Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Physics Letters A 312 (2003) 263–267

www.elsevier.com/locate/pla

Unified theory of coherence and polarization of random


electromagnetic beams
Emil Wolf
Department of Physics and Astronomy and the Institute of Optics, University of Rochester, Rochester, NY 14627, USA
Received 9 April 2003; accepted 24 April 2003
Communicated by P.R. Holland

Abstract
The usual treatment of polarization of randomly varying electromagnetic waves is based on the well-known Stokes parameter
or, equivalently, on 2 × 2 coherence matrices (also called polarization matrices). Neither of them is adequate to elucidate the
recently discovered changes of the state of polarization which a light beam may undergo as the beam propagates.
We present a unified theory of coherence and polarization of random electromagnetic beams which brings out clearly the
intimate relationship which exists between these two phenomena and which makes it possible to predict the changes in the
state of polarization of a partially coherent electromagnetic beam on propagation. We illustrate the analysis by showing that a
completely unpolarized beam may be spatially fully coherent.
 2003 Elsevier Science B.V. All rights reserved.

PACS: 42.25.Ja; 42.25.Kb; 43.50.Ar

1. Introduction meters are simple linear combinations of correlations


that may exist between two mutually orthogonal com-
ponents of the electric vector perpendicular to the di-
Despite the great deal of literature that exists about
rection of propagation of a fluctuating electromagnetic
polarization of light and other random electromagnetic
plane wave. This fact became evident from a well-
radiation, the underlying theory has hardly advanced
known paper of N. Wiener [3] on the generalized har-
since 1858 when G.G. Stokes [1] introduced four pa-
monic analysis and from an investigation of Wolf [4]
rameters which now bear his name, to characterize the
when they introduced the so-called coherency matri-
state of polarization of a light wave. His theory ac-
ces which, though closer in spirit to the formalism of
tually predates the formulation of Maxwell’s electro-
quantum mechanics, did not provide a generalization
magnetic theory of light. Only after Maxwell’s theory
of Stokes’ treatment.
was formulated and the theory of random processes
A severe restriction of Stokes’ theory is the fact that
began to be used in the analysis of random optical
his parameters only contain information about correla-
fields [2] it has become apparent that the Stokes para-
tions between the Cartesian components of the fluctu-
ating electric field vector at one point in space at the
E-mail address: ewlupus@pas.rochester.edu (E. Wolf). same instant of time [5]. For this reason the theory

0375-9601/03/$ – see front matter  2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0375-9601(03)00684-4
264 E. Wolf / Physics Letters A 312 (2003) 263–267

cannot predict any changes which the state of polar- {E(r, ω)} ≡ {Ei (r, ω)} (i = x, y), be the statistical
ization of a light beam may undergo as the beam prop- ensemble of the spectral component of frequency ω
agates. That there may be such changes has only re- of the fluctuation electric field E(r, ω) at a point r
cently attracted attention [6]. These investigations in- in space. The existence of such an ensemble is well
dicate, mainly by means of various examples, what known in the so-called coherence theory in the space-
might be expected on general grounds, namely, that frequency domain ([8, Section 4.7]). We may then
there is an intimate relationship between polarization characterize the second-order coherence properties of
properties of a random electromagnetic beam and its the beam by a 2 × 2 cross-spectral density matrix1
coherence properties. However, in order to express  
such a relationship in a satisfactory manner it is nec- W ≡ Wij (r1 , r2 , ω) = Ei∗ (r1 , ω)Ej (r2 , ω) (1)
essary to characterize spatial coherence quantitatively, where the asterisk denotes the complex conjugate.
by means of an appropriate “degree of coherence”. A Suppose that an opaque screen A is placed in the
measure of this kind is well known in the scalar theory path of the beam at right angle to the axis, covering the
of random waves (see, for example, [7, pp. 564 and plane z = z0 and containing small openings at points
586], and [8, pp. 163 and 171]), but no satisfactory Q1 ≡ (ρ 1 , z0 ) and Q2 ≡ (ρ 2 , z0 ) (see Fig. 1). We
measure of this kind is available within the framework will determine the spectral density (spectrum) of the
of electromagnetic theory [9].
In the present Letter we first introduce such a mea-
sure which might be called the spectral degree of co-
herence of a random, statistically stationary, electro-
magnetic beam [9], based on the changes which, as we
will show, the spectrum of such a beam undergoes, in
general, when two small portions of it are isolated and
are then superposed in a Young’s interference exper-
iment. Using this definition of the spectral degree of
coherence and a somewhat refined version of the usual
expression for the degree of polarization we show how
these quantities and the changes which they undergo
on propagation may be calculated from a unified the-
ory.
We illustrate our analysis by an example which
shows that a completely unpolarized electromagnetic
beam may be spatially fully coherent within the frame- Fig. 1.
work of second-order coherence theory of electromag-
netic beams. It seems likely that the outputs of some 1 This matrix must be distinguished from the 2 × 2 coherence
lasers are of this kind. (or polarization) matrix J used in the elementary theory of partial
polarization ([7, Section 10:9]; [8, Section 6.2]). Unlike the cross-
spectral density matrix W the matrix J depends on only one and not
on two spatial arguments, a fact which makes it not applicable for
2. The spectral interference law for random
studying the propagation properties of the field. The matrix J may
electromagnetic beams be derived from the more general cross-spectral density matrix W
by means of the formula
Consider a random, wide-sense stationary, electro- 
magnetic field. We assume that the field is a beam, J(ρ) = W(ρ, ρ, ω) dω
which propagates close to the z-direction. This re-
striction to beams implies that the Cartesian compo- where ρ is the two-dimensional position vector of a point in any
transverse plane z = constant, the integration extending over the
nents of the field vectors along the z-direction may bandwidth of the light.
be neglected. Let us choose mutually orthogonal x- Here and elsewhere we write, for the sake of clarity E(r, ω)
and y-axes perpendicular to the z-direction and let rather than E(P, ω), etc.
E. Wolf / Physics Letters A 312 (2003) 263–267 265

transmitted light at a point P on a screen B parallel to The formula (5) then gives S = S (1) , where
A and some distance away from it. Assuming that the
|K1 |2
angles of incidence and diffraction are small we may S (1) (P, ω) = S(Q1 , ω). (6)
represent the electric field at P by a statistical ensemble R12
{E(P, ω)}, whose typical realizations are given in
S (1) (P, ω) evidently represents the spectral density at
terms of the realizations E(Q1 , ω) and E(Q2 , ω) of the
P of the light reaching this point only through the
field at the two pinholes by the formula
pinhole at Q1 . A strictly similar expression is obtained
eikR1 eikR2 for the spectral density, say S (2) (P, ω), at P if the light
E(P, ω) = K1 E(Q1 , ω) + K2 E(Q2 , ω) , reached that point only through the pinhole, at Q2 .
R1 R2
Using these expressions and making use of the fact
(2)
that
where R1 and R2 are the distances Q1 P and Q2 P,    ∗
respectively, and Tr W(Q2 , Q1 , ω) = Tr W(Q1 , Q2 , ω) ,
i which follows from the definition (1) of the cross-
KJ = − dAj (j = 1, 2), (3) spectral density matrix, one readily obtains from
λ
Eq. (5) the following expression for the spectral
k = 2π/λ, λ = 2πc/ω being the wavelength associ-
density at the point P:
ated with the frequency ω, c being the vacuum speed
of light and dAj are the areas of the pinholes. The ex- S(P, ω) = S (1) (P, ω) + S (2) (P, ω)
pression (3) for Kj follows from elementary diffrac-  
tion theory (cf. [7, p. 46, Eq. (14)]). + 2 S (1) (P, ω) S (2) (P, ω)
 
Let us now consider the spectral density S(P, ω) of × Re µ(Q1 , Q2 , ω)eik(R2 −R1 ) , (7)
the field at the point P. We may identify the spectral
where Re denotes the real part,
density with the average value of the electric energy
density. Apart from an inessential proportionality µ(Q1 , Q2 , ω)
constant which depends on the choice of units, S(P, ω) Tr W(Q1 , Q2 , ω)
is given by the expression =  , (8)
  Tr W(Q1 , Q1 , ω) Tr W(Q2 , Q2 , ω)
S(P, ω) = E∗ (P, ω) · E(P, ω)
and Tr W(Qj , Qj , ω) = S(Qj , ω) (j = 1, 2), is the
= Tr W(P, P, ω), (4)
spectral density of the light at the pinholes at Qj .
where Tr W is the trace of the cross-spectral density If, as is usually the case, S (2) (P, ω) ≈ S (1) (P, ω),
matrix (1). If we substitute from Eq. (2) into Eq. (1) the formula (7) reduces to
and use Eq. (4) we obtain for the spectral density at P   
the formula S(P, ω) = 2S (1)(P, ω) 1 + Re µ(Q1 , Q2 , ω)eiδ

= 2S (1)(ω) 1 + µ(Q1 , Q2 , ω)
|K1 |2 |K2 |2  
S(P, ω) = 2
S(Q1 , ω) + S(Q2 , ω) × cos α(Q1 , Q2 , ω) + δ , (7a)
R1 R22
eik(R2 −R1 ) where α(Q1 , Q2 , ω) is the argument (phase) of µ(Q1 ,
+ K1∗ K2 Tr W(Q1 , Q2 , ω) Q2 , ω) and δ = (R2 − R1 ).
R1 R2
The formula (7) shows that the spectrum at a
e−ik(R2 −R1 ) point P in the observation plane B is not just the sum
+ K1 K2∗ Tr W(Q2 , Q1 , ω) .
R1 R2 of the spectra of the two beams reaching that point but
(5) differs from it by the presence of the third term, which
obviously represents the effect of interference. We
This formula may be rewritten in a physically more may, therefore, refer to the formula (7) as the spectral
significant form by noting first that if the pinhole at Q2 interference law for the superposition of random
was closed, then dA2 = 0 and, consequently, K2 = 0. electromagnetic beams. It is of exactly the same form
266 E. Wolf / Physics Letters A 312 (2003) 263–267

as the spectral interference law for scalar wavefields the traces which appear in the expression (8) and then
([7, p. 587, Eq. (14)], or [8, p. 173, Eq. (4.3-54)]), evaluate the spectral degree of coherence µ(r1 , r2 , ω)
the only difference being the different definition of by the use of that formula.
the factor µ(Q1 , Q2 , ω). In the scalar theory the factor Let us now turn to the (spectral) degree of polariza-
is known as the spectral degree of coherence. We tion P(r) at a point P(r) in the beam. It may be defined
will, therefore, call the corresponding factor, defined by the expression2
by Eq. (8), the spectral degree of coherence of the
electric field at the points Q1 and Q2 [10]. It is 4 Det W(P, P, ω)
P= 1− . (11)
not difficult to prove by using certain non-negative [Tr W(P, P, ω)]2
definitions conditions which the cross-spectral density
tensor W(Q1 , Q2 , ω) satisfies ([8, Section 6.6.1]) that Just as with the spectral degree of coherence one
can determine how the spectral degree of polarization
0  µ(Q1 , Q2 , ω)  1. (9) changes as the beam propagates, by first solving
Eqs. (10) and then evaluating the expression (11).
It is clear from Eq. (7a) that the interference term will In a similar way one can study changes of other
be absent when µ = 0 and that it will have maximum quantities that characterize more completely the state
value (giving rise to interference fringes with the high- of polarization of the beams, i.e., those representing
est contrast) when |µ| = 1. Hence µ = 0 represents the polarized portion of the beam at the point P(r).
complete spatial incoherence and the other extreme,
|µ| = 1, represents complete spatial coherence of the
spectral component of frequency ω, of the fluctuating 4. An example
electric field at the two pinholes.
Methods have been described in the literature and
We will conclude this analysis by a simple example
some of them implemented for measuring the spectral which will illustrate the subtleties involving the rela-
degree of coherence of scalar fields [11]. The same tionship between coherence and polarization.
methods can be used to measure the spectral degree of
Consider the cross-spectral density matrix (with δij
coherence (8) of the electric field, the only difference denoting the Kronecker symbol)
being that the measurements have now to be carried
out by first placing polarizers at the pinholes to W(Q1 , Q2 , ω) = 12 I0 (ω)δij (12a)
transmit components of the electric field along one 1
of the transverse directions (x, say) and then making = 2 I0 (ω) 0
, (12b)
1
measurements with the polarizers transmitting the 0 2 I0 (ω)
component along the other direction (y). where I0 (ω) is independent of position. On substitut-
ing from Eq. (12b) into the formula (11) we see at once
that
3. Propagation of coherence and of polarization
P(Q1 , ω) = P(Q2 , ω) = 0. (13)
The cross-spectral density tensor W is known to Next let us substitute from Eq. (12b) into Eq. (8).
satisfy in free space the two Helmholtz equations (see One then finds that
[7, Section 6.6.3])
µ(Q1 , Q2 , ω) ≡ 1. (14)
∇12 W(r1 , r2 , ω) + k 2 W(r1 , r2 , ω) = 0, (10a)
Eq. (13) implies that such a beam is completely
∇22 W(r1 , r2 , ω) + k 2 W(r1 , r2 , ω) = 0, (10b) unpolarized at each pinhole. Eq. (14) shows that it
where ∇12 and ∇22 are the Laplacian operators acting
with respect to r1 and r2 , respectively. These equations 2 The formula (11) is a refinement of the usual expression for

make it possible to determine how the spectral degree the degree of polarization, viz., P = 1 − 4 Det J(ρ)/[Tr J(ρ)]2 ,
of coherence changes on propagation. One would first where J(ρ) is the usual coherence matrix of the elementary theory
determine, by the use of these equations, the values of of partial polarization, mentioned in footnote 1.
E. Wolf / Physics Letters A 312 (2003) 263–267 267

is completely spatially coherent at the pinholes. It is E. Collett, Polarized Light: Fundamentals and Applications,
possible that some lasers generate beams of this kind. Marcel Dekker, New York, 1993.
[3] N. Wiener, Acta Math. 55 (1930), Section 9.
[4] E. Wolf, Nuovo Cimento 13 (1959) 1165.
[5] In spite of these limitations the Stokes parameters and co-
Acknowledgements herency matrices are frequently used not only in classical op-
tics but also in quantum treatments of radiation and scattering.
This work was stimulated by interesting experi- See, for example, U. Fano, J. Opt. Soc. Am. A 39 (1949) 859;
ments of Dr. Aristide Dogariu and his students and U. Fano, Phys. Rev. 93 (1954) 121;
G.S. Agarwal, Opt. Commun. 37 (1981) 349.
other collaborators at the University of Central Florida, [6] Of the many papers on this subject the following two are of
on inverse scattering problems with light of different special relevance to the present investigation: D.F.V. James,
states of polarization. J. Opt. Soc. Am. A 11 (1994) 1641;
I am obliged to Dr. Sergey Ponomarenko for some F. Gori, M. Santarsiero, S. Vacalvi, R. Borghi, G. Guattari, Pure
helpful comments relating to the analysis presented in Appl. Opt. 7 (1998) 941.
[7] M. Born, E. Wolf, Principles of Optics, Cambridge Univ. Press,
this Letter. Cambridge, 1999.
This research was supported by the US Air Force [8] L. Mandel, E. Wolf, Optical Coherence and Quantum Optics,
Office of Scientific Research under Grant No. F49620- Cambridge Univ. Press, Cambridge, 1995.
03-1-0138, by the Engineering Research Program [9] An earlier definition proposed by B. Karczewski, Phys. Lett. 5
of the Office of Basic Energy Sciences at the US (1963) 191, Nuovo Cimento 30 (1963) 906, did not attract
much attention, probably because the quantity he proposed
Department of Energy under grant No. DE-FG02-ER could not easily be measured.
45992, and by the Defense Advance Research Project [10] Essentially the same quantity was introduced rather formally
Agency under Grant MDA 972011043. and in a somewhat different context by W.H. Carter, E. Wolf,
Phys. Rev. A 36 (1987) 1258, Eq. (6.17).
[11] D.F.V. James, E. Wolf, Opt. Commun. 145 (1998) 1;
S.S. Titus, A. Wasan, J.S. Vaishya, H.C. Kandpal, Opt. Com-
References mun. 175 (2000) 45;
V.N. Kumar, D.N. Rao, J. Mod. Opt. 48 (2001) 1455;
[1] G.G. Stokes, Trans. Cambridge Philos. Soc. 9 (1852) 399; G. Popescu, A. Dogariu, Phys. Rev. Lett. 88 (2002) 183902.
G.G. Stokes, Mathematical and Physical Papers, Cambridge
Univ. Press, Cambridge, Vol. 3, p. 233.
[2] For modern accounts of the theory and applications of polar-
ized light see, for example, C. Brosseau, Fundamentals of Po-
larized Light, Wiley, New York, 1998;

You might also like