Soliton Microcombs in Integrated Chalcogenide Microresonators

Photonic integrated microcombs have enabled advanced applications in optical communication, microwave synthesis, and optical metrology, which in nature unveil an optical dissipative soliton pattern under cavity-enhanced nonlinear processes. The most decisive factor of microcombs lies in the photonic material platforms, where materials with high nonlinearity and in capacity of high-quality chip integration are highly demanded. In this work, we present a home-developed chalcogenide glasses-Ge25Sb10S65 (GeSbS) for the nonlinear photonic integration and for the dissipative soliton microcomb generation. Compared with the current integrated nonlinear platforms, the GeSbS features wider transparency from the visible to 11 um region, stronger nonlinearity, and lower thermo-refractive coefficient, and is CMOS compatible in fabrication. In this platform, we achieve chip-integrated optical microresonators with a quality (Q) factor above 2 x 10^6, and carry out lithographically controlled dispersion engineering. In particular, we demonstrate that both a bright soliton-based microcomb and a dark-pulsed comb are generated in a single microresonator, in its separated fundamental polarized mode families under different dispersion regimes. The overall pumping power is on the ten-milliwatt level, determined by both the high Q-factor and the high material nonlinearity of the microresonator. Our results may contribute to the field of nonlinear photonics with an alternative material platform for highly compact and high-intensity nonlinear interactions, while on the application aspect, contribute to the development of soliton microcombs at low operation power, which is potentially required for monolithically integrated optical frequency combs.


Introduction
Integrated nonlinear photonics have combined nonlinear optics with state-of-the-art photonic integration and have unprecedentedly promoted light-matter interactions in terms of efficiency, bandwidth, and coherence [1][2][3][4][5][6] . They have enabled revolutionary techniques including chip-integrated optical frequency combs (OFCs) 5,7 , ultra-high bandwidth electrooptical modulation 2 , and chip-scale quantum optics 8 . Of particular interest is the dissipative soliton-based OFCs in photonic integrated microresonators, which can be generated at a low pump power 9 and have a broad bandwidth with fully coherent laser lines, benefiting from cavity-enhanced nonlinear efficiency and the lithographically controlled dispersion engingeering 7 . To date, they have enabled various advanced applications, including massive parallel optical telecommunication [10][11][12] , low-noise microwave synthesis 13 , parallel LiDAR 14,15 , photonic neuromorphic computing 16 , and other photonic functionalities 3,17,18 for chipscale time and frequency measurements.

. Crystalline-based platforms including AlN and
LiNbO3 introduce extra quadratic nonlinearities to generate the ultra-broadband soliton combs, which are potentially required for the self-referencing of microcombs 22,30 .
Particularly, most recently, monolithic microcomb chips have been developed based on the hybrid integration of such high-Q microresonators with III-V lasing chips, which are prone to be fully integrated in realistic systems 31,32 .
Yet, limitations are also realized in operating such platforms for soliton combs with low pump power, compact and flexibility in high volume microcomb-chip fabrication. For instance, additional processes in the fabrication of Si3N4 thick films for bright soliton microcombs are required, including crack-mitigation, chemical-mechanical polishing, and high temperature anneals (exceeding 1100 o C) strategies 33,34 . Crystalline platforms require the wafer-bonding process to be integrated on insulator substrates. In this way, the flexibility of performing dispersion engineering is reduced, and the yield of devices may be degraded as well 6,35 . Additionally, semiconductor resonators usually feature a high thermorefractive coefficient (TOC), which may hinder the existence of soliton combs 24,29 . As such, there exists a continuous motivation of seeking advanced material platforms, which could patch up the above issues and can be alternative to present platforms. In the meantime, while the current soliton microcombs are mostly designed and operated in the telecom band, it is equally essential for the material to open access in new wavelength regions for a broader range of microcomb applications, such as precise spectroscopy in the mid-infrared (MIR) 36 .
The material is Ge25Sb10S65 (GeSbS) that inherits the ultrabroad transmission window with absence of TPA, large refractive index and Kerr nonlinearity [37][38][39][40][41][42] of ChGs, and shows flexibility in photonic integration on silicon-based chips [43][44][45] . In the meantime, with modified compounds, GeSbS could overcome existing problems of As-based ChGs, and features an improved laser damage threshold (LDT) with a reduced TOC. We fabricate chipintegrated GeSbS microresonators with an intrinsic Q-factor above 2×10 6 , and demonstrate the generation of both the bright dissipative soliton microcombs and the dark pulse microcombs. In particular, the two microcomb regimes can be implemented in a single microresonator in the two fundamental polarized mode families that are lithographically engineered to induce normal and anomalous dispersive effects. Given a high Q-factor and high nonlinearity of the microresonator, the soliton comb is supported with a low pump power on the 10-mW level, which is prone to implementing monolithic frequency comb chips for integrated metrology applications, such as compact dual-comb spectrometers and MIR spectroscopy.

Results
Nonlinear GeSbS photonics for soliton microcomb generation.

Fabrication of high Q-factor GeSbS microresonators.
The lack of sufficiently high Q in ChG platforms has limited the ability to harness their dramatic material properties for nonlinear optics applications 9 . The surface roughness is critical for achieving high Q-factor photonic integrated devices 34  Based on the abovementioned processes, the top and sidewall surfaces of the microresonator could still be smooth after the resist removing, see Fig. 2f. As a result, a Qfactor of more than 10 6 was achieved, in contrast to the Q-factor of GeSbS microresonator with the same geometric dimensioning using the previously reported RIE-ICP etching recipe was only 10 5 level 47 . Moreover, the cross-section of the waveguide presents an almost vertical sidewall (see Fig. 2g), which is beneficial for accurate geometry dispersion control. GHz) spaced 1 st sidebands for or TE00 and TM00 modes, respectively.

Characterization of high Q-factor GeSbS microresonators.
Next, we systematically studied the dispersion, Q-factors, and OPO threshold powers based on GeSbS microresonators 48 . To further ensure accurate dispersion control of the waveguide, refractive index variation during the annealing process was also considered in device design, as shown in SI Fig. S1. Normal material GVD was exhibited in the telecom band (ca. -400 ps/nm/km at 1550 nm) by measuring the linear indices of GeSbS film, which had to tailor the waveguide dimensions to achieve strongly anomalous geometric GVD. In The intrinsic Q-factor of our microresonators is typically above 10 6 for both mode families, see Fig. 3b and 3g. In a number of microresonator samples on the same fabrication batch, all the resonances within the measurement range (1510-1630 nm) are characterized, which show a mean value of Q-factor of ca. 1.97×10 6 for the TE00 mode family and ca.
1.38×10 6 for TM00 mode family, see Fig. 3d and 3i. The pump threshold for OPO in such high-Q microresonators was also characterized by measuring the output powers of the first generated FWM sidebands with different input power, which for a pair of selected TE00 and

/ 22
TM00 modes are 1.3 mW and 3.5 mW, respectively. Moreover, in a 20 μm-radius microresonator, the measured threshold power can be as low as 0.78 ± 0.1 mW (see SI Fig.   S3 ), as it is scaled by the free spectral range (FSR) of the resonator: where n is the linear, n2 is the nonlinear refractive index, ν 0 is the pump frequency ν FSR is FSR of the resonator, Aeff is the effective mode area of the resonator, the coupling factor κ= κ ex κ i = Q i Q c ⁄ ⁄ , κex is the coupling rate, and κi is the intrinsic rate of the resonator, respectively. Given that n 2 has been experimentally measured to be 1.3×10 -18 m 2 /W (which is almost five times higher than that of Si3N4), we noticed that the calculated threshold power is in good agreement with the measured value, namely the estimated power is 0.72 mW (TE00, 20 μm-radius), 1.4 mW (TE00, 100 μm-radius) and 3.4 mW (TM00, 100 μmradius), respectively.

Generation of soliton microcombs in GeSbS microresonators
We measured the mode-locked soliton comb operations to demonstrate the advance of our GeSbS platform for microcomb formations. Because high thermo-optical instability is challenging for accessing soliton in resonators 35,50 , a lower TOC (dn/dT , where T is temperature) is beneficial. We measured the thermal-optic shift of the resonance frequency when heating the entire chip, see Fig. 4a. From the fitting of temperature-frequency data,

/ 22
we determined TOC utilizing the equation: Such a TOC value is comparable to that of Si3N4 (2.4×10 -5 K -1 ) and is around one order of magnitude lower than that of AlGaAs (3.6×10 -4 K -1 ) and GaN (ca. 10 -4 K -1 ). We applied the laser tuning method to form the soliton microcomb. The experimental setup for soliton generation and characterization is shown in SI Fig. S4. During the laser scan process, primary comb separated by 16-FSR (FSR = 197 GHz) and modulation instability (MI) combs were first observed when the pump light was located at the blue-detuned side of the resonance (see Fig. 4d, top and middle). A "soliton step" was observed when the laser was tuned to the red-detuned side, indicating the formation of the dissipative solitons in the cavity 51 , see Fig. 4b. By stopping the laser frequency on the step and slightly adjusting the detuning, a single-soliton-state microcomb was observed with a wavelength span from 1440 nm to 1680 nm (see Fig. 4d, bottom). Simultaneously, the drastic reduction of intensity noise of output light shows the transition to a soliton regime with low noise (see Fig. 4c).
The soliton duration can be estimated from its 3-dB bandwidth as 61 fs by fitting the final spectrum of a single soliton state with a sech 2 envelope. We also observed the Raman effectinduced soliton red spectral shift (ca. 1.2 THz in the present case) concerning the pump line (the green arrow in Fig. 4d) 52 . As a result, the lower TOC coefficient and low comb operation power persuade the stable soliton microcombs generations in our GeSbS microresonator without complex laser tuning schemes or auxiliary lasers 24 . Moreover, given that the TE00 mode family is in the normal dispersion regime, we further test the possibility of dark pulse-based comb generation in the GeSbS microresonator. Compared with the bright soliton comb in the anomalous dispersion regime, the dark-pulse comb could feature a higher conversion efficiency, while its generation usually requires a localized anomalous effect to initiate the sideband comb modes. By careful analysis of our microresonators, we notice that there exist several localized mode coupling positions in the mode spectrum of the resonator, as indicated by FEM mode 15 / 22 calculations, by the coupling between the fundamental TE00 mode (solid line) and highorder modes (dashed lines), see Fig. 5a. The mode coupling would lead to avoided mode crossings (AMXs) on the mode spectrum and introduce localized anomalous effect to the mode where the coupling occurs 1 . In this way, we may choose to pump an AMX mode for dark-pulse comb generation using the laser tuning scheme.
In experiment, the TE00 mode of the microresonator shows an overall normal GVD of D2/2π = -8.23 MHz, and the selected AMX mode is at 1543 nm. The comb generation was observed by manually tuning the laser from the blue side to the red side of the resonance when the pump laser was launched into the bus waveguide with 25 mW (see Fig. 5b). The comb power trace shows three steps during the laser tuning process, associated with the different states of the microcomb in normal dispersion region 53 . Figure 5c shows the evolution of the output optical spectra. At stage I, 3-FSR spaced comb lines were initially observed. Next to stage II, the bandwidth was increased with further tuning the pump frequency, and a flat wing at 1578 nm was formed. Finally, when the pump laser was stopped at stage III, we obtained the microcomb covering the range from 1510 nm to 1590 nm with a repetition rate of ca. 200 GHz. As a critical factor, the measure power conversion experimental result, see Fig. 5c and 5d. In addition, the duty cycle of the temporal darkpulse pattern is simulated to be around 27.4%, which is also close to the measured conversion efficiency 54 , see Fig. 5d. Moreover, the comb was assessed on the lowfrequency spectrum in the radio-frequency (RF) domain and showed a low-noise figure.

Conclusion
We have presented an integrated nonlinear photonics platform for soliton microcombs based on our home-developed GeSbS microresonators, which featured high Q-factor up to ca. 2.3×10 6 , higher nonlinearity, low TOC, and high LDT. The OPO pump threshold as low as 780 μW was attained. We achieved completely different dispersion curves by precise dispersion engineering, with TE00 being normal dispersion and TM00 being anomalous dispersion, respectively. Moreover, both a bright soliton-based microcomb and a darkpulsed comb were realized in a single microresonator, in its separated fundamental polarized mode families with the ten-milliwatt pump power level. Our results pave the way that our homemade GeSbS is a robust and compact integrated nonlinear platform and reveals potential system-level applications by heterogeneously full-integrated on silicon, such as MIR frequency combs, and optical frequency synthesizers.

Materials and Method
Chalcogenide film deposition. High  evaporation method was used to deposit the Ge20Sb10S65 film on Si substrates with a 3 μm SiO2 layer in a vacuum chamber at a base pressure of 7×10 -6 Pa. The substrates were mounted on a rotatable hold and pretreated using Ar plasma to improve the adhesion between the films and substrates. The evaporation rate was set to approximately 5-6 nm/min. Fig. 2a (YOKOGAWA AQ6370D). The second was detected by a photodetector, which was monitored by an electrical spectrum analyzer (ESA) (Agilent N9030A) for determining the intensity noise of the generated comb. The third part was received by a photodetector, which was monitored by the oscilloscope for recording the power trace of pump light.

Supporting Information
Supporting information is available from the corresponding author upon reasonable request.