1444 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 53, NO.

4, APRIL 2005

Three-Dimensional Diffraction by Infinite

Conducting and Dielectric Wedges Using
a Generalized Total-Field/Scattered-Field
FDTD Formulation
Jiuan-Her Chang and Allen Taflove, Fellow, IEEE

Abstract—We extend the generalized total-field/scattered-field modeling of three-dimensional (3-D) diffraction by infinite con-
formulation of the finite-difference time-domain method to permit ducting and dielectric wedges. While a number of recent papers
efficient computational modeling of three-dimensional (3-D) have been published dealing with improvements on the TF/SF
diffraction by infinite conducting and dielectric wedges. This
new method allows: 1) sourcing a numerical plane wave having formulation (for example [3], [4]), the present work is the first to
an arbitrary incident angle traveling into, or originating from, a extend the TF/SF interface into the 3-D perfectly matched layer
perfectly matched layer absorbing boundary and 2) terminating (PML) absorbing boundary region. This work is an advance rel-
the infinite wedge inside the perfectly matched layer with neg- ative to the two-dimensional (2-D) technique reported in [1] in
ligible reflection. We validate the new method by comparing its that, here, all six vector electromagnetic field components are
results with the analytical diffraction coefficients for an infinite
3-D right-angle perfect electric conductor wedge obtained using accounted for at various FDTD space lattice points along and
the uniform theory of diffraction. Then, we apply the new method immediately adjacent to the TF/SF interface. Furthermore, we
to calculate numerical diffraction coefficients for a 3-D infinite demonstrate how to calibrate the PML in 3-D to numerically
right-angle dielectric wedge, covering a wide range of incident and obtain the incident field components needed to implement the
scattering angles. Finally, we show means to compactly store the TF/SF algorithm where the TF/SF interface is embedded within
calculated diffraction coefficients in a manner which permits easy
interpolation of the results for arbitrary incidence and observation the PML. In essence, we show how to efficiently setup and use
angles. an auxiliary 3-D FDTD space lattice to obtain the necessary in-
cident wave data within the PML.
Index Terms—Diffraction, finite-difference time-domain
(FDTD) method, wedges. Our new method allows: 1) sourcing a numerical plane wave
having an arbitrary incident angle traveling into, or originating
from, a perfectly matched layer absorbing boundary and 2)
I. INTRODUCTION terminating the infinite wedge inside the perfectly matched
layer with negligible reflection. We validate the new method by
W IRELESS communication systems ideally provide con-
tiguous coverage for mobile users in the geographical
areas served. A combination of software planning tools and
comparing its results with the analytical diffraction coefficients
for an infinite 3-D right-angle perfect electric conductor (PEC)
on-site measurements is used to determine the location and type wedge obtained using the uniform theory of diffraction (UTD)
of radio equipment that is required to achieve this goal. In urban [5]. Then, we apply the new G-TF/SF method to calculate
environments, where cells are small, planning tools usually em- numerical diffraction coefficients for a 3-D infinite right-angle
ploy deterministic prediction models. Here, accurate estimation dielectric wedge, covering a wide range of incident and scat-
of radio wave diffraction from building corners and edges be- tering angles. Finally, we show means to compactly store the
comes very important. However, existing analytical methods calculated diffraction coefficients in a manner which permits
are generally inadequate in calculating electromagnetic wave easy interpolation of the results for arbitrary incidence and
diffraction from such material wedges. Inaccurate diffraction observation angles. This storage/interpolation technique is also
models could lead to significant errors in RF coverage predic- a significant advance relative to [1].
tions for cellular systems in urban environments.
In this paper, we extend the generalized total-field/scattered- II. GENERAL DESCRIPTION OF THE 3-D G-TF/SF FDTD
field (G-TF/SF) formulation [1] of the finite-difference time- FORMULATION
domain (FDTD) method [2] to permit efficient computational The G-TF/SF formulation of the FDTD method permits ac-
curate modeling of an infinite material wedge inside a compact
3-D FDTD grid to efficiently obtain numerical diffraction coef-
Manuscript received June 22, 2004; revised October 6, 2004. This work was
performed under a grant from Motorola, Incorporated, Schaumburg, IL, to the ficients. As shown in Fig. 1, the G-TF/SF boundary is located
Motorola/Northwestern University Center for Communications. in part within the PML [6] absorbing boundary region of the
The authors are with the Department of Electrical and Computer Engineering, FDTD grid.
Northwestern University, Evanston, IL 60208 USA (e-mail: [email protected]
western.edu). Fig. 2 shows the six faces of the 3-D TF/SF interface of
Digital Object Identifier 10.1109/TAP.2005.846359 Fig. 1, where a 3-D right-angle wedge is located inside the
0018-926X/$20.00 © 2005 IEEE
CHANG AND TAFLOVE: 3-D DIFFRACTION BY INFINITE CONDUCTING AND DIELECTRIC WEDGES 1445

Fig. 1. G-TF/SF formulation of FDTD extends the material wedge into the PML to suppress spurious diffraction. Only diffraction from the desired corner is
observed.

G-TF/SF boundary and partially embedded within the PML. Rewriting the above equation in a more convenient form, we
An all Berenger split-field PML formulation is assumed for have
simplicity. The portion of the G-TF/SF surface located inside
the PML is indicated by shading. For the special field points
along this surface that lie in free space, we use the well-known (3a)
update equations for the conventional TF/SF boundary [2].
The special field points in the PML absorbing region require Similarly, at this same face of the TF/SF surface in the PML, we
special update equations, to be discussed in the following have
section.

(3b)
III. SPECIAL UPDATE EQUATIONS FOR G-TF/SF BOUNDARY
IN PML REGION In an analogous manner, we proceed to the other five faces of
the TF/SF surface in the PML.
Referring to Fig. 2(a), we first consider the face of TF/SF Face in PML: [Regions I–VI in Fig. 2(b)]
the TF/SF surface inside the PML region, i.e., Regions I and
III–VI, for the E-field updates. Here, the special update equation
for is given by:
TF/SF Face in PML: [Regions I and III–VI, in
Fig. 2(a)] (4a)

(1)
(4b)
TF/SF Face in PML: [Regions I–IV in Fig. 2(c)]
Since ,
therefore

(5a)

(2) (5b)
1446 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 53, NO. 4, APRIL 2005

Fig. 2. Six faces of the total-field/scattered-field interface of Fig. 1. (a) j =j face; (b) j =j face; (c) k =k face; (d) k =k face; (e) i =i face; (f) i =i
face.

TF/SF Face in PML: [Regions I–IV in Fig. 2(d)] TF/SF Face in PML: [Regions I and III–VI in Fig. 2(e)]

(6a)
(7a)

(6b) (7b)
CHANG AND TAFLOVE: 3-D DIFFRACTION BY INFINITE CONDUCTING AND DIELECTRIC WEDGES 1447

TF/SF Face in PML: [Regions I–VI in Fig. 2(f)] Face in PML: [Adjacent to regions I–IV in
Fig. 2(c)]

(8a)
(12a)

(8b)
(12b)
Equations (3)–(8) represent the complete set of special elec-
tric-field update equations required to implement the 3-D gen-
Face in PML: [Adjacent to regions I–IV in
eralized TF/SF boundary of Figs. 1 and 2.
Fig. 2(d)]
Next, we consider the face adjacent to the TF/SF
surface inside the PML for the required H-field updates. Here,
the special update equation for is given by:
Face in PML: [Adjacent to regions I and III–VI
in Fig. 2(a)]
(13a)

(9)
(13b)

Since Face in PML: [Adjacent to regions I and III–VI

, therefore in Fig. 2(e)]

(10a)
(14a)
Similarly, at this same face adjacent to the TF/SF surface in the
PML, we have

(14b)
(10b)
Face in PML: [Adjacent to regions I–VI in
In an analogous manner, we proceed to the other five faces ad- Fig. 2(f)]
jacent to the TF/SF surface in the PML:
Face in PML: [Adjacent to regions I–VI in
Fig. 2(b)]

(15a)

(11a)

(15b)

Equations (10)–(15) represent a complete set of special mag-

netic-field update equations required to implement the general-
(11b) ized TF/SF boundary of Figs. 1 and 2. These equations, as well
1448 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 53, NO. 4, APRIL 2005

as (3)to (8) can be implemented as long as we know the appro- for all PML depths in the PML region. Correspond-
priate incident fields in the PML. Note that all of these equa- ingly, in another two preliminary runs we obtain ,
tions, knowledge of the total incident field components , , and for all PML depths
, etc., and not the individual split incident field is required. ( , ) in the and PML region.
We can now obtain from , ,
IV. CALIBRATION PROCESS , , , and for all
G-TF/SF boundary segments in PML, and all angles of inci-
In this section, we discuss the method to obtain the incident
dence. For example, referring to Fig. 2; consider the
E and H field components in the PML region that are required
face:
in (3)–(8) and (10)–(15).
We cannot assume a perfect exponential decay of the incident For ,
wave in the PML region. For accuracy, this decay must be ob- Region I:
tained numerically in a calibration process

(16)
(18a)
Here, represents the required incident E or H field com-
ponent at an observation point in the PML region, is
the corresponding free space incident field (which can be ob-
where
tained using the table look-up procedure of [2]); is an
Region II:
appropriate multiplying factor to be determined in the calibra-
tion process and is the direction of the incident plane
wave.
For a given FDTD grid configuration and an arbitrary (18b)
, is obtained by conducting preliminary
FDTD runs to calibrate the performance of the PML. In these Region III:
runs, we illuminate the desired PML region of the grid with a
pulsed incident plane wave having a desired center frequency,
, and full-width at half-maximum (FWHM) bandwidth, .
We record the amplitude of the E and H field components (18c)
within the PML, at each required depth, .
We also compute the amplitude of the corresponding incident
electric and magnetic fields, , in free where
space. Then, for a wave impinging upon a given PML region, Region IV:
we define the attenuation factor, , as

(17a)

Similarly, for a wave originating within a given PML region, we

define the amplification factor, , as

(17b)
Note that is obtained in the prelimi-
(18d)
nary FDTD run by illuminating the given PML region with a
plane wave incident at . where and where
We now summarize the calibration procedure. For each re- Region V:
quired incident direction , we set up three prelimi-
nary FDTD calibration runs to obtain in ,
, PML regions. In the preliminary FDTD runs we launch
an approximate plane wave impinging upon the local PML re-
gion of interest by exciting a nearby parallel TF/SF boundary
plane. For example, to obtain the calibration for a plane wave
penetrating into the PML region, we use a nearby TF/SF (18e)
boundary having only an face, and observe penetration
into the PML at points away from the edges of the face. Here,
in one preliminary run, we obtain and
CHANG AND TAFLOVE: 3-D DIFFRACTION BY INFINITE CONDUCTING AND DIELECTRIC WEDGES 1449

where
Region VI:

(18f)

where and
For ,
Region I:

(19a)

where Fig. 3. Snapshot visualizations of a Gaussian pulsed sinusoidal plane wave

Region II: launched within an empty 3-D total-field zone that extends into PML. (a)
Horizontal cut through the space lattice. (b) Vertical cut through the space
lattice.

where and
(19b) Region V:

Region III:

(19c)

(19e)
where where
Region IV: Region VI:

(19f)
(19d) where and .
1450 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 53, NO. 4, APRIL 2005

Fig. 4. Snapshot visualizations of a Gaussian pulsed sinusoidal plane wave

illuminating a right-angle PEC wedge that extends into PML. (a) Horizontal cut
through the space lattice. (b) Vertical cut through the space lattice.

Fig. 5. Comparison of G-TF/SF FDTD and UTD results for the diffraction
The of , , , , faces coefficient of a right-angle PEC wedge. (a) At 850 MHz as a function of
can be obtained by analogy [7]. We use the above equations and scattering angle  for fixed = 30 ,  = 55 , and  = 35 .
(b) As a function of frequency for  = 30 ,  = 165 ,  = 55 ,
(16) to obtain the incident E and H field components at any point and  = 65 .
in the PML region. We then use the special update (3)–(8) and
(10)–(15) to implement the G-TF/SF boundary in the PML.
B. Right-Angle PEC Wedge in Total-Field Zone
V. NUMERICAL RESULTS FOR THREE DIMENSIONAL We next apply the G-TF/SF technique to calculate the diffrac-
SCATTERING tion coefficients for an infinite right-angle PEC wedge. The 3-D
geometry of the wedge in the FDTD grid and sample field vi-
A. Empty Total-Field Zone sualizations are shown in Fig. 4. Here, the incident illumina-
We first demonstrate how the G-TF/SF formulation al- tion is a Gaussian pulsed plane wave with
lows launching a numerical plane wave within an empty 3-D and . The diffracted wave observation points
total-field zone that extends into PML. Fig. 3 is a snapshot are marked in the scattered-field region outside of the TF/SF
visualization of a Gaussian pulsed sinusoidal plane wave prop- boundary. FDTD grid resolution equals , where is the
agating at , . The center frequency is wavelength at 850 MHz.
and the full width at half maximum (FWHM) Fig. 5(a) and (b) compares the G-TF/SF FDTD-computed
bandwidth . Here, three preliminary calibra- diffraction coefficients for the infinite PEC wedge of Fig. 4 with
tion runs were required to obtain , , the UTD asymptotic results for the corresponding infinite wedge
, , , and for all (see [8] and Appendices A and B). Fig. 5(a) shows the variation
PML depths ( , , ). From Fig. 3, we see that the propa- of the amplitude of the soft and hard diffraction coefficients at
gating wave generated by the G-TF/SF technique maintains its as a function of the observation angle at
plane nature despite the presence of the PML. Further, the wave a fixed and a fixed observation distance
is well confined within the TF zone, with little external leakage. from the diffracting corner. Here, the incident wave propagates
CHANG AND TAFLOVE: 3-D DIFFRACTION BY INFINITE CONDUCTING AND DIELECTRIC WEDGES 1451

Fig. 7. G-TF/SF FDTD calculated soft diffraction coefficient data set for the
3-D " = 6 dielectric wedge of Fig. 6. (a) Discrete values calculated at 10
increments of  and  in the range (15   75 , 165 
  285 ) for  = 65 . (b) MATLAB interpolation of (a) visualized at
1 increments of  and  .

VI. DEVELOPMENT OF A DIFFRACTION COEFFICIENT LIBRARY

Fig. 6. G-TF/SF FDTD diffraction coefficients for a right-angle lossless VIA INTERPOLATION
dielectric wedge (" = 6). 850-MHz incident plane wave at  = 55 . (a)
Soft diffraction coefficients. (b) Hard diffraction coefficients.
Finally, we present an efficient means to store G-TF/SF
FDTD-calculated diffraction coefficient data for a wide range
at ( , ). Fig. 5(b) shows the variation of of incident and diffracted wave angles. We exploit the capability
the amplitude of the soft and hard diffraction coefficients as a of MATLAB to fit an interpolating hypersurface to multidimen-
function of frequency at a fixed observation point ( , sional data, in our case the calculated variation of the diffraction
, ) over the frequency range 700 MHz to coefficient as a function of the complete set of incident-wave
1.3 GHz for the incident plane wave propagating at ( , and diffracted-wave propagation angles. As an example, we
). Both figures show very good agreement of the consider the same 3-D dielectric wedge of Fig. 6.
G-TF/SF FDTD results and the asymptotic UTD calculations. Fig. 7(a) is a 3-D surface visualization of a set of soft diffrac-
tion coefficient data for this dielectric wedge calculated using
G-TF/SF FDTD at 10 increments of and in the range
C. Right-Angle Dielectric Wedge in Total-Field Zone and for .
Fig. 7(b) shows the corresponding visualization generated by
We now demonstrate the capability of the G-TF/SF FDTD MATLAB using a three-dimensional cubic surface interpola-
technique to calculate 3-D diffraction coefficients for an infinite tion of the data of Fig. 7(a) at 1 increments of and .
lossless right-angle dielectric wedge. The FDTD grid geometry Fig. 8(a) and (b) shows the corresponding visualizations for
remains that of Fig. 4, but the finite PEC wedge is replaced by the same dielectric wedge for the hard diffraction coefficients.
a finite lossless dielectric wedge of permittivity . Finally, Fig. 9 visualizes the soft diffraction coefficient data for
Fig. 6(a) and (b) shows multiple curves for the G-TF/SF incident angles and for
FDTD-calculated soft and hard diffraction coefficients at fixed angles and .
for plane waves at incident angles We see that the problem of generating a library of diffraction
and observation angles . In both cases coefficients could be greatly simplified. Namely, the G-TF/SF
. We ensure that the distance from the vertex of FDTD method would be used to calculate the diffraction coeffi-
the wedge to each observation point is greater than , where cients for a particular material wedge over a relatively coarsely
is the wavelength at . sampled set of incident and diffracted wave angles. This small
1452 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 53, NO. 4, APRIL 2005

Fig. 9. G-TF/SF FDTD calculated soft diffraction coefficient data set for the
3-D " = 6 dielectric wedge of Fig. 6. (a) Discrete values calculated at 10
Fig. 8. G-TF/SF FDTD calculated hard diffraction coefficient data set for the increments of  and  in the range (15  75 , 25   
3-D " =
6 dielectric wedge of Fig. 6. (a) Discrete values calculated at 10 75 ) for fixed angles ( = 185 ,  = 30 ). (b) MATLAB interpolation
increments of  and  in the range (15    75 , 165  of (a) visualized at 1 increments of  and  .
 285 ) for  = 65 . (b) MATLAB interpolation of (a) visualized at
1 increments of  and  .
APPENDIX A
UNIFORM THEORY OF DIFFRACTION FOR AN INFINITE
set of data would be supplied to a field engineer to be imported to RIGHT-ANGLE PEC WEDGE
a MATLAB package which would subsequently interpolate the
FDTD data to any required resolution. In this manner, the burden In Fig. 10(a), a plane wave is incident at an oblique angle to
of performing the interpolation would fall upon the well-known the PEC wedge. This generates a cone of diffracted rays starting
and widely available commercial software MATLAB. from at an angle with respect to the edge. All the
observation point lying on this cone obeys the Keller’s law
of diffraction [9], . By choosing appropriate ray-fixed
VII. CONCLUSION coordinates, the diffracted-ray field is obtained as [5], [8]
We have shown that 3-D numerical diffraction coefficients for
(A1)
infinite right-angle material wedges can be efficiently calculated
using a generalized total-field/scattered-field FDTD technique. where the incident field components are
This approach permits modeling an infinite material wedge in-
side a compact FDTD grid which contains only the volume in (A2a)
the immediate vicinity of the diffracting corner. Our numerical (A2b)
validation results for the PEC wedge show very good correspon-
The uniform diffraction coefficients and are given by
dence with asymptotic UTD solution. Calculation of diffraction
(A3a) and (A3b) at the bottom of the next page, where
coefficients for the infinite right-angle material wedge having
, , is Fresnel’s transition function, and
arbitrary permittivity and/or loss is a trivial extension of the PEC
.
wedge case. For the results shown here, the wedge size is re-
duced by about 4:1 in each dimension relative to the conven-
APPENDIX B
tional TF/SF method. This yields an approximate 64:1 reduc-
3-D DIFFRACTION COEFFICIENTS DERIVED USING FDTD
tion in computer memory and running time. Finally, MATLAB
provides a powerful and straightforward means to compactly Fig.10(b) shows the 3-D geometry of the scatterer used in
store and interpolate the numerical diffraction coefficients over the FDTD modeling and the relationship between the FDTD
a broad range of propagation angles of the incident and scat- coordinate system and the ray-fixed coordinate system. This
tered waves. Future work involves extension and validation of figure shows the edge-fixed plane of incidence with the
the technique for nonright-angle wedges. ray-fixed unit vectors and parallel and perpendicular to it,
CHANG AND TAFLOVE: 3-D DIFFRACTION BY INFINITE CONDUCTING AND DIELECTRIC WEDGES 1453

Fig. 10. (a) 3-D geometry of an infinite right-angle wedge illuminated by a plane wave at oblique incidence. The edge-fixed plane of incidence and diffraction,
the ray-fixed coordinate system, and (b) the FDTD coordinate system are shown.

respectively. Also shown is the edge-fixed plane of diffraction and (A3b), respectively. This indicates that the FDTD-com-
with the ray-fixed unit vectors and parallel and per- puted diffraction coefficients, and , can be
pendicular to it, respectively. The edge-fixed spherical angles found using
made by the incident ray and the diffracted ray are and (B1a)
, respectively. The FDTD coordinate system ( , , )
is parallel to the three orthogonal edges of the scatterer. (B1b)
We find the diffracted-field impulse response of the scatterer where is the distance of the observation point from the scat-
numerically using FDTD. By illuminating the wedge with a tering edge and . The Fourier transform has been
pulsed plane wave having an electric field (E-field) component defined using the convention. The factor, , in the
parallel to the plane of incidence, we obtain the diffracted-field above equation arises from the nature of the Green’s function in
impulse response polarized parallel to the plane of two dimensions. and can be obtained using
diffraction. An analogous procedure is performed with the
incident E-field component perpendicular to the plane of inci-
(B2a)
dence, yielding polarized perpendicular to the plane of
diffraction. The Fourier transforms of these diffracted-field im-
pulse responses, and , give the corresponding
spectra of the diffracted fields. The diffracted-ray field given by (B2b)
the UTD in terms of the scalars and is shown in (A3a)

(A3a)

(A3b)
1454 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 53, NO. 4, APRIL 2005

where and are, re- [6] J. P. Berenger, “A perfectly matched layer for the absorption of electro-
spectively, the components of the diffracted field parallel and magnetic waves,” J. Computational Phys., vol. 114, pp. 185–200, 1994.
[7] J. Chang, “Diffraction models for accurate RF propagation predictions
perpendicular to the plane of diffraction. in urban environments,” Ph.D. dissertation, Northwestern Univ., 2004.
and are, respectively, the components of [8] C. A. Balanis, “Geometrical theory of diffraction,” in Advanced Engi-
neering Electromagnetics. New York: Wiley, 1989, ch. 13.
the incident field at the point of diffraction, parallel to and per- [9] J. B. Keller, “Geometrical theory of diffraction,” J. Opt. Soc. Amer., vol.
pendicular to the plane of incidence. Here, the angles 52, no. 2, pp. 116–130, Feb. 1962.
represent the direction of the incident plane wave illumination in
the FDTD coordinate system. Further, the angles repre-
sent the direction of the diffracted ray from the point of diffrac-
tion to the observation point . Jiuan-Her Chang received the B.S. degree in
physics from National Taiwan Normal University,
In order to compute the dot products in (B2), we express the Taiwan, R.O.C., in 1993, the M.S. degree in physics,
ray-fixed unit vectors and in terms of the FDTD and the Ph.D. degree in electrical and computer en-
unit vectors ( , , ) as follows: gineering from Northwestern University, Evanston,
IL, in 1998 and 2004, respectively.
Since 1999, she has been a Research Assistant in
the Electrical and Computer Engineering Department
at Northwestern University.
(B3a)
(B3b)

(B3c) Allen Taflove (F’90) was born in Chicago, IL, on

June 14, 1949. He received the B.S., M.S., and Ph.D.
(B3d) degrees in electrical engineering from Northwestern
University, Evanston, IL, in 1971, 1972, and 1975,
From (B1) and (B2), it is clear that the FDTD computed respectively.
diffraction coefficients are obtained using [7] After nine years as a Research Engineer at the IIT
Research Institute, Chicago, IL, he returned to North-
western in 1984. Since 1988, he has been a Professor
in the Department of Electrical and Computer Engi-
neering, McCormick School of Engineering, North-
western. Since 1972, he has pioneered basic theoret-
(B4a) ical approaches and engineering applications of finite-difference time-domain
(FDTD) computational electrodynamics. He coined the FDTD acronym in a
1980 IEEE paper, and in 1990 was the first person to be named a Fellow of IEEE
in the FDTD area. Currently, FDTD is one of the most powerful and widely used
methods for solving Maxwell’s equations to model linear and nonlinear elec-
tromagnetic wave interactions with electrically large and complex structures. In
(B4b) 1995, he authored a popular textbook on this subject Computational Electro-
dynamics: The Finite-Difference Time-Domain Method (Boston, MA: Artech
House, 1995), which is now in its second edition (2000), with a third edition
is planned for 2005. In total, he has authored or co-authored four books, 20
book chapters and magazine articles, 92 refereed journal papers, approximately
200 conference papers and abstracts, and 14 U.S. patents. These publications
REFERENCES
resulted in his being included in ISIHighlyCited.com, the Institute of Scien-
[1] V. Anantha and A. Taflove, “Efficient modeling of infinite scatterers tific Information’s compilation of the most-cited researchers worldwide. He has
using a generalized total-field/scattered-field FDTD boundary partially been the thesis adviser of 18 Ph.D. recipients who hold professorial or technical
embedded within PML,” IEEE Trans. Antennas Propag., vol. 50, no. 10, staff positions at major institutions including research universities and national
pp. 1337–1349, Oct. 2002. labs. Currently, he serves as the faculty Master of Northwestern’s 140-student
[2] A. Taflove and S. Hagness, Computational Electrodynamics: The Finite- Slivka Residential College of Science and Engineering. He also serves as the
Difference-Time-Domain Method. Boston, MA: Artech House, 2000. Faculty Advisor to Northwestern’s Undergraduate Design Competition, the stu-
[3] T. Martin and L. Pettersson, “Dispersion compensation for Huygens’ dent chapters of the Eta Kappa Nu and Tau Beta Pi engineering honor societies,
sources and far-zone transformation in FDTD,” IEEE Trans. Antennas and McCormick’s Honors Program in Undergraduate Research. His efforts on
Propag., vol. 48, no. 4, pp. 494–501, Apr. 2000. behalf of students at all levels were recognized by Northwestern when he was
[4] C. D. Moss, F. L. Teixeira, and J. A. Kong, “Analysis and compensation named a C. D. McCormick Professor of Teaching Excellence in 2000. His re-
of numerical dispersion in the FDTD method for layered, anisotropic search interests span much of the electromagnetic spectrum. He and his students
media,” IEEE Trans. Antennas Propag., vol. 50, no. 9, pp. 1174–1184, are currently modeling electrodynamic phenomena ranging from geophysically
Sep. 2002. induced extremely low frequency wave propagation about the entire Earth to
[5] R. G. Kouyoumjian and P. Pathak, “A uniform geometrical theory of the lasing behavior of aggregates of micron-scale zinc-oxide particles exhibiting
diffraction for an edge in perfectly conducting surface,” Proc. IEEE, vol. four-level quantum-system characteristics. The principle that “Maxwell’s equa-
62, no. 11, pp. 1448–1461, November 1974. tions work from dc to light” is vividly demonstrated in his laboratory every day.