Parts | Days | Selection | Search | Updates | Downloads | Help
AKjDPG: Arbeitskreis junge DPG
AKjDPG 3: Hacky Hour I (joint session AGI/SOE/AKjDPG)
AKjDPG 3.4: Talk
Wednesday, March 20, 2024, 11:00–11:30, MAR 0.011
Quantum Many Body Simulations with TeNPy — •Johannes Hauschild — Technical University Munich, Germany
Matrix product state (MPS) based algorithms like the density matrix renormalization group (DMRG) are established as the state-of-the-art method for simulations of quantum many body systems in 1D, for example Heisenberg and Hubbard type models. In fact, MPS are so successfull that they are routinely used for 2D systems as well, by mapping thin long cylinder geometries to 1D. Generalizations of MPS to natively 2D tensor network states in the form of PEPS or isoTNS provide an alternative route for competitive results, especially for cases where quantum monte carlo methods suffer from the sign problem.
I will present version 1.0 of TeNPy, the “Tensor Network Python” package that I started developing half a decade ago. The major goal has been to make MPS and tensor network simulations accessible not only to experts of the field but also new users, by excellent documentation, and balancing speed of the code with flexibility to define new models and algorithms. Indeed, TeNPy has been accepted well by the community with over 250 papers acknowledging its use and code contributions from various groups. After a (very) brief introduction to the main ideas behind the algorithms, I will show small examples for typical use cases of TeNPy. I will further discuss our ongoing efforts and first benchmarks to adapt TeNPy and the implemented algorithms to GPU-based calculations, and how we plan to incorporate the conservation of non-abelian symmetries.
Keywords: Tensor Networks; Quantum Many Body Systems; Matrix Product States; Density Matrix Renormalization Group