arXiv:2610.12844
Improved classical optimization of fermionic Hamiltonians beyond Gaussian states
Ojas Parekh
Abstract
We give a polynomial-time randomized classical algorithm that approximates the largest eigenvalue $λ_{\max}$ of a traceless quartic fermionic Hamiltonian $H$ on $n$ modes. The algorithm achieves energy at least $λ_{\max}/O(\sqrt n)$ and outputs either a fermionic Gaussian state or a polynomially sparse state in the Fock basis. This is achieved through a new upper bound on $λ_{\max}$ in terms of the maximum energy of a Gaussian state and $\sqrt{2^{-n}\operatorname{Tr}(H^2)}$.
AI Summary
The paper studies polynomial-time classical approximation of the maximum eigenvalue of a traceless quartic fermionic Hamiltonian on $n$ modes. It improves the general energy guarantee from $\Omega(1/(n\log n))$ to $\Omega(1/\sqrt n)$ by supplementing Gaussian states with sparse non-Gaussian states. Quadratic terms can be included through an existing reduction using one additional mode. The randomized algorithm returns, with probability at least $2/3$, a state with energy at least $\lambda_{\max}(H)/(c_G\log(2n)+80\sqrt n)$, where $c_G$ is universal. Its output is either Gaussian or supported on at most $t+1$ Fock basis strings, where $t$ is the number of nonzero quartic coefficients; either representation permits polynomial-time energy evaluation. The central bound is $\lambda_{\max}(H)\le G(H)+5\sqrt n\,\sigma$, where $G(H)$ is the optimal Gaussian energy and $\sigma=\sqrt{2^{-n}\operatorname{Tr}(H^2)}$. The proof approximates every state's quartic moment matrix by that of a Gaussian mixture within operator-norm error $2$, then uses physical moment-matrix constraints to obtain Frobenius-norm error $5\sqrt{6n}$. A separate construction filters random Fock states with $16\sigma I+H$ to achieve energy at least $\sigma/16$. Comparing this candidate with an existing $\Omega(1/\log n)$ approximation to $G(H)$ yields the guarantee. The result does not establish a constant-factor approximation or an $\Omega(1/\sqrt n)$ guarantee using Gaussian states alone. The latter question, its polynomial-time attainability, and possible improvements through SDP hierarchies remain open.
GPT-6 Astra was used extensively for this work. I proposed research directions for approximation algorithms for homogeneous quartic fermionic Hamiltonians, and GPT developed additive bounds
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.