Project: A high-performance tensor network library for classical and quantum many body physics

2020-01-01 – 2023-12-31

Abstract

The goal of this project is to develop open source libraries to facilitate the use of tensors and tensor networks in the context of many body physics, that supports the latest trends in high performance computing such as GPU computing and automatic differentiation Tensor network algorithms have proven useful in the context of both quantum many body phyics and field theories and classical statistical systems (statistical mechanics) They all require the contraction of networks of tensors (multi-dimensional arrays), which themselves have typically some kind of block sparsity due to symmetries in the physical system Our key goals are supporting general symmetries (eg non- abelian, fermions) and providing state of the art efficiency (multithreaded and with GPU support) in combination with user friendly and well documented code Here we choose the programming language Julia,  motivated by its proven ability to produce efficient code, while at the same time being a high level, scientifically oriented language that is easy to get started with and with much functionality available In particular, the paradigm of Julia is that there is no need to rewrite computationally intensive parts of the code in a low level language, and as such, the step from user to contributor is small, a goal that we also try to accomplish with our libraries