arXiv:2610.27462
Systematic Truncated Wigner Approximation for Open Spin Systems Reveals Long-Range Order Far from Equilibrium
Igor Timofeev, Dawid A. Hryniuk, Marzena H. Szymańska
Abstract
Long-range interactions can protect dissipative order against fluctuations, but determining when they do so in two dimensions remains out of reach for existing many-body methods. We develop a continuous SU(2)-covariant truncated-Wigner framework that maps the problem onto stochastic dynamics with one Bloch sphere per site. Derived systematically from the exact Stratonovich-Weyl evolution of the Lindblad equation, it generates both drift and reservoir noise for arbitrary spin $S$. Benchmarks against exact and tensor-network calculations validate its description of non-equilibrium dynamics and steady states. Applied to 400-site two-dimensional dissipative anisotropic XY lattices, our method reveals physics missed by mean-field and perturbative approaches: vortex-antivortex proliferation destroys the predicted staggered-XY phase for nearest-neighbour coupling, whereas algebraically decaying interactions stabilize an extended $\mathbb{Z}_2$-ordered phase. We obtain the first beyond-mean-field steady-state phase diagram of the long-range interacting model, revealing reentrant loss of order at strong coupling and closure of the ordered region beyond $α\approx 3.5$. Our new method provides a scalable route to strongly correlated driven-dissipative systems with long-range interactions, higher-dimensional geometries, spin-boson couplings, and spins beyond one half, relevant to modern experiments with magnetic atoms and polar molecules in optical lattices and tweezer arrays.
AI Summary
The paper addresses how to simulate correlated steady states of two-dimensional dissipative spin lattices with long-range interactions, and whether those interactions stabilize order absent for nearest-neighbour coupling. It derives a continuous SU(2)-covariant truncated-Wigner approximation from the Stratonovich–Weyl representation of Lindblad dynamics. Expanding the spin star product through second order in $\epsilon=(2S+1)^{-1}$ gives Hamiltonian and dissipative drift plus reservoir diffusion, represented by stochastic trajectories with two coordinates per spin. The construction accommodates arbitrary $S$; stochastic sampling requires positive-semidefinite diffusion. Benchmarks include exact dissipative Lipkin–Meshkov–Glick dynamics and published iPEPO results for the nearest-neighbour anisotropic XY model. Reported steady-state magnetization errors are approximately $2\%$ for $S=10$, $15\%$ for $S=1/2$ in the single-spin benchmark, and up to $10\%$ relative to iPEPO for the lattice. On $20\times20$ lattices, nearest-neighbour simulations show exponentially decaying correlations and vortex–antivortex proliferation, consistent with suppression of mean-field staggered-XY order. For interactions $r^{-\alpha}$, a finite-system correlation map indicates an extended $\mathbb{Z}_2$-ordered region: at $\alpha=3$, it spans roughly $0.5\lesssim J/\Gamma\lesssim2$, with reentrant disorder at stronger coupling. No ordered region is resolved above $\alpha\approx3.5$. The approximation improves with increasing spin and cannot exactly represent negative steady-state Wigner distributions through positive trajectory densities. Signed initial-state sampling becomes impractical for large lattices. The long-range phase map uses 30 trajectories per point at finite time and size; the reported correlation crossover is plausibly affected by finite size.
Central claim may not hold: The claimed long-range ordered phase is inferred from a normalized collective correlator on one 20×20 lattice without finite-size scaling establishing a nonzero thermodynamic order parameter. The abstract's closure near α≈3.5 is stronger than the body's finding that no ordered region is resolved above that exponent at the accessible system size. The reported algebraically decaying correlations do not establish the nonzero asymptotic correlation limit required for true long-range order. The spin-1/2 long-range phase boundaries lack independent benchmarking or truncation-error control, despite appreciable discrepancies in the supplied benchmarks.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.