Chip-Scale Soliton Laser Advances

• Chip-scale soliton lasers are integrated photonic devices that generate stable frequency combs via dissipative Kerr solitons or active cavity mechanisms.
• They leverage platforms like Si3N4, InP/Si, SiC, and TFLN with innovative processes to lower thresholds, enhance efficiency, and achieve electronic-rate repetition.
• Applications span coherent communications, spectroscopy, and LiDAR while addressing challenges in soliton stabilization and nonlinear dynamics control.

Searching arXiv for recent and foundational papers on chip-scale soliton lasers and related integrated soliton microcomb architectures. A chip-scale soliton laser is an integrated photonic source in which a soliton waveform—or the frequency comb generated by that waveform—is produced on chip, either by driving a high-$Q$ microresonator into the dissipative Kerr soliton regime or by forming dissipative solitons directly in an active semiconductor cavity. In the literature, the term spans laser-pumped $\mathrm{Si_3N_4}$ or SiC microresonators, heterogeneously integrated InP/Si/$\mathrm{Si_3N_4}$ systems, electro-optic–Kerr hybrids, and monolithic semiconductor ring lasers supporting Nozaki-Bekki or bright solitons. The common objective is the replacement of bulky bench-top comb sources by compact photonic engines that provide mutually coherent comb lines, electronically relevant repetition rates, or turnkey electrically driven ultrafast pulse generation (Xiang et al., 2021, Opačak et al., 2023, Hu et al., 16 Oct 2025).

1. Emergence of the field

An early milestone was the demonstration of a monolithic comb generating electronic-rate soliton pulses in a silica-on-silicon wedge microresonator. That platform operated with a 3 mm diameter resonator, an FSR of 22 GHz, intrinsic quality factors near $4 \times 10^8$, and soliton repetition tones around 21.92 GHz and 22.01 GHz, thereby placing dissipative Kerr solitons squarely inside the bandwidth of standard microwave instrumentation. The same work reported pump powers typically around 200 mW, a measured parametric-oscillation threshold slightly larger than 1 mW, pulse widths of about 130 fs from spectral fitting, and phase noise about −100 dBc/Hz at 10 kHz offset, establishing a silicon-compatible route toward electronically detectable soliton microcombs (Yi et al., 2015).

The subsequent maturation of $\mathrm{Si_3N_4}$ photonics reoriented the field toward planar integration, lower threshold, and broader spectral reach. Near-infrared operation at $1~\mu$m extended soliton Kerr combs to the biological imaging window, including octave-spanning spectra from 776 nm to 1630 nm and femtosecond pulse durations of about 27 fs and 17 fs in different states. In parallel, photonic Damascene reflow enabled ultralow-power single-soliton formation in 99 GHz resonators with only 9.8 mW input power and 6.2 mW in the waveguide, while preserving electronically detectable repetition rates compatible with coherent communication and ranging (Karpov et al., 2017, Liu et al., 2018).

The field then diversified in two directions. One direction pursued increasingly integrated pump architectures—packaged hybrid lasers, self-injection-locked DFBs, and heterogeneously integrated InP/Si/$\mathrm{Si_3N_4}$ stacks—so that the pump laser became part of the photonic module rather than an external laboratory instrument. The other direction moved beyond passive Kerr cavities toward directly electrically driven soliton sources, including monolithic ring quantum-cascade lasers generating Nozaki-Bekki solitons and hybrid III-V/TFLN cavities directly emitting mode-locked microcombs (Raja et al., 2019, Xiang et al., 2021, Ling et al., 2023).

2. Soliton physics and classes of operation

In passive microresonator implementations, the central mechanism is the dissipative Kerr soliton. A continuous-wave pump drives a lossy, dispersive, Kerr-nonlinear cavity; in the anomalous-dispersion regime, the intracavity field can self-organize into a stable circulating pulse whose comb spacing is set by the resonator free spectral range. The standard integrated-dispersion description is

$$D_{\rm int}(\mu)=\omega_\mu-(\omega_0+D_1\mu) =\sum_{i>1}\frac{D_i}{i!}\mu^i,$$

and the comb repetition rate satisfies $f_{\mathrm{rep}\approx D_1/(2\pi)}$. In this framework, stable DKS operation is associated with red detuning and a $\mathrm{sech}^2$-like spectral envelope (Karpov et al., 2017).

The same physical picture underlies many chip-scale microcomb devices, but not all chip-scale soliton lasers are passive Kerr resonators. In monolithic ring semiconductor lasers, the soliton may be an intrinsically dissipative structure of the active cavity itself. In a ring QCL around 8.2 μm, tuning the laser bias current alone produces Nozaki-Bekki solitons, which are traveling localized dark pulses governed by the complex Ginzburg-Landau equation

$\mathrm{Si_3N_4}$0

Their experimental signatures include $\mathrm{Si_3N_4}$1 phase jumps around the primary mode, a reconstructed single localized dark pulse, and a $\mathrm{Si_3N_4}$2 temporal phase slip across the pulse (Opačak et al., 2023).

A distinct active-cavity regime is represented by driven bright solitons on a mid-infrared laser chip. There, a coherent drive field is injected into an active ring resonator above threshold, and fast optical bistability—rather than a saturable absorber or active gain modulation—supports stable bright pulses near 8.3 μm with about 1 ps duration at GHz repetition rates. The theoretical description is framed by the generalized Lugiato–Lefever equation, which places active laser cavities and passive Kerr resonators inside a common driven-dissipative formalism (Kazakov et al., 2024).

The literature therefore suggests that the phrase “chip-scale soliton laser” denotes a family of related sources rather than a single device topology. Bright DKSs in passive rings, bright solitons in active driven resonators, and dark Nozaki-Bekki solitons in semiconductor ring lasers are all represented within that family (Opačak et al., 2023, Kazakov et al., 2024).

3. Materials platforms and integration strategies

Silicon nitride has been the dominant passive platform because it combines wide transparency, Kerr nonlinearity, and dispersion engineering with CMOS-compatible processing. Photonic Damascene reflow improved sidewall smoothness and scattering loss, enabling intrinsic quality factors above $\mathrm{Si_3N_4}$3 in 99 GHz resonators and fiber-chip-fiber transmission around 40%, corresponding to 63% per facet. These process gains directly lowered the power needed for soliton formation and simplified access to the soliton state (Liu et al., 2018).

A more advanced silicon architecture heterogeneously integrates three photonic functions on a single wafer: an InP/Si DFB semiconductor laser, a thermo-optic Si phase tuner, and an ultralow-loss $\mathrm{Si_3N_4}$4 microresonator. Implemented on a 100-mm-diameter silicon substrate, this stack yields thousands of devices per wafer using DUV stepper lithography, photonic Damascene processing, CMP, wafer bonding, and heterogeneous III-V integration. The $\mathrm{Si_3N_4}$5 resonator provides anomalous group-velocity dispersion in the telecom C band, 100 GHz FSR, and intrinsic quality factor about $\mathrm{Si_3N_4}$6, while the 1.8-mm-long InP/Si DFB laser supplies sufficient on-chip power for soliton generation (Xiang et al., 2021).

Other material systems broaden the operating envelope. In 4H-silicon carbide-on-insulator, submicron-confinement microrings achieve average intrinsic $\mathrm{Si_3N_4}$7 million and highest measured $\mathrm{Si_3N_4}$8 million, while exploiting both high Kerr nonlinearity and non-negligible $\mathrm{Si_3N_4}$9. In this platform, the fundamental TE$\mathrm{Si_3N_4}$0 mode is engineered to anomalous dispersion for octave-spanning soliton formation, whereas the higher-order TE$\mathrm{Si_3N_4}$1 mode is engineered to normal dispersion for auxiliary-laser cooling (Zheng et al., 18 Dec 2025).

Thin-film lithium niobate has supported two different integration paths. One path couples an on-chip LNOI pulse generator to a $\mathrm{Si_3N_4}$2 microring, producing a two-chip but fully photonic-chip-based EO–Kerr system. The other embeds a high-$\mathrm{Si_3N_4}$3 resonator directly inside a III-V/TFLN laser cavity. In the former case, a 35.5 GHz pulse train from LNOI drives a high-$\mathrm{Si_3N_4}$4 $\mathrm{Si_3N_4}$5 resonator; in the latter, resonantly enhanced EO modulation, Kerr nonlinearity, and optical gain act within a single hybrid cavity to produce direct comb emission (Niu et al., 21 May 2025, Ling et al., 2023).

At the highest current level of integration reported in the dataset, an InP / thin-film lithium niobate hybrid external-cavity semiconductor laser directly emits ultrafast soliton microcombs. The TFLN circuit contains the nonlinear mode-locking resonator, while the InP reflective semiconductor optical amplifier supplies gain. The reported resonator parameters are intrinsic optical $\mathrm{Si_3N_4}$6 about 2.7 million, loaded $\mathrm{Si_3N_4}$7 about 1 million, and FSR of 200 GHz, with mode-locked output at 0.8 to 3 THz repetition rates (Hu et al., 16 Oct 2025).

4. Soliton access, locking, and stabilization

The major technical bottleneck in chip-scale soliton lasers has been controlled entry into the red-detuned soliton existence range despite thermo-optic drift and laser noise. Several access strategies recur across the literature. One early communication-oriented demonstration used an auxiliary-laser-heating scheme to reliably access and stabilize the single-soliton state in a high-$\mathrm{Si_3N_4}$8 silicon nitride microring with 191.31 GHz native comb spacing. Another low-power $\mathrm{Si_3N_4}$9 platform showed that solitons in 99 GHz resonators could be accessed through simple, slow laser piezo tuning, with soliton steps lasting several hundred microseconds to a millisecond (Geng et al., 2018, Liu et al., 2018).

When compact semiconductor pumps are used, current tuning and self-injection locking become central. A packaged 99-GHz $4 \times 10^8$0 microcomb driven by an ultra-compact, low-noise laser accessed single-soliton states by changing the laser diode current, with no fast actuator such as piezo scan or electro-optic tuning stage. The same general principle is refined in heterogeneously integrated InP/Si/$4 \times 10^8$1 devices, where the resonator is directly coupled to the laser without an isolator and Rayleigh-scattered backward light re-enters the laser cavity, enabling self-injection locking. In that geometry, the on-chip phase tuner current $4 \times 10^8$2 deterministically controls the optical phase relation between the forward laser field and the backward-scattered resonator field (Raja et al., 2019, Xiang et al., 2021).

The nonlinear theory of self-injection locking proved essential because linear models could not explain the large red detunings needed for soliton formation. A coupled laser–microresonator model including self-phase modulation, cross-phase modulation, and backscattering showed that Kerr nonlinearity produces a positive nonlinear detuning shift,

$4 \times 10^8$3

which moves the locked operating point into the soliton-supporting domain. Experimentally, this framework was validated in 30.6 GHz and 35.4 GHz self-injection-locked systems, including real-time measurements of nonlinear tuning curves and direct observation of solitons on both forward and backward current sweeps (Voloshin et al., 2019).

More recent work has pushed stabilization toward deterministic or self-correcting behavior. In 4H-SiCOI, placing the auxiliary laser in the normal-dispersion TE$4 \times 10^8$4 mode prevents modulation-instability comb generation in the cooling channel, yielding a soliton existence range of $4 \times 10^8$5 and 100% success rate for single-soliton access over 100 consecutive sweeps. Complementing this experimentally, numerical bifurcation analysis of a bi-directionally coupled semiconductor laser–microresonator system showed that feedback from the backscattered field dynamically corrects the laser frequency, stabilizing 1-soliton states over parameter ranges unavailable in the uni-directional LLE picture (Zheng et al., 18 Dec 2025, Bengel et al., 24 Apr 2026).

5. Quantitative performance and system demonstrations

The most explicit system-level use of chip-scale soliton lasers as multiwavelength carriers appears in coherent communications. A silicon-nitride DKS comb with 191.31 GHz spacing was converted through Nyquist pulse modulation into a hybrid Kerr–EO comb with effective spacing about 12.75 GHz, enabling 180 CO-OFDM bands of 12.75 Gbaud 8-QAM data over 50 km standard single-mode fiber. The reported results were 6.885 Tb/s total bitrate within 2.295 THz comb bandwidth, spectral efficiency 2.625 bit/Hz/s, and 175 of the 180 bands below the 7% FEC threshold. By exactly matching adjacent comb-family spacing to 12.75 GHz, two neighboring OFDM bands reached mean SNRs of 11.10 dB and 10.34 dB, with BERs of $4 \times 10^8$6 and $4 \times 10^8$7, demonstrating guard-interval-free channel stitching based on mutual coherence (Geng et al., 2018).

At the device level, performance has been benchmarked in terms of linewidth, threshold, repetition rate, and packaging robustness. The packaged 99-GHz $4 \times 10^8$8 microcomb reported a 19 nm 3-dB optical bandwidth, inferred pulse duration 131 fs, and a heterodyne beatnote with 9 kHz Lorentzian linewidth and 61 kHz Gaussian linewidth. The fiber-chip coupling remained stable for more than 30 hours after curing, and a single-soliton state was maintained for more than an hour in a packaged device under a higher-power pump configuration. In self-injection-locked integrated systems, DFB linewidths were reduced from 119 kHz free-running to 1.1 kHz under SIL in a 30.6 GHz device, while an InP/Si/$4 \times 10^8$9 laser soliton microcomb reduced the free-running DFB linewidth from about 60 kHz to about 25 Hz in the self-injection-locked single-soliton state, with nearest comb lines around 200–300 Hz (Raja et al., 2019, Voloshin et al., 2019, Xiang et al., 2021).

Efficiency has become a defining metric in recent architectures. An EO–Kerr hybrid using an LNOI pulse source and a $\mathrm{Si_3N_4}$0 resonator achieved 43.9% pump-to-soliton conversion efficiency under steady-state conditions, with soliton threshold about 1 mW on-chip pump power when 1.6 ps pulses were produced by cascaded modulators. A different integrated approach, described as an electrically empowered microcomb laser, reported individual comb linewidth down to 600 Hz, whole-comb tuning rate exceeding $\mathrm{Si_3N_4}$1 Hz/s, and 100% of the optical power contributing directly to comb generation. The most aggressive all-electric hybrid InP/TFLN soliton laser reported 3-dB bandwidth exceeding 3.4 THz, pulse width below 90 fs, repetition rates from 0.8 to 3 THz, comb linewidth down to 53 Hz, and soliton generation threshold as low as 1 V and 75 mA (Niu et al., 21 May 2025, Ling et al., 2023, Hu et al., 16 Oct 2025).

6. Applications, conceptual boundaries, and open problems

Chip-scale soliton lasers have been developed as enabling sources for coherent transceivers, datacenter interconnects, parallel coherent LiDAR, RF photonics, optical frequency synthesis, and photonics-assisted signal processing. In spectroscopy, two single-soliton $\mathrm{Si_3N_4}$2 microcombs generated on the same chip from a single 1550-nm laser enabled scanning dual-comb spectroscopy over 37.5 THz, with measured resolution $\mathrm{Si_3N_4}$3 kHz and molecular spectroscopy of $\mathrm{Si_3N_4}$4HCN over a 2.3 THz overtone band. In the near infrared, octave-spanning combs covering 776 nm to 1630 nm overlap the biological imaging window and alkali-vapor transitions, while mid-infrared active-chip bright solitons target the 4–12 μm range relevant to sensing and spectroscopy (Xiang et al., 2021, Lin et al., 2020, Karpov et al., 2017, Kazakov et al., 2024).

A recurring conceptual ambiguity concerns what should count as a soliton laser. Some architectures remain externally pumped Kerr microresonators, although the pump source may be compact, packaged, or hybrid integrated. Others are genuinely electrically driven lasers in which gain, mode locking, and nonlinear broadening coexist in one cavity. The surveyed literature suggests that both usages persist, and that the distinction matters technically because the control variables, efficiency limits, residual pump background, and noise transfer pathways differ between externally pumped DKS systems and directly emitting laser combs (Maier et al., 2022, Ling et al., 2023).

A second common misconception is that octave-spanning bandwidth alone completes the route to self-referencing. The SiCOI work explicitly notes that self-referencing still requires stronger out-coupled power at the dispersive-wave frequencies, despite deterministic generation of a single soliton comb spanning 136 to 307 THz with dispersive waves around 138 THz and 302 THz. Future improvements identified there include broader bandwidth couplers, flatter anomalous dispersion, improved $\mathrm{Si_3N_4}$5, and refined dispersion engineering (Zheng et al., 18 Dec 2025).

Open problems also include the controlled exploitation of nonstationary or multistable soliton dynamics. A recent octave-spanning $\mathrm{Si_3N_4}$6 device showed that a single DKS can be driven into chaotic group velocity hopping through phase-modulated Kerr-induced synchronization with an externally injected reference laser. The dynamics were captured by a second-order Adler equation, and the system exhibited a positive maximal Lyapunov exponent,

$\mathrm{Si_3N_4}$7

together with random transitions between distinct repetition rates. This suggests that chip-scale soliton lasers are not only stable comb engines but also controlled nonlinear oscillators that can be steered between precision metrology, communications, and randomness-oriented regimes (Moille et al., 11 Sep 2025).