Linear Algebra for Describing Polarization and Polarizing Elements

Alex Miles, September 2, 2014

Last edit: September 5, 2014

1 Matrices as Operators
So far we’ve been using matrices to describe linear systems of simultaneous equations, however they can also
be used to describe maps or operations. A mapping will take an input vector, here denoted ~x, and yield an
output vector, denoted ~x0 .

M~x = ~x0
A simple example of this sort of action would be a matrix that “scales the x-component by 5, while
scaling the y component by 3”, which would look like this:
      
5 0 5 0 1 5
M= , =
0 3 0 3 1 3
A more complicated mapping, such as “rotate the input vector by π/2” is similarly easy once we know
the general form of the 2D rotation matrix:
   
cos(θ) − sin(θ) 0 −1
R(θ) = , R(π/2) =
sin(θ) cos(θ) 1 0
    
0 −1 1 −1
=
1 0 1 1
As has been discussed, it is important to note that matrix operations do not, in general, commute. For
example (left) rotating by 90 degrees then scaling the components separately is not equivalent to applying
the scaling then rotating.
          
5 0 1 0 5 0 1 0
R(π/2) = , R(π/2) =
0 3 0 3 0 3 0 5

2 Polarization
Recalling from electromagnetism that energy flows in the direction of the Poyting vector, S ~∝E ~ × B,
~ meaning
that (in linear media) the electric field must oscillate perpendicular to the direction of propagation. This still
permits two of freedom in the perpendicular plane, and materials both on the macroscopic level (reflection
coefficients) and the microscopic level (atomic transitions) are sensitive to how the field is configured. It is
useful to be able to describe the field using a vector, and the mappings or operations that various optical
elements cause as matrices. It is useful here to describe the function of several common polarization elements
that we will shortly express as matrices.

• Linear polarizers pass a given linear polarization while attenuating the orthogonal component.
• Circular polarizers pass a given circular polarization while attenuating the opposite component.
• Linear retarders impart a phase difference between the two orthogonal linear polarizations states.

– ∆φ = π, (half-wave) will rotate the polarization incident on it by 2θ where θ is the angle between
the fast axis of the plate and the incident linear polarization.
– ∆φ = π/2, (quarter-wave) converts ±45 degree linear polarized light to left and right hand
circularly polarized light.

1
3 Jones Calculus
First formulated in 1941, Jones calculus utilizes a 2x1 vector to describe the polarization of light, and 2x2
matrices to describe the action of an element upon that light. It is important to note that Jones calculus
can only be used to describe light that is fully polarized and coherent.

Jones calculus uses x̂ and ŷ as its basis states, and this means that all Jones vectors and matrices must
be defined with respect to some axes. Linear polarization of any angle can be described as a super-position
of these two basis states.
       
1 0 1 1 1 1
x̂ = , ŷ = , p̂+45 = √ , p̂−45 = √
0 1 2 1 2 −1
The most useful feature of Jones vectors are that they describe not only the amplitude of the two
components, but also the relative phase of the two components. The direction and frequency of the wave
are factored out.

E0x ei(kz−ωt+φx ) E0x eiφx

     
E(t)x
= = ei(kz−ωt)
E(t)y E0y ei(kz−ωt+φy ) E0y eiφy
This allows us to describe right and left circular polarizations easily.
   
1 1 1 1 1 1
R̂ = √ = √ (x̂ − iŷ), L̂ = √ = √ (x̂ + iŷ)
2 −i 2 2 i 2
Similarly, it can occasionally be useful to phrase x̂ and ŷ in terms of circular polarizations (e.g. when
your system has circular diattenuation and retardance, but you input linear light).
1 i
x̂ = √ (R̂ + L̂), ŷ = √ (R̂ − L̂)
2 2
There are two distinct classes of elements we must be able to describe: diattenuators, elements that
preferentially block one polarization and pass another, and retarders, elements that yield a phase difference
between two polarization components. First let us look at the most common components, linear polarizers.
Note the scaling factor on the ±45 degree case.
     
1 0 0 0 1 1 ±1
LPh = , LPv = , LP±45 =
0 0 0 1 2 ±1 1
Similarly, we can construct right and left circular polarizers by utilizing those states (and their conjugates)
as columns:
   
1 1 i 1 1 −i
CPr = , CPl =
2 −i 1 2 i 1
Retarder elements that have no diattenuation take on the following general form, where the relative phase
∆φ = φx − φy . By plugging in values we obtain the matrices for quarter and half-wave plates.

eiφx
     
0 iπ/4 1 0 cos(2θ) sin(2θ)
LR = , LRλ/4,vertical = e , LRλ/2 (θ) =
0 eiφy 0 −i sin(2θ) − cos(2θ)

To find the matrix for any rotated element, a standard rotation matrix can be applied. To find the
rotated version of a matrix the rotation matrix must be applied on both sides with a negated argument on
the right.
 
cos(θ) − sin(θ)
R(θ) =
sin(θ) cos(θ)

2
 Mrotated = R(θ)Moriginal R(−θ)
For example, a horizontal polarizer becomes a vertical polarizer if rotated 90 degrees.
π   
π
 
1 0 0 0
R R − =
2 0 0 2 0 1

3.1 Examples: Jones Calculus

Imagine we have input light that is horizontally polarized (along x̂). If we multiply it by the matrix for a
vertical polarizer we see that it is entirely attenuated.
    
0 0 1 0
=
0 1 0 0
However, if we insert a 45-degree polarizer one would expect that we should get some light through,
as the input is not orthogonal to the first polarizer, nor is that resulting state perpendicular to the final
polarizer.
          
0 0 1 1 1 1 0 0 1 1 1 0
= =
0 1 2 1 1 0 0 1 2 1 2 1

It seems that we get 1/4th of the light through (as must square the field to find the intensity). Now, let
us be entirely general. What happens when we rotate a linear polarizer arbitrarily?
Horizontal LP
z
 }| {    
1 0 cos(θ) − sin(θ) 1 0 cos(θ) sin(θ)
LP (θ) = R(θ) R(−θ) =
0 0 sin(θ) cos(θ) 0 0 − sin(θ) cos(θ)
| {z }
Rotated LP

cos2 (θ)
    
cos(θ) − sin(θ) cos(θ) sin(θ) cos(θ) sin(θ)
LP (θ) = =
sin(θ) cos(θ) 0 0 sin(θ) cos(θ) sin2 (θ)

We see that if we put in a horizontally polarized input, we obtain the expected cosine-squared dependence
of Malus’s Law.

cos2 (θ)
   
1
LP (θ) =
0 cos(θ) sin(θ)
It can easily be seen that Jones calculus permits one to take an arbitrarily large sequence of elements,
rotated at arbitrary angles or otherwise, and reduce them to a single 2x2 matrix with predictable action on
any expressible input polarization.

4 Mueller Calculus
First proposed in 1943, Mueller calculus allows us to analyze incoherent and partially-polarized light. The
vector that describes the light is now 4 elements long. While the elements are less intuitive than those of the
Jones vector, they are experimentally convenient. Any element of the stokes vector can be determined by
taking two measurements with either a linear or circular polarizers. It is important to note that the Mueller
approach neglects phase and works directly with intensities.

3
 |Ex |2 + |Ey |2
      
S0 I  Intensity of the light
 S1   |Ex |2 − |Ey |2

  pI cos(2ψ) cos(2χ))  
Linear-axial component
 S2  =  2< Ex Ey∗   pI cos(2ψ) sin(2χ)  ⇒  Linear 45-degree component
  
= 

−2= Ex Ey∗


S3 pI sin(2χ) Circular component


The above definition requires understanding of two concepts, the first being degree of polarization,
denoted p. A state that is only partially polarized may have p < 1. We can relate this quantity using the
relation below. It can be seen that if the length of the lower 3-element vector is equal to this element then
we have perfect polarization, p = 1.
p
S12 + S22 + S32
p=
S0
Secondly we need to introduce the Poincaré sphere (pronounced “Pwahn-car-ay”). This unit sphere exists
in a space where the x, y, z coordinate axes are instead S1 , S2 , S3 . The polarization state is represented by
a vector with one end rooted at the origin and with a length p. This means that perfectly polarized states
lie on the surface of the sphere, while the interior represents partially polarized states. The parameter ψ
tracks the azimuth as the state varies between S2 = −1 to S2 = +1 corresponding to ±45 degree linear
while χ tracks the altitude as the state varies from S3 = −1 to S3 = +1 corresponding to variation between
left-hand circular and right-hand circular.

The equator contains all linear polarizations, while circular polarizations lie at the two poles. All other
states are elliptical polarized. In this illustration one can imagine the action of a half-wave plate to be
traversal in the azimuthal direction (moving along the equatorial direction at whichever latitude you start
at) while the action of a quarter-wave plate is to alter the altitude
Common polarization states must be expressed differently, but also that we can express unpolarized light:

           
1 1 1 1 1 1
 1   −1   0   0   0   0 
x̂ = 
 0  , ŷ =  0
   , p̂±45 =
 ±1  , R̂ = 
   , L̂ =   , unpolarized =  
 0   0   0 
0 0 0 1 −1 0

Similarly, optical elements must be represented by significantly different matrices in Mueller calculus.

     
1 1 0 0 1 −1 0 0 1 0 ±1 0
1 1 1 0 0  1 −1 1 0 0   0 0 0 0 
LPh =  , LPv =  , LP±45 =
 ±1

2 0 0 0 0  2 0 0 0 0  0 1 0 
0 0 0 0 0 0 0 0 0 0 0 0

Quarter-wave and half-wave plates with the fast axis oriented vertically can be expressed similarly:

4
    
1 0 0 0 1 0 0 0
 0 1 0 0   0 1 0 0 
LRλ/4 = , LRλ/2 = 
 0 0 0 −1   0 0 −1 0 
0 0 1 0 0 0 0 −1