Parametrically Driven Cavity Solitons

• PDCS are localized optical pulses driven by a phase-sensitive parametric process, enabling two stable, binary phase states.
• They are modeled by the PDNLSE with implementations in both quadratic-cubic and pure-Kerr systems, exhibiting sech-shaped profiles and bound-state dynamics.
• Experimental realizations show reduced phase noise and enhanced quantum squeezing, paving the way for applications in random number generation and metrology.

Parametrically driven cavity solitons (PDCSs) are localized dissipative optical pulses sustained in nonlinear resonators by a phase-sensitive parametric drive rather than by direct coherent injection at the soliton carrier frequency. In the 2021 fibre-resonator experiment they were termed parametric cavity solitons (PCSs) and were excited around twice their carrier frequency in a degenerate optical parametric oscillator; in later pure-Kerr work, PDCSs were realized under bichromatic driving, with the soliton carrier located between two pumps and sustained by degenerate four-wave mixing (Englebert et al., 2021, Leonhardt et al., 2022, Moille et al., 2023). Across these implementations, the defining features are the conjugate-field driving term, phase-sensitive locking, and a binary phase structure that permits two solitons of opposite phase to coexist.

1. Concept and distinction from conventional cavity solitons

Conventional temporal cavity solitons in Kerr resonators are sustained by a monochromatic coherent drive near the carrier frequency of the intracavity field. In the standard mean-field description, the forcing appears as an additive external term,

$$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + S,$$

so the soliton is directly phase locked to the driving laser and typically exists as a unique stable localized state for fixed parameters (Englebert et al., 2021).

By contrast, PDCSs are driven through a parametric process. In the quadratic-cubic implementation, the resonator is pumped near twice the signal frequency, and the reduced equation becomes

$$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$

In the pure-Kerr implementation, two coherent pumps at $\omega_\pm$ generate an effective midpoint signal at $\omega_0 \approx (\omega_+ + \omega_-)/2$, and the soliton is sustained by degenerate parametric four-wave mixing rather than by a direct central pump (Englebert et al., 2021, Leonhardt et al., 2022).

Several qualitative differences follow from this change in forcing mechanism. The 2021 work emphasizes that parametrically driven solitons are backgroundless, possess a $\mathbb{Z}_2$ symmetry, and can occur as two phase-locked states of opposite phase. In the pure-Kerr bichromatically driven case, the comb has a sech-like envelope but lacks a pump at its center, which makes spectral separation and extraction easier than in single-pumped dissipative Kerr solitons (Englebert et al., 2021, Mendez et al., 5 May 2026). A recurrent misconception in the field is that PDCSs require simultaneous $\chi^{(2)}$ and $\chi^{(3)}$ nonlinearities; the 2022 and 2023 studies show that they also arise in pure Kerr resonators under bichromatic driving (Leonhardt et al., 2022, Moille et al., 2023).

2. Governing equations and reduced descriptions

The central reduced model for PDCS physics is the parametrically driven nonlinear Schrödinger equation (PDNLSE). In the pure-Kerr bichromatic setting, the starting point is a generalized Lugiato–Lefever equation for the intracavity envelope $E(t,\tau)$,

$$\frac{\partial E(t,\tau)}{\partial t} = \left[-1 + i\left(|E|^2-\Delta_0\right) + iD\!\left(i\frac{\partial}{\partial \tau}\right)\right]E + S_+ e^{-i\Omega_p \tau + i a t} + S_- e^{+i\Omega_p \tau - i a t},$$

where $t$ is slow time, $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$0 is fast time, $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$1 is the dispersion operator, $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$2 are the pump amplitudes, $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$3 is the pump offset from the signal, and $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$4 is a phase-mismatch parameter (Leonhardt et al., 2022).

After separating the stationary pump-frequency fields and assuming that the signal spectrum does not overlap the pumps, the dynamics reduce to

$$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$5

with

$$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$6

and effective parametric coefficient

$$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$7

An important structural point is that the desynchronization parameter $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$8 cancels in the product $$\dfrac{\partial u}{\partial T}= \left(-1+i(|u|^2-\Delta)+i\dfrac{\partial^2}{\partial \tau^2}\right)u + \mu u^*.$$9, so it does not enter the reduced PDCS dynamics (Leonhardt et al., 2022).

The 2026 multimode quantum treatment adopts an equivalent normalized form,

$\omega_\pm$0

where $\omega_\pm$1 is the normalized integrated dispersion and $\omega_\pm$2 is the parametric drive coefficient (Mendez et al., 5 May 2026). In the quadratic-cubic fibre implementation, the 2021 experiment showed that the same PDNLSE accurately describes solitons excited around twice their carrier frequency (Englebert et al., 2021).

These formulations make explicit why PDCSs are phase sensitive. The conjugate-field term $\omega_\pm$3 or $\omega_\pm$4 encodes pairwise conversion processes—down-conversion in the degenerate OPO realization and degenerate four-wave mixing in the pure-Kerr realization. This is also the origin of the analogy drawn in the literature to a parametrically driven damped pendulum (Englebert et al., 2021).

3. Existence conditions, symmetry, and formation dynamics

The 2021 analysis identifies the principal thresholds and bifurcations of the PDNLSE. The degenerate OPO threshold occurs at

$\omega_\pm$5

corresponding to a pitchfork bifurcation of the trivial state. For negative detuning, the pitchfork is supercritical and the trivial state becomes modulationally unstable beyond $\omega_\pm$6; for positive detuning, the pitchfork is subcritical, and the unstable homogeneous branch folds at the saddle-node bifurcation $\omega_\pm$7 at $\omega_\pm$8 (Englebert et al., 2021).

The same work gives an exact solitary-wave solution,

$\omega_\pm$9

with

$\omega_0 \approx (\omega_+ + \omega_-)/2$0

There are two soliton branches, each of which can occur with either of two opposite phases. The two branches connect at the saddle-node bifurcation $\omega_0 \approx (\omega_+ + \omega_-)/2$1 at $\omega_0 \approx (\omega_+ + \omega_-)/2$2; solutions with $\omega_0 \approx (\omega_+ + \omega_-)/2$3 are always unstable; and Hopf bifurcations on the main soliton branch permit breathing or oscillatory localized states (Englebert et al., 2021).

In the pure-Kerr bichromatic case, PDCS existence requires anomalous dispersion at the signal frequency, near phase matching of the degenerate four-wave-mixing process, and approximately homogeneous stationary pump fields. The 2022 theoretical study states the phase-matching condition as $\omega_0 \approx (\omega_+ + \omega_-)/2$4, or equivalently $\omega_0 \approx (\omega_+ + \omega_-)/2$5, and emphasizes that fourth-order dispersion can supply the required nontrivial pump separation. Stable solitons occur when $\omega_0 \approx (\omega_+ + \omega_-)/2$6 and $\omega_0 \approx (\omega_+ + \omega_-)/2$7 are of order unity (Leonhardt et al., 2022).

The soliton phase structure is set by the parametric symmetry. In the pure-Kerr reduced model, the exact PDCS solution has the form

$\omega_0 \approx (\omega_+ + \omega_-)/2$8

with

$\omega_0 \approx (\omega_+ + \omega_-)/2$9

The two allowed phase states correspond to the $\mathbb{Z}_2$0 symmetry of the PDNLSE and are the direct analogue of the two phase states of a parametric oscillator (Leonhardt et al., 2022).

A distinct formation problem arises in doubly resonant degenerate micro-optical parametric oscillators supporting both $\mathbb{Z}_2$1 and $\mathbb{Z}_2$2 nonlinearities. The 2020 study shows that deterministic single PD-DKS generation occurs only above a threshold group velocity mismatch. For $\mathbb{Z}_2$3 pm/V, stable deterministic single PD-DKS generation near perfect phase matching requires $\mathbb{Z}_2$4; at phase mismatch $\mathbb{Z}_2$5, the GVM threshold is reduced by almost an order of magnitude. In that system, a perturbative effective cubic nonlinearity generated by cascaded quadratic processes destabilizes multisolition states and drives soliton annihilation, thereby selecting a single PD-DKS (Nie et al., 2020).

4. Experimental realizations and empirical signatures

The first experimental demonstration of Kerr resonator PDCSs was reported in an all-fibre singly resonant degenerate OPO. The cavity incorporated a 27 cm periodically poled fibre, 21 m of standard single-mode fibre, and 52 cm of erbium-doped fibre. It was synchronously pumped with 650 ps flat-top pulses at 775 nm, generated by frequency doubling a narrow-linewidth 1550 nm laser, while the intracavity EDF was pumped at 1480 nm. Solitons were excited by scanning the driving laser frequency through resonance with a 10 W peak drive corresponding to $\mathbb{Z}_2$6 and a scan rate of 230 kHz/ms; a counter-propagating frequency-shifted control beam and PID feedback loop were then used to stabilize the detuning (Englebert et al., 2021).

The observed signatures were those expected from PDNLSE theory: a resonance response during the scan, a sudden drop in average output power on soliton formation, and a small soliton step. After stabilization, the experiment yielded a resolution-limited pulse circulating every round trip, with autocorrelation and spectrum matching the analytic PDNLSE soliton. The reconstructed pulse was a background-free sech-shaped pulse of about 3.6 ps duration. Coherent detection further showed that different solitons could lock to different phases, confirming the coexistence of two opposite-phase localized states (Englebert et al., 2021).

A second major realization was achieved in a chip-integrated silicon nitride microring resonator. The device had thickness 690 nm, waveguide width 850 nm, radius 23 $\mathbb{Z}_2$7m, round-trip length $\mathbb{Z}_2$8, FSR about 1 THz, intrinsic $\mathbb{Z}_2$9 of $\chi^{(2)}$0, loaded $\chi^{(2)}$1 of $\chi^{(2)}$2, finesse about 3000, and resonance linewidth about 300 MHz. The bichromatic drive used one pump near 314 THz ($\chi^{(2)}$3 nm) and one near 192 THz ($\chi^{(2)}$4 nm), each with about 150 mW on-chip power, targeting a signal frequency near 253 THz ($\chi^{(2)}$5 nm) (Moille et al., 2023).

The experimentally observed signatures in the microcavity included initial non-degenerate parametric oscillation, transition to a smooth broadband comb centered between the pumps, a step-like drop in detected power, and a narrow heterodyne beat note consistent with coherent soliton formation. Additional combs appeared around the pump frequencies with the same line spacing as the central PDCS comb but a constant offset; the measured offset of about 50 GHz $\chi^{(2)}$6 GHz agreed with the theoretical prediction of 49 GHz. Numerical modeling reproduced the smooth comb envelope, a dispersive-wave peak near 210 THz, and a predicted pulse width of about 24 fs (FWHM). In a two-soliton state, suppression of the degenerate-frequency comb line by about 40 dB corresponded to a relative phase near 0.992\pi, which was interpreted as a direct signature of the $\chi^{(2)}$7 phase structure (Moille et al., 2023).

5. Interactions, bound states, and generalized localized structures

PDCS interactions are not restricted to local pulse overlap. In the pure-Kerr bichromatically driven theory, when all waves are resonant and walk-off is finite, the solitons back-act on the intracavity pump-frequency fields and generate localized depletion regions. These depletion structures drift away from the solitons and mediate long-range interactions between well-separated PDCSs. With moderate walk-off, the interaction can be attractive and can even lead to spontaneous generation of additional PDCSs; with larger walk-off, the interaction is much weaker. Third-order dispersion further induces dispersive radiation, recoil, and drift, as in ordinary Kerr cavity solitons (Leonhardt et al., 2022).

The static bound-state problem has been analyzed directly in the parametrically driven, damped nonlinear Schrödinger equation. The 2024 study shows that the stationary ODEs for symmetric two-hump PDCS bound states can be transformed into a Schrödinger-like eigenvalue problem with a self-consistent effective potential. Numerically, this potential is very close to a three-soliton reflectionless potential, and the symmetric two-hump solution is well approximated by a three-eigenfunction structure with discrete eigenvalues

$\chi^{(2)}$8

fixed by the pumping $\chi^{(2)}$9 and damping $\chi^{(3)}$0. In this description, the bound state is not a naive two-soliton molecule but a nonlinear superposition involving two even eigenfunctions and an internal odd mode (Bogdan et al., 2024).

A broader generalization appears in the multicolor bound soliton molecule of a dispersion-engineered Kerr cavity. That state is sustained by a generalized Lugiato–Lefever equation with oscillatory dispersion around zero and consists of several spectrally separated sech components that are temporally overlapped and phase locked. The binding mechanism combines phase-matched inter-soliton four-wave mixing, Cherenkov-radiation transfer, and group-velocity matching. Pumping near 194 THz in the example studied yields three locked spectral regions near 104 THz, 194 THz, and 284 THz, with a temporally coincident pulse of FWHM $\chi^{(3)}$1 fs and a common spectral phase offset of $\chi^{(3)}$2. The authors explicitly characterize this object as a multicolor, strongly interacting PDCS-like bound state rather than an isolated single-frequency PDCS (Luo et al., 2015).

6. Noise, quantum regime, and applications

The classical and quantum noise properties of pure-Kerr PDCSs differ materially from those of conventional monochromatically driven cavity solitons. In the 2026 noise analysis, if the two pumps carry phase fluctuations $\chi^{(3)}$3, then the PDCS phase fluctuation is

$\chi^{(3)}$4

and therefore

$\chi^{(3)}$5

For uncorrelated pumps with equal variance $\chi^{(3)}$6, this gives

$\chi^{(3)}$7

so the central PDCS comb line has half the phase-noise variance of either pump. For perfectly anti-correlated pumps, the phase noise cancels completely. The same study finds that PDCSs exhibit lower quantum-limited timing jitter than comparable conventional cavity solitons because $\chi^{(3)}$8 across the stable PDCS parameter space, and it reports stronger two-mode squeezing for all mode pairs studied (Shamailov et al., 28 May 2026).

The first multimode quantum description of pure-Kerr PDCSs was presented in 2026. Below threshold ($\chi^{(3)}$9), the system behaves as a parametric amplifier below oscillation and supports both single-mode and two-mode squeezing. Above threshold ($E(t,\tau)$0), true PDCS solutions appear and the dominant new feature is the emergence of quantum dispersive waves (QDWs), described as the quantum analogue of soliton Cherenkov radiation. In the quartic-dispersion system studied, the squeezed supermodes become strongly localized near zero-crossing modes around $E(t,\tau)$1 or $E(t,\tau)$2; when the quartic term is removed and the dispersion is purely quadratic, these QDW features disappear. The work reports squeezing levels of up to about 20 dB for experimentally routine parameters and models intrinsic loss through a beam-splitter relation,

$E(t,\tau)$3

so strong squeezing requires strong overcoupling, with numerics performed at 99% overcoupling (Mendez et al., 5 May 2026).

Applications follow directly from the binary phase structure and the spectral geometry of PDCSs. The 2021 fibre experiment showed that a spontaneously formed PCS has a 50% chance of locking into either of the two opposite phase states and used this property to generate 4-bit random numbers in a proof-of-principle experiment. The same phase multiplicity motivates PDCS-based random number generators and Ising machines; in the fibre-OPO platform, a 40 GHz phase modulation was estimated to yield a roughly three-orders-of-magnitude increase in spin-spin couplings compared with the state of the art (Englebert et al., 2021). The later pure-Kerr works connect PDCSs to nonlinear dynamics studies and metrology, noise-sensitive photonic applications, frequency-comb-based communications, squeezed-comb quantum photonics, and frequency-domain continuous-variable photonics, while also stressing the practical advantage that the squeezed light is easier to spectrally isolate because the pumps are far from the soliton spectrum (Mendez et al., 5 May 2026, Shamailov et al., 28 May 2026).

Taken together, these results establish PDCSs as a distinct soliton class within driven-dissipative photonics: they are localized states of the PDNLSE family, supported either by quadratic down-conversion or by bichromatic Kerr four-wave mixing, characterized by phase bistability, capable of forming bound and multicolor states, and increasingly relevant for low-noise comb generation, random-bit encoding, and multimode quantum squeezing (Englebert et al., 2021, Moille et al., 2023).