QuCumber is an open-source Python package that implements neural-network quantum state reconstruction of many-body wavefunctions from measurement data such as magnetic spin projections, orbital occupation number. Given a training dataset of measurements, QuCumber discovers the most likely quantum state compatible with the measurements by finding the optimal set of parameters of a restricted Boltzmann machine (RBM).
.
AkaiKKR is a first-principles all-electron code package that calculates the electronic structure of condensed matters using the Green’s function method (KKR). It is based on the density functional theory and is applicable to a wide range of physical systems. It can be used to simulate not only periodic crystalline solids, but also used to calculate electronic structures of impurity systems and, by using the coherent potential approximation (CPA), random systems such as disordered alloys, mixed crystals, and spin-disordered systems.
An open-source application for molecular dynamics to simulate biopolymers such as proteins and nuclear acids. This application can perform high-speed molecular dynamics simulation by hybrid parallel computing maintaining high-accuracy energy conservation. This application also support high-speed calculation of long-range interaction based on the particle mesh Ewald method. The code is released under GPL lisense.
A general-purpose application for molecular dynamics simulation equipped with many tools. This package was originally developed for biomolecules (peptides, proteins, nuclear acids, etc.), and the current version can perform molecular dynamics simulation for various systems such as solutions, crystals, membranes, and so on. It supports several sampling methods and calculation of free energy. It also supports various computing environments including both serial and parallel computers.
An open-source application for ab initio quantum chemical calculation. This application performs electronic structure calculation of molecules by the Hartree-Fock, density functional, many-body perturbation, configuration interaction theories, and so on. Even though this application is freeware, it succeeds in maintaining high-quality and high-performance codes by active development, and has a number of world-wide users. It histrically shares core programs with GAMESS-UK.
A low-energy solver for a wide ranger of quantum lattice models (multi-orbital Hubbard model, Heisenberg model, Kondo-lattice model) by using variational Monte Carlo method. User can obtain high-accuracy wave functions for ground states of above models. Users flexibly choose the correlation factors in wavefunctions such as Gutzwiller, Jastrow, and doublon-holon binding factors and optimize more the ten thousand variational parameters. It is also possible to obtain the low-energy excited states by specifying the quantum number using the quantum number projection.
Application for performing first-principles simulations with an implicit solvent model. The code is released as a patch to VASP. The user can perform molecular dynamics as well as reaction analysis using e.g., nudged elastic band method.
RSPACE is a first-principles code package based on a real-space finite-difference pseudo-potential method. It computes electronic states with high-speed and high precision in aperiodic systems of surfaces, solid interfaces, clusters, nanostructures, and so forth. It provides large-scale computing for semiconductor devices of nanostructure surface and interface reactions, calculation of transport properties in semi-infinite boundary conditions, and a massively parallel computing using the space partitioning method.
Open-source software for building computational physics applications. Common C++ auxiliary modules required for various methods in computational physics such as the quantum Monte Carlo method are prepared. This software helps to build reusable codes and to reduce development time for complex computational science applications. It also supports parallel programming based on MPI or OpenMP.
Open source software for constructing the Allegro potential model based on E(3)-equivariant graph neural networks and using the potential model for molecular dynamics simulations. The code depends on NequIP and can be run in a similar manner. Allegro scales better than NequIP since it doesn’t rely on message passing and the architecture is strictly local with respect to atom-wise environments.