arXiv:2610.03696
Maximum-Entropy Extension of Quantum Correlation Functions from Short Real-Time Dynamics
Filippo Maria Gambetta, Sabrina Yue Wang, Raul A. Santos
Abstract
Correlation functions are central to understanding the behavior of quantum systems, with their spectra giving direct access to quasiparticle and collective excitations. However, the long-time evolution required for sufficient spectral resolution is challenging to reach for both classical and quantum simulations. In this work, we introduce a principled approach for extending time series of quantum correlation functions and reconstructing their spectral densities, based on their connection with covariance sequences of stationary stochastic processes. This connection enables us to apply Burg's maximum-entropy principle to find the least-committal extension compatible with the measured values of the correlation function, which corresponds to an all-pole autoregressive model whose coefficients solve a set of Yule--Walker equations. The resulting extension is positive semi-definite by construction, guaranteeing the positivity of the spectrum and the automatic satisfaction of its total spectral weight sum rule. The approach applies to both scalar and matrix-valued correlation functions, including those with highly structured or continuous spectra. The generalization to the matrix case makes it possible to trade simulation time for additional measured correlation functions, enabling accurate time-series extension and spectral reconstruction from short-time data, even in the presence of noise.
AI Summary
The paper addresses reconstruction of quantum correlation spectra when simulations provide only short real-time trajectories. It treats sampled stationary correlators as covariance lags of a stationary stochastic process, rather than as signal realizations. For $n$ samples with a strictly positive-definite Toeplitz Gram matrix, Burg’s maximum-entropy-rate principle selects a Gaussian autoregressive process of order at most $n-1$, determined by Yule–Walker equations. A matrix-valued extension uses block equations to constrain auto- and cross-correlators jointly. The resulting infinite extensions preserve positive semidefiniteness, spectral positivity, and total spectral weight, with stable autoregressive dynamics. Numerical tests use an open, 11-site Fermi–Hubbard chain at $U/t_h=4$, comparing scalar and matrix maximum-entropy reconstructions against exact dynamics. Using all 121 spatial correlators sampled at $\Delta t=0.2/t_h$ through $t=3/t_h$, the matrix method gives relative $L^2$ errors of 0.35 for the displayed space-time dynamics and 0.34 for the broadened local spectra, versus 0.84 and 0.53 for the scalar construction. Central-site spectral error falls below 10% with training through $t=4/t_h$. Noise tests combine block-Toeplitz positive-semidefinite denoising with bootstrap analysis, including Gaussian noise of standard deviation 0.1. The extension is a maximum-entropy choice among generally infinitely many compatible continuations, not a guarantee of recovering the true dynamics. Near-singular or noisy Gram matrices require denoising or regularization; scalar and few-channel fits can be fragile. More channels require additional correlator measurements and manipulation of $nM\times nM$ matrices. Evidence is confined to this finite-chain example, and spectral comparisons use Lorentzian broadening $\eta=0.1t_h$.
Error found: The scalar and matrix direct spectral formulas have a frequency-sign error: with the paper's Fourier conventions, their AR polynomials must be evaluated at e^{+iωΔt}, not e^{−iωΔt}. The displayed maximum-entropy spectra satisfy the total-weight sum rule over one Nyquist interval, not over the entire real line as written, because they are periodic in frequency. The numerical discussion claims perfect matrix spectral agreement despite reporting a relative L2 error of 0.34, and similarly overstates the space-time reconstruction with reported error 0.35.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.