Development and Validation of a Method for Individualization of Middle Petroleum Distillates and Kerosene Ignitable Liquids

• NCJ Number: 240686
• Author(s): J. Graham Rankin, Ph.D; Peter Harrington, Ph.D.
• Date Published: November 2012
• Length: 109 pages
• Grant Number(s): 2008-DN-BX-K146
• Agencies: NIJ-Sponsored
• Publication Type: Report (Study/Research)

Annotation

This project applied target compound analysis from high-resolution gas chromatography/mass spectrometry (GCMS) in individualizing gasoline to medium petroleum distillates (MPD) and kerosene, which may be used as accelerants in arson cases.

Abstract

The intent was to test analytical tools for use in forensic comparisons between ignitable liquid residues in fire debris and ignitable liquids in the possession of arson suspects. The Fuzzy Rule-building Expert System (FuRES) models produced correct classification rates greater than 90 percent for discriminating between samples. Projected Difference Resolution (PDR) mapping technique was applied to measure the quantitative differences among the ignitable liquid (IL) samples by their GCMS profiles. PDR mapping, a new method for characterizing complex data-sets was consistent with the FuRES classification result. For MPD, the Kendal's-tau metric for "association/no association" demonstrated "association/ no association" true positive rates in excess of 95 percent and 5-percent false positive rates. In the course of conducting this research, a large set of kerosene and MPD samples were collected and analyzed by high-resolution GCMS. Researchers selected a number of candidate target compounds common to most samples that elute in pairs relatively close in elution time and have repeatable peak area ratios within samples, but are variable between samples of different origin. Multivariate statistical methods were used to determine those candidate peak ratios that are best for discriminating between samples for kerosene and MPD. The "false positive rate" was determined when comparing all the samples in the dataset. The robustness of the method was determined for evaporated and simulated burned samples. 7 figures, 3 tables, and 19 references