arXiv:2609.34699
Log-Euclidean Rényi Conditional Mutual Information
Roberto Rubboli, Amir Arqand, Mark M. Wilde
Abstract
Quantum conditional mutual information (QCMI) is a fundamental measure of conditional independence, with central applications throughout quantum information theory and many-body physics. Despite its importance, finding a fully quantum Rényi generalization that satisfies all desirable structural properties has remained an open question. In this work, we introduce a quantum Rényi conditional mutual information possessing these desirable properties: it is monotone under local quantum channels, additive under tensor products, monotone in the Rényi parameter, and converges to the QCMI in the limit $α\to 1$. We establish bounds on this quantity in terms of relative entropies to the Petz-recovered state, showing that it provides a distinct measure of approximate conditional independence. As a consequence, we derive recovery-based upper and lower bounds on the QCMI in terms of quantum generalizations of the Kullback-Leibler divergence, improving upon the best known results in certain regimes. Finally, we establish bounds in terms of relative entropies to the sets of Markov and separable states and derive Rényi chain rules, which appear to be novel even in the classical setting.
AI Summary
Quantum conditional mutual information (QCMI) characterizes quantum Markov chains, but Rényi extensions face competing requirements of data processing, additivity and parameter monotonicity. This paper introduces a log-Euclidean version, $I_\alpha^\flat(A:B|C)_\rho=(\alpha-1)^{-1}\log\operatorname{Tr}\exp[\alpha\log\rho_{ABC}+(1-\alpha)(\log\rho_{AC}+\log\rho_{BC}-\log\rho_C)]$, and a variant optimized over auxiliary marginal states. The non-optimized quantity obeys data processing under channels on $A$ and $B$ for $0<\alpha<1$, is additive and increasing in $\alpha$ for all $\alpha>0$, and converges to QCMI at $\alpha=1$. Both variants vanish exactly on quantum Markov chains for $0<\alpha<1$. The optimized variant obeys data processing in that range and is additive for $\alpha\in[1/2,2]$. Gibbs variational formulas reduce these properties to ordinary relative entropy and QCMI; optimized additivity additionally uses convexity, minimax arguments and fixed-point optimality conditions. Multivariate trace inequalities yield lower and upper recovery bounds involving rotated Petz maps. At $\alpha\to1$, these bound QCMI between rotation-averaged divergences involving Keyl’s rate function and a complementary quantum KL extension. The lower bound strengthens the corresponding fidelity bound but is not universally ordered with the measured-relative-entropy bound. Further results include Rényi chain inequalities and bounds involving distances to Markov and separable states. Data processing is not established above $\alpha=1$; optimized additivity remains open below $1/2$. Recovery upper bounds assume full rank because extensions by approximation may be path dependent. The dual quantity fails data processing. Operational applications, including strong-converse problems and Rényi squashed entanglement, remain future directions.
ChatGPT was used for proofreading, mathematical and bibliographic checks, and assistance with some derivations
Central claim may not hold: The abstract leaves local-channel monotonicity unqualified, whereas the paper establishes the complete structural package only for 0<α≤1, not throughout the non-optimized quantity's domain α>0. The abstract does not mention that its recovery-based upper bound on QCMI requires a full-rank tripartite state, unlike the lower bound.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.