Hello-QM/catgo-LRG / server/catgo/workflow/skills/catgo-gibbs-pipeline
catgo-gibbs-pipeline
VerifiedCompute adsorption/reaction Gibbs free energies, free-energy diagrams, and electrochemical overpotentials (HER/ORR/OER/CO2RR/NRR) with VASP. The per-species pipeline is geo_opt → freq → Gibbs; CHE references; ΔG → η. Use whenever a study's target is a FREE energy (not raw DFT energy) — overpotential, free-energy diagram, ΔG of adsorption/reaction, limiting potential. Pairs with the catgo-campaign skill for orchestration.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.7 KB |
| examples/orr-pt111.md | markdown | 1.8 KB |
| references/method.md | markdown | 5.4 KB |
| scripts/build_freq_inputs.py | script | 4.7 KB |
references/method.md
View on GitHub# Gibbs / overpotential — method detail
Deep reference for the `catgo-gibbs-pipeline` skill. Read when you need the exact thermo,
freq setup, or pathway formulas.
## Free energy
G = E_DFT + ZPE − TΔS (+ thermal/pressure terms for gas-phase molecules). Raw geo_opt
energies are NOT free energies — ZPE and TΔS come from vibrational frequencies, so every
species entering a ΔG needs a freq calc on its relaxed geometry.
## Per-species pipeline (auto-advance, not a barrier)
For each species (adsorbate intermediate AND molecular reference):
1. geo_opt → relaxed E (force-converged: `FORCES: max atom` < |EDIFFG|, not just dE).
2. freq (from CONTCAR) → ZPE + TΔS.
3. Gibbs G.
Freq fires as THAT species' geo_opt converges — pipeline per species, don't wait for siblings.
## freq setup
Use **`catgo freq-inputs`** (wraps `scripts/build_freq_inputs.py`). Key choices (cheap + robust):
- **K = Γ only (1×1×1)** for ALL freq — freq needs only forces, and for an adsorbate the
surface is frozen, so a dense k-mesh is wasted. Γ ⇒ run on the gamma build `vasp_gam`.
- **No NCORE / NPAR** — IBRION=5 finite differences are most robust with no band
parallelization (default NCORE=1); NCORE>1 can corrupt the Hessian on some builds.
- **shared node, ~32 cores** — freq is small; don't take an exclusive 128-core node.
- **Adsorbates** — fix all surface atoms, free ONLY the adsorbate atoms (harmonic adsorbate
modes; surface phonons cancel in ΔG). `IBRION=5 NFREE=2 POTIM=0.015 NSW=1 EDIFF=1e-6`,
ISMEAR=1/SIGMA=0.2 (metal), ISPIN=2. → `catgo freq --mode adsorbed` → ZPE + TΔS_vib.
- **Molecules** — all atoms free, ISMEAR=0/SIGMA=0.05. → `catgo freq --mode gas` →
ZPE + thermal + rotational/translational entropy (ideal gas).
- Use real modes only; shift small/imaginary modes per the freq_cutoff convention
(~50 cm⁻¹).
## Gas-phase references (compute convention)
Small molecules (H2, H2O, O2, CO, …): Γ-point only + the gamma build `vasp_gam` on a
`shared` node (~32 cores) — never an exclusive 128-core node for a few-atom molecule.
Slabs: `vasp_std` + a k-mesh on `compute`.
- **H2O reference at 0.035 bar, NOT 1 bar.** In CHE the water reference is LIQUID; approximate it by gas-phase H2O at its 300 K equilibrium vapor pressure ≈ **0.035 bar** (`catgo freq --mode gas --P 0.035`). At 1 bar the translational entropy is ~0.086 eV too large → G(H2O) ~0.086 eV too high. Only H2O gets 0.035 bar; H2 stays 1 bar (CHE standard). It shifts every H2O-releasing/consuming step; whether it moves η depends on whether the limiting step involves H2O.
- **Match ISPIN between geo_opt and freq.** Closed-shell molecules (H2, H2O): ISPIN=1 for BOTH — ISPIN=2 can break H2 into two spin-polarized H atoms (dissociation / corrupt finite-diff Hessian → imaginary stretch). Open-shell (O2, NO, radicals): ISPIN=2. `catgo freq-inputs --gas --ispin <n>` to set it.
## CHE references (avoid O2)
Computational Hydrogen Electrode: G(H⁺ + e⁻) = ½ G_H2 at U = 0, pH 0. Each (H⁺ + e⁻)
transfer shifts a state by +eU. Derive O/OH chemical potentials from H2O + H2:
- μ_O = G_H2O − G_H2
- μ_OH = G_H2O − ½ G_H2
Do NOT compute O2 (triplet / over-binding error) — the whole point of CHE.
## Adsorbate solvation (don't skip for ORR/OER)
Vacuum *OH (and *OOH) are UNDER-bound — water H-bonding stabilizes them, so a vacuum free
energy makes ΔG(*OH) too high → step *O→*OH too flat → η too large. This is NOT a wrong
adsorption site: bare-DFT binding ENERGIES (ΔE) already match literature; the gap is in the
free-energy CORRECTION. **Caveat on the number:** Nørskov 2004 folds solvation + ZPE +
entropy into ONE empirical step correction fit to their detailed results — it does NOT print
an isolated solvation value, so don't cite "~0.5 eV solvation (Nørskov 2004)". The commonly
used ~0.3–0.5 eV *OH/*OOH solvation comes from later explicit-water studies. **Quantify it
by COMPUTING, not by plugging a hand-waved number:** explicit co-adsorbed H2O, or an implicit
solvent (VASPsol). (Empirically, bare-DFT Pt(111) ORR η ~0.9–1.0 V vs experimental ~0.45 V
is consistent with a ~0.5 eV *OH stabilization, but treat that as a result to verify, not an
input.)
## ORR (4e, associative) pathway, U = 0, pH 0
- ΔG1: O2 + * + (H⁺+e⁻) → *OOH (anchored via 4×1.23 = 4.92 eV total to 2 H2O)
- ΔG2: *OOH + (H⁺+e⁻) → *O + H2O
- ΔG3: *O + (H⁺+e⁻) → *OH
- ΔG4: *OH + (H⁺+e⁻) → * + H2O
With G_state = E + ZPE − TΔS for each adsorbed state, referenced to clean slab + the CHE
potentials. Here the ORR ΔG_i are the **reduction** step free energies — NEGATIVE when
downhill. **Limiting potential** U_L = **−**max_i(ΔG_i)/e (the negative sign is essential —
ORR steps are negative, unlike the OER convention below); **overpotential** η = 1.23 − U_L
= 1.23 + max_i(ΔG_i)/e. **Sanity check (assert it in code):** an ideal catalyst has every
ΔG_i = −1.23 eV → U_L = 1.23 V → η = 0; a formula that returns η = 2.46 there has the sign
wrong (this exact bug doubled a Pt(111) η from 0.96 → 1.50). Limiting step = the one with
the **largest (least-negative)** ΔG_i.
## OER
Reverse of ORR (water oxidation): same four intermediates, η = max(ΔG_i)/e − 1.23.
## HER
ΔG_H* = G(*H) − G(*) − ½ G_H2. Ideal at ΔG_H* ≈ 0; η descriptor from |ΔG_H*|.
## Outputs
A ΔG_i table, the limiting step, η, and a free-energy diagram (G vs reaction coordinate at
U = 0 and at U = U_L). Write them into the campaign's `analysis/`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |