Octave-spanning dissipative Kerr soliton frequency combs in $Si_3N_4$ microresonators

1.9 MB·

by  Martin H. P. Pfeiffer

Octave-spanning dissipative Kerr soliton frequency combs in Si N microresonators
3 4
Martin H. P. Pfeiffer1, Clemens Herkommer1,2, Junqiu Liu1, Hairun Guo1,
Maxim Karpov1, Erwan Lucas1, Michael Zervas1, Tobias J. Kippenberg1∗
1. École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland and
2. Technische Universität München (TUM), D-80333, München, Germany
Octave-spanning,self-referencedfrequencycombsareappliedindiversefieldsrangingfrompreci-
sionmetrologytoastrophysicalspectrometercalibration. Theoctave-spanningopticalbandwidthis
typicallygeneratedthroughnonlinearspectralbroadeningoffemtosecondpulsedlasers. Inthepast
decade, Kerr frequency comb generators have emerged as novel scheme offering chip-scale integra-
tion, high repetition rate and bandwidths that are only limited by group velocity dispersion. The
recentobservationofKerrfrequencycombsoperatinginthedissipativeKerrsoliton(DKS)regime,
7 along with dispersive wave formation, has provided the means for fully coherent, broadband Kerr
1 frequency comb generation with engineered spectral envelope. Here, by carefully optimizing the
0 photonic Damascene fabrication process, and dispersion engineering of Si N microresonators with
3 4
2 1THzfreespectralrange,weachievebandwidthsexceedingoneoctaveatlowpowers(O(100mW))
b for pump lasers residing in the telecom C-band (1.55µm), as well as in the O-band (1.3µm). Pre-
e cise dispersion engineering enables emission of two dispersive waves, increasing the power in the
F spectral ends of the comb, down to a wavelength as short as 850nm. Equally important, we find
thatforTHzrepetitionratecombstates,conventionalcriteriaappliedtoidentifyDKScombstates
8
fail. Investigatingthecoherenceofgenerated,octave-spanningKerrcombstatesweunambiguously
2
identify DKS states using a response measurement. This allows to demonstrate octave-spanning

DKS comb states at both pump laser wavelengths of 1.3µm and 1.55µm including the broadest
]
s DKS state generated to date, spanning more than 200THz of optical bandwidth. Octave-spanning
c DKS frequency combs can form essential building blocks for metrology or spectroscopy, and their
ti operationat1.3µmenablesapplicationsinbiologicalandmedicalimagingsuchasKerrcombbased
p optical coherence tomography or dual comb coherent anti-stokes Raman scattering.
o
.
s
c I. INTRODUCTION quency combs feature high coherence across their band-
i widthandasmoothspectralenvelopethatcanbenumer-
s
y Optical frequency combs with coherent optical band- ically predicted with high accuracy using the Lugiato-
h widths of one octave or more are required for many ap- Lefever equation or in frequency domain via coupled
p
plications, such as precision spectroscopy [1], optical fre- mode approaches [8–10]. Furthermore, higher order
[
quency synthesis [2] or astrophysical spectrometer cal- GVD causes soliton Cherenkov radiation, an oscillatory

2 ibration [3]. Conventionally these spectra are synthe- tailinthetemporalsolitonpulseprofile,correspondingto
v sized by nonlinear spectral broadening of a pulsed laser adispersivewave(DW)inthespectraldomain,whichex-
4 [1]. However, ensuring coherent spectral broadening, a tendsthespectralcombbandwidthintothenormalGVD
9
smooth spectral envelope, and sufficient power in the regime [11]. Soliton Cherenkov radiation in microres-
5
8 spectralendsforagivenpulsesourcecanbechallenging, onators is a fully coherent process and has e.g. allowed
0 in particular for high repetition rate pulse sources. Kerr self-referencingandstabilizationwithoutadditionalspec-
. frequency combs [4] have emerged as alternative scheme tral broadening via the 2f-3f-method [12]. These prop-
1
0 that enables compact form factor, high repetition rates erties have made DKS states the preferred operational
7 and broadband optical frequency combs, that are even lownoisestates[7,13,14]ofKerrfrequencycombgener-
1 amenable to wafer-scale integration with additional elec- ationwithagrowingnumberofapplicationshavingbeen
: trical or optical functionality. The spectral bandwidth demonstrated, including e.g. terabit coherent commu-
v
i of Kerr frequency combs, being independent of a spe- nications [15], dual comb spectroscopy [16, 17] and low
X cific material gain and primarily determined by the res- noise microwave generation [18]. So far DKS formation
r onator’s group velocity dispersion (GVD), has reached has been observed in a variety of resonator platforms
a octave span shortly after the principle’s first demonstra- [8, 18–22] among which planar silicon nitride (Si N )
3 4
tion [5, 6], although in the high noise operation regime waveguide resonators have gained significant attention.
[7]. Only recently, the observation of dissipative Kerr Allowing CMOS-compatible, wafer-scale fabrication and
soliton (DKS) formation in microresonators has enabled exhibiting low linear and nonlinear optical losses in the
thecontrolledexcitationoffullycoherentKerrfrequency telecom wavelength region, Si N microresonators bear
3 4
combs [8]. realisticpotentialforapplications[23]. Theaccuratecon-
When operated in the single soliton state, DKS fre- trol of waveguide dimensions during microfabrication is
prerequisite to precise GVD engineering and thus tailor-
ingoftheKerrfrequencycombbandwidth[10]. Dualdis-
persive wave emission extending the spectral bandwidth
∗ tobias.kippenberg@epfl.ch
2
(a) 10 μm (b) intracavity position (rad) photonic Damascene process, is important. Moreover,
0 � 2� the bus waveguide cross section needs to be engineered
u.) for high ideality coupling [26]. The fundamental mode
er (a. 2 families of the microresonator devices used in this work
22.7 μm w 10 have a typical internal linewidth of κ /2π = 100MHz
o 0
FSR = 1 THz y p and thus a loaded finesse of F ≈104 at critical coupling.
avit 100 While a high microresonator finesse is beneficial to
ac lower the power threshold of Kerr comb generation, the
ntr microresonator GVD determines mostly the power re-
i
−0.5 0 0.5 quirements to reach a certain spectral bandwidth. Mi-
time (ps) croresonator GVD is conveniently expressed in its inte-
(c)
wavelength (nm) grated form around a central pump frequency ω :
0
20001800 1600 1400 1200 1000
10 60
D µ2 D µ3
m) octave span 40 D Dint(µ)≡ωµ−(ω0−D1µ)= 22! + 33! +...
B 0 in
power (d−10 20t/2 (GHzπ Dthinmt(oµd)eisωµthreelraetsivoenatnoctehefrecqeunternaclypduemvpiatmioondeofω0thferoµm-
0 ) theequidistantgriddefinedbytherepetitionrateD1/2π.
−20 DKSformationrequiresanomalousGVD(D >0)asthe
150 200 250 300 associatedresonancefrequencydeviationisc2ompensated
frequency (THz)
by the nonlinear phase shift induced by the soliton. In
Figure 1. Octave-spanning dissipative Kerr soliton frequency thecaseofdominantquadraticGVD,the3dBbandwidth
comb generation in a dispersion engineered Si N microres- of the resulting characteristic sech2 soliton spectral en-
3 4
onator. (a) Scanning electron microscope image of a Si N velope, and thus the temporal pulse width, is a function
3 4
microresonatorwith1THzfreespectralrangeduringfabrica- ofthecavitydetuningandD [8]. Lowpositivevaluesof
2
tion. The waveguide structures are highlighted in blue. (b) D result in larger optical bandwidths for a given pump
Simulated temporal intracavity power profile of an octave- 2
power and detuning. Beyond this, dispersive wave for-
spanning DKS frequency comb excited with 75mW pump
mation offers an effective way to coherently enlarge the
power for κ/2π = 200MHz using the Lugiato-Lefever equa-
bandwidth into the normal GVD with locally enhanced
tion. The two dispersive waves form an overlapping trailing
comb tooth power in the spectral ends [19, 27]. Dis-
and leading wave pattern. (c) The according spectral intra-
persive waves are emitted in resonances which are phase
cavitypowerdistribution(blue)andtheintegrateddispersion
profile D /2π (red) engineered for dispersive wave emission matchedbythesolitoninducednonlinearphaseshiftand
int
around 1µm and 2µm wavelengths. the spectral positions of DWs are thus approximately
given by the linear phase matching criterion D = 0.
int
For dominant, negative fourth-order dispersion D two
4
to both sides is attractive for low power octave-spanning DW positions can be engineered, causing dual dispersive
DKSgenerationbutparticularlychallengingtorealizeas wave emission.
it requires control of higher order GVD. Most microres- We design the microresonator waveguide cross section
onators used for DKS generation to date, did not have based on finite element (FEM) simulations of the res-
specificallyengineeredGVD,resultinginlimitedspectral onator’s TE mode family dispersion. Dispersive wave
00
bandwidth, far below one octave. formation is engineered to occur around wavelengths of
Here we present Si N microresonators with a free 1µmand2µmspanninganoctavebandwidtharoundthe
3 4
spectral range (FSR) of 1THz which allow for DKS fre- pump line at 1.3µm or 1.55µm. Full 3D finite difference
quency comb generation with bandwidths exceeding one time domain simulations are employed to engineer the
octave. The wide FSR reduces the total power require- bus waveguide cross section for high ideality coupling to
ments to cover one octave bandwidth and benefits the theTE modefamily[26]. Next,wesimulatetheoctave-
00
nonlinear conversion efficiency in the DKS state [24]. spanningDKSgenerationusingaLugiato-Lefevermodel
Such microresonators are thus also of interest for wave- [9, 28] for a critically coupled Si N microresonator with
3 4
length regions where no high power pump sources exist. linewidth κ/2π = 200MHz driven by a pump laser of
We use the recently developed photonic Damascene pro- 75mW power. Figure 1(b), (c) show the intracavity field
cess [25] for wafer-scale fabrication of such high-Q mi- and the spectral envelope of the simulated single DKS
croresonator devices with high yield. Figure 1(a) shows state. Ascanbeseenthetwodispersivewavesarepresent
a scanning electron microscope (SEM) picture of the mi- inthetemporalintracavityfieldastrailingandfollowing
croresonator device including a straight bus waveguide. wave patterns which fill the complete cavity. Further-
For small microresonator radii (r ≈23µm) the coupling more, close examination reveals that the position of the
section forms a relatively large fraction of the total cir- higher frequency dispersive wave is slightly offset from
cumference and can cause parasitic loss. Thus, void-free the linear phase-matching point, indicating the approxi-
fabrication of narrow coupling gaps, as provided by the mative nature of the criterion D (µ)=0.
int
3
(a) (c) (e)
SiN
radial non-uniformity SiO2 10 μm
Si3N4 SiO2 Si
local deviation of removal rate
(b) local non-uniformity due to filler pattern
Deflection (µm) −−−−864200000 a f t aearftf etberar d cpekopsloiisdsheitii norgenmoval (d) (f) (g)
−100
20 40 60 80 SiN
position on wafer (mm) SiO2 200 nm
Figure2. Originsofnon-uniformplanarizationinthephotonicDamasceneprocess. (a)Schematicrepresentationofwaferbow
and local loading causing non-uniformity during chemical mechanical planarization. (b) Measurement of wafer bow evolution
during planarization for a 525µm thick 4” wafer with ∼ 1µm thick Si N thin film. (c) Photograph of a 4” wafer after
3 4
planarization showing non-uniformity through colored interference patterns. (d) Optical microscope image of a 1THz FSR
microresonator surrounded by a “checkerboard” filler pattern. The Si N waveguides are highlighted in blue. (e) SEM image
3 4
of a non-uniformly planarized bus waveguide (indicated in blue). The neighboring filler pattern patches have different loading
andthuscausemoreSi N (darkshadedareas)toremainoverthebuswaveguideincertainareas. (f)Crosssectionofthebus
3 4
waveguideusedtocouplelightina1THzFSRmicroresonator. Thedimensions(0.5µm×0.67µm)arechosentoprovidehigh
idealitycouplingtothemicroresonator’sfundamentalTE modefamily. (g)Opticalmicroscopeimageofwaveguidestructures
00
surrounded by optimized filler pattern.
In the following, we first discuss the challenges of reported photonic Damascene process, employed in the
precise waveguide dimension control using the photonic present work, solves these problems by depositing the
Damascene process. Leveraging optimized fabrication Si N thin film on a prestructured substrate and sub-
3 4
procedures, we achieve control over the dispersive wave sequent removal of the excess material using chemical
positions resulting in octave-spanning comb generation mechanical planarization (CMP). A dense filler pattern
based on 1.3µm and 1.55µm pump lasers. The coher- surrounding the waveguide structures effectively lowers
ence of the generated Kerr frequency combs is stud- the thin film’s tensile stress and enables crack free fabri-
ied and a response measurement is used to unambigu- cation even for micron thick waveguides. Moreover, the
ously identify DKS formation. Finally, we study octave- filler pattern helps to control and unify the material re-
spanning single DKS formation for 1.3µm and 1.55µm moval rate across the wafer during CMP.
pumpwavelengthsfeaturingthebroadestKerrsolitonfre- Although the level of control on the waveguide width
quency comb generated to date, exceeding 200THz in is similar for Damascene and subtractive processing
bandwidth. schemes(bothrelyingontheprecisionoflithographyand
etching processes) the waveguide height control is sub-
stantially eased in the subtractive fabrication scheme: a
II. PRECISION DISPERSION ENGINEERING typical thin film deposition non-uniformity of 3% across
USING THE PHOTONIC DAMASCENE a 4” wafer results in a maximal 25nm height deviation
PROCESS for 800nm thick Si N waveguides. For the Damascene
3 4
processthevariationofthefinalwaveguidethicknessde-
Dispersion engineering for octave-spanning Kerr fre- pends on the uniformity of the dry etch process used to
quency comb generation requires stringent waveguide structure the waveguide trenches and most importantly
width and height control on the order of 10nm, that is on the uniformity of the CMP process. While optical in-
challenging to meet. As detailed below via simulations terferometrycangivepreciselocalinformationonthere-
(showninFigure3),especiallythepositionofthedisper- movalrateduringCMP,especiallythewaferbowandthe
sive wave features is very sensitive to variations of the local loading lead to height variations (see Figure 2(a)).
waveguide dimensions. Although DKS frequency comb This reduces the yield of devices with dimensions within
generators have been fabricated using conventional sub- the tolerance range for octave-spanning Kerr comb gen-
tractive processing, several challenges arise for the fab- eration,andcausesuncontrolledspectralpositionsofthe
rication of high confinement Si N waveguides for non- dispersive waves.
3 4
linear applications. In particular, these are cracking of Figure2(b)showstheevolutionofthebowduringpla-
the highly stressed Si N thin film and void formation narization of a 525µm thick 4” wafer measured using a
3 4
at the microresonator coupling gap [25]. The recently thin film stress measurement tool. After deposition, the
4
continuousSi N filmonthewaferbacksidecausesaposi-
3 4 (a) wavelength (nm)
tivedeflectionasitstensilestressishigherthanthestress
20001800 1600 1400 1200 1000
on the wafer front side which is reduced by the prestruc-
m h = 0.73µm, w = 1.425µm
ttuortehdesCuMrfaPcei.nvTerhtesrtehmeobvoawl otfotmheorbeatchkasnid−e 9S0i3µNm4.pTrihoer GHz) 40 = 1.3 µ hh == 00..7745µµmm,, ww == 11..442255µµmm
sounbtsheequfreonnttpsliadneaarnizdatcioonntisntueopurselymroevleasxetshetheexwceasfserSbi3oNw4. πD/2 (int 20 λ pump h = 0.74µm, w = 1.375µm
This bow variation leads to non-uniformity with radial 0
symmetry visible as the colored interference ring pat-
(b)
tern in Figure 2(c). Furthermore, local non-uniformity w=1.425 µm
0
in the central wafer region can also be observed, which
originates from different removal rates due to loading −20
variations. Loading refers here to the amount of excess −40
Si N that is in direct contact with the polishing pad
3 4 −60
and depends on the local structure density. Such local
variations can also occur on much smaller scales. This w=1.375 µm
is exemplified in Figure 2(e) for an area around a bus dB) 0
waveguide (indicated in blue) after CMP. Depending on er ( −20
w
theloadingoftheneighboringfillerpattern, moreorless o
p −40
dSii3oNxi4de(d(aSrikOer)sphraedfeodrmar(elaig)histerremshaaidneindgaarbeao)v.eStuhcehsilloicaodn- rel. −60
2
ingdependentnon-uniformityleadstolocalvariationsof
w=1.325 µm
the waveguide height, and therefore needs to be mini- 0
mized. Finally, during landing when most excess Si3N4 −20
is removed and mostly the SiO preform is in contact
2 −40
with the polishing pad, the material removal rate can
drastically change. Thus, the polishing endpoint is hard −60
to predict and the overall mean waveguide height differs
150 200 250 300 350
from the target value. frequency (THz)
Here we apply an optimized CMP process that uses
Figure 3. Wafer-scale dispersion engineering of octave-
higher pressure to equalize the wafer bow and a thicker
spanning Kerr frequency combs. (a) Simulated integrated
wafersubstrate(700µm)thatreducestheamountofini- dispersion D (ν)/2π of fully cladded Si N microresonators
tial wafer bow to −50µm. The removal rates for Si3N4 with 82° sideinwtall angle and different wid3th4s and heights for
and SiO are chosen to be similar, limiting the uncer- generating dual dispersive wave DKS states. For a pump
2
taintyduringlanding. Inordertoreducelocalloadingef- wavelengthof1.3µmdispersivewavescanbeexcitedinalin-
fects,anovelfillerpattern(seeFigure2(g))isusedwhich earapproximationatpositionsclosetothezero-pointsofthe
homogenizes the loading while providing sufficient stress integrated dispersion. (b) Unstable modulation instability,
release for crack-free fabrication. The dispersion control high-noise Kerr frequency combs generated in three different
samples fabricated on the same wafer at different positions
achieved using this optimized CMP process is summa-
which are indicated in the inset. The spectral position of
rizedinFigure3. BasedonFEMsimulationsofmicrores-
the two dispersive waves are indicated by gray arrows. The
onator GVD shown in Figure 3(a), the target waveguide
samples have by design a difference of 50nm in waveguide
height of 0.74µm was chosen to enable octave-spanning width which enables tuning of the high frequency dispersive
DKS frequency comb generation in the resonator’s TE
00 waveposition,demonstratingtheexcellentdimensioncontrol
mode family for a 1.3µm pump laser. The simulation of the Damascene process.
predictsdispersivewaveformationaround1µmand2µm
for a 1.425µm wide, fully SiO cladded waveguide with
2
82° sidewall angle. As shown in Figure 3(a), a waveg-
allow to measure higher order dispersion terms. Figure
uide height deviation of only 10nm causes a shift of the
4(a) shows the setup used for DKS frequency comb gen-
low frequency dispersive wave position by 5THz (i.e. ca.
eration. A 1.3µm external cavity diode laser (ECDL) is
∆λ=100nminwavelength). Achangeinthewaveguide amplifiedusingataperedsemiconductoramplifier(SOA)
width does not strongly influence the position of the low
to a maximum power of 0.75W. Laser light is coupled
frequencydispersivewavebutchangesthepositionofthe
into the photonic chip using lensed fibers and inverse
high frequency dispersive wave.
tapered waveguides with about ∼ 3dB loss per facet.
The dispersion of different samples is evaluated based During the laser tuning into the TE mode family res-
00
on the spectral envelopes of the generated frequency onance, the transmission through the device as well as
combs. The low values of microresonator GVD (i.e. thegeneratedcomblightarerecordedusingphotodiodes
D /2π ∼O(20MHz) ) and the bandwidth limitations of (f = 1GHz). Up to three optical spectrum analyz-
2 3dB
ourprecisiondiodelaserspectroscopysetup[29], didnot ers (OSA) are used to capture the generated broadband
5
comb spectra. Figure 3(b) shows octave spanning un- and avoided modal crossings [31], DKS state identifica-
stable modulation instability (uMI) comb states excited tion purely based on the spectral envelope is unreliable.
in three different microresonator chips. The sample po- In contrast, a recently introduced response measure-
sition on the 4” wafer is provided as inset. As can be ment technique [32] allows to unambiguously identify
seen, the position variation of the low frequency disper- DKS comb states, and is applied here for the first time
sive wave for the three samples is 5.5THz around to the to octave spanning soliton states. Figure 4(a) shows the
designed target value of 150THz. In comparison with setup used to investigate the coherence of comb states
thesimulationsinFigure3(a)thisindicatesavariationof generatedwitha1.3µmand1.55µmECDL.The1.55µm
less than ±10nm in waveguide height around the target ECDL is amplified using an erbium-doped fiber ampli-
value of 0.74µm. Moreover, the width variation allows fier (EDFA) and in order to perform the response mea-
the position of the high frequency dispersive wave to be surement, a phase modulator with 10GHz bandwidth
tuned. A waveguide width reduction of 50nm in the de- is added before the amplifier. A vector network ana-
sign moves the position of the high frequency dispersive lyzer (VNA), probing the Kerr comb state’s response,
wave by ∼ 25THz in good agreement with simulations. drives the EOM and receives the signal via a photodiode
These measurements reveal actual lithographic control with 25GHz bandwidth. We note that this measure-
overthedispersivewaveposition, ratherthanpostfabri- ment was only possible using the 1.55µm pump laser
cation selection. The photonic Damascene process with due to the lack of an appropriate phase modulator for
optimized planarization step thus offers sufficient preci- the 1.3µm pump path. A heterodyne beat note of the
sion in the control over waveguide dimensions to allow comb with a fiber laser at 1064nm is recorded using an
for wafer-scale fabrication of microresonators with engi- electrical spectrum analyzer (ESA). All presented Kerr
neered dispersion properties over a full octave of band- frequency comb states are excited using a simple laser
width. tuning method, applying a linear voltage ramp provided
byanarbitraryfunctiongenerator(AFG)tocontrollably
tune into a comb state [8, 32].
III. COHERENCE AND SOLITON DKS formation is accompanied by characteristic step
IDENTIFICATION featuresthatoccursimultaneouslyinthegeneratedcomb
and transmitted light trace [8]. However, in the exper-
For applications of the octave-spanning Kerr comb iments reported here in Figure 4(b)-(d), we found that
state coherence is a key requirement. Common meth- abrupt changes in the transmitted power do not neces-
ods to classify the coherence of Kerr comb states are the sarily originate from soliton formation. Instead, abrupt
measurements of the low frequency amplitude noise of changes between different non-solitonic comb states, as
the generated comb light, the repetition rate beat-note, well as resonance splitting due to waveguide surface
aswellasheterodynebeat-noteswithreferencelasers[7]. roughness or near-by resonances of other mode fami-
Amongthedifferentlownoisecombstates[7,13,14],the lies can cause similar features. Figure 4(b) shows the
DKS state is particularly attractive as it provides high optical spectrum of a comb state generated by tuning
coherenceacrossitsentirebandwidth. SolitonKerrcomb the 1.3µm pump laser with P = 125mW in the
bus
states have in particular been identified via the charac- bus waveguide into the step feature visible in Figure
teristic transmission steps in the generated comb light, 4(d). The spectrum spans an octave and is highly struc-
via the red-detuned nature of the pump laser, as well as tured, featuring individual sharp lines at the dispersive
thecharacteristicsolitonspectralenvelope. Althoughthe wave positions. Moreover, the beat-note of an individ-
spectra contain no direct information on coherence, for ual comb tooth with a fiber laser at 1064nm (see Fig-
microresonators with dominant (positive) quadratic dis- ure 4(e)) shows low noise characteristics i.e. > 20dB
persion D the single soliton state exhibits a character- signal-to-noise ratio in a 100kHz resolution bandwidth
2
istic spectral sech2 envelope. For broadband Kerr combs (RBW). Based on these measurements, the generated
the spectral envelope can exhibit dispersive waves, due comb state would agree with properties associated with
tohigherordermicroresonatordispersion. Althoughdis- a multi-soliton state. However, close examination of the
persive wave features are also present in the uMI state generated spectral lines, shown in Figure 4(c), reveals
spectra, in the DKS state their width reduces with a si- that certain comb lines have a splitting of ∼ 8GHz, in-
multaneous increase in the individual comb tooth power compatible with a DKS comb state and associated with
and a shift of the dispersive wave maximum [19]. Other merging of individual subcombs [7]. Importantly, com-
spectral signatures of the DKS regime include the soli- monly employed [7, 33] low frequency amplitude noise
ton Raman self-frequency shift, that leads to a redshift measurements,andevenlocalheterodynebeat-notemea-
of the soliton’s sech2 envelope with respect to the pump surements,areinsufficientdiscriminators,unlesstheyare
[30]. TheRamanself-frequencyshiftisabsentintheuMI performedoversufficientlywidebandwidthtodetectthe
comb state, but can be masked in the DKS state by the subcomb spacing (here (cid:29) 1GHz). Indeed, for the mea-
soliton recoil due to dispersive wave formation. For oc- surements above, no noise < 1GHz is observed, which
tavespanningsolitonKerrcombswithTHzmodespacing would lead to an erroneous soliton state identification.
anddispersivewaveemissionduetomicroresonatorGVD A more reliable method is using the response of the
6
(a) FBG
VNA
PD OSC
EDFA ESA
1550nm ECDL Fiber laser
EOM DUT
FPC
wavelength (nm)
AFG
(f)
21001900 1700 1500 1300 1100
P = 175 mW
(i) no soliton state bus
SOA OSA 0
1300nm ECDL
−20
−40
(b) wavelength (nm)
−60
21001900 1700 1500 1300 1100
20 P = 350 mW
B) 0 no soliton state Pbus= 125 mW B) 0 (ii) no soliton state bus
d d
er ( −20 1064 nm er ( −20
w w
po −40 po −40
rel. −60 rel. −60
P = 320 mW
140 160 180 200 220 240 260 280 300 320 (iii) multi-soliton state bus
0
(c) frequency (THz) 160 THz
180.7 180.75 210.7 210.75 255.7 255.75 −20
m)
B −40 ~8 GHz −40 ~8 GHz −40 ~8 GHz −40
d
er ( −60 −60 −60 −60
w
po −80 −80 −80 140 160 180 200 220 240 260 280 300 320
frequency (THz)
(d) (e) (g)
−60 4
arbitrary units −022 tracnotsnumvneiitnrttege dvd ol illgtigahhgtte power (dBm)−−−−109870000 rbw = 100 kHz lin. mag. (a.u.)0123 (i) (ii) (iii)
0.032 0.034 0.036 0.038 0.04 −50 0 50 0 2 4 6 8 2 4 6 8 2 4 6 8
time (s) f − 0.95GHz (MHz) frequency (GHz)
Figure 4. Distinction of octave-spanning multi-soliton states and non-solitonic states. (a) Schematic of the setup used to
generate octave-spanning Kerr combs using 1.3µm or 1.55µm pump lasers. In both cases an arbitrary waveform generator
(AFG) provides the voltage ramp to tune the external cavity diode seed laser (ECDL) into a resonance of the device under
test (DUT). Either an erbium-doped fiber amplifier (EDFA) or a semiconductor tapered amplifier (SOA) is used to amplify
the seed laser before coupling onto the photonic chip using lensed fibers. An oscilloscope (OSC) records the transmitted and
converted comb light power during the scan. Several optical spectrum analyzers (OSA) are used to capture the full spectra
of the excited comb state. A 1064nm fiber laser allows the recording of heterodyne beat-notes using a fast photodiode and
an electrical spectrum analyzer (ESA). The comb state response is measured by a vector network analyzer (VNA) that drives
an electro-optic phase modulator (EOM) and receives part of the transmitted light on a photodiode with 25GHz bandwidth.
(b) Highly structured octave-spanning comb state excited by tuning the 1.3µm pump laser into the step-like feature (which,
importantly, does not originate from soliton formation) highlighted in (d) with gray arrows. The red boxes mark comb teeth
which reveal a splitting of ∼8GHz upon close examination in (c), demonstrating subcomb formation. (d) Oscilloscope trace
of the voltage ramp (blue), the generated comb light (green) and the transmitted light signal (red) (e) Narrow heterodyne
beat note of a 1064nm fiber laser with the comb tooth marked with a green arrow in (b) of a non-solitonic comb state. (f)
Threecombstatesexcitedinthreedifferentmicroresonatorswiththesamewaveguidedimensions. (g)Responsemeasurements
associated with the comb states in (f). The position of the cavity (C-res.) and the soliton (S-res.) resonance is indicated.
comb state to a phase modulation of the pump laser. important quantity determining the soliton properties.
It was shown that the phase modulation response func- Figure 4(f) shows three comb states recorded from three
tion of DKS states consists of two characteristic reso- microresonators on the same photonic chip having the
nancesoriginatingfromthecirculatingsolitonpulses(S- same waveguide dimensions but varying bus-resonator
resonance)andthecavityresponse(C-resonance)[32,34]. distances. Again, the microresonator dispersion was de-
Furthermore, the measurement allows to infer the ef- signed for dispersive wave emission around 1µm and
fective pump laser detuning within the DKS state, an 2µm, here upon pumping with a 1.55µm laser. We ex-
7
cite different comb states and analyze their correspond-
(a) wavelength (nm)
ing response functions shown in Figure 4(g). State (i) is
2200 2000 1800 1600 1400 1200
a uMI state and the response function exhibits a single 20
pcloieepasek.aSantndadtaea(sienie)omhna-inzsegarloymsbuhacachrkpmgrdooirusepnsedtrrsnuivoceitsuewreaadvteshpfiegeachtturfarreelqecunoevmne--- ower (dB)−200 PPbbuuss== 243555 mmWW ~FsWeHchM2 = 24.85 THz lin. magnitude (a.u.)0240 2 4 6 8
p frequency (GHz)
pared to state (i). However, the response measurement el. −40
revealsafundamentallydifferentresponsethanexpected r−60
for a DKS state. Also, state (iii) has a strongly struc-
140 160 180 200 220 240 260 280
tured spectral envelope with similarities to the envelope
frequency (THz)
of state (ii) but its response function shows two overlap-
(b) wavelength (nm)
ping resonances. The lower frequency S-resonance orig-
20001800 1600 1400 1200 1000 800
inates from the circulating soliton pulses and has a re- 20
dhuigcheedrafmrepqluiteundcye,cwomhopsaerefdreqtouetnhceycianvditiycaCte-sretshoneadnecteuant- er (dB)−200 Pbu2s0=0 2T4H5z mW ~FsWeHchM2 = 19.9 THz
ing. Moreover, the separation of both peaks diminishes w
o
urepsopnecltaisveerlyb.luWeedtehtuusniidnegnatinfdysintactreea(siieis)afosrarmeduldtei-tsuonliitnogn, rel. p−−6400
state. The multi-soliton state leads to an increased con-
version efficiency compared to the single soliton case, re- 150 200 250 300 350
frequency (THz)
sulting in negligible residual pump power in the present
case. Moreover,weobserveanotableshiftofthehighfre-
Figure 5. Octave-spanning single soliton generation at
quency dispersive wave peak position (indicated by gray 1.55µm and 1.3µm pump wavelength. (a) The same sin-
arrows). gleDKSstateshownfortwodifferentpumplaserpowersand
cavity detunings. Inset: Response measurement of the single
DKS states shown in (a) revealing the respective pump laser
cavity detuning. The S- and C-resonances are separated by
IV. OCTAVE SPANNING SPECTRA VIA ∼ 2.5GHz and ∼ 5GHz, respectively. (b) Single DKS span-
SINGLE SOLITON GENERATION ning 200THz in optical bandwidth with a dispersive wave at
850nmexcitedusinga1.3µmpumplaser. Fitsofbothstates
Once excited, a multi-soliton state can often be con- with a spectral sech2 envelope are shown in red. Pump pow-
verted into a single soliton state by soliton switching ers in the bus waveguide and fitted spectral bandwidths are
upon “backward tuning” of the pump laser [32]. Figure noted.
5 shows two examples of octave-spanning single soliton
generation obtained via this technique using 1.3µm or
to date, to the best of our knowledge. The DKS comb
1.55µm pump lasers. We note that not all multi-soliton
states excited were observed to switch to a lower soliton spans a total bandwidth of ∼ 200THz using a pump
number upon backward tuning, but some changed into laser at 1.3µm providing only 245mW power in the bus
non-solitonic states. Moreover, it is observed that both waveguide. The fitted 3dB bandwidth of 19.9THz is
similar to the value of the single DKS state shown in
single soliton combs are excited in direct vicinity of an
Figure 5(a) even though the required pump power is sig-
avoided modal crossing causing strong local deviations
nificantly lower. This demonstrates the strong influence
from the characteristic sech2 shaped spectral envelope
of dispersion on the power requirements for broadband
[35].
comb generation and underlines the necessity for precise
The soliton spectrum shown in Figure 5(a) features
dispersion control.
nodispersivewavesindicatingadominantquadraticfac-
tor of the anomalous GVD. The response measurement
shown in the inset shows a double resonance signature
and reveals a cavity detuning of ∼ 4.5GHz. The pump V. CONCLUSION
powerisincreasedto455mWinthebuswaveguidewhile
maintaining the single DKS state, which allows for a In conclusion, we have demonstrated DKS generation
larger soliton existence range and octave bandwidth at in Si N microresonators with 1THz FSR and spec-
3 4
a cavity detuning of ∼ 7GHz. This cavity detuning tral bandwidths exceeding one octave, both at 1.55µm
is significantly higher than previously published values and 1.3µm pump wavelength. The photonic Dama-
for Si N microresonators indicating a strong nonlinear scene process with optimized planarization step allows
3 4
phase shift due to the high peak intensity of the soli- for high yield, wafer-scale fabrication of microresonators
ton pulse [32]. The spectral envelope is fitted using a with precisely controlled dispersive wave positions. Our
sech2 shape and a 3dB bandwidth of 24.85THz is ex- work moreover revealed that for comb states generated
tracted, corresponding to a 12.7fs pulse. In Figure 5(b) in Si N microresonators with 1THz FSR, conventional
3 4
wepresentthebroadest singleDKScombstatepublished criteria of coherence can fail. We carefully investi-
8
gatedthecoherenceofgeneratedcombstatesandunam- This publication was supported by Contract HR0011-
biguously identified octave-spanning multi-DKS states 15-C-0055fromtheDefenseAdvancedResearchProjects
based on their unique phase modulation response signa- Agency (DARPA), Defense Sciences Office (DSO), fund-
ture. Finally, we demonstrated single soliton generation ingfromAirForceOfficeofScientificResearch,AirForce
with record bandwidth of 200THz generated with only Material Command, USAF under Award No. FA9550-
245mW of pump power. 15-1-0099, the European Union’s Horizon 2020 research
Our findings demonstrate the technological readiness and innovation programme under MIRCOMB: Marie
ofthephotonicDamasceneprocessandintegratedSi N Sklodowska-Curie IF grant agreement No. 709249 and
3 4
frequency comb generators for a wide range of appli- theSwissNationalScienceFoundationundergrantagree-
cations. The ability to generate ultra-short pulses also ment No.161573 and 163864. M.K. acknowledges the
with a 1.3µm pump laser, as demonstrated here, opens support from the Marie Curie Initial Training Network
up applications in biological and medical imaging as this FACT.
wavelength represents a compromise between low tissue
scattering and absorption due to water. Moreover, such
DKS frequency combs with designed spectral envelopes
may in the future enable chip integration and miniatur-
ization of self-referenced frequency combs [36] and dual
comb CARS [37].
This research was carried out concurrently with the
Acknowledgements workfromK.S.atNISTintheframeworkoftheproject
DODOS that is available online under [33] and also re-
Si N microresonator samples were fabricated in the ports octave spanning Kerr soliton combs with dual dis-
3 4
Center of MicroNanoTechnology (CMi) at EPFL. persive waves.
[1] T.Udem,R.Holzwarth, andT.W.Hänsch,Nature416, [13] X.Xue,Y.Xuan,Y.Liu,P.-H.Wang,S.Chen,J.Wang,
233 (2002). D.E.Leaird,M.Qi, andA.M.Weiner,NaturePhotonics
[2] R. Holzwart, T. Udem, T. W. Hänsch, J. Knight, W. J. 9, 594 (2015).
Wadsworth, andP.Russel,PhysicalReviewLetters85, [14] K. Saha, Y. Okawachi, B. Shim, J. S. Levy, R. Salem,
2264 (2000). A. R. Johnson, M. A. Foster, M. R. E. Lamont, M. Lip-
[3] T. Steinmetz, T. Wilken, C. Araujo-Hauck, son, and A. L. Gaeta, Optics Express 21, 1335 (2013).
R.Holzwarth,T.W.Hänsch,L.Pasquini,A.Manescau, [15] P. Marin-Palomo, J. N. Kemal, M. Karpov, A. Ko-
S.D’Odorico,M.T.Murphy,T.Kentischer,W.Schmidt, rdts, J. Pfeifle, M. H. P. Pfeiffer, P. Trocha, S. Wolf,
andT.Udem,Science321,1335(2008),arXiv:0809.1663. V.Brasch,R.Rosenberger,K.Vijayan,W.Freude,T.J.
[4] P. Del’Haye, A. Schliesser, O. Arcizet, T. Wilken, Kippenberg, andC.Koos,arXiv1610.01484,13(2016),
R. Holzwarth, and T. J. Kippenberg, Nature 450, 1214 arXiv:1610.01484.
(2007). [16] A. Dutt, C. Joshi, X. Ji, J. Cardenas, Y. Okawachi,
[5] P. Del’Haye, T. Herr, E. Gavartin, M. L. Gorodetsky, K. Luke, A. L. Gaeta, and M. Lipson, arXiv
R. Holzwarth, and T. J. Kippenberg, Physical Review 1611.07673, 1 (2016), arXiv:1611.07673.
Letters 107, 063901 (2011). [17] M.-G.Suh, Q.Yang, K.Yang,X.Yi, andK.J.Vahala,
[6] Y.Okawachi,K.Saha,J.S.Levy,Y.H.Wen,M.Lipson, Science 354, 600 (2016).
and A. L. Gaeta, Optics Letters 36, 3398 (2011). [18] W.Liang,D.Eliyahu,V.S.Ilchenko,A.A.Savchenkov,
[7] T. Herr, K. Hartinger, J. Riemensberger, C. Y. Wang, A. B. Matsko, D. Seidel, and L. Maleki, Nature Com-
E.Gavartin,R.Holzwarth,M.L.Gorodetsky, andT.J. munications 6, 7957 (2015).
Kippenberg, Nature Photonics 6, 480 (2012). [19] V. Brasch, M. Geiselmann, T. Herr, G. Lihachev,
[8] T. Herr, V. Brasch, J. D. Jost, C. Y. Wang, N. M. Kon- M. H. P. Pfeiffer, M. L. Gorodetsky, and T. J. Kip-
dratiev,M.L.Gorodetsky, andT.J.Kippenberg,Nature penberg, Science 351, 357 (2016).
Photonics 8, 145 (2014), arXiv:1211.0733. [20] C.Joshi,J.K.Jang,K.Luke,X.Ji,S.A.Miller,A.Klen-
[9] S. Coen, H. G. Randle, T. Sylvestre, and M. Erkintalo, ner, Y. Okawachi, M. Lipson, and A. L. Gaeta, Optics
Optics Letters 38, 37 (2013), arXiv:1211.1697. Letters 41, 1 (2016), arXiv:1603.08017.
[10] Y.Okawachi,M.R.E.Lamont,K.Luke,D.O.Carvalho, [21] X. Yi, Q.-F. Yang, K. Y. Yang, M.-G. Suh, and K. J.
M. Yu, M. Lipson, and A. L. Gaeta, Optics Letters 39, Vahala, Optica 2, 1078 (2015).
3535 (2014), arXiv:arXiv:1307.1037. [22] P.-H. Wang, J. A. Jaramillo-Villegas, Y. Xuan, X. Xue,
[11] N. Akhmediev and M. Karlsson, Physical Review A 51, C. Bao, D. E. Leaird, M. Qi, and A. M. Weiner, Optics
2602 (1995). Express 24, 10890 (2016), arXiv:arXiv:1601.05036.
[12] V. Brasch, E. Lucas, J. D. Jost, M. Geiselmann, and [23] D.J.Moss,R.Morandotti,A.L.Gaeta, andM.Lipson,
T. J. Kippenberg, Light: Science & Applications 6, 1 Nature Photonics 7, 597 (2013).
(2017), arXiv:1605.02801. [24] C.Bao,L.Zhang,A.B.Matsko,Y.Yan,Z.Zhao,G.Xie,
9
A. M. Agarwal, L. C. Kimerling, J. Michel, L. Maleki, Pfeiffer, V. Brasch, G. Lihachev, V. E. Lobanov, M. L.
and A. E. Willner, Optics Letters 39, 6126 (2014). Gorodetsky, and T. J. Kippenberg, Nature Physics 13
[25] M. H. P. Pfeiffer, A. Kordts, V. Brasch, M. Zervas, (2016), 10.1038/nphys3893.
M.Geiselmann,J.D.Jost, andT.J.Kippenberg,Optica [33] Q.Li,T.C.Briles,D.A.Westly,T.E.Drake,J.R.Stone,
3, 20 (2016), arXiv:1511.05716. R.B.Ilic,S.A.Diddams,S.B.Papp, andK.Srinivasan,
[26] M.H.P.Pfeiffer,J.Liu,M.Geiselmann, andT.J.Kip- arXiv 1611.09229, 16 (2016), arXiv:1611.09229.
penberg,arXiv1608.06607,1(2016),arXiv:1608.06607. [34] E. Lucas, J. D. Jost, and T. J. Kippenberg, arXiv
[27] C. Milián and D. V. Skryabin, Optics Express 22, 3732 1609.02723, 16 (2016), arXiv:1609.02723.
(2014). [35] T. Herr, V. Brasch, J. D. Jost, I. Mirgorodskiy, G. Li-
[28] L. A. Lugiato and R. Lefever, Physical Review Letters hachev,M.L.Gorodetsky, andT.J.Kippenberg,Phys-
58, 2209 (1987). ical Review Letters 113, 1 (2014), arXiv:1311.1716.
[29] J.Liu,V.Brasch,M.H.P.Pfeiffer,A.Kordts,A.Kamel, [36] J.E.Bowers,A.Beling,D.J.Blumenthal,A.Bluestone,
H. Guo, M. Geiselmann, and T. J. Kippenberg, Optics S. M. Bowers, T. C. Briles, L. Chang, S. A. Diddams,
Letters 41, 3134 (2016). G. Fish, H. Guo, T. J. Kippenberg, T. Komljenovic,
[30] M.Karpov,H.Guo,A.Kordts,V.Brasch,M.H.P.Pfeif- E.Norberg,S.B.Papp,M.H.P.Pfeiffer,K.Srinivasan,
fer, M. Zervas, M. Geiselmann, and T. J. Kippenberg, L.Theogarajan,K.J.Vahala, andN.Volet,2016IEEE
Physical Review Letters 116, 103902 (2016). InternationalFrequencyControlSymposium,IFCS2016
[31] Q.-F.Yang,X.Yi,K.Y.Yang, andK.J.Vahala,Optica - Proceedings (2016), 10.1109/FCS.2016.7546782.
3, 1132 (2016), arXiv:1606.00954. [37] T. Ideguchi, S. Holzner, B. Bernhardt, G. Guelachvili,
[32] H. Guo, M. Karpov, E. Lucas, A. Kordts, M. H. P. N. Picqué, and T. W. Hänsch, Nature 502, 355 (2013),
arXiv:1302.2414.