Author ORCID Identifier

https://orcid.org/0000-0003-0536-2329

Semester

Summer

Date of Graduation

2025

Document Type

Dissertation

Degree Type

PhD

College

Eberly College of Arts and Sciences

Department

Forensic and Investigative Science

Committee Chair

Tatiana Trejos

Committee Member

Luis Arroyo

Committee Member

Glen Jackson

Committee Member

Cedric Neumann

Abstract

Glass is a trace material that is commonly encountered during investigations of violent crimes. When glass is recovered at crime scenes, it can be used to establish links between suspects, victims, and the scene itself. The most discriminatory form of analysis for glass evidence is elemental analysis, and micro-X-ray fluorescence spectrometry (µXRF) is becoming an increasingly common technique used for this purpose. Recent advances in µXRF technology, such as the introduction of silicon drift detectors (SDD) and improved polycapillary optics are promising in enhancing the capabilities for the examination of glass evidence in forensic investigations. However, along with these advances comes a need to adapt to the new technology. For example, it becomes necessary to reevaluate the standard analytical methods and comparison criteria for the latest instrumentation and improve objectivity in interpreting spectral data. Additionally, the ubiquity of cellular phones in everyday life has made them increasingly prominent in crimes. Still, the current body of research surrounding this type of glass is limited, necessitating the development of methods before it can be incorporated into forensic services.

This research aimed to address these gaps in the forensic glass analysis discipline by establishing an extensive µXRF database of glass from various sources, including vehicles, residential and commercial buildings, and cellular devices. The µXRF technique was tested through the analysis of extremely thin fragments of glass that are routinely observed in forensic casework and contemporary formulations and coatings of architectural and portable electronic device (PED) glasses. This study also proposed the development of a spectral similarity metric to increase objectivity and enable greater collaboration and data sharing among law enforcement agencies.

This project generated a physical and digital database comprising over 160 sources of contemporary soda-lime and aluminosilicate glasses with known origins of the manufacturing sources. These sources were used to create more than 2,700 glass fragments, which were all analyzed using µXRF instrumentation, and a subset was compared across multiple laboratories. The within- and between-source variations were evaluated to provide revised recommendations to the ASTM E 2926-25 Standard Test Method for Forensic Comparison of Glass Using µXRF, regarding the collection, sampling, analysis, and interpretation of glass when using µXRF with SDD detectors. The spectral comparison comprised spectral overlay, comparison intervals of selected elemental ratios, and the Similarity Contrast Angle Ratio (SCAR). The use of a modified 3s comparison criterion for soda-lime glass resulted in false exclusion rates of less than 3% and false inclusion rates of less than 0.5%. A modified 5s comparison criterion was necessary for aluminosilicate glass, resulting in false exclusion rates of less than 4% and false inclusion rates of less than 0.2%. The use of the SCAR quantitative metric also showed promise in evaluating spectral similarity, with misclassification rates of less than 3%. Since the µXRF with SDD offered improved sensitivity and precision, it enabled the analysis of smaller samples and faster acquisition times. The µXRF technique was stress-tested through the analysis of thin soda-lime glass fragments, ranging in thickness from 10 to 50 µm, to simulate cases where only minute fragments are recovered. The most substantial finding of this study was that modern µXRFs are suitable for analyzing glass as small as 10 µm, thereby expanding capabilities in casework. However, when using thin fragments, heavier elements were more challenging to detect in thinner fragments than in their full-thickness counterparts. Since these elements are the most discriminatory in soda-lime glass, forensic scientists must exercise caution when evaluating this type of sample size.

An interlaboratory study was also conducted to evaluate the performance of different µXRF instruments with eight participants across the United States. This study showed remarkable performance in discriminating between sets of different glass sources and correctly associating glass sets from a common source through spectral overlay comparisons. The use of a modified 3s comparison criterion for element ratio comparisons resulted in zero false inclusions and a false exclusion rate of less than 5%. Regardless of instrument configuration, same-source comparisons yielded low SCAR values, close to one. In contrast, different-source comparisons mostly resulted in SCAR values much greater than one, with the higher the value, the more distinctive the elemental profiles of the compared items. The level of agreement in SCAR values among participants indicated that this quantitative metric could be used to support the analyst’s opinion in a more objective manner, opening up the opportunity to share µXRF data among participants with less influence on the instrumental configurations employed. Furthermore, SCAR provided a promising proxy for the weight of the evidence when used as an input for calculating score likelihood ratios (SLRs).

This study also expanded the forensic examination of glass to modern PED glass. The findings demonstrated that the chemical and optical properties of PED glass are substantially different from those of typical soda-lime glass, requiring modifications in the methodologies. For example, the refractive index values of PED glass fell outside of the typical calibration ranges of silicon oils A, B, or C used for soda-lime and borosilicate glass. A modified method using a mixture of A-B and B-C oils is proposed here as a modification to ASTM E1967 Standard Test Method for the Automated Determination of Refractive Index of Glass Samples Using the Oil Immersion Method and a Phase Contrast Microscope. Moreover, due to inherent variations observed across a cell phone pane, specific recommendations for RI and elemental µXRF analysis are presented here for sampling, analysis, and data comparison criteria. It is recommended that 30 measurements from at least 10 fragments be collected to properly characterize the refractive index of a known sample, thereby reducing error rates. For refractive index measurements, the false exclusion and inclusion rates were less than 3 % and 9 %, respectively, when using a range overlap criterion. For µXRF measurements, using a modified 5s comparison criterion for element ratio comparisons resulted in low false exclusion and false inclusion rates (less than 4.0 % and 0.5 %, respectively). In this dataset of 45 phones from six manufacturers, the combination of refractive index and µXRF yielded a 99.9 % discrimination of glass originating from different sources. The SCAR metric was also effective in interpreting data from aluminosilicate glasses. The study provides insights into the informative elements to characterize PED glass and differentiate it from soda-lime glass.

Altogether, this study significantly expanded the knowledge base in the forensic analysis and interpretation of glass using modern µXRF technology with SDD detectors. The combination of the largest existing digital database of µXRF spectral data and interlaboratory testing enabled a comprehensive evaluation of error rates, providing a foundation for recommendations on necessary adaptations to existing standard test methods and the creation of new methods for examining newer formulations, such as aluminosilicate glasses from mobile devices.

Comments

Revised Version - 26 July 2025

Download

Share

COinS