Professional Documents
Culture Documents
Barakat 1993
Barakat 1993
Richard Barakat*
Aiken Computation Laboratory, Division of Applied Sciences, Harvard University, Cambridge, Massachusetts 02138
Received February 3, 1992; accepted July 2, 1992; revised manuscript received July 27, 1992
The four laws of Arago-Fresnel governing the interference of polarized light were determined experimentally.
Analytical proofs are derived directly from first principles. The technique used for the proofs is that of the
spectral representation of stationary stochastic processes in conjunction with the Jones matrix description of
a polarizer.
complete polarization and complete unpolarization. The We require interaction of two beams characterized by
first person to grasp the importance of Stokes's research dZa((o) and Zb(wO)- It can be shown that if dZii((o) =
was Verdet,' 5 who made a significant simplification and dUji(w) + idVj (co), where I = a,b andj = 1,2, then
reinterpretation of his research. In fact, Verdet's version
is the one that is now commonly attributed to Stokes. As (dUja( )dVkb(&)) = (dja(co')dUkb()t))
usual, Lord Rayleigh 6 was quick to understand and ap-
preciate their research. However, we will use not the
= '1/2qjk(o)8(c' - )dw,
temporal Stokes parameters but rather their cousins, the (dUja(W)dUkb(t)) = (dVja(d)dVkb()) = 0,
spectral Stokes parameters of Wiener 7 in the interpreta-
tion of Barakat, 8" 9 who used the theory of stationary (dUji(o)') dVii (ok = 0 - (2.5)
stochastic processes in the context of the spectral- Here jk(co), withj $ k, is the cross-power spectral-density
representation theorem, rather than generalized harmonic function; it is generally complex valued.
analysis, for orthogonal increments. Technical reasons For thermal light, such as we are concerned with in this
for using the spectral-representation theorem rather than paper, then sj(t) is Gaussian, implying that dZj(&) is also
generalized harmonic analysis are discussed in Ref. 20. Gaussian. Thus dUj(c) and dV(co) not only are uncorre-
There are several ancillary issues, mainly of a mathe- lated but are also statistically independent.
matical character, that must be addressed before we can When E(t) is subjected to a linear operation, the input
prove the laws. Section 2 is devoted to such a discussion. and output dZ's are related through a 2 2 complex-valued
Inasmuch as we are going to be propagating various matrix J(w) that characterizes the optical system; thus
waves through polarizing devices, we must work in the
spectral domain rather than in the time domain in order dZ, 0(w&) dZi(,W)
to use multiplication operations rather than convolution 0 (2.6)
dZ2 (co) = J( ) dZ2 (W)
operations. This point has been emphasized in Ref. 18
and more recently (and in greater detail) in Ref. 21. or
dZ(Q) = J(Q)dZ(co). (2.7)
2. PRELIMINARIES
Every zero-mean, stationary stochastic process that is Note that we are working in the spectral domain, not in
mean-square continuous [such as the electromagnetic the time domain. If we were to work in the time domain,
field components ej(t) generated by thermal (Gaussian) then E(t) and E(t) would be related by a convolution op-
light] possesses the spectral representation 2 0 22 eration, not by a multiplication operation as connects dZ'
and dZi. Additionally, Eq. (2.7) holds on a sample realiza-
tion basis only; we must eventually deal with the observ-
E (t) = )(t = I exp(itwo)dZ(co) able power spectral density and cross-power spectral
density of the output.
The only matrix Jo) that we require is obviously that
= exp(it,) dZi(w) (2.1)
dZ 2(wO) of a polarizer as given by Ref. 23:
So(eO) = kb11(uO)
+ 22(0o) - 0, So = S51 = S2 = S3 = 0; (3.3)
S1(0) = 011(0) - 022(00), i.e., the light is unpolarized.
The optical path length of beams (AC) and (BC) are
S2((o) = 012 (() + ck 22 *(0J),
given by eOAC = kL, and ecTBC = kL 2, respectively, where k
S3((O) = i 2 (w) - j42*(0)- (2.13) is the mean wave number of the radiation at frequency o.
Since only the relative phases are important, we set
They are real because F(co) is Hermitian, positive semi-
definite. Consequently, e = eIBC - Ac = k(L2 - L,). (3.4)
1 So(w) + Si(c) S2 (0w) - iS3(Q) Note that e is also a function of ao.
(2.14)
4>Z = 2 S2 (0w) + iS3(wO) So(Z) - S,() The relation between the instantaneous field at the re-
ceiving plane that is due to beams (AC) and (BC) and
These parameters should not be confused with the aver-
the instantaneous field at the thermal source is
age values of the temporal Stokes parameters that are dis-
cussed in most texts (e.g., Born and Wolf 24). dZAc = dZe,
Although we are going to propagate light through vari-
ous polarizers, we cannot use the relation 8 dZBc = exp(ie)dZ,. (3.5)
The factor (1 + cos e) represents the interference be- for the spectral densities of the beams at point C on the
tween the two beams that results in fringes, in spite of receiving plane.
the fact that the incident radiation from the thermal The cross-spectral-density function of the two beams
source is unpolarized. The corresponding Stokes parame- 0,ac,epc is obtainable from the product (dZac and dZqBc*).
ters (at point C) are However, this product vanishes with ensemble averaging
because
SO = (1 + cos e)0,
(dZ1(0o)dZ 2 *(0w)) a 0, (4.10)
sIc = S2 c = S3 c = 0, (3.9)
as discussed in Section 2. Consequently,
so that the light is still unpolarized.
The really important point of Young's experiment is that 0kicC,aPC 0. (4.11)
interference fringes are created in spite of the fact that
the light is unpolarized. In the usual accounts of Young's The corresponding spectral Stokes parameters SjC at the
experiment, the light is taken to be linearly polarized and receiving plane are given by
then is shown to have interference fringes; the fact that
the source is unpolarized is not discussed. The present S = raC + k,9C,
mathematical machinery that uses the spectral represen- S = eaC - oC
tation theorem is ideally suited to the description of this
experiment. S2 = caaC,o-f6C + tkoao 3 C*,
For recent work on Young's experiment, the reader is S3 = iaC,73C - O.C, o6C*. (4.12)
referred to the recent work of James and Wolf. 2627
Consequently,
4. PROOFS OF THE FIRST AND S0 = /2[cos 2 a(1 + sin 2qica cos e)
SECOND LAWS 2
+ sin qlp(1 + sin 2p cos 010,
Proofs of the first two laws can be established from a SI = 1/2[COS
2
q1a( + sin 2,. cos e)
single configuration, as shown in Fig. 2. The polarizer at
a is set at ql,, = 0 (thereby generating a vertically polar- - sin2 qi3(1 + sin 2 q cos E)]+,
ized beam), while the polarizer at 18 is set at qip = 7/2 S2 = S3 C 0. (4.13)
(thereby generating a horizontally polarized beam). Both
beams then propagate to the polarizer at and then to In general, then, intensity Soc is modulated by the inter-
point C on the receiving plane. ference term cos e, with the strength of the modulation
The various spectral components at point C on the re- governed by tp,. and qi. The light at C is linearly polar-
ceiving plane are related to those at the thermal source by ized because S3 = 0, with degree of polarization
C= isici.
dZ.-aC = EJ()Ja(0)dZu, (4.1) P
dZoc = J(t/)J,,(7r/2)dZ a,
S0C (4.14)
(4.2)
where A proof of the first law follows by the setting of
1 0 a = i, (4.15)
(4.3)
- e so that the pass planes of the polarizers are parallel to
each other, as required by Arago and Fresnel. The
On calculating the norms of the vectors, we have Stokes parameters in Eqs. (4.13) reduce to
IdZ.CI2 = cos 2 41a(l + sin 2qa cos E)IdZiaI 2 , (4.4) So' = (1 + sin 24i cos e)4,
2 2
IdZOcl = sin p( + sin 2 cos e)IdZ 2 ,1 . (4.5) Si = (cos 24i) (1 + sin 2i cos E)0,
The ensemble average is now taken. Since S2 = S = 0, (4.16)
184 J. Opt. Soc. Am. A/Vol. 10, No. 1/January 1993 Richard Barakat
2
Ipac = 1/2 Cos Adyda,
SLIT
PLANE qloc = 1/2 sin Mya,
* SOURCE -B RECEIVING
PLANE
R
PLANE-
S' = 1/2 cos 2ya ,
S2 = 1/2 sin 2qfrO,
Fig. 3. Experimental configuration of polarizers for proof of the
third law: qpa = r/4, t,, = 0, and q} = 7r/l. s3 = O. (5.6)
Obviously there is no interference, because So' is a con-
while the degree of polarization becomes stant; this is the statement of the third law of Arago and
Fresnel. The light at C is linearly polarized with
Pc = Icos 2qil 1/2. (4.17)
P 1. (5.7)
Note that the maximum intensity at the receiving plane
is obtained at the setting i = r/4. This is exactly the
same result that we obtain from the simplified theory (see 6. PROOF OF THE FOURTH LAW
Ref. 11, p. 436). To prove the fourth law we consider the configuration that
The second law follows by the setting of the angles so Arago and Fresnel used; see Fig. 4.
that the pass planes of the polarizers are now perpendicu- The instantaneous field immediately after the light
lar to each other: beam passes through polarizers or and a (with ff== =T/4
PC = sin 2
a COS el 1 2. (4.20) 1 dZ,, + dZ 2 a| 1 0
doa2 0 2 dZ,. + dZ 2c.
Note that the degree of polarization for this case depends (6.3)
on both Hia,and e, whereas in the first case it depends only
on A. As a result of the combined beams, the instantaneous
field just before polarizer y is
5. PROOF OF THE THIRD LAW 1 dZ1 + dZ2 a_ 1 dZi + dZ 2 l
(6.4)
The configuration employed by Arago and Fresnel for the 2 dZi, + dZ2 f, 2 exp(ie)(dZ1, + dZ2 ,)
third law is shown in Fig. 3. The polarizer at a is set at
HPa = 0, the polarizer at (3is set at Adp = r/2, and the pass
and immediately after the combined beams pass through
plane of the polarizer at y is allowed to be arbitrary. polarizer , the instantaneous field is
The instantaneous fields from beams (a) and (B) just
before the polarizer at y are dZ~ J. (a )1 7 dZli + Z cy (6.5)
2exp(ie)dZ,,. + 2 dIZ2 ,
dZ, = Ja(0)dZa, (5.1)
dZa = Jp(7r/2)dZa, (5.2) * SLIT
* PLANE
while at the point C on the receiving plane,
THERMAL * A C
dZ.c = J,(4i) [dZya. + dZ,0 . (5.3)
SOURCE & [I UT
The two components of dZ c are
dZ,,c = cos 2 ydZl, + cos Miyr
sin Miy exp(ie)dZ 2,r,
dZ2,C = cos 2 Mi,sin qrzdZic + sin 2 f, exp(ie)dZ 2 f. (5.4)
Fig. 4. Experimental configuration of polarizers for proof of the
These expressions lead to fourth law: (, = 0, ik = r/2, and p,, is arbitrary.
Richard Barakat Vol. 10, No. 1/January 1993/J. Opt. Soc. Am. A 185
Following the now familiar procedures, we have 6. E. Mach, The Principles of Physical Optics, An Historical
and Philosophical Treatment (Methuen, London, 1926),
Chaps. 10-12.
2
4'iC = 1/2(COS B/, + 2 cos3 yi,sin B cos E)tk, 7. Z. Buchwald, The Rise of the Wave Theory of Light (U. of
Chicago Press, Chicago, Ill., 1989), Chap. 7.
02,C = 1/2(COS
2
Biy + 2 cos B1,sin 3 By, cos E)0, 8. G. Cantor, "The reception of the wave theory of light in
Britain," in Historical Studies in the Physical Sciences,
aC,2.aC = /4(sin 24ik(1 + sin 2ir, cos e)4y. (6.6) R. McCormach, ed. (Princeton U. Press, Princeton, N.J.,
1975), Vol. 6, pp. 109-132.
9. T. Preston, Theory of Light, 5th ed. (Macmillan, London,
The corresponding Stokes parameters at point C on the 1928), p. 348.
receiving plane are 10. R. W Ditchburn, Light, 2nd ed. (Interscience, New York,
1964), Chap. 12.
11. M. Francon, "Interferences, diffraction et polarisation," in
SoC = /2(1 + sin 24li cose)(k, Handbuch der Physik, S. FlUgge, ed. (Springer-Verlag,
Berlin, 1956), Vol. 24, Sec. 4.
Si' = 1/2(1 + sin 2ir, cos e)cos 24i,0, 12. T. P. Mathieu, Optics I. Electromagnetic Optics (Pergamon,
Oxford, 1975), pp. 178-179.
S2 = /2(1 + sin 2qli cos e)sin 24i,0,, 13. J. R. Partington, An Advanced Treatise on Physical Chemis-
try 4 Physical Optics (Macmillan, London, 1946).
S3 = 0. (6.7) 14. G. Stokes, "On the composition and resolution of streams of
polarized light from different sources," Trans. Cambridge
Philos. Soc. 9, 399-426 (1952). Reprinted in Mathematical
We still have interference fringes, but they can be made to and PhysicalPapers (Cambridge U. Press, Cambridge, 1901),
disappear if we let Bi,approach 45°. This is the statement Vol. 3.
of the fourth law. Note that the light is linearly polarized 15. E. Verdet, Legons d'Optique Physique (L'Imprimerie Impdri-
ale, Paris, 1869), Vol. 1. See also "Etude sur la constitution
because S3 C = 0; furthermore, the degree of polarization de la lumiere non polarisee et de la lumiere partiellement po-
at c is P, 1, even though Pa 0. larisee," Ann. Scientif. 'Ecole Normale Supgrieure 2, 291-
318 (1865).
16. Lord Rayleigh, "Wave theory of light," in Encyclopaedia
7. SUMMARY Brittanica(1888), Vol. 24. Reprinted in Collected Scientific
Papers (Cambridge U. Press, Cambridge, 1920). See Vol. 3,
I have derived the four Arago-Fresnel laws of the inter- pp. 140-147.
ference of polarized light directly from first principles 17. N. Wiener, "Coherency matrices and quantum theory,"
and included the thermal source as part of the system. J. Math. Phys. (M.I.T.) 7, 109-115 (1928). See also "General-
Reference is made to Mach, 6 as well as to older texts, for a ized harmonic analysis," Acta Math. 55, 118-245 (1930).
18. R. Barakat, "Theory of the coherency matrix for light of arbi-
detailed discussion of the consequences of these laws. trary spectral bandwidth," J. Opt. Soc. Am. 53, 317-323
The methods developed here are readily applied to other (1963). The coherency matrix referred to is a spectral-
examples of these types of problem. For example, density matrix, not a correlation matrix.
Stokes 4 makes reference to some experiments conceived 19. R. Barakat, "N-fold polarization measures and associated
by Herschel as well as by Dove.28 Evidently the Herschel thermodynamic entropy of N partially coherent pencils of
radiation," Opt. Acta 30, 1171-1182 (1983).
experiments were never carried out; it would be of some 20. J. L. Doob, Stochastic Processes (Wiley, New York, 1953),
interest to do so. Chap. 11.
21. C. Brosseau and R. Barakat, "Jones and Muller-Jones ma-
*The author is also with the Electro-Optical Research trices for random media," Opt. Comm. 84, 127-132 (1991).
Center, Tufts University, Medford, Massachusetts 02155. See Sec. 2.
22. A. M. Yaglom, Stationary Random Functions (Dover, New
York, 1962), Chap. 2.
REFERENCES 23. R. Azzam and N. Bashara, Ellipsometry and PolarizedLight
(North-Holland, New York, 1977).
1. J. W Herschel, "Light," in Encyclopaedia Metropolitana 24. M. Born and E. Wolf, Principlesof Optics, 6th ed. (Pergamon,
(1841), Vol. 4, pp. 341-586. Copies of this article were pri- Oxford, 1980), Chap. 10.
vately printed and distributed in 1827. 25. T. Young, "Experiments and calculations relative to physical
2. F. Arago and A. Fresnel, "L'action que les rayons de lumibre optics," Phil. Trans. R. Soc. London 94, 1-16 (1804). Re-
polarisde exercent les uns sur les autres," Ann. Chim. Phys. printed in H. Crew, ed., The Wave Theory of Light (American,
2, 288-304 (1819). New York, 1900).
3. A. Fresnel, Oeuvres Compl6tes de Augustin Fresnel (L'Impri- 26. D.James and E. Wolf, "Some new aspects of Young's interfer-
merie Impgriale, Paris, 1866). See Vol. 1, p. 509. ence experiment," Phys. Lett. A 157, 6-10 (1991).
4. F Arago, Oeuvres Compldtes de FrancoisArago (Gide, Paris, 27. D. James and E. Wolf, "Spectral changes produced in Young's
1854-1858). See Vol. 10, p. 132. interference experiment," Opt. Comm. 81, 150-154 (1991).
5. F Arago and A. Fresnel, "Memoir on the action of rays of po- 28. P. Dove, "On the phenomena exhibited by polarized light
larized light upon each other," in The Wave Theory of Light, when its plane of polarization is rapidly rotated," Philos.
H. Crew, ed. (American, New York, 1900), pp. 145-157. Mag. 30, 465-471 (1847).