A Report on
Potentially Low-Cost Widely Tunable Laser Consisting of a
Semiconductor Optical Amplifier Connected Directly to a
Silica Waveguide Grating Router
By
C.R. Doerr, L.W. Stulz, R. Pafchek, K. Dreyer, and L. Zhang
Submitted to
Professor Roger Dorsinville
at
The City College of New York
of
The City University of New York
Submitted by
Brandon J George
on
May 21th, 2008
CITY COLLEGE OF NEW YORK
Table of Contents
1. INTRODUCTION –
1
2. NEED FOR A LOW-COST WIDELY TUNABLE LASER –
1
3. CONCEPT OF THIS TUNABLE LASER –
2
Concept of Tuning Lasers –
2
Matching SOA to Silica Waveguide –
4
Coupling to Star Coupler from SOA –
4
Waveguide Grating Router –
5
Thermo-optic Phase Shifters –
6
Coupling to Output Fiber from Output Star Coupler –
6
4. DESIGN OF TUNABLE LASER –
7
5. RESULTS –
9
6. CONCLUSION –
10
7. APPENDIX –
11
FROM TEMPLATE
1. Introduction –
Modern networks depend on optical networks to carry large bandwidths of
information and those optical networks depend on wavelength-division-multiplexed
systems. Widely tunable lasers are needed for theses systems. Several types of widely
tunable lasers currently exist but are difficult to produce and therefore expensive, have
moving parts or are not completely solid state. C.R. Doerr, L.W. Stulz, R. Pafchek, K.
Dreyer, and L. Zhang’s solution to this problem was to simply glue a low-cost Indium
Phosphide (InP) Semiconductor Optical Amplifier (SOA) to a low-cost silica waveguide
grating router (WGR) with thermo-optic phase shifters. By this the Doerr, Stulz, Pafchek,
Dreyer, and Zhang hoped to create a relatively simple low-cost solution that is
completely solid state to the problem of the alternative expensive and precise widely
tunable amplifiers.
2. Need for a Low-Cost Widely Tunable Laser –
Modern optical networks use wavelength-division-multiplexing to carry large
bandwidths of information. Wavelength-division-multiplexing (WDM) works by carrying
multiple signals down the same fiber by using different wavelengths of light to carry each
signal.
Fig. 2-1 An example of WDM
There are solutions to these types of widely tunable lasers needed but each have
their problems. One solution is the distributed Bragg reflector laser employing gratingassisted couplers; however this requires complex InP growth, long testing and calibration,
1
sensitive control, an external wavelength monitor, and precise alignment. Another type is
the vertical-cavity surface-emitting laser with a movable top mirror. These require optical
pumping and have moving parts. Other less common types are an array of fixedwavelength lasers coupled together with a power combiner, the multi-frequency laser
(MFL) consisting of SOA’s monolithically integrated with a WGR.
So while there are solutions to widely tunable lasers there is still a need for a lowcost widely tunable laser. Since the laser presented here is built with a low-cost InP SOA
and a low-cost silica WGR simply glued together with no moving or precise parts it
provides a promising solution.
3. Concept of this Tunable Laser –
Fig. 3-1 Proposed Laser Design
Concept of Tuning Lasers –
Fig. 3-2 Grating Used to Tune Laser Cavity
2
The selection of a certain frequency within the gain bandwidth of the laser
medium is taken by choosing the length of the laser cavity to meet the following equation
when the gain medium does not fill the entire cavity size.
c

1

 2  C d  L   L L 
  n 
Where L is the index of refraction of the gain material, C is the index of refraction for
the other space of the cavity, d is the length of the cavity, L is the length of the gain
medium, and n is an integer. By choosing a specific length of a cavity we are effectively
selecting a single longitudinal mode.
Tunable lasers can work by varying the length of the laser cavity and therefore the
frequency. Tuning the laser cavity is not usually accomplished by actually moving one of
the mirrors but rather by adding a Fabry-Perot etalon into the cavity or adding dispersion
prisms into the cavity. The tunable laser proposed and built by Doerr, Stulz, Pafchek,
Dreyer, and Zhang shown in Fig. 3-1 essentially works on the same principle as the
tunable laser cavity shown in Fig. 3-2, which is another way to tune a laser cavity and
works on grating tuning.
Tuning of the cavity shown in Fig. 3-2 works by changing the angle of the grating
to the beam of light. The grating adds interferometric properties to the cavity, meaning
only certain wavelengths can add constructively in the cavity. The dispersive effects of
the diffraction grating essentially create a narrow wavelength pass-band for intracavity
transmission and can restrict oscillation of the laser to a single longitudinal mode.
The laser cavity of Fig. 3-1 starts at (17) and ends at (30) which is also serves as
the output mirror. The separate lengths of arms (34) are chosen to select one frequency in
the InP SOA’s gain bandwidth using the same principle of diffraction grating that Fig. 32 uses. The individual grating arms (34) effectively select a single wavelength and that
wavelength can be tuned by changing the length of each arm by applying a phase shift
with each phase shifter (38). With no phase shifter power, the selected wavelength is
chosen to be at the lower end of the SOA’s gain bandwidth so when the phase shifters are
turned on the cavity selects higher and higher wavelengths as the cavity gets effectively
longer.
3
Matching SOA to Silica Waveguide –
Attaching an InP SOA directly to a silica waveguide is not as easy as simply
gluing the two together. For one reason, spot-size (mode) is usually much smaller in
SOA’s than in silica waveguide since the index of refraction in silica is only around 1.45
while InP SOA has an index of refraction around 3.3 which results in a relatively small
spot-size. There are ways to convert spot-size between two materials such as lenses or a
horizontal and vertical tapered section at the output of the SOA, which is similar to the
following, but in two dimensions.
Fig. 3-3 Example of Spot Size Converter
Although both of these solutions will work they are both difficult to do and therefore are
expensive. The tapered section will likely affect the performance of the SOA.
Converting the spot-size can be achieved in one dimension by rotating the SOA
90° to the silica waveguide. This couples the horizontal polarized light in the SOA to the
vertical in the silica. We only need to couple the horizontally polarized light since our
SOA only has gain for horizontal polarization. Also a high numerical aperture star
coupler will help couple the spot-sizes.
Coupling to Star Coupler from SOA –
4
Fig. 3-4 Star Coupler and Output Waveguides
Spot-size conversion at the interface between the SOA and the silica waveguide is
also accomplished by using a high numerical aperture star coupler. A star coupler is a
radial arrangement of waveguides. The high numerical aperture star coupler distributes
the light into the many waveguides when the light from the SOA hits the lower refractive
index of the silica, it begins to spread quickly, which then gets guided into the many
different output arms of the star coupler. The many arms at the output of the star coupler
serve as the waveguide grating router.
Waveguide Grating Router –
In the WGR the light travels via rectangular optical waveguides. The light
traveling through the WGR arms is bound inside the core of the waveguide since the core
has a higher index of refraction than the cladding’s index of refraction. The difference in
refractive index is due to different doping levels in each. Since the core has a higher
index of refraction than the cladding the light is confined to the core by total internal
reflection. The difference between the refractive index of the core and that of the
cladding is called the delta of a waveguide.
The many different path lengths of the waveguide grating router are analogous to
adding a diffraction grating in the cavity of a laser. Shown later is exactly how the length
of each arm is chosen but they are chosen with the idea of simulating a diffraction grating
using multiple path lengths. When the spread of wavelengths get split into the 10 arms,
get shifted by different amounts, and then join back together, only one frequency is
designed to be passed while the others are rejected. This in conjunction with the thermooptic phase shifters, the arms can be lengthened by a specific amount, which selects a
5
higher wavelength. By having this “adjustable filter” in the laser cavity one can select a
specific wavelength to lase at.
Thermo-optic Phase Shifters –
Each 3 mm long phase shifter in the center of each grating arm works on the
thermo-optic effect. The thermo-optic effect comes from the change in refractive index of
silicon by applying heat and changing the temperature. The thermo-optic coefficient for
silicon is
dn
 1.86  10 4 K 1
dT
So by adding a heater to each phase shifter we can increase the refractive index, the phase
in each grating arm, and therefore increase the length in each grating arm. For example if
we would like a -phase shift in our 3 mm phase shifter we must increase the temperature
by ~1.4 °C. Now we can tune from C  1, 555nm to higher wavelengths simply by
applying heat to the phase shifters instead of incorporating cumbersome and precise
moving parts.
Coupling to Output Fiber from Output Star Coupler –
After the phase shifters, the grating arms need to converge again, so there is
another star coupler to join the separate paths again. The numerical aperture of this star
coupler is chosen to couple the spot-size at the output of the second star coupler directly
into the output fiber. The output port of the second star coupler is coupled through a
partially reflective facet (plane surface). This serves to define the laser cavity from facet
(17) throught the SOA, star couplers, grating arms, and phase shifters to facet (30) in Fig.
3-1.
6
4. Design of Tunable Laser –
To construct the laser built in Fig. 3-1 the first task is to build or acquire an InP
SOA. The inventors, for convenience, used a structure for making a monolithically
integrated multiple frequency laser. The cross section of the amplifier is shown below.
Fig. 4-1 Amplifier Cross Section
The SOA consists of 4 compressively strained buried quantum wells sitting on a 0.46 m
thick quaternary slab. The SOA is ~900 m. This SOA was designed for a multiple
frequency laser and is being used here for its ease of integration. It is not designed
specifically for this laser design and therefore has a few performance sacrifices. The five
main aspects of the laser that have been sacrificed are poor optical mode overlap with the
gain material, high contact resistance with the wire bond, gain begins to level off at ~90
mA of injection current, its lateral width is a narrow 1.5 m where we would like ~6 m
to match the silica, and there was no angling an the anti-reflection (AR) coated facet.
Also the two SOA facets are cleaved, one is coated with an AR coating matching to the
silica and the other is uncoated.
The WGR has ten grating arms and the length of arm m, is chosen by the
following formula.
2

M  1  



L m   round   m    m 
 A  C


2   

 

7
Where M is the total number of arms, C is the zero-phase-shift-power wavelength
(chosen to be the lower wavelength in the gain of the SOA since we can only tune to
higher wavelengths), A is the starting grating order (chosen to be ~1000 to simplify
calculations),  is the chirp parameter (which is a chirp in the grating period).  is chosen
from the following formula when A~1000.
1
 unchirped
2
 SOA   chirped 
Where  SOA is the bandwidth of the SOA and  unchirped is the unchirped free spectral
range of the waveguide grating router. In this design C  1.555  m ,
unchirped  200GHz , A  948 at C and   0.0296 . From the preceding equation and
the following bandwidth equation we can find chirped .
chirped 
2
c
 chirped
Using the parameters given we find that chirped
27nm .
The WGR can tune the wavelength from the zero-phase-shift-power wavelength
C to the highest wavelength in the SOA’s gain bandwidth by applying a parabolic phase
shift distribution via the phase shifters. The phase shifter setting in arm m to focus grating
order q and channel p is
2

M  1 

 m   2  pm  ( p  q)  m 
  mod 2

2  

q is any integer from 0 (which corresponds to the lower wavelength the SOA’s gain
bandwidth) to ~14 (which corresponds approximately to the upper wavelength of the
SOA’s gain bandwidth). p is any real number between -.5 to .5 which tunes around the
wavelength chosen by q. The modulo is used to reduce the power needed in each phase
shifter, since     2 .
The procedure to assemble the laser was as follows. First the output fiber is glued
directly to the output waveguide. This results with a 43% reflectivity and also serves as
the output mirror and defines the cavity. Next the silica chip was glued to a copper block,
which was then glued to a thermoelectric cooler. Then all 11 phase shifters (one on each
grating arm and one on the output waveguide) were attached via wire bonds to an
8
electrical connector. The next step was to prepare the SOA, which was soldered to a
small copper block and wire-bonded to the submount in which wires were attached to. As
stated before the SOA was then rotated 90° and glued to the silica chip using active
alignment (aligning with SOA powered on).
5. Results –
The threshold for lasing at 20 °C is ~50 mA. The thermo-optic phase shifter
efficiency is 2/750 mW; meaning it takes 750 mW to shift one arm 2. This means the
total power dissipation by the phase shifter in each arm can be as high as ~4 W. In the
inventors trials, the 20 °C could not be held so the thermoelectric heat sink had to be
cooled to take the measurements. The following figure shows the measured spectra of the
laser output for various values of q applied to the phase shifters. The SOA gain peak is
~35 nm and we can see that the tuning range of the laser is ~25 nm with approximately
50 mW of output power.
Fig 5-1 Laser Output with Various Phase Shifts Applied
There are a few improvements that Doerr, Stulz, Pafchek, Dreyer, and Zhang note
could improve performance of the laser. First, eliminate the SOA/glue reflection at the
SOA/WGR interface by angling the SOA waveguide. Second, Use an SOA that is purely
9
optimized for high saturation output power and good high-temperature performance
instead of one that is designed purely for full integration. Third, use trenched thermooptic phase shifters to reduce their power consumption since the most power was lost in
the phase shifters and there was significant amount of waste heat generated in the WGR
and using trenched phase shifters could reduce the power dissipated from ~4 W down as
low as 0.5 W. And last, use higher delta (difference between core and cladding indices of
refraction) silica waveguides to increase the cavity mode spacing and facilitate the
vertical mode matching to the SOA lateral mode.
6. Conclusion –
The low-cost widely tunable laser consisting of a SOA connected directly to a
silica WGR as proposed by Doerr, Stulz, Pafchek, Dreyer, and Zhang demonstrated a
tuning range of wavelengths around 25 nm. The design had a wide range of advantages
over the other widely tunable lasers currently available. This laser required much simpler
InP processing and less sensitive alignment and calibration than the distributed Bragg
reflector laser, no moving parts unlike the bulk-optic external cavity laser and verticalcavity surface-emitting laser, and the other which had complicated processing and
alignment. Doerr, Stulz, Pafchek, Dreyer, and Zhang believe if the improvements stated
above are applied to the low-cost tunable laser than the laser will have high output power
and be a practical solution for WDM systems.
10
7. Appendix –
This is a code for a Matlab m-file that I wrote for this project to simulate the WGR and
thermo-optic phase shifters used in this project. It seems to give good results and is
helpful to understand the relationship between q and p with the frequency selected by the
cavity. The function plots the output vs frequency in GHz. To use the function in Matlab
simply call “ wgr( q , p );” where q and p are the values you’d like to use to tune the
frequency of the cavity. Example: q = 0 and p = 0 is called as wgr(0,0); and corresponds
to the zero-phase-shift-power wavelength of 1550 nm. Also q = 14 and p = 0 tunes to
around the upper wavelengths. A digital copy of this code can be found in m format at
http://
CODE STARTS HERE:
function f = wgr(q, p)
X(200000)=0; % allocates space for frequency domain
X(190500:193600)=1; % sets frequencies in gain bandwidth to 1
n=(1:200000);
A=948; % sets starting grading order used in project
% this block phase shifts in each arm by modulating in frequency domain
m=1;
Y1(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=2;
Y2(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=3;
Y3(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=4;
Y4(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=5;
Y5(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=6;
11
Y6(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=7;
Y7(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=8;
Y8(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=9;
Y9(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
m=10;
Y10(n)=X(n).*exp(-j*(round((m+.0296*(m(10+1)/2)^2)*A)*(2*pi*1550*n/(3e8))+2*pi*((p*m)+(p+q)*.0296*(m-11/2)^2)));
% adds the split and shifted signals back together.
Y=1/10*(Y1+Y2+Y3+Y4+Y5+Y6+Y7+Y8+Y9+Y10);
np=(190000:194000); % sets region to plot
plot(np,Y(np));
12