Reservoir-Engineered Optical Parametric Oscillator

• Reservoir-engineered OPOs are optical parametric oscillators whose environments are deliberately modified to enhance selected nonlinear interactions while suppressing unwanted modes.
• They employ techniques such as mode-spectrum engineering in Kerr microresonators and squeezed-reservoir injection to lower thresholds and control coherence.
• Experimental and theoretical advances show these systems deliver single-mode operation, high spectral purity, and improved performance for quantum metrology and nonlinear photonics.

Searching arXiv for the specified topic and cited papers to ground the article in current literature. A reservoir-engineered optical parametric oscillator (OPO) is an OPO whose effective environment is deliberately redesigned so that selected parametric interactions are enhanced while competing nonlinear channels, decoherence pathways, or unwanted modes are suppressed. In the cited literature, this designation covers at least two distinct but related constructions. In one, the “reservoir” is the set of cavity modes available to Kerr four-wave mixing, and photonic-crystal functionalization plus dispersion design reshape that mode spectrum so that only a single signal-idler pair is phase matched and grows above threshold (Brodnik et al., 10 Apr 2025). In the other, the “vacuum reservoir” coupled to an OPO is replaced by a broadband squeezed vacuum generated by a second OPO, so that the main OPO experiences enhanced parametric coupling and reduced noise in a squeezed basis (Tian et al., 8 Jul 2025). A broader theoretical antecedent treats wavelength-scale, multimode resonators as structured, lossy, non-Hermitian environments whose quasi-normal modes and radiation channels determine OPO threshold, supermode selection, and phase transitions (Jahani et al., 2020).

1. Conceptual scope and definitions

The conventional OPO converts pump photons into lower-frequency photons through parametric interaction. In a Kerr microresonator this appears as degenerate four-wave mixing,

$2\omega_p \rightarrow \omega_s + \omega_i,$

with mode indices satisfying

$m_s + m_i = 2m_p.$

In a standard Kerr microresonator with smooth dispersion and nearly constant free spectral range, broadband parametric gain makes many pairs $(m_s,m_i)$ nearly phase matched. As pump power increases, multiple channels can reach threshold, producing clustered combs, modulation-instability sidebands, cascaded four-wave mixing, soliton states, or chaos (Brodnik et al., 10 Apr 2025).

Reservoir engineering, in this context, denotes deliberate control over the environment that mediates or constrains those nonlinear processes. In the microresonator implementation, the reservoir is the dense set of cavity modes coupled by the Kerr nonlinearity; the objective is to deactivate nearly all of them and activate only one targeted OPO channel (Brodnik et al., 10 Apr 2025). In the squeezed-lasing implementation, the reservoir is the electromagnetic vacuum environment at the output coupler; the objective is to replace ordinary vacuum by a squeezed bath with controlled squeezing strength and angle, thereby enhancing the desired parametric interaction and suppressing spontaneous-emission-related noise (Tian et al., 8 Jul 2025).

The wavelength-scale QNM framework extends the idea to multimode, open resonators. Although the paper does not use the phrase “reservoir engineering,” it explicitly describes the cavity plus radiation channels as a structured environment encoded in complex mode frequencies, losses, and nonlinear overlap integrals. This suggests a broader definition in which a reservoir-engineered OPO is any OPO whose threshold and oscillating supermodes are controlled by tailoring the mode spectrum, dissipation, and intermode coupling of an open photonic system (Jahani et al., 2020).

2. Mode-spectrum reservoir engineering in Kerr microresonators

In the photonic-crystal Kerr OPO, the central strategy is to engineer both the background dispersion and a single local spectral perturbation. Geometric dispersion, set by ring-waveguide width $w$ and thickness $t$, is chosen so that the bare resonator has weakly normal group-velocity dispersion around the pump; this broadly suppresses degenerate OPO and modulation instability. A photonic-crystal bandgap is then imposed on one selected signal mode, shifting that mode alone into phase matching while leaving the rest of the mode spectrum inactive (Brodnik et al., 10 Apr 2025).

The cavity frequencies are expressed through the integrated dispersion,

$$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$

and the phase-matching error is written as

$\delta_\nu = 2\nu_p - \nu_s - \nu_i.$

For a chosen geometry, $\delta_\nu(\mu_s)$ is known. The photonic crystal produces coherent backscattering between clockwise and counter-clockwise modes at a selected azimuthal order $m_{\text{PhC}}$, lifting the degeneracy and generating a blue-shifted and a red-shifted standing-wave pair with splitting $\gamma$. The design rule is

$m_s + m_i = 2m_p.$0

so that the blue-shifted branch moves upward by $m_s + m_i = 2m_p.$1, exactly compensating the dispersion-induced mismatch for one target signal mode (Brodnik et al., 10 Apr 2025).

The result is a dressed spectrum in which only one triplet $m_s + m_i = 2m_p.$2 is phase matched. All other candidate signal modes do not experience a photonic-crystal splitting, remain detuned by the weakly normal group-velocity dispersion, and therefore see negligible parametric gain. The paper describes this as activating only a single OPO interaction within broadband parametric gain, yielding a continuously single-mode OPO with one signal and one idler line and side-mode suppression ratios greater than $m_s + m_i = 2m_p.$3 even far above threshold (Brodnik et al., 10 Apr 2025).

The dynamics are modeled with a frequency-domain Lugiato-Lefever equation extended by a mode-dependent photonic-crystal term,

$m_s + m_i = 2m_p.$4

where $m_s + m_i = 2m_p.$5 is nonzero only for the target mode. In the paper’s interpretation, this modifies the linear Hamiltonian by selectively shifting the chosen signal resonance, thereby altering which eigenmode of the driven-dissipative Kerr system first becomes unstable (Brodnik et al., 10 Apr 2025).

3. Squeezed-reservoir engineering and squeezed lasing

The second major realization of a reservoir-engineered OPO uses two semimonolithic PPKTP-based OPOs. OPO1 is operated below threshold and generates a strong phase-squeezed vacuum at $m_s + m_i = 2m_p.$6; OPO2 is the main lasing OPO, pumped near threshold, whose vacuum reservoir at the output coupler is replaced by the squeezed vacuum emitted by OPO1 (Tian et al., 8 Jul 2025). The experiment uses a single $m_s + m_i = 2m_p.$7 single-frequency continuous-wave laser as the fundamental source, one second-harmonic-generation cavity to generate $m_s + m_i = 2m_p.$8 pump light, and two OPOs based on $m_s + m_i = 2m_p.$9 PPKTP crystals.

The theoretical starting point for OPO2 is

$(m_s,m_i)$0

The external squeezed reservoir is characterized by squeezing parameter $(m_s,m_i)$1 and squeezing angle $(m_s,m_i)$2, with the squeezed-basis transformation

$(m_s,m_i)$3

When the phase relation is set to $(m_s,m_i)$4 and the cavity detuning is $(m_s,m_i)$5, the effective parametric coupling becomes

$(m_s,m_i)$6

This is the key reservoir-engineering result: in the squeezed basis, the parametric interaction is exponentially enhanced by the externally injected squeezed bath (Tian et al., 8 Jul 2025).

The corresponding Heisenberg-Langevin dynamics couple OPO2 to a squeezed input field rather than ordinary vacuum. The squeezed reservoir has correlations

$(m_s,m_i)$7

with

$(m_s,m_i)$8

For a phase-squeezed input,

$(m_s,m_i)$9

The paper’s physical interpretation is that the engineered squeezed bath both enhances coherent parametric amplification and suppresses the undesired noise associated with spontaneous photon emission and two-photon damping (Tian et al., 8 Jul 2025).

Experimentally, OPO1 produces approximately $w$0 phase-quadrature squeezing at $w$1, corresponding to $w$2. After a net propagation loss of about $w$3, this becomes about $w$4 at the OPO2 input, corresponding to $w$5. Under these conditions, the threshold of OPO2 drops from approximately $w$6 with an ordinary vacuum reservoir to approximately $w$7 with the squeezed reservoir, and the emitted field is a bright, narrow-linewidth, squeezed laser with $w$8 amplitude-quadrature squeezing at $w$9, linewidth $t$0, and milliwatt-level brightness (Tian et al., 8 Jul 2025).

4. Multimode, quasi-normal-mode, and non-Hermitian formulations

A general theory for wavelength-scale and multimode OPOs is developed in a quasi-normal-mode basis. The electric field at carrier frequency $t$1 is expanded as

$t$2

where $t$3 is the slowly varying envelope of mode $t$4, $t$5 is the normalized quasi-normal mode, and $t$6 is the decay rate (Jahani et al., 2020). This framework avoids the spatial slowly-varying-envelope approximation, which the paper states breaks down at wavelength scale.

For a single-mode degenerate OPO, the envelope equations are

$t$7

$t$8

The nonlinear coefficient

$t$9

captures spatial overlap between the pump mode, the signal-idler modes, and the nonlinear tensor $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$0 (Jahani et al., 2020). Near threshold, neglecting pump depletion, the threshold condition becomes

$$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$1

The multimode case is formulated through a gain matrix. For the $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$2-th signal mode,

$$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$3

and, in compact form,

$$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$4

Diagonalizing $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$5 yields complex eigenvalues $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$6, whose imaginary parts give parametric gain and whose real parts give the signal-idler frequency splitting $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$7. The oscillation threshold is defined by the first supermode for which $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$8 (Jahani et al., 2020).

This theory shows that multimode interaction can reduce threshold far beyond a simple mode-counting argument. In the AlGaAs-sphere example, including full multimode coupling reduces the actual threshold by a factor of $$\omega(\mu) = \omega_0 + D_1\mu + \frac{D_2\mu^2}{2} + \frac{D_3\mu^3}{6} + \dots = \omega_0 + D_1\mu + D_{\text{int}}(\mu),$$9 relative to the best single-mode magnetic-dipole threshold, even though only four fundamental multipolar modes are considered. The paper attributes this to strong off-diagonal couplings $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$0, which can exceed the diagonal self-couplings $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$1, so that the oscillating field is a collective supermode rather than a single cavity mode (Jahani et al., 2020).

The same non-Hermitian framework predicts phase transitions in parametric gain. As pump power or pump wavelength is varied, the eigenvalue with maximum gain can change identity, and eigenvalues can also coalesce. The paper interprets these transitions as changes between non-degenerate and degenerate OPO states, with discontinuities or divergences in derivatives of the maximum gain. It further argues that operating near these critical points can be used for enhanced sensing (Jahani et al., 2020).

5. Performance, thresholds, and tunability

In the photonic-crystal Kerr implementation, the high-power device operated with a pump near $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$2, signal near $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$3, and idler near $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$4. The idler power exceeded $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$5, the signal power exceeded $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$6, and the total on-chip OPO output exceeded $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$7. The threshold power was $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$8 on chip after facet loss, and the on-chip pump power was estimated up to approximately $\delta_\nu = 2\nu_p - \nu_s - \nu_i.$9. Over this range, the output increased monotonically with pump power, the output spectrum showed only pump, signal, and idler, and the side-mode suppression ratio remained greater than $\delta_\nu(\mu_s)$0 with no mode hopping (Brodnik et al., 10 Apr 2025).

The same work gives a coupling-dependent efficiency model, adapted from Stone et al. and Pérez et al. For a conventional OPO,

$\delta_\nu(\mu_s)$1

and, for threshold,

$\delta_\nu(\mu_s)$2

For the photonic-crystal OPO, the standing-wave character of the signal is included by replacing $\delta_\nu(\mu_s)$3 with $\delta_\nu(\mu_s)$4. The measured coupling values for the highest-power device were $\delta_\nu(\mu_s)$5, $\delta_\nu(\mu_s)$6, and $\delta_\nu(\mu_s)$7. The paper argues that an optimum coupling point exists because larger $\delta_\nu(\mu_s)$8 improves out-coupling and $\delta_\nu(\mu_s)$9 but also raises threshold (Brodnik et al., 10 Apr 2025).

Tunability in that platform is demonstrated lithographically. Four devices with identical waveguide geometry, $m_{\text{PhC}}$0 and $m_{\text{PhC}}$1, but different photonic-crystal parameters $m_{\text{PhC}}$2, produced signal wavelengths around $m_{\text{PhC}}$3, $m_{\text{PhC}}$4, $m_{\text{PhC}}$5, and $m_{\text{PhC}}$6, with corresponding idlers on the opposite side of the pump. The design procedure uses finite-element simulation of cold-cavity integrated dispersion, a hot-cavity correction including an effective pump-resonance shift of about $m_{\text{PhC}}$7 from self-phase and cross-phase modulation, selection of the target azimuthal index $m_{\text{PhC}}$8, and calibration of $m_{\text{PhC}}$9 to satisfy $\gamma$0 (Brodnik et al., 10 Apr 2025).

In the squeezed-lasing implementation, threshold and output depend strongly on the injected squeezing parameter. The measured threshold decreases from approximately $\gamma$1 at $\gamma$2 to approximately $\gamma$3 at $\gamma$4, consistent with the scaling

$\gamma$5

At fixed squeezed injection with $\gamma$6, the normalized pump ratio $\gamma$7 yielded about $\gamma$8 output at $\gamma$9 and about $m_s + m_i = 2m_p.$00 at $m_s + m_i = 2m_p.$01. Above $m_s + m_i = 2m_p.$02, mode competition and phase instabilities appeared, so the device was operated below that point to maintain single-mode squeezed lasing (Tian et al., 8 Jul 2025).

The quadrature squeezing is measured at an $m_s + m_i = 2m_p.$03 sideband, where the detection is shot-noise limited. With $m_s + m_i = 2m_p.$04 and $m_s + m_i = 2m_p.$05, the amplitude-quadrature squeezing reached $m_s + m_i = 2m_p.$06, corresponding to

$m_s + m_i = 2m_p.$07

At fixed $m_s + m_i = 2m_p.$08, increasing $m_s + m_i = 2m_p.$09 from $m_s + m_i = 2m_p.$10 to $m_s + m_i = 2m_p.$11 improved the measured squeezing from about $m_s + m_i = 2m_p.$12 to about $m_s + m_i = 2m_p.$13. The linewidth decreased from about $m_s + m_i = 2m_p.$14 to about $m_s + m_i = 2m_p.$15, approaching the linewidth of the seed laser, which was benchmarked at about $m_s + m_i = 2m_p.$16 by delayed self-heterodyne and about $m_s + m_i = 2m_p.$17 by the $m_s + m_i = 2m_p.$18-separation method (Tian et al., 8 Jul 2025).

6. Relation to phase matching, instability suppression, and broader significance

A central theme across these works is that OPO behavior is not fixed solely by the intrinsic nonlinearity; it is controlled by the linear and dissipative environment in which the nonlinear interaction occurs. In the Kerr microresonator case, weakly normal group-velocity dispersion deactivates the broadband reservoir of unstable modes, and a single photonic-crystal bandgap reactivates only one target signal mode. The paper contrasts this with conventional Kerr resonators, where self-phase and cross-phase modulation, regular mode spacing, and approximate phase matching across many mode pairs enable cluster combs, mode hopping, and chaos (Brodnik et al., 10 Apr 2025). A common oversimplification is that Kerr microresonator OPOs are intrinsically comb-prone or chaos-prone; the reported single-mode, high-power operation indicates that this need not hold when the optical mode environment is engineered lithographically (Brodnik et al., 10 Apr 2025).

In the squeezed-lasing case, the common trade-off between bright coherent emission and quadrature squeezing is modified by replacing the vacuum bath with a squeezed bath. The reported system retains laser-like brightness and a $m_s + m_i = 2m_p.$19 linewidth while producing $m_s + m_i = 2m_p.$20 squeezing, which the paper interprets as simultaneous preservation of coherence and quantum properties in the main OPO (Tian et al., 8 Jul 2025). A second oversimplification is that reservoir engineering in OPOs refers only to dissipative-bath design; the microresonator work shows a mode-spectrum version of the same principle, whereas the squeezed-lasing work uses explicit bath replacement (Brodnik et al., 10 Apr 2025, Tian et al., 8 Jul 2025).

The wavelength-scale theory adds a third qualification: high $m_s + m_i = 2m_p.$21 alone does not determine OPO threshold. The explicit threshold expression depends on both decay rates and nonlinear overlap, and the multimode results show that strong off-diagonal couplings can reduce threshold by a factor larger than the number of interacting modes (Jahani et al., 2020). The same paper establishes a direct connection between second-harmonic-generation efficiency and OPO threshold,

$m_s + m_i = 2m_p.$22

near double resonance, offering a practical design proxy for miniaturized OPOs (Jahani et al., 2020).

The immediate significance reported in these studies is correspondingly diverse. The photonic-crystal Kerr platform targets wavelength-variable, low-noise sources with custom output wavelengths and high spectral purity (Brodnik et al., 10 Apr 2025). The squeezed-reservoir platform targets squeezed lasing for quantum metrology and quantum optics, and the authors also identify quantum communication and coupling to atoms, spins, mechanical oscillators, or other cavities as potential directions (Tian et al., 8 Jul 2025). The wavelength-scale theory identifies phase-transition-assisted sensing, low-threshold miniaturized OPOs, and a route toward compact classical and quantum nonlinear photonics (Jahani et al., 2020). Taken together, these results define reservoir-engineered OPOs as a class of parametric oscillators in which the accessible mode spectrum, the bath statistics, or the non-Hermitian photonic environment is shaped deliberately to control threshold, spectral purity, multimode competition, coherence, and quantum noise.