of 7
arXiv:1607.08222v1 [physics.optics] 27 Jul 2016
Microresonator Soliton Dual-Comb Spectroscopy
Myoung-Gyun Suh
1
,
, Qi-Fan Yang
1
,
, Ki Youl Yang
1
, Xu Yi
1
, and Kerry J. Vahala
1
,
1
T. J. Watson Laboratory of Applied Physics, California Institute of
Technology, Pasadena, California 91125, USA.
These authors contributed equally to this work.
Corresponding author: vahala@caltech.edu
Rapid characterization of optical and vibrational
spectra with high resolution can identify species
in cluttered environments and is important for
assays and early alerts. In this regard, dual-
comb spectroscopy has emerged as a powerful
approach to acquire nearly instantaneous Raman
and optical spectra with unprecedented resolu-
tion. Spectra are generated directly in the electri-
cal domain and avoid bulky mechanical spectrom-
eters. Recently, a miniature soliton-based comb
has emerged that can potentially transfer the
dual-comb method to a chip platform. Unlike ear-
lier microcombs, these new devices achieve high-
coherence, pulsed mode locking. They generate
broad, reproducible spectral envelopes, which is
essential for dual-comb spectroscopy. Here, dual-
comb spectroscopy is demonstrated using these
devices. This work shows the potential for in-
tegrated, high signal-to-noise spectroscopy with
fast acquisition rates.
Since their demonstration in the late 1990s
1–3
, opti-
cal frequency combs have revolutionized precision mea-
surements of time and frequency and enabled new tech-
nologies such as optical clocks
3
, low-noise microwave
generation
4
and dual-comb spectroscopy
5–12
, while also
adding performance capablibility to methods like coher-
ent LIDAR
13–15
. In spectroscopic applications, frequency
comb systems exist across a broad spectral range span-
ning ultraviolet to infrared, making them well suited for
measurement of diverse molecular species. At the same
time, the method of dual-comb spectroscopy leverages
the coherence properties of combs for rapid, broad-band
spectral analysis with high accuracy
16
.
In parallel with advancements in frequency comb
applications, the past decade has witnessed the ap-
pearance of a miniature optical frequency comb or
microcomb
17,18
. These microcombs have been demon-
strated across a range of emission bands using several
dielectric materials
17,19–24
. Under continuous-wave laser
pumping, the combs are initiated by way of paramet-
ric oscillation
25,26
and are broadened by cascaded four-
wave mixing
17,18
to spectral widths that can encompass
an octave of spectrum
22
. Four-wave mixing in the ultra-
fast intraband gain medium of quantum cascade lasers
(QCL) has also been shown to create frequency modu-
lation (FM) combs
27
. These FM systems have been ap-
plied to demonstrate dual-comb spectroscopy in the mid
infrared
28
. Also, heterodyne of two conventional micro-
combs in the mid infrared has been demonstrated, a key
Test
Sample
3 mm
f
n
=f
c1
+nf
r1
Soliton I
f
n
=f
c2
+nf
r2
Soliton II
f
n
=
f
c
+n
f
r
RF Domain
Optical Domain
Absorption spectrum
Dual-Comb
Spectroscopy
CW Laser I
Microresonator I
CW Laser II
Microresonator II
Reference
Signal
FIG. 1:
Microresonator-based dual-comb spec-
troscopy
. Two soliton pulse trains with slightly differ-
ent repetition rates are generated by continuous optical
pumping of two microresonators. The pulse trains are
combined in a fiber bidirectional coupler to produce
a signal output path that passes through a test sam-
ple as well as a reference output path. The output of
each path is detected to generate an electrical interfero-
gram of the two soliton pulse trains. The interferogram
is Fourier transformed to produce comb-like electrical
spectra having spectral lines spaced by the repetition
rate difference of the soliton pulse trains. The absorp-
tion features of the test sample can be extracted from
this spectrum by normalizing the signal spectrum by
the reference spectrum. Also shown is the image of two,
silica wedge disk resonators. The disks have a 3 mm
diameter and are fabricated on a silicon chip.
step towards dual-comb spectroscopy
29
. Direct hetero-
dyne detection of two QCL FM combs in the laser current
has also been shown
30
.
A major advancement in microcombs has been the
realization of soliton mode-locking
31–35
. Soliton micro-
combs feature dissipative Kerr solitons that leverage the
Kerr nonlinearity to both compensate dispersion and to
overcome cavity loss by way of parametric gain
36
. Un-
like earlier microcombs, this new device provides phase-
locked femtosecond pulses with well-defined, repeatable
spectral envelopes, which is important for dual-comb
spectroscopy. Their pulse repetition rate is detectable
and has excellent phase noise characteristics
32
. In this
work, we demonstrate dual-comb spectroscopy using this
new platform. The dual-comb source spans over 30 nm
with 22 GHz optical spectral resolution and the interfer-
ogram spectra feature high signal-to-noise. Also, precise
2
Dual soliton generation setup
PD
Spectroscopy setup
PC
a
b
c
50/50
400
Power (dBm)
-20
-40
-60
-80
TEC
EDFA
AOM
μ
Disk
FBG
CW laser
PD
PC
Feedback Loop
EDFA
AOM
μ
Disk
FBG
CW laser
PD
PC
Feedback Loop
PD
Power (dBm)
-20
-40
-60
-80
Wavelength (nm)
1500
1520
1540
1560
1580
1600
Power (20 dB / div)
Power (20 dB / div)
Frequency offset (kHz)
-400
-200
0
200
21.9842 GHz
21.9815 GHz
RBW
500 Hz
RBW
500 Hz
Soliton I
Soliton II
OSA
Oscilloscope
ESA
Pump
Pump
Servo
Servo
90/10
90/10
d
e
Signal
Reference
Gas Cell
Waveshaper
FIG. 2:
Detailed experimental setup and soliton comb characteriza
tion
.
a,
Continuous-wave (CW) fiber
lasers are amplified by erbium-doped fiber amplifiers (EDFA) and coup
led into high-Q silica wedge microresonators
via tapered fiber couplers. An acousto-optic modulator (AOM) is us
ed to control pump power to trigger soliton
generation in the microresonators. Polarization controllers (PC) a
re used to optimize resonator coupling. A fiber
Bragg grating (FBG) removes the transmitted pump power in the so
liton microcomb. The pump laser frequency is
servo controlled to maintain a fixed detuning from the microcavity re
sonance by holding the soliton average power
to a fixed setpoint. An optical spectrum analyzer (OSA) monitors t
he spectral output from the microresonators.
The two soliton pulse streams are combined in a bidirectional coupler a
nd sent to a gas cell (or a WaveShaper) and
a reference path. The interferograms of the combined soliton puls
e streams are generated by photodetection (PD)
and recorded on an oscilloscope. The repetition rates of the soliton
pulse streams are also monitored by an electri-
cal spectrum analyzer (ESA). The temperature of one resonato
r is controlled by a thermoelectric cooler (TEC) to
tune the optical frequency difference of the two solitons.
b-c,
Optical spectra of the microresonator soliton pulse
streams.
d-e,
Electrical spectra showing the repetition rates of the soliton pulse
streams. The rates are given in the
legends.
microfabrication control enables close matching of the
repetition rates so that over 4 THz of optical bandwidth
is measured within 500 MHz of electrical bandwidth.
A schematic view of the dual comb experimental setup
is provided in Fig. 1. Two soliton trains having dif-
ferent repetition rates (∆
f
r
=
f
r
1
f
r
2
) are generated
from distinct microresonators and then combined using
a directional coupler. One of the combined streams is
coupled through a gas cell of molecules whose absorp-
tion spectrum is to be measured. The other combined
stream functions to provide a reference. The slight dif-
ference in repetition rates of the soliton streams creates
a periodically time-varying interferogram in the detected
current with a period 1
/
f
r
. Fourier transform of this
time-varying signal reveals the interfering soliton comb
spectra, now shifted to radio-frequency rates. The signal
spectrum containing the molecular absorption informa-
tion is then normalized using the reference spectrum to
reveal the spectral absorption of the gas cell.
Figure 2a gives further details on the experimen-
tal setup. Solitons are generated and stabilized in
two microresonators using the active-capture/locking
technique
37
. The microresonators are pumped at
1549.736 nm and 1549.916 nm using two amplified fiber
3
a
d
0.0
1.0
0.5
1545
1550
1555
1560
1565
Absorption
Wavelength (nm)
1545
1550
1555
1560
1565
0.0
1.0
0.5
Absorption
Wavelength (nm)
e
200
300
400
500
600
700
800
Radiofrequency (MHz)
Intensity (10 dB / div)
2.6 MHz
500
Voltage (V)
0.4
0
-0.4
-0.8
0.8
Time (ns)
0
100
200
300
400
386 ns
200
300
400
500
600
700
800
Radiofrequency (MHz)
Intensity (10 dB / div)
Intensity (10 dB / div)
Intensity (10 dB / div)
Radiofrequency (MHz)
b
800
200
300
400
500
600
700
c
490
500
510
Radiofrequency (MHz)
MHz (+
502.25
MHz)
0
-1
1
< 50 kHz
FIG. 3:
Measured electrical interferogram and spectra. a,
The detected interferogram of the reference soli-
ton pulse train.
b,
Typical electrical spectrum obtained by Fourier transform of the
temporal interferogram in 3a.
To obtain the displayed spectra, ten spectra each are recorded o
ver a time of 20
μ
s and averaged.
c,
Resolved (mul-
tiple and individual) comb lines of the spectrum in 3b are equidistantly se
parated by 2.6 MHz, the difference in
the soliton repetitation rate of the two microresonators. The linew
idth of each comb line is
<
50 kHz and set by
the mutual coherence of the pumping lasers.
d-e,
Fourier-transform (black) of the signal interferogram produce
d
by coupling the dual-soliton pulse trains through the WaveShaper (s
ee Fig. 2a) with programmed absorption func-
tions (spectrally flat and sine-wave). The obtained dual-comb abso
rption spectra (red) are compared with the pro-
grammed functions (blue curves) from 1545 nm to 1565 nm.
lasers (Orbits Lightwave), but in principle, pumping from
a single laser is possible. The difference frequency of the
pumps was determined to be 22.5 GHz by detecting their
electrical beat note and measurement on a spectrum an-
alyzer. After amplification, each pump laser is coupled
to an acousto-optic frequency modulator (AOM). The
frequency-shifted output of the AOM is used to provide
a controllable optical pumping power that is required
for soliton triggering
37
. The pump light is then evanes-
cently coupled into the silica microresonator via a fiber
taper
38,39
. Residual pumping light that is transmitted
past each resonator is filtered using a fiber Bragg grating
(FBG). After the FBG, a 90/10 tap is used to monitor the
soliton power for feedback control of the pump laser fre-
quency so as to implement soliton locking
37
. The optical
spectra of the individual soliton streams was monitored
using a Yokogawa optical spectrum analyzer. Additional
precision calibration of the spectra was possible using a
Wavelength References Clarity laser locked to a molec-
ular absorption line. Typical soliton optical spectra are
presented in Figs. 2b-c and feature the characteristic
sech
2
envelope observed in this case over a 60 nm wave-
length span. The detected electrical spectrum for each
soliton source is also shown in Figs. 2d-e. The narrow
spectral lines measured with a resolution bandwidth of
500 Hz have a signal-to-noise greater than 75 dB showing
4
that corresponding repetition rates are extremely stable.
The high-Q resonators used in this work are described
elsewhere
40
. Briefly, they are silica wedge devices fabri-
cated on a silicon wafer using a combination of lithogra-
phy and wet/dry etching. The unloaded quality factor of
the microresonators exceeds 300 million, and the gener-
ated solitons have repetition rates determined primarily
by the diameter of the devices (3 mm). The repetition
rate difference of the two microcomb devices is controlled
by varying the silica resonator etching time
40
.
The optical outputs from the stabilized soliton sources
are combined and coupled into two paths as shown in
Fig. 2a. One path contains a 16.5 cm-long 300 Torr
H
13
CN gas cell manufactured by Wavelength References,
Inc. which functions as the test sample in the measure-
ment. The other path is coupled directly to a photode-
tector and functions as the reference. The test sample
path also contained an alternate path in which a Finisar
WaveShaper was inserted. The WaveShaper required an
erbium fiber amplifier to compensate its insertion loss.
As demonstrated below, the WaveShaper allowed syn-
thesis of arbitray spectral transmission profiles to fur-
ther verify the dual comb operation. Detection to gen-
erate the interferograms used u2t photodetectors with
bandwidths of 50 GHz. Temperature control of one of
the microresonators was used to tune the relative optical
frequency difference of the two solitons streams. In the
measurements this difference was held below 1 GHz, al-
lowing the observation of the temporal interferogram on
an oscilloscope (bandwidth 1 GHz). The spectrum of the
photocurrent signals was also measured to determine the
soliton repetition rates (see Figs. 2d-e) using an electrical
spectrum analyzer (Rhode Schwartz) with a bandwidth
of 26 GHz.
The reference interferogram produced by detection of
the lower path in Fig. 2a and as recorded on the os-
cilloscope is shown in Fig. 3a. It has a period of 386
ns, corresponding to a soliton repetition rate difference
of 2.6 MHz. This relatively small repetition rate differ-
ence was possible by precise lithographic control of the 22
GHz soliton repetition rate. It was possible to fabricate
disks with even more closely matched repetition rates (
<
100kHz). Figure 3b shows the calculated Fourier trans-
form of the interferogram. The small repetition rate dif-
ference on the much larger 22 GHz soliton repetition rate
makes it possible compress an optical span of 4 THz (1535
nm to 1567 nm) into 500 MHz of electrical spectrum. The
measured wavelength span is actually narrower than the
observable wavelength span of the original soliton pulse
streams and is limited by the photodetector noise. The
interferogram spectrum has a signal-to-noise ratio (SNR)
in excess of 30 dB near the central lines. A zoom-in of the
spectrum (multi- and single-line) is provided in Fig. 3c.
The electrical comb lines are equidistantly separated by
2.6 MHz and have a full-width-half-maximum linewidth
less than 50 kHz, limited by the mutual coherence of the
independent fiber pump lasers. The pump laser line in a
dissipative Kerr soliton is also a comb tooth in the soliton
Scanning laser
Dual comb
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
0.0
0.2
1540
1545
1550
1555
1560
Wavelength (nm)
a
b
1545
1550
1555
0.0
0.2
0.4
0.6
0.8
1.0
Wavelength (nm)
Absorption
Absorption
Absorption
Residual
σ = 0.0254
FIG. 4:
Measured molecular absorption spectra.
a,
Absorption spectrum of 2
ν
3
band of H
13
CN mea-
sured by direct power transmission using a wavelength-
calibrated scanning laser (see Methods section) and
comparison to the microresonator-based dual-comb
spectrum. The residual difference between the two
spectra is shown in green.
b,
Overlay of the directly
measured optical spectrum and the dual-comb spec-
trum showing line-by-line matching. The vertical po-
sitions of the two spectra are adjusted to compensate
insertion loss.
optical spectrum. As a result, the frequency jitter in each
pump is transferred as an overall shift on the resulting
soliton comb. Externally locking the two combs should
reduce the observed linewidth in the interferogram spec-
tum.
It is interesting to note the placement of the pump
lines toward the high frequency side (near 550 MHz) of
the spectral maximum in the interferogram spectrum (see
Fig. 3b). In the optical spectra (Figs. 2b-c) the pump
is blue detuned relative to the soliton spectral maximum
(this occurs on account of the Raman self-frequency-shift
of the soliton
41–43
). This spectral landmark shows that
the relative spectral placement of the soliton combs is
such that high optical frequencies are mapped to high in-
terferogram frequencies. It is also interesting to note how
certain non-idealities in the soliton spectra map into the
interferogram spectrum. Specifically, there are avoided-
mode-crossing induced Fano-like spurs
32
in the soliton
5
optical spectra (Figs. 2b-c) occurring near 1535 nm and
this generates a corresponding feature at 750 MHz in Fig.
3b.
As an initial test of the dual-comb source, artificial
absorption spectra were programmed in a Finisar Wave-
Shaper 1000S and then measured as dual-comb spec-
tra. In Figs. 3d-e, electrical spectra Fourier-transformed
from the signal interferograms after coupling through the
WaveShaper are shown. The two programmed functions
are a spectrally flat 3 dB absorption and a sine-wave
absorption having a 4 dB amplitude. The absorption
spectra, obtained by normalizing the signal and reference
electrical spectra, are compared with the programmed
functions in Figs. 3d-e. The ability to reconstruct these
synthetic spectral profiles clearly demonstrates the repro-
ducibility of solitonic spectral profile.
Finally, the absorption spectrum of the H
13
CN 2
ν
3
band is studied. In Fig. 4a, the measured dual-comb
absorption spectrum from 1538 nm to 1562 nm is shown
in red and compared with a directly measured absorp-
tion spectrum shown in blue. Both absorption spectra
are normalized. The direct measurement was performed
by a wavelength-calibrated scanning laser (see Methods
section). Sampling-induced choppiness of the dual-comb
spectrum is caused by the relatively coarse spectral reso-
lution of the solitons in comparison to the spectral scale
of the H
13
CN absorption lines. Nonetheless, the charac-
teristic envelope of H
13
CN 2
ν
3
band is clearly resolved.
The residual difference between the two absorption spec-
tra is shown in green and the calculated standard devi-
ation is 0.0254. Furthermore, a line-by-line overlay of
the measured optical and dual-comb spectra is shown in
Fig. 4b to visually confirm the wavelength precision and
absorption intensity accuracy of the dual-comb source.
In principle, a finer comb spacing (lower repetition
frequency) soliton source is possible. Non-soliton mi-
crocombs having mode spacings as narrow as 2.4 GHz
have been demonstrated using the silica wedge resonator
platform
23
. Modulating the microcombs by an integer
factor of the repetition frequency using electro-optical
modulators is another possible way to create a finer spec-
tral comb grid. On the other hand, larger mode spac-
ings could allow studies of fast dynamic processes such as
chemical reactions and rapid measurements of the broad
absorption features in liquids or solids
44,45
.
In conclusion, two soliton microcombs featuring highly
balanced microwave repetition rates were used as a dual-
comb spectroscopy system to measure the absorption
spectrum of the 2
ν
3
band of H
13
CN. This is the first
demonstration of a microresonator soliton-based dual-
comb spectroscopy system. The dual-comb source has
a high SNR and spans over 30 nm in optical C-band. Us-
ing fiber nonlinear broadening or internal (resonator) dis-
persive wave generation, it should be possible to greatly
extend this spectral span
12,46
. With careful engineering
of the resonator dispersion
47
it should also be possible
to cover other spectral ranges within the transmission
window of silica. More generally, a wide range of mate-
rials are available for microcombs enabling access to mid
infrared spectra. With further improvements, it should
also be possible to realize chip-based dual-comb coher-
ent anti-Stokes Raman spectroscopy (CARS). The inte-
gration with other on-chip devices
48
makes soliton-based
microcombs well suited for possible realization of a dual-
comb spectroscopic system-on-a-chip.
Methods
The H
13
CN absorption spectrum in Fig. 4a is obtained
by coupling an external cavity diode laser (ECDL) into
the H
13
CN gas cell and scanning the laser while moni-
toring the transmitted optical power. A separate signal
is also tapped from the ECDL to function as a refer-
ence. The relative wavelength change of the ECDL dur-
ing the scan is calibrated using a fiber Mach-Zehnder
interferometer and absolute calibration is obtained using
a reference laser which is locked to a molecular absorp-
tion line (Wavelength References Clarity laser). The sig-
nal passing through the gas cell and the reference trans-
missions are recorded simultaneously, and the absorption
spectrum in Fig. 4a is extracted by dividing the signal
transmission by the reference transmission.
Acknowledgments
The authors thank Nathan New-
bury at the National Institute of Standards and Technol-
ogy (NIST) and Giacomo Scalari at ETH Zurich for help-
ful comments on this manuscript. The authors gratefully
acknowledge support from the Defense Advanced Re-
search Projects Agency (DARPA) under the PULSE and
SCOUT programs, the National Aeronautics and Space
Administration (NASA) and the Kavli Nanoscience In-
stitute (KNI).
Author Contributions
Experiments were conceived by
all co-authors. Analysis of results was conducted by
MGS, QFY and KJV. MGS and QFY performed mea-
surements with assistance from XY. KYY fabricated de-
vices. All authors participated in writing the manuscript.
Author Information
Correspondence and requests
for materials should be addressed to KJV (va-
hala@caltech.edu ).
1
Jones, D. J.
et al.
Carrier-envelope phase control of fem-
tosecond mode-locked lasers and direct optical frequency
synthesis.
Science
288
, 635–639 (2000).
2
Holzwarth, R.
et al.
Optical frequency synthesizer for
precision spectroscopy.
Physical review letters
85
, 2264
(2000).
3
Diddams, S. A.
et al.
An optical clock based on a single
trapped 199hg+ ion.
Science
293
, 825–828 (2001).
4
Fortier, T. M.
et al.
Generation of ultrastable microwaves
via optical frequency division.
Nature Photonics
5
, 425–