HAL Id: jpa-00207388
https://hal.archives-ouvertes.fr/jpa-00207388
Submitted on 1 Jan 1973
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
On birefringence phenomena, associated with total reflection
B. Julia, A. Neveu
To cite this version:
B. Julia, A. Neveu. On birefringence phenomena, associated with total reflection. Journal de Physique,
1973, 34 (5-6), pp.335-340. �10.1051/jphys:01973003405-6033500�. �jpa-00207388�
LE JOURNAL DE PHYSIQUE
ON BIREFRINGENCE PHENOMENA, ASSOCIATED WITH TOTAL REFLECTION
B. JULIA and A. NEVEU
Laboratoire de
Physique Théorique
et HautesEnergies, Orsay (*) (Reçu
le 10 novembre1972)
Résumé. 2014 Nous étudions les
propriétés
debiréfringence
de la réflexion totale d’un faisceau lumineuxmonochromatique,
sur leplan
deséparation
de deux milieuxhomogènes isotropes,
par des méthodes de
déphasage.
Ceciexplique
à la fois l’effet Goos-Hänchenlongitudinal
danslequel
une source de lumièremonochromatique
nonpolarisée
donne deux images depolarisations rectilignes orthogonales,
et l’effet transverse étudié récemment parImbert,
pourlequel
les deuximages
sontpolarisées
circulairement. Nous donnons une méthodesimple
pour déterminer lespolarisations
et lespositions
desimages
d’une sourcecylindrique.
Abstract. 2014 We
study by phase-shift
methods thebirefringence properties
of total reflection of alight
beam at theseparation plane
between twoisotropic homogeneous
media. This includes the Goos-Hänchenlongitudinal
effect in wich a sourceof unpolarized light
issplit
into twoimages,
each of them
linearly polarized,
as well as the transverse shiftrecently investigated experimentally by Imbert,
where the effect is between left andright
circularpolarizations.
We give ageneral simple
method fordetermining
thepolarizations
andpositions
of theimages.
Classification Physics Abstracts
08.10 - 08.20 - 18.10
Introduction. - The existence of
birefringence
pro-perties
of total reflection oflight
on theseparation plane
of twoisotropic homogeneous
media has beenknown for some time
[1 ],
andexperimental
evidencewas first demonstrated
by
Goos and Hanchez[2].
These authors considered a
large
number of successive total reflections of apencil
oflight
between twoparallel planes,
and found that a source ofunpola-
rized
light
issplit
into twoimages,
oneimage being polarized parallel
to the incidenceplane,
the otherperpendicular. Using
a differentapparatus,
Imbert[3]
has
recently
found evidence for asplitting
betweenright
and left circularpolarizations (see
also[10]) :
inhis
experiment,
the successive total reflections of thepencil
ofunpolarized light
takeplace
on the sides of aregular
dielectricprism.
These
effects,
which are of the order of one wave-length
at eachreflection,
have beencomputed by
various methods
[4], [6], [7]
among which[4]
the useof
Poynting
vectors is verypopular.
In ouropinion however,
if thePoynting
vector may be useful toobtain an order of
magnitude
of theeffect,
it cannotprovide
acompletely
consistent treatment of theproblem :
inparticular
because it does not tellclearly
which
polarization
states arerelevant ;
another(*) Laboratoire associé au Centre National de la Recherche
Scientifique. Postal address : Laboratoire de Physique Théorique
et Hautes Energies, Bâtiment 211, Université de Paris-Sud, 91405 Orsay, France.
unphysical point
in thecomputation
of the various effects with thePoynting
vector is the use ofplane
waves of a finite
width, neglecting
what may occurnear the
edges.
For these reasons, anapproach using phase-shifts
seems safer to us, inparticular
becauseit is more
closely
related both to the exact solution ofMaxwell’s
equations,
and to the mechanism of for- mation ofimages
ingeometrical optics.
In this paper, we show that one can
compute
all effects which have beenexperimentally
observedby using
the exact solutions of Maxwell’sequations
forthe reflection of
polarized plane
waves on the sepa- rationplane
of twohomogeneous isotropic media, together
with somesimple geometrical optics approxi- mations,
which turn out to bealways
valid in expe- rimental situations. We deal with the usuallongitu-
dinal Goos-Hânchen effect in section
1,
and with Imbert’s transverse shift in sections 2 and 3. It turns out that thegeometrical arrangement
of theexperimental apparatus
is of crucialimportance
in the determination of thepolarizations
of the twoimages
of an unpola-
rized source.
1.
Longitudinal
shift(Goos-Hânchen).
- The lon-gitudinal
shift is the easiest tocompute, owing
to thesimpler geometrical
structure of thesystem :
we consider(Fig. 1)
a rectilinear sourceof light
S(narrow
slit for
instance), emitting
apencil
oflight
of smallaperture
Brn. Thispencil
istotally
reflected on theplane
surface between the vacuum and anisotropic
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01973003405-6033500
336
medium with index n. This
separation plane
isparallel
to the source ;
Oy
is its intersection with theplane
inwhich
figure
1 is drawn.Let i be the mean
angle
of incidence of the narrowpencil
emittedby
S. We candecompose
thispencil
into a linear
superposition
ofplane
waves, which haveslightly
different incidenceangles
onOy.
Eachplane
wave can then be reflected on the surface
Oy,
accord-ing
to Maxwell’stheory.
We then consider the set of the reflectedplane
waves. In theapproximation
ofgeometrical optics (to
bejustified later),
these reflectedwaves recombine
by superposition
to apencil which,
inlowest order in Brn, comes from the virtual
cylindrical
source 1 : 1 is
just
the center of curvature of theenvelope
of the reflectedplane
waves. 1 will ingeneral
be different from
S’,
thesymmetric
of S withrespect
to
Oy,
and itsposition
willdepend
on thepolarization,
thus
giving
rise to the effect.It is
straight
forward tocompute
the location of I : its coordinates in the(S’ X,
S’Y) system (see Fig. 1)
are :
where À is the
wavelength
of thelight
in vacuo, and5(i)
the
phase
shift of aplane
wave at totalreflection,
ascomputed
from Maxwell’stheory. Going
to the expres- sion of 5 in terms of i and n forpolarizations
perpen- dicular andparallel
to the incidenceplane [5],
onefinds the
explicit expressions :
B"1’)
As i
varies, I1
andIII
define two caustic surfaceswhich are the
images
of Sthrough
ouranisotropic optical system.
Several remarks are in order : formulae
(2)
and(4)
do not seem to exist in the literature on the
subject.
Our results for the
experimentally
observed quan-tities
L1 LL
andAll1
agree with those of Hora[6]
andBoulware
[7]
who also derives them via aphase-shift analysis. However, they
differby
a factorcos2 i/cos2 ij,
withsin il
=1/n
from the formula of Renard[8]
andImbert
[3] who, however,
derive themusing
a heuristicargument
withPoynting
vectors.°
Since the
image
of a beam with a finiteaperture
Brn isFIG. 1. - Goos-Hânchen Reflection
a finite
portion
of the causticsurface,
theseparation
of
polarizations
iscomplete only
if the distance between the twoimages
islarger
than theirspreading
out :
If this condition is not
satisfied,
one can still see the Goos-Hânchen effectby using light polarized,
perpen- dicular orparallel,
to the incidenceplane.
Let us now
briefly
discuss thevalidity
of theapproxi-
mation of
geometrical optics
used in our derivationof formulae
(1)-(4). Indeed,
it is inprinciple
unsafeto use
geometrical optics
to obtain a result like(1)-(4)
which is of the order of a
wavelength,
that is to say of the same order ofmagnitude
as diffraction.However,
in the
experimental apparatus,
the beam is reflected N times between two reflectionplanes
likeOy,
in orderto
amplify
the effect. In this process, it is clear that indeedphase-shifts
add up, whereas themagnitude
ofdiffraction,
which is setby
theaperture
of thebeam,
remains constant. Thanks to thismultiplying effect,
the two
images
aresplit by N(L11.L -
11111),
which canbe made much
larger
than diffraction.Hence,
whereasit would be
meaningless
for asingle reflection,
ourapproximation
ofgeometrical optics
isjustified
for thefull
experiment.
2. Transverse shift and
approximate analysis
ofImbert’s
experiment.
- In[3],
Imbert describes a niceexperiment
which shows that total reflection can alsogive
rise to asplitting
between the two circularpola-
risations. He also
computes
this effectby using
thePoynting
vector of the evanescent wave. What wewant to show in this section is that the effect can be
computed by using phases, just
as thelongitudinal
effect of section 1. The
origin
of this new effect oncircular
polarizations
is more subtle than the effect ofsection
1,
and restsessentially
on the rathercomplex
geometrical
structure of Imbert’sexperiment.
In this
section,
we shallonly give
anapproximate
treatment of the
effect, leaving
thecomplete analysis
for the next section.
Indeed,
we shall now show thatthe
geometry
of Imbert’sexperiment implies
that thesplitting
of the two circularpolarizations
isalready present,
even if thephase
shifts at each reflection arezero,
independent
from the incidenceangle
for bothlinear
polarizations, perpendicular
andparallel
to theincidence
plane.
Since in theexperiment
of[3]
allincidence-angles
are very close to thelimiting angle,
the
phase
shifts of both transverse andparallel polariza-
tions are indeed
quite
small.However, they
varyrapidly
when the incidenceangle varies,
so that thetreatment of this section is
only approximate
in thatsense : we shall find that it accounts for about two thirds of the
experimental result ;
a more refinedtreatment, taking
into account the variation ofphase shifts,
will bepresented
in the next section.We now describe
(Fig. 2)
thegeometry
of the expe-riment, following
Imbert’s notations whenever pos- sible. The source S is a line in the horizontalplane
H.It emits a
cylindrical
narrowpencil
oflight
with theangle
0’ withrespect
to the horizontalplane
H. Thelight
emittedby
S istotally
reflected on the faces of avertical
prism
ofglass
of index n = 1.8. The cross-section of this
prism
is anequilateral triangle.
The...
FIG. 2. - Imbert’s Apparatus
(perspective)
arrangement
is such that thelight
follows a helicalpath
in theprism.
The view of thatpath
from under is drawn onfigure
3. If thephase
shifts at each reflectionare zero, a
linearly-polarized plane-wave
remainslinearly polarized. However,
the direction of thepola-
rization vector is in
general
différent for each ray of the reflected beam.FiG. 3. - Imbert’s Apparatus (from under) This contrasts with the result of an ideal metallic
reflection,
where one of thephase
shifts is zero, and the other is 7:, in which case there is no effect.It is then easy to
get
convinced that thepolarizations
which should be used are not the linear
polarizations,
but the circular
polarizations
whichdiagonalize
theS-matrix,
in Imbert’sexperiment.
Infact,
after onereflection
they
are conserved for all the rays of the beam.For all rays issued from the source, let us take the vector no
orthogonal
to the source and to the beamas the
origin
ofphases.
For the firstreflection, phases
are to be counted with
respect
to the normal ni to the incidenceplane.
This vector nidepends
on0’,
theangle
of the ray and of the horizontalplane.
LetqJ 1 ((}’)
be the
angle
between no and ni, and n2 thesymmetric
of no with
respect
to the reflectionplane.
Ifcot(- cvt)
is the
phase
of aright (left)
circular wave withrespect
to no before the
reflection,
it is rot - 2qJ 1 ( -
rot - 2qJ 1)
with
respect
to n2 after the reflection.Hence,
total reflection induces for each circularpolarization
aphase
shift which
depends
on the ray of the beam.For the second
reflection,
theimages Si
ofS,
andn2 of no
play
the same role as S and no for the first reflection.By studying
thegeometry
of thesystem,
one also finds that the
angle
ç between two successive incidenceplanes
is ç = 2 qJ 1.Hence,
after N reflec-tions,
thephase
shift of aright (left)
circular ray isNqJ((}’) (- Nç(0’)). By
the sameargument
as for thelongitudinal
shift of section1,
one finds that the sourceis then
split
into twoimages, right
and leftpolarized,
which are
separated
from each otherby 2 d,
with338
The
experimental set-up
of[3] gives :
so that one finds
(1)
or,
numerically,
This is to be
compared
with Imbert’sexperimental
value
[3] ]
We thus see that our
approximate
treatment whichneglects
the variation of l5 with the incidenceangle gives
an effectwhich, though
toosmall,
is of thecorrect order of
magnitude.
3.
Complète description
of the transverse shift. - Let us now come to a morerigorous
treatment of thetransverse
shift,
in which we shall take into account the variation ofphase
shifts with the incidenceangle.
In the first two
preceding sections,
we have used theapproximation
ofgeometrical optics
to reconstruct theimages
from theenvelope
of theoutgoing plane
waves.However,
we aredealing
with asystem
in whichpola-
rization
plays
a crucialrole,
and standardgeometrical optics
do not tell us whichpolarizations
of the out-going
waves should be used to determine theposition
and nature of the
images.
Let us see, in an
example,
the kind ofproblems
oneruns into when one does not choose the
polarization
to which one
applies geometrical optics appropriately :
in section
1,
we couldimagine
to send an incoherentlight
on theapparatus,
and at theexit,
topick
up agiven
circularpolarization (say, right)
with a filter.The
envelope
method of section 1applied blindly
tothis
outgoing
set ofright circularly-polarized plane
waves would then
give
asingle image,
half waybetween the two
images
oflinearly polarized light.
This would contradict the result of section
1,
which isthat one should observe
precisely
theright
circularcomponent
of these twoimages.
Thisparadox
iseasily
solved : the twoimages
found in section 1 areobservable if
they
are wellseparated
withrespect
to diffraction effects : the beam must have alarge enough aperture,
and theintensity relatively
constant over thisaperture.
One then finds that over such anaperture
the filteredright
circularlight
would vary a lot inintensity, passing through
zero many times.Hence,
such anamplitude
cannot beinterpreted
ascoming
from asingle image,
butprecisely
as the inferencepattern
(1) Ly is the shift normal to the slit S ; it corresponds to a shift L’ y
= cos Ly qJ /2
normal to the incidence plane, orthat result agrees with Schilling’s [8]. It can be shown to hold for any prism whose cross-section is a regular polygon
produced by
theright
circularcomponent
of the twolinearly polarized images
of section 1.From this
example,
we can induce thegeneral
cri-terion for the choice of
polarizations
to which thegeometrical optics
treatment of section 1 and 2 can beapplied ;
we restrict ourselves tocylindrical beams,
which is
enough
if the source is a slit or a rectilinearobject
as in Imbert’sexperiment (2).
In this case,considering
the incident beam as asuperposition
ofplane
waves, the action of theoptical system
on eachplane
wave can be describedby
atwo-by-two
matrix B :if the
polarization
vector of the incidentplane
waveis
a , BP/ the polarization
vector of theoutgoing plane
wave is
B a .
For acylindrical
incidentbeam,
thematrix B
depends only
on oneparameter, 0’,
whichis the
angle
of the rays of the beam with some referenceplane.
It is clear from the discussion of thepreceding
section that the
polarization
statesa BP/
to which onecan
apply geometrical optics
are those such that theoutgoing polarization
statesB(e’) a W
areindependent
from
0’, except
for an overallphase,
when 0’ spans the beam : for any other choice of the incidentpolariza- tion,
thepolarization
would varyrapidly
over theoutgoing beam,
so thatsimple geometrical optics
would be
inapplicable.
In otherwords,
one shouldlook for the
polarizations
suchthat,
at the first orderin e
(3),
(2) We hope to deal with the more complex case of spherical
waves in a later publication.
(3) Here we derive rigorously a formulation in terms of geo- metrical optics from Maxwell’s equations (the crude approxi-
mation À = 0 would give J = 0).
1) We must divide the general problem into two problems
with one parameter (the eikonal) each.
2) We can write
([9]
p.119)
E and H in the general formomitting the phase factor e-irot ,(e and h are complex vectors).
3) In a homogeneous medium of index n, Maxwell’s equa- tions become
-
with
and the analogous equations with h. We use the eikonal 8 for geometrical optics : it means K = 0. We can also impose
L
= 0i. e. polarization is constant along each « ray » [9].
Conclusion : We must then have Ae = 0 to verify Maxwell’s equations
more (
preciselyÂ2 lael « lel 1 . )
That is the case if e(r) = Cte, that is if the only variation of E is in the eikonalor
( BP/ ) is thus an eigenvector
of B-1 B’,
with eigenva-
lue ia. B
being
aunitary matrix, a
is real. Thiseigen-
value is then related to the
displacement
L1 of theimage
of the sourceby
the formulaA
general polarization
state of the incident beam should then bedecomposed
as asuperposition
of thetwo
polarizations al , (Pl @ ("2) l’2
which solve eq. (6)
andone then observes two
images,
for such ageneral pola- rization,
withpolarizations B (fil a 1 and B a2 P2
whichare
separated by L11
1 -d 2.
It is nowstraightforward
toapply
those ideas to Imbert’sexperiment :
at eachreflection inside the
prism,
the relevant matrix A includes notonly
thephase
shiftsgiven by
the solu-tions of Maxwell’s
equations,
but also takes care of the rotation 9 of successive incidenceplanes
due tothe
peculiar geometry
of theexperiment.
One thenfinds :
where
ôjj (b.1)
is thephase
shift of aplane
wavepola-
rized
parallel (perpendicular)
to the incidenceplane, b = t (bll
-b .1). qJ
and ô are functions of B’through
the
geometry
offigure
2. The total transition matrix of theapparatus
iswhere N is the number of reflections inside the
prism.
Introducing
theangle
uby
one finds
where
Using
Imbert’sexperimental
values :and the
simplifying
fact thatone
easily
finds that theeigenstates of B-1 B’
areindeed the left and
right polarizations
when ô L---0,
and that the distance between the two
circularly polarized images
isOne can translate this in terms of the
displacement
per reflection :From formula
(8),
andusing
the fact that b N0,
onefinds
In this
equation, dç/d0’
is thegeometrical
termalready given by
section 2. One also has in Imbert’sexperi-
ment :
and,
ibeing
close to thelimiting angle ii,
one can usethe
approximate
formulae :Hence
(4)
or
numerically,
This value may be
compared
with Imbert’s result :The ratio of the two values is 1.03 for n = 1.8 and i
= il
ingood agreement
withexperimental
results.Conclusion. - We have shown that
simple
argu- mentsusing
classicaloptics
and a carefulanalysis
ofthe
geometrical
structure of Imbert’sexperiment
canyield
a directcomputation
andinterpretation
of hisresults in terms of the
phase
shifts at total reflectioncomputed directly
from Maxwell’sequations :
thesemethods also
apply
to thelongitudinal
Goos-Hâncheneffect,
andexplain
how asingle phenomenon (total
internal
reflection)
from asingle-unpolarized
source(4) That result was independently derived by D. Boulware [7],
340
can
give
twoimages
which are eitherlinearly pola-
rized
(Goos-Hânchen)
orcircularly polarized (Imbert).
Acknowledgements. -
We aregrateful
toPr. C. Bouchiat for a discussion which is at the
origin
of this
work,
and to Dr. C. Imbert and 0. Costa deBeauregard
for many informations about their expe- riment and their calculations.NOTE : While this work was in progress, we received
a
preprint by
D. Boulware[7]
in which aphase-shift analysis
of Imbert’sexperiment
is alsoproposed.
References
[1]
PICHT, J., Ann.Physik
3(1929) 433 ; Physik
Z. 30(1929)
905.[2] Goos,
F. andHÄNCHEN,
H., Ann.Physik
1(1947) 333 ;
5
(1949)
251.[3]
IMBERT,C., Phys.
Rev. D 5(1972)
787.[4]
RENARD, R. H., J.Opt.
Soc. Am. 54(1964)
1190.IMBERT,
C.,
loc. cit. and further references therein.COSTA de BEAUREGARD,
O.,
Nouv. Rev.Opt. Appl.
3(1972)
191.AGUDIN,
J.,Phys.
Rev. 171(1968)
1385.CHIN, K. W. and
QUIN,
J.J.,
Amer. J.Phys.
40(1972)
1847.RICARD, J., Nouv. Rev.
Opt. Appl.
2(1970)
273.ZAGRODZINSKI, J., Universita di Pisa
preprint.
[5] LANDAU,
L. and LIFSHITZ, E.,Electrodynamique
desMilieux Continus. Ed. de Moscou
(1969).
[6]
HORA,Optik
17(1960)
409.[7]
BOULWARE, D.,University
ofWashington preprint, (July 1972).
[8]
SCHILLING, H., Ann.Physik
16(1965)
122.[9]
BORN-WOLF,Principles of Optics,
second ed.Perga-
mon Press
(1964).
[10]
PAVAJEAU,J.,
C. R. Hebd. Séan. Ac. Sci. 263(1966)
276.