Machine learning assisted LIBS classification of burnt and unburnt paper samples: A forensic perspective

Area of Science:

• Forensic Science
• Analytical Chemistry
• Spectroscopy

Background:

• Arson investigation requires analyzing fire-affected materials.
• Laser-Induced Breakdown Spectroscopy (LIBS) offers potential for chemical analysis of burnt items.
• Paper documents are common in forensic investigations.

Purpose of the Study:

• To explore the forensic potential of LIBS for analyzing burnt paper.
• To evaluate LIBS for identifying elemental composition changes after burning.
• To assess machine learning's role in LIBS-based forensic classification.

Main Methods:

• LIBS spectra were recorded for unburnt paper samples using a Nd: YAG laser (532 nm).
• Plasma parameters (temperature, electron number density) were analyzed to validate local thermodynamic equilibrium.
• Paper samples were burnt using different ignition sources (lighter, candle, gas stove).
• Machine learning algorithms (PCA, LDA, QDA, Linear SVM) were applied to LIBS data.

Main Results:

• LIBS analysis revealed consistent elemental composition across burnt paper samples regardless of ignition source.
• Principal Component Analysis (PCA) showed slight sample discrimination.
• Supervised algorithms (LDA, QDA, Linear SVM) achieved 100% classification accuracy.

Conclusions:

• Machine learning-assisted LIBS is a highly effective tool for forensic applications.
• LIBS can analyze elemental composition for forensic casework.
• This technique shows promise for identifying burnt paper and potentially other materials.