You are on page 1of 9

PHYSICAL REVIEW B 68, 045121 2003

One-dimensional photonic crystal: The Kronig-Penney model

Subodha Mishra and S. Satpathy
Department of Physics & Astronomy, University of Missouri, Columbia, Missouri 65211, USA
Received 31 October 2002; published 30 July 2003
We formulate the photonic band-structure problem for a one-dimensional photonic crystal in terms of the
reflection and transmission coefficients, obtaining a transcendental photonic band equation. The reflection and
the transmission coefficients may be evaluated by using the standard transfer-matrix method. The structure of
the equation reveals the existence of gaps, analogous to the Kronig-Penney model in the electronic band-
structure problem. As an example, the photonic band equation is solved for the simple case of the Kronig-
Penney dielectric structure, consisting of alternating slabs of refractive indices n 1 and n 2 .

DOI: 10.1103/PhysRevB.68.045121 PACS numbers: 42.70.Qs

I. INTRODUCTION In the following, the dielectric constant (r ) is translation-

ally invariant along the y and z directions, varying only along
The idea of photonic crystals was demonstrated by the the x direction, in a periodic manner. See Fig. 1. Mean-
early experiments of Yablonovitch and Gmitter1 and the while, the magnetic permeability is taken to be uniform
flurry of theoretical works that followed.2 These crystals are throughout this paper, being equal to its vacuum value
microstructured materials in which the dielectric constant is (r) 0 , which is an excellent approximation for practical
periodically modulated with the result that the multiple inter- systems of interest.
ference of light scattered from different unit cells in the crys- Combining now the Maxwells equations, we get
tal produces a photonic band structure, where gaps may oc- the second-order differential equation for the displacement
cur. The existence of a photonic gap leads to a number of :
field D
physical properties, of both fundamental interest and for po-
tential device applications.
r / r 2 0 D
. 3
In this paper, we derive a transcendental equation for the
photonic band structure, akin to the solution of the Kronig- The dielectric can be broken down into segments of constant
Penney model arising in the electronic band-structure prob- as indicated in Fig. 1, so that the electric field E satisfies
lem. The equation, written in terms of the reflection and the equation in each segment,
transmission coefficients of the unit cell of the crystal, allows
us to derive several general properties of the photonic band 2n 2

2E ,
E 4
structure. The method presented here is an alternative proce- c2
dure to the plane-wave-expansion method, commonly used
to solve the photonic band-structure problem. Some general where c is the vacuum speed of light and n is the refractive
relations between the reflection and transmission coefficients index for the segment.
for the electromagnetic waves are also presented in the Ap-


The Maxwells equations for an electromagnetic em

wave propagating through a dielectric medium with no free
charge or current are given by

E i B ,

i D


.B 0,
FIG. 1. Variation of the refractive index n(x) for the one-
where the time dependence of the em field has been taken to dimensional photonic crystal and its segmentation into regions of
(r ,t)E (r )e i t and B (r ,t)B (r )e i t . Furthermore,
be E constant n shown in the lower part. The lower part of the figure
we also have shows the refractive index of one unit cell, attached to the
asymptotic regions of the constant refractive index n 0 outside the
r r E r .
D 2 cell. The quantities n(x) and n (x) are identical inside the unit cell.

0163-1829/2003/684/0451219/$20.00 68 045121-1 2003 The American Physical Society


III. PHOTONIC BAND EQUATION at the cell boundary. This will ensure that the Bloch sum of
A. Bloch periodicity
such functions is well behaved whenever n(x) is smoothly
In view of the fact that we have translational symmetry The tail may be chosen such that it satisfies the Maxwells
along directions parallel to the slab and Bloch periodicity equation, Eq. 8, for some n(x) in the tail region, but it is
along the direction perpendicular to the slab, the electric field not necessary that it does so since the tails will cancel out.
may be written as It is convenient to choose the tails of (x) to satisfy the
Maxwells equation with a constant n 0 n(xA/2), i.e., the
r e ik .r F x ,
E 5 same refractive index as its value at the boundaries of the
central cell. That way we can compute the transmission and
F m e ik mA xmA ,
B 6
reflection coefficients easily, which are needed later. The re-
fractive index n (x) that (x) sees in the entire space is
indicated in the lower part of Fig. 1, while the upper part of
where a Bloch form has been taken for F (x) by adding the the same figure indicates the refractive index n(x) seen by
functions (xmA), centered in the mth cell, after multiply- the function F (x). The refractive indices n(x) and n (x) are
ing them with the appropriate phase factors. Any function for the same in the central cell, differing from each other only
(x) is good enough as far as the Bloch symmetry is con- outside the cell.
cerned, but it has to be constructed such that F (x) satisfies To find a solution, (k ,k B ), let us try a specific guess
the Maxwells equations everywhere. Here, k B is the Bloch solution for the given values of k and k B . First we con-
momentum with /Ak B /A and A is the lattice con- struct the auxiliary function (x) for the central cell with
stant, x is normal to the slabs, and y and z are parallel to the
these k and from the differential equation, Eq. 8, in
slabs. The solution is thus characterized by the symmetry
terms of which the electric field F (x) for the entire periodic
labels (k ,k B ) and given a specific value of these, our goal is
structure will be given by the Bloch sum Eq. 6. This will be
to find . This would give us all possible solutions for the
a solution for the entire dielectric structure, provided that the
em fields photonic band structure.
Bloch sum of the tails of the auxiliary functions cancel ev-
Putting the expression for the electric field 5 into the
erywhere. In that case, the Bloch sum will satisfy the Max-
wave Eq. 4, one finds that the equation for the field F (x) is
wells equation everywhere, because only the heads of the
auxiliary functions survive, which by explicit construction

d 2F satisfy the Maxwells equation within the central cell. The
k x,i F x , 7
dx 2 condition that the tails will cancel in the central cell and
therefore in all other cells is that
where k x,i 2 n 2i /c 2 k 2 . Thus k is fixed, while k x,i varies
from segment to segment, according to the value n i corre-
sponding to the ith segment.
e ik B mA xmA 0 9

B. Tail cancellation
for all values of x in the central cell. The idea of the tail
We first obtain the solution of the Maxwells equation by cancellation is in fact well known from the Korringa-Kohn-
the method of tail cancellation, which although a bit te- Rostoker Greens function and the muffin-tin orbital formu-
dious in one dimension as compared to the method using the lations of the electronic band-structure theory.3
boundary condition matching, is quite powerful for higher- We now turn to the construction of (x), the explicit form
dimensional systems. The derivation below illustrates the of which is needed just for the tail, in order to apply the tail
method and can be generalized to two- and three- cancellation condition. Now, since the geometry of our prob-
dimensional photonic crystals. Similar generalization involv- lem is such that TE and TM modes do not mix,4 and further,
ing the tail cancellation is used in the formulation of the since in regions with constant (x), E 0 implies that the
muffin-tin orbitals method in the electronic band-structure E fields are transverse to the direction of propagation, we can
work with the two modes separately, with the mode index
The auxiliary functions (x) are to be determined such 1 and 2. For each of the modes, we have two solutions,
that the Bloch sum Eq. 6 satisfies Eq. 7. Let us choose it
in the following way: In the central cell defined as A/2 right and left propagating, respectively, which we call 1 (x)
xA/2), (x) satisfies Eq. 7, so that and 2 (x). Written in terms of the reflection and the trans-
mission coefficients, r and t , the four independent solu-
d 2 x tions for the auxiliary functions (x) satisfying Eq. 8 are
k x,i x , 8 then
dx 2

while outside the central cell, the tail of (x) is such that
the function is well behaved continuous and differentiable
1 x e ik x x r e ik x x e ,
t e ik x x
e ,


2 x t e ik x x e ,
e ik x x
r e ik x x
e ,


k 2x 2 /c 2 n 20 k 2 11
has the fixed value corresponding to the tail region, and the
unit vector e indicates the direction of the electric field
corresponding to the polarization . Note that since the val-
ues of and are same at the left and the right cell bound- FIG. 2. Sketch of the unit cell for obtaining the photonic band
aries for a periodic crystal, the transmission coefficients are equation using the boundary condition matching for the TE mode.
the same, while the reflection coefficients r and r differ by Electric-field coefficients (A and B) on the left and right sides of
a phase factor see Eq. A21. the cell are valid in the small shaded region, where the dielectric
constants may be taken to be unchanged.
Note also that the above expressions are only for the tails,
since, at this point, we dont really care how the auxiliary
electric fields look inside the central cell. The electric field Eliminating now the unknowns A and B from Eq. 13
may be obtained everywhere by integration of the Maxwells and after some algebra, we get the following transcendental
equations, Eq. 8, once the eigenfrequency has been de- equation for :

t 2 r r e ik x A
Since the TE and the TM modes dont mix, the most e ik x A cos k B A . 15
general solution for (x) is written as a linear combination of 2t 2t
the two independent solutions for each mode, Using Eq. A21 of the Appendix, we write the left and the
right transport coefficients in terms of their magnitudes and
x A 1 x B 2 x , 12 phases:
where, again, 1 and 2, corresponding to the TE and TM
t t t e i ,
modes, respectively. Substituting the expression for (x) in
the tail cancellation condition Eq. 9 and equating the coef- r r e i,
ficients of e ik x x in the resulting expression to be zero, we
obtain the conditions that the coefficients A and B must r r e i(2 ) . 16
Putting this in Eq. 15, we get
* B r S
A S t S * 0,
cos k x A
A r S B t S S
* 0, 13 cos k B A . 17

where S S(k B k x ) with S(k) m1 exp(iAmk). This is the central equation of the paper, the solution of
The sum S(k) is over a series of oscillating terms. The which gives the photonic band structure. It is very similar to
oscillation can be traced to the fact that the plane-wave-like the equation appearing in the Kronig-Penney model for the
tails in Eq. 10 continue undamped to infinity. If we keep a electronic case;5,6 however, unlike the electronic case, we
finite number of terms N in the summation, then the second now have two modes corresponding to the two polarizations
term in the numerator of the result S(k) e ikA of light, 1,2. The band structure is expressed here in
e ik(N1)A /(1e ikA ) oscillates rapidly between 1 and terms of the complex transmission coefficient t for the unit-
1 as N with the average value zero. It turns out that cell dielectric structure embedded in a uniform dielectric of
taking this average value yields the correct answer for the refractive index n 0 on either side. As we shall see later ex-
problem at hand. A more careful way of evaluating the sum plicitly, the dependence of n 0 drops out of the photonic band
is to take the limit equation as it must; however, it is needed to define the tran-
N mission coefficient.
e ikA
S k lim
N, 0 n1
e (ik )nA
1e ikA
, 14
C. Photonic band equation from boundary matching
where the limit has been taken in such a way that A1 and The photonic band equation can be obtained more simply
NA1. Physically this corresponds to a small damping by matching the boundary condition. Consider first the TE
term e x in the auxiliary functions, Eq. 10, such that the mode. This is fine since for a stratified medium the TE and
amplitudes of the plane-wave tails damp out at infinity but TM modes dont mix.4 Referring to Fig. 2, the electric fields
do not change appreciably over the length of a unit cell. The in the immediate neighborhood on the left and right sides of
same situation arises in the solution of the Kronig-Penney the unit cell may be expressed as
model by the tail cancellation method in the electronic prob-

lem as well.5 E L e ik r A L e ik x x B L e ik x x z,


E R e ik r s 11A L s 12B L e ik x x

s 21A L s 22B L e ik x x z, 18
where the right coefficients A R and B R have been written in
terms of A L and B L using the scattering matrix Appendix
A, and the wave vector k x is related to the frequency via
Eq. 11. In this formulation, the refractive index n 0 entering
in that equation is the refractive index at the cell boundary as
indicated in the shaded area in Fig. 2.
Now, the continuity conditions at the boundary, viz., that
E and H are continuous, together with the Maxwells
equation relating E to B and the Blochs theorem, lead to
the two equations:

R A e ik B A E L 0 ,

R /dx A e ik B A dE L /dx 0 .
dE 19
Substituting these into Eq. 18, we get the eigenvalue equa-

s 11e ik x A
s 21e ik x A
s 12e ik x A
s 22e ik x A

B L e
ik B A
BL . 20
FIG. 3. Plot of the function f ( )cos / t , for a slab dielec-
The Bloch factor e ik B A is therefore an eigenvalue of the 22 tric structure for the case of normal incidence k 0. As indicated
matrix,7,8 the structure of which is such that see the Appen- by the photonic band equation, Eq. 17, those values of are
dix if is an eigenvalue then so is *. And, since the trace allowed for which the function lies between 1. For other frequen-
is unchanged by diagonalization, one gets cies, there is no Bloch solutions so that forbidden gaps in the fre-
quency spectrum are formed, the first few of which have been
s 11e ik x A s 22e ik x A 2cos k B A . 21 shaded in the figure. Lower part of the figure shows an enlargement
in the lower-frequency region. Dielectric parameters are 13 for
Expressing now the elements of the S matrix appearing in the x 0.2A, and 1 for 0.2A x A in the unit cell of the crystal.
above equation in terms of the transport coefficients Eqs.
A9 and A17, viz., inc being the angle of incidence at a reference point say,
the edge of the unit cell, xA/2 with refractive index n 0 ).
s 11s 22
* 1/t * , 22
Since k is a constant throughout the structure, so is Ksin
we obtain the same photonic band equation as Eq. 17. For Eq. 24, from which it follows that nsin is also the same
the TM mode, the same equation may be derived by working throughout the system. This is nothing but Snells law, which
field and following similar arguments.
with the B is seen here to be a consequence of the translational symme-
The problem then boils down to solving for , given the
symmetry labels Bloch momentum k and k B , or,
A. Structure of the equation equivalently, given inc and k B . The solution of the tran-
The photonic band-structure Eq. 17 is of the form scendental Eq. 17 is illustrated in Fig. 3, which shows the
occurrence of gaps in the photonic spectrum.
f k ,k B , 0. 23
B. Computation of the transmission coefficient
Another auxiliary variable which is quite physical is the
angle of incidence defined as In order to solve the photonic band equation, Eq. 17, we
need to compute the transmission coefficient t . This is most
k simply computed by using the transfer-matrix method as out-
sin , 24 lined below, a method that is quite familiar in the treatment
of light transmission in layered media.7,911
where K n/c. Since the refractive index varies along x, so We first divide the cell into a large number N of segments
do the quantities K and . With this, the photonic band equa- along x such that the refractive index in each segment may
tion has the form be considered to be a constant. For each of the two modes
TE or TM, the em fields at the left and right end of the ith
f inc ,k B , 0, 25 segment are related via the 22 transfer matrix Mi ,


, 26
in terms of r and t, and a connection made with the barred
quantities. Following these lines, one finds that the barred
and the unbarred coefficients differ only by a phase factor:
with the numerical value,

t t e ik x A , rre ik x (2x 0 ) , 31
cos i isin i / i
Mi i sin where the unit cell extends between xx 0 and xx 0 A. In
i i cos i ,
terms of the barred transmission coefficients for the two po-
larizations ,
i d i k i TE or TM ,
t t e i ,


i k i /
n i /c 2 / k i /
TE ,
TM ,
the photonic gap Eq. 17 is then written as

with cos k B A . 33
k i n i /c 2 k 2 1/2, 27
While the photonic band equation Eq. 33 is quite physi-
where n i is the refractive index of the ith segment, d i is its cal, being analogous to the corresponding equation in the
thickness ( i1 d i A), and c is the vacuum speed of light. electronic case, and it provides the connection between the
The overall transfer matrix corresponding to the entire transmission coefficient and the photonic bands, it may be
unit cell, treated as a multisegment stack consisting of N simplified somewhat for practical calculations by writing it
segments, is the product of the individual transfer matrices: in terms of the transfer matrix M. We first note that the

diagonal elements of M are real, while the off-diagonal el-
m 11 m 12 N ements are purely imaginary, a form that originates from the
M m
21 m 22
lim Mi . 28 matrix form for the individual segment Mi and is retained as
i1 N we multiply such matrices to obtain the final M. Therefore,
The transmission and the reflection coefficients for the entire in the expression for M Eq. 28, m 11 and m 22 are real,
multilayer stack may be written in terms of the following while m 12 and m 21 are purely imaginary. Using this fact in
two quantities: evaluating the left part of the photonic band equation Eq.
33 by plugging in the transmission coefficient Eq. 29,
t 2 0 / 0 m 11 20 m 12m 21 0 m 22 , we find that the photonic band equation may alternatively be
written as
0 m 11 20 m 12m 21 0 m 22
r , 29 1/2TrMcos k B A . 34
0 m 11 20 m 12m 21 0 m 22
The equation may also be obtained directly from the bound-
where ary condition matching.10 The three equivalent equations for
the photonic band structure Eqs. 17, 33, and 34 are the
0 1/ n 0 /c 2 k 2 1/2 30 central equations of the paper.
corresponds to the tail region, n 0 being the refractive index It is clear from Eq. 34 that the band structure is inde-
pendent of the constant n 0 , since it never appears in the
at the two edges of the unit cell, xA/2. The quantities t
expression for M. Secondly, the band structure is indepen-
and r are in fact the transmission and reflection coefficients,
dent of the choice of the unit cell. This is simply because the
respectively, if we define them according to their values at
trace of a product of matrices is unchanged under cyclic
the left and the right boundaries see, e.g., Pedrotti and
permutation of the matrices, Tr(ABC)Tr(CAB), and in
Pedrotti9. We have, in contrast, defined t and r according to
our formulation, choosing a different unit cell amounts to
their asymptotic forms, Eq. 10.
multiplication of the same set of transfer matrices, without
Note that the transfer matrices depend on the parameters
altering the sequence, e.g., M1 M2 M3 M4 vs M2
of the slab only, without reference to the dielectric in which
M3 M4 M1 , if the unit cell is divided into four seg-
it is embedded, while the reflection and the transmission co-
efficients depend on the refractive index of the embedding
medium (n 0 ), the dependence entering via the quantity 0 in
Eq. 29. C. Weak scatterers
The barred and the unbarred coefficients (t, r, r, and t ) We consider now the limiting case of a weak scatterer and
are related quite simply. Since for the construction of the normal incidence (k 0). For a weak scatterer defined as
auxiliary functions (x) in Eq. 10, we have taken the di- t 1, r 0, and 0, the transcendental Eq. 17 shows
electric constant to be fixed (x) ( A/2 ) for x A/2 ], that the band gaps are centered around the wave vector
the asymptotic forms extend right up to the unit-cell bound- k 0 Am , m being an integer. Band states begin with nearby
ary. Thus the em fields at the cell boundaries can be written k x values of k k 0 , so that


cos k 0 A A t , 35

since at these values the right-hand side of Eq. 17 becomes

plus or minus one. Solving this, with the assumption that r
does not vary strongly in the gap region, one finds the value
for the gap to be

c k k /n 0 r, 36
n 0A

where r is the reflection coefficient for some average fre-

quency in the gap region. Thus the band gap scales linearly
with the magnitude of the reflection coefficient in the limit
that it is small.
In the case of the strong scatterers, the transmission coef-
ficient as a function of the frequency is small much of the
time except for sharp resonances. Photonic bands will be
narrow, forming in the region in which the resonances occur,
i.e., t( ) 1, with the bandwidth controlled by the sharp
variation of the complex t( ) with . Apart from these FIG. 4. Photonic band structure for the same slab dielectric
bands, only in rare cases will the cosine term in the numera- structure considered in Fig. 3 for both normal (k 0; left panel
tor in Eq. 17 conspire with a small t in the denominator to and slanted (k 0; right panel incidence. For the left panel, k
produce a ratio of magnitude less than unity to provide ad- 0 and the Bloch momentum k B varies along the x axis, while for
ditional bands away from the resonances. the right panel, we have k B 0 and varying k . For the slanted
incidence, the TE and the TM modes have different frequencies and
D. Numerical results they are indicated by solid and dashed lines, respectively.

We now turn to the calculation of the photonic band struc-

ture for a specific example. We consider a Kronig-Penney modes. However, for normal incidence k 0, the two equa-
slab dielectric, consisting of alternate slabs of dielectric con- tions become identical:
stants n 2 and n 1 , of thicknesses a and b, respectively, with
the unit-cell size Aab. The ratio cos / t appearing in
the photonic band equation Eq. 33 is easily obtained by
constructing and multiplying two M matrices following Sec.

n1 n2

n2 n1
sin 1 sin 2 cos 1 cos 2 cos k B A ,

IV B. 40
There are just two M matrices involved corresponding to
the two slabs, viz., so that for the case of normal incidence, TE and TM modes

are degenerate.
cos 1 isin 1 / 1 The photonic band structure obtained by solving the tran-
M1 i sin cos 1 , 37 scendental Eq. 39 is plotted in Fig. 4. For normal incidence
1 1
(k 0), the TE and the TM modes are degenerate, as might
be expected, while for slanted incidence (k 0), they are
not. The results shown in Fig. 4 agree completely with earlier

results for the same structure obtained by using the plane-
cos 2 isin 2 / 2
wave-expansion method.12
M2 i sin cos 2 , 38
2 2

which are obtained from Eq. 27. Multiplying the two ma-
trices and taking the trace of the resulting matrix MM1 In conclusion, we have obtained a transcendental equation
M2 , we get the photonic band equation for the photonic band structure for a one-dimensional photo-
nic crystal in terms of the reflection and the transmission


1 2
2 1
sin 1 sin 2 cos 1 cos 2 cos k B A .
coefficients, analogous to the solution of the Kronig-Penney
model in the electronic structure problem. The photonic band
39 equation allows us to extract several general features of the
photonic band structure and illustrates the origin of the band
The gammas and the deltas for the two slabs are defined in gaps. The case of the Kronig-Penney dielectric structure,
Eq. 27 and note that they are different for the TE and the with slabs of refractive index n 1 alternating with slabs of n 2 ,
TM modes, leading to two different equations for the two was solved as an illustration of the method.


e ik y y A L e ik x x B L e ik x x z, x,

e ik y y A R e ik x x B R e ik x x z, x,
E A3

where 2 n 21 /c 2 k 2x k 2y , 2 n 22 /c 2 k 2x k 2y , and we have

chosen the y direction along k fixed throughout the struc-
ture owing to the translational symmetry along directions
parallel to the slab and the direction of polarization of E to
be along z.
Since these expressions are asymptotic forms of the same
solution of a linear differential equation, there must be a
linear relation between the coefficients, which defines the
FIG. 5. Reflection and transmission coefficients for light inci-
scattering matrix:
dent from the left and the right. The coefficients are defined with

reference to the plane-wave solutions in the asymptotic regions x
AR s 11 s 12 AL
B R s 21 s 22 BL . A4

We acknowledge support of this work by the U. S. De- It is clear from Eq. A1 that if E is a solution, then so is E
partment of Energy Grant Nos. DE-FG02-00ER45818. In addition, it also follows from Eq. A1 that if E e iy
f (x)z is a solution, then so is the function E e iy
APPENDIX: REFLECTION AND TRANSMISSION f (x)z. The asymptotic forms, obtained by taking the com-
COEFFICIENTS plex conjugate of Eq. A3 and changing the sign of k y ,
In this Appendix, we outline the relationship between the
transmission and reflection coefficients for the electromag-
e ik y y A L* e ik x x B L* e ik x x z,
netic waves incident on a dielectric. Consider the dielectric
medium with the dielctric constant and the magnetic perme- e ik y y A R* e ik x x B R* e ik x x z,
E A5
ability varying along x: (x) and (x) varying from one set
of constants, 1 , 1 as x, to another, 2 , 2 as x differ from Eq. A3 only in the nomenclature of the constant
, as in Fig. 5. In the absence of any free charge or coefficients. We therefore have

current, the electric field obeys the Maxwell wave equation
in the asymptotic region (x): B R* s 11 s 12 B L*
A R* s 21 s 22 A L* . A6
2n 2

2E ,
E A1
c2 Taking the complex conjugate and comparing with Eq. A4,
we obtain the relations between the S-matrix elements:
where we have assumed a harmonic time dependence
(r ,t)E (r )e i t , with c the vacuum speed of light, and n
E s 11s 22
the refractive index, n c. Not just in the asymptotic
region, but in all space one can in fact write down the above s 12s 21
*. A7
equation for E , provided one divides the given (x), (x)
into piecewise segments of constants and .
2. TE mode: Reflection and transmission coefficients.
A similar wave equation for the magnetic field can be
written, viz., Meanwhile, let the reflection and transmission coefficients
for a wave incident from the left be denoted by r and t, while
2n 2 be let r and t denote the corresponding quantites for a

2B .
B A2
c2 wave incident from the right as indicated in Fig. 5. For the
wave incident from the left, the asymptotic form of the elec-
With our geometry, the TE and TM modes dont mix and it is tric field is given by
for the TE
convenient to work with the wave equation for E
e ik y y e ik x x re ik x x z,
mode and the wave equation for B for the TM mode.

1. TE mode: Asymptotic forms and the scattering matrix. E e ik y y te ik x x z. A8

Considering first the TE mode, the asymptotic form of the This is a special case of the general asymptotic solution A3
electric field is given, both for x and for x, by with A L 1, B L r, A R t, and B R 0. Putting these values
a sum of two plane-wave components into Eq. A4, the S matrix can be written as


1 t

1 r 2 t * r
tr *
t* 1/t *
r * /t *
r /t
1/t .

Substituting the right coefficients A R , B R in terms of
A L , B L , and the S matrix, we find

k x k x
s 2 s 12 2 , A16
The second equality comes from an alternative form of the S 1 2 11
matrix expressed in terms of r and t , by considering an em which in turn leads to the relation
wave incident from the right. Equating the two alternative
forms of the S matrix Eq. A9 element by element, we k x 1 2
obtain r2 t 1. A17
k x 2
1 r 2 tt * , Using the analogous equation for r and t , plus the fact that
r r which follows from Eq. A10, we find that
r * r
t *

. A10
k x 1
k x 2
t. A18

3. TE mode: Conservation of energy

4. TE mode: Relationship between the left and the right
Consider now the conservation of energy current. In a
transport coefficients: r,t and r ,t
lossless and source-free medium, the conservation of energy
implies that Consider em waves of equal amplitude incident from both
sides of the dielectric. Applying the energy conservation
ReS .dA
0, A11
A14, we find that

kx k x
where S 21 (E H
* ) is the Poynting vector and the integral 1 rt 2 r t 2 1 , A19
1 2
has been taken on a closed surface.
The Poynting vector can be easily computed for the which leads to the condition

asymptotic forms of the em fields. For example, for the E
field in Eq. A3, we have k x 2
Re rt * r t * 0. A20
k x 1
S ik x A L 2 B L 2 A L B L* e 2ik x x c.c.
2 The most general form of the transport coefficients satisfying
all relations between them is that
jk y A L 2 B L 2 A L B L* e 2ik x x c.c. . A12
r r e i ,
This form is valid everywhere and not just in the asymptotic
region, provided that we make the segmentation into regions t t e i ,
of constant and . Inspection of Eq. A12 shows that

ReS x r r e i(2 ) ,

ReS y
t k x 1
k x 2
te i. A21
0. A13
5. TE mode: Symmetric dielectric
Applying the conservation relation A11 to a rectangular
box, with faces normal to the Cartesian coordinates, and us- The above equations are valid for both a symmetric and
ing the relation A13, we find that nonsymmetric dielectric. Consider now the symmetric di-
electric, (x) (x) and (x) (x), on which an em
i ReS const A14 wave is incident from both the left and right with equal am-
plitudes. From symmetry, the energy current is zero every-
in all space. This is easily shown to be true for the TM mode where. Computing the current in the asymptotic region x
as well. , we obtain
The conservation relation A14 applied to the asymptotic
solution A3 and the resulting expression A12 for S leads i ReS
1 rt 2 0, A22
to the condition that 2 1
which leads to the condition that
kx k x
A L 2 B L 2 A R 2 B R 2 . A15
1 2 Re rt * 0, A23


i.e., rt * purely imaginary, where we have used the relations 6. TM mode

tt Eq. A21 and r 2 t 2 1 Eq. A17. Equation field.
For the TM mode, we choose to work with the B
A23 together with Eq. A21 leads to the result
The Maxwells wave equation is
rr i r e i ,
2n 2
2B .
B A26
tt t e , A24 c2
for the symmetric case. In addition, from Eq. A17, we also
have With B along z, this equation is exactly the same as the
corresponding Eq. A1 for the E field with E along z.
r 2 t 2 1, A25 Therefore all asymptotic properties of the differential equa-
Note that if the dielectric structure is symmetric but only tion are identical. Therefore, all relations between the trans-
about the point xx 0 i.e., (x 0 x) (xx 0 ) and simi- port coefficients r,t and r ,t derived for the TE case also
larly for ], then the em waves e ikx incident from the left hold true for the TM case. Note, however, that since E and B
and e ikx incident from the right do not produce a zero cur- satisfy different boundary conditions at the interface between
rent, because the two incident waves approach the structure two dielectric media, the numerical values of the reflection
with different phases. Therefore, Eqs. A24 and A25 are and transmission coefficients and the S matrix will be differ-
not valid in that case. ent in general.

E. Yablonovitch and T. J. Gmitter, Phys. Rev. Lett. 63, 1950 7
P. Yeh, A. Yariv, and C. S. Hong, J. Opt. Soc. Am. 67, 423 1977;
1989. P. Yeh, Optical Waves in Layered Media Wiley, New York,
Z. Zhang and S. Satpathy, Phys. Rev. Lett. 65, 2650 1990; K. M. 1988, Chap. 6.
Leung and Y. F. Liu, ibid. 65, 2646 1990; K. M. Ho, C. T. 8
See, for example, G. Grosso and G. P. Parravicini, Solid State
Chan, and C. M. Soukoulis, ibid., 65, 3152 1990. Physics Academic, London, 2000.
H. L. Skriver, The LMTO Method Springer, New York, 1983; O. 9
F. L. Pedrotti and L. S. Pedrotti, Introduction to Optics, 2nd ed.
K. Andersen, Phys. Rev. B 12, 3060 1975. Prentice-Hall, Englewood Cliffe, NJ, 1993, p. 395.
M. Born and E. Wolf, Principles of Optics, 7th ed. Cambridge 10
N. Liu, Phys. Rev. B 55, 4097 1997.
University, Cambridge, England, 1999, Sect. 1.5. 11
J. B. Pendry, J. Phys.: Condens. Matter 8, 1085 1996.
S. Mishra and S. Satpathy, Am. J. Phys. 69, 512 2001. 12
J. D. Joannapoulous, R. D. Meade, and J. N. Winn, Photonic
N. W. Ashcroft and N. D. Mermin, Solid State Physics Saunders,
Crystrals Princeton University, Princeton, NJ, 1995.
Philadelphia, 1976, p. 148.