arXiv:2609.17988
The quantum Mpemba effect in symmetric random Clifford circuits
Shion Yamashika, Filiberto Ares, Pasquale Calabrese
Abstract
The quantum Mpemba effect is the counterintuitive phenomenon whereby a quantum state initially farther from equilibrium relaxes faster than one initially closer to it. We demonstrate this effect in random Clifford circuits with a conserved $\mathrm{U}(1)$ charge, using the entanglement asymmetry to characterize relaxation through dynamical symmetry restoration. Exact numerical simulations show that an initially more asymmetric state can become locally more symmetric than an initially less asymmetric one. We explain this behavior by mapping the dynamics onto a charge-conserving quantum automaton, in which the decay of the entanglement asymmetry is controlled by the statistics of encounters between two particle species evolving according to a symmetric simple exclusion process. This mapping yields an analytic expression for the entanglement asymmetry and provides a simple microscopic mechanism for the quantum Mpemba effect: stronger initial symmetry breaking corresponds to a denser particle configuration, which enhances the frequency of particle encounters and thereby accelerates symmetry restoration.
AI Summary
The paper asks whether a quantum Mpemba effect occurs in charge-conserving random Clifford circuits and what microscopic process controls it. It studies a one-dimensional brickwork circuit of uniformly sampled two-qubit Clifford gates commuting with $Z_i+Z_{i+1}$. Initial product states contain a randomly distributed fraction $m$ of $|0\rangle$ sites, with the remainder in $|+\rangle$. Exact stabilizer simulations, averaged over 1,000 initial configurations and circuit realizations, show crossings of relaxation curves: initially more asymmetric states can become less asymmetric at later times. The observable is entanglement asymmetry under independent on-site $Z$ rotations, rather than only the global conserved $\mathrm{U}(1)$ symmetry. Its ensemble average satisfies $\mathbb{E}[\Delta S]=N_A(1-m)-\mathbb{E}[S(\rho_A)]$. The relevant stabilizer dynamics are mapped to a quantum automaton with equally probable $I$, CZ, SWAP, and SWAP–CZ gates. A two-species exclusion-process description relates purity decay to encounters between particles initially inside and outside the subsystem. Their density is $n=(1-m)/2$. The resulting large-time, large-subsystem estimate is $\mathbb{E}[\Delta S(t)]\simeq\max\{0,N_A(1-m)+\sqrt{2t\ln t}\log_2[(1+m)/2]\}$. Higher particle density accelerates encounters and accounts for the curve crossings. The same prediction extends to arbitrary product stabilizer states, interpreting $m$ as the computational-basis fraction. The analytic expression uses approximate phase cancellation, a characteristic encounter threshold, and replacement of an averaged logarithm by the logarithm of an average. The circuit’s late-time states are not generic Gibbs states. Robustness to measurements and extension beyond Clifford dynamics remain open.
Central claim may not hold: The analytic asymmetry formula uses an uncontrolled replacement of an average logarithm by the logarithm of an average and an uncontrolled sharp encounter threshold. The numerical observable measures symmetry under independent on-site rotations, so its crossings do not establish crossings of entanglement asymmetry for the conserved total U(1) charge. The supplement computes the variance of a SWAP-transported stirring label but identifies it with an order-preserving SSEP particle, leaving its justification of the front coefficient invalid. The auxiliary domain-wall construction does not establish that its indexed-particle dynamics reproduce encounter times across the original randomly populated empty interval.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 8+10 pages, 3+1 figures