Integrated Optics

Chapter 11
Physics 208, Electro-optics
Peter Beyersdorf

Document info ch 11. 1

 Dielectric Waveguides

Optical waveguides
“patterned” onto class
or crystals form the
basis for optical
integrated circuits
that can perform most
of the functions of
free space optics, and
some functions that
free space optics
cannot.

ch 11. 2
Wave Equation in a Waveguide

Consider an isotropic material with translational

symmetry along z and a refractive index profile in
the x-y plane such as
nc y
ns x

Ampere’s law and Faraday’s law can be written in

the form
! ×H
∇ ! !
= iω#0 n2 (x, y)E
! ×E
∇ ! = −iωµ0 H !

and will have solutions of the form

E! ! y)ei(ωt−βz)
= E(x,
!! ! ! ! ! ! !!
H != !H(x,
! ! y)e
! i(ωt−βz)
! ! ! ! where β kz ch 11. 3
Solutions in the Substrate
∂2E
The wave equation !∇ !E −! µ!! !2 =! 0
2 can be written as
∂t
! "
ω2 2
2
n (x, y) − α2 (x, y) − β 2 E$ = 0
c

where α2(x,y) n2(x,y)[k0x2+k0y2] is the transverse

component of the propagation vector such that
! √
! y) = ! 0 )e in(x,y)α0 x2 +y 2
E(x, E(α
α0

For a guided mode the fields should go to zero far

from the waveguide core, thus α must be imaginary
making E an exponentially decaying function in x
and y. Thus ω2
β2 > n2s
c2 ch 11. 4
 Solutions in the Core

Since the fields decay exponentially to zero far from

the core in the substrate, they must have a maximum
value in the core (i.e. a point where the field gradient
is zero, ∇E=0, and the laplacian is negative, ∇2E<0)

Thus with our phasor notation where ∇2→α2+β2 and

with the constraint on α(x,y) and β from the wave
equation
! "
ω2 2
2
n (x, y) − α2 (x, y) − β 2 E$ = 0
c
in the core we must have
2 ω2 2
β < 2 nc
c

ch 11. 5
Orthogonality of Modes
The longitudinal propagation constant β must
satisfy 2 2
ω 2 2 ω 2
n < β < n
c2 s c2 c

any number of modes, each with a different β

satisfying this condition can propagate
simultaneously, without interacting (i.e. without
exchanging power).

ch 11. 6
Slab Waveguide Example
Consider a one dimensional slab waveguide, as is
commonly used in solid state lasers. Taking
n2>n3≥n1 i.e. we have a high index core for
guiding the radiation

What do the waveguide modes look like?

ch 11. 7
 Slab Waveguide Example

∂2
ial 1 2
E(x, y) + (k02 n21 − β 2 )E(x, y) = 0
∂x
∂2
ial 2 E(x, y) + (k0
2 2
n 2 − β 2
)E(x, y) = 0
∂x2
∂2
ial 3 2
E(x, y) + (k02 n23 − β 2 )E(x, y) = 0
∂x

in material 1, 2 and 3
respectively

β larger than nsk0

β too small for field is not physical
to decay at ±∞ ch 11. 8
 Guided Modes
Consider a symmetric slab waveguide
(ns n3=n1) with nc=n2 with thickness d
For guided modes ω2 2 2 ω2 2
n < β < 2 nc
c2 s c

In the substrate the field is

exponentially decaying and has the
form

E(x > d) = E+ e−q(x−d)

E(0 < x < d) = Ec [A cos hx + B sin hx]
E(x < 0) = E− eqx

h α(x,y) in the core

q iα(x,y) in the substrate ch 11. 9
 TE Modes
If you consider a ray zig-zagging
through the slab, the plane of zig-
zagging defines a polarization
direction. For a mode with an
electric field transverse to this plane
(a TE mode) the field can be
expressed as
Ey (x, y, z, t) = Ey (x)ei(ωt−βz)
and must be continuous at the
interfaces giving
! q "
E(x > d) = C cos hd + sin hd e−q(x−d)
! h "
q
E(0 < x < d) = C cos hx + sin hx
h
ch 11.10
E(x < 0) = Ceqx
 TE Modes
Plugging each of
! q " ns
E(x > d) = C cos hd + sin hd e−q(x−d)
nc
! h "
q ns
E(0 < x < d) = C cos hx + sin hx
h
E(x < 0) = Ceqx

into the corresponding wave equations

∂2
E(x, y) + (k0
2 2
n s − β 2
)E(x, y) = 0
∂x2
∂2
2
E(x, y) + (k02 n2c − β 2 )E(x, y) = 0
∂x
∂2
2
E(x, y) + (k02 n2s − β 2 )E(x, y) = 0
∂x
!
in the three regions gives h = n2c k02 − β 2
!
q = β 2 − n2s k02
ch 11. 11
 TE Modes
The H field must also be continuous
ns
across the interfaces. From nc
Faraday’s law ns

! ×E
∇ ! = −iωµ0 H
!

with ! = Ey (x)ei(ωt−βz) ĵ
E
! " # $
dEy (x) i(ωt−βz)
gives ! = i
H iβ Ey (x)ei(ωt−βz) î + e k̂
ωµ0 dx
This is the magnitude of
The magnetic field has a
the E field which is
component along the
already continuous across
direction of propagation
the boundaries

ch 11. 12
 TE Modes
Requiring H(z) be continuous across
ns
the boundaries gives nc
ns
q2
h sin hd − q cos hd = q cos hd + sin hd
h

where h and q are functions of β.

Thus guided modes can only exist
for discrete values of β which
satisfy this mode condition
For light to leak out of core
β non physical (kz>k)
β<nsko ns θt sk 0
n
nc
k0

θi
nc

β
unconfined modes, not
ch 11.13
propagating along z
 TM Modes
Analogous analysis of TM modes
gives similar results. In terms of the
magnetic field amplitude:
! "
n2s h q
Hy (x > d) = C cos hd + sin hd e−q(x−d)
n2c q h
! 2 "
n h q
Hy (0 < x < d) = C s2 cos hx + sin hx
nc q h
n2s h qx
Hy (x < 0) = Ce
n2c q

subject to the mode constraint

! "2
n2c n2c n2c q2
h sin hd − 2 q cos hd = 2 q cos hd + sin hd
ns ns n2s h

ch 11.14
 Cutoff Frequencies
Waves below a “cutoff”
frequency will not be guided
by the waveguide. Roughly
speaking the waveguide
dimensions must be larger
than a wavelength. The
requirements for guiding the
ωc
mth mode is
d m
≥ !
λ 2 n2c − n2s

d 1
For!! λ! <!!n!2 −! n2! only a single mode can be guided
c s

ch 11.15
 Mode Structure
Mode profiles in a symmetric 1D waveguide

Given a mode profile, how do you determine its

mode number?

ch 11. 16
Integrated EOM devices
Two issues with free-space EOM devices can be
largely eliminated by integrating the devices into
waveguides

Half wave voltage can be much lower due to

the small gap between electrodes allowing a
much larger field

Interaction length can be much longer

because the beam does not spread out as it
propagates electrode

waveguide
ch 11.17
Dielectric Tensor Perturbation

Consider a perturbation to the dielectric tensor

created by an externally applied field E.
#
η ! = η + ∆η = "0 "−1 + rE
# = "0 (" + ∆")−1
η ! = η + ∆η = "0 "−1 + rE

!0 !−1 (1 + ∆!!−1 )−1 "

= !0 !−1 + rE
"
!0 !−1 (1 − ∆!!−1 ) = !0 !−1 + rE
−!0 !−1 ∆!!−1 = rE "
!(rE)!
∆! = −
!0

this perturbation affects the propagation of modes

in the waveguide. In an isotropic material
∆! = −!0 n4 (x)(rE) ch 11.18
 Mode Coupling

Consider the coupling between two modes (m and n)

introduced by this perturbation. For efficient
coupling the modes must have similar propagation
constants (βm≈βn) thus they must be orthogonally
polarized with the same mode number (m=n).
Considering only these two modes
the filed can be written as
! TE TM
"
! r, t) =
E(! TE
Am Em (x)e−iβm z + Bn EnT M (x)e−iβn z eiωt

where Am and Bn will be functions of

z in the presence of mode coupling
introduced by the external electric
field
ch 11.19
 Mode Coupling Equations
! TE TM
"
! r, t) =
E(! TE
Am Em (x)e−iβm z + Bn EnT M (x)e−iβn z eiωt

Coupled mode analysis (see slide 9.21) leads to a

relation between Am(a) and Bm(z)
dAm
= −iκmn Bn ei∆βz
dz
dBn
= −iκ∗mn Am e−i∆βz
dz

with
∆β = βm
TE
− βnT M
and ω
! ∞
κmn = ∗T E
Em · #EnT M dx
4 −∞

“overlap integral”
ch 11.20
 Overlap Integral

With the overlap integral

! ∞
ω
κmn = ∗T E
Em · #EnT M dx
4 −∞

and the fields for the slab waveguide already

studied where ETE is along y and ETE is along x it
is the ε6 term that couples the modes:
!6 = −!0 n4 (x)(r6k Ek )

if the modes are well confined so that they see

an index n(x) that is primarily nc then
1
κmn ≈ − n3c (x)k0 r6k Ek
2
ch 11.21
Phase Matched Coupling
The solutions to the mode amplitudes (subject to
Am(0)=A0 and Bn(0)=0) when the modes are phase
matched (Δβ=0) is

Am = A0 cos(κmn z)
Bn = −iA0 sin(κmn z)

as with phase matching for acousto-optic

interactions or second harmonic generation,
imperfect phase matching reduces the maximum
amount of power coupling between the modes

ch 11.22
Phase Mismatched Coupling
If Δβ≠0 the modes drift out of phase reducing
(and eventually inverting) the transfer of energy
between modes, resulting in less than 100%
conversion efficiency

Periodically reversing the sign of the

perturbation can “quasi-phase match” the
interaction allowing 100% conversion efficiency

ch 11.23
 Directional Couplers
Consider the coupling of two modes from
different waveguides that are parallel to each
other but fully separated. Power can be coupled
form one to another in a process called
“Directional coupling”

ch 11.24
 Waveguide Coupling
The electric field in a structure with two
parallel waveguides (a and b) can be written

!
E(x, !
y, z, t) = A(z)Ea (x, y)e
i(ωt−βa z) !
+ B(z)Eb (x, y)e
i(ωt−βb z)

and we express square of the refractive index as

the sum of three parts, that of the surrounding
cladding (ns), that of core a (na), and core b (nb)
n2 (x, y) = n2s (x, y) + ∆n2a (x, y) + ∆n2b (x, y)

ch 11.25
 Coupled Modes
The eigenmodes in each uncoupled waveguide
obey the wave equation
n2 ∂ 2
2
∇ Ea,b = 2 2 Ea,b
c ∂t
giving
! "
∂2 ∂2 ω2 # 2 $
+ − β 2
a,b Ea,b (x, y) = − n (x, y) + ∆n 2
(x, y) Ea,b (x, y)
∂x2 ∂y 2 c2 s a,b

introducing the coupling between the waveguides

this becomes (for waveguide a)
! ! 0
""
∂2
∂2
dA d A 2
ω2 # $
+ − βa
2
+ 2iβ a + 2 Ea (x, y) = − A n2
s (x, y) + A ∆n 2
a (x, y) + B ∆n 2
b (x, y) Ea (x, y)
∂x2 ∂y 2 dz dz c2

and likewise for waveguide b Coupling term

with a b
ch 11.26
 Differential Equations
This coupled mode analysis leads to the following
equations for the mode amplitudes A and B

dA
= −iκab Bei(βa −βb )z − iκaa A
dz
dB
= −iκba Ae−i(βa −βb )z − iκbb B
dz
Modification to βa
with 1
!
κab ≡ ω#0 Ea∗ · ∆n2a (x, y)Eb dxdy due to Δnb
4
!
1
κba ≡ ω#0 Eb∗ · ∆n2b (x, y)Ea dxdy
4 Modification to βb
!
1 due to Δna
κaa ≡ ω#0 Ea∗ · ∆n2b (x, y)Ea dxdy
4
!
1
κbb ≡ ω#0 Eb∗ · ∆n2a (x, y)Eb dxdy
4
ch 11.27
 Alternative Form
Letting β’a→βa+κaa and β’b→βb+κbb gives a
simpler form of coupled differential equations

dA ! !
= −iκab Bei(βa −βb )z
dz
dB ! !
= −iκba Ae−i(βa −βb )z
dz
For a symmetric waveguide (κ=κab=κba) this has
the same form as the equations governing the
acoustooptic interaction or second harmonic
generation

ch 11.28
Directional Coupler Solutions
With only a single input (A(0)=A0, B(0)=0) the
power in the waveguides at z is is Pa(z)=A(z)*A(z)
and Pb(z)=B*(z)B(z) and has the form

Pa (z) = P0 − Pb (z)
% 
& '
! 2
κ2 ∆β
Pb (z) = P0 ! " sin2  κ2 + z
∆β !
2 2
κ2 + 2

ch 11.29
 Practical Applications
For typical parameters, κ-1≈2mm

Uses for this type of coupler include

Amplitude modulator (Vary output of waveguide a or b by

varying Δβ via EO effect)

EO switch (adjust Δβ via EO effect to change power

coupling from 0 to 100%)

Wideband frequency filter (based on λ dependence of κ)

Narrow band frequency filter (based on resonant

enhancement [as described in Pradeep’s talk])

Frequency division multiplexer

Integrated “beam splitter” ch 11. 30

 References
Yariv & Yeh “Optical Waves in Crystals” chapter
11

ch 11.31