mace-screening-and-relaxation

You can use MACE-based rapid screening to relax or evaluate structure batches, pruning low-value candidates before committing to DFT. It applies mace_relax_batch for geometry relaxation and mace_sp_batch for energy-only ranking to shortlist structures.

Pre-DFT screening with MACE allows you to process a slab and adsorbate candidate pool quickly. By running batch relaxations first, you generate concise handoff evidence and avoid unnecessary VASP compute cost on unsuitable structures.

Reproducible handoffs require explicit control of model, head, dispersion, and relax_lattice options. Keeping dispersion settings consistent is essential for accurate adsorption-energy comparisons across your batched structure outputs.

Yes, MACE batch screening enforces separate input and output trees. You must place the output_root outside the input_dir, ensuring batch_state_rel and batch_summary_rel artifacts are cleanly staged for downstream analysis.

Explicit model and dispersion choices are required to guarantee reproducible handoffs to downstream DFT workflows. Consistent dispersion settings are specifically needed for reliable adsorption-energy calculations across pruned structure batches.

Yes, you can use mace_sp_batch to perform single-point energy evaluations for rapid ranking instead of full relaxation. This allows fast energy-based pruning of structure batches before sending selected candidates to DFT.