~ Optical and Quantum Electronics 34: 435~-43, 2002.

9 2002 Kluwer Academic Publishers. Printed in the Netherlands. 435

Wave-vector diagrams for two-dimensional photonic

crystals

OTTOKAR LEMINGER
T-Nova, Technologiezentrum, Am Kavalleriesand3, 64295 Darmstadt, Germany
(E-mail: [email protected])

Abstract. Photonic crystals exhibit band gaps, meaning that electromagnetic fields cannot propagate in
them for specific ranges of wavelengths and directions. The calculation of band structure diagrams has
been intensively studied and is now well understood. In contrast to that, so-called wave-vector diagrams
(i.e. dispersion surfaces, depicting the loci of all relevant wave vectors at a fixed wavelength) are less known
and used. In principle, they show how the effective index of the structure depends on the direction of
propagation. A method to calculate explicitly wave-vector diagrams for two-dimensional photonic crystals
is derived which leads finally to quadratic eigenvalue problems. Results for square and triangular lattices
are presented and some applications are discussed.

Key words: doubly periodic structures, photonic crystals, wave-vector diagrams

1. Introduction

Photonic crystals are currently the subject of intense theoretical and experi-
mental efforts (Kurizki and Haus 1994; Joannopoulos et al. 1995). They
exhibit band gaps, meaning that electromagnetic fields cannot propagate in
such crystals for specific ranges of wavelengths and directions. The calcula-
tion of photonic band structure diagrams is well understood and they appear
in almost everyone relevant paper.
In contrast to that, so-called wave-vector diagrams of photonic crystals (i.e.
dispersion surfaces, depicting the loci of all relevant wave vectors at a fixed
wavelength) are less known and used. They have been extensively studied for
singly periodic structures corresponding to one-dimensional (1D) photonic
crystals (Russell 1986; Russell et al. 1995) using a transfer matrix relating
the fields on one side of the investigated structure to those on the other.
Wave-vector diagrams for singly and doubly periodic structures were ob-
tained experimentally by Zengerle (1987). For two-dimensional (2D) or
three-dimensional (3D) photonic crystals the transfer-matrix approach was
implemented as a finite-element method (Pendry and MacKinnon 1992) but
only band structure diagrams and transmission coefficients of a square lattice
are presented there. Wave-vector diagrams for a 3D photonic crystal appear
(for example Kosaka et al. 1998; Kosaka et al. 1999; Notomi 2000) without
details of calculation. Presumably they were interpolated from the calculated
band-structure diagram data mentioned there.
436 o. LEMINGER
Wave-vector diagrams illustrate, in principle, how the effective index of the
structure depends on the direction of propagation. It can be shown (Yeh
1979) that the direction of the phase velocity is identical with that of the wave
vector but the direction of the group velocity (and of the energy flow) is
normal to the dispersion surfaces in wave-vector diagrams. This spatial an-
isotropy is the basis of many interesting effects (beam steering, self-colli-
mating, ultra-refractive optics).
A method, to a large extent analytical, to calculate explicitly wave-vector
diagrams for 2D photonic crystals is derived which uses a plane wave ex-
pansion and leads finally to a quadratic eigenvalue problem. Results for square
and triangular lattices are presented and some applications are discussed.

2. Wave propagation in 2D photonic crystals

A 2D photonic crystal is periodic along two directions in the xy plane and

homogeneous in the z direction. In this paper, specimens are investigated
consisting of a square or triangular lattice of dielectric columns with circular
cross section in air (see Fig. 1). Only in-plane propagation (i.e. in the xy
plane) is considered.
The wave equations for the z components of the field are (with the wave
number k = 2~/2):

Fig. 1. A 2D photonic crystal consisting of dielectric columns with circular cross-section in air. The insets
show a cut in the xy plane through a (a) square lattice (b) triangular lattice.
2D PHOTONIC CRYSTALS WAVE-VECTOR DIAGRAMS 437

For TM modes: O2E~ ~2E~-}- k2rt2(x, y)Ez

aX2 -}- -~y2 = 0 (la)

For TE modes: (n_2(x,a<)

y) = 0

(lb)

For the field pattern we take the Floquet-Bloch expansion

--oo +oo
Ez(or Hz)= eiav Z ~ Cm,nexp(i(mb, + rib2)" T) (2)
m=--oG/'t~ cx)

with F = ( x , y ) , fi = (fix, fly) and the reciprocal lattice vectors ~31, [)2.
An analogous plane wave expansion for the periodic refractive-index dis-
tribution
q-oc +oc
"• = Z Z fm,n exp (i(mbl + r/[32)" r') (3)
g / / ~ - - OO n ~ - - O O

can be calculated analytically. In doing this, no and ni denote the refractive

indices of air and the columns, respectively, r the radius of the columns, a the
lattice constant and f~ = 2rr/a. For the square lattice we have bi = f~G,
~32 • ~"~e'yand obtain

f0,0 = r/~-1- (r/~ -- n2)7~ ( r ) 2

fm.-- n2--n~ r d l ( 2 r c ~ s (4)

V#~ q- n 2 a ', a/

For the triangular lattice it is t~l,2 = ~'-~(ex :~ (1/V/3)~) and we get

2" 2re 2
joo:. +<

_ 2 n2-n 2 J1 2rc (m+n) 2 + ~ ( m - (5)

These formulas hold for TM modes. For TE modes it is sufficient to

replace the refractive indices in them by their reciprocal values.
438 o. LEMINGER
3. Eigenvalue equations

Let us for example start with the TM modes of a square lattice. Inserting the
expansions (2) and (3) into Equation (la), one gets after lengthy algebra the
equations

[(fix "at-m~"~)2 -~" (fly -'~ n~'~)2]r = k2 E Z fm-i'n-jci'j

i j
m,n . . . . - 2 , - 1 , 0 , 1,2,3,... (6)

They can be processed in two different ways. In band-structure calculations,

we take a particular normalised vector k = (kx, ky) = (flx/ft, fly/~) and get
from Equation (6)

[(kx + m) 2 + (ky + n)2]Cm,n = (coa/27te)2~-~, Zfm_i,n_jCi, j (7)

i j
This is a generalised symmetric eigenvalue problem for the dimensionless
frequency. If this eigenvalue calculation is performed for all values of k in the
Brillouin zone, we get the well-known band-structure diagrams.
In wave-vector diagram calculations, we take at a chosen fixed wavelength
2 (or frequency co) aparticular direction of the vector fl, characterised by an
angle q~. It i s fix = Ill cos q~, fly = Ifilsin q~, and with the normalised param-
eters A = Ill~k, q = 2/a we get from Equation (6)

[A2 + 2A(mcos q~+ n sin ~p)q + (m2 + n2)q2]em,n = Z E fm-i,n-jCi,j (8)

i j

This is a quadratic eigenvalue problem for A, the normalised length of the

vector ft. If this eigenvalue calculation is performed for all dii'ections q~ and
we plot in polar coordinates p = Ifll(~p), we get the wave-vector diagram at
the wavelength 2.
The quadratic eigenvalue equation (8) holds for TM modes of a square
lattice. The other cases have been treated analogously with the following
results:
TE modes of a square lattice

Cm,n : E Z {A2 + A[(m+i)cosq~+(n+j)sinq~]q

i j
+ (mi + nj)q2}fm_i,,_jci,j (9)

TM modes of a triangular lattice

2D PHOTONIC CRYSTALS WAVE-VECTOR DIAGRAMS 439

{A2 + 2A[(m + n)cos q~ + (m - n) sin ~p/x/3]q + [(m + n) 2

+ (m - n)2/3]q2}Cm,n = Z Z fm-i'n-jCi'j (10)
i j

TE modes of a triangular lattice

Cm,n = ZZ{A 2+ A[(m+n+i+j)cos~o+(m-n+i-j)sin~p/x/3]q

i j

+ [(m + n)(i + j) + (m - n)(i - j)/3]q2}fm_i,n_jCi,j (11)

4. Calculation of wave-vector diagrams

For the quadratic eigenvalue problems in Equations (8)-(11) there exist no

implemented routines in commercial software packages. Fortunately, they
can be turned into linear problems using the expansion of Gfinther (1876) see
e.g. Press et al. (1992). The quadratic eigenvalue problem (AZA + AB+
~7) 9~ = 0, where ~], /~, ~7 are n x n matrices, is equivalent to introducing
an additional unknown vector S and solving the 2n x 2n generalised linear
eigenvalue problem

The Gfinther expansion has, however, two disadvantages: the resulting linear
eigenvalue problem is non-symmetric (hence pairs of complex conjugated
eigenvalues appear) and its size doubles.
To solve numerically the eigenvalue problems in Equations (8)-(11), which
are infinite-dimensional, they must be truncated. If we take M plane waves in
the expansions (3) and (4) along each reciprocal-lattice vector direction, we
get finally non-symmetric L x L matrices with a size L = 2(2M + 1)2. In order
to calculate the eigenvalues, for convenience's sake standard routines from
the NAG Fortran Library were used. There exist of course more suitable
methods for computing only a few eigenvalues of large matrices (see e.g.
Johnson and Joannopoulos (2001) and the references there) and in future
work such iterative eigensolvers will be used.
Large discontinuities in the refractive-index distribution at the boundaries
between the dielectric columns and air may cause poor convergence of the
plane wave expansion and the use of appropriate smoothing methods can
accelerate the convergence (e.g. Meade et al. 1993; Johnson and Joanno-
poulos 2001). I have computed only the eigenvalues and not the eigenvectors
belonging to them, and test calculations show that choosing M - - 6 (i.e.
L = 338) yields a sufficient accuracy.
440 o. LEMINGER

1.65
1.61

1.55
1 .so-"

Fig. 2. Wave-vector diagrams for T M modes in the Brillouin zone of a square lattice (a = 0.8 pm,
r = 0.16 [am, no = 1, nl = 3.2) at the wavelengths 1.45, 1.50, 1.55, 1.61, 1.65 [am.

In this way, a great number of wave-vector diagrams has been calculated. Of

course, we need only plot them within the corresponding Brillouin zone. Some
characteristic examples are shown for a square lattice in Figs. 2 and 3, and for
a triangular lattice in Figs. 4 and 5. In each figure, several wave-vector curves
are plotted for the same structure at different wavelengths. They can be in-
terpreted as horizontal cuts through band-structure surfaces at different levels.
To my knowledge, the only other published method for the explicit calcu-
lation of wave-vector diagrams for 2D photonic crystals (Pendry and MacK-
innon 1992) uses transfer matrices relating the fields on the sides of the unit
cell of the investigated periodic structure and a finite-element discretisation. It
has the advantage of being able to employ unstructured meshes with non-
uniform resolution. For relatively simple lattice structures, however, it is al-
ways possible to calculate the expansion coefficients in Equation (3) analyti-
cally and thus considerably reduce the computational complexity.

5. Application

It is possible to get directly from the wave-vector diagrams some qualitative

notions about the directional dependence of light propagation at different
2D P H O T O N I C C R Y S T A L S W A V E - V E C T O R D I A G R A M S 441

Fig. 3. Wave-vector diagrams for TE modes in the Brillouin zone of a square lattice (a = 0.8 ~tm,
r = 0.16 ~m, no = 1, nl = 3.2) at the wavelengths 1.80, 1.90, 2.20, 2.60, 3.30 ~m.

wavelengths. For example, in Fig. 2 (square lattice, TM modes) there are

curves plotted for five wavelengths between 1.45 and 1.65 ~tm. We see that
along the directions 0 ~ or 90 ~ propagation always occurs, but along the
direction 45 ~ only up to 1.61 ~tm. At the wavelength 1.65 ~tm along the di-
rection 45 ~ no propagation should be possible (band-gap).
I have checked these predictions with a numerical method based on rig-
orous scattering theory (Felbacq et al. 1994) and calculated the transmission
of a normally incident TM Gaussian beam through a slab of a square-lattice
photonic crystal with the same material parameters as those in Fig. 2. When
the slab was cut along the x axis (direction 0~ transmission occurs at both
1.45 and 1.65 lam wavelengths. In contrast to that, a slab cut along the
direction 45 ~ acts as a wavelength filter: transmission occurs at 1.45 lain but
the beam with a wavelength of 1.65 ~tm is totally reflected and has zero
transmission. Thus the qualitative predictions from the wave-vector diagram
have been verified numerically.
More interesting applications of wave-vector diagrams for explaining su-
perprism and self-collimating phenomena are shown by Kosaka et al. (1998,
1999).
442 o. LEMINGER

Fig. 4. Wave-vector d i a g r a m s for T M m o d e s in the Brillouin zone of a triangular lattice (a = 0.8 gm,
r = 0.2 ~tm, no = 1, nl = 2.6) at the wavelengths 1.45, 1.50, 1.55, 1.60, 1.65, 1.70 gm.

Fig. 5. Wave-vector d i a g r a m s for T E m o d e s in the Brillouin zone of a triangular lattice (a = 0.8 p.m,
r = 0.2 lam, no = 1, nl = 2.6) at the wavelengths 1.0, 1.1, 1.2, 1.3, 1.4 pan.
2D PHOTONIC CRYSTALS WAVE-VECTOR DIAGRAMS 443

6. Conclusion

Wave-vector diagrams describe visually the properties of photonic crystals at

a fixed wavelength in dependence on the direction of propagation. A method
of their explicit calculation for 2D photonic crystals is presented which leads
to quadratic eigenvalue problems. As well as the transfer matrix method it
has the ability to easily use frequency-dependent dielectrics, without self-
consistent schemes or perturbation theory. A generalisation to 3D photonic
crystals ought to incorporate the additional problem of maintaining the
transversality of the fields.
From wave-vector diagrams quick qualitative predictions about the
propagation of light beams are possible.

References

Felbacq, D., G. Tayeb and D. Maystre. J. Opt. Soc. Am. A 11 2526, 1994.
Gfinther, S. Zeitschriftf~ir Mathematik 21 185, 1876.
Joannopoulos, J.D., R.D. Meade and J.N. Winn. Photonic crystals, Princeton University Press, Princeton,
1995.
Johnson S.G. and J.D. Joannopoulos. Opt. Express 8 173, 2001.
Kosaka H., T. Kawashima, A. Tomita, M. Notomi, T. Tamamura, T. Sato and S. Kawakami. Phys. Rev.
B 58 R10096, 1998.
Kosaka H., T. Kawashima, A. Tomita, M. Notomi, T. Tamamura, T. Sato and S. Kawakami. Appl. Phys.
Lett. 74 1212, 1999.
Kurizki G. and J.W. Haus. J. Mod. Optics (Special issue) 41, 1994.
Meade R.D., A. Rappe, K. Brommer, J. Joannopoulos and O, Alerhand. Phys. Rev. B 48 8434, 1993.
Notomi M. Phys. Rev. B 62 10 696, 2000.
Pendry J.B. and A. MacKinnon. Phys. Rev. Lett. 69 2772, 1992.
Press W.H., S. Teukolsky, W. Vetterling and B. Flannery. Numerical Recipes in FORTRAN, Cambridge
University Press, New York, 455, 1992.
Russell P.St.J., Phys. Rev. A 33 3232, 1986.
Russell P.St.J., T. Birks and F. Lloyd-Lucas. Cor~fined Electrons and Photons, Plenum Press, NewYork,
585, 1995.
Yeh, P. J. Opt. Soc. Am. 69 742, 1979.
Zengerle, R. J. Mod. Optics 34 1589, 1987.