Progress In Electromagnetics Research Symposium, Beijing, China, March 2327, 2009

197

Higher Order Finite-difference Frequency-domain Analysis of

Two-dimensional Photonic Crystals with Arbitrary Shapes
Yen-Chung Chiang
National Chung Hsing University, Taiwan

Abstract A higher order finite-difference frequency-domain (FDFD) method is proposed

for the analysis of the band diagrams of two-dimensional (2-D) photonic crystals (PCs). This
improved formulation is based on Taylor series expansion (TSE), local coordinate transformation
(LCT), boundary condition matching (BCM), and the generalized Douglas (GD) scheme. The
fourth-order convergence can be achieved with little additional computational cost compared
with that of a second-order scheme. This proposed scheme can deal with piecewise homogeneous
structures with slanted dielectric interfaces.
1. INTRODUCTION

Since Eli Yablonovitch [1] and Sajeev John [2] published two milestone papers on photonic crystals
(PCs) in 1987, these new categories of materials suggest that we might be able to control the
properties of light propagation just as we do in semiconductor devices. The PCs are characterized
by their photonic band structures. Due to the periodicity in the geometry, waves may be forbidden
to propagate in all directions for some frequencies, that is, the photonic band gaps. We may design
various devices based on this characteristic by properly introducing defects between the periodic
structures. Thus the calculation of the band diagram is a fundamental task for analyzing the PCs
and the design of devices.
Among the proposed methods for the calculation of the band diagrams of PCs, the most commonly used methods are the plane-wave expansion (PWE) method [3] and the finite-difference
time-domain (FDTD) method [4]. Aside from these two methods, Yang [5] proposed a finitedifference frequency-domain (FDFD) scheme for analyzing the band structures of 2-D PCs, and he
also demonstrated the sparsity of the resulting matrix and the efficiency of this method. But such
scheme is suitable only for irregular structures with interfaces parallel to the axes but not for those
with arbitrary shapes. To overcome the difficulty of adopting FDFD method in such structures
with slanted interfaces, we have developed a systematic procedure for deriving a suitable finite
difference scheme [6]. We will extend this method to higher order formulas with the help of the
generalized Douglas scheme.

( m 1,n +1)

( m ,n +1)

a
( m1,n )

( m ,n
)

( m +1,n +1)

( m+1,n )
n^
^t

y^

( m 1,n 1)
(a)

(b)

Figure 1: (a) Cross section of a 2-D PC with the square lattice, and
(b) the unit cell used in the calculation.

( m ,n 1)

( m+1,n 1)

x^
Figure 2: Cross section of the problem under consideration for a linearly slanted interface.

2. FORMULATION

The cross-section of a 2-D PC consisting of cylindrical dielectric rods is shown in Figure 1(a) for
the square lattice case, where a is the lattice distance, and r is the radius of the element rods.

PIERS Proceedings, Beijing, China, March 2327, 2009

198

When we calculate the band structures of the 2-D PC, we only need to calculate the problem over
a unit cell area, which is surrounded by dashed lines in Figure 1(a). The unit cell is re-drawn with
the meshes generally used in the FD method and corresponding periodic boundary conditions as
shown in Figure 1(b). The periodic boundary conditions PBC1 and PBC2 can be expressed as
PBC1: (x + a, y) = exp (jkx a) (x, y) ,
PBC2: (x, y + a) = exp (jky a) (x, y) ,

(1)
(2)

where kx and ky are the wavenumbers in the x and y directions, respectively.

Consider a linearly slanted step-index interface between two dielectric regions, with permittivities L and R , respectively, as shown in Figure 2. The basic procedure to derive our higher order
finite-difference scheme is to find the relation between two grid points and the steps are similar to
those provided in [6]. Referring to Figure 2 and using the point (m, n) as an example, we need first
express the fields at its neighbor points, for example the point (m+1, n+1), as the expansion of the
field and its derivatives at point (m, n). This can be done using steps illustrated in Equation (3):
TSE

LCT

BCM

LCT

TSE

|(m+1,n+1) |R;xy |R;nt |L;nt |L;xy |(m,n) .

(3)

Since the relations up to second order can be found in [6], we will not show them explicitly
here. To make it more concise, we also skip the relatively easier high order equations for Taylor
series expansion (TSE) and local coordinate transformation (LCT). As for the boundary condition
matching, we just list two third order equations as following:

3 Ez
3 Ez
Ez
2
,
(4)
=
+ k0 (L R )
n3 R
n3 L
n L
and

3 Hz
R 3 Hz
Hz
2
.
=
+ k0 (L R )
n3 R
L n3 L
n L

(5)

Other required higher order relations can easily be found from existing lower order boundary
conditions and proper combination of basic boundary conditions.
Once we obtain the relations between mesh points, we can express them in a 9-by-9 matrix
form. Performing the matrix inversion will reach the second order formulas provided in [6]. Then
we adopt the GD scheme introduced in [7] for treating the high order terms. Using second order
formula for Ez as an example, assume the formula obtained from the process provided in [6] can
be expressed as following with the residual third order terms
00

2 Ez
2 Ez
3 Ez
3 Ez
00
Dxx + Dyy
Ez =
+
+
g
+
g
,
1
2
x2
y 2
x3
y 3

(6)

00 and D 00 denote the second order finite difference formulas for differentiation with respect
where Dxx
yy
to x and y, respectively. After utilizing the GD scheme, we will obtain

2
2
+
x2 y 2

Ez =

00 + D 00 + g D 000 + g D 000
Dxx
2 xxy
1 xyy
yy
0
0
1 + g1 Dx + g2 Dy

Ez .

(7)

Of course, the final equation for calculating the band structures, that is, Equation (3) in [6] should
be modified as

2
[A]
[B] [] = 0.
(8)
c
where [] is the vector whose elements are the fields on the mesh nodes, [A] is the characteristic
matrix with entries related to the coefficients in the numerator part of Equation (7), and [B] is no
longer an identical matrix as in [6] but with entries related to the coefficients in the denominator
part of Equation (7).

Progress In Electromagnetics Research Symposium, Beijing, China, March 2327, 2009

199

3. NUMERICAL RESULTS

To validate our high order finite-difference scheme, we choose the above-mentioned 2-D PC with
square lattice as an example. The cross section of the PC is as shown in Figure 1(a), which is
formed by parallel alumina rods in the air. The relative permittivity of the alumina is Al = 8.9
and that of the air is air = 1. The radius of the alumina rod is r = 0.2a, where a is the lattice
distance. Each point along the boundary of the first Brillouin zone shown as the middle inset in
Figure 3(a) determines the values of kx and ky we use in Equations (1) and (2), and thus in (9).
1

TE mode

Normalized frequency (a/2c)

Normalized frequency (a/2c)

1
0.8
0.6

This method
PWE method

0.4
M

0.2

TM mode

0.8
0.6
This method
PWE method

0.4
0.2

(a)

(b)

Figure 3: The band diagram of the (a) TE modes and (b) TM modes for the 2-D PC.
10

6
TE first band

TM first band
4
Relative error (%)

Relative error (%)

0
This work
Method of [5]

-5

Method of [6]
-10
0

500

1000
1500
2000
Number of grid points

2
0
-2

This work

-4

Method of [5]
Method of [6]

-6
-8
0

2500

500

(a)

1000
1500
2000
Number of grid points

2500

(b)

5
TM second band

TE second band
0
Relative error (%)

Relative error (%)

0
-1
-2

This work
Method of [5]

-3

Method of [6]

-5

This work
Method of [5]
Method of [6]

-10

-4
-5
0

500

1000
1500
2000
Number of grid points

(c)

2500

-15
0

500

1000
1500
2000
Number of grid points

2500

(d)

Figure 4: The convergence properties of our method for the (a) TE first band, (b) TM first band, (c) TE
second band, and (d) TM second band compared with those of the methods in [5] and [6].

The calculated band diagrams of the TE and TM modes are plotted in Figures 3(a) and (b).
The band structures marked with red circles are the results obtained using our proposed finitedifference scheme with 30 30 grid points, and the solid blue lines are the results obtained using
the MIT Photonic-Bands (MPB) package [8] based on the PWE method with 128 128 resolution.

200

PIERS Proceedings, Beijing, China, March 2327, 2009

It can be shown that they match each other quite well for both polarizations. As indicated in [5],
the resultant matrix is quite sparse (quasi-band diagonal), thus it is easy to be implemented with
direct QR procedure and it also reduces the memory requirement.
To examine the efficiency of our new scheme, we fix the wavenumber vector k at the M point
in the first Brillouin zone as in the middle inset of Figure 3(a), i.e., k = (/a, /a, 0), and search
the eigen frequencies using different numbers of grid points. We also use the scheme provided
in [5] and [6] to calculate the convergence behavior under the same conditions for comparison. The
computed results are plotted as shown in Figures 4(a)(d). The relative error here refers to the
difference between the result and the reference result obtained with the grid points being 2500. It
shows that as the number of total grid points is larger than 1000, all four bands of our method
can achieve a relative error less than about 0.5%. Compared with the methods of [5] and [6], our
method can provide better convergence behavior.
4. CONCLUSIONS

We have derived a higher order finite-difference scheme for solving the band diagram of the 2-D
PCs that can deal with structures with arbitrary shapes. Our results can match very well with
those obtained by PWE method, and it is shown that our scheme can improve convergence behavior
significantly.
ACKNOWLEDGMENT

This work was supported in part by the National Science Council of the Republic of China under
Grant NSC97- 2221-E-005-091-MY2.
REFERENCES

1. Yablonovitch, E., Inhibited spontaneous emission in solid-state physics and electronics,

Physics Review Letters, Vol. 58, No. 20, 20592062, 1987.
2. Johm, S., Strong localization of photons in certain disordered dielectric superlattices, Physics
Review Letters, Vol. 58, No. 23, 24862489, 1987.
3. Ho, K. M., C. T. Chan, and C. M. Soukoulis, Existence of a photonic gap in periodic dielectric
structures, Physics Review Letters, Vol. 65, No. 25, 31523155, 1990.
4. Qiu, M. and S. He, A nonorthogonal finite-difference time-domain method for computing the
band structure of a two-dimensional photonic crystal with dielectric and metallic inclusions
Journal of Applied Physics, Vol. 87, No. 12, 82688275, 2000.
5. Yang, H. Y. D., Finite difference analysis of 2-D photonic crystals, IEEE Trans. Microw.
Theory Tech., Vol. 44, No. 12, 26882695, 1996.
6. Chiang, Y.-C., Y.-P. Chiou, and H.-C. Chang, Finite-difference frequency-domain analysis of
two-dimensional photonic crystals with curved dielectric interfaces, IEEE MTT-S International Microwave Symposium Digest, 19511954, Honolulu, Hawaii, USA, June 2007.
7. Chiou, Y.-P., Y.-C. Chiang, and H.-C. Chang, Improved three-point formulas considering
the interface conditions in the finite-difference analysis of step-index optical devices, IEEE
Journal of Lightwave Technology, Vol. 18, No. 2, 243251, 2000.
8. Johnson, S. G. and J. D. Joannopoulos, The MIT photonic-bands package home page [on
line], Available: http://ab-initio.mit.edu/mpb/.