Ranging Atom Probe Spectra to Reduce Measurement Bias

Published

July 24, 2024

Author(s)

Frederick Meisenkothen, David Newton, Karen DeRocher, Mark McLean

Abstract

Atom probe tomography (APT) is emerging as an essential characterization tool in a wide variety of science and engineering fields. For example, APT is helping to foster nascent fields, such as nano-geology [1], and is increasingly being adopted by the semiconductor industry. APT provides the unique capability to access 3-dimensional information about both the chemical and the isotopic distributions within a nano-scale sample of material. However, APT spectra are complex, and there is currently no community-wide consensus on how best to attribute regions of interest in the spectrum to specific ion species (i.e., ion ranging). Ion ranging is thus an active area of research within the community. In this work, we summarize our findings regarding some limited sources of measurement bias in APT quantitative analysis and discuss ion ranging strategies that can improve accuracy and repeatability. Accurate and repeatable single-element isotopic analysis can be achieved by employing a three-step approach — i.e., (1) data filtering, (2) adaptive peak fitting, and (3) calibration [1-3]. Generally, the largest contribution to measurement bias in single-element isotopic analysis is associated with multi-hit detection events and post-acquisition redefinition of detection events. Working with predominantly single-hit data can thus greatly improve accuracy. Adaptive peak fitting — based on the premise, that isotopic variants (e.g., 28Si2+, 29Si2+, 30Si2+) within a single ion species have the same peak form — uses a machine learning-based approach to determine the relative peak intensities of isotopic variants within a single ion species. The repeatability provided by adaptive peak fitting makes it possible to use reference materials and a standards-based analysis approach to further improve accuracy in some circumstances. Our adaptive peak fitting code has been ported to a user-friendly online application for public release. Figure 1(a) shows the graphical user interface along with example peak fitting results for a set of Si2+ peaks in Figure 1 (b)-(d). The three-step approach developed for single-element isotopic analysis cannot be used for complex multi-element analyses, since peak forms cannot be assumed to be identical across the spectrum. Further, using single-hit data to remove some artifacts can introduce other composition measurement bias. Therefore, an alternative ranging methodology must be developed. To this end, a ranging study was conducted to compare different common ranging strategies [4]. A script was used to systematically and incrementally sample peaks at different height fractions (full width at "x" maximum, FWxM) from a well-characterized natural mineral specimen as well as other materials. Figure 2(a) shows a schematic of the systematic sampling process. Systematic sampling demonstrates how peak shape and ranging method can interact to bias quantitative analysis results, as shown in Figure 2(b). Therefore, peaks must be sampled in proper areal proportion to one another to obtain the most accurate analysis result. Full integrated peak intensity, if obtainable, should minimize peak shape-related composition bias, emphasizing the need for the APT community to develop robust and adaptable peak fitting algorithms for APT spectral analysis.

Citation

Microscopy and Microanalysis

Volume

30

Pub Type

Journals

Download Paper

https://doi.org/10.1093/mam/ozae044.027

Local Download

Keywords

atom probe, mass spectrometry, peak fitting, isotopic analysis

Citation

Meisenkothen, F. , Newton, D. , DeRocher, K. and McLean, M. (2024), Ranging Atom Probe Spectra to Reduce Measurement Bias, Microscopy and Microanalysis, [online], https://doi.org/10.1093/mam/ozae044.027, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=957539 (Accessed December 3, 2024)