Академический Документы
Профессиональный Документы
Культура Документы
www.elsevier.com/locate/optcom
Abstract
We present a new approach for analyzing band structures in one-dimensional Kerr-nonlinear photonic crystals. It
combines the nite-dierence time-domain method, based on the numerical simulation of oscillating dipole radiation,
with the Kerr-nonlinear model. The approach is applied to analyze the band structures in Kerr-nonlinear one-dimensional photonic crystals as a function of the intensity of the oscillating dipole. We nd that these bands are dynamically
red-shifted with regard to the bands in linear one-dimensional photonic crystals. The proposed approach will help to
understand such phenomena as intensity-driven optical limiting and all-optical switching with Kerr-nonlinear photonic
crystals.
2004 Elsevier B.V. All rights reserved.
PACS: 42.65.Pc; 42.65; 42.70.Qs
Keywords: One-dimensional nonlinear photonic crystal; Finite-dierence time-domain method; Oscillating dipole; Band structure;
Dynamic red-shift
1. Introduction
One-dimensional linear photonic crystals (1-D
LPC) are periodic structures that contain alternating layers of material with high and low dielectric
0030-4018/$ - see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.optcom.2004.05.022
214
nonlinear optical skin eect and self-induced transparency eects were studied. Tran [21] was the rst
who calculated band structures in Kerr-nonlinear
photonic crystals. The notable feature of his approach is the calculation of the electric eld from
a cubic equation derived from the constitutive
equation for a Kerr-nonlinear medium. We have
recently applied the Trans approach to calculate
band structures in Kerr-nonlinear photonic crystal
slabs [22].
In this paper, we present a new approach for
analyzing band structures in 1-D NLPC. It combines the FDTD method, based on the numerical
simulation of oscillating dipole radiation, with
the Kerr-nonlinear model. Such FDTD method
was rst proposed by Sakoda [23] and successfully
used to calculate band structures in metallic systems. With this method, the photonic crystals are
excited by an embedded oscillating dipole. The
Kerr-nonlinear model used is borrowed from
[21,24,25].
The paper is organized as follows. First, we discuss the theoretical considerations of the 1-D
NLPC. In the second section, we present the
FDTD method and the Kerr-nonlinear model,
which are used to analyze band structures. The
next section discusses the results. In it, the energy
spectrum plots calculated for the 1-D LPC and
1-D NLPC are identied and the band structures
in the 1-D LPC calculated with the FDTD and
the plane wave expansion methods are compared.
We then go on to discuss the band structures in the
1-D NLPC for dierent amplitudes of the oscillating dipole. We demonstrate the validity of our approach by comparing the results with analogical
results. Finally, we discuss the physical origins of
the red-shift of the band structures.
2. Theory
Here, we consider the FDTD method based on
the numerical simulation of oscillating dipole radiation and the Kerr-nonlinear model. The oscillating dipole is embedded in the 1-D NLPC, whose
general geometry is illustrated in Fig. 1. It consists
of alternating layers of materials with high (e1) and
215
~ ~
where ~
E~
r; t and H
r; t are the electric and mag~ r; t is the electric displacement;
netic elds; D~
~
r; t and ~
P M ~
r; t are the polarization elds of
P E ~
the electric and magnetic dipoles; c is the speed
of light in vacuum and e~
r is the position-dependent dielectric constant. In order to solve Eqs.
(2) and (3), we need so-called constitutive equation
~ r; t to ~
that relates D~
E~
r; t. For the linear
medium, it is simply
~
D~
r; t e~
r ~
E~
r; t:
Therefore only the region where 0 6 y 6 C is considered for calculation and in the y-direction the
1-D NLPC is truncated by the symmetric boundary condition (BC). The periodic BC truncates
the structure in the x-direction. In the z-direction,
the structure is assumed to be innite. The oscillating dipole is situated at ~
r0 x0 ; y 0 . The parameters nx and ny denote the number of subcells, which
divide the unit cell (the period a). The parameters
Dx and Dy characterize the dimensions of the single subcell. All these parameters will be dened in
the following section.
For a description of the FDTD method, we
start with Maxwells equations (Gaussian units)
1 o ~
r~
E~
r; t
H ~
r; t 4p~
P M ~
r; t ;
2
c ot
~ ~
rH
r; t
1 o ~
D~
r; t 4p~
P E ~
r; t ;
c ot
In the nonlinear case, we will have a dierent constitutive equation. For the Kerr-nonlinear medium
(the layers with a high dielectric constant), the dielectric constant depends on the electric eld ~
E~
r; t
and the Kerr coecient v(3) [21,24,25]
2
e~
r e1 v3 ~
E~
r; t :
5
Now Eq. (5) can be substituted into Eq. (4), which
is resulted in the constitutive equation for the
Kerr-nonlinear medium
2
~
D~
r; t e1 v3 ~
E~
r; t:
6
E~
r; t ~
The polarization elds of the electric and magnetic dipoles can be expressed in the explicit form
as:
~
P E ~
r ~
r0 eixt ;
r; t ~
el d~
~
P E ~
r ~
r0 eixt ;
r; t ~
hl d~
where ~
el and ~
hl are the amplitudes of the electric
and magnetic dipoles, ~
r0 indicates the position of
the dipoles within the photonic crystal and x is
the angular frequency of the oscillation; i refers
to the imaginary unit and d~
r ~
r0 denotes the Dirac delta function.
The electromagnetic energy density W emitted
per unit time by the oscillating dipole at ~
r0 can
be calculated by using the following expression
[26]:
2 i
1 h~
~~
~ ~
W
E~
r D
r H
r :
9
8p
In accordance with the Bloch theorem [23], the periodic BC can be written as
216
~
~~
~~
U
r ~
a; t U
r; t eik~a ;
10
11
~x; C U
~x; C Dy :
U
12
1 o
H z 4phlz dx x0 dy y 0 eixt
c ot
oEy oEx
;
ox
oy
13
oH z
1 o
Dx 4pelx dx x0 dy y 0 eixt
;
c ot
oy
14
i oH
1 oh
z
Dy 4pely dx x0 dy y 0 eixt
:
c ot
ox
15
Ey
Dx
e1 v3 jEx j2
Dy
2 :
e1 v3 Ey
3. Numerical method
In order to numerically solve Eqs. (13)(17),
one of the existing nite-dierence schemes [28]
can be used to approximate both derivatives in
space and time. We will use the scheme proposed
by K. Yee, which is the standard for solving Maxwells equations. The nite-dierence expressions
corresponding to Eqs. (13)(17) are
1
1
1
2
;
m
H p
z
2
2
1
1
1
2
;
m
H p
l
z
2
2
"
p
Ey l 1; m 12 Epy l; m 12
Dt
Dx
#
Epx l 12 ; m 1 Epx l 12 ; m
Dy
4pixhlz Dt
dll0 dmm0 exp ixpDt;
18
DxDy
1
p1
Dx
l ;m
2
1
Dt
1
1
p
p12
Dx l ; m
Hz
l ;m
2
Dy
2
2
4pixe
Dt
1
1
1
l
x
2
H p
dll0 dmm0
l ;m
z
2
2
DxDy
16
17
exp ixpDt;
19
1
l; m
2
1
Dt
1
1
p
p12
Hz
Dy l; m
l ;m
2
Dx
2
2
1
1
1
2
H p
l ;m
z
2
2
4pixely Dt
dll0 dmm0 exp ixpDt;
20
DxDy
Dxp1 l 12 ; m
1
p1
l ;m
21
Ex
2 ;
2
e1 v3 Ep l 1 ; m
Dyp1
Eyp1
1
l; m
2
p1
l; m 12
2 ;
e1 v3 Epy l; m 12
Dy
22
217
where index p refers to a grid point in time and indices l and m denote x and y, respectively. The expressions dll0 and dmm0 denote the position of the
dipole on the spatial grid. We normalize the angular frequency and the wave vector as xa/2pc and
ka/2p, respectively. Because of this normalization,
the values of Dx and Dy are now calculated as 2pa/
nx and 2pa/ny, where nx and ny are the number of
the subcells in the unit cell in the x- and y-directions, respectively. If the symmetric boundary condition is set in the y-direction, Dy is dened as pa/
ny. The temporal step Dt is calculated as 2p/(xNt),
where Nt is the total number of temporal steps.
The Courant
stability condition [28] is satised
q
2
1 X
1
1
W
E x l ; m Dx l ; m
8p l;m0
2
2
1
1
Ey l; m
Dy l; m
2
2
1
1
H 2z l ; m
:
23
2
2
has nished for all k-points (it is enough to take into account about 1015 k-points for a wave vector),
the accumulated electromagnetic energy density
distribution W is a function of the angular frequency
for the specied wave vector. The band structure is
obtained from the resonance peaks of this electromagnetic energy density distribution.
In Fig. 3, we present the ow chart of the sequence of operations inside the main body of the
FDTD algorithm. Here j denotes the number of
the current temporal step and Nt is the total num-
218
Fig. 4. The accumulated electromagnetic energy density distribution for 1-D LPC (a) and 1-D NLPC (b) vs. the number of
oscillating cycles. ., s, e and h denote the accumulated
electromagnetic energy after 10, 20, 50 and 100 oscillation
cycles, respectively. In both cases the amplitude of the magnetic
dipole is hlz 100 arb: units.
219
220
Fig. 7. The band structures in the 1-D LPC (dashed line) and
1-D NLPC (solid line) calculated with the proposed approach
and oscillating dipole amplitudes of hlz 100 arb: units (a) and
hlz 450 arb: units (b). The open circle refers to the bands
calculated in 1-D NLPC by using the Fourier method [11].
The value of De increases as the intensity increases if v(3) > 0 and decreases if v(3) < 0. As the magnetic dipole excites the structure, the value of De
changes and the bands of the band structure dynamically shift. This process is the basis for intensity-driven optical limiting and all-optical
switching. The results show that for the positive
Kerr coecient the value of De increases and
the bands dynamically red-shift. This red-shift increases as the intensity of the oscillating dipole in-
5. Conclusions
221
We have presented a new approach for analyzing band structures in one-dimensional Kerr-nonlinear photonic crystals. This approach was used
to analyze the band structures in Kerr-nonlinear
one-dimensional photonic crystals as a function
of the intensity of the oscillating dipole. We found
that the bands of these band structures are dynamically red-shifted with regard to the bands of band
structures in linear one-dimensional photonic crystals. We have discussed the physical origins of this
red-shift and showed that the red-shift increases as
the incident intensity of the oscillating dipole increases. The proposed approach was validated
and found to be useful for understanding such
phenomena as intensity-driven optical limiting
and all-optical switching with Kerr-nonlinear photonic crystals.
Acknowledgements
References
222