Adsorbate Placement

SkillAI & models

The adsorbateplace task type places adsorbate molecules on surface slabs. It uses ferrox (Rust) findadsorptionsites to locate surface sites, then the CatGo placement engine (utils/adsorbateplacement.py) for Rodrigues rotation, overlap detection, and multi-dentate support.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Adsorbate Placement skill

About this capability

Use when the user asks to place an adsorbate molecule on a surface, find adsorption sites, or set up a surface+adsorbate model for DFT.

What this skill tells your AI

The instructions your AI receives, as published by hello-qm/catgo-lrg in .claude/skills/structure-adsorbate/SKILL.md and read by ahel’s review.

Overview

The adsorbate_place task type places adsorbate molecules on surface slabs. It uses ferrox (Rust) find_adsorption_sites to locate surface sites, then the CatGo placement engine (utils/adsorbate_placement.py) for Rodrigues rotation, overlap detection, and multi-dentate support.

Task Type: adsorbate_place

  • Type: adsorbate_place (local task, no HPC needed)
  • Engine: ferrox site finder + CatGo placement engine
  • Outputs: structure (slab+adsorbate as JSON)

Parameters

ParameterTypeDefaultDescription
structureJSONrequiredSlab structure input
speciesstr"OH"Adsorbate species name
sitestr"ontop"Site type: "ontop", "bridge", "hollow", or "all"
heightfloat2.0Height above surface in Angstroms
site_indexint0Which site of the given type to use (0 = first)

Supported Adsorbate Species

SpeciesAtomsBinding AtomNotes
OHO, HOHydroxyl, O-H = 0.96 A
OOOAtomic oxygen
OOHO, O, HO1Hydroperoxo, key OER intermediate
HHHAtomic hydrogen
H2OO, H, HOWater molecule
COC, OCCarbon monoxide, C-O = 1.13 A
CO2C, O, OCCarbon dioxide, linear
N2N, NNDinitrogen, N-N = 1.10 A
NHN, HNImide
NH2N, H, HNAmino
NH3N, H, H, HNAmmonia
CHOC, H, OCFormyl
COOHC, O, O, HCCarboxyl

Site Types

Siteferrox TypeCoordinationDescription
ontopatop1-foldDirectly above one surface atom
bridgebridge2-foldBetween two surface atoms
hollowhollow33-foldAbove threefold hollow site
allatop (default)1-foldAuto-selects ontop

Discussion Checkpoints

🔴 Must discuss with user:

  • Adsorbate species — determines the chemistry being studied; wrong species = wrong intermediate in the reaction pathway
  • Adsorption site (ontop/bridge/hollow) — different sites have different binding energies; for screening, test all three and report the most stable

🟡 Recommend confirming:

  • Height above surface (default: 2.0 A) — too close triggers repulsion during geo_opt, too far causes adsorbate to fly away; use 1.5-1.8 A for atomic O, 2.0-2.5 A for molecular species
  • Site index (default: 0) — which specific site of the given type; call catgo_analyze(action="adsorption_sites") first to see available sites
  • Multi-dentate orientation — for OOH, COOH, and other multi-atom adsorbates, the binding orientation matters; verify with catgo_view after placement

🟢 Safe defaults:

  • Collision detection enabled (ferrox automatic)
  • Automatic site finding via ferrox find_adsorption_sites
  • Binding atom orientation follows species convention (O down for OH, C down for CO)

MCP Workflow: Place Adsorbate on Slab

Step 1: Generate slab (or use existing)

{"tool": "catgo_fetch", "arguments": {
  "action": "crystal", "formula": "Pt", "provider": "mp"
}}
{"tool": "catgo_structure", "arguments": {
  "action": "slab",
  "miller_index": [1, 1, 1],
  "min_slab_size": 12.0,
  "min_vacuum_size": 15.0
}}
{"tool": "catgo_structure", "arguments": {
  "action": "supercell",
  "scaling": [2, 2, 1]
}}

Step 2: Find adsorption sites

{"tool": "catgo_analyze", "arguments": {
  "action": "adsorption_sites"
}}

This returns all available sites (ontop, bridge, hollow) with coordinates.

Step 3: Build workflow with adsorbate placement

Using the workflow engine with adsorbate_place node:

{"tool": "catgo_workflow", "arguments": {
  "action": "batch",
  "workflow_id": "wf_123",
  "operations": [
    {"op": "add_node", "node_type": "slab_gen", "label": "slab1",
     "params": {"miller": [1, 1, 1], "layers": 4, "vacuum": 15.0}},
    {"op": "add_node", "node_type": "adsorbate_place", "label": "ads1",
     "params": {"species": "OH", "site": "ontop", "height": 2.0, "site_index": 0}},
    {"op": "add_node", "node_type": "geo_opt", "label": "go1",
     "params": {"software": "vasp", "ENCUT": 520, "freeze_mode": "layers", "freeze_layers": 2}},
    {"op": "connect", "from_id": "<structure_input_id>", "to_id": "slab1"},
    {"op": "connect", "from_id": "slab1", "to_id": "ads1",
     "from_handle": "structure", "to_handle": "structure"},
    {"op": "connect", "from_id": "ads1", "to_id": "go1",
     "from_handle": "structure", "to_handle": "structure"}
  ]
}}

Step 4: PENDING_REVIEW -- verify adsorbate position

The user should inspect the structure before submitting geo_opt. Check:

  • Adsorbate is at the correct site (ontop/bridge/hollow)
  • Height above surface is reasonable (1.5-2.5 A for most species)
  • No atom overlaps or unrealistic bond lengths
  • Binding atom orientation is correct (e.g., C down for CO, O down for OH)
{"tool": "catgo_view", "arguments": {"action": "get_state"}}

Step 5: Submit for DFT optimization

{"tool": "catgo_workflow", "arguments": {
  "action": "run",
  "workflow_id": "wf_123",
  "run_config": {"cluster": "expanse", "partition": "shared", "walltime": "04:00:00"}
}}

Python API

from catgo.workflow import Workflow

wf = Workflow("OH on Pt(111)")

# Bulk input
inp = wf.add_task("structure_input", structure=pt_bulk_json)

# Cut slab
slab = wf.add_task("slab_gen",
    structure=inp.output.structure,
    miller=(1, 1, 1),
    layers=4,
    vacuum=15.0)

# Place adsorbate
ads = wf.add_task("adsorbate_place",
    structure=slab.output.structure,
    species="OH",
    site="ontop",
    height=2.0,
    site_index=0)

# PENDING_REVIEW: user should verify adsorbate position before geo_opt

# Geometry optimization
opt = wf.add_task("geo_opt",
    structure=ads.output.structure,
    software="vasp",
    ENCUT=520,
    freeze_mode="layers",
    freeze_layers=2)

wf.submit()

Complete OER Workflow Example

The oxygen evolution reaction (OER) has four intermediates: *OH, *O, *OOH, and clean slab. Each needs geo_opt + freq + gibbs_energy.

from catgo.workflow import Workflow

wf = Workflow("OER on IrO2(110)")

# Shared bulk input
inp = wf.add_task("structure_input", structure=iro2_bulk_json)
slab = wf.add_task("slab_gen",
    structure=inp.output.structure,
    miller=(1, 1, 0), layers=4, vacuum=15.0)

# --- Clean slab branch ---
slab_opt = wf.add_task("geo_opt", structure=slab.output.structure,
    software="vasp", ENCUT=520, system_name="clean_slab",
    freeze_mode="layers", freeze_layers=2)
slab_freq = wf.add_task("freq", structure=slab_opt.output.structure,
    software="vasp", freeze_mode="layers", freeze_layers=2)
slab_gibbs = wf.add_task("gibbs_energy",
    energy=slab_opt.output.energy,
    frequencies=slab_freq.output.frequencies,
    phase="adsorbed")

# --- *OH branch ---
oh_ads = wf.add_task("adsorbate_place", structure=slab.output.structure,
    species="OH", site="ontop", height=2.0)
oh_opt = wf.add_task("geo_opt", structure=oh_ads.output.structure,
    software="vasp", ENCUT=520, system_name="OH_ads",
    freeze_mode="layers", freeze_layers=2)
oh_freq = wf.add_task("freq", structure=oh_opt.output.structure,
    software="vasp", freeze_mode="layers", freeze_layers=2)
oh_gibbs = wf.add_task("gibbs_energy",
    energy=oh_opt.output.energy,
    frequencies=oh_freq.output.frequencies,
    phase="adsorbed")

# --- *O branch ---
o_ads = wf.add_task("adsorbate_place", structure=slab.output.structure,
    species="O", site="ontop", height=1.8)
o_opt = wf.add_task("geo_opt", structure=o_ads.output.structure,
    software="vasp", ENCUT=520, system_name="O_ads",
    freeze_mode="layers", freeze_layers=2)
o_freq = wf.add_task("freq", structure=o_opt.output.structure,
    software="vasp", freeze_mode="layers", freeze_layers=2)
o_gibbs = wf.add_task("gibbs_energy",
    energy=o_opt.output.energy,
    frequencies=o_freq.output.frequencies,
    phase="adsorbed")

# --- *OOH branch ---
ooh_ads = wf.add_task("adsorbate_place", structure=slab.output.structure,
    species="OOH", site="ontop", height=2.0)
ooh_opt = wf.add_task("geo_opt", structure=ooh_ads.output.structure,
    software="vasp", ENCUT=520, system_name="OOH_ads",
    freeze_mode="layers", freeze_layers=2)
ooh_freq = wf.add_task("freq", structure=ooh_opt.output.structure,
    software="vasp", freeze_mode="layers", freeze_layers=2)
ooh_gibbs = wf.add_task("gibbs_energy",
    energy=ooh_opt.output.energy,
    frequencies=ooh_freq.output.frequencies,
    phase="adsorbed")

# --- Gas-phase references (H2O, H2) ---
h2o_inp = wf.add_task("structure_input", structure=h2o_gas_json)
h2o_opt = wf.add_task("geo_opt", structure=h2o_inp.output.structure,
    software="vasp", ENCUT=520, ISMEAR=0, KPOINTS=[1,1,1],
    system_name="H2O_gas")
h2o_freq = wf.add_task("freq", structure=h2o_opt.output.structure,
    software="vasp")
h2o_gibbs = wf.add_task("gibbs_energy",
    energy=h2o_opt.output.energy,
    frequencies=h2o_freq.output.frequencies,
    phase="gas")

h2_inp = wf.add_task("structure_input", structure=h2_gas_json)
h2_opt = wf.add_task("geo_opt", structure=h2_inp.output.structure,
    software="vasp", ENCUT=520, ISMEAR=0, KPOINTS=[1,1,1],
    system_name="H2_gas")
h2_freq = wf.add_task("freq", structure=h2_opt.output.structure,
    software="vasp")
h2_gibbs = wf.add_task("gibbs_energy",
    energy=h2_opt.output.energy,
    frequencies=h2_freq.output.frequencies,
    phase="gas")

# Free energy diagram
fed = wf.add_task("free_energy_diagram",
    gibbs_values={
        "clean": slab_gibbs.output.gibbs,
        "OH": oh_gibbs.output.gibbs,
        "O": o_gibbs.output.gibbs,
        "OOH": ooh_gibbs.output.gibbs,
        "H2O": h2o_gibbs.output.gibbs,
        "H2": h2_gibbs.output.gibbs,
    },
    step_order=["clean", "OH", "O", "OOH", "O2"])

wf.submit()

DAG Structure (Single Adsorbate)

bulk_crystal --> slab_gen --> adsorbate_place --> [PENDING_REVIEW] --> geo_opt

DAG Structure (OER)

                                /--> *OH  --> geo_opt --> freq --> gibbs --\
bulk --> slab_gen --> slab -----+--> *O   --> geo_opt --> freq --> gibbs ---+--> free_energy_diagram
                          \    \--> *OOH --> geo_opt --> freq --> gibbs --/
                           \--> clean_slab --> geo_opt --> freq --> gibbs -/

Comparing Multiple Sites

To compare adsorption at different sites (ontop, bridge, hollow), create separate branches from the same slab:

for site_type in ["ontop", "bridge", "hollow"]:
    ads = wf.add_task("adsorbate_place",
        structure=slab.output.structure,
        species="OH",
        site=site_type,
        height=2.0,
        site_index=0)
    opt = wf.add_task("geo_opt",
        structure=ads.output.structure,
        software="vasp", ENCUT=520,
        system_name=f"OH_{site_type}")

Common Pitfalls

  1. Always build the slab and supercell BEFORE placing adsorbates. Adding adsorbates to a 1x1 slab gives unphysically high coverage.
  2. The initial height matters: too close triggers repulsion, too far may cause the adsorbate to fly away during geo_opt. Use 1.5-2.5 A for most species.
  3. OOH tends to dissociate into O + OH during relaxation on some surfaces. Use tight EDIFFG and monitor the trajectory.
  4. For bridge and hollow sites, the adsorbate is placed between atoms automatically by the ferrox site finder. Do not manually calculate midpoints.
  5. After placement, always verify with catgo_view that no atoms overlap and the geometry looks reasonable before submitting to DFT.
  6. When using site_index, call catgo_analyze with action: "adsorption_sites" first to see available sites and their indices.
  7. The site="all" option defaults to ontop (atop) sites. For specific site types, always pass the explicit type name.

Signals

GitHub stars
196
Forks
23
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
adsorbate-placement-hello-qm
Source
github.com/hello-qm/catgo-lrg