arXiv:2609.35159
Weakly interacting fermionic Gibbs states are Gaussian mixtures and classically simulable
Zhengyi Han, Yuanchen Zhao, Zi-Wen Liu
Abstract
Identifying regimes in which interacting fermionic systems admit efficient classical simulation is a crucial task in quantum many-body physics. In this work, we rigorously establish the Gaussianity and classical simulability of sufficiently weakly interacting fermions. For the Gibbs state of a local interacting fermionic Hamiltonian $H=H_0+V$ at any finite temperature and in any spatial dimension, where $H_0$ is quadratic and $V$ is the interaction term, we show that there exists a positive interaction strength below which the Gibbs state is a convex mixture of fermionic Gaussian states. Moreover, based on this structure, we construct a polynomial-time classical sampler for the Gibbs state via Metropolis updates on a tree whose leaves are Gaussian states.
AI Summary
The paper asks whether weak interactions provide a general regime for classically simulating fermionic thermal states beyond high temperature. It considers $H=H_0+V$ on finite-dimensional lattices, with bounded finite-range quadratic $H_0$ and even finite-range interactions of local strength $g$. Pairing terms, disorder, and interactions beyond quartic order are allowed; particle-number conservation and special symmetries are not required. For every fixed finite inverse temperature $\beta$, it establishes a system-size-independent threshold $g_{\rm mix}>0$ below which the Gibbs state is exactly a convex mixture of physical fermionic Gaussian states. Below a potentially smaller threshold $g_{\rm samp}>0$, a randomized classical algorithm outputs a Gaussian state $\tau$ satisfying $\|\mathbb E\tau-\rho_\beta\|_1\le\epsilon$ in $\operatorname{poly}(n,\log(1/\epsilon))$ time. Both thresholds scale as $e^{-\Theta(\beta)}$, with other local parameters fixed. Applications include occupation-number sampling, nonlocal observables with polynomially sparse Majorana expansions, additive estimation of reduced-state moments, and subsystem von Neumann entropies for $O(\log n)$ modes. The proof factors out quadratic imaginary-time evolution and recursively removes Majorana labels, controlling interaction coefficients locally to construct a positive Gaussian decomposition. The sampler uses soft quadratic pinning, a connected-cumulant expansion with Wick/Pfaffian and BKAR forest bounds to estimate internal trace weights, and Metropolis updates on a polynomial-depth decomposition tree. The guarantees apply at fixed $\beta$, not uniformly toward zero temperature, and the sufficient interaction thresholds may be very small. Optimal thresholds, long-range interactions, and extensions to broader interaction regimes remain open; convex-Gaussianity alone does not imply efficient sampling.
generative AI tools provided substantial assistance in developing the proof details under extensive instructions from the authors.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.