Slab Generation

Overview

A slab model represents a crystal surface: a finite number of atomic layers with vacuum above and below. Required for any surface chemistry calculation.

The slab_gen task type uses ferrox (Rust) generate_slab internally for fast slab cutting. The c axis of the output slab is always perpendicular to the ab plane (the surface plane).

Task Type: slab_gen

• Type: slab_gen (local task, no HPC needed)
• Engine: ferrox (Rust) surfaces.generate_slab
• Outputs: structure (slab as JSON)

Parameters

Parameter	Type	Default	Description
structure	JSON	required	Bulk crystal structure input
miller	tuple	(1, 1, 0)	Miller index (h, k, l) for surface orientation
layers	int	4	Number of atomic layers in the slab
vacuum	float	15.0	Vacuum thickness in Angstroms
thickness	float	10.0	Minimum slab thickness in Angstroms

Miller Index Quick Reference

Surface	Miller Index	Structure Type	Common Use
FCC (111)	(1,1,1)	Close-packed	Pt, Pd, Au, Cu catalysis
FCC (100)	(1,0,0)	Square surface	Open face, higher activity
FCC (110)	(1,1,0)	Ridged surface	Step edges
BCC (110)	(1,1,0)	Close-packed	Fe, W surfaces
BCC (100)	(1,0,0)	Open surface	Fe catalysis
Rutile (110)	(1,1,0)	Most stable	TiO2, RuO2 catalysis
Perovskite (001)	(0,0,1)	AO or BO2 terminated	SrTiO3, LaCoO3

Discussion Checkpoints

🔴 Must discuss with user:

• Miller index — determines surface orientation and active site geometry; e.g., FCC(111) is close-packed while (110) has ridged surface with step edges
• Number of layers — too few layers (< 4) gives unconverged surface energy; production calculations need 5-6 layers minimum
• Termination choice — critical for oxides (e.g., RuO2(110) O-terminated vs metal-terminated); wrong termination invalidates the surface chemistry model

🟡 Recommend confirming:

• Vacuum thickness (default: 15 A) — increase to 20 A for charged surfaces or large dipole corrections; 12 A acceptable with LDIPOL
• Supercell size — 1x1 slab has unphysical adsorbate-adsorbate interactions; expand to at least 2x2 for adsorbate studies
• Fixed layers (default: freeze bottom 2) — adjust based on total slab thickness; for 6-layer slab, freeze bottom 3

🟢 Safe defaults:

• c perpendicular to ab orientation (ferrox guarantees this)
• Vacuum along c direction
• layers = 4 for initial testing
• vacuum = 15.0 A

MCP Workflow

Step 1: Fetch bulk crystal

{"tool": "catgo_fetch", "arguments": {
  "action": "crystal", "formula": "Pt", "provider": "mp"
}}

Step 2: Generate slab (interactive viewer)

{"tool": "catgo_structure", "arguments": {
  "action": "slab",
  "miller_index": [1, 1, 1],
  "min_slab_size": 12.0,
  "min_vacuum_size": 15.0
}}

If multiple terminations are returned, the tool lists them. Select the desired termination:

{"tool": "catgo_structure", "arguments": {
  "action": "slab",
  "miller_index": [1, 1, 1],
  "min_slab_size": 12.0,
  "min_vacuum_size": 15.0,
  "termination_index": 0
}}

Step 3: PENDING_REVIEW -- verify the slab

The user should inspect the slab before proceeding. Check:

• Correct surface termination (especially for oxides)
• Slab thickness is adequate (5+ layers for production)
• No dangling bonds or polar surface artifacts
• Vacuum gap is sufficient (15+ A)

{"tool": "catgo_view", "arguments": {"action": "get_state"}}

Step 4: Make supercell for adsorbate calculations

A 1x1 slab is usually too small (periodic image interactions). Expand to at least 2x2 in the surface plane. Never scale in z (vacuum direction):

{"tool": "catgo_structure", "arguments": {
  "action": "supercell",
  "scaling": [2, 2, 1]
}}

Step 5: Submit geometry optimization

{"tool": "catgo_workflow_engine", "arguments": {
  "action": "add_task",
  "params": {
    "task_type": "geo_opt",
    "software": "vasp",
    "ENCUT": 520,
    "freeze_mode": "layers",
    "freeze_layers": 2,
    "system_name": "Pt111_slab"
  }
}}

Python API

from catgo.workflow import Workflow

wf = Workflow("Pt(111) slab")

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

# Cut slab using ferrox
slab = wf.add_task("slab_gen",
    structure=inp.output.structure,
    miller=(1, 1, 1),
    layers=4,
    vacuum=15.0,
    thickness=12.0)

# PENDING_REVIEW: user should check the slab before submitting geo_opt
# Verify correct termination, adequate thickness, and no polar artifacts

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

wf.submit()

Workflow Engine (catgo_workflow) Batch Example

Using the graph-based workflow editor with batch operations:

{"tool": "catgo_workflow", "arguments": {
  "action": "batch",
  "workflow_id": "wf_123",
  "operations": [
    {"op": "add_node", "node_type": "slab_gen", "label": "slab1",
     "params": {"miller": [1, 1, 0], "layers": 4, "vacuum": 15.0}},
    {"op": "add_node", "node_type": "geo_opt", "label": "go1",
     "params": {"software": "vasp", "ENCUT": 520}},
    {"op": "connect", "from_id": "<structure_input_id>", "to_id": "slab1"},
    {"op": "connect", "from_id": "slab1", "to_id": "go1",
     "from_handle": "structure", "to_handle": "structure"}
  ]
}}

DAG Structure

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

Slab Thickness Guidelines

Application	thickness (A)	Approx. Layers (FCC)
Quick test	8.0	3-4
Production adsorption	12.0	5-6
Accurate work function	15.0	7-8
Subsurface diffusion	18.0+	9+

Vacuum Thickness

• 15 A minimum for standard DFT (prevents periodic image interaction)
• 20 A for charged surfaces or large dipole corrections
• With dipole correction (LDIPOL, IDIPOL=3): 12 A can be sufficient

Layer Freezing

For a 4-layer slab:

• Layers 1-2 (bottom): freeze during geo_opt and freq
• Layers 3-4 (top): free to relax

In VASP, constrained atoms use Selective Dynamics (T/F flags). CatGo handles this via freeze_mode="layers" in the geo_opt/freq task.

Oxide Slabs

For oxides (TiO2, RuO2, IrO2), the slab may have multiple terminations. For rutile (110):

• O-terminated: bridging oxygen rows on surface (most common, most stable)
• Metal-terminated: exposed cation rows (CUS sites)

Select based on experimental relevance and thermodynamic stability.

Common Pitfalls

1. Always start from a BULK crystal. Cutting a slab from an already-cut slab produces garbage.
2. The c axis is always perpendicular to the ab plane in the output slab. Vacuum is along the c direction. Never use supercell scaling in z (e.g., [2,2,2] would double the vacuum AND the slab, wasting compute).
3. For non-cubic systems, verify the Miller index orientation is correct. Hexagonal systems use 3-index notation (h,k,l) not 4-index (h,k,i,l).
4. Oxide slabs may be polar (e.g., ZnO(0001)). Polar slabs have a net dipole and require either reconstruction, passivation, or dipole correction. Check with catgo_view after generation.
5. After supercell expansion, atom indices change. Re-identify surface atoms before placing adsorbates.
6. For doped slabs, ALWAYS generate the slab from pristine bulk FIRST, then dope the slab. Doping bulk before slabbing replicates the dopant once per bulk repeat in slab thickness, giving unrealistically high concentrations.