See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/243524096

On energy definition in electromagnetism: An analogy with viscoelasticity

Article  in  The Journal of the Acoustical Society of America · February 1999

DOI: 10.1121/1.426253

CITATIONS READS
13 169

1 author:

José M Carcione
OGS Istituto Nazionale di Oceanografia e di Geofisica Sperimentale
476 PUBLICATIONS   7,950 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wave dispersion and attenuation View project

Seismic Modeling View project

All content following this page was uploaded by José M Carcione on 21 January 2014.

The user has requested enhancement of the downloaded file.

On energy definition in electromagnetism: An analogy
with viscoelasticity
J. M. Carcionea)
Osservatorio Geofisico Sperimentale, P.O. Box 2011 Opicina, 34016 Trieste, Italy

~Received 15 March 1998; accepted for publication 7 October 1998!

Electromagnetic media as, for instance, imperfect dielectrics, semiconductors, and materials with
magnetic losses, have time- ~and frequency-! dependent permittivity, magnetic permeability, and
electric conductivity, and, therefore, energy dissipation and pulse distortion occurs. The
electromagnetic Umov–Poynting’s theorem is reinterpreted in light of the theory of viscoelasticity
in order to define the stored and dissipated energy densities in the time domain. A simple dielectric
relaxation model equivalent to a viscoelastic mechanical model illustrates the analogy that identifies
electric field with stress, electric induction with strain, dielectric permittivity with reciprocal bulk
modulus, and resistance with viscosity. © 1999 Acoustical Society of America.
@S0001-4966~99!04801-8#
PACS numbers: 43.20.Bi, 43.20.Wd @DEC#

INTRODUCTION dielectric that does not satisfy the Kramers–Krönig relations.

Energy-balance equations are important for characteriz-
ing the energy stored and the transport properties in a field.
However, the definition of stored ~free! energy and energy
dissipation rate is controversial, both in electromagnetism1
and viscoelasticity.2 The problem is particularly intriguing in
the time domain, since different definitions may give the
same time-average value for harmonic fields. This ambiguity
is not present when the constitutive relation can be described
in terms of springs and dashpots ~e.g., Refs. 3 and 4!. That is,
when the system can be defined in terms of internal variables
and the relaxation function has an exponential form.5
Christensen6 and Golden and Graham7 give a general expres-
sion of the viscoelastic energy densities which is consistent
with the mechanical model description.
Formal analogies, in the mathematical sense, exist be-
tween electromagnetism and other fields like mechanics8 and
viscoelasticity. In this case, Carcione and Cavallini,9 for in-
stance, show an analogy for vector wavefields and material
properties that allows the acoustic and electromagnetic prob-
lems to be solved with the same analytical methodology. In
this work, the stored electric and magnetic energies are de-
fined in terms of the viscoelastic expression by using the
analogy. The theory is applied to a simple dielectric relax-
ation process that is mathematically equivalent to the Zener
model or standard linear solid viscoelastic rheology.10
It is important to point out that the theory cannot be
applied to the whole range of electromagnetic ~e.m.! prob-
lems, since a mathematical analogy may not necessarily im-
ply a physical analogy. In viscoelasticity, the real part of the
complex moduli ~describing pure deformation modes! is
positive, and the presence of intrinsic dissipation implies ve-
locity dispersion and vice versa ~Kramers–Krönig
relations11,12!. These properties preclude the use of the theory
for an ionized gas whose complex permittivity may, in some
cases, be negative ~e.g., Ref. 13!, or for a lossless dispersive
a!
Electronic mail: carcione@gems755.ogs.trieste.it
The definition of energy is important in a large number
of applications where it is necessary to know how the energy
transferred by the e.m. field is related to the strength of the
field. This context involves the whole electrical, radio, and
optical engineering, where the medium can be assumed di-
electrically and magnetically linear. In particular, the Debye
model has been applied to bioelectromagnetics14,15 in the
analysis of the response of biological tissues, and to geo-
physics, in the simulation of ground-penetrating radar wave
propagation through wet soils.16,17,18
I. MAXWELL’S EQUATIONS AND CONSTITUTIVE
RELATIONS
Maxwell’s equations for isotropic dispersive media, in
the absence of external electric and magnetic currents, are
]B
“3E52 , ~1!
]t
]D
“3H5 1J , ~2!
]t
where E is the electric field, H is the magnetic field, D is
the electric induction, B is the magnetic induction, and J is
the conduction current. The symbol 3 denotes the vector
product.
For time-harmonic fields with time dependence exp(ivt),
where v is the angular frequency, Eqs. ~1! and ~2! read
“3E52i v B, ~3!
“3H5i v D1J, ~4!
respectively, where E, D, H, and B are the corresponding
time-harmonic fields.
We consider constitutive relations D~E! and B~H! of
the form
]E
D5 e * ~5!
]t

626 J. Acoust. Soc. Am. 105 (2), Pt. 1, February 1999 0001-4966/99/105(2)/626/7/$15.00 © 1999 Acoustical Society of America 626
and respectively. The time-average energy densities are such that
]H 1
4 Re~ e ! u Eu 2 5Re~ u e ! stored ~di!electric
B5 m * ~6!
]t
2 1
2 v 21 Im~ s ! u Eu 2 5Im~ u s ! stored electric
~i.e., nonmoving media!, where e and m are the dielectric
permittivity and magnetic permeability functions, and * de-
1
4 Re~ m ! u Hu 2 5Re~ u m ! stored magnetic
notes time convolution. Similarly, generalized Ohm’s law ~17!
1
4 Im~ e ! u Eu 2 5Im~ u e ! dissipated ~di!electric
can be written as
]E
1
2 v 21 Re~ s ! u Eu 2 5Re~ u s ! dissipated electric
J 5s* , ~7!
]t 1
4 Im~ m ! u Hu 5Im~ u m !
2
dissipated magnetic,
where s is the conductivity function. where
For harmonic fields, the constitutive relations read
u s 5 21 v 21 J* –E5 21 v 21 s * u Eu 2 ~18!
D5F
]e
]t
E, S D B5F
]m
]t
H,S D ~8! is the ~complex! electric energy density.

and

J5F S D
]s
]t
E, ~9!
III. ENERGY DEFINITION IN ELECTROMAGNETISM

Time-average energies for harmonic fields are precisely

where F ~•! is the Fourier transform operator. For conve- defined. Let us consider, for instance, the formulation of the
nience, the medium properties are denoted by the same sym- energy balance equation given in Oughstun and Sherman.1
bols, in both the time and the frequency domains, i.e., Upon taking the scalar product of Eq. ~1! with H and Eq.

S D
~2! with E and taking the difference, Oughstun and Sherman
]e
F → e ,...,etc... ~Ref. 1, Eq. 2.2.5! obtain
]t
]U
2“–P 5J –E1 , ~19!
II. UMOV–POYNTING’S THEOREM FOR HARMONIC ]t
FIELDS
where P 5E3H is the Umov–Poynting vector and U is
The scalar product of the complex conjugate of Eq. ~4! the total energy per unit volume given by
with E, use of “•(E3H* )5(“3E)•H* 2E•(“3H* ), U5U e 1U m , ~20!
and substitution of Eq. ~3! gives Umov–Poynting’s theorem
for harmonic fields, with

2“–P5 21 J* –E22i v ~ u e 2u m ! , ~10! ]Ue ]D

5E• , ~21!
]t ]t
where
the time rate of ~di! electric energy density, and
P5 12 E3H* ~11!
is the complex Umov–Poynting vector, and ]Um ]B
5H• , ~22!
]t ]t
u e 5 E–D* ,
1
4 u m 5 B–H* ,
1
4 ~12!
the time rate of magnetic energy density.
are the harmonic ~complex! ~di!electric and magnetic energy The time average of the scalar product of two harmonic
densities. The symbol * denotes complex conjugate. vector fields, with the same oscillation frequency, is given by
Substitution of the constitutive relations ~8! and ~9! into
Eq. ~10! yields ^ A–B & 5 21 Re~ A•B* ! . ~23!
2i“–P52 v ~ e T* u Eu 2 2 m u Hu 2 ! , ~13! Note that taking the time average of Eqs. ~19! yields
where 2 ^ “–P & 5 1
Re~ J* •E! 12 v @ Im~ u e ! 2Im~ u m !# , ~24!
2
i
e T5 e 2 s. ~14! which can be obtained from Eq. ~10! by taking its real part.
v Equation ~24! gives the balance of the rate of dissipated en-
Taking real and imaginary parts of ~13! gives ergy density and is equivalent to Eq. ~16!.
We now review some of the statements discussed by
2 Im~ “–P! 5 v @ Re~ e T ! u Eu 2 2Re~ m ! u Hu 2 # Oughstun and Sherman.1
5power energy density, ~15! ~1! Poynting’s theorem provides a mathematically consis-
2 Re~ “–P! 5 v @ Im~ e T ! u Eu 1Im~ m ! u Hu #
2 2 tent formulation of energy flow ~p. 24!. This does not
preclude the existence of an alternative formulation. For
5rate of dissipated energy, ~16! instance, Jeffreys19 gives an alternative energy balance,

627 J. Acoust. Soc. Am., Vol. 105, No. 2, Pt. 1, February 1999 J. M. Carcione: Energy in lossy electromagnetic media 627
 implying a new interpretation of the Poynting vector ~see density is by no means the only possible one ~see
also the interesting discussion in Robinson20 and Rabotnov,22 p. 72!. Moreover, as we shall see below, the
Jeffreys21!. general expression of the free energy is not uniquely deter-
~2! ... it cannot be definitely concluded that the time rate of mined by the relaxation function.
energy flow at a point is uniquely given by the value of Differentiating W e yields
the Poynting vector at that point, for one may add to the
Poynting vector any solenoidal vector field without af-
fecting the statement of conservation of energy... ~p. 26!.
]We ]D
]t
5
]t
• E 2`
t
K ~ t2 t 2 ,0! D 8 ~ t 2 ! d t 2

E E
In fact, as the authors state, there is no strictly valid
justification for the accepted interpretation of the Poyn- 1 t t ]
1 K ~ t2 t 1 ,t2 t 2 ! D 8 ~ t 1 !
ting vector. 2 2` 2` ]t
~3! ... one cannot, in general, express the electric energy
•D 8 ~ gt 2 ! d t 1 d t 2 . ~28!
density and the dissipation separately in terms of the
dielectric permittivity and electrical conductivity of a The constitutive relation ~5! can be rewritten as
dispersive medium ~p. 31!. In the general case @i.e., the
time-domain equation ~19!#, it is not possible to separate ]D
E5 b * , ~29!
the stored energies from the dissipated energies. Relating ]t
u e and U e gives @see Eq. ~2.2.38!#,

Im~ue!5
1 ]Ue
2v ]t K L, ~25!
where b (t) is the dielectric impermeability function, satisfy-
ing

and no relationship of this type for the real part of u e . ]e ]b

5d~ t !, e ` b ` 5 e 0 b 0 51, e ~ v ! b ~ v ! 51,
The same reasoning applies to the magnetic energy. We ]t * ]t
present in the next section an alternative definition where ~30!
energy can be, in principle, separated between stored and
dissipated. with the subindices ` and 0 corresponding to the limits t
~4! For time-harmonic fields, the separation is shown in Eq. →0 and t→`, respectively. If
~17!. However, there is no relation between the time- b ~ t ! 5K ~ t,0! H ~ t ! , ~31!
average energies defined in Eq. ~17! and the time aver-
ages of U e , U m ~p. 36!. In the next section, a link be- where H(t) is the Heaviside function, then,
tween harmonic energy densities and transient energy
densities is obtained.
Note that Oughstun and Sherman @Ref. 1, Eq. ~2.1.19!#
E t

2`
K ~ t2 t 2 ,0! D 8 ~ t 2 ! d t 2 5E ~ t ! , ~32!

use an exp(2ivt) time dependence. and ~28! becomes

IV. UMOV–POYNTING’S THEOREM FOR TRANSIENT ]D ]We

E– 5 1D e , ~33!
FIELDS ]t ]t
Poynting’s theorem ~19!, omitting Oughstun and Sher- where
man’s interpretation of the energies, is

2“–P 5J –E1E•
]D
]t
1H•
]B
]t
. ~26!
D e ~ t ! 52
1
2
E E 2`
t

2`
t ]
]t
K ~ t2 t 1 ,t2 t 2 !

Let us consider a stored ~di!electric ~free! energy density of 3D 8 ~ t 1 ! –D 8 ~ t 2 ! d t 1 d t 2 ~34!

the form
is the rate of dissipation of ~di! electric energy density. Note

W e~ t ! 5
1
2
E E
t

2`
t

2`
K ~ t2 t 1 ,t2 t 2 !
that the relation ~31! does not determine the stored energy,
i.e., this can not be obtained from the constitutive relation.
However, if we assume that
3D 8 ~ t 1 ! –D 8 ~ t 2 ! d t 1 d t 2 , ~27!
K ~ t, t 1 ! 5G ~ t1 t 1 ! , ~35!
where the prime denotes the first-order derivative with re-
spect to the argument. Hunter ~Ref. 3, p. 545! and Golden such that
and Graham ~Ref. 7, p. 12! define a similar form for the
linear viscoelastic case. The underlying assumptions are that b ~ t ! 5G ~ t ! H ~ t ! , ~36!
the dielectric properties of the medium do not vary with time this choice will suffice to determine K, and
~nonaging material!, and, as in the lossless case, the energy
density is quadratic in the electric field. Moreover, the ex-
pression includes a dependence on the history of the electric W e~ t ! 5
1
2
E E t

2`
t

2`
G ~ 2t2 t 1 2 t 2 ! D 8 ~ t 1 !
field. However, it is important to note that the above assump-
tion as to the structure of the formula for the free energy –D 8 ~ t 2 ! d t 1 d t 2 , ~37!

628 J. Acoust. Soc. Am., Vol. 105, No. 2, Pt. 1, February 1999 J. M. Carcione: Energy in lossy electromagnetic media 628
 D e ~ t ! 52 E E
2`
t

2`
t
G 8 ~ 2t2 t 1 2 t 2 ! D 8 ~ t 1 ! H–
]B ]Wm
]t
5
]t
1D m , ~46!

–D 8 ~ t 2 ! d t 1 d t 2 . ~38! where

Equation ~35! is consistent with the corresponding theory

implied by mechanical models.6 Breuer and Onat23 discuss
W m~ t ! 5
1
2
E E
2`
t

2`
t
F ~ 2t2 t 1 2 t 2 ! B 8 ~ t 1 !
some realistic requirements from which K(t, t 1 ) must have
the reduced from G(t1 t 1 ). –B 8 ~ t 2 ! d t 1 d t 2 , ~47!

E E
Let us calculate the time average of the stored energy t t
density for monochromatic fields. Although D(2`) does D m ~ t ! 52 F 8 ~ 2t2 t 1 2 t 2 ! B 8 ~ t 1 !
2` 2`
not vanish, the transient contained in ~37! vanishes for suf-
ficiently large times, and this equation can be used to com- –B 8 ~ t 2 ! d t 1 d t 2 , ~48!
pute the average of time-harmonic fields. The change of vari-
ables t 1 →t2 t 1 and t 2 →t2 t 2 yields are the stored magnetic energy density and rate of dissipation
of magnetic energy density, respectively, such that
W e~ t ! 5
1
2
EE 0
` `

0
G ~ t 1 1 t 2 ! D 8 ~ t2 t 1 !
H5 g *
]B
, g ~ t ! 5F ~ t ! H ~ t ! , ~49!
]t
–D 8 ~ t2 t 2 ! d t 1 d t 2 . ~39!
with g the magnetic impermeability function.
Using ~23!, the time average of Eq. ~39! is The rate of dissipated electric energy density can be de-
fined as
^ W e & 5 41 v 2 u Du 2

3 EE` `
G ~ t 1 1 t 2 ! cos@ v ~ t 1 2 t 2 !# d t 1 d t 2 .
D s ~ t ! 52 E E
t

2`
t

2`
S ~ 2t2 t 1 2 t 2 ! E 8 ~ t 1 !
0 0
–E 8 ~ t 2 ! d t 1 d t 2 , ~50!
~40!
where
A new change of variables u5 t 1 1 t 2 and v 5 t 1 2 t 2 gives
]E
1
^ W e & 5 v 2 u Du 2
8
EE `

0
u

2u
G ~ u ! cos~ v v ! du d v
J 5s*
]t
, s ~ t ! 5S ~ t ! H ~ t ! . ~51!

Formally, the stored energy density due to the electric cur-

1
5 v u Du 2
4
E `

0
G ~ u ! sin~ v u ! du. ~41!
rents out of phase with the electric field, satisfies
]Ws
5J –E2D s . ~52!
From Eq. ~36!, and using integration by parts, we have that ]t

Re FF S DG ]b
]t
5Re@ b ~ v !#
In terms of the energy densities, Eq. ~26! becomes

2“–P 5
]
~ W s 1W e 1W m ! 1D s 1D e 1D m , ~53!
E ` ]t
5G ~ ` ! 1 v @ G ~ t ! 2G ~ ` !# sin~ v t ! dt.
0 and the correspondences with the averaged time-harmonic
values are
~42!
Using the property
^ W e & 5Re~ u e ! , ^ W m & 5Re~ u m ! ,
^ D e & 52 v Im~ u e ! ,
v E0
`
sin~ v t ! dt51, ~43!
^ D s & 52 v Re~ u s ! ,
~54!

we obtain ^ D m & 522 v Im~ u m ! .

Re@ b ~ v !# 5 v E `
G ~ t ! sin~ v t ! dt. ~44!
Note that ^ J –E & is equal to the rate of dissipated energy
density ^ D s & , and that

K L
0
]We
Substituting Eq. ~44! into Eq. ~41!, and since E5 b ( v )D, we 50, ~55!
finally get ]t
in contrast with Oughstun and Sherman’s result ~25!. The
^ W e & 5 41 u Du 2 Re@ b ~ v !# 5Re~ u e ! . ~45!
same property holds for the stored electric and magnetic en-
A similar calculation shows that ^ D e & 52 v Im(ue). ergy densities.
Similarly, the magnetic term on the right-hand side r.h.s. There are other alternative time-domain expressions for
of Eq. ~26! can be recasted as the energy densities whose time-average values coincide

629 J. Acoust. Soc. Am., Vol. 105, No. 2, Pt. 1, February 1999 J. M. Carcione: Energy in lossy electromagnetic media 629
 with h a parameter that introduces the dissipation. Note that
e 0 is the static ~low-frequency! permittivity and e `
5 e 0 t E / t D 5 e 2 , e 0 is the optical ~high-frequency! permit-
tivity.
We have that

F S
e ~ t ! 5 e 0 12 12
tE
tD D G
exp~ 2t/ t D ! H ~ t ! , ~62!

b ~ t ! 5G ~ t ! H ~ t ! 5
1
e0 F S
12 12
tD
tE D
exp~ 2t/ t E ! H ~ t !G
~63!

FIG. 1. This electric circuit is equivalent to a purely dielectric relaxation

and
process, where e 1 and e 2 are the capacitances, h is a resistance, E is the
electric field, and D is the electric induction. e ~ v ! 5 b 21 ~ v ! 5 e 0 S 11i v t E
11i v t DD. ~64!

with those given in Eqs. ~54!, but fail to match the energy in Equation ~64! can be rewritten as
the time domain. For instance, the following definition,
e 02 e `
W e8 5 21 E–D, ~56! e~ v !5 e `1 . ~65!
11i v t D
as the stored ~di! electric energy density, and For instance, the permittivity ~65! describes the response of

D e8 5
1
2SE–
]D
]t
2D•
]E
]t
, D ~57!
polar molecules, such as water, to the e.m. field.16,25
Substituting Eq. ~63! into Eq. ~29! and defining the in-
ternal variable
as the rate of dissipation, satisfy Eq. ~53! and ^ W 8e & 5 ^ W e &
and ^ D e8 & 5 ^ D e & . However, W e8 is not equal to the energy
stored in the capacitors for the Debye model given in the
j ~ t ! 5 f exp~ 2t/ t E ! H ~ t ! * D, f5
1
e 0t ES12
tD
tE D
,
next section @see Eqs. ~69! and ~71!#. ~66!
yields
V. EXAMPLE 1
E5 D1 j , ~67!
e`
It is well known that the Debye model used to describe
the behavior of dielectric materials24 is mathematically where j satisfies
equivalent to the Zener or standard linear solid model used in
viscoelasticity. The following example uses this model to ]j j
5 f D2 . ~68!
illustrate the concepts presented in the previous section. ]t tE
A. Debye-type dielectric model The ~di!electric energy density is that stored in the ca-
pacitors,
Let us consider a capacitor C 2 in parallel with a series
connection between a capacitor C 1 and a resistance R. This 1 1
circuit obeys the following differential equation: W e5 D 1 –D 1 1 D –D , ~69!
2e1 2e2 2 2

U1 t U
]U 1
5
]t C S
I1 t I
]I
]t
, D ~58! where D 1 and D 2 are the respective electric inductions.
Since D 2 5 e 2 E, D5D 1 1D 2 , and e ` 5 e 2 , we obtain
where U5 ] V/ ] t, I is the current, V is the voltage, and D 1 52 e ` j , ~70!

C5C 1 1C 2 , t U 5R S 1
1
1
C1 C2 D 21
, t I 5C 1 R. ~59! where Eq. ~67! has been used. Note that the internal variable
is closely related to the electric field acting on the capacitor
From the point of view of a pure dielectric process, we in series with the dissipation element. Substitution of the
identify U with E and I with D ~see Fig. 1!. Hence, the electric fields into Eq. ~69!, and after some calculations,
dielectric relaxation model is yields

E1 t E
]E 1
5
]t e0 S
D1 t D
]D
]t
, D ~60! W e5
e`
2 FS e`
e 02 e `Dj • j 1E–E . G ~71!

where Let us verify that Eq. ~37! is in agreement with Eq. ~71!.

S D 21 From Eqs. ~63! and ~66! we have

1 1 1
e 05 e 11 e 2 , tE5 1 , t D 5 e 1 / h , ~61!
h e1 e2 G ~ t ! 5 e 21
0 2 ft E exp~ 2t/ t E ! . ~72!

630 J. Acoust. Soc. Am., Vol. 105, No. 2, Pt. 1, February 1999 J. M. Carcione: Energy in lossy electromagnetic media 630
Replacing Eq. ~72! into Eq. ~37!, and after some algebra,
yields

W e5
1
2e0
1
D–D2 ft E exp 2
2
t
tE F S D
H~ t !*
]
]t
2
D~t ! , G
~73!
where the exponent 2 means the scalar product. Using Eqs.
~66! and ~68! gives
1 1
W e5 D–D2 ~ ft E D2 j ! • ~ ft E D2 j ! .
2e0 2 ft E
~74!
Using e ` t D 5 e 0 t E , and a few calculations, shows that the
expression in ~74! is equal to the stored energy density ~71!.
This equivalence can also be obtained by avoiding the use of FIG. 2. The Zener viscoelastic model is mathematically equivalent to the
internal variables. However, the introduction of these vari- electric circuit represented in Fig. 1. k 1 and k 2 are the springs constants, n is
ables is a requirement to obtaining a complete differential the viscosity of the dashpot, E is the strain, and T is the stress.
formulation of the e.m. equations. This formulation is the
basis of most simulation algorithms.17,18 respectively, and the relaxation times are
The rate of dissipated energy density is
t E 5 b 0 ~ ab 0 2 b ! 21 , t T 51/a ~ t E. t T !. ~81!
1 ]D1 ]D1
D e5 • , ~75! The relaxation function is
h ]t ]t
which from Eqs. ~70! and ~68! becomes G ~ t ! 5 a 21 @ ab 0 2 b 1 b exp~ 2 a t !# . ~82!

D e5
1
h tES D
e` 2
~ ft E D2 j ! • ~ ft E D2 j ! . ~76!
Assuming that the potential energy is stored in the springs,
we have that
W e 5 21 ~ k 1 E 21 1k 2 E 22 ! , ~83!
Taking into account the previous calculations, it is easy to
show that substitution of Eq. ~72! into Eq. ~38! gives Eq. where E 1 and E 2 are the dilatations of the springs. Since T
~76!. 5k 2 E 2 and E5E 1 1E 2 , and using ~77!, we obtain
E 1 52 j / b 0 , E 2 5E1 j / b 0 . ~84!

B. Zener viscoelastic model Note that the internal variable is closely related to the dila-
tation on the spring that is in parallel with the dashpot. Sub-
Fabrizio and Morro ~Ref. 5, p. 42! define a viscoelastic stitution of the dilatations into Eq. ~83! yields

FS D G
solid with internal variables, for which the free ~stored! en-
ergy density can be uniquely determined. For simplicity, we 1 a 1 2 1
W e5 2 j 1 ~ b E1 j ! 2 . ~85!
only consider dilatational deformations, since this choice 2 b b0 b0 0
does not affect the analysis of the problem. If T is the hydro- For isothermal processes, the free energy coincides with
static stress, E is the dilatation ~trace of the strain tensor! and the stored energy. In general, the free energy is not unique
j is the internal variable, the constitutive relation and growth ~Ref. 5, p. 57!. However, for exponential relaxation functions
equation are having one internal variable, like the Zener model, the free
T5 b 0 E1 j ~77! energy is unique.26
Day’s free energy for Zener systems ~Ref. 5, p. 61! is
and
]j
]t
52 a j 2 b E, ~78!
1
c D5 G `E 21
2
1
2
~ G 0 2G ` ! 21/2 H
respectively, where a, b 0 , and b are real and positive con-
stants. Integration of ~78! yields
3 E 0
`
G 8 ~ s !@ E ~ t ! 2E ~ t2s !# ds J 2
, ~86!

j ~ t ! 52 b exp~ 2 a t ! H ~ t ! * E ~ t ! . ~79! where G 8 is the first-order derivative of the relaxation func-

tion. Since
It can be easily shown that Eqs. ~77! and ~78! correspond to
a Zener mechanical model consisting of a spring of constant
k 1 5 b 0 (211 ab 0 / b ) in parallel with a dashpot of viscosity E 0
`
G 8 ~ s ! ds52 b / a , E 0
`
G 8 ~ s ! E ~ t2s ! ds5 j ,
n 5 b 20 / b , together in series connection with a spring of con-
stant k 2 5 b 0 ~Fig. 2!. and using Eq. ~80!, Eq. ~86! becomes
The relaxed and unrelaxed moduli are
G `5 b 02 b / a , G 05 b 0 , ~80!
c D5
1
2 S b
b 02 E 21
a
b
2a
a
E1 j
b D S D 2
5W e . ~87!

631 J. Acoust. Soc. Am., Vol. 105, No. 2, Pt. 1, February 1999 J. M. Carcione: Energy in lossy electromagnetic media 631
This equation differs from the particular form obtained by
Fabrizio and Morro ~Ref. 5, p. 61! when only dilatational
deformations are considered. The cause is a notation error in
the book’s equation.27
The rate of energy density dissipated in the dashpot is
harmonic fields. However, after the transient regime ~large
times after application of the field!, the time average of W e
gives ^ u e & , and, for Debye-type dielectric processes ~expo-
nential relaxation functions!, W e is the energy stored in the
capacitors.

S D
Day’s free energy @Eq. ~90!# is unique for exponential
]E1 2
D e5 n , ~88! dielectric functions with one internal variable. Actually,
]t Day’s free energy is a particular form of Eq. ~37!. Both en-
which from Eqs. ~84! and ~78! becomes ergies coincide for Debye-type dielectric processes.
1
D e5 ~ a j 1 b E !2. ~89!
b
C. Mathematical analogy 1
K. E. Oughstun and G. C. Sherman, Electromagnetic Pulse Propagation
in Causal Dielectrics ~Springer-Verlag, New York, 1994!.
The mathematics of the viscoelastic problem is the same 2
G. Caviglia and A. Morro, Inhomogeneous Waves in Solids and Fluids
as for the dielectric relaxation model previously introduced, ~World Scientific, Singapore, 1992!.
since the mathematical equivalence identifies E with stress T 3
S. C. Hunter, Mechanics of Continuous Media ~Wiley, New York, 1983!.
4
and D with strain E. Then, from the analogy between stress F. Cavallini and J. M. Carcione, ‘‘Energy balance and inhomogeneous
plane-wave analysis of a class of anisotropic viscoelastic constitutive
and electric field, strain and electric induction, internal vari- laws,’’ in Waves and Stability in Continuous Media, edited by S. Rionero
ables, and bulk modulus and dielectric impermeability, and T. Ruggeri ~World Scientific, Singapore, 1994!, pp. 47–53.
Day’s free energy ~86! in the electromagnetic case reads
5
M. Fabrizio and A. Morro, Mathematical Problems in Linear Viscoelas-

HS D
ticity, Studies in Applied Mathematics Vol. 12 ~SIAM, Philadelphia,
21/2
1 1 1 1 1992!.
c D5 D–D1 2 6
R. M. Christensen, Theory of Viscoelasticity. An Introduction ~Academic,
2e` 2 e0 e` New York, 1971!.

E J
7
2 J. M. Golden and G. A. C. Graham, Boundary Value Problems in Linear
`
Viscoelasticity ~Springer-Verlag, New York, 1988!.
3 G 8 ~ s !@ D ~ t ! 2D ~ t2s !# ds , ~90! 8
P. Hammond, Energy Methods in Electromagnetism ~Clarendon, Oxford,
0
1981!.
9
where the exponent 2 means the scalar product. J. M. Carcione and F. Cavallini, ‘‘On the acoustic-electromagnetic anal-
ogy,’’ Wave Motion 21, 149–162 ~1995!.
Comparing Eqs. ~72! and ~82! we obtain the equiva- 10
C. Zener, Elasticity and Anelasticity of Metals ~University of Chicago,
lences Chicago, 1948!.
11
H. A. Kramers, ‘‘La diffusion de la lumiere par les atomes,’’ Atti Congr.
b 1 1 1 1 Intern. Fisica, Como 2, 545–557 ~1927!.
5 2 , b 05 , a5 . ~91!
a e` e0 e` tE 12
R. Krönig, ‘‘On the theory of the dispersion of x-rays,’’ J. Opt. Soc. Am.
12, 547–557 ~1926!.
Substituting these equivalences into Eq. ~85!, the electro- 13
L. P. Felsen and N. Marcuvitz, Radiation and Scattering of Waves
magnetic stored energy density ~71! is obtained. ~Prentice-Hall, Englewood Cliffs, NJ, 1973!.
14
T. M. Roberts and P. G. Petropoulus, ‘‘Asymptotics and energy estimates
On the other hand, applying the analogy and substituting for electromagnetic pulses in dispersive media,’’ J. Opt. Soc. Am. A 13,
the equivalences ~91! into Eq. ~89!, yields the electromag- 1204–1217 ~1996!.
netic rate of dissipated energy density ~76!. 15
P. G. Petropoulos, ‘‘The wave hierarchy for propagation in relaxing di-
The complete correspondence between the dielectric and electrics,’’ Wave Motion 21, 253–262 ~1995!.
16
G. Turner and A. F. Siggins, ‘‘Constant Q attenuation of subsurface radar
the viscoelastic models is pulses,’’ Geophysics 59, 1192–1200 ~1994!.
17
J. M. Carcione, ‘‘Ground-penetrating radar: Wave theory and numerical
fields properties
simulations in lossy anisotropic media,’’ Geophysics 61, 1664–1677
E ↔ T e` ↔ b 21
0 18
~1996!.
T. Xu and G. A. McMechan, ‘‘GPR attenuation and its numerical simu-
D ↔ E tD ↔ tE lation in 2.5 dimensions,’’ Geophysics 62, 403–414 ~1997!.
19
C. Jeffreys, ‘‘A new conservation law for classical electrodynamics,’’
E1 ↔ T1 tE ↔ tT , SIAM ~Soc. Ind. Appl. Math.! Rev. 34, 386–405 ~1993!.
~92!
h n 21
20
E2 ↔ T2 ↔ F. N. H. Robinson, ‘‘Poynting’s vector: Comments on a recent paper by
Clark Jeffreys,’’ SIAM ~Soc. Ind. Appl. Math.! Rev. 36, 633–637 ~1994!.
D1 ↔ E1 e1 ↔ k 21
1
21
C. Jeffreys, ‘‘Response to a Commentary by F. N. H. Robinson,’’ SIAM
~Soc. Ind. Appl. Math.! Rev. 36, 638–641 ~1994!.
D2 ↔ E2 e2 ↔ k 21
2
22
Y. N. Rabotnov, Elements of Hereditary Solid Mechanics ~Mir, Moscow,
1980!.
j ↔ j 23
S. Breuer and E. T. Onat, ‘‘On the determination of free energy in linear
viscoelastic solids,’’ Z. Angew. Math. Phys. 15, 184–190 ~1964!.
where the symbols can be identified in Figs. 1 and 2. 24
A. R. Von Hippel, Dielectrics and Waves ~Wiley, New York, 1962!.
25
P. Debye, Polar Molecules ~Dover, New York, 1929!.
VI. REMARKS 26
D. Graffi, and M. Fabrizio, ‘‘Non unicità dell’energia libera per materiali
viscoelastici,’’ Atti Accad. Naz. Lincei, Cl. Sci. Fis. Mat. Nat. Rend.
It is important to keep in mind that Eq. ~37! holds when LXXXIII, 209–214 ~1989!.
27
D(t→`)50, an assumption that is violated by time- A. Morro, personal communication.

632 J. Acoust. Soc. Am., Vol. 105, No. 2, Pt. 1, February 1999 J. M. Carcione: Energy in lossy electromagnetic media 632

View publication stats