Lecture notes on

Polarization and its matrix representation (v. 1.5)

Julio César Gutiérrez Vega

Photonics and Mathematical Optics Group
Tecnológico de Monterrey, México
[Link]
February 23, 2011

God used beautiful mathematics in creating the world.

Paul Dirac

These lecture notes are devoted to elementary theory of polarization of light waves and its matrix repre-
sentation. The material is intended for undergraduate students of a …rst course of Electrodynamics and/or
Optics.

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Polarization of a monochromatic plane wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1 Sense of rotation of the electric vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Geometrical properties of the polarization ellipse . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Determining the parameters fE0x ; E0y ; g from the parameters fEmax ; Emin ; g . . . . . . . . . . 4
2.4 Linear and circular polarizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Visualizing the polarization in space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3 Jones description of the polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1 Representation of the polarization states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.1 Orthogonal polarizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1.2 The linear and the circular polarization basis . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1.3 The linear and the circular polarization ratios . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Representation of the optical devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.1 Polarization eigenmodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.2 Jones matrices of elementary polarization devices . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.3 Expressing the Jones matrix in other polarization basis . . . . . . . . . . . . . . . . . . . 13
3.3 Illustrative applications of the Jones calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4 Stokes-Mueller description of the polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.1 Statistical derivation of the Stokes parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Physical interpretation of the Stokes parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3 De…nition of the Stokes parameters in terms of di¤erent parameters of the polarization ellipse . . 19
4.3.1 Standard de…nition in terms of parameters fE0x ; E0y ; g . . . . . . . . . . . . . . . . . . . 20
2 2
4.3.2 De…nition in terms of the observables Emax ; Emin ; . . . . . . . . . . . . . . . . . . . . 20

1
2 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

4.3.3 De…nition in terms of the wave intensity and the inclination and ellipticity angles fI; ; g:
the Poincaré sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.3.4 De…nition in terms of the wave intensity and the retardance and rectangle angles fI; ; #g:
the observable Poincaré sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3.5 Relations between angles f ; g and angles f ; #g . . . . . . . . . . . . . . . . . . . . . . 23
4.3.6 Stokes parameters in terms of the Pauli spin matrices . . . . . . . . . . . . . . . . . . . . 23
4.4 Propagation of the Stokes vectors through polarization devices: Mueller calculus . . . . . . . . . 23
4.5 Relationship between Mueller and Jones matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5.1 Calculating the Mueller matrix from the Jones matrix . . . . . . . . . . . . . . . . . . . . 25
4.5.2 Calculating the Jones matrix from the Mueller matrix . . . . . . . . . . . . . . . . . . . . 26
5 Unpolarized and partially polarized light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.1 Degree of polarization of a wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2 Understanding the partial polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3 Illustrative applications of the Mueller calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.3.1 Optical activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6 Measurements of the polarization parameters of a wave . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.1 Measuring the Stokes parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.2 Determination of the Jones vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7 General treatment of the elliptical birefringence: Propagation through a birefringent medium . . . . . 32
7.1 Jones formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.2 Stokes-Mueller formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.2.1 Relation with the optical …ber parameters: coupled mode theory . . . . . . . . . . . . . . 33
7.2.2 Linear and circular birefringence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
8 The Pancharatnam-Berry phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
A Appendix: Derivation of the polarization ellipse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
B Appendix: Derivation of the parameters fEmax ; Emin ; g of the polarization ellipse . . . . . . . . . . . 35
C Appendix: Rotations in a two-component Jones vector formalism . . . . . . . . . . . . . . . . . . . . . 36
D Appendix: Polarization of harmonic waves in three-dimensions . . . . . . . . . . . . . . . . . . . . . . 37

1 Introduction
The polarization of light is determined by the time variation of the direction of the electric …eld vector E (r; t)
at position r: Ordinary sources of light, such as the sun or incandescent light bulbs, produce thermal radiation
which we describe as unpolarized and incoherent light. Light of this sort is a chaotic jumble of independent
disturbances, each with its own traveling direction, its own frequency, and its own state of polarization.
For purely monochromatic light, the three components of E (r; t) vary sinusoidally with time with amplitudes
and phases that are generally di¤erent, so that at each position r the endpoint of the vector E (r; t) moves in
a plane tracing an ellipse in general. The plane, orientation and shape of the polarization ellipse generally
vary with position. For paraxial beams, light propagates along directions that lie within a narrow cone centered
about the optical axis (typically the z axis). Paraxial beams are approximately transverse electromagnetic and
the vector E (r; t) lies in the transverse plane (x; y) : In this case, the polarization plane is thus approximately
parallel for all points in space.
Ideal monochromatic plane waves are examples of fully polarized waves. In general, waves in nature are
said to be partially polarized. The limit case of the unpolarized light occurs when the electric …eld of light
vibrates randomly without exhibiting any preponderance along a particular linear direction, helicity or state of
polarization. It was demonstrated that a partially polarized …eld can be uniquely decomposed into a sum of
two …elds, one corresponding to a fully unpolarized and the other to a fully polarized …eld. This …nding led to
the introduction of the concept of degree of polarization for two-dimensional electromagnetic …elds, de…ned as
the ratio of the intensity contained in the polarized part to that of the total …eld. The degree of polarization of
a …eld can be measured by a simple experiment involving a compensator and a polarizer. Since then the degree
of polarization has taken its place as a central parameter in characterizing the polarization state of …elds
Lecture notes on Polarization 3

In dealing with problems involving polarized light, it is often necessary to determine the e¤ect of various
types of optical devices on the state of polarization of a light beam. For quantitative calculations, several matrix
methods have been developed and applied for many years. The matrix methods are based on the fact that the
e¤ect of the device is to perform a linear transformation (represented by a matrix) on the vector representation
of a polarized wave. The most common forms of matrix approaches to polarization are the Stokes-Mueller
calculus and the Jones calculus, but the coherency-matrix formulation is also a powerful tool for dealing with
problems involving partially polarized light.
These lecture notes are devoted to elementary theory of polarization of light waves and its matrix represen-
tation. The material is intended for undergraduate students of a …rst course of Optics and/or Electrodynamics.
We give here a description of the Jones and the Stokes-Mueller descriptions of polarization, indicating how they
are used and the di¤erent types of problems for which they are helpful. A treatment of the 2 2 coherence–
matrix theory goes beyond the scope of these introductory notes, so we refer the interested reader to classical
books [1, 2].

2 Polarization of a monochromatic plane wave

We begin by writing the electric vector of a monochromatic plane wave traveling in positive z direction in terms
of its Cartesian components along the x b and y
b axes

E (z; t) = E0x cos (kz !t + b

x) x + E0y cos kz !t + y b:
y (1)

where E0x ; E0y ; x ; y 2 R:

Let us now consider the nature of the curve which the end point of the electric vector describes at a given
point in space. This curve is the locus of the points whose coordinates (Ex ; Ey ) are

Ex (z; t) = E0x cos (kz !t + x) ; (2a)

Ey (z; t) = E0y cos kz !t + y : (2b)

Figure 1: Polarization ellipses for the electric and magnetic …elds of a monochromatic plane wave.

By combining Eqs. (2) to eliminate the term (kz !t) we obtain (see derivation in Appendix A)
2 2
Ex Ey Ex Ey
+ 2 cos = sin2 ; (3)
E0x E0y E0x E0y
where
y x (4)
4 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Figure 2: Elliptical polarization for several values of the phase di¤erence. The arrows show the time evolution
of the electric vector at a given point in space.

is the phase shift between the Cartesian components of the electric …eld.
Equation (3) is often referred to as the polarization ellipse because it corresponds to an ellipse inscribed
into a rectangle whose sides are parallel to the axes (x; y) and whose lengths are 2E0x and 2E0y [see Fig. 1(a)].
The electric plane wave Eq. (1) is then said to be elliptically polarized.
It is clear from Eq. (3) that the state of polarization of the wave depends only on three independent
parameters, namely, the amplitudes E0x ; E0y and the phase shift :
On the other hand, according to the Faraday law, the magnetic …eld of the plane wave Eq. (1) is orthogonal
to the electric …eld, that is

1 E0y E0x
B (z; t) = k E= cos kz !t + y b+
x cos (kz !t + b;
x) y (5)
! c c
where c = !=k is the speed of light. Similar to the electric …eld, the magnetic vector is elliptically polarized and
its polarization ellipse is rotated 90 degrees respect to the ellipse of the electric …eld, as shown in Fig. 1(b).

2.1 Sense of rotation of the electric vector

At a given point in the space, the electric vector E (z; t) rotates with angular frequency !. When viewed against
the direction of propagation, the sense of rotation is determined by the phase di¤erence according to the
following table:

Condition Sense of rotation 1 Polarization terminology

sin > 0 Counterclockwise Left Positive helicity
sin < 0 Clockwise Right Negative helicity

The terminology left and right is traditionally used in the context of optics. However, in modern physics,
the wave is said to have positive or negative helicity. The latter description has the advantage of resembling the
projection of the angular momentum on the z axis.
Figure 2 shows several examples of elliptical polarization for di¤erent values of the phase shift :

2.2 Geometrical properties of the polarization ellipse

A geometrical analysis of Eq. (3) reveals the following properties of the polarization ellipse:

1. The ellipse touches the sides of the rectangle at points ( E0x ; E0y cos ) and ( E0x cos ; E0y ) :

2. The ellipse is the same for both ; i.e. it is invariant under a change in the sign of :
1 It is important to emphasize that the senses of rotation should be changed if one uses the traveling-wave form (!t kz) rather
than (kz !t) for the plane wave.
Lecture notes on Polarization 5

3. The inclination angle 2 [ =2; =2] between the x-axis and the direction of the major axis of the ellipse
is given by (see derivation in Appendix B)
!
1 2E0x E0y
= arctan 2 2 cos : (6)
2 E0x E0y

To avoid the ambiguity of the standard arctan function (which returns a principal value in the range
[ =2; =2]), the arctan function in Eq. (6) should be treated like the atan2 function which computes the
arctan of y=x given y and x, i.e. the angle between the positive x-axis and the radius to the point (x; y).
The atan2 function returns a principal value in the range ( ; ].

4. Thus, when > 0 the major axis of the ellipse falls on the …rst and third quadrants. When < 0; fall on
the second and fourth quadrants.

5. The semi-major and semi-minor axes of the polarization ellipse give the maximum Emax and minimum
Emin values of the electric …eld. The explicit values of Emax and Emin are (see derivation in Appendix B)

2
2
E0x cos2 + E0y
2
sin2 + 2E0x E0y cos sin cos ;
Emax = (7)
2
cos sec 2 E0x 2
sin2 sec 2 E0y2
;

2
2
E0x sin2 + E0y
2
cos2 2E0x E0y cos sin cos ;
Emin = 2 2 (8)
sin sec 2 E0x + cos2 sec 2 E0y2
:

By convention the value of Emax always is positive, but the value of Emin can be positive or negative
depending if the state of polarization has positive or negative helicity, respectively.
Equations (7) and (8) can be written in a convenient matrix notation as follows:

2
Emax cos2 sin2 2
E0x
= sec 2 : (9)
2
Emin sin2 cos2 2
E0y

From these relations it can be veri…ed that the intensity of the wave can be calculated with
2 2 2 2
I = Emax + Emin = E0x + E0y : (10)

The polarization ellipse can be determined by either the set of parameters fE0x ; E0y ; g or fEmax ; Emin ; g :

2.3 Determining the parameters fE0x ; E0y ; g from the parameters fEmax ; Emin ; g
Conversely, if the maximum and minimum electric …elds and the inclination of the polarization ellipse are known
(i.e. Emax ; Emin and are given) the above formulae enable the determination of the amplitudes E0x ; E0y and
the phase di¤erence : Inverting (9) we obtain

2
E0x cos2 sin2 2
Emax
= : (11)
2
E0y sin2 cos2 2
Emin

Once determined E0x and E0y ; the phase di¤erence is calculated with Eq. (6), that is
!
2 2
E0x E0y
= arccos tan 2 ; (12)
2E0x E0y

where the sign is chosen to specify the desired sense of rotation.

6 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Figure 3: Time and space evolutions of the electric vector for a left- and right-handed polarized plane waves
with traveling convention (kz !t) :

2.4 Linear and circular polarizations

Two meaningful special cases of the elliptic polarization are of particular importance, namely, when the polar-
ization ellipse degenerates into a straight line or a circle.

1. Linear polarization occurs when

y x =m ; (m = 0; 1; 2; :::) : (13)

Then Eq. (3) reduces to the equation of a straight line

m E0y
Ey = ( 1) Ex ; (14)
E0x

and we say that E (z; t) is linearly polarized. In this case, the direction of the electric vector is the same
is all points in space for any time.

2. Circular polarization occurs when

E0x = E0y = E0 ; (15)

y x =m ; (m = 1; 3; 5; :::) : (16)
2

Then Eq. (3) reduces to the equation of a circle Ex2 + Ey2 = E02 ; and the electric …eld components reduce
to
Ex = E0 cos (kz !t + x ) ; Ey = E0 sin (kz !t + x ) ; (17)

where the signs + and corresponds to right- and left-handed polarizations. The sense of rotation of the
circularly polarized electric vector is the same as for the elliptic polarization.
Lecture notes on Polarization 7

Figure 4: Spatial distribution of a set of plane wavefronts of (a) a vertical linear and (b) a right-handed circularly
polarized wave.

2.5 Visualizing the polarization in space

Figures 1 and 2 show the variation of the electric vector at a …xed point in space as the time is increasing.
Alternatively, we can …x the time and observe the propagation of the electric vector as the coordinate z increases.
Figure 3 shows both situations for right and left polarizations. Note that in the traveling convention (kz !t)
the standard right-hand rule applies to the visualization in space.
The spatial distribution of the electric vector can be observed in Fig. 4 for a right-handed wave. The black
arrows illustrate the time rotation of each electric vector at the particular point in space. It should be pointed
out that if one "rides along" with the wave, then the vectors do not change in either direction or magnitude,
because the quantity (kz !t) remains constant. This holds for any type of polarization.

3 Jones description of the polarization

R. C. Jones introduced in 1941 an elegant and useful matrix formalism to describe the polarization of mono-
chromatic light [5]. In this section we review the fundamentals of Jones representation of the polarization.

3.1 Representation of the polarization states

Consider again the monochromatic electromagnetic wave of frequency ! traveling in the positive z direction
Eq. (1). The polarization state of this wave is completely characterized by the complex amplitudes Ex =
E0x exp (i x ) and Ey = E0y exp i y : This information can be conveniently represented as a 2 1 column
vector
Ex E0x ei x cos # ei x
E= = i y = E0 ; E0x ; E0y ; x ; y 2 R; (18)
Ey E0y e sin # ei y
where
2 2 1=2
# = arctan (E0y =E0x ) ; and E0 = E0x + E0y : (19)
This vector is called the Jones vector or the Maxwell column.
De…nition 1 Adjoint vector. The adjoint of the Jones vector is de…ned by the adjoint (i.e. conjugate
transpose) vector
Ey = Ex ; Ey = E0x e i x E0y e i y ; (20)
where the asterix denotes complex conjugate.
8 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Table 1: Standard normalized Jones and Stokes vectors for some polarization states.
Polarization state Jones vector Stokes vector (Transpose)
Linear state cos T
L = 1 cos 2 sin 2 0
angle with x axis sin
1 T
Linear horizontal ! L0 ^=
x 1 1 0 0
0
0 T
Linear vertical " L90 ^=
y 1 1 0 0
1
Linear at + 45 % 1 1 T
L 45 =p 1 0 1 0
Linear at 45 & 2 1
Circular state
1 1 T
Left-handed or positive helicity ^
c p 1 0 0 1
2 i
Right-handed or negative helicity

De…nition 2 Inner product. The inner product between two polarization states E1 and E2 is de…ned by
D E E1x
Ey2 ; E1 E2x ; E2y = E1x E2x + E1y E2y : (21)
E1y

The inner product of the vector with itself gives the intensity of the optical …eld, namely

cos # ei x
I = Ey ; E = E02 cos # e i x ; sin # e i y = E02 2 R 0: (22)
sin # ei y

For simplicity, it is often convenient to work with normalized Jones vectors, so the condition E0 = 1 can
assumed without loss of generality.
The standard normalized Jones vectors for some special polarization states are provided in Table 1.

3.1.1 Orthogonal polarizations

TwoDpolarization
E states E1 and E2 are said to be orthogonal if the inner product between E1 and E2 is zero,
y
i.e. E2 ; E1 = 0: Two typical examples of orthogonal polarization states are given by the linearly polarized
waves in the x and y directions (see Table 1), i.e.

1
^y ; x
y ^ = 0 1 = 0; (23)
0

and the right and left circularly polarization waves (see Table 1), i.e.
D E 1 1 1
^cy+ ; ^
c =p 1 i p = 0: (24)
2 2 i

Expressing an arbitrary polarization as a superposition of two orthogonal polarizations

Two orthogonal polarization states E1 and E2 form a basis to expand an arbitrary polarization E, that is

E = a1 E1 + a2 E2 : (25)

If E1 and E2 are normalized, then the expansion weights are the inner products
D E D E
a1 = Ey1 ; E ; a2 = Ey2 ; E : (26)
Lecture notes on Polarization 9

3.1.2 The linear and the circular polarization basis

In general, an elliptically polarized wave E can be expressed in terms of two orthogonal polarization states
(elliptical in the most general case). For example, if we choose the basis of linear polarizations (^
x; y
^) or the
basis of circular polarizations ^
c+ ; ^
c ; we have

E = Ex x
^ + Ey y
^; or E = E+ ^
c+ + E ^
c ; (27)

where (Ex ; Ey ) and (E+ ; E ) are the expansion coe¢ cients in the basis of linear and circular polarization,
respectively. The basis of circular polarizations ^
c+ ; ^
c is often referred to as the rotating frame. The relation
between both basis is as follows:

1. If the coe¢ cients (Ex ; Ey ) are known, the corresponding circular coe¢ cients (E+ ; E ) are
D E 1 1
Ex
E+ = ^cy+ ; E = p 1 i = p (Ex iEy ) ; (28a)
2 Ey 2
D E 1 Ex 1
E = ^cy ; E = p 1 i = p (Ex + iEy ) : (28b)
2 Ey 2

Therefore
Ex iEy Ex + iEy
E = Ex x
^ + Ey y
^= p c+ +
^ p c :
^ (29)
2 2
The relation can be written in matrix form as follows:

E+ 1 1 i Ex
=p : (30)
E 2 1 i Ey

Linear polarization: Using Eqs. (28) it is easy to see that a linearly polarized wave with plane of
cos
polarization making an angle with the x axis, i.e. E = ; can be expanded in terms of circular
sin
polarized waves as follows:
e i ei
E = cos x ^ + sin y ^=p ^ c+ + p ^ c : (31)
2 2
If the linear polarization of E is horizontal or vertical (i.e. = 0; = =2) then we conclude
p p
^ = (^
x c+ + ^
c ) = 2; ^ = i( ^
y c+ + ^
c ) = 2: (32)

2. Conversely, if the coe¢ cients in the circular basis E+ and E of E are known, the coe¢ cients of E in the
basis of linear polarizations are

1 E+ E+ + E
Ex = ^y ; E = p 1 1
x = p ; (33a)
2 E 2
i E+ E+ E
Ey = ^y ; E = p 1
y 1 =i p : (33b)
2 E 2

Therefore
E+ + E E+ E
E = E+ ^
c+ + E ^
c = p ^+i
x p ^:
y (34)
2 2
In matrix form we write
Ex 1 1 1 E+
=p : (35)
Ey 2 i i E
10 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

3.1.3 The linear and the circular polarization ratios

For a wave E expressed in terms of the linear polarization basis, E = Ex x
^ +Ey y
^, we de…ne the linear polarization
ratio as
Ey
p i 2 C: (36)
Ex
The quantity p gives information of the relation between the linearly polarized components of the wave. If the
reference frame is rotated an angle ; the components Ex and Ey will change as shown in Fig. 5. In the rotated
frame, the new components are
Ex0 cos sin Ex
= : (37)
Ey0 sin cos Ey
Then, the transformation rule for the linear polarization ratio under a rotation is
Ey0 Ey cos Ex sin p i tan
p0 = i =i = : (38)
Ex0 Ex cos + Ey sin 1 ip tan

Figure 5: The Cartesian components of the vector E are not constant under a rotation of the reference frame.

Similarly, for a wave E expressed in terms of the circular polarization basis E = E+^
c+ + E ^ c ; we de…ne
the circular polarization ratio as
E+
q 2 C: (39)
E
The quantity q gives information of the relation between the circular polarized components of the wave. If one
rotates the reference frame an angle with respect to the principal axes, one gets
0
E+
q0 = 0 = q ei2 : (40)
E
Hence the quantity jqj is independent of the reference frame and therefore has an intrinsic meaning.
If jqj > 1; the helicity of the wave is more positive than negative, thus the wave is left-handed polarized.
Analogously, If jqj < 1; the wave is right-handed polarized. Finally, for jqj = 1 both helicities cancel out and
the wave is linearly polarized.
From Eq. (29) or (34) it is easy to see that the relationship between both linear and the circular polarization
ratios is given by the bilinear transformations
1 p 1 q
q= ; p= ; (41)
1+p 1+q
from which we obtain
2
j1 pj 1 + jpj2 2 Re p
jqj2 = 2 = : (42)
j1 + pj 1 + jpj2 + 2 Re p
Therefore we can see that
Lecture notes on Polarization 11

1. the half-plane Re p > 0 is mapped onto the region jqj < 1; i.e. the region of the right-handed polarization.
The special case q = 0; correspond to the right-handed circular polarization.

2. the half-plane Re p < 0 is mapped onto the region jqj > 1; i.e. the region of the left-handed polarization.
The special case q = 0; correspond to the right-handed circular polarization.

3. the line Re p = 0 is mapped onto the circle jqj > 1; i.e. the locus of linear polarization.

3.2 Representation of the optical devices

Consider an optical system that alters the polarization of an input wave but maintains its plane-wave nature.
The system is assumed to be lossless and linear, so that the principle of superposition holds. If the input and
output waves are represented by the Jones vectors E1 and E2 , then the transformation of the waves is described
by the matrix relation

E2 = JE1 ; (43a)
E2x J11 J12 E1x
= : (43b)
E2y J21 J22 E1y

1
The matrix J, called Jones matrix, describes the optical system and is unitary, i.e. J = Jy :

Rotated devices. If an optical element is rotated about the optical axis by angle ; the Jones matrix J ( )
for the rotated element is constructed from the matrix J for the unrotated element by the transformation

J( ) = S ( ) J S ( ); (44)

where
cos sin
S( ) ; (45)
sin cos

is the rotation matrix. Equation (45) is indeed the Jones matrix of a rotator.

Jones matrix for a linear birefringent device with two orthogonal transmission factors. Consider
an optical device with di¤erent complex transmission factors fx and fy along two …xed orthogonal directions,
let us say x and y: The Jones matrix is given by

fx 0 jfx j ei x 0
= ; (46)
0 fy 0 jfy j ei y

and depends only on the nature of the optical element.

If now the device is rotated an angle ; the new Jones matrix is given by

cos sin fx 0 cos sin fx cos2 + fy sin2 (fx fy ) sin cos

= : (47)
sin cos 0 fy sin cos (fx fy ) sin cos fy cos2 + fx sin2

The rotated Jones matrix describes a large set of common optical devices, including for example ideal [fx = 1;
fy = 0] and non-ideal [jfx j ' 1; jfy j ' 0] polarizers, linear phase retarders [jfx;y j = 1] ; etc. taking onto account
the possible orientation of the device respect to the x; y axes.

Cascaded devices. If a polarized wave is sent through a train of optical devices (each with its own
rotation), then the overall transformation matrix is given by the matrix product

J = Jn Jn 1 J2 J1 : (48)
12 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Table 2: Jones and Mueller matrices for ideal polarizers

Polarizer Jones matrix 2 Mueller matrix 3
1 cos 2 sin 2 0
Linear polarizer
cos2 sin cos 166 cos 2 cos 2
2 cos 2 sin 2 077
axis at with P = 4
sin cos sin2
2 sin 2 cos 2 sin 2 sin2 2 05
the horizontal
0 0 0 0
2 3
1 1 0 0
Linear polarizer 1 0 1 6 7
P0 PX = 61 1 0 07
horizontal ! 0 0 2 40 0 0 05
0 0 0 0
2 3
1 1 0 0
Linear polarizer 0 0 166 1 1 0 07
7
P 90 PY = 4
vertical " 0 1 2 0 0 0 05
0 0 0 0
2 3
1 0 1 0
Linear polarizer
1 1 1 166 0 0 0 07
7
at + 45 % P 45 =
2 1 1 2 4 1 0 1 05
at 45 &
0 0 0 0
2 3
1 0 0 1
Circular polarizer
1 1 i 1660 0 0 07
7
Left or positive helicity CP
2 i 1 2 0 0 0 05
4
Right or negative helicity
1 0 0 1

3.2.1 Polarization eigenmodes

Consider an optical system characterized by the Jones matrix

J11 J12
J= : (49)
J21 J22

In general, the polarization state of the output wave is di¤erent of the polarization state of the input wave. An
eigenstate v of the optical system is a particular polarization state that if is passed through the system the
output state is proportional to the input state, that is

J v = v; (50)

where is a constant called the eigenvalue (complex in general). For a matrix of size 2 2; the eigenstates v
and eigenvalues are given by
q
1 2
= J11 + J22 (J11 J22 ) + 4J12 J21 ; (51)
2
J12
v = ; (52)
J11

where is an arbitrary constant that can be set to normalize the eigenvectors v .

If the matrix J is Hermitian, i.e. J12 = J21 ; the eigenmodes are orthogonal, i.e. hv+ ; v i = 0: In this way,
the eigenmodes of the system can be used as an expansion basis, so that an arbitrary input wave E1 may be
expanded as a superposition of normal modes E1 = A+ v+ + A v , where the coe¢ cients A are calculated
with Eqs. (26). In this case, the output of the system is directly

E2 = J E1 = J (A+ v+ + A v ) = A+ + v+ +A v : (53)
Lecture notes on Polarization 13

3.2.2 Jones matrices of elementary polarization devices

De…nition 3 Polarizer. A polarizer is an optical device that converts an input electromagnetic wave with
unde…ned or arbitrary polarization into an output wave with a well-de…ned state of polarization, e.g. linear,
circular, etc.
An ideal linear polarizer is a device that is completely transparent to wave components parallel to the
transmission axis of the polarizer, and complete opaque to light in orthogonal direction to the transmission
axis, see Fig. 6(a). The linear polarizers are manufactured in several ways taking advantage of physical
properties like the double refraction and the anisotropic optical absorption (i.e. dichroism), among others [4,
Chap. 8].

Figure 6: (a) Ideal linear polarizer with transmission axis (t.a.) at angle with the horizontal. (b) Ideal
quarter-wave plate can transform a linearly polarized wave into a circular one.

De…nition 4 Retarder. A retarder modi…es the polarization of an incident wave by changing the phase
di¤ erence of the two orthogonally polarized waves that form the incident wave. Depending of a circularity
parameter ", the retarders are classi…ed as follows:

A linear retarder (" = 0) introduces a phase di¤erence between two orthogonal Cartesian directions …xed
in the material. These particular directions are referred to as the fast and the slow axes of the retarder.
Linear retarders are made of birefringent materials and are by far the most common type of retarder. If
the input wave is linearly polarized and its polarization plane coincides with the fast or the slow axis,
then its state of polarization is not changed by the retarder. Linear retarders with retardation of 90 and
180 are often called quarter-wave and half-wave plates, respectively, see Fig. 6(b).

A circular retarder (" = =2) introduces a phase di¤erence between the two orthogonal right and left
circular polarizations. In other words, if the input wave is expressed as a superposition of a right and
a left circularly polarized waves, each component will su¤er a di¤erent retardation. If the input wave is
circularly polarized, then its state of polarization is not changed by the retarder. Circular retarders are
sometimes called rotators.

An elliptical retarder ( =2 < " < =2; " 6= 0) ; introduces a phase di¤erence between the two particular
orthogonal elliptical polarizations. These polarizations states are de…ned by the circularity parameter ":
Two or more linear retarders in series are equivalent to an elliptical retarder in general.

De…nition 5 Polarization rotator. A rotator rotates the plane of polarization of a linearly polarized wave
by a …xed angle, maintaining its linearly polarized nature.
If a polarization rotator is placed between two polarizers, the amount of transmitted light depends on the
rotation angle.
Jones matrices for some polarizers and retarders are provided in Tables 2 and 3, respectively. Remember
that, in many applications, overall factors do not matter because they do not a¤ect the state of polarization.
14 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Table 3: Linear retarder with retardance = y x and fast axis at angle with the horizontal
Jones matrix 2 Mueller matrix 3
" i # 1 0 0 0
e 2 cos2 + e i 2 sin2 i2 sin 2 sin cos 60 cos2 2 + C4 sin2 H4 sin2 2 H2 sin 7
R = 6 2 7
40 H4 sin2 2 cos2 2 C4 sin2 2 C2 sin 5
i2 sin 2 sin cos e i 2 cos2 + ei 2 sin2
0 H2 sin C2 sin cos
C2 cos 2 ; C4 cos 4 ; H2 sin 2 ; H4 sin 4

Table 4: Quarter-wave plate ( = =2) with fast axis at angle with the horizontal
Quarter-wave plate Jones matrix 2 Mueller matrix 3
1 0 0 0
1 + i cos 2 i sin 2 60 cos 2
2 sin 2 cos 2 sin 2 7
R90 = p12 6 7
i sin 2 1 i cos 2 40 sin 2 cos 2 sin2 2 cos 2 5
0 sin 2 cos 2 0
2 3
1 0 0 0
Quarter-wave plate 60 1 0
0 ;90 e i =4 0 6 077
horizontal fast axis ! R90 = 4
0 e i =4 0 0 0 15
vertical fast axis "
0 0 1 0
2 3
1 0 0 0
Quarter-wave plate 60 0 0
1 1 i 6 17
7
Fast axis at + 45 % R9045 = p 40 0 1 0 5
2 i 1
Fast axis at 45 &
0 1 0 0

3.2.3 Expressing the Jones matrix in other polarization basis

Jones calculus can be developed for an arbitrary basis, although the most usual choice is a laboratory Cartesian
coordinate system in which light propagates along the positive z axis and the optical elements have plane and
parallel surfaces lying in the x-y plane. Thus the transmission relation is given by

E2x J J12 E1x

E2 = = 11 = JE1 ; (54)
E2y J21 J22 E1y

In this expression, the electric …elds are expressed in terms of the standard linear polarization basis. However,
as discussed in Sect. 3.1.1, E could be expressed in terms of any other orthogonal polarization basis, i.e.
E = Uu ^+Vv ^:
If T represents the linear transformation between the components (Ex ; Ey ) in the Cartesian basis and the
components (U; V ) in the (^ u; v
^) basis, that is

Ex U T T12 U
=T = 11 ; (55)
Ey V T21 T22 V

then Eq. (54) can be recasted as

1
U2 U1 T T12 J11 J12 T11 T12 U1
=T 1
JT = 11 : (56)
V2 V1 T21 T22 J21 J22 T21 T22 V1

The matrix
1
1 T11 T12 J11 J12 T11 T12
K=T JT= ; (57)
T21 T22 J21 J22 T21 T22
corresponds to the Jones matrix of the device in the polarization basis (^
u; v
^).
Lecture notes on Polarization 15

Table 5: Half-wave plate ( = ) with fast axis at angle with the horizontal
Half-wave plate Jones matrix 2 Mueller matrix 3
1 0 0 0
Half-wave plate ( = ) 60 cos 4
cos 2 sin 2 6 sin 4 077
Fast axis at angle R180 = 40 sin 4
sin 2 cos 2 cos 4 05
with the horizontal
0 0 0 1
2 3
1 0 0 0
Half-wave plate 60 1 0
1 0 6 07 7
Fast axis horizontal R0180;90 = 40 0
0 1 1 05
vertical
0 0 0 1
2 3
1 0 0 0
Half-wave plate 60
45 0 1 6 1 0 07 7
Fast axis at + 45 % R180 = 4
1 0 0 0 1 05
Fast axis at 45 &
0 0 0 1

Table 6: Jones and Mueller matrices for circular retarders

Circular retarder Jones matrix 2 Mueller matrix 3
1 0 0 0
Circular retarder. Retardance 60 cos
cos ( =2) sin ( =2) 6 sin 077
Left or positive helicity CR = 4
sin ( =2) cos ( =2) 0 sin cos 05
Right or negative helicity
0 0 0 1

Example. For the case of the circular polarization basis we have

Ex 1 1 1 E+
=p : (58)
Ey 2 i i E

Therefore the Jones matrix is given by

1
1 1 1 J11 J12 1 1 1 J11 + J22 + i (J12 J21 ) J11 J22 + i ( J12 J21 )
K= = : (59)
2 i i J21 J22 i i 2 J11 J22 + i (J12 + J21 ) J11 + J22 + i ( J12 + J21 )

The input-output relation takes the form

E2+ K11 K12 E1+

= ; (60)
E2 K21 K22 E1

where the vectors are the amplitudes of circular polarization basis.

3.3 Illustrative applications of the Jones calculus

The Jones representation allows to analyze the operations of polarization devices by simple application of matrix
algebra. In Fig. 7 we show some typical operations of quarter–and half–wave retarders over linear and circular
polarized waves.

Example 1 (Circular polarization from linear polarization) A quarter-wave ( =2) retarder with hori-
zontal fast axis transforms a linearly polarized wave at +45 into a right-handed circular state, see Fig. 6(a).

1 0 1 1 ei =4 1
ei =4
p = p : (61)
0 i 2 1 2 i
16 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Example 2 (Linear polarization from circular polarization) A quarter-wave ( =2) retarder with vertical
fast axis transforms a left-handed circular state into a linearly polarized wave at 45 , see Fig. 6(b).

i =4 1 0 1 1 e i =4 1
e p = p : (62)
0 i 2 i 2 1

Example 3 (Rotating 90 the linear polarization.) A linearly polarized wave at arbitrary with the hor-
izontal is rotated 90 by a half-wave retarder ( ) whose fast axis makes an angle of 45 with the polarization
of the linearly polarized wave, see Fig. 6(c).

cos 2 4 sin 2 4 cos sin cos 2

= = : (63)
sin 2 4 cos 2 4
sin cos sin 2

Example 4 (Inverting the handedness of the circular polarization) A half-wave ( ) retarder with fast
axis at arbitrary angle with the horizontal inverts the handedness of the circular polarization, see Fig. 6(d).

cos 2 sin 2 1 1 e i2 1
p = p : (64)
sin 2 cos 2 2 i 2 i

Figure 7: Operations of the quarter-wave ( =2) and the half-wave ( ) retarders. Fast and slow axes are labeled
with letters F and S, respectively.

Example 5 (Malus law) Malus law states that when a perfect polarizer is placed in a linearly polarized wave,
the intensity, I, of the wave that passes through is given by I = I0 cos2 ; where I0 is the initial intensity and
is the angle between the light’s initial polarization direction and the axis of the polarizer. The demonstration
of this law is quite simple using the Jones formalism. Let us assume that a linear horizontally polarized wave
incides over a polarizer with transmission axis at with the horizontal, the output wave is then given by

cos2 sin cos 1 cos2

= : (65)
sin cos sin2 0 sin cos
Lecture notes on Polarization 17

The intensity of the output wave takes the value

cos2
I = cos2 sin cos = cos4 + sin2 cos2 = cos2 ; (66)
sin cos

which con…rms the Malus law.

Example 6 (E¤ect of the circular retarders) Consider the e¤ ect of a circular retarder with retardance
on linearly polarized …elds. Let the input …eld be oriented at an angle with respect to the x axis. The output
wave is
cos ( =2) sin ( =2) cos cos ( =2)
= : (67)
sin ( =2) cos ( =2) sin sin ( =2)
From this, we see that the e¤ ect of a circular retarder is to rotate linearly polarized waves by an angle =2 and
the intensity is not changed. This is reason why the circular retarder are often called rotators.

Example 7 (Re‡ection and transmission at the interfase of two dielectric media) Consider the clas-
sical problem of the re‡ection and transmission of an electromagnetic wave at the planar boundary between two
dielectrics with refractive indices n1 and n2 : As it is well known, the re‡ection and transmission Fresnel coe¢ -
cients, rT M ; rT E ; tT M ; tT E ; of the TM and TE components of the incident wave are functions of the incidence
angle : If we identify the TM component as "horizontal" and the TE component as "vertical", then we can
apply the Jones formalism by de…ning the re‡ection and transmission matrices

rT M ( ) 0 tT M ( ) 0
; : (68)
0 rT E ( ) 0 tT E ( )

The Jones vectors of the re‡ected and transmitted waves are then

ErT M r ( ) 0 EiT M EtT M t ( ) 0 EiT M

= TM ; = TM : (69)
ErT E 0 rT E ( ) EiT E EtT E 0 tT E ( ) EiT E

Consider, for example, the case of normal incidence. For = 0; the re‡ection matrix is

(n 1) = (n + 1) 0 1 n 1 0
= ; (70)
0 (1 n) = (n + 1) 1+n 0 1

where n = n2 =n1 is the relative refractive index. Suppose that the incident wave is right circularly polarized, the
Jones vector of the re‡ected wave is then

1 n 1 0 1 n 1 1
= : (71)
1+n 0 1 i 1+n i

Thus the re‡ected wave is left circularly polarized, and its amplitude is changed by the factor (n 1) = (n + 1) :
This reversal of the handedness of the circular polarization is independent of the value of n:

4 Stokes-Mueller description of the polarization

The description of polarization in terms of the polarization ellipse is useful. It allows to describe any state of
completely polarized waves by means of a single equation [Eq. (3)]. However, under some circumstances this
representation can be incomplete for several reasons:

1. The polarization ellipse is only applicable to describing waves that are completely polarized.

2. At a given point in space, the electric vector traces out an ellipse whose time period is of order 10 15 sec.
for visible light. This period is clearly too short to allow us to follow the tracing of the ellipse as the wave
propagates and also prevents us from following the polarization ellipse in the time domain.
18 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

3. The polarization ellipse is an amplitude description of polarized waves that can not be observed and
measure in a direct way. The measurable quantities are the time average of the square of the …eld
amplitudes, i.e. the intensities.

It is then clear the necessity of establishing an e¢ cient description of polarization based on observable
(measurable) quantities that also allows the possibility of describing partial-polarized and complete unpolarized
waves.

4.1 Statistical derivation of the Stokes parameters

The Stokes parameters for describing the polarization of light in terms of measurable quantities were introduced
by G. G. Stokes in 1852 [9]. There are several ways for introducing the Stokes parameters. In this section, we
follow a standard procedure which directly relates the parameters of the polarization ellipse used in the past
section with the new Stokes parameters [6].
At a given point in space, consider the following orthogonal components of the electric vector of a wave not
necessarily monochromatic

Ex (t) = E0x (t) cos [ !t + x (t)] ; (72a)

Ey (t) = E0y (t) cos !t + y (t) ; (72b)

where E0x (t) and E0y (t) are the instantaneous amplitudes and x (t) and y (t) are the instantaneous phase
factors. At all times the amplitudes and phase factors ‡uctuate slowly compared to the rapid vibrations of
the cosine functions. As we saw in the past section, removal of the term !t in Eqs. (72) leads to the familiar
polarization ellipse which is valid, in general, only at a given instant of time

Ex2 (t) Ey2 (t) Ex (t) Ey (t)

2 (t) + 2 (t) 2 cos (t) = sin2 (t) ; (73)
E0x E0y E0x (t) E0y (t)

where (t) = y (t) x (t) :

If the wave is monochromatic, the amplitudes and phases are constant for all time, so (73) reduces to

Ex2 (t) Ey2 (t) Ex (t) Ey (t)

2 + 2 2 cos = sin2 : (74)
E0x E0y E0x E0y

To transform Eq. (73) into an intensity (or observable) representation, we …rst take a time average of the
time–dependent quantities in Eq. (73),

Ex2 (t) Ey2 (t) hEx (t) Ey (t)i

2 + 2 2 cos = sin2 ; (75)
E0x E0y E0x E0y

where the angle brackets denote the time average of two time-dependent quantities de…ned by
Z
1
hEx (t) Ey (t)i = lim Ex (t) Ey (t) dt; (76)
!1 0

with being the time of measurement.

2 2
To evaluate time averages, we now multiply by 4E0x E0y to remove the denominator, we get
2 2
4E0y Ex2 (t) + 4E0x
2
Ey2 (t) 8E0x E0y hEx (t) Ey (t)i cos = (2E0x E0y sin ) : (77)

Using Eq. (76), the time averages of the above terms take the values
1 2 1 2 1
Ex2 (t) = E ; Ey2 (t) = E ; hEx (t) Ey (t)i = E0x E0y cos : (78)
2 0x 2 0y 2
Replacing these values we get
2 2 2 2
4E0x E0y (2E0x E0y cos ) = (2E0x E0y sin ) : (79)
Lecture notes on Polarization 19

2 2
In order to describe both the intensity and the polarization, the term E0x +E0y should be appear in the equation.
4 4
If E0x + E0y is added to both sides, the equation becomes

2 2 2 2 2 2 2 2
E0x + E0y E0x E0y (2E0x E0y cos ) = (2E0x E0y sin ) : (80)

Each term in this equation can be identi…ed with a Stokes parameter 2 . Arranging the parameters in a
column vector we have
2 3 2 2 2
3
S0 E0x + E0y
6S1 7 6 E0x 2 2
E0y 7
S=6 7=6
4S2 5 42E0x E0y cos 5.
7 (81)
S3 2E0x E0y sin

This vector is called the Stokes vector for an elliptically polarized plane wave. Therefore, Eq. (80) can be
written as
S02 = S12 + S22 + S32 : (82)

Some remarks are in order:

We see that the Stokes parameters are simply the observables of the polarization ellipse.

The Stokes vector has all real elements and gives information about intensity properties of the beam.
Thus it is not able to handle problems involving phase changes or combinations of two beams that are
coherent .

Equality (82) is valid only for fully polarized light.

4.2 Physical interpretation of the Stokes parameters

To gain physical insight of the Stokes parameters, let us …rst express the monochromatic wave E in terms of the
linear polarization basis (^x; y
^), the circular polarization basis (^
c+ ; ^
c ), and the basis of 45 linearly polarized
waves (^u; v
^) (i.e. positive u
^ points in direction of 45 and positive v ^ points in direction of 135 ) we have

E = Ex x
^ + Ey y
^; E = E+ ^
c+ + E ^
c ; E = Eu u
^ + Ev v
^: (83)

The inner products

x; Ei = Ex = E0x ei x ;
h^ y; Ei = Ey = E0y ei y ;
h^
c+ ; E = E+ = E0+ ei
^ + ; c ; E = E = E0 ei
^ ; (84)
u; Ei = Eu = E0u ei u ;
h^ v; Ei = Ev = E0v ei v ;
h^

are the amplitude components of E with linear polarization in the x and y directions, with circular polarization
in the positive and negative directions, and with linear polarization in the 45 and 135 directions, respectively.
The squares of these amplitudes give a measure of the intensity of each type of polarization.
In terms of the linear polarization basis (^
x; y
^), the Stokes parameters of a monochromatic wave are
2 2 22
S0 = jEx j + jEy j = E0x + E0y ;
2 22 2
S1 = jEx j jEy j = E0x E0y ; (85)
S2 = 2 Re (Ex Ey ) = 2E0x E0y cos y x ;
S3 = 2 Im (Ex Ey ) = 2E0x E0y sin y x :
2 The notation for the Stokes parameters is not uniform. Stokes himself used fA; B; C; Dg ; other labelings are fI; Q; U; V g and

fI; M; C; Sg : Here we adopt the most common notation used by Born and Wolf [1].
20 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

In terms of the linear polarization basis (^

u; v
^), the Stokes parameters of a monochromatic wave are
2 2 2 2
S0 = jEu j + jEv j = E0u + E0v ;
S1 = 2 Re (Eu Ev ) = 2E0u E0v cos ( v u) ; (86)
2 2 2 2
S2 = jEu j jEv j = E0u E0v ;
S3 = 2 Im (Eu Ev ) = 2E0u E0v sin ( v u) :

In terms of the circular polarization basis (^

c+ ; ^
c ), the Stokes parameters of a monochromatic wave are
2 2 2
S0 = jE+ j + jE j = E0+ + E02 ;
S1 = 2 Re E+ E = 2E0+ E0 cos + ; (87)
S2 = 2 Im E+ E = 2E0+ E0 sin + ;
2 2 2
S3 = jE+ j jE j = E0+ E02 :

From these expressions we can get the following physical interpretation of the Stokes parameters:

1. S0 is the intensity of the optical wave.

2. S1 describes the preponderance of the intensity of linearly horizontal polarized light over linearly vertical
polarized light. Actually, S1 may describe the degree of plane polarization with respect to two arbitrary
orthogonal axes.
3. S2 describes the preponderance of the intensity of the +45 linearly polarized light over 135 linearly
polarized light.
4. S3 describes the preponderance of the intensity of positive (left-handed) circularly polarized light over
negative (right-handed) circularly polarized light.

The Stokes vectors for some special polarization states of light can be found in Table 1.
Actually, S1 may describe the degree of plane polarization with respect to two arbitrary orthogonal axes and
S2 the degree of plane polarization with respect to a set of axes oriented at 45 to the right of the previous one.

4.3 De…nition of the Stokes parameters in terms of di¤erent parameters of the

polarization ellipse
In addition to Eqs. (85)-(87), the Stokes parameters can be also de…ned in terms of the di¤erent parameters of
the polarization ellipse, see Fig. 8. The following four cases are relevant.

Figure 8: Relevant parameters of the polarization ellipse.

Lecture notes on Polarization 21

4.3.1 Standard de…nition in terms of parameters fE0x ; E0y ; g

If the polarization ellipse is characterized by the set of parameters fE0x ; E0y ; g ; the Stokes parameters are
directly given by Eq. (81), namely
2 3 2 2 2
3
S0 E0x + E0y
6S1 7 6 2
E0x E0y 2 7
S=6 7 6
4S2 5 = 4
7:
5 (88)
2E0x E0y cos
S3 2E0x E0y sin
The inverted relations are given by
r r
S0 + S1 S0 S1 S3 S2 + iS3
E0x = ; E0y = ; = arctan ; ei = p 2 : (89)
2 2 S2 S2 + S32

2 2
4.3.2 De…nition in terms of the observables Emax ; Emin ;
Consider now that the polarization ellipse is characterized by the set of parameters fEmax ; Emin ; g ; where
the positive (negative) sign of Emin denotes positive (negative) helicity of the wave. Replacing the relations
between fEmax ; Emin ; g and fE0x ; E0y ; g [Eq. (11)] into Eq. (88) we obtain after some simpli…cations
2 3 2 2 2
3
S0 Emax + Emin
6S1 7 6 Emax 2 2
Emin cos 2 7
S=6 7=6
4S2 5 4 Emax Emin sin 2 5 :
2 2
7 (90)
S3 2Emax Emin
Solving for fEmax ; Emin ; g we get
q q
2 1 2 1 S2
Emax = S0 + S12 + S22 ; Emin = S0 S12 + S22 ; 2 = arctan : (91)
2 2 S1

4.3.3 De…nition in terms of the wave intensity and the inclination and ellipticity angles fI; ; g:
the Poincaré sphere
Let us consider now that the polarization ellipse is characterized by the intensity I and the inclination and
ellipticity angles f ; g ; see Fig. 8. where the intensity and ellipticity angle are de…ned by
Emin h i
2 2
I = Emax + Emin > 0; = arctan 2 ; : (92)
Emax 4 4
Factoring out the intensity from Eq. (90) we have the partial results
2 2
2
Emax 2
Emin 1 Emin =Emax 1 tan2
2 = 2 = = cos 2 ; (93)
2
Emax + Emin 2
1 + (Emin =Emax ) 1 + tan2
2Emax Emin 2Emin =Emax 2 tan
2 = 2 =E 2 = = sin 2 : (94)
2
Emax + Emin 1 + (Emin max ) 1 + tan2
Therefore the Stokes parameters can be written as
2 3 2 3
S0 I
6S1 7 6 I cos 2 cos 2 7
S=6 7 6
4S2 5 = 4 I cos 2 sin 2
7:
5 (95)
S3 I sin 2
Solving for fI; ; g we get
!
1 S3 h i 1 S2 h i
I = S0 ; = arctan p 2 ; ; = arctan 2 ; : (96)
2 S12 + S22 4 4 2 S1 2 2
22 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Figure 9: (a) The Cartesian axes of a Poincaré sphere are given by the Stokes parameters fS1 ; S2 ; S3 g. (b) The
Poincaré sphere for representing the states of polarization of a plane monochromatic wave.

The expressions for the Stokes parameters [Eq. (95)] in terms of fI; ; g suggest a geometrical visualization
of the states of polarization: fS1 ; S2 ; S3 g may be regarded as the Cartesian coordinates of a point P on a
sphere of radius S0 such 2 and 2 are the azimuthal and equatorial angles of a spherical coordinate system,
see Fig. 9(a). Each point on the surface of the sphere corresponds to a particular state of polarization of a
monochromatic wave, and vice versa. Consequently
The equatorial plane (S3 = 0) is the locus of linearly polarized waves.
The upper hemisphere (S3 > 0) corresponds to polarization sates with positive (left-handed) helicity. The
lower hemisphere (S3 < 0) corresponds to polarization sates with negative (right-handed) helicity.
North and south poles (S1 = S2 = 0) represent positive and negative circular polarization states, respec-
tively.
Mirror symmetry of the polarization state S with respect to the plane of equator preserves the shape of
the polarization ellipse but inverts the sense of rotation.
Mirror symmetry of the polarization state S1 with respect to the origin yields the orthogonal state, in
other words, if two polarization states S1 and S2 are diametrically opposed, i.e.
2 3 2 3
S0 S0
6S1 7 6 S1 7
S1 = 6 7
4S2 5 ; S2 = 6 7
4 S2 5 ; (97)
S3 S3
then are orthogonal. This result can be corroborated by noting from Eqs. (89) that the Jones vectors of
S1 and S2 are
2 1=2
3 2 1=2
3
S0 + S1 S0 S1
6 7 6 7
6 2 7 6 2 7
E1 = 6
6 S
!7 ;
7 E = 6
6
!7; (98)
S2 + iS3 7
1=2 2 1=2
4 0 S1 p
S2 + iS3 5 4 S0 + S1
p 5
2 S22 + S32 2 S22 + S32
and that the inner product hE2 ; E1 i vanishes.
Lecture notes on Polarization 23

4.3.4 De…nition in terms of the wave intensity and the retardance and rectangle angles fI; ; #g:
the observable Poincaré sphere
In the traditional description of the Poincaré sphere [Eq. (95)], the change of the relative phase between the
Cartesian components of the electric …eld leads to a simultaneous change in the inclination and ellipticity
angles. It is desirable to have a parameterization for which the change in corresponds to a simple transformation
on the surface of the Poincaré sphere. This property can be achieved by de…ning the Stokes parameters in terms
of the angle itself and the angle
E0y h i
# = arctan 2 0; : (99)
E0x 2
see Fig. 8. Factoring out the intensity from Eq. (88) we have the partial results
2 2 2 2
E0x E0y 1 E0y =E0x 1 tan2 #
2 + E2 = = = cos 2#; (100)
E0x 0y
2 =E 2
1 + E0y 0x 1 + tan2 #
2E0x E0y 2E0y =E0x 2 tan #
2 + E2 = = = sin 2#: (101)
E0x 0y
2 =E 2
1 + E0y 0x 1 + tan2 #

Figure 10: Poincaré sphere in terms of the angles f ; #g :

Therefore the Stokes parameters can be written as

2 3 2 3
S0 I
6S1 7 6 I cos 2# 7
S=6 7 6
4S2 5 = 4
7:
5 (102)
I sin 2# cos
S3 I sin 2# sin

Solving for fI; ; g we get

p !
S22 + S32 S3
I = S0 ; 2# = arctan ; = arctan : (103)
S1 S2

Equations (102) suggest to construct a Poincaré sphere to visualize the behavior of the polarization states
as a function of f ; #g : In this case f2#; g play the roles of the polar and the azimuthal angles of a spherical
coordinate system with axes (x; y; z) ! fS2 ; S3 ; S1 g ; see Fig. 10.
24 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Each point on the surface of the sphere corresponds to a particular state of polarization of a monochromatic
wave, and vice versa. It is important to emphasize that the spheres in Figs. 9 and 10 are not two di¤erent
spheres, actually they are the same sphere. The di¤erence falls in the fact that we are using di¤erent angular
parameters to specify a point on the sphere. In Fig. 10(b), the linearly polarized waves appears along the prime
meridian = 0: The north and south poles corresponds to horizontally and vertically polarized waves. The
sphere in Fig. 10 is often refereed by some authors to as the observable polarization sphere [7, Chap. 7].

4.3.5 Relations between angles f ; g and angles f ; #g

The de…nitions (95) and (102) enable us to determine the relations between the angles f ; g and f ; #g :
Equating both sets of equations, we get

cos 2# = cos 2 cos 2 ; (104)

sin 2# cos = cos 2 sin 2 ; (105)
sin 2# sin = sin 2 : (106)

The combinations of these relations yield

tan 2 = tan 2# cos ; (107)

tan 2 = sin 2 tan : (108)

4.3.6 Stokes parameters in terms of the Pauli spin matrices

The electric …eld of a fully polarized wave is represented by the Jones vector

Ex
E= ; Ey = Ex Ey : (109)
Ey

The Stokes parameters may be expressed in a more compact form in terms of the Pauli matrices which we
take to be 3
1 0 1 0 0 1 0 i
0 ; 1 ; 2 ; 3 : (110)
0 1 0 1 1 0 i 0
We have
Sj = Ey jE = Tr (J j) ; J EEy ; (111)
where Tr (A) stands for the trace of the matrix A; i.e. the sum of the elements on the main diagonal. Explicitly
2 3 2 y 3
S0 E 0 E
6 S1 7 6 Ey E 7
S=6 7 6
4 S2 5 = 4 Ey
1 7. (112)
2 E 5
S3 Ey 3 E

4.4 Propagation of the Stokes vectors through polarization devices: Mueller cal-
culus
In the same way that Jones vectors can be propagated through an optical systems using matrix products, the
e¤ect of a polarization device on a Stokes vector can be easily calculated with matrix algebra. This technique
is customarily called Mueller calculus and constitutes a matrix method for manipulating Stokes vectors. Light
which is unpolarized or partially polarized must be treated using Mueller calculus, while fully polarized light
can be treated with either Mueller calculus or the simpler Jones calculus.
3 Note that our de…nition of the Pauli matrices difers with the standard ordering of the Pauli matrices in quantum mechanics

which is ( 0 ; 2 ; 3 ; 1 ) : This is because we are working with Jones vectors expressed in Cartesian basis vectors rather than spinors
expressed in circular polarization basis, although it is clear that neither convention is reallv more fundamental than the other. It
is somewhat more practical to associate 3 with the z axis rather than with the x axis
Lecture notes on Polarization 25

The e¤ect of a particular optical element is represented by a Mueller matrix ; which is a 4 4 matrix
2 3
m11 m12 m13 m14
6m21 m22 m23 m24 7
M=6 4m31 m32 m33
7: (113)
m34 5
m41 m42 m43 m44

The matrix contains 16 real parameters, but there are only 7 independent parameters. The matrix contains no
information about absolute phase, but it handles partially polarized and unpolarized light without modi…cation.
If a wave with Stokes vector Sin passes through an optical element with Mueller matrix M, then the Stokes
vector Sout of the output wave is given by the matrix product

Sout = M Sin : (114)

Rotated devices. If an optical element is rotated about the optical axis by angle ; the Mueller matrix
M ( ) for the rotated element is constructed from the matrix M for the unrotated element by the transformation

M( ) = W ( ) M W ( ); (115)

where 2 3
1 0 0 0
60 cos 2 sin 2 07
W( ) 6 7; (116)
40 sin 2 cos 2 05
0 0 0 1
T
is the rotation matrix. If a coordinate system the values of a Stokes vector are S0 S1 S2 S3 ; then, in a
rotated system of coordinates, the Stokes parameters of the same beam are
2 3
S0
6 S1 cos 2 + S2 sin 2 7
6 7
4 S1 sin 2 + S2 cos 2 5 : (117)
S3

Thus the parameters S0 and S3 are invariant under rotations.

Cascaded devices. If a polarized wave is sent through a train of optical devices (each with its own
rotation), then the overall transformation matrix is given by the matrix product

M = Mn Mn 1 M2 M1 : (118)

Physical realizability. The following are four necessary conditions for physical realizability of a Mueller
matrix [7, 20, Sect. 13.7]

Tr M MT 4m211 ; (119a)
m11 jmi;j j ; (119b)
m211 2
; (119c)
4 4
!
2
X X mjk m1k
(m11 ) m1;j ; (119d)
j=2 k=2

1=2
where = m212 + m213 + m214 :
The Mueller matrices for some ideal common devices have been included in Tables 2 to 6.
For every matrix in Jones formalism there is a matrix in Mueller calculus, but the converse in not true. For
example, a depolarizer can be described in Mueller calculus, but there is no matrix for such a device in Jones
calculus.
26 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

4.5 Relationship between Mueller and Jones matrices

For fully polarized waves, calculations by the Mueller and Jones calculi are completely equivalent leading always
to the same …nal results, so there must be a close relationship between the two approaches. In this section we
show how the elements of a Mueller matrix can be calculated from the known elements of a Jones matrix and
vice versa, provided that the original matrices represent physically possible polarizing devices [19].

4.5.1 Calculating the Mueller matrix from the Jones matrix

Let S , Sout and Ein , Eout be the Stokes vectors and the corresponding Jones vectors at the input and output
in

of a device characterized by the Mueller M and J Jones matrices. The equivalent input-output transformations
are 2 out 3 2 3 2 in 3
S0 m11 m12 m13 m14 S0
6S1out 7 6m21 m22 m23 m24 7 6S1in 7 Exout J11 J12 Exin
6 out 7 = 6 7 6 in 7 ; out = : (120)
4S2 5 4m31 m32 m33 m34 5 4S2 5 Ey J21 J22 Eyin
S3out m41 m42 m43 m44 S3in
Inserting Eqs. (112) for the Stokes parameters in terms of the Jones elements, we obtain
2 3 2 yin in
3 2 yin 3
m11 m12 m13 m14 E 0 E E (m11 0 + m12 1 + m13 2 + m14 3) Ein
6m21 m22 m23 7 6
m24 7 6 E yin
1 E
in 7 6 Eyin (m21 0 + m22 1 + m23 2 + m24 3) E
in 7
S =6
out
4m31 m32 m33 5 4 yin in
7 = 6 yin
5 4 E
7
in 5 :
m34 E 2 E (m31 0 + m32 1 + m33 2 + m34 3) E
yin in
m41 m42 m43 m44 E 3 E Eyin (m41 0 + m42 1 + m43 2 + m44 3) E
in

(121)
Substituting the values of the Pauli matrices, the terms in parenthesis become
mj1 + mj2 mj3 imj4
w(j) mj1 0 + mj2 1 + mj3 2 + mj4 3 = ; j = f1; 2; 3; 4g : (122)
mj3 + imj4 mj1 mj2

Thus, Sout can be written in terms of Ein and Eout as follows:

2 yout 3 2 yin 3
E 0 E
out E w(1) Ein
6 E yout
E out 7 6 Eyin w(2) Ein 7
Sout = 6 1
4 Eyout 2 Eout
7=6
5 4 Eyin
7: (123)
w(3) Ein 5
Eyout 3 Eout Eyin w(4) Ein
On the other hand, the Jones transformation of the device is
J11 J12
Eout = J Ein = Ein : (124)
J21 J22

Multiplying both sides by Eyout j and taking into account that the Hermitian (i.e. the complex conjugate)
of Eout is given by Eyout = Eyin Jy we get
Eyout j Eout = Eyout j J Ein = Eyin Jy j J Ein = Eyin Jy jJ Ein ; (125)
where we have used the associative property of the matrix multiplication. Explicitly we write
2 yout 3 2 yin 3
E 0 E
out E Jy 0 J Ein
6 E yout
1 E
out 7 6 E yin
Jy 1 J Ein 7
Sout = 64 E
7=6
5 4 E
7: (126)
yout
2 E
out yin
Jy 2 J Ein 5
yout out in
E 3 E Eyin Jy 3 J E
The left sides of Eqs. (123) and (126) are identical, so the right sides are identical, therefore we can establish
the following equality
w(j) = Jy j 1 J; (127a)
mj1 + mj2 mj3 imj4 J11 J21 J11 J12
= j 1 : (127b)
mj3 + imj4 mj1 mj2 J12 J22 J21 J22
Lecture notes on Polarization 27

This is a system of four equations for the unknowns mj1 ; mj2 ; mj3 and mj4 whose values may be calculated by
suitable summations and substrations of the elements of the matrix Jy j 1 J; thus
(j) (j) (j) (j) (j) (j) (j) (j)
w11 + w22 w11 w22 w12 + w21 w12 w21
mj1 = ; mj2 = ; mj3 = ; mj4 = i : (128)
2 2 2 2
After replacing the values of the Pauli matrices for j = f1; 2; 3; 4g we obtain the four systems of equations

m11 + m12 m13 im14 J J + J21 J21 J11 J12 + J21 J22
= 11 11 ; (129a)
m13 + im14 m11 m12 J12 J11 + J22 J21 J12 J12 + J22 J22
m21 + m22 m23 im24 J J J21 J21 J11 J12 J21 J22
= 11 11 ; (129b)
m23 + im24 m21 m22 J12 J11 J22 J21 J12 J12 J22 J22
m31 + m32 m33 im34 J J + J21 J11 J11 J22 + J21 J12
= 11 21 ; (129c)
m33 + im34 m31 m32 J12 J21 + J22 J11 J12 J22 + J22 J12
m41 + m42 m43 im44 J J J11 J21 J21 J12 J11 J22
= i 21 11 ; (129d)
m43 + im44 m41 m42 J22 J11 J12 J21 J22 J12 J12 J22

Finally, adding and subtracting suitable pairs of equations in (129) we get the Mueller from the Jones matrix
elements, namely

m11 = (J11 J11 + J21 J21 + J12 J12 + J22 J22 ) =2; (130a)
m12 = (J11 J11 + J21 J21 J12 J12 J22 J22 ) =2; (130b)
m13 = (J11 J12 + J21 J22 + J12 J11 + J22 J21 ) =2; (130c)
m14 = i (J11 J12 + J21 J22 J12 J11 J22 J21 ) =2 (130d)

m21 = (J11 J11 J21 J21 + J12 J12 J22 J22 ) =2; (130e)
m22 = (J11 J11 J21 J21 J12 J12 + J22 J22 ) =2; (130f)
m23 = (J11 J12 J21 J22 + J12 J11 J22 J21 ) =2; (130g)
m24 = i (J11 J12 J21 J22 J12 J11 + J22 J21 ) =2 (130h)

m31 = (J11 J21 + J21 J11 + J12 J22 + J22 J12 ) =2; (130i)
m32 = (J11 J21 + J21 J11 J12 J22 J22 J12 ) =2; (130j)
m33 = (J11 J22 + J21 J12 + J12 J21 + J22 J11 ) =2; (130k)
m34 = i (J11 J22 + J21 J12 J12 J21 J22 J11 ) =2 (130l)

m41 = i ( J11 J21 + J21 J11 J12 J22 + J22 J12 ) =2; (130m)
m42 = i ( J11 J21 + J21 J11 + J12 J22 J22 J12 ) =2; (130n)
m43 = i ( J11 J22 + J21 J12 J12 J21 + J22 J11 ) =2; (130o)
m44 = (J11 J22 J21 J12 J12 J21 + J22 J11 ) =2 (130p)

These expressions were …rst reported by Parke [15].

4.5.2 Calculating the Jones matrix from the Mueller matrix

The Jones elements are complex in general. For the derivation, it is convenient to express them in polar form,
i.e.
J11 J12 jJ11 j jJ12 j exp [i ('12 '11 )]
= exp (i'11 ) : (131)
J21 J22 jJ12 j [i ('21 '11 )] jJ22 j [i ('22 '11 )]
28 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

The magnitudes of the Jones elements can be calculated from Eqs. (129). For example, by adding the
elements (1; 1) of (129a) and (129b) we may obtain jJ11 j : The results are [16]
2
jJ11 j = (m11 + m12 + m21 + m22 ) =2; (132a)
2
jJ12 j = (m11 m12 + m21 m22 ) =2; (132b)
2
jJ21 j = (m11 + m12 m21 m22 ) =2; (132c)
2
jJ22 j = (m11 m12 m21 + m22 ) =2: (132d)

For the angles, it is enough to …nd the di¤erence between one of them (let us choose '11 ), and the other
three, because adding a phase factor is irrelevant for the Jones matrix.
Adding Eqs. (130c) and (130g)

m13 + m23 = J11 J12 + J12 J11 = 2 jJ11 j jJ12 j cos ('12 '11 ) ; (133)

therefore using Eqs. (132) we obtain

m13 + m23
cos ('12 '11 ) = h i1=2 : (134)
2 2
(m11 + m21 ) (m12 + m22 )

Similarly, adding m14 + m24 we get

m14 + m24
sin ('12 '11 ) = h i1=2 : (135)
2 2
(m11 + m21 ) (m12 + m22 )

Combining the expressions for the cosine and sine we …nally get

m13 + m23 i (m14 + m24 )

exp [i ('12 '11 )] = h i1=2 : (136)
2 2
(m11 + m21 ) (m12 + m22 )

The other two angles can be found in the same way, the results are

m31 + m32 + i (m41 + m42 )

exp [i ('21 '11 )] = h i1=2 : (137)
2 2
(m11 + m21 ) (m12 + m22 )
m33 + m44 + i (m43 + m34 )
exp [i ('22 '11 )] = h i1=2 (138)
2 2
(m11 + m21 ) (m12 + m22 )

These formulae are valid while the Mueller matrix can be represented as a Jones matrix. Remember that
to every Jones matrix which can be written down there corresponds a Mueller matrix, but the inverse situation
is not true in general. This happens, for example, if one considers the matrix of an ideal depolarizer, for which
m11 = 1 and all other elements vanish.

5 Unpolarized and partially polarized light

Stokes parameters not only describes completely polarized light but also unpolarized light and partially polarized
light as well.

5.1 Degree of polarization of a wave

Unpolarized light (often called natural light) is a light that is una¤ected (except possibly by a constant attenua-
tion factor) when passed through a retarder and/or a linear polarizer even if either or both elements are rotated
Lecture notes on Polarization 29

around their longitudinal axis. In other words, at a given point, the electric …eld of unpolarized light vibrates
randomly without exhibiting any preponderance along a particular linear direction or circular helicity, that is
2 2
E0x + E0y = I; (139)
2 2
E0x E0y = 0; hcos i = 0; hsin i = 0: (140)

Therefore, for normalized intensity, the Stokes vector reads as

2 3 2 3
S0 1
6S1 7 607
S=6 7 6 7
4S2 5 = 405 : (141)
S3 0

Thus, the Stokes parameters can be used to describe the extreme states of polarized light: completely
polarized light and unpolarized light. Within the framework of the Poincaré sphere, the unpolarized light
corresponds to the origin.
The intermediate situation between completely polarized and unpolarized light is called partially polarized
light. In general, for a partially polarized wave

S02 S12 + S22 + S32 ; (142)

where the equality S02 = S12 + S22 + S32 is valid only for completely polarized light.
The degree of polarization is de…ned by
p
Ipol S12 + S22 + S32
DOP = p = = 2 [0; 1] ; (143)
Itot S0
p
where Itot = S0 is the light intensity and Ipol = S12 + S22 + S32 . If p = 0 then the light is unpolarized and
p = 1, then the light is completely polarized (elliptically polarized).
In general, a partially polarized light can be decomposed into a completely polarized wave and unpolarized
wave. The intensity of the polarized wave is Ipol = pS0 and the intensity of the unpolarized light is Itot Ipol =
(1 p) Itot : Therefore we can write the Stokes vector for partially polarized light as an incoherent superposition
of unpolarized an completely polarized light
2 3 2 3 2 3
S0 pS0 (1 p) S0
6S1 7 6 S1 7 6 0 7
S=6 7 pol
4S2 5 = S + S
unpol
=6 7 6
4 S2 5 + 4
7
5 (144)
0
S3 S3 0

The polarization state of partially polarized light travelling through an optical system can be analyzed with
the Mueller calculus explained in the past section.

5.2 Understanding the partial polarization

To gain physical insight into the partial polarization, it is instructive to see the time variation of a fully polarized
wave E0 with frequency ! 0 that is perturbed by N weak waves with random amplitudes, phases and frequencies.
At a given point in space, the electric vector E (t) is given by the superposition

N
X
Ex (t) A0 Aj
E (t) = = E0 (t) + Epert (t) = e i! 0 t
+ e i! j t
; (145)
Ey (t) B0 ei 0
Bj ei j
j=1

where, in general, the parameters Aj ; Bj and j will vary with time.

30 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

Let us now to assume that the fully polarized wave E0 is characterized by the following Jones and Stokes
vectors: 2 3
2
1 6 0 7
E0 (t) = i50 e i2 t ; S0 = 6 7
41:2865 ; (146)
1e
1:532
which correspond to a left-handed elliptical polarized wave with frequency f = 1 Hz, see the red curve in Fig.
11(a).

Figure 11: Time evolution of a partially polarized wave.

For numerical purposes we sum N = 500 perturbing waves with random amplitudes and phases randomly
(uniformly) distributed in the ranges fAj ; Bj g 2 ( 0:04; 0:04) and j 2 [0; 2 ); respectively. For the further
discussion, the parameters Aj ; Bj and j are assumed to be constant in the course of time. We also assume
that frequencies ! j of the perturbing waves are normally distributed with mean at ! 0 and adjustable standard
deviation .
Figure 11(a) shows the behavior of E (t) (blue line) for the case when = 0; i.e. when all perturbing
waves have the same frequency ! 0 : Because the amplitudes and phases of the perturbing waves are random, the
resulting …eld E di¤ers from E0 ; however it is still a fully polarized wave. This result holds because the Aj ; Bj
and j do not vary with time, thus the superposition of waves is coherent.
In Figs. 11(b) to 11(d) we illustrate the evolution of E (t) for the spectral widths =! 0 = f0:05; 0:1; 0:2g. In
these cases, the wave E (t) is polychromatic and thus it is partially polarized. The blue ellipses correspond to
the fully polarized component after making the separation E = Epol + Eunpol : The Stokes parameters of E (t)
were calculated averaging the components Ex2 (t), Ey2 (t) ; and (t) over 20 cycles of the wave of E0 (t) : The
Stokes vectors of E (t) and Epol are included in the following table:

=! 0 S p Spol Jpol
T T 1:262
0 2:445 0:740 1:350 1:900 1.000 2:445 0:740 1:350 1:900
0:924 ei54:61
T T 1:057
0:05 2:749 0:462 2:139 1:578 0.981 2:698 0:462 2:139 1:578
1:257 ei36:41
T T 1:321
0:10 3:459 0:442 1:624 2:539 0.885 3:046 0:442 1:624 2:539
1:141 ei57:4
T T 0:883
0:20 2:120 0:004 1:188 1:005 0.734 1:556 0:004 1:188 1:005
0:881 ei40:24

The partial polarization illustrated in Fig. 11 is a consequence of that the perturbing waves have di¤erent
frequencies ! j . However, it is important to remark that even in the case that all perturbing waves have the
same frequency ! j = ! 0 ; partial polarization arises if the amplitudes Aj ; Bj or phases j vary with time. Fully
polarized waves requires a coherent superposition of constituent waves.
Lecture notes on Polarization 31

5.3 Illustrative applications of the Mueller calculus

5.3.1 Optical activity
A crystal exhibiting optical activity which rotates the plane of polarization an angle to the right has the
interaction matrix 2 3
1 0 0 0
60 cos 2 sin 2 07
6 7: (147)
40 sin 2 cos 2 05
0 0 0 1
The simplicity of these interaction matrices shows the ease of treating polarized light in the formalism of the
Stokes parameters.

6 Measurements of the polarization parameters of a wave

Stokes parameters are intensities and they are real quantities. Stated another way, the light is completely
speci…ed by four real numbers, which represent four actual measurements that could be performed. Consider
T
a wave E with arbitrary state of polarization whose Stokes vector S = S0 S1 S2 S3 is going to be
determined. In general the wave is partially polarized. The classical method to determine the polarization state
of E is directly inspired in the physical interpretation of the Stokes parameters. Actually, all what we need is
a linear polarizer, a quarter-wave plate and an irradiance meter insensitive to polarization.
Before discussing the experimental procedure to measure the Stokes parameters of a light wave, it is conve-
nient to review …rst the e¤ect of the system (quarter-wave-retarder)–polarizer on the wave E.

If the wave E is passed through a polarizer (transmission axis at p ), the output wave is
2 32 3 2 3
1 cos 2 p sin 2 p 0 S0 Ip
166cos 2 p cos2 2 p cos 2 p sin 2 p 07 6S1 7 6Ip cos 2
76 7 = 6
7
p7
; (148)
2 4 sin 2 p cos 2 p sin 2 p sin2 2 p 05 4S2 5 4 Ip sin 2 p5
0 0 0 0 S3 0

where
Ip ( p ) = (S0 + S1 cos 2 p + S2 sin 2 p ) =2; (149)
is the output intensity as a function of p:

If the wave E is passed through a quarter-wave retarder (fast axis at r) and later through a polarizer
(transmission axis at p ), the output wave is
2 32 32 3 2 3
1 cos 2 p sin 2 p 0 1 0 0 0 S0 Irp
166cos 2 p cos2 2 p 1
2 sin 4 p 07 6
7 60 cos2 2 r 1
2 sin 4 r sin 2 r 7 6S1 7 6Irp cos 2
76 7 = 6
7
p7
; (150)
2 4 sin 2 p
1
2 sin 4 p sin2 2 p 0 40
5 1
2 sin 4 r sin2 2 r cos 2 r 5 4S2 5 4 Irp sin 2 p5
0 0 0 0 0 sin 2 r cos 2 r 0 S3 0

where
1
Irp ( r ; p) = [S0 + (S1 cos 2 r + S2 sin 2 r ) cos (2 p 2 r ) + S3 sin (2 p 2 r )] ; (151)
2
is the output intensity as a function of ( r ; p) :

6.1 Measuring the Stokes parameters

From Eqs. (149) and (151) we can derive the following procedure to measure the Stokes parameters of a light
wave. In general, six intensity measurements are needed to …nd the Stokes parameters of the wave.
32 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

1. Pass the wave E through a polarizer with horizontal transmission axis. From Eq. (149) the intensity
measurement at the output Ix gives

Ip (0) = Ix = (S0 + S1 ) =2: (152)

2. Pass the wave E through a polarizer with vertical transmission axis. From Eq. (149) the intensity
measurement at the output Iy gives

Ip (90 ) = Iy = (S0 S1 ) =2: (153)

3. Pass the wave E through a polarizer with transmission axis at +45 with the x axis. From Eq. (149) the
intensity measurement at the output Iu gives

Ip (45 ) = Iu = (S0 + S2 ) =2: (154)

4. Pass the wave E through a polarizer with transmission axis at 45 with the x axis. From Eq. (149) the
intensity measurement at the output Iv gives

Ip ( 45 ) = Iv = (S0 S2 ) =2: (155)

5. Pass the wave E through a quarter-wave plate with horizontal fast axis and later through a linear polarizer
with transmission axis at 45 with the x axis. From Eq. (151) the intensity measurement at the output
I+ gives
Irp (0; 45 ) = I+ = (S0 + S3 ) =2: (156)

6. Pass the wave E through a quarter-wave plate with horizontal fast axis and later through a linear polarizer
with transmission axis at 45 with the x axis. From Eq. (151) the intensity measurement at the output
I gives
Irp (0; 45 ) = I = (S0 S3 ) =2: (157)

Therefore, combining the last expressions we obtain the Stokes parameters of the wave E in terms of the
intensity measurements, namely

S0 = Ix + Iy = Iu + Iv = I+ + I ; (158)
S1 = Ix Iy ; (159)
S2 = Iu Iv ; (160)
S3 = I+ I : (161)

The fact that the total intensity of the beam S0 can be calculated with three equations [Eq. (158)] reduces the
set of six intensities fIx ; Iy ; Iu ; Iv ; I+ ; I g to a set of four independent intensities. The procedure to determine
the Stokes parameters can be therefore reduced to only four measurements. For example, if we measure the
total intensity IT of the beam, then the Stokes parameters can be determined with

S0 = IT ; (162a)
S1 = 2Ix IT ; (162b)
S2 = 2Iu IT ; (162c)
S3 = 2I+ IT ; (162d)

where Ix ; Iu ; and I+ are measured with the procedures explained above. The method based on Eqs. (162)
constitutes the standard procedure of the so-called four-channel Stokes polarimeters [7].
Lecture notes on Polarization 33

6.2 Determination of the Jones vector

T
Once measured the Stokes parameters of a light wave, i.e. S = PS0 ; S1 ; S2 ; S3 ; the corresponding
Jones vector
E0x
E= ; fE0x ; E0y ; g 2 R; (163)
E0y ei
can be calculated with Eq. (89), namely
r r
S0 + S1 S0 S1 S3 S2 + iS3
E0x = ; E0y = ; = arctan ; ei = p 2 : (164)
2 2 S2 S2 + S32

7 General treatment of the elliptical birefringence: Propagation

through a birefringent medium
Any birefringent medium is characterized by two orthogonal polarization eigenstates (in general elliptical) that
are prserved while propagating through the medium (intensity may change, but not the polarization state).
Each eigenstate travels with di¤erent velocity through the birefringent medium. Because the polarization
state of an arbitrary input wave can be regarded as a superposition of polarization eigenstates, the polarization
state will change along the medium as the phase di¤erence between the two polarization eigenstates (travelling at
di¤erent phase velocities) accumulates. The length of material required to bring the two polarization orthogonal
modes back in phase and reproduce the original polarization state is the beat length
2 0
LB = = ; (165)
k0 jn1 n2 j jn1 n2 j
where 0 is the wavelength in free-space and n1 and n2 are the refractive indexes of the two orthogonal eigen-
states.
Special cases of the birefringence are as follows:

1. If the polarization eigenstates are two linearly polarized waves oriented along the two …xed Cartesian axes
of the medium, the birefringence is linear.
2. If the polarization eigenstates are the two orthogonal circularly polarized waves, the birefringence is
circular.
3. If the polarization eigenstates are the two orthogonal elliptical polarized waves, the birefringence is ellip-
tical. This is the general situation.

Evolution on the Poincaré sphere. The evolution of an arbitrary polarization state as it propagates
through a birefringent medium can be traced on the Poincaré sphere. The two polarization eigenstates have
orthogonal polarizations and so are placed at opposite points on the sphere. The axis-diameter connecting the
orthogonal eigenstates is called the principal axis of birefringence. An arbitrary input polarization will rotate
on the Poincaré sphere about the principal axis of birefringence, so as to trace a circle as in Fig. 1. The
polarization of the light will process around this circle as it propagates along the material and describes the
property of optical retardation.
If light propagates from one medium to another whose principal axes of birefringence are misaligned from
the …rst medium by an angle then we rotate the polarization state on the sphere east or west by twice the
angle of misalignment 2 , before propagating light through the second medium as described in (1).

7.1 Jones formalism

Consider a device characterized by the two orthogonal eigenstates

Ax Ay
A1 = ; A2 = ; fAx ; Ay g 2 C; (166)
Ay Ax
34 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

with corresponding eigenvalues 1 and 2 ; where f 1 ; 2 g 2 C:

By matrix algebra it can be shown that if a matrix J has the two orthogonal eigenvectors A1 and A2 with
eigenvalues 1 and 2 , then the matrix is
1 1 Ax Ax + 2 Ay Ay ( 1 2 ) Ax Ay
J= : (167)
Ax Ax + Ay Ay ( 1 2 ) Ax Ay 2 Ax Ax + 1 Ay Ay

Thus in order to determine the Jones matrix of any optical element, one need only to determine the Jones
vectors for the eigenstates of the element and the corresponding eigenvalues.
Examples
T T
1. Linear polarizer with horizontal transmission axis: A1 = 1 0 ; A2 = 0 1 ; 1 = 1; 2 = 0;

1 0
J= : (168)
0 0

T T
2. Linear retarder with A1 = cos sin ; A2 = sin cos ; 1 = ei x ; 2 = ei y ;

ei x cos2 + ei y sin2 ei x ei y sin cos

J= : (169)
ei x ei y sin cos ei y cos2 + ei x sin2

This matrix can be rewritten in a symmetric form using phase di¤erence ; we have
" i #
e 2 cos2 + e i 2 sin2 i2 sin 2 sin cos
J= ; (170)
i2 sin 2 sin cos e i 2 cos2 + ei 2 sin2

which is a standard expression of the Jones matrix of a linear retarder.

7.2 Stokes-Mueller formalism

Let
T
e1;2 = 1;
^ s1 ; s2 ; s3 ; (171)
be Stokes vectors of the fast and slow polarization eigenstates of a medium. If is the angle retardance between
both eigenstates, the Mueller matrix for this material is given by [17, 18]
2 3
1 0 0 0
60 s 2
+ (1 s2
) cos s s (1 cos ) + s3 sin s1 s3 (1 cos ) s2 sin 7
M ( ; s1 ; s2 ; s3 ) = 6 1 1
40 s1 s2 (1 cos ) s3 sin
1 2 7:
s22 + (1 s22 ) cos s2 s3 (1 cos ) + s1 sin 5
0 s1 s3 (1 cos ) + s2 sin s2 s3 (1 cos ) s1 sin s23 + (1 s23 ) cos
(172)
This matrix describes the most general case of elliptical birefringence without losses and depolarization.

7.2.1 Relation with the optical …ber parameters: coupled mode theory
Consider the propagation of polarized light in a birefringent medium. In the weakly guiding approximation, the
electric …eld can be written as

E (x; y; z) = ax (z) Ex (x; y) + ay (z) Ey (x; y) ; (173)

where ax (z) and ay (z) are the amplitudes of the modes The coupling process can be described [17, 18] by a
set of two …rst-order di¤erential equations on the hypothesis that the variations be small versus z. This set is
written as follows 2 3
dax (z)
6 dz 7 11 K ax (z)
4 5= i K ax (z)
; (174)
dax (z) 22
dz
Lecture notes on Polarization 35

where K = KR + iKI and jj are the coupling coe¢ cients. In the lossless case, the evolution of the Stokes
parameters is given by
2 3
dS1 (z)
6 dz 7 2 32 3
6 7 0 2KI 2KR S1 (z)
6 dS2 (z) 7 4 5 4S2 (z)5 ;
6 7= 2KI 0 (175)
6 dz 7
4 5 2KR 0 S3 (z)
dS3 (z)
dz
where = 22 11 :
The polarization eigenstates of the …ber can be related to the coupling coe¢ cients according to
q
2KR 2KI 2 + 4 jKj2 ;
s1 = ; s2 = ; s3 = ; = (176)

where is the phase di¤erence per unit length of the two polarization eigenstates.

7.2.2 Linear and circular birefringence

Linear birefringence. Let us suppose a monomode optical …bre in which there is a linear birefringence
(radians per unit length) expressed in radians per length unit, so that its fast axis makes an angle with respect
to the horizontal x-axis, the parameters are given by
T
e1;2 = 1;
^ cos 2 ; sin 2 ; 0 ; (177)

11 = 22 = cos 2 ; K= sin 2 ; = cos 2 : (178)

2 2
Taking = z; the Mueller matrix takes the form
2 3
1 0 0 0
60 cos2 + sin2 cos 4 sin 4 sin2 2 sin 2 sin 7
M =6 40
2 2 7: (179)
sin 4 sin2 2 cos2 2 sin2 2 cos 4 cos 2 sin 5
0 sin 2 sin cos 2 sin cos

Circular birefringence. Let us suppose a monomode optical …bre in which there is a circular birefringence
(radians per unit length) expressed in radians per length unit, the parameters are given by
T
e1;2 = 1; 0; 0;
^ 1 ; (180)

11 = 22 = 0; K=i ; = 0: (181)
2
Taking = z; the Mueller matrix takes the form
2 3
1 0 0 0
60 cos ( z) sin ( z) 07
M =6 40
7 (182)
sin ( z) cos ( z) 05
0 0 0 0

8 The Pancharatnam-Berry phase

A Appendix: Derivation of the polarization ellipse
The electric vector components along the Cartesian axes are written as

Ex = E0x cos (kz !t + x) ; (183a)

Ey = E0y cos kz !t + y : (183b)
36 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

In order to eliminate the term (kz !t) we rewrite these equations in the form
Ex
= cos (kz !t) cos x sin (kz !t) sin x; (184a)
E0x
Ey
= cos (kz !t) cos y sin (kz !t) sin y: (184b)
E0y

We now multiply the …rst and second equations in (184) by sin y and sin x respectively, and substract
them to get
Ex Ey
sin y sin x = cos (kz !t) sin y x : (185a)
E0x E0y
Similarly, we multiply Eqs. (184) by cos y and cos x respectively, and substract them to get

Ex Ey
cos y cos x = sin (kz !t) sin y x : (185b)
E0x E0y

Squaring and adding Eqs. (185) gives the equation for the polarization ellipse
2 2
Ex Ey Ex E y
+ 2 cos = sin2 ; (186)
E0x E0y E0x E0y

where y x is the phase shift between both Cartesian components of the electric …eld.

B Appendix: Derivation of the parameters fEmax ; Emin ; g of the po-

larization ellipse
Starting from the equation for the polarization ellipse,
2 2
Ex Ey Ex E y
+ 2 cos = sin2 ; (187)
E0x E0y E0x E0y

we will show that the value of the angle between the major axis of the ellipse and the x-axis, and the maximum
and minimum values of the electric …eld are given by
2E0x E0y
tan 2 = 2 2 cos ; (188)
E0x E0y
2
Emax = (cos2 2
sec 2 )E0x (sin2 2
sec 2 )E0y ; (189)
2 2 2 2
Emin = (sin sec 2 )E0x + (cos sec 2 )E0y2 : (190)

First, we change the equation of the ellipse to polar coordinates by making the substitution Ex = E cos ,
Ey = E sin . We obtain " #
2 cos
2
sin2 sin cos
E 2 + 2 2 cos = sin2 : (191)
E0x E0y E0x E0y
We rewrite this expression using double-angle trigonometric identities, obtaining
" ! !#
cos 1 1 1 1
2
E sin 2 + cos 2 2 2 + 2 + 2E 2 = sin2 : (192)
E0x E0y 2E0x 2E0y 2E0x 0y

Note that E 2 is a function of . We will use di¤erentiation with respect to to …nd the extrema of E 2 ,
which is equivalent to …nding the extrema of E. First, we de…ne F ( ) = E 2 ( ) and G( ) such that we can write
Eq. (192) as
F ( )G( ) = sin2 : (193)
Lecture notes on Polarization 37

By using implicit di¤erentiation, we …nd that F 0 ( )G( ) + F ( )G0 ( ) = 0. The extrema of F take place at
angles such that F 0 ( ) = 0. By substitution, it follows that F ( )G0 ( ) = 0. Since F ( ) = E 2 ( ) must be
positive for any , we conclude that G0 ( ) = 0.
Now we obtain an expression for G0 ( ),
! !
cos 1 1 1 1
G( ) = sin 2 + cos 2 2 2 + 2 + 2E 2 ; (194)
E0x E0y 2E0x 2E0y 2E0x 0y
!
0 2 cos 1 1
G ( ) = cos 2 sin 2 2 2 = 0: (195)
E0x E0y E0x E0y

Then we solve the last equation for tan 2 to …nd

2E0x E0y
tan 2 = 2 2 cos ; (196)
E0x E0y

which is Eq. (188).

Note that this equation gives the angle between the x-axis and the axis containing either the maximum
or the minimum value of the electric …eld. It has two relevant solutions:
!
1 2E 0x E 0y
1 = tan 1 2 2 cos ; (197)
2 E0x E0y
!
1 1 2E0x E0y
2 = tan 2 2 cos + : (198)
2 E0x E0y 2

Each of these solutions may represent the angle for the major or minor axis depending on the particular values
of E0x , E0y , and .
Now, suppose that for a particular ellipse we know the angle between the major axis and the x-axis. We
can solve Eq. (188) for cos and also obtain an expression for sin2 using the identity sin2 = 1 cos2 . To
obtain the maximum value of the electric …eld Emax , we take Eq. (191) and substitute = and our expressions
for cos and sin2 . After some algebraic manipulation, the left-hand side becomes
" #
2 2 2 2
2
E 0y cos E 0x sin
Emax 2 E 2 cos 2 ; (199)
E0x 0y

and the right-hand side,

2
(E0y cos2 2
E0x sin2 )(E0x
2
cos2 2
E0y sin2 )
2 2 2
: (200)
E0x E0y cos 2
2
We solve for Emax to …nd
2
Emax = (cos2 2
sec 2 )E0x (sin2 2
sec 2 )E0y : (201)
2 2
To obtain Emin we just need to replace for + =2 in the expression for Emax and simplify. The result is
2
Emin = (sin2 2
sec 2 )E0x + (cos2 2
sec 2 )E0y : (202)

C Appendix: Rotations in a two-component Jones vector formalism

Let EA ; EB be the two-component polarization states of the input and output …eld of a polarization device
characterized by the Jones J matrix.
In the three-dimensional Poincaré space the corresponding states are represented by A; B.
38 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

The states EA ; EB can be represented in a Cartesian basis (i.e. a Jones vector)

Ax A0x exp (i x)
EA = = ; y x; (203a)
Ay A0y exp (i y)
T
A = [A0 ; A1 ; A2 ; A3 ]T = A20x + A20y ; A20x A20y ; 2A0x A0y cos ; 2A0x A0y sin ; (203b)

or, equivalently, in a circular polarization basis (i.e. a spinor)

E+ ^
c+ 1 (Ax iAy ) ^ c+
EA = =p : (204)
E ^c 2 (Ax + iAy ) ^
c
Cartesian representation are more common in Optics while spinor representation in Quantum Mechanics.
We will develop here the relations in Cartesian formalism.
The Pauli matrices are given by 4
1 0 1 0 0 1 0 i
0 ; x ; y ; z : (205)
0 1 0 1 1 0 i 0

Consider the transformation of the state EA into the state EB (represented as Jones vectors)

EB = x EA : (206)

In the Poincaré space, this operation represents a rotation of radians of the normalized state A around the x
axis. In the same way for y and z :
Consider now that we wish to rotate the state A around the rotation axis Q ^ = [Q1 ; Q2 ; Q3 ] by an angle :
In the Jones space the trasformation is denoted by

EB = R ( ) EA ; (207)

where the rotation operator R ( ) is given by

R ( ) = cos 0 i sin ^ ;
Q (208)
2 2
with = [ x ; y ; z ] being the vector of Pauli matrices. Explicitly, we have ^ = Q1
Q x + Q2 y + Q3 z:
The rotation matrix can be written in an exponential form as

R ( ) = cos 0 i sin ^ = exp

Q i ^ :
Q (209)
2 2 2
From the relation EB = R ( ) EA , it is clear that the rotation operator R ( ) can be multiplied by an
arbitrary phase factor and the Poincaré state B preserves unchanged.

D Appendix: Polarization of harmonic waves in three-dimensions

At a given point r = (x1 ; x2 ; x3 ) ; the oscillation of a real time-harmonic wave (with angular frequency !) in
three dimensions is given by

E (r; t) = E1 (r; t) u
^ 1 + E2 (r; t) u
^ 2 + E3 (r; t) u
^3 ; (210)

where

E1 (r; t) = a1 (r) cos [g1 (r) !t] ; g1 (r) = k r + 1; (211a)

E2 (r; t) = a2 (r) cos [g2 (r) !t] ; g2 (r) = k r + 2; faj ; gj g 2 R; (211b)
E3 (r; t) = a3 (r) cos [g3 (r) !t] ; g3 (r) = k r + 3; (211c)
4 Note
that our de…nition difers with the standard ordering of the Pauli matrices in quantum mechanics which is ( 0 ; 2 ; 3 ; 1 ) :
It is somewhat more practical to associate 3 with the z axis rather than with the x axis, although it is clear that neither convention
is reallv more fundamental than the other.
Lecture notes on Polarization 39

and f^ u1 ; u
^2 ; u
^ 3 g are unit vectors of a completely arbitrary Cartesian orthogonal reference frame at r =
(x1 ; x2 ; x3 ) :
By expanding the cosine functions the wave E (r; t) can be rewritten in the form

E (r; t) = p (r) cos !t + q (r) sin !t; (212)

where

p (r) = a1 (r) cos g1 (r) u

^ 1 + a2 (r) cos g2 (r) u
^ 2 + a3 (r) cos g3 (r) u^ 3 2 R; (213)
q (r) = a1 (r) sin g1 (r) u
^ 1 + a2 (r) sin g2 (r) u
^ 2 + a3 (r) sin g3 (r) u
^ 3 2 R; (214)

are real and time-independent vectors.

In general, as time varies, the end point of E (r; t) describes a curve. To see this we note from Eq. (212)
that, with varying time, the end point describes a curve in the plane speci…ed by p (r) and q (r) : Since E is
periodic, the curve must be closed. Actually, Eq. (212) is the parametric equation of an ellipse.
For convenience we now introduce the complex notation. We write E (r; t) in the form
i!t
E (r; t) = Re U (r) e ; (215)

where U (r) is the complex vector

U (r) = p (r) + i q (r) : (216)
Now we set
U = p + iq = (a + i b) exp (i ) ; 2 R; (217)
where is any scalar. Solving for a and b

a = p cos + q sin ; (218)

b = p sin + q cos : (219)

Let us choose " so that the vectors a and b are perpendicular to each other and let jaj jbj ; thus " must satisfy
the equation
a b = (p cos + q sin ) ( p sin + q cos ) = 0; (220)
i.e.
2p q
tan 2 = : (221)
p2 q2

References
[1] M. Born and E. Wolf, Principles of Optics (Cambridge University Press, 7th ed., 1999).
[2] L. Manel and E. Wolf, Optical coherence and quantum optics (Cambridge University Press, 1995).
[3] E. Wolf, “Coherence properties of partially polarized electromagnetic radiation”, Nuovo Cimento 13, 1165–
1181 (1959).
[4] E. Hetch, Optics (Addisson Wesley, 4th ed., 2002).
[5] R. C. Jones, “New calculus for the treatment of optical systems,” J. Opt. Soc. Am. 31, 488–493 (1941).
[6] E. Collett, “The description of polarization in classical physics,” Am. J. Phys. 36, 713–725 (1968).
[7] E. Collet, Polarized light in …ber optics (Polawave, 2003).
[8] A. Gerrard and J.M. Burch, Introduction to Matrix Methods in Optics (John Wiley & Sons, 1975).
[9] G. Stokes, “On the composition and resolution of streams of polarized light from di¤erent sources,”Trans.
Cambridge Phil. Soc. 9, 399–416 (1852).
40 Julio C. Gutiérrez-Vega, Lecture notes on Polarization

[10] H. Poincaré, Théorie Mathématique de la Lumière, Vol. 2, (1892) Chap. 12.

[11] J. D. Jackson, Classical Electrodynamics (Wiley, 3rd ed. 1998).
[12] W. H. McMaster, “Matrix Representation of Polarization,” Rev. Mod. Phys, 33, 8–28 (1961)
[13] R. W. Schmieder, “Stokes-algebra formalism,” J. Opt. Soc. Am. 59, 297–302 (1969).

[14] D. S. Kliger and J. W. Lewis, Polarized Light in Optics and Spectroscopy (Academic Press, 1990)
[15] N. G. Parke, “Optical algebra,” J. Math. Phys. 28, 131–139 (1949).
[16] Oriol Arteaga and Adolf Canillas, "Analytic inversion of the Mueller-Jones polarization matrices for ho-
mogeneous media," Opt. Lett. 35, 559-561 (2010)

[17] P Olivard , P Y Gerligand , B Le Jeune , J Cariou and J Lotrian , “Measurement of optical …bre parameters
using an optical polarimeter and Stokes-Mueller formalism,”J. Phys. D: Appl. Phys. 32, 1618–1625 (1999)
[18] Giorgio Franceschetti and Cynthia Porter Smith, "Representation of the polarization of single-mode …bers
using Stokes parameters," J. Opt. Soc. Am. 71, 1487–1491 (1981)
[19] Donald G. M. Anderson and Richard Barakat, "Necessary and su¢ cient conditions for a Mueller matrix
to be derivable from a Jones matrix," J. Opt. Soc. Am. A 11, 2305–2319 (1994)
[20] José J. Gil, "Characteristic properties of Mueller matrices," J. Opt. Soc. Am. A 17, 328–334 (2000)
Lecture notes on Polarization 41

Additional matrices
In this appendix we include the general expression for an elliptical retarder with retardance

Phase retarder Jones matrix Mueller matrix

2 3
Elliptical retarder with ei x cos2 + ei y sin2 0 0 0 0
retardances x and y ; ei x ei y e i" sin cos 60 0 0 07
6 7 Hetch 8:59
rotated with horizontal. ei x ei y ei" sin cos 40 0 0 05
" = circularity parameter ei x sin2 + ei y cos2 0 0 0 0