arXiv:2610.20078
Strong subadditivity of topological entropy on the torus: Rényi order and non-Abelian structure
Yi Liang
Abstract
We classify canonical Rényi strong subadditivity on an unpunctured torus for finite unitary modular categories. For all pure logical states and admissible region pairs, validity at 0<alpha<1 is equivalent to Abelianity; at alpha=1 it is universal. Each finite non-Abelian group G has a pure untwisted D(G) fixed-point ground state with an order-two violation under ordinary edge partial traces. A specified local-filter D(S_3) family changes reduced spectra; its fixed-cut deficit tends to the negative fixed-point value over a small size-independent interval in a controlled joint limit.
AI Summary
The paper asks when the canonical topological Rényi entropy on an unpunctured torus obeys spatial strong subadditivity for every pure logical state and admissible region pair. This functional records boundary topology and shared flux; it is not generally an ordinary reduced-density-matrix entropy. For finite unitary modular tensor categories, universal validity at each $0<\alpha<1$ is equivalent to Abelianity, while at $\alpha=1$ it holds universally. Every non-Abelian category admits violations at every finite $\alpha>1$. For a fixed Abelian category, high-order validity is instead equivalent to an equal-order entropic uncertainty bound for its modular Fourier transform; high-order violations therefore do not diagnose non-Abelianity. The negative results follow from strict curvature of the shared-flux contribution in annulus count when occupied sectors have unequal quantum dimensions. The positive results combine a subsurface-geometric reduction with Fourier uncertainty, including an established weighted theorem at order one. Ordinary edge partial traces reproduce violations in specified untwisted $D(G)$ fixed-point ground states for every finite non-Abelian group. For $D(S_3)$, one order-two deficit is $\log(16/29)$. An explicit local-filter family changes reduced spectra while satisfying $|\Delta_2-\log(16/29)|\le4N_xL e^{-\min(W,L)}$ on a small size-independent filter interval, with a separately established uniformly gapped local parent. The microscopic statements use specified states, refined edge tensor factors and cuts. Filtered persistence requires a joint large-system/separation limit; it does not establish whole-phase stability. The Rényi continuation is explicitly chosen, not inferred uniquely from integer replicas, and high-order thresholds for individual Abelian theories remain unresolved.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 20 pages, including appendices