Finite-Difference Time-

Domain (FDTD) Simulations

of Electromagnetic Wave
Propagation Using a
Spreadsheet
DAVID W. WARD, KEITH A. NELSON
Department of Chemistry, Massachusetts Institute of Technology, Cambridge MA 02139

Received 18 March 2005; accepted 18 May 2005

ABSTRACT: We describe a simple and intuitive implementation of the method of finite-

difference time-domain (FDTD) simulations for propagating electromagnetic waves using
Microsoft Excel. The method overcomes the usual obstacles of unfamiliarity with program-
ming languages as it relies on little more than the cut and paste features of Excel. ß2005 Wiley
Periodicals, Inc. Comput Appl Eng Educ 13: 213
221, 2005; Published online in Wiley InterScience
(www.interscience.wiley.com); DOI 10.1002/cae.20055

Keywords: electromagnetic simulation; FDTD; spreadsheets in education

INTRODUCTION studying and their intuitive understanding of wave

Introductory electricity and magnetism is a notor-
iously difficult subject for students and instructors
alike. Even students with strong analytical skills often
find little connection between the material they are
Correspondence to D. W. Ward (david@davidward.org).
Contract grant sponsor: National Science Foundation; contract
grant number: CHE-0212375.
Contract grant sponsor: MRSEC; contract grant number: DMR-
0213282.
Contract grant sponsor: Cambridge-MIT Institute; contract
grant number: CMI-001.
ß 2005 Wiley Periodicals Inc.
mechanics and propagation. As a result, they may not
derive the full benefits of this understanding in their
studies. Instructors also find frustration in the need to
cover a large body of material without the time or
resources required to guide students in their explora-
tion of electromagnetic waves with computational or
graphical tools that would build on and develop
further their intuition with the subject. Meanwhile, at
the advanced undergraduate and graduate levels, and
especially in research, such tools now play a central
role in generating and illuminating new results. The
benefits of electromagnetic wave simulation tools at
the introductory level would be manifest, but lack

213
214 WARD AND NELSON
of mathematical and programming skills generally
keeps such capabilities out of the hands of beginning
students.
Here we outline a simple demonstrative example
of the method of finite-difference time-domain
(FDTD) simulations of electromagnetic wave pro-
pagation appropriate for undergraduates in an intro-
ductory electricity and magnetism course or advanced
high school science students using Microsoft Excel,
a robust software package that offers advanced
graphics, numeric, and animation capabilities while
requiring minimal computer experience [1]. Though
it is commercial software, almost every student has
access to it. The algorithm is based on Maxwell’s
modifications to Faraday’s, Ampere’s, and Gauss’s
law, which are fundamental electromagnetic princi-
ples covered in freshman and high school physics
courses. FDTD enjoys the predictive power of
Maxwell’s equations without requiring students to
derive the wave equation, which requires a fair degree
of mathematical sophistication. The implementation
presented here further increases the accessibility to
students by removing the computer science element
from the algorithm, leaving a robust and facile tool
that empowers the student with the ability to tinker.
Further, since the algorithm is not limited to homo-
genous media, it allows students to explore cutting-
edge topics in physics and engineering like photonic
bandgap materials. Only two rows in the Excel
spreadsheet, one for initial electric field and another
for initial magnetic field, need be specified in order to
initiate the simulation, and graphs of the spatially and
temporally evolving waveforms can be updated in real
time as the simulation is carried out. We begin with a
review of the basic tenets of the FDTD method, pro-
ceed to an outline for the implementation in Microsoft
Excel, propose some suggested exercises, and con-
clude with a discussion on the results of employing the
spreadsheet algorithm in the classroom.
METHODS
FDTD Basics
The starting points for an FDTD simulation are
Maxwell’s equations, which are repeated here for the
case of one-dimensional free space propagation in
time (t) and space (z) with no sources or sinks for
magnetic or electric fields, B or E, respectively, and in
addition the constitutive relations defining the mag-
netic intensity H, and electric displacement D.
@Ex 1 @Hy
¼
ð1Þ
@t "0 @z
@Hy 1 @Ex
¼
ð2Þ
@t
0 @z
rE ¼0 ð3Þ
rB¼0 ð4Þ
Though, the equations themselves are probably
foreign to most students, the physical phenomona they
describe, however, are not. Thus, they are best pre-
sented to students by describing each one in terms of
the physical phenomoena it describes, that is, the
magnetic field induced by a current, electric fields
emanating from charge centers, etc. That propagating
solutions exist for this system of equations is not
easily shown. An entire lecture could be devoted to
deriving the wave equation; however, an intuitive
understanding of why there are such solutions can
be established more simply. The waves come about
because a charge moved somewhere. The way a
charge makes itself known to other charges in its
vicinity is through the electromagnetic force (Lorentz
force law). So every time a charge moves, a distur-
bance much like ripples on a pond, send word to
its neighbors so they do not bump into each other.
That the electromagnetic ripples do not need a pond
is very Zen. The waves are simply a consequence
of the first two Maxwell Equations 1 and 2. An
electric field induces a magnetic field, which induces
an electric field, ad infinitum. In general, FDTD is a
useful learning tool because it simulates waves pro-
pagating in the same causal manner as it really
happens.
If we choose the x-direction for the polarization
of the electric field and the z-direction for direction of
propagation, then it follows that the magnetic field is
y-polarized as indicated in the subscripts on the fields
above. To discretize the equations of propagation,
central difference approximations are taken for the
derivatives of time and space. Temporal steps are in-
dexed by the integer n and related to continuous time
by the relation t ¼ n
t, and spatial steps are indexed
by the integer k and related to continuous space by the
relation z ¼ k
z. The temporal discretization method
using central differences can then be designated for
variable X as
@X DX Xðn þ 1=2Þ
Xðn
1=2Þ
!  ð5Þ
@t Dt Dt
and spatial discretization as
@X DX Xðk þ 1=2Þ
Xðk
1=2Þ
!  ð6Þ
@z Dz Dz
 FDTD SIMULATIONS 215

In FDTD, we need only consider the two curl
Equations 1 and 2 above because the divergence
conditions 3 and 4 can be satisfied implicitly by in-
terleaving of the electric and magnetic field compo-
nents in space in what has come to be known as a
Yee cell [2]. A consequence of the spatial interleaving
is that the fields must also be interleaved in time,
known as ‘‘leapfrog,’’ since the temporal response of
one field is proportional to the spatial variation of the
other at the previous time step. Implementing
Equations 1 and 2 as indicated in Equations 5 and 6
and interleaving spatially and temporally yields the
difference equation version of Maxwell’s equations:
where c0 is the speed of light in vacuum. If the fields
start to diverge or other pecularities arise, the culprit is
usually in the choice of space and time step sizes.
The computer algorithm implementation of this
procedure, based on Equations 9 and 10, requires two
one-dimensional arrays in the spatial coordinate, one
for Ex and one for Hy, to be allocated in memory. After
imposing an initial condition, in the form of an initial
electric and magnetic field, each time step is calcu-
lated by the following prescription:
Dt
t ¼ t0 þ n
2



~ x ½k þ 1 fHy ½k
1
Hy ½kg
1 1
nþ n
2 ~ x ½k ¼ E
Ex 2 ðkÞ
Ex ðkÞ Hyn k þ 12
Hyn k
12 E
¼
ð7Þ 2
Dt "0 Dz
Dt


 nþ
1
nþ
1 t ¼ t0 þ n
Hynþ1
k þ 2
Hyn k þ 12
1
Ex 2 ðk þ 1Þ
Ex 2 ðkÞ 2
¼

Dt
0 Dz 1 ~
Hy ½k ¼ Hy ½k þ fE ~
x ½k
Ex ½k þ 1g ð13Þ
ð8Þ 2
At a given position in space, the field at each current where we have adopted the normalized fields
time step can be calculated from the previous values ~ ¼ pffiffiffiffiffiffiffiffiffi
(E "0
0 E) to simplify the code, and included
of the fields. Solving for the latest field at the latest the stability conditions, Equations 11 and 12, within
time step in Equations 7 and 8 yields: the constant preceding the curl [4]. The algorithm
(Eq. 13) is iterated for the desired number of time
1 1
nþ n
Dt steps. Note that iteration of Equation 13 results
Ex 2 ðkÞ ¼ Ex 2 ðkÞ

"0 Dz in an implicit time formulation, that is time is not
    
1 1 made explicit in the equations for the field and only
Hyn k þ
Hyn k
ð9Þ appears in the algorithm above for bookkeeping
2 2
purposes.
    1
1 1 Dt nþ
Hynþ1 k þ ¼ Hyn k þ
fEx 2 ðk þ 1Þ
2 2
0 Dz FDTD Spreadsheet Algorithm
1
nþ

Ex 2 ðkÞg ð10Þ The FDTD methodology for one dimension can be
implemented in spreadsheet format, executed here in
In this fashion, known as Euler forward, the Microsoft Excel, using simple cell formulas and cut-
solution of Maxwell’s equations proceeds in much and-paste features. This has an advantage over other
the same way that we envision electromagnetic wave pedagogical approaches [5,6,7] in that no program-
propagation—an electric field induces a magnetic ming experience is required. The starting point is the
field, which induces and electric field, ad infinitum. algorithm in Equation 13 where columns represent
The time step and grid size parameters are chosen spatial steps and pairs of rows (one for E and one for
based on the propagating wave frequency and wave- H) represent time steps. The algorithm is illustrated
length. For stability, the general rule is that at least ten graphically in the center section of Figure 1. Since the
grid points sum to less than the smallest wavelength electric field value from one prior cell is needed to
lmin present in the simulation, and the Courant compute the curl, Ex begins at column one and ends at
condition then determines the time step [3]: column k þ 1. Similarly, the magnetic field begins
at column zero but ends at column k, since its value at
min one subsequent cell is needed when computing the
Dz  ð11Þ
10 curl. The first two rows (t ¼ 0) are for the initial
conditions. The algorithm computes the remaining
Dz cells. The student need only type in the formulas for E
Dt  ð12Þ
2c0 and H for the second time step and the first two
216 WARD AND NELSON

Figure 1 Graphical illustration of the spreadsheet implementation of FDTD. Center:

Components of the spreadsheet fields and a graphical map of the field update de-
pendencies. Exterior: Illustration of the implementation steps in the text.

columns; cut and paste may be used to complete the
remaining spatial columns and temporal rows.
Though a recipe is given below, the best way
to illustrate the algorithm is visually with a small
problem space, that is few rows and columns. The
electric field is always up and to the left. The magnetic
field is always up and to the right. Displaying a
representation similar to that in the figure and tracing
through the mediating cells several times serves to
internalize the algorithm in the student. Highlighting
the relevant cells also facilitates this, exploiting the
pattern recognition capabilities of the human brain.
The recipe for implementing the algorithm is as
follows:
(1) As a visual aid, enter the spatial coordinate
(1, 2, 3, etc.) in the topmost row of the
spreadsheet starting at column B. After typing
in the first two spatial coordinates, select
cells B and C and drag the ‘‘fill’’ handle to
encompass as large a problem space as
needed.
(2) Type ‘Ex0’ in cell A2 and ‘Hy0’ in cell A3,
where the number following the field compo-
nent refers to the time step. Drag the ‘‘fill’’
handle down the column, as in 1, to include
the desired number of time steps.
(3) As a visual aid, select and highlight the
leftmost and rightmost columns of the pro-

blem space with the ‘‘fill color’’ color tab on
the Excel toolbar.
(4) Fill in all of rows 2 and 3 (time step 0) to the
end of the problem space with zeros and
highlight to indicate that this is the initial
condition for the fields.
(5) In the leftmost column and first time step for
Ex, type ‘‘¼B2.’’ The cell will take on the
value stored in cell B2. Do the same for the
rightmost column, but use the rightmost
column letter instead of ‘B.’ This imposes a
perfect electric conductor (PEC)-radiation
boundary condition on the problem space,
which means that impinging waves will re-
flect from the problem space boundaries as
they would from a mirror.
(6) In the second spatial position of Ex for
time step one (cell C4), type the formula
‘¼C2þ0.5(B3
C3)’ and press enter.
(7) Select the cell from the previous step and drag
the ‘‘fill’’ handle to the last un-highlighted
spatial column in the row.
(8) In the first spatial position of Hy for time
step one (cell B5) type the formula
‘¼B3þ0.5(B4
C4)’ and press enter. These
last three steps define the computing steps of
the algorithm.
(9) Select the cell from the previous step, and
drag the ‘‘fill’’ handle to the last un-high-
lighted spatial column in the row.
(10) Finally, select all the columns for time step 1
(Ex and Hy ), and drag the ‘‘fill’’ handle
through the last time step in the simulation.
Test that it works by entering a 1 in the initial
condition regions and see that all cells auto-
matically update; note that the only valid
regions for initialization are at the zero
time step (Ex and/or Hy ) and between the
first and last columns of the problem space
(highlighted).
The FDTD Excel code will now update all
fields in response to the initial conditions entered
by the user. To graphically visualize the simula-
tion output, a single row can be selected and
graphed using the insert > chart > (xy)scatter menu
item. This outputs the spatial field pattern at the
moment in time designated by the row number. To
graph temporal evolution at a single point in space,
it is recommended to make a new field mesh with
onlyone field component (either Ex or Hy ) on a
separate sheet and graph from that; otherwise, both
the E and H fields will appear in the graphs. This can
be done using cut and paste. An example spatial graph
FDTD SIMULATIONS 217
of two counter propagating pulses is illustrated in
Figure 3.
Once the algorithm is in place, it is advisable to let
students experiment by introducing a single electric
field value in the initial conditions row. This simulates
the sudden appearance of a charge, as if it swept
through the plane in the simulation in a single time
step, leaving only the spatial extent of its field as proof
that it was ever there. Students can be told to think of
it like dropping a stone in a pond. The electromagnetic
disturbance can be visualized directly in the subse-
quent rows of the spreadsheet or by outputting the cell
contents to a graph. The simulations in its raw form,
may be confusing if they do not understand that one
axis is space and the other time. Graphing the output
and comparing with the row/column content seems to
help. Ensure that they pay close attention to the
reflections on the left and right hand side of the
problem space. If the graphical algorithm is inter-
nalized, they should quickly recognize why one
reflection is inverted and the other is not. This is
discussed more below.
APPLICATIONS OF THE ALGORITHM
Power Flow
The direction of propagation of light is dictated by
the phase between the electric and magnetic fields.
The familiar right-hand rule, applied by curling E into
H, indicates the direction of power flow. Using our
spreadsheet code, this can be demonstrated by
initializing both the electric and magnetic fields as
sinusoidal waveforms modulated by a Gaussian
envelope. Introducing a phase shift of either 0 or p
in the magnetic field allows the direction of propaga-
tion to be controlled. If the electric (Ex ) and magnetic
(Hy ) fields are both positive at the same points in space
and time, then the pulse travels in the positive z
direction; if the signs are opposite, then propagation is
in the opposite direction as illustrated in Figure 2. It is
advisable to introduce the topic of power flow by
drawing the electric and magnetic component of a
light wave and illustrating the consequences on the
area of overlap between them when the electric and
magnetic field become out of phase. Mention of the
extreme case of no power flow when they are com-
pletely out of phase and unattenuated power flow
when they are completely in phase is helpful in
making this point. Experimentation with the phase
of the two components with FDTD can reinforce the
concepts of wave attenuation and directionality of
propagation.
218 WARD AND NELSON

Figure 2 Illustration of the relationship between the phases of E and H and the direction
of propagation. (a) Left-traveling pulse generated by introducing a p=2 phase shift in the
initial H field with respect to E. (b) Right-traveling pulse generated by initializing E and
H in phase with each other. Both graphs show the fields after the same (arbitrary) number
of time steps after generation. The smaller pulses in (a) and (b) are remnants of imperfect
initial conditions. The arrow indicates the direction of propagation and the dotted line
down the center demarks the starting point of the pulse.

Resonator Modes nodes at the boundaries, as illustrated in Figure 3.

Some of the most interesting applications of electro-
dynamics occur when constraints are placed on the
fields. The PEC radiation boundary conditions form
the walls of a resonator when the wavelength in the
problem space (resonator width) is of similar size to
the problem space itself. With PEC walls, no light
wave with a wavelength longer than twice the re-
sonator width will propagate. This is referred to as
a cutoff wavelength. Also, since the electric field
must be zero at the boundary and only nodes in a
plane wave may satisfy this, only wavelengths that are
half-integer multiples of the resonator width can
propagate. Resonator modes can be investigated by
setting the initial conditions to a plane wave with
Seeing the destructive interference responsible for
wave-guide cutoff is off considerable value to the
students intuition.
Transmission and Reflection
By introducing a relative permittivity "r into Equation
9, which changes the index of refraction and hence
the wave propagation speed, the spreadsheet code
can illustrate reflection and transmission at a material
interface. In terms of the normalized fields, the
constant 1=2 becomes 1=2er . Small errors in the
transmitted and reflected fields are to be expected, but
these will decrease as the number of mesh points per
wavelength increases. Figure 4 illustrates this pro-
cess for an air (er ¼ 1)/material (er ¼ 2) interface.
Students are often confused that the electric field
increases when passing from a high to low index
material, or decreases for the converse case. Simulat-
ing this allows them to gain some physical insight,
realizing that this is not so much related to the
terse radiation boundary conditions derived from
Maxwell’s equations, as it is simply related to whether
or not the reflection inverts at the interface. To aid
their understanding, the analogy of wave reflections in
thick and thin strings or a string bound to a wall may
be introduced.

Figure 3 Lowest energy symmetric and anti-sym- FURTHER EXERCISES AND

metric (bold) cavity mode for a 50 unit wide resonant SUGGESTED ACTIVITIES
cavity with perfect electric conductor (PEC) walls.
Figures display the spatial distribution of the electric The spreadsheet code for the examples is available
and magnetic fields in the cavity after 200 time steps. online [8], but it is recommended that students

Figure 4 Incident wave approaching an interface
(dotted line), and transmitted and reflected waves
moving away from it. Time is in units of integer
time steps. The field amplitudes are consistent with
the transmission (T) and reflection (R) coeffici-
pffiffiffiffi pffiffiffiffi pffiffiffiffi pffiffiffiffi pffiffiffiffi
ent T ¼ ð2 e1 Þ=ð e1 þ e2 Þ, R ¼ ð e1
e2 Þ=
pffiffiffiffi pffiffiffiffi
ð e1 þ e2 Þ, where in each medium the index of
pffiffiffiffi
refraction is n ¼ er .
augment the code on their own because writing and
tinkering with the code is the best way to understand
how it works. The following recommended exercises
are listed in order of increasing difficulty, or in a
manner such that each exercise depends only on
changes already administered. These exercises not
only serve to reinforce the understanding of the FDTD
algorithm, but also to foster understanding of funda-
mental electromagnetic phenomena.
Fundamental
(1) Use trigonometric and exponential mathema-
tical functions to generate the initial conditions
and reproduce the results in Figure 2.
(2) Study the cavity modes of a one-dimensional
resonator, Figure 3, by introducing standing
waves into the initial conditions. What happens
when the wavelength becomes longer than
the problem space size (the width of the
resonator)?
(3) Experiment with radiation boundary condi-
tions. Change the radiation boundary condition
from PEC to perfect magnetic conductor (Hy is
zero at the problem space boundaries analo-
gous to a PEC). What is different about
reflections from these two kinds of surfaces?
(4) Implement a periodic boundary condition, in
which waves exiting through the edge of the
problem space on one side, emerge from the
opposite edge back into the problem space.
FDTD SIMULATIONS 219
(5) Reproduce the reflection and transmission simu-
lation in Figure 4 and then change the interface
from air/material to material/air. What changes?
Advanced
(1) Experiment with spatially periodic variation in
the relative permittivity. This should produce
frequency-dependent filtering, and for some
conditions should produce ‘‘photonic band-
gap’’ materials with no propagation in certain
frequency regions.
(2) Introduce loss into the simulation by including
non-zero conductivity in Maxwell’s equations
and implementing it into the FDTD code.
(3) Implement an absorbing boundary condition, in
which waves exiting through the edge of the
problem space pass through without reflecting
and without re-entering the problem space.
CONCLUSION
The spreadsheet algorithm described above was used
in an independent activities period (IAP) course on
electromagnetic wave simulation. Most of the student
participants were not yet acquainted with Maxwell’s
equations or the electromagnetic wave equation,
though they were familiar with Ampere’s, Faraday’s,
and Coulomb’s laws. Since IAP is not required for the
fulfillment of any degree plan at MIT, academic topics
usually do not attract a large number of students.
Instead most students use the time either for relaxation
or to learn a new hobby, which is the focus of most
of the IAP courses. The course was advertised by
posting fliers on campus with ‘‘Are you a visual
thinker?’’ printed prominently at the top. Several
fliers were used, each featuring a different current
application of electromagnetic wave propagation, for
example photonic bandgap materials, negative re-
fraction, optical guided-wave circuits, etc. Similar
messages were disseminated through email to fresh-
men pre-registered for introductory electricity and
magnetism courses in the upcoming term. Student
turnout for the IAP course was almost ten times larger
than is typical. Most of them, 80%, were freshmen.
The remainder was composed of undergraduate
and graduate students in electrical engineering and
physics. Part of the reason for the high turnout is
likely due to the current popularity of photonics in
academics and industry. Surveys conducted at the end
of the course, however, indicated that the strongest
motivation was the focus on visual thinking. The
survey also indicated a high satisfaction level with the
220 WARD AND NELSON
understanding gained through the ‘visual thinking’
pedagogical approach.
The spreadsheet algorithm was only used on the
first day of the course—a MATLAB implementation
of the algorithm was used for the remainder. Past
experiences with teaching FDTD through the usual
code-intensive implementations of the algorithm
indicate that the benefit of starting with the spread-
sheet is an increase in understanding of the algorithm
itself, especially with regard to the staggering of the
electric and magnetic fields. Students that learned by
the visual algorithm do not seem to encounter the
same difficulty in implementing the staggering in C or
MATLAB-based FDTD as do students that started
with the code-intensive algorithm. The latter tend to
randomly permute the starting and ending index of
the spatial field arrays until a reasonable output is
obtained—mindlessly tweaking the code until an
acceptable answer appears. This becomes a serious
problem when a spatially dependent material para-
meter is introduced because then instead of just the
problem space boundaries there are boundaries at
every material interface. A problem unique to those
introduced through the spreadsheet is a tendency to
stick with the explicit formulation of time by
maintaining an array for it in memory rather than
using it implicitly with a for loop for time. This can be
overcome by simply pointing out how one implements
implicit time.
At the end of the course, a student evaluation
form was issued. Among other things, the students
were asked to evaluate the effectiveness of the
instructional software on a scale of 1 to 7 (1 being
worst and 7 being best). The class average among the
32 respondents (of a class of 40) was 5.7, where
the lowest score assigned was a 4 and the highest a 7.
The students were also specifically asked about how
useful the spreadsheet algorithm was to them in the
course, on the same scale. The class average was 6.3
with the lowest score being 5 and the highest 7.
Written feedback, in the form of general comments,
was also requested. Over 85% of the responses
mentioned something positive about the visual aspect
of the spreadsheet, with only one responding that they
did not find it beneficial.
In conclusion, we have formulated a simple
implementation of the FDTD method of propagating
electromagnetic waves using basic features available
in Microsoft Excel, and we have presented some
illustrative examples. The implementation was tested
on freshmen at MIT during its 2004 Independent
Activities Period in a short course [8] that met for 2 h
on each of 2 consecutive days. This brief introduc-
tion was sufficient for students to gain substantial
proficiency in elementary simulations like those
illustrated above. The class enjoyed unusually high
attendance, especially among students preparing for
or recently involved in electromagnetism coursework.
Instructor (D.W.W.) observation and student feedback
after the course indicated considerable satisfac-
tion with insights and proficiency gained from the
visual approach offered by the spreadsheet algorithm.
Several student responses pinpointed the graphical
form of the spreadsheet-based algorithm as the key
merit of the approach. The benefit of keeping track of
a small number of unit cells in order to see how the
algorithm works is useful for beginning and inter-
mediate students of computer simulation, numerical
recipes, and electricity and magnetism. The automatic
updating of Excel graphs makes this an effective
learning tool for students interested in the basics
of electromagnetic wave propagation, as the conse-
quences of changes in initial conditions are illustrat-
ed graphically and instantly. Use of the graphical
algorithm could be used to introduce the notion of
electromagnetic wave propagation without having to
derive the wave equation. All that is needed are the
fundamental physical laws manifest in Maxwell’s
equations and a basic understanding of a derivative
and an integral. Knowledge of differential equations is
not necessary. This approach is also useful in teaching
FDTD, which is a powerful tool in electrical engi-
neering. The methodology learned in this rapid
introduction may stimulate student interest in more
advanced FDTD simulation techniques and their
broad research applications [9,10].
ACKNOWLEDGMENTS
This work was supported in part by the National
Science Foundation Grant no. CHE-0212375 and
MRSEC award no. DMR-0213282, and by the
Cambridge-MIT Institute Grant no. CMI-001.
REFERENCES
[1] L. Webb, Spreadsheets in physics teaching, Phys Educ
28 (1993), 77
82.
[2] K. S. Yee, Numerical solution of initial boundary value
problems involving Maxwell’s equations in isotropic
media, IEEE Trans Antennas Propagat, AP-14 (1966),
302
307.
[3] A. Taflove and S. Hagness, Computational electro-
dynamics: The finite difference time domain method,
2nd ed., Artech House, Boston, 2000.
[4] D. M. Sullivan, Electromagnetic simulations using the
FDTD method, IEEE Press, New York, 2000.

[5] R. J. Luebbers, K. S. Kunz, and K. A. Chamberlin,
An interactive demonstration of electromagnetic wave
propagation using time-domain finite differences,
IEEE Trans Educ 33 (1990), 60
68.
[6] A. W. Glisson and A. Z. Elsherbeni, An interactive 1D
Matlab FDTD code for education, Comput Appl Eng
Educ 9 (2001), 136
147
[7] W. K. Leung, Y. Chen, and Y. S. Lee, EM wave
simulation: An animated electromagnetic wave teach-
BIOGRAPHIES
David W. Ward received his BS degree in
physics in 1999 from the College of Charles-
ton and his PhD in physical chemistry from
the Massachusetts Institute of Technology in
2005. Dr. Ward is a cofounder of the field of
polaritonics, which is an intermediate fre-
quency regime between electronics and
photonics where signals are manifest as
admixtures of photons and phonons. He is
presently a postdoctoral fellow at Harvard University. His interests
are in terahertz technology, negative refraction, nanophotonics,
single molecule spectroscopy, and finite-difference time-domain
and molecular dynamics simulations. Dr. Ward received the
Qauttrochi scholarship as an undergraduate and was a finalist for
the Truman Fellowship. He is a member of the American Physical
Society, the Optical Society of America, and the Materials Research
Society. His teaching experience ranges from high school to
graduate level education.
FDTD SIMULATIONS 221
ing package, Comput Appl Eng Educ 9 (2001),
208
219.
[8] http://nelson.mit.edu/iap2004.html
[9] D. W. Ward, E. R. Statz, N. S. Stoyanov, and K.A.
Nelson, Simulation of Phonon-Polariton Propagation in
Ferroelectric LiNbO3 Crystals, in Engineered Porosity
for Microphotonics and Plasmonics, MRS Symposium
Proceedings, Vol. 762, 2003, pp. c11.60.1
c11.60.6.
[10] http://fdtd.davidward.org/
Keith A. Nelson received the BS degree in
chemistry and PhD in physical chemistry in
1976 and 1981 from Stanford University. He
then did postdoctoral work at UCLA. He has
been on the faculty of the MIT Department
of Chemistry since 1982. His main interests
are in time-resolved spectroscopy of struc-
tural and chemical rearrangements in con-
densed materials, coherent control and pulse
shaping, and terahertz generation and control. Professor Nelson has
received a Presidential Young Investigator Award, Alfred P. Sloan
Fellowship, Coblenz Prize, and Japan Society for Promotion of
Science Fellowship. He is a fellow of the American Physical
Society.