mat-disorder

To generate ordered crystal structures from disordered CIF files with partial occupancies, use an Ewald-energy based ordering transformation with ALGO_FAST. This samples stoichiometry-consistent orderings and ranks candidates by electrostatic repulsion to output clean configurations.

Yes, you can use pymatgen to convert fractional occupancy sites for cluster expansion training. The process generates ordered candidates from disordered inputs and outputs ASE structures suitable for CE and MLIP training pipeline preparation.

The best way to sample ordered candidates from partially occupied structures for phase stability studies is using an Ewald-energy based solver. It prioritizes lower electrostatic repulsion configurations via ALGO_FAST while sampling beyond the absolute minimum to provide diverse candidates.

Yes, this ordering workflow requires a specific target atom count for supercell expansion. The transformation tunes supercell expansion to reach a target number of atoms per structure, ensuring the generated ordered configurations match your desired calculation scale.

Disordered crystal structures need stoichiometry-consistent ordering before ground-state property calculations because partial occupancies create fractional sites unsuitable for direct energy computation. Generating ordered candidates preserves the original stoichiometry while providing clean atomic configurations for accurate relaxation.

CIF or POSCAR-like input files preserve occupancy information for generating ordered atomic configurations. These formats maintain fractional site occupancy data required by the Ewald-energy ordering transformation to sample and rank stoichiometry-consistent ordered structures.