1/28/2020

Advanced Electromagnetics:
21st Century Electromagnetics

Nonlinear &
Anisotropic Materials

Lecture Outline

• Nonlinear materials
• Anisotropic materials
• Tensor rotation and diagonalization (i.e. unrotation)
• Dispersion relation and index ellipsoids
• Appendix
• Rotation matrix for rotating 𝑎⃗ onto 𝑏
• Rotation matrix for rotating 𝑎⃗, 𝑏, 𝑐⃗ onto 𝑎⃗ , 𝑏 , 𝑐⃗ .

Slide 2

1
 1/28/2020

Nonlinear
Materials

Slide 3

Nonlinear Materials
All materials are nonlinear; some just have stronger nonlinear behavior than others.
For radio frequencies, materials tend to breakdown before they exhibit nonlinear
properties. Nonlinear properties are commonly exploited in optics.
In general, the polarization 𝑃 𝜔 of a material is a nonlinear function of the electric field
𝐸 𝜔 and can be expressed in scalar form as…

P     0  e    E     0  e   E 2     0  e3   E 3   
1 2

Linear response Nonlinear response

This is what we have done so far.

Slide 4

2
 1/28/2020

“Potential Well” Description

V is the “push” required to displace
the charge by a distance r.
V  r2

r
Linear materials have a parabolic potential well.
This leads to a resonator that oscillates as a perfect sinusoid.

Slide 5

“Potential Well” for Nonlinear Materials

This type of resonance does not
oscillate as a sinusoid. It can be seen
Nonlinear for as a series of sinusoids.
large r.

Linear “parabolic”
region for small r.

As bound charges are pushed very hard, the restoring force

is no longer linear.
The oscillator is effectively resonating at
This results in anharmonic oscillation. multiple frequencies at the same time.

Slide 6

3
 1/28/2020

Nonsymmetric Potentials

This is a non-symmetrical potential and

therefore exhibits a preferred direction
for polarization.

The result is a rectified signal along with

potentially other frequency content.

DC component

Slide 7

Applications of Nonlinear Materials

(1) Materials
• Ordinary linear materials

(2) Materials
• Rectification – can create a DC potential from an optical wave
• Frequency doubling (second harmonic generation) – can make lasers at otherwise
impossible wavelengths. Lacks inversion symmetry
• Parametric mixing – can provide sum and difference frequencies
• Pockel’s effect – can introduce birefringence from an applied electric field.

(3) Materials
• Kerr effect – field dependent dielectric constant
• Third harmonic generation – can generate very short wavelength waves.
• Raman scattering – the mechanical vibration of a molecule can shift the frequency of the
wave
• Brillouin scattering – dielectric response changes with applied pressure
• Two photon absorption – two photons are absorbed simultaneously where as a single Has inversion symmetry
photon would not be absorbed.

Slide 8

4
 1/28/2020

Anisotropic
Materials

Slide 9

Anisotropic Materials
In some materials, charges are more easily displaced along certain directions. For this
reason, the material can become polarized in a direction slightly different than the applied
field. In this case, the susceptibility is a tensor quantity.
Imagine pushing at an angle against a sliding glass door.
The door has a preferred direction for displacement.
Real materials are not this dramatic!

 
P     0   e   E  

Isotropic materials are good for simple transmission of electromagnetic waves.

Anisotropic materials are good for controlling and manipulating the waves.

Slide 10

5
 1/28/2020

Atomic Scale Picture

Charges will oscillate differently in

these two directions. Therefore, the
susceptibility changes as a function of
direction in the lattice.

large 

small 
Slide 11

Tensor Relations for Anisotropic Materials

The material polarization is now expressed as:
 Px       xx    xy    xz     Ex   
     Materials can become polarized in directions that
 Py      0   yx    yy    yz     E y    are slightly different than the electric field!
 Pz      zx    zy    zz     Ez    
  
This is most often treated through a dielectric tensor.
r     I    e   

The constitutive relation between 𝐸 and 𝐷 is then

 Dx     xx   xy   xz      Ex   
    
 Dy      0  yx     yy   yz     E y    
 Dz     zx   zy   zz      Ez   
  
The dielectric tensor has Hermitian symmetry for the general lossy case.
 ij   *ji
This means the field components are no longer independent. 
But…it presents many new possibilities!!
Slide 12

6
 1/28/2020

Principle Axes
It is always possible to choose a coordinate system such that the dielectric tensor becomes
diagonal.

xˆ aˆ
yˆ  bˆ 𝑎⃗, 𝑏, and 𝑐⃗ are called the Principal Axes of the crystal.
𝑎⃗, 𝑏, and 𝑐⃗ are not necessarily at 90° to each other.
zˆ cˆ

 Da   a   0 0   Ea    
    
 Db      0  0 b   0   Eb   
 Dc     0 0 c     Ec   
Alternative Description: There are only three degrees of freedom for 3D tensors.
Numbers can only occur in the off-diagonal elements when the tensor is rotated.
Slide 13

Maxwell’s Equations with Anisotropic Materials

Maxwell’s equations remain unchanged.
  
  E   j B D  0
  
  H  j D B  0

The constitutive relations now include tensors

 
D    E
 
B    H  Dx   xx  xy  xz   Ex 
 D    
 yy  yz   E y 
 y   yx
 Dz    zx  zy  zz   Ez 

Slide 14

7
 1/28/2020

Symmetry and Anisotropy

Anisotropy Crystal Symmetry Dielectric Properties
Isotropic Cubic
 0 0 
 0  0 
 
 0 0  

Uniaxial Hexagonal
 O 0 0 
Tetragonal 0  0
 O 
 0 0  E 
 E   O positive birefringence
 E   O negative birefringence

Biaxial Monoclinic
a 0 0
Triclinic 0
Orthorhombic 

b 0 
 0 0 c 
a  b  c
Slide 15

Tensor Rotation
and
Diagonalization

Slide 16

8
 1/28/2020

Definition of a Rotation Matrix

Rotation matrix [R] is defined as
 
 b   R    a
b

 a The rotation matrix should not change the
amplitude. This implies that [R] is unitary.
 
b a

 R   R
H 1

 R  R    R   R   I
H H

17

Derivation of a 2D Rotation Matrix


 Start with vector a at angle q.

b a  ax xˆ  a y yˆ
  a  cos q xˆ  sin q yˆ 
 a
q Add angle  to rotate the vector

b  a  cos q    xˆ  sin q    yˆ 
Apply trig identities
   cos q cos   sin q sin   xˆ 
b  a 
   sin q cos   cos q sin   yˆ 
 bx   cos   sin    a x    ax cos   a y sin   xˆ 
b     
sin  cos    a y 
 y  
 
    a y cos   ax sin   yˆ 
 R   
 
18

9
 1/28/2020

3D Rotation Matrices
Rotation matrices for 3D coordinates can be written directly from the previous result.


  b
  b 
a a a



b
1 0 0   cos  0 sin    cos   sin  0
 Rx     0 cos   sin    Ry      0
 1 0   Rz      sin  cos  0 
0 sin  cos     sin  0 cos    0 0 1 
19

Tensor Rotation
Tensors are rotated using the same rotation matrices, but in a slightly different manner than
vectors.

  R     Rz       Rz   
1

 


20

10
 1/28/2020

Combinations of Rotations
Suppose it is desired to first rotate about the x-axis by some angle, second rotate about
the y-axis by some angle, and third rotate about the z-axis by some angle.

For vectors, this would be done as…

 
b   Rz   Ry   Rx  a

For tensors, this would be done as…

1
  R     Rz   Ry   Rx   Rx 1  Ry   Rz 
1
     

21

Composite Rotation Matrix

We can combine multiple rotation matrices into a single composite rotation matrix [R].

 R    Ry   Rz  Rx  This equation implies we will rotate first about the x-axis,

second about the z-axis, and third about the y-axis.

Vector rotation using the composite rotation matrix is

 
b   R a
Tensor rotation using the composite rotation matrix is

  R     R   R 1
 
22

11
 1/28/2020

Animation of Tensor Rotation

23

Order of Rotations Matter

The order that the rotation matrices are multiplied controls the order that the rotations
are performed.
 
b   Rz  45    Ry  220    Rx 10   a
1. First rotates about the x-axis by 10°.
2. Second rotates about the y-axis by 220°.
3. Third rotates about the z-axis by -45°.

In general, different results are obtained when the order of rotation is changed.

 Ry   Rx    Rx   Ry 
24

12
 1/28/2020

Numerical Examples (1 of 2)
1 0 0 
Given:  r   02 0 
0 0 3

Rotate about x-axis by 20°

1 0 0  1 0 0 
 Rx  20    0 0.9397 0.3420   Rx  20    r    0 2.1170 0.3214 
0 0.3420 0.9397   0 0.3214 2.8830 

Rotate about y-axis by 45°

 0.7071 0 0.7071 2 0 1 
 Ry  45     0 1 0   Ry  45    r    0 2 0 
  
 0.7071 0 0.7071  1 0 2 

Rotate about z-axis by 60°

0.5000 0.8660 0   1.7500 0.4330 0 
 Rz  60    0.8660 0.5000 0   Rz  60    r    0.4330 1.2500 0 
 0 0 1   0 0 3 

25

Numerical Examples (2 of 2)
1 0 0 
Given:  r   0 2 0
 0 0 3 

Rotate first about x-axis by 20° and second about the y-axis by 45°
 0.7071 0.2418 0.6645   1.9415 0.2273 0.9415 
 Ry  45  Rx  20     0 0.9397 0.3420   Ry  45  Rx  20    r    0.2273 2.1170 0.2273
  
 0.7071 0.2418 0.6645   0.9415 0.2273 1.9415 
Rotate first about z-axis by 60° and second about the y-axis by 45°
 0.3536 0.6124 0.7071  2.3750 0.3062 0.6250 
 Ry  45  Rz  60     0.8660 0.5000 0   Ry  45  Rz  60    r    0.3062 1.2500 0.3062 
  
 0.3536 0.6124 0.7071  0.6250 0.3062 2.3750 
Rotate first about x-axis by 20°, second about the y-axis by 45°, and third about the z-axis by 60°
 0.3536 0.6929 0.6284   2.2699 0.0377 0.6676 
   Rz  60  Ry  45  Rx  20    r    0.0377 1.7886 0.7017 
 Rz  60  Ry  45 Rx  20     0.6124 0.6793 0.4044   
 0.7071 0.2418 0.6645   0.6676 0.7017 1.9415 

26

13
 1/28/2020

Tensor Diagonalization (1 of 2)
A tensor can always be diagonalized along its principle axes 𝑎, 𝑏, and 𝑐̂ .
Principle Axes:
 a 0 0
 a is along aˆ
    0 b 0 
Convention:
 0 0  c   b is along bˆ a  b  c
 c is along cˆ
But suppose a general nine-element tensor is given.

 xx  xy  xz 
 rot    yx  yy  yz 

How are the diagonal elements a, b, c and the principle axes 𝑎, 𝑏, and 𝑐̂ determined?
 zx  zy  zz 


27

Tensor Unrotation (2 of 2)
The original tensor [] is related to the rotated tensor [rot] through the composite
rotation matrix [R].

 rot    R   R 
1

Calculate the eigen-vectors and eigen-values of [rot].

 R  is the eigen-vector matrix of  rot 

 rot  
  is the eigen-value matrix of  rot 

28

14
 1/28/2020

Determining Principle Axes (1 of 2)

What are the principle axes of [rot]?
To find out, put the original principal axes 𝑎, 𝑏, and 𝑐̂ into a matrix.
 ax bx cx 

 P   ay by c y 
  
 az bz cz 
  
aˆ bˆ cˆ
Now rotate them according to [R] so they correspond to that of [rot].
 
 Prot    R   P  A tensor is not being rotated here so rotation is not performed as [R][P][R]-1.
   

29

Determining Principle Axes (2 of 2)

For the special case of orthorhombic symmetry, the principle axes start off as

1 0 0 

 P   0 1 0
   
0 0 1 
  
aˆ bˆ cˆ
[P] is just the identity matrix here. The principle axes of the rotated tensor are then


 Prot    R  I    R  Here is a second interpretation of the eigen-vector matrix [R]. These are
  the principle axes of the rotated tensor (for orthorhombic symmetry).

30

15
 1/28/2020

Directly Setting the Principle Axes

The orientation may be directly known in terms of the desired crystal axes 𝑎, 𝑏, and 𝑐̂ .
Place these principal directions into a matrix and interpret it as the composite rotation matrix.

 ax bx cx 
 R   a y by c y 
 az bz cz 
  
aˆ bˆ cˆ

The rotated tensor can be directly calculated from [R].

 rot    R   R 
1

31

Dispersion
Relation &
Index Ellipsoids
Slide 32

16
 1/28/2020

The Wave Vector

The wave vector (wave momentum) is a vector quantity that conveys two pieces
of information:
1. Wavelength and Refractive Index – The magnitude of the wave vector tells us the spatial
period l (wavelength) of the wave inside the material. When the free space wavelength is
known, |k| conveys the material’s refractive index n (more to be said later).
 2 2 n
k   l0  free space wavelength
l l0
2. Direction – The direction of the wave is
perpendicular to the wave fronts
(more to be said later).

k  ka aˆ  kbbˆ  kc cˆ

Slide 33

Dispersion Relations
The dispersion relation for a material relates the wave vector 𝑘 to frequency 𝜔.
Essentially, it conveys the refractive index 𝑛 as a function of direction through a material.

It is derived by substituting a plane wave solution into the wave equation.

For an ordinary linear homogeneous and isotropic (LHI) material, the dispersion relation is:

ka2  kb2  kc2  k02 n 2

ka2  kb2  kc2
This can also be written as:
2
 k02  0
n
Slide 34

17
 1/28/2020

Index Ellipsoids
From the previous slide, the dispersion relation for a LHI material was:

ka2  kb2  kc2  k02 n 2

This defines a sphere called an “index ellipsoid.” ĉ
The vector connecting the origin to a point on
the surface of the sphere is the k-vector for that
index ellipsoid
direction. Refractive index is calculated from this.

k  k0 n

For LHI materials, the refractive index is the

same in all directions. b̂
Think of this as a map of the refractive â
index as a function of the wave’s direction
through the medium.
Slide 35

How to Derive the Dispersion Relation (1 of 2)

The wave equation in a linear homogeneous anisotropic material is:
 
    E  k02 0  r  E  0 We are ignoring the magnetic response here.

The solution to this equation is still a plane wave, but our allowed values for 𝑘 (modes) are more complicated.
    
E  E0 e  jk r E0  Ea aˆ  Eb bˆ  Ec cˆ
Substituting this solution into the wave equation leads to the following relation:
     
 
k k  E0  k E0  k02  r  E0  0
This equation has the form:      aˆ       bˆ       cˆ  0
Each (•••) term has the form:  Ea   Eb   Ec  0
Each vector component must be set to zero independently.
aˆ component:  Ea   Eb   Ec  0 Matrix form…
bˆ component:  Ea   Eb   Ec  0      Ea 

cˆ component:  Ea   Eb   Ec  0      Eb   0
     Ec 
Slide 36

18
 1/28/2020

How to Derive the Dispersion Relation (2 of 2)

Solutions for 𝑘 are the eigen-values of the big matrix and derived by setting the determinant to zero.

  


det      0
   
This leads to the following general equation:

ka2 k2 k2
2  2 b  2 c 1
k  k02 na2 k  k02 nb2 k  k02 nc2

It can also be shown that given the wave vector 𝑘, the polarization of the electric field 𝐸 is:

  k
 
kb
 
kc

E0    2 a  aˆ  
2
 bˆ  
2
 cˆ
 2 2   2 2   2 2 
k  k0 na   k  k0 nb   k  k0 nc 
     
Slide 37

Generalized Dispersion Relation

Given the dielectric tensor…,
 a 0 0   na2 0 0
 
 r   0 b 0    0 2
n
B 0
 0 0  c   0 0 nC2 

The general form of the dispersion relation is:

ka2 kb2 kc2

2  2  2 1
k  k02 na2 k  k02 nb2 k  k02 nc2
This can be written in a more useful form as:
 2  ka2 kb2 kc2  2  kb2  kc2 ka2  kc2 ka2  kb2  4
k  2 2  2 2  2 2   k0  2
    k0  1
 nb nc na nc na nb   na nb2 nc2 
Slide 38

19
 1/28/2020

Dispersion Relation for Uniaxial Crystals

Uniaxial crystals have
na  nb  nO nO  ordinary refractive index
nc  nE nE  extraordinary refractive index
The general dispersion relation reduces to:

 ka2  kb2  kc2 2   k a  kb

2 2
kc2  This has two solutions
 2
 k0   2
 2  k02   0 corresponding to the two
 nO nE nO
     polarizations (TE and TM).
Sphere Ellipse
Ordinary Wave Extraordinary Wave

This has two solutions corresponding to the two polarizations (TE and TM).
This first solution is the same as for an isotropic material. It acts like it is propagating through a isotropic
material with index nO so it is called the “ordinary wave.”
The second solution is an ellipsoid. Depending on its direction, the effective refractive index will be
somewhere between nO and nE.

Slide 39

Index Ellipsoids for Uniaxial Crystals (1 of 2)

Ordinary Wave  ka2  kb2  kc2 2  k a  kb
2 2
kc2  Extraordinary Wave
 2
 k 0  2
 2
 k02   0
 nO  nE nO 

 ka2  kb2 kc2 

 ka2  kb2  k c2    2  k02   0
  k02   0 2
nO2  nE nO 
 

nO  nE nO  nE
Slide 40

20
 1/28/2020

Index Ellipsoids for Uniaxial Crystals (2 of 2)

ĉ
Observations
nO
• Both solutions share a common axis.
• This “common” axis looks isotropic with refractive index n0
regardless of polarization.
nE nE • Since both solutions share a single axis, these crystals are
called “uniaxial.”
• The “common” axis is called:
o Optic axis
nO o Ordinary axis
o C axis
o Uniaxial axis
b̂ • Deviation from the optic axis will result in two separate
possible modes.
â
nO
Slide 41

Dispersion Relation for Biaxial Crystals (1 of 2)

Biaxial crystals have all unique refractive indices. Most texts adopt the convention where
na  nb  nc
The general dispersion relation cannot be reduced. 

ĉ Notes and Observations

• The convention 𝑛 < 𝑛 < 𝑛 causes the optic axes to
lie in the 𝑎-𝑐̂ plane.

optic axes • The two solutions can be envisioned as one balloon

inside another, pinched together so they touch at only
four points.

• Propagation along either of the optic axes looks

isotropic, thus the name biaxial.
b̂
â Slide 42

21
 1/28/2020

Dispersion Relation for Biaxial Crystals (2 of 2)

There are three special cases when the biaxial case can be simplified.
These conditions can be produced in practice by launching electromagnetic waves at the
proper orientation.

 kb2 kc2 
ka  0 :  k  k  k n   2  2  k02   0
2
b
2
c
2 2
0 a
 nc nb 
Each special case has two separate
 k2 k2 
kb  0 :  k  k  k n   a2  c2  k02   0
2 2 2 2 solutions corresponding to the two
a c 0 b polarizations (TE and TM).
 nc na 
k 2
k 2

kc  0 :  ka2  kb2  k02 nc2   a2  b2  k02   0
 nb na 

Slide 43

Dispersion Surfaces of Magnetoelectric Materials

Magnetoelectric materials can exhibit up to 16 singularities.
     
D    E    H and B    H    E

Alberto Favaro and Friedrich W. Hehl, “Light propagation in local and linear media: Fresnel-Kummer
wave surfaces with 16 singular points,” arXiv preprint arXiv:1510.05566 (2015).
Slide 44

22
 1/28/2020

Direction of Power Flow

Isotropic Materials Anisotropic Materials
y  y
P 
P
 
k k
x x

Phase propagates in the direction of 𝑘. Therefore, the refractive index derived from |𝑘| is best described as the phase
refractive index. Velocity here is the phase velocity.
Power propagates in the direction of the Poynting vector which is always normal to the surface of the index ellipsoid. From
this, we can define a group velocity and a group refractive index.
Slide 45

Illustration of versus

Raymond C. Rumpf "Engineering the

Dispersion and Anisotropy of Periodic
Electromagnetic Structures," Solid State
Physics, Vol. 66, pp. 213-300, 2015.

Negative refraction into

a photonic crystal.

Slide 46

23
 1/28/2020

Double Refraction in Anisotropic Materials

Isotropic Materials Anisotropic Materials
y y

 
k1 k1
x  x
ke
 
k2 k0

Anisotropic materials have two index ellipsoids – one for each polarization.
Wave energy can split between the two produce an ordinary and an extraordinary wave.
Slide 47

Appendix

Slide 48

24
 1/28/2020

Rotation Matrix [R] that

Rotates Onto
Normalize Vectors
 
a ˆ b
aˆ   b 
a b

Algorithm

v  aˆ  bˆ  1 c   2
  R    I    v    2 
v 
s v i.e. sinq ab  s 
c  aˆ  bˆ i.e. cosq ab
 0 v3 v2  [R] is such that

v    v3 0 v1   R  aˆ  bˆ
 v2 v1 0 
49

Rotation Matrix [R] that

Rotates Onto

 ax bx cx   ax bx cx 

 S1    a y by cy 
  S2   ay by cy 

 az bz cz   az bz cz 

 R    S2  S1 
1

50

25