1/253

Back
Close
Optical Waveguides (OPT568)
Govind P. Agrawal
Institute of Optics
University of Rochester
Rochester, NY 14627
c _2008 G. P. Agrawal
2/253

Back
Close
Introduction
Optical waveguides conne light inside them.
Two types of waveguides exist:
Metallic waveguides (coaxial cables,
useful for microwaves).
Dielectric waveguides (optical bers).
This course focuses on dielectric waveguides
and optoelectronic devices made with them.
Physical Mechanism: Total Internal Reection.
3/253

Back
Close
Total Internal Reection
Refraction of light at a dielectric interface is governed by
Snells law: n
1
sin
i
= n
2
sin
t
(around 1620).
When n
1
> n
2
, light bends away from the normal (
t
>
i
).
At a critical angle
i
=
c
,
t
becomes 90

(parallel to interface).
Total internal reection occurs for
i
>
c
.
4/253

Back
Close
Historical Details
Daniel Colladon Experimental Setup John Tyndall
TIR is attributed toJohn Tyndall (1854 experiment in London).
Book City of Light (Je Hecht, 1999) traces history of TIR.
First demonstration in Geneva in 1841 by Daniel Colladon
(Comptes Rendus, vol. 15, pp. 800-802, Oct. 24, 1842).
Light remained conned to a falling stream of water.
5/253

Back
Close
Historical Details
Tyndall repeated the experiment in a 1854 lecture at the suggestion
of Faraday (but Faraday could not recall the original name).
Tyndalls name got attached to TIR because he described the ex-
periment in his popular book Light and Electricity (around 1860).
Colladon published an article The Colladon Fountain in 1884 to
claim credit but it didnt work (La Nature, Scientic American).
A sh tank and a laser pointer can be
used to demonstrate the phenomenon
of total internal reection.
6/253

Back
Close
Dielectric Waveguides
A thin layer of high-index material is sandwiched between two layers.
Light ray hits the interface at an angle = /2
r
such that n
0
sin
i
= n
1
sin
r
.
Total internal reection occurs if >
c
= sin
1
(n
2
/n
1
).
Numerical aperture is related to maximum angle of incidence as
NA = n
0
sin
max
i
= n
1
sin(/2
c
) =
_
n
2
1
n
2
2
.
7/253

Back
Close
Geometrical-Optics Description
Ray picture valid only within geometrical-optics approximation.
Useful for a physical understanding of waveguiding mechanism.
It can be used to show that light remains conned to a waveguide for
only a few specic incident angles angles if one takes into account
the GoosHanchen shift (extra phase shift at the interface).
The angles corresponds to waveguide modes in wave optics.
For thin waveguides, only a single mode exists.
One must resort to wave-optics description for thin waveguides
(thickness d ).
8/253

Back
Close
Maxwells Equations
E =
B
t
H=
D
t
D = 0
B = 0
Constitutive Relations
D =
0
E+P
B =
0
H+M
Linear Susceptibility
P(r, t) =
0
_

(r, t t
/
)E(r, t
/
)dt
/
9/253

Back
Close
Nonmagnetic Dielectric Materials
M= 0, and thus B =
0
H.
Linear susceptibility in the Fourier domain:

P() =
0
()

E().
Constitutive Relation:

D =
0
[1+()]

E
0
()

E.
Dielectric constant: () = 1+().
If we use the relation = (n+ic/2)
2
,
n = (1+Re )
1/2
, = (/nc)Im.
Frequency-Domain Maxwell Equations:

E = i
0

H, (

E) = 0

H = i
0

E,

H= 0
10/253

Back
Close
Helmholtz Equation
If losses are small, n
2
.
Eliminate H from the two curl equations:

E =
0

2
n
2
()

E =

2
c
2
n
2
()

E = k
2
0
n
2
()

E.
Now use the identity

E (

E)
2

E =
2

E = 0 only if n is independent of r (homogeneous medium).

We then obtain the Helmholtz equation:

E+n
2
()k
2
0

E = 0.
11/253

Back
Close
Planar Waveguides
x
y
z
n
s
n
c
n
1
Core
Substrate
Cover
Core lm sandwiched between two layers of lower refractive index.
Bottom layer is often a substrate with n = n
s
.
Top layer is called the cover layer (n
c
,= n
s
).
Air can also acts as a cover (n
c
= 1).
n
c
= n
s
in symmetric waveguides.
12/253

Back
Close
Modes of Planar Waveguides
An optical mode is solution of Maxwells equations satisfying all
boundary conditions.
Its spatial distribution does not change with propagation.
Modes are obtained by solving the curl equations
E = i
0
H, H=i
0
n
2
E
These six equations solved in each layer of the waveguide.
Boundary condition: Tangential component of E and H be
continuous across both interfaces.
Waveguide modes are obtained by imposing
the boundary conditions.
13/253

Back
Close
Modes of Planar Waveguides
E
z
y

E
y
z
= i
0
H
x
,
H
z
y

H
y
z
= i
0
n
2
E
x
E
x
z

E
z
x
= i
0
H
y
,
H
x
z

H
z
x
= i
0
n
2
E
y
E
y
x

E
x
y
= i
0
H
z
,
H
y
x

H
x
y
= i
0
n
2
E
z
Assume waveguide is innitely wide along the y axis.
E and H are then y-independent.
For any mode, all led components vary with z as exp(iz). Thus,
E
y
= 0,
H
y
= 0,
E
z
= iE,
H
z
= iH.
14/253

Back
Close
TE and TM Modes
These equations have two distinct sets of linearly polarized solutions.
For Transverse-Electric (TE) modes, E
z
= 0 and E
x
= 0.
TE modes are obtained by solving
d
2
E
y
dx
2
+(n
2
k
2
0

2
)E
y
= 0, k
0
=

0
= /c.
Magnetic eld components are related to E
y
as
H
x
=

0
E
y
, H
y
= 0, H
z
=
i

0
dE
y
dx
.
For transverse magnetic (TM) modes, H
z
= 0 and H
x
= 0.
Electric led components are now related to H
y
as
E
x
=

0
n
2
H
y
, E
y
= 0, E
z
=
i

0
n
2
dH
y
dx
.
15/253

Back
Close
Solution for TE Modes
d
2
E
y
dx
2
+(n
2
k
2
0

2
)E
y
= 0.
We solve this equation in each layer separately using
n = n
c
, n
1
, and n
s
.
E
y
(x) =
_
_
_
B
c
exp[q
1
(x d)]; x > d,
Acos(px ) ; [x[ d
B
s
exp[q
2
(x +d)] ; x <d,
Constants p, q
1
, and q
2
are dened as
p
2
= n
2
1
k
2
0

2
, q
2
1
=
2
n
2
c
k
2
0
, q
2
2
=
2
n
2
s
k
2
0
.
Constants B
c
, B
s
, A, and are determined from the boundary
conditions at the two interfaces.
16/253

Back
Close
Boundary Conditions
Tangential components of E and H continuous across any interface
with index discontinuity.
Mathematically, E
y
and H
z
should be continuous at x =d.
E
y
is continuous at x =d if
B
c
= Acos(pd ); B
s
= Acos(pd +).
Since H
z
dE
y
/dx, dE
y
/dx should also be continuous at x =d:
pAsin(pd ) = q
1
B
c
, pAsin(pd +) = q
2
B
s
.
Eliminating A, B
c
, B
s
from these equations, must satisfy
tan(pd ) = q
1
/p, tan(pd +) = q
2
/p
17/253

Back
Close
TE Modes
Boundary conditions are satised when
pd = tan
1
(q
1
/p) +m
1
, pd + = tan
1
(q
2
/p) +m
2

Adding and subtracting these equations, we obtain

2 = m tan
1
(q
1
/p) +tan
1
(q
2
/p)
2pd = m +tan
1
(q
1
/p) +tan
1
(q
2
/p)
The last equation is called the eigenvalue equation.
Multiple solutions for m = 0, 1, 2, . . . are denoted by TE
m
.
Eective index of each TE mode is n = /k
0
.
18/253

Back
Close
TM Modes
Same procedure is used to obtain TM modes.
Solution for H
y
has the same form in three layers.
Continuity of E
z
requires that n
2
(dH
y
/dx) be continuous
at x =d.
Since n is dierent on the two sides of each interface,
eigenvalue equation is modied to become
2pd = m +tan
1
_
n
2
1
q
1
n
2
c
p
_
+tan
1
_
n
2
1
q
2
n
2
s
p
_
.
Multiple solutions for dierent values of m.
These are labelled as TM
m
modes.
19/253

Back
Close
TE Modes of Symmetric Waveguides
For symmetric waveguides n
c
= n
s
.
Using q
1
= q
2
q, TE modes satisfy
q = ptan(pd m/2).
Dene a dimensionless parameter
V = d
_
p
2
+q
2
= k
0
d
_
n
2
1
n
2
s
,
If we use u = pd, the eigenvalue equation can be written as
_
V
2
u
2
= utan(um/2).
For given values of V and m, this equation is solved to nd p =u/d.
20/253

Back
Close
TE Modes of Symmetric Waveguides
Eective index n = /k
0
= (n
2
1
p
2
/k
2
0
)
1/2
.
Using 2 = m tan
1
(q
1
/p) +tan
1
(q
2
/p)
with q
1
= q
2
, phase = m/2.
Spatial distribution of modes is found to be
E
y
(x) =
_
B

exp[q([x[ d)]; [x[ > d,

Acos(px m/2) ; [x[ d,
where B

= Acos(pd m/2) and the lower sign is chosen for

x < 0.
Modes with even values of m are symmetric around
x = 0 (even modes).
Modes with odd values of m are antisymmetric around
x = 0 (odd modes).
21/253

Back
Close
TM Modes of Symmetric Waveguides
We can follow the same procedure for TM modes.
Eigenvalue equation for TM modes:
(n
1
/n
s
)
2
q = ptan(pd m/2).
TM modes can also be divided into even and odd modes.
22/253

Back
Close
Symmetric Waveguides
TE
0
and TM
0
modes have no nodes within the core.
They are called the fundamental modes of a planar waveguide.
Number of modes supported by a waveguide depends on the V
parameter.
A mode ceases to exist when q =0 (no longer conned to the core).
This occurs for both TE and TM modes when V =V
m
= m/2.
Number of modes = Largest value of m for which V
m
>V.
A waveguide with V = 10 supports 7 TE and 7 TM modes
(V
6
= 9.42 but V
7
exceeds 10).
A waveguide supports a single TE and a single TM mode when
V < /2 (single-mode condition).
23/253

Back
Close
Modes of Asymmetric Waveguides
We can follow the same procedure for n
c
,= n
s
.
Eigenvalue equation for TE modes:
2pd = m +tan
1
(q
1
/p) +tan
1
(q
2
/p)
Eigenvalue equation for TM modes:
2pd = m +tan
1
_
n
2
1
q
1
n
2
c
p
_
+tan
1
_
n
2
1
q
2
n
2
s
p
_
Constants p, q
1
, and q
2
are dened as
p
2
= n
2
1
k
2
0

2
, q
2
1
=
2
n
2
c
k
2
0
, q
2
2
=
2
n
2
s
k
2
0
.
Each solution for corresponds to a mode with eective index
n = /k
0
.
If n
1
> n
s
> n
c
, guided modes exist as long as n
1
> n > n
s
.
24/253

Back
Close
Modes of Asymmetric Waveguides
Useful to introduce two normalized parameters as
b =
n
2
n
2
s
n
2
1
n
2
s
, =
n
2
s
n
2
c
n
2
1
n
2
s
.
b is a normalized propagation constant (0 < b < 1).
Parameter provides a measure of waveguide asymmetry.
Eigenvalue equation for TE modes in terms V, b, :
2V

1b = m +tan
1
_
b
1b
+tan
1
_
b+
1b
.
Its solutions provide universal dispersion curves.
25/253

Back
Close
Modes of Asymmetric Waveguides
0 1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Normalized frequency, V
N
o
r
m
a
l
i
z
e
d

p
r
o
p
a
g
a
t
i
o
n

c
o
n
s
t
a
n
t
,

b
m = 0
1
2
3
4
5
Solid lines ( = 5); dashed lines ( = 0).
26/253

Back
Close
Mode-Cutoff Condition
Cuto condition corresponds to the value of V for which mode
ceases to decay exponentially in one of the cladding layers.
It is obtained by setting b = 0 in eigenvalue equation:
V
m
(TE) =
m
2
+
1
2
tan
1

.
Eigenvalue equation for the TM modes:
2V

1b = m +tan
1
_
n
2
1
n
2
s
_
b
1b
_
+tan
1
_
n
2
1
n
2
c
_
b+
1b
_
.
The cuto condition found by setting b = 0:
V
m
(TM) =
m
2
+
1
2
tan
1
_
n
2
1
n
2
c

_
.
27/253

Back
Close
Mode-Cutoff Condition
For a symmetric waveguide ( =0), these two conditions reduce to
a single condition, V
m
= m/2.
TE and TM modes for a given value of m have the same cuto.
A single-mode waveguide is realized if V parameter of the waveguide
satises
V k
0
d
_
n
2
1
n
2
s
<

2
Fundamental mode always exists for a symmetric waveguide.
An asymmetric waveguide with 2V < tan
1

does not support

any bounded mode.
28/253

Back
Close
Spatial Distribution of Modes
E
y
(x) =
_
_
_
B
c
exp[q
1
(x d)]; x > d,
Acos(px ) ; [x[ d
B
s
exp[q
2
(x +d)] ; x <d,
Boundary conditions: B
c
= Acos(pd ), B
s
= Acos(pd +)
A is related to total power P =
1
2
_

z (EH)dx:
P =

2
0
_

[E
y
(x)[
2
dx =
A
2
4
0
_
2d +
1
q
1
+
1
q
2
_
.
Fraction of power propagating inside the waveguide layer:
=
_
d
d
[E
y
(x)[
2
dx
_

[E
y
(x)[
2
dx
=
2d +sin
2
(pd )/q
1
+sin
2
(pd +)/q
2
2d +1/q
1
+1/q
2
.
For fundamental mode 1 when V /2.
29/253

Back
Close
Rectangular Waveguides
Rectangular waveguide connes light in both x and y dimensions.
The high-index region in the middle core layer has a nite width 2w
and is surrounded on all sides by lower-index materials.
Refractive index can be dierent on all sides of a rectangular waveg-
uide.
30/253

Back
Close
Modes of Rectangular Waveguides
To simplify the analysis, all shaded cladding regions are assumed to
have the same refractive index n
c
.
A numerical approach still necessary for an exact solution.
Approximate analytic solution possible with two simplications;
Marcatili, Bell Syst. Tech. J. 48, 2071 (1969).
Ignore boundary conditions associated with hatched regions.
Assume core-cladding index dierences are small on all sides.
Problem is then reduced to solving two planar-waveguide problems
in the x and y directions.
31/253

Back
Close
Modes of Rectangular Waveguides
One can nd TE- and TM-like modes for which either E
z
or H
z
is
nearly negligible compared to other components.
Modes denoted as E
x
mn
and E
y
mn
obtained by solving two coupled
eigenvalue equations.
2p
x
d = m +tan
1
_
n
2
1
q
2
n
2
2
p
x
_
+tan
1
_
n
2
1
q
4
n
2
4
p
x
_
,
2p
y
w = n +tan
1
_
q
3
p
y
_
+tan
1
_
q
5
p
y
_
,
p
2
x
= n
2
1
k
2
0

2
p
2
y
, p
2
y
= n
2
1
k
2
0

2
p
2
x
,
q
2
2
=
2
+ p
2
y
n
2
2
k
2
0
, q
2
4
=
2
+ p
2
y
n
2
4
k
2
0
,
q
2
3
=
2
+ p
2
x
n
2
3
k
2
0
, q
2
5
=
2
+ p
2
x
n
2
5
k
2
0
,
32/253

Back
Close
Effective-Index Method
Eective-index method appropriate when thickness of a rectangular
waveguide is much smaller than its width (d w).
Planar waveguide problem in the x direction is solved rst to obtain
the eective mode index n
e
(y).
n
e
is a function of y because of a nite waveguide width.
In the y direction, we use a waveguide of width 2w such that n
y
=n
e
if [y[ < w but equals n
3
or n
5
outside of this region.
Single-mode condition is found to be
V
x
= k
0
d
_
n
2
1
n
2
4
< /2, V
y
= k
0
w
_
n
2
e
n
2
5
< /2
33/253

Back
Close
Design of Rectangular Waveguides
In (g) core layer is covered with two metal stripes.
Losses can be reduced by using a thin buer layer (h).
34/253

Back
Close
Materials for Waveguides
Semiconductor Waveguides: GaAs, InP, etc.
Electro-Optic Waveguides: mostly LiNbO
3
.
Glass Waveguides: silica (SiO
2
), SiON.
Silica-on-silicon technology
Laser-written waveguides
Silicon-on-Insulator Technology
Polymers Waveguides: Several organic
polymers
35/253

Back
Close
Semiconductor Waveguides
Useful for semiconductor lasers, modulators, and photodetectors.
Semiconductors allow fabrication
of electrically active devices.
Semiconductors belonging to III
V Group often used.
Two semiconductors with dier-
ent refractive indices needed.
They must have dierent
bandgaps but same lattice
constant.
Nature does not provide such
semiconductors.
36/253

Back
Close
Ternary and Quaternary Compounds
A fraction of the lattice sites in a binary semiconductor (GaAs, InP,
etc.) is replaced by other elements.
Ternary compound Al
x
Ga
1x
As is made by replacing a fraction x of
Ga atoms by Al atoms.
Bandgap varies with x as
E
g
(x) = 1.424+1.247x (0 < x < 0.45).
Quaternary compound In
1x
Ga
x
As
y
P
1y
useful in the wavelength
range 1.1 to 1.6 m.
For matching lattice constant to InP substrate, x/y = 0.45.
Bandgap varies with y as E
g
(y) = 1.350.72y +0.12y
2
.
37/253

Back
Close
Fabrication Techniques
Epitaxial growth of multiple layers on a base substrate (GaAs or InP).
Three primary techniques:
Liquid-phase epitaxy (LPE)
Vapor-phase epitaxy (VPE)
Molecular-beam epitaxy (MBE)
VPE is also called chemical-vapor
deposition (CVD).
Metal-organic chemical-vapor deposition (MOCVD) is often used in
practice.
38/253

Back
Close
Quantum-Well Technology
Thickness of the core layer plays a central role.
If it is small enough, electrons and holes act as if they are conned
to a quantum well.
Connement leads to quantization of energy bands into subbands.
Joint density of states acquires a staircase-like structure.
Useful for making modern quantum-well, quantum wire, and
quantum-dot lasers.
in MQW lasers, multiple core layers (thickness 510 nm) are
separated by transparent barrier layers.
Use of intentional but controlled strain improves performance
in strained quantum wells.
39/253

Back
Close
Doped Semiconductor Waveguides
To build a laser, one needs to inject current into the core layer.
This is accomplished through a pn junction formed by
making cladding layers p- and n-types.
n-type material requires a dopant with an extra electron.
p-type material requires a dopant with one less electron.
Doping creates free electrons or holes within a semiconductor.
Fermi level lies in the middle of bandgap for undoped
(intrinsic) semiconductors.
In a heavily doped semiconductor, Fermi level lies inside
the conduction or valence band.
40/253

Back
Close
pn junctions
Fermi level continuous across the
pn junction in thermal equilib-
rium.
A built-in electric eld separates
electrons and holes.
Forward biasing reduces the built-
in electric eld.
An electric current begins to ow:
I = I
s
[exp(qV/k
B
T) 1].
Recombination of electrons and
holes generates light.
(b)
(a)
41/253

Back
Close
Electro-Optic Waveguides
Use Pockels eect to change refractive index of the core layer with
an external voltage.
Common electro-optic materials: LiNbO
3
, LiTaO
3
, BaTiO
3
.
LiNbO
3
used commonly for making optical modulators.
For any anisotropic material D
i
=
0

3
j=1

i j
E
j
.
Matrix
i j
can be diagonalized by rotating the coordinate system
along the principal axes.
Impermeability tensor
i j
= 1/
i j
describes changes induced by an
external eld as
i j
(E
a
) =
i j
(0) +

k
r
i jk
E
a
k
.
Tensor r
i jk
is responsible for the electro-optic eect.
42/253

Back
Close
Lithium Niobate Waveguides
LiNbO
3
waveguides do not require an epitaxial growth.
A popular technique employs diusion of metals into a LiNbO
3
sub-
strate, resulting in a low-loss waveguide.
The most commonly used element: Titanium (Ti).
Diusion of Ti atoms within LiNbO
3
crystal increases refractive
index and forms the core region.
Out-diusion of Li atoms from substrate should be avoided.
Surface atness critical to ensure a uniform waveguide.
43/253

Back
Close
LiNbO
3
Waveguides
A proton-exchange technique is also used for LiNbO
3
waveguides.
A low-temperature process (200

C) in which Li ions are replaced

with protons when the substrate is placed in an acid bath.
Proton exchange increases the extraordinary part of refractive index
but leaves the ordinary part unchanged.
Such a waveguide supports only TM modes and is useful for some
applications because of its polarization selectivity.
High-temperature annealing used to stabilizes the index dierence.
Accelerated aging tests predict a lifetime of over 25 years at a tem-
perature as high as 95

C.
44/253

Back
Close
LiNbO
3
Waveguides
Electrodes fabricated directly on the surface of wafer (or on
an optically transparent buer layer.
An adhesion layer (typically Ti) rst deposited to ensure that
metal sticks to LiNbO
3
.
Photolithography used to dene the electrode pattern.
45/253

Back
Close
Silica Glass Waveguides
Silica layers deposited on top of a Si substrate.
Employs the technology developed for integrated circuits.
Fabricated using ame hydrolysis with reactive ion etching.
Two silica layers are rst deposited using ame hydrolysis.
Top layer converted to core by doping it with germania.
Both layers solidied by heating at 1300

C (consolidation process).
Photolithography used to etch patterns on the core layer.
Entire structure covered with a cladding formed using ame hydrol-
ysis. A thermo-optic phase shifter often formed on top.
46/253

Back
Close
Silica-on-Silicon Technique
Steps used to form silica waveguides on top of a Si Substrate
47/253

Back
Close
Silica Waveguide properties
Silica-on-silicon technology produces uniform waveguides.
Losses depend on the core-cladding index dierence
= (n
1
n
2
)/n
1
.
Losses are low for small values of (about 0.017 dB/cm
for = 0.45%).
Higher values of often used for reducing device length.
Propagation losses 0.1 dB/cm for = 2%.
Planar lightwave circuits: Multiple waveguides and optical
components integrated over the same silicon substrate.
Useful for making compact WDM devices (55 cm
2
).
Large insertion losses when a PLC is connected to optical bers.
48/253

Back
Close
Packaged PLCs
Package design for minimizing insertion losses.
Fibers inserted into V-shaped grooves formed on a glass substrate.
Glass substrate connected to the PLC chip using an adhesive.
A glass plate placed on top of V grooves is bonded to the PLC chip
49/253

Back
Close
with the same adhesive.
50/253

Back
Close
Silicon Oxynitride Waveguides
Employ Si substrate but use SiON for the core layer.
SiON alloy is made by combining SiO
2
with Si
3
N
4
, two dielectrics
with refractive indices of 1.45 and 2.01.
Refractive index of SiON layer can vary from 1.452.01.
SiON lm deposited using plasma-enhanced chemical vapor
deposition (SiH
4
combined with N
2
O and NH
3
).
Low-pressure chemical vapor deposition also used
(SiH
2
Cl
2
combined with O
2
and NH
3
).
Photolithography pattern formed on a 200-nm-thick chromium layer.
Propagation losses typically <0.2 dB/cm.
51/253

Back
Close
Laser-Written Waveguides
CW or pulsed light from a laser used for writing waveguides in
silica and other glasses.
Photosensitivity of germanium-doped silica exploited to enhance
refractive index in the region exposed to a UV laser.
Absorption of 244-nm light from a KrF laser changes refractive index
by 10
4
only in the region exposed to UV light.
Index changes >10
3
can be realized with a 193-nm ArF laser.
A planar waveguide formed rst through CVD, but core layer is
doped with germania.
An UV beam focused to 1 m scanned slowly to enhance n se-
lectively. UV-written sample then annealed at 80

C.
52/253

Back
Close
Laser-Written Waveguides
Femtosecond pulses from a Ti:sapphire laser can be used to write
waveguides in bulk glasses.
Intense pulses modify the structure of silica through
multiphoton absorption.
Refractive-index changes 10
2
are possible.
53/253

Back
Close
Silicon-on-Insulator Technology
Core waveguide layer is made of Si (n
1
= 3.45).
A silica layer under the core layer is used for lower cladding.
Air on top acts as the top cladding layer.
Tightly conned waveguide mode because of large index dierence.
Silica layer formed by implanting oxygen, followed with annealing.
54/253

Back
Close
Polymer Waveguides
Polymers such as halogenated acrylate, uorinated polyimide, and
deuterated polymethylmethacrylate (PMMA) have been used.
Polymer lms can be fabricated on top of Si, glass, quartz,
or plastic through spin coating.
Photoresist layer on top used for reactive ion etching of the core
layer through a photomask.
55/253

Back
Close
Optical Fibers
Contain a central core surrounded by a lower-index cladding
Two-dimensional waveguides with cylindrical symmetry
Graded-index bers: Refractive index varies inside the core
56/253

Back
Close
Total internal reection
Refraction at the airglass interface: n
0
sin
i
= n
1
sin
r
Total internal reection at the core-cladding interface
if >
c
= sin
1
(n
2
/n
1
).
Numerical Aperture: Maximum angle of incidence
n
0
sin
max
i
= n
1
sin(/2
c
) = n
1
cos
c
=
_
n
2
1
n
2
2
57/253

Back
Close
Modal Dispersion
Multimode bers suer from modal dispersion.
Shortest path length L
min
= L (along the fiber axis).
Longest path length for the ray close to the critical angle
L
max
= L/sin
c
= L(n
1
/n
2
).
Pulse broadening: T = (L
max
L
min
)(n
1
/c).
Modal dispersion: T/L = n
2
1
/(n
2
c).
Limitation on the bit rate
T < T
B
= 1/B; BT < 1; BL <
n
2
c
n
2
1

.
Single-mode bers essential for high performance.
58/253

Back
Close
Graded-Index Fibers
Refractive index n() =
_
n
1
[1(/a)

]; < a,
n
1
(1) = n
2
; a.
Ray path obtained by solving
d
2

dz
2
=
1
n
dn
d
.
For = 2, =
0
cos(pz) +(
/
0
/p)sin(pz).
All rays arrive simultaneously at periodic intervals.
Limitation on the Bit Rate: BL <
8c
n
1

2
.
59/253

Back
Close
Fiber Design
Core doped with GeO
2
; cladding with uorine.
Index prole rectangular for standard bers.
Triangular index prole for dispersion-shifted bers.
Raised or depressed cladding for dispersion control.
60/253

Back
Close
Silica Fibers
Two-Stage Fabrication
Preform: Length 1 m, diameter 2 cm; correct index prole.
Preform is drawn into ber using a draw tower.
Preform Fabrication Techniques
Modied chemical vapor deposition (MCVD).
Outside vapor deposition (OVD).
Vapor Axial deposition (VAD).
61/253

Back
Close
Fiber Draw Tower
62/253

Back
Close
Plastic Fibers
Multimode bers (core diameter as large as 1 mm).
Large NA results in high coupling eciency.
Use of plastics reduces cost but loss exceeds 50 dB/km.
Useful for data transmission over short distances (<1 km).
10-Gb/s signal transmitted over 0.5 km (1996 demo).
Ideal solution for transferring data between computers.
Commonly used polymers:
polymethyl methacrylate (PMMA), polystyrene
polycarbonate, poly(peruoro-butenylvinyl) ether
63/253

Back
Close
Plastic Fibers
Preform made with the interfacial gel polymerization method.
A cladding cylinder is lled with a mixture of monomer (same
used for cladding polymer), index-increasing dopant, a chemical for
initiating polymerization, and a chain-transfer agent.
Cylinder heated to a 95

C and rotated on its axis for a period of

up to 24 hours.
Core polymerization begins near cylinder wall.
Dopant concentration increases toward core center.
This technique automatically creates a gradient in the core index.
64/253

Back
Close
Microstructure Fibers
New types of bers with air holes in cladding region.
Air holes reduce the index of the cladding region.
Narrow core (2 m or so) results in tighter mode connement.
Air-core bers guide light through the photonic-crystal eect.
Preform made by stacking silica tubes in a hexagonal pattern.
65/253

Back
Close
Fiber Modes
Maxwells equations in the Fourier domain lead to

E+n
2
()k
2
0

E = 0.
n = n
1
inside the core but changes to n
2
in the cladding.
Useful to work in cylindrical coordinates , , z.
Common to choose E
z
and H
z
as independent components.
Equation for E
z
in cylindrical coordinates:

2
E
z

2
+
1

E
z

+
1

2
E
z

2
+

2
E
z
z
2
+n
2
k
2
0
E
z
= 0.
H
z
satises the same equation.
66/253

Back
Close
Fiber Modes (cont.)
Use the method of separation of variables:
E
z
(, , z) = F()()Z(z).
We then obtain three ODEs:
d
2
Z/dz
2
+
2
Z = 0,
d
2
/d
2
+m
2
= 0,
d
2
F
d
2
+
1

dF
d
+
_
n
2
k
2
0

m
2

2
_
F = 0.
and m are two constants (m must be an integer).
First two equations can be solved easily to obtain
Z(z) = exp(iz), () = exp(im).
F() satises the Bessel equation.
67/253

Back
Close
Fiber Modes (cont.)
General solution for E
z
and H
z
:
E
z
=
_
AJ
m
(p)exp(im)exp(iz) ; a,
CK
m
(q)exp(im)exp(iz); > a.
H
z
=
_
BJ
m
(p)exp(im)exp(iz) ; a,
DK
m
(q)exp(im)exp(iz); > a.
p
2
= n
2
1
k
2
0

2
, q
2
=
2
n
2
2
k
2
0
.
Other components can be written in terms of E
z
and H
z
:
E

=
i
p
2
_

E
z

+
0

H
z

_
, E

=
i
p
2
_

E
z

H
z

_
,
H

=
i
p
2
_

H
z

0
n
2

E
z

_
, H

=
i
p
2
_

H
z

+
0
n
2

E
z

_
.
68/253

Back
Close
Eigenvalue Equation
Boundary conditions: E
z
, H
z
, E

, and H

should be continuous
across the corecladding interface.
Continuity of E
z
and H
z
at = a leads to
AJ
m
(pa) =CK
m
(qa), BJ
m
(pa) = DK
m
(qa).
Continuity of E

and H

provides two more equations.

Four equations lead to the eigenvalue equation
_
J
/
m
(pa)
pJ
m
(pa)
+
K
/
m
(qa)
qK
m
(qa)
__
J
/
m
(pa)
pJ
m
(pa)
+
n
2
2
n
2
1
K
/
m
(qa)
qK
m
(qa)
_
=
m
2
a
2
_
1
p
2
+
1
q
2
__
1
p
2
+
n
2
2
n
2
1
1
q
2
_
p
2
= n
2
1
k
2
0

2
, q
2
=
2
n
2
2
k
2
0
.
69/253

Back
Close
Eigenvalue Equation
Eigenvalue equation involves Bessel functions and their derivatives.
It needs to be solved numerically.
Noting that p
2
+q
2
= (n
2
1
n
2
2
)k
2
0
, we introduce the dimensionless
V parameter as
V = k
0
a
_
n
2
1
n
2
2
.
Multiple solutions for for a given value of V.
Each solution represents an optical mode.
Number of modes increases rapidly with V parameter.
Eective mode index n =/k
0
lies between n
1
and n
2
for all bound
modes.
70/253

Back
Close
Effective Mode Index
Useful to introduce a normalized quantity
b = ( nn
2
)/(n
1
n
2
), (0 < b < 1).
Modes quantied through () or b(V).
71/253

Back
Close
Classication of Fiber Modes
In general, both E
z
and H
z
are nonzero (hybrid modes).
Multiple solutions occur for each value of m.
Modes denoted by HE
mn
or EH
mn
(n =1, 2, . . .) depending on whether
H
z
or E
z
dominates.
TE and TM modes exist for m = 0 (called TE
0n
and TM
0n
).
Setting m = 0 in the eigenvalue equation, we obtain two equations
_
J
/
m
(pa)
pJ
m
(pa)
+
K
/
m
(qa)
qK
m
(qa)
_
= 0,
_
J
/
m
(pa)
pJ
m
(pa)
+
n
2
2
n
2
1
K
/
m
(qa)
qK
m
(qa)
_
= 0
These equations govern TE
0n
and TM
0n
modes of ber.
72/253

Back
Close
Linearly Polarized Modes
Eigenvalue equation simplied considerably for weakly guiding bers
(n
1
n
2
1):
_
J
/
m
(pa)
pJ
m
(pa)
+
K
/
m
(qa)
qK
m
(qa)
_
2
=
m
2
a
2
_
1
p
2
+
1
q
2
_
2
.
Using properties of Bessel functions, the eigenvalue equation can
be written in the following compact form:
p
J
l1
(pa)
J
l
(pa)
=q
K
l1
(qa)
K
l
(qa)
,
where l = 1 for TE and TM modes, l = m1 for HE modes, and
l = m+1 for EH modes.
TE
0,n
and TM
0,n
modes are degenerate. Also, HE
m+1,n
and EH
m1,n
are degenerate in this approximation.
73/253

Back
Close
Linearly Polarized Modes
Degenerate modes travel at the same velocity through ber.
Any linear combination of degenerate modes will travel without
change in shape.
Certain linearly polarized combinations produce LP
mn
modes.
LP
0n
is composed of of HE
1n
.
LP
1n
is composed of TE
0n
+ TM
0n
+ HE
2n
.
LP
mn
is composed of HE
m+1,n
+ EH
m1,n
.
Historically, LP modes were obtained rst using a simplied analysis
of ber modes.
74/253

Back
Close
Fundamental Fiber Mode
A mode ceases to exist when q = 0 (no decay in the cladding).
TE
01
and TM
01
reach cuto when J
0
(V) = 0.
This follows from their eigenvalue equation
p
J
0
(pa)
J
1
(pa)
=q
K
0
(qa)
K
1
(qa)
after setting q = 0 and pa =V.
Single-mode bers require V < 2.405 (rst zero of J
0
).
They transport light through the fundamental HE
11
mode.
This mode is almost linearly polarized ([E
z
[
2
[E
x
[
2
)
E
x
(, , z) =
_
A[J
0
(p)/J
0
(pa)]e
iz
; a,
A[K
0
(q)/K
0
(qa)]e
iz
; > a.
75/253

Back
Close
Fundamental Fiber Mode
Use of Bessel functions is not always practical.
It is possible to approximate spatial distribution of HE
11
mode
with a Gaussian for V in the range 1 to 2.5.
E
x
(, , z) Aexp(
2
/w
2
)e
iz
.
Spot size w depends on V parameter.
76/253

Back
Close
Single-Mode Properties
Spot size: w/a 0.65+1.619V
3/2
+2.879V
6
.
Mode index:
n = n
2
+b(n
1
n
2
) n
2
(1+b),
b(V) (1.14280.9960/V)
2
.
Connement factor:
=
P
core
P
total
=
_
a
0
[E
x
[
2
d
_

0
[E
x
[
2
d
= 1exp
_

2a
2
w
2
_
.
0.8 for V = 2 but drops to 0.2 for V = 1.
Mode properties completely specied if V parameter is known.
77/253

Back
Close
Fiber Birefringence
Real bers exhibit some birefringence ( n
x
,= n
y
).
Modal birefringence quite small (B
m
=[ n
x
n
y
[ 10
6
).
Beat length: L
B
= /B
m
.
State of polarization evolves periodically.
Birefringence varies randomly along ber length (PMD) because of
stress and core-size variations.
78/253

Back
Close
Fiber Losses
Denition: P
out
= P
in
exp(L), (dB/km) = 4.343.
Material absorption (silica, impurities, dopants)
Rayleigh scattering (varies as
4
)
79/253

Back
Close
Losses of Plastic Fibers
Large absorption losses of plastics result from vibrational modes of
molecular bonds (CC, CO, CH, and OH).
Transition-metal impurities (Fe, Co, Ni, Mn, and Cr) absorb strongly
in the range 0.61.6 m.
Residual water vapors produce strong peak near 1390 nm.
80/253

Back
Close
Fiber Dispersion
Origin: Frequency dependence of the mode index n():
() = n()/c =
0
+
1
(
0
) +
2
(
0
)
2
+ ,
where
0
is the carrier frequency of optical pulse.
Transit time for a ber of length L : T = L/v
g
=
1
L.
Dierent frequency components travel at dierent speeds and arrive
at dierent times at output end (pulse broadening).
81/253

Back
Close
Fiber Dispersion (continued)
Pulse broadening governed by group-velocity dispersion:
T =
dT
d
=
d
d
L
v
g
= L
d
1
d
= L
2
,
where is pulse bandwidth and L is ber length.
GVD parameter:
2
=
_
d
2

d
2
_
=
0
.
Alternate definition: D =
d
d
_
1
v
g
_
=
2c

2

2
.
Limitation on the bit rate: T < T
B
= 1/B, or
B(T) = BL
2
BLD < 1.
82/253

Back
Close
Material Dispersion
Refractive index of of any material is frequency dependent.
Material dispersion governed by the Sellmeier equation
n
2
() = 1+
M

j=1
B
j

2
j

2
j

2
.
83/253

Back
Close
Waveguide Dispersion
Mode index n() = n
1
() n
W
().
Material dispersion D
M
results from n
1
() (index of silica).
Waveguide dispersion D
W
results from n
W
() and depends on
core size and dopant distribution.
Total dispersion D = D
M
+D
W
can be controlled.
84/253

Back
Close
Dispersion in Microstructure Fibers
Air holes in cladding and a small core diameter help to shift ZDWL
in the region near 800 nm.
Waveguide dispersion D
W
is very large in such bers.
Useful for supercontinuum generation using mode-locking pulses
from a Ti:sapphire laser.
85/253

Back
Close
Higher-Order Dispersion
Dispersive eects do not disappear at =
ZD
.
D cannot be made zero at all frequencies within the pulse spectrum.
Higher-order dispersive eects are governed by
the dispersion slope S = dD/d.
S can be related to third-order dispersion
3
as
S = (2c/
2
)
2

3
+(4c/
3
)
2
.
At =
ZD
,
2
= 0, and S is proportional to
3
.
86/253

Back
Close
Commercial Fibers
Fiber Type and A
eff

ZD
D (C band) Slope S
Trade Name (m
2
) (nm) ps/(km-nm) ps/(km-nm
2
)
Corning SMF-28 80 13021322 16 to 19 0.090
Lucent AllWave 80 13001322 17 to 20 0.088
Alcatel ColorLock 80 13001320 16 to 19 0.090
Corning Vascade 101 13001310 18 to 20 0.060
TrueWave-RS 50 14701490 2.6 to 6 0.050
Corning LEAF 72 14901500 2 to 6 0.060
TrueWave-XL 72 15701580 1.4 to 4.6 0.112
Alcatel TeraLight 65 14401450 5.5 to 10 0.058
87/253

Back
Close
Polarization-Mode Dispersion
Real bers exhibit some birefringence ( n
x
,= n
y
).
Orthogonally polarized components of a pulse travel at dierent
speeds. The relative delay is given by
T =

L
v
gx

L
v
gy

= L[
1x

1y
[ = L(
1
).
Birefringence varies randomly along fiber length (PMD) because of
stress and core-size variations.
RMS Pulse broadening:

T
(
1
)
_
2l
c
L D
p

L.
PMD parameter D
p
0.0110 ps/

km
PMD can degrade the system performance considerably (especially
for old bers).
88/253

Back
Close
Pulse Propagation Equation
Optical Field at frequency at z = 0:

E(r, ) = xF(x, y)

B(0, )exp(iz).
Optical eld at a distance z:

B(z, ) =

B(0, )exp(iz).
Expand () is a Taylor series around
0
:
() = n()

c

0
+
1
() +

2
2
()
2
+

3
6
()
3
.
Introduce Pulse envelope:
B(z, t) = A(z, t)exp[i(
0
z
0
t)].
89/253

Back
Close
Pulse Propagation Equation
Pulse envelope is obtained using
A(z, t) =
1
2
_

d()

A(0, )exp
_
i
1
z +
i
2

2
z()
2
+
i
6

3
z()
3
i()t
_
.
Calculate A/z and convert to time domain by replacing
with i(A/t).
Final equation:
A
z
+
1
A
t
+
i
2
2

2
A
t
2

3
6

3
A
t
3
= 0.
With the transformation t
/
=t
1
z and z
/
= z, it reduces to
A
z
/
+
i
2
2

2
A
t
/2

3
6

3
A
t
/3
= 0.
90/253

Back
Close
Pulse Propagation Equation
If we neglect third-order dispersion, pulse evolution is governed by
A
z
+
i
2
2

2
A
t
2
= 0.
Compare with the paraxial equation governing diraction:
2ik
A
z
+

2
A
x
2
= 0.
Slit-diraction problem identical to pulse propagation problem.
The only dierence is that
2
can be positive or negative.
Many results from diraction theory can be used for pulses.
A Gaussian pulse should spread but remain Gaussian in shape.
91/253

Back
Close
Major Nonlinear Effects
Self-Phase Modulation (SPM)
Cross-Phase Modulation (XPM)
Four-Wave Mixing (FWM)
Stimulated Brillouin Scattering (SBS)
Stimulated Raman Scattering (SRS)
Origin of Nonlinear Eects in Optical Fibers
Third-order nonlinear susceptibility
(3)
.
Real part leads to SPM, XPM, and FWM.
Imaginary part leads to SBS and SRS.
92/253

Back
Close
Self-Phase Modulation (SPM)
Refractive index depends on intensity as
n
/
j
= n
j
+ n
2
I(t).
n
2
= 2.610
20
m
2
/W for silica bers.
Propagation constant:
/
= +k
0
n
2
P/A
eff
+P.
Nonlinear parameter: = 2 n
2
/(A
eff
).
Nonlinear Phase shift:

NL
=
_
L
0
(
/
)dz =
_
L
0
P(z)dz = P
in
L
eff
.
Optical eld modies its own phase (SPM).
Phase shift varies with time for pulses (chirping).
93/253

Back
Close
SPM-Induced Chirp
SPM-induced chirp depends on the pulse shape.
Gaussian pulses (m = 1): Nearly linear chirp across the pulse.
Super-Gaussian pulses (m = 1): Chirping only near pulse edges.
SPM broadens spectrum of unchirped pulses; spectral narrowing
possible in the case of chirped pulses.
94/253

Back
Close
Nonlinear Schr odinger Equation
Nonlinear eects can be included by adding a nonlinear term to the
equation used earlier for dispersive eects.
This equation is known as the Nonlinear Schr odinger Equation:
A
z
+
i
2
2

2
A
t
2
= i[A[
2
A.
Nonlinear parameter: = 2 n
2
/(A
eff
).
Fibers with large A
eff
help through reduced .
Known as large eective-area ber or LEAF.
Nonlinear eects leads to formation of optical solitons.
95/253

Back
Close
Cross-Phase Modulation (XPM)
Refractive index seen by one wave depends on the intensity of other
copropagating channels.
E(r, t) = A
a
(z, t)F
a
(x, y)exp(i
0a
z i
a
t)
+A
b
(z, t)F
b
(x, y)exp(i
0b
z i
b
t)],
Propagation constants are found to be modied as

/
a
=
a
+
a
([A
a
[
2
+2[A
b
[
2
),
/
b
=
b
+
b
([A
b
[
2
+2[A
a
[
2
).
Nonlinear phase shifts produced become

NL
a
=
a
L
eff
(P
a
+2P
b
),
NL
b
=
b
L
eff
(P
b
+2P
a
).
The second term is due to XPM.
96/253

Back
Close
Impact of XPM
In the case of a WDM system, total nonlinear phase shift is

NL
j
= L
eff
_
P
j
+2

m,=j
P
m
_
.
Phase shift varies from bit to bit depending on the bit pattern in
neighboring channels.
It leads to interchannel crosstalk and aects system performance
considerably.
XPM is also benecial for applications such as optical switching,
wavelength conversion, etc.
Mathematically, XPM eects are governed by two coupled NLS
equations.
97/253

Back
Close
Four-Wave Mixing
FWM converts two photons from one or two pump beams into two
new frequency-shifted photons.
Energy conservation:
1
+
2
=
3
+
4
.
Degenerate FWM: 2
1
=
3
+
4
.
Momentum conservation or phase matching is required.
FWM eciency governed by phase mismatch:
= (
3
) +(
4
) (
1
) (
2
).
In the degenerate case (
1
=
2
),
3
=
1
+, and
4
=
1
.
Expanding in a Taylor series, =
2

2
.
FWM becomes important for WDM systems designed with low-
dispersion bers.
98/253

Back
Close
FWM: Good or Bad?
FWM leads to interchannel crosstalk in WDM systems.
It can be avoided through dispersion management.
On the other hand . . .
FWM can be used benecially for
Parametric amplication
Optical phase conjugation
Demultiplexing of OTDM channels
Wavelength conversion of WDM channels
Supercontinuum generation
99/253

Back
Close
Brillouin Scattering
Scattering of light from acoustic waves (electrostriction).
Energy and momentum conservation laws require

B
=
p

s
and k
A
=k
p
k
s
.
Brillouin shift:
B
=[k
A
[v
A
= 2v
A
[k
p
[ sin(/2).
Only possibility: = for single-mode bers
(backward propagating Stokes wave).
Using k
p
= 2 n/
p
,
B
=
B
/2 = 2 nv
A
/
p
.
With v
A
= 5.96 km/s and n = 1.45,
B
11 GHz near 1.55 m.
Stokes wave grows from noise.
Not a very ecient process at low pump powers.
100/253

Back
Close
Stimulated Brillouin Scattering
Becomes a stimulated process at high input power levels.
Governed by two coupled equations:
dI
p
dz
=g
B
I
p
I
s

p
I
p
,
dI
s
dz
= +g
B
I
p
I
s

s
I
s
.
Brillouin gain has a narrow Lorentzian spectrum ( 20 MHz).
101/253

Back
Close
SBS Threshold
Threshold condition: g
B
P
th
L
eff
/A
eff
21.
Eective ber length: L
eff
= [1exp(L)]/.
Eective core area: A
eff
5080 m
2
.
Peak Brillouin gain: g
B
510
11
m/W.
Low threshold power for long bers (5 mW).
Most of the power reected backward after the SBS threshold.
Threshold can be increased using
Phase modulation at frequencies >0.1 GHz.
Sinusoidal strain along the ber.
Nonuniform core radius or dopant density.
102/253

Back
Close
Stimulated Raman Scattering
Scattering of light from vibrating molecules.
Scattered light shifted in frequency.
Raman gain spectrum extends over 40 THz.
Raman shift at Gain peak:
R
=
p

s
13 THz).
(a)
(b)
103/253

Back
Close
SRS Threshold
SRS governed by two coupled equations:
dI
p
dz
=g
R
I
p
I
s

p
I
p
dI
s
dz
= g
R
I
p
I
s

s
I
s
.
Threshold condition: g
R
P
th
L
eff
/A
eff
16.
Peak Raman gain: g
R
610
14
m/W near 1.5 m.
Threshold power relatively large ( 0.6 W).
SRS is not of concern for single-channel systems.
Leads to interchannel crosstalk in WDM systems.
104/253

Back
Close
Fiber Components
Fibers can be used to make many optical components.
Passive components
Directional Couplers
Fiber Gratings
Fiber Interferometers
Isolators and Circulators
Active components
Doped-Fiber Ampliers
Raman and Parametric Ampliers
CW and mode-locked Fiber Lasers
105/253

Back
Close
Directional Couplers
Port 1
Port 2
Port 1
Port 2
Core 1
Core 2
Coupling region
Four-port devices (two input and two output ports).
Output can be split in two dierent directions;
hence the name directional couplers.
Can be fabricated using bers or planar waveguides.
Two waveguides are identical in symmetric couplers.
Evanescent coupling of modes in two closely spaced waveguides.
Overlapping of modes in the central region leads to power transfer.
106/253

Back
Close
Theory of Directional Couplers
Coupled-mode theory commonly used for couplers.
Begin with the Helmholtz equation
2

E+ n
2
k
2
0

E = 0.
n(x, y) =n
0
everywhere except in the region occupied by two cores.
Approximate solution:

E(r, ) e[

A
1
(z, )F
1
(x, y) +

A
2
(z, )F
2
(x, y)]e
iz
.
F
m
(x, y) corresponds to the mode supported by the each waveguide:

2
F
m
x
2
+

2
F
m
y
2
+[n
2
m
(x, y)k
2
0

2
m
]F
m
= 0.
A
1
and A
2
vary with z because of the mode overlap.
107/253

Back
Close
Coupled-Mode Equations
Coupled-mode theory deals with amplitudes A
1
and A
2
.
We substitute assumed solution in Helmholtz equation, multiply by
F

1
or F

2
, and integrate over xy plane to obtain
d

A
1
dz
= i(

1
)

A
1
+i
12

A
2
,
d

A
2
dz
= i(

2
)

A
2
+i
21

A
1
,
Coupling coecient is dened as

mp
=
k
2
0
2
_ _

( n
2
n
2
p
)F

m
F
p
dxdy,
Modes are normalized such that
__

[F
m
(x, y)[
2
dx dy = 1.
108/253

Back
Close
Time-Domain Coupled-Mode Equations
Expand

m
() in a Taylor series around the carrier frequency
0
as

m
() =
0m
+(
0
)
1m
+
1
2
(
0
)
2

2m
+ ,
Replace
0
by i(/t) while taking inverse Fourier transform
A
1
z
+
1
v
g1
A
1
t
+
i
21
2

2
A
1
t
2
= i
12
A
2
+i
a
A
1
,
A
2
z
+
1
v
g2
A
2
t
+
i
22
2

2
A
2
t
2
= i
21
A
1
i
a
A
2
,
where v
gm
1/
1m
and

a
=
1
2
(
01

02
), =
1
2
(
01
+
02
).
For a symmetric coupler,
a
= 0 and
12
=
21
.
109/253

Back
Close
Power-Transfer Characteristics
Consider rst the simplest case of a CW beam incident on one of
the input ports of a coupler.
Setting time-dependent terms to zero we obtain
dA
1
dz
= i
12
A
2
+i
a
A
1
,
dA
2
dz
= i
21
A
1
i
a
A
2
.
Eliminating dA
2
/dz, we obtain a simple equation for A
1
:
d
2
A
1
dz
2
+
2
e
A
1
= 0,
e
=
_

2
+
2
a
( =

12

21
).
General solution when A
1
(0) = A
0
and A
2
(0) = 0:
A
1
(z) = A
0
[cos(
e
z) +i(
a
/
e
)sin(
e
z)],
A
2
(z) = A
0
(i
21
/
e
)sin(
e
z).
110/253

Back
Close
Power-Transfer Characteristics
0 1 2 3 4 5 6
Normalized Distance
0.0
0.5
1.0
P
o
w
e
r

F
r
a
c
t
i
o
n
0.1
1
4
Even though A
2
= 0 at z = 0, some power is transferred to the
second core as light propagates inside a coupler.
Power transfer follows a periodic pattern.
Maximum power transfer occurs for
e
z = m/2.
Coupling length is dened as L
c
= /(2
e
).
111/253

Back
Close
Symmetric Coupler
Maximum power transfer occurs for a symmetric coupler (
a
= 0)
General solution for a symmetric coupler of length L:
A
1
(L) = A
1
(0)cos(L) +iA
2
(0)sin(L)
A
2
(L) = iA
1
(0)sin(L) +A
2
(0)cos(L)
This solution can be written in a matrix form as
_
A
1
(L)
A
2
(L)
_
=
_
cos(L) i sin(L)
i sin(L) cos(L)
__
A
1
(0)
A
2
(0)
_
.
When A
2
(0) = 0 (only one beam injected), output elds become
A
1
(L) = A
1
(0)cos(L), A
2
(L) = iA
2
(0)sin(L)
A coupler acts as a beam splitter; notice 90

phase shift for the

cross port.
112/253

Back
Close
Transfer Matrix of a Coupler
Concept of a transfer matrix useful for couplers because a single
matrix governs all its properties.
Introduce = P
1
(L)/P
0
= cos
2
(L) as a fraction of input power
P
0
remaining in the same port of coupler.
Transfer matrix can then be written as
T
c
=
_

i

1
i

_
.
This matrix is symmetric to ensure that the coupler behaves the
same way if direction of light propagation is reversed.
The 90

phase shift important for many applications.

113/253

Back
Close
Applications of Directional Couplers
Simplest application of a ber coupler is as an optical tap.
If is close to 1, a small fraction of input power is transferred to
the other core.
Another application consists of dividing input power equally between
the two output ports ( =
1
2
).
Coupler length L is chosen such that L = /4 or L = L
c
/2.
Such couplers are referred to as 3-dB couplers.
Couplers with L = L
c
transfer all input power to the cross port.
By choosing coupler length appropriately, power can be divided be-
tween two output ports in an arbitrary manner.
114/253

Back
Close
Coupling Coefcient
Length of a coupler required depends on .
Value of depends on the spacing d between two cores.
For a symmetric coupler, can be approximated as
=
V
2k
0
n
1
a
2
exp[(c
0
+c
1

d +c
2

d
2
)] (

d = d/a).
Constants c
0
, c
1
, and c
2
depend only on V.
Accurate to within 1% for values of V and

d in the range 1.5
V 2.5 and 2

d 4.5.
As an example, 1 cm
1
for

d = 3 but it reduces to 0.01 cm
1
when

d exceeds 5.
115/253

Back
Close
Supermodes of a Coupler
Are there launch conditions for which no power transfer occurs?
Under what conditions

A
1
and

A
2
become z-independent?
d

A
1
dz
= i(

1
)

A
1
+i
12

A
2
,
d

A
2
dz
= i(

2
)

A
2
+i
21

A
1
,
This can occur when the ratio f =

A
2
(0)/

A
1
(0) satises
f =

12
=

21

2
.
This equation determines for supermodes

=
1
2
(

1
+

2
)
_

2
a
+
2
.
116/253

Back
Close
Supermodes of a Coupler
Spatial distribution corresponding to two eigenvalues is given by
F

(x, y) = (1+ f
2

)
1/2
[F
1
(x, y) + f

F
2
(x, y)].
These two specic linear combinations of F
1
and F
2
constitute the
supermodes of a ber coupler.
In the case of a symmetric coupler, f

=1, and supermodes

become even and odd combinations of F
1
and F
2
.
When input conditions are such that a supermode is excited, no
power transfer occurs between two cores of a coupler.
When light is incident on one core, both supermodes are excited.
Two supermodes travel at dierent speeds and develop a relative
phase shift that is responsible for periodic power transfer between
two cores.
117/253

Back
Close
Effects of Fiber Dispersion
Coupled-mode equations for a symmetric coupler:
A
1
z
+
i
2
2

2
A
1
T
2
= iA
2
A
2
z
+
i
2
2

2
A
2
T
2
= iA
1
(1)
GVD eects negligible if coupler length L L
D
= T
2
0
/[
2
[.
GVD has no eect on couplers for which L
D
L
c
.
L
D
exceeds 1 km for T
0
> 1 ps but typically L
c
< 10 m.
GVD eects important only for ultrashort pulses (T
0
< 0.1 ps).
Picosecond pulses behave in the same way as CW beams.
Pulse energy transferred to neighboring core periodically.
118/253

Back
Close
Dispersion of Coupling Coefcient
Frequency dependence of cannot be ignored in all cases:
()
0
+(
0
)
1
+
1
2
(
0
)
2

2
,
Modied coupled-mode equations become
A
1
z
+
1
A
2
T
+
i
2
2

2
A
1
T
2
+
i
2
2

2
A
2
T
2
= i
0
A
2
,
A
2
z
+
1
A
1
T
+
i
2
2

2
A
2
T
2
+
i
2
2

2
A
1
T
2
= i
0
A
1
.
Approximate solution when
2
= 0 and
2
= 0:
A
1
(z, T) =
1
2
_
A
0
(T
1
z)e
i
0
z
+A
0
(T +
1
z)e
i
0
z

,
A
2
(z, T) =
1
2
_
A
0
(T
1
z)e
i
0
z
A
0
(T +
1
z)e
i
0
z

,
Pulse splits into two subpulses after a few coupling lengths.
119/253

Back
Close
Fiber Gratings
Silica bers exhibit a photosensitive eect.
Refractive index can be changed permanently when ber is exposed
to UV radiation.
Photosensitivity was discovered in 1978 by chance.
Used routinely to make ber Bragg gratings in which mode index
varies in a periodic fashion along ber length.
Fiber gratings can be designed to operate over a wide range of
wavelengths.
Most useful in the wavelength region 1.55 m because of its
relevance to ber-optic communication systems.
Fiber gratings act as a narrowband optical lter.
120/253

Back
Close
Bragg Diffraction
Bragg diraction must satisfy the phase-matching condition
k
i
k
d
= mk
g
, k
g
= 2/.
In single-mode bers, all three vectors lie along ber axis.
Since k
d
=k
i
, diracted light propagates backward.
A ber grating acts as a reector for a specic wavelength for which
k
g
= 2k
i
, or = 2 n.
This condition is known as the Bragg condition.
121/253

Back
Close
First Fiber Grating
In a 1978 experiment, Hill et al. launched blue light from an argon-
ion laser into a 1-m-long ber.
Reected power increased with time and became nearly 100%.
Mechanism behind grating formation was understood much later.
The 4% reection occurring at the ber ends creates a standing-
wave pattern.
Two-photon absorption changes glass structure changes and alters
refractive index in a periodic fashion.
Grating becomes stronger with time because it enhances the visi-
bility of fringe pattern.
By 1989, a holographic technique was used to form the fringe pat-
tern directly using a 244-nm UV laser.
122/253

Back
Close
Photosensitivity of Fibers
Main Mechanism: Formation of defects in the core of a Ge-doped
silica ber.
Ge atoms in ber core leads to formation of oxygen-decient bonds
(SiGe, SiSi, and GeGe bonds).
Absorption of 244-nm radiation breaks defect bonds.
Modications in glass structure change absorption spectrum.
Refractive index also changes through KramersKronig relation
n(
/
) =
c

_

0
()d

/2
.
Typically, n is 10
4
near 1.5 m, but it can exceed 0.001 in
bers with high Ge concentration.
123/253

Back
Close
Photosensitivity of Fibers
Standard telecommunication bers not suitable for forming Bragg
gratings (<3% of Ge atoms results in small index changes.
Photosensitivity can be enhanced using dopants such as phosphorus,
boron, and aluminum.
n > 0.01 possible by soaking ber in hydrogen gas at high pres-
sures (200 atm).
Density of GeSi oxygen-decient bonds increases in hydrogen-soaked
bers.
Once hydrogenated, ber needs to be stored at low temperature to
maintain its photosensitivity.
Gratings remain intact over long periods of time.
124/253

Back
Close
Fabrication Techniques
A dual-beam holographic technique is used commonly.
Cylindrical lens is used to expand UV beam along ber length.
Fringe pattern formed on ber surface creates an index grating.
Grating period related to
uv
as =
uv
/(2sin).
can be varied over a wide range by changing .
Wavelength reected by grating is set by = 2 n.
125/253

Back
Close
Fabrication Techniques
Several variations of the basic technique have been developed.
Holographic technique requires a UV laser with excellent temporal
and spatial coherence.
Excimer lasers used commonly have relatively poor beam quality.
It is dicult to maintain fringe pattern over ber core over a dura-
tion of several minutes.
Fiber gratings can be written using excimer laser pulses.
Pulse energies required are close to 40 mJ for 20-ns pulses.
Exposure time reduced considerably, relaxing coherence
requirements.
126/253

Back
Close
Phase-Mask Technique
Commercial production makes use of a phase-mask technique.
Phase mask acts as a master grating that is transferred to the ber
using a suitable method.
A patterned layer of chromium is deposited on a quartz substrate
using electron-beam lithography and reactive ion etching.
Demands on the temporal and spatial coherence of UV beam are
much less stringent when a phase mask is used.
Even a non-laser source such as a UV lamp can be used.
Quality of ber grating depends completely on the master phase
mask.
127/253

Back
Close
Phase-Mask Interferometer
Phase mask can also be used to form an interferometer.
UV laser beam falls normally on the phase mask and is diracted
into several beams through RamanNath scattering.
The zeroth-order is blocked or cancelled with a suitable technique.
Two rst-order diracted beams interfere on ber surface and form
a fringe pattern.
Grating period equals one-half of phase mask period.
This method is tolerant of any beam-pointing instability.
Relatively long gratings can be made with this technique.
Use of a single silica block that reects two beams internally forms
a compact interferometer.
128/253

Back
Close
Point-by-Point Fabrication
Grating is fabricated onto a ber period by period.
This technique bypasses the need of a master phase mask.
Short sections (w<[) of ber exposed to a single high-energy UV
pulse.
Spot size of UV beam focused tightly to a width w.
Fiber moved by a distance w before next pulse arrives.
A periodic index pattern can be created in this manner.
Only short ber gratings (<1 cm) can be produced because of time-
consuming nature of this method.
Most suitable for long-period gratings.
129/253

Back
Close
Grating Theory
Refractive index of ber mode varies periodically as
n(, z) = n() +n
g
(z) =

m=
n
m
exp[2im(z/)].
Total eld

E in the Helmholtz equation has the form

E(r, ) = F(x, y)[

A
f
(z, )exp(i
B
z) +

A
b
(z, )exp(i
B
z)],
where
B
= / is the Bragg wave number.
If we assume

A
f
and

A
b
vary slowly with z and keep only nearly
phase-matched terms, we obtain coupled-mode equations

A
f
z
= i()

A
f
+i

A
b
,

A
b
z
= i()

A
b
+i

A
f
,
130/253

Back
Close
Coupled-Mode Equations
Coupled-mode equations look similar to those obtained for couplers
with one dierence: Second equations has a negative derivative
This is expected because of backward propagation of A
b
.
Parameter () = ()
B
measures detuning from the Bragg
wavelength.
Coupling coecient is dened as
=
k
0
__

n
1
[F(x, y)[
2
dxdy
__

[F(x, y)[
2
dxdy
.
For a sinusoidal grating, n
g
= n
a
cos(2z/), n
1
= n
a
/2 and
= n
a
/.
131/253

Back
Close
Time-Domain Coupled-Mode Equations
Coupled-mode equations can be converted to time domain by
expanding () as
() =
0
+(
0
)
1
+
1
2
(
0
)
2

2
+
1
6
(
0
)
3

3
+ ,
Replacing
0
with i(/t), we obtain
A
f
z
+
1
v
g
A
f
t
+
i
2
2

2
A
f
t
2
= i
0
A
f
+iA
b
,

A
b
z
+
1
v
g
A
b
t
+
i
2
2

2
A
b
t
2
= i
0
A
b
+iA
f
,

0
= (
0

B
)/v
g
and v
g
= 1/
1
is the group velocity.
When compared to couplers, The only dierence is the sign ap-
pearing in the second equation.
132/253

Back
Close
Photonic Bandgap
In the case of a CW beam, the general solution is

A
f
(z) = A
1
exp(iqz) +A
2
exp(iqz),

A
b
(z) = B
1
exp(iqz) +B
2
exp(iqz),
Constants A
1
, A
2
, B
1
, and B
2
satisfy
(q)A
1
= B
1
, (q+)B
1
=A
1
,
(q)B
2
= A
2
, (q+)A
2
=B
2
.
These relations are satised if q obeys
q =
_

2
.
This dispersion relation is of paramount importance for gratings.
133/253

Back
Close
Dispersion Relation of Gratings
-6 -2 2 6
-6
-2
2
6
q/
/
If frequency of incident light is such that < < ,
q becomes purely imaginary.
Most of the incident eld is reected under such conditions.
The range [[ is called the photonic bandgap or stop band.
Outside this band, propagation constant of light is modied by the
grating to become
e
=
B
q.
134/253

Back
Close
Grating Dispersion
Since q depends on , grating exhibits dispersive eects.
Grating-induced dispersion adds to the material and waveguide dis-
persions associated with a waveguide.
To nd its magnitude, we expand
e
in a Taylor series:

e
() =
g
0
+(
0
)
g
1
+
1
2
(
0
)
2

g
2
+
1
6
(
0
)
3

g
3
+ ,
where
g
m
=
d
m
q
d
m

_
1
v
g
_
m
d
m
q
d
m
.
Group velocity V
G
= 1/
g
1
=v
g
_
1
2
/
2
.
For [[ , optical pulse is unaected by grating.
As [[ approaches , group velocity decreases and becomes zero at
the edges of a stop band.
135/253

Back
Close
Grating Dispersion
(cm
1
)
-20 -10 0 10 20

2
g

(
p
s
2
/
c
m
)
-600
-400
-200
0
200
400
600
10 cm
1
5 1
Second- and third-order dispersive properties are governed by

g
2
=
sgn()
2
/v
2
g
(
2

2
)
3/2
,
g
3
=
3[[
2
/v
3
g
(
2

2
)
5/2
.
GVD anomalous for > 0 and normal for < 0.
136/253

Back
Close
Grating as an Optical Filter
What happens to optical pulses incident on a ber grating?
If pulse spectrum falls entirely within the stop band,
pulse is reected by the grating.
If a part of pulse spectrum is outside the stop band, that part
is transmitted by the grating.
Clearly, shape of reected and transmitted pulses will be quite
dierent depending on detuning from Bragg wavelength.
We can calculate reection and transmission coecients for each
spectral component and then integrate over frequency.
In the linear regime, a ber grating acts as an optical lter.
137/253

Back
Close
Grating Reectivity
Reection coecient can be calculated from the solution

A
f
(z) = A
1
exp(iqz) +r(q)B
2
exp(iqz)

A
b
(z) = B
2
exp(iqz) +r(q)A
1
exp(iqz)
r(q) =
q

=

q+
.
Reection coecient r
g
=

A
b
(0)

A
f
(0)
=
B
2
+r(q)A
1
A
1
+r(q)B
2
.
Using boundary condition

A
b
(L) = 0, B
2
=r(q)A
1
exp(2iqL).
Using this value of B
2
, we obtain
r
g
() =
i sin(qL)
qcos(qL) i sin(qL)
.
138/253

Back
Close
Grating Reectivity
-10 -5 0 5 10
Detuning
0.0
0.2
0.4
0.6
0.8
1.0
R
e
f
l
e
c
t
i
v
i
t
y
-10 -5 0 5 10
Detuning
-15
-10
-5
0
5
P
h
a
s
e
(a) (b)
L = 2 (dashed line); L = 3 (solid line).
Reectivity approaches 100% for L = 3 or larger.
= 2n
1
/ can be used to estimate grating length.
For n
1
10
4
, = 1.55 m, L > 5 5 mm to yield L > 2.
139/253

Back
Close
Grating Apodization
(a)
(b)
Reectivity sidebands originate from a FabryPerot cavity formed
by weak reections occurring at the grating ends.
An apodization technique is used to remove these sidebands.
Intensity of the UV beam across the grating is varied such that it
drops to zero gradually near the two grating ends.
increases from zero to its maximum value in the center.
140/253

Back
Close
Grating Properties
(a)
(p)
80-ps pulses transmitted through an apodized grating.
Pulses were delayed considerably close to a stop-band edge.
Pulse width changed because of grating-induced GVD eects.
Slight compression near = 1200 m
1
is due to SPM.
141/253

Back
Close
Nonuniform Gratings
Grating parameters and become z-dependent in a nonuniform
grating.
Examples of nonuniform gratings include chirped gratings, phase-
shifted gratings, and superstructure gratings.
In a chirped grating, optical period n changes along grating length.
Since
B
= 2 n sets the Bragg wavelength, stop band shifts along
the grating length.
Mathematically, becomes z-dependent.
Chirped gratings have a much wider stop band because it is formed
by a superposition of multiple stop bands.
142/253

Back
Close
Chirped Fiber Gratings
Linearly chirped gratings are commonly used in practice.
Bragg wavelength
B
changes linearly along grating length.
They can be fabricated either by varying physical period or by
changing n along z.
To change , fringe spacing is made nonuniform by interfering
beams with dierent curvatures.
A cylindrical lens is often used in one arm of interferometer.
Chirped ber gratings can also be fabricated by tilting or stretching
the ber, using strain or temperature gradients, or stitching multiple
uniform sections.
143/253

Back
Close
Chirped Fiber Gratings
Useful for dispersion compensation in lightwave systems.
Dierent spectral components reected by dierent parts of grating.
Reected pulse experiences a large amount of GVD.
Nature of GVD (normal vs. anomalous) is controllable.
144/253

Back
Close
Superstructure Gratings
Gratings have a single-peak transfer function.
Some applications require optical lters with multiple peaks.
Superstructure gratings have multiple equally spaced peaks.
Grating designed such that varies periodically along its length.
Such doubly periodic devices are also called sampled gratings.
Such a structure contain multiple grating sections with constant
spacing among them.
It can be made by blocking small regions during fabrication such
that = 0 in the blocked regions.
It can also be made by etching away parts of a grating.
145/253

Back
Close
Fiber Interferometers
Two passive componentscouplers and gratingscan be combined
to form a variety of ber-based optical devices.
Four common ones among them are
Ring and FabryPerot resonators
Sagnac-Loop interferometers
MachZehnder interferometers
Michelson interferometers
Useful for optical switching and other WDM applications.
146/253

Back
Close
Fiber-Ring Resonators
Fiber coupler
Input Output
Made by connecting input and output ports of one core of a direc-
tional coupler to form a ring.
Transmission characteristics obtained using matrix relation
_
A
f
A
i
_
=
_

i

1
i

_
_
A
c
A
t
_
.
After one round trip, A
f
/A
c
= exp[L/2 +i()L]

ae
i
where a = exp(L) 1 and () = ()L.
147/253

Back
Close
Transmission Spectrum
The transmission coecient is found to be
t
r
()

T
r
e
i
t
=
A
t
A
i
=

e
i
1

ae
i
e
i(+)
.
Spectrum shown for a = 0.95 and = 0.9.
If a = 1 (no loss), T
r
= 1 (all-pass resonator) but phase varies as

t
() = + +2tan
1

sin
1

cos
.
148/253

Back
Close
All-Pass Resonators
Frequency dependence of transmitted phase for all-pass resonators
can be used for many applications.
Dierent frequency components of a pulse are delayed by dierent
amounts near a cavity resonance.
A ring resonator exhibits GVD (similar to a ber grating).
Since group delay
d
= d
t
/d, GVD parameter is given by

2
=
1
L
d
2

t
d
2
.
A ber-ring resonator can be used for dispersion compensation.
If a single ring does not provide enough dispersion, several rings can
be cascaded in series.
Such a device can compensate dispersion of multiple channels.
149/253

Back
Close
FabryPerot Resonators
Grating Grating
Fiber
(a)
(b)
Fiber
Fiber
Mirror
(c)
Coupler
Coupler
Use of couplers and gratings provides an all-ber design.
Transmissivity can be calculated by adding contributions of succes-
sive round trips to transmitted eld.
150/253

Back
Close
Transmission Spectrum
Transmitted eld:
A
t
=A
i
e
i
(1R
1
)
1/2
(1R
2
)
1/2
_
1+

R
1
R
2
e
i
R
+R
1
R
2
e
2i
R
+

,
Phase shift during a single round trip:
R
= 2()L.
When R
m
= R
1
= R
2
, A
t
=
(1R
m
)A
i
e
i
1R
m
exp(2i
R
)
.
Transmissivity is given by the Airy formula
T
R
=

A
t
A
i

2
=
(1R
m
)
2
(1R
m
)
2
+4R
m
sin
2
(
R
/2)
.
Round-trip phase shift
R
= (
0
)
r
, where
r
is the
round-trip time.
151/253

Back
Close
Free Spectral Range and Finesse
Sharpness of resonance peaks quantied through the nesse
F
R
=
Peak bandwidth
Free spectral range
=

R
m
1R
m
,
Free spectral range
L
is obtained from phase-matching condition
2[( +2
L
) ()]L = 2.

L
= 1/
r
, where
r
= 2L/v
g
is the round-trip time.
FP resonators are useful as an optical lter with periodic passbands.
Center frequencies of passbands can be tuned by changing physical
mirror spacing or by modifying the refractive index.
152/253

Back
Close
Sagnac Interferometers
Made by connecting two output ports of a ber coupler to form
a ber loop.
No feedback mechanism; all light entering exits after a round trip.
Two counterpropagating parts share the same optical path and
interfere at the coupler coherently.
Their phase dierence determines whether input beam is
reected or transmitted.
153/253

Back
Close
Fiber-Loop Mirrors
When a 3-dB ber coupler is used, any input is totally reected.
Such a device is called the ber-loop mirror.
Fiber-loop mirror can be used for all-optical switching by exploiting
nonlinear eects such as SPM and XPM.
Such a nonlinear optical loop mirror transmits a high-power signal
while reecting it at low power levels.
Useful for many applications such as mode locking, wavelength con-
version, and channel demultiplexing.
154/253

Back
Close
Nonlinear Fiber-Loop Mirrors
Input eld splits into two parts: A
f
=

A
0
, A
b
= i

1 A
0
.
After one round trip, A
/
f
= A
f
exp[i
0
+i([A
f
[
2
+2[A
b
[
2
)L],
A
/
b
= A
b
exp(i
0
+i([A
b
[
2
+2[A
f
[
2
)L].
Reected and transmitted elds after ber coupler:
_
A
t
A
r
_
=
_

i

1
i

_
_
A
/
f
A
/
b
_
.
Transmissivity T
S
[A
t
[
2
/[A
0
[
2
of the Sagnac loop:
T
S
= 12(1)1+cos[(12)P
0
L],
If ,= 1/2, ber-loop mirror can act as an optical switch.
155/253

Back
Close
Nonlinear Transmission Characteristics
At low powers, little light is transmitted if is close to 0.5.
At high powers, SPM-induced phase shift leads to 100% transmis-
sion whenever [12[P
0
L = (2m1).
Switching power for m = 1 is 31 W for a 100-m-long ber loop
when = 0.45 and = 10 W
1
/km.
It can be reduced by increasing loop length or .
156/253

Back
Close
Nonlinear Switching
Most experiments use short optical pulses with high peak powers.
In a 1989 experiment, 180-ps pulses were injected into a 25-m
Sagnac loop.
Transmission increased from a few percent to 60% as peak power
was increased beyond 30 W.
Only the central part of the pulse was switched.
Shape deformation can be avoided by using solitons.
Switching threshold can be reduced by incorporating a ber ampli-
er within the loop.
157/253

Back
Close
Nonlinear Amplifying-Loop Mirror
If amplier is located close to the ber coupler, it introduces an
asymmetry benecial to optical switching.
Even a 50:50 coupler ( = 0.5) can be used for switching.
In one direction pulse is amplied as it enters the loop.
Counterpropagating pulse is amplied just before it exits the loop.
Since powers in two directions dier by a large amount, dierential
phase shift can be quite large.
Transmissivity of loop mirror is given by
T
S
= 12(1)1+cos[(1 G)P
0
L].
Switching power P
0
= 2/[(G1)L].
158/253

Back
Close
Nonlinear Amplifying-Loop Mirror
Since G 30 dB, switching power is reduced considerably.
Such a device can switch at peak power levels below 1 mW.
In a 1990 experiment, 4.5 m of Nd-doped ber was spliced within
a 306-m ber loop formed with a 3-dB coupler.
Switching was observed using 10-ns pulses.
Switching power was about 0.9 W even when amplier provided
only 6-dB gain.
A semiconductor optical amplier inside a 17-m ber loop produced
switching at 250 W with 10-ns pulses.
159/253

Back
Close
Dispersion-Unbalanced Sagnac Loops
Sagnac interferometer can also be unbalanced by using a ber whose
GVD varies along the loop length.
A dispersion-decreasing ber or several bers with dierent disper-
sive properties can be used.
In the simplest case Sagnac loop is made with two types of bers.
Sagnac interferometer is unbalanced as counterpropagating waves
experience dierent GVD during a round trip.
Such Sagnac loops remain balanced for CW beams.
As a result, optical pulses can be switched to output port while any
CW background noise is reected.
Such a device can improve the SNR of a noisy signal.
160/253

Back
Close
XPM-Induced Switching
XPM can also be used for all-optical switching.
A control signal is injected into the Sagnac loop such that it prop-
agates in only one direction.
It induces a nonlinear phase shift through XPM in that direction.
In essence, control signal unbalances the Sagnac loop.
As a result, a low-power CW signal is reected in the absence of a
control pulse but is transmitted in its presence.
As early as 1989, a 632-nm CW signal was switched using intense
532-nm picosecond pump pulses with 25-W peak power.
Walk-o eects induced by group-velocity mismatch aect the de-
vice. It is better to us orthogonally polarized control at the same
wavelength.
161/253

Back
Close
MachZehnder Interferometers
A MachZehnder (MZ) interferometer is made by connecting two
ber couplers in series.
Such a device has the advantage that nothing is reected back
toward the input port.
MZ interferometer can be unbalanced by using dierent path lengths
in its two arms.
This feature also makes it susceptible to environmental uctuations.
162/253

Back
Close
Transmission Characteristics
Taking into account both the linear and nonlinear phase shifts,
optical elds at the second coupler are given by
A
1
=

1
A
0
exp(i
1
L
1
+i
1
[A
0
[
2
L
1
),
A
2
= i
_
1
1
A
0
exp[i
2
L
2
+i(1
1
)[A
0
[
2
L
2
],
Transmitted elds from two ports:
_
A
3
A
4
_
=
_

2
i

1
2
i

1
2

2
__
A
1
A
2
_
.
Transmissivity of the bar port is given by
T
b
=
1

2
+(1
1
)(1
2
)2[
1

2
(1
1
)(1
2
)]
1/2
cos(
L
+
NL
),

L
=
1
L
1

2
L
2
and
NL
= P
0
[
1
L
1
(1
1
)L
2
].
163/253

Back
Close
Transmission Characteristics
Nonlinear switching for two values of
L
.
A dual-core ber was used to make the interferometer (L
1
= L
2
).
This conguration avoids temporal uctuations occurring invariably
when two separate ber pieces are used.
164/253

Back
Close
XPM-Induced Switching
Switching is also possible through XPM-induced phase shift.
Control beam propagates in one arm of the MZ interferometer.
MZ interferometer is balanced in the absence of control,
and signal appears at port 4.
When control induces a phase shift through XPM, signal is di-
rected toward port 3.
Switching power can be lowered by reducing eective core area
A
eff
of ber.
165/253

Back
Close
Michelson Interferometers
Can be made by splicing Bragg gratings at the output ports of
a ber coupler.
It functions like a MZ interferometer.
Light propagating in its two arms interferes at the same coupler
where it was split.
Acts as a nonlinear mirror, similar to a Sagnac interferometer.
Reectivity R
M
=
2
+(1)
2
2(1)cos(
L
+
NL
).
Nonlinear characteristics similar to those of a Sagnac loop.
Often used for passive mode locking of lasers (additive-pulse
mode locking).
166/253

Back
Close
Isolators and Circulators
Isolators and circulators fall into the category of nonreciprocal
devices.
Such a device breaks the time-reversal symmetry inherent in optics.
It requires that device behave dierently when the direction of light
propagation is reversed.
A static magnetic eld must be applied to break time-reversal sym-
metry.
Device operation is based on the Faraday eect.
Faraday eect: Changes in the state of polarization of an optical
beam in a magneto-optic medium in the presence of a magnetic
eld.
167/253

Back
Close
Faraday Effect
Refractive indices of some materials become dierent for RCP and
LCP components in the presence of a magnetic eld.
On a more fundamental level, Faraday eect has its origin in the
motion of electrons in the presence of a magnetic eld.
It manifests as a change in the state of polarization as the beam
propagates through the medium.
Polarization changes depend on the direction of magnetic eld but
not on the direction in which light is traveling.
Mathematically, two circularly polarized components propagate with

= n

(/c).
Circular birefringence depends on magnetic eld as
n = n
+
n

= K
F
H
dc
.
168/253

Back
Close
Faraday Rotator
Relative phase shift between RCP and LCP components is
= (/c)K
F
H
dc
l
M
=V
c
H
dc
l
M
,
where V
c
= (/c)K
F
is the Verdet constant.
Plane of polarization of light is rotated by an angle

F
=
1
2
.
Most commonly used material: terbium gallium garnet with Verdet
constant of 0.1 rad/(Oe-cm).
Useful for making a device known as the Faraday rotator.
Magnetic eld and medium length are chosen to induce 45

change
in direction of linearly polarized light.
169/253

Back
Close
Optical Isolators
GRIN
Lens
GRIN
Lens
Faraday
Isolator
Polarizer Analyzer
Fiber
Fiber
Direction of
Magnertic
Field
Optical analog of a rectifying diode.
Uses a Faraday rotator sandwiched between two polarizers.
Second polarizer tilted at 45

from rst polarizer.

Polarization-independent isolators process orthogonally polarized com-
ponents separately and combine them at the output end.
Commercial isolators provide better than 30-dB isolation in a com-
pact package (4 cm5 mm wide).
170/253

Back
Close
Optical Circulators
A circulator directs backward propagating light does to another port
rather than discarding it, resulting in a three-port device.
More ports can be added if necessary.
Such devices are called circulators because they direct light to dif-
ferent ports in a circular fashion.
Design of optical circulators becomes increasingly complex as the
number of ports increases.
171/253

Back
Close
Active Fiber Components
No electrical pumping possible as silica is an insulator.
Active components can be made but require optical pumping.
Fiber core is often doped with a rare-earth element to realize optical
gain through optical pumping.
Active Fiber components
Doped-Fiber Ampliers
Raman Ampliers (SRS)
Parametric Ampliers (FWM)
CW and mode-locked Fiber Lasers
172/253

Back
Close
Doped-Fiber Ampliers
Core doped with a rare-earth element during manufacturing.
Many dierent elements such as erbium, neodymium, and ytter-
bium, can be used to make ber ampliers (and lasers).
Amplier properties such as operating wavelength and gain band-
width are set by the dopant.
Silica ber plays the passive role of a host.
Erbium-doped ber ampliers (EDFAs) operate near 1.55 m and
are used commonly for lightwave systems.
Yb-doped ber are useful for high-power applications.
Yb-doped ber lasers can emit > 1 kW of power.
173/253

Back
Close
Optical Pumping
Optical gain realized when a doped ber is pumped optically.
In the case of EDFAs, semiconductor lasers operating near 0.98-
and 1.48-m wavelengths are used.
30-dB gain can be realized with only 1015 mW of pump power.
Eciencies as high as 11 dB/mW are possible.
174/253

Back
Close
Amplier Gain
Gain coecient can be written as
g() =
g
0
(P
p
)
1+(
0
)
2
T
2
2
+P/P
s
.
T
2
is the dipole relaxation time (typically <1 ps).
Fluorescence time T
1
can vary from 1 s10 ms depending on the
rare-earth element used (10 ms for EDFAs).
Amplication of a CW signal is governed by dP/dz = g()P.
When P/P
s
1, solution is P(z) = P(0)exp(gz).
Amplier gain G is dened as
G() = P
out
/P
in
= P(L)/P(0) = exp[g()L].
175/253

Back
Close
Gain Spectrum
For P P
s
, small-signal gain is of the form
g() =
g
0
1+(
0
)
2
T
2
2
.
Lorentzian Gain spectrum with a FWHM
g
=
1
T
2
.
Amplier gain G() has a peak value G
0
= exp(g
0
L).
Its FWHM is given by
A
=
g
_
ln2
ln(G
0
/2)
_
1/2
.
Amplier bandwidth is smaller than gain bandwidth.
Gain spectrum of EDFAs has a double-peak structure with a
bandwidth >35 nm.
EDFAs can provide amplication over a wide spectral region
(15201610 nm).
176/253

Back
Close
Amplier Noise
All ampliers degrade SNR of the amplied signal because of
spontaneous emission.
SNR degradation quantied through the noise gure F
n
dened as
F
n
= (SNR)
in
/(SNR)
out
.
In general, F
n
depends on several detector parameters related to
thermal noise.
For an ideal detector (no thermal noise)
F
n
= 2n
sp
(11/G) +1/G 2n
sp
.
Spontaneous emission factor n
sp
= N
2
/(N
2
N
1
).
For a fully inverted amplier (N
2
N
1
), n
sp
= 1.
51-dB gain realized with F
n
= 3.1 dB at 48 mW pump power.
177/253

Back
Close
Amplier Design
EDFAs are designed to provide uniform gain over the entire
C band (15301570 nm).
An optical lter is used for gain attening.
It often contains several long-period ber gratings.
Two-stage design helps to reduce the noise level as it permits to
place optical lter in the middle.
Noise gure is set by the rst stage.
178/253

Back
Close
Amplier Design
A two-stage design is used for L-band ampliers operating
in the range 15701610 nm.
First stage pumped at 980 nm and acts as a traditional EDFA.
Second stage has a long doped ber (200 m or so) and is pumped
bidirectionally using 1480-nm lasers.
An optical isolator blocks the backward-propagating ASE.
Such cascaded ampliers provide at gain with relatively low noise
level levels.
179/253

Back
Close
Raman Ampliers
A Raman amplier uses stimulated Raman scattering (SRS) for
signal amplication.
SRS is normally harmful for WDM systems.
The same process useful for making Raman ampliers.
Raman ampliers can provide large gain over a wide bandwidth in
any spectral region using a suitable pump.
Require long ber lengths (>1 km) compared with EDFAs.
Fiber used for data transmission can itself be employed as a
Raman-gain medium.
This scheme is referred to as distributed Raman amplication.
180/253

Back
Close
Raman Ampliers
Similar to EDFAs, Raman ampliers must be pumped optically.
Pump and signal injected into the ber through a ber coupler.
Pump power is transferred to the signal through SRS.
Pump and signal counterpropagate in the backward-pumping con-
guration often used in practice.
Signal amplied exponentially as e
gL
with
g() = g
R
()(P
p
/A
eff
).
181/253

Back
Close
Raman Gain and Bandwidth
(b) (a)
Raman gain spectrum g
R
() has a broad peak located near 13 THz.
The ratio g
R
/A
eff
is a measure of Raman-gain eciency and depends
on ber design.
A dispersion-compensating ber (DCF) can be 8 times more e-
cient than a standard silica ber.
Gain bandwidth
g
is about 6 THz.
Multiple pumps can be used make gain spectrum wider and atter.
182/253

Back
Close
Single-Pump Raman Amplication
Governed by a set of two coupled nonlinear equations:
dP
s
dz
=
g
R
A
eff
P
p
P
s

s
P
s
,
dP
p
dz
=

s
g
R
A
eff
P
p
P
s

p
P
p
,
=1 depending on the pumping conguration.
In practice, P
p
P
s
, and pump depletion can be ignored.
P
p
(z) = P
0
exp(
p
z) in the forward-pumping case.
Signal equation is then easily integrated to obtain
P
s
(L) = P
s
(0)exp(g
R
P
0
L
eff
/A
eff

s
L) G(L)P
s
(0),
where L
eff
= [1exp(
p
L)]/
p
.
183/253

Back
Close
Bidirectional Pumping
In the case of backward-pumping, boundary condition becomes
P
p
(L) = P
0
.
Solution of pump equation becomes P
p
(z) = P
0
exp[
p
(Lz)].
Same amplication factor as for forward pumping.
In the case of bidirectional pumping, the solution is
P
s
(z) G(z)P
s
(0) = P
s
(0)exp
_
g
R
A
eff
_
z
0
P
p
(z)dz
s
L
_
,
where P
p
(z) = P
0
f
p
exp(
p
z) +(1 f
p
)exp[
p
(Lz)].
P
0
is total power and f
p
is its fraction in forward direction.
Amplier properties depend on f
p
.
184/253

Back
Close
Bidirectional Pumping
Change in signal power along a 100-km-long Raman amplier as f
p
is varied in the range 0 to 1.
In all cases, g
R
/A
eff
= 0.7 W
1
/km,
s
= 0.2 dB/km,
p
= 0.25
dB/km, and G(L) = 1.
Which pumping conguration is better from a system standpoint?
185/253

Back
Close
Forward or Backward Pumping?
Forward pumping superior from the noise viewpoint.
Backward pumping better in practice as it reduces nonlinear eects
(signal power small throughout ber link).
Accumulated nonlinear phase shift induced by SPM is given by

NL
=
_
L
0
P
s
(z)dz = P
s
(0)
_
L
0
G(z)dz.
Increase in
NL
because of Raman amplication is quantied by
the ratio
R
NL
=

NL
(pump on)

NL
(pump o)
= L
1
eff
_
L
0
G(z)dz.
This ratio is smallest for backward pumping.
186/253

Back
Close
Multiple-Pump Raman Amplication
Raman ampliers need high pump powers.
Gain spectrum is 2025 nm wide but relatively nonuniform.
Both problems can be solved using multiple pump lasers at
suitably optimized wavelengths.
Even though Raman gain spectrum of each pump is not very at,
it can be broadened and attened using multiple pumps.
Each pump creates its own nonuniform gain prole over a
specic spectral range.
Superposition of several such spectra can create relatively at gain
over a wide spectral region.
187/253

Back
Close
Example of Raman Gain Spectrum
(b) (a)
Five pump lasers operating at 1,420, 1,435, 1,450, 1,465, and
1,495 nm are used.
Individual pump powers chosen to provide uniform gain over a
80-nm bandwidth (top trace).
Raman gain is polarization-sensitive. Polarization problem is solved
using two orthogonally polarized pump lasers at each wavelength.
It can also be solved by depolarizing output of each pump laser.
188/253

Back
Close
Example of Raman Gain Spectrum
Measured Raman gain for a Raman amplier pumped with 12 lasers.
Pump powers used (shown on the right) were below 100 mW for
each pump laser.
Pump powers and wavelengths are design parameters obtained by
solving a complex set of equations.
189/253

Back
Close
Noise in Raman Ampliers
Spontaneous Raman scattering adds noise to the amplied signal.
Noise is temperature dependent as it depends on phonon
population in the vibrational state.
Evolution of signal is governed by
dA
s
dz
=
g
R
2A
eff
P
p
(z)A
s

s
2
A
s
+ f
n
(z, t),
f
n
(z, t) is modeled as a Gaussian stochastic process with
f
n
(z, t) f
n
(z
/
, t
/
)) = n
sp
h
0
g
R
P
p
(z)(z z
/
)(t t
/
),
n
sp
() = [1exp( h/k
B
T)]
1
, =
p

s
.
190/253

Back
Close
Noise in Raman Ampliers
Integrating over amplier length, A
s
(L) =
_
G(L)A
s
(0) +A
sp
:
A
sp
=
_
G(L)
_
L
0
f
n
(z, t)
_
G(z)
dz, G(z) =exp
_
_
z
0
[g
R
P
p
(z
/
)
s
] dz
/
_
.
Spontaneous power added to the signal is given by
P
sp
= n
sp
h
0
g
R
B
opt
G(L)
_
L
0
P
p
(z)
G(z)
dz,
B
opt
is the bandwidth of the Raman amplier (or optical lter).
Total noise power higher by factor of 2 when both polarization
components are considered.
191/253

Back
Close
Noise in Raman Ampliers
Noise gure of a Raman amplier is given by
F
n
=
P
sp
Gh
0
f
= n
sp
g
R
B
opt
f
_
L
0
P
p
(z)
G(z)
dz.
Common to introduce the concept of an eective noise gure
as F
eff
= F
n
exp(
s
L).
F
eff
can be less than 1 (negative on the decibel scale).
Physically speaking, distributed gain counteracts ber losses and
results in better SNR compared with lumped ampliers.
Forward pumping results in less noise because Raman gain is con-
centrated toward the input end.
192/253

Back
Close
Parametric Ampliers
Make use of four-wave mixing (FWM) in optical bers.
Two pumps (at
1
and
2
) launched with the signal at
3
.
The idler eld generated internally at a frequency

4
=
1
+
2

3
.
Signal and idler both amplied through FWM.
Such a device can amplify signal by 3040 dB if a phase-matching
condition is satised.
It can also act as a wavelength converter.
Idler phase is reverse of the signal (phase conjugation).
193/253

Back
Close
Simple Theory
FWM is described by a set of 4 coupled nonlinear equations.
These equations must be solved numerically in general.
If we assume intense pumps (negligible depletion), and treat pump
powers as constant, signal and idler elds satisfy
dA
3
dz
= 2i[(P
1
+P
2
)A
3
+

P
1
P
2
e
i
A

4
],
dA

4
dz
= 2i[(P
1
+P
2
)A

4
+

P
1
P
2
e
i
A
3
],
P
1
=[A
1
[
2
and P
2
=[A
2
[
2
are pump powers.
= [ 3(P
1
+P
2
)]z represents total phase mismatch.
Linear part =
3
+
4

2
, where
j
= n
j

j
/c.
194/253

Back
Close
Signal and Idler Equations
Two coupled equations can be solved analytically as they are linear
rst-order ODEs.
Notice that A
3
couples to A

4
(phase conjugation).
Introducing B
j
= A
j
exp[2i(P
1
+P
2
)z], we obtain the following
set of two equations:
dB
3
dz
= 2i

P
1
P
2
e
iz
B

4
,
dB

4
dz
= 2i

P
1
P
2
e
iz
B
3
,
Phase mismatch: = +(P
1
+P
2
).
= 0 is possible if pump wavelength lies close to ZDWL but in the
anomalous-dispersion regime of the ber.
195/253

Back
Close
Parametric Gain
General solution for signal and idler elds:
B
3
(z) = (a
3
e
gz
+b
3
e
gz
)exp(iz/2),
B

4
(z) = (a
4
e
gz
+b
4
e
gz
)exp(iz/2),
a
3
, b
3
, a
4
, and b
4
are determined from boundary conditions.
Parametric gain g depends on pump powers as
g =
_
(P
0
r)
2
(/2)
2
, r = 2

P
1
P
2
/P
0
, P
0
= P
1
+P
2
.
In the degenerate case, single pump provides both photons
for creating a pair of signal and idler photons.
In this case P
1
= P
2
= P
0
and r = 1.
Maximum gain g
max
= P
0
occurs when = 0.
196/253

Back
Close
Single-Pump Parametric Ampliers
A single pump is used to pump a parametric amplier.
Assuming P
4
(0) = 0 (no input at idler frequency), signal and idler
powers at z = L are
P
3
(L) = P
3
(0)[1+(1+
2
/4g
2
)sinh
2
(gL)],
P
4
(L) = P
3
(0)(1+
2
/4g
2
)sinh
2
(gL),
Parametric gain g =
_
(P
0
)
2
(/2)
2
.
Amplication factor G
p
=
P
3
(L)
P
3
(0)
= 1+(P
0
/g)
2
sinh
2
(P
0
L).
When phase matching is perfect ( = 0) and gL 1
G
p
= 1+sinh
2
(P
0
L)
1
4
exp(2P
0
L).
197/253

Back
Close
Single-Pump Parametric Ampliers
2 1.5 1 0.5 0 0.5 1 1.5 2
0
5
10
15
20
25
30
35
40
Signal Detuning (THz)
G
a
i
n

d
B
)
P
0
= 1 W
0.8
0.6
G
p
as a function of pump-signal detuning
s

p
.
Pump wavelength close to the zero-dispersion wavelength.
500-m-long ber with = 10 W
1
/km and
2
=0.5 ps
2
/km.
Peak gain is close to 38 dB at a 1-W pump level and occurs when
signal is detuned by 1 THz from pump wavelength.
198/253

Back
Close
Single-Pump Parametric Ampliers
Experimental results agree with simple FWM theory.
500-m-long ber with = 11 W
1
/km.
Output of a DFB laser was boosted to 2 W using two EDFAs.
It was necessary to broaden pump spectrum from 10 MHz to >1 GHz
to suppress SBS.
199/253

Back
Close
Dual-Pump Parametric Ampliers
Z
1
Z
2
Z
1 -
Z
2 -
Z
1 +
Z
2 +
Z
Z
0
1- 1+ 2- 2+
Z
1
Z
2
Z
1 -
Z
2 -
Z
1 +
Z
2 +
Z
Z
0
1- 1+ 2- 2+
Pumps positioned on opposite sides of ZDWL.
Multiple FWM processes general several idler bands.
Degenerate FWM :
1
+
1

1+
+
1
.
Nondegenerate FWM:
1
+
2

1+
+
2
.
Additional gain through combinations

1
+
1+

2
+
2
,
2
+
1+

1
+
2+
.
200/253

Back
Close
Dual-Pump Parametric Ampliers
60 40 20 0 20 40 60
0
5
10
15
20
25
30
35
40
Signal Detuning (nm)
A
m
p
l
f
i
e
r

G
a
i
n

(
d
B
)
500 mW
300 mW
200 mW
Examples of gain spectra at three pump-power levels.
A 500-m-long ber used with = 10 W
1
/km, ZDWL = 1570 nm,

3
= 0.038 ps
3
/km, and
4
= 110
4
ps
3
/km.
Two pumps at 1525 and 1618 nm (almost symmetric
around ZDWL) with 500 mW of power.
201/253

Back
Close
Dual-Pump Parametric Ampliers
Wavelength (nm)
G
a
i
n

(
d
B
)
Measured gain (symbols) for pump powers of 600 and 200 mW at
1,559 and 1,610 nm, respectively.
Unequal input pump powers were used because of SRS.
SBS was avoided by modulating pump phases at 10 GHz.
Theoretical t required inclusion of Raman-induced transfer
of powers between the pumps, signal, and idlers.
202/253

Back
Close
Polarization effects
Parametric gain is negligible when pump and signal are orthogonally
polarized (and maximum when they are copolarized).
Parametric gain can vary widely depending on SOP of input signal.
This problem can be solved by using two orthogonally polarized
pumps with equal powers.
Linearly polarized pumps in most experiments.
Amplier gain is reduced drastically compared with
the copolarized case.
Much higher values of gain are possible if two pumps
are chosen to be circularly polarized.
203/253

Back
Close
Polarization effects
Two pumps at 1,535 and 1,628 nm launched with 0.5 W powers.
Gain reduced to 8.5 dB for linearly polarized pumps but increases
to 23 dB when pumps are circularly polarized.
Reason: Angular momentum should be conserved.
204/253

Back
Close
Fiber Lasers
Any amplier can be converted into a laser by placing it inside a
cavity designed to provide optical feedback.
Fiber lasers can use a FabryPerot cavity if mirrors are butt-coupled
to its two ends.
Alignment of such a cavity is not easy.
Better approach: deposit dielectric mirrors onto the polished ends
of a doped ber.
Since pump light passes through the same mirrors, dielectric
coatings can be easily damaged.
A WDM ber coupler can solve this problem.
Another solution is to use ber gratings as mirrors.
205/253

Back
Close
Ring-Cavity Design
A ring cavity is often used for ber lasers.
It can be made without using any mirrors.
Two ports of a WDM coupler connected to form a ring cavity.
An isolator is inserted for unidirectional operation.
A polarization controller is needed for conventional bers that do
not preserve polarization.
206/253

Back
Close
Figure-8 Cavity
Ring cavity on right acts as a nonlinear amplifying-loop mirror.
Nonlinear eects play important role in such lasers.
At low powers, loop transmissivity is small, resulting in large cavity
losses for CW operation.
Sagnac loop becomes transmissive for pulses whose peak power
exceeds a critical value.
A gure-8 cavity permits passive mode locking without any active
elements.
207/253

Back
Close
CW Fiber Lasers
EDFLs exhibit low threshold
(<10 mW pump power) and
a narrow line width (<10 kHz).
Tunable over a wide wavelength
range (>50 nm).
A rotating grating can be used
(Wyatt, Electon. Lett.,1989).
Many other tuning techniques
have been used.
(a)
(b)
208/253

Back
Close
Multiwavelength Fiber Lasers
EDFLs can be designed to emit light at several wavelengths
simultaneously.
Such lasers are useful for WDM applications.
A dual-frequency ber laser was demonstrated in 1993 using a
coupled-cavity conguration.
A comb lter (e.g., a FabryPerot lter) is often used for this pur-
pose.
In a recent experiment, a ber-ring laser provided output at 52
channels, designed to be 50 GHz apart.
209/253

Back
Close
Mode-Locked Fiber Lasers
Saturable absorbers commonly used for passive mode locking.
A saturable Bragg reector often used for this purpose.
Dispersion and SPM inside bers play an important role and should
be included.
15 cm of doped ber is spliced to a 30-cm section of standard ber
for dispersion control.
Pulse widths below 0.5 ps formed over a wide range of average GVD
(
2
=2 to 14 ps
2
/km).
Harmonic mode locking was found to occur for short cavity lengths.
210/253

Back
Close
Nonlinear Fiber-Loop Mirrors
Nonlinear amplifying-loop mirror (NALM) provides mode locking
with an all-ber ring cavity.
NALM behaves like a saturable absorber but responds at femtosec-
ond timescales.
First used in 1991 and produced 290 fs pulses.
Pulses as short as 30 fs can be obtained by compressing pulses in a
dispersion-shifted ber.
211/253

Back
Close
Nonlinear Polarization Rotation
Mode locking through intensity-dependent changes in the SOP in-
duced by SPM and XPM.
Mode-locking mechanism similar to that used for gure-8 lasers:
orthogonally polarized components of same pulse are used.
In a 1993 experiment, 76-fs pulses with 90-pJ energy and 1 kW of
peak power generated.
212/253

Back
Close
Planar Waveguides
Passive components
Y and X Junctions
Grating-assisted Directional Couplers
MachZehnder Filters
Multimode Interference Couplers
Star Couplers
Arrayed-waveguide Gratings
Active components
Semiconductor lasers and ampliers
Optical Modulators
Photodetectors
213/253

Back
Close
Y Junctions
A three-port device that acts as a power divider.
Made by splitting a planar waveguide into two branches bifurcating
at some angle .
Similar to a ber coupler except it has only three ports.
Conceptually, it diers considerably from a ber coupler.
No coupling region exists in which modes of dierent waveguides
overlap.
214/253

Back
Close
Y Junctions
Functioning of Y junction can be understood as follows.
In the junction region, waveguide is thicker and supports higher-
order modes.
Geometrical symmetry forbids excitation of asymmetric modes.
If thickness is changed gradually in an adiabatic manner, even
higher-order symmetric modes are not excited.
As a result, power is divided into two branches.
Sudden opening of the gap violates adiabatic condition, resulting in
some insertion losses for any Y junction.
Losses depend on branching angle and increase with it.
should be below 1

to keep insertion losses below 1 dB.

215/253

Back
Close
Four-Port Couplers
(a) (b)
(c) (d)
Spacing between waveguides reduced to zero in coupled Y junctions.
Waveguides cross in the central region in a X coupler.
In asymmetric X couplers, two input waveguides are identical
but output waveguides have dierent sizes.
Power splitting depends on relative phase between two inputs.
If inputs are equal and in phase, power is transferred to wider core;
when inputs are out of phase, power is transferred to narrow core.
216/253

Back
Close
Grating-Assisted Directional Couplers
An asymmetric directional coupler with a built-in grating.
Little power will be transferred in the absence of grating.
Grating helps to match propagation constants and induces power
transfer for specic input wavelengths.
Grating period = 2/[
1

2
[.
Typically, 10 m (a long-period grating).
A short-period grating used if light is launched in opposite direc-
tions.
217/253

Back
Close
MachZehnder Switches
Two arm lengths equal in a symmetric MZ interferometer.
Such a device transfers its input power to the cross port.
Output can be switched to bar port by inducing a phase shift in
one arm.
Phase shift can be induced electrically using a thin-lm heater (a
thin layer of chromium).
Thermo-optic eect is relatively slow.
Much faster switching using electro-optic eect in LiNbO
3
.
218/253

Back
Close
MachZehnder Filters
An asymmetric MZI acts as an optical lter.
Its output depends on the frequency of incident light.
Transfer function H() = sin().
is the additional delay in one arm of MZI.
Such a lter is not sharp enough for applications.
A cascaded chain of MZI provides narrowband optical lters.
In a chain of N cascaded MZIs, one has the freedom of adjusting
N delays and N+1 splitting ratios.
This freedom can be used to synthesize optical lters with arbitrary
amplitude and phase responses.
219/253

Back
Close
Cascaded MachZehnder Filters
Transmission through a chain of N MZIs can be calculated with the
transfer-matrix approach. In matrix form
F
out
() = T
N+1
D
N
T
N
D
2
T
2
D
1
T
1
F
in
,
T
m
is the transfer matrix and D
m
is a diagonal matrix
T
m
=
_
c
m
is
m
is
m
c
m
_
D
m
=
_
e
i
m
0
0 e
i
m
_
.
c
m
= cos(
m
l
m
) and s
m
= sin(
m
l
m
) and 2
m
=
m
.
Simple rule: sum over all possible optical paths. A chain of two
cascaded MZI has four possible paths:
t
b
() = ic
1
c
2
s
3
e
i(
1
+
2
)
+ic
1
s
2
s
3
e
i(
1

2
)
+
i
3
s
1
c
2
s
3
e
i(
1
+
2
)
+is
1
s
2
s
3
e
i(
1
+
2
)
220/253

Back
Close
Multimode Interference Couplers
MMI couplers are based on the Talbot eect: Self-imaging of ob-
jects in a medium exhibiting periodicity.
Same phenomenon occurs when an input waveguides is connected
to a thick central region supporting multiple modes.
Length of central coupling region is chosen such that optical eld is
self-imaged and forms an array of identical images at the location
of output waveguides.
Such a device functions as an 1N power splitter.
221/253

Back
Close
Multimode Interference Couplers
Expand input eld into mode
m
(x) as A(x, z) =

C
m

m
(x).
Field at a distance z: A(x, z) =

C
m

m
(x)exp(i
m
z).
Propagation constant
m
for a slab of width W
e
:

2
m
= n
2
s
k
2
0
p
2
m
, where p
m
= (m+1)/W
e
.
Since p
m
k
0
, we can approximate
m
as

m
n
s
k
0

(m+1)
2

2
2n
s
k
0
W
2
e
=
0

m(m+2)
3L
b
,
Beat length L
b
=

1

4n
s
W
2
e
3
.
Input eld is reproduced at z = 3L
b
.
Multiple images of input can form for L < 3L
b
.
222/253

Back
Close
Star Couplers
Some applications make use of N N couplers designed with N
input and N output ports.
Such couplers are known as star couplers.
They can be made by combining multiple 3-dB couplers.
A 88 star coupler requires twelve 3-dB couplers.
Device design becomes too cumbersome for larger ports.
223/253

Back
Close
Star Couplers
Compact star couplers can be been made using planar waveguides.
Input and output waveguides connected to a central region.
Optical eld diracts freely inside central region.
Waveguides are arranged to have a constant angular separation.
Input and output boundaries of central slab form arcs that are cen-
tered at two focal points with a radius equal to focal distance.
Dummy waveguides added near edges to ensure a large periodic
array.
224/253

Back
Close
Theory Behind Star Couplers
An innite array of coupled waveguides supports supermodes in the
form of Bloch functions.
Optical eld associated with a supermode:
(x, k
x
) =

m
F(x ma)e
imk
x
a
.
F(x) is the mode prole and a is the period of array.
k
x
is restricted to the rst Brillouin zone: /a < k
x
< /a.
Light launched into one waveguide excites all supermodes within
the rst Brillouin zone.
225/253

Back
Close
Theory Behind Star Couplers
As waveguides approach central slab, (x, k
x
) evolves into a freely
propagating wave with a curved wavefront.
k
x
/
s
, where
s
is the propagation constant in the slab.
Maximum value of this angle:

BZ
k
max
x
/
s
= /(
s
a).
Star coupler is designed such that all N waveguides are within
illuminated region: Na/R = 2
BZ
, where R is focal distance.
With this arrangement, optical power entering from any input
waveguide is divided equally among N output waveguides.
Silica-on-silicon technology is often used for star couplers.
226/253

Back
Close
Arrayed-Waveguide Gratings
AWG combines two N M star couplers through an array of M
curved waveguides.
Length dierence between neighboring waveguides is constant.
Constant phase dierence between neighboring waveguides pro-
duces grating-like behavior.
227/253

Back
Close
Theory Behind AWGs
Consider a WDM signal launched into an input waveguides.
First star coupler splits power into many parts and directs them into
the waveguides forming the grating.
At the output end, wavefront is tilted because of linearly varying
phase shifts.
Tilt is wavelength-dependent and it forces each channel to focus
onto a dierent output waveguide.
Bragg condition for an AWG:
k
0
n
w
(l) +k
0
n
p
a
g
(
in
+
out
) = 2m,
a
g
= garting pitch,
in
= pa
i
/R, and
out
= qa
o
/R.
228/253

Back
Close
Fabrication of AWGs
AWGs are fabricated with silica-on-silicon technology.
Half-wave plate helps to correct for birefringence eects.
By 2001, 400-channel AWGs were fabricated .
Such a device requiring fabrication of hundreds of waveguides on
the same substrate.
229/253

Back
Close
Semiconductor Lasers and Ampliers
Semiconductor waveguides useful for making lasers operating
in the wavelength range 4001600 nm.
Semiconductor lasers oer many advantages.
Compact size, high eciency, good reliability.
Emissive area compatible with bers.
Electrical pumping at modest current levels.
Output can be modulated at high frequencies.
First demonstration of semiconductor lasers in 1962.
Room-temperature operation rst realized in 1970.
Used in laser printers, CD and DVD players, and telecommunication
systems.
230/253

Back
Close
Operating Principle
Forward biasing of a pn junction produces free electrons and holes.
Electron-hole recombination in a direct-bandgap semiconductor
produces light through spontaneous or stimulated emission.
231/253

Back
Close
Basic Structure
Active layer sandwiched between p-type and n-type cladding layers.
Their bandgap dierence connes carriers to active layer.
Active layers larger refractive index creates a planar waveguide.
Single-mode operation require layer thickness below 0.2 m.
Cladding layers are transparent to emitted light.
Whole laser chip is typically under 1 mm in each dimension.
232/253

Back
Close
Advanced Laser Structures
A waveguide is also formed in the lateral direction.
In a ridge-waveguide laser, ridge is formed by etching top cladding
layer close to the active layer.
SiO
2
ensures that current enters through the ridge.
Eective mode index is higher under the ridge because of low re-
fractive index of silica.
233/253

Back
Close
Buried Heterotructure Laser
Active region buried on all sides by cladding layers of lower index.
Several dierent structures have beeb developed.
Known under names such as etched-mesa BH, planar BH, double-
channel planar BH, and channelled substrate BH lasers.
All of them allow a relatively large index step (n > 0.1) in lateral
direction.
Single-mode condition requires width to be below 2 m.
Laser spot size elliptical (21 m
2
).
Output beam diracts considerably as it leaves the laser.
A spot-size converter is sometimes used to improve coupling
eciency into a ber.
234/253

Back
Close
Control of Longitudinal Modes
L
Single-mode operation requires lowering of cavity loss for a specic
longitudinal mode.
Longitudinal mode with the smallest cavity loss reaches threshold
rst and becomes the dominant mode.
Power carried by side modes is a small fraction of total power.
235/253

Back
Close
Distributed Feedback Lasers
Feedback is distributed throughout cavity length in DFB lasers.
This is achieved through an internal built-in grating
Bragg condition satised for = 2 n.
236/253

Back
Close
Distributed Bragg reector Lasers
End regions of a DBR laser act as mirrors whose reectivity is
maximum for a wavelength = 2 n.
Cavity losses are reduced for this longitudinal mode compared with
other longitudinal modes.
Mode-suppression ratio is determined by gain margin.
Gain Margin: excess gain required by dominant side mode to reach
threshold.
Gain margin of 35 cm
1
is enough for CW DFB lasers.
Larger gain margin (>10 cm
1
) needed for pulsed DFB lasers.
Coupling between DBR and active sections introduces losses
in practice.
237/253

Back
Close
Fabrication of DFB Lasers
Requires advanced technology with multiple epitaxial growths.
Grating is often etched onto bottom cladding layer.
A fringe pattern is formed rst holographically on a photoresist
deposited on the wafer surface.
Chemical etching used to change cladding thickness in a periodic
fashion.
A thin layer with refractive index n
s
< n < n
a
is deposited on the
etched cladding layer, followed with active layer.
Thickness variations translate into periodic variations of mode index
n along the cavity length.
A second epitaxial regrowth is needed to make a BH device.
238/253

Back
Close
239/253

Back
Close
Coupled-Cavity Structures
240/253

Back
Close
Tunable Semiconductor Lasers
Multisection DFB and DBR lasers developed during the 1990s to
meet conicting requirements of stability and tunability.
In a 3-section device, a phase-control section is inserted between
the active and DBR sections.
Each section can be biased independently.
Current in the Bragg section changes Bragg wavelength through
carrier-induced changes in mode index.
Current injected into phase-control section aects phase of feedback
from the DBR.
Laser wavelength can be tuned over 1015 nm by controlling these
two currents.
241/253

Back
Close
Tuning with a Chirped Grating
Several other designs of tunable DFB lasers have been developed.
In one scheme, grating is chirped along cavity length.
Bragg wavelength itself then changes along cavity length.
Laser wavelength is determined by Bragg condition.
Such a laser can be tuned over a wavelength range set by the grating
chirp.
In a simple implementation, grating period remains uniform but
waveguide is bent to change n.
Such lasers can be tuned over 56 nm.
242/253

Back
Close
Tuning with a superstructure Grating
Much wider tuning range possible using a superstructure grating.
Reectivity of such gratings peaks at several wavelengths.
Laser can be tuned near each peak by controlling current in phase-
control section.
A quasi-continuous tuning range of 40 nm realized in 1995.
Tuning range can be extended further using a 4-section device in
which two DBR sections are used.
Each DBR section supports its own comb of wavelengths but spac-
ing in each comb is not the same.
Coinciding wavelength in the two combs becomes the output wave-
length that can be tuned widely (Vernier eect).
243/253

Back
Close
Tuning with a Directional Coupler
A fourth section is added between the gain and phase sections.
It consist of a directional coupler with a superstructure grating.
Coupler section has two vertically separated waveguides of dierent
thickness (asymmetric directional coupler).
Grating selectively transfers a single wavelength to passive
waveguide in the coupler section.
A tuning range of 114 nm was produced in 1995.
244/253

Back
Close
Vertical-Cavity Surface-Emitting Lasers
VCSELs operate in a single longitudinal mode simply because mode
spacing exceeds the gain bandwidth.
VCSELs emit light normal to active-layer plane.
Emitted light is in the form of a circular beam.
245/253

Back
Close
VCSEL Fabrication
Fabrication of VCSELs requires growth of hundreds of layers.
Active region in the form of one or more quantum wells.
It is surrounded by two high-reectivity (>99.5%) mirrors.
Each DBR mirror is made by growing many pairs of alternating
GaAs and AlAs layers, each /4 thick.
A wafer-bonding technique is sometimes used for VCSELs operating
in the 1.55-m wavelength.
Chemical etching used to form individual circular disks.
Entire two-dimensional array of VCSELs can be tested without sep-
arating individual lasers (low cost).
Only disadvantage is that VCSELs emit relatively low powers.
246/253

Back
Close
Semiconductor Optical Ampliers
Reection feedback from end facets must be suppressed.
Residual reectivity must be <0.1% for SOAs.
Active-region stripe tilted to realize such low feedback.
A transparent region between active layer and facet also helps.
247/253

Back
Close
Gain Spectrum of SOAs
Measured gain spectrum exhibits ripples.
Ripples have origin in residual facet reectivity.
Ripples become negligible when G

R
1
R
2
0.04.
Amplier bandwidth can then exceed 50 nm.
248/253

Back
Close
Polarization Sensitivity of SOAs
Amplier gain dierent for TE and TM modes.
Several schemes have been devised to reduce polarization sensitivity.
249/253

Back
Close
SOA as a Nonlinear Device
Nonlinear eects in SOAs can be used for switching, wavelength
conversion, logic operations, and four-wave mixing.
SOAs allow monolithic integration, fan-out and cascadability, re-
quirements for large-scale photonic circuits.
SOAs exhibit carrier-induced nonlinearity with n
2
10
9
cm
2
/W.
Seven orders of magnitude larger than that of silica bers.
Nonlinearity slower than that of silica but fast enough to make
devices operating at 40 Gb/s.
Origin of nonlinearity: Gain saturation.
Changes in carrier density modify refractive index.
250/253

Back
Close
Gain Saturation in SOAs
Propagation of an optical pulse inside SOA is governed by
A
z
+
1
v
g
A
t
=
1
2
(1i
c
)g(t)A,
Carrier-induced index changes included through
c
.
Time dependence of g(t) is governed by
g
t
=
g
0
g

g[A[
2
E
sat
,
For pulses shorter than
c
, rst term can be neglected.
Saturation energy E
sat
= h(
m
/
g
) 1 pJ.
251/253

Back
Close
Theory of Gain Saturation
In terms of =t z/v
g
, A =

Pexp(i), we obtain
P
z
= g(z, )P(z, ),

z
=
1
2

c
g(z, ),
g

= g(z, )P(z, )/E

sat
.
Solution: P
out
() = P
in
()exp[h()] with h() =
_
L
0
g(z, )dz.
dh
d
=
1
E
sat
[P
out
() P
in
()] =
P
in
()
E
sat
(e
h
1).
Amplication factor G = exp(h) is given by
G() =
G
0
G
0
(G
0
1)exp[E
0
()/E
sat
]
,
G
0
= unsaturated amplier gain and E
0
() =
_

P
in
()d.
252/253

Back
Close
Chirping Induced by SOAs
Amplier gain is dierent for dierent parts of the pulse.
Leading edge experiences full gain G
0
because amplier is not yet
saturated.
Trailing edge experiences less gain because of saturation.
Gain saturation leads to a time-dependent phase shift
() =
1
2

c
_
L
0
g(z, )dz =
1
2

c
h() =
1
2

c
ln[G()].
Saturation-induced frequency chirp

c
=
1
2
d
d
=

c
4
dh
d
=

c
P
in
()
4E
sat
[G() 1],
Spectrum of amplied pulse broadens and develops multiple peaks.
253/253

Back
Close
Pulse Shape and Spectrum
(a) (b)
A Gaussian pulse amplied by 30 dB. Initially E
in
/E
sat
= 0.1.
Dominant spectral peak is shifted toward red side.
It is accompanied by several satellite peaks.
Temporal and spectral changes depend on amplier gain.
Amplied pulse can be compressed in a ber with
anomalous dispersion.