This is the authors' abstract. We don't add key points for this paper.
Ranging is the process to define lower and upper bounds of a peak in a mass histogram for an element/ion. In a perfect system with high mass resolving power, Gaussian peak shapes, no noise, and no direct peak overlaps, one could calculate local and global compositions with perfect accuracy. In all real-world atom probe tomography (APT) analyses however, there are often multiphase materials that have peak overlaps, peak shape differences, differences in background noise, etc. As there are currently no standards governing the ranging of APT data, two people analyzing the same dataset may get different answers [1, 2]. To investigate ranging techniques, a multiphase simulation of a boron doped silicon and nickel diffusion couple was created with variable mass resolving power, background noise levels and peak shapes for the various elements using the region of interest (ROI) simulation utility in AP Suite 6.3.1 (See Figure 1). Ranging was first done ‘by hand’ identifying peaks and then starting and stopping the ranges at approximately two-times the local background noise. Reasons for differences in expected composition include background subtraction, peak overlaps, and potential errors due to the undefined ranging ‘by hand’ method. Region 1 and Region 3 compositional variances have measured compositional differences as high as 0.3 at% (Ni) due to Ni and B ranges counting background noise while spectrally no peaks exist. Region 2 includes differences as high as 2.2 at%. due to the nickel/silicon peak overlap (28/29 Da). Decomposition of peaks reduces this difference to 0.1 at%. Note that decomposition of peaks works volumetrically and regions that have direct peak overlaps (Figure 2, Region 2) need to be addressed separately. AP Suite 6.3.1 includes a new feature called detached ROI ranging, which allows ranging definitions to be segmented by volumes defined by geometrical ROI selection tools or by isoconcentration surfaces. The improved compositional accuracy, especially in Regions 1 and 3, are demonstrated in Table 1. Peaks that are easily identifiable and share the same shape can be ranged in a consistent manner to get accurate and precise results. Complex spectra with molecular species often have peaks with variable shapes. Therefore, accurate and precise ranging of complex spectra requires not only accurate identification of peaks and unique range definitions for distinct volumes, but also variable ranging rules which account for the peak shape variations that can exist in real (not simulated) APT data. Machine learning and algorithmically driven ranging can help overcome these problems and provide valuable insights to aid in quantification (e.g., Figure 3) in both simulated and real APT data. Simulated APT data with three ROIs including different peak shapes and backgrounds. A one-dimensional composition profile through the simulated regions labeled with the simulated compositions. A single range definition was used, demonstrating errors due to not segmenting the distinct volumes (Regions 1, 3), or accounting for direct peak overlaps (Region 2). Target and measured compositions for each region shown in Figure 1. Incremental sampling of ranges at different height fractions quantitatively shows the peaks all have a similar form in this simulated spectrum.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Ulfig et al. (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: