DESCRIPTION
Data set used to learn a representation of the reactive potential energy surface is saved in a compressed numpy binary file (.npz). The ab initio properties are obtained using MOLPRO (2019) at the CASPT2/cc-pVDZ level of theory.
The .npz file contains a python dictionary with seven numpy arrays:
N: Number of atoms in structure (num_data,)
R: Cartesian Coordinates of atoms (in Angstrom [A]), (num_data, N, 3)
Q: Total charge (in elementary charges [e]), (num_data,)
D: Dipole moment vector (in elementary charges times Angstrom [eA]), (num_data, 3)
E: Potential energy with respect to free atoms (in electronvolt [eV]), (num_data,)
F: Forces acting on atoms (in electronvolt per Angstrom [eV/A]), (num_data, N, 3)
Z: Atomic number of atoms, (num_data, N)
ACCESS DATA SET:
The data sets can be accessed using python:
>>> data = np.load("dataset.npz")
The different keywords of the python dictionary can be listed using
>>> data.files
>>>['N', 'E', 'Q', 'D', 'Z', 'R', 'F']
and the individual entries can be loaded using the appropriate
keyword, e.g. for the energy
>>> energies = data["E"]
Simulations were carried out with the pyCHARMM API linked to c47a2.
The PhysNet codes are available at https://github.com/MMunibas/PhysNet