Pontus–Mpemba Effect: Protocol Acceleration

• The Pontus–Mpemba effect is a protocol-induced relaxation anomaly where an auxiliary preparatory step helps a system reach its target state faster than direct relaxation.
• It leverages a two-step method to suppress the slowest spectral mode via an intermediate generator, effectively bypassing the bottleneck in dynamics.
• This effect has been demonstrated in classical, quantum, and many-body systems, with applications in cavity-QED, dissipative chains, and other engineered setups.

The Pontus–Mpemba effect is a protocol-dependent relaxation anomaly in which two identical systems, prepared in the same initial state and evolving toward the same target state, reach that target faster under a two-step protocol than under direct relaxation, even after the time spent in the preparatory step is included. In contrast to the ordinary Mpemba effect, which compares different initial states under a common final generator, the Pontus formulation compares different dynamical paths from a common starting point. In the modern literature it is formulated for classical and quantum Markovian dynamics, and has been extended to multi-step, continuous-control, and closed-system settings (Nava et al., 20 May 2025, Peluso et al., 19 Feb 2026).

1. Definition and conceptual scope

The ordinary Mpemba effect concerns two initial states—typically labeled “cold” and “hot”—that relax under the same post-quench dynamics, with the hotter state sometimes reaching the target faster. The Pontus–Mpemba effect instead compares two protocols acting on the same initial state ${\bf S}$. In the direct process, the system is quenched immediately to the final generator and reaches the target ${\bf F}$ in time $t_{\rm SF}$. In the two-step process, the system first evolves toward an auxiliary state under a different generator, is interrupted at an intermediate state ${\bf I}$, and then relaxes under the final generator, with total time

$T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$

A Pontus–Mpemba effect occurs when

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

This formulation makes the preparation stage part of the observable speed-up rather than treating it as external to the comparison (Nava et al., 20 May 2025).

The terminology follows “a strategy of fishermen in Pontus described by Aristotle,” used to motivate the idea that a detour can shorten total arrival time. In this sense the Pontus–Mpemba effect is not an initial-condition effect but a protocol effect: the system is deliberately driven away from the direct relaxation path in order to improve the eventual approach to the same target (Nava et al., 20 May 2025, Peluso et al., 19 Feb 2026).

A common misconception is to treat Pontus–Mpemba phenomena as merely a renamed quantum Mpemba effect. The distinction is sharper. In the ordinary effect, the anomaly is encoded in a non-monotonic dependence of the slow-mode amplitude on the initial state; in the Pontus case, the anomaly is encoded in how an auxiliary generator reshapes the state before the final relaxation stage begins. The two notions are related but not identical.

2. General Markovian and spectral formulation

For classical systems, the state is a probability vector $p(t)$ satisfying a master equation $dp/dt=W p$. For open quantum systems, the state is a density matrix $\rho(t)$ satisfying a Lindblad equation

$$\frac{d\rho}{dt} = \mathcal L[\rho] = -\,i[H,\rho] +\sum_j \gamma_j \Bigl( L_j \rho L_j^\dagger -\tfrac12\{L_j^\dagger L_j,\rho\} \Bigr).$$

Under the target dynamics, deviations from the final steady state can be expanded in right eigenmodes ${\bf F}$0,

${\bf F}$1

with the slowest nonzero mode ${\bf F}$2 determining late-time relaxation (Nava et al., 20 May 2025).

In this framework, the direct protocol has slow-mode amplitude

${\bf F}$3

while the two-step protocol produces an intermediate state ${\bf F}$4 with slow-mode amplitude

${\bf F}$5

The necessary and sufficient condition for a Pontus–Mpemba speed-up is

${\bf F}$6

The mechanism is therefore spectral: the preparatory stage is useful only insofar as it sufficiently suppresses the overlap with the bottleneck mode of the final generator (Nava et al., 20 May 2025).

Later quantum realizations preserve this logic while changing the physical origin of the suppression. In dissipative quasiperiodic chains, the two-step stage redistributes Liouvillian weights ${\bf F}$7 so that ${\bf F}$8, while leaving the Liouvillian spectrum itself unchanged (Song et al., 17 Feb 2026). In a dissipative tight-binding chain with Liouvillian skin effect, the coherent first step transfers the excitation so that the state has much smaller overlap with the slow boundary-localized left eigenmode, again without changing the asymptotic gap ${\bf F}$9 (Longhi, 20 Jan 2026). These cases show that the central object is not the gap alone but the pair $t_{\rm SF}$0.

3. Taxonomy and terminology

The general theory distinguishes three classes of Pontus–Mpemba effect. In weak type-I PME, the auxiliary trajectory stays closer to the target than the direct one throughout the preparation interval, so the advantage is visible already at intermediate times. In weak type-II PME, the auxiliary trajectory is not always closer, but the two distance curves cross once before the switch time, and the intermediate state is still closer to the target than the direct trajectory at the same elapsed time. In strong PME, the auxiliary trajectory may initially move away from the target, with $t_{\rm SF}$1, and the intermediate state can even be farther from the target than the initial state, yet the total arrival time is still shorter than for direct relaxation (Nava et al., 20 May 2025).

This classification should be kept distinct from the older notion of the strong Mpemba effect in ordinary relaxation theory. There, the defining property is the exact vanishing of the slowest-mode amplitude for a special initial temperature,

$t_{\rm SF}$2

so that relaxation proceeds with rate $t_{\rm SF}$3 or faster. The number of such special temperatures defines the Mpemba index, and its parity is a topological property in Markovian dynamics (Klich et al., 2017). By contrast, strong PME does not require a zero of the slow mode as a function of initial temperature; it concerns a protocol that can transiently worsen the distance to the target and still improve total completion time (Nava et al., 20 May 2025).

This terminological distinction matters because several quantum papers use “Pontus–Mpemba” to denote protocol-induced acceleration, while the ordinary strong effect remains tied to spectral zeros in families of initial equilibrium states. Conflating the two obscures whether the anomaly comes from state preparation, generator switching, or both.

4. Representative quantum realizations

A substantial part of the literature studies the Pontus–Mpemba effect in open quantum systems, where the auxiliary step can be engineered by coherent control, temporary dissipation changes, or temporary thermalization.

System	Two-step protocol	Mechanism
Dissipative tight-binding chain	Coherent swap, then relaxation	Suppressed overlap with slow skin mode
Jaynes–Cummings cavity QED	Weak-loss Rabi swap, then strong-loss quench	Loads fast dissipative mode
Quasiperiodic dephasing chain	Thermal intermediate state, then dephasing	Redistribution of Liouvillian weights

In the dissipative tight-binding chain with asymmetric incoherent hopping $t_{\rm SF}$4 and coherent boundary coupling $t_{\rm SF}$5, the Liouvillian skin effect localizes right and left eigenmodes at opposite boundaries when $t_{\rm SF}$6. Direct relaxation from $t_{\rm SF}$7 has slow-mode amplitude $t_{\rm SF}$8. The two-step protocol first implements a coherent swap $t_{\rm SF}$9 with ${\bf I}$0, approximately sending ${\bf I}$1, and then restores the original Liouvillian. The resulting slow-mode amplitude is ${\bf I}$2, and the skin effect gives ${\bf I}$3 for ${\bf I}$4. The spectral gap ${\bf I}$5 is unchanged, but the prefactor is exponentially suppressed; the effect disappears at ${\bf I}$6, where the skin effect vanishes (Longhi, 20 Jan 2026).

In cavity quantum electrodynamics, the effect was proposed in the Jaynes–Cummings model with photon loss, restricted to the single-excitation sector ${\bf I}$7. The direct protocol uses a large cavity decay rate ${\bf I}$8 throughout, so the atomic excitation decays via the slow Purcell mode,

${\bf I}$9

The two-step protocol first evolves for $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$0 under $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$1, with $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$2, so that a nearly coherent Rabi swap prepares an almost pure photon state $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$3. After quenching to $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$4, the subsequent relaxation overlaps almost entirely with the fast mode,

$T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$5

Because $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$6, the quenched trajectory overtakes the direct one in the overdamped regime. Optical cavity-QED and circuit-QED implementations were explicitly identified as experimentally accessible (Longhi, 7 May 2026).

In dissipative quasiperiodic chains subject to local dephasing, the target steady state is the maximally mixed infinite-temperature state $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$7. The direct protocol relaxes from an initial eigenstate of the quasiperiodic Hamiltonian under pure dephasing. The two-step protocol first couples the system for time $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$8 to a weak auxiliary thermal bath to generate an intermediate state $T_{\rm total}=t_{\rm SI}+t_{\rm IF}.$9, then turns off that bath and resumes pure dephasing. Since the post-switch Liouvillian is identical in both cases, any acceleration must come from modified weights $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$0, especially a reduced $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$1. The reported numerical results show the effect for both localized and extended eigenstates, robustness under power-law long-range hopping, and $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$2–$t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$3 for the parameters studied (Song et al., 17 Feb 2026).

5. Extensions: many-body, closed-system, and optimal-control variants

Beyond single-switch open-system protocols, Pontus–Mpemba effects have been connected to metastability and dynamical phase transitions. In dissipative lattice fermions described as a dissipative Gross–Neveu model, a dynamical phase transition is signaled by $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$4 while $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$5 remains finite, producing a metastable window $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$6. A direct quench across such a region inherits the slow $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$7 bottleneck. A two-step protocol can avoid it by choosing an auxiliary point so that both sub-quenches retain finite $t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$8, yielding

$t_{\rm SI}+t_{\rm IF}<t_{\rm SF}.$9

Here the Pontus acceleration is tied to bypassing metastable waiting times rather than merely to a local overlap reduction in a fixed Liouvillian spectrum (Nava et al., 11 Sep 2025).

The control-theoretic generalization considers $p(t)$0-step and continuous protocols governed by time-inhomogeneous Lindblad master equations. In the $p(t)$1 limit one obtains continuous Pontus–Mpemba protocols, with optimization formulated through Euler–Lagrange equations or the Pontryagin Minimum Principle for the Bloch-vector dynamics

$p(t)$2

This framework reveals a crossover between sudden-quench and quasi-static limits. In the sudden limit the protocol reduces to the standard two-step case; in the quasi-static limit the system follows the instantaneous attractor and no speed-up occurs. The optimized regime lies in between, where time-dependent dissipation rates produce dynamically generated shortcuts. Allowing transiently negative rates introduces non-Markovianity, but the reported results do not show a strict one-to-one correlation between non-Markovianity and gain (Peluso et al., 19 Feb 2026).

Pontus–Mpemba effects have also been extended beyond open-system relaxation. In a disordered spin-$p(t)$3 chain with a $p(t)$4-symmetric Hamiltonian $p(t)$5 and symmetry-breaking Hamiltonian $p(t)$6, the same initial tilted ferromagnetic state is evolved either directly under $p(t)$7 or first under $p(t)$8 and then under $p(t)$9. The effect appears in both real-time and imaginary-time dynamics, is most pronounced for small tilt angles, is suppressed for large tilts and for tilted antiferromagnetic states, and remains robust under finite-size scaling from exact diagonalization and TEBD (Yu et al., 2 Sep 2025). A plausible implication is that protocol-induced acceleration is not intrinsically tied to dissipation, provided the first stage broadens the state’s support in the degrees of freedom that control the later bottleneck.

An even broader generalization occurs in quantum resource theories. For coherence, imaginarity, non-Gaussianity, and magic, direct “cooling” under free circuits can be compared with “preheating + cooling,” where the preheating stage is a resource-generating circuit and the cooling stage is a free circuit. The reported result is that all four resources exhibit a Pontus–Mpemba effect, even though only coherence and imaginarity show an ordinary quantum Mpemba effect in the same framework (Aditya et al., 26 Sep 2025). This shifts the concept from thermal relaxation to resource dissipation and quantum complexity.

6. Relation to ordinary Mpemba effects and broader significance

The ordinary Mpemba effect is now widely understood spectrally: the anomaly appears when a farther initial state has a smaller overlap with the slowest relaxation mode. In Markovian models this logic underlies both weak and strong ordinary Mpemba effects, including the strong condition $dp/dt=W p$0 and the associated Mpemba index (Klich et al., 2017). In overdamped Langevin dynamics on a double-well potential, the strong ordinary effect occurs when the initial well occupation matches the bath occupation,

$dp/dt=W p$1

equivalently suppressing the slow inter-well mode; the same work also expresses the weak effect through mean first-passage times (Walker et al., 2022).

The Pontus–Mpemba effect inherits this spectral perspective but turns it into a control problem. Instead of selecting a favorable initial temperature, one engineers a favorable intermediate state. Instead of asking whether a hotter preparation happens to suppress the bottleneck mode, one asks whether an auxiliary generator can do so fast enough that the detour pays for itself. This suggests that Pontus–Mpemba protocols are an operational generalization of ordinary Mpemba logic from state families to trajectory design.

Several open directions follow directly from the current literature. One is the optimization of switching time and the design of protocols beyond a single switch (Peluso et al., 19 Feb 2026, Yu et al., 2 Sep 2025). A second is extension to non-Markovian reservoirs and alternative dissipation channels (Song et al., 17 Feb 2026, Peluso et al., 19 Feb 2026). A third is analytic control of many-body and complexity-theoretic realizations, where the role of spectral form factors, monitored dynamics, or quasiparticle structure remains unresolved (Aditya et al., 26 Sep 2025). A fourth is experimental systematics: cavity QED, circuit QED, optical lattices, photonic synthetic lattices, and engineered dissipative chains have all been identified as candidate platforms, but the decisive observable is not merely a gap measurement; it is the protocol-dependent redistribution of spectral weight onto and away from the slowest mode (Longhi, 7 May 2026, Longhi, 20 Jan 2026).

In this form, the Pontus–Mpemba effect occupies a distinct position within nonequilibrium relaxation theory. It is not primarily a statement about “hotter” versus “colder” states, but about whether a temporary excursion under a different generator can create an effective shortcut to the same final steady state. The unifying theme across current realizations is precise: the asymptotic spectrum may remain fixed, yet carefully chosen preparation dynamics can make the slowest mode largely irrelevant to the observed relaxation time.