MateriApps
A Portal Site of Materials Science Simulation

now 326 Apps

Inquiry / Application Request
Detailed search

peps-torch

  • Openness:3 ★★★
  • Document quality:2 ★★☆

peps-torch is a python library for calculation of quantum many-body problems on two dimensional lattices. Variational principles calculation is used with an infinite PEPS (iPEPS) as the trial wave function. Therefore, the ground state is obtained in the form of the element tensor of the iPEPS.  The energy of the trial state is estimated by the corner transfer matrix method (CTM), and its gradient with respect to the element tensor is computed through automatic differentiation provided by pytorch.  Functions/classes for exploiting the system’s symmetry are provided for reducing the computational cost if possible. While general models and lattices are not supported, many examples of stand-alone codes would make it relatively easy for users to write their own codes to suit their needs. pytorch is required.

Information
Last Update:2022/11/07

Official site

https://github.com/jurajHasik/peps-torch

License

MIT License

Availability

PyTorch
(optional) YAST, scipy 1.3
Core Developers

by Juraj Hasik, Glen Bigan Mbeng
with contributions by Wei-Lin Tu, Seydou-Samba Diop, Sen Niu, Yi Xi
Target substance/model

quantum spin models
arbitrary rectangular unit cells
Abelian symmetric tensors
real/complex tensors
Physical quantities that can be computed

Stand-alone examples for quantum spin models produce various quantities including the energy and the one-site observables (such as the magnetization).
Methodology

infinite projective entangle pair state (iPEPS)
cornar transfer matrix method (CTM)
variational method
automatic differentiation
Parallelization

GPGPU parallelization via pytorch
Related App
Related keywords
TOP