arXiv:2609.20020
Corner entanglement scaling with projected entangled pair states
Chloé Van Bastelaere, Rui-Zhen Huang, Laurens Vanderstraeten
Abstract
Entanglement scaling provides a powerful probe of universal properties at quantum critical points. In two dimensions, contributions originating from a corner-shaped bipartition exhibit a universal scaling, which is determined by the underlying conformal field theory. We develop a method to extract this corner entanglement entropy from projected entangled pair states directly in the thermodynamic limit. When applied to models at a quantum critical point, we show that the corner contribution exhibits scaling with the effective correlation length, in agreement with the hypothesis of finite-entanglement scaling. Our results for the corner coefficients are consistent with other methods, demonstrating the efficiency of our method for diagnosing strongly-correlated quantum critical points in two dimensions.
AI Summary
The paper addresses how to extract universal corner contributions to entanglement entropy from infinite PEPS, without fitting or subtracting an extensive area-law term. It focuses on the second Rényi entropy for a $\pi/2$ corner and tests whether its subleading contribution follows $S_{c,2}=c_2\log\xi+\mathrm{const}$ under finite-entanglement scaling. The construction combines the PEPS bulk–boundary correspondence with boundary MPS and corner-shaped contraction environments. Linear transfer-channel eigenvalues determine the area-law contribution, while a separate corner contraction gives $S_{c,2}$ directly in the thermodynamic limit. For each optimized PEPS, the boundary-MPS bond dimension $\chi$ is varied to obtain the effective correlation length $\xi$ and fit the logarithmic slope. At criticality, the reported coefficients are $c_2=-0.0050(1)$ for the transverse-field Ising model using $D=3$, $-0.00503(2)$ for a PEPS approximation to the three-dimensional classical Ising transfer-matrix fixed point, and $-0.012(2)$ for the bilayer XY model using $D=5$. These agree with existing estimates. For a variational chiral spin-liquid PEPS, the corner contribution instead saturates despite increasing correlation length. The extraction is sensitive to fitting windows: small-$\chi$ environments are inaccurate, while fixed-$D$ correlation-length saturation clusters large-$\chi$ data and destabilizes slopes. Reported uncertainties characterize fitting rather than full physical systematic errors. Demonstrations are restricted to $n=2$ and right-angle corners; extending to higher Rényi indices entails exponentially growing contraction tensors.
Error found: The critical-point benchmarks demonstrate correlation-length scaling by varying the contraction dimension at fixed PEPS bond dimension, without an increasing-bond-dimension study controlling its systematic accuracy. The quoted Ising corner-coefficient error bars are fit uncertainties, not demonstrated total uncertainties incorporating fitting-window and finite-bond-dimension effects.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.