English

Area laws and tensor networks for maximally mixed ground states

Quantum Physics 2025-06-24 v2 Other Condensed Matter Computational Complexity Information Theory math.IT

Abstract

We show an area law in the mutual information for the maximally-mixed state Ω\Omega in the ground space of general Hamiltonians, which is independent of the underlying ground space degeneracy. Our result assumes the existence of a `good' approximation to the ground state projector (a good AGSP), a crucial ingredient in previous area-law proofs. Such approximations have been explicitly derived for 1D gapped local Hamiltonians and 2D frustration-free locally-gapped Hamiltonians. As a corollary, we show that in 1D gapped local Hamiltonians, for any ε>0\varepsilon>0 and any bi-partition LLcL\cup L^c of the system, \begin{align*} \mathrm I_{\max}^\varepsilon (L:L^c)_{\Omega} \le \mathrm O \big( \log (|L|\log(d))+\log(1/\varepsilon)\big), \end{align*} where L|L| represents the number of sites in LL, dd is the dimension of a site and Imaxε(L:Lc)Ω \mathrm I_{\max}^\varepsilon (L:L^c)_{\Omega} represents the ε\varepsilon-\emph{smoothed maximum mutual information} with respect to the L:LcL:L^c partition in Ω\Omega. From this bound we then conclude I(L:Lc)ΩO(log(Llog(d)))\mathrm I (L:L^c)_\Omega \le \mathrm O\big(\log(|L|\log(d))\big) -- an area law for the mutual information in 1D systems with a logarithmic correction. In addition, we show that Ω\Omega can be approximated in trace norm up to ε\varepsilon with a state of Schmidt rank of at most poly(L/ε)\mathrm{poly}(|L|/\varepsilon), leading to a good MPO approximation for Ω\Omega with polynomial bond dimension. Similar corollaries are derived for the mutual information of 2D frustration-free and locally-gapped local Hamiltonians.

Keywords

Cite

@article{arxiv.2310.19028,
  title  = {Area laws and tensor networks for maximally mixed ground states},
  author = {Itai Arad and Raz Firanko and Rahul Jain},
  journal= {arXiv preprint arXiv:2310.19028},
  year   = {2025}
}

Comments

38 pages, version 2