arXiv:2609.30149
Sharp universal death of entanglement threshold for Pauli Hamiltonians
Bobak T. Kiani
Abstract
We determine the exact universal high-temperature separability threshold for Pauli Hamiltonians of bounded degree $Δ\ge2$. If every coefficient in the Hamiltonian has magnitude at most one and each term has overlapping support with at most $Δ$ other terms, the Gibbs state is a mixture of product Pauli eigenstates whenever \[ β\le z_Δ:= \operatorname{arctanh}\left[\max_{0\le x \le 1}x\left(\frac{1-x}{1+x}\right)^{Δ-1} \right]. \] For every $β>z_Δ$, a finite commuting Hamiltonian with maximum overlap degree at most $Δ$ has an entangled Gibbs state. At any fixed $β<z_Δ$ strictly below the threshold, a classical polynomial-time algorithm produces samples from a distribution over product Pauli eigenstates approximating the Gibbs state in trace distance.
AI Summary
The paper asks for the exact temperature above which every bounded-overlap Pauli Hamiltonian has a fully separable Gibbs state. For $H=\sum_i J_iP_i$, assume $|J_i|\le J$ and each term overlaps at most $\Delta\ge2$ other terms. It proves separability whenever $\beta J\le z_\Delta$, where $z_\Delta=\operatorname{arctanh}[\max_{0\le x\le1}x((1-x)/(1+x))^{\Delta-1}]$. The decomposition uses only product single-qubit Pauli eigenstates. The threshold is sharp: for every $\beta J>z_\Delta$, a finite commuting Pauli Hamiltonian satisfying the same degree bound has an entangled Gibbs state. Examples are $z_2=\tfrac14\log2\approx0.1733$ and $z_3\approx0.0904$. The proof reorganizes an inclusion–exclusion expansion into connected clusters, shifts their operators into a separable cone, and reduces positivity of the remaining coefficients to a polymer independence polynomial. Anticommutation cancellations and a comparison with colored trees establish positivity. Sharpness follows from commuting stabilizer Hamiltonians with two qubits per tree edge: a partial-transpose eigenvalue obeys a tree recursion whose fixed-point boundary gives the same threshold. For weighted interactions, separability holds at $\beta\Lambda\le1/(2e)$, with $\Lambda=\max_i\sum_{j\in N[i]}|J_j|$; this constant is uniformly optimal. At $\beta J\le(1-\delta)z_\Delta$, zero-free partition-function approximation yields a classical product-state sampler with trace-distance error $\epsilon$ and polynomial runtime for fixed $\Delta,\delta$. The sharpness construction has term locality growing with $\Delta$, leaving optimal fixed-locality thresholds open. Sampling is not established at the threshold, and the results do not identify sharp rapid-mixing or physical partition-function zero-free thresholds.
A draft of this manuscript was generated with ChatGPT 6 and heavily revised by me.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.