arXiv:2610.00474
Algebraic Relaxation of Strong-to-Weak Symmetry-Breaking Order via Charge-Sector Broadening
Zi-Xi Fang, Kai Zhang
Abstract
Strong-to-weak spontaneous symmetry breaking (SW-SSB) is a distinctive phenomenon of mixed quantum states, characterized by nonlinear long-range order associated with strong symmetry. In realistic open quantum systems, however, strong symmetry is macroscopically fragile and can be explicitly reduced to weak symmetry through local charge exchange with the environment. In this paper, we investigate how the SW-SSB order decays under such strong-symmetry-breaking Liouvillian dynamics. For discrete $\mathbb{Z}_2$ symmetry, a local symmetry-breaking perturbation drives an exponential decay of the Rényi-1 long-range correlator, revealing the fragility of discrete SW-SSB order. By contrast, for continuous $U(1)$ symmetry, local charge exchange broadens the distribution over symmetry-charge sectors and leads to parametrically slower algebraic relaxation. We establish a rigorous algebraic lower bound on the connected Rényi-1 long-range correlation, demonstrating the robustness of this slow relaxation. In the thermodynamic limit, we further uncover a two-stage dynamical scaling: $\tilde{R}_1\sim t^{-1}$ at early times and $\tilde{R}_1\sim t^{-1/z}$ at long times, with $z$ set by the underlying hydrodynamic dynamics. As a numerical verification, we consider fermionic chains with $z=2$ and $z=4$, which are excellent agreement with the theoretical predictions.
AI Summary
The paper asks how nonlinear long-range correlations associated with strong-to-weak spontaneous symmetry breaking relax when a reservoir exchanges symmetry charge with only a finite spatial region. It uses a connected Rényi-1 correlator, subtracting the product of nonlinear one-point terms, to track residual order after exact strong symmetry is lost. For a solvable $\mathbb Z_2$ dephasing model, local parity exchange gives exactly $\widetilde R_1(t)=e^{-2\kappa t}$. For $U(1)$-symmetric fermionic dynamics initialized in a fixed-charge maximally mixed state, local particle gain and loss broaden the total-charge distribution. An auxiliary state retaining the exact sector weights but maximally mixing within each sector gives $\widetilde R_1^{\rm aux}\simeq\nu(1-\nu)/(4\sigma_Q^2)$ for a broad Gaussian charge distribution. A conditional lower bound relates the actual correlator to this expression through the sector-averaged squared fidelity to the maximally mixed states. One-dimensional hydrodynamic analysis predicts $\sigma_Q^2\sim t$ initially and $\sigma_Q^2\sim t^{1/z}$ asymptotically, with crossover time $t^*\sim(D/\kappa^z)^{1/(z-1)}$. The corresponding auxiliary-correlator decays are $t^{-1}$ and $t^{-1/z}$. Simulations of diffusive and dipole-conserving chains, with $L=100$ and $50$, find variance exponents approximately $0.48$ and $0.27$, respectively. The algebraic lower envelope requires the fidelity prefactor to remain bounded away from zero; connecting it to the stated auxiliary expression additionally uses weak sector dependence of fidelity. The scaling analysis assumes hydrodynamic relaxation and a broad, approximately Gaussian distribution before finite-size saturation. Extensions to other Rényi indices are derived; slower conditional-mutual-information relaxation is observed numerically but lacks an analytical theory.
Central claim may not hold: The claimed rigorous algebraic lower bound requires an unproved uniformly nonvanishing sector fidelity and an approximate replacement of fidelity-reweighted charge overlaps. The analytical auxiliary-state decay laws and a conditional lower bound do not establish the asserted universal decay exponents of the actual connected correlator. The printed long-time charge-variance coefficient contains an extra factor 1/kappa at balanced half filling and omits the general susceptibility dependence. The covariance closure cannot represent the specified initial covariance and lacks error control sufficient to establish the late-time variance law generally.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 22 pages, 5 figures