Skip to content

Advanced Configuration

This guide covers advanced options for fine-tuning Graph ID generation.

Neighbor Detection Strategies

MinimumDistanceNN (Default)

from pymatgen.analysis.local_env import MinimumDistanceNN
from graph_id import GraphIDMaker

maker = GraphIDMaker(nn=MinimumDistanceNN())

CrystalNN

from graph_id_cpp import CrystalNN, GraphIDGenerator

gen = GraphIDGenerator(nn=CrystalNN())

from pymatgen.analysis.local_env import CrystalNN
from graph_id.core.graph_id import GraphIDGenerator

gen = GraphIDGenerator(nn=CrystalNN())

Distance Clustering

Uses DBSCAN clustering on interatomic distances:

from graph_id.core.distance_clustering_graph_id import DistanceClusteringGraphID

gen = DistanceClusteringGraphID(
    rank_k=3,    # Number of distance clusters to consider
    cutoff=6.0   # Maximum distance cutoff (Å)
)

graph_id = gen.get_id(structure)

Traversal Depth Control

Dynamic Depth (Default)

Depth is calculated from the graph diameter:

depth = diameter_factor × graph_diameter + additional_depth

from graph_id_cpp import GraphIDGenerator

gen = GraphIDGenerator(
    diameter_factor=2,     # Multiplier for diameter
    additional_depth=1     # Extra depth added
)

Fixed Depth

Override dynamic calculation with a fixed value:

from graph_id import GraphIDMaker

# Use exactly depth=5 for all structures
maker = GraphIDMaker(depth=5)

Topology-only Mode

Generate IDs based purely on connectivity, ignoring element types:

from graph_id_cpp import GraphIDGenerator

gen = GraphIDGenerator(topology_only=True)
topo_id = gen.get_id(structure)

Applications:

  • Finding isostructural compounds
  • Structure type classification
  • Comparing across element substitutions

Wyckoff Position Mode

Include crystallographic symmetry information:

from graph_id_cpp import GraphIDGenerator

gen = GraphIDGenerator(
    wyckoff=True,
    symmetry_tol=0.1  # Tolerance for symmetry detection
)

wyckoff_id = gen.get_id(structure)

The ID incorporates: - Space group number - Wyckoff letter for each site

Site Sequence Reduction

For structures with repeated motifs, reduce the sequence:

from graph_id import GraphIDMaker

maker = GraphIDMaker(reduce=True, depth=6)

This divides repeated site patterns. It is useful for:

  • Supercell comparisons
  • Detecting structural relationships