Académique Documents
Professionnel Documents
Culture Documents
Abstract
The finite element method has become a preeminent simulation technique in electromagnetics. For problems involving anisotropic media and metamaterials, proper algorithms should be
developed. It has been proved that discretizing in quadratic triangular elements may lead to an
improved accuracy. Here we present a collection of elemental matrices evaluated analytically for
quadratic triangular elements. They could be useful for the finite element method in advanced
electromagnetics.
I.
INTRODUCTION
The finite element method is a numerical technique for obtaining approximate solutions to
boundary-value problems of mathematical physics. The method was developed and applied
extensively for the analysis of electromagnetic problems [1, 2, 3, 4, 5].High-order vector finite
elements have been developed which make it possible to obtain highly accurate and efficient
solutions of vector wave equations [2]. When applied to problems involving anisotropic
media and metamaterials, since the most of available commercial packages cannot be applied,
proper algorithms should be developed [6]. An improved accuracy results with discretization
in quadratic triangular elements [2]. The evaluation of elemental matrices, although is a
simple algebra, is rather cumbersome. Here we present a collection of elemental matrices
evaluated analytically for quadratic triangular elements.
II.
BASIC RELATIONS
We consider an optical medium with an arbitrary cross section in the xy plane. With
a time dependence of the form exp(jt), where is the angular frequency, from Maxwells
equations the following vectorial wave function is derived
([p] ) k02 [q] = 0
(1)
where k0 is the free-space wavenumber, [p], [q] are related to the permittivity and permeability tensors, and denotes either the electric E or the magnetic H field. Dividing the cross
section into quadratic triangular elements, as shown in Fig.1, we expand the transverse
components x , y and the axial components z in each element as [2]
T
x [U ] [t ]e
= y = [V ]T [t ]e
T
z
j[N ] [z ]e
(2)
where [t ]e is the transverse tangential field, [z ]e is the longitudinal nodal field of each
element, [U ], [V ], [N ] are shape function vectors, and T denotes a transpose. The shape
function [N ] has six components [N ] = [N1 N2 N3 N4 N5 N6 ]T which are expressed in terms of
area coordinates L1 , L2 , L3 ,
N1 = L1 (2L1 1)
N2 = L2 (2L2 1)
N3 = L3 (2L3 1)
N4 = 4L1 L2
N5 = 4L2 L3
N6 = 4L3 L1
The area coordinates Lk (k = 1, 2, 3) are given by
L
a
b
c
1 1 1
1
1
L2 =
a2 b2 c2
2Ae
a3 b 3 c 3
L3
(3)
(4)
(5)
where xk , yk are the Cartesian coordinates of the corner points 1 to 3 of the triangle, and
the subscripts k, l, m progress cyclically around the three corners. The coefficients ak , bk , ck
are given by
ak = x l y m x m y l
bk = yl ym
ck = x m x l
2
(6)
l1 b2 L1
lbL
2 3 2
l
b
L
1
3 1 3
[U ] =
2Ae l1 b1 L2
l2 b2 L3
lcL
2 3 2
l
c
L
1
3 1 3
[V ] =
2Ae l1 c1 L2
l2 c2 L3
l3 b3 L1
with |lk | =
l1 c2 L1
(7)
l3 c3 L1
p
b2m + c2m where the subscripts k, l, m progress cyclically around the three corners
of the triangle, but lk includes a sign which is defined differently in [2, 5, 7]. When vector
shape functions such as those given above are employed to represent a vector field in a
finite element solution of vector wave equation, it is necessary to consider evaluation of the
associated elemental matrices. These integrals can be evaluated analytically for triangular
elements. In case of simple triangular elements, analytical relations for elemental matrices
can be found in literature [1]. For quadratic triangular elements the integration is generally
more involved. Analytical relations presented in [2] are rather complex. Different forms
of elemental matrices could be useful in order to check their correctness when they are
implemented in a developed algorithm. In the following we present a collection of elemental
matrices for quadratic triangular elements.
III.
ELEMENTS
We used the following integration formula for the area coordinates [3]
ZZ
i!j!k!
Li1 Lj2 Lk3 dxdy =
2Ae
(i, j, k = 0, 1, 2, 3, . . . )
(i + j + k + 2)!
e
(8)
c1 + c2 + c3 = 0
(9)
RR
e
in [2, 3, 4]
6 1 1 0 4 0
ZZ
Ae
1
T
[N ][N ] dxdy =
180 0
e
0 4
1 6 4 0 0
0 4 32 16 16
0 0 16 32 16
6 1 0
4 0
e
3b21
4
16 16 32
b1 b2
4
3b22
4
b1 b3
4
b1 b2
b1 b3
b1 b3
2b2 b3
b1 b3
2b1 b2
b1 b3
2b2 b3
2b1 b2
2
2
2(b1 + b3 + b1 b3 )
c1 c3
b1 b2
b2 b3
b1 b2
b2 b 3
4
4
2
3b
b1 b3 b2 b3
3
0
b2 b 3
4
4
4
b1 b2 b1 b2
0 2(b21 + b22 + b1 b2 )
2b1 b3
0 b2 b3 b2 b3
2b1 b3
2(b22 + b23 + b2 b3 )
ZZ
(10)
1
[N ] [N ]T
dxdy =
x x
3Ae
ZZ
(11)
[N ] [N ]T
1
dxdy =
y y
3Ae
3c21
4
c1 c2
4
3c22
4
c1 c3
4
c1 c2
c1 c2
c2 c3
c1 c2
c2 c3
4
4
2
c1 c3 c2 c3 3c3
0
c2 c3
4
4
4
c1 c2 c1 c2
0 2(c21 + c22 + c1 c2 )
2c1 c3
0 c2 c3 c2 c3
2c1 c3
2(c22 + c23 + c2 c3 )
c1 c3
c1 c3
2c2 c3
2c1 c2
c1 c3
2c2 c3
2c1 c2
2
2
2(c1 + c3 + c1 c3 )
0
(12)
Note that Eqs.(11) and (12) are given also in [2, 3, 4], but here they are simplified by using
Eq.(9).
ZZ
e
[N ] [N ]T
1
dxdy =
y x
3Ae
b2 c1
4
3b1 c1
4
b3 c1
4
b2 c 1
0
b1 c2 3b2 c2 b3 c2
b1 c 2
b3 c 2
4
4
4
b1 c3 b2 c3 3b3 c3
0
b2 c 3
4
4
4
b1 c 2 b2 c 1
0 b1 c 1 + b2 c 2 + b3 c 3
b3 c 1 + b1 c 3
0 b2 c 3 b3 c 2
b 1 c 3 + b3 c 1
b1 c 1 + b2 c 2 + b3 c 3
b1 c 3
b3 c 1
b2 c3 + b3 c2
b2 c 1 + b1 c 2
2b1 b2 b3 b1 + 2b2
b
1
ZZ
1
[N ]T
b1
dxdy =
[N ]
x
30 3b1
e
b1
b3 c 1
b1 c 3
b3 c 2 + b2 c 3
b1 c 2 + b2 c 1
b1 c 1 + b 2 c 2 + b3 c 3
b2 b3
b1 + 2b3
b1 b3
2b2 b3 b2 + 2b1
(13)
b2 + 2b3
(14)
2c1 c2 c3 c1 + 2c2
c
1
ZZ
[N ]T
1
c1
[N ]
dxdy =
y
30 3c1
e
c1
3c1
ZZ
[U ]
e
c2 c3
c1 + 2c3
c1 c3
c2 + 2c3
(15)
[N ]T
1
dxdy =
x
12Ae
l1 b1 b2
0
l2 b2 b3
0
0
0
l3 b1 b3
0
l1 b1 b2
0
0
0
l2 b2 b3
l3 b1 b3
0
0
l1 b2 (b1 + 2b2 )
l1 b2 (b2 + b3 )
l1 b2 (b1 + 2b3 )
l2 b3 (b1 + b3 )
l3 b1 (b1 + b2 )
l3 b1 (b3 + 2b2 ) l3 b1 (b3 + 2b1 )
l2 b3 (b2 + 2b1 )
l2 b3 (b2 + 2b3 )
(16)
ZZ
[V ]
e
[N ]T
1
dxdy =
x
12Ae
l1 b1 c2
l1 c2 (b1 + 2b2 )
0
l2 b2 c3
0
0
0
l3 b3 c1
0
l1 b2 c1
0
0
0
l2 b3 c2
l3 b1 c3
0
0
ZZ
l2 c3 (b2 + 2b1 )
l2 c3 (b2 + 2b3 )
l3 c1 (b1 + b2 )
l3 c1 (b3 + 2b2 )
l1 c2 (b1 + 2b3 )
l2 c3 (b1 + b3 )
l3 c1 (b3 + 2b1 )
l1 c1 (b1 + b3 )
l2 c2 (b3 + 2b1 )
(17)
1
[N ]T
dxdy =
[U ]
y
12Ae
e
l1 b2 c1
l1 b2 (c1 + 2c2 )
l1 b2 (c2 + c3 )
l1 b2 (c1 + 2c3 )
l2 b3 (c1 + c3 )
0
l3 b1 c3
l3 b1 (c1 + c2 )
l3 b1 (c3 + 2c2 ) l3 b1 (c3 + 2c1 )
l1 b1 c2
0
l1 b1 (c2 + 2c1 ) l1 b1 (c2 + 2c3 ) l1 b1 (c1 + c3 )
0
l2 b2 c3 l2 b2 (c1 + c2 ) l2 b2 (c3 + 2c2 ) l2 b2 (c3 + 2c1 )
l2 b3 c2
0
0
0
l3 b3 c1
ZZ
[V ]
e
l1 c2 (b2 + b3 )
l2 b3 (c2 + 2c1 )
(18)
[N ]T
1
dxdy =
y
12Ae
l1 c1 c2
l1 c2 (c1 + 2c2 )
0
l2 c2 c3
0
0
0
l3 c1 c3
0
l1 c1 c2
0
0
0
l2 c2 c3
l3 c1 c3
0
0
ZZ
l2 b3 (c2 + 2c3 )
[U ][U ]T dxdy =
e
l1 c2 (c2 + c3 )
l1 c2 (c1 + 2c3 )
l2 c3 (c1 + c3 )
l3 c1 (c1 + c2 )
l3 c1 (c3 + 2c2 ) l3 c1 (c3 + 2c1 )
l2 c3 (c2 + 2c1 )
l2 c3 (c2 + 2c3 )
(19)
48Ae
2l12 b22
l1 l2 b2 b3
l1 l2 b2 b3
l1 l3 b1 b2
l12 b1 b2
l1 l2 b22
2l22 b23
l2 l3 b1 b3
2l1 l2 b1 b3
l22 b2 b3
l1 l3 b1 b2
l2 l3 b1 b3
2l32 b21
l1 l3 b21
2l2 l3 b1 b2
l12 b1 b2
2l1 l2 b1 b3
l1 l3 b21
2l12 b21
l1 l2 b1 b2
l1 l2 b22
l22 b2 b3
2l2 l3 b1 b2
l1 l2 b1 b2
2l22 b22
2l1 l3 b2 b3
l2 l3 b23
l32 b1 b3
l1 l3 b1 b3
l2 l3 b2 b3
2l1 l3 b2 b3
l2 l3 b23
l32 b1 b3
l1 l3 b1 b3
l2 l3 b2 b3
2 2
2l3 b3
(20)
ZZ
[V ][V ]T dxdy =
e
2l12 c22
48Ae
l1 l2 c2 c3
l1 l3 c1 c2
l12 c1 c2
l1 l2 c22
2l1 l3 c2 c3
llcc
2l22 c23
l2 l3 c1 c3 2l1 l2 c1 c3 l22 c2 c3
12 2 3
llcc
l1 l3 c21 2l2 l3 c1 c2
l2 l3 c1 c3
2l32 c21
13 1 2
l2 c1 c2 2l1 l2 c1 c3 l1 l3 c2
2l12 c21
l1 l2 c1 c2
1
1
l1 l2 c2
2l22 c22
l22 c2 c3 2l2 l3 c1 c2 l1 l2 c1 c2
2
2l1 l3 c2 c3 l2 l3 c23
l32 c1 c3
l1 l3 c1 c3
l2 l3 c2 c3
ZZ
[U ][V ]T dxdy =
e
2l12 b2 c2
l2 l3 c23
2
l3 c1 c3
l1 l3 c1 c3
l2 l3 c2 c3
2l32 c23
48Ae
l1 l2 b2 c3
l1 l3 b2 c1
l12 b2 c1
l1 l2 b2 c2 2l1 l3 b2 c3
llbc
2
2
l2 l3 b3 c1 2l1 l2 b3 c1 l2 b3 c2 l2 l3 b3 c3
2l2 b3 c3
12 3 2
llbc
l1 l3 b1 c1 2l2 l3 b1 c2 l32 b1 c3
l2 l3 b1 c3
2l32 b1 c1
13 1 2
l2 b1 c2 2l1 l2 b1 c3 l1 l3 b1 c1
2l
b
c
l
l
b
c
l
l
b
c
1
1
1
2
1
2
1
3
1
3
1
1
2
l1 l2 b2 c2 l2 b2 c3 2l2 l3 b2 c1 l1 l2 b2 c1
l2 l3 b2 c3
2l2 b2 c2
2
2
2
2l1 l3 b3 c2 l2 l3 b3 c3 l3 b3 c1
l1 l3 b3 c1
l2 l3 b3 c2
2l3 b3 c3
a11
a
12
ZZ
T
[U ] [U ]
1
a13
dxdy =
y
16A3e a14
e y
a15
a16
a11 = l12 b22 c21 ,
a12 = l1 l2 b2 b3 c1 c2 ,
a14 = l12 b1 b2 c1 c2 ,
a22 = l22 b23 c22 ,
(21)
a13 = l1 l3 b1 b2 c1 c3 ,
a16 = l1 l3 b2 b3 c21 ,
a24 = l1 l2 b1 b3 c22 ,
a25 = l22 b2 b3 c2 c3 ,
a26 = l2 l3 b23 c1 c2 ,
a34 = l1 l3 b21 c2 c3 ,
a35 = l2 l3 b1 b2 c23 ,
a36 = l32 b1 b3 c1 c3 ,
a45 = l1 l2 b1 b2 c2 c3 ,
a46 = l1 l3 b1 b3 c1 c2 ,
a56 = l2 l3 b2 b3 c1 c3 ,
(23)
a15 = l1 l2 b22 c1 c3 ,
a23 = l2 l3 b1 b3 c2 c3 ,
(22)
(24)
The matrix
[V ] [V ]T
x
e x
RR
a12 = l1 l2 c2 c3 b1 b2 ,
a14 = l12 c1 c2 b1 b2 ,
a22 = l22 c23 b22 ,
a13 = l1 l3 c1 c2 b1 b3 ,
a16 = l1 l3 c2 c3 b21 ,
a15 = l1 l2 c22 b1 b3 ,
a23 = l2 l3 c1 c3 b2 b3 ,
a24 = l1 l2 c1 c3 b22 ,
a25 = l22 c2 c3 b2 b3 ,
a26 = l2 l3 c23 b1 b2 ,
a34 = l1 l3 c21 b2 b3 ,
a35 = l2 l3 c1 c2 b23 ,
a36 = l32 c1 c3 b1 b3 ,
a45 = l1 l2 c1 c2 b2 b3 ,
a46 = l1 l3 c1 c3 b1 b2 ,
a56 = l2 l3 c2 c3 b1 b3 ,
a11
a
21
ZZ
[U ] [V ]T
1
a31
dxdy =
x
16A3e a41
e y
a51
a61
a11 = l12 b1 b2 c1 c2 ,
a14 = l12 b22 c21 ,
a12 = l1 l2 b22 c1 c3 ,
a15 = l1 l2 b2 b3 c1 c2 ,
a21 = l1 l2 b1 b3 c22 ,
a22 = l22 b2 b3 c2 c3 ,
a24 = l1 l2 b2 b3 c1 c2 ,
a31 = l1 l3 b21 c2 c3 ,
a32 = l2 l3 b1 b2 c23 ,
a34 = l1 l3 b1 b2 c1 c3 ,
a41 = l12 b21 c22 ,
a51 = l1 l2 b1 b2 c2 c3 ,
a54 = l1 l2 b22 c1 c3 ,
a61 = l1 l3 b1 b3 c1 c2 ,
a45 = l1 l2 b1 b3 c22 ,
a52 = l22 b22 c23 ,
a55 = l22 b2 b3 c2 c3 ,
(26)
a43 = l1 l3 b1 b3 c1 c2 ,
a46 = l1 l3 b21 c2 c3 ,
a53 = l2 l3 b2 b3 c1 c3 ,
a56 = l2 l3 b1 b2 c23 ,
a62 = l2 l3 b2 b3 c1 c3 ,
a65 = l2 l3 b23 c1 c2 ,
a35 = l2 l3 b1 b3 c2 c3 ,
a42 = l1 l2 b1 b2 c2 c3 ,
a44 = l12 b1 b2 c1 c2 ,
a64 = l1 l3 b2 b3 c21 ,
(25)
a66 = l32 b1 b3 c1 c3 .
(27)
[1] M. Koshiba and K. Inoue, Simple and efficient finite-element analysis of microwave and optical
waveguides, IEEE Trans. Microwave Theory Tech., vol. 40, no. 2, pp. 371377, 1992.
[2] M. Koshiba, S. Maruyama, and K. Hirayama, A vector finite element method with the
high-order mixed-interpolation-type triangular elements for optical waveguiding problems, J.
Lightw. Technol., vol.12, no. 3, pp. 495502, 1994.
[3] K. Kawano and T. Kitoh, Introduction to optical waveguide analysis. Solving Maxwells equations and the Schrodinger equation, John Wiley & Sons, New York 2001.
[4] J. Jin, The finite element method in electromagnetics, Second Edition, John Wiley & Sons,
New York 2002.
[5] H. Deng, Design and characterization of silicon-on-insulator passive polarization converter with
finite-element analysis, Ph. D. thesis, University of Waterloo, Ontario, Canada, 2005.
[6] Y. W. Kwon and H. Bang, The finite element method using Matlab, CRC Press, New York
2000.
[7] D. R. Tanner and A. F. Peterson, Vector expansion functions for the numerical solution of
Maxwells equations, Microwave and Optical Technology Letters, vol. 2, no. 9, pp. 331334,
1989.