Wave propagation
in periodic
nonlinear
dielectric
superlattices
D. Hennig
Freie Universitiit Berlin Fachbereich Physik, Institut jiir Theoretische Physik, Arnimallee 14, 14195 Berlin,
Germany, and Computational Physics Laboratory, Department of Physics, University of North Texas,
Demon, Texas 76203
H. Gabriel
Freie Universitiit Berlin, Fachbereich Physik, Institut fir ‘Theoretische Physik, Arnimallee 14, 14195 Berlin,
Germany
G. P. Tsironis
Computational Physics Laboratory, Department of Physics, University of North Texas, Denton, Texas
76203, and Research Center of Crete and Physics Department, University of Crete, l? 0. Box 1527, Heraklion
71110, Crete, Greece
M. Molina
Computational Physics Laboratory, Department of Physics, University of North Texas, Denton, Texas 76203
(Received 4 March 1994; accepted for publication 1 April 1994)
We investigate wave propagation in a superlattice consistent of dielectric material with a nonlinear
Kerr coefficient. We find gaps in the propagating properties of the medium that depend critically on
the injected wave power. This property can be used for transmission of information.
Novel phenomena such as photo& band gaps and possible light localization occur when electromagnetic (EM)
waves propagate in dielectric superlattices.1-6 In an approximation where only the scalar nature of the EM wave is taken
into account, wave propagation in a periodic or disordered
medium resembles the dynamics of an electron in a crystal
lattice. As a result, photonic bands and gaps arise in the
periodic lattice case whereas EM wave localization is theoretically possible in the disordered case. In the latter case and
when the dielectric medium is one dimensional, Andersontype optical localization has been predicted.r3* In an ordered
dielectric superlattice, on the other hand, photon band gaps
have been demonstrated for various realistic configurations.5
One issue that has not been widely addressed yet is the possibility of using superlattices with nonlinear dielectric properties and, in particular, the- nonlinear Kerr effect, to condevices with
desired transmission
struct optical
characteristics.277This is the problem we are addressing in
the present letter.
We consider the propagation of plane EM waves in the
scalar approximation in a one-dimensional continuous linear
dielectric medium. In the medium we embed periodically
small dielectric regions that have non-negligible third order
nonlinear susceptibility xC3’(Fig. 1). We will assume for simplicity that the width of the nonlinear dielectric regions is
much smaller than the distance between two adjacent ones.
We are thus led to a model for a periodic nonlinear superlattice and the propagation of a plane wave injected on one side
of the structure can be described through the following nonlinear Kronig-Penney equation:
wu(z)= -
d%(z)
dz2
N
+aC
&z-n)
it=1
X(1+Xlu(z)12)u(z).
(1)
In Eq. (l), u is the complex amplitude of an incoming plane
wave with frequency 6.1along direction Z, CYis proportional to
2934
Appt. Phys. Lett. 64 (22), 30 May 1994
the dielectric constant of the dielectric in each superlattice
slab and A is a nonlinear coefficient that incorporates ti3) and
the input wave power.’ The series of equidistant delta functions represent the effect of the periodic nonlinear dielectric
medium on the wave propagation. Straightforward manipulations similar to the ones used in the standard linear
Kronig-Penney problem lead to the following nonlinear difference equation for UjZU(j)‘-rl
2 COS(k)+Ct(l+k]ZLj]z)~
Uj+l+Uj-*=
sin(k)
i
i
uj,
(2)
where k is the wave number associated with the frequency
o(k)
=2 cos k. A local transformation to polar coordinates
and a subsequent grouping of pairs of adjacent variables
u,- , U, turns Eq. (2) to the following two-dimensional map
M:lf
&+,=-[
2 cosik)+aj
1+;
xWn+d--x,7
zn+l=z,--
1 x;-x;+l
2 wn+z,
h(lY,tznjj~]
(3)
’
(4)
x,=2r,r,-1
cos(O,-On-,),
where W,=V!~
z,,=<-G-r
with u,=r,
exp(itQ and J is the conserved current, i.e., J=2r,rnmI
sin(&-/3,-r).
The map M can contain bounded and diverging orbits.
The former ones correspond to transmitting waves whereas
the latter correspond to waves with amplitude escaping to
infinity and hence do not contribute to wave transmission.
The structure on the phase plane (x,,z,) is organized by a
hierarchy of periodic orbits surrounded by quasiperiodic orbits. As the value of X increases some periodic orbits become
unstable leading to stochasticity. This corresponds to passage
from a transmitting to a nontransmitting region. In Fig. 2 we
show one orbit corresponding to the period-4 Poincare-
0003-6951/94/64(22)/2934/3/$6.00
0 1994 American Institute of Physics
Downloaded 03 Feb 2002 to 128.210.126.195. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
2l-
N
-I-
FIG. 1. Aperiodic dielectric superlattice with nonlinear susceptibility due to
the Kerr effect. The dark regions denote the dielectric slabs with nonlinear
properties. The periodic value of the nonlinear coefficient X is approximated
by the periodic delta functions.
Birkhoff resonance zone. In Fig. 2(a) the regular periodic
orbit surrounding the four fixed points corresponds to wave
transmission through the super-lattice. In Fig. 2(b) on the
other hand, the same trajectory is shown for a larger incident
, wave amplitude. We observe that a thin chaotic layer has
developed that surrounds the separatrix but also some scat-
O-
-2-3 I
-14
-12
-10
B
x
-i3
14
t
PIG. 2. Orbit of the map corresponding to the period-4 Poincare-Birkhoff
chain. Parameters: k=4.7, X=1 and the incoming wave intensity IROlis in (a)
1.6 and in (b) 1.7. In both (a) and (b) the same trajectory is plotted. We note
the chaotic nature of the trajectory in (b) leading to nonpropagating waves.
1.00
0.75
+#
0.50
0.25
0
0
1
2
3
4
5
6
7
6
7
0
1
2
3
0
1
2
iki
4
5
6
5
6
T
0
1
2
3
k4
5
’
7
6
PIG. 3. (a) Transmission coefficient as a function of the wave number k for A equal to (a) zero (linear case), (b) 0.2, (c) 1.0, and (d) 4.0. The value of the linear
coefficient is o=l and the amplitude of the injected wave is taken as unity.
Hennig et a/.
2935
Appl. Phys. Lett., Vol. 64, No. 22, 30 May 1994
Downloaded 03 Feb 2002 to 128.210.126.195. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
The “band structure” shown in Fig. 4 can be obtained
directly from the tight-binding-like Eq. (2). In the linear
case, i.e., for X=0, the allowed propagating band states are
obtained from the well-known Kronig-Penney condition
1 SW) sl
cost k) + z a-
k
.
a
0
1
In the nonlinear Kronig-Penney case, on the other hand, this
condition for propagation gets modified leading to
2
cos(k)+;
4
0
K
sin(k)
Gl.
3
2
1
0
0
12
3k4
5
e
FIG. 4. Amplitude of injected wave R. as a function of the wave number k
for X equal to (a) zero (linear case) and (b) 1.0. The value of the linear
coefficient is o= 1. The superlattice has 4000 nonlinear slabs. Transmitting
solutions correspondto the blank area whereas diverging solutions are indicated by the hatched area.
tered points escaping to larger z values are visible. This trajectory corresponds to nonpassing plane waves.
In order to investigate directly the transmission properties of the injected plane waves in the nonlinear periodic
superlattice, we iterate numerically the discrete nonlinear
equation of Eq. (2). For the initial condition [ua,ur]
=[l, exp(ik)] we compute the transmitted wave amplitude T
for a superlattice with 10” nonlinear planes for different nonlinearity parameter A and wave number k. In Fig. 3 we plot
the transmission coefficient t=]a2 as a function of the input
wave number k for various nonlinearity values X. There are
clear transmission gaps whose width (in k space) dependson
X. We note that with increasing X the width of each gap
increases while in addition more gaps in the range between
two gaps develop. This process of gap creating starts in the
low energy range and extends with further increased X also
to the high energy region. Finally, above critical X values
neighboring gaps merge leading to a cancellation of transparency.
In Fig. 4 we plot the injected amplitudes for the linear
(x=oj and nonlinear (X#O) casesas a function of the wave
number k. We note that the typical linear band gaps [dark
areas in Fig. 4(a)] become exceedingly complicated when
nonlinearity becomes nonzero [hatched region in Fig. 4(b)].
A region was considered transmitting whenever the transmission coefficient was different from zero. In particular, we
observe the occurrence of new gaps in previously perfectly
transmitting regions. Furthermore, the width of the passing
regions (white regions) shrinks with increasing injected wave
energy. The diagram was obtained by taking a grid of 500
values of k and 250 values ]Ro] and iterating Eq. (2) on each
individual point of the grid over the N=104 sites.
2936
a(l++]“)-
Appl. Phys. Lett., Vol. 64, No. 22, 30 May 1994
The explicit occurrence of the wave intensity in the inequality (6) causesa broadening of the instability regions and also
produces new instability tongues in the region of former allowed bands for h values above a critical one. Furthermore,
merging of neighboring instability regions is possible leading
to an enhancedparameter instability.
The results we presented for the least transmitting case
of c+O are also sustained qualitatively in the more transmitting and physically relevant for dielectrics case of a<O. In
general, the presence of nonlinearity in the dielectric super- ’
lattice planes alters substantially the transmission properties
of the waves. In particular, when the nonlinear coefficient h
is increased new nontransmitting regions appear adjacent to
the regular instability regions. Consequently, for a given
wave number k, an appropriate change of the input power of
the wave (corresponding to a change in X) can switch the
wave from a transmitting to a nontransmitting region. Is is
then possible by a simple amplitude modulation of the incoming wave to transmit binary information to the other side
of the transmission line in the forms of zeros (non transmitting region) and ones (transmitting region). Since this transmitting capability depends critically on the properties of the
nontransmitting regions, further study of these regions under
the nonscalar wave approximation is currently under way.
We acknowledge partial support from the US Air Force
Phillips Labs.
‘S. John, Phys. Rev. Lett. 53, 2169 (1984).
‘S. John, Phys. Rev. L&t. 58, 2486 (1987).
‘E. Yablonovitch, Phys. Lett. 58, 2059 (1987).
4S. John and R. Rangarajan, Phys. Rev. B 38, 10101 (1988).
‘K. M. Ho, C. T. Chan, and C. M. Soukoulis, Phys. Rev. Lett. 65, 3152
(1990).
6M. Sigalas, C. M. Soukoulis, E. N. Economou, C. T. Chan, and K. M. Ho,
Phys. Rev. B 48, 121 (1993).
‘W. Chen and D. L. Mills, Phys. Rev. Lett. 58, 160 (1987); Phys. Rev. B
36, 6269 (1987).
‘MM.I. Molina, W. D. Deering, and G. P. ‘lhironis, Physica D 66,135 (1993).
‘J. BelIisard, A. Formoso, R. Lima, and D. Testard, Phys. Rev. B 26,3024
(1982).
I’D. Wiirtz, M. P. Soerensen,and T. Schneider, Helvetica Physica Acta 61,
345 (1988).
‘ID. Hennig, H. Gabriel, G. P. Tsironis, and M. I. Molina (unpublished).
Hennig et al.
Downloaded 03 Feb 2002 to 128.210.126.195. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp