Appendix F
Codirectional Coupling
We consider a structure with two high-index cores, supporting two modes. In the
case of weak coupling, the two modes are close to the modes that would be support-
ed by the cores in isolation. However, it is more accurate to use the two modes of
the combined structure as a basis.
A square grating is introduced as a perturbation to the original structure, which
contains a layer with an average refractive index as shown in Figure F.1. The index
perturbation is written as a Fourier series, with ling = n, - n2 being the index differ-
ence of the layers forming the grating:
lin cos (21TZ)
-4 z:». -A + higher-order terms
lin(x, z) = 1T 2 (F.!)
{
o if x < 0 or if x > dg
Neglecting any y-dependence and using lin(x, Z)2 : : : : 2navlin(x, z), where nav = (n} +
n2)/2, we seek a solution to the field equation
iJ2E(x, z) iJ2E(x, z) 2 2
az2 + ax2 + ko[n (x) + 2navlin(x, z)]E(x, z) = 0 (F.2)
where n2(x) is the x-dependence of the index variation in the unperturbed structure.
We assume that the solution to (F.2) is of the form
E(x, z) = R(z)ER(x)exp(-if3RZ) + S(z)ES<x)exp(-if3sz) (F.3)
where ER(x) and ES<x) are the normalised field distributions for the two modes of
the unperturbed structure with the propagation constants f3R and f3s, respectively.
R(z) and S(z) are the z-dependent amplitude factors.
Inserting (F.3) into (F.2), neglecting terms containing the second derivatives of
Rand S with respect to z, writing the index variation in (F.!) as the sum of two ex-
Tunable Laser Diodes and Related Optical Sources, 2nd ed. By Buus, Amann, and Blumenthal 367
ISBN 0-471-20816-7 © 2004 by the Institute of Electrical and Electronics Engineers.
368 CODIRECT/ONAL COUPLING
x ~A~
I I
: L : L :
I A I B I
I , I
~n(x/z)==(n 2-~ )/2
.,'-
I I I
~"(
~/
~n(x/z)==(nl-n 2)/2
.z •z .z
(0) (b) (c)
Figure F.1 Grating with period A forming a codirectional coupler structure. (a) Unperturbed
structure. (b) Structure with grating. (c) Index perturbation caused by the grating.
ponentials, and using the fact that to first order the x-dependence of the modes is
still given by ER(x) and EsCx), we get
dR dS
2if3R-ER(x)exp(-if3RZ) + 2if3s-EsCx)exp(-if3sz) ==
dz dz
2 konavdn
1T
2 l (.
g exp J T
2m ) + exp ( -J.T
2m )]
g(x)[R(z)ER(x)exp(-if3RZ) + S(z)EsCx)exp(-if3sz)] (F.4)
where
if 0::; x ::; dg
g(X)=g (F.5)
if x < 0 or if x > dg
Multiplying (F.4) by ER(x) and EsCx) in tum, integrating over x, and using the ortho-
normality of ER(x) and EsCx) leads to
f3R dR
dz -_ - i1T konavdn
2
g
l (. Tttz )
exp J T + exp -JT
( . Tttz )]
and
f3s dS
dz == - i1T k1Jnavdng[ r:
( 2m) + exp (2m
exp -iT )l
REFERENCES 369
Retaining only terms which are nearly phase matched, f3R - f3s ~ 27T/A, defining
the detuning 0 according to
(F.8)
and using f3R ~ f3s ~ kon av with ko == 27T/A, gives
~ 2~n l~
- == -j--g S(z)exp(2joz) ER(x)Es(x)dx (F.9)
dz A 0
and
dS 2~n l~
- == -j--g R(z)exp(-2joz) ER(x)Es(x)dx ([Link])
dz A 0
The coupling coefficient K is defined as
K ==
is»
zi::«
Id g
ER(x)Es(x)dx ([Link] )
A 0
and we introduce the new variables
9l{z) == R(z)exp(-joz) (F.12)
$(z) == S(z)exp(joz) (F.13)
After taking out the common phase factors we finally get
dfll
- + jofll== -jK$ (F.14)
dz
d$
- - jo$== -jKfll (F.15)
dz
These equations are seen to be identical to (7.65) and (7.66). For a more general,
detailed, and unified treatment of codirectional and contradirectional coupling, see
Section 6.4 in [1].
REFERENCE
[1] Yariv, A., and Yeh, P., Optical Waves in Crystals, New York, Wiley, 1984.