Development and evaluation of methods for objective comparison of x-ray fluorescence spectra

Area of Science:

• Analytical Chemistry
• Forensic Science
• Materials Science

Background:

• X-ray Fluorescence (XRF) is a powerful technique for elemental analysis.
• Objective comparison of XRF spectra is crucial for forensic investigations and material characterization.
• Existing methods for spectral comparison may lack quantitative rigor and transparency.

Purpose of the Study:

• To develop and validate a quantitative similarity measure for XRF spectral comparisons.
• To assess the variability within and between different sources of electrical tape samples.
• To provide a more objective framework for reporting spectral data in forensic science.

Main Methods:

• Characterization of 94 rolls of electrical tape using XRF.
• Application of Fast Fourier Transform for noise filtering and weighted regression for baseline correction.
• Calculation of Spectral Contrast Angle Ratios (SCAR) for pairwise spectral comparisons (n=4561).
• Utilizing cluster analysis and a random forest algorithm for sample classification and automated comparison.

Main Results:

• The SCAR metric effectively captures spectral variability between different samples and within same-source samples.
• A threshold was proposed to minimize erroneous associations or exclusions based on SCAR values.
• Cluster analysis and random forest algorithms demonstrated the utility of SCAR for sample classification and automated comparison.

Conclusions:

• SCAR provides a statistically robust and objective method for comparing XRF spectra.
• This approach enhances transparency and supports forensic examiner decision-making.
• The SCAR metric can serve as a model for objective spectral data analysis in diverse scientific fields.