SCOcopilot

Spin-Crossover copilot with Equivariant Graph Convolutional Neural Networks

This library is based on the work published in J. Chem. Theory Comput. 21 3913–3921 (2025). Please ensure that the following external libraries are installed beforehand:

• pyTorch
• NumPy
• ASE
• RDKit
• PyG
• Matplotlib

Query

The library accepts SMILES strings as follows:

from scocopilot import query

spin_crossover = query()

spin_crossover("[Co](NCCCC)(NCCCC)(SCCC)(SCCC)")

Alternatively, a file with Cartesian coordinates may also serve as input:

from scocopilot import query

spin_crossover = query()

spin_crossover("path/to/file.xyz")

Default units are eV, but can be changed with the option units="kJ/mol" or units="kcal/mol". It also is possible to deactivate the verbose environment:

from scocopilot import query

spin_crossover = query()

spin_gap, standard_deviation = spin_crossover("path/to/file.xyz", units="kJ/mol", verbose=False)

Changing the device from CPU (default) to GPU is done during the creation of the object:

from scocopilot import query

spin_crossover = query(device="cpu")