Home Search Collections Journals About Contact us My IOPscience

The Simon - Mukunda polarization gadget

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1996 Eur. J. Phys. 17 279

([Link]

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: [Link]
This content was downloaded on 18/01/2014 at 12:30

Please note that terms and conditions apply.

Eur. J. Phys. 17 (1996) 279–284. Printed in the UK 279

The Simon–Mukunda polarization

gadget

V Bagini†, R Borghi†, F Gori†, M Santarsiero†, F Frezza‡, G Schettini‡

and G Schirripa Spagnolo§
† Dipartimento di Fisica, Università degli Studi di Roma ‘La Sapienza’, Piazzale A Moro, 2-00185, Roma, Italy
‡ Dipartimento di Ingegneria Elettronica, Università degli Studi di Roma ‘La Sapienza’, Via Eudossiana,
18-00184, Roma, Italy
§ Dipartimento di Energetica, Università di L’Aquila, Località Monteluco di Roio, 67040 Roio Poggio,
L’Aquila, Italy

Received 28 December 1995, in final form 16 April 1996

Abstract. The universal gadget recently proposed by Simon Résumé. Le gadget universel, proposé récemment par Simon
and Mukunda to synthesize any non-absorbing optical et Mukunda pour synthétiser chaque élément optique
element acting on the polarization of a wave is explained by pas-absorbent qui agit sur la polarisation d’une onde, est
elementary means. This different approach also leads to expliqué par des moyens élémentaires. Cette différente
alternative synthesis procedures. approche conduit aussi à des procédés alternatifs de synthèse.

1. Introduction implementation of alternative gadgets using QWPs and

rotators.
Simon and Mukunda discussed some years ago the
following problem: can an arbitrary non-absorbing
optical element, acting on the polarization of a wave, 2. Preliminaries
be synthesized using only quarter wave and half wave
plates (QWPs and HWPs, respectively)? They were We recall here a few basic concepts about Jones
able to prove (Simon and Mukunda 1989) that such a vectors and matrices. A deeper discussion can be
synthesis is possible if two QWPs and two HWPs are found in Swindell (1975) and Hecht (1987), while
used. Later, they refined their result showing (Simon a generalization to partially coherent light can be
and Mukunda 1990) that the synthesis process could seen in Mandel and Wolf (1995). We shall refer to
be made by using two QWPs and only one HWP. coherent monochromatic radiation, omitting a temporal
Simon and Mukunda reached such conclusions using the factor exp(−iωt), where ω is the angular frequency.
powerful and elegant tools of group theory. For readers The wavefront is supposed to be plane, at least
that are not familiar with this theory the full appreciation approximately. The general polarization state can be
of Simon and Mukunda’s works can be difficult. In specified by means of a column vector, say v, of the
this paper we intend to show how the Simon–Mukunda form
 
gadget can be explained in an elementary way. We a exp(iδx )
v= x , (1)
shall see that the most important feature of the gadget ay exp(iδy )
can be physically understood taking into account the where the two column elements are projections on the
effect of the rotation of a HWP on circularly polarized x- and y-axes of a given reference frame of a vector
radiation. This not only clarifies the gadget physical representing the electromagnetic wave, e.g. the electric
basis, but also allows us to connect the phenomena field. The positive quantities ax , ay are the amplitudes,
under consideration to other interesting effects, in which while the real quantities δx , δy are initial phases at the
rotating anisotropic elements produce frequency shifts considered point. As is known, the polarization state is
(Crane 1969, Sommargren 1975, Hu 1983, Bagini et not affected by the multiplication of ax , ay by a common
al 1994, Pippard 1994). The mathematical tools of factor. Therefore a normalization can be made by setting
our analysis are Jones matrices (Jones 1941), whose equal to one, in suitable units, the quantity ax2 +ay2 , which
basic properties are briefly recalled in the following is proportional to the radiation intensity. Furthermore,
section. As an additional bonus, our results suggest the polarization does not change if the same contribution to

0143-0807/96/050279+06$19.50 c 1996 IOP Publishing Ltd & The European Physical Society
280 V Bagini et al

the phase is added to δx and δy . Accordingly, we could where the star denotes complex conjugation. It may be
always set δx equal to zero and make the corresponding worthwhile noting that the set of matrices representing
adjustment to the phase constant δy . unitary transformations forms a group, usually indicated
Linear polarization states are characterized by a phase by U (2) (Joshi 1982). Special unitary transformations
difference δx − δy equal to 0 or π . The case δx − δy = π form a group too, usually denoted by SU (2). Because of
deserves a comment. This is that, if δx = 0 (arbitrarily), the constraint appearing in equation (7), SU (2) elements
then δy = −π . The x- and y-components of the are specified by three real parameters. This is why such
wave thus become, respectively, ax exp[i(kz − ωt)] and a group is said to be a three-parameter group.
ay exp[i(kz−ωt −π )]. Hecht’s comment (Hecht 1987) is There are several components that are able to change
that −π has been added to the phase of the x-component polarization without altering the intensity. The most
to get the phase of the y-component leading the x- important examples are, basically, wave plates and
component by π . rotators. The corresponding Jones matrices are
If θ is the angle between the direction of polarization  
and the x-axis, the corresponding Jones vector, indicated 1 0
M0 (ϕ) = ,
by lθ , is 0 exp(iϕ)
   
cos α − sin α
lθ =
cos θ
. (2) R(α) = , (8)
sin θ sin α cos α
In particular, we shall be interested in the cases θ = π/4 respectively, where the index 0 means that the wave
and θ = −π/4, in which equation (2) becomes plate axes coincide with the coordinate ones. The effect
1 1 1  1  of M0 is to increase the phase of the y-component of
lπ/4 = √ , l−π/4 = √ . (3) an amount ϕ with respect to the x-component. The
2 1 2 −1 effect of R is to rotate the field vector by an angle α.
Circular polarization states, which are characterized It can be easily controlled; in fact, if the field vector
by x- and y-components with the same amplitude and is rotated by such an angle, its x- and y-components
phase difference δx − δy = ±π/2, will be indicated by are obtained from the old ones through multiplication
cr and cl , depending on whether polarization is right or by R(α). Here α is assumed to be positive when it
left circular. The corresponding Jones vectors are represents a counterclockwise rotation in the xy-plane
1 1 1  1  (seen from positive z). Note that M0 can be written
cr = √ , cl = √ . (4)  
2 i 2 −i exp(−iϕ/2) 0
M0 (ϕ) = exp(iϕ/2) , (9)
We look now at the variation with time of the successive 0 exp(iϕ/2)
positions in the xy-plane from the side where z is
i.e. it is a special unitary matrix, apart from a phase
negative (Yariv and Yeh 1984). The convention to
factor. R(α) is also special unitary.
assess the sense of rotation of the electric field vector is
not unique (Born and Wolf 1991). We could look at the An important point for the following developments is
xy-plane from the side where z is positive. The terms the change in the Jones matrix of an optical component
righthanded and lefthanded could be interchanged. when the component is rotated through an angle α in the
The effect of an optical component acting on the xy-plane. Its matrix A before rotation must be replaced
polarization can be described by means of a suitable by A0 , where
2 × 2 Jones matrix, say A. The vector v 0 , expressing A0 = R(α)AR(−α). (10)
the polarization after passing through the component, is
v 0 = Av, i.e. in an explicit way, In particular, as it is physically obvious, the matrix
 0    
ax exp(iδx0 ) ζ ζ ax exp(iδx ) associated with a rotator is unchanged. In order
0 0 = 1 2 , (5) to prove equation (10), one can proceed as follows.
ay exp(iδy ) ζ3 ζ4 ay exp(iδy )
where ζj (j = 1, . . . , 4) are the complex elements of First, the x- and y-components of the incoming field
the A matrix. are found in a rotated reference frame whose axes
We are interested in optical components that do not are aligned to those of the optical component. This
change the wave intensity. In this case it is easily seen is obtained by multiplying the x- and y-components
that the following condition, in the original frame by R(−α). In fact, rotating
the frame through the angle α is formally equivalent
|det A| = 1, (6) to rotating the vector through the angle −α while
has to be satisfied, where det stands for determinant. It keeping the frame unchanged. The matrix A then
is then said that the A matrix characterizes a unitary describes the action of the optical component. Finally,
transformation. In particular, if det A is equal to one, we simply have to go back to the original frame
the transformations are said to be special unitary. It by means of the matrix R(α). We shall indicate
is not difficult to show that the corresponding matrices by M(ϕ, α) the matrix of a plate characterized by a
take on the form delay ϕ rotated by an angle α. In particular, M(ϕ, 0)
  coincides with M0 (ϕ). Applying equation (10) to
ζ1 ζ2
A= ∗ ∗ , |ζ1 |2 + |ζ2 |2 = 1, (7) equation (8), by simple calculations it is found that
−ζ2 ζ1
 Universal polarization gadget 281

4. Synthesized compensator and rotator

M(ϕ, α) = exp(iϕ/2)
 
cos ϕ2 − i sin ϕ2 cos 2α −i sin ϕ2 sin 2α
× . (11) The compensator, or variable wave plate (Born and Wolf
−i sin 2 sin 2α
ϕ
cos 2 + i sin 2 cos 2α
ϕ ϕ
1991), proposed by Simon and Mukunda (Simon and
For HWPs and QWPs, ϕ is π and π/2, respectively. Mukunda 1990), is made of two QWPs and one HWP.
The corresponding matrices, to be denoted by H (α) and Three different sequences of such elements are possible.
Q(α), are The simplest one to explain is, with obvious notations,
  Q–H–Q. It is clear that, if the axes of all the plates
cos 2α sin 2α
H(α) = , have the same direction, the total phase lag is 2π and
sin 2α − cos 2α no polarization change takes place. Indeed, it is easily
exp(iπ/4) checked by means of equation (13) that the product
Q(α) = √
2 matrix is
   
1 − i cos 2α −i sin 2α Q0 H0 Q0 =
1 0
× . (12) , (19)
−i sin 2α 1 + i cos 2α 0 1
In particular we have, for α=0, i.e. the identity matrix. Let us suppose now that the
    HWP is rotated by an angle α and study the action of
1 0 1 0
H0 = , Q0 = . (13) the system on arbitrarily polarized radiation. Before
0 −1 0 i
examining the matrix representation of the device, let
We conclude this section by observing that H(α) can be us try to understand its effect using the results we have
written in either of the following forms, just seen. Any incoming field can be represented by
H(α) = R(2α)H0 , (14) means of lπ/4 and l−π/4 in the form
H(α) = H0 R(−2α). (15) v = b1 lπ/4 + b2 l−π/4 , (20)
where b1 and b2 are suitable coefficients. It is then
sufficient to study the effect of the system on lπ/4 and
3. Some effects of wave plates on l−π/4 light separately. When the field b1 lπ/4 passes
through the first QWP, the field b1 cr (see equation (16))
In this section we show how, using Jones calculus, some is produced. This gives rise to b1 cl exp(2iα) for the
well known effects of wave plates are found, together effect of the rotated HWP (equation (18)). Finally, the
with the effect which is the physical basis of the gadget second QWP produces b1 lπ/4 exp(2iα) (equation (17)).
we study in this paper. By use of equations (3), (4) and Therefore, lπ/4 light impinging on the system comes
(13), we obtain out with unchanged polarization and amplitude, but
Q0 lπ/4 = cr , Q0 l−π/4 = cl . (16) with a phase change of 2α. This is an example
of the Pancharatnam phase, that is the phase taken
This result means that linear polarization at an angle by the radiation field as its polarization changes in a
π/4 (−π/4) with the x-axis is transformed into circular cyclic way (Pancharatnam 1956, Chyba et al 1988,
right (left) polarization by a QWP whose axes are the Agarwal and Simon 1990, Aravind 1992). In an
coordinate ones. Conversely, we easily find the result analogous way, l−π/4 light retains its polarization and
Q0 cr = lπ/4 , Q0 cl = l−π/4 (17) amplitude, but undergoes a phase change of −2α. In
conclusion, a phase difference 4α is introduced between
i.e. the rule of transformation of circularly into linearly the components lπ/4 and l−π/4 . Therefore the system is
polarized light. It is also well known that a HWP equivalent to a 4α wave plate, as is better seen in a
converts right circular light into left circular light and reference frame whose x-axis is along lπ/4 . In such a
vice versa. This is easily checked on applying H0 (see frame, the three plates appear to be rotated by π/4. The
equation (13)) to the vectors (4). Maybe it is less known validity of the previous analysis can be verified by using
how the conversion takes place if the HWP is rotated. Jones calculus. By means of equation (12), we find
Using equations (4) and (12), we find that π  π  π  i  1 −i 
H(α)cr = exp(2iα)cl , H(α)cl = exp(−2iα)cr , Q H +α Q =
4 4 4 2 −i 1
  
(18) − sin 2α cos 2α 1 −i
×
i.e. that the effect of rotation is to introduce a phase cos 2α sin 2α −i 1
 
factor. It is important to note that the phase change has exp(−2iα) 0
= . (21)
opposite sign for the two types of circular polarization. 0 exp(2iα)
As we shall see, this phenomenon is essential to explain By comparison with equation (9), we conclude that
the compensator proposed by Simon and Mukunda. π  π  π 
It is also the basis of the frequency shifts produced Q H +α Q = exp(−2iα)M0 (4α). (22)
by rotating anisotropic elements (Chyba et al 1988, 4 4 4
Agarwal and Simon 1990, Aravind 1992, Bagini et al This proves that any wave plate can be synthesized,
1994). apart from a phase factor, by the sequence Q–H–Q.
282 V Bagini et al

The plate adjustment needs only angular regulation where a phase factor has been omitted in M0 .
of the HWP. This is very useful for the practical Performing the matrix products, we see that the relations
implementation of the device. Furthermore, the phase
delay is linear in α and only one eighth of a turn is ζ1 cos ε − ζ2∗ sin ε = exp(−iϕ/2) cos χ ,
sufficient to make the delay vary of π , which is the −ζ1 sin ε − ζ2∗ cos ε = exp(iϕ/2) sin χ (29)
meaningful range of delays. It is physically obvious
that the equivalence has to remain valid if the whole have to hold. Equations (29), solved with respect to ζ1
device is rotated by an angle γ . In other terms, the and ζ2 , give
equality
 exp(−iϕ/2) cos ε cos χ − exp(iϕ/2) sin ε sin χ = ζ1 ,
π  π   π
Q γ+ H γ + +α Q γ + exp(−iϕ/2) sin ε cos χ + exp(iϕ/2) cos ε sin χ = −ζ2∗ .
4 4 4
= exp(−2iα)M(4α, γ ) (23) (30)
has to hold. This is easy to prove using equation (10). We have to prove that, for any choice of ζ1 and ζ2 , under
It has still to be shown that the compensator can the condition |ζ1 |2 + |ζ2 |2 = 1, a triplet of real numbers
be synthesized by the sequences Q–Q–H and H–Q–Q ϕ, ε, χ exists, solving equations (30). This can be seen
too. We shall discuss this subject later. Now we are in a simple way. Setting
interested in the synthesis of a rotator using two QWPs
and also one HWP. It appears in equation (14) that such ζ1 = a + ib, ζ2 = c + id,
a synthesis is possible, if we recall that two cascaded a 2 + b2 + c2 + d 2 = 1, (31)
QWPs produce a HWP, i.e. Q0 Q0 = H0 . Then, on
multiplying equation (14) from the right by Q0 Q0 , we equations (30) become
obtain ϕ ϕ
cos(ε + χ ) cos = a, cos(ε − χ ) sin = −b,
H(α)Q0 Q0 = R(2α). (24) 2 2
ϕ ϕ
Alternatively, the sequence Q–Q–H can be used, sin(ε + χ) cos = −c, sin(ε − χ ) sin = −d.
according to the relation 2 2
(32)
Q0 Q0 H(α) = R(−2α), (25)
These equations are easily solved with respect to ϕ, ε,
which is easily checked. χ. In fact we can set
ϕ p ϕ p
cos = a 2 + c2 , sin = b2 + d 2 . (33)
5. The general anisotropic optical component 2 2

From an optical point of view, as non-absorbing These relations agree with the condition a 2 + b2 + c2 +
anisotropic elements can only introduce phase differ- d 2 = 1 and account for the fact that the physically
ences between the x- and y-components and/or rotate meaningful range for ϕ is [0, π ]. From equations (32)
the polarization plane, one can envisage that any ele- ε and χ are easily obtained.
ment, possibly produced by cascaded devices, is equiv-
alent to a sequence of only two elements, i.e. a suitably
oriented wave plate and a rotator. Reasonable as it may 6. The Simon–Mukunda universal gadget
sound at the physical level, such an equivalence has
to be proved. In mathematical terms, one has to show We have seen in equation (26) that, apart from phase
that the A matrix describing the system can always be factors, we can always set
written as the product of a M matrix and a R matrix.
Neglecting possible phase factors, we shall refer to ma- A = M(ϕ, ε)R(λ). (34)
trices of the form (7). We have to show that, for any
given A, it is possible to find three real parameters ϕ, ε According to equation (23), equation (34) can be written
  
π
and λ such that
π +ϕ
A = M(ϕ, ε)R(λ) = R(ε)M0 (ϕ)R(λ − ε), (26) A = exp(iϕ/2)Q ε + H ε+
4 4
or, in an equivalent way, setting χ = λ − ε,  π
×Q ε + R(λ). (35)
R(−ε)A = M0 (ϕ)R(χ ). (27) 4
Equation (27), written in an explicit way, gives H and Q can be exchanged (see equation (A5) in the
   appendix) obtaining
cos ε sin ε ζ1 ζ2
 π  ϕ
− sin ε cos ε −ζ2∗ ζ1∗ π
   A = exp(iϕ/2)Q ε + Q ε+ +
exp(−iϕ/2) 0 cos χ − sin χ  4
 4 2
= ,
0 exp(iϕ/2) sin χ cos χ π +ϕ
×H ε + R(λ). (36)
(28) 4
 Universal polarization gadget 283

Furthermore, H and R can be compacted (see 7. An alternative gadget using rotators

equation (A10)). Accordingly, equation (36) becomes
 π  π ϕ
Let us consider now a different identity,
A = exp(iϕ/2)Q ε + Q ε+ + π  π 
 4  4 2
exp(iα)Q R(α) = M0 (2α)Q , (43)
π +ϕ λ 4 4
×H ε + − . (37)
4 2
whose validity is easily checked by using equations (22)
This proves the Simon and Mukunda general result: and (24). As all the identities given so far, it has
apart from phase factors, any element A, specified by the been derived (with a method that differs from ours)
parameters ϕ, ε, λ, can be synthesized by two QWPs by Simon and Mukunda, but it has some interesting
and a HWP. The order of the elements can be easily consequences that were not employed in the original
modified. In fact, using equation (A6), we can give papers. Multiplying both sides of equation (43) by
equation (37) the form Q(−π/4) from the right or from the left, we obtain
   π   π
π π +ϕ λ
A = exp(iϕ/2)Q ε + H ε+ − Q R(α)Q − = exp(−iα)M0 (2α), (44)
 
4 4 2 4π 4 
π
π Q − M0 (2α)Q = exp(iα)R(α). (45)
×Q ε + − λ . (38) 4 4
4
Finally, using equation (A6) of the appendix once again, These relations show that, apart from a phase factor, a
equation (38) can be put in the form rotator is changed into a wave plate and vice versa when
  such elements are sandwiched between two crossed
π +ϕ λ QWPs. Equation (45) has also been derived in a recent
A = exp(iϕ/2)H ε + −
4 2 paper (Ye 1995) where an experimental realization was
 π ϕ   π  also presented. Let us now refer to equation (44).
×Q ε + + − λ Q ε + − λ . (39) Notice that, in order to synthesize a 2α wave plate, the
4 2 4
rotator must give exactly the rotation α. In other terms,
In section 4, we have shown that a compensator, or a fixed rotator can be used to synthesize only one wave
variable wave plate, can be synthesized by the sequence plate. The question arises of whether it is possible to
Q–H–Q. Indeed, letting in equation (38) obtain a greater flexibility by rotating one (or both) of
the QWPs. It is easy to calculate that
λ = 0, ϕ = 4α, ε = γ, (40)
π   π
equation (23) is found again. We are now able to Q R(α)Q γ −
4   4 
answer the question posed in section 4, i.e. how can the π π
compensator be synthesized by the sequences Q–Q–H =Q R(α + γ )Q − R(−γ )
4 4
and H–Q–Q. Making the substitutions of equation (40)
= exp[−i(α + γ )]M0 (2α + 2γ )R(−γ ), (46)
into equation (37), we obtain
 π  π  having used equation (44). As one can see, any phase
M(4α, γ ) = exp(2iα)Q γ + Q γ + + 2α lag can be obtained acting on γ . The price to be paid
  4 4
π is that the synthesized plate is preceded by a rotator (it
×H γ + + α . (41) must be kept in mind that, in matrix multiplication, the
4
rightmost term represents the first element encountered
The same substitutions, performed in equation (39), give by the light). An analogous procedure is used to prove
 π  the identity
M(4α, γ ) = exp(2iα)H γ + + α
 π  π
 π  4 π Q ε+ R(α)Q − = exp[−i(α − ε)]
×Q γ + + 2α Q γ + . (42) 4 4
4 4 ×R(ε)M0 (2α − 2ε), (47)
Taking into account equations (23), (41) and (42), we
conclude that the compensator can be synthesized by providing us with a synthesis process in which the
any combination involving two QWPs and a HWP. rotator R(ε) follows the plate. Equations (46) and
However, it is worthwhile noting that the sequence (47) have some practical usefulness because obviously
given in equation (23) seems the most useful for a the effect of the rotation can be compensated by a
practical implementation of the device, as it requires suitable rotation of part of the apparatus. Notice that
only angular regulation of the HWP. The alternative equations (46) and (47) also hold in the case α = 0.
gadgets in equations (41) and (42) need control over two Therefore any wave plate, apart from a rotation, can be
relative angles and their realization is somehow trickier. synthesized using only two QWPs.
284 V Bagini et al

8. Conclusions account equation (A2), equation (A3) is generalized in

the following way,
The Simon–Mukunda gadget is a simple device that can
H(γ )Q(γ + α) = Q(γ − α)H(γ ), (A4)
be usefully exploited in the laboratory for synthesizing
any non-absorbing anisotropic element. As we have or, changing the variables,
shown its operation can be explained through elementary
H(γ )Q(β) = Q(2γ − β)H(γ ), (A5)
considerations. An additional result of the technique
used in the present paper is that, apart from a rotation, Q(β)H(γ ) = H(γ )Q(2γ − β). (A6)
any wave plate can be synthesized using only two Another pair of identities that derives from equation (14)
quarter-wave plates. We finally note that, following an is
inverse procedure with respect to Simon and Mukunda, α 
the acquired familiarity with combination of optical R(α)H0 = H , (A7)
elements could be used as a didactical tool to explore 2α
some properties of the group SU (2), whose structure H0 R(α) = H − . (A8)
can be represented in terms of Jones matrices. 2
Rotating by an angle γ , as R(α) remains unchanged
under any rotation, we finally obtain
Acknowledgments  α
R(α)H(γ ) = H γ + , (A9)
 2
The present research has been supported by MURST α
and INFM. H(γ )R(α) = H γ − . (A10)
2

Appendix. Some useful identities

References
We discuss here some simple identities that allow us
to change the positions of HWP and QWP. We begin Agarwal G S and Simon R 1990 Phys. Rev. A 42 6924–7
Aravind P K 1992 Opt. Commun. 94 191–4
with the following observation. If the Jones matrices Bagini V, Gori F, Santarsiero M, Frezza F, Schettini G and
representing three optical elements A, B and C verify Schirripa Spagnolo G 1994 Eur. J. Phys. 15 71–8
the relation Born M and Wolf E 1991 Principles of Optics 6th edn
A = BC, (A1) (Oxford: Pergamon)
Chyba T H, Wang L J, Mandel L and Simon R 1988 Opt.
then also the matrices obtained by rotating every element Lett. 13 562–4
by one and the same angle γ verify an analogous Crane R 1969 Appl. Opt. 8 538–42
relation, i.e. Hecht E 1987 Optics 2nd edn (Reading, MA:
Addison-Wesley)
A0 = B0 C0 . (A2) Hu H Z 1983 Appl. Opt. 22 2052–6
Jones R C 1941 J. Opt. Soc. Am. 31 488
This property is physically obvious. It means that, if two Joshi A W 1982 Elements of Group Theory for Physicists
cascaded optical elements are equivalent to a third one, (New Delhi: Wiley Eastern)
the equivalence cannot depend on the reference frame. Mandel L and Wolf E 1995 Optical Coherence and Quantum
We have already used this idea to obtain equation (23) Optics (Cambridge: Cambridge University Press)
from equation (22). In a formal way, equation (A2) is Pancharatnam S 1956 Proc. Ind. Acad. Sci. A 44 247–62
proved by multiplying equation (A1) by R(γ ) from the Pippard A B 1994 Eur. J. Phys. 15 79–80
left and by R(−γ ) from the right and inserting between Simon P and Mukunda N 1989 Phys. Lett. 138A 474–80
B and C the identity operator R(−γ )R(γ ). Let us —— 1990 Phys. Lett. 140A 165–9
Sommargren G E 1975 J. Opt. Soc. Am. 65 960–1
consider now the equation Swindell W 1975 Polarized Light (Stroudsburg, PA: Dowden,
H0 Q(α) = Q(−α)H0 . (A3) Hutchinson and Ross)
Yariv A and Yeh P 1984 Optical Waves in Crystals (New
As it is easily checked using equations (12) and (13), York: Wiley)
this is an identity that holds for any α. Taking into Ye C 1995 Opt. Eng. 10 3031–5