Beyond Calabrese-Cardy Scaling: Exceptional-Point Sensitivity from the de Sitter RT Surface
Kuang-Hung Chou
Abstract
Entanglement entropy at one-dimensional criticality typically follows the Calabrese-Cardy scaling. In non-Hermitian critical chains near exceptional points, however, we show that the biorthogonal entropy of a finite system retains an additional sensitivity to a small energy gap \(Δ\) even when \(Δ< 1/L\). On top of the usual Calabrese-Cardy term, we find an interval-independent contribution \(S res=(ΔL)\), visible as a vertical offset and detectable even for a one-site subsystem. This behavior has no Hermitian analogue: a sub-finite-size gap is effectively invisible to entanglement in unitary critical chains, whereas here the entropy continues to resolve such a gap through its dependence on \(ΔL\). We interpret the result within the de Sitter geometry generated by non-unitary continuous multiscale entanglement renormalization: because the dS extremal surface reaches the IR endpoint, entanglement necessarily retains the endpoint contribution. On a finite ring, a regular circuit cannot terminate at a one-site product state and instead leaves an entangled two-site IR state. Computing the entanglement of the IR state recovers the same \((ΔL)\) term, identifying this additional long-range entanglement as the residual entropy left after finite-depth disentangling.
Create a lesson
Related papers
Trading Circuit Depth for Pulse Sparsity in Chromatic Dynamical Decoupling
Amy F. Brown, Daniel A. Lidar
Optimal spectrum estimation
Ainesh Bakshi, Apoorv Vikram Singh, Xinyu Tan
Non-Abelian sheaf quantum LDPC codes: good and magical
Zimu Li, Fuchuan Wei, Zhengyi Han et al.
Learning and interpreting policies for simultaneous entanglement requests in quantum networks
Leon Rode, Sumeet Khatri, Supartha Podder
Sharp universal death of entanglement threshold for Pauli Hamiltonians
Bobak T. Kiani
Proper Agnostic Learning of Matrix Product States and Tree Tensor Networks
Constantin Cedillo Vayson de Pradenne, Jordan Cotler