arXiv:2610.11235
Markov length can diverge in systems whose universal physics is spatially Markovian
Yu-Hsueh Chen, Tarun Grover
Abstract
Conditional mutual information (CMI) probes spatial non-Markovianity and plays a central role in a recently developed framework for defining mixed-state phases based on local reversibility. This framework sharply distinguishes exponentially decaying from algebraically decaying CMI, since the latter obstructs equivalence to a state with finite Markov length. Here we show that states that flow to the same renormalization-group (RG) fixed point can nevertheless exhibit qualitatively different CMI. Starting from a critical Gibbs state of a finite-range local classical Hamiltonian, whose CMI vanishes beyond the interaction range, we show that an RG-irrelevant perturbation that breaks detailed balance generically generates algebraically decaying CMI. Crucially, these power-law tails are cutoff-suppressed: although they lead to an infinite Markov length at every fixed lattice cutoff, they vanish as a positive power of the cutoff in the continuum limit. We construct solvable models that exhibit this phenomenon and predict that the Ising-symmetric critical point of Toom's cellular automata generically exhibits such behavior. We also find analogous cutoff-suppressed CMI in the high-temperature paramagnetic phase of the long-range Ising paramagnet, in contrast with the genuine power-law CMI at the finite-temperature critical point in this same model that survives the continuum limit. We also derive a general result that in one spatial dimension, CMI decays faster than the inverse square of the buffer length if and only if the long-distance physics is Markovian.
AI Summary
The paper asks whether divergent Markov length necessarily indicates non-Markovian universal physics. It distinguishes continuum-surviving algebraic conditional mutual information (CMI) from cutoff-suppressed tails: both imply $\xi_M=\infty$ at fixed lattice spacing, but the latter vanish in the continuum limit. This separates RG equivalence from mixed-state phase equivalence based on locally reversible channels. For classical Langevin dynamics with a critical finite-range Gibbs steady state, a quadratic CMI expansion combined with the McLennan–Zubarev steady-state correction gives a susceptibility formula. Scaling arguments predict that a generic irrelevant detailed-balance-breaking perturbation with eigenvalue $\omega<0$ generates $I(A:C|B)\propto(\xi_{\rm cross}/l_B)^{2|\omega|}\Phi(l_B/l)$. Gaussian precision-kernel calculations provide explicit realizations and crossover exponents. A solvable two-field, two-dimensional Toom-like model yields $I\sim(\xi_{\rm cross}/l_B)^2(l/l_B)^4$, supported numerically on a $128\times128$ lattice. For Toom’s Ising-symmetric critical point, the predicted cutoff-suppressed tail uses an estimated $\omega\approx-1.3\pm0.2$. For translation-invariant one-dimensional states, an interval-entropy expansion relates CMI decay faster than $l_B^{-2}$ to a Markovian scaling limit. The long-range Ising model illustrates the distinction: high-temperature perturbation theory gives exponent $2(1+\sigma)>2$, whereas its critical Gaussian regime gives $I\simeq\sigma^2(l/l_B)^2/8$. The analysis assumes UV-finite CMI; the one-dimensional equivalence also assumes a controlled entropy expansion. Generic algebraic tails and the Toom prediction rely on scaling and nonzero operator-overlap arguments, rather than a general proof. The Langevin susceptibility is leading-order perturbation theory, and the critical Ising calculation is restricted to the Gaussian regime $0<\sigma<d/2$.
Central claim may not hold: The generic production of algebraic CMI is supported by a perturbative scaling argument and solvable examples, but remains conditional on projected critical overlaps, scaling-function asymptotics, and control beyond leading perturbative order. The general one-dimensional 'if and only if' statement lacks a justification connecting its small-interval Taylor limit to the fixed-ratio continuum limit, particularly for the converse implication.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 5 pages + appendices