Gallery
The examples below showcase SBC on systems that are hard for distance- or composition-based clustering: heterostructures, curved sheets with impurities, grain boundaries and point defects. Each example builds a structure with ASE, runs SBC, verifies that the expected clusters are found, and renders the original system (left) next to the subsystems that SBC identifies (right). The full, self-contained source for every example is linked underneath it.
Layered heterostructure
A Cu(111) surface, a MoS2 monolayer and a graphene sheet stacked on top of one another. SBC separates the heterostructure into its three components based on their distinct repeating unit cells.
Warped sheet with adsorbates
A graphene sheet that has been curved by a sinusoidal ripple, with a few carbon dioxide molecules adsorbed a few ångström above it. A looser position tolerance lets SBC follow the curvature and recover the graphene as a single cluster, while the adsorbed CO2 molecules are left unclustered.
Grain boundary
Two grains of the same FCC copper, rotated with respect to each other and joined along a tilt boundary. The grains are chemically identical, so only the difference in their lattice orientation tells them apart — exactly what SBC keys on to return them as two separate clusters.
Bulk with a point defect
A diamond-silicon crystal with a single caesium atom substituted at an interior site (the large caesium atom is visible through the open diamond channels in the original). SBC returns the surrounding bulk as one cluster and discards the lone impurity, since a single atom has no translational symmetry of its own.
