Conformal Properties of Hyperinvariant Tensor Networks

Authors:
  • MSMatthew Steinberg
    Matthew Steinberg
  • JPJavier Prior
    Javier Prior

Abstract
Since the discovery of the famous Ryu-Takayanagi formula, the AdS/CFT correspondence has seen a surge in interest from the quantum-information community. Tensor networks have emerged in parallel as a useful tool for simulating certain aspects of this correspondence. Two of these ans\"atze, the so-called "perfect tensor networks" (or HaPPY) and the "multiscale entanglement renormalization ansatz" (or MERA), have become paradigms of a field, realizing distinct but related aspects of AdS/CFT. Recently, a tensor-network ansatz was introduced which combines the advantages of both of these ans\"atze. These networks, known as "hyperinvariant tensor networks", have not been demonstrated as a viable tensor-network class due to several practical considerations, most of which are related to issues of boundary quasiperiodicity which arise in tensor networks based on regular tessellations of hyperbolic manifolds. In this paper, we discuss these challenges with the aim of explaining current difficulties on more technical grounds; in the process, we propose two new tensor decompositions for hMERA that include antisymmetric tensors, demonstrating a potential flexibility for describing physical systems in the network. Additionally, we provide numerical evidence which suggests that the Ising CFT could be reliably captured, provided that modifications are made to the current tensor decompositions in both the {7,3}- and {5,4}-parameterizations. We conclude by analyzing the current tensor decompositions in hyperinvariant tensor networks. This analysis is anticipated to aid in the simulation of conformal field theories in tensor networks which are based on regular tessellations of negative-curvature hyperbolic manifolds.