adsorption-screening

name	adsorption-screening
description	Use this skill for adsorption-candidate generation, execution, and first-pass ranking when screening one adsorbate across one slab family with reproducible site provenance and thermochemistry-ready metadata.

adsorption-screening

Overview

Use this skill to build, run, and rank a controlled adsorption screen while preserving site labels and adsorbate atom indices for later thermochemistry.

Quick Start

1. Start from one termination-reviewed slab and one canonical adsorbate file.
2. Prefer an orthogonal adsorption slab; if the slab is non-orthogonal, record the reason before placement.
3. Enumerate sites first; keep the site JSON as the ledger.
4. Generate only the adsorption candidates you actually plan to relax.
5. Prepare and dispatch one consistent adsorption batch.
6. Rank with a task-specific pymatgen parser and keep ads_indices metadata for any later frequency correction work.

Allowed tools

• enumerate_adsorption_sites
• place_adsorbate
• generate_batch_adsorption_structures
• vasp_prepare
• remote_submission
• remote_submission_batch
• vaspkit_adsorbate_thermo_correction
• execute

Workflow

1. Preserve adsorption provenance

• Always enumerate sites before broad placement.
• Verify slab provenance before site enumeration: selected termination, termination provenance summary, and orthogonal setting should be known. If the task only provides an unlabeled slab, state that limitation and inspect or regenerate the slab family before quantitative screening.
• Treat generate_batch_adsorption_structures as a convenience wrapper over site enumeration plus placement, not as a separate scientific abstraction.

2. Build a controlled candidate set

• Use one adsorbate geometry source across the whole screen.
• Use an orthogonal, c-oriented slab for new adsorption model generation unless the user explicitly requests native-cell adsorption. Non-orthogonal adsorption cells are allowed only as an intentional exception, not as an unnoticed tool default.
• If only one site is under study, use place_adsorbate directly and keep the chosen site label explicit.
• Use the structure-visual-inspection skill script when collisions or orientation ambiguity need a visual check before DFT.

3. Run one comparable relaxation stage

• Prepare adsorption candidates with one consistent slab regime and method policy.
• Keep clean slab, gas reference, and adsorbed systems method-aligned if adsorption energies will be compared later.

4. Rank, then branch

• Use a focused pymatgen parser to detect failed or unconverged candidates before ranking.
• For adsorption-energy ranking, read E0 from structured parser output by default rather than other OUTCAR energy fields.
• Run vaspkit_adsorbate_thermo_correction only on the retained adsorption states that genuinely need thermal corrections.

Method-critical defaults

• Do not begin adsorption ranking from a slab whose termination provenance has not been reviewed or documented.
• For newly built adsorption slabs, the project preference is orthogonal=true; keep that choice consistent across clean slab, adsorbed structures, and any reference calculations derived from the slab.
• Do not compare adsorption energies if the clean slab or gas reference was prepared under a different spin, dispersion, or DFT+U contract.
• Preserve and report ads_indices; they are part of the downstream thermochemistry contract.
• Do not silently accept truncated candidate batches from the wrapper if the ranking depends on full site coverage.

Output Contract

Return:

• slab provenance: selected termination, provenance-review status or uncertainty, and orthogonal setting
• site JSON path
• generated adsorption structure root
• retained candidate list
• adsorption ranking/parser summary path
• adsorbate thermochemistry artifact path when used

References

• Use the primitive skills adsorption-site-screening and adsorbate-and-intermediate-generation when only the front half of the workflow is needed.