Coordinate Transformation & Invariance in Electromagnetism

Steven G. Johnson, notes for the course 18.369 at MIT Created April 2007; updated March 10, 2010
It is a remarkable fact [1] that Maxwells equations under any coordinate transformation can be written in an identical Cartesian form, if simple transformations are applied to the materials ( and ), the elds (E and H), and the sources ( and J). This result has numerous useful and/or beautiful consequences, from designs of invisibility cloaks [2], to a simple derivation of PML absorbing boundaries [3], to enabling analyses of bent and twisted waveguides in terms analogous to a quantum Stark eect [4] , to providing a simple way of applying numerical methods designed for Cartesian coordinates to other coordinate systems [1]. Here, we review the proof in a compact form (from [5]), generalized to arbitrary anisotropic media. (Most previous derivations seem to have been for isotropic media in at least one coordinate frame [1], or for coordinate transformations with purely diagonal Jacobians J where Jii depends only on xi [3], or for constant ane coordinate transforms [6].) We will show that Maxwells equations take on the same form (14) in the primed coordinate system, with replaced by , if we make the transformations: E = (J T )1 E, H = (J T )1 H, J J T , det J J J T , det J JJ , det J , det J (6) (7)

(8)

(9)

J =

(10)

(11)

Summary of the Result

Maxwells equations in Cartesian coordinates x are written (in natural units 0 = 0 = 1): E +J t H E = t (E) = H = (H) = 0, (1) (2) (3) (4)

where J and are the usual free current and charge densities, respectively, and (x) and (x) are the 3 3 relative permittivity and permeability tensors, respectively. Now, suppose that we make some (dierentiable) coordinate transformation x x (usually chosen to be nonsingular, with some exceptions [2]). Let J denote the 3 3 Jacobian matrix: xi Jij = . (5) xj 1

where J T is the transpose. Note that, even if we start out with isotropic materials (scalar and ), after a coordinate transformation we in general obtain an isotropic materials (tensors and ). For example, if x = sx for some scale factor s = 0, then = /s and = /s, which is precisely the material scaling required to keep e.g. the eigenfrequencies xed under a rescaling of a structure. Note also that if s = 1, i.e. a coordinate inversion, then we set E = E, H = H, = and = , and the system switches handed-ness (ipping the sign of the refractive index). [A more common alternative choice in that case would be to set H = H, transforming H as a pseudovector [7], while keeping and unchanged. This corresponds to sprinkling a few factors of sign(det J ) in the above equations, which we are free to do as long as the sign is constant.]

Proof
We will proceed in index notation, employing the Einstein convention whereby repeated indices are summed over. Eq. (1) is now expressed: a Hb
abc

coordinates is also straightforward. Gauss Law, eq. (3), becomes = a ab Eb = Jia i ab Jjb Ej
1 = Jia i (det J )Jak kj Ej 1 = (det J )i ij Ej + (a Jak det J )kj Ej

= cd

= (det J )i ij Ej , (21) where abc is the usual Levi-Civita permutation tensor and a = /xa . Under a coordinate which gives ( E ) = for = / det J , corresponding to eq. (11). Similarly for eq. (4). x change x x , if we let Jab = xa be the b Here, we have used the fact that (non-singular) Jacobian matrix associated with 1 the coordinate transform (which may be a funca Jak det J = a anm kij Jin Jjm /2 = 0, (22) tion of x), we have from the cofactor formula for the matrix inverse, a = Jba b . (13) and recalling that a Jjb abc = 0 from above. In particular, note that = 0 = 0 and Furthermore, as in eqs. (67), let J = 0 J = 0, so a non-singular coordinate Ea = Jba Eb , (14) transformation preserves the absence (or presence) of sources.d Ha = Jba Hb . (15) Hence, eq. (12) becomes Jia i Jjb Hj
abc

Ed + Jc t

(12)

References
El + Jc . t

(16) [1] A. J. Ward and J. B. Pendry, Refraction and geometry in Maxwells equations, J. Modern Optics, vol. 43, no. 4, pp. 773793, 1996. Here, the Jia i = a derivative falls on both the Jjb and Hj terms, but we can eliminate the [2] J. B. Pendry, D. Schurig, and D. R. Smith, former thanks to the abc : a Jjb abc = 0 because Controlling electromagnetic elds, Science, a Jjb = b Jja . Then, again multiplying both vol. 312, pp. 17801782, 2006. sides by the Jacobian Jkc , we obtain [3] F. L. Teixeira and W. C. Chew, GenE eral closed-form PML constitutive tensors to Jkc Jjb Jia i Hj abc = Jkc cd Jld l + Jkc Jc t match arbitrary bianisotropic and dispersive (17) linear media, IEEE Microwave and Guided Noting that Jia Jjb Jkc abc = ijk det J by deWave Lett., vol. 8, no. 6, pp. 223225, 1998. nition of the determinant, we nally have [4] S. G. Johnson, M. Ibanescu, M. Skorobo1 Jkc Jc E Jkc cd Jld l + (18) gatiy, O. Weisberg, T. D. Engeness, M. Soldet J t det J jai, S. A. Jacobs, J. D. Joannopoulos, or, back in vector notation, and Y. Fink, Low-loss asymptotically singlemode propagation in large-core OmniGuJ J T E ide bers, Optics Express, vol. 9, no. 13, H = +J, (19) det J t pp. 748779, 2001. i Hj
ijk

= cd Jld

where J = J J/ det J according to (10). Thus, [5] C. Kottke, A. Farjadpour, and S. G. Johnson, we see that we can interpret Amperes Law in Perturbation theory for anisotropic dielecarbitrary coordinates as the usual equation in tric interfaces, and application to sub-pixel Euclidean coordinates, as long as we replace the smoothing of discretized numerical methmaterials etc. by eqs. (68). By an identical ods, Phys. Rev. E, vol. 77, p. 036611, 2008. argument, we obtain [6] I. V. Lindell, Methods for Electromagnetic J J T H Fields Analysis. Oxford, U.K.: Oxford Univ. , (20) E = Press, 1992. det J t which yields the transformation (9) for . [7] J. D. Jackson, Classical Electrodynamics. The transformation of the remaining diverNew York: Wiley, third ed., 1998. gence equations into equivalent forms in the new 2