Electromagnetic Green Functions Using Dierential Forms

Karl F. Warnick and David V. Arnold

Short Title: Electromagnetic Green Forms

Department of Electrical and Computer Engineering

459 Clyde Building
Brigham Young University
Provo, UT, 84602

Warnick, et al. 1 September 14, 1995

 Abstract| In this paper we redevelop the scalar and dyadic Green functions of
electromagnetic theory using dierential forms. The Green dyadic becomes a dou-
ble form, which is a dierential form in one space with coecients that are forms in
another space, or a dierential form-valued form. The results presented here corre-
spond closely with the usual dyadic treatment, but are clearer and more intuitive.
Many of the usual expressions using green functions in vector notation require a
surface normal with the Green forms the surface normal is unnecessary. We illus-
trate the formalism by computing scattering from a randomly rough conducting
surface and deriving the Green form for a dielectric half-space. We also dene the
interior derivative, which is equivalent to the coderivative but for a constant met-
ric has a computational rule dual to that of the exterior derivative and simplies
calculation in coordinates. This work makes available some of the tools that have
not yet been presented in the language of dierential forms but are essential in
applied electromagnetics.

Warnick, et al. 2 September 14, 1995

 1. INTRODUCTION

In this paper we treat Green function methods in electromagnetic (EM) eld theory using
the calculus of dierential forms. The calculus of dierential forms has been applied to EM
theory by Deschamps 1], Baldomir 2], Schleifer 3], Thirring 4], Burke 5, 6], Bamberg 7],
Ingarden and Jamiolkowksi 8], Parrott 9] and others.
Several authors have advocated the use of the calculus of forms in engineering EM theory,
but some important tools for applied problems have not been developed. In 10] the authors
presented a representation of EM boundary conditions using dierential forms. In this
work we develop another tool well suited for practical use, the Green form in the (3+1)
representation.
As proposed by Thirring 4], the EM Green function becomes a double form. Double
forms are dened by de Rham in 11]. Green forms are treated in the mathematics literature
(see 12] and its references), and Thirring gives the time-dependent Green form for electro-
dynamics in Minkowski spacetime. Our Green form has the same components as the Green
dyadic in Kong 13] and therefore is easily related to the usual methods in applied electro-
magnetics. We derive expressions for the Green double form in terms of the scalar Green
function, the electric eld due to a surface current density and the Stratton-Chu formula.
The use of dierential forms makes the results presented here clearer in certain ways
than the usual vector and dyadic treatment. In obtaining expressions for observed elds
in terms of the Green forms, the product rule for the exterior derivative takes the place of
several vector identities. This makes the derivation much cleaner. The dyadic expression for
observed elds due to tangential elds along the surface of an observation region using the
Green dyadic includes a surface normal. With the corresponding expression using the Green
form, the surface normal is unnecessary.
In this paper we also dene the interior derivative, which is equivalent to the standard
coderivative 7], but simplies calculations in coordinates when the metric is constant. The
computational rule which we propose is dual to that of the exterior derivative.
In Sec. 2 we review operations on forms and treat double forms briey. In Sec. 3 we solve
Maxwell's laws of electromagnetics in terms of the Green double form and the scalar Green

Warnick, et al. 3 September 14, 1995

function. Finally, in Sec. 4 we illustrate the method by computing scattered elds from a
rough conducting surface and deriving the Green form for a dielectric half-space. This work
shows that the calculus of dierential forms can be used in all applications to which Green
functions and dyadics are suited.

2. DEFINITIONS

In this section we give denitions and notation to be used in Sec. 3 to derive the Green
forms. We dene the interior and exterior derivatives, the interior and exterior products and
the Laplace-de Rham operator. Double 1-forms are also introduced in this section.

2.1 Operators
The exterior derivative d is dened in 7] and elsewhere. It can be represented formally as
d @x@ i dxi ^ (1)
where x1 : : : xn are coordinates on an n-dimensional space and the summation convention is
used. The exterior product ^ is the antisymmetrized tensor product, so that dxi ^ dxj =
; dxj ^ dxi and dxi ^ dxi = 0. (Often the wedge between dierentials is dropped there is an
implied wedge between the dierentials in the integrand of any multiple integral.) The partial
@y dy ^ dx+ @z dz ^ dx
derivatives @x@ act on the coecients of a form, so that in R3, d(f dx) = @f @f
i

since dx ^ dx = 0. We dene the interior derivative in the euclidean metric similarly,

d @ dxi (2)
@xi
where is the interior product dened in 10]. In R3 we have d (f dx) = @f
@x since dx dx = 1
and dy dx = dz dx = 0.
The interior product is dened to be the contraction of a vector with a k-form (which is
a totally antisymmetric (0k ) tensor). In this paper we use the euclidean metric, so we can
extend this denition to the interior product of a 1-form and k-form easily since vectors and
1-forms have the same components. The interior product of dxi and dxi1 ^  ^ dxi is zero
j k

Warnick, et al. 4 September 14, 1995

for ij not equal to any of i1 : : : ik , otherwise it is (;1)j;1dxi1 ^  ^ dxi ;1 ^ dxi +1 ^  ^ dxi
j j k

for 1  j  k. Thus, by (2) the interior derivative of a form is computed by moving each
dierential in turn to the leftmost position by alternating the sign of the form each time two
dierentials are swapped, removing that dierential and taking the corresponding partial
derivative.
The interior derivative is equivalent1 up to a sign to the coderivative dened in 4, 7] and
elsewhere,
d ! = (;1)k+1 ?;1 d ? ! (3)
where k is the degree of ! and ? is the Hodge star operator. In R3 with the euclidean metric,
?1 = dx dy dz, ? dx = dy ^ dz, ? dy = dz ^ dx, ? dz = dx ^ dy and ?;1 = ?. Note that
the interior derivative contains the sign (;1)k+1 naturally. For a nonconstant metric, such
as would arise in curvilinear coordinates, (3) replaces (2) as the denition of the interior
derivative.
The interior derivative is easier to compute with than the coderivative, as illustrated by
the following example. We rst use the coderivative to nd
; ? d ? (D1 dy dz + D2 dz dx + D3 dx dy)
@ @ @
= ; ? ( dx + dy + dz) ^ (D1 dx + D2 dy + D3 dz)
@x @y @z
= ; ? (D1x dx ^ dx + D1y dy ^ dx + D1z dz ^ dx
+ D2x dx ^ dy + D2y dy ^ dy + D2z dz ^ dy
+ D3x dx ^ dz + D3y dy ^ dz + D3z dz ^ dz )
= (D2z ; D3y ) dx + (D3x ; D1z ) dy + (D1y ; D2x ) dz:
Using the denition of the interior derivative we compute the same result immediately,
d (D1 dy dz + D2 dz dx + D3 dx dy) = D1y dz ; D1z dy + D2z dx ; D2x dz + D3x dy ; D3y dx
where we have removed each dierential in turn, after it moving to the left if necessary using
the antisymmetry of the exterior product, and taken the corresponding partial derivative.
The Laplace-de Rham operator  is
 = d d + dd (4)

Warnick, et al. 5 September 14, 1995

which is a generalization of the vector operator r2. With the euclidean metric,  becomes
(!)i = Pj @x@22 !i where the subscript i indexes components of !. On 1-forms, (4) is equiv-
j

alent to the euclidean vector identity r2 = ;r  r  +rr.

The generalized Stokes theorem is
Z Z
V
d! = @V
! (5)
where ! is a p-form and V is a p + 1 dimensional region with @V as its boundary. Also, the
interior product of two arbitrary forms a and b satises
a b = ?(?b ^ a) (6)
where ? is the Hodge star operator.

2.2 Double Forms

A double form 11] is a dierential form in one space with coecients that are forms in
another space. The double forms that we will use in this paper are associated with R3  R3 0
where R3 is the observation space and R3 0 is the source space. We will use 1-form valued
1-forms, or double 1-forms, which can be written in general
G = G11 dx dx0 + G12 dx dy0 + G13 dx dz 0
+ G21 dy dx0 + G22 dy dy0 + G23 dy dz0
+ G31 dz dx0 + G32 dz dy0 + G33 dz dz 0:
Between the primed and unprimed dierentials there is an implied tensor product rather
than an exterior product. The coecients are functions Gij (r r0) of both the observation
and source coordinates.
A double form can be used as a transformation kernel (if its coecients vanish suciently
quickly at innity). If we x a double 1-form G, we have the transformation from R3 0 to R3
given by the volume integral Z
! = G ^ ?!0 (7)
where !0 is a 1-form and ! is the transform of !0 due to the kernel G. The exterior product

Warnick, et al. 6 September 14, 1995

yields a 3-form in dx0 dy0 dz0 which is integrated over R3 0. The unprimed dierentials remain,
resulting in the 1-form ! in the observation space.
Components of a dyadic are of the form x^i x^j with no prime, which does not show explicitly
the relationship the dyadic can provide to the source and observation spaces. The action of
a double form G as a kernel from the source to the observation space is clearly reected in
the product dxi dxj 0 of primed and unprimed dierentials in each component.
We introduce the identity kernel I where I is the double form
dx dx0 + dy dy0 + dz dz0 (8)
or dr dr0 + rd r0d0 + r sin d r0 sin 0 d 0 in spherical coordinates.  is the three-dimensional
R
Dirac delta function (x ; x0)(y ; y0)(z ; z0 ). Using this kernel, I ^ ?!0 = !0j(xyz) = !,
so that the transformation takes !0 from source to observation space without otherwise
changing its components.

3. THE EM GREEN FORMS

In this section we derive expressions for the electric eld 1-form at an observation point
due to applied sources and elds using the Green forms. We consider time-harmonic (e;i!t )
elds in an isotropic medium of permittivity
and permeability .
We write Faraday's and Ampere's laws as
dE = i!B (9)
d B = i!
E ;  ? J (10)
where E is the electric eld intensity 1-form, B is the magnetic ux density 2-form and J is
the electric current density 2-form. The constitutive relations are D =
? E and B =  ? H ,
where D is the electric ux density 2-form and H is the magnetic eld intensity 1-form.
By taking the interior derivative of (9) and substituting (10), we obtain
(d d + k2 )E = ;i! ? J (11)

Warnick, et al. 7 September 14, 1995

where k2 = !2
. The Green double 1-form G for Eq. (11) then satises
(d d + k2)G = ;I: (12)
Here and below, all derivatives will act on primed coordinates unless otherwise noted, but
to avoid clutter, the derivatives will remain unprimed.
Let V 0 be a volume containing source current density given by the 2-form J 0 . Outside V 0,
the electric eld due to the sources is E , a 1-form in the observation space. Using the unit
kernel I we can write Z
E = 0 I ^ ?E 0 V
(13)
Substituting Eq. (12) into (13), we obtain
Z
E= d ? dG ^ E 0 ; k2 G ^ ?E 0 ] (14)
V0

where we have used d dG = (;1)1 ? d ? dG and moved a ? across the exterior product (if
 and are both p-forms, then ? ^ =  ^ ? , as can be veried easily in coordinates).
Using the product rule for the exterior derivative, d( ^  ) = d ^  + (;1)deg  ^ d , Eq.
(14) becomes
Z
E= G ^ (d ? dE 0 ; k2 ? E 0) + d(?dG ^ E 0 + G ^ ?dE 0)]: (15)
V0

After applying the star operator to Eq. (11) and using the denition of the interior derivative,
we can insert J 0 into (15),
Z
E= i!G ^ J 0 + d(?dG ^ E 0 + G ^ ?dE 0)]: (16)
V0

Applying the generalized Stokes theorem and using Faraday's law, we nd that
Z Z
E = i! V0
G ^ J0 + @V 0
(i!G ^ H 0 + ?dG ^ E 0 ) (17)
where the second term takes into account elds on the surface @V 0 due to sources outside of
V 0.
The integrals in (17), like all integrals of dierential forms, can be integrated by the
method of pullback 7]. This method is completely general and allows forms to be integrated
conveniently over parameterized regions.

Warnick, et al. 8 September 14, 1995

 The expression corresponding to (17) using the dyadic Green function G is
Z Z
E = i! V0
G  J0dv0 ; @V 0
(i!G  ^n  H0] + r  G  ^n  E0])ds0
where n^ is an outward surface normal. Because E 0 and H 0 are 1-forms rather than 2-
forms, their exterior product with G behaves dierently than the exterior product G ^ J 0 .
Components of E 0 and H 0 tangent to @V 0 naturally do not contribute to the surface integral
in (17). Thus, the surface normal is eliminated and a simpler expression results. Also,
Green's theorem is used in deriving the dyadic result. Green's theorem on forms is an
immediate consequence of the product rule for the exterior derivative and the generalized
Stokes theorem.

3.1 The Scalar Green Function

The scalar Green function g satises the wave equation for an elementary source ;,
( + k2 )g = ; (18)
It can easily be shown that gI satises Eq. (18) for the source ;I . Substituting gI for g
and ;I for ; in (18), expanding the Laplace-de Rham operator and rearranging gives
d d(gI ) + k2 (gI + 1 dd gI ) = ;I: k2 (19)
Since dd = 0, this can be rewritten as
d d(gI + 12 dd gI ) + k2 (gI + 12 dd gI ) = ;I:
k k (20)
By comparison with Eq. (12), we see that
G = (1 + k12 dd )gI (21)
up to a solution of (d d + k2)H = 0 where H is a double 1-form. This freedom is used to
satisfy boundary conditions. The double 1-form I could be included in the denition of g,
but we prefer to leave it out so that g is the usual scalar Green function.
By pullback we can transform Eq. (18) to the spherical coordinate system. Noting that
g0 in free space is spherically symmetric, we nd that
1 d2 rg + k2g = ;(r): (22)
r dr2 0 0

Warnick, et al. 9 September 14, 1995

The solution of this dierential equation is the usual result
g0 = 4er :
ikr
(23)
From this we can compute G0 ,
G0 = g0I + k12 @@rg20 dr dr0
2
(24)
which becomes g0(rd r0d0 + r sin d r0 sin 0d 0) in the far eld.
It is easily veried that if g is symmetric in r and r0 (as is the case for reciprocal media),
the derivatives in the rst term of Eq. (21) can be taken to act on umprimed rather than
primed coordinates. Thus, we can write using (17) after neglecting sources outside of V 0,
1 Z
E = i!(1 + dd ) gI ^ J k2 V0
(25)
where the derivatives act on unprimed coordinates. The Lorentz gauge is d A = i!
 ,
where is the scalar electric potential and A is the magnetic vector potential 1-form. In the
R
Lorentz gauge, E = i!A ; d together with (25) imply that A =  V 0 gI ^ J .
For a region V 0 containing no sources, Eq. (17) becomes
Z
E= (i!G ^ H 0 + ?dG ^ E 0 ): (26)
@V 0

Substituting (21) and using dd = 0 gives

Z
E= (i!gI ^ H 0 + i! dd gI ^ H 0 + ?dgI ^ E 0): (27)
@V 0 2 k
Using the product rule for the exterior derivative, (27) can be rewritten as
Z
E= (0 i!gI ^ H 0 + i! d(d gI ^ H 0) ; d gI ^ dH 0] + ?dgI ^ E 0 ): (28)
@V 2 k
The second term vanishes by the generalized Stokes theorem. Using Ampere's law, we obtain
the Stratton-Chu formula,
Z
E= i!gI ^ H 0 + (?dgI ) ^ E 0 ; (d gI ) ^ ?E 0] (29)
@V 0

which again eliminates the dot and cross products with a surface normal found in the usual
vector expression 13].

Warnick, et al. 10 September 14, 1995

3.2 Fields Due to a Surface Current
We can nd the observation elds due to a surface current using either the rst or the second
term of Eq. (17). We will do the computation both ways. The rst method illustrates how
the direction normal to a surface can be integrated out using the interior product and a
singular integrand. The second method uses the boundary conditions in 10] to arrive at the
same result.
Suppose that the electric current density is given by J 0 (f ) where the dierentiable func-
tion f (x0 y0 z0) is zero only along the surface @V 0 , J 0 is a 2-form parallel to @V 0 and (f ) is the
q
Dirac delta function df df(f (x0 y0 z0 )). Graphically, the surfaces of the 1-form df and
the surface normal 1-form n = q df are parallel to @V 0 because f is constant along @V 0.
df df
The tubes of the 2-form J 0 are also parallel to @V 0, so that J 0 must contain a factor n. We can
thus decompose J 0 into the product n^T where T is a 1-form satisfying n T = 0. The interior
product distributes over the exterior product by  ( ^  ) = (  ) ^  + (;1)deg  ^ (  ).
It follows that n ^ (n J 0 ) = J 0. Substituting this into the rst term of (17), we nd that
Z
E = i! V0
G ^ (n ^ n J 0)(f ):
Integrating along the direction perpendicular to @V 0 eliminates n(f ), so that
Z
E = i! @V 0
G ^ Js0 (30)
where Js is the surface current 1-form dened by (f )Js = n J (see reference 10] the surface
current can be dened equivalently to be Js = ;v s where v is the velocity eld of the
surface charge density 2-form s).
Alternately, if there is a tangential magnetic eld H 0 along the boundary of V 0, we can
can place by equivalence a surface current density J 0(f ) on @V 0 satisfying the appropriate
boundary condition for vanishing magnetic eld inside V 0. Since the integral in the second
term of (17) is along @V 0, we can apply the boundary projection operator n n^ to the
integrand without aecting the integral,
Z
E = i! @V 0
n (n ^ G ^ H 0):
After interchanging n and G using the antisymmetry of the exterior product, we distribute

Warnick, et al. 11 September 14, 1995

the interior product n to obtain
Z
E = ;i! @V 0
f(n G) ^ n ^ H 0 ; G ^ n (n ^ H 0)]g (31)
Since the 2-form n ^ H 0 is parallel to the surface @V 0 and n G is a 0-form in the primed
coordinates, the rst term of (31) drops out. By the boundary condition n (n ^ H 0) = Js0
the second term reduces to Eq. (30).
There is an interesting dierence between Eq. 30 and the usual vector result. In the
R
integral @V 0 G  Js0 ds0, the vector Js0 must lie on the surface @V 0 or the integration yields
an incorrect result. In (30), Js can be arbitrary because its normal component does not
contribute to the integral.
As noted in 10], the surface current density 1-form diers from the usual vector quantity
Js in that it includes the geometry of the boundary more naturally, making the computation
of elds due to surface sources using the Green form slightly simpler than with vectors.

4. APPLICATIONS

In this section we give two elementary applications of the Green form as exercises in the
use of this method: on-axis scattering from a rough surface and the derivation of the Green
form for a dielectric half-space. The rst example illustrates the simplication that results
due to the absence of surface normals in the Green form expressions. The second is intended
as a reference to show how Green forms are manipulated in solving a standard problem.

4.1 Rough Surface Scattering

Consider a rough, perfectly conducting surface z =  (x y). For a horizontally polarized
plane wave propagating in the ;z direction towards the surface with the magnetic eld
Hi = E00 e;ikz dy, the approximate (physical optics) total tangential eld at the boundary is
Ht0 = 2Hitjz=(x0y0 ) (32)
Since H 0 in Eq. (17) need not be tangential to @V 0, we can substitute 2Hi0 in place of 2Hit0 .
0
Using the free-space scalar Green function g0 = e 4( z; ) for the scattered electric eld in the
ik z z

Warnick, et al. 12 September 14, 1995

z direction, the Green double form G becomes (I ; dz dz0 )g0. The third term of (17) drops
out since Et0 is approximately zero, so that
Z
E (z) = i! G ^ H 0
Z ik
z ; (x0 y0 )]
= i! ( dx dx0 + dy dy0) e 4z
^ 2E0 e;ik(x0y0) dy0
Z 0

= ikE0 e dx dx0 dy0e;i2k(x0y0)

ikz

2z
where the integration is over the region of illumination.
Obtaining this result using vector notation requires the computation of a surface normal.
The surface normal contains a factor due to the curvature of the surface that cancels a factor
in the dierential surface area element dS . The present method eliminates the need to take
into account such geometrical factors that are ultimately extraneous to the problem. While
this is only a slight simplication computationally, it shows that the calculus of forms is not
only equivalent to vector analysis but is more natural for these types of problems.

4.2 Half-Space Green Forms

Following the dyadic treatment in Kong 13], we determine the Green form G10 from source
point in region 0, and observation point in region 1. Taking the three-dimensional Fourier
transform of g = e4k jr(r;;rr0j) and integrating over kz , we nd that for z < 0,
i
0

i Z
g(r) = (2)2 dkx dky 2k1 ei(k x+k y;k0 z)
x y
(33)z

0z
q
where k0z = k2 ; kx2 ; ky2 and the source is at r0 = 0. Now, if K = x^kx + y^ky ; z^k0z and
K = kx dx + ky dy ; k0z dz, then
dd eiKr( dx dx0 + dy dy0 dz dz0 ) = di(kx dx0 + ky dy0 ; k0z dz 0)eiKr]
= diK0 eiK r]
= ;KK 0 eiKr:
Using this result to apply G = (1 ; k12 dd )gI to Eq. (33), we nd that
Z 0 eiKr
G(r) = 8i 2 dkx dky (I ; KK
k2 ) k : (34)
0z

Warnick, et al. 13 September 14, 1995

In order to simplify this expression, we note that I can be rewritten in the new orthonormal
basis of 1-forms
we = ? jK ^ dz
K ^ dzj
= kyqdx2; kx2dy
kx + ky
wh = ? wejK^jK
q2 2
= ; k (
1z x
q 2 2y + dz kxk+ ky
k dx + k dy )
k kx + ky
wk = jK Kj :
The source basis we0 , wh0 , wk0 is dened similarly but with primed dierentials. In this basis,
0
I ; KK
k2 becomes we we + wh wh . Translating the source to r we have the Green double form
0 0 0
from the source region to region 1,
G10 (r r0 ) =
i Z dk dk ei(K1 r;Kr0) T T E w (;k ) w0 (;k ) + T T M w (;k ) w0 (;k )]
8 2 x y
k0z e 1z e 0z h 1z h 0z

(35)
where K1 = x^kx + y^dy ; z^k1z . The coecients T T E and T T M are found by matching the
tangential components dz dz ^G and
1 dz dz ^?dG, where ?d acts on unprimed coordinates,
of G00 and G10 at the boundary of regions 0 and 1. Using stationary phase to evaluate the
Fourier integral as kr ! 1,
G10(r r0) = 4er e;Kr0 T10T E we(;k1z ) we0 (;k0z ) + kk1 T10T M wh(;k1z ) wh0 (;k0z )]
ikr
(36)
which is the desired result. Writing the current density 2-form J 0 in the source basis shows
that terms in wh0 ^ wk0 produce TE waves, whereas terms in wk0 ^ we0 produce TM waves.

5. CONCLUSION

In this paper we show that the dyadic Green function can be replaced by a double form.
The derivations presented here are more straightforward than their vector-dyadic counter-
parts because the product rule for the exterior derivative and the generalized Stokes theorem
Warnick, et al. 14 September 14, 1995
replace unwieldy vector identities. The expressions obtained are also simpler than those using
vectors and dyadics because surface normals required in the vector formulation are absent.
Over the past few decades, various authors have contended that the calculus of dierential
forms brings greater clarity and conciseness to basic electromagnetic theory than vector
analysis. With vector analysis, one must force the geometry of the physics into a much
smaller set of quantities (scalars, vectors and dyadics) than is available with dierential
forms. With the geometry of a problem more naturally represented by dierential forms,
the expressions themselves often suggest the next step in a derivation or lead to a useful
physical interpretation. We nd that this clarity and conciseness extends to the method of
Green forms as well.
Possible extensions of this work include propagation in anisotropic media, by treating
the material properties as metrics and thereby embedding the permittivity and permeability
tensors in the star operator and interior derivative.

Acknowledgements. This material is based in part upon work supported under a National
Science Foundation Graduate Fellowship to KFW. The authors also gratefully acknowledge
the eort of the referees in carefully reviewing our paper and providing insightful and helpful
comments.

Warnick, et al. 15 September 14, 1995

References

1] G. A. Deschamps, \Electromagnetics and dierential forms," IEEE Proc., vol. 69,
pp. 676{696, June 1981.
2] D. Baldomir, \Dierential forms and electromagnetism in 3-dimensional Euclidean space
R3," IEE Proc., vol. 133, pp. 139{143, May 1986.
3] N. Schleifer, \Dierential forms as a basis for vector analysis|with applications to
electrodynamics," Am. J. Phys., vol. 51, pp. 1139{1145, Dec. 1983.
4] W. Thirring, Classical Field Theory, vol. II. New York: Springer-Verlag, 1978.
5] W. L. Burke, \Manifestly parity invariant electromagnetic theory and twisted tensors,"
J. Math. Phys., vol. 24, pp. 65{69, Jan. 1983.
6] W. L. Burke, Applied Dierential Geometry. Cambridge: Cambridge University Press,
1985.
7] P. Bamberg and S. Sternberg, A Course in Mathematics for Students of Physics, vol. II.
Cambridge: Cambridge University Press, 1 ed., 1988.
8] R. S. Ingarden and A. Jamiolkowksi, Classical Electrodynamics. Amsterdam, The
Netherlands: Elsevier, 1 ed., 1985.
9] S. Parrott, Relativistic Electrodynamics and Dierential Geometry. New York: Springer-
Verlag, 1987.
10] K. F. Warnick, R. H. Selfridge, and D. V. Arnold, \Electromagnetic boundary conditions
using dierential forms," IEE Proc., vol. 142, no. 4, pp. 326{332, 1995.
11] G. de Rham, Dierentiable Manifolds. New York: Springer-Verlag, 1984.
12] Q.-K. Lu, \Green forms to intrinsic metric of a ball," Sci. in China Ser. A, vol. 32,
pp. 129{141, Feb. 1989.
13] J. A. Kong, Electromagnetic Wave Theory. New York: John Wiley & Sons, 2 ed., 1990.
14] H. Federer, Geometric Measure Theory. New York: Springer-Verlag, 1969.

Warnick, et al. 16 September 14, 1995

Footnote (First sentence of the third paragraph of Sec. 2.1)

1. We prefer the term \interior derivative" to the usual \coderivative" for pedagogical
reasons.
The denition (2) also provides a computational advantage over (3) when extended to a
noneuclidean (but constant) metric, such as would arise when modelling material properties.
The interior product is dened to be the contraction of a vector and p-form. Formally, the
index of the \1-form" @x@ dxi must be raised using the inverse metric gij , so that the interior
i

derivative becomes for a constant but otherwise arbitrary metric,

d = gij @x@ i dxj
where the dxj are a basis for the vector space dual to the space of 1-forms. It can be shown
that this expression is equivalent to the coderivative for the same metric (the star operator
is metric-dependent see 4]). A non-reciprocal medium yields a non-symmetric gij , which is
not a metric, but can still be used in the denition of the interior derivative.
Federer 14] also denes an interior derivative, but it takes p-vectors to (p ; 1)-vectors
and is metric independent.

Warnick, et al. 17 September 14, 1995