1/13/2020

Computational Science:
Introduction to Finite-Difference Time-Domain

Review of Electromagnetics
and Introduction to FDTD

Lecture Outline
• Review
• Maxwell’s equations
• Physical boundary conditions
• The constitutive relations
• Parameter relations
• see Balanis Chapter 1
• Introduction to FDTD
• Flow of Maxwell’s equations
• Finite-difference approximations
• The update equation
• The FDTD algorithm…for now 
Slide 2

1
 1/13/2020

Review of Maxwell’s
Equations and
Electromagnetics

Slide 3

Maxwell’s Equations

2
 1/13/2020

Four Field Terms in Maxwell’s Equations

    
  H  J  D t   E   B t


Electric Field Quantities

E  V m
+ = D
Electric field intensity
Initial electric push
Displaced Charges C m  2


Magnetic Field Quantities

H  A m

Magnetic field intensity

+ = B
Initial magnetic push
Tilted Magnetic Dipoles
 Wb m  2

Slide 5

Gauss’s Law
  Dx Dy Dz
  D  v D 
x

y

z

Electric fields diverge from positive charges and

converge on negative charges.

If there are no charges,

+ - electric fields must form loops.

3
 1/13/2020

Gauss’s Law for Magnetism

  B By Bz
B  0 B  x 
x y

z

Magnetic fields always form loops.

Consequence of Zero Divergence

The divergence theorems force the 𝐷 and 𝐵 fields to be perpendicular to the propagation direction of a plane wave.


D  0
 
 
  de  jk r  0
 
   k D
  d  jk  d  0 
no charges
k
 

k d  0 k


B  0
 
 
  be  jk r  0
 
   kB
  b  jk  b  0 
k
no charges
  
k b  0 k
8

4
 1/13/2020

Ampere’s Law with Maxwell’s Correction


  D   H z H y   H x H z   H y H x 
 H  J   H  
 y

z
 aˆ x  
  z

x  aˆ y   x  y
 
 aˆ z

t
Circulating magnetic fields induce currents and/or time varying electric fields. Currents
and/or time varying electric fields induce circulating magnetic fields.

Faraday’s Law of Induction


 B   E E y   Ex E z   E y Ex 
 E    E   z 
 y
 aˆ x  
z   z

x  aˆ y   x  y  aˆ z
  
t
Circulating electric fields induce time varying magnetic fields.
Time varying magnetic fields induce circulating electric fields.

10

5
 1/13/2020

How Waves Propagate

Start with an
oscillating
electric field.

Slide 11

How Waves Propagate

This induces a
circulating
magnetic field.
 
  H  j E

Slide 12

6
 1/13/2020

How Waves Propagate

Now let’s
examine the
magnetic field
on axis.

Slide 13

How Waves Propagate

This induces a  
circulating
electric field.
  E   j H

Slide 14

7
 1/13/2020

How Waves Propagate

Now let’s
examine the
electric field
on axis.

Slide 15

How Waves Propagate

This induces a
circulating
magnetic field.
 
  H  j E
Slide 16

8
 1/13/2020

How Waves Propagate

…and so on…

Slide 17

Starting Point for Electromagnetic Analysis

Divergence Equations

Curl Equations
  D
  H  J 
B  0 t
  What produces fields

  D  v  B
 E  
t

Constitutive Relations
 
D  t     t    E  t 
   means convolution How fields interact
   means tensor with materials

B  t      t    H  t 
18

9
 1/13/2020

Maxwell’s Equations in Cartesian Coordinates (1 of 4)

Vector Terms
  
E  Ex aˆ x  E y aˆ y  Ez aˆ z H  H x aˆ x  H y aˆ y  H z aˆ z J  J x aˆ x  J y aˆ y  J z aˆ z
 
D  Dx aˆ x  Dy aˆ y  Dz aˆ z B  Bx aˆ x  By aˆ y  Bz aˆ z

Divergence Equations
 
 D  0  B  0
Dx Dy Dz Bx By Bz
  0   0
x y z x y z

Slide 19

Maxwell’s Equations in Cartesian Coordinates (2 of 4)

Constitutive Relations
 
D    E

Dx aˆ x  Dy aˆ y  Dz aˆ z    xx E x   xy E y   xz Ez  aˆ x   yx E x   yy E y   yz E z  aˆ y    zx Ex   zy E y   zz Ez  aˆ z

Dx   xx Ex   xy E y   xz Ez
Dy   yx E x   yy E y   yz Ez
Dz   zx Ex   zy E y   zz Ez

Bx   xx H x   xy H y   xz H z
 
B    H By   yx H x   yy H y   yz H z
Bz   zx H x   zy H y   zz H z
Slide 20

10
 1/13/2020

Maxwell’s Equations in Cartesian Coordinates (3 of 4)

Curl Equations

 B
 E  
t
 E z E y   E x Ez   E y Ex   
   aˆ x     aˆ y     aˆ z    Bx aˆ x  By aˆ y  Bz aˆ z 
 y z   z x   x y  t  

 E z E y   E x Ez   E y Ex  Bx By B

   aˆ x     aˆ y     aˆ z   aˆ x  aˆ y  z aˆ z
 y z   z x   x y  t t t

E z E y B E y Ex B
  x E x E z By   z
y z t   x y t
z x t

Slide 21

Maxwell’s Equations in Cartesian Coordinates (4 of 4)

Curl Equations

  D
 H  J 
t
 H z H y   H x H z   H y H x  
   aˆ x     aˆ y     aˆ z   J x aˆ x  J y aˆ y  J z aˆ z    Dx aˆ x  Dy aˆ y  Dz aˆ z 
 y z   z x   x y  t

 H z H y   H x H z   H y H x   Dx   Dy   Dz 

   aˆ x     aˆ y     aˆ z   J x   aˆ x   J y   aˆ y   J z   aˆ z
 y z   z x   x y   t   t   t 

H z H y D H y H x D
  Jx  x H x H z Dy   Jz  z
y z t   Jy  x y t
z x t

Slide 22

11
 1/13/2020

Alternative Form of Maxwell’s Equations in Cartesian

Coordinates (1 of 2)
Alternate Curl Equations

 E
  H   
t
 H z H y   H x H z   H y H x   Ex E y E 
   aˆ x     aˆ y   x  y  aˆ z    xx   xy   xz z  aˆ x
 y z   z x      t t t 
 E E y E 
   yx x   yy   yz z  aˆ y
H z H y E E y E  t t t 
   xx x   xy   xz z
y z t t t  E E y E 
   zx x   zy   zz z  aˆ z
H x H z E E y E   t  t t 
   yx x   yy   yz z
z x t t t
H y H x E E E
  zx x   zy   zz z
y

x y t t t

Slide 23

Alternative Form of Maxwell’s Equations in Cartesian

Coordinates (2 of 2)
Alternate Curl Equations

 H
  E   
t
 Ez E y   Ex Ez   E y Ex   H x H y H z 




 aˆ x     aˆ y   x  y  aˆ z     xx t   xy t   xz t  aˆ x
 y z   z x     
 H x H y H z 
   yx   yy   yz  aˆ y
Ez E y H x H y H  t t t 
    xx   xy   xz z
y z t t t  H x H y H z 
   zx   zy   zz  aˆ z
Ex E z H x H y H z  t t t 
    yx   yy   yz
z x t t t
E y Ex H x H H z
   zx   zy   zz
y

x y t t t

Slide 24

12
 1/13/2020

Tensors
Tensors are a generalization of a scaling factor where the direction of a vector can be
altered in addition to its magnitude.

 
Scalar Relation  V aV

 
V  
a V  Tensor Relation

a axy axz  Vx 

  xx 
 a V  a yx a yy a yz  Vy 
 
 azx azy azz  Vz 

Slide 25

The Constitutive Relations

Linear, isotropic and non-dispersive materials:
 
D t    E t 
We will use this in our FDTD codes
Dispersive materials:
 
D t    t   E t  The point here is that you can wrap all of the
complexities associated with modeling
strange materials into this single equation.
The implementation of Maxwell’s equations in
Anisotropic materials:
  FDTD will never change. This will make your

D  t     E  t  code more modular and easier to modify. It

may not be as efficient as it could be though.

Nonlinear materials:
D  t    0  e  E  t    0  e  E 2  t    0  e  E 3  t  
1 2 3

26

13
 1/13/2020

Anisotropic Materials
A generalized tensor for permittivity is written as

 xx  xy  xz   ij  how much of E j contributes to Di

  
D  t    yx  yy  yz  E  t 
We see that E and d can be in different directions
 zx  zy  zz 
 when the permittivity is anisotropic.

It greatly simplifies a finite-difference method to consider only diagonal tensors.

That is, all of the off-diagonal terms will be set to zero.

 xx 0 0 Dx  t    xx Ex  t 
 
D  t    0  yy 0  E  t   D y  t    yy E y  t 
 0 0  zz  Dz  t    zz Ez  t 

Special Note: There are only three degrees of freedom for the tensor components. The nine elements cannot be chosen arbitrarily. It is
always possible to choose a coordinate system that makes the tensor diagonal. The off diagonal terms only arise when the chosen
coordinate system does not match the crystal axes of the anisotropic material. The simplification above restricts us to only be able to
model anisotropic materials that align perfectly with our x, y, and z axes. 

27

Simplifying Maxwell’s Equations


1. Assume no charges or current sources: v  0 J 0
    
B  0   H  D t D  t     t    E  t 
    
D  0   E   B t B  t      t    H  t 

Convolution becomes
2. Assume linear, isotropic, and non-dispersive materials: simple multiplication
    
B  0   H  D t D t     E t 
    
D  0   E   B t B t     H t 

3. Sometimes the constitutive relations are substituted into Maxwell’s equations:


  E
    H  t   0   H   Note: It is helpful to retain μ and ε and not replace
t
  them with refractive index n.
   E  t   0  H
  E  
t
28

14
 1/13/2020

Physical Interpretation of and

𝐸 – Electric Field
• A disturbance produced around charges or in
the presence of a time-varying magnetic field. V + d
• Think of 𝐸 as a “push”
-
• Units are volts per meter (V/m) V
E
d
𝐷 – Electric Displacement Field
• “D” stands for displacement
• Includes 𝐸 , but also accounts for displaced charges in a material (material polarization)
• Equivalent to flux density
• Think of 𝐷 as displaced charge
• Units are dipole moments per unit
volume (C·m/m3), or just (C/m2)
• We can make 𝐸 look like an equivalent displaced
charge through 𝐷 = 𝜀 𝐸 .

29

Physical Interpretation of and

𝐻 – Magnetic Field
• A disturbance produced around currents or in
the presence of a time-varying electric field.
• Think of 𝐻 as a magnetic “push”
• Units are amperes per meter (A/m)
𝐵 – Magnetic Displacement Field
• Includes 𝐻 , but also accounts for tilted magnetic dipoles in a material (magnetization)
• Equivalent to flux density
• Think of 𝐵 as reoriented magnetic dipoles
• Units are magnetic dipole moments per unit volume (W·m/m3), or just (W/m2)
• We can make 𝐻 look like an equivalent reoriented dipole through 𝐵 = 𝜇 𝐻.

30

15
 1/13/2020

Physical Boundary Conditions

1 and 1  2 and  2

E1,T E2,T
Tangential components of E and H
are continuous across an interface.
H1,T H 2,T

Fields normal to the interface are

E  2 E2,N
discontinuous across an interface. 1 1,N These are more complicated
Note: Normal components of D boundary conditions,
and B are continuous across the 1 H1,N 2 H 2,N physically and analytically.
interface.

Tangential component of the wave

vector is continuous across an k1,T k2,T
interface.
31

Parameter Relations

Slide 32

16
 1/13/2020

The Dielectric Constant, r

The permittivity 𝜀 is a measure of how well a material stores electric energy. A
circulating magnetic field induces an electric field at the center of the
circulation in proportion to the permittivity.

 E
 H  
t

The dielectric constant 𝜀 of a material is its permittivity relative to the

permittivity of free space 𝜀 .

   0 r  0  8.854187817 10 12 F m

1  r   r is the relative permittivity or dielectric constant

Slide 33

Table of Dielectric Constants

Constantine A. Balanis, Advanced Engineering Electromagnetics, Wiley, 1989.

Slide 34

17
 1/13/2020

The Relative Permeability, r

The permeability 𝜇 is a measure of how well a material stores magnetic
energy. A circulating electric field induces a magnetic field at the center of the
circulation in proportion to the permeability.

 H
  E  
t

The relative permeability 𝜇 of a material is its permeability relative to the

permeability of free space 𝜇 .
0  1.256637061 106 H m
  0 r
1  r   r is the relative permeability

Slide 35

Table of Permeabilities

Constantine A. Balanis, Advanced Engineering Electromagnetics, Wiley, 1989. Slide 36

18
 1/13/2020

The Refractive Index

The permittivity and permeability appear in Maxwell’s equations so they are the most
fundamental material properties. However, it is difficult to determine physical meaning
from them in terms of how waves propagate (i.e. speed, loss, etc.). In this case, the
refractive index is a more meaningful quantity.

n  r r
Most materials exhibit a negligible magnetic response and the refractive index and dielectric
constant are related through

n2   r Hint: one of the most common mistakes made in this course is

using values of refractive index directly as permittivity.

Slide 37

Material Impedance
The impedance 𝜂 of a material quantifies the relation between the electric and magnetic
field of a wave travelling through that material. It is the most fundamental quantity that
causes reflections and scattering.
 
 E H
The impedance can be written relative to the free space impedance as

r 0
  0 0   376.73031346177 
r 0
This shows that the electric field is around two to three orders of magnitude larger than the
magnetic field.
Slide 38

19
 1/13/2020

versus
𝜔 is the angular frequency measured in radians per second. It relates more
directly to phase and 𝑘. Think cos 𝜔𝑡 .

𝑓 is the ordinary frequency measured in cycles per second.

It relates most directly to time 𝑡. Think cos 2 𝜋𝑓𝑡 and 𝜏 = 1/𝑓.

  2 f
Slide 39

Wavelength and Frequency

The frequency f and free space wavelength l0 are related through

c0  f l0 c0  299792458 m
s
 speed of light in vacuum

Inside a material, the wave slows down according to the refractive

index as follows.
c0
v
n
The frequency is the most fundamental parameter because it is fixed. Inside a material, the
wave slows down so the wavelength is reduced.
The free space wavelength l0 is often used interchangeably with
v fl frequency f. This is most common in optics.

Slide 40

20
 1/13/2020

Sign Convention
How do you define forward wave propagation? +z

Sign convention for this course

Quantity - bz +bz
  j t  kz     j t  kz 
Wave Solution E  E0 e  E  E0 e 

Dielectric Function      j       j 

Refractive Index N  n  j N  n  j

Slide 41

Summary of Parameter Relations

Permittivity Permeability
   0 r   0  r
 0  8.854187817  1012 F m 0  1.256637061 106 H m

Refractive Index Impedance

n  r r   0  r  r
0  0  0  376.73031346177 

Wave Velocity Frequency and Wavelength

c0   2 f Wave Number
v
n c0  f l0 2
c0  299792458 m s k0 
l0
Exact Slide 42

21
 1/13/2020

Duality Between E-D and H-B

Electric Field Magnetic Field

E H
D B
P M
ε μ

43

Introduction to
Finite-Difference
Time-Domain

Slide 44

22
 1/13/2020

Flow of Maxwell’s Equations


 B  t   
  E t    B  t      t    H  t 
t
A circulating E field induces a A B field induces an H field in
change in the B field at the center proportion to the permeability.
of circulation.


   D  t 
D  t     t    E  t    H t  
t
A D field induces an E field in A circulating H field induces a
proportion to the permittivity. change in the D field at the center
of circulation.

Note: In reality, this all happens simultaneously. In FDTD, it follows this flow.
Slide 45

Flow of Maxwell’s Equations Inside Linear, Isotropic and Non-Dispersive Materials

In materials that are linear, isotropic and non-dispersive we have

   t         t      

In this case, the flow of Maxwell’s equations reduces to

 
 H  t   E  t 
  E  t      H t   
t t
A circulating E field induces a A circulating H field induces a
change in the H field at the center change in the E field at the center
of circulation in proportion to the of circulation in proportion to the
permeability. permittivity.

Slide 46

23
 1/13/2020

Finite-Difference Approximations

df1.5 f 2  f1
 second-order accurate
first-order derivative
dx x
This derivative is defined to df
exist at the mid point between
f1 and f2. dx
f2
f1
x

Slide 47

Stable Finite-Difference Equations

Each term in a finite-difference equation must exist at the same point in time and space.
f  x 
Example:  f  x  0 Given f  0  , f  x  , f  2x  ,
x
Exists at x  x 2 Exists at x
f  x  x   f  x 
 f  x  0
Your simulation will be
unstable (i.e. explode).
x
Exists at x  x 2 Exists at x  x 2
f  x  x   f  x   x  f(x) is only known at integer
 f x 0 multiples of x. How do we
x  2  calculate f(x+x/2)?

Exists at x  x 2 Exists at x  x 2
f  x  x   f  x  f  x  x   f  x 
 0
x 2
Slide 48

24
 1/13/2020

Approximating the Time Derivative (1 of 3)

An intuitive first guess at approximating the time derivatives in Maxwell’s equations is:


Exists at t
 Exists at t  t2
 H  t   H  t  t   H  t 
  E t   
t
  E t   
t


Exists at t
 Exists at t  t 2
 E  t   E  t  t   E  t 
  H t   
t
  H t   
t

This is an unstable formulation.

Slide 49

Approximating the Time Derivative (2 of 3)

We adjust the finite-difference equations so that each term exists at the same point in time.

   
H  t   H  t  t  H  t   H  t  t 
  
 H  t  2 2
  E t      E t   
t t

    
 E  t  H  t   H  t  t  E  t  t   E  t 
  H t    
t 2 t

This works, but we will be doing more calculations

than are necessary.

Is there a simpler approach?

Slide 50

25
 1/13/2020

Approximating the Time Derivative (3 of 3)

Stagger 𝐸 and 𝐻 in time so that 𝐸 exists at integer time steps (0, t, 2t, …) and 𝐻 exists at
half time steps (t/2, t+t/2, 2t+t/2,…).
 
  H H
 H  t  t t 2 t t 2
  E t      E  
t t t

 
  E E
 E  t 
  H t     H  t t t
t t t 2 t

We will handle the spatial derivatives in × next

lecture in a very similar manner.

Slide 51

Derivation of the Update Equations

The “update equations” are the equations used inside the main FDTD loop to calculate the
field values at the next time step.

They are derived by solving our finite-difference equations for the fields at the future time
values.

    

  E  
t
H
t t 2
H
t
t t 2 H
t t 2
H
t t 2

t

  E
t


    

 H
t t 2

E
t t

t
E
t E
t t
E 
t
t
 H
  t t 2 
Slide 52

26
 1/13/2020

Anatomy of the FDTD Update Equation

Update To speed up simulation, we
coefficient calculate these before iterating.

  
E
t t
E 
t
 H
t   t t 2 
Field at the Field at the Curl of the “other” field
future time previous time at an intermediate time
step. step. step
Slide 53

The FDTD Algorithm…for now 

Initialize Fields to Zero  
  H t 0 E 0
EH 0
t t
2

yes
Done? Finished!

no
 
H E
Loop over time

Update H from E t
t
t

  
2
 
H
t t 2
H
t t 2

t

  E
t
 
H
E
t t
H

E
t
t
2

3 t
t t t
2
 
E H 3 t
t  2 t t
Update E from H  
2

t    E

  
H
t
5 t
t t  2 t

E   H
2
E  
t t t t t 2 E H 5 t
t 3 t t
2


Slide 54

27