Phonopy Skill (Jinzhe Zeng Group)

Overview

The Phonopy Skill is an instruction resource in the computational-chemistry-agent-skills collection, under analysis/phonopy/SKILL.md. It guides an assistant through phonon workflows built around phonopy, while separating displacement generation and analysis from the engine that calculates forces. It is not the upstream Phonopy package or an independently running agent. The instructions describe force-provider options including VASP, Quantum ESPRESSO, and ML force fields; they do not establish tested integrations.

The workflow starts with a valid structure and its unit-cell or primitive-cell context. Users specify the force backend, supercell, displacement amplitude, and desired outputs. Band calculations additionally require a path definition and sampling choices; DOS and thermal calculations require mesh settings, with temperature settings for thermal properties. The Skill directs the assistant to generate displaced supercells, preserve their correspondence with force files, verify dataset completeness, and assemble FORCE_SETS or force constants. Expected deliverables include the displacement task layout, assembly status, requested analysis files, and a record of assumptions and unresolved decisions.

The instructions require both a valid structure and displaced-supercell forces or precomputed force constants, and direct the assistant to request missing inputs rather than fabricate them. Backend execution and cluster submission belong to separate workflows. Scientific limitations remain explicit: small supercells, inconsistent units, insufficient convergence, and ambiguous band paths can compromise interpretation. Relevant long-range or non-analytic correction settings must also be reported. These are documented workflow requirements, not evidence of validated scientific results.

Key Features

• Defines a backend-independent division of responsibility: the Skill organizes displacements and phonon analysis, while a separate provider evaluates forces.
• Collects structure, supercell, displacement, primitive-cell, symmetry, mesh, and band-path settings appropriate to the requested analysis.
• Requires traceable displacement-to-force-file mapping and checks force-data completeness before force-constant assembly.
• Guides assembly of `FORCE_SETS` or force constants and subsequent band-structure, DOS, and thermal-property analysis.
• Requires explicit reporting of assumptions, unresolved scientific choices, convergence caveats, and relevant non-analytic correction settings.
• Sets boundaries against fabricated force data and direct cluster submission, with handoff guidance for backend and submission workflows.

Use Cases

• Suggested evaluation: organize a finite-displacement band-structure workflow for a supplied periodic structure, with force calculations delegated to a selected DFT backend.
• Suggested evaluation: assess whether an ML force-provider workflow returns complete, consistently mapped forces suitable for phonon DOS and thermal analysis.
• Suggested evaluation: audit a phonon task for missing force files, unclear band-path conventions, inconsistent units, and unresolved convergence settings before interpreting outputs.

How to Use

1. Read the named Skill in the collection. Treat it as workflow instructions; the stated prerequisites are phonopy and a separate force-provider workflow.
2. Supply a real structure file and cell context. Identify whether displaced-supercell forces or precomputed force constants are available, and resolve missing required inputs before proceeding.
3. Specify the analysis objective, backend, supercell, and displacement amplitude. Make primitive-cell and symmetry choices explicit; add a band path or mesh and temperature settings as appropriate.
4. Follow the Skill’s displacement workflow and hand displaced structures to the chosen backend. Keep each displacement linked to its force output; use separate backend or submission instructions for execution.
5. Check completeness and unit conventions before assembling FORCE_SETS or force constants. Request the desired analysis outputs, then inspect assumptions, convergence caveats, imaginary modes, and relevant long-range settings. Treat this inspection as evaluation, not proof of scientific validity.