Supporting Information
Santis et al. 10.1073/pnas.1400184111
Frequency Noise Measurement
Lasers are characterized with regard to their temporal coherence
by measuring their frequency noise using the configuration of
Fig. S1.
Lasers under test are mounted on a stage cooled by a ther-
moelectric cooler (TEC), and the output power is coupled into
a lensed fiber. After passing through a pair of optical isolators
(ISO) to suppress backreflections, the polarization of the signal is
rotated (PC) for maximum amplification by a booster optical
amplifier (BOA; Thorlabs BOA 1004), used to increase the
measurement
’
s dynamic range. Phase fluctuations in the laser
field are converted to intensity
fluctuations measured on a high-
speed photodetector (PD; HP 119
82A). This conversion is accom-
plished using a Mach
–
Zehnder interferometer (MZI) with sub-
coherent time delay (
τ
=
1
:
2
ns, free
spectral range
=
847
MHz)
biased at quadrature (
ωτ
=
π
2
). The output voltage of the photode-
tectorismonitoredonanoscilloscope(OSC)totracktheMZI
’
s
state relative to quadrature, where
as its spectrum is resolved with an
rf spectrum analyzer (SA; Agilen
t 4395A). The electrical power
spectral density
W
SA
(here in decibels per hertz) recorded on the
spectrum analyzer is related to the frequency noise power spectral
density
S
ν
(in hertz
2
per hertz, using standard units) via
S
ν
ð
f
Þ
=
R
L
4
π
2
τ
2
υ
2
G
10
W
SA
=
10
;
[S1]
where
R
L
is the input impedance of the spectrum analyzer and
υ
G
is a gain factor that corresponds to the excursion amplitude
over one fringe of the MZI. Our measurement range is limited to
offset frequencies below around 150
MHz to stay in the flat region
of the MZI frequency response. An MZI with a bigger free-spec-
tral range would not provide enough frequency-to-amplitude gain
to keep the measurement above the noise floor of our spectrum
analyzer.
The unpackaged high-
Q
hybrid lasers are susceptible to am-
bient temperature fluctuations that cause the center frequency to
jitter. To keep the MZI at quadrature long enough for a high-
resolution scan of the spectrum analyzer, we stabilize the lasers
against temperature fluctuations by implementing negative
electronic feedback. Part of the photodetector
’
soutputistap-
ped to create a correction term that is fed into the current
source
’
s (CS; ILX LDX-3620) external modulator (EXT
MOD) and finally back into the laser. The loop bandwidth is
kept below 100
Hz to ensure that noise at frequencies higher
than this cutoff remain unaffected by any artificial noise sup-
pression. To confirm that this indeed is the case, we measured
a commercial, packaged laser with and without feedback. The
commercial laser is a JDSU DFB semiconductor laser (CQF-
935/908). Fig. S2 shows the frequency noise power spectral
density of the JDSU laser measured with and without feedback.
The spectrum measured with feedback exhibits significant
suppression at low offset frequencies
ð
<
100
Hz
Þ
, as intended.
The excess noise in the intermediate frequency range
ð
1
kHz to 1
MHz
Þ
in the case of feedback, is attributed to noise
injected by the electronics of the external modulation module of
the current source. At high offset frequencies
ð
>
1
MHz
Þ
,the
two spectra are nearly identical, confirming that the feedback
stabilization does not affect the part of the frequency noise
spectrum from which the spectral linewidth is extracted.
For every measurement of the frequency noise, we also track
the level of all other noise sources pertinent to the measurement.
Fig. S3 shows a representative frequency noise power spectral
density spectrum of a high-
Q
hybrid laser in decibel-milliwatts
per hertz, plotted against the laser
’
s intensity noise, the photo-
detector
’
s dark current noise, and spectrum analyzer
’
s noise
sensitivity. Intensity noise is measured using the setup of Fig. S1
with the MZI replaced by a variable attenuator, used to keep
optical power incident on the photodetector constant for the two
measurements. For all pump currents, the laser frequency noise
is found to be well above
ð
>
10
dB
Þ
the respective intensity noise
level. Also, for all pump currents, except those close to thresh-
old, the intensity noise is below the dark noise level, which sets
the measurement noise floor.
Comparison with the Control Laser
In addition to using the JDSU DFB laser for calibration of the
noise measurement, we also use it as a benchmark of comparison
against our high-
Q
hybrid lasers. Fig. S4 presents the frequency
noise spectrum of a high-
Q
hybrid laser with a linewidth of 18
kHz
measured at 4
:
5
×
I
th
ð
160
mA
Þ
, plotted against the spectrum of
the JDSU laser taken at two different pump currents.
The JDSU laser exhibits a clear white noise floor in both cases,
with spectral linewidths of 500
kHz at 4
×
I
th
ð
100
mA
Þ
and
160
kHz at 12
×
I
th
ð
300
mA
Þ
.
We also compare the two lasers
’
linewidth dependence on
pump current. Immediately above threshold
ð
1
:
2
×
I
th
Þ
, the high-
Q
hybrid laser exhibits sub-megahertz
–
scale linewidth whereas
the JDSU has a linewidth of about 10
MHz.
III-V Wafer Structure
The III-V wafer used in this work was purchased from Archcom
Technology. The wafer structure may be found in Table S1.
Santis et al.
www.pnas.org/cgi/content/short/1400184111
1of4
ISO
ISO
PC
τ
ΜΖΙ
BOA
OSC
SA
+
-
REF
EXT MOD
CS
SP
LPF
TEC
PD
LASER
Fig. S1.
Experimental setup for the measurement of laser frequency noise.
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
4
10
5
10
6
10
7
10
8
10
9
10
10
10
11
Offset Frequency (Hz)
Frequency Noise PSD (Hz
2
/Hz)
w/o Feedback
w/ Feedback
Fig. S2.
Frequency noise spectrum of the commercial JDSU DFB laser with and without feedback.
10
3
10
4
10
5
10
6
10
7
10
8
−160
−150
−140
−130
−120
−110
−100
Offset Frequency (Hz)
Power Spectral Density (dbm/Hz)
PD Dark Noise Floor
SA Noise Sensitivity
HQ Hybrid Frequency Noise (160 mA)
HQ Hybrid Intensity Noise (160 mA)
Fig. S3.
Power spectral density of all noise sources in the measurement, demonstrating that the frequency noise is stronger than other noise sources in this
measurement.
Santis et al.
www.pnas.org/cgi/content/short/1400184111
2of4
10
3
10
4
10
5
10
6
10
7
10
8
10
3
10
4
10
5
10
6
10
7
10
8
10
9
Offset Frequency (Hz)
Frequency Noise PSD (Hz
2
/Hz)
HQ Hybrid (I = 4.5
I
th
= 160 mA)
JDSU (I = 12
I
th
= 300 mA)
I
JDSU (I = 4
th
= 100 mA)
Fig. S4.
Comparison of the frequency noise spectrum between the high-
Q
hybrid laser and the JDSU laser, measured using the same setup.
10
−1
10
0
10
1
10
2
10
4
10
5
10
6
10
7
Offset Current (I/I
th
−1)
Spectral Linewidth (Hz)
JDSU DFB
HQ Hybrid
Fig. S5.
Calculated linewidth of the high-
Q
hybrid and JDSU DFB lasers. The solid line is a calculated fit to
ð
I
−
I
th
Þ
−
1
.
Santis et al.
www.pnas.org/cgi/content/short/1400184111
3of4
Table S1. III-V wafer structure
Layer name
Material
Height, nm Doping, cm
−
3
Index
Substrate
InP
n
3.1
Buffer
InP
500
n
=
1
×
10
18
3.1
p-Contact
In
0.53
Ga
0.47
As
200
p
>
1
×
10
19
3.43
Cladding
InP
1,500
p
=
1
×
10
18
3.1
→
5
×
10
17
SCL
1.15Q InGaAsP
40
3.33
1.25Q InGaAsP
40
3.3755
QWs
×
5 (1% comp. strain)
InGaAsP
7
3.53
QW barriers
×
4 (3% ten. strain)
InGaAsP
10
3.3755
SCL
1.25Q InGaAsP
40
3.3755
1.15Q InGaAsP
40
3.33
Cladding
InP
110
n
=
1
×
10
18
3.1
Superlattice
ð
×
2
Þ
In
0.85
Ga
0.15
As
0.327
P
0.673
7.5
3.25
InP
7.5
n
=
1
×
10
18
3.1
Bonding layer
InP
10
n
=
1
×
10
18
3.1
The top two layers (substrate and buffer) are removed after wafer bonding. QW, quantum well; SCL, separate
confinement layer.
Santis et al.
www.pnas.org/cgi/content/short/1400184111
4of4