Quantum Pontus-Mpemba Effect

• Quantum Pontus-Mpemba effect is a protocol-dependent acceleration mechanism whereby an auxiliary preparatory stage reconfigures mode overlaps to speed up relaxation.
• The two-step protocol compares direct evolution with an indirect route through an auxiliary steady state, enabling a faster transition even under identical final dynamics.
• Experimental realizations in cavity QED, Liouvillian skin chains, and dephasing channels highlight its practical significance in advanced quantum control.

The quantum Pontus-Mpemba effect denotes a protocol-dependent acceleration of relaxation in which a quantum system reaches a target state sooner through a two-step evolution than through a direct route. In the formulation introduced for Markovian classical and quantum systems, one compares a direct process from an initial state ${\bf S}$ to a target state ${\bf F}$ with an indirect process ${\bf S}\to{\bf A}\to{\bf F}$, and the effect is present when the total time of the indirect route is shorter even after the preparation stage is included. This operational refinement distinguishes Pontus-Mpemba phenomena from the ordinary Mpemba effect, which compares only relaxation from different initial states under a common final dynamics. Quantum versions have since been developed for open Lindblad systems, time-inhomogeneous dissipative control, cavity QED, non-normal Liouvillian chains, quasiperiodic dephasing problems, closed many-body real-time and imaginary-time dynamics, and resource-theoretic measures of quantum complexity (Nava et al., 20 May 2025, Peluso et al., 19 Feb 2026, Longhi, 7 May 2026, Yu et al., 2 Sep 2025, Aditya et al., 26 Sep 2025).

1. Protocol definition and distinction from the ordinary quantum Mpemba effect

In the standard quantum Mpemba setting, two distinct initial states evolve under the same final generator, and the anomaly is that the state initially farther from equilibrium relaxes faster. The Pontus generalization instead compares two protocols starting from the same initial state. One copy follows the direct route

${\bf S}\to{\bf F}$

in time $t_{\rm SF}$, while the other is first driven toward an auxiliary steady state ${\bf A}$, switched at time $t_{\rm SI}$ to the target dynamics, and then reaches ${\bf F}$ after an additional time $t_{\rm IF}$. The Pontus-Mpemba effect is defined by

$t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$

In the quantum formulation, the second-copy state is piecewise

${\bf F}$0

and the monitoring function is typically the trace distance

${\bf F}$1

Because the preparation cost is explicit, the Pontus criterion is more stringent than the ordinary Mpemba criterion and more directly tied to state-preparation protocols and control design (Nava et al., 20 May 2025).

This formulation also clarifies the scope of “quantum Pontus-Mpemba effect.” In open systems, the two-step route is usually implemented by changing Lindblad generators or dissipation rates; in closed systems, the first stage is instead a transient evolution under a different Hamiltonian, after which the system is switched to the final symmetric Hamiltonian. The common structure is that both protocols share the same target dynamics in the last stage, but differ in how the state is steered before that stage begins (Nava et al., 20 May 2025, Yu et al., 2 Sep 2025).

2. Spectral mechanism: overlap engineering in Liouville space

Across the quantum Mpemba literature, the decisive variable is not initial distance alone but the decomposition of the initial state into relaxation modes. For Markovian open systems with unique steady state ${\bf F}$2, the state admits a Liouvillian-mode expansion of the form

${\bf F}$3

where the amplitudes are overlaps with left eigenmodes. The generic quantum Mpemba mechanism is that a seemingly farther state can relax faster if it has smaller overlap with the slowest decaying mode, and the strong version occurs when that overlap vanishes exactly. In the strong case,

${\bf F}$4

so the asymptotic decay is controlled by the next mode, producing an exponential speedup. More generally, recent work emphasizes that Mpemba behavior is an overlap effect with relaxation eigenmodes rather than a distance effect (Das, 10 Dec 2025, Zhang et al., 2024).

Quantum Pontus-Mpemba protocols exploit the same structure, but by engineering an intermediate state rather than by choosing a different initial state ab initio. The preparatory stage is designed to reduce the coefficient of the slowest mode, enhance overlap with a fast mode, or geometrically steer the state around a slow region of state space. This explains why Pontus acceleration can arise even when the final generator and asymptotic decay spectrum are unchanged: the protocol modifies the coefficients ${\bf F}$5, not necessarily the eigenvalues ${\bf F}$6. In open systems with non-normal Liouvillians, this coefficient engineering can be highly sensitive to boundary localization, exceptional points, or dissipative anisotropy; in closed systems it can instead proceed through transient symmetry breaking and redistribution among symmetry sectors (Longhi, 20 Jan 2026, Song et al., 17 Feb 2026, Yu et al., 2 Sep 2025).

3. Markovian classification and multi-step generalizations

For Markovian classical or quantum systems coupled to multiple reservoirs, all Pontus-Mpemba effects were classified into three classes. The classification is expressed through the behavior of the distance to the target along the auxiliary trajectory. In weak type-I PME,

${\bf F}$7

so the auxiliary route is always closer to the target. In weak type-II PME,

${\bf F}$8

so a crossing can occur. In strong PME,

${\bf F}$9

meaning that the intermediate state is even farther from the target than the starting state, yet the total two-step protocol still wins. The geometric interpretation is a velocity field on state space: the direct route may pass through a low-velocity region, whereas the auxiliary route bypasses it (Nava et al., 20 May 2025).

Later work generalized this logic to time-inhomogeneous Lindblad equations and multi-step protocols. In an open two-level system with time-dependent Hamiltonian and rates, the continuous protocol is described by

${\bf S}\to{\bf A}\to{\bf F}$0

The performance is quantified by a gain function

${\bf S}\to{\bf A}\to{\bf F}$1

The analysis shows that positive gain appears in an intermediate crossover regime between the quasi-static and sudden-quench limits. In the quasi-static limit ${\bf S}\to{\bf A}\to{\bf F}$2, while in the sudden-quench limit ${\bf S}\to{\bf A}\to{\bf F}$3. Time-dependent rates can generate non-Markovian behavior when some ${\bf S}\to{\bf A}\to{\bf F}$4 become negative, but the reported conclusion is that non-Markovianity is neither necessary nor sufficient for a Pontus speed-up (Peluso et al., 19 Feb 2026).

4. Representative open-system realizations

Several concrete open-system realizations make the mechanism explicit.

Setting	Two-step protocol	Reported mechanism
Cavity QED (Longhi, 7 May 2026)	Weak-loss stage, then quench to strong cavity loss	Coherent transfer to photonic fast mode
Liouvillian skin chain (Longhi, 20 Jan 2026)	Coherent boundary swap, then same target Liouvillian	Reduced overlap with slow boundary-localized mode
Quasiperiodic dephasing chain (Song et al., 17 Feb 2026)	Finite-temperature prethermalization, then dephasing	Spectral-weight redistribution without changing decay spectrum

In cavity QED, the model is the Jaynes-Cummings Hamiltonian with cavity photon loss and, typically, ${\bf S}\to{\bf A}\to{\bf F}$5. Starting from the initial atom-cavity state ${\bf S}\to{\bf A}\to{\bf F}$6, the direct strong-loss protocol with ${\bf S}\to{\bf A}\to{\bf F}$7 relaxes mainly through the slow Purcell channel, whose decay rate is

${\bf S}\to{\bf A}\to{\bf F}$8

whereas the fast cavity channel has

${\bf S}\to{\bf A}\to{\bf F}$9

The Pontus protocol first sets ${\bf S}\to{\bf F}$0 for

${\bf S}\to{\bf F}$1

approximately half a vacuum Rabi period, thereby transferring the excitation close to ${\bf S}\to{\bf F}$2. After a sudden quench to large ${\bf S}\to{\bf F}$3, the state overlaps strongly with the fast-decaying mode, and the system reaches ${\bf S}\to{\bf F}$4 faster than under constant high loss. The paper emphasizes that the effect does not arise merely because dissipation is increased; it arises because the quench changes which Liouvillian eigenmodes are populated. The same work identifies an exceptional point at

${\bf S}\to{\bf F}$5

separating underdamped and overdamped regimes, and argues that both optical and circuit QED are experimentally accessible platforms (Longhi, 7 May 2026).

In the dissipative tight-binding chain with asymmetric incoherent hopping, the Liouvillian skin effect localizes right and left relaxation modes near opposite boundaries when ${\bf S}\to{\bf F}$6. The direct protocol evolves the boundary excitation ${\bf S}\to{\bf F}$7 under the target Liouvillian with ${\bf S}\to{\bf F}$8. The Pontus protocol first turns off the bulk terms by setting

${\bf S}\to{\bf F}$9

and evolves for

$t_{\rm SF}$0

thereby swapping the excitation to $t_{\rm SF}$1 before restoring the target Liouvillian. Both routes have the same asymptotic decay rate, governed by $t_{\rm SF}$2, but the two-step route has a smaller coefficient of the slow boundary-localized mode and therefore relaxes faster transiently. The effect disappears when $t_{\rm SF}$3, i.e. when the skin effect vanishes (Longhi, 20 Jan 2026).

In dissipative quasiperiodic chains with local dephasing, the target Lindbladian drives all states to the maximally mixed infinite-temperature steady state

$t_{\rm SF}$4

The two-step protocol first couples the system to a finite-temperature bosonic bath satisfying detailed balance and drives it toward

$t_{\rm SF}$5

then switches to the original dephasing stage. The reported mechanism is again spectral: the protocol leaves the Liouvillian eigenvalues unchanged but suppresses overlap with the slowest decay modes. The acceleration persists for initially localized and extended eigenstates and remains robust in the long-range hopping extension (Song et al., 17 Feb 2026).

5. Closed-system and resource-theoretic variants

A distinct line of work formulated a quantum Pontus-Mpemba effect for closed many-body dynamics. In a disordered spin-$t_{\rm SF}$6 chain with

$t_{\rm SF}$7

the symmetric Hamiltonian is the $t_{\rm SF}$8-preserving point $t_{\rm SF}$9, while ${\bf A}$0 breaks the symmetry. The two-step protocol evolves first under the symmetry-breaking Hamiltonian ${\bf A}$1 and then under the symmetric Hamiltonian ${\bf A}$2. In real time, the diagnostic is the entanglement asymmetry

${\bf A}$3

and in imaginary time the diagnostic is the energy expectation under normalized projection dynamics. The effect is strongest for tilted ferromagnetic initial states with small tilt angle, while larger tilts and tilted antiferromagnetic (Néel) states suppress it. The stated mechanism is that the initial asymmetric stage broadens the charge distribution and destroys the Hilbert subspace imprint bottleneck, making subsequent evolution under the symmetric Hamiltonian more efficient. Numerical evidence for ${\bf A}$4 and ${\bf A}$5 was reported as stable in the thermodynamic limit (Yu et al., 2 Sep 2025).

This closed-system variant is continuous with earlier work on the ordinary quantum Mpemba effect in charge-preserving random circuits, where tilted ferromagnets were found to restore symmetry faster than less tilted ferromagnets, whereas tilted antiferromagnets did not show the effect. There the diagnostic was again entanglement asymmetry, and the proposed mechanism was the spreading of nonconserved operators in terms of conserved densities (Turkeshi et al., 2024). The Pontus construction adds a preparatory stage and thereby turns symmetry-sector engineering into a genuine two-step protocol.

A further generalization places Pontus-Mpemba physics inside quantum resource theories. In random circuit models built from free dynamics, coherence, imaginarity, non-Gaussianity, and magic were analyzed through subsystem resource monotones. The reported result is that coherence and imaginarity exhibit an ordinary quantum Mpemba effect for resourceful product-state initializations, while all four resources exhibit a Pontus-Mpemba effect under a preheating stage followed by free evolution. In that setting, the preparatory step redistributes the resource so that the subsequent free dynamics drains local resource content more efficiently (Aditya et al., 26 Sep 2025).

6. Diagnostics, common misconceptions, and current directions

Quantum Pontus-Mpemba effects are diagnosed by several inequivalent observables, reflecting the fact that the phenomenon is protocol dependent rather than tied to a unique scalar notion of temperature. Open-system studies commonly use trace distance and Hilbert-Schmidt distance to the target state. The cavity-QED proposal evaluates

${\bf A}$6

and

${\bf A}$7

while closed-system studies use entanglement asymmetry or the energy expectation during imaginary-time projection. Resource-theoretic variants replace thermodynamic distances by resource monotones. The choice of diagnostic therefore tracks the physical content of the target state: thermal, symmetric, infinite-temperature, ground-state, or locally resource-free (Longhi, 7 May 2026, Yu et al., 2 Sep 2025, Aditya et al., 26 Sep 2025).

Several misconceptions are explicitly ruled out by the current literature. First, the effect is not a simple consequence of larger dissipation: in cavity QED, the fast protocol succeeds because the first stage transfers amplitude into the photonic fast mode, not because the second stage merely increases ${\bf A}$8 (Longhi, 7 May 2026). Second, the effect is not governed by initial distance alone: both the generic Markovian quantum Mpemba analysis and the quasiperiodic-chain Pontus analysis identify mode overlap, rather than bare distance, as the decisive quantity (Das, 10 Dec 2025, Song et al., 17 Feb 2026). Third, Pontus acceleration need not alter the asymptotic decay rate: the Liouvillian-skin realization is explicitly a same-gap, same-generator effect that changes only the transient coefficient of the slow mode (Longhi, 20 Jan 2026). Fourth, non-Markovianity is not the defining resource; continuous Pontus protocols can show positive gain in both Markovian and non-Markovian sectors (Peluso et al., 19 Feb 2026).

Current directions broaden the concept rather than changing its core mechanism. Multi-step and continuous protocols show that optimal acceleration occurs in an intermediate regime between adiabatic following and sudden quenches, where dynamically generated shortcuts become available (Peluso et al., 19 Feb 2026). Open quantum systems with dynamical phase transitions provide another route: a long metastable time window preceding the transition can be exploited to accelerate the Pontus protocol toward a predesignated target state (Nava et al., 11 Sep 2025). Experimentally, the most concrete near-term route remains cavity and circuit QED, where real-time tuning of ${\bf A}$9 and $t_{\rm SI}$0, variable couplers, and rapid loss quenches are already part of the proposed implementation landscape (Longhi, 7 May 2026). As the field now stands, the quantum Pontus-Mpemba effect is best understood as a protocol-engineering principle: temporary detours through auxiliary dynamics can suppress slow-mode weight, exploit non-normal spectral geometry, or reorganize symmetry content, thereby shortening the full preparation-plus-relaxation time to the target state.