Quantification of structural information in atom probe tomography using statistical learning techniques

The purpose of this thesis is to develop data-driven methods to extract structural information using Atom Probe Tomography (APT). Statistically robust methods to identify spatial relationships between atomic positions in three-dimensional APT data are developed. Spatial uncertainties of a single atom are analyzed. The role of crystallographic orientations on the impact of spatial resolutions in APT is explored.;Using Spatial Distribution Maps (SDMs) obtained from Atom Probe data; the classification capability of singular value decomposition methods to classify Atom Probe data into structurally relevant, noise and aberrations is demonstrated. The feature extraction power of singular value decomposition is utilized to de-convolute spatial resolutions of a single atom from the SDM data. The ability of this methodology to identify lateral resolutions and trajectory aberrations at various crystallographic locations within the APT data is demonstrated. This technique is extended to develop a benchmark methodology to analyze spatial resolutions in APT using tungsten as a standard specimen. This work has laid the foundation to utilize statistical learning methods as a spatial metrology tool to analyze Atom Probe data. The results of this work provide exciting new opportunities to further understanding of structural analysis of materials using Atom Probe.