Pure Appl. Opt. 7 (1998) 941–951.

Printed in the UK PII: S0963-9659(98)91420-9

Beam coherence–polarization matrix

F Gori†, M Santarsiero†, S Vicalvi‡, R Borghi§ and G Guattari§

† Dipartimento di Fisica, Università degli Studi Roma Tre, and Istituto Nazionale per la Fisica
della Materia, Via della Vasca Navale 84, I-00146 Rome, Italy
‡ Dipartimento di Fisica, Università La Sapienza, P le Aldo Moro 2, I-00185 Rome, Italy
§ Dipartimento di Ingegneria Elettronica, Università degli Studi Roma Tre, and Istituto Nazionale
per la Fisica della Materia, Via della Vasca Navale 84, I-00146 Rome, Italy

Received 5 February 1998, in final form 25 May 1998

Abstract. We present an approach for describing the properties of a quasi-monochromatic,

beam-like field that is both partially polarized and partially coherent from the spatial standpoint.
It is based on the use of a single 2 × 2 matrix, called the beam coherence–polarization matrix,
whose elements have the form of mutual intensities. This approach, which can be viewed as an
approximate form of Wolf’s general tensorial theory of coherence, appears to be very simple, yet
it is able to cover significant aspects of the beam behaviour that would not be accounted for by
a scalar theory or by a local polarization matrix approach. A peculiar interference law applying
to mutual intensities is derived. We show through simple examples how this approach leads
to distinguish fields that would appear identical in a scalar treatment or in a local polarization
matrix description. Hints for extensions are given.

1. Introduction

The study of the polarization properties of light is of the utmost importance in optics, both
conceptually and for applications [1–3]. It is a fascinating subject that often leads to results
of great elegance and significance. As a matter of fact, some of the most intriguing topics
in classical and quantum optics, such as Pancharatnam’s phase [4–6] and the optical version
of the EPR paradox [1, 7], to name only two, have to do with polarization. Therefore, it is
no surprise that this subject, with its extraordinary richness, keeps on attracting the attention
of scientists [8–10]. We are interested here in partially polarized light, which includes of
course completely polarized and completely unpolarized light as particular cases. In the
simplest situation we deal with a quasi-monochromatic plane wave. Polarization properties
can then be described by a 2 × 2 matrix known as the coherence matrix or the polarization
matrix [1, 11]. Here, we shall adopt the latter name. Essentially, the polarization matrix
is a set of four numbers (two of them being real and the other two complex conjugate)
that give the self- and mixed correlations, at equal times, of the Cartesian components of
the electric field of the wave. Equivalently, we can use Stokes parameters [1]. This type
of description can also be used whenever the light field, without being a plane wave, is in
the form of a well collimated, uniform beam [1]. In this case one assumes that the field is
approximately the same at any point in a beam section. This implies that the polarization
is constant and that spatial coherence is complete across a transverse plane. Indeed, it
is amazing how many different phenomena can be handled by this simple model. At
present, however, beams that depart from the previous hypotheses are becoming of interest
[12–14]. One then wonders which approach is to be used. An answer was furnished

0963-9659/98/050941+11$19.50
c 1998 IOP Publishing Ltd 941
942 F Gori et al

long ago by Wolf [15] through his general tensorial theory of the electromagnetic field.
Generally speaking, this is a rather demanding approach since it involves the use of three
different 3 × 3 matrices† whose elements are functions of six (scalar) space variables.
However, a drastic simplification can be made if the beam structure of the field is taken
into account. In this case, the electromagnetic field is essentially transverse with respect
to the propagation axis and this allows us to neglect the longitudinal vector components.
In addition, we may base our description on one matrix only, much in the same spirit as
in the ordinary use of the polarization matrix. We then pass from three 3 × 3 matrices
to a single 2 × 2 matrix. For reasons to be discussed later, we shall refer to it as the
beam coherence–polarization matrix (or BCP matrix, for the sake of brevity). A brief
sketch of it has been presented in [16]. This type of approach has already been used by
James, who showed how the polarization of a Gaussian Schell-model beam can change
upon propagation [17]. Furthermore, partially polarized, partially coherent beams have
been studied recently by Martı́nez-Herrero et al by using an approach based on the Wigner
distribution function [18].
Here we discuss the basis of an approach based on the BCP matrix. The elements
of such a matrix are similar to mutual intensities [1]. Thanks to symmetry properties
only three of the matrix elements are required, each of them obeying well known wave
equations. After a brief discussion of the main properties of the BCP matrix we shall
examine its use in the analysis of experiments in which anisotropic optical elements (either
spatially uniform or not) are inserted in the beam path. This will lead us to establish an
interference law of a new type in that it applies to mutual intensity functions rather than
to optical intensities only. As an example of the use of the present approach we shall
discuss three beams that would be indistinguishable in the scalar treatment, while exhibiting
measurable differences in the BCP matrix description. Throughout this paper we will draw
from the theory of the ordinary polarization matrix as well as from the general tensorial
approach by Wolf. The reader can consult chapter 6 of [1] for a complete exposition of
the present state of the vectorial theory of coherence. Explicit references to such work
will be made at specific points. A few hints for extensions will be given in the final
section.

2. The beam coherence–polarization matrix

Many partially coherent light fields discussed in the scalar theory of coherence propagate
in the form of beams, i.e. they remain essentially concentrated around a mean axis without
a large angular spread. This applies to light fields generated by lasers oscillating on several
transverse modes [19–21] as well as to partially coherent fields obtained through synthesis
procedures [22, 23]. Conditions to be satisfied for beam-like propagation are discussed in
[1]. Essentially, both the coherence length across the source and the linear dimensions of
the source itself must be large with respect to the wavelength‡. Typical angular widths of
a beam are of the order of a few milliradians or less. With similar angular apertures the
components of the electric and magnetic vectors along the beam axis have amplitudes two
or three orders of magnitude smaller than the amplitudes of the vectors themselves. We can
then neglect, to a good approximation, the longitudinal vector components. We shall refer
to this as the quasi-transversality approximation.
† Actually four matrices appear in the general theory, but two of them are equivalent.
‡ It may be useful to recall that filtered sunlight arriving at the Earth’s surface has a coherence (transverse) length
of the order of 100 wavelengths [1].
 Beam coherence–polarization matrix 943

There is another important point to be made. In the usual matrix theory of well
collimated, uniform, quasi-monochromatic beams only one vectorial quantity† is used to
describe the polarization properties of the field. Under the assumption of stationarity for
the light field, the essential property of such a vector is that the time average of its square
must be representative of the response signal of the detector used in the experiments. Such
a single-vector description of the wave field has proved to be fully satisfactory. Physically,
this can be explained by noting that, since the field is constituted by a set of plane waves
with slightly different directions, the Poynting vector can be assumed to be proportional to
the square of the vectorial quantity used to describe the field, as for a single plane wave.
The same holds true for the cases of our interest. Accordingly, we shall use an identical
approach.
Let us then introduce the covariance matrix of the electric field vector [1]. Since we
want to consider a typical cross section of the beam, we adopt the following system of
coordinates. The z-axis is along the effective direction of propagation of the beam. Then,
at any plane z = constant, we denote by r1 and r2 the position vectors, lying in that plane, of
two typical points. In the quasi-transversality approximation only the x- and y-components
of the electric vector are considered and the corresponding covariance matrix across the
plane z = constant, at time delay τ , is
!
(E) (E)
0xx (r 1 , r 2 , z; τ ) 0xy (r 1 , r 2 , z; τ )
0̂ (E) (r1 , r2 , z; τ ) = (E) (E)
(1)
0yx (r1 , r2 , z; τ ) 0yy (r1 , r2 , z; τ )

where


(E)
0αβ (r1 , r2 , z; τ ) = Eα∗ (r1 , z; t) Eβ (r2 , z; t + τ ) . (2)

Here and in the following, α and β denote the subscripts x or y. The time variable is
denoted by t. Ex and Ey are the components of the complex electric field vector [1] in the
x- and y-directions, respectively, and the asterisk denotes a complex conjugate. The field
is assumed to be stationary and ergodic. Angular brackets indicate a time average. Here
and below, we denote a matrix by a symbol surmounted by a caret.
We assume the field to be quasi-monochromatic with mean frequency ν̄ and effective
bandwidth 1ν. Furthermore, we suppose that the time delays τ introduced by the
experimental apparatus are small compared to the coherence time 1/1ν. In this case it
can be easily shown that the following approximation holds:
(E)
0αβ (r1 , r2 , z; τ ) = Jαβ (r1 , r2 , z) exp (−2πiν̄τ ) (3)

where


Jαβ (r1 , r2 , z) = Eα∗ (r1 , z; t) Eβ (r2 , z; t) (4)

are the elements of the covariance matrix at equal times. Consequently, the polarization
and spatial coherence properties are specified by the matrix
!
Jxx (r1 , r2 , z) Jxy (r1 , r2 , z)
Jˆ (r1 , r2 , z) = . (5)
Jyx (r1 , r2 , z) Jyy (r1 , r2 , z)

† Typically the electric vector is used but, depending on the detection process, other vectorial quantities, e.g. the
vector potential [1], could be more suitable.
944 F Gori et al

Equation (5) can be written in a more  compact form by introducing the row vector
E (r, z; t) = Ex (r, z; t) Ey (r, z; t) , thus giving [1]
Jˆ (r1 , r2 , z) = hE † (r1 , z; t) E(r2 , z; t)i (6)
where the dagger denotes the Hermitian conjugate.
Matrix (5) shares with the ordinary polarization matrix [1] the fact that it refers to light
beams, but differs from it since its elements are functions of two position vectors instead
of being constant parameters of the beam. This is because the present matrix formulation
does not rely on the hypothesis of complete spatial coherence across the beam section.
On the other hand, the general Wolf coherence matrices for the electromagnetic field also
have elements depending on two position vectors, but differ from (5) since they have nine
elements instead of four. This is because the general matrix formulation does not require
the hypothesis that the field is beam-like. Therefore, we can distinguish matrix (5) from
other coherence matrices by adopting the name beam coherence–polarization matrix (BCP
matrix).
Of course the BCP matrix includes the ordinary polarization matrix as the particular
case in which Jαβ is independent of the space coordinates for all points within the beam.
In turn, the BCP matrix can be thought of as a particular case of the general coherence
matrix of the electric vector when the z-component of the field can be neglected and the
quasi-monochromaticity hypothesis applies.
As shown by equation (4), the elements of Jˆ have the same structure as the mutual
intensity used in the scalar theory of coherence [1, 11]. More precisely, Jxx (Jyy ) is the
mutual intensity that would characterize the beam if the y-component (x-component) of
the field were eliminated (for example, by a linear polarizer). On the other hand, Jxy and
Jyx account for the correlations that exist at two distinct points between the x- and the
y-components. Only one of these functions is needed since it follows from equation (4)
that
∗
Jyx (r1 , r2 , z) = Jxy (r2 , r1 , z). (7)
The optical intensity at a point (r, z) is made up of two contributions corresponding to
the orthogonal components of the electric vector. Disregarding a proportionality factor, as
in scalar theory, we shall compute the optical intensity I (r, z) as

I (r, z) = Jxx (r, r, z) + Jyy (r, r, z) = Tr Jˆ(r, r, z) (8)
where Tr denotes the trace.
Using the Schwarz inequality it can be easily proved that
 
Jαβ (r1 , r2 , z)2 6 Jαα (r1 , r1 , z) Jββ (r2 , r2 , z). (9)
This allows us to introduce a normalized version of the elements of the BCP matrix through
the definitions
Jαβ (r1 , r2 , z)
jαβ (r1 , r2 , z) = p . (10)
Jαα (r1 , r1 , z) Jββ (r2 , r2 , z)
 
Equation (9) implies that jαβ  6 1. The diagonal elements jxx and jyy are the complex
degrees of coherence for the x- and y-components of the field, respectively. As for jxy ,
it represents the degree of cross-coherence between the x-component at (r1 , z) and the y-
component at (r2 , z) (and a similar meaning can be given to jyx ). We shall see later on
that jxy and jyx are measurable through suitable interference experiments.
It is worthwhile to remark that Jxy (r1 , r2 , z) is not locally connected to Jxx (r1 , r2 , z)
and Jyy (r1 , r2 , z) except when r1 = r2 (see equation (9)). This means, for example, that
 Beam coherence–polarization matrix 945

Jxy (r1 , r2 , z) can be different from zero even if (at the same points) Jxx and Jyy vanish.
There is, however, an important global connection that can be derived from the general
semi-definiteness conditions obeyed by the coherence matrices [1]. Denoting by f1 (r) and
f2 (r) two arbitrary well behaved functions, the following inequality holds:
ZZ
 ∗
f1 (r1 ) f1 (r2 ) Jxx (r1 , r2 , z) + f2∗ (r1 ) f2 (r2 ) Jyy (r1 , r2 , z)

+f1∗ (r1 ) f2 (r2 ) Jxy (r1 , r2 , z) + f2∗ (r1 ) f1 (r2 ) Jyx (r1 , r2 , z) dr1 dr2 > 0
(11)
where the integrals are extended to the beam section. Using equation (7) it is seen at once
that equation (11) can be written as
ZZ
 ∗
f1 (r1 )f1 (r2 ) Jxx (r1 , r2 , z) + f2∗ (r1 ) f2 (r2 ) Jyy (r1 , r2 , z)
 
+2 Re f1∗ (r1 )f2 (r2 ) Jxy (r1 , r2 , z) dr1 dr2 > 0 (12)
where Re{ } stands for the real part. Letting f2 = 0 (f1 = 0) we see that Jxx (Jyy ) is
a semi-definite positive kernel. Furthermore, equation (12) shows that the functions Jαβ
cannot be chosen independently from one another.
The local polarization properties of the beam at a typical point of a cross section
are specified by the BCP matrix with equal arguments. We shall refer to it as the local
polarization matrix. Further, the degree of polarization is given by [1, 11]
v 
u
u
t 4 det Jˆ(r, r, z)
P (r, z) = 1 − 
2 (13)
Tr Jˆ(r, r, z)
where det stands for determinant. It should be noted that P will, in general, be different
from one point to another.

3. Interference and propagation of partially coherent, partially polarized beams

Let us discuss how the BCP matrix changes when some anisotropic optical element is
inserted in the beam path. We suppose that such an element is a flat transparency, i.e. that
its thickness can be neglected. Then it can be represented by a Jones matrix [1, 24] of the
form
 
a (r) b (r)
T̂ (r) = (14)
c (r) d (r)
which differs from the usual form only because the matrix elements can be functions of
the position across the element plane. Simple examples are offered by a linear polarizer
covering only a portion of the beam section or by a lens made of a birefringent material.
When the anisotropic element is inserted in the beam path at a certain plane z = constant,
the BCP matrix of the field emerging from it, say Jˆ0 , can be evaluated through the relation
Jˆ0 (r1 , r2 , z) = T̂ † (r1 ) Jˆ(r1 , r2 , z) T̂ (r2 ). (15)
Equation (15) can be derived by the same procedure used for the ordinary polarization
matrix and indeed is quite similar to a relation holding for the latter [1], except that the
matrices multiplying Jˆ on the right and on the left have different arguments.
A relation of the form (15) also holds for determining the form taken by Jˆ when the
axes x and y are rotated, provided that T̂ represents the rotation matrix.
946 F Gori et al

Let us use equations (8) and (15) for evaluating the intensity emerging from a uniform
linear polarizer, such as a sheet of polarizing film. If the transmission axis of the polarizer
forms an angle α (anti-clockwise) with the x-axis, the matrix has the form
 2 
C CS
T̂ = (16)
CS S 2
where C = cos α and S = sin α and the dependence on r has been dropped because the
polarizer is uniform. On inserting from equation (16) into (15) and evaluating the trace of
Jˆ0 we find that the intensity, say Iα (r, z), after the polarizer is

Iα (r, z) = Jxx (r, r, z) C 2 + Jyy (r, r, z) S 2 + 2 Re Jxy (r, r, z) CS. (17)
This relation is also similar to the one that is found using the ordinary polarization matrix
[1]. Again, the difference is that here all the involved quantities depend on the position.
Let us now give a brief discussion on the role played by the BCP matrix in the analysis
of the classical Young interference experiment. Since in our procedure we have simply to
rephrase a type of analysis carried out in a full way in several references (see e.g. [1, 11, 25])
we can omit most details of the calculation. Let the Young mask be located at z = zm with
pinholes at r1 and r2 . Further, let z = zp be the observation plane. The observed optical
intensity is the sum of two contributions due to the (orthogonal) x- and y-components of
the field.
We then easily find
 
I (r, zp ) = |K1 |2 Jxx (r1 , r1 , zm ) + Jyy (r1 , r1 , zm )
+ |K2 |2 [Jxx (r2 , r2 , zm ) + Jyy (r2 , r2 , zm )]
  
+2 Re K1∗ K2 Jxx (r1 , r2 , zm ) + Jyy (r1 , r2 , zm ) exp(−i2π ν̄τ ) . (18)
The quantities K1 and K2 are purely imaginary numbers that depend on the size of the
pinholes and on the geometry [1, 11]. The time delay τ is given by
q q 
1
2
2
τ= (r − r1 ) + zp − zm − (r − r2 ) + zp − zm
2 2
(19)
c
where c is the speed of light. Equation (18) can be written in the more compact form

I (r, zp ) = |K1 |2 I (r1 , zm ) + |K2 |2 I (r2 , zm ) + 2 Re K1∗ K2 Jeq (r1 , r2 , zm ) exp(−i2π ν̄τ )
(20)
where we used equation (8) and we have defined an equivalent mutual intensity at the
pinholes as
Jeq (r1 , r2 , zm ) = Jxx (r1 , r2 , zm ) + Jyy (r1 , r2 , zm ). (21)
It is seen that equation (20) is exactly of the same form that one obtains in the scalar theory
provided that we take as the mutual intensity the sum of the corresponding functions for
the x- and y-components of the field vector. This is physically very sound. After all, if
no anisotropic element is used, we should see the superposition of the interference patterns
produced by the x- and y-components of the field at the pinholes. It may be worthwhile
to note that rule (21) for the equivalent mutual intensity is the same as what we would
obtain in scalar theory for the superposition of two uncorrelated fields. Here, however, the
origin is different because Jxy (r1 , r2 , zm ) will, in general, be different from zero. Rule (21)
simply stems from the geometrical orthogonality of the two components of the field vector.
Just as in the case of the intensity distribution across the beam (see equation (17)), the role
of anti-diagonal elements of the BCP matrix appears when some optical anisotropy occurs.
 Beam coherence–polarization matrix 947

As a simple example, suppose the Young mask is covered with a uniform linear polarizer
set at an angle ϕ with respect to the x-axis. On evaluating the intensity distribution at the
observation plane we find
Iϕ (r, zp ) = |K1 |2 Iϕ (r1 , zm ) + |K2 |2 Iϕ (r2 , zm )

+2 Re K1∗ K2 Jϕ,eq (r1 , r2 , zm ) exp(−i2π ν̄τ ) (22)
where Iϕ is computed through equation (17) and the equivalent mutual intensity is given by

Jϕ,eq (r1 , r2 , zm ) = Jxx (r1 , r2 , zm ) C 2 + Jyy (r1 , r2 , zm ) S 2 + 2 Re Jxy (r1 , r2 , zm ) CS.
(23)
This result deserves some comment. We have found that the Young interference pattern
seen after the polarizer can be cast in the usual form if we introduce a suitable equivalent
mutual intensity. The latter is furnished by equation (23), which represents a new type of
interference law. Indeed what is going to interfere in equation (23) for r1 6= r2 is not a set
of optical intensities but of mutual intensities. In other words, the correlation between the
fields emerging at r1 and r2 from the polarizer is determined by an interplay among the
self-correlations of the x- and the y-components and the mixed correlation between them.
For example, it may well happen that while Jxx , Jyy and Jxy are all different from zero,
the equivalent mutual intensity vanishes. Paradoxically, the overall effect of interference,
in such cases, is to destroy correlation between the fields at the pinholes. Needless to say,
equation (23) includes equation (17) as a particular case for r1 = r2 .
We finally note that, setting ϕ = 0 (π/2), Jϕ,eq becomes equal to Jxx (Jyy ). The
diagonal elements of the BCP matrix can then be obtained from intensity measurements
(see equation (22)).
The similarity between any element of the BCP matrix and the ordinary mutual intensity
allows us to exhaust the propagation problem with a few words. Starting from the knowledge
of the BCP matrix at a certain plane z = z1 , the evaluation at any successive plane
z = z2 (z2 > z1 ) can be made by using, for each matrix element, the formulae holding for
the propagation of the mutual intensity. We simply note that, while in certain cases it will
be possible to use exact formulae, in most cases the very hypothesis of beam propagation
will suggest the use of paraxial propagation formulae. Of course the propagation problem
could not be treated using the local polarization matrix only.

4. Generalized Young interference experiments

It is of some interest to extend the previous formalism to treat Young interference when the
two pinholes are covered by different anisotropic elements. As we shall see, this gives the
key for measuring the off-diagonal element of the BCP matrix. Let T̂j be the matrix of the
anisotropic element placed onto the j th pinhole (j = 1, 2). Further, let E 0 (rj , zm ; t) be the
electric field at point rj on the transverse plane immediately after the pinhole j (j = 1, 2).
Following Mandel and Wolf [1], we can write
E 0 (r1 , zm ; t) = E(r1 , zm ; t) T̂1
(24)
E 0 (r2 , zm ; t) = E(r2 , zm ; t) T̂2
where E(rj , zm ; t) is the electric field impinging on the Young interferometer at rj .
Following analogous lines as for the previous section, the BCP matrix at the point (r, zp )
is given by (see equation (6))
Jˆ(r, r, zp ) = |K1 |2 hE 0† (r1 , zm ; t) E 0 (r1 , zm ; t)i + |K2 |2 hE 0† (r2 , zm ; t) E 0 (r2 , zm ; t)i

+2 Re K1∗ K2 hE 0† (r1 , zm ; t) E 0 (r2 , zm ; t)i exp(−i2π ν̄τ ) . (25)
948 F Gori et al

Taking equations (24) and (6) into account, we obtain

† †
Jˆ(r, r, zp ) = |K1 |2 T̂1 Jˆ(r1 , r1 , zm ) T̂1 + |K2 |2 T̂2 Jˆ(r2 , r2 , zm ) T̂2
 ∗ †
+2 Re K1 K2 T̂1 Jˆ(r1 , r2 , zm ) T̂2 exp(−i2π ν̄τ ) (26)
which represents a generalized interference law. Obviously, equation (26) leads to
equation (22) in the particular case when both T̂1 and T̂2 coincide with the matrix of a
linear polarizer set at one and the same angle ϕ.
A suitable choice of T̂1 and T̂2 can lead to selecting the off-diagonal elements of the
BCP matrix. As a simple example let us consider the case in which T̂1 and T̂2 are given by
   
1 0 0 0
T̂1 = T̂2 = . (27)
0 0 1 0
Such matrices describe an x-oriented linear polarizer (T̂1 ) and a y-oriented linear polarizer
followed by a −π/2 rotator (T̂2 ). On inserting equation (27) into (26) and taking the trace
of both sides, we obtain
I (r, zp ) = |K1 |2 Jxx (r1 , r1 , zm ) + |K2 |2 Jyy (r2 , r2 , zm )

+2 Re K1∗ K2 Jxy (r1 , r2 , zm ) exp(−i2π ν̄τ ) . (28)
This shows how a modified Young interferometer may be used to measure Jxy .

5. Examples

In this section we will show through simple examples how physically different beams can
appear identical in a scalar description while their differences are revealed by the BCP
matrix. To this end, let us consider three field distributions specified, at the plane z = 0,
by the following sets of Cartesian components:
  2

 r r
 Ex (r, 0; t) = a(t) √ exp − 2 exp(iϑ)

(1)
v 2 v
 2 (29)

 r r

 Ey (r, 0; t) = b(t) √ exp − 2 exp(−iϑ)
(1)
v 2 v
  2

   r r
 Ex (r, 0; t) = a(t) exp(iϑ) + b(t) exp(−iϑ)

(2)
exp − 2
2v v
 2 (30)

   r r

 Ey (r, 0; t) = i a(t) exp(iϑ) − b(t) exp(−iϑ)
(2)
exp − 2
2v v
  2

 r r

 Ex (r, 0; t) = a(t) v exp − v 2 cos ϑ
(3)

 2 (31)

 r r

 Ey (r, 0; t) = b(t) exp − 2 sin ϑ
(3)
v v
where r and ϑ are the polar coordinates associated with r and v is a positive constant. The
functions a(t) and b(t) are uncorrelated quasi-monochromatic signals such that
h|a|2 i = h|b|2 i = I0 ha ∗ bi = 0 (32)
with positive I0 . We shall refer to the field corresponding to equation (29) (equations (30)
and (31)) as field 1 (2, 3). Each of the functions appearing in equation (29) will be
recognized as a particular Laguerre–Gauss mode with spot size v, taken at its waist [26].
 Beam coherence–polarization matrix 949

Both modes exhibit a vortex, but the first has charge 1, whereas the second has charge −1
[27]. Field 1 is thus the incoherent superposition of these two modes polarized along the x-
and the y-axis, respectively. Similarly, field 2 is obtained through (incoherent) superposition
of the same two modes except that they are now circularly polarized with opposite helicity.
Finally, field 3 is the incoherent superposition of linearly polarized Laguerre–Gauss modes
with angular amplitude modulation.
Using equations (29)–(31) and taking equation (32) into account, the elements of the
PCB matrix are easily evaluated for fields 1–3 and turn out to be
  2 

 r1 r2 r1 + r22

 J (1)
(r , r , 0) = I exp − + i (ϑ − ϑ )
 xx 1 2 0
2v 2 v2
2 1
  ∗ (33)

 J (1) (r1 , r2 , 0) = Jxx (1)
(r1 , r2 , 0)
 yy

 J (1) (r , r , 0) = 0
xy 1 2
  2 

 r1 r2 r1 + r22
 Jxx (r1 , r2 , 0) = I0

(2)
exp − cos (ϑ2 − ϑ1 )

 2v 2 v2

(2)
Jyy (r1 , r2 , 0) = Jxx (2)
(r1 , r2 , 0) (34)

  2

 2

 r1 r2 r1 + r2
 Jxy (r1 , r2 , 0) = −I0 2 exp − sin (ϑ2 − ϑ1 )
(2)
2v v2
  2 

 r1 r2 r1 + r22

 J (3)
(r , r , 0) = I exp − cos ϑ2 cos ϑ1


xx 1 2 0
v2 v2
  2 
r1 r2 r1 + r22 (35)
 Jyy (r1 , r2 , 0) = I0 2 exp −

(3)
sin ϑ2 sin ϑ1

 v v 2


 J (3) (r , r , 0) = 0.
xy 1 2

On inserting equations (33)–(35) into equations (21) and (8) we can evaluate the
equivalent mutual intensities and the optical intensity distributions associated with fields 1–
3. It is easily seen that the results are the same for all of them. More explicitly, we
obtain
 2 
r1 r2 r + r2
(j )
Jeq (r1 , r2 , 0) = I0 2 exp − 1 2 2 cos (ϑ2 − ϑ1 ) (j = 1, 2, 3) (36)
v v
 
r2 2r 2
I (r, 0) = I0 2 exp − 2
(j )
(j = 1, 2, 3). (37)
v v
Using the well known propagation laws of the Laguerre–Gauss modes, the mutual intensities
and the optical intensities associated with fields 1–3 could be easily computed at any plane
z = constant. Without doing this, we simply remark that the field distributions specified
by equations (29)–(31) give rise to shape-invariant beams in the sense of coherence theory
[28]. This essentially means that, at any plane z = constant, the mutual intensity remains
of the same functional form as across the plane z = 0, except for a transverse scale factor
and a quadratic phase factor depending on r22 − r12 . As a consequence, the beams produced
by fields 1–3 behave as identical in the realm of scalar coherence theory. In other words,
they cannot be distinguished from one another if only intensity and/or scalar coherence
measurements are performed (without inserting anisotropic elements on the beam path). In
contrast, they will appear different if the elements of their BCP matrices are measured. One
might guess that, without introducing the BCP matrix, differences among the three beams
would be revealed by the use of the local polarization matrix. This is partly true. As a
950 F Gori et al

matter of fact, fields 1, 2 could be distinguished from field 3 because the former are not
polarized, while the latter has a variable degree of polarization (as can be easily seen through
equations (13) and (33)–(35)). On the other hand, fields 1 and 2 remain indistinguishable
unless the BCP matrix description is adopted.

6. Conclusions

For a general optical field, a complete description of the correlations among the
electromagnetic vectors requires the use of a tensorial theory of coherence [1]. However,
for quasi-monochromatic, beam-like fields, a much simpler, albeit approximate, approach
can be based on the BCP matrix. As we saw in this paper, it can capture several significant
features of beams that are both partially polarized and partially coherent from the spatial
standpoint. In particular, it leads us to discover a peculiar interference law that applies to
mutual intensities and not only to optical intensities. Furthermore, it affords a simple tool
for describing Young’s interferometers modified through insertion of anisotropic elements.
Propagation problems can be easily tackled since each BCP matrix element behaves as
a mutual intensity. In this connection, it should be remarked that, differently from the
examples dealt with in this paper, in which we deliberately chose shape-invariant fields, the
propagation process can lead to drastic changes of the BCP matrix elements.
For the sake of brevity and simplicity, we limited ourselves to quasi-monochromatic
fields, further assuming all delays of interest to be small compared to the coherence time.
The removal of such limitations as well as the introduction of suitable cross-spectral densities
should not offer any difficulties.
The scalar theory of coherence applied to beam-like fields has led to the discovery of
a host of interesting models and phenomena. Let us simply quote the Schell–Gauss beams
[1, 29] and the occurrence of the Wolf effect [1, 30]; similarly, phenomena describable by
the ordinary polarization matrix fill volumes [2, 3]. The BCP matrix, which is the adaption
of the tensorial Wolf theory [1] to beam-like fields, establishes a link between these two
topics. Therefore, it should prove to be useful in several applications.
Nonetheless, the application of this type of formalism to radiation fields not describable
by the beam model requires some caution. The need for a complete vectorial treatment for
certain cases, such as black-body radiation [31], has been demonstrated (see for example
[1, 32]).

References

[1] Mandel L and Wolf E 1995 Optical Coherence and Quantum Optics (Cambridge: Cambridge University
Press)
[2] Shurcliff W A 1962 Polarization Light (Cambridge, MA: Harvard University Press)
[3] Azzam R M A and Bashara N M 1986 Ellipsometry and Polarized Light (Amsterdam: North-Holland)
[4] Pancharatnam S 1956 Proc. Ind. Acad. Sci. A 44 247
[5] Berry M V 1984 Proc. R. Soc. A 392 45
[6] Schmitzer H, Klein S and Dultz W 1993 Phys. Rev. Lett. 71 1530
[7] Aspect A, Dalibord J and Roger G 1982 Phys. Rev. Lett. 49 1804
[8] Simon P and Mukunda N 1990 Phys. Lett. A 140 165
[9] Barakat R 1996 Opt. Commun. 123 443
[10] Tudor T 1997 J. Opt. Soc. Am. A 14 2013
[11] Born M and Wolf E 1980 Principles of Optics (Oxford: Pergamon)
[12] Erdogan T and Hall D G 1990 J. Appl. Phys. 68 1435
[13] Tidwell S C, Ford D H and Kimura W D 1992 Opt. Eng. 31 1527
[14] Greene P L and Hall D G 1996 J. Opt. Soc. Am. A 13 962
 Beam coherence–polarization matrix 951

[15] Wolf E 1954 Nuovo Cimento 12 884

[16] Gori F 1998 Opt. Lett. 23 241
[17] James D F V 1994 J. Opt. Soc. Am. A 11 1641
[18] Martı́nez-Herrero R, Mejı́as P M and Movilla J M 1997 Opt. Lett. 22 206
[19] Gori F 1980 Opt. Commun. 34 301
[20] Starikov A and Wolf E 1982 J. Opt. Soc. Am. 72 923
[21] Gase R 1991 J. Mod. Opt. 38 1107
[22] Simon R and Mukunda N 1993 J. Opt. Soc. Am. A 10 95
[23] Friberg A T, Tervonen E and Turunen J 1994 J. Opt. Soc. Am. A 11 1818
[24] Jones R C 1941 J. Opt. Soc. Am. 31 488
[25] James D F V and Wolf E 1998 Opt. Commun. 145 1
[26] Siegman A E 1986 Lasers (Mill Valley, CA: University Science Books)
[27] Indebetouw G 1993 J. Mod. Opt. 40 73
[28] Gori F 1984 Coherence and Quantum Optics ed L Mandel and E Wolf (New York: Plenum) p 363
[29] Foley J T and Zubairy M S 1978 Opt. Commun. 26 297
Saleh B E A 1979 Opt. Commun. 30 135
Farina J D, Narducci L M and Collett E 1980 Opt. Commun. 32 203
Carter W H and Wolf E 1981 Opt. Acta 28 227
Carter W H and Wolf E 1981 Opt. Acta 28 245
Indebetouw G 1989 J. Mod. Opt. 36 251
Sundar K, Mukunda N and Simon R 1995 J. Opt. Soc. Am. A 12 560
[30] Wolf E and James D F V 1996 Rep. Prog. Phys. 59 771
Wolf E 1986 Phys. Rev. Lett. 56 1370
Dačić Z and Wolf E 1988 J. Opt. Soc. Am. A 5 1118
Kandpal H C, Vaishya J S and Joshi K C 1989 Opt. Commun. 73 169
Gori F, Marcopoli G-L and Santarsiero M 1990 Opt. Commun. 81 123
Palma C and Cincotti G 1997 J. Opt. Soc. Am. A 14 1885
[31] Jaiswal A K, Agrawal G P and Metha C L 1973 Nuovo Cimento 15b 295
[32] James D F V 1994 Opt. Commun. 109 209