Modularity Optimization for Fast Automated Detection of Solute Clusters in Atom Probe Tomography

The current work describes the use of some unsupervised Machine Learning Methods to uncover patterns in Atom Probe Tomography (APT) scans of microstructures. Data reconstructed from an experiment comprises of the following variables: position in x; y; z coordinates and mass-to-charge ratio. While the geometry of the atoms are provided by the x; y; z positions, the mass-to-charge ratio provides us information about the chemistry of the atom. The analysis performed on few datasets have been reported in the following sections.