catgo-gibbs-pipeline — adsorption/reaction free energies & overpotential

TL;DR: G = E_DFT + ZPE − TΔS. Every species in a ΔG needs geo_opt and freq. Per-species pipeline: geo_opt → freq → Gibbs; assemble ΔG → η.

When to use

Any study whose target is a free energy — overpotential η, free-energy diagram, ΔG of adsorption/reaction, limiting potential (HER/ORR/OER/CO2RR/NRR, adsorption thermo). Raw geo_opt energies are NOT enough: add ZPE + TΔS from frequencies.

Core rules (the why)

• Free energy needs freq. Skipping freq gives wrong ΔG and wrong η.
• Per species, pipeline not barrier. Each species' freq fires as ITS geo_opt converges — don't wait for siblings, don't wait for a reminder.
• Convergence by force, not dE: FORCES: max atom < |EDIFFG|.
• CHE, not O2: G(H⁺+e⁻)=½G_H2; derive O/OH from H2O+H2; never compute O2.
• Gas refs cheap: Γ-only + vasp_gam on shared; slabs vasp_std + k-mesh on compute.

Bundled resources

• catgo freq-inputs — build a VASP freq input from a relaxed CONTCAR/POSCAR (adsorbate: fix surface, free adsorbate atoms; gas: all free). Runs from ANY directory: catgo freq-inputs --structure <CONTCAR> --out <dir> [--gas | --free-elements O,H]. Wraps scripts/build_freq_inputs.py (call the script directly only if catgo isn't on PATH; resolve it as <this skill base>/scripts/build_freq_inputs.py, never a hardcoded ~/.claude/... path).
• references/method.md — full thermo + freq setup + CHE/ORR/OER/HER pathway formulas + the η definition. Read for the exact equations.
• examples/orr-pt111.md — a worked ORR-on-Pt(111) instance (build freq, assemble η).

Flow

1. geo_opt each species (force-converged) → E.
2. catgo freq-inputs from each CONTCAR → submit freq → catgo freq --mode adsorbed|gas → ZPE + TΔS → G.
3. Assemble ΔG_i along the pathway → U_L = max ΔG_i/e → η = |1.23 − U_L|; free-energy diagram + ΔG table. (Building blocks: the vasp-freq, gibbs, her, oer, energy-diagram skills.)

In a CatGo campaign, wire this into plan.md and let the loop fire freq per species — see the catgo-campaign + catgo-campaign-loop skills.