10.1098/ rspa.2002.

1054

Dynamic Eshelby tensor and potentials

for ellipsoidal inclusions
By T h o m a s M. M ic h eli ts c h1 , H u a j ia n G a o1
a n d Valery M. L e v i n2
1
Max Planck Institute for Metals Research, Department of Theory of
Mesoscopic Phenomena, Heisenbergstrasse 3, 70569 Stuttgart, Germany
2
Instituto Mexicano del Petr¶oleo, Eje Central L¶azaro C¶ardenas, No. 152,
Col. San Bartolo Atepehuacan, C.P. 07730, M¶ exico, D.F., Mexico
([email protected]; [email protected]; [email protected])

Received 10 May 2002; accepted 30 July 2002; published online 3 February 2003

The dynamic Eshelby inclusion problem for an ellipsoidal inclusion in a three-

dimensional in­ nite elastic isotropic medium is considered. The dynamic Eshelby
tensor is expressed in terms of solutions of the Helmholtz equation (Helmholtz
potentials). A compact formulation for the components of the dynamic Eshelby
tensor is derived for the inside region of the inclusion. For spheroidal inclusions,
one-dimensional integrals similar to the elliptic integrals are obtained. The approach
leads to closed-form expressions in cases such as spheres and continuous ­ bres coin-
ciding with those given in 1990 by Mikata & Nemat-Nasser and in 2002 by Michelitsch
et al . by employing other techniques. For the inside region of the inclusion, the static
limit is performed in closed form and coincides with Eshelby’s classical 1957 result.
Keywords: Helmholtz equation; dynamic Green’s function;
dynamic Eshelby tensor; dynamic inclusion problem

1. Introduction
The determination of the elastodynamic ­ elds due to ellipsoidal inclusions and inho-
mogeneities are of fundamental interest in a wide range of physical and engineering
problems in the mechanics of micro-inhomogeneous solids. For instance, in the non-
destructive evaluation (NDE), the knowledge of dynamic ­ elds is of special interest
(e.g. Pao 1978; Miklowitz & Achenbach 1978; Fu 1982). For modelling the dynamic
properties of materials, the time-harmonic solution (frequency{space representation)
of the inclusion problem provides useful information, since any time dependence of
the transformation strain may be represented by the superposition of time-harmonic
transformation strains (Mikata & Nemat-Nasser 1990). Especially if the phase trans-
formation is reversible, knowledge of the time-harmonic solution is of high practical
importance (Marshall & James 1986).
Throughout this paper we denote as an inclusion a region containing a distribution
of eigenstrains with the same moduli as the matrix. In the case of isotropic media,
elastodynamic ­ elds can be expressed in terms of the solution of the Helmholtz
equation (e.g. Fu & Mura 1982; Michelitsch et al . 2002).

Proc. R. Soc. Lond. A (2003) 459, 863{890 °

c 2003 The Royal Society
863
864 T. M. Michelitsch and others

In static elasticity, the inclusion problem was ­ rst solved by Eshelby (1957) for
ellipsoidal inclusions in an in­ nite isotropic medium. Analogous tensors can be
de­ ned for a piezoelectric transversely isotropic medium. For spheroidal inclusions,
closed-form expressions of the Eshelby tensors were derived by Michelitsch & Levin
(2000) and Levin et al. (2000) and independently by Mikata (2001), following di¬er-
ent approaches.
In the framework of dynamics, the inclusion and inhomogeneity problems were
only slightly touched upon due to their much higher mathematical complexity. In the
case of spherical inclusions, closed-form expressions for dynamic Eshelby tensors were
given by Mikata & Nemat-Nasser (1990). The case of a two-dimensional quasiplanar
piezoelectric inclusion problem with a time-harmonic eigenstrain and eigenelectric
­ eld was solved in closed form only recently by Levin & Michelitsch (2001) and
Levin et al . (2002) and by Michelitsch et al . (2002). In three-dimensional elasticity,
the problem of harmonic wave interaction with transforming inhomogeneities was
treated by Mikata & Nemat-Nasser (1991).
For isotropic elastic media, the dynamic ­ elds due to a time-harmonically trans-
forming inclusion can be expressed by solutions of an inhomogeneous Helmholtz
equation. This solution (the Helmholtz potential) is de­ ned as a convolution of the
source density with the Green function of the Helmholtz equation. A formulation
of the volume integrals associated with the Helmholtz potential of ellipsoidal unit-
density source distributions was given by Fu & Mura (1982) in terms of in­ nite
series.
The subject of this paper is the dynamic variant of Eshelby’s inclusion problem for
an ellipsoidal inclusion. The paper is organized as follows. In x 2 we derive expressions
for the dynamic displacement ­ elds of an ellipsoidal inclusion. The displacements
are represented in terms of Helmholtz potentials. By using these expressions, the
dynamic Eshelby tensor is de­ ned in the same way as in static elasticity (Mikata &
Nemat-Nasser 1990).
In x 3 we derive general integral representations for the Helmholtz potential of
an ellipsoidal source region S, for the inside and the outside of S. Two alternative
formulations are given using an integral transformation. We show that the imaginary
part of this solution, which is a solution of the homogeneous Helmholtz equation, is
covered by the same expression for the inside and outside of S.
In x 4 the Helmholtz potentials of spheres and continuous ­ bres are derived in
closed form. Here we reobtain the expressions given by Mikata & Nemat-Nasser
(1990), Levin & Michelitsch (2001), Levin et al . (2002) and Michelitsch et al . (2002).
In x 5 expressions for the components of the dynamic Eshelby tensor are given for
the inside region of an ellipsoidal inclusion. All components which are non-vanishing
in the static limit are expressed in terms of three types of integrals. This compact
representation makes the approach useful for numerical applications.
The static limiting case coinciding with Eshelby’s (1957) classical result is consid-
ered in x 5 b.

2. Dynamic inclusion problem

The inclusion is assumed to have the same material characteristics (elastic moduli
and mass density) as the matrix material. The linear constitutive relations in such

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 865

a material have the form

¤
¼ ij = Cijrs (° rs ¡ ° rs ); (2.1)
¤
where ¾, ², ² and C denote stress, strain, eigenstrain and the tensor of elastic
constants, respectively, with
° ij = 12 (@i uj + @j ui ): (2.2)
The inclusion S is assumed to undergo a time-dependent spatially uniform eigen-
strain ²¤ of the form
²¤ = f (t)£ s (r)²0 ; (2.3)
where r = (xi ) = (x; y; z) denotes Cartesian spatial coordinates £ s (r) the charac-
= S) and ²0 is
teristic function of the inclusion (£ s (r) = 1, r 2 S, £ s (r) = 0 and r 2
a constant tensor. The function f (t) is an arbitrary function of time. We assume the
absence of external body forces. The equations of motion are then given by
@2
» u i = @j ¼ ij ; (2.4)
@t2
where » denotes the mass density of the material. With (2.1) and (2.2), this equation
assumes the form
(T (r) + » ! 2 1)u(r; !) + f ¤ (r; !) = 0 (2.5)
in the Fourier-transformed space-frequency representation, where Tij (r) = Cikjl @k @l ,
1 is the 3 £ 3 unity tensor and f ¤ (r; !) is the e¬ective force density that induces the
same displacement ­ eld u in the undisturbed matrix as the inclusion and is obtained
as
fi¤ (r; !) = ¡ f~(!)Cijrs ° 0rs @j £ s (r); (2.6)
~
where f (!) is the Fourier transform of function f (t) which is de­ ned by
Z + 1
f~(!) = ei!t f (t) dt: (2.7)
¡1

Since the arbitrary time dependence of the eigenstrain a¬ects the space{frequency
representation of the induced ­ elds only by a multiplier f~(!)y, we omit this factor
corresponding to the !-representation of a ¯ (t)-type time dependence of the eigen-
strain. The displacement ­ eld u(r; !) can be represented as a convolution with the
dynamic Green function by
Z
^ ¡ r 0 ; !)f ¤ (r 0 ; !) d3 r0 :
u(r; !) = G(r (2.8)

^ is de­ ned by
This Green function G
^
(T (r) + » ! 2 1)G(r; !) + 1¯ 3 (r) = 0; (2.9)
where ¯ 3 (r) denotes the spatial ¯ -function. For our goal of determining u in com-
pact form it is convenient to introduce a tensor function G(r; !) determined by the
characteristic function £ s (r) of the inclusion. This function is de­ ned by
(T (r) + » ! 2 1)G(r; !) + 1£ s (r) = 0: (2.10)
y A step function f (t) = £(t) time dependence of the eigenstrain, for example, describes a sudden
uniform switching process corresponding to a martensitic transformation.

Proc. R. Soc. Lond. A (2003)

866 T. M. Michelitsch and others

Obviously, G and G^ are then related by

Z Z
G(r; »; !) = G(r^ ¡ r 0 ; !)£ s (r0 ) d3 r0 = ^ ¡
G(r r0 ; !) d3 r 0 ; (2.11)
S
where » indicates a set of geometric characteristics of the inclusion. We shall call
the tensor function G of (2.11) the dynamic potential of the inclusion. For isotropic
material and inclusions of arbitrary shapes, the dynamic potential function (2.11)
has the form (e.g. Levin & Michelitsch 2001)
1
G(r; »; !) = [­ 2 g(r; »; ­ 2 )1 ¡ r « rfg(r; »; ­ 1 ) ¡ g(r; »; ­ 2 )g]; (2.12)
» !2 2
with the two frequencies ­ given by
s
r r
! » ! »
­ 1 = =! ; ­2 = =! ; (2.13)
c1 ¶ + 2· c2 ·
corresponding to one longitudinal and two transversal acoustic waves in an isotropic
three-dimensional medium, where ¶ and · denote the Lam´ e constants. The func-
tions (Helmholtz potentials) g(r; »; ­ i ) are the dynamic analogue of the Newtonian
potentials in statics for the source region of unit density represented by the inclusion.
The Helmholtz potentials are the solutions of the scalar inhomogeneous Helmholtz
equation
(¢ + ­ 2 )g(r; »; ­ ) + £ s (r) = 0: (2.14)
Taking into account equations (2.6), (2.8) and (2.11), we can represent the dis-
placement ­ eld u(r; !) in the form
0
ul (r; »; !) = ¡ Ckjrs ° rs @j G kl (r; »; !): (2.15)
1
For the strain ° ij = 2 (@i uj + @j ui ) we then can write
0
° il (r; »; !) = ¡ Ckjrs ° rs (Pijkl (r; »; !))(il) ; (2.16)
where (il) indicates symmetrization with respect to the subscript `il’. Here we intro-
duce
Pijkl (r; »; !) = @i @j Gkl (r; »; !): (2.17)
Using (2.16) we de­ ne the dynamic Eshelby tensor S analogously to statics by
(Mikata & Nemat-Nasser 1990)
0
° il (r; »; !) = Silrs (r; »; !)° rs ; (2.18)
where the dynamic Eshelby tensor is given by
Silrs (r; »; !) = ¡ Ckjrs (Pijkl (r; »; !))(il) : (2.19)
So far our derivation holds for inclusions of arbitrary shape. In the next section we
will derive a compact integral formulation for the Helmholtz potential g(r; »; ­ ) of
an ellipsoidal source region. The solution of this problem is the crucial part to derive
a convenient representation for the components of the dynamic Eshelby tensor of an
ellipsoidal inclusion (x 5).
Unlike in statics, in the dynamic case Pijkl is a spatially non-uniform tensor func-
tion inside an ellipsoidal inclusion. When we expand Pijkl in a series with respect to
!, the zero order in ! corresponds to the static limit leading to the classical result
of Eshelby (1957). This is con­ rmed explicitly in x 5 b.

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 867

a3
y
a2
a1
a1 x
a2

a3

Figure 1. Ellipsoidal source region with semi-axes ai .

3. Helmholtz potential of an ellipsoidal source region

Here we derive integral formulations for the inhomogeneous solution of Helmholtz
equation for a source distribution of unit density inside an ellipsoidal source region.
Such a density distribution is conveniently represented by a characteristic func-
tion £ s (r) = £ (1 ¡ p), where £ (¹ ) denotes the Heaviside step function, de­ ned by
£ (¹ ) = 0 for ¹ < 0 and £ (¹ ) = 1 for ¹ > 0. The ellipsoidal region is characterized by

x2 y2 z2
+ + = p2 ; (3.1)
a21 a22 a23

where ai denotes the semi-axes of the ellipsoid. The parameters p < 1 (r 2 S) and
p > 1 (r 2 = S) characterize the inside and outside region of the ellipsoid S, respec-
tively. The origin r = 0 of our Cartesian coordinate system is located in the centre
of the ellipsoid (­ gure 1).
The Helmholtz potential we are looking for is the solution of the inhomogeneous
Helmholtz equation
(¢ + ­ 2 )g(r; ai ; ­ ) + £ (1 ¡ p) = 0: (3.2)

Here we introduced ­ = (! + i° )=c with the angular frequency ! and an in­ nites-
imal damping constant ° ! 0.y Obviously, the Helmholtz potential de­ ned in (3.2)
is given by
Z
g(r; ai ; ­ ) = g^(r ¡ r0 ; ­ )£ s (r 0 ) d3 r 0 ; (3.3)

where we introduced the Green function (the fundamental solution)

µ i­ r ¶
e
g^(r; ­ ) =
4º r
y The in¯nitesimal damping constant ² ! +0 regularizes the problem and leads to causal behaviour
(see, for example, Levin & Michelitsch 2001).

Proc. R. Soc. Lond. A (2003)

868 T. M. Michelitsch and others

of the Helmholtz equation. For our evaluation it is convenient ­ rst to consider the
r{t representation g(r; ai ; t) which ful­ ls the wave equation
µ µ ¶2 ¶
1 @
¢¡ + ° g(r; ai ; t) + £ (1 ¡ p) ¯ (t) = 0: (3.4)
c2 @t
With the spatial Fourier transformation, g(r; ai ; t) can be represented in the form
(Levin & Michelitsch 2001; Michelitsch et al . 2002)
Z
1
g(r; ai ; t) = eik¢r £ ~ s (k; ai )g(k; t) d3 k; (3.5)
(2º )3
where Z 1
1 e¡i!t sin(ckt)
g(k; t) = d! = c£ (t) e¡° t : (3.6)
2º ¡1 k2 ¡ 2
(! + i° ) =c2 k
The above expression is the causal Green function of an in­ nitesimally damped
harmonic oscillator of eigenfrequency ck. The causality is indicated by the step func-
tion £ (t). The representations g(r; ai ; t) and the desired solution g(r; ai ; ­ ) of (3.2)
are related via the Fourier transformation
Z 1
g(r; ai ; ­ ) = g(r; ai ; t) ei!t dt; (3.7)
0

where the lower integration limit t = 0 is a consequence of the causality of g(r; ai ; t).
In (3.5) we have introduced the Fourier transform £ ~ s (k; ai ) of the characteristic
function £ s of the source region de­ ned as
Z
£ ~ s (k; ai ) = e¡ik¢r £ s (r) d3 r: (3.8)

For an ellipsoid with semi-axes ai , this integral yields (Appendix B)

µ ¶
~ 4º a1 a2 a3 4º a1 a2 a3 d sin(¶ ks)
£ s (k; ai ) = (sin ks ¡ ks cos ks) = ¡ ; (3.9)
(ks)3 (ks)3 d¶ ¶
q ¶ = 1

where s(k)^ = a2 k^ 2 + a2 k^ 2 + a2 k^2 and k = k k ^ with jkj = k. With the above expres-
1 1 2 3 1 3
sion, we can write
ca1 a2 a3 e¡° t £ (t)
g(r; ai ; t) = ¡
2º 2
½ Z ^ Z 1 ^ ¾
@ 1 d« (k) eik k¢r sin(ckt) sin ¶ ks
£ dk (3.10)
@¶ ¶ jkj=
^ 1 s3 (k)^ 0 k2 ¶ = 1
^ = 1. Since we
for (3.5), where d« denotes the surface element of the unit sphere jkj
^
integrate over the unit sphere jkj = 1 in equation (3.10), we add vanishing contribu-
tions (odd functions) to obtain
ca1 a2 a3 e¡° t £ (t)
g(r; ai ; t) = ¡
2º 2
½ Z ^ Z 1 ^ ^ ¾
@ 1 d« (k) 1 eik(k¢r + ¶ s¡ct)
¡ eik(k¢r + ¶ s+ ct)
£ Re dk :
@¶ ¶ jkj=
^ 1 s3 (k)^ 2 0 k2 ¶ = 1
(3.11)

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 869

The k integral yields (Appendix C)

Z 1 ^ ^
1 eik(k¢r + ¶ s¡ct)
¡ eik(k¢r + ¶ s+ ct)
I = 2 Re dk
0 k2
^ + ct + ¶ s(k^i )j ¡
= 14 º (jr ¢ k ^ + ¶ s(k)
jr ¢ k ^ ¡ ctj): (3.12)
To evaluate the integral (3.11) with (3.12) we put xi = pai ni (n2i = 1) and take
into account the fact that k ¢ r = psi ni = ps cos ® with
q
^
si = ki ai ; s = s21 + s22 + s33 :
In view of (3.12) we have to distinguish between the two cases p < 1 (r 2 S) and
p > 1 (r 2= S). We denote the corresponding Helmholtz potentials as g in and g ou t ,
respectively. We begin with the inside region.
Case I: p < 1 (r 2 S). Solution inside the ellipsoid. In view of (3.12) we
observe that the inequality k ¢ r + s > 0 holds on the whole unit sphere jkj ^ = 1.
^ ^
Only the term r ¢ k + ¶ s(k) ¡ ct of (3.12) changes the sign. Let us introduce time
^ + ¶ s ¡ ct1 = 0, which gives t1 = (¶ s + r ¢ k)=c
t1 , determined by r ¢ k ^ > 0 (¶ = 1).
Equation (3.12) can then be rewritten as
I(r; s; t) = 12 cº [£ (t1 ¡ t)(t ¡ t1 ) + t1 ]: (3.13)
Since we are interested in the r{­ representation only, it is convenient to introduce
Z
2c 1 1 ^
J= dt ei!t e¡° t I(r; a; t) = 2 (ei­ ¶ s ei­ r ¢k ¡ 1): (3.14)
º 0 ­
Then (3.11) takes in the r{­ representation the form
½ Z ^ ¾
a1 a2 a3 d 1 ^ J(k; r; ¶ )
g in (r; ai ; ­ ) = ¡ d« (k) : (3.15)
4º d¶ ¶ jkj=
^ 1 s3 (k^i ; ai ) ¶ = 1

We then ­ ndy
½Z ^ ¾
a1 a2 a3 d« (k) ^ ^
g in (r; ai ; ­ ) = [(1 ¡ i­ s(k^i ; ai )) ei­ s(ki ;ai ) ei­ r ¢k ¡ 1] : (3.16)
4º ­ 2 ^ j= 1
jk s3 (k^i ; ai )
By putting k^i = (K ^ i =ai )s with s = (K
^ 2 =a2 )¡1=2 and using the transformation rela-
i i
tion (A 6)z
Z Z µ ^ ¶
^ f (k)
^ = 1 ^ s 3 f Ki s ;
d« (k) d« (K) (3.17)
^ 1
jkj= a1 a2 a3 jK^ j= 1 ai
we obtain (3.16) in the alternative representation
½ Z ¾
1 1 ^ (1 ¡ i­ s) ei­ s(1+ ^ i =ai )
g in (r; ai ; ­ ) = 2 d« (K) xi K
¡ 1 : (3.18)
­ 4º jK^ j= 1
This representation of the Helmholtz potential for the inside region is used to
derive the components of the Eshelby tensor in x 5. For the sake of completeness we
derive next the Helmholtz potential outside the source region.
^ ^
y Since s is an even function on the unit sphere in this expression, ei­ r ¢ k can be replaced by cos ¯r ¢ k.
z Corresponding to the coordinate transformation ki = Ki =ai in the k-space.

Proc. R. Soc. Lond. A (2003)

870 T. M. Michelitsch and others

Case II: p > 1 (r 2= S). Solution outside the ellipsoid. In (3.11) we now have
^ = 1, namely k
to distinguish between two regions on the unit sphere jkj ^¢r+s>0
^
(region 1) and k ¢ r + s < 0 (region 2). Correspondingly we decompose
g ou t
= g1ou t
+ g2ou t : (3.19)
^ ¢ r + s > 0).
Region 1 (k
The term in the ­ rst absolute value on the right-hand side of (3.12) is always
^ ¢ r + s)=c. Hence we obtain
positive, whereas the second changes the sign at t1 = (k
½Z ^
a1 a2 a3 d« (k)
g1ou t (r; ai ; ­ )=
4º ­ 2 ^ + s(k
k¢r ^ i ;ai )>0 s3 (k^i ; ai )
¾
^ ^
£ [(1 ¡ i­ s(k^i ; ai )) ei­ s(ki ;ai ) ei­ r ¢k ¡ 1] : (3.20)

^ ¢ r + s < 0).
Region 2 (k
By using (3.12) in this case we ­ nd
½Z ^
a1 a2 a3 d« (k)
g2ou t (r; ai ; ­ ) = ¡
4º ­ 2 ^ + s(k
k¢r ^ i ;ai )<0 s3 (k^i ; ai )
¾
^ ^
£ [(1 + i­ s(k^i ; ai )) e¡i­ s(ki ;ai ) e¡i­ r ¢k ¡ 1] : (3.21)

Applying (3.17) yields the convenient representations

Z
ou t 1 ^ [(1 ¡ i­ s) ei­ s(1+ ^ i =ai )
xi K
g1 (r; ai ; ­ ) = d« (K) ¡ 1] (3.22)
4º ­ 2 1+ xi K^ i =ai >0
and
Z
1 ^ [(1 + i­ s) e¡i­ s(1+ ^ i =ai )
g2ou t (r; ai ; ­ ) = ¡ 2
d« (K) xi K
¡ 1]; (3.23)
4º ­ ^ i =ai <0
1+ xi K

where the integrations are performed on the unit sphere jKj ^ = 1 over region 1
^ ^ i =pai = cos ³ .
(cos ³ > ¡ 1=p) and region 2 (cos ³ < ¡ 1=p) when we put K ¢ n = xi K
Moreover, the representation
t¤
g ou t
= g in + g2ou t
+ g2ou (3.24)
is useful, where g in and g2ou t are given by equations (3.18) and (3.23), respectively,
and `¤’ denotes complex conjugation. When approaching the surface of the ellipsoid
from the outside region p ! 1 + 0, we observe that (3.23) then tends to zero. This
can be seen by considering the size of region 2, which is given by
Z µ ¶
^ 1
lim d« (K) = 2º lim 1 ¡ = 0: (3.25)
p! 1+ 0 1+ pK ^ ¢n <0 p! 1+ 0 p
As a consequence we ­ nd
lim g2ou t
= 0; (3.26)
p! 1+ 0

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 871

since the integrand of (3.23) is ­ nite on the complete unit sphere jKj ^ = 1, especially
ou t
in region 2. Moreover, we observe in (3.23) that g2 is non-zero only for p > 1,
i.e. in the outside region, since 1 + pK ^ ¢ n < 0 is never valid for p < 1. If p < 1,
^
the entire unit sphere jKj = 1 is covered by region 1 (1 + pK ^ ¢ n > 0) and (3.22)
coincides with (3.18), whereas (3.23) is vanishing. Hence we can conclude that the
Helmholtz potential g = £ (1 ¡ p)g in + £ (p ¡ 1)g ou t is a continuous function when
crossing the boundary p = 1 of the ellipsoid. It is worthwhile to note that the
imaginary part of g ou t and g in of (3.16) coincide, namely
Z ^
a1 a2 a3 d« (k) ^¢r
Im g(r; ai ; ­ ) = (sin ­ s ¡ ­ s cos ­ s) cos ­ k (3.27)
4º ­ 2 jk^ j= 1 s3 (k^i ; ai )

or, with (3.9) 1 ,

Z
­ ^ £ ~ s (­ k;
^ ai ) cos ­ k
^ ¢ r:
Im g(r; ai ; ­ ) = d« (k) (3.28)
(4º )2 ^ 1
jkj=

Equations (3.27) and (3.28) hold for r both inside and outside the ellipsoid. We
note that (3.28) can also be obtained directly, by evaluating the imaginary part of
Z
1 ik¢r £ ~ s (k; ai )
g(r; ai ; ­ ) = e d3 k (3.29)
(2º )3 k 2 ¡ (! + i° )2 =c2
by taking into account the fact that
1 º
Im = (¯ (k ¡ ­ )¡ ¯ (k + ­ )): (3.30)
k2 ¡ (! + i° )2 =c2 2k
Moreover, we observe that Im g is a solution of the homogeneous Helmholtz equa-
tion (¢ + ­ 2 ) Im g = 0.

4. Special cases
Before we focus our attention on the dynamic Eshelby tensor we devote this section
to some special cases which allow us to ­ nd the Helmholtz potentials in closed form.

(a) Sphere
First of all we consider the simplest case, namely a spherical source region with
radius q
ai = a = s = a2i k^i2 ;
where p = r=a and r = rn.
Case I: p < 1 (r 2 S). Equation (3.18) takes the form
½ Z ¾
1 (1 ¡ i­ a)ei­ a ^ ei­ rn ¢K^ ¡
g in (r; ai ; ­ ) = 2 d« (K) 1 ; (4.1)
­ 4º ^
jK j= 1

where Z
^
^ ) ei­ rn ¢K = 4º sin ­ r
d« (K : (4.2)
^ j= 1
jK r

Proc. R. Soc. Lond. A (2003)

872 T. M. Michelitsch and others

Thus · ¸
in 1 i­ a sin ­ r
g (r; a; ­ ) = 2
e (1 ¡ i­ a) ¡ 1 : (4.3)
­ ­r
This expression coincides with the result given by Mikata & Nemat-Nasser (1990).
Case II: p > 1 (r 2 = S). We evaluate equations (3.22) and (3.23), which take the
following forms by introducing spherical coordinates ³ , ’ with n ¢ K ^ = cos ³ after
the ’ integration,
½ Z º ¡³ 0 Z º ¡³ 0 ¾
ou t 1 i­ a i­ r cos ³
g1 (r; a; ­ ) = (1 ¡ i­ a)e e sin ³ d³ ¡ sin ³ d³ (4.4)
2­ 2 0 0

and
½ Z º Z º ¾
1
g2ou t (r; a; ­ ) = ¡ (1 + i­ a)e ¡i­ a
e¡i­ r cos ³
sin ³ d³ ¡ sin ³ d³ ;
2­ 2 º ¡³ 0 º ¡³ 0
(4.5)
where ³ 0 = cos ¡1 (a=r). These integrals yield
½ ¾
1 ei­ a i­ r a
g1ou t (r; a; ­ ) = 2
(1 ¡ i­ a) (e ¡ e¡i­ a ) ¡ ¡ 1 (4.6)
2­ i­ r r
and
½ ¾
1 e¡i­ a i­ a a
g2ou t (r; a; ­ )= (1 + i­ a) (e ¡ ei­ r
)¡ +1 : (4.7)
2­ 2 i­ r r
The Helmholtz potential g ou t
= g1ou t
+ g2ou t
then yields
ei­ r
g ou t (r; a; ­ ) = (sin ­ a ¡ ­ a cos ­ a); (4.8)
r­ 3
again coinciding with the results of Mikata & Nemat-Nasser (1990). g ou t has the
form g ou t = g^(r; ­ )£ ~ s (a; ­ ), where ei­ r =(4º r) is the Green function of the Helmholtz
equation and £ ~ s (a; ­ ) = (4º =­ 3 )(sin ­ a ¡ ­ a cos ­ a) is the Fourier transform of the
characteristic function of the sphere (B 5). We shall see in the next subsection that
such a property is also true for the continuous ­ bre.
The corresponding Eshelby tensor of a spherical inclusion can be generated
straightforwardly from equations (4.3) and (4.8), respectively, by using equations
(2.17){(2.19). These expressions were already given by Mikata & Nemat-Nasser
(1990) and are therefore not repeated here.

(b) Continuous cylindrical ¯bre

We derive here the Helmholtz potential of the cylindrical continuous ­ bre as
a limiting
p case. In this case we have a = a1 = a2 , a3 ! 1, p ! » =a, where
» = x2 + y 2 . To perform this limiting case from the above integral representations
((3.18) and (3.19) onwards) it is convenient to use spherical coordinates, namely
(K^1; K
^ 2; K
^ 3 ) = (sin ³ cos ¬ ; sin ³ sin ¬ ; cos ³ ), so that s = a= sin ³ . Let us next con-
sider ­ rst the Helmholtz potential of the inside region.

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 873

Case I: p < 1 (r 2 S). In spherical coordinates, equation (3.18) assumes

asymptotically (a3 ! 1) the form
½ Z 2º Z º µ ¶ ¾
1 1 i­ a
g in (» ; a; ­ ) = 2 ei­ » cos ’ d’ ei­ a= s in ³ sin ³ 1 ¡ d³ ¡ 1 : (4.9)
­ 4º 0 0 sin ³
The ’ integral yields a zero-order Bessel function J0 (­ » ); thus we can write
1
g in (» ; a; ­ ) = 2
(J0 (­ » )I(­ a) ¡ 1); (4.10)
­
and by taking into account the symmetry of the sine function with respect to 12 º we
can write Z º µ ¶
i­ a= s in ³ i­ a
I(­ a) = e sin ³ 1 ¡ d³ : (4.11)
º =2 sin ³
By introducing the substitution cosh ¿ = 1= sin ³ , the above integral yields
Z 1
(1)
I(­ a) = ¡ i­ a ei­ a cos h ¿ cosh ¿ d¿ = 12 (º i­ a)H1 (­ a); (4.12)
0
(1)
where H1 denotes the Hankel function. Thus, we ­ nally obtain for (4.9) the asymp-
totic solution
1 (1)
g in (» ; a; ­ ) = 2 ( 12 (º i)­ aH1 (­ a)J0 (­ » ) ¡ 1) (4.13)
­
for inside the cylindrical ­ bre, coinciding with the expression obtained by employing
another approach in Levin & Michelitsch (2000), Levin et al. (2002) and Michelitsch
et al . (2002).
Case II: p > 1 (r 2 = S). In order to evaluate this case we rewrite equations
(3.22) and (3.23) in the form (3.24). Performing g2ou t over the area 1 + pK ^ ¢n<1
leads to
·Z º + ’0 Z ³ 0 (’) µ ¶
1 ­a
g ou t = g in ¡ Re d’ ei­ » cos ’
ei­ a= s in ³
1 ¡ i
º ­2 º º =2 sin ³
Z º + ’0 Z ³ 0 (’) ¸
£ sin ³ d³ ¡ d’ sin ³ d³ ; (4.14)
º º =2
¡1 ¡1
where ’0 = cos (a=» ) and ³ 0 (’) = sin (¡ a=(» cos ’)) are introduced. Again, with
cosh ¿ = 1= sin ³ , after partial integration equation (4.14) assumes the more conve-
nient form
·Z º + ’0 Z ¿ 0 (’)
ou t in ia i­ » cos ’
g =g + d’ e ei­ a cos h ¿ cosh ¿ d¿
º ­ º 0
Z º + ’0 Z ¿ 0 (’) ¸
¡i­ » cos ’ ¡i­ a cos h ¿
¡ d’ e e cosh ¿ d¿ ; (4.15)
º 0
¡1
where ¿ 0 (’) = cosh (¡ (» =a) cos ’). In Appendix D we evaluate these integrals
explicitly and arrive at
º ai (1)
g ou t (r; a; ­ ) = J1 (­ a)H0 (­ » ) = g^(» ; ­ )£ ~ s (­ ; a); (4.16)
2­

Proc. R. Soc. Lond. A (2003)

874 T. M. Michelitsch and others

coinciding with the expressions derived in Levin & Michelitsch (2000), Levin et al .
(1)
(2002) and Michelitsch et al. (2002). H0 and J1 denote the Hankel and Bessel
functions, respectively. In analogy to the result for a sphere (4.8) we observe that
(1)
g ou t is again the product of the Green function g^(» ; ­ ) = 14 iH0 (­ » ) of the two-
dimensional (2D) Helmholtz equationy and the Fourier transform of the characteristic
function of the continuous ­ bre
2º aJ1 (­ a)
£ ~ s (­ ; a) =
­
(see equation (B 7)).
The results g(» ; ­ ; a) (4.13) and (4.16) for the continuous ­ bre can also be obtained
by a 2D calculation (Levin & Michelitsch 2000; Levin et al. 2002; Michelitsch et al .
2002) which can be derived from (3.5) by using
Z 1
1
¯ (z) = eik3 z dk3 : (4.17)
2º ¡1
It can easily be shown that the g(» ; ­ ; a) of equations (4.13) and (4.16) ful­ ls the
2D Helmholtz equation
· 2 ¸
@ @2 2
+ 2 +­ g(» ; ­ ; a) + £ (1 ¡ p) = 0; (4.18)
@x2 @y
where £ (1 ¡ p) is the characteristic function of the circle with radius a (p = » =a).
The components of the Eshelby tensor can again be generated straightforwardly in
terms of spatial derivatives of equations (4.13) and (4.16), respectively, via equations
(2.17){(2.19).

5. Dynamic Eshelby tensor inside an ellipsoidal inclusion

This section is devoted to deriving integral expressions for the dynamic Eshelby
tensor for the inside region (p < 1) of the ellipsoidal inclusion. To this end we use
equation (2.19), which has the form
Silrs (r; a; !) = ¡ Ckjrs (Pijkl (r; a; !))(il) ; (5.1)
where we introduced a = (a1 ; a2 ; a3 ). With (3.18) we obtain P in the form
Pijkl (r; a; !)
Z ^iK^j · µ ^ kK
^l ¶
1 ^ K ^ 1 ^ r) ¯ K ^
=¡ d« (K) s2 (K) f (­ 2 ; ai ; K; kl ¡ s2 (K)
4º ^ j= 1
jK ai aj C66 ak al
^ ^ ¸
1 ^ r) Kk Kl s2 (K)
^
+ f (­ 1 ; ai ; K;
C11 ak al
(5.2)
with µ ^ i xi ¶
K
^ r) = (1 ¡
f (­ ; a; K; i­ s)e i­ s
cos ­ s (5.3)
ai
y (@ 2 =@x2 + @ 2 =@y2 + ¯ 2 ) g^(½; ¯) + ±(x) ±(y) = 0.

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 875

and
"s #¡1
^2
K ^2
K ^2
K
^ =
s(K) 1
+ 2
+ 3
:
a21 a22 a23

Equation (5.2) can be rewritten as

· (2) µ ¶¸
Dij ¯ kl 1 (1) 1 (2)
Pijkl (r; ai ; !) = ¡ + J ¡ J (5.4)
C66 C11 ijkl C66 ijkl

where (Di =: Dii , no sum over i)y

Z · ^2 ^ 22 ^ 32 ¸¡1
1 ^ ^ 2 ^ K1 K K
Di (­ ; a; r) = d« (K) (Ki f (­ ; a; K; r)) 2 + 2 + 2 (5.5)
4º a2i ^ j=
jK 1 a1 a2 a3
(s) (s)
with Di = Di (­ s ; a; r) and Jijrs = Jijrs (­ s ; a; r) (superscripts s = 1; 2 indicate
the frequencies of (2.13))z and
Z
1 ^ (K ^iK
^jK
^ rK
^ s f (­ ; a; K;
^ r))
Jijrs (­ ; a; r) = d« (K)
4º ai aj ar as jK^ j= 1
µ ^2 ^2 ^ 2 ¶¡2
K1 K K
£ + 2 + 23
2
: (5.6)
a21 a2 a3

In the static limit ­ ! 0 (f ! 1) the components of this totally symmet-

ric tensor are non-vanishing if only even powers of K ^ i in (5.6) occur, namely
Jij =: Jiijj = Jijij = ¢ ¢ ¢ . In the following we employ Voigt’s notation for the
components of the Eshelby tensor.{ Unlike the static case, the dynamic Eshelby
tensor has 36 non-vanishing components (Sij 6= Sji , i; j = 1; : : : ; 6). In the
static limit only 12 of these components are non-vanishing, namely fS11 ; S22 ; S33 ;
S12 ; S21 ; S13 ; S31 ; S23 ; S32 ; S44 ; S55 ; S66 g. Since we focus in x 5 b on the static limit,
we con­ ne ourselves here to the evaluation of these components. Using (5.1), we ­ nd
for these components of the dynamic Eshelby tensor the expressions
µ (2) (1) (2) ¶ µ (1) (1) (2) (2) ¶
D1 J J11 D1 ¡ J11 D1 ¡ J11
S11 = C11 + 11 ¡ + C12 ¡ (5.7)
C66 C11 C66 C11 C66
with its counterparts S22 , S33 , and
µ (2) (1) (2) ¶ µ (1) (2) ¶ µ (1) (2) ¶
D1 J J11 J12 J12 J13 J13
S12 = C12 + 11 ¡ + C11 ¡ + C12 ¡ (5.8)
C66 C11 C66 C11 C66 C11 C66

with the counterparts S21 ; S13 ; S31 ; S23 ; S32 , and

µ (1) (2) ¶
1 (2) (2) J12 J12
S66 = 2 (D1
+ D2 ) + 2C66 ¡ (5.9)
C11 C66
y Note that Di (¯ = 0) = Di , corresponding to the static case, equation (E 5).
(s)
z Jij rs (¯s = 0) = Jij rs , de¯ned in equation (5.28) below, corresponds to the static limiting case.
{ S11 = S1111 , S12 = S1122 , S66 = S1212 : : : .

Proc. R. Soc. Lond. A (2003)

876 T. M. Michelitsch and others

with the counterparts S44 ; S55 . We note that the dynamic Eshelby tensor is a spa-
tially non-uniform function of r, a and ­ and has the same (orthotropic) symmetry
properties as in the static case. With C11 = ¶ + 2· , ¶ = C12 , C66 = · and
E¸ E
¶ = ; · =
(1 + ¸ )(1 ¡ 2¸ ) 2(1 + ¸ )
and by using the relations
(s) (s) (s) (s)
Ji1 + Ji2 + Ji3 = Di ; (5.10)

we obtain
1 (2) (1) (2) (1) (2)
S11 = [2(1 ¡ ¸ )D1 + ¸ D1 ¡ J11 + (1 ¡ 2¸ )(J11 ¡ J11 )]: (5.11)
(1 ¡ ¸ )

The counterparts S22 , S33 are obtained from (5.11) by replacing subscripts (11)
by (22), (33), and using
1 (1) (2) (1) (2)
S12 = [¸ D1 ¡ J12 + (1 ¡ 2¸ )(J12 ¡ J12 )]: (5.12)
(1 ¡ ¸ )

The counterparts S21 , S13 , S31 , S23 , S32 are given by (5.12) when replacing the
subscripts (12) by (21), (13), (31), (23), (32) and taking into account the symmetry
(s) (s)
Jij = Jji . Finally, we ­ nd

(2) (2) 1 (2) (1) (2)

S66 = 12 (D1 + D2 ) + [¡ J12 + (1 ¡ 2¸ )(J12 ¡ J12 )]; (5.13)
(1 ¡ ¸ )
where the counterparts S44 , S55 are obtained by simultaneously replacing the sub-
scripts f(66); (12)g by f(44); (23)g, f(55); (13)g, respectively, in (5.13). In (5.11){
(5.13), only three types of integrals occur, namely (5.5),
Z µ ^2 ^2 ^ 2 ¶¡2
(s) 1 ^ ^ 4 ^ K1 K K
J11 (­ s ; a; r) = d« (K) (K1 f (­ s ; a; K; r)) + 2 + 23
2
;
4º a41 ^ j=
jK 1 a21 a2 a3
(5.14)
Z µ ^2 ^ 22 ^ 32 ¶¡2
(s) 1 ^ ^ 2 ^2 ^ K1 K K
J12 (­ s ; a; r) = d« (K) (K1 K2 f (­ s ; a; K; r)) + 2 + 2
4º a21 a22 ^ j=
jK 1 a21 a2 a3
(5.15)

and their cyclic counterparts. In general, integrals (5.5), (5.14) and (5.15) have to be
evaluated numerically. The remaining components of the dynamic Eshelby tensor can
be obtained straight-forwardly by using (5.1). In the static limit these components
are vanishing. The static limiting case ­ s ! 0 is easily obtained from expressions
(s) (s)
(5.11){(5.13) when we put lim­ ! 0 Di = Di and lim­ s ! 0 Jij = Jij . Integrals Di
and Jij are evaluated explicitly in x 5 b. Speci­ cally, it is explicitly shown there that
the static limiting case of expressions (5.11){(5.13) coincides with Eshelby’s (1957)
expressions.

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 877

(a) Spheroidal inclusion

Here we consider spheroidal inclusions, that is a1 = a2 = a; a3 . In this case the
(s) (s)
integrals Di and Jij can be reduced to one-dimensional integrals similar to elliptic
integrals. As in the previous subsection, we con­ ne our work here to the inside region
of the inclusion. We ­ nd the integral
" Z p
1
in 1 1 J0 (­ » u=(u + a2 ))
g = ¡ 1+ a
2 3
du
­ 2
0 (u + a23 )3=2
à s ! à s ! µ ¶#
u + a23 u + a23 ­ az
£ 1¡ i­ a exp i­ a cos p (5.16)
u + a2 u + a2 u + a2

by putting u = a23 tan2 ³ in cylindrical coordinates (x; y; z) = (» cos ’0 ; » sin ’0 ; z)

for the Helmholtz potential. Integrals Di and Jij take the form
p às !
Z 1 2 )=(u + a2 ))g 2
u expfi­ a( (u + a 3 u + a
D1 = 14 a3 1 ¡ i­ a 3
0 (u + a23 )3=2 (u + a2 ) u + a2
µ ¶· µ r ¶ µ r ¶ ¸
­ az u u
£ cos p J0 ­ » ¡ J 2 ­ » cos 2’ 0 du;
u + a2 u + a2 u + a2
(5.17)
p às !
Z 1
1 u expfi­ a( (u + a23 )=(u + a2 ))g u + a23
D 2 = 4 a3 1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 ) u + a2
µ ¶· µ r ¶ µ r ¶ ¸
­ az u u
£ cos p J0 ­ » + J2 ­ » cos 2’0 du;
u + a2 u + a2 u + a2
(5.18)
p às !
Z 1
1 2 expfi­ a( (u + a23 )=(u + a2 ))g u + a23
D 3 = 2 a a3 1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 ) u + a2
µ ¶ µ r ¶
­ az u
£ cos p J0 ­ » du (5.19)
u + a2 u + a2

and
p às !
2 Z 1 2 )=(u + a2 )g 2
1 u expfi­ a (u + a 3 u + a 3
J11 = 16 a3 1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 )2 u + a2
µ ¶· µ r ¶ µ r ¶
­ az u u
£ cos p J4 ­ » cos 4’ 0 ¡ 4J 2 ­ »
u + a2 u + a2 u + a2
µ r ¶¸
u
£ cos 2’0 + 3J0 ­ » du;
u + a2
(5.20)

Proc. R. Soc. Lond. A (2003)

878 T. M. Michelitsch and others
p às !
Z 1 2
1 u expfi­ a (u + a23 )=(u + a2 )g u + a23
J22 = a
16 3
1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 )2 u + a2
µ ¶· µ r ¶ µ r ¶
­ az u u
£ cos p J4 ­ » cos 4’0 + 4J2 ­ »
u + a2 u + a2 u + a2
µ r ¶¸
u
£ cos 2’0 + 3J0 ­ » du;
u + a2
(5.21)
p às !
Z 1
1 4 expfi­ a (u + a23 )=(u + a2 )g u + a23
J33 = 2
a a3 1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 )2 u + a2
µ ¶ µ r ¶
­ az u
£ cos p J0 ­ » du; (5.22)
u + a2 u + a2
p às !
Z 1
1 u2 expfi­ a (u + a23 )=(u + a2 )g u + a23
J12 = a
16 3
1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 )2 u + a2
µ ¶· µ r ¶ µ r ¶ ¸
­ az u u
£ cos p J 0 ­ » ¡ J 4 ­ » cos 4’ 0 du;
u + a2 u + a2 u + a2
(5.23)
p às !
Z 1 2 )=(u + a2 )g 2
1 2 u expfi­ a (u + a 3 u + a 3
J13 = 4 a3 a 1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 )2 u + a2
µ ¶· µ r ¶ µ r ¶ ¸
­ az u u
£ cos p J0 ­ » ¡ J2 ­ » cos 2’0 du;
u + a2 u + a2 u + a2
(5.24)
p às !
Z 1 2 )=(u + a2 )g
1 2 u expfi­ a (u + a 3 u + a23
J23 = 4 a3 a 1 ¡ i­ a
0 (u + a23 )3=2 (u + a2 )2 u + a2
µ ¶· µ r ¶ µ r ¶ ¸
­ az u u
£ cos p J0 ­ » + J 2 ­ » cos 2’0 du:
u + a2 u + a2 u + a2
(5.25)

We observe that the symmetries D1 = D2 and J11 = J22 , J13 = J23 are absent in
(5.17){(5.25). The above integrals determine completely the full set of those compo-
nents of the Eshelby tensor which are non-vanishing in the static limit for a spheroidal
inclusion.

(b) The static limiting case

This subsection is devoted to the static limit of the dynamic Eshelby tensor for
the inside region of the inclusion for an arbitrary ellipsoid with semi-axes ai . By
using the abbreviations introduced in Appendix E ((E 4) and (E 5)) we can ­ rst of

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 879

all write, for the potential function of (2.12) in the limit ­ ! 0,

µ ¶
R0 (r; ai ) 1 1 1
G0ij (r; ai ) = ¯ ij + ¡ (¯ ij (C ¡ 12 (a2i )Di ) ¡ Jijrs xr xs ); (5.26)
C66 2 C11 C66
where no sum over index i is performed and
R0 = 12 [(a21 ¡ x2 )D1 + (a22 ¡ y 2 )D2 + (a23 ¡ z 2 )D 3 ]:
Here the integrals
Z µ ^2 ^2 ^ 2 ¶¡1
1 ^ 2 ^ K1 K K
Di (a1 ; a2 ; a3 ) = (Ki d« (K)) + 2 + 23
2
; (5.27)
4º a2i ^
^
jK j= 1 a 2
1 a2 a3
Z µ ^2 ^2 ^ 2 ¶¡2
1 ^ iK
^j K
^ rK^ s d« (K))
^ K1 K 2 K 3
Jijrs (a1 ; a2 ; a3 ) = (K + +
4º ai aj ar as jK^ j= 1 a21 a22 a23
(5.28)
occur corresponding to the static limit of (5.5), (5.14) and (5.15). They are explic-
itly evaluated in Appendix E and can be expressed in terms of elliptic integrals
(Appendix E a). The above tensor Jijrs is totally symmetric and has non-zero com-
ponents if only even powers of K ^ i occur (Jij =: Jiijj = Jijij = ¢ ¢ ¢ ). This tensor has
the property (5.10), namely Ji1 + Ji2 + Ji3 = Di . We ­ nd
Z µ ¶
a1 a2 a3 1 du 1 a22 a23
J11 = p +
4 0 (u + a21 )(u + a22 )(u + a23 ) (u + a21 ) u + a22 u + a23
½ ¾
1 a22 a23
= (D 1 ¡ D 2 ) + (D 1 ¡ D 3 )
2 a22 ¡ a21 a23 ¡ a21
and
Z
a1 a2 a3 1 du u
J12 = p 2 )(u + a2 )
4 0
2 2 2
(u + a1 )(u + a2 )(u + a3 ) (u + a 1 2
½ 2 2 ¾
1 a + a2
= D1 + D2 + 12 (D1 ¡ D2 ) ;
4 a1 ¡ a22
together with their cyclic counterparts. The Di are fully expressed in terms of elliptic
integrals and de­ ned by (E 24) and (E 25) in Appendix E a.
From equation (5.2) and taking into account in (5.3) that f ! 1 + O(­ 2 ), we ­ nd
0
for the static part Pijkl = @i @j G0kl (r; a) the expression
Z ^ ^
0 1 ^ Ki Kj s2 (K) ^
Pijkl (r; ! ! 0) = Pijkl =¡ d« (K)
4º jK^ j= 1 ai aj
· µ ^kK ^l ¶ ^kK^l ¸
1 K ^ 1 K ^
£ ¯ kl ¡ s2 (K) + s2 (K) (5.29)
C66 ak al C11 ak al
or (equation (5.4) for ! ! 0)
· µ ¶ ¸
0 Di ¯ ij ¯ kl 1 1
Pijkl =¡ + ¡ Jijkl (5.30)
C66 C11 C66

Proc. R. Soc. Lond. A (2003)

880 T. M. Michelitsch and others

(no sum over i). The static Eshelby tensor is de­ ned by the limit ­ ! 0 of (5.1),
0 0 0
where the only non-vanishing components are S1111 ; S1122 ; S1212 : : : with their cyclic
counterparts. In the following we employ Voigt’s notation for the components of the
Eshelby tensor and skip the superscript zero.y When we use C11 = ¶ + 2· , ¶ = C12 ,
C66 = · with ¶ = E¸ =[(1 + ¸ )(1 ¡ 2¸ )] and · = E=[2(1 + ¸ )] we obtain
µ ¶
C11 D1 1 1
S11 = + ¡ (C11 J11 + C12 (J12 + J13 ))
C66 C11 C66
1
= [(2 ¡ ¸ )D1 ¡ J11 ]
(1 ¡ ¸ )
· ¸
1 a21 a21
= 1 + (1 ¡ 2¸ )D1 + 2 (D1 ¡ D2 ) + 2 (D1 ¡ D3 ) ; (5.31)
2(1 ¡ ¸ ) a1 ¡ a22 a1 ¡ a23
µ ¶
C12 D1 1 1 1
S12 = + ¡ (C11 J12 + C12 (J11 + J13 )) = (¸ D1 ¡ J12 )
C66 C11 C66 (1 ¡ ¸ )
· ¸
1 a2
= ¡ (1 ¡ 2¸ )D1 + 2 2 2 (D2 ¡ D1 ) ; (5.32)
2(1 ¡ ¸ ) a1 ¡ a2
µ ¶
D1 + D2 1 1 D1 + D2 J12
S66 = + 2C66 J12 ¡ = ¡
2 C11 C66 2 (1 ¡ ¸ )
· 2 2
¸
1 a + a2
= (1 ¡ 2¸ )(D1 + D2 ) + 12 (D2 ¡ D1 ) ; (5.33)
4(1 ¡ ¸ ) a1 ¡ a22
together with their counterparts. Expressions (5.31), (5.32) and (5.33) coincide with
Eshelby’s classical results of 1957 (Eshelby 1957, eqns (3.7){(3.14)).

6. Conclusions
The dynamic Eshelby tensor of a harmonically transforming ellipsoidal inclusion
was derived in terms of compact integral formulations for the inside region. Closed-
form representations were given for several special cases (sphere, continuous ­ bre)
and one-dimensional quadratures for the case of spheroidal inclusions. Unlike in
the static case, the dynamic Eshelby tensor is non-uniform inside an ellipsoidal
inclusion and has 36 non-vanishing components. The crucial part of the deriva-
tion was to ­ nd an appropriate approach for the Helmholtz potential of an ellip-
soidal source distribution of unit density. The formulation is in close analogy to
Eshelby’s static 1957 approach. In the static limit our approach yields Eshelby’s
(1957) classical results straightforwardly. Moreover, our approach also allows us
to formulate the dynamic Eshelby tensor for the outside region of an ellipsoidal
inclusion.
The approach may be useful in application to a wide range of dynamic prob-
lems in the physics of solids. For instance the dynamic Eshelby tensor is a
key quantity for the determination of e¬ective dynamic characteristics in micro-
inhomogeneous materials, such as dynamic moduli, mean ­ eld wave speed, disper-
sion e¬ects, and attenuation factor due to the random arrangement of inhomo-
geneities.
y S11 = S1111 , S12 = S1122 , S66 = S1212 ; : : : .

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 881

Financial support of the Deutsche Forschungsgemeinschaft (DFG), project no. GA 732/1-1, is

gratefully acknowledged.

Appendix A.
Here we derive the integral transformation (3.17). To this end we consider an integral
over the unit sphere Z
I= ^ f (k):
d« (k) ^ (A 1)
^ 1
jkj=

Let us assume that f is a homogeneous function of degree zero in k with f (k) =

^ depending only on k
f (k) ^ = k=ky and take into account the fact that

rk ¢ [kf ^ = 2 f (k):
^ (k)] ^ (A 2)
k
We can transform (A 1) into a volume integral
Z
2 ^
I= d3 k f (k): (A 3)
k2 <1 k
Let us now introduce new coordinates according to ki = Ki =ai with
"s #¡1 q
^i
K K^2 K
k^i = s(K)
^ ; ^ =
s(K) i
2
= a2i k^i2 ; k = :
ai ai s
Then (A 3) assumes the form
Z ^ µ
d3 K 2s(K) ^i ¶
K
I= ^
f s(K) : (A 4)
Ki2 =a2i <1 a1 a2 a3 K ai
This integral can be rewritten as
Z µ ^ i ¶ Z s(K^ )
2 ^ ^ ^ K
I= d« (K) s(K)f s(K) K dK (A 5)
a1 a2 a3 jK^ j= 1 ai 0
or µ
Z
1
Z ^i ¶
K
^ f (k)
d« (k) ^ = ^ 3 ^ ^
d« (K) s (K)f s(K) : (A 6)
^ j= 1
jk a1 a2 a3 ^ j=
jK 1 ai
Using (A 6), we ­ nd the following useful transformations:
f = s¡3 :
Z ^ Z
d« (k) 1 ^ = 4º
= d« (K) ; (A 7)
2 ^ 2 2 ^ 2 2 ^ 2
jk j= 1 (a1 k1 + a2 k2 + a3 k3 )
^ 3=2 a a a ^
1 2 3 jK j= 1 a1 a2 a3

f = s¡1 :
Z ^ Z µ ^2 ^2 ^ 2 ¶¡1
d« (k) 1 ^ K1 K K
q = d« (K) + 2 + 23
2
:
^ j= 1 a1 a2 a3 ^ j= 1 a21 a2 a3
jk a21 k^12 + a22 k^22 + a23 k^32 jK

(A 8)
^ de¯ned on the unit sphere
y For the following steps it is important to note that any function f(k)
can be continued within the unit sphere by a function f (k=k) which is independent on k.

Proc. R. Soc. Lond. A (2003)

882 T. M. Michelitsch and others

Appendix B.
Here we calculate the Fourier transform of the characteristic function £ (1 ¡ p) of
the ellipsoid (3.1) which is de­ ned by (3.8):
Z
~
£ (k; ai ) = e¡ik¢r £ s (r) d3 r: (B 1)

For our convenience we introduce the following parametrization of the ellipsoid

r = p(a1 n1 ; a2 n2 ; a3 n3 ); (B 2)
where ni characterizes the unit sphere with n21 + n22 + n23 = 1. Taking into account
the fact that d3 r = a1 a2 a3 p2 dp d« (n) and r ¢ k = w ¢ n with wi = pki ai , which is a
constant vector in the r-space so that
q
w ¢ n = pai ki ni = pk (a2i k^i2 ) cos ® :
| {z } | {z }
jw j (w =jw j)¢n

Since w = wi = pki ai denotes a ­ xed axis in the r-space, we can choose the spatial-
coordinate system so that its 3-axis is parallel to w; thus d« = sin ® d® d’ and
Z 1 Z 2º Z º
~
£ (k; ai ) = a1 a2 a3 2
£ (1 ¡ p)p dp sin ® d® d’ e¡ikps cos ® ; (B 3)
0 0 0
q
where s = s(k^i ; ai ) = (a2i k^i2 ) and with u = cos ® ,y leading to (3.9):
Z 1 Z 1
£ ~ (k; ai ) = 2º a1 a2 a3 p2 dp e¡ikpsu du
0 ¡1
µ ¶
4º a1 a2 a3 4º a1 a2 a3 d sin(ks)
= (sin ks ¡ ks cos ks) = ¡ : (B 4)
k 3 s3 k3 s ds s
In the case of a sphere with radius ai = a and s = a (B 4) becomes
4º
£ ~ (k; a) = (sin ka ¡ ka cos ka): (B 5)
k3
It is also worthwhile to consider the limiting case of the continuous ­ bre, that is
a3 ! 1 and a1 = a2 = a. Because of (4.17), the Fourier transform (B 4) must then
assume the asymptotic form
£ ~ (k; ai ) = 2º ¯ (k3 )£ ~ 2D (k;
· a); (B 6)
· a) is the corresponding 2D Fourier transform of the characteristic func-
where £ 2D (k;
tion of a circle with radius a and yields
Z ·
· a) = 1
£ ~ 2D (k;
·
e¡ik¢·r £ (1 ¡ p) d2 r· =
2º aJ 1 (ka)
; (B 7)
(2º ) 2 ·
k
where p = » =a and r· and k· indicate plane vectors (k1 ; k2 ) and (x; y), respectively. We
con­ rm this relation directly by putting k3 = (ak=a3 ) cosh ¿ and integrating (B 4) in
~ ai ) ! 4
y Note that £(0; 3 ¼a1 a2 a3 (which equals the volume of an ellipsoid) when k ! 0.

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 883

a small interval k3 2 [¡ ° ; ° ]. For a3 ! 1 we then obtain

Z ° ½Z 1 ¾
1 ~ 4 d¿ · · cosh ¿ )
I= £ (k; ai ) dk3 = · 2 Im 2 eika cos h ¿ (1 ¡ ika
2º ¡° k 0 cosh ¿
½Z 1 ¾
4 · 2º ia (1) ·
= · 2 Im d¿ eika cos h ¿ cosh ¿ = Im · H1 (ka)
k 0 k
2º a · = £ ~ 2D (k;
· a);
= · J1 (ka) (B 8)
k
con­ rming (B 6).

Appendix C.
Here we derive integral (3.12), which is de­ ned by
Z 1 ^ ^
1 eik(k ¢r + s¡ct) ¡ eik(k¢r + s+ ct)
I = 2 Re dk: (C 1)
0 k2
Because of
Z 1
1
= ye¡ky dy; (C 2)
k2 0
Z 1 Z 1
^ ^
I = 12 Re y dy [e¡k(y¡i(k¢r + s¡ct)) ¡ e¡k(y¡i(k ¢r + ct+ s))
] dk (C 3)
0 0

and
Z 1 · ¸
1 1 1
I= 2
Re y dy ¡ ; (C 4)
0 y¡ iA y¡ iB
where
A=k ^ ¢ r + s ¡ ct; B=k ^ ¢ r + s + ct: (C 5)
Further evaluation yields
Z 1 · ¸ Z 1 · ¸
1 1 A B
I(r; a; t) = 12 Re y dy ¡ = 12 Re i dy ¡
0 y ¡ iA y ¡ iB 0 y ¡ iA y ¡ iB

= 1
2 Re(iA ln(y ¡ iA) ¡ iB ln(y ¡ iB))j10

= 14 º (jBj ¡ ^ ¢ r + s + ctj ¡
jAj) = 14 º (jk ^¢r+s¡
jk ctj): (C 6)
1 1
Here we have used ln(¡ iA) = ln(expf¡ 2 iº
sgn(A)gjAj) = ln jAj ¡ 2 iº
sgn(A).

Appendix D.
In order to derive equation (4.16) from our approach, we consider the integrals of
equation (4.15) explicitly, namely
· Z º + ’0 Z ¿ 0 (’)
ia
J = d’ ei­ » cos ’ ei­ a cos h ¿ cosh ¿ d¿
º ­ º 0
Z º + ’0 Z ¿ 0 (’) ¸
¡ d’ e¡i­ » cos ’ e¡i­ a cos h ¿ cosh ¿ d¿ ; (D 1)
º 0

Proc. R. Soc. Lond. A (2003)

884 T. M. Michelitsch and others

where cos ’0 = a=» (» > a), cosh ¿ 0 (’) = ¡ (» =a) cos ’ > 1 for ’ 2 [º ¡ ’0 ; º + ’0 ].
In Michelitsch et al . (2002, eqns (A.9) in Michelitsch et al . (2002)) it is shown that
g can be represented in the form
º ai (1)
g ou t (» ; a; ­ ) = J1 (­ a)H0 (­ » )
2­
Z 1 Z 2º Z 2º
a i­ ½
= e d½ Re d’ 1 d’2 cos ’2
(2º )2 0 0 0
½ ¾
» ½
£ ln cos ’2 ¡ cos ’1 + + i° ; (D 2)
a a

where ° ! 0+ is an in­ nitesimal positive constant. The last integral yields g in and
g ou t for » < a and » > a, respectively. Since we are now focused on g ou t , let » > a.
Equation (D 2) contains the integral
Z 2º
I0 = Re d’ cos ’ lnfcos ’ + A(½ ; ’1 )g; (D 3)
0

where A = ¡ (» =a) cos ’1 + ½ =a + i° . Since » > a, all three cases A = cosh ¿ > 1,
¡ 1 < A = cos Á < 1, and A = ¡ cosh ¿ < ¡ 1 occur and (D 3) yields I0 = 2º e¡¿ ,
I0 = 2º cos Á and I0 = ¡ 2º e¡¿ , respectively. Hence we can summarize these cases
to

I0 = 2º £ (’; ¡ ’0 ; ’0 )
p
£ f£ (½ ; 0; ½ 2 )(A + A2 ¡ 1) + £ (½ ; ½ 2 ; ½ 1 )A
p
+ £ (½ ; ½ 1 ; 1)(A ¡ A2 ¡ 1)g
+ 2º (£ (’; ’0 ; º ¡ ’0 ) + £ (’; º + ’0 ; 2º ))
p
£ [A£ (½ ; 0; ½ 1 ) + (A ¡ A2 ¡ 1)£ (½ ; ½ 1 ; 1)]
p
+ 2º (A ¡ A2 ¡ 1)£ (’; º ¡ ’0 ; º + ’0 ); (D 4)

where A(½ 1 ) = 1, ½ 1 = a + » cos ’ and A(½ 2 ) = ¡ 1, ½ 2 = ¡ a + » cos ’. In (D 4) we

introduce the characteristic function £ (x; x1 ; x2 ) which is de­ ned by £ (x; x1 ; x2 ) = 1
for x 2 [x1 ; x2 ] and £ (x; x1 ; x2 ) = 0 for x 2= [x1 ; x2 ]. Thus
Z 2º Z 1
ou t a
g (» ; a; ­ ) = d’ ei­ ½ I0 (’; ½ ) d½ : (D 5)
(2º )2 0 0

Taking into account (D 4) we can split this integral into

g ou t (» ; a; ­ )
½ Z ’0 ·Z ½ 2 (’)
a
= ¡ d’ ei­ ½ e¡¿ (½ ;’)
d½
2º ¡’0 0
Z ½ 1 (’)
Z 1 ¸
i­ ½ i­ ½ ¡¿ (½ ;’)
+ e A(½ ; ’) d½ + e e d½
½ 2 (’) ½ 1 (’)

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 885
Z º ¡’0 · Z ½ 1 (’) Z 1 ¸
+ d’ ei­ ½ A(½ ; ’) d½ + ei­ ½ e¡¿ (½ ;’) d½
’0 0 ½ 1 (’)
Z 2º ¡’0 ·Z ½ 1 (’)
Z 1 ¸
i­ ½ i­ ½ ¡¿ (½ ;’)
+ d’ e A(½ ; ’) d½ + e e d½
º + ’0 0 ½ 1 (’)
Z º + ’0 Z 1 ¾
+ d’ ei­ ½ e¡¿ (½ ;’)
d½ :
º ¡’0 0
(D 6)
We have to observe the following cases, namely ½ > ½ 1 . We then put
½ = a cosh ¿ + » cos ’;
where ¿ (½ 1 ) = 0. The second case is ½ 6 ½ 2 . Then we put ½ = ¡ a cosh ¿ + » cos ’,
where ¿ (½ 2 ) = 0. In the last integral of (D 6) we observe that cosh ¿ (½ = 0) =
¡ (» =a) cos ’. In the ­ rst integral of (D 6) we put cosh ¿ (½ = 0) = +(» =a) cos ’y.
Then the integrals (D 6) can be summarized, after some rearrangements, as
½ Z 2º Z 1
ou t a (½ ¡ » cos ’) i­ ½
g (» ; a; ­ ) = d’ e d½
2º 0 0 a
Z 2º Z 1
¡ d’ sinh ¿ (’; ½ )ei­ ½ d½
0 ½ 1 (’)
Z ’0 Z ½ 2 (’)

+ d’ sinh ¿ (’; t)ei­ ½ d½

¡’0 0
Z º + ’0 Z ½ 1 (’)
¾
i­ ½
¡ d’ sinh ¿ (’; t)e d½ (D 7)
º ¡’0 0
or
g ou t (» ; a; ­ )
½ Z 2º Z 1 ¾
1 2 2 1 i­ » cos ’ 2 i­ a cos h ¿
= 2 ¡ 1¡ ­ a e d’ sinh ¿ e d¿
­ 2º 0 0
½Z ’0 Z cos h ¡ 1
((» =a) cos ’)
a2
+ e i­ » cos ’
d’ e¡i­ a cos h ¿
sinh2 ¿ d¿
º 0 0
Z º + ’0 Z cos h ¡ 1
(¡(» =a) cos ’) ¾
i­ » cos ’ i­ a cos h ¿ 2
+ e d’ e sinh ¿ d¿ :
º 0
(D 8)
The ­ rst line of this equation assumes g in (equation (4.13)). Therefore we obtain
for (D 1) of (4.15) by using (D 2) after partial integrations
½ Z º + ’0 Z cos h ¡ 1 (¡(» =a) cos ’)
ia i­ » cos ’
J = e d’ ei­ a cos h ¿ cosh ¿ d¿
º ­ º 0
Z º + ’0 Z cos h ¡ 1 (¡(» =a) cos ’) ¾
¡i­ » cos ’ ¡i­ a cos h ¿
¡ e d’ e cosh ¿ d¿
º 0
y Corresponding to A = cosh Á, ¿ > ¿1 and A = ¡ cosh Á, ¿ < ¿2 , where A = (¿ ¡ ½ cos ’)=a.

Proc. R. Soc. Lond. A (2003)

886 T. M. Michelitsch and others

= g ou t (» ; a; ­ ) ¡ g in (» ; a; ­ )
1 (1) (1)
= [1º
2 2
ia­ (J1 (­ a)H0 (­ » ) ¡ H1 (­ a)J0 (­ » )) + 1]; (D 9)
­
where ’0 = cos ¡1 (a=» ) (» > a). We observe that J is real. This is due to the fact
that Im g in = Im g ou t . Moreover, we observe that J is vanishing for » 6 a. Then
expression (D 9)3 holds only for » = a and yields the well-known result for the
Wronskian (e.g. Abramovitz & Stegun 1964)
(1) (1) 2
H1 (¹ )J0 (¹ ) ¡ H0 (¹ )J1 (¹ ) = : (D 10)
º i¹

Appendix E.
Here we consider the static limit of the Helmholtz potential of (3.18) (inside region)
which we denote as g0 (r; ai ). Generally the static limit of the Helmholtz potential is
the Newtonian potential, which ful­ ls the Laplace equation
¢g0 (r; ai ) + £ (1 ¡ p) = 0; (E 1)
where £ (1 ¡ p) = £ s (r) is the characteristic function of the ellipsoid and can be
represented by the integral
Z
g0 (r; ai ) = g^0 (jr ¡ r0 j)£ s (r 0 ) d3 r0 : (E 2)

Here g^0 (r) = 1=(4º r) denotes the Green function of the Laplace operator.
Let us consider equation (3.18) (inside the ellipsoid) and obtain by expanding the
integrand up to powers ­ 2
g0in (r; ai ) = C(ai ) ¡ 1
D (a )x x ;
2 ij i i j
(E 3)
where · ^2
1
Z
K1 K^2 K^ 2 ¸¡1
C(a1 ; a2 ; a3 ) = ^
d« (K) + 2 + 23
2
: (E 4)
8º ^
^
jKj= 1 a21 a2 a3
Tensor Dij = ¯ ij D i
(no sum with respect to i) is diagonal with
Z · ^2 ^ 22 ^ 32 ¸¡1
1 ^ 2 ^ K1 K K
Di (a1 ; a2 ; a3 ) = Ki d« (K) + 2 + 2 : (E 5)
4º a2i jKj=
^
^ 1 a21 a2 a3

It was to be expected that g0 of (E 3) is a quadratic function inside the ellipsoid.

This property is a consequence of the Eshelby (1957) theorem, which only holds in
statics. With
D1 + D2 + D3 = 1; (E 6)
it is immediately con­ rmed that g0in ful­ ls Laplace equation (E 1) (p < 1) inside
the ellipsoid. When we furthermore use the relation (because of K ^ 12 + K
^ 22 + K
^ 33 = 1)
between the integrals Di (a1 ; a2 ; a3 ) and C(a1 ; a2 ; a3 ) according to
1 2
2 (a1 D1
+ a22 D2 + a23 D3 ) = C: (E 7)

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 887

Then relation (E 3) can be written in the form

g in (r; ai ) = 12 (a21 ¡ x2 )D1 + 12 (a22 ¡ y 2 )D 2 + 12 (a23 ¡ z 2 )D3 ; (E 8)

which corresponds to the Newtonian potential of an ellipsoid in the inside region

given by Eshelby (1957, eqn (3.9)). The integrals Di can be expressed by elliptic
integrals F and E of the ­ rst and second kinds as given in Eshelby’s paper (Eshelby
1957, eqn (3.9); see also Appendix E a).
For a sphere with radius ai = a we obtain C(a) = a2 =2 and Di (a) = D(a) = 13 in
accordance with the static limit of expression (4.3).
To evaluate Di for the general elliptic case we use the spherical coordinates
(K^ 1; K
^ 2; K
^ 3 ) = (sin ³ cos ’; sin ³ sin ’; cos ³ ) and obtain
Z º Z 2º
1 2 d’
D3 = 2
cos ³ sin ³ d³ ; (E 9)
4º a3 0 0 A(³ ) ¡ B(³ ) cos 2’
where
µ ¶ µ ¶
1 2 1 1 cos 2 ³ 1 2 1 1
A= 2
sin ³ 2
+ 2 + ; B= sin ³ ¡ : (E 10)
a1 a2 a23 2 a22 a21
With a1 > a2 > a3 it is guaranteed that A > B > 0 so that cosh ¿ = A=B is a
possible transformation. The ’ integration then yields
Z 2º
d’ 2º 2º
= = p
0 A(³ ) ¡ B(³ ) cos 2’ B sinh ¿ A ¡ B2
2

2º a23
= p ; (E 11)
cos2 ³ (1 + (a23 =a21 ) tan2 ³ )(1 + (a23 =a22 ) tan2 ³ )
and, with (E 11), the integral (E 9) assumes that
Z º =2 ·µ ¶µ ¶¸¡1=2
a23 a23
D3 = sin ³ d³ 1+ tan2 ³ 1+ tan2 ³ : (E 12)
0 a21 a22
By putting u = a23 tan2 ³ we ­ nally arrive aty
Z 1
1 du
D i = 2 a1 a2 a 3 p ; (E 13)
0 (u + ai ) (u + a21 )(u + a22 )(u + a23 )
2

coinciding with Eshelby’s standard-form expression (Eshelby 1957, eqn (3.8)).z

Because
a23
= u + a23
cos2 ³
we ­ nally obtain
Z 1
1 du
C= 4 a1 a2 a3
p : (E 14)
0 (u + a21 )(u + a22 )(u + a23 )

y By cyclic permutation of subscripts 1, 2, 3 we obtain the corresponding expressions for all Di .

z Our D1 corresponds to (1=4¼)Ia in (3.8) of Eshelby’s (1957) paper.

Proc. R. Soc. Lond. A (2003)

888 T. M. Michelitsch and others

(a) Relation to elliptic integrals

Let us consider here brie®y how Di and C can be related to the elliptic integrals F
and E of the ­ rst and second kind, respectively. They are de­ ned as (e.g. Abramovitz
& Stegun 1964)
Z ’
d³
F (’; k) = p (E 15)
0 1 ¡ k 2 sin2 ³
and
Z ’ p
E(’; k) = 1¡ k 2 sin2 ³ d³ ; (E 16)
0

where ’ and k (0 < k < 1) are called their amplitude and modulus, respectively. By
putting u = 1= sin2 ³ ¡ 1= sin2 ’, integral F assumes the form
Z 1
1 du
F (’; k) = p (E 17)
2 0 (u + ¹ 1 )(u + ¹ 22 )(u + ¹ 32 )
2

and
Z 1
1 (u + ¹ 22 ) du
E(’; k) = p ; (E 18)
2 0 (u + ¹ 12 ) (u + ¹ 12 )(u + ¹ 22 )(u + ¹ 32 )
where
2 1 2 1 1
¹ 1 = ; ¹ 2 = ¡ k2 ; ¹ 2
3 = ¡ 1 = cot2 ’: (E 19)
sin2 ’ sin2 ’ sin2 ’
2 2
Because 0 < k < 1, we observe that ¹ 1 >¹ 2 > ¹ 32 . The coe¯ cients ¹ i ful­ l
2 2 2 2
¹ 1 ¡ ¹ 3 = 1; ¹ 1 ¡ ¹ 2 = k2: (E 20)

We can map the parameters ¹ i onto the parameters p ai by ¹ i = ¶ ai with the only
restriction a1 > a2 > a3 . The multiplier ¶ = 1= a21 ¡ a23 is determined by (E 20)
being uniform for all i. p The modulus k is determined by the second equation of
(E 20)pand yields k = (a21 ¡ a22 )=(a21 ¡ a23 ). From (E 19) it follows that ’ =
¡1
sin 1 ¡ (a3 )=(a1 ). By replacing u with ¶ 2 u in (E 17) and (E 18), F and E are
2 2

expressed in terms of the new parameters a1 > a2 > a3 , namely

q Z 1
du
F (’; k) = 12 a21 ¡ a23 p (E 21)
0 (u + a1 )(u + a22 )(u + a23 )
2

and, taking into account

2
u+¹ 2 k2
2
= 1¡ 2;
u+¹ 1 u+¹ 1
Z 1
1 2 du
E(’; k) = F (’; k) ¡ 2 (a1 ¡ a23 )3=2 k 2 p :
0 (u + a21 ) (u + a21 )(u + a22 )(u + a23 )
(E 22)

Proc. R. Soc. Lond. A (2003)

 Dynamic Eshelby tensor for ellipsoidal inclusions 889

With equations (E 14) and (E 13) we arrive at

a1 a2 a3
C= p 2 F (’; k) (E 23)
2 a1 ¡ a23
and
a1 a2 a3
D1 = p [F (’; k) ¡ E(’; k)]: (E 24)
(a21 ¡ a22 ) a21 ¡ a23
Equations (E 23) and (E 24) correspond to Eshelby’s expressions (Eshelby 1957,
eqns (3.8) and (3.9)).
From equations (E 7) and (E 6) we obtain D2 and D3 . Together with (E 23) and
(E 24) we ­ nally obtain
1
D3 = [2C ¡ a22 + D1 (a22 ¡ a21 )] (E 25)
a23 ¡ a22
or
· p ¸
a1 a2 a3 a2 a21 ¡ a23
D3 = p ¡ E(’; k) (E 26)
(a22 ¡ a23 ) a21 ¡ a23 a1 a3

and D2 = 1 ¡ D1 ¡ D3 . Equation (E 26) is again in accordance with eqn (3.9) 2 of

Eshelby’s (1957) paper.

References
Abramovitz, M. & Stegun, I. A. 1964 Handbook of mathematical functions, p. 375. New York:
Dover.
Eshelby, J. D. 1957 The determination of the elastic ¯eld of an ellipsoidal inclusion, and related
problems. Proc. R. Soc. Lond. A 241, 376{396.
Fu, L. S. 1982 Mechanics aspects of NDE by sound and ultrasound. Appl. Mech. Rev. 35,
1047{1057.
Fu, L. S. & Mura, T. 1982 Volume integrals associated with the inhomogeneous Helmholtz
equation. Wave Motion 4, 141{149.
Levin, V. M. & Michelitsch, T. 2001 Scattering of accoustoelectric waves on a cylindrical inho-
mogeneity in the transversely isotropic piezoelectric medium. Proc. SPIE 4333, 279{290.
Levin, V. M., Michelitsch, T. & Sevostianov, I. 2000 Spheroidal inhomogeneity in a transversely
isotropic piezoelectric medium. Arch. Appl. Mech. 70, 673{693.
Levin, V. M., Michelitsch, T. M. & Gao, H. 2002 Propagation of electroacoustic waves in the
transversely isotropic piezoelectric medium reinforced by randomly distributed cylindrical
inhomogeneities. Int. J. Solids Struct. 39, 5013{5051.
Marshall, D. B. & James, M. R. 1986 Reversible stress induced martensitic transformation in
ZrO2 . J. Am. Ceram. Soc. 69, 215{217.
Michelitsch, T. & Levin, V. M. 2000 Inclusions and inhomogeneities in electroelastic media with
hexagonal symmetry. Eur. Phys. J. B 14, 527{533.
Michelitsch, T. M., Levin, V. M. & Gao, H. 2002 Dynamic Green’ s functions of a quasi-plane
piezoelectric medium with inclusion. Proc. R. Soc. Lond. A 458, 2393{2415.
Mikata, Y. 2001 Explicit determination of piezoelectric Eshelby tensors for a spheroidal inclusion.
Int. J. Solids Struct. 38, 7045{7063.
Mikata, Y. & Nemat-Nasser, S. 1990 Elastic ¯eld due to a dynamically transforming spherical
inclusion. J. Appl. Mech. 57, 845{849.

Proc. R. Soc. Lond. A (2003)

890 T. M. Michelitsch and others

Mikata, Y. & Nemat-Nasser, S. 1991 Interaction of a harmonic wave with a dynamically trans-
forming inhomogeneity. J. Appl. Phys. 70, 2071{2078.
Miklowitz, J. & Achenbach, J. (eds) 1978 Modern problems in elastic wave propagation. Wiley.
Pao, Y. H. 1978 Elastic waves and non-destructive testing of materials. Applied Mechanics
Division, vol. 29. New York: American Society of Mechanical Engineers.

Proc. R. Soc. Lond. A (2003)