arXiv:2609.33815
Quantum data processing equality
Tai-Hsuan Yang
Abstract
The quantum data processing inequality asserts that quantum relative entropy is monotonically non-increasing under quantum channels. Here, we promote this fundamental inequality to an exact equality via boundary and surface Poisson integral representations across a complex strip. The relative-entropy loss is resolved into two independent, nonnegative physical mechanisms: an operational Petz recovery component governing state reconstructibility, and a boundary term measuring the mismatch of relative modular dynamics. We further express the loss as the measured relative entropy between the original state and its averaged Petz recovery state plus a nonnegative variational remainder. For the Belavkin--Staszewski (BS) loss, analogous integral representations yield a decomposition into negative log Uhlmann fidelity with an averaged reconstruction plus a nonnegative variational remainder. Furthermore, we establish recovery bounds at any prescribed modular parameter, including the canonical unrotated Petz map, with optimal square-root scaling in the relative entropy loss. The prefactors depend on distinct-eigenvalue counts or logarithmic spectral quantities, improving upon the inverse-power dependence on small eigenvalues in previous bounds.
AI Summary
The paper studies whether relative-entropy data processing admits an exact nonnegative remainder representation, and how entropy loss controls recovery at a prescribed Petz rotation rather than only after averaging. Its setting is faithful finite-dimensional states and arbitrary quantum channels. For $\Delta=D(\rho\|\sigma)-D(\Phi(\rho)\|\Phi(\sigma))$, it derives an exact two-boundary Poisson integral on a complex strip. One boundary measures relative-modular-flow mismatch; the other controls rotated Petz recovery error. Both include nonnegative Schwarz defects for general channels. Green’s identity gives an equivalent weighted surface integral of the squared derivative of an analytic matrix family, plus boundary defects. Tilting this family by a positive test operator yields $\Delta=D_{\mathrm M}(\rho\|\tau)+\inf_{\omega>0}\mathcal P(\omega)$, with $\tau$ the standard averaged rotated-Petz reconstruction and $\mathcal P\geq0$. Finite-frequency evaluation inequalities convert the integrated errors into prescribed-rotation bounds: $\|\rho-\mathcal R_s(\Phi(\rho))\|_1\leq2\sqrt{C_+(s)\Delta}$. Here $C_+(s)=\cosh^2(\pi s)\min\{2(r_\rho+r_{\Phi(\sigma)}),C_{\rm eval}\}$, where $r_a$ counts distinct eigenvalues and $C_{\rm eval}\leq16(1+\|\log\sigma\|_\infty+\|\log\Phi(\sigma)\|_\infty)$. This includes the unrotated Petz map; a qubit erasure example establishes optimality of the square-root exponent, not the prefactors. For conditional mutual information, the Schwarz defects vanish, giving exact modular-flow/reconstruction identities. Parallel Belavkin–Staszewski identities produce a fidelity bound with an explicit variational remainder, using a generally nonlinear, subnormalized reconstruction. Extensions to non-faithful states, infinite dimensions, and von Neumann algebras are left open. The pointwise constants retain spectral-count or logarithmic spectral dependence.
The results presented in this work were obtained using GPT-5.6 Sol Pro and GPT-6 Astra Pro.
Error found: The abstract omits the faithful, finite-dimensional scope of the proved identities, while extensions beyond that scope are explicitly left to future work. The arbitrary-channel setup uses logarithms and inverse powers of output states that can be singular even for faithful inputs, requiring an unstated restriction to their common output support.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 25 pages, 1 table, including appendices, improved exposition