Vous êtes sur la page 1sur 92

Photonic

Materials
Frank Güell
Daniel Navarro

1
TABLE OF CONTENTS

3. Fundamentals of waveguides
· 1D and 2D waveguides: fundamentals and
experimental characterization.

Reference: Silicon photonics: an introduction


Graham T. Reed andAndrew P. Knights
The Ray Optics Approach to Describing Planar Waveguides

n2 θ2 Et

n1 θ1 θ1

Ei Er

Light rays refracted and reflected at the interface of two media

n1 sin θ1 = n 2 sin θ2

n2

n1

θc
n1 sin θ1 = n 2
n2
sin θc =
n1

n1

n2
Total internal reflection at two interfaces demonstrating 
the concept of a waveguide
Reflection Coefficients

Er = r.Ei
where ‘r’ is a complex reflection coefficient, which is
polarisation dependant

The Transverse Electric (TE) condition is defined as the


condition when the electric fields of the waves are
perpendicular to the plane of incidence.

Correspondingly, the Transverse Magnetic (TM) condition


occurs when the magnetic fields are perpendicular to the
plane of incidence.

n2 Et
interface
n1

Ei Er

Circles indicate that the electric fields are vertical (i.e. coming out of 
screen)
Orientation of electric fields for TE incidence at the interface
between 2 media.
The reflection coefficients rTE and rTM, are described by the
Fresnel formulae

For TE polarisation:

n1 cos θ1 − n 2 cos θ2
rTE =
n1 cos θ1 + n 2 cos θ2

For TM polarisation:
n 2 cos θ1 − n1 cos θ2
rTM =
n 2 cos θ1 + n1 cos θ2

Using Snell’s Law:

n1 cos θ1 − n 22 − n12 sin 2 θ1


rTE =
n1 cos θ1 + n 22 − n12 sin 2 θ1

And 

n22 cosθ1 − n1 n22 − n12 sin2 θ1


rTM =
n22 cosθ1 + n1 n22 − n12 sin2 θ1
If θ1 is greater than θc 
a − jb e − jα
r = e(
jφ )
= = jα = e − j 2α
a + jb e
and
r =1
Where φTE and φTM are given by:

2
⎛n ⎞
sin 2 θ1 − ⎜ 2 ⎟
φTE = 2 tan −1 ⎝ n1 ⎠
cos θ1
Which are negative indeed
and

n12 2
2
sin θ1 − 1
n2
φTM = 2 tan −1
n2
cos θ1
n1
But r relates reflected fields.  Power is described by 
the Poynting Vector

1 2 εm 2
S= E = E
Z μm

And reflected power is related by reflectance R:

Sr E2r
R = = 2 = r2
Si Ei

where E is electric field, εm is the permittivity of the


medium, μm is the permeability of the medium, and Z is the
impedance of the medium
Phase of a propagating wave and its wavevector

Let
E = E 0 exp[ j( kz ± ωt )]
and

H = H 0 exp[ j( kz ± ωt )]

Where z is the direction of propagation

Therefore, phase φ is

φ = kz ± ωt
The phase varies with time (t), and with distance (z). These
variations are quantified by taking the time derivative and the
spatial derivatives:
∂φ
= ω = 2πf
∂t
∂φ
=k
∂z
where ω is angular frequency (rads/sec), and f is frequency (Hz).
k is the wavevector (propagation constant) in the direction
of the wavefront. It is related to wavelength, λ, by:


k=
λ

In free space k = k0, and

k = nk 0

Hence in free space,


k0 =
λ0
Modes of a planar waveguide

y                                                
x                                       
n3 z

n1 k0n1
h

n2
The wavevector in a planar waveguide

We can decompose the wavevector k, into two components,


in the y and z directions.

θ1
k=n1k0
ky= n1k0cosθ1

kz= n1k0sinθ1

The relationship between propagation constants in the y, z, 
and wavenormal directions
We require self consistency condition. As the wave reflects
twice it reproduces itself. The total phase shift must be a
multiple of 2π, hence:

AC − AB = 2h cos θ1

2 k 0 n1h cos θ1 − φu − φl = 2 mπ

, since we have seen that reflection coefficients give a phase


change upon reflection. We have refered to these phase
shifts as φu, and φl respectively.
y                                                
x                                       
n3 z

n1
h

n2
The planar waveguide

Thus light propagates in discrete modes described by the 
polarisation and the mode number.
E.g. TE0, TE1, TM0, etc
Each mode will have a unique propagation constant in the 
y and z directions

2 k 0 n1 h cos θ 1 − ϕ u − ϕ l = 2 m π
The number of modes is limited by satisfaction of the 
requirements of total internal reflection.
The Symmetrical planar waveguide

In the symmetrical planar waveguide, n2 = n3, and hence 
φu=φl.  Therefore for TE polarisation, the equation 
becomes:

⎡ ⎛ n ⎞
2

⎢ sin θ1 − ⎜ ⎟ ⎥
2 2

⎢ ⎝ n1 ⎠ ⎥ = 2 mπ
2 k 0 n1h cos θ1 − 4 tan −1 ⎢ ⎥
cos θ1
⎢ ⎥
⎢⎣ ⎥⎦

This can be rearranged as :

⎡ ⎛ n2 ⎞ ⎤
2

⎢ sin θ1 − ⎜ ⎟ ⎥
2

⎡ k0n1h cosθ1 − mπ ⎤ ⎢ ⎝ n1 ⎠ ⎥
tan⎢ ⎥ =⎢ ⎥
⎣ 2 ⎦ cosθ1
⎢ ⎥
⎣⎢ ⎥⎦

Later we will solve this equation for angle θ1
We can find the approximate number of modes supported by 
the waveguide as follows:

The minimum value that θ1 can take is θc. i.e.

n2
sin θc =
n1

Hence the right hand side of the previous equation reduces


to zero and the equation becomes:

k 0 n1h cos θc − m max π


=0
2
rearranging for m, the mode number,

k 0 n1h cos θc
m max =
π

Number of modes = [mmax]int +1, since the lowest order 
mode (usually called the fundamental mode), has a mode 
number m=0. Note that the symmetrical waveguide is 
never cut‐off.
The Asymmetrical planar waveguide

y                                                
x                                       
n3 z

n1
h

n2
Propagation in an asymmetric planar waveguide

n2 ≠ n3, and φu ≠ φl , hence 

⎡ ⎛ n ⎞
2
⎤ ⎡ ⎛ n ⎞
2

⎢ sin θ1 − ⎜ ⎟ ⎥
2 2
⎢ sin θ1 − ⎜ ⎟ ⎥
2 3

[k0n1h cosθ1 − mπ] = tan−1 ⎢⎢ ⎝ 1 ⎠ ⎥ + tan−1 ⎢ ⎝ n1 ⎠ ⎥


n
cosθ1 ⎥ ⎢ cosθ1 ⎥
⎢ ⎥ ⎢ ⎥
⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦

Note that there is not always a solution for m=0, hence the 
asymmetrical guide may be cut‐off.
Solving the eigenvalue equation for symmetrical and
asymmetrical waveguides
Let n1 = 1.5, n2 = 1.49, n3 = 1.40, λ0= 1.3μm, and h = 0.3μm,
and TE polarisation.

33
k0n1hcosθ1 θu θl
Phase  2φu
Change f( θ ) 22
φ θ2( θ ) φu +φl φu
(radians)θ1( θ ) φl
θ3( θ )
1
θ4( θ )
1
Solution for θ1

0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 π
0        0.2        0.4       0.6        0.8        1         1.2       1.4      π/2
θ 2

Fundamental mode angle θ1 (radians)

Solution of the eigenvalue equation for m=0

Note that the asymmetrical waveguide is cut‐off, whereas the 
symmetrical waveguide is not 
Now consider a silicon waveguide:

n1 = 3.5 (silicon), n2 = 1.5 (silicon dioxide), n3 = 1.0 (air),


λ0= 1.3μm, and h = 0.15μm.
33
θu θl
k0n1hcosθ1
Phase  f( θ ) 22
Changeθ2( θ )
φ θ1( θ ) 2φl
(radians)
θ3( θ )
1
φu +φl
θ4( θ )
1

φu
φl
0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 π
0        0.2        0.4       0.6        0.8        1         1.2       1.4      π/2
θ 2

Fundamental mode angle θ1 (radians)

Solution of the eigenvalue equation for m=0 
(silicon‐on –insulator)

Note that the symmetrical and asymmetrical waveguides are 
similar
Monomode Conditions

Consider the TE polarisation eigenvalue equation for a


symmetrical waveguide:

⎡ ⎛ n2 ⎤

2

⎢ sin θ1 − ⎜ ⎟ ⎥
2

k n h cos θ1 − mπ ⎤ ⎢ ⎝ n1 ⎠ ⎥
tan ⎡⎢ 0 1 =
⎥⎦ ⎢ ⎥
⎣ 2 cos θ1
⎢ ⎥
⎢⎣ ⎥⎦

For the second mode, m=1, and the equation becomes:

k 0 n1h cos θc − π ⎤
tan ⎡⎢ ⎥⎦ = 0
⎣ 2
i.e.

π λ
cos θc = = 0
k 0 n1h 2 n1h
Hence for monomode conditions

⎛ λ ⎞
θc < cos −1 ⎜⎜ 0 ⎟⎟
⎝ 2 n1h ⎠
Effective Index

We have seen that propagation constants in the z and y 
directions are:
k z = n1k 0 sin θ1

and

k y = n1k 0 cos θ1

kz is also known as β, the propagation constant in the


direction of the waveguide
Now define a parameter N, called the effective index of the
mode, such that:

N = n1 sin θ1

i.e.

k z = β = Nk 0
The lower bound on β is determined by the larger of the
critical angles of the waveguide, usually at the lower
interface:

β ≥ n1k0 sinθl = k0n2

The upper bound on β is governed by the maximum value of θ,


which is 90o. In this case β = k =n1k0. Hence:

k 0 n1 ≥ β ≥ k 0 n 2

Substituting for β = k =Nk0:

n1 ≥ N ≥ n 2
M-line technique
Si detector

ϑ0 n0

n1

ϑ1

n2
ϑ2

ns
M-line technique
Detector

λ=633nm

5
Intensity in the detector (a.u.)

3
1.12 1.16 1.20 1.24 1.28 1.32 1.36 1.40
effective refractive index
M-line technique

neff
1.15 1.20 1.25 1.30
1.2
Normalised intensity

0.9
0.6
TEexp
0.3
TE sim
1.0

0.5 TM exp
TM sim

1.4 1.239 no
1.255
ne
nmat

1.2
drift=1.3%
1.0 1.15

0 2000 4000 6000 8000 10000 12000

thickness (nm)

Nice agreement between experiments and


simulations
Just a little electromagnetic theory

Starting with Maxwell’s equations, if we assume a loss‐
less, non conducting medium, limit ourselves to 
propagation in the z direction, and consider one 
polarisation at a time (TE or TM),we can derive a scalar 
equation describing wave propagation in our planar 
waveguide:

∂2Ex ∂2Ex ∂2 Ex
+ 2 = μ mεm 2
∂y 2
∂z ∂t

where εm is the permittivity of the waveguide, μm is the


permeability of the waveguide, and in this case the electric
field polarised in the x direction corresponds to TE
polarisation. This is called the scalar wave equation.
Since there is only electric field in the x direction the general
solution of this field as:

E x = E x ( y)e − jβz e jωt

This means that there is a field directed (polarised) in the x


direction, with a variation in the y direction yet to be
determined, propagating in the z direction with propagation
constant β, with sinusoidal (ejωt) time dependence. The
waveguide configuration is:

y     x                                       
n3 z
y=h/2
h
y=0
n1 k0n1
y=‐h/2
n2

Propagation in an asymmetric planar waveguide
Solution of the wave equation provides us with
expressions of the electric fields in the core and
claddings:
In the upper cladding: h
− k yu ( y − )
E ( y) = E e
x u
2 For y≥(h/2)

In the core we will have

− jk yc y
E x ( y) = E c e For -(h/2) ≤y≤ (h/2)

In the lower cladding we will have:

h
k yl ( y + )
E x ( y) = E l e 2 For y ≤ -(h/2)

k 2 = k 2 n 2 − β2
Where                              
yi 0 i

Note that n and ky are written ni and kyi because they can
now represent any of the three media (core, upper cladding,
or lower cladding), by letting i=1, 2, or 3.
Propagation constants again

In solving the wave equation we assumed field solutions


with propagation constants in the core, and the upper and
lower claddings, but we didn’t discuss why the propagation
constants took the form they did.

Our general solution was of the form:

− k y y − jβ z jωt
Ex = E c e e e

However, in the three media, core, upper cladding and lower


cladding, the propagation constant ky took different forms.
In the claddings ky was a real number, whereas in the core
ky was an imaginary number

•Total internal reflection occurs
•field decays in cladding
•part of the field propagates in the cladding
Mode Profiles

Now that we have field solutions for modes in the planar


waveguide, we can plot the field distribution, Ex(y) or the
intensity distribution, ⎟Ex(y)⎟2. Consider the following
waveguide:

n1 = 3.5, n2 = 1.5, n3 = 1.0, λ0= 1.3μm, and h = 1.0μm.

For even functions, the field in the core is a cosine


function, and in the claddings, an exponential decay.
Therefore, solutions for m=0, and m=2 are even functions,
and are plotted below. The field distribution is plotted for
m = 0, and the intensity distribution is plotted for m = 2.
1
1

0.80.8

0.6
NormalisedEC0( y ) 0.6
Electric 
field Ex(y) EU0(0.4
y2 )

EL0( y1 )
0.4

0.2

0.2

0
‐1                 ‐0.5                   0                   0.5                  1
1 10
6
5 10
7
0 5 10
7
1 10
6

y , y2 , y1
Distance in the y direction ( μm)

Electric field profile of the fundamental mode (m=0)
1 1

0.80.8

0.6
Normalised  0.6
IC2
(y)
Intensity 
⎥Ex(y)⎥2 IU2
(y2)
0.4
IL2
(y1)
0.4

0.2
0.2

0 6 7 7 6
‐1                 ‐0.5                   0                   0.5                  1
110 510 0 510 110
y,y2y1
,
Distance in the y direction 
(μm)

Intensity profile of the 2nd even mode (m = 2)
Confinement factor

How much of the power propagates inside the core?  
We can define a confinement factor Γ:

h/2
∫ E ( y)dy
2
x
Γ= −h / 2

∫ E ( y)dy
2
x
−∞
3.0 2D waevguides
Silicon on Insulator Waveguides

silicon substrate
SiO2
Surface guiding layer

w
n3

h n1
r

SiO2(n2)
Modes of two dimensional waveguides

Modes are designated

x y
E p ,q E p,q
or

HEp,q EHp,q
P maxima in the x direction, q maxima in the y direction

Hence Fundamental mode:

x y
E1,1 or
E1,1
Sometimes labelled

x y
E 0,0 E0,0
The Effective Index Method of analysis

y
n3
x

n1 h

First solve: Then solve

n3
n1 h n3  neff n3

n2
w
The Effective Index Method of analysis

EXAMPLE
3.5μm

n3 =1.0
5μm
3μm n1 =3.5

n2=1.5

Find effective index Nwg?
Solution

Firstly decompose the rib structure into vertical and horizontal


planar waveguides This means we must first solve the
horizontal planar waveguide. Since the polarisation is TE, we
use the asymmetrical TE eigenvalue equation:

⎡ ⎛ n ⎞
2
⎤ ⎡ ⎛ n ⎞
2

⎢ sin θ1 − ⎜ ⎟ ⎥
2 2
⎢ sin θ1 − ⎜ ⎟ ⎥
2 3

−1 ⎢ ⎝ n 1 ⎠ ⎥ −1 ⎢ ⎝ n1 ⎠ ⎥
[k 0 n1h cos θ1 − mπ] = tan ⎢ ⎥ + tan ⎢ ⎥
cos θ1 cos θ1
⎢ ⎥ ⎢ ⎥
⎢⎣ ⎥⎦ ⎣⎢ ⎥⎦
n3 =1.0

n1= 3.5 h = 5μm

n2 = 1.5

First planar waveguide of the decomposed rib structure

This gives a fundamental mode propagation angle of 
87.92o (1.53456 radians), and neffg = n1sinθ1 = 3.4977
We now need to solve the second planar waveguide of the
decomposed rib structure:

(i) First find the effective index of the planar regions


either side of the core.

(ii) Solve the asymmetrical TE eigenvalue equation


again, for r = 3μm.

This yields a propagation angle of 86.595o, and an effective 
index for the planar region of neffp = 3.4938.

neffp = 3.4938 neffg=  neffp = 3.4938


3.4977

w = 3.5μm

Second planar waveguide of the decomposed rib structure
We can now solve the second decomposed planar
waveguide of the decomposed rib structure, using the
effective indices just calculated. This time the symmetrical
TM eigenvalue equation should be used:

⎡ 2 ⎤
⎢ ⎛ n effg ⎞
⎜ ⎟ sin 2 θwg − 1 ⎥
⎢ ⎜n ⎟ ⎥
k 0 n effg w cos θ wg −1
= 2 tan ⎢ ⎝ effp ⎠ ⎥
⎢ ⎛ n effp ⎞ ⎥
⎜ ⎟ cos θwg
⎢ ⎜n ⎟ ⎥
⎢⎣ ⎝ effg ⎠ ⎥⎦

Hence θwg is be found. The solution is θwg= 88.285o, which


corresponds to an effective index of Nwg= 3.496. Knowing
the effective index we can evaluate all of the propagation
constants of the waveguide. For example we are
particularly interested in β:

β = k 0 N wg = 16.897μm −1

Note:   High confinement in silicon
Different degrees of confinement horizontally & 
vertically 
Large single mode rib waveguides

Firstly consider a 5μm planar waveguide:
n3 =1.0

n1= 3.5 h = 5μm

n2 = 1.5

5μm planar waveguide

From section 1, the approximate number of modes can be 
found as:
k 0 n1h cos θc
m max =
π
Since

1 .5
θc = sin −1 = 25.4 o
3.5
2 x 3.5x 5x10 −6 cos θc
[ m max ]int = = 24
λ0
and hence 25 modes are supported (including the m=0 mode).
Alternatively consider the thickness required to make such a 
waveguide single mode.  From section 1:

⎛ λ ⎞
θc < cos −1 ⎜⎜ 0 ⎟⎟
⎝ 2 n1h ⎠

Rearranging for h, and substituting for cos θc:

⎛ ⎞
⎜ ⎟
⎜ λ0 / 2 ⎟
h<⎜ 2 ⎟
⎜ ⎛ n2 ⎞ ⎟
n
⎜ 1 1 − ⎜ ⎟ ⎟
⎝ ⎝ n1 ⎠ ⎠

Therefore, for single mode operation in an SOI waveguide, at a


wavelength of 1.3μm, h needs to be less than approximately 0.2μm.
It is therefore a little surprising that rib waveguides with 
cross sectional dimensions of several microns are 
routinely described as single mode waveguides.

The answer to this apparent paradox is that, if the


waveguide is correctly designed, higher order modes leak
out of the waveguide over a very short distance, as
demonstrated by Soref et al, in 1991.

Firstly the authors defined a rib waveguide in terms of


some normalising parameters:
2aλ

n0

2bλ n1
2brλ

n2

Rib Waveguide definitions


The authors limited their analysis to waveguides in which
0.5 ≤ r < 1.0, because for r ≥ 0.5, the effective index of
‘vertical modes’ in the planar region either side of the rib,
becomes higher than the effective index of all vertical
modes in the rib, other than the fundamental.

They then used an effective index approach to define a


parameter related to the aspect ratio of the rib waveguide,
a/b. They then found the limiting condition such that the
EH01and HE01 just failed to be guided. This resulted in a
condition for the aspect ratio:

a r
≤ 0 .3 +
b 1 − r2

In order to demonstrate this they simulated excitation of higher
order modes and watched them leak out of the waveguide. 
Pogossian et al took the experimental data of Rickman &
Reed, and fitted an equation of the form:
a r
≤c+
b 1− r2
c being approximately 0, resulting in a modified equation:
a r

b 1− r2

a/b

r
The singlemode condition
Refractive index and loss coefficient in optical waveguides

Complex refractive index can be defined as:


n ' = n R + jn I
And we have expressed a propagating field as:

E = E 0e j( kz − ωt )
Substituting for propagation constant and complex 
refractive index:

E = E 0 e j( k 0n 'z − ωt ) = E 0 .e jk 0n R z .e − k 0n I z .e − jωt
1
The term exp(-k0nIz) is often re-designated exp( − α) .
2
The term α is called the loss coefficient. The factor of ½ is
included in the definition above because α is an intensity
loss coefficient. Therefore we can write :

− αz
I = I 0e
The benchmark for acceptable waveguide loss is of the order 
of 1dB/cm.  Typical losses for SOI waveguides are in the range 
0.1 ‐ 0.5 dB/cm
The contributions to loss in an optical waveguide
Losses result from scattering, absorption, and radiation:

Scattering

Scattering in an optical waveguide can result from two


sources: volume scattering and interface scattering

Volume scattering follows either a λ‐3 dependence, or a 
λ‐1 dependence, as a consequence of the type and 
concentration of scattering centres.

Interface scattering has been modelled by many authors, 
but a reasonably accurate and attractively simple model 
was produced by Tien:

⎛ ⎞
⎜ ⎟
2
⎛ 1

cos3 θ ⎜ 4 πn1 ( σ 2u + σ ) ⎟ ⎜
2 2
1 ⎟
αs = ⎜
l

2 sin θ ⎜ λ0 ⎟ ⎜h+ 1 + 1 ⎟
⎝ ⎠ ⎜ k yu k yl ⎟⎠

where σu is the rms roughness for the upper waveguide
interface, σl is the rms roughness for the lower waveguide
interface, kyu is the decay constant in the upper cladding, kyl is
decay constant in the lower cladding, and h is the waveguide
thickness
Example of interface scattering
Consider a planar waveguide:

n1 = 3.5; n2 = 1.5; n3 = 1.0; h = 1.0μm, and the operating


wavelength λ0 = 1.3μm. Let us compare the scattering loss of
two different modes of the waveguide, say the TE0 and the TE2
modes.

The propagation angles, θ1 of these two are 80.8o (TE0), and


60.7o (TE2). Thus the decay constants in the claddings are
given by: k yi = β − k 0 n i
2 2 2 2

Therefore we can evaluate the decay constants for each mode are
as follows:
TE0 TE2

k yu = β 2 − k 02 n 32 = 15.98 μm-1 k yu = β 2 − k 02 n 32 = 13.94 μm


-1

k yl = β 2 − k 02 n 22 = 15.04 μm-1 k yl = β 2 − k 02 n 22 = 12.85 μm


-1
If we now let both σu and σl equal 1nm, we can evaluate the
scattering loss for each mode, for each nanometre of rms
roughness at each interface.

For the TE0 mode,

⎛ ⎞
⎞ ⎜ ⎟
2
⎛ 1
cos θ ⎜ 4 πn1 ( σ u + σ ) ⎟ ⎜
3 2 2 2
1 ⎟ = 0.04cm −1
αs = l

2 sin θ ⎜⎜ λ0 ⎟ ⎜
⎟ h+ 1 + 1 ⎟
⎝ ⎠ ⎜ k yu k yl ⎟
⎝ ⎠
This is equivalent to a loss of 0.18 dB/cm.

For the TE2 mode


⎛ ⎞
⎜ ⎟
2
⎛ 1

cos3 θ ⎜ 4 πn1 ( σ 2u + σ ) ⎟
2 2
⎜ 1 ⎟ = 1.33cm −1
αs = l

2 sin θ ⎜⎜ λ0 ⎟
⎟ ⎜
h +
1
+
1 ⎟
⎝ ⎠ ⎜ k yu k yl ⎟
⎝ ⎠

This is equivalent to a loss of 5.79 dB/cm.


Absorption

The two main potential sources of absorption loss for


semiconductor waveguides are band edge absorption and
free carrier absorption. If we operate at a wavelength well
away from the band edge, the former is negligible.

Changes in free carrier absorption can be described by


Drude-Lorenz equation:

e 3λ20 ⎛ N e Nh ⎞
Δα = 2 3 ⎜⎜ + * 2 ⎟

4 π c ε0 n ⎝ μ e ( m ce ) μ h ( m ch ) ⎠
* 2

where e is the electronic charge; c is the velocity of light in


vacuum; μe is the electron mobility; μh is the hole
mobility; m*ce is the effective mass of electrons; m*ch is
the effective mass of holes; Ne is the free electron
concentration; ; Nh is the free hole concentration; ε0 is the
permittivity of free space; and λ0 is the free space
wavelength.
Silicon
λ=1.3μm

1018 cm‐3 electrons 


and holes introduces 
an additional loss of ~ 
2.5cm‐1.  i.e 10.86 
dB/cm!

Additional loss of silicon due to free carriers
Radiation Losses in optical waveguides

Ideally negligible.  Possibility of radiation via leaky 
modes or curvature at too fast a rate.
Coupling to the Optical circuit

Coupling light to a waveguide is non trivial.  The 
main methods are:
Input beam
Input 
beam

waveguide waveguide

(a) prism coupling (b) grating 
coupling

Input  waveguide
waveguide
bea
m
lens

Optical  (d) end‐fire coupling
fibre
(c) butt coupling

Four techniques for coupling light to optical waveguides
Grating couplers
Grating couplers allow individual mode selection.  
In order to couple light to the waveguide, the 
propagation constants of the exciting beam and the 
waveguide must be matched.

Consider a light ray incident upon a waveguide 
surface:

n3 θa z

n1 β

n2
Light incident upon the surface of a waveguide

In medium n3 the propagation constant is k0n3.  The z 
directed propagation constant in medium n3 will be:

k z = k 0 n 3 sin θa
Therefore the phase-match condition will be
β = k z = k 0 n 3 sin θa
where β is the waveguide propagation constant

But β ≥ k0n3.  Therefore the condition can never be met, 
since sinθa will be less than unity.  This is why a grating is 
required to couple light into the waveguide.

The periodic nature of the grating causes a periodic


modulation of the effective index of the waveguide. For an
optical mode with propagation constant βW when the grating
is not present, the modulation results in a series of possible
propagation constants, βp given by

2 pπ
βp = βW +
Λ

where Λ is the period of the grating, and p = ±1, ±2, ±3, etc.
Only the negative values of p can result in a phase match.
It is usual to fabricate the grating such that only p = -1
results in a phase match with a waveguide mode.
Therefore the waveguide propagation constant becomes:


βp = βW −
Λ

And the phase match condition becomes:


βW − = k 0 n 3 sin θa
Λ

Writing βW in terms of the effective index , N,


k0N − = k 0 n 3 sin θa
Λ
and substituting for k0

λ
Λ=
N − n 3 sin θa
If medium n3 is air, n3 = 1 and

λ
Λ=
N − sin θa

Because the refractive index of silicon is large, the period of


grating couplers in silicon is of the order of 400nm. The highest
coupling efficiency from gratings in silicon to date has been
reported by Ang et al., who reported an output coupling efficiency
of approximately 70% for rectangular gratings and 84% for a non
symmetrical profile. One of their devices is shown below:

Waveguide coupler fabricated in an SOI waveguide
Butt coupling and End fire coupling

The efficiency with which the light is coupled into the


waveguide is a function of (i) how well the fields of the
excitation and the waveguide modes match; (ii) the degree
of reflection from the waveguide facet; (iii) the quality of
the waveguide endface; (iv) and the spatial misalignment of
the excitation and waveguide fields.

Overlap of excitation and waveguide fields

The overlap integral Γof two fields E and ε, is given by

∞ ∞
∫ dy ∫ E.ε.dx
Γ= −∞ −∞
1
∞ ∞ ∞ ∞
⎡ dy E 2 dx. dy ε 2 dx ⎤ 2

⎢⎣ −∫∞ −∫∞ ∫
−∞

−∞ ⎥⎦

The factor Γ lies between 0 and 1, and therefore represents


the range between 0 coupling and 100% coupling due to
field overlap.
The problem is often simplified to the overlap of two 
gaussian functions. Let 
⎡ ⎛ x 2 y 2 ⎞⎤
E = exp ⎢ − ⎜⎜ 2 + 2 ⎟⎟⎥
⎢⎣ ⎝ ωx ωy ⎠⎥⎦

This represents a waveguide field with 1/e widths in the x


and y direction of 2ωx and 2ωy respectively. Similarly let

⎡ (x 2 + y 2 )⎤
ε = exp ⎢ −
⎣ ω02 ⎥⎦

This represents a circularly symmetrical input beam.

Using the mathematical identity for a definite integral:


π
∫ exp[− r x ].dx =

2 2

0 2r
The overlap integral reduces to:

2 ⎡ 1 ⎤ 2

⎢ ⎥
ω0 ⎣ ωx ωy ⎦
Γ= 1 1

⎡1 1⎤ ⎡1
2
1⎤ 2

⎢ ω2 + ω2 ⎥ ⎢ ω2 + ω2 ⎥
⎣ x 0⎦ ⎣ y 0⎦
Since the overlap integral describes the coupling efficiency of
the field profiles, the power coupling efficiency is given by:

4 ⎡ 1 ⎤
⎢ ⎥
ω20
⎣ ω ω ⎦
Γ2 =
x y

⎡1 1 ⎤⎡ 1 1⎤
+ +
⎢ ω2 ω2 ⎥ ⎢ ω2 ω2 ⎥
⎣ x 0 ⎦⎣ y 0⎦

A few examples show how coupling efficiency varies:

ω0 ωx ωy Γ Γ2 Loss due to
Γ2 (dBs)

5μm 5μm 5μm 1.0 1.0 0

10μm 10μm 5μm 0.894 0.8 0.97

20μm 16μm 3μm 0.535 0.286 5.4

20μm 1μm 1μm 0.1 0.01 20

5μm 5μm 3μm 0.939 0.882 0.55

5μm 4.8μm 4.9μm 0.999 0.999 0.004


Reflection from the waveguide facet

Using the Fresnel equations from section 1. 
Reflection coefficient for TE polarisation was:

n1 cos θ1 − n 2 cos θ2
rTE =
n1 cos θ1 + n 2 cos θ2
Similarly, the reflection coefficient rTM was:

n 2 cos θ1 − n1 cos θ2
rTM =
n 2 cos θ1 + n1 cos θ2
Using Snells Law rTE reduces to:

− sin( θ1 − θ2 )
rTE =
sin( θ1 + θ2 )

Hence the reflectivity R is given by:

sin 2 ( θ1 − θ2 )
R TE = rTE =
2

sin 2 ( θ1 + θ2 )
Similarly RTM can be found to be

tan 2 ( θ1 − θ2 )
R TM = rTM 2
=
tan 2 ( θ1 + θ2 )

The two functions are plotted below for an air/silicon


interface (i.e. n1= 1.0, n2 =3.5).

1 1

0.80.8
Reflectance

0.60.6 TE
RTE( θ1 )

RTM( θ1 )
0.40.4

TM
0.20.2

0 0
0      0.2       0.4      0.6       0.8      1.0      1.2      1.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4
θ1
Incident angle θ1 (radians)

Reflection at an air/silicon interface
At normal incidence (θ1=0), the reflection of both TE and TM
polarisations is the same. Furthermore, end-fire coupling
introduces light at near normal incidence. Consequently the
approximation is usually made that the Fresnel reflection at the
waveguide facets is that due to normal incidence. In this case,
reflectivity is:
2
n − n2
R= 1
n1 + n 2

For a silicon/air interface this reflection is approximately


31%, which introduces an additional loss of 1.6dBs. A loss
of 1.6dBs for each facet of the waveguide is considerable,
and is reduced in commercial devices by the use of anti-
reflection coatings.
An anti‐reflection coating has a thickness of λ/4, reducing or 
eliminating the reflection. For normal incidence, the net 
reflectivity R is given by: 

2 2
n1n 2 − n
R= ar

n1n 2 + n 2
ar

where nar is the refractive index of the anti-reflection


coating. R will be zero if:

n1n 2 = n ar2

For a silicon/air interface, nar needs to be approximately


1.87.

For silicon nitride (Si3N4), n = 2.05.

For silicon oxynitride (SiOxNy), n ranges from 1.46 (SiO2)


to 2.05 (Si3N4)
The quality of the waveguide endface

Three main options are available for endface preparation of


semiconductor waveguides: cleaving; polishing; etching.

Polishing is probably the most common method of


preparing a waveguide facet. The sample endface is lapped
with abrasive materials with sequentially decreasing grit
sizes, and can result in an excellent surface finish, although
‘rounding’ of the endface is a common failing.

Endfaces can be prepared by chemical or dry etching. The


details are beyond the scope of this course, but suffice to say
it is a technique that can be developed to a sufficiently high
level for commercial applications.
“The edge problem”
polished edge

clived edge
Spatial misalignment of the excitation and waveguide
fields

Alignment is of critical importance due to the very small


dimensions involved. Recall the overlap of exciting and
mode fields:
∞ ∞
∫ dy ∫ E.ε.dx
Γ= −∞ −∞
1
∞ ∞ ∞ ∞
⎡ dy E 2dx. dy ε 2dx ⎤ 2

⎢⎣ −∫∞ −∫∞ ∫
−∞

−∞ ⎥⎦

where

⎡ ⎛ x 2 y 2 ⎞⎤ ⎡ (x 2 + y 2 )⎤
E = exp ⎢ − ⎜⎜ 2 + 2 ⎟⎟⎥ ε = exp ⎢ −
ω02 ⎥⎦
and
⎣⎢ ⎝ ωx ωy ⎠⎦⎥ ⎣

Let us introduce and offset, A,  into one of the field equations:

i.e.

⎡ (x 2 + ( y − A ) 2 )⎤
ε = exp ⎢ − ⎥
⎣ ω02 ⎦
This expression can be re-written as:

⎡ (x 2 + ( y 2 + A 2 − 2 Ay) )⎤
ε = exp ⎢ − ⎥
⎣ ω02 ⎦

Subsequently, the overlap integral equation can be


manipulated into the form:

∞ ∞

⎡ A 2
⎤ ∫ dy ∫ E.εdx
Γ I = exp ⎢ − 2 2⎥
. −∞ −∞

⎣ ω + ω ⎦⎡ ∞ ∞ ∞ ∞
1
y 0
ε ⎤2
⎢⎣ −∫∞ −∫∞ ∫ ∫
2 2
dy E dx . dy dx
−∞ −∞ ⎥⎦
i.e.

⎡ A2 ⎤
Γ = exp ⎢ − 2
I
2⎥
Γ
⎣ ωy + ω0 ⎦
Therefore we can evaluate the term

⎡ A2 ⎤
offset = exp ⎢ − 2 2⎥
⎣ ω y + ω0 ⎦
Offset2 is also evaluated as it represents to additional loss
when considering power transmission (coupling).

The terms are evaluated below for a range of offsets A,


when ωy = ω0 = 5μm
1
1

0.8
Offset  and Offset2

0.6
C(A )

D(A )
0.4
Offset
Offset2
0.2

0 0
2 .10 4 .10 6 .10 8 .10 1 .10 1.2 .10 1.4 .10
6 6 6 6 10          12           14     
0             2            4             6            8          5 5 5
0
0 A −
Offset A (μm) 15⋅ 10

The effect of an offset A on the electric field 
overlap and the power coupling efficiency

An offset of A=2μm produces ‘offset2 of 0.85 ≡ 0.7dB


Measurement of propagation loss is integrated
optical waveguides

There is often confusion between insertion loss and


propagation loss

Insertion loss and propagation loss

The insertion loss of a device, is the total loss associated


with introducing that element into a system, and includes
the inherent loss and the coupling losses.

Alternatively, the propagation loss is the loss associated


with propagation in the waveguide alone – i.e.
measurement of loss coefficient, α.

There are three main experimental techniques associated


with waveguide measurement. These are (i) the cut–back
method; (ii) the Fabry-Perot resonance method; and (iii)
scattered light measurement.
Insertion losses in a waveguide:
−α prop L −α prop L
I out = I o e = I inCe
⎛ I in ⎞ ⎛1⎞
α coupling (dB) = −10 log ⎜ ⎟ = −10 log ⎜ ⎟
⎝ Io ⎠ ⎝C ⎠

Origin of the coupling losses


Dimension fiber-waveguide (modal mismatch).

Intrinsic Numerical Aperture

Fresnel Reflection

Misalignement

Extrinsic Defects on the facets

Scattering
Insertion losses in a waveguide:
−α prop L −α prop L
I out = I o e = I inCe
1 I in ⎛1⎞
= → −10 log ⎜ ⎟ = α coupling (dB)
C Io ⎝C ⎠

Origin of the propagation losses


Absorption

Scattering

Radiation
Positioning system
–For the fiber.
• Nanopositioning system with Piezoelectric actuators
(precision ∼1nm)

–For the sample


• Micropositioning system
The cut-back method

The cut back method is conceptually simple. A waveguide


of length L1 is excited by one of the coupling methods
mentioned, and the output power from the waveguide, I1,
and the input power to the waveguide, I0 are recorded. The
waveguide is then shortened to another length, L2, and the
measurement repeated to determine I2. Hence:

I1
= exp(−α(L1 − L 2 ))
I2

i.e.

⎛ 1 ⎞ ⎡ I2 ⎤
α = ⎜⎜ ⎟⎟ ln ⎢ ⎥
⎝ L1 − L 2 ⎠ ⎣ I1 ⎦
Cut-back technique
I = I0 e-αL

Laser in
Waveguide

L1 Fotodiode

L
L1
Cut-back technique
I = I0 e-αL

Laser in
Waveguid
e

L2 Fotodiode

L2 L1
L
Cut-back technique

I = I0 e-αL

Laser
Waveg
in
uide

L3 Fotodi
ode

L3 L2 L1
L
Cut-back technique

I = I0 e-αL

Laser in Waveg
uide

L Fotodio
4
de
ln (I)= ln(I0) - αL
I
ln(I)
ln(I)=ln(I0)=ln(Iin)+ln(C)
Measurable
Estimaton of
the coupling losses

L
L=0 L4 L3 L2 L1
The accuracy of the technique can be improved by taking
multiple measurements and plotting a graph

Optical 
Loss
(dBs)

Propagation length 
(cm)
Note that the data presented is now insertion loss for each
device length, not propagation loss. Hence the loss at zero
propagation loss represents coupling loss.

A useful variation of the cut-back method, is to carry out an


insertion loss measurement for a single waveguide length,
and calculate the coupling. Whilst this technique is less, it
has the enormous advantage of being non-destructive
Similar technique

Losses from a curve

Propagation losses

Equivalent to cut-back but without the coupling losses


problems
The Fabry-Perot resonance method

An optical waveguide with polished end faces (facets), is


similar in structure to the cavity of a laser, and may be
regarded as a potentially resonant cavity. Such a cavity is
called a Fabry-Perot cavity. The optical intensity
transmitted through such a cavity, It, is related to the
incident light intensity, I0, by the well known equation:

It (1 − R ) 2 e − αL
=
I 0 (1 − Re −αL ) 2 + 4 Re −αL sin 2 ( φ )
2
where R is the facet reflectivity, L is the waveguide
length, α is the loss coefficient, and φ is the phase
difference between successive waves in the cavity
This transfer function has a maximum value when φ=0 (or
multiples of 2π), and a minimum value when φ=π. i.e. :

Imax (1 − R)2 e−αL


=
I0 (1 − Re−αL )2
Imin (1 − R)2 e−αL (1 − R)2 e−αL
= −αL 2 −αL
=
I0 (1 − Re ) + 4 Re (1 + Re−αL )2
Therefore the ratio of the maximum intensity to minimum
intensity, ζ, is:

I max (1 + Re − αL ) 2
ζ= =
I min (1 − Re −αL ) 2

Which we can rearrange as:

1 ⎡ 1 ζ − 1⎤
α = − ln ⎢
L ⎣ R ζ + 1⎥⎦

If we know the reflectivity, R, if we can measure ζ, the 
loss coefficient can be evaluated.
Therefore if we can sweep through a few cycles of 2π, ζ
can be measured.   Such cycling can be achieved 
thermally, or by varying the wavelength of the light 
source.
1
1

0.8
R=0.1

FP1 ( θ ) 0.6
ItFP2
/I0( θ )
FP3 ( θ ) R=0.31
0.4

0.2
Imax
Imin R=0.5
0 0
0 5 10 15 20 25
0 φ (radians)
θ 8⋅ π

Plot of the Fabry-Perot transfer function for


three different mirror reflectivities
0.65

0.60

0.55
intensity(arb. units)

0.50

0.45

0.40

0.35

0.30

0.25

1550.00 1550.05 1550.10 1550.15 1550.20 1550.25


wavelength(nm)

Fabry‐Perot scan of a single mode 
waveguide
0.9

0.8
intensity(arb. units)

0.7

0.6

0.5

0.4
1550.00 1550.05 1550.10 1550.15 1550.20 1550.25

wavelength(nm)

Fabry‐Perot scan of a multimode 
waveguide
Scattered Light Measurement

The measurement of scattered light from the surface of a


waveguide can be used to determine the loss. The assumption is
that the amount of light scattered is proportional to the
propagating light. Therefore, the rate of decay of scattered light
with length will mimic the rate of decay of light in the waveguide.

However, it is clear that light is only scattered significantly if the


loss of the waveguide is high, so this approach has limited uses.
Scattering light collection

Si3N4 channel λ =780nm


α = 0.1 ± 0.07 dB/cm
Ln Intensity [a.u.]

The sample has to be quite lossy

0 5 10 15 20 25 30 35
Length [mm]
Examples of micromachining: 

Optical Fibre in a V‐groove

Vous aimerez peut-être aussi