surface-and-termination-screening

name	surface-and-termination-screening
description	Use this skill for turning one relaxed bulk reference into a controlled slab/termination screening set, including slab generation, freezing policy, optional lateral expansion, and standardized VASP ranking runs.

surface-and-termination-screening

Overview

Use this skill to screen surface terminations without quietly changing slab geometry, fixing policy, or execution settings between candidates.

Quick Start

1. Start from one accepted relaxed bulk reference.
2. Generate all relevant slab terminations with one fixed thickness/vacuum and orthogonality contract.
3. Audit the exposed layers and surface-atom coordination of every termination before selecting survivors.
4. For adsorption handoff, prefer orthogonal=true unless a non-orthogonal native surface cell is explicitly required.
5. Apply one explicit freezing rule, then rank the survivors with one uniform slab relax or static batch.

Allowed tools

• build_slab
• fix_atoms_by_layers
• fix_atoms_by_height
• fix_atoms_by_indices
• supercell
• vasp_prepare
• remote_submission
• remote_submission_batch
• execute

Workflow

1. Enumerate the termination family

• Use build_slab once per Miller index and keep slab thickness, vacuum, orthogonality, and reduction policy fixed during the screen.
• Do not promote termination_index=0 just because it is first. Check each emitted termination for exposed species/layer sequence, stoichiometry, polarity or dipole concerns, duplicate geometry, and whether it matches the intended surface chemistry.
• After cutting the slab, identify the top and bottom surface layers with one consistent geometric rule and numerically compare their coordination distributions across the termination family. Unless the user requests a particular reactive, defective, or under-coordinated termination, prefer candidates whose surface atoms retain relatively higher coordination and fewer severely under-coordinated sites.
• Treat a large population of coordination-number-one (CN=1) surface atoms as a dangling-atom warning. Do not promote that termination without checking the neighbor criterion, slab thickness, cut/orientation, polarity, and visual geometry; normally reject or deprioritize it because such surfaces are empirically associated with high surface energy. This is a screening heuristic, not a substitute for a calculated surface energy.
• When the survivor will be used for adsorption placement, set orthogonal=true during slab generation by default. Use orthogonal=false only for a stated scientific reason and keep that setting fixed across the termination family.
• If the task is surface-energy oriented, surface get_symmetry_slab=true explicitly.

2. Apply one freezing strategy

• Use layer-based freezing by default for simple slabs.
• Switch to height- or index-based fixing only when the slab geometry or later adsorbate bookkeeping requires it.
• If lateral coverage needs more spacing, apply supercell after the fixing policy is understood.

3. Inspect before paying for DFT

• Use the structure-visual-inspection skill script when termination identity or vacuum/fixing mistakes are visually ambiguous.
• Inspect enough representative terminations to verify the intended exposed layer, vacuum direction, and any low-coordination warning before preparing relax inputs. Use visual inspection only as corroboration; preserve the numerical atom indices and coordination evidence behind any dangling-atom claim.
• Do not send obviously malformed or duplicated slabs into the VASP batch.

4. Run one controlled ranking stage

• Use one consistent vasp_prepare preset/regime across the termination set.
• Dispatch as one batch and parse convergence/energy with a focused pymatgen parser before selecting survivors.
• When ranking by total energy, use E0 as the default energy field from structured parser output.

Method-critical defaults

• Keep thickness, vacuum, k-point density, spin treatment, and dispersion policy fixed across the termination screen unless the workflow explicitly studies those variables.
• Use one stated neighbor algorithm/cutoff and one stated surface-layer definition for every candidate. Report coordination distributions by element and surface side; do not compare coordination numbers produced by different heuristics as if they were equivalent.
• Termination provenance review is not optional for slab modeling; report what terminations were generated, which were retained or rejected, and why. Treat the review as an evidence-backed selection record, not an absolute proof of surface identity from one structure file.
• For adsorption-ready survivors, report the orthogonal setting and prefer orthogonal=true.
• Do not compare slabs if some candidates were relaxed with cell degrees of freedom and others were not.
• Preserve selective dynamics through any supercell expansion and report that the mask survived.

Output Contract

Return:

• termination set or retained termination IDs
• termination provenance summary, including retained/rejected rationale and remaining uncertainty
• surface-coordination audit, including the neighbor criterion, surface-layer definition, per-element coordination summary, and count/indices of any CN=1 surface atoms
• orthogonal setting used for the family
• representative slab artifact path(s)
• freezing policy used
• batch ranking/parser summary path

References

• Hand off accepted slabs to adsorption-screening rather than combining slab enumeration and adsorption placement in one campaign.