Using alpha hulls to automatically and reproducibly detect edge clusters in atom probe tomography datasets

An automated way to accurately and reproducibly identify edge clusters within atom probe tomography datasets has been developed. The alpha-hull algorithm is used to generate a concave alpha-shape around an atom probe dataset. Information from core-linkage cluster searches is used, in combination with the calculated alpha-shape, to determine which clusters are on the edge of the dataset. The potential effects that not removing edge clusters may have on calculated cluster sizes, number densities and compositions is discussed. The viability of the methodology is demonstrated via application to real datasets, one of which was a non-standard shape. The sensitivity of the method to user parameter selection is explored. Sampling fractions >0.1% and an alpha value, used to make the alpha shape, greater than twice the maximum measured nearest neighbour distance were found to be suitable. [Display omitted] •Novel method developed for automatically detecting edge clusters in atom probe datasets.•Application to simulated data and its effect on measurement accuracy is assessed.•Successfully applied to real data – including datasets with complex shapes.