Kerr Oscillator: Nonlinear Dynamics

• Kerr oscillators are nonlinear modes whose frequency depends on intensity via a quartic self-interaction, underlying both optical Kerr effects and quantum dynamics.
• In classical settings, they exhibit self-phase modulation, multistability, and chaos, enabling state switching and synchronization in systems like optical fibers and superconducting circuits.
• Quantum realizations produce non-Gaussian phase-space structures and stabilized cat states, offering a robust platform for noise-biased bosonic quantum computation.

A Kerr oscillator is a nonlinear mode whose frequency depends on its own intensity or occupation number through a quartic self-interaction. In nonlinear optics, it models the optical Kerr effect associated with the third-order polarizability $\chi^{(3)}$, often summarized by an intensity-dependent refractive index $n(I)=n_0+n_2 I$; in quantum settings, it is a single bosonic mode with a Kerr Hamiltonian such as $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$. Across these realizations, the defining feature is the same: self-phase modulation generated by a cubic nonlinearity in the classical amplitude equations, or equivalently a quartic term in the oscillator energy. This common structure underlies multistability, parametric oscillation, cat-state stabilization, topological control, and chaos in systems ranging from optical fibers and microresonators to superconducting circuits and trapped ions (Sliwa et al., 2011, Ding et al., 2024).

1. Physical definition and mathematical forms

In nonlinear-optical language, a Kerr oscillator is a model of an optical mode governed by the optical Kerr effect, where the medium’s refractive properties change with intensity because of the third-order nonlinear polarizability $\chi^{(3)}$. The same physics is routinely described as self-phase modulation and cross-phase modulation in fibers, couplers, and related devices (Sliwa et al., 2011). A closely related quantum definition is a single bosonic mode with quartic self-interaction, for which a “bare” Kerr Hamiltonian can be written as

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$

with $a$ the annihilation operator, $\omega_a$ the mode frequency, and $K$ the Kerr nonlinearity (Ding et al., 2024).

Several equivalent Hamiltonian conventions appear in the literature. One common form is

$H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$

while another uses $H=\omega a^\dagger a+\chi(a^\dagger a)^2$ with $n(I)=n_0+n_2 I$0 (Guo et al., 2024). In a classical amplitude description, the single driven-damped Kerr oscillator studied in nonlinear optics is generated from

$n(I)=n_0+n_2 I$1

leading, after phenomenological damping is added, to

$n(I)=n_0+n_2 I$2

(Sliwa et al., 2011). This form makes explicit that the Kerr term changes the instantaneous oscillation frequency by an amount proportional to intensity.

Parametric variants generalize the Kerr oscillator by adding multiphoton drives. A two-photon driven Kerr oscillator in a rotating frame can be written as

$n(I)=n_0+n_2 I$3

where $n(I)=n_0+n_2 I$4 is the detuning and $n(I)=n_0+n_2 I$5 the two-photon drive amplitude (Ruiz et al., 2022). The Kerr-cat Hamiltonian used in superconducting-circuit implementations takes the form

$n(I)=n_0+n_2 I$6

which stabilizes a double-well metapotential in phase space (Ding et al., 2024). Higher-order generalizations also exist, including the three-photon Kerr parametric oscillator

$n(I)=n_0+n_2 I$7

whose discrete symmetry is $n(I)=n_0+n_2 I$8 rather than $n(I)=n_0+n_2 I$9 (Bruno et al., 20 May 2026).

2. Classical dynamics, periodic orbits, and multistability

The classical Kerr oscillator is a paradigmatic driven nonlinear system because the same cubic nonlinearity that shifts the frequency also creates multiple attractors and strong sensitivity to initial conditions. For the single driven-damped model

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$0

the conservative solution at $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$1 is

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$2

showing explicitly that the oscillation frequency depends on the initial intensity $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$3 (Sliwa et al., 2011).

In the driven-damped case, special periodic solutions exist when the pump frequency matches the nonlinear frequency and the initial amplitude is chosen appropriately. For example, with $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$4, $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$5, $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$6, $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$7, and $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$8, the paper reports the periodic solution

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2$9

corresponding to the phase-plane circle

$\chi^{(3)}$0

Numerical evolution from generic initial conditions reveals convergence to one of two coexisting periodic attractors, one at radius $\chi^{(3)}$1 and another at $\chi^{(3)}$2 (Sliwa et al., 2011).

A distinctive feature of this system is the basin geometry. The basins of attraction show a spiral-like structure, and the two circular attractors exhibit “mutual interpenetration”: each attractor lies partly in its own basin and partly in the other’s basin. The paper describes these as “semistable attractors,” because switching between them can be induced by placing the initial condition on a portion of the other attractor’s basin (Sliwa et al., 2011). This suggests that Kerr oscillators are naturally suited to switching tasks in which state selection is performed by initial-condition preparation or transient parameter detuning.

Coupling enriches this structure. For two coupled Kerr oscillators with nonlinear cross-coupling,

$\chi^{(3)}$3

the driven-damped equations acquire cross-phase-modulation terms, and the attractor structure of one subsystem depends on the initial condition of the other (Sliwa et al., 2011). For the parameter set

$\chi^{(3)}$4

the periodic solutions are

$\chi^{(3)}$5

but additional stable and unstable solutions coexist, and the stability of the $\chi^{(3)}$6 attractor in the $\chi^{(3)}$7 subsystem changes when $\chi^{(3)}$8 is changed from $\chi^{(3)}$9 to $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$0 (Sliwa et al., 2011). A plausible implication is that Kerr couplers permit control of one mode by preparing the partner mode, rather than only by adjusting external drive parameters.

3. Chaos, beats, and stability diagnostics

Kerr oscillators support deterministic chaos in both single-mode and multimode settings. In the single classical oscillator, the cited work defines “chaotic beats” as signals showing either chaotic envelopes with a stable fundamental frequency or almost regular collapses and revivals with small chaotic perturbations. These beats are induced by modulating the drive frequency according to

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$1

which yields the nonautonomous equation

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$2

(Sliwa et al., 2011).

Lyapunov-exponent maps provide the standard stability diagnostic. The largest Lyapunov exponent is defined as

$H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$3

and positive $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$4 indicates chaos (Sliwa et al., 2011). In the single oscillator, chaos appears mainly for weak damping, especially for $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$5. One explicit off-resonant example uses $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$6, $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$7, $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$8, $H/\hbar=\omega_a a^\dagger a+K a^{\dagger 2}a^2,$9, $a$0, and $a$1, for which the Lyapunov spectrum is $a$2; a resonant example with $a$3, $a$4, and $a$5 yields $a$6 (Sliwa et al., 2011).

Coupling markedly enlarges chaotic regions. For two coupled Kerr oscillators, Lyapunov maps in the $a$7 plane show much broader chaotic regimes than in the single-oscillator case, and the full Lyapunov spectrum $a$8 reveals hyperchaos when two exponents are positive. In the cited example, hyperchaos occurs for $a$9, while divergence sets in for $\omega_a$0 (Sliwa et al., 2011). With $\omega_a$1 and both oscillators initialized on their periodic circles, the spectrum

$\omega_a$2

signals hyperchaotic beats generated without explicit external modulation, purely through nonlinear energy exchange (Sliwa et al., 2011).

Three coupled Kerr oscillators extend this picture. For triangular and sandwich coupling topologies, the dynamics is governed by

$\omega_a$3

with self-Kerr terms, pairwise cross-Kerr terms, and periodic drives on each mode. The corresponding equations form six nonautonomous first-order ODEs in the real quadratures. Stability analysis based on the full Lyapunov spectrum shows transitions from regular dynamics to quasi-periodicity, chaos, hyperchaos, and chaotic beats as damping is reduced and coupling is increased (Chmielewski et al., 23 Jun 2025). In the triangular configuration, for $\omega_a$4, $\omega_a$5, $\omega_a$6, $\omega_a$7, $\omega_a$8, $\omega_a$9, $K$0, and $K$1, all Lyapunov exponents remain negative for $K$2, remain negative but closer to zero for $K$3, approach bifurcation at $K$4, and become positive at $K$5 (Chmielewski et al., 23 Jun 2025). This suggests that damping engineering is as important as nonlinearity engineering when Kerr oscillators are used in dynamical hardware.

4. Quantum Kerr oscillators and phase-space structure

In quantum mechanics, the Kerr oscillator is the simplest continuous one-dimensional anharmonic system in which the Hamiltonian is quadratic plus quartic in the oscillator energy. A phase-space formulation is especially natural because Kerr evolution shears initially Gaussian Wigner functions into non-Gaussian crescents and, at special times, into cat- or kitten-like structures (Guo et al., 2024, Oliva et al., 2018). The quantum Kerr evolution operator is

$K$6

and the phase rotation rate depends on radius in phase space, exactly as in the classical self-phase-modulation picture (Guo et al., 2024).

A particularly explicit formulation is available in terms of the Wigner current. For the Kerr Hamiltonian

$K$7

the Wigner distribution $K$8 satisfies the continuity equation

$K$9

In the units $H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$0, $H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$1, $H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$2, the exact current is

$H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$3

so $H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$4 is tangent to circles in phase space (Oliva et al., 2018). The classical part corresponds to the phase-space velocity

$H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$5

with angular velocity

$H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$6

while the quantum correction introduces a Laplacian term that suppresses classical shear (Oliva et al., 2018).

The paper interprets this suppression as an effective quantum “viscosity.” A local measure is built from the vorticity of the quantum part of the current,

$H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$7

and the global shear-suppression measure is

$H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$8

Classically, shear causes indefinite filamentation; quantum mechanically, the Kerr current polarizes across structures in $H=\omega a^\dagger a+\frac{K}{2}a^{\dagger 2}a^2,$9 and halts further refinement at the Zurek scale (Oliva et al., 2018). A plausible implication is that the same nonlinearity that generates interference fringes also self-limits the fine-scale phase-space flow that would classically diverge.

The driven quantum Kerr oscillator further shows collapse-and-revival structure. For

$H=\omega a^\dagger a+\chi(a^\dagger a)^2$0

the paper derives, for an initial coherent state $H=\omega a^\dagger a+\chi(a^\dagger a)^2$1,

$H=\omega a^\dagger a+\chi(a^\dagger a)^2$2

with an autocorrelation

$H=\omega a^\dagger a+\chi(a^\dagger a)^2$3

For $H=\omega a^\dagger a+\chi(a^\dagger a)^2$4, this exhibits fractional and full revivals whose period decreases as $H=\omega a^\dagger a+\chi(a^\dagger a)^2$5 increases (Bolandhemmat et al., 2023). The same work reports that the Mandel parameter remains

$H=\omega a^\dagger a+\chi(a^\dagger a)^2$6

for Kerr-evolved coherent states, so the number statistics stays Poissonian even though the phase-space distribution becomes strongly non-Gaussian (Bolandhemmat et al., 2023).

5. Parametric Kerr oscillators, cats, and encoded quantum information

Two-photon driving converts the Kerr oscillator into a parametric device with a double-well phase-space structure and a nearly degenerate cat manifold. In the detuned two-photon driven model

$H=\omega a^\dagger a+\chi(a^\dagger a)^2$7

special detunings

$H=\omega a^\dagger a+\chi(a^\dagger a)^2$8

produce multiple exact degeneracies between even and odd sectors. At such points there are $H=\omega a^\dagger a+\chi(a^\dagger a)^2$9 perfectly degenerate pairs of levels, including the two ground states, and the associated localized states remain confined to the two lobes of the double well (Ruiz et al., 2022). The paper argues that these extra degeneracies suppress bit-flip channels because leakage into the lowest excited manifolds no longer accumulates relative phase between parity sectors (Ruiz et al., 2022).

A related experimentally realized system is the squeezed Kerr oscillator

$n(I)=n_0+n_2 I$00

or, with detuning,

$n(I)=n_0+n_2 I$01

Its metapotential has minima at $n(I)=n_0+n_2 I$02 with

$n(I)=n_0+n_2 I$03

and its ground manifold is spanned by the cat states

$n(I)=n_0+n_2 I$04

The same work reports that the lifetime of the coherent-state components increases in steps as $n(I)=n_0+n_2 I$05 increases, a pattern interpreted as pairwise “spectral kissing” in the excited-state spectrum (Frattini et al., 2022). In particular, the paper states that QND readout fidelities greater than $n(I)=n_0+n_2 I$06 are achieved and the phase-flip lifetime is enhanced by more than two orders of magnitude while universal control is retained (Frattini et al., 2022).

The Kerr-cat qubit realizes these ideas in a hardware-efficient bosonic code. Under

$n(I)=n_0+n_2 I$07

the metapotential has two minima at coherent states $n(I)=n_0+n_2 I$08 with

$n(I)=n_0+n_2 I$09

The logical manifold is spanned by even and odd cat states

$n(I)=n_0+n_2 I$10

while the bit-flip channel is exponentially suppressed and phase flips scale with the mean photon number (Ding et al., 2024). In the reported experiment, a SNAIL-based Kerr-cat qubit coupled parametrically to a high-$n(I)=n_0+n_2 I$11 storage cavity achieves a conditional displacement rate

$n(I)=n_0+n_2 I$12

with a $n(I)=n_0+n_2 I$13 gate, a Kerr strength

$n(I)=n_0+n_2 I$14

and, after frequency-selective dissipation is activated, storage-cavity $n(I)=n_0+n_2 I$15 consistent with $n(I)=n_0+n_2 I$16 at large $n(I)=n_0+n_2 I$17, indicating negligible idle dephasing (Ding et al., 2024). This suggests that the Kerr oscillator is not merely a nonlinear mode but a controllable protected manifold for bosonic quantum information.

6. Networks, platforms, and broader applications

Kerr oscillators appear in multiple hardware classes, and their role changes with the driving architecture. In integrated nonlinear optics, Kerr parametric oscillators can be built from silicon nitride microresonators. A dual-pumped degenerate Kerr oscillator in a Si$n(I)=n_0+n_2 I$18N$n(I)=n_0+n_2 I$19 ring operates through four-wave mixing with

$n(I)=n_0+n_2 I$20

and, in the degenerate regime, exhibits two stable phase states separated by $n(I)=n_0+n_2 I$21, analogous to a $n(I)=n_0+n_2 I$22 degenerate OPO (Okawachi et al., 2015). The reported device has free spectral range $n(I)=n_0+n_2 I$23, loaded quality factor $n(I)=n_0+n_2 I$24, and degenerate oscillation is observed for pump wavelengths $n(I)=n_0+n_2 I$25 and $n(I)=n_0+n_2 I$26 with combined on-chip pump power $n(I)=n_0+n_2 I$27 (Okawachi et al., 2015).

The same binary-phase principle extends to chip-scale Kerr DOPOs with richer nonlinear dynamics. In a silicon nitride microring pumped symmetrically about a central mode, the three-mode equations

$n(I)=n_0+n_2 I$28

support a pitchfork bifurcation into the two $n(I)=n_0+n_2 I$29-shifted phase states, followed at higher power by a Hopf bifurcation to self-sustained MHz oscillations, then by period doubling and a numerical route to chaos (Trinchão et al., 18 May 2026). The reported device has radius $n(I)=n_0+n_2 I$30, loaded quality factor

$n(I)=n_0+n_2 I$31

linewidth

$n(I)=n_0+n_2 I$32

and experimentally observed Hopf oscillations near

$n(I)=n_0+n_2 I$33

(Trinchão et al., 18 May 2026).

Kerr oscillators also form programmable computational networks. In the KPO picture, each mode encodes an Ising spin in the sign of a coherent state $n(I)=n_0+n_2 I$34. Two capacitively coupled Josephson KPOs implement an effective coupling

$n(I)=n_0+n_2 I$35

whose sign and magnitude are controlled by the relative pump phase $n(I)=n_0+n_2 I$36 (Yamaji et al., 2022). Experimentally, the same-phase probability reaches approximately $n(I)=n_0+n_2 I$37 at $n(I)=n_0+n_2 I$38, vanishes near $n(I)=n_0+n_2 I$39, and flips parity at $n(I)=n_0+n_2 I$40 (Yamaji et al., 2022). This suggests that external microwave phase can serve as a Hamiltonian-programming knob in KPO-based Ising hardware.

More elaborate networks realize higher-order interactions. Four-body couplings between Kerr parametric oscillators can be generated using only linear capacitive couplers, producing an effective interaction

$n(I)=n_0+n_2 I$41

and, at the spin level,

$n(I)=n_0+n_2 I$42

A four-KPO circuit demonstrates the phase-controlled four-spin parity constraint and quantum annealing in the Lechner–Hauke–Zoller architecture (Kawakami et al., 29 Nov 2025). This broadens the encyclopedic meaning of “Kerr oscillator” from a single nonlinear mode to a modular unit for programmable many-body Hamiltonians.

Other realizations continue the same theme. A tapered ion trap generates an effective mechanical Kerr oscillator for the radial motional mode with

$n(I)=n_0+n_2 I$43

where both the Kerr coefficient and the squeezing rate are set by trap geometry and drive parameters (Nikolova et al., 31 Mar 2025). A Kerr-induced synchronized soliton microcomb yields a terahertz voltage-controlled oscillator whose repetition rate is locked through optical frequency division with

$n(I)=n_0+n_2 I$44

and measured synchronization bandwidth

$n(I)=n_0+n_2 I$45

(Javid et al., 2024). These examples indicate that the Kerr oscillator concept is now shared by nonlinear optics, superconducting circuits, trapped ions, and photonic frequency-comb engineering.

The unifying point is that Kerr nonlinearity supplies a minimal anharmonic mechanism for intensity-dependent phase evolution. In classical devices this yields switching, multistability, chaos, and synchronization; in quantum devices it yields protected cat manifolds, noise-biased encodings, tunable degeneracies, and programmable many-body couplings. A plausible synthesis is that the Kerr oscillator has become a standard nonlinear primitive: simple enough to admit exact or reduced descriptions, yet rich enough to support the full spectrum from periodic motion to fault-tolerant bosonic control (Sliwa et al., 2011, Ding et al., 2024).