arXiv:2609.20753
Quantum Entropy Contraction and Factorization from Hypercontractivity
Li Gao, Lijun Wang
Abstract
We prove that hypercontractivity implies entropy contraction for a single quantum channel, without a detailed balance condition. For primitive quantum Markov semigroups that are KMS-symmetric with respect to a faithful invariant state \(σ\), we obtain the modified Log-Sobolev bound $α_1\geq \fracλ{(2+\log\|σ^{-1}\|_\infty)}$ where $λ$ is the spectral gap. This removes the assumption of \(L_p\)-regularity for the comparison through the log-Sobolev constant. As an application, we show that the hypercontractivity of an average of two conditional expectations implies the approximate tensorization of relative entropy.
AI Summary
The paper studies when weighted noncommutative hypercontractivity yields relative-entropy contraction, and how contraction relates to approximate tensorization for conditional expectations. It works in finite dimensions with a faithful invariant state $\sigma$. For a single $\sigma$-preserving channel, $\|\Phi\|_{p\to q,\sigma}\leq1$ implies $D(\Phi_*\rho\|\sigma)\leq(p/q)D(\rho\|\sigma)$ without detailed balance. This implication follows from existing results through KMS duality; the proof interpolates the adjoint norm estimate toward $L_1$ and differentiates. A weighted centering argument combines hyperboundedness with strict centered $L_2$ contraction to produce a defect-free hypercontractive estimate. For primitive quantum Markov semigroups with positive KMS coercivity $\lambda$, the resulting bound is $\alpha_1\geq\alpha_H/2\geq\lambda/[2+\log\|\sigma^{-1}\|_\infty]$, where entropy decays at rate $2\alpha_1$. Under KMS symmetry, $\lambda$ is the spectral gap. This removes the $L_p$-regularity assumption previously needed for the comparison to modified log-Sobolev inequalities. A separate bound uses one hyperbounded endpoint, analytic interpolation, and the initial slope of a local hypercontractive curve. For two state-preserving conditional expectations, entropy contraction of their average with coefficient $\eta<1$, relative to the intersection expectation, implies approximate tensorization with constant $3/[2(1-\eta)]$. The proof controls the averaging loss using the quantum Jensen–Shannon metric; the implication also holds completely. Limitations include the smallest-eigenvalue dependence, nonoptimal tracial bounds, and the scalar-intersection restriction for the stated hypercontractivity-to-tensorization corollary. Primitivity alone provides only delayed estimates: a qubit example has delayed entropy contraction but zero modified log-Sobolev constant.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 15 pages; comments are welcome