ever-works/orgs/computational-chemistry

Computational Chemistry

Compute energetics and electronic-structure properties at a level of theory matched to the question, and report results with the caveats the method demands.

Procedure

1. Define the property and accuracy needed. Relative energies, reaction barriers, geometries, or spectra each imply a different method and basis. Expensive is not automatically correct.
2. Choose the level of theory deliberately. Select the method (and basis set, where applicable) with justification, and state its known failure modes for this class of system.
3. Prepare the input geometry sensibly — reasonable starting structure, correct charge and spin state, and a clear conformer treatment.
4. Optimize and verify. Confirm geometry optimizations converged to genuine minima or transition states (frequency check where relevant), not saddle points mistaken for minima.
5. Check numerical convergence. Basis-set adequacy, grid and SCF convergence; a loosely converged calculation gives a confident wrong number.
6. Benchmark where possible. Compare against experimental data or a higher-level reference for a related system to bound the error.
7. Report relative, not just absolute, quantities. Energy differences are usually more reliable and more meaningful than raw totals; state the reference.
8. Document the method fully so another chemist could reproduce and judge it.

Quality bar

• Level of theory is justified and its limitations for the system are named.
• Optimizations are verified as true minima/transition states.
• Numerical convergence is checked and stated.
• Results are benchmarked or their error is otherwise bounded.

Output format

A results report: property and required accuracy, level of theory with justification, verified geometries, convergence and benchmark notes, and relative energies with their reference state.