2

Ring Resonators: Theory and Modeling

One of the rst papers dealing with the simulation of an integrated ring
resonator for a bandpass lter has been published in 1969 by Marcatili. The
layout of the channel dropping lter which he proposed is shown in Fig. 2.1.
This can be regarded as the standard conguration for an integrated ring
resonator channel dropping lter. Two straight waveguides also known as the
bus or the port waveguides are coupled either by directional couplers through
the evanescent eld or by multimode interference (MMI) couplers to the ring
resonator. A simpler conguration is obtained, when the second bus or port
waveguide is removed. Then the lter is typically referred to as notch lter
because of the unique lter characteristic. In the following chapter, the ring
resonator simulation model is described beginning with the basic notch con-
guration and adding more bus waveguides and ring resonators to eventually
build a multiple coupled ring resonator lter. Dierent types of ring resonator
simulation models will be explained, so as to be able to chose from a range of
models which best suit the need.

2.1 Single Ring Resonators

2.1.1 Ring Structure

The Basic Conguration

The basic conguration, which consists of unidirectional coupling between a

ring resonator with radius r and a waveguide, is described in Fig. 2.2, based
on Yariv (2002a, b). Dening that a single unidirectional mode of the res-
onator is excited, the coupling is lossless, single polarization is considered,
none of the waveguide segments and coupler elements couple waves of dier-
ent polarization, the various kinds of losses occurring along the propagation of
4 2 Ring Resonators: Theory and Modeling

Fig. 2.1. Ring resonator channel dropping lter

Fig. 2.2. Model of a single ring resonator with one waveguide

light in the ring resonator lter are incorporated in the attenuation constant,
the interaction can be described by the matrix relation:




Et1 t Ei1
= . (2.1)
Et2 t Ei2

The complex mode amplitudes E are normalized, so that their squared magni-
tude corresponds to the modal power. The coupler parameters t and depend
on the specic coupling mechanism used. The denotes the conjugated com-
plex value of t and , respectively.
The matrix is symmetric because the networks under consideration are
reciprocal. Therefore  2  2
  + t  = 1. (2.2)
In order to further simplify the model, Ei1 is chosen to be equal to 1. Then
the round trip in the ring is given by

Ei2 = ej Et2 , (2.3)

 2.1 Single Ring Resonators 5

where is the loss coecient of the ring (zero loss: = 1) and = L/c,
L being the circumference of the ring which is given by L = 2r, r being the
radius of the ring measured from the center of the ring to the center of the
waveguide, c the phase velocity of the ring mode (c = c0 /ne ) and the xed
angular frequency = kc0 , c0 refers to the vacuum speed of light. The vacuum
wavenumber k is related to the wavelength through: k = 2/. Using the
vacuum wavenumber, the eective refractive index ne can be introduced
easily into the ring coupling relations by
2 ne
= k ne = , (2.4)

where is the propagation constant. This leads to
L kc0 L 2 ne 2r r
= = = k ne 2r = = 4 2 ne . (2.5)
c c
From (2.1) and (2.3) we obtain

+ t ej
Et1 = , (2.6)
t + ej

Ei2 = , (2.7)
t + ej

Et2 = . (2.8)
1 t ej
This leads to the transmission power Pt1 in the output waveguide, which is
2
2 2 + |t| 2 |t| cos ( + t )
Pt1 = |Et1 | = 2 , (2.9)
1 + 2 |t| 2 |t| cos ( + t )

where t = |t| exp (jt ), |t| representing the coupling losses and t the phase
of the coupler.
The circulating power Pi2 in the ring is given by
2
2 2 (1 |t| )
Pi2 = |Ei2 | = 2 . (2.10)
1 + 2 |t| 2 |t| cos ( + t )

On resonance, (+t ) = 2m, where m is an integer, the following is obtained:

2
2 ( |t|)
Pt1 = |Et1 | = 2 (2.11)
(1 |t|)

and
2
2 2 (1 |t| )
Pi2 = |Ei2 | = 2 . (2.12)
(1 |t|)
6 2 Ring Resonators: Theory and Modeling

A special case happens when = |t| in (2.11), when the internal losses are
equal to the coupling losses. The transmitted power becomes 0. This is known
in literature as critical coupling, which is due to destructive interference.
In using the above equations, it is possible to get a good idea of the behav-
ior of a simplied basic ring resonator lter conguration consisting of only
one waveguide and one ring. The wavelength-dependent lter characteristic
for a ring resonator conguration with a radius of r = 148 m with matched
coupling and loss coecient, derived using (2.1)(2.11), is shown in Fig. 2.3.
This model can be extended to suit the requirement of various types of ring
resonator congurations.
The next conguration which is discussed is the basic ring resonator add
drop conguration, consisting of one input, one output waveguide and the ring
resonator. The four ports of the ring resonator are referred to in the following
as input port, throughput port, drop port and add port (Fig. 2.4).
The ring resonator simulation model has been updated according to
Fig. 2.4. For simplication Ei1 is as dened before equal to 1. The throughput
mode amplitude in the rst waveguide is given by

1 1 t2 1/2 ej|t1 | +|1 |2 =1

t1 t2 1/2
2
2 2
t1 t2 ej ej
Et1 = t1 + = . =
1 t1 t2 1/2
2 ej1 t1 t2 1/2
1 t1 t2 ej 2 ej

(2.13)
In this calculation, 1/2 and 1/2 are used which are the half round trip loss
2
and phase, respectively. It is = 1/2 and = 21/2 .
Now, the mode amplitude in the ring has to pass the second coupler as
can be seen from the schematic to become the new dropped mode amplitude
Et2 . The dropped mode amplitude in the second waveguide is then given by:

1 2 1/2 ej1/2
Et2 = . (2.14)
1 t1 t2 ej

Fig. 2.3. Notch type ring resonator lter characteristic

 2.1 Single Ring Resonators 7

t1
Ei1 Et1

1 1
t1
Er

t2
2 2

Et2 Ei2
t2

Fig. 2.4. Model of a basic adddrop single ring resonator lter

At resonance, the output power from the drop port is given by

   
2 2
1 |t1 | 1 |t2 |
2
Pt2Resonance = |Et2Resonance | = (2.15)
(1 |t1 t2 |)2

The throughput port mode amplitude Et1 (2.13) will be zero at resonance for
identical symmetrical couplers t1 = t2 if = 1, which indicates that the wave-
length on resonance is fully extracted by the resonator. The value of = 1
can only be achieved by the implementation of gain incorporated in the ring
resonator to compensate the waveguide losses. The value of the loss coe-
cient is xed in a purely passive ring resonator. A possibility of achieving
minimum intensity (Pt1 = 0) at resonance of the output transmission Pt1 at
the throughput port is to adjust the coupling parameters t1 , t2 to the loss
coecient . From (2.13) we obtain
 
 t1 
=   . (2.16)
t2

If the ring resonator is lossless ( = 1), then the couplers have to be symmetric
in order to achieve minimum intensity. The transmission of a lossless ring
resonator add drop lter with radius of r = 148 m is shown in Fig. 2.5.
There are dierent kinds of requirements on the simulation of various
kinds of ring resonator congurations. Starting of with the given equations
satises most basic models. The ring model can for example be divided
8 2 Ring Resonators: Theory and Modeling

Fig. 2.5. Adddrop ring resonator lter characteristic

into more segments to account for dierent materials or modied waveguide

paths. Examples of calculated models can be found in Rabus (2002) and
Michelotti et al. (2004).

Ring Resonator Parameters

Ring resonator lters can be described by certain gures of merit which are
also generally used to describe optical lters. One important gure is the dis-
tance between resonance peaks, which is called the free spectral range (FSR).
A simple approximation can be obtained for the FSR by using the propaga-
tion constant from (2.4), neglecting the wavelength dependency of the eective
refractive index
ne
= +k . (2.17)

This leads to the FSR , which is the dierence between the vacuum wave-
lengths corresponding to two resonant conditions.

1
2 2
FSR = = . (2.18)
L ne L

Note that (2.18) is for the resonant condition next to a resonance found for
the used propagation constant.
If the wavelength dependence of the eective index can not be neglected,
it can be incorporated in the following way to obtain a modied version of
(2.17).
k
= ng , (2.19)

where ng is the group refractive index, which is dened as:

ne
ng = ne . (2.20)

 2.1 Single Ring Resonators 9

The group refractive index can be used instead of the eective index whenever
appropriate avoiding the approximation and obtaining more accurate values.
The modied FSR is then given by

2
FSR = = . (2.21)
ng L

The next parameter of importance is the resonance width which is dened

as the full width at half maximum (FWHM) or 3 dB bandwidth 2 of the
resonance lineshape. Using the expressions for the drop port (2.14) and (2.15)
 
 ej1/2 2 2 2
1 |1 | |2 |
 1 2 1/2 
  = . (2.22)
 1 t1 t2 ej  2 (1 |t1 t2 |)2

Assuming that the coupling coecients are real, lossless, and without a phase
term, (2.22) can be written as
 2  2
1 2 1/2 1 1 2 1/2
2 = . (2.23)
1 2t1 t2 cos () + (t1 t2 ) 2 (1 t1 t2 )2

Then
2 2
2 (1 t1 t2 ) = 1 2t1 t2 cos () + (t1 t2 ) . (2.24)
For small , using the real part of the series expansion of the Euler formula

2
cos () = 1 . (2.25)
2
Therefore
2
(1 t1 t2 )
2 = . (2.26)
t1 t2
This equation can further be simplied if the loss in the ring is negligible and
the coupling is symmetric (t = t1 = t2 ) to

2
(1 t2 ) 1 t2
= 2
= . (2.27)
t t

Using (2.5) and (2.17) to translate into the wavelength domain

2 1 t2
2 = . (2.28)
Lne t
The expression which is commonly used can be obtained by assuming weak
coupling and 
2 2
FWHM = 2 = . (2.29)
Lne
10 2 Ring Resonators: Theory and Modeling

Another parameter which can now be directly calculated from the parameters
in the previous chapter is the nesse F of the ring resonator lter. It is dened
as the ratio of the FSR and the width of a resonance for a specic wavelength
(FWHM):
FSR t 1
F = = = 2. (2.30)
FWHM 2 1 t2
A parameter which is closely related to the nesse is the quality factor Q of a
resonator, which is a measure of the sharpness of the resonance. It is dened
as the ratio of the operation wavelength and the resonance width
ne L t ne L
Q= = = F. (2.31)
2 1 t2
The quality factor can also be regarded as the stored energy divided by the
power lost per optical cycle.
The intensity in the ring resonator can be much higher than that in the
bus waveguides, as the traveling wave in the ring resonator interferes construc-
tively at resonance with the input wave and thus the amplitude builds up. In
addition to this intensity increase, the eld also experiences a phase-shift of
an integral multiple of 2 in one round trip. The intensity enhancement or
buildup factor B is given by (for a conguration shown in Fig. 2.2)
   2
 Ei2 2  

B=  
=  . (2.32)
Ei1  t + ej 

For a conguration like Fig. 2.4, the buildup factor is given by

   2
 Er  2  1 
B=   =  . (2.33)
Ei1  1 t1 t2 ej 

On resonance, the intensity enhancement factor is

 
 1 2
B=   . (2.34)
1 t1 t2 

For a lossless resonator and setting 1 = 2 = which is  1, B can be

written as
1 (2.30) F
B = 2 . (2.35)

This equation directly relates the intensity enhancement factor B to the
nesse F .
For the all-pass conguration and on resonance, the buildup factor is given
by (t1 = t2 = t)
1+t
B= . (2.36)
1t
 2.1 Single Ring Resonators 11

Ring resonators can be used for nonlinear optical devices since the intensity
in the resonator can be much higher than in the bus waveguide. Examples of
devices utilizing nonlinearities in ring resonators are given in Sects. 5.7 and 5.8.

The Time-Dependent Relations

Before going on to extend the previous model incorporating additional ring

resonators, the time-dependent relations are adapted from Little et al. (1997b)
for the basic ring resonator adddrop conguration. Like in the previous
model, it is assumed that the ring supports a traveling wave of amplitude
E(t), P (t) represents the total power owing through any cross section of
the ring waveguide at time t. The ring is regarded as an oscillator of energy
amplitude e(t), normalized so that p(t) represents the total energy stored in
the ring. Energy and power in the ring are related through
2 2r
p (t) = |e (t)| = P (t) , (2.37)
vg
where vg is the group velocity. The ring resonator has a resonant frequency
of R and amplitude decay time-constant of 1/ . The decay rate and the
power exciting the ring resonator are related to each other. The decay rate
includes the power coupled to the transmitted wave 1/tr , the power lost
due to intrinsic eects 1/ie , and power coupled to the output waveguide
1/t2 . This leads to
1 1 1 1
= + + . (2.38)
tr ie t2
The time rate of change in ring energy can then be written as


d 1
e = jR e Ei1 . (2.39)
dt
The relation between the coupling parameter and the decay rates of the
transmitted wave 1/tr and the output waveguide 1/t2 , is determined by
power conservation. The case is considered, when the ring resonator is excited
to an energy of |e0 |2 , no output waveguide is present and no input wave Ei1 .
The ring energy, with no intrinsic loss, then decays as follow:


2 2 2t
|e (t)| = |e0 | exp . (2.40)
tr
From these set of eqs. (2.372.40), the power transfer characteristic for the
drop port can be calculated with the input wave Ei1 being proportional to
exp(jt)
2
j
tr
e= E . (2.41)
1 i1
j ( R ) +

12 2 Ring Resonators: Theory and Modeling

From this equation, we can calculate the transmitted wave at the throughput
port
1 2
j ( R ) + + j
tr
Et1 = Ei1 e = Ei1 . (2.42)
1
j ( R ) +

Finally the drop port power transfer characteristic is obtained by using the
equation for power conservation (note, Ei2 = 0):
4
2 2 2 2 2 t2 tr 2
|Et2 | = |Ei1 | |Et1 | = |e| = |Ei1 | (2.43)
t2 2 1
( R ) +

This equation can be simplied further if both waveguides couple equally to
the ring, then t2 = tr .

The Z-Transform

Another approach to simulate ring resonator lters is by using the

Z-transformation to describe the spectral and temporal response of ring
resonator lters, described in detail in Madsen and Zhao (1999). Z-transform
relationships for basic optical elements were rst developed for ber optic
lters in Moslehi et al. (1984). This technique is used in pole-zero diagrams
to design ring resonator lters in Kaalund and Peng (2004). In the following
chapter, the Z-transform for the basic adddrop conguration (consisting
of one ring resonator and two waveguides) is given which shall serve as a
starting point for the calculation of more complex devices. Z-transforms
are discussed extensively in the aforementioned literature on digital signal
processing, therefore only a brief introduction is given here.
The Z-transform is an analytic extension of the discrete-time Fourier trans-
form (DTFT) which converts a discrete time signal into a complex-variable
frequency signal

H (z) = h (n) z n , (2.44)

n=

where z is a complex variable and h(n) is the impulse response of a lter or

the values of a discrete signal. Each term z n represents a delay. Of particular
interest is when the absolute value of |z| = 1. This is called the unit circle in
the complex plane where pole and zero locations of the function H(z), which
is evaluated along z = exp(j), are plotted. In the case of ring resonator l-
ters, |z| = 1 corresponds to resonant frequencies. A complete roundtrip of the
unit circle corresponds to the FSR of the lter. Poles and zeros are related
to the lters frequency spectrum by their position on the complex plane.
A zero positioned on the unit circle results in zero transmission at the fre-
quency corresponding to the angle of this zero. A pole on the other hand on
 2.1 Single Ring Resonators 13

the unit circle causes unity transmission at the corresponding frequency. As

poles and zeros move away from the unit circle, their eect on the magnitude
spectrum is reduced.
A linear discrete system with input signals x has the following output
signal:

y (n) = b0 x (n) + b1 x (n 1) + + bM x (n M ) a1 y (n 1) ,
aN y (n N ) . (2.45)

The Z-transform for this type of lter is then given by


M
m
bm z
m=0 B (z)
H (z) = = . (2.46)
N A (z)
1+ an z n
n=1

A(z) and B(z) are M th and N th-order polynomials. The zeros zm and poles
pn of H(z) can be derived from the roots of the polynomials as follows:


M
z N M (z zm )
m=1 B (z)
H (z) = = , (2.47)

N A (z)
(z pn )
n=1

where is the gain of the lter. In passive lters, the transfer function
can never be greater than 1, so has a maximum value determined by
max {|H(z)|z=exp(j) } = 1 for these types of lters.
A ring resonator has a response which can be expressed in the form


1
H (z) = an z n = . (2.48)
n=0
1 az 1

The basic conguration of an adddrop ring resonator lter (Fig. 2.4) is the
simplest lter with a single pole response.
The sum of all optical paths for the drop port is given by

Et2 (z) = 1 2 z 1 1 + t1 t2 z 1 + Ei1 (z) . (2.49)

Using the Taylor series expansion, the equation can be simplied to give the
drop port transfer function:

Et2 (z) 1 2 z 1
H21 (z) = = . (2.50)
Ei1 (z) 1 t1 t2 z 1

There is a single pole at t1 t2 . As the matrix is symmetric, the transfer

function H21 (z) is equal to H12 (z).
14 2 Ring Resonators: Theory and Modeling

The sum of all optical paths for the throughput port resulting in its transfer
function H11 (z) is given by
  
Et1 (z) = t1 21 t2 z 1 1 + t1 t2 z 1 + Ei1 (z) (2.51)
 2 1

1 t2 z
= t1 Ei1 (z)
1 t1 t2 z 1
 
t1 t2 z 1
= Ei1 (z) ,
1 t1 t2 z 1
Et1 (z) t1 t2 z 1
H11 (z) = = .
Ei1 (z) 1 t1 t2 z 1
Similarly, the transfer function H22 (z) can be derived to be
Et2 (z) t2 t1 z 1
H22 (z) = = . (2.52)
Ei2 (z) 1 t1 t2 z 1
The obtained results for each transfer function can be expressed in two dier-
ent matrix forms. The rst form relates the input ports to the output ports
and is called the scattering matrix (see also (2.1)) which is given by



Et1 (z) Ei1 (z)
= SRR (z) , (2.53)
Et2 (z) Ei2 (z)

t1 t2 z 1 1 2 z 1
1 t1 t2 z 1 ,
1 t1 t2 z 1
SRR (z) = z 1
.
1 2 t2 t1 z 1
1 t1 t2 z 1 1 t1 t2 z 1
The second form relates the quantities in one plane to the ones in another
plane as can be seen in Fig. 2.6. This type of matrix is called the transfer
matrix which is given by



Ei1 (z) Ei2 (z)
= RR (z) , (2.54)
Et1 (z) Et2 (z)


1 1 t1 t2 z 1 t2 + t1 z 1
RR (z) = 1 .
1 2 z 1 t1 t2 z 1 + 1 2 z 1
The scattering matrix form is used to express the implication of power con-
servation and reciprocity. The transfer matrix form is used for describing
multistage lters. It is therefore also referred to as the chain matrix. This
type of transfer matrix is suitable to describe multiple serially coupled ring
resonators. The transfer matrix to be used for describing multiple parallel
coupled resonators has the form (Grover et al. 2002)



Ei1 (z)  Et1 (z)
= RR (z) , (2.55)
Et2 (z) Ei2 (z)


1
1 1 t1 t
2 z 1 2 z 1
RR (z) = .
t1 t2 z 1 1 2 z 1 t1 t2 z 1
 2.1 Single Ring Resonators 15

Fig. 2.6. Z-transform layout of an adddrop ring resonator lter

Using the dierent simulation techniques described in this chapter it is

possible to build a simulation model for all relevant lateral and vertically
coupled ring resonator lter congurations.

Incorporating Loss in Ring Resonator Filters

In the previous simulation models, loss is basically included by insertion of the

parameter in the formulas. Loss in ring resonator lters scales the spectral
curves without changing them signicantly if the loss is not too large. Dierent
types of losses have been considered and described in literature to account for
various fabrication methods and tolerances. A few examples will be given
in the following paragraph to give a starting point for the calculation and
simulation of losses in ring resonators.
A loss which is most obvious is the radiation loss which occurs in the curved
section of ring resonators (Chin and Ho 1998). The radiation loss will become
large for a large surface roughness. The surface roughness on the other hand
can induce contradirectional coupling which can degrade the performance of a
ring resonator lter and can even cause a splitting of the resonant peak (Little
et al. 1997a). The calculation of the radiation and scattering losses in ring
resonators is presented in detail in Rabiei (2005). Here a perturbative method
for the calculation of losses due to an arbitrary variation in the refractive index
of the core or due to an arbitrary variation in the shape of ring resonator is
described.
Loss in ring resonator devices can also be used advantageously as was
demonstrated using (2.16). An analysis of how loss and gain can be used to
tune, trim, or compensate the wavelength response of ring resonator lters is
described in Little and Chu (2000).
In the following chapter the basic model of a ring resonator is extended to
analyze a racetrack shaped ring resonator.
16 2 Ring Resonators: Theory and Modeling

2.1.2 Racetrack-Shaped Resonators

The equations in the previous chapter described the general behavior of a

basic ring resonator lter. However to simulate certain ring resonator cong-
urations, this set of equations has to be extended. In a waveguide-coupled
ring resonator lter, the coupling gap size is determined by the amount of
coupling required and the coupling length available. Using lateral coupling, a
larger gap is desirable as it increases the fabrication tolerance. In this case,
for a given coupling coecient, the gap size can be enlarged if the coupling
distance is increased to obtain a racetrack-shaped ring resonator (Fig. 2.7).
The coupling distance can be increased by using a lateral or vertical geom-
etry (Chin and Ho 1998). In the case of the racetrack shaped ring resonator,
the coupling distance is approximately the length of the straight sections (if no
coupling occurs before the actual coupler region starts), which are tangential
to the circular arcs at output and input ports. At the transitions between the
curved and the straight sections, the mode will change adiabatically between
the radial mode in a curved waveguide and the normal mode in a straight
waveguide. As the straight sections are lengthened the radius of the curved
sections must be reduced if the total cavity length is to remain constant. This
cannot always be achieved in practical applications.
In the case of racetrack resonators, it is important to consider the variation
of phase dierence which occurs in the coupling region in both couplers
between t and . The phase dierence is length-dependent and aects the
output characteristics, not only in the magnitude but also in the resonant
conditions. This phase dierence can be implemented in (2.13) and (2.14)
using the method described in Lee et al. (2004).

Fig. 2.7. A racetrack-shaped ring resonator lter with integrated platinum resistors
on top of the curved sections and the coupler (contact pads visible)
 2.2 Double Ring Resonators 17

For the throughput port:

 
t1 ejt1 t2 1/2
2
ej t1 t1 ej2t1 1 1 ej22 ejt2
Et1 = . (2.56)
1 t1 t2 1/2
2 ej ejt1 ejt2

For the drop port

1 2 1/2 ej1/2 ej1 ej2
Et2 = . (2.57)
1 t1 t2 ej ejt1 ejt2
The resonant condition ( + t1 + t2 ) = 2m has changed slightly com-
pared to (2.9). The output transmission at resonance is no longer indepen-
dent of t and which can be seen comparing the relevant terms in (2.50)
and (2.51). This leads to a signicant performance change and conrms the
importance of taking into account the phase dierence when analyzing race-
track shaped ring resonator congurations. The phase dierence is not only of
importance in directional couplers, but also in MMI couplers. An analysis of
a racetrack shaped ring resonator adddrop lter having two MMI couplers is
given in Caruso and Montrosset (2003). The MMI length should be selected in
order to satisfy the /2 relation between the phases of the transmission and
the coupling terms to obtain a symmetric transmission characteristic of the
throughput and of the drop port. If the /2 relation is not satised, a strong
asymmetry in the transmission characteristic will be obtained.
A further description of the implementation of directional and MMI cou-
plers in ring resonators is given in Sect. 4.1.

2.2 Double Ring Resonators

Double ring resonators oer the possibility of realizing a box-like lter char-
acteristic which is favorably used in optical networks. This is not the only
advantage, but also from the point of characterization, two rings if coupled
in series have the drop port in the same direction as the input port, which
is also convenient for interconnection of many 2 2 devices. Serially and
parallel coupled ring resonator congurations have been described in detail
in Chu et al. (1999a); Little et al. (1997b, 2000a); Melloni (2001) and Emelett
and Soref (2005).
In the serially coupled conguration, each ring resonator is coupled to one
another, and a signal that is to be dropped from the input port to the drop
port must pass sequentially through each resonator. Because of this sequen-
tial power transfer, all resonators must be precisely resonant at a common
wavelength. The resulting resonant line shape in the series conguration is
determined physically by the separations between the ring resonators. In the
parallel-coupled conguration, all resonators are coupled to both the input
and drop port waveguides, but usually not directly to one another (the res-
onators can also be coupled to one another resulting in a wavelength selective
18 2 Ring Resonators: Theory and Modeling

reector as is described in Sects. 2.2.2 and 2.3.1). The resonators are instead
indirectly coupled to each other by the optical path lengths along the input
and output waveguides that interconnect them. These lengths determine the
details of the resonant line shapes. An optical signal in the parallel cong-
uration passes through all ring resonators simultaneously. This softens the
requirement that the resonances of each ring have to be precisely identical.
Nonaligned resonant frequencies instead lead to multiple peaks, or ripple in
the lineshape.
The ready to use transfer functions of serially and parallel coupled double
ring resonators will be described in the following sections.

2.2.1 Serially Coupled Double Ring Resonator

The schematic of a serially coupled double ring resonator is depicted in

Fig. 2.8.
From this model and using the same procedure as in Sect. 2.1.1 the elds
depicted in Fig. 2.8 can be calculated as follows:
1
E1a = 1 Ei1 + t1 1 ej 2 E1b , (2.58)

E1b = t2 1 e j 21
E1a 2 2 e j 21
E2b , (2.59)

j 22 j 22
E2a = 2 1 e E1a + t2 2 e E2b , (2.60)
2
E2b = 3 Ei2 + t3 2 ej 2 E2a (2.61)

Fig. 2.8. Two ring resonators coupled in series

 2.2 Double Ring Resonators 19
1
Et1 = t1 Ei1 + 1 1 ej 2 E1b , (2.62)

j 22
Et2 = t3 Ei2 + 3 2 e E2a , (2.63)

where 1,2 = R11/2 ,R21/2 represent the half round trip loss coecients of
ring resonator one and two respectively. From (2.58) to (2.63) the general
expressions for the transfer functions for the throughput and the drop port
can be derived. A simplied form can be obtained by assuming a coupler
without losses and symmetric coupling behavior, setting t = t and =
(note that the phase factor j has not been introduced into the assumption
and can be added if required) which gives the ready to use amplitude forms
for the throughput port (Ei2 = 0)
 
Et1 t1 21 1 ej1 t3 2 ej2 t2
= (2.64)
Ei1 1 t3 t2 2 ej2 t2 t1 1 ej1 + t3 t1 1 2 ej1 ej2
and for the drop port:
1 2
Et2 3 2 1 1 2 ej 2 ej 2
= . (2.65)
Ei1 1 t3 t2 2 ej2 t2 t1 1 ej1 + t3 t1 1 2 ej1 ej2
For realizing a double ring resonator with maximally at response, rst the
input/output waveguide ring coupling coecient 1 (3 ) has to be determined.
To simplify the model further, it is dened that the input/output waveguide-
ring coupling coecients 1 = 3 . The calculation of the coupling coecients
to obtain the appropriate coupling values between the two ring resonators in
order to achieve maximally at response can be made according to Emelett
and Soref (2005), Little et al. (1997b) by using the geometry and index prole
shown in Fig. 2.9, where two coupled waveguides of width 2wp and 2wq with
indexes of np and nq , surrounded by a cladding of n0 , at the plane of smallest
separation 2s0 , dened as the center to center gap are shown.
The approximate coupling coecient is then given by
  
0 cos kxp,q wq  2  R [q (wq 2s0 )]
=    np n02
e
2 Pp Pq kx2p + q2 q
    
q cos kxp wp sinh (q wp ) + kxp sin kxp wp cosh (q wp ) . (2.66)

Fig. 2.9. Index prole and geometry of coupled waveguides

20 2 Ring Resonators: Theory and Modeling

Using


p,q 1
Pp,q = wp,q + , (2.67)
20 p,q

kxp,q = n2p,q k 2 p,q
2 , (2.68)

p,q = p,q2 n2 k 2 , (2.69)
0

where Pp,q is the mode power, kxp,q is the transverse propagation constant of
the core, and p,q is the decay constant in the cladding, p,q is the propagation
constant, is the circular frequency, 0 is the permittivity of free space, all
within waveguide p or q. The refractive index n0 of the surrounding media
is set equal to 1 (air). The coecient R is dened as the eective radius of
curvature of the ring and is given by:
r1 r2
R= , (2.70)
r1 + r2
where r1,2 represents the radius of ring one and two respectively. The radii of
the rings are chosen to satisfy the 2 phase shift condition with the completion
of one round trip in the ring resonator which is given by:

mm = 2r1,2 ne . (2.71)

In order to realize a at passband, the analysis of the power loss ratio,

which is the ratio of total input power to power present at the detected port,
is required. The power loss ratio in polynomial form is given by
  
4 
2 
 Ei  2 1 1 41
PLR =   = 1 + 4 4 + 4
22 + 2
2 2 2
,
Et2  1 2 2 4
(2.72)
where is the frequency deviation which is related to the resonant frequency
m by
= m . (2.73)
The coecient is the fractional power coupled. It is given by

21 vg1 2 vg1 vg2

21 = and 22 = 2 2 . (2.74)
2r1 4 r1 r2
The goal of realizing a maximally at response is obtained when the power
loss ratio is zero at the dened resonance. This is the case for

22 = 0.25041 . (2.75)

Assuming identical rings, this leads to the coupling coecient

22 = 0.25041 . (2.76)
 2.2 Double Ring Resonators 21

Using these equations, it is possible to design a double ring resonator lter

with maximally at response for the drop port.
A serially coupled double ring resonator opens up the possibility of
expanding the FSR to the least common multiple of the FSR of individual
ring resonators. This is done by choosing dierent radii in the double ring
resonator. In the case of dierent radii, the light passing through the double
ring resonator is launched from the drop port when the resonant conditions of
both single ring resonators are satised. The FSR of the double ring resonator
with two dierent radii is expressed by

FSR = N FSR1 = M FSR2 , (2.77)

which leads to
FSR1 FSR2
FSR = |M N| , (2.78)
|FSR1 FSR2 |
where N and M are natural and coprime numbers.
The use of two ring resonators with dierent radii opens the possibility
to realize a larger FSR than would be achieved using only a single ring res-
onator. The transmission characteristic of the throughput port has mainly a
Lorentzian shape. A box-like lter response using two dierent radii can only
be realized by using two parallel coupled double ring resonators (R1 = R2 ).
The use of such congurations as optical lters is limited by unwanted addi-
tional resonant peaks. Investigations on these types of lters have been per-
formed in Suzuki et al. (1995), Sorel et al. (1999). Dierent types of waveguide
coupled ring resonator congurations to expand the free spectral range have
been analyzed in Hidayat et al. (2003).

2.2.2 Parallel Coupled Double Ring Resonator

The schematic of a parallel coupled double ring resonator is shown in Fig. 2.10.
From this model, the elds in Fig. 2.10 can be calculated as follows:
1
E1a = 1 Ei1 + t1 1 ej 2 E1b , (2.79)
1
 2

E1b = t2 1 ej 2 E1a 2 ejW 4 2 ej 2 E2a t4 Ei2 , (2.80)
 1
 2
E2a = 3 ejW t1 Ei1 + 1 1 ej 2 E1b + t3 2 ej 2 E2b , (2.81)
2
E2b = 4 Ei2 + t4 2 ej 2 E2a , (2.82)
 1
 2
Et1 = t3 ejW t1 Ei1 + 1 1 ej 2 E1b + 3 2 ej 2 E2b , (2.83)
 2
 1
Et2 = t2 ejW t4 Ei2 + 4 2 ej 2 E2a + 2 1 ej 2 E1a (2.84)

W = kW nWe , (2.85)
22 2 Ring Resonators: Theory and Modeling

Fig. 2.10. Parallel coupled double ring resonator

where W is the phase shift introduced by the segment of length with eec-
tive refractive index nWe of the input waveguide joining both ring resonators.
From (2.79) to (2.85) the general expressions for the transfer functions
for the throughput and the drop port can be derived. A simplied form can
again be obtained by assuming couplers and bus waveguides without losses
and symmetric coupling behavior, setting t = t and = (note that
the phase factor j has not been introduced into the assumption and can be
added if required) which gives the amplitude forms for the throughput port
(Ei2 = 0)
 
Et1 jW

jW 21 f (a + bc) a + bc
= t3 t1 e + t3 1 1 e h+ + 3 t4 22 ej2 . (2.86)
Ei1 1d 1d

And for the drop port

 
Et2
j 21 2 j1 f (a + bc) 2 a + bc
= 1 2 1 e + 2 t1 1 e h+ + t2 4 2 ejW 2 ,
Ei1 1d 1d
(2.87)
where
3 t1 ejW
a= , (2.88)
1 t3 t4 22 ej2
1
1 1 ej 2
b= , (2.89)
1 t3 t4 22 ej2
1
1 t2 1 ej 2
c= (2.90)
1 t1 t2 12 ej1
 2.2 Double Ring Resonators 23
2
2 4 2 ejW 2
f= , (2.91)
1 t1 t2 12 ej1
1
1 t2 1 ej 2
h= . (2.92)
1 t1 t2 12 ej1
d=bf (2.93)

Similar to the serially coupled double ring resonator, it is possible to increase

the overall FSR of this conguration by adjusting the length of the waveguide
joining the two ring resonators.
The parallel conguration can be treated as a grating. Constructive inter-
ference of the reected waves from each ring resonator is obtained by choosing
to be equal to an odd multiple of a quarter wavelength.
0
= (2m + 1) , (2.94)
4nWe
where 0 is the center wavelength of the passband.
If the length is chosen such that the FSR of the ring resonators and the
FSR of the grating obey the following condition:

FSR = NRing FSRRing = MGrating FSRGrating . (2.95)

The Vernier eect (Griel 2000b) causes the transmission peaks of the ring
resonators within the overall obtained FSR to be suppressed, which results in
a larger FSR than would be achieved for a single ring resonator. The distance
between the resonators can be calculated in this case using:
Mgrating ne
= r. (2.96)
NRing nWe
These set of equations can of course be transferred to multiple coupled parallel
ring resonator congurations as will be shown in Sect. 2.3.

Parallel Coupled Double Ring Resonator with Coupling Between

the Two Ring Resonators

A special double ring resonator conguration is obtained, when coupling

between the two ring resonators is allowed (Fig. 2.11). The reective prop-
erties of such a device have been analyzed in Chremmos and Uzunoglu (2005)
and the following transfer functions are based on this reference. A similar
wavelength reective lter consisting of four interacting ring resonators has
been presented in Poon, Scheuer, Yariv (2004e) which is going to be described
in Sect. 2.3.1.
Using the elds and coupling coecient in Fig. 2.11, it is possible to derive
the general expressions for describing the transfer characteristics of this type
of conguration.
24 2 Ring Resonators: Theory and Modeling

Fig. 2.11. Double ring resonator with inter ring coupling

A simplied expression for the reectivity at port Ei1 (denoted by the

symbol as the superscript) is obtained for the lossless case and symmetric
coupling coecients between the ring resonator and the bus waveguide ( =
1 = 3 ):

2    2
2 2 t2 1 + t2
 2 4 cos
  2t 2t
E i1  =  2
. (2.97)
   2
t2 1 + t2 2 2
cos 2
+
2t 2t

The phase factor for the distance between the rings does not appear and
therefore does not have an inuence on the transfer characteristic. This is
a special property of the double coupled conguration and does not apply
for multiple serially coupled ring resonators. Dierent reectivity proles can
be realized using appropriate coupling coecients. Weakly coupled ring res-
onators lead to single reection peaks in the reectivity function where the
height of the peak depend on the value of the coupling coecients. In order to
realize a maximally at response with a single peak, the coupling coecients
have to obey the following equation:

2
2 =  . (2.98)
2 1 + t2 + 2t

The corresponding FWHM is given by


2
FWHM = 4 sin1 3 . (2.99)
22 t
The expression for the reectivity at port Ei1 incorporating loss parameter
is given by
 2.3 Multiple Coupled Resonators 25
   2
2 2 2 + 1
 2 2t
 
E i1  =    2   2 2 2 (2.100)
t2 1 + 2 t2 2 1 2 t 2
cos +
2t 2t
      2
t2 1 + 2 t2 t2 1 1 2 t2
2
cos +
2t 2t 1 + 2

 .
1 2
2

+ sin
1 + 2

So far ready to use transfer functions for single and double ring resonator
conguration have been presented. In the following chapter, dierent calcula-
tion methods are presented to derive the transfer function of dierent types
of multiple coupled ring resonators.

2.3 Multiple Coupled Resonators

The use of multiple vertical or lateral coupled ring resonator congurations
opens up the possibility to realize custom designed transmission functions
and thus is suitable for a wide variety of device implementations (Schwelb
and Frigyes 2001). Vertical and lateral ring resonator architectures where the
rings are either coupled in series or in parallel have been implemented in
the past. A detailed theoretical analysis of a vertically stacked multiple cou-
pled ring resonator, where the ring resonators are on top of each other is
presented in Sumetsky (2005). The use of multiple coupled ring resonator
congurations together with other photonic devices like grating couplers or
MachZehnder interferometers increases the functionality and transmission
characteristic even further (Weiershausen and Zengerle 1996). One of the main
targets of realizing optical lters is to tailor the passband shape. In analogy to
electronic lter design, Butterworth (1930) and Chebyshev type of optical l-
ters are preferred. Butterworth lters are maximally at and do not have any
ripples in the passband. The shape of the transfer function does not change for
higher order lters except that the roll o becomes steeper as the order of the
lter increases. Chebyshev lters have a steeper roll o than Butterworth type
of lters and have ripples either in the passband or stopband which distin-
guishes Chebyshev type I and type II lters. The square magnitude response
of a Butterworth lter has the form (Madsen and Zhao 1999)

2 1
|HN (x)| =
2N , (2.101)
x
1+
x0

where x0 is the 3 dB cuto frequency.

26 2 Ring Resonators: Theory and Modeling

The response of Chebyshev type I and II lters have the form (Madsen and
Zhao 1999)
Type I
1
|HN (x)| =
2

(2.102)
x
1 + y 2 TN2
xC
$  
cos N cos1 x for |x| 1
TN (x) =   ,
cosh N cosh1 x for |x| > 1

Type II
1
|HN (x)| =
2

, (2.103)
2 xS
T
N xP
1 + y2
2  xS 
TN
x

where TN (x) is the N th-order Chebyshev polynomial, y determines the pass-

band ripples, xC is the center frequency of the lter, xP and xS dene the
changeover region on the passband and stopband, respectively.

2.3.1 Serially Coupled Ring Resonators

One of the rst papers to present a calculation method for serially coupled
ring resonator synthesis is Orta et al. (1995). The method is based on the Z-
transformation using the transfer matrix (see (2.54)). The Z-transformation
has also been used in Madsen and Zhao (1996) to simulate and fabricate a
serially coupled ring resonator lter. The transfer matrix method has also
been used in Melloni and Martinelli (2002) and Poon et al. (2004c) for
simulating and designing various serially coupled ring resonator congura-
tions. Serially coupled ring resonators are also referred to as coupled-resonator
optical waveguides (CROW) Poon et al. (2004d). A model for deriving the
transfer functions of serially coupled ring resonators based on the time-
dependent calculation as described in Sect. 2.1.1 is presented in Little et al.
(1998b). Another method of simulating the transfer function of serially coup-
led ring resonators is by using a characteristic matrix approach presented in
Chen et al. (2004a). These dierent methods for analyzing serially coupled
ring resonators (Fig. 2.12) will be used in the following to derive the transfer
functions.
The Z-transformation is used to start with in the beginning. In order to
describe a serially coupled ring resonator, the lter can be broken down into
two components, a symmetrical directional coupler and a pair of uncoupled
guides as shown in Fig. 2.13 (Orta et al. 1995). Another way to describe the
transfer function of a directional coupler is to use a chain matrix, which is
given by
 2.3 Multiple Coupled Resonators 27

Fig. 2.12. Photograph of a serially coupled triple ring resonator

Fig. 2.13. Components of a serially coupled ring resonator lter

 
csc (k ) ejlk cot (k )
HkC jlk
= je , (2.104)
ejlk cot (k ) e2jlk csc (k )

where k is referred to as the coupling angle and lk is the length of the coupler
(note: this is not the coupling length). The chain matrix of the uncoupled
guides is given by  
R jdk 1 0
Hk = e , (2.105)
0 e2jdk
where dk is the length of the ring waveguides. The chain matrix for the entire
system is then expressed by
%  csc (k ) z 1 ejk cot (k )   csc (0 ) cot (0 ) 
H= , (2.106)
cot (k ) z 1 ejk csc (k ) cot (0 ) csc (0 )
k=N,1
28 2 Ring Resonators: Theory and Modeling

where
zejk = ej(lk +lk1 +2dk ) . (2.107)
As previously described in Sect. 2.1.1 the elements of the matrix H are N
degree polynomials. The chain matrix can also be compared to the chain
matrix derived in (2.54). Of interest is again the transfer function for the
throughput and the drop port, both related to the signal at the input port.
The polynomials are given by

Et1
N
= H11 (z) = ak z k (2.108)
Ei1
k=0

and
Et2
N
= H21 (z) = bk z k . (2.109)
Ei1
k=0

The transfer functions of the lter are now based on the denition of these
polynomials. Assuming, that the lter is lossless, the scattering matrix is
unitary for |z| = 1. On |z| = 1, z = z 1 , then the relationship between
(2.108) and (2.109) can be written as follows:



1 1
H11 (z) H11
= 1 + H21 (z) H21 . (2.110)
z z

Using this equation, it is possible to calculate H11 (z) for a given H21 (z).
The coupling angles k (k = 0 . . . N ) and the phase shift k (k = 1 . . . N ) are
calculated as follows. First, a superscript is introduced which denotes elements
belonging to the structure formed by the rst n+1 coupler. Then the elements
[N 1]
of the chain matrix relative to the rst (N 1) ring resonators, H11 and
[N 1] [N ] [N ]
H21 are related to H11 and H21 through
   [N 1]   [N ] 
csc (N ) z 1 ejN cot (N ) H11 H11
cot (N ) z 1 ejN csc (N ) H21 [N 1] = [N ] . (2.111)
H21

This leads to the equations for the coecients of the polynomials (2.108) and
(2.109), which are given by
[N 1] [N ] [N ]
ak = ak csc (N ) bk cot (N ) (2.112)

and & '

[N 1] [N ] [N ]
bk = ak+1 cot (N ) + bk+1 csc (N ) ejN , (2.113)

where k = 1. . .(N 1) and

[N 1] [N 1]
aN = b1 = 0. (2.114)
 2.3 Multiple Coupled Resonators 29

Then the coupling angle of the N th coupler is given by

[N ] [N ]
b0 aN
cos (N ) = [N ]
= [N ]
. (2.115)
a0 bN

Equation (2.115) is satised if the ratio is real. Next step is to determine the
phase shift k using (2.115) and (2.113). All other coecients follow the same
procedure.
In order to start this calculation method and derive a transfer lter char-
[N ]
acteristic, the polynomial H21 has to be determined. As stated before in
Sect. 2.3, a bandpass lter characteristic is preferred. When using the Z-
transform, the resonant frequencies of the ring resonator lter are all placed
on the circumference |z| = 1. For a Butterworth type lter, the zeros are
located on z = 1. The zeros for a Chebyshev type of lter are given by
    
z0k = e
2j arccos sin( FWHM
4 ) cos (2k1)
2N
,
k = 1 . . . N. (2.116)

This is an intuitive approach to design ring resonator lters, however it is

challenging to extract the poles of the frequency response.
The time-dependent relations which have been presented for a single ring
resonator can be extended to multiple serially coupled resonators. The
response of the rst ring (counted from the bottom of the ring lter) can be
written as (Little et al. 1998b)

Ei1
e1 = , (2.117)
1
2

jR1 +
N 1
2

jR2 . . . +
jRN



1



Rn j , n = 1, N


n

1 1
n = Rn j j , n = 1, N ,

n tr





1 2

Rn j j , N = 1
n tr
where is the optical frequency of the input wave, Rn is the resonant fre-
quency of ring resonator n and n is the complex frequency deviation. The
energy decay rate of each ring is given by n . The response of resonator N
can be expressed by a product of continued fractions:

%
N
eN = Ei1 Tn , (2.118)
n=1
30 2 Ring Resonators: Theory and Modeling

where
n1
TN = ,
n1
2

jn +
n2
2

jn1 +
1
2

jn2 . . . +
jR1

T1 = .
jR1
The coecients for maximally at and Chebyshev lter characteristics are
given in the following table for two to six serially coupled ring resonators
(Little et al. 1997b):

Table 2.1. Coecients for maximally at and Chebyshev lter characteristics

N maximally at Chebyshev
2
2 1 = 0.250 4
2
1 = 0.250
2
(1 + 2y)
 
2
3 2 2
1 = 2 = 0.125
4
2 2
1 = 2 = 0.125
4
1 + 1.5y 3
4 2
1 = 2
3 = 0.100
4

2
2 = 0.040 4

5 2
1 = 2
4 = 0.0955
4

2
2 = 2
3 = 0.040 4

6 2
1 = 2
5 = 0.0915
4

2
2 = 4 = 0.02454
2

2
3 = 0.01794
y determines the passband ripples of the Chebyshev lter.

Another method of calculating the transfer functions of serially coupled

ring resonator lters is by using a so called characteristic matrix approach,
which has been presented in Chen et al. (2004a). The method is based on
dening the optical paths which an optical signal travels form the input port
to the drop port. There are several paths and the shortest path is called the
common path or the zeroth-order path which is dened to be unity. Looking at
a serially coupled triple ring resonator, the light using the rst order paths for
example has three choices, one additional ring resonator path and the common
path. The mth order transfer function is related to the m + 1th order transfer
function by


H1,m+1 T11 T12 T13 H1,m
H m+1 = H2,m+1 = T21 T22 T23 H2,m = T H m , (2.119)
H3,m+1 T31 T32 T33 H3,m
tij are the transfer functions of the additional path in ring resonator i, with
previous path ends in ring resonator j. T is called the characteristic matrix.
 2.3 Multiple Coupled Resonators 31

Using the coecients for the ring resonators as described previously for sym-
metric coupling (t = t and = ), T is given by


j1 22
t1 t2 1 e 1 2 0

t 2 2 
TDropPort = 3
t t e j2
1 1 . (2.120)
2 3 2
t 2
 3
t3 t4 3 ej3 1 1 1

The general expression for the characteristic matrix for a lter consisting of
n ring resonators is given by


j1 22
t t
1 2 1 e 1 0 . . . 0

t22 

t2 t3 2 ej2 1 1 3 0 . . . 0
2
2
TDropPort =
 t3  . (2.121)
.
.. 1 ... 1 ... 0 

. . . 1 . . . . . . 1 n
2

t2n
jn
 
tn tn+1 n e 1 ... ... 1 1

The general expression for the transfer function for the drop port is then
given by

/ / j k
k k e k
k k
HDropPort = ,
|UM atrix TDropPort |
k = 1 . . . n, (2.122)

where UMatrix is the unit matrix.

Filter synthesis using (2.121) and (2.122) can be done by solving a matrix
eigenvalues equation. The poles of a lter function correspond to unity minus
the eigenvalues of TDropPort . The eigenvalues and power coupling ratios for
a maximally at transfer function are given in Table 2.2 assuming lossless
devices.
Using the aforementioned techniques, it is possible to nd the transfer
functions for arbitrary serially coupled ring resonators. Each method has its
pros and cons and therefore it is left to the designer to choose the right
method which seems convenient to solve the problem. When designing ring
resonator lters, it is appropriate to evaluate the technological feasibility of
certain parameters like coupling coecients and loss in order to easily transfer
the design to fabrication.
A special serially coupled ring resonator design (Fig. 2.14) has been pro-
posed and analyzed in Poon, Scheuer, Yariv (2004e). The following calculation
method is based on this literature.
32 2 Ring Resonators: Theory and Modeling

Table 2.2. Eigenvalues and power coupling ratios for a maximally at transfer
function
N eigenvalues (= 1 poles) power coupling ratios
2 0.63 + 0.32i; 0.63 0.32i 0.5; 0.2; 0.5
3 0.73 + 0.42; 0.71; 0.73 0.42i 0.5; 0.14; 0.14; 0.5
4 0.78 + 0.45i; 0.77 + 0.17i; 0.77 0.17i; 0.5; 0.13; 0.09; 0.13; 0.5
0.78 0.45i
5 0.81 + 0.46i; 0.8 + 0.26i; 0.81; 0.8 0.5; 0.13; 0.08; 0.08; 0.13; 0.5
0.26i; 0.81 0.46i
10 0.86 + 0.48i; 0.85 + 0.42i; 0.86 + 0.33i; 0.5; 0.12; 0.07; 0.07; 0.07; 0.07; 0.07;
0.88 + 0.21i; 0.9 + 0.07i; 0.9 0.07i; 0.07; 0.07; 0.12; 0.5
0.88 0.21i; 0.86 0.33i; 0.85 0.42i;
0.86 0.48i
(Chen et al. 2004a)

Eiin Etin

Eiin Etin
Eiin Etin

Eiin Etin

EtN Et0
EtN1 EtN Et0 E
t1

EtN1 EiN Ei0

Et1
EiN Ei0

EiN1 Ei1 Ei1

Eiin

Ei1
Ei1 Et1

Et1

Fig. 2.14. Ring resonator conguration with interring coupling

Depending on the number of rings used the lter can either be reecting or
nonreecting. The lter acts as a reector for an odd number of rings (N 3),
whereas a nonreecting lter is obtained for an even number of rings (N 4).
The transfer functions are derived by using a transfer matrix method, starting
in dening a vector xn , which represents the eld component in ring resonator
n 1: & '
T
xn = E t E i E i E t (2.123)
n
 2.3 Multiple Coupled Resonators 33

The arrows used as superscripts are as described earlier for the double ring the
direction of the propagating eld, clockwise or anticlockwise without mixing
of waves between the two directions. The coupling between the ring resonators
can be represented by the following 4 4 matrix (n 0):

t 1
0 0

1 t
0 0

xn+1 = x = MP xn .
t 1
(2.124)
n
0 0

1 t
0 0

Assuming only the phase matched waves are being coupled (LCoupler  ).
The vector xn is related to xn by the following propagation matrix:

0 0 0 ejr
0 0 ejr(2) 0
xn =
0 ejr(2)
x = MQ xn , (2.125)
0 0 n
ejr 0 0 0
(N 2)
= 2 .
N
Combining (2.124) and (2.125)

xn+1 = MP xn = MP MQ xn = MT xn . (2.126)

For an N type ring resonator conguration with N > 2, (2.126) yields:

xN = MT N 1 MP x0 = MA x0 . (2.127)

Only the components of vector xin hold the transfer functions, therefore the
transfer functions of the ring resonator conguration are derived using the
relation for the coupling to the external waveguide and the phase relations in
the rst resonator:
xin = MPin xin , (2.128)

E t0 = E tin ejr 2 E  t0 = E tin ejr 2 ,

E  i0 = E iN ejr(2) E  i0 = E iN ejr(2) ,

E tN = E iin ejr 2

E tN = E iin ejr 2 .
Using (2.127) and (2.128), xN and x0 can be expressed by elements of xin .
Equation (2.127) can then be rewritten as

mMPin xin = MTN 1 MP wMPin xin (2.129)

xin = MB xin ,
34 2 Ring Resonators: Theory and Modeling


where m and w represent matrices which express the terms E in , E in and Ei0 ,


Ei0 , using components of vector xin respectively. The matrix MB can then be
expressed by
1 1
MB = MPin m MTN 1 MP wMPin . (2.130)
If only one input is considered as in the previous ring resonator examples, xin
can be expressed by & '
T
xin = 0 Ei 1 Et . (2.131)
in in

The reection and transmission function Eiin and Etin can then be derived
using elements of matrix MB and solving the following matrix equation:

MB4,2 MB4,3
1 MB 1 E  iin
4,4 = 1 MB4,4 . (2.132)
MB2,4  MB2,3
1 E tin
1 MB2,2 1 MB2,2
The following equations are also valid for the solutions:

MB3,2 E  iin + MB3,3 + MB3,4 E  tin = 1, (2.133)

 
MB1,2 E iin + MB1,3 + MB1,4 E tin = 0.

In using the above equations, it is possible to calculate the reectance and the
transmittance spectra of this type of serially interring coupled ring resonator.

2.3.2 Parallel Coupled Ring Resonators

Parallel coupled ring resonators have been addressed in literature and

the transfer functions have been derived by several methods. One of the
advantages of parallel coupled ring resonators over serially coupled ring
resonators is that their transfer functions are less sensitive to fabrication
tolerances, as each ring resonator can compensate errors in any of the others.
Two congurations are discussed, one where the ring resonators are coupled to
two input/output waveguides and the other, where all share one input/output
waveguide.

Coupled to Two Input/Output Waveguides

As can be seen from Fig. 2.15, parallel coupled ring resonators coupled to two
waveguides share the same input, throughput, drop, and add port.
The synthesis of the transfer functions of parallel coupled ring resonators
using a recursive algorithm is presented in Little et al. (2000a). A com-
plex matrix formalism employing racetrack ring resonator lters is derived in
Griel (2000a). A technique using simple closed-form formulas to determine
 2.3 Multiple Coupled Resonators 35

Fig. 2.15. Photograph of a parallel coupled triple ring resonator

the Q factor of each involved ring resonator which leads to the coupling
coecients is demonstrated in Melloni (2001). These dierent methods for
analyzing the transfer functions will be used in the following.
The model which has been presented in Fig. 2.10 can be easily extended for
multiple parallel coupled ring resonators. The distance between the uncoupled
ring resonators is chosen so that the transfer functions of each ring resonator
add in phase. The transfer function derived by the recursive algorithm used
in Little et al. (2000a) is given by

Et Eti n Eto n
Tn = = Rn 1 j2W n1 , (2.134)
Ei Rn Tn1 e
in on
Rn =
1 2 1  2 ,
j + i,o
n + i,o
n
2 2
 i,o 2
n
Eti,o n = Rn
1  i,o 2 1  i,o 2
,
j + n + n
2 2
where the indices i, o correspond to the through responses of each ring res-
onator n in the waveguides joining the ring resonators and is the frequency
deviation away from resonance. The term is related to the coupling coe-
cient (2.74):
i,o i,o vg n
n = n . (2.135)
2rn
The recursion is started with T1 = R1 . Loss can also be incorporated into this
algorithm by substituting j with j + 1/ .
In this type of synthesis, the transfer function is directly related to the cou-
pling coecient to obtain any desired lter shape. In the methods described
earlier, the transfer functions are rational polynomials related to frequency,
wavelength or z where the coecients of the polynomials are adjusted for
any specic lter function. In using the recursive algorithm, it is possible to
36 2 Ring Resonators: Theory and Modeling

suppress unwanted sidelobes by apodization, which is realized by adjusting

the coupling coecients properly.
A straight forward method of designing parallel coupled ring resonator
lters is presented in Melloni (2001). The lter shape can be calculated by
using the given specications for bandwidth, FSR, out of band rejection and
selectivity. The formulas for a parallel N coupled ring resonator lter are as
follows:
FSR
Qn = (2.136)
gn F W HM


2n 1
gn = 2 sin .
2N
The coupling coecients are given by
 
2 4Qn 2
n = 1+ 1 . (2.137)
2Qn 2 2
A specic lter characteristic can be derived by using this method and the
relations described in Sect. 2.1.1.
For deriving the transfer functions for parallel coupled racetrack shaped
ring resonators, the matrix formalism described in Griel (2000a) is used.
The parallel coupled lter can be broken down into two components, similar
to the method shown for the serially coupled ring resonator lter, into a ring
resonator coupled to two waveguides and into a pair of uncoupled guides as
shown in Fig. 2.16.
The transfer matrix for the ring resonator coupled to two waveguides is
given by

a1n a2n b2n bn
0 0
a2n a2n
Ei Ei
1 bn
E i 0 0
= a1n a1n
E i
. (2.138)


E t bn a1n a2n b2n E t
0 0
a1n a1n
Et n bn 1 E t n+1
0 0
a2n a2n

Fig. 2.16. Schematic of a parallel coupled ring resonator lter

 2.3 Multiple Coupled Resonators 37

The transfer matrix for the pair of uncoupled waveguides is given by

j L
Ei e W Wn
0 0 0 E
i
E i 0 e jW LW n
0 0
= E i
.
0 0 e jW LW n
0
E t E t
0 0 0 ejW LW n
Et n E t n+1
(2.139)

The coecients a1n , a2n and bn are given by

1n 2n ej(2W LCn +R rn )
bn = , (2.140)
1 t1n t2n e2j(W LCn +R rn )
t2n ejW LCn t1n ej(3W LCn +2R rn )
a1n = ,
1 t1n t2n e2j(W LCn +R rn )
t1n ejW LCn t2n ej(3W LCn +2R rn )
a2n = .
1 t1n t2n e2j(W LCn +R rn )
The arrows used as the superscripts indicate whether the eld is propagating
to the left or the right direction. Multiplying the transfer matrices in an alter-
nating rhythm, the transfer functions for a parallel N coupled ring resonator
lter can be derived. Note that in the calculation the propagation constant
for the uncoupled waveguides and for the ring resonators is assumed to be the

same for each one, respectively. The drop port E t1 and the throughput E iN
responses from the resulting matrix H, where h represent the elements of the

matrix, are then given by (assuming only one input, E i1 = 1):
h3,2 1
E t1 = E iN = , (2.141)
h2,2 h2,2

where E tN = E i1 = 0.
As in the recursive algorithm mentioned earlier in this chapter, the cou-
pling coecients can also be varied to control the bandwidth of the lter and
thus reducing the sidelobes by apodization.
Several congurations to obtain a box-like lter response using an array
of parallel coupled ring resonators are presented in Ma et al. (2005). A special
conguration consisting of a parallel coupled double ring resonator (with two
waveguides), serially coupled with a single ring resonator has been analyzed
in Okamoto et al. (2003).

Coupled to One Input/Output Waveguide

The basic ring resonator lter conguration consisting of a ring and a wave-
guide is used to realize a multiple parallel coupled lter. Another term for
38 2 Ring Resonators: Theory and Modeling

parallel coupled ring resonators with only one input/output waveguides is

side-coupled, integrated, space sequence of resonators (SCISSORS) (Heebner
et al. 2002). The transfer functions of these types of lters can be derived using
the methods and formulas described earlier in this chapter. In this chapter
the focus lies on the optical properties of this kind of device which have been
analyzed in detail in Heebner et al. (2002) and Pereira et al. (2002). The
following expressions are based on this literature.
In order to describe linear propagation eects, a pulse is considered, which
travels through a single ring resonator, obtaining a frequency-dependent phase
shift that results in a delay or distorts the pulse shape (see also section The
Time-Dependent Relations). The eective propagation constant for a res-
onator spacing of is then given by
ne ()
0 () = + , (2.142)
c0
where () is the phase shift which a eld acquires when traversing a ring
resonator. It is derived from the transfer function of a single ring resonator
(Fig. 2.2) where the throughput eld is related to the input eld (2.9).
Et1 () = ej() Ei1 () , (2.143)


t sin ()
() = + () + 2 arctan .
1 t cos ()
The transfer function of a single ring resonator can be expanded into two terms
using the Taylor series, representing the transmitted (expanded about the
normalized detuning 0 ) and the exponential phase shift (expanded about the
transmitted phase shift of the carrier 0 ). The transfer function is given by
$
  n 0

j n
1 dm 
H () = ej = ej0 1 +  ( 0 )m
. (2.144)
n!
n=1
m! dm 
m=1 0

Using this equation, the eld at some point zl+1 expressed by the eld at
another point zl which is only a small distance z away is given by (the trans-
mitted phase shift induced by each ring resonator is assumed to be distributed
over the separation , leading to an eective propagation constant which is
independent of propagation distance):
&   '
ne 0 0
El+1 () = exp j + z
c0
$   n 0

jn ne

1 z dm 
1+ z + ( 0 )m El () .
n! c0 m! dm 
n=1 m=1 0

(2.145)

The Fourier transform of (2.145) leads to a dierence equation for the transfer
function expanded to two Tailor series (transmitted and exponential phase
shift)
 2.3 Multiple Coupled Resonators 39

Al+1 (t) =

 
m n

j n ne
1 z dm  j
Al (t) + j z + Al (t)
n=1
n! c0 t m=1 m! d m 0 FSR t
(2.146)

For weak coupling, (2.146) can be rewritten using

Al+1 (t) Al (t) z0 dA

(2.147)
z dz
leading to
 
m 
dA ne
1 1 dm  j
= +j A. (2.148)
dz c0 t m=1
m! m 0 FSR t

This equation yields the group-velocity reduction and group-velocity disper-

sion including higher order dispersion. The group-velocity reduction, which is
proportional to the inverse of the rst frequency derivative of the propagation
constant, is given by the term:


 de ne 1 d =0,t1 ne 4r
e = = + 0 1+ F , (2.149)
d c0 d c0

where F is the nesse and r the radius of the ring resonator (see (2.30)).
The group-velocity dispersion, which is proportional to the second frequency
derivative of the eective propagation constant, is given by:
2

 de 1 d2 0 = F 3 3 3F 2
e = = 2 . (2.150)
d 2 d 2 4 FSR2
The dispersion maxima are obtained at a detuning of 0 = F 3 .
Higher-order dispersion is derived in a similar way as described before.
The expression for the third-order dispersion is given by the term:

 1 d3 0 =0,t1 4 F3
e = ,
. (2.151)
d 3 3 FSR3
All orders of dispersion have to be taken into account if the pulse bandwidth
corresponds approximately to the resonance bandwidth given by the FWHM.
Ring resonators are versatile devices and do not only exhibit linear proper-
ties, but can also be used to realize nonlinear eects. This can be accomplished
by using a material system like InGaAsP or GaAs for example. The eective
nonlinear propagation constant is given by (assuming that the nonlinearity of
the bus waveguide does not contribute signicantly and can be neglected):
2
1 d 1 d d d |Ei2 | 0 =0,t1 nl 8r 2
nl e = 2 = F . (2.152)
d |Ei1 | d d |Ei2 |2 d |Ei1 |2
40 2 Ring Resonators: Theory and Modeling

The linear propagation (2.148) can then be modied to:

 
dA ne
1 1 dm 
= +j
dz c0 t m=1
m! m 0

m 
2 j
nl 2rB |A| + A, (2.153)
F SR t

where B is the intensity enhancement or buildup factor (see (2.32)(2.36)).

This conguration of side-coupled ring resonators exhibits strong disper-
sive and nonlinear properties and also supports solitons. As fabrication tech-
nologies mature, integration of these kinds of devices becomes feasible.
The following chapter highlights the realization and characterization of
ring resonator devices in dierent material systems.
http://www.springer.com/978-3-540-68786-3