Journal of Electromagnetic Waves and Applications

ISSN: 0920-5071 (Print) 1569-3937 (Online) Journal homepage: www.tandfonline.com/journals/tewa20

Electromagnetic Wave Propagation in Linearly

Dispersive Media Using Higher-Order Weno Scheme

S. Sardeshpande & A. Chatterjee

To cite this article: S. Sardeshpande & A. Chatterjee (2009) Electromagnetic Wave Propagation
in Linearly Dispersive Media Using Higher-Order Weno Scheme, Journal of Electromagnetic
Waves and Applications, 23:16, 2135-2142, DOI: 10.1163/156939309790109270

To link to this article: https://doi.org/10.1163/156939309790109270

Published online: 03 Apr 2012.

Submit your article to this journal

Article views: 51

View related articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tewa20
J. of Electromagn. Waves and Appl., Vol. 23, 2135–2142, 2009

ELECTROMAGNETIC WAVE PROPAGATION IN LIN-

EARLY DISPERSIVE MEDIA USING HIGHER-ORDER
WENO SCHEME

S. Sardeshpande and A. Chatterjee

Department of Aerospace Engineering
Indian Institute of Technology Bombay
Mumbai 400 076, India

Abstract—Higher-order Weighted Essentially Non-Oscillatory

(WENO) based spatial discretization is used to numerically model
electromagnetic wave propagation in linearly dispersive media. An
Auxiliary Differential Equation (ADE) approach is used to account
for effect of the dispersive medium. Results are presented for propa-
gation of plane pulsed wave in single and multi-pole Debye medium.
Higher-order accuracy can be easily maintained, including at material
interfaces involving discontinuous change in material property, mak-
ing WENO schemes an attractive option for modeling electromagnetic
wave propagation in layered media.

1. INTRODUCTION

Numerical simulation of electromagnetic wave propagation in linearly

dispersive media is of interest in many engineering fields. Examples of
linearly dispersive medium are plasma, biological material like tissue
and muscle, ice, snow, radar absorbing material (RAM), optical fibers
etc. [1]. Applications are widespread and can vary from analysis of
RAM coatings for reduced radar cross section of aircraft to modeling
interaction of electromagnetic fields with biological substances. The
latter, in turn, can be useful in disease detection as well as in analyzing
exposure to unwanted electromagnetic radiation. The popular Finite
Difference Time Domain (FDTD) [2] method, which solves Maxwell’s
equations in time-domain, is commonly used for numerically simulating
electromagnetic wave propagation in linearly dispersive media by
including an appropriate model for the dispersive medium. FDTD
Corresponding author: A. Chatterjee ([email protected]).
2136 Sardeshpande and Chatterjee

modeling for linearly dispersive media as reviewed in [1], can be based

on Auxiliary Differential Equation (ADE), recursive-convolution and
Z-transform methods. Accuracy considerations have also led to the
use of higher-order FDTD methods for modeling wave propagation in
linearly dispersive media [3, 4]. FDTD methods are based on Cartesian
grids, and can have difficulties dealing with complex geometries in
simulations of practical interest. The Finite Volume Time Domain
(FVTD) method based on solving the integral form of the time-domain
Maxwell’s equations can be used to tackle complex geometries with
relative ease. Higher-order characteristic based schemes, originally
developed for time-dependent Euler equations in gas dynamics, are
usually adapted to solve the time-domain Maxwell’s equations in
the FVTD framework since both represent systems of hyperbolic
conservation laws [2, 5]. In the present work, we simulate one-
dimensional electromagnetic wave propagation in Debye medium using
a higher-order Weighted Essentially Non-Oscillatory (WENO) [6, 7]
based spatial discretization. Debye dielectric, the representative linear
dispersive medium considered in the present work, is often used to
model electromagnetic wave interaction with water based substances
including biological material such as tissues [1]. The one-dimensional
treatment presented here, can be easily extended to multidimensions
in a FVTD framework. Arbitrary higher-order accurate spatial
accuracy can be obtained using WENO based spatial discretization,
originally proposed to solve gas dynamic problems containing both
shock waves and complex smooth structures. WENO schemes are an
extension of the successful Essentially Non-Oscillatory (ENO) [8, 9]
family of schemes, which allow for higher-order accuracy in smooth
parts of the solution along with sharp non-oscillatory resolution of
discontinuities in the flowfield. In the present context, a WENO
based spatial discretization allows higher-order accuracy to be easily
incorporated, including at material interfaces involving discontinuous
change in material properties. The ADE approach followed in present
simulations, allows effect of dispersive media to be relatively easily
added to FVTD formulations for electromagnetic wave propagation
in free space or perfect dielectrics. The ADE approach for modeling
dispersive media requires additional ordinary differential equations
to be solved to update polarization vector and polarization current
in addition to solving the time-domain Maxwell’s equations with an
unchanged Faraday’s law.
EM wave in linearly dispersive media 2137

2. GOVERNING EQUATIONS AND NUMERICAL

SCHEME

The time-domain Maxwell’s equations incorporating an M -pole Debye

model can be expressed as [1]
 1 M
∂E
0 ∞ = ∇×H− [(sk − ∞ ) 0 E − Pk ] (1)
∂t τk
k=1
∂B
= −∇ × E (2)
∂t
E, H, P, respectively, are electric field, magnetic field and polarization
vectors. ∞ is infinite-frequency relative permittivity, sk the kth
zero-frequency relative permittivity also known as the static relative
permittivity, 0 the free-space permittivity and τk the kth relaxation
time. Equation (1) is Ampere’s law modified for Debye medium and
Equation (2) is the unchanged Faraday’s law. The accompanying
ADEs for updating the polarization vector P, in turn, are
∂Pk 1
= [(sk − ∞ ) 0 E − Pk ] . (3)
∂t τk
For one-dimensional wave propagation, the above time-domain
Maxwell’s equations form a 1D system of hyperbolic conservation laws
with an added source term, and are written as
∂u ∂f (u)
+ = s(u) (4)
∂t ∂x
where
   
Ey Hz /0 ∞
u= , f (u) = (5)
Hz Ey /μ0
with the source term
 M 
− k=1 1
[(sk − ∞ ) 0 Ey − Pk ]
s(u) = τk 0 ∞ . (6)
0
Plane wave propagation is in the Cartesian x direction, with Ey and
Hz electric and magnetic field components. Subscripts x, y, z indicate
components along respective Cartesian directions. The polarization
vector in the accompanying ADEs similarly consists of a single
component in the y-direction. The WENO reconstruction is applied
to characteristic variables v = R−1 u where
⎡  ⎤
0 ∞
1 − μ0 1
R−1 = ⎣  ⎦ (7)
2  
0 ∞
μ0 1
2138 Sardeshpande and Chatterjee

with μ0 the free-space permeability. The matrix of right eigenvectors

for the Jacobian matrix in the 1D system of hyperbolic conservation
laws are given by  
− 0μ0∞ μ0
R= 0 ∞ (8)
1 1
with diagonal matrix of eigenvalues
− √c∞ 0
Λ= √c
(9)
0 ∞
√
where c = 1/ 0 μ0 . The state vector u and the polarization vector are
co-located at cell centres xi with i = 1, N . The WENO reconstruction
±
procedure is used to obtain vi+1/2 on either side of a cell interface
xi+1/2 . Details of the WENO reconstruction procedure are widely
available in literature including Refs. [6, 7]. Results presented in
this work are based on a fifth-order WENO reconstruction with the
numerical flux function at cell interface xi+1/2 based on the Godunov
flux [6]. The numerical flux function for the pth characteristic equation
at the cell interface is expressed as
⎧
⎨ a≤v
min f p (v p ) if a ≤ b
p ≤b c
p
Fc,i+1/2 (a, b) = (10)
⎩ max fcp (v p ) if a > b.
p
b≤v ≤a
p− p+
a = vi+1/2 , b = vi+1/2 , are the reconstructed pth characteristic
variables on either side of the cell interface. fcp (v p ) the pth
characteristic flux equals λp v p with λp the pth eigenvalue. The
numerical flux function at cell interface xi+1/2 is recovered from the
characteristic form as Fi+1/2 = RFc,i+1/2 . This process, is also
equivalent to defining the numerical flux function based on the exact
solution of the Riemann problem at the cell interface, with initial
Riemann data based on a higher-order WENO reconstruction. Time
integration is based on a third-order Runge-Kutta method [10], with
residual for the ith cell expressed as
(Fni+1/2 − Fni−1/2 )
L(u(n) ) = − + s(un ) (11)
Δx
where Δt is the time increment and Δx the uniform grid spacing. The
polarization vector is considered to be frozen at time level n in the
intermediate Runge-Kutta time steps. The polarization vectors are
updated separately, by simply adapting the formulation suggested by
EM wave in linearly dispersive media 2139

Young [1] for use in FDTD based frameworks as

 
Pkn+1 − Pkn 1 n Pkn+1 + Pkn
= (sk − ∞ ) 0 Ey − (12)
Δt τk 2
for the k = 1, 2, . . . , M poles considered.

3. NUMERICAL EXAMPLES

3.1. Single-Pole Debye Medium

The normal interaction of a pulsed plane wave, with the air-water
interface considered here, is also solved in a FDTD framework in
[11]. The discretization, time step and initial conditions are taken
to be identical to that in [11] and mentioned here for the sake of
completeness. 1000 uniform cells are used to discretize the one-
dimensional space with the first 499 cells representing freespace and
the next 501 water. Water is modeled as a single-pole Debye medium
with s = 81, ∞ = 1.8 and τ = 9.4e−12. The cell width Δx = 37.5 µm
and time step Δt = 0.0625 ps. The incident electric field is in the
form of a Gaussian pulse described in [11]. The plot for the initial
electric field against cell numbers is shown in Fig. 1. Plots of
electric field at 480 and 1200 time steps are, respectively, shown in
Fig. 2 and Fig. 3. Computations utilized a Mur absorbing boundary
conditions [12]. The results agree well with FDTD results in [11]. The
reflection coefficient at the air-water interface was also calculated and
this involved collecting the total electric field at air-water interface for
8000 time steps. The time history of the reflected field was obtained
by subtracting the incident field before conversion to the frequency
domain using a discrete Fourier transform. The reflection coefficient

1 1
Electric Field Intensity (kV/m)
Electric Field Intensity (kV/m)

0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0
0
-0.2
-0.2
-0.4
-0.6 -0.4
-0.8 -0.6
-1 -0.8
-1
0 200 400 600 800 1000 0 200 400 600 800 1000
Cell Number Cell number

Figure 1. Initial electric field. Figure 2. Total electric field

after 480 time steps.
2140 Sardeshpande and Chatterjee

finally obtained by dividing the reflected field by the incident field

compares well with the exact solution as shown in Fig. 4.

3.2. Five-pole Debye Medium

One-dimensional scattering from a dielectric slab consisting of a five-
pole Debye dielectric also used to model muscle is solved next. This
problem is also solved in the FDTD context in [13]. The time
step, discretization and initial conditions are identical to that in [13].
The parameters for the five-pole Debye model model are ∞ = 4.3,
s1 = 8e5, s2 = 8.19e4, s3 = 1.19e4, s4 = 3.63e1, s5 = 5.10e1,
τ1 = 2.31 ms, τ2 = 3.70µs, τ3 = 238 ns, τ4 = 692 ps, τ5 = 7.96 ps.
The physical and computational space consists of 1000 cells with the

1 1
Magnitude of Reflection Coefficient
Electric Field Intensity (kV/m)

0.8 0.95
Exact
0.6 WENO-5
0.9
0.4
0.2 0.85
0 0.8
-0.2 0.75
-0.4
0.7
-0.6
-0.8 0.65
-1 0.6
0 200 400 600 800 1000 0 10 20 30 40 50 60
Cell Number Frequency (GHz)

Figure 3. Total electric field Figure 4. Computed magnitude

after 1200 time steps. of reflection coefficient at air-
water interface compared with
exact results.
1 0.2
0.9 0.1 WENO-5
Exact
0.8 0
Electric Field

0.7 -0.1
Electric Field

0.6 -0.2
0.5 -0.3
0.4 -0.4
0.3 -0.5
0.2 -0.6
0.1 -0.7
0 -0.8
0 200 400 600 800 1000 0 200 400 600 800 1000
Position (Cell Number) Position (Cell Number)

Figure 5. Initial electric field. Figure 6. The exact and

computed and electric field after
750 time steps.
EM wave in linearly dispersive media 2141

dielectric slab resident between cells 500 and 750 and rest of the domain
as freespace. Δx = 0.119 mm and Δt = 0.199 ps. The incident field
is a Gaussian pulse, as in [13], and the initial electric field is shown in
Fig. 5. The computed electric field again compares well with the exact
solution as shown in Fig. 6.

4. CONCLUSIONS

A higher-order WENO based solution of electromagnetic wave

propagation in single and multi-pole Debye media is presented. The
treatment presented in one dimension can be easily extended to
multidimensions in the context of a FVTD framework for solving the
Maxwell’s equations. Arbitrary higher-order accuracy can be easily
maintained, including at material interfaces involving discontinuous
change in material property, with WENO based spatial discretizations
making it an attractive option for solution of electromagnetic wave
propagation in heavily layered media. Time integration of ADEs in the
present work follows that proposed earlier for FDTD based methods
and could possibly be developed further to be made more compatible
within a FVTD framework.

REFERENCES

1. Young, J. L. and R. O. Nelson, “A summary and systematic

analysis of FDTD algorithms for linearly dispersive media,” IEEE
Antennas and Propagation Magazine, Vol. 43, No. 1, 61–77, 2001.
2. Yee, K. S. and J. S. Chen, “The finite-difference time domain
(FDTD) and finite-volume time-domain (FVTD) methods in
solving Maxwell’s equations,” IEEE Transactions on Antennas and
Propagation, Vol. 45, No. 3, 354–363, 1997.
3. Young, J., “A higher order FDTD method for EM propagation in
a collisionless cold plasma,” IEEE Transactions on Antennas and
Propagation, Vol. 44, No. 9, 1283–1289, 1996.
4. Prokopidis, K. P., E. P. Kosmidou, and T. D. Tsiboukis, “An
FDTD algorithm for wave propagation in dispersive media using
higher-order schemes,” Journal of Electromagnetic Waves and
Applications, Vol. 18, No. 9, 1171–1194, 2004.
5. Shang, J. S., “Characteristic-based algorithms for solving the
Maxwell equations in the time domain,” IEEE Antennas and
Propagation Magazine, Vol. 37, No. 3, 15–25, 1995.
6. Shu, C. W., “Essentially non-oscillatory and weighted essentially
non-oscillatory schemes for hyperbolic conservation laws,” Ad-
2142 Sardeshpande and Chatterjee

vanced Numerical Approximation of Nonlinear Hyperbolic Equa-

tions, B. Cockburn, C. Johnson, C.-W. Shu, and E. Tadmor (eds.),
Lecture Notes in Mathematics, Vol. 1697, 325–343, Springer, New
York, 1998.
7. Jiang, G.-S. and C.-W. Shu, “Efficient implementation of weighted
ENO schemes,” Journal of Computational Physics, Vol. 126, No. 1,
202–228, 1996.
8. Harten, A. and S. Osher, “Uniformly higher-order accurate
nonoscillatory schemes, I,” SIAM Journal on Numerical Analysis,
Vol. 24, No. 2, 279–309, 1987.
9. Harten, A., B. Engquist, S. Osher, and S. Chakravarthy,
“Uniformly higher-order accurate nonoscillatory schemes, III,”
Journal of Computational Physics, Vol. 71, No. 2, 231–303, 1987.
10. Gottlieb, S. and C. Shu, “Total variation diminishing Runge-
Kutta schemes,” Mathematics of Computation, Vol. 67, No. 221,
73–85, 1998.
11. Luebbers, R., F. Hunsberger, K. Kunz, R. Standler, and
M. Schneider, “A frequency-dependent finite-difference time-
domain formulation for dispersive materials,” IEEE Transactions
on Electromagnetic Compatibility, Vol. 32, No. 3, 222–227, 1990.
12. Mur, G., “Absorbing boundary conditions for the finite-
difference approximation of the time-domain electromagnetic-field
equations,” IEEE Transactions on Electromagnetic Compatibility,
Vol. 23, No. 4, 377–382, 1981.
13. Young, J., “Propagation in linear dispersive media: Finite
difference time-domain methodologies,” IEEE Transactions on
Antennas and Propagation, Vol. 43, No. 4, 422–426, 1995.