Quantum Mpemba effect in chaotic systems with conservation laws
Thomas Martin Müller, Silvia Pappalardi, Rosario Fazio · 2026
Read Quantum Mpemba effect in chaotic systems with conservation laws by Thomas Martin Müller, Silvia Pappalardi, Rosario Fazio (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-11876.txt.
Introduction
Müller, Pappalardi, and Fazio demonstrate a quantum Mpemba effect in two chaotic spin chains with conservation laws. Closed chaotic systems settle into a Gibbs ensemble after a quench, with late relaxation set by conservation laws and hydrodynamics. Pairs of starting states can therefore reach the same final ensemble at very different speeds. In their chains, a state initially closer to equilibrium relaxes more slowly than one starting farther away, giving a simple, robust realisation of the effect.
Catalog record
| CREATOR | Thomas Martin Müller, Silvia Pappalardi, Rosario Fazio |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26541 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.11876
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quantum Mpemba effect in chaotic systems with conservation laws Thomas Martin Müller, Silvia Pappalardi, Rosario Fazio Abstract Closed chaotic quantum systems relax after a quench into a Gibbs ensemble. At late times, the relaxation speed is determined by their conservation laws and hydrodynamics. As a result, there exist pairs of initial states which thermalize to the same ensemble, yet exhibit drastically different hydrodynamic relaxation. We show in two chaotic spin chains how this enables a simple and robust realization of the quantum Mpemba effect: a system initially closer to equilibrium relaxes slower than one that starts farther away, despite both approaching the same final state. Originally, the Mpemba effect refers to the counterintuitive observation that an initially hot sample of water freezes quicker than an initially colder one 48 , i.e. closer to the freezing point. More recently, a closely related phenomenon has been discussed, where a classical 41 , or a quantum system 15 ; 4 initialized further – according to an appropriately chosen measure 59 – from its stationary state can relax faster than one initialized in closer proximity to the final steady state, see Fig. 1 a). If this occurs in a quantum system, we call it quantum Mpemba effect (QME) 3 ; 70 . Systems where the QME was found include closed integrable 4 ; 51 ; 66 ; 56 ; 3 ; 6 ; 5 ; 12 ; 16 ; 68 ; 13 , chaotic 60 ; 55 ; 36 ; 1 ; 11 ; 71 ; 2 ; 29 ; 67 ; 62 ; 69 ; 39 ; 35 and many-body localizing 38 quantum systems as well as open 15 ; 52 ; 53 ; 72 ; 17 ; 18 ; 19 ; 47 ; 55 ; 46 ; 26 ; 36 ; 28 ; 40 ; 37 ; 74 ; 58 ; 64 ; 20 ; 22 ; 10 , and monitored dynamics 25 . The QME has been experimentally observed in quenches in synthetic quantum matter 32 ; 73 ; 65 ; 19 . The underlying mechanism determining the QME is tightly bound to the specific dynamics that governs the quantum system. In a Lindbladian setting with few degrees of freedom, the QME can be constructed by tuning the overlap of the initial state with the slowest decaying mode to zero. Then, the leading behavior at late times is governed by the second-slowest decaying mode, letting the system relax exponentially faster than a generic initial state 15 . In integrable models, typically one can analytically solve the system using a mapping to quasi-particles and predict the QME in a measure quantifying the restoration of a symmetry, the entanglement asymmetry 4 . For chaotic many-body systems, the QME has been seen numerically 4 ; 71 ; 2 ; 29 ; 60 ; 55 ; 36 , but a full physical explanation is still lacking. Filling this gap is the goal of this work. In this Letter, we present a general mechanism on how some specific initial states undergo faster thermalization than a generic initial state, which explains the QME in chaotic many-body systems. We study systems with a conservation law, where thermalization occurs algebraically instead of exponentially, due to hydrodynamic long-time tails 49 ; 42 ; 23 ; 44 . The appearing power-laws can be modified for specific initial states, where the fluctuations of initial and stationary state are in agreement 44 , which results in the robust and systematic appearance of the QME in the hydrodynamic regime, as shown in Fig. 2 . Our numerical observations will be further corroborated by analytic results using a classical hydrodynamic description 44 . Figure 1: Mpemba effect in the hydrodynamic regime. a) Sketch of the relaxation from different initial states (i)-(iii) into a unique stationary state in Hilbert space. Even though all initial states relax into the same state, they may have drastically different relaxation speeds, marked by the arrows. In b), c) we sketch the corresponding relaxation of a witness of the distance from the stationary state to zero in the hydrodynamic regime. b) If an initial state (i) further from equilibrium relaxes slower than one closer (ii), there is no crossing and no Mpemba effect. c) If another state (iii) closer than (ii) to equilibrium also relaxes more slowly, this gives rise to a crossing marked by the star. General setting – The Mpemba effects rests on three critical ingredients: (I) A system setup and generator of the dynamics. In our case, this is a spin chain undergoing chaotic symmetric Floquet or Hamiltonian dynamics. (II) Different initial states, which relax under the given dynamics into a stationary state. An optional additional constraint that we find important is to require that the various initial states locally relax into the same unique stationary state. This is typically not taken into consideration in earlier work on unitarily evolving systems 51 ; 66 ; 56 ; 3 ; 6 ; 5 ; 12 ; 16 ; 68 ; 4 ; 71 ; 2 ; 29 ; 60 ; 55 ; 36 . In this work, we always pick initial states that locally relax into the infinite temperature state. (III) A (local or quasi-local) measure or witness that quantifies the distance between the instantaneous state of the system and the stationary state. Commonly used distance quantifiers are the trace distance, fidelity 54 or entanglement asymmetry 4 , which we check in our numerical simulations. If the distance witness of one state is initially larger than the one of the other state, but there is a time t_{\text{M}} at which this order swaps, we call this QME (see Fig. 1 ). It turns out that this scenario is realized whenever the speed at which relaxation occurs in the chosen distance measure depends on the initial state. Hence, in order to observe the Mpemba effect, we need to find situations where two initial states relax into the same stationary state under the same dynamics at qualitatively different speed. Models – We consider two different cases of chaotic dynamics of a spin- \frac{1}{2} chain described by a (pure) state \ket{\psi} . In both models, we consider periodic boundary conditions in a system of N sites, i=1\dots N , and are interested in the thermodynamic limit N\to\infty . 1) \mathrm{U}(1) symmetric Floquet dynamics 44 – The state after n periods 1 1 1 We identify the discretized time as t=t_{n}=n . evolves according to \ket{\psi(t_{n+1})}=\hat{U}_{\text{FL}}\ket{\psi(t_n)} with \hat{U}_{\text{FL}}=e^{-i\hat{H}_{\gamma}}e^{-i\hat{H}_{\beta}^{\text{e}}}e^{-i\hat{H}_{\beta}^{\text{o}}}e^{-i\hat{H}_{\alpha}}, (1) and \hat{H}_{\alpha}=-\alpha\sum_{i}\hat{Z}_{i}\hat{Z}_{i+1} , \hat{H}_{\beta}^{\text{e/o}}=-\beta\sum_{i\text{ even/odd}}(\hat{X}_{i}\hat{X}_{i+1}+\hat{Y}_{i}\hat{Y}_{i+1}) , \hat{H}_{\gamma}=-\gamma\sum_{i}\hat{Z}_{i}\hat{Z}_{i+2} 2 2 2 We choose the parameters \alpha=2,\beta=0.25,\gamma=1 44 . 2) Mixed field Ising (MFI) model 11 ; 33 ; 7 ; 9 ; 43 – We evolve the state according to the Hamiltonian \hat{H}_{\text{MFI}}=\sum_{i}(J\hat{Z}_{i}\hat{Z}_{i+1}+h_{x}\hat{X}_{i}+h_{z}\hat{Z}_{i}). (2) We take J=1 , h_{x}=(5+\sqrt{5})/8 , h_{z}=(1+\sqrt{5})/4 . Both models have a single conserved quantity \hat{Q}=\sum_{i}\hat{Q}_{i} , where [\hat{Q},\hat{U}]=0 . In the Floquet model, \hat{Q}=\sum_{i}\hat{Z}_{i} , the magnetization is conserved. The MFI model conserves energy, \hat{Q}=\hat{H}_{\text{MFI}} . Then, by the eigenstate thermalization hypothesis 24 , they are both expected to evolve into a state that is locally indistinguishable from a Gibbs state fixed by the expectation value of the respective conserved quantity \hat{Q} in the initial state. Hence, if we pick