Engineered Two-Photon Loss (ETPL)

• Engineered Two-Photon Loss (ETPL) is a reservoir-engineering tool that uses the a² jump operator to preserve parity by acting solely within even or odd photon-number sectors.
• It stabilizes non-Gaussian cat states in driven nonlinear oscillators, enabling autonomous state generation and mitigating errors in quantum systems.
• ETPL smooths metrological dynamics by converting irregular oscillations from single-photon loss into a steady, monotonic decay, thereby enhancing sensing performance.

Engineered Two-Photon Loss (ETPL) is a reservoir-engineering tool that induces the loss of photon pairs from a bosonic mode via the jump operator $L = a^2$. In Lindblad form, ETPL appears as $\kappa_2 \mathcal{D}[a^2]\rho$, and its defining feature is parity preservation: because $a^2$ changes excitation number by $\pm 2$, it acts within fixed even or odd photon-number sectors rather than coupling them. In driven nonlinear oscillators, this property allows ETPL to generate and stabilize non-Gaussian even-parity cat states, to confine dynamics to a quantum manifold spanned by coherent states of opposite phase, and, in the presence of unavoidable single-photon loss (SPL), to mitigate the loss-induced degradation of metrological resources by transforming hard-to-track damped oscillations into a smooth, monotonic decay (Yang et al., 22 Apr 2026, Leghtas et al., 2014).

1. Definition, formalism, and parity structure

ETPL is most naturally formulated for a bosonic mode $a$ governed by an open-system master equation of the form

$$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$

where $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$, $\kappa_1$ denotes intrinsic single-photon loss, and $\kappa_2$ is the engineered two-photon loss rate (Kapit, 2017). In the two-photon-driven Kerr resonators studied for quantum metrology, the corresponding model is

$$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$

with zero detuning $\kappa_2 \mathcal{D}[a^2]\rho$0, and the Lindblad dynamics are

$\kappa_2 \mathcal{D}[a^2]\rho$1

where $\kappa_2 \mathcal{D}[a^2]\rho$2 is the SPL rate and $\kappa_2 \mathcal{D}[a^2]\rho$3 is the ETPL rate (Yang et al., 22 Apr 2026).

The central algebraic feature is parity conservation under the two-photon processes. The parity operator is $\kappa_2 \mathcal{D}[a^2]\rho$4. Even cat states and odd cat states are

$\kappa_2 \mathcal{D}[a^2]\rho$5

Since $\kappa_2 \mathcal{D}[a^2]\rho$6 preserves parity, ETPL does not induce even–odd switching; by contrast, a single-photon jump $\kappa_2 \mathcal{D}[a^2]\rho$7 flips parity. This distinction is the basis for ETPL’s use in cat-state stabilization, bosonic error mitigation, and dissipative sensing protocols (Yang et al., 22 Apr 2026, Leghtas et al., 2014).

A standard reformulation in cat-code settings uses an engineered jump operator $\kappa_2 \mathcal{D}[a^2]\rho$8, whose dark states satisfy $\kappa_2 \mathcal{D}[a^2]\rho$9 and are the even and odd cat states. The review literature presents this as the canonical dissipative route to a cat manifold, while also emphasizing the simpler ETPL limit with collapse operator proportional to $a^2$0 (Kapit, 2017).

2. Nonlinear oscillator models and steady-state manifolds

A particularly important ETPL setting is the two-photon-driven Kerr resonator with three model variants: TPD-Kerr with SPL ($a^2$1, $a^2$2), TPD-Kerr-ETPL ($a^2$3, $a^2$4), and TPD-ETPL ($a^2$5, $a^2$6) (Yang et al., 22 Apr 2026). In the metrological study, the numerical examples use $a^2$7, $a^2$8, and vary $a^2$9 and $\pm 2$0 (Yang et al., 22 Apr 2026).

With $\pm 2$1, the ETPL master equation

$\pm 2$2

has pure steady states in the manifold spanned by $\pm 2$3 with

$\pm 2$4

Because both $\pm 2$5 and $\pm 2$6 contain only even powers of $\pm 2$7, parity is conserved. From a vacuum initial state, the system relaxes to the even-parity cat $\pm 2$8; from $\pm 2$9 it relaxes to $a$0 (Yang et al., 22 Apr 2026).

The experimentally realized storage-mode ETPL model of “Confining the state of light to a quantum manifold by engineered two-photon loss” takes the effective form

$a$1

with

$a$2

Neglecting Kerr and single-photon loss, the semiclassical steady-state condition yields a manifold spanned by $a$3 and $a$4 with

$a$5

This is the sense in which ETPL can confine the state of light to a quantum manifold rather than stabilize only a single fixed state (Leghtas et al., 2014).

3. Loss mitigation mechanism: parity confinement, spectral smoothing, and cat stabilization

In the SPL-only TPD-Kerr model, the master equation admits a quantum-jump representation

$a$6

The non-Hermitian $a$7 has approximately degenerate coherent-state eigenmodes $a$8. In the regime $a$9,

$$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$0

and $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$1 (Yang et al., 22 Apr 2026).

The dynamical consequence is a parity-flip mechanism. Each SPL jump sends an even cat into an odd cat according to $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$2. Stochastic parity flips therefore drive the state back and forth between even and odd manifolds, producing long-lived damped oscillations in both the quantum Fisher information gain and the squeezing level. The steady state is approximately

$$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$3

with vanishing interference in the Wigner function and degraded metrological resources (Yang et al., 22 Apr 2026).

Adding ETPL changes the Liouvillian in three specific ways. First, ETPL introduces a jump $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$4 that preserves parity and an effective imaginary Kerr-like damping $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$5 that selects a preferred amplitude. Second, ETPL confines the dynamics to fixed-parity manifolds, so it does not create the even–odd switching that underlies the irregular oscillations. Third, ETPL increases the Liouvillian damping of the nonstationary modes responsible for underdamped oscillations under SPL, converting the trajectories of $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$6 and $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$7 into a smooth, monotonic decay (Yang et al., 22 Apr 2026).

This mechanism is often summarized by three phrases used in the metrological analysis: parity confinement, spectral smoothing, and cat-state stabilization. The first reflects the action of $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$8 within a fixed parity sector; the second refers to enhanced damping of oscillatory Liouvillian modes; the third refers to the dissipative selection of a cat amplitude through the balance of $$\dot{\rho} = -i[H,\rho] + \kappa_1 \mathcal{D}[a]\rho + \kappa_2 \mathcal{D}[a^2]\rho + \cdots,$$9 and $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$0 (Yang et al., 22 Apr 2026).

4. Metrological consequences in two-photon-driven resonators

The metrological setting considered for ETPL is displacement sensing along an optimal quadrature

$$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$1

with

$$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$2

The quantum Fisher information is defined as

$$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$3

where $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$4 are eigenvalues and eigenvectors of $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$5, and the maximized value is $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$6. A coherent state yields $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$7. Squeezing is quantified through the minimum quadrature variance $$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$8 and the squeezing level

$$\mathcal{D}[O]\rho = O\rho O^\dagger - \tfrac{1}{2}(O^\dagger O\rho + \rho O^\dagger O)$$9

with $\kappa_1$0 and $\kappa_1$1 indicating squeezing (Yang et al., 22 Apr 2026).

For the TPD-Kerr model with $\kappa_1$2, $\kappa_1$3 rises to $\kappa_1$4 at $\kappa_1$5, then quickly drops and enters long-lived damped oscillations persisting beyond $\kappa_1$6, with $\kappa_1$7 by that time. The non-oscillatory high-sensitivity interval, using the threshold $\kappa_1$8, lasts only from $\kappa_1$9 to $\kappa_2$0. The squeezing $\kappa_2$1 exhibits an early peak of $\kappa_2$2 at $\kappa_2$3, followed by strong antisqueezing, $\kappa_2$4, and irregular oscillations that remain predominantly negative for long times (Yang et al., 22 Apr 2026).

For the TPD-Kerr-ETPL hybrid with $\kappa_2$5, even weak ETPL, $\kappa_2$6–$\kappa_2$7, suppresses oscillation amplitude, while moderate ETPL, $\kappa_2$8–$\kappa_2$9, eliminates oscillations entirely and yields a smooth decay. With $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$0, the non-oscillatory $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$1 window extends from $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$2–$$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$3 to $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$4–$$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$5; with $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$6, it extends to $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$7–$$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$8. For squeezing, the practical window $$H = \varepsilon(a^{\dagger 2} + a^2) - K a^{\dagger 2} a^2,$$9 extends from $\kappa_2 \mathcal{D}[a^2]\rho$00–$\kappa_2 \mathcal{D}[a^2]\rho$01 without ETPL to $\kappa_2 \mathcal{D}[a^2]\rho$02–$\kappa_2 \mathcal{D}[a^2]\rho$03 for $\kappa_2 \mathcal{D}[a^2]\rho$04 (Yang et al., 22 Apr 2026).

The Wigner-function snapshots reveal a temporal hierarchy of quantum resources. At early times, $\kappa_2 \mathcal{D}[a^2]\rho$05–$\kappa_2 \mathcal{D}[a^2]\rho$06, Gaussian squeezing dominates, and $\kappa_2 \mathcal{D}[a^2]\rho$07 grows in tandem with $\kappa_2 \mathcal{D}[a^2]\rho$08. At intermediate times, $\kappa_2 \mathcal{D}[a^2]\rho$09–$\kappa_2 \mathcal{D}[a^2]\rho$10s, non-Gaussian even-parity cat states dominate: although $\kappa_2 \mathcal{D}[a^2]\rho$11 has decayed toward $\kappa_2 \mathcal{D}[a^2]\rho$12, $\kappa_2 \mathcal{D}[a^2]\rho$13 remains high and slowly decays. At long times, residual SPL contaminates parity and coherence, and both $\kappa_2 \mathcal{D}[a^2]\rho$14 and $\kappa_2 \mathcal{D}[a^2]\rho$15 approach zero as the state relaxes toward a classical mixture. ETPL delays and smooths this decay (Yang et al., 22 Apr 2026).

An important limiting case is the TPD-ETPL model with $\kappa_2 \mathcal{D}[a^2]\rho$16. For $\kappa_2 \mathcal{D}[a^2]\rho$17–$\kappa_2 \mathcal{D}[a^2]\rho$18, its behavior closely matches the hybrid TPD-Kerr-ETPL case at the same $\kappa_2 \mathcal{D}[a^2]\rho$19. This demonstrates that strong ETPL with $\kappa_2 \mathcal{D}[a^2]\rho$20 dominates the dynamics, making Kerr negligible for the metrological trajectory (Yang et al., 22 Apr 2026).

5. Physical implementation and experimental demonstrations

A standard implementation of ETPL uses a lossy ancillary “buffer” mode $\kappa_2 \mathcal{D}[a^2]\rho$21 coupled to a memory mode $\kappa_2 \mathcal{D}[a^2]\rho$22 through the two-photon exchange Hamiltonian

$\kappa_2 \mathcal{D}[a^2]\rho$23

with the resonance condition $\kappa_2 \mathcal{D}[a^2]\rho$24. In the regime $\kappa_2 \mathcal{D}[a^2]\rho$25, adiabatic elimination yields an effective ETPL jump $\kappa_2 \mathcal{D}[a^2]\rho$26 with rate

$\kappa_2 \mathcal{D}[a^2]\rho$27

This rate is tunable via flux and mode matching in superconducting circuits (Yang et al., 22 Apr 2026).

The 2014 experimental realization of ETPL in a superconducting microwave resonator implemented a closely related mechanism. The storage and readout modes were coupled through a four-wave mixing process driven at

$\kappa_2 \mathcal{D}[a^2]\rho$28

and adiabatic elimination of the rapidly damped readout mode yielded

$\kappa_2 \mathcal{D}[a^2]\rho$29

The measured parameters included storage $\kappa_2 \mathcal{D}[a^2]\rho$30, readout $\kappa_2 \mathcal{D}[a^2]\rho$31, storage lifetime $\kappa_2 \mathcal{D}[a^2]\rho$32, readout lifetime $\kappa_2 \mathcal{D}[a^2]\rho$33, $\kappa_2 \mathcal{D}[a^2]\rho$34, pump $\kappa_2 \mathcal{D}[a^2]\rho$35, $\kappa_2 \mathcal{D}[a^2]\rho$36, and $\kappa_2 \mathcal{D}[a^2]\rho$37, with $\kappa_2 \mathcal{D}[a^2]\rho$38 (Leghtas et al., 2014).

Starting from vacuum in that experiment, the Wigner function showed clear squeezing at $\kappa_2 \mathcal{D}[a^2]\rho$39, interference fringes and Wigner negativity at $\kappa_2 \mathcal{D}[a^2]\rho$40 with average photon number $\kappa_2 \mathcal{D}[a^2]\rho$41 and measured parity $\kappa_2 \mathcal{D}[a^2]\rho$42, and loss of negativity by $\kappa_2 \mathcal{D}[a^2]\rho$43 as single-photon loss converted the cat into a classical mixture centered near $\kappa_2 \mathcal{D}[a^2]\rho$44 (Leghtas et al., 2014).

Superconducting circuits provide the most developed ETPL platform. Reported implementations achieve $\kappa_2 \mathcal{D}[a^2]\rho$45–$\kappa_2 \mathcal{D}[a^2]\rho$46, often more than $\kappa_2 \mathcal{D}[a^2]\rho$47 larger than SPL, with $\kappa_2 \mathcal{D}[a^2]\rho$48–$\kappa_2 \mathcal{D}[a^2]\rho$49, using Josephson nonlinearities, three-wave mixing, and autoparametric resonance (Yang et al., 22 Apr 2026). The review literature also places ETPL in a broader reservoir-engineering context that includes strongly damped auxiliary modes, parametric pumping of Josephson elements, and effective incoherent rates of the form

$\kappa_2 \mathcal{D}[a^2]\rho$50

with on-resonance scaling $\kappa_2 \mathcal{D}[a^2]\rho$51 in the familiar single-auxiliary-mode model (Kapit, 2017). Optomechanics and trapped ions are described as additional feasible platforms: two-phonon loss via red detuning on the two-phonon sideband in membrane-in-the-middle or self-oscillation optomechanical setups, and two-phonon loss via sideband cooling on the red second sideband in trapped ions (Yang et al., 22 Apr 2026).

6. Relation to autonomous stabilization, error correction, and terminological boundaries

ETPL is a paradigmatic example of engineered dissipation used as a resource rather than treated solely as a source of decoherence. The review “The upside of noise: engineered dissipation as a resource in superconducting circuits” presents ETPL as a central reservoir-engineering primitive for state generation, state stabilization, and autonomous quantum error correction, with the rate hierarchy

$\kappa_2 \mathcal{D}[a^2]\rho$52

identified as the characteristic “ratcheting” condition in several autonomous protocols (Kapit, 2017).

In cat encodings, ETPL stabilizes the oscillator manifold while single-photon loss flips parity and maps the logical manifold to an error manifold. The review further argues that any dissipative mechanism which can correct photon loss and imposes an energy gap $\kappa_2 \mathcal{D}[a^2]\rho$53 is very likely to automatically suppress dephasing, with the approximate scaling

$\kappa_2 \mathcal{D}[a^2]\rho$54

for low-frequency $\kappa_2 \mathcal{D}[a^2]\rho$55 phase noise (Kapit, 2017). This suggests that ETPL is relevant not only for cat-state preparation but also for biased-noise bosonic memories and autonomous error suppression.

At the same time, ETPL has limits that recur across the literature. It does not eliminate SPL; very long times still see decay to classical mixtures. Excessively strong $\kappa_2 \mathcal{D}[a^2]\rho$56 can reduce peak squeezing and peak QFI, so the ETPL rate must be balanced against the two-photon drive and any Kerr nonlinearity. In the metrological study, the practical guideline is to use $\kappa_2 \mathcal{D}[a^2]\rho$57–$\kappa_2 \mathcal{D}[a^2]\rho$58 and ensure $\kappa_2 \mathcal{D}[a^2]\rho$59, with conservative operational thresholds such as $\kappa_2 \mathcal{D}[a^2]\rho$60 or $\kappa_2 \mathcal{D}[a^2]\rho$61 (Yang et al., 22 Apr 2026).

A common terminological confusion arises from uses of the phrase “engineered two-photon” that do not correspond to canonical ETPL of a bosonic mode. In “Steady state entanglement of two superconducting qubits engineered by dissipation,” the engineered two-photon ingredient is a two-tone microwave drive that realizes an effective two-photon transition on each transmon, $\kappa_2 \mathcal{D}[a^2]\rho$62 via virtual excursions through $\kappa_2 \mathcal{D}[a^2]\rho$63, and the natural single-photon resonator loss $\kappa_2 \mathcal{D}[a^2]\rho$64 then yields an effective correlated-qubit dissipator

$\kappa_2 \mathcal{D}[a^2]\rho$65

that pumps the system into the singlet state. The paper explicitly distinguishes this from cat-state stabilization with a genuine cavity dissipator $\kappa_2 \mathcal{D}[a^2]\rho$66 (Reiter et al., 2013). The distinction is conceptually important: canonical ETPL refers to a bosonic two-photon loss channel with jump operator $\kappa_2 \mathcal{D}[a^2]\rho$67, whereas some dissipative protocols use engineered two-photon drives to generate different effective dissipators.

Within that boundary, ETPL’s defining role is consistent across the cited literature: it preserves parity, shapes the Liouvillian rather than only the Hamiltonian, stabilizes non-Gaussian manifolds, and, when combined with suitable driving, enables autonomous confinement and tracking of quantum resources in the presence of realistic single-photon loss (Yang et al., 22 Apr 2026, Kapit, 2017).