Titolo della tesi: Innovative spectroscopic approaches in forensic analytical chemistry
Forensic science, also known as criminalistics, is defined as the application of scientific methodologies to traditional judicial investigations, often described as science in the service of the law. The primary objective is to determine and characterize the origin of a sample to establish relationships that can link it to a person, event, or place. In the field of forensic analysis, one of the key goals for competent authorities is to optimize the time required to obtain real-time analytical answers.
For the analysis of evidence, forensic science employs two types of tests: screening or first-level tests, commonly used for their cost-effectiveness and quick response, though lacking the precision and accuracy required for official methods; and second-level tests, involving instrumental techniques. These include chromatographic methods such as high-performance liquid chromatography (HPLC) and gas chromatography (GC), often coupled with detectors like photodiode array detectors (DAD), flame ionization detectors (FID), or mass spectrometers (MS). However, these analytical techniques have limitations due to their destructive nature, costs associated with solvent waste, and lengthy analysis times. Additionally, they pose challenges for in situ analysis implementation.
The aim of this PhD project was the development and validation of new analytical platforms based on MicroNIR spectroscopy combined with Chemometrics for forensic investigations on various complex samples taken under seizure. Recognizing the potential of Near Infrared Region (NIR) spectroscopy, an agreement between the Carabinieri Scientific Investigations Group (RaCIS) and the Chemistry Department of 'Sapienza' University led to a joint research plan. This plan aimed to devise a specific analytical protocol for samples taken and placed under seizure. Close collaboration with the Carabinieri Scientific Investigation Departments (RIS) of Messina facilitated the analysis of objects under forensic investigation, allowing the validation of models developed in laboratory simulations on real samples. The uniqueness of this research project lies in its experimental design, enabling the transfer of a tool developed by scientific research to law enforcement authorities—an avenue typically inaccessible to civilians without crime scene access. Notably, the RIS of Messina acquired the same instrumental apparatus, ensuring the reproducibility and transferability of the acquisition method for validation.
The added value of the portable NIR analytical approach lies in its ability to bring the laboratory to the 'field,' enabling on-site acquisitions and providing reliable real-time information. This advancement could empower law enforcement personnel to make timely decisions about whether to proceed with investigations, thereby relegating traditional analysis performed by forensic laboratories to a secondary, comparative role. The success of this research project could redefine the traditional laboratory's role, allowing for the elimination of unnecessary analysis. The results obtained demonstrated rapid and accurate non-destructive identification, eliminating the need for sample pretreatment, and avoiding the use of solvents. Furthermore, the simplicity of the approach makes it accessible even to non-experts, requiring only training for operators.
The studies conducted over the past three years of my PhD program have been united by a single common thread: the development and optimization of MicroNIR/Chemometrics platforms to support law enforcement personnel in forensic investigations. For each project, the initial step involved studying different matrices for the calibration of the spectral response. Subsequent steps included in-depth literature studies to identify the most descriptive range of wavelengths in the electromagnetic spectrum for use in predictive models. The main areas of application explored in these years include the following points:
1. MicroNIR/Chemometrics for bloodstains identification
During the first year of doctoral work, the attention was focused on the development and optimization of a MicroNIR/Chemometrics platform designed for the rapid and automatic identification of blood traces, regardless of the deposition surface. The optimized model demonstrated the capability to discriminate between genuine bloodstains and potential false positives found at crime scenes.
To investigate the variability of the matrix and real-life circumstances, blood from three volunteers and five red substances easily mistaken for blood were deposited on seven different surfaces immediately after collection. Various operating conditions were explored, including different deposition volumes to recreate blood traces in three distinct forms and varying acquisition distances. Spectra recording was conducted both in contact and at appropriate distances, serving the dual purpose of assessing the instrument's detection capabilities in contactless mode and, more importantly, avoiding specimen contamination.
Each investigated distance yielded a valid and appreciable response. Considering the assumptions made, working in contactless mode was deemed optimal to ensure the repeatability of the preliminary spectroscopic technical assessment and prevent contamination of the trace, which could undergo additional investigations.
2. MicroNIR/Chemometrics for the determination of drugs
During the second and third year of research, chemometric models were developed to quantify the ∆9-tetrahydrocannabinol and cannabidiol content in samples of cannabis and hashish, as well as the characterisation of cocaine samples seized by law enforcement officers. The samples were initially processed with reference methods, and the results were compared with those obtained by NIR spectroscopic analysis.
The considered matrices produced NIR spectra that were inherently complex and challenging to interpret. To extract the most significant information from them, multivariate analysis was employed. The first step involved an exploratory investigation using Principal Component Analysis (PCA), which graphically revealed correlations between samples and between variables. Building upon the promising PCA results, Partial Least Squares Regression (PLSr) was used to construct efficient prediction models. The models' performances were evaluated by studying relevant figures of merit and predicting a set of unknown samples.