scimate-ai/hpc-skills/hpc-gaussian

HPC Gaussian

Treat Gaussian as a staged quantum-chemistry workflow built around a valid input deck and explicit checkpoint policy.

Start

1. Read references/workflow-manual.md before creating or editing a Gaussian input file.
2. Read references/input-and-link0-matrix.md when mapping Link 0 directives, route keywords, title, charge, multiplicity, and geometry blocks.
3. Read references/geometry-charge-spin.md when the structure, charge, multiplicity, or fragment logic is uncertain.
4. Read references/method-basis-and-jobtype.md when choosing the method, basis set, and job type.
5. Read references/scf-opt-freq-and-solvation.md when choosing SCF, optimization, frequency, or solvent settings.
6. Read references/checkpoint-restart-and-artifacts.md when designing checkpoint, restart, formatted-checkpoint, or cube workflows.
7. Read references/cluster-execution-playbook.md when staging a Gaussian workflow for scheduler-backed cluster execution.
8. Read references/error-recovery.md when parsing, SCF, optimization, frequency, or restart behavior fails.

Additional References

Load these on demand:

• references/error-pattern-dictionary.md for structured SCF, geometry, and restart failure signatures

Reusable Templates

Use assets/templates/ when a concrete starting input is needed, especially:

• sp_water.gjf
• opt_freq_water.gjf
• gaussian-g16-slurm.sh

Guardrails

• Do not invent Gaussian keywords or Link 0 directives.
• Do not treat charge and multiplicity as cosmetic metadata.
• Do not run restart-sensitive workflows without an explicit checkpoint policy.
• Do not copy route sections from unrelated systems without checking method, basis, and job intent.

Outputs

Summarize:

• job type
• method and basis selection
• charge and multiplicity
• checkpoint or restart policy
• expected key outputs and next stage