New strategies for portable XRF spectrum processing applied to historical glasses

Abstract

Portable X-ray fluorescence (p-XRF) is a powerful, non-destructive tool widely used to analyze cultural heritage objects. However, limitations such as noise, drift, and spectral overlaps may affect the accuracy of elemental identification and quantification. This study presents a comprehensive methodology for processing and analyzing p-XRF spectra, aiming to improve the reliability and accuracy of elemental identification in cultural heritage materials. The methodology incorporates key steps such as spectral alignment, energy calibration, smoothing, drift correction, signal detection, deconvolution, and integration.The approach was successfully applied to reference materials and historical glass objects from La Real Fábrica de Cristales de La Granja, which exhibit a wide range of elemental concentrations. Furthermore, the methodology was adapted to deal with spectra acquired from various portable XRF instruments, demonstrating broad compatibility. Notable improvements include increased positive and negative predictive values for element identification, especially for challenging low-intensity signals such as Na, Mg, and Al, compared to the direct instrument report. Limits of detection were also established, demonstrating reliability across a range of elements. Key advancements include reduced noise, enhanced linearity (R²), and the preservation of spectral proportionality, ensuring robust and reproducible results. This approach reduces uncertainty in element identification, improves the accuracy in concentration determination and facilitates the use of p-XRF by researchers and conservators. A user-friendly implementation of this procedure, including example datasets and scripts, is provided in the Supplementary Material. Additionally, an interactive Shiny application was included to allow users to perform basic spectral processing in a straightforward and accessible manner.