference 6th, Palm Springs, California 5 - 7 April 107. F. M. A. CARPAY and J. VAN DEN BOOMGARD,
1965 (AIAA, New York, 1965) pp. 285-290. ActaMet. 19 (197 !) 1279.

2139
108. F. M. A. CARPAY and W. A. CENSEI Scripta Met. 8 137. M. STRONGIN, F. KAMMERER, D.H.
(1974) 11. DOUGLASS, JUN. and M. H. COHEN, Phys: Rev.
109. F. W. PEAKER, in "Techniques of Polymer Charac- Letters 19 (1967) 121.
terization", edited by P.W. ALLEN (Butterworths, 138. I. D. LIVINGSTON, J. AppL Phys. 38 (1967) 2408.
London, 1959) pp. 131- 170. 139. B. J. MADDOCK, Composites 1 (1969) 104.
110. P. DEBYE, J. Phys. Chem. 51 (1947) 18. 140. M. N. WILSON, ibid 1 (1970) 341.
111. Idem, J. Appl. Phys. 15 (1948) 338. 141. M. N. WILSON, C. R. WALTERS, J. D. LEWIN and
112. R. D. MAURER, ibid 33 (1962) 2132. P. F. SMITH, J. Phys. D. 3 (1970) 1517.
113. L. PROD'HOMME, Verres Refract. 22 (1968) 604. 142. C. R. WALTERS, ]. Phys. D. 3 (1970) 1547.
114. R. S. STEIN and M. B. RHODES, J. Appl. Phys. 31 143. M. N. WILSON and C. R. WALTERS ibid 3 (1970)
(1960) 1873. 1547.
115. R. E. PRUD'HOMME and R. S. STEIN, J. Polymer 144. P. F. SMITH, M. N. WILSON and A. H. SPURWAY,
Sci., Polymer Phys. Ed. (USA) 11 (1973) 1357. ibid 3 (1970) 1561.
116. D. R. CLARKE, J. Mater. Sci. 10 (1975) 172. 145. A. H. SPURWAY, J.D. LEWlN and P. F. SMITH,
117. A. J. SIEVERS, Proceedings of the Conference on In ibid 3 (1970) 1572.
Situ Composites, September 5 - 8 (1972) Lakeville, 146. P.S. SWARZ and C.P. BEAN, General Electric
Connecticut, Vol. III (Publication NMAB 308-III, Research and Development Centre, New York,
Nat. Acad. Sci., Washington, D.C. 1973) Report 6 8 - C - 0 3 9 (1968).
pp. 129-141. 147. S. L. WlPF, Phys. Rev. 161 (1967) 404.
118. A. KELLER, E. PEDEMONTE and F.M. 148. P. F. CHESTER, Proceedings of the 1st Cryogenic
WlLLMOUTH, Nature, Lond. 225 (1970) 538. Engineering Conference, 9 - 1 3 April (1967) Tokyo
119. J. DLUGOSZ, A. KELLER and E. PEDEMONTE, and Kyoto, Japan (Heywood-Temple, London,
Kolloid-Z. Z. Polym. 242 (1970) 1125. 1968), pp. 147-149.
120. R. G. C. ARRIDGE and M. J. FOLKES, J. Phys. D. 149. P. F. CHESTER, Repts. Prog. Phys. 30 Part II
(AppL Phys.) 5 (1972) 344. (1967) 561.
121. J. BROWN, "Progress in Dielectrics", Vol. 2 edited 150. H. R. HART, Proceedings of Summer Study on
by J. B. BIRKS and J. H. SCHULMAN (Heywood, Superconducting Devices and Accelerators, 10 J u n e -
London, 1960) pp. 193-225. 19 July (1968) Brookhaven National Laboratory
122. C. KITTEL, "Introduction to Solid State Physics", (Brookhaven National Laboratory, Upton, New
3rd Edn. (Wiley, New York, 1966) p. 488. York, 1969) pp. 571-600.
123. J. H. WERNICK, Ann. Rev. Mat. Sci. 2 (1972) 607. 151. H. LONDON, Phys. Letters 6 (1963) 162.
124. J. E. GOULD, "Cobalt Alloy Permanent Magnets", 152. R. HANCOX, Proc. lEE 113 (1966) 1221.
Cobalt Monograph Series 'Centre d'Information du 153. S. L. WIPF, Proceedings of Summer Study on
Cobalt, Brussels, 1971). Superconducting Devices and Accelerators, 10
125. P. R. SAHM and F. HOFER, Z. Angew. Phys. 30 J u n e - 1 9 July (1968) Brookhaven National Labora-
(1970) 95. tory, (Brook Upton, New York, 1969) pp, 511-543.
126. H. WAGINI and M. WILHELM, Z. Naturforschung 154. A. S. YUE, Y.W. CHANG and M.P. MATHUR,
21a (1966) 329. Proceedings of the Conference on In Situ Com-
127. D. L. ALBRIGHT, G.P. CONRAD II and R.W. posites, September 5 - 8 (1972) Lakeville, Con-
KRAFT, J. AppL Phys. 38 (1967) 2919. necticut, Vol. III (Publication NMAB 308-1II, Nat.
128. M.R. JACKSON, R.N. TAUBER and R.W. Acad. Sci., Washington, D.C. 1973) pp. 7 9 - 8 3 .
KRAFT, Or.AppL Phys. 39 (1968) 4452. 155. A. GUINIER, C. PETIPAS, G. SAUVAGE; J.
129. J. D. LIVINGSTON, ibid 41 (1970) 197. BAIXERAS, J.M. DUPART, G. FOURNET, G.
130. J.M. NOOTHOVEN VAN GOOR and J.H. FONTAINE, M. G. BLANCHIN, M. TURPIN and R.
ZIJLSTRA, ibid 39 (1968) 5471. RACEK, ibid pp. 85-96.
131. J. C. BOULBES, R.W. KRAFT, M. R. NOTIS and 156. W. A. LITTLE, Canacl. J. Phys. 37 (1959) 334.
C. D. GRAHAM, Proceedings of the Conference on 157. A. C. ANDERSON and R. B. RAUCH, J. Appl. Phys.
In Situ Composites, September 5 - 8 (1972) 41 (1970) 3648.
LakeviUe, Connecticut, Vol. III (Publication NMAB 158. K.W. GARRETT and H. M. ROSENBERG, J. Phys.
308-111, Nat. Acad. Sci., Washington, D.C. 1973) D. 7 (1974) 1247.
pp. 6 1 - 7 8 . 159. R. E. MEREDITH and C. W. TOBIAS, J. Appl. Phys.
132. J.J. de JONG, J~ M, G. SMEETS and H.B. 31 (1960) 1270.
HAANSTRA, J. AppL Phys. 29 (1958) 297. 160. R. L. HAMILTON and O. K. CROSSER, lnd. Eng.
133. D, A. r and R. KOSSOWSKY, Met. Trans. Chem. Fund. 1 (1962) 187.
2 (1971) 1523. 161. Y. S. TOULOUKIAN, R. W. POWELL, C. Y. HO and
134. R. KOSSOWSKY and D. A. COLLING , "Science of P.G. KLEMENS, "Thermophysical Properties of
Advanced Materials and Process Engineering", Vol. Matter" Vol. 2 Thermal Conductivity -
15 (Society of Aerospace Material and Process Nonmetallic Solids (IFI/Plenum, New York, 1970).
E n g i n e e r s - Western Periodicals, North Hollywood, 162. P. G. KLEMENS and I.N. GREENBERG, jr. Appl.
California 1969) pp. 491-500. Phys. 44 (i973) 2992.
135. J. H.P. WATSON, Phys. Rev. 148 (1966) 223. 163. W. K. LIEBMANN and E.A. MILLER, ibid 34
136. Idem, Phys. Letters 25A (1967) 326. (1963) 2653.

2140
164. A. YUE, Met. Soc. AIME, Fall Meeting, Philadelphia, 167. E. STEPHENS and E. J. EVANS, Phil. Mag. 7 (1929)
Pa., Oct. 1969. 161.
165. L. M. LEVINSON, Proceedings of the Conference on 168. N. J. STEPANOW, Z. anorg, u. allgem. Chem. 78
In Situ Composites, September 5 - 8 (1972) (1912) 1.
Lakeville, Connecticut, Vol. III (Publication NMAB 169. "Handbook of Chemistry and Physics", 49th Edn.,
3 0 8 - I l i , Nat. Acad. Sci., Washington D.C. 1973) edited by R. C. Weast (The Chemical Rubber Co.
pp. 9 7 - 1 0 9 . Cleveland, Ohio, 1968).
166. C.T. ELLIOTT and S. E. R. HISCOCKS, Brit. J.
AppL Phys. (s Phys. 19) Ser. 2 2 (1969) 1083. Received 10 November 1975 and accepted 1 March 1976.

2141