May 10, 2006 / Vol. 4, No. 5 / CHINESE OPTICS LETTERS 285
A quasi-vector f inite difference mode solver for
optical waveguides with step-index profiles
Jinbiao Xiao (肖肖肖金金金标标标), Mingde Zhang (张张张明明明德德德), and Xiaohan Sun (孙孙孙小小小菡菡菡)
Lab of Photonics and Optical Communications, Department of Electronic Engineering, Southeast University, Nanjing 210096
Received May 8, 2005
A finite difference scheme based on the polynomial interpolation is constructed to solve the quasi-vector
equations for optical waveguides with step-index profiles. The discontinuities of the normal components of
the electric field across abrupt dielectric interfaces are taken into account. The numerical results include
the polarization effects, but the memory requirement is the same as in solving the scalar wave equation.
Moreover, the proposed finite difference scheme can be applied to both uniform and non-uniform mesh
grids. The modal propagation constants and field distributions for a buried rectangular waveguide and
a rib waveguide are presented. Solutions are compared favorably with those obtained by the numerical
approaches published earlier.
OCIS codes: 130.2790, 230.7370, 130.0130, 000.4430.
Mode solver for optical waveguides is a key issue in pho-
tonic device design. Only a few simple waveguide ge-
ometries, however, can be solved analytically, e.g., the
one-dimensional slab waveguide or the circular core op-
tical fiber. Therefore, the use of numerical (or approxi-
mate) analysis becomes necessary. To perform this task
with accuracy, various kinds of numerical techniques
[1,2]
,
such as finite difference method (FDM)
[3]
, finite ele-
ment method (FEM)
[4]
, and mapped Galerkin method
(MGM)
[5,6]
, have been proposed. Among them, FDM is
an attractive candidate because of the simplicity of its
implementation and the sparsity of its resulted matrix.
To date, FDM has been employed to solve both scalar
[7]
and vector
[8]
wave equations for optical waveguides with
step-index
[7,8]
even arbitrary-index
[3]
profiles. In the for-
mer case, it is valid only for optical waveguides with very
small refractive index difference (say, weakly-guiding
waveguides). In the latter case, the optical waveguides
with large refractive index difference in regions of high
field intensity can be analyzed, but there is a large in-
crease in computational time and memory
[1,2]
. Instead,
the mode solver based on quasi-vector wave equation is
a go od candidate for optical waveguides, in which the
polarization effects of the guided modes are considered,
while the memory requirement is the same as that in
solving the scalar wave equation. As a result, the com-
putational time is moderate. In addition, Taylor series
expansion (TSE) is often used to approximate the re-
sulted quasi-vector wave equation
[1,2]
. However, the TSE
is not easily adaptable to non-uniform grid sizes and ex-
tended to account for the discontinuities of the normal
components of the electric field across abrupt dielectric
interfaces.
In this letter, a modified finite difference scheme is con-
structed for quasi-vector analysis of optical waveguides
with step-index profiles. The polynomial interpolation
is employed to convert the quasi-vector wave equation
into the finite difference equation in which the disconti-
nuities above-mentioned are taken into account. More-
over, three adjacent grid points are used to approximate
each differential operator, so the solution is more accu-
rate than that uses two adjacent grid points. In addition,
the present scheme can be applied to both uniform and
non-uniform mesh grids.
The quasi-vector wave equation based on electric fields
derived from the Maxwell’s equations can be written as
[2]
A
xx
u
x
= β
2
u
x
, (1a)
A
yy
u
y
= β
2
u
y
, (1b)
with
A
xx
u
x
=
∂
∂x
·
1
n
2
∂(n
2
u
x
)
∂x
¸
+
∂
2
u
x
∂y
2
+ k
2
0
n
2
u
x
, (2a)
A
yy
u
y
=
∂
2
u
y
∂x
2
+
∂
∂y
·
1
n
2
∂(n
2
u
y
)
∂y
¸
+ k
2
0
n
2
u
y
, (2b)
where k
0
is the wave number in free space, n = n(x, y)
is the refractive index profile of the guiding medium,
β = k
0
n
eff
is the propagation constant and n
eff
is the
effective index. u
x
and u
y
are the electric components
in x- and y-direction, respectively. The above two equa-
tions are the scalar wave equations with polarization
correction which correspond to the quasi-transverse elec-
tric (TE) and the quasi-transverse magnetic (TM) wave
equations, respectively. This assumption is accurate for
three classes of waveguides
[1]
: 1) weakly guiding waveg-
uides with arbitrary shape and small difference in re-
fractive indices between core and cladding or substrate;
2) arbitrary refractive index profile waveguides with an
elongated or slab-like cross section; and 3) rectangular
core waveguides with arbitrary core-cladding refractive
index operated in the far-from cutoff region.
Figure 1 shows the finite difference mesh used in our
approach for a rib waveguide. The structure is scanned
with small rectangular sub-regions of size ∆x × ∆y. In-
side each sub-region the refractive index is constant, so
the discontinuities of the refractive index profile occur
only at the boundaries between adjacent sub-regions. In
1671-7694/2006/050285-04 http://www.col.org.cn