Point process statistics in atom probe tomography

Thomas Philippe (1) , Sébastien Duguay (1) , G. Grancher (2) , D. Blavette (1)

1 GPM - Groupe de physique des matériaux
2 LMRS - Laboratoire de Mathématiques Raphaël Salem

We present a review of spatial point processes as statistical models that we have designed for the analysis and treatment of atom probe tomography (APT) data. As a major advantage, these methods do not require sampling. The mean distance to nearest neighbour is an attractive approach to exhibit a non-random atomic distribution. A χ2test based on distance distributions to nearest neighbour has been developed to detect deviation from randomness. Best-fit methods based on first nearest neighbour distance (1NN method) and pair correlation function are presented and compared to assess the chemical composition of tiny clusters. Delaunay tessellation for cluster selection has been also illustrated. These statistical tools have been applied to APT experiments on microelectronics materials. \textcopyright 2012.

Mots clés: Atom probe tomography; Clustering; Delaunay tessellation; Nearest neighbour; Pair correlation function; Spatial point process

Domaines: Physique [physics]

Thomas Philippe, Sébastien Duguay, G. Grancher, D. Blavette. Point process statistics in atom probe tomography. Ultramicroscopy, 2013, 132, pp.114--120. ⟨10.1016/j.ultramic.2012.10.004⟩. ⟨hal-01928863⟩