ASSOCIATION OF FIRED CARTRIDGE RESIDUES TO UNBURNED SMOKELESS POWDERS USING GC-MS AND MULTIVARIATE STATISTICAL PROCEDURES By Rebecca L. Boyea Michigan State University in partial fulfillment of the requirements Forensic Science – Master of Science A THESIS Submitted to for the degree of 2020 ABSTRACT ASSOCIATION OF FIRED CARTRIDGE RESIDUES TO UNBURNED SMOKELESS POWDERS USING GC-MS AND MULTIVARIATE STATISTICAL PROCEDURES By Rebecca L. Boyea Forensic analysis of smokeless powders has historically focused on the analysis of unburned powder or gunshot residue. The analysis of fired cartridge residues and subsequent statistical association to the corresponding unburned powder has only recently been investigated. Previous work in our laboratory employed liquid chromatography-time of flight-mass spectrometry (LC-TOFMS) and chemometric procedures to investigate association of fired cartridge residues to the corresponding unburned powders.2 While successful association was achieved for some powders, LC-TOFMS is not readily available in forensic laboratories. A widely available alternative is gas chromatography-mass spectrometry (GC-MS). The work presented here demonstrates the use of GC-MS for the analysis of unburned powder and fired cartridge residues, followed principal components analysis (PCA) and hierarchical cluster analysis (HCA) to investigate association and differentiation of fired cartridge residues to the corresponding unburned powder. Both PCA and HCA resulted in distinct groupings of the unburned powders, based largely on the abundance of ethyl centralite and dibutyl phthalate. Despite variability and decreased abundances observed in all fired cartridge residues, successful association of the fired cartridges to the corresponding unburned powder was possible but was limited by the original composition of the unburned powder. Overall, this work demonstrates that GC-MS and chemometric procedures are effective tools for the association of fired cartridge residues and unburned powders. ACKNOWLEDGEMENTS I would first like to thank my advisor, Dr. Ruth Smith, for her guidance and support throughout my time at MSU. I truly appreciate all the advice you have given me over these past few years, whether it was related to research, presentations, interviews, writing, life, or careers. I would also like to thank Dr. Sanja Kutnjak Ivkovich for serving on my committee, and Dr. Brian Hunter for both serving on my committee as well as for his assistance in sample collection. I am incredibly lucky to have so many wonderful female role models to turn to. Dr. Victoria McGuffin, thank you for your insight and for always encouraging me to remember my analytical chemistry training. Dr. Kathryn Severin, I am so grateful for the instrumentation experience that I gained while working with you. Thank you for teaching me, and for trusting me. To my fellow forensic chemists (Natasha, Emma, Hannah, Bri, Amber, Otyllia, and Amanda) what an adventure this has been. From near-missed flights and football games to Christmas parties and Costco trips, thank you for laughing and learning with me. I am forever thankful for the love, support, patience, and encouragement that my friends and family have given me throughout this journey. To Morgan, and to Olivia, Carol, and Steve, thank you for making Michigan feel a little more like home. I appreciate it more than you know, and all of my favorite Michigan memories involve you. And of course (though he could not care less) I am thankful for my little furball of chaos. Over 90% of this thesis was written with Salem sleeping by my side, and the rest was written despite him attacking my fingers as I typed. Finally, to Mom, Dad, and Adam – a girl really couldn’t ask for a better support system than you. Thank you for giving me the world. iii TABLE OF CONTENTS LIST OF TABLES ......................................................................................................................... vi LIST OF FIGURES ...................................................................................................................... vii 1. Introduction ................................................................................................................................. 1 1.1 Composition and Manufacturing of Smokeless Powders ..................................................... 1 1.2 Methods of Analysis for Smokeless Powders ....................................................................... 8 1.2.1 Morphology and Inorganic Composition of Smokeless Powders .................................. 8 1.2.2 Organic Composition of Smokeless Powders ................................................................ 9 1.2.3 Methods of Data Preprocessing ................................................................................... 12 1.3 Multivariate Statistical Procedures ..................................................................................... 13 1.3.1 Principal Components Analysis ................................................................................... 13 1.3.2 Hierarchical Cluster Analysis ...................................................................................... 15 1.3.3 Application of Multivariate Statistical Procedures for Smokeless Powder Association ............................................................................................................................................... 18 1.4 Research Objectives ............................................................................................................ 19 REFERENCES ......................................................................................................................... 20 2. Materials and Methods .............................................................................................................. 26 2.1 Reference Materials and Solvents ....................................................................................... 26 2.1.1 Preparation of Reference Standards ............................................................................. 26 2.2 Unburned Smokeless Powder ............................................................................................. 28 2.2.1 Smokeless Powder Sample Set .................................................................................... 28 2.2.2 Extraction of Unburned Smokeless Powder ................................................................ 29 2.3 Fired Cartridge Residue ...................................................................................................... 30 2.3.1 Collection of Fired Cartridges ..................................................................................... 30 2.3.2 Extraction of Fired Cartridge Residue ......................................................................... 31 2.4 Gas Chromatography-Mass Spectrometry (GC-MS) Analysis .......................................... 32 2.5 Data Processing ................................................................................................................... 33 2.6 Multivariate Statistical Analysis ......................................................................................... 34 2.6.1 Principal Components Analysis (PCA) ....................................................................... 34 2.6.2 Hierarchical Cluster Analysis (HCA) .......................................................................... 35 APPENDIX ............................................................................................................................... 36 REFERENCES ......................................................................................................................... 38 3. Association and Differentiation of Unburned Powders using Multivariate Statistical Analysis of Organic Compounds ................................................................................................................. 40 3.1 Introduction ......................................................................................................................... 40 3.2 Organic Composition of Unburned Smokeless Powders by GC-MS ................................. 41 3.3 Classification of Unburned Powders based on Organic Composition ................................ 49 3.3.1 Organic Composition of Unburned Powders, as determined by GC-MS .................... 50 3.3.2 Comparison of Organic Profiles of Unburned Powder: GC-MS and LC-TOFMS ..... 55 iv 3.4 Association and Differentiation of Unburned Powders using Principal Components Analysis .................................................................................................................................... 57 3.4.1 Principal Components Analysis of Unburned Powders, based on GC-MS data ......... 58 3.4.2 Comparison of PCA of Smokeless Powders using GC-MS to PCA using LC-TOFMS data ........................................................................................................................................ 67 3.5 Association and Differentiation of Unburned Powders using Hierarchical Cluster Analysis .................................................................................................................................................. 71 3.6 Summary of the Association and Differentiation of Unburned Powders ........................... 77 APPENDIX ............................................................................................................................... 79 REFERENCES ......................................................................................................................... 92 4. Association and Discrimination of Fired Cartridge Residues to Unburned Powders using Multivariate Statistical Analysis of Organic Compounds ............................................................ 94 4.1 Introduction ......................................................................................................................... 94 4.2 Comparison of the Organic Composition of Fired Cartridge Residues to the corresponding Unburned Powders .................................................................................................................... 95 4.3 Principal Components Analysis for the Association of Fired Cartridge Residues to Unburned Powders .................................................................................................................. 106 4.3.1 Principal Components Analysis of a Subset of Unburned Powders .......................... 106 4.3.2 Principal Components Analysis of Fired Cartridge Residues .................................... 109 4.3.3 Comparison of PCA of Smokeless Powders using GC-MS to PCA using LC-TOFMS data ...................................................................................................................................... 117 4.4 Hierarchical Cluster Analysis for the Association of Fired Cartridge Residues to Unburned Powders ................................................................................................................................... 118 4.4.1 Hierarchical Cluster Analysis of a Subset of Unburned Powders ............................. 118 4.4.2 Hierarchical Cluster Analysis of Fired Cartridge Residues ....................................... 120 4.4.3 Comparison of HCA performed using GC-MS data to HCA performed using LC- TOFMS data ........................................................................................................................ 128 4.5 Summary ........................................................................................................................... 129 APPENDIX ............................................................................................................................. 131 REFERENCES ....................................................................................................................... 138 5. Conclusions and Future Work ................................................................................................ 140 5.1 Conclusions ....................................................................................................................... 140 5.2 Future Work ...................................................................................................................... 141 REFERENCES ....................................................................................................................... 145 v LIST OF TABLES Table 1.1 Compound Categories, adapted from Goudsmits et al.14 ............................................... 7 Table 1.2 Hypothetical distance matrix for six samples, A-F. ..................................................... 15 Table 1.3 Distance matrix after cluster EF is formed .................................................................. 16 Table 2.1 Commercial Smokeless Powder Samples Selected For Analysis ................................ 29 Table 3.1 Unburned Smokeless Powder Profiles based on Category 1 Compounds ................... 51 Table 3.2 Unburned Smokeless Powder Profiles based on Category 1 and 2 Compounds ......... 52 Table A3.1 Composition of Unburned Powders, as determined by GC-MS ............................... 80 Table A3.2 Approximate Retention Times of Compounds .......................................................... 81 Table A3.3 Chemical Composition of Unburned Smokeless Powders, as determined by LC-TOFMS9 ................................................................................................................................. 91 Table 4.1 Organic Profiles of MG-nL Unburned and Fired Cartridge Residues, as determined by GC-MS .......................................................................................................................................... 97 Table 4.2 Organic Profiles of Rem-L Unburned and Fired Cartridge Residues, as determined by GC-MS ........................................................................................................................................ 100 Table 4.3 Organic Profiles of Win-L Unburned and Fired Cartridge Residues, as determined by GC-MS ........................................................................................................................................ 102 Table 4.4 Organic Profiles of Win-nL Unburned and Fired Cartridge Residues, as determined by GC-MS ........................................................................................................................................ 104 vi LIST OF FIGURES Figure 1.1 Common smokeless powder morphologies (a) tubular, (b) ball, and (c) disc .............. 3 Figure 1.2 Example distribution system for smokeless powders ................................................... 5 Figure 1.3 Resultant HCA Dendrogram ...................................................................................... 17 Figure 3.1 Representative chromatogram of unburned MG-nL, showing two peaks provisionally identified as nitroglycerin ............................................................................................................. 42 Figure 3.2 Mass spectra of the two peaks provisionally identified as nitroglycerin, at retention times (a) 10.6 min and (b) 11.0 min .............................................................................................. 43 Figure 3.3 Literature mass spectra of (a) ethylene glycol dinitrate and (b) propylene glycol dinitrate2 ........................................................................................................................................ 45 Figure 3.4 Nitrosation of diphenylamine ..................................................................................... 48 Figure 3.5 Composition of unburned powders as determined by GC-MS, compared to composition determined by LC-TOFMS. ..................................................................................... 56 Figure 3.6 Principal components analysis (PCA) of unburned powders using Category 1 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 ......................... 59 Figure 3.7 Principal components analysis (PCA) of unburned powders using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 ......................... 62 Figure 3.8 Principal components analysis (PCA) of unburned powders using Category 1 and 2 compounds (a) scores plot of PC1 vs PC3 and (b) loadings plot of PC1 vs PC3 ......................... 63 Figure 3.9 Principal components analysis (PCA) of unburned powders using LC-TOFMS data (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC29 ........................................... 68 Figure 3.10 HCA dendrogram of Unburned Powders using Category 1 Compounds. Colors correspond to the legend of Figure 3.6. ........................................................................................ 72 Figure 3.11 Hierarchical cluster analysis (HCA) dendrogram of unburned powder extracts using Category 1 and 2 compounds. Colors correspond to the legend of Figure 3.7. ............................ 75 Figure A3.1 Representative mass spectrum of nitroglycerin ....................................................... 81 Figure A3.2 Representative mass spectrum of 2,4-dinitrotoluene ............................................... 82 Figure A3.3 Representative mass spectrum of diphenylamine .................................................... 82 vii Figure A3.4 Representative mass spectrum of methyl centralite ................................................. 83 Figure A3.5 Representative mass spectrum of ethyl centralite .................................................... 83 Figure A3.6 Representative mass spectrum of dibutyl phthalate ................................................. 84 Figure A3.7 Representative mass spectrum of 2-nitrodiphenylamine ......................................... 84 Figure A3.8 Representative mass spectrum of 1-methyl-3,3-diphenylurea (Akardite II) ........... 85 Figure A3.9 Representative mass spectrum of 4-nitrodiphenylamine ......................................... 85 Figure A3.10 Representative Chromatogram of Unburned Powder 22-O (1) ............................. 86 Figure A3.11 Representative Chromatogram of Unburned Powder 22-O (2) ............................. 86 Figure A3.12 Representative Chromatogram of Unburned Powder 44-N ................................... 87 Figure A3.13 Representative Chromatogram of Unburned Powder 44-O ................................... 87 Figure A3.14 Representative Chromatogram of Unburned Powder AA-O (1) ........................... 88 Figure A3.15 Representative Chromatogram of Unburned Powder AA-O (2) ........................... 88 Figure A3.16 Representative Chromatogram of Unburned Powder AA-O (3) ........................... 89 Figure A3.17 Representative Chromatogram of Unburned Powder BZR-nL ............................. 89 Figure A3.18 Representative Chromatogram of Unburned Powder MG-O ................................ 90 Figure A3.19 Representative Chromatogram of Unburned Powder SB-N .................................. 90 Figure 4.1 Representative chromatograms of MG-nL (a) unburned powder and (b) fired cartridge residue ............................................................................................................................ 98 Figure 4.2 Representative chromatograms of Rem-L (a) unburned powder and (b) fired cartridge residue ......................................................................................................................................... 101 Figure 4.3 Representative chromatograms of Win-L (a) unburned powder and (b) fired cartridge residue ......................................................................................................................................... 103 Figure 4.4 Representative chromatograms of Win-nL (a) unburned powder and (b) fired cartridge residue .......................................................................................................................... 105 viii Figure 4.5 Principal components analysis (PCA) of the unburned powder subset using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 .......... 107 Figure 4.6 Principal components analysis (PCA) of fired cartridge residues using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 .................... 110 Figure 4.7 Principal components analysis (PCA) scores plots of the unburned powder subset and MG-nL fired cartridge residue (a) fired MG-nL cartridge 4 and (b) fired MG-nL cartridge 2 .. 113 Figure 4.8 Principal components analysis (PCA) scores plots of unburned powder subset and one Win-nL fired cartridge residue ............................................................................................. 114 Figure 4.9 Principal components analysis (PCA) scores plot of the unburned powder subset and one fired cartridge residue using Category 1 and 2 compounds (a) fired Win-L and (b) the first fired Rem-L residue .................................................................................................................... 116 Figure 4.10 Hierarchical cluster analysis (HCA) dendrogram of unburned powder subset, based on GC-MS data ........................................................................................................................... 119 Figure 4.11 Hierarchical cluster analysis (HCA) dendrogram of the unburned powder subset with all fired cartridge residues (indicated by black dots) introduced ........................................ 121 Figure 4.12 Hierarchical cluster analysis (HCA) dendrograms of the unburned powder subset with fired MG-nL (a) cartridge 1 and (b) cartridge 4 introduced (indicated by black dot) ........ 124 Figure 4.13 Exemplar HCA dendrogram of the unburned powder subset with one fired Win-nL cartridge introduced (indicated by black dot) ............................................................................. 125 Figure 4.14 Exemplar HCA dendrograms of the unburned powder subset with (a) one fired Win-L cartridge and (b) Rem-L cartridge 1 introduced (indicated by black dot) ...................... 127 Figure A4.1 Total organic composition of the powders included in the subset of unburned powders, as determined by GC-MS ............................................................................................ 132 Figure A4.2 Total organic composition of all fired cartridge residues, as determined by GC-MS ..................................................................................................................................................... 133 Figure A4.3 Principal components analysis of the unburned powder subset and successive fired cartridge residues using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 ....................................................................................................... 135 Figure A4.4 Loadings plots for PCA of fired cartridge residue MG-nL C4 .............................. 135 Figure A4.5 Loadings plots for PCA of fired cartridge residue MG-nL C2 .............................. 136 Figure A4.6 Loadings plots for PCA of fired cartridge residue Win-nL C3 ............................. 136 ix Figure A4.7 Loadings plots for PCA of fired cartridge residue Win-L C1 ............................... 137 Figure A4.8 Loadings plots for PCA of fired cartridge residue Rem-L C1 ............................... 137 x 1. Introduction Firearm-related violence is a prominent and expensive public health problem, accounting for more than 32,000 deaths per year and more than $48 billion annually in medical and work costs in the United States alone.1 From 2010 to 2018, the Centers for Disease Control and Prevention estimated that there were more than 1.17 million cases of firearm violence in the United States, of which approximately 319,240 involved fatalities.2 Evidence collected as a part of a firearm-related incident may include a spent ammunition cartridge, the projectile itself, or trace evidence resulting from the explosive charge. In cases where the projectile is not recovered or is too damaged for comparison, the analysis of fired cartridge residue and gunshot residue (GSR) is another method to evaluate ammunition evidence. 1.1 Composition and Manufacturing of Smokeless Powders Modern ammunition is composed of the primer, propellant (commonly black or smokeless powder), and bullet that are all encased in a metal cartridge. When the firing pin of a firearm strikes the primer, this creates a spark and ignites the primer. The ignited primer then causes the propellant to ignite. As the powder burns, the gaseous products of combustion cause pressure to rapidly build up in the enclosed cartridge. When the pressure can no longer be contained in the cartridge, the bullet is ejected and propelled down the barrel of the firearm.3 Both black powder and smokeless powder are classified as low explosives, indicating these powders deflagrate rather than detonate. While detonation involves decomposition of a substance via the propagation of supersonic shock waves, deflagration involves combustion via subsonic heat transfer.4 1 The propellant in modern ammunition and many improvised explosive devices is known as smokeless powder. Smokeless powders, named because they produce much less smoke than traditional black powder, largely replaced black powder in the mid-19th century.5 The primer for smokeless powders is typically composed of an initiating explosive (typically lead styphnate), an oxidizing component (such as barium nitrate), and a fuel source (antimony sulfate). An increase in concern regarding the toxicity of lead has resulted in the development of non-leaded primers, also called non-toxic primers. In 1986, the German company RWS® (now RUAG Ammotec) developed Sintox, one of the first non-toxic primers. Sintox contained diazole (2-diazo-4,6-dinitrophenol) in place of lead styphnate as a primary explosive material, zinc peroxide as an oxidizer, and titanium as a fuel source.6 Other non-toxic primer mixtures include diazole with tetrazine, barium nitrate, potassium nitrate, nitrocellulose or a nitrate ester such as pentaerythritol tetranitrate (PETN).5 Smokeless powders contain one or more energetic materials. When nitrocellulose is the only explosive component, the powders are known as single-based powders. Powders containing nitrocellulose and nitroglycerin are double-based powders, and powders that contain nitrocellulose, nitroglycerin and nitroguanidine are triple-based powders. Triple based powders are generally reserved for military use and are not widely available for civilian use. The morphology of smokeless powder kernels can sometimes indicate whether a powder is single- or double-based. Most powders with tubular morphology are single-based, whereas ball powders and disc powders are generally double-based (Figure 1.1).7 Other common powder morphologies include lamella, flattened ball, and flake.8,9 2 a) b) c) Figure 1.1 Common smokeless powder morphologies (a) tubular, (b) ball, and (c) disc In addition to the explosive component(s), smokeless powders also consist of various organic additives that are used to modify burn characteristics of the propellant. Deterrents such as ethyl centralite, dibutyl phthalate and Akardite II are added to slow the burning rate of the powder, resulting in a more consistent, controlled burn of the propellant. Diphenylamine is a common stabilizer. Diphenylamine prevents or decreases the rate of autocatalytic decomposition by combining with nitrogen acids and oxides that are produced by the decomposition of nitrocellulose.10 Plasticizers such as dibutyl phthalate and dinitrotoluene are integrated into the powders to assist with flexibility; this helps prevent kernel fracture during combustion, which would result in increased surface area and contribute to inconsistent burn rates.11 Many additives can serve multiple purposes; for example, ethyl centralite can act as a stabilizer and a plasticizer in addition to being used as a deterrent. The various additives are incorporated during the manufacturing process. The primary method of producing smokeless powders begins by dissolving nitrocellulose in an organic solvent such as acetone or ether-alcohol.11 Different additives are then mixed with the nitrocellulose.7 For non-ball double-based powders, nitroglycerin is also added at this time. Adding nitroglycerin increases the explosive power, but also increases chemical stability.11 The resulting plastic dough is then formed into the desired shapes, followed by the careful evaporation of solvent. Single-based powders are generally shaped by extrusion and then cut to the desired length, while double-based powders are pressed into blocks and then can be cut or 3 shaped by extrusion.12 The evaporation process must be carried out slowly to prevent any warping, as changes to grain shape will impact the ballistic properties of the powder.11 The remaining solid grains need to retain some elasticity, which help to prevent fracture of the grains during combustion. The manufacture of ball powder is more specialized.8 Nitrocellulose is mixed with solvent and the various additives (primarily stabilizers) to form a dough. The dough is shaped into spheres and the solvent evaporated and then nitroglycerin is impregnated into the ball powders.7 Regardless of manufacturing method, the grains of smokeless powder are screened for consistency of size and shape, and then coated with deterrents, flash suppressants, and graphite before batch mixing to promote homogeneity.7 During the screening process, batches of smokeless powder that are deemed unsatisfactory are recycled and re-processed as part of a different lot of smokeless powders.7 Surplus powders can also be re-used in this manner.8 This introduces compositional heterogeneity in kernels that are produced within a batch, as well as between different batches from the same manufacturer.13 Consequently, while the propellant in different cartridges within a single box of ammunition contains the same compounds, the percentages of the various compounds in each cartridge may differ slightly due to unavoidable heterogeneity of the manufacturing process.13 After a batch of smokeless power is manufacture, the product can be distributed in many different ways (Figure 1.2).7,8 Of a particular batch of smokeless powder, a majority will be sold to original equipment manufacturers (OEMs) and immediately put into ammunition.8 Some powder manufacturers have in-house packaging facilities that prepare the ammunition for sale, while other manufacturers sell the powder to different re-packagers such as Hogdon®. A smaller 4 proportion of the batch is sold in cannisters to hunting clubs or to commercial retail outlets for re-loading by personal users.8 s r e r u t c a f u n a M r e d w o P Domestic and Foreign Militaries Re-packagers, Powder Manufacturers with packaging facilities Wholesalers Retail Outlets, Gun Clubs ) n g i e r o F & c i t s e m o D ( OEMs Figure 1.2 Example distribution system for smokeless powders Powder manufacturers may change their smokeless powder formulation over time, re-packagers may buy ammunition from different manufacturers, and a particular batch of smokeless powder may not be sold to a single distributor. Due to this distribution system, two different boxes of the same apparent brand of ammunition may have different smokeless powder compositions. Conversely, two different ammunition brands may contain smokeless powders with the same composition. As such, identification of a particular brand of smokeless powder may not be possible, and in fact not necessarily beneficial in a forensic investigation. Rather, analysis should focus on comparisons of the actual composition of the powders, with the goal of associating similar compositions of powders rather than associating to a particular brand.9 Some compounds in smokeless powders also have other applications or are common environmental contaminants. For example, nitroglycerin is a common medication prescribed as a 5 vasodilator, and diphenylamine is a common pollutant that has been detected in rubbers and on produce, in addition to its application as an explosives stabilizer.14 While the individual detection of such compounds may not indicate the presence of explosive residues, added confidence can be gained when multiple of these compounds are detected together, or with the addition of other compounds commonly found in explosive materials. Previously, only ethyl centralite and methyl centralite were considered to be unique identifiers of smokeless powders, but this resulted in false negatives.15 Following a review of smokeless powder literature, Goudsmits et al. proposed three categories of organic compounds, based on the perceived evidentiary value of each compound (Table 1.1).14 Category 1 consisted of compounds that were considered to have the highest forensic value, as the compounds have either minimal or restricted applications unrelated to explosives residues. Category 2 contained compounds that are also commonly found in explosive residues, but their use outside of explosives is less restricted than the compounds in Category 1. Finally, Category 3 contained compounds that have been associated with explosive residue but have additional common applications. 6 Table 1.1 Compound Categories, adapted from Goudsmits et al.14 Category Category 1 Compounds that are very strongly associated with GSRs with very restricted applications unrelated to GSR Category 2 Compounds that are strongly associated with GSRs, but which have less restricted applications unrelated to GSR Category 3 Compounds that are associated with GSR, but which are detected less frequently and have less restricted applications unrelated to GSR Compounds Nitroglycerin Nitroguanidine Methyl Centralite Ethyl Centralite 2,4-Dinitrotoluene Akardite II 2-Nitrodiphenylamine 4-Nitrodiphenylamine Diphenylamine + nitrated derivatives Nitrocellulose Other nitrotoluenes (2-NT, 2,6-DNT, TNT, etc.) Other diphenylamine derivatives (N-nitrosodiphenylamine, 2,4-dinitrodiphenylamine, etc.) Triacetin The categories were originally developed to serve as indicators of compounds that could be considered as confirmatory for the presence of gunshot residue. However, the prevalence and application of the different compounds can also be considered when examining the composition of unburned powders and fired cartridge residues. In addition to serving as confirmatory markers for the presence of smokeless powders, powders can be compared based on the compounds that have the most evidentiary value. 7 1.2 Methods of Analysis for Smokeless Powders 1.2.1 Morphology and Inorganic Composition of Smokeless Powders Analysis of smokeless powders can include examination of the unburned powder, fired cartridge residue remaining in the ammunition case after the weapon has been discharged, and gunshot residue deposited on the shooter or target. One of the earliest methods to differentiate smokeless powders is based on morphology of unburned or partially burned powder.7 Morphology and micrometry of unburned powders are helpful in combination with other forms of analysis, but alone, do not provide enough information to differentiate among powders. Additionally, single-kernel analysis is time consuming, and morphologies are not necessarily specific to a particular manufacturer. There is also a certain degree of heterogeneity in kernel shape due to changes that occur during the manufacturing process, namely during the evaporation of solvent.11 In addition to morphology, the inorganic components of gunshot residue have been extensively studied. Scanning electron microscopy coupled to energy dispersive X-Ray spectroscopy (SEM-EDS) is most commonly used to identify gunshot residue, based on the presence of spherical particles of lead, barium and antimony originating from the primer.16,17 The inorganic material has also been studied using inductively coupled plasma optical emission spectrometry (ICP-OES) and atomic absorption spectrometry to compare different brands of ammunition.18 However, the characteristic inorganic material is not present in unburned powder, as the propellant powder has not been exposed to the primer. Consequently, SEM-EDS is not helpful for the analysis of unburned powder. Additionally, as primer formulations progress to eliminate toxic components such as lead and other heavy metals, these characteristic particles are no longer present and thus are no longer helpful in identifying the presence of gunshot residue.17 8 1.2.2 Organic Composition of Smokeless Powders To compensate for the changing primer formulations, the organic components of smokeless powders have also been studied. Rather than solely identifying a particle as being consistent with gunshot residue, examination of organic composition is also applicable to unburned powder and fired cartridge residue. A standard methodology for the analysis of the organic components of smokeless powder has not been established, but the most common methods include liquid chromatography (LC) or gas chromatography (GC). These methods of separation are often coupled to mass spectrometry to allow identification of separated compounds. An advantage of LC is that it does not require elevated temperatures, which is beneficial in the analysis of thermally labile compounds. Reversed-phase separations of smokeless powder samples are commonly performed using a C18 column with a mobile phase gradient.9,19–25 McCord et al. reported a reversed-phase gradient separation that resulted in improved separation of geometric isomers such as dinitrotoluenes and nitrodiphenylamines, which was challenging under isocratic conditions.26 Different detection systems have been coupled to LC for the analysis of smokeless powders, namely ultraviolet (UV) detection and mass spectrometry.9,19–32 Mass spectrometry can employ various ionization methods, with most of the literature on LC-MS applications to smokeless powders focusing on electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI).20–23,28–30 Popular additives such as ethyl centralite, dibutyl phthalate, Akardite II and diphenylamine have successfully been detected by LC-MS. Various nitrated derivatives of diphenylamine have also been detected by LC-MS.30,33 However, while the majority of compounds in smokeless powders are detected using positive ionization, 9 nitroglycerin and 2,4-dinitrotoluene were only detected following additional analysis using negative ionization.23,34 Gas chromatography (GC) has also been used for the analysis of smokeless powders, often performed using non-polar stationary phases, such as 100% dimethylpolysiloxane or 5% phenyl, 95% methylpolysiloxane. Splitless injections are common due to the potential presence of minor components that might not be detected with a split injection.9 A consideration of smokeless powder analysis is that the compounds present are thermally labile and can degrade rapidly when exposed to elevated temperatures. This is of particular concern when high GC inlet temperatures are used. Degradation of compounds in the inlet could lead to degradation products being identified as characteristic of the powder, when they are actually just an artifact of the instrument parameters. In addition to the creation of decomposition products, the abundance of the original compound would decrease or be depleted entirely. When comparing smokeless powder samples, these differences could result in incorrect association or differentiation. To address this problem, limited research has been performed using programmed temperature vaporizing (PTV) injectors, which is preferred for the analysis of thermally labile compounds.9,33–36 A PTV injector starts at a cooler temperature than a normal injection temperature, and then involves rapid heating to volatilize the analytes onto the GC column. The more volatile compounds can thus be introduced onto the column at a lower temperature, which helps to prevent thermal degradation in the injection port. However, GC-MS instrumentation in forensic laboratories do not often have PTV injectors, as this feature increases instrument cost. Instead, an injection temperature of approximately 175 °C is commonly used in the analysis of smokeless powders by GC.9 10 While other detection methods have been studied, the most common system involves the coupling of gas chromatography with mass spectrometry. Gas chromatography-mass spectrometry (GC-MS) is widely available in crime laboratories and is thus an attractive option for forensic applicability. Extensive research has been performed on the use of GC-MS for the analysis of smokeless powders.37–48 Research often focuses on the analysis of gunshot residue, which may be collected as forensic evidence in cases of suspected firearm violence and is currently the only form of smokeless powder evidence to have an established, widely accepted protocol for analysis (i.e., SEM-EDS). Though the sample preparation processes may differ (e.g., headspace extraction rather than solvent extraction), GC-MS analysis is not limited to gunshot residue but is also applicable to unburned powders and fired cartridge residues. One recent example by Goudsmits et al. employed GC-MS to analyze the organic composition of unburned powders and corresponding gunshot residue, followed by SEM-EDS of the same gunshot residue samples.38 As the gunshot residue samples were subjected to solid- phase microextraction (SPME) rather than solvent extraction, the methodology allowed for the determination of both organic and inorganic composition of the same gunshot residue sample. However, as non-toxic ammunition becomes more common, the characteristic inorganic compounds and morphology of gunshot residue are not present, and thus inorganic analysis may not be beneficial. When considering application to forensic laboratories, GC-MS has many advantages. GC-MS instrumentation is widely available and the theory is well understood. Additionally, GC-MS information is available for over 800 unburned powders in the Smokeless Powders Database that was developed by the Explosives Database Committee of the Technical Working Group for Fire and Explosions (TWGFEX).49 11 1.2.3 Methods of Data Preprocessing Recently, the focus of GC-MS analysis has progressed to include new and interesting ways of using the resultant chromatographic and mass spectral data to build profiles or draw comparisons among different powders. When investigating patterns within a data set, it is important to minimize any non-sample variance. McIlroy et al. investigated the impact of various pretreatment procedures on multivariate statistical analyses of chromatographic data, and determined that retention time alignment and normalization of the data allowed for the greatest improvement on multivariate association.50 Background subtraction and smoothing offered only minimal improvement.50 Normalization processes can be used to correct for systematic differences due to normal instrumental variances such as variations in flow rate, temperature and syringe dwell time.50 One method of normalization involves dividing every peak by the abundance or height of a specific peak in the chromatogram. This specific peak is often an internal standard, which must be sufficiently non-volatile so that evaporation does not affect the concentration of the added internal standard during analysis. It is assumed that the internal standard is present at a constant concentration across all samples, so that the variation in abundance or peak height is due only to instrumental variance. Some types of data also benefit from a scaling procedure. Scaling can be used to adjust the relative importance of variables in a data set. One such scaling method is known as Pareto scaling, which is a scaling method based on data dispersion.51 For each variable, individual values are mean-centered and then divided by the square root of the standard deviation of that particular variable. Each variable is scaled independently of the other variables. Pareto scaling is useful for data sets that contain variables that have values much larger than other variables. 12 1.3 Multivariate Statistical Procedures 1.3.1 Principal Components Analysis Principal components analysis (PCA) can be used to visualize differences among samples in a data set.52 As an unsupervised method of data reduction, PCA is an unbiased way to reduce the number of original variables to a fewer number of variables that contribute most to the variance observed in the data. For GC-MS of smokeless powders, the variables are the peaks present in the chromatograms, corresponding to compounds detected in the samples. The data set is thus comprised of an abundance measurement of each compound for each sample analyzed. Before PCA is performed, the data are first mean-centered. For each variable, the mean abundance of the variable across all samples in the data set is subtracted from the abundance of the variable in each sample. This produces a data set that has a mean abundance of zero for each variable. Values that have a negative abundance after mean-centering indicate that the abundance was below the average of the data set. After the data have been mean-centered, a covariance matrix is calculated, and the eigenvalues and eigenvectors of the covariance matrix are determined. The eigenvalues describe the proportion of variance accounted for by each principal component (PC). Eigenvectors are weighting coefficients of each variable for each PC. Each PC is a linear combination of the original variables and the corresponding weighting coefficients. An example can be seen in Equation 1.1, where !!and !"represent two independent variables, and "! and "" represent the corresponding weighting coefficients. The total number of variables is represented by #. $="!!!+""!"+⋯"#!# 13 Equation 1.1 The linear combination of variables is used to calculate the score for each individual sample on a particular PC. This process is repeated for each PC, where the maximum number of PCs is equal to the number of samples or variables, whichever is smaller. The first PC always describes the maximum variance of the data set, while the second PC is orthogonal to the first and describes the second most variance of the data set. Each subsequent PC is orthogonal to the previous PCs. The results of PCA can be visualized using three types of plots – the scree plot, scores plots, and loadings plots. The scree plot is a line plot of the eigenvalues of each PC and is used to visualize the proportion of variance described by each PC. The scores plots are derived from the eigenvectors. Scores plots are represented as scatterplots, with each axis corresponding to a specific PC (e.g., PC1 vs PC2). The scores are the new coordinates of each of the samples on a particular PC. The positioning of a sample on each PC can be explained by examining the loadings plots, which are related to the eigenvalues. Each PC has a corresponding loadings plot that describes the contributions of each variable to the overall positioning on that PC. Examining the weighting of each variable on a particular PC can help to explain the positioning of samples on the scores plot. Samples with similar chemical compositions will have similar scores, resulting in similar positioning on the scores plot. Scores plots are thus helpful for visualizing similarities and differences among different samples that contain hundreds of variables. However, the identification of groups on the scores plot can sometimes be subjective. Another multivariate statistical procedure that can be used to investigate association of samples is hierarchical cluster analysis (HCA). 14 1.3.2 Hierarchical Cluster Analysis While PCA involves the calculation of a covariance matrix, HCA first determines a distance matrix that contains the distance from every sample to every other sample in the multidimensional data space. With agglomerative HCA, each sample begins as a separate cluster, followed by the sequential merging of similar clusters until all samples merge to form one large cluster. Similarity is based on the distance between the two samples, as visualized by the distance matrix. There are multiple methods by which distance can be measured. Euclidean distance is the metric commonly used, where distance is measured as a straight line drawn between two points (X1, Y1) and (X2, Y2) (Equation 1.2). The two samples that are the closest together (the shortest distance apart) are the most similar. (=)(+!−+")"+(.!−.")" Equation 1.2 In the example below, samples E and F are the closest together (Table 1.2). These two samples would then form cluster “EF”. Table 1.2 Hypothetical distance matrix for six samples, A-F. A B C D E F A 0 0.139 0.402 0.462 0.448 0.402 B 0 0.297 0.323 0.318 0.269 C 0 0.256 0.130 0.141 D 0 0.136 0.117 E F 0 0.054 0 15 After the formation of cluster EF, the distance matrix is recalculated. There are different ways to measure the distance between the cluster and the remaining samples. Some examples of linkage methods include single linkage, average linkage, and complete linkage. In complete linkage, the Euclidean distance is measured between the farthest pairs, and the clusters that have the overall shortest distance are joined together. The new distance matrix after the formation of cluster EF (Table 1.3) then illustrates that the next cluster would join sample D to cluster EF, forming cluster (EF, D). Table 1.3 Distance matrix after cluster EF is formed D EF A 0 0.139 B 0 C 0.402 0.297 0 A B C D 0.462 0.323 0.256 0 EF 0.448 0.318 0.141 0.136 0 16 This process is repeated until all samples belong to the same cluster. The resulting clusters are shown as a dendrogram (Figure 1.3). ) % ( l e v e L y t i r a l i m S i 0 — 50 — 100 — A B C D E F Figure 1.3 Resultant HCA Dendrogram The height at which different samples cluster is also represented as a measure of similarity (Equation 1.3). Similarity Level = 1 - Individual Distance Maximum Distance Equation 1.3 In the example above, the maximum distance is 0.462 and the individual distance between samples E and F is 0.054. The similarity level at which samples E and F cluster is then /1 − $.$&' $.'("2, or approximately 0.88. This can also be represented as percent similarity, 88%. The similarity level is a useful metric to compare the association between different samples. 17 1.3.3 Application of Multivariate Statistical Procedures for Smokeless Powder Association Both PCA and HCA have been used to investigate association of smokeless powders. Lennert and Bridge analyzed unburned powder by both GC-MS and direct analysis in real time mass spectrometry (DART-MS), and determined that the two techniques allowed for similar association and differentiation when the resultant data were subjected to PCA.53 Perez et al. used mass spectral data obtained from laser electrospray mass spectrometry (LEMS) of smokeless powders to investigate brand association.15,54 Both unburned powder and fired cartridge residues were separately analyzed, but association of the fired cartridge residue to the corresponding unburned powders was not investigated. Principal components analysis resulted in the successful differentiation of multiple brands of ammunition based on the mass spectral data.15,54 Reese et al. analyzed 18 unburned smokeless powders and the corresponding fired cartridge residues by liquid chromatography-time-of-flight mass spectrometry (LC-TOFMS), and the resultant chromatographic data was subjected to PCA.23,30 Successful association was achieved for some of the powders, but was limited by the original composition of the unburned powders.23 This investigation illustrated the potential for association of fired cartridge residue to the correct unburned powder. Reese et al. also investigated association of fired cartridge residues to the corresponding unburned powders using HCA of LC-TOFMS data.23 Results were similar to those achieved using PCA, with association success limited by the original composition of the unburned powders.23 Hierarchical cluster analysis was also used by Lennert and Bridge to investigate association of unburned powders following GC-MS and DART-TOFMS analysis, with both techniques resulting in successful association based on chemical composition of the unburned powders.53,55 However, fired cartridge residues were not collected, so the impact of the firing process on association was not investigated. 18 1.4 Research Objectives The specific goals of this thesis were (1) to analyze unburned powders and fired cartridge residues by GC-MS, (2) to apply multivariate statistical procedures (PCA, HCA) to the resultant chemical profiles to investigate statistical association of fired cartridge residues to the corresponding unburned powders, and (3) investigate the variability of the firing process and the subsequent impact on association success. Investigating the variability of the firing process included monitoring the variability of abundances of the observed compounds among multiple fired rounds of the same ammunition, as well as identifying any products that are formed as a result of the firing process. Principal components analysis was used to investigate differentiation of unburned powders and fired cartridge residues, while hierarchical cluster analysis was used to investigate association of fired cartridge residues to unburned powders. Differentiation of unburned powders using Category 1 compounds was also compared to differentiation achieved using all 14 compounds detected. 19 REFERENCES 20 REFERENCES (1) Fowler, K. A.; Dahlberg, L. L.; Haileyesus, T.; Annest, J. L. Firearm Injuries in the United States. Preventive Medicine 2015, 79, 5–14. https://doi.org/10.1016/j.ypmed.2015.06.002. (2) Centers for Disease Control and Prevention, National Center for Injury Prevention and Control. CDC Web-based Injury Statistics Query and Reporting System (WISQARS) https://www.cdc.gov/injury/wisqars/index.html (accessed Feb 24, 2020). (3) Meng, H.; Caddy, B. Gunshot Residue Analysis—A Review. J. Forensic Sci. 1997, 42 (4), 14167J. https://doi.org/10.1520/JFS14167J. (4) International Ammunition Technical Guidelines 01.40:2015(E). United Nations Office for Disarmament Affairs (UNODA) February 1, 2015. (5) Wallace, J. S. Chemical Analysis of Firearms, Ammunition, and Gunshot Residue, 2nd ed.; International Forensic Science and Investigation Series; CRC Press LLC, 2008. (6) Hagel, R.; Redecker, K. Sintox - A New, Non-Toxic Primer Composition by Dynamit Nobel AG. Propellants Explos. Pyrotech. 1986, 11 (6), 184–187. https://doi.org/10.1002/prep.19860110606. (7) Heramb, R. M.; McCord, B. R. The Manufacture of Smokeless Powders and Their Forensic Analysis: A Brief Review. Forensic Science Communications 2002, 4 (2). (8) National Research Council. Black and Smokeless Powders: Technologies for Finding Bombs and the Bomb Makers; National Academies Press: Washington, D.C., 1998. https://doi.org/10.17226/6289. (9) Moorehead, W. Characterization of Smokeless Powders. In Forensic Analysis on the Cutting Edge; Blackledge, R. D., Ed.; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2007; pp 241–268. https://doi.org/10.1002/9780470166932.ch10. (10) Schroeder, W. A.; Malmberg, E. W.; Fong, L. L.; Trueblood, K. N.; Landerl, J. D.; Hoerger, E. Chromatographic Investigations of Smokeless Powder. Ind. Eng. Chem. 1949, 41 (12), 2818–2827. https://doi.org/10.1021/ie50480a037. (11) Hough, A. Manufacture of Smokeless Powder Propellants. US1916763. (12) Eugeniusz Bazylewicz-Kniazykowski; Kazimierz Partyka. Method of Producing Smokeless Gunpowder. US1889462. (13) Reardon, M. R.; MacCrehan, W. A.; Rowe, W. F. Comparing the Additive Composition of Smokeless Gunpowder and Its Handgun-Fired Residues. J. Forensic Sci. 2000, 45 (6), 14871J. https://doi.org/10.1520/JFS14871J. 21 (14) Goudsmits, E.; Sharples, G. P.; Birkett, J. W. Preliminary Classification of Characteristic Organic Gunshot Residue Compounds. Science & Justice 2016, 56 (6), 421–425. https://doi.org/10.1016/j.scijus.2016.06.007. (15) Perez, J. J.; Flanigan, P. M.; Brady, J. J.; Levis, R. J. Classification of Smokeless Powders Using Laser Electrospray Mass Spectrometry and Offline Multivariate Statistical Analysis. Anal. Chem. 2013, 85 (1), 296–302. https://doi.org/10.1021/ac302661k. (16) Chang, K. H.; Jayaprakash, P. T.; Yew, C. H.; Abdullah, A. F. L. Gunshot Residue Analysis and Its Evidential Values: A Review. Australian Journal of Forensic Sciences 2013, 45 (1), 3–23. https://doi.org/10.1080/00450618.2012.691546. (17) Hogg, S. R.; Hunter, B. C.; Waddell Smith, R. Elemental Characterization and Discrimination of Nontoxic Ammunition Using Scanning Electron Microscopy with Energy Dispersive X-Ray Analysis and Principal Components Analysis. J Forensic Sci 2016, 61 (1), 35–42. https://doi.org/10.1111/1556-4029.12881. (18) Yañez, J.; Paz Farías, M.; Zúñiga, V.; Soto, C.; Contreras, D.; Pereira, E.; Mansilla, H. D.; Saavedra, R.; Castillo, R.; Sáez, P. Differentiation of Two Main Ammunition Brands in Chile by Regularized Discriminant Analysis (RDA) of Metals in Gunshot Residues. Microchemical Journal 2012, 101, 43–48. https://doi.org/10.1016/j.microc.2011.10.003. (19) Taudte, R. V.; Roux, C.; Bishop, D.; Blanes, L.; Doble, P.; Beavis, A. Development of a UHPLC Method for the Detection of Organic Gunshot Residues Using Artificial Neural Networks. Anal. Methods 2015, 7 (18), 7447–7454. https://doi.org/10.1039/C5AY00306G. (20) Gassner, A.-L.; Weyermann, C. LC–MS Method Development and Comparison of Sampling Materials for the Analysis of Organic Gunshot Residues. Forensic Science International 2016, 264, 47–55. https://doi.org/10.1016/j.forsciint.2016.03.022. (21) Zhao, X.; Yinon, J. Identification of Nitrate Ester Explosives by Liquid Chromatography– Electrospray Ionization and Atmospheric Pressure Chemical Ionization Mass Spectrometry. Journal of Chromatography A 2002, 977 (1), 59–68. https://doi.org/10.1016/S0021- 9673(02)01349-3. (22) Xu, X.; van de Craats, A. M.; de Bruyn, P. C. A. M. Highly Sensitive Screening Method for Nitroaromatic, Nitramine and Nitrate Ester Explosives by High Performance Liquid Chromatography---Atmospheric Pressure Ionization---Mass Spectrometry (HPLC-API-MS) in Forensic Applications. J. Forensic Sci. 2004, 49 (6), 1–10. https://doi.org/10.1520/JFS2003279. (23) Reese, K. L.; Jones, A. D.; Waddell Smith, R. Characterization of Smokeless Powders Using Multiplexed Collision-Induced Dissociation Mass Spectrometry and Chemometric Procedures. Forensic Science International 2017, 272, 16–27. (24) Benito, S.; Abrego, Z.; Sánchez, A.; Unceta, N.; Goicolea, M. A.; Barrio, R. J. Characterization of Organic Gunshot Residues in Lead-Free Ammunition Using a New Sample Collection Device for Liquid Chromatography–Quadrupole Time-of-Flight Mass 22 Spectrometry. Forensic Science International 2015, 246, 79–85. https://doi.org/10.1016/j.forsciint.2014.11.002. (25) DeTata, D.; Collins, P.; McKinley, A. A Fast Liquid Chromatography Quadrupole Time-of- Flight Mass Spectrometry (LC-QToF-MS) Method for the Identification of Organic Explosives and Propellants. Forensic Science International 2013, 233 (1–3), 63–74. https://doi.org/10.1016/j.forsciint.2013.08.007. (26) Wissinger, C. E.; McCord, B. R. A Gradient Reversed Phase HPLC Procedure for Smokeless Powder Comparison. J. Forensic Sci. 2002, 47 (1), 15219J. https://doi.org/10.1520/JFS15219J. (27) Andrasko, J. Characterization of Smokeless Powder Flakes from Fired Cartridge Cases and from Discharge Patterns on Clothing. J. Forensic Sci. 1992, 37 (4), 13289J. https://doi.org/10.1520/JFS13289J. (28) Mathis, J. A.; McCord, B. R. Mobile Phase Influence on Electrospray Ionization for the Analysis of Smokeless Powders by Gradient Reversed Phase High-Performance Liquid Chromatography-ESIMS. Forensic Science International 2005, 154 (2–3), 159–166. https://doi.org/10.1016/j.forsciint.2004.10.008. (29) Mathis, J. A.; McCord, B. R. Gradient Reversed-Phase Liquid Chromatographic- Electrospray Ionization Mass Spectrometric Method for the Comparison of Smokeless Powders. Journal of Chromatography A 2003, 988 (1), 107–116. https://doi.org/10.1016/S0021-9673(02)02055-1. (30) Reese, K. L. Association and Differentiation of Smokeless Powders Utilizing Non-Targeted Mass Spectrometry and Multivariate Statistical Analysis, Michigan State University, 2016. (31) Laza, D.; Nys, B.; Kinder, J. D.; Kirsch-De Mesmaeker, A.; Moucheron, C. Development of a Quantitative LC-MS/MS Method for the Analysis of Common Propellant Powder Stabilizers in Gunshot Residue. J Forensic Sci 2007, 52 (4), 842–850. https://doi.org/10.1111/j.1556-4029.2007.00490.x. (32) Thomas, J. L.; Lincoln, D.; McCord, B. R. Separation and Detection of Smokeless Powder Additives by Ultra Performance Liquid Chromatography with Tandem Mass Spectrometry (UPLC/MS/MS) ,. J Forensic Sci 2013, 58 (3), 609–615. https://doi.org/10.1111/1556- 4029.12096. (33) Gallidabino, M. D.; Barron, L. P.; Weyermann, C.; Romolo, F. S. Quantitative Profile– Profile Relationship (QPPR) Modelling: A Novel Machine Learning Approach to Predict and Associate Chemical Characteristics of Unspent Ammunition from Gunshot Residue (GSR). Analyst 2019, 144 (4), 1128–1139. https://doi.org/10.1039/C8AN01841C. (34) Joshi, M.; Rigsby, K.; Almirall, J. R. Analysis of the Headspace Composition of Smokeless Powders Using GC–MS, GC-ΜECD and Ion Mobility Spectrometry. Forensic Science International 2011, 208 (1–3), 29–36. https://doi.org/10.1016/j.forsciint.2010.10.024. 23 (35) Yinon, J. Trace Analysis of Explosives in Water by Gas Chromatography—Mass Spectrometry with a Temperature-Programmed Injector. Journal of Chromatography A 1996, 742 (1–2), 205–209. https://doi.org/10.1016/0021-9673(96)00261-0. (36) Munch, J. W. Method 529: Determination of Explosives an Drelated Compounds in Drinking Water by Solid Phase Extraction and Capillary Gas Chromatography/Mass Spectrometry. Environmental Protection Agency September 2002. (37) Dalby, O.; Butler, D.; Birkett, J. W. Analysis of Gunshot Residue and Associated Materials-A Review. Journal of Forensic Sciences 2010, 55 (4), 924–943. https://doi.org/10.1111/j.1556-4029.2010.01370.x. (38) Goudsmits, E.; Blakey, L. S.; Chana, K.; Sharples, G. P.; Birkett, J. W. The Analysis of Organic and Inorganic Gunshot Residue from a Single Sample. Forensic Science International 2019, 299, 168–173. https://doi.org/10.1016/j.forsciint.2019.03.049. (39) Williamson, R.; Gura, S.; Tarifa, A.; Almirall, J. R. The Coupling of Capillary Microextraction of Volatiles (CMV) Dynamic Air Sampling Device with DART-MS Analysis for the Detection of Gunshot Residues. Forensic Chemistry 2018, 8, 49–56. https://doi.org/10.1016/j.forc.2018.01.005. (40) Muller, D.; Levy, A.; Vinokurov, A.; Ravreby, M.; Shelef, R.; Wolf, E.; Eldar, B.; Glattstein, B. A Novel Method for the Analysis of Discharged Smokeless Powder Residues. J Forensic Sci 2007, 52 (1), 75–78. https://doi.org/10.1111/j.1556-4029.2006.00309.x. (41) Perr, J.; Furton, K.; Almirall, J. Gas Chromatography Positive Chemical Ionization and Tandem Mass Spectrometry for the Analysis of Organic High Explosives. Talanta 2005, 67 (2), 430–436. https://doi.org/10.1016/j.talanta.2005.01.035. (42) Gallidabino, M.; Romolo, F. S.; Bylenga, K.; Weyermann, C. Development of a Novel Headspace Sorptive Extraction Method To Study the Aging of Volatile Compounds in Spent Handgun Cartridges. Anal. Chem. 2014, 86 (9), 4471–4478. https://doi.org/10.1021/ac5004568. (43) Tarifa, A.; Almirall, J. R. Fast Detection and Characterization of Organic and Inorganic Gunshot Residues on the Hands of Suspects by CMV-GC–MS and LIBS. Science & Justice 2015, 55 (3), 168–175. https://doi.org/10.1016/j.scijus.2015.02.003. (44) Sauzier, G.; Bors, D.; Ash, J.; Goodpaster, J. V.; Lewis, S. W. Optimisation of Recovery Protocols for Double-Base Smokeless Powder Residues Analysed by Total Vaporisation (TV) SPME/GC-MS. Talanta 2016, 158, 368–374. https://doi.org/10.1016/j.talanta.2016.04.048. (45) Stevens, B.; Bell, S.; Adams, K. Initial Evaluation of Inlet Thermal Desorption GC–MS Analysis for Organic Gunshot Residue Collected from the Hands of Known Shooters. Forensic Chemistry 2016, 2, 55–62. https://doi.org/10.1016/j.forc.2016.10.001. 24 (46) Dennis, D.-M. K.; Williams, M. R.; Sigman, M. E. Assessing the Evidentiary Value of Smokeless Powder Comparisons. Forensic Science International 2016, 259, 179–187. https://doi.org/10.1016/j.forsciint.2015.12.034. (47) Dennis, D.-M. K.; Williams, M. R.; Sigman, M. E. Investigative Probabilistic Inferences of Smokeless Powder Manufacturers Utilizing a Bayesian Network. Forensic Chemistry 2017, 3, 41–51. https://doi.org/10.1016/j.forc.2016.12.001. (48) Weyermann, C.; Belaud, V.; Riva, F.; Romolo, F. S. Analysis of Organic Volatile Residues in 9mm Spent Cartridges. Forensic Science International 2009, 186 (1–3), 29–35. https://doi.org/10.1016/j.forsciint.2009.01.005. (49) National Center for Forensic Science, Smokeless Powders Database. http://www.ilrc.ucf.edu/powders/ (accessed Aug 22, 2019). (50) McIlroy, J. W.; Smith, R. W.; McGuffin, V. L. Assessing the Effect of Data Pretreatment Procedures for Principal Components Analysis of Chromatographic Data. Forensic Science International 2015, 257, 1–12. https://doi.org/10.1016/j.forsciint.2015.07.038. (51) van den Berg, R. A.; Hoefsloot, H. C.; Westerhuis, J. A.; Smilde, A. K.; van der Werf, M. J. Centering, Scaling, and Transformations: Improving the Biological Information Content of Metabolomics Data. BMC Genomics 2006, 7 (1), 142. https://doi.org/10.1186/1471-2164-7- 142. (52) Waddell Smith, R. Chemometrics. In Forensic Chemistry: Fundamentals and Applications; Siegel, J., Ed.; John Wiley & Sons, Ltd: West Sussex, UK, 2016; pp 469–503. (53) Lennert, E.; Bridge, C. Analysis and Classification of Smokeless Powders by GC–MS and DART-TOFMS. Forensic Science International 2018, 292, 11–22. https://doi.org/10.1016/j.forsciint.2018.09.003. (54) Perez, J. J.; Watson, D. A.; Levis, R. J. Classification of Gunshot Residue Using Laser Electrospray Mass Spectrometry and Offline Multivariate Statistical Analysis. Anal. Chem. 2016, 88 (23), 11390–11398. https://doi.org/10.1021/acs.analchem.6b01438. (55) Lennert, E.; Bridge, C. M. Rapid Screening for Smokeless Powders Using DART-HRMS and Thermal Desorption DART-HRMS. Forensic Chemistry 2019, 13, 100148. https://doi.org/10.1016/j.forc.2019.100148. 25 2. Materials and Methods 2.1 Reference Materials and Solvents Ethyl centralite (1,3-diethyl-1,3-diphenyl-urea, 99%), methyl centralite (1,3-dimethyl-1,3-diphenyl-urea, 99%), 2,4-dinitrodiphenylamine (98%), 2,4-dinitrotoluene (97%), and 2-naphthol (99%) were purchased from Sigma Aldrich (St. Louis, MO, USA). 2,5-Dimethylnitrobenzene was purchased from Tokyo Chemical Industry Co. (Singapore, >99%). Diphenylamine (sulfate salt) was purchased from Eastman Organic Chemicals (Rochester, NY, USA). Dibutyl phthalate and n-Heptadecane were purchased from Alfa Aesar (Heysham, LA, USA, >99%). Dichloromethane (CH2Cl2, HPLC grade) was purchased from VWR Chemicals (Radnor, Pennsylvania) and acetone ((CH3)2CO, CHROMASOLV grade) was purchased from Sigma Aldrich (St. Louis, MO, USA). 2.1.1 Preparation of Reference Standards A reference standard mixture was prepared by weighing approximately 10 mg of each 2,4-dinitrodiphenylamine, ethyl centralite, diphenylamine, methyl centralite, 2-naphthol, and 2,4-dinitrotoluene into a clean 13x100 mm borosilicate glass vial, followed by 10 µL each dibutyl phthalate and 2,5-dimethylnitrobenzene. A 900 µL aliquot of (CH3)2CO was added to the vial, which was capped and briefly vortexed. The standard mixture was soaked in (CH3)2CO for 90 min, and the vial was vortexed for 15 s every 15 min. After 90 min, the (CH3)2CO extract was dried completely under nitrogen gas, leaving a nitrocellulose film. Following (CH3)2CO extraction, CH2Cl2 was added in three aliquots to the vial containing the nitrocellulose film. After each 500 µL aliquot of CH2Cl2 was added, the vial was vortexed for 10 s and then allowed to sit for 10 min. The supernatant was then transferred to a 26 clean 13x100 mm borosilicate glass vial. This process was repeated two more times, giving a total volume of 1.5 mL CH2Cl2 extract collected in a clean vial.1 The extracts were then dried completely under nitrogen gas and stored at 4 °C. Prior to analysis, extracts for GC-MS analysis were reconstituted with 1.5 mL CH2Cl2 containing 15 µM heptadecane as an internal standard. Extraction reproducibility tests were performed using available reference standards. However, the impact of nitrocellulose and nitroglycerin was unable to be investigated due to unavailability of reference standards. Seven standard mixes were subjected to the same extraction procedures as the samples. The standard mix contained 2,4-dinitrotoluene, ethyl centralite, methyl centralite and dibutyl phthalate. The maximum abundance of each compound was compared using relative standard deviations (RSDs, Equation 2.1), where s is the standard deviation and !̅ is the average abundance of a specific compound. 456=100/)*̅2 Equation 2.1 Instrument replicates (multiple injections of the same sample preparation) had RSDs of less than 3%. With the exception of ethyl centralite, the RSDs among the different preparations were less than 10%. Ethyl centralite was the most variable, with an RSD of approximately 19%. These experiments indicated that additional variance observed in the samples is likely due to the nitrocellulose present in the unburned powders. 27 2.2 Unburned Smokeless Powder Ammunition of various calibers and manufacturers was purchased by previous graduate students.2,3 The unburned powder was previously removed from five individual cartridges of each ammunition type using an inertia-based bullet puller. Five cartridges of each ammunition type were analyzed. For further information, refer to Reese et al.1 2.2.1 Smokeless Powder Sample Set The smokeless powder sample set consisted of powders that were purchased within the past 10 years, as well as powders that were of unknown age but were at least 15 years old. Purchased ammunition was denoted ‘new’, as the origin, place and date of purchase were known to be within the past 10 years. Ammunition of unknown age was denoted ‘aged’ and was obtained from Dr. Brian Hunter. Some aged ammunition was received in the original packaging whereas other cartridges were not contained in a single storage box. It was assumed that aged cartridges found in the original packaging were produced at the same time, whereas the ‘loose’ aged cartridges were not assumed to have originated from the same box of ammunition. The manufacturer, caliber, primer type, and approximate age for each of the ammunition types are summarized in Table 2.1. For complete manufacturing information, refer to Reese (2015).2 28 Table 2.1 Commercial Smokeless Powder Samples Selected For Analysis Abbreviation Manufacturer AEFed-L Horn-L PMC-L 44-N Rem-L† SB-N Win-L† AA-N BZR-nL MG-nL† Rem-nL SB-nL Win-nL† MG-O 44-O AA-O 22-O PMC-O Federal Hornady PMC PMC Remington Sellier and Bellot Winchester Winchester AA Blazer Magtech Remington Sellier and Bellot Winchester Magtech PMC Winchester AA CCI PMC Caliber 9 mm 9 mm 9 mm 0.44 9 mm 7.62x39 mm 9 mm 12-Gauge 9 mm 9 mm 9 mm 9 mm 9 mm 7.62x39 mm 0.44 12-Gauge 0.22 LR 9 mm Primer Type Lead Lead Lead Lead Lead Lead Lead Lead No Lead No Lead No Lead No Lead No Lead Lead Lead Lead Lead Lead Approx. Age <10 years <10 years <10 years <10 years <10 years <10 years <10 years <10 years <10 years <10 years <10 years <10 years <10 years > 15 years > 15 years > 15 years* > 15 years* > 15 years* * Cartridges were not obtained in the original packaging. † Fired cartridge residue was also collected 2.2.2 Extraction of Unburned Smokeless Powder Approximately 10 mg of unburned smokeless powder kernels were weighed into a 13x100 mm borosilicate glass vial. A 900 µL aliquot of (CH3)2CO was added to the vial, which was capped and briefly vortexed. The unburned powder kernels were soaked in (CH3)2CO for 90 minutes, and the vial was vortexed for 15 seconds every 15 minutes. After 90 minutes, the (CH3)2CO extract was dried completely under nitrogen gas, leaving a nitrocellulose film. 29 Following (CH3)2CO extraction, CH2Cl2 was added in three aliquots to the vial containing the nitrocellulose film. After each 500 µL aliquot of CH2Cl2 was added, the vial was vortexed for 10 s and then allowed to sit for 10 min. The supernatant was then transferred to a clean 13x100 mm borosilicate glass vial. This process was repeated two more times, giving a total volume of 1.5 mL CH2Cl2 extract collected in a clean vial.1 The CH2Cl2 extract was then divided into two clean GC vials, with 400 µL CH2Cl2 extract in each vial. One vial was used for initial GC-MS analysis, and the other vial was reserved for future analysis. The remaining 700 µL was stored in a separate capped vial. The extracts were then dried completely under nitrogen gas and stored at 4 °C. Prior to analysis, extracts for GC-MS analysis were reconstituted with 1.5 mL CH2Cl2 containing 15 µM heptadecane (C17) as an internal standard. 2.3 Fired Cartridge Residue 2.3.1 Collection of Fired Cartridges A subset of four different types of ammunition (Win-L, Win-nL, Rem-L, and MG-nL) was fired using a XD Mod.2 sub-compact 9mm handgun from Springfield Armory. The cartridges chosen for fired cartridge residue analysis were from the same box as the corresponding unburned powder previously analyzed. For each type of ammunition, 10 cartridges were fired. The first five cartridges were fired in succession without cleaning the barrel. For the second set of five cartridges, the barrel was cleaned between each shot. Cleaning the barrel consisted of inserting a bore brush with solvent into the muzzle to remove particulate matter, followed by swabbing the interior of the barrel with a cotton patch until the patch came out clean. The weapon was not disassembled for the barrel cleaning between 30 rounds of the same ammunition. However, the firearm was field stripped between each different type of ammunition fired; this involves partially disassembling the firearm for more extensive cleaning. All shots were fired through a hole cut in a cotton pillowcase, designed to catch the expelled fired cartridge in a controlled environment rather than allowing the cartridge to fall to the ground. The fired cartridge was transferred into a labeled scintillation vial using forceps and the vial was capped immediately. 2.3.2 Extraction of Fired Cartridge Residue A 700 µL aliquot of (CH3)2CO was added directly to the fired cartridge to dissolve any remaining nitrocellulose. This aliquot was then immediately transferred to a clean 13x100 mm borosilicate glass vial, and a second 700 µL aliquot was added to the cartridge. After 10 min, the second aliquot was transferred to the vial and the (CH3)2CO extract was dried under nitrogen. A 500 µL aliquot of CH2Cl2 was added to the film remaining in the vial, vortexed for 10 s, and soaked for 10 min. The CH2Cl2 extract was then transferred to a second clean 13x100 mm borosilicate glass vial. This process was repeated two more times, resulting in 1.5 mL CH2Cl2 extract collected in the second vial. The CH2Cl2 extract was dried under nitrogen and reconstituted with 200 µL of CH2Cl2 containing 15 µM heptadecane as an internal standard. The final solution was transferred to a GC vial with a glass insert to account for the lower volume. 31 2.4 Gas Chromatography-Mass Spectrometry (GC-MS) Analysis The reference mixture, unburned powder, and fired cartridge residue extracts were analyzed on the same instrument under equivalent parameters. All extracts were analyzed in replicate on an Agilent 7890 gas chromatograph coupled to an Agilent 5975 mass spectrometer with an Agilent 7693 autosampler (Agilent Technologies, Santa Clara, CA). The temperature program was based on that used by the FBI, as indicated on the online Smokeless Powders Database maintained by the National Center for Forensic Science (NCFS).4 A capillary column containing a 5%-diphenyl-95%-dimethyl polysiloxane stationary phase (DB-5MS, 30 m x 0.25 mm x 0.25 μm, Agilent Technologies, Santa Clara, CA) was used for all analyses. Ultra-high purity helium with a nominal flow rate of 1.2 mL/min was used as a carrier gas. The injection temperature was 170 °C with a splitless injection of 1 µL. The initial oven temperature was 45 °C. Following a three-minute hold at 45 °C, the temperature increased at a rate of 15 °C/min to 150 °C, followed by a 30 °C/min increase to a final oven temperature of 265 °C and a final hold of 2.70 min. The transfer line was maintained at 280 °C. Electron ionization at 70 eV was used, and the scan range was set to include m/z 35-400 with a scan rate of 3.89 scans/s. The temperature of the ion source was 230 °C, and the temperature of the quadrupole mass analyzer was 150 °C. A three-minute solvent delay was employed to extend the life of the filament. 32 2.5 Data Processing The chromatogram of each sample extract was examined using ChemStation (Version E.02.01.1177, Agilent Technologies). Retention times and mass spectra were compared to reference standards collected under equivalent conditions for identification. When a suitable reference standard was not available, the NIST/EPA/NIH Mass Spectral Library was used for provisional identification. In total, nine compounds were identified. Five compounds were identified following comparison of retention time and mass spectra to reference materials: 2,4-dinitrotoluene, diphenylamine, ethyl centralite, methyl centralite, and dibutyl phthalate. An additional four compounds commonly associated with smokeless powder composition (nitroglycerin, 2-nitrodiphenylamine, 4-nitrodiphenylamine, 1-methyl-3,3-diphenylurea) were provisionally identified based on comparison to the library search results. The chromatographic data were exported from ChemStation, imported into Microsoft Excel (Version 16.28, Microsoft Corporation, Redmond, WA), and normalized to the maximum abundance of the C17 internal standard. These data were imported into Origin (version 9.0 OriginLab Corporation, Northampton, MA), where a normalized chromatogram was generated for each extract. Each resultant each data file contained an abundance value at every time point in the 16.5 min chromatographic run, resulting in over 3000 data points for a single extract. However, a large majority of the 3000 data points are not chemically informative, as they describe the abundance of the baseline or other chemically uninformative noise such as column bleed. For multivariate statistical analysis, the chromatographic data was reduced to include only the maximum normalized abundance of each of the nine compounds identified previously. Each compound was thus considered a separate variable. 33 2.6 Multivariate Statistical Analysis Prior to multivariate statistical analysis, the unburned powder data set was Pareto scaled to account for the dominance of ethyl centralite and dibutyl phthalate observed in the analyzed powders. For each compound, the average normalized abundance and standard deviation was calculated across all unburned powder extracts. Pareto scaling involves mean-centering the data using the average normalized abundance and then dividing each value by the square root of the standard deviation (Equation 2.2), !8=*,*̅√) where !8 represents the final Pareto scaled value and ! represents the normalized abundance to be scaled, while !̅ and 9 represent the average normalized abundance and standard deviation of a Equation 2.2 given compound, respectively.5 The scaled data were imported into R (version 3.6.1)6, an open source coding platform used for statistical computing, for principal components analysis (PCA) and into Pirouette (v. 4.0) for hierarchical cluster analysis (HCA). The fired cartridge residue data set was Pareto scaled in the same manner. All multivariate statistical analyses were performed using normalized, Pareto scaled data sets. 2.6.1 Principal Components Analysis (PCA) The unburned powder data were subjected to PCA, using the R script given in Appendix A2.1. The resultant loadings and scores were imported into Microsoft Excel (Version 16.28, Microsoft Corporation, Redmond, WA) for visualization and interpretation. The fired cartridge residue data set was subjected to PCA in the same manner. Principal components analysis was also performed on a subset of unburned powders, corresponding to the four powders selected for 34 fired cartridge residue analysis. To investigate association of fired cartridge residues to the corresponding unburned powders, a single fired cartridge residue was included with the unburned powder subset and PCA was performed. This process was repeated for each of the fired cartridge residues to allow for the association of each fired cartridge to be examined individually. 2.6.2 Hierarchical Cluster Analysis (HCA) The unburned powder data set was subjected to agglomerative hierarchical cluster analysis (HCA) using Pirouette (v. 4.0). After HCA of the unburned data set was completed, a single fired cartridge was added to the unburned powder subset and HCA was repeated. This process was repeated, separately introducing each fired cartridge residue to the unburned powder subset and observing where each fired residue clustered. The unburned powder data set was also subjected to HCA in R, using the Euclidean distance method and complete-linkage clustering. The R script used to perform HCA in R is given in Appendix A2.2. The dendrograms produced using the R script were consistent to the Pirouette comparisons. 35 APPENDIX 36 APPENDIX A2.1 R Script for PCA #Each column is a separate variable (Variable names along top row. No special characters or sp aces.) Each row is a separate sample (Sample names down first column. No special characters o r spaces.) data=read.table("ubpdiscrete.txt",header=TRUE) pca<-prcomp(data,scale=FALSE) options(max.print=1000000000) #stops R from truncating data # loadings<-print(pca) #loadings variance<-summary(pca) #variance scores<-pca$x #scores # write.table(scores,"/Users/Becca/R Working Directory/scores.txt",sep="/") A2.2 R Script for HCA #Each column is a separate variable (Variable names along top row. No special characters or sp aces.) Each row is a separate sample (Sample names down first column. No special characters o r spaces.) # data=read.table("ubpdiscrete.txt",header=TRUE) d<-dist(data,method="euclidean",diag=FALSE,upper=FALSE,p=2) #distance matrix hc<-hclust(d) #performes complete linkage HCA plot(hc,hang=-1,labels=FALSE) #remove labels=false if want labels on plot rect.hclust(hc,k=7,border=2:5) #k defines number of clusters, border defines colors #exporting distmatrix<-as.matrix(d) #to get distance matrix in an exportable format write.table(distmatrix,"/Users/Becca/R Working Directory/distmatrix.txt",sep="/") #saves to txt file # height<-as.matrix(hc$height) #distance between clusters (?) merge<-as.matrix(hc$merge) # ?? order<-as.matrix(hc$order) #Left to Right order of samples along dendrogram # sub_grp<-cutree(hc,k=7) #define a cutoff to cut tree into a defined number of groups table(sub_grp) #output tells how many samples in each subgroup/cluster 37 REFERENCES 38 REFERENCES (1) Reese, K. L.; Jones, A. D.; Smith, R. W. Characterization of Smokeless Powders Using Multiplexed Collision-Induced Dissociation Mass Spectrometry and Chemometric Procedures. Forensic Sci. Int. 2017, 272, 16–27. (2) Reese, K. L. Association and Differentiation of Smokeless Powders Utilizing Non-Targeted Mass Spectrometry and Multivariate Statistical Analysis, Michigan State University, 2016. (3) Hogg, S. R. Analysis of Lead-Free Ammunition by Scanning Electron Microscopy Using Energy Dispersive X-Ray Spectroscopy and Discrimination of Samples Using Principal Components Analysis, Michigan State University, 2016. (4) National Center for Forensic Science, University of Central Florida. Smokeless Powder Reference Collection http://www.ilrc.ucf.edu/powders/ (accessed Aug 22, 2019). (5) van den Berg, R. A.; Hoefsloot, H. C.; Westerhuis, J. A.; Smilde, A. K.; van der Werf, M. J. Centering, Scaling, and Transformations: Improving the Biological Information Content of Metabolomics Data. BMC Genomics 2006, 7 (1), 142. https://doi.org/10.1186/1471-2164-7- 142. (6) R: The R Project for Statistical Computing https://www.r-project.org/ (accessed Aug 22, 2019). 39 3. Association and Differentiation of Unburned Powders using Multivariate Statistical Analysis of Organic Compounds 3.1 Introduction Unburned smokeless powder extracts were analyzed in replicate by gas chromatography-mass spectrometry (GC-MS) to generate organic compound profiles. Following categories set forth by Goudsmits et al., two data sets were created from the nine identified compounds. The first data set contained the abundances of Category 1 compounds (nitroglycerin, methyl centralite and ethyl centralite), while the second (Category 1 and 2 compounds) contained the abundances of all nine compounds identified. To investigate similarities and differences among the unburned powders, the two data sets were separately subjected to principal components analysis (PCA) and hierarchical cluster analysis (HCA). From PCA of powders based on Category 1 compounds, the unburned powders were separated into two groups, primarily based on differences in abundance of ethyl centralite. From PCA of powders based on Category 1 and 2 compounds, the unburned powders were separated into four groups, primarily based on differences in abundances of ethyl centralite, dibutyl phthalate, 2,4-dinitrotoluene, and diphenylamine. Hierarchical cluster analysis (HCA) of the unburned powders resulted in groupings consistent with the those identified by PCA. Inclusion of the additional compounds in Category 2 increased differentiation of the unburned powders over Category 1, primarily due to the addition of dibutyl phthalate, 2,4-dinitrotoluene and diphenylamine. Overall, successful association and differentiation of unburned powders was achieved based on differences in organic compound profiles and abundances of major compounds. 40 3.2 Organic Composition of Unburned Smokeless Powders by GC-MS All unburned powder extracts were analyzed in replicate by GC-MS as described in Chapter 2. The organic composition of each unburned powder extract as determined by GC-MS can be found in Table A3.1. A total of nine compounds were identified: nitroglycerin, dibutyl phthalate (DBPH), ethyl centralite (EC), diphenylamine (DPA), 2-nitrodiphenylamine (2nDPA), 4-nitrodiphenylamine (4nDPA), Akardite II (AKII), methyl centralite (MC), and 2,4-dinitrotoluene (2,4-DNT). Heptadecane (C17) was added to each extract as an internal standard (I.S.). Approximate retention times for all compounds can be found in Table A3.2 with the corresponding mass spectra in Figures A3.1-A3.9. Representative chromatograms for each powder can be found in Figures A3.10-A3.19. Nitroglycerin is a common explosive component in double-based powders and also acts as a plasticizer. With the exception of the five cartridges of MG-O unburned powder, nitroglycerin was present in every powder analyzed (Table A3.1). The maximum normalized abundance of nitroglycerin was detected in AEFED-L, at an abundance that was approximately 32% of the maximum abundance of dibutyl phthalate, which was the most abundant compound detected in any of the 18 powders. For every double-based powder, two peaks provisionally identified as nitroglycerin were present (Figure 3.1). The first peak had an approximate retention time of 10.6 min, while the second peak had an approximate retention time of 11.0 min. The mass spectra of both peaks were dominated by only two ions - m/z 46 and m/z 76 (Figure 3.2). 41 UBP_MG-nL_Cartridge1 EC NG I.S. 10 5 0 e c n a d n u b A d e z i l a m r o N 4 8 12 time (min) 16 Figure 3.1 Representative chromatogram of unburned MG-nL, showing two peaks provisionally identified as nitroglycerin 42 Nitroglycerin (1) a) 1.0 46 y t i s n e t n I . l e R 0.8 0.6 0.4 0.2 0.0 76 50 100 150 200 250 300 Nitroglycerin (2) m/z b) 1.0 46 y t i s n e t n I . l e R 0.8 0.6 0.4 0.2 0.0 76 50 100 150 200 250 300 m/z Figure 3.2 Mass spectra of the two peaks provisionally identified as nitroglycerin, at retention times (a) 10.6 min and (b) 11.0 min 43 The characteristic ions of nitroglycerin are limited to m/z 46 and 76, which are shared by other nitrated glycerol derivatives such as ethylene glycol dinitrate (EGDN) and propylene glycol dinitrate (PGDN) (Figure 3.3). Reference materials for these explosive components were not available, and thus the peaks could not be definitively identified. Rather, the peaks were provisionally identified as nitroglycerin due to the prevalence of nitroglycerin in smokeless powders, while EGDN and PGDN are not commonly used in commercial ammunition. Subsequent analysis of a representative subset of the unburned powder extracts on two additional GC-MS instruments showed only one peak attributed to nitroglycerin, at a retention time of approximately 10.9 minutes. As a result, the additional detected peak was determined to be an artifact of the instrument rather than the unburned powder samples. Based on a comparison of retention times between the different instruments and columns, the peak with a retention time of 11.0 min was provisionally identified as nitroglycerin, and the earlier eluting peak was deemed not chemically relevant to the samples. Over time, the gold-plated seal at the base of the inlet liner collects sample residue and other debris. Contamination of the gold-plated inlet seal of the GC disrupts the inert coating and can result in the formation of active sites, which can lead to adsorption of analytes.1 While it is possible that adsorption resulted in extreme peak splitting, the two peaks are fairly well-resolved, indicating there might be a different reason. Another possible explanation is that the first peak at tR 10.6 min corresponds to EGDN. Further studies to determine the identity of this first peak were not possible, as only one peak was detected when a subset of samples was analyzed on two additional instruments, as previously discussed. 44 a) b) 46 46 76 76 Figure 3.3 Literature mass spectra of (a) ethylene glycol dinitrate and (b) propylene glycol dinitrate2 45 Additionally, nitroglycerin is susceptible to rapid decomposition when exposed to elevated temperatures, which could result in the loss of a nitrate group, producing EGDN. To investigate thermal degradation in the inlet, a subset of unburned powder extracts was also analyzed on an Agilent 7010 GC/triple quadrupole MS/MS (GC-TQMS), equipped with a Programmed Temperature Vaporizer (PTV) inlet. The extracts were first analyzed with an isothermal inlet temperature of 170°C, followed by a variable temperature inlet program. The initial inlet temperature was 100°C, with a 40°C/min ramp to 170°C. A comparison of the analyses using the isothermal and variable temperature inlet programs demonstrated that starting at a lower inlet temperature did not result in the detection of any additional compounds. The GC-TQMS analysis also only resulted in one peak attributed to nitroglycerin, further supporting the conclusion that the additional nitroglycerin peak present in the initial analyses was an artifact of the primary GC-MS instrument, not of the unburned powder extracts. In addition to nitroglycerin, the unburned powders contained various additives that serve as plasticizers, burn deterrents, and stabilizers. Dibutyl phthalate is added to smokeless powder formulations to serve as a plasticizer and was present in every powder analyzed. Extracts of 44-N and 44-O contained the highest abundances of dibutyl phthalate of all the powders analyzed. Dibutyl phthalate was present at the highest abundance of all compounds detected, resulting in a maximum normalized abundance of 100%. Akardite II (1-methyl-3,3-diphenylurea) is added as to serve as a stabilizer and was present in nine of the powders analyzed. Unburned AEFED-L contained the highest abundance of Akardite II, with a maximum normalized abundance that was 28% of the maximum abundance of dibutyl phthalate. Methyl centralite can act both as a stabilizer and a burn deterrent and was only detected in Win-nL and Rem-nL, at a maximum normalized abundance of less than 0.3% of the maximum abundance of dibutyl phthalate. 46 Ethyl centralite is also added as a stabilizer and burn deterrent and was present in 13 of the 18 powders analyzed. Ethyl centralite was the second most abundant compound detected, with a maximum normalized abundance that was 91% of the maximum abundance of dibutyl phthalate. Ethyl centralite is known to be a more effective stabilizer for powders that contain NG.3 The powders that contained the highest abundance of ethyl centralite (BZR-nL, MG-nL, SB-N, and two cartridges of AA-O) all contained nitroglycerin. Conversely, MG-O, which did not contain nitroglycerin, contained little to no ethyl centralite but instead had the highest abundance of diphenylamine of all the unburned powders, with a maximum normalized abundance that was 55% of the maximum abundance of dibutyl phthalate. Diphenylamine is also used as a stabilizer in smokeless powders and acts as a nitric oxide scavenger. Over time, nitrocellulose and nitroglycerin decompose to produce nitric oxide (NO), which upon contact with air and moisture oxidizes to produce NO2.4 Diphenylamine binds free molecules of NO2, which helps to prevent auto-catalytic decomposition in the powders. When diphenylamine binds a molecule of NO2, N-nitrosodiphenylamine is formed. This nitrated product can then be oxidized and undergo transfer reactions, producing 2- and 4-nitrodiphenylamine (Figure 3.4).5 47 H N NO N Diphenylamine N-nitrosodiphenylamine H N O2N 4-nitrodiphenylamine NO2 H N 2-nitrodiphenylamine Figure 3.4 Nitrosation of diphenylamine With the exception of BZR-nL and MG-nL, diphenylamine and 2-nitrodiphenylamine were present in all powders, regardless of age. Conversely, 4-nitrodiphenylamine was only detected in six of the unburned powders and was not always detected in all five cartridges of a powder. As 2- and 4-nitrodiphenylamine are produced as powders age, older powders were expected to contain higher abundances of these nitrated diphenylamine derivatives. However, the older powders did not contain consistently higher abundances of the derivatives; the nitrated derivatives 2- and 4-nitrodiphenylamine were detected in comparable abundances in powders that were greater than 15 years old as well as those that were less than 10 years old. This indicates that even the newer powders had begun the process of nitrocellulose and nitroglycerin decomposition, resulting in the nitration of diphenylamine. 2,4-dinitrotoluene is added as a deterrent and plasticizer and was detected in five of the 18 unburned powders, with a maximum normalized abundance that was 44% of the maximum abundance of dibutyl phthalate. With the exception of SB-nL (a ‘new’ powder, less than 10 years 48 old), 2,4-dinitrotoluene was only present in aged powders that were greater than 15 years old. The absence of 2,4-dinitrotoluene in most of the newer powders was consistent with findings by the National Research Council (NRC), which determined that isomers of dinitrotoluene have become less frequently used, due to initial regulations by the Environmental Protection Agency (EPA) in the 1980s.3 As of 2011, the EPA classified dinitrotoluene as a pollutant, and section 311(b)(2)(A) of the Federal Water Pollution Control Act designated common isomers of dinitrotoluene as hazardous substances.6 Newer formulations contain additives such as ethyl centralite, Akardite II, and dibutyl phthalate as alternate plasticizers and deterrents. 3.3 Classification of Unburned Powders based on Organic Composition Goudsmits et al. proposed three categories of compounds used for the confirmation of the presence of gunshot residue (GSR).7 Category 1 compounds were compounds that were very strongly associated with GSR and had very restricted applications unrelated to GSR. The four compounds in Category 1 were nitroglycerin, nitroguanidine, ethyl centralite, and methyl centralite.7 A second category included compounds that were strongly associated with GSR but had less restricted applications unrelated to GSR.7 Category 2 compounds were 2,4-dinitrotoluene, Akardite II, 2-nitrodiphenylamine, and 4-nitrodiphenylamine. Diphenylamine was also included in Category 2, but only if its nitrated-derivatives (e.g. 2-nitrodiphenylamine and 4-nitrodiphenylamine) were also present. Diphenylamine has many common environmental uses and thus diphenylamine alone was not considered a Category 2 compound. The third category contained compounds that were associated with GSR and had less restricted applications unrelated to GSR, but were less commonly detected.7 Category 3 compounds included compounds such as other nitrotoluenes, nitrocellulose, and triacetin. 49 The 18 unburned powders analyzed herein contained both Category 1 and 2 compounds but no Category 3 compounds. Additionally, as nitroguanidine was not present in the unburned powders, Category 1 consisted of only three compounds – nitroglycerin, methyl centralite, and ethyl centralite. Category 1 and Category 2 together encompass all compounds detected in this work, with the exception of dibutyl phthalate, which was not included in any category defined by Goudsmits et al.7 The argument for the exclusion of dibutyl phthalate was based on the high prevalence of phthalates in non-GSR applications, such as plasticizers and many polymeric materials.7 The high prevalence of phthalates in the environment lessens the ability of dibutyl phthalate to serve as confirmatory for the presence of GSR. However, dibutyl phthalate was present in every powder analyzed in this work, and has often been included in composition comparisons of smokeless powders.8 Though not originally included as a compound in the original categories proposed by Goudsmits et al., the abundance of dibutyl phthalate was highly varied among the powders analyzed in this work and could be used to differentiate among powders. In some powders, dibutyl phthalate was one of the least abundant compounds, while in 44-N and 44-O, dibutyl phthalate was not only the most abundant compound present in the extracts, but the abundance of dibutyl phthalate in these powders was the highest abundance observed across all compounds detected. Thus, for the purposes of this work, Category 2 comparisons included dibutyl phthalate. 3.3.1 Organic Composition of Unburned Powders, as determined by GC-MS The unburned powder extracts were categorized into different chemical profiles based on the presence or absence of organic compounds in each extract, first based only on the three Category 1 compounds (Table 3.1). 50 Table 3.1 Unburned Smokeless Powder Profiles based on Category 1 Compounds Profile Powder Identity* 1 2 3 4 MG-O 44-N 44-O PMC-L PMC-O (3) 22-O PMC-O (2) AA-O SB-nL Horn-L AA-N Rem-L AEFED-L Win-L SB-N BZR-nL MG-nL Rem-nL Win-nL NG ✓ ✓ ✓ MC ✓ EC ✓ ✓ * Parentheses indicate number of cartridges, if cartridges were chemically different ✓ indicates compound was detected in the sample When only Category 1 compounds were considered, the unburned powders could be divided into four separate profiles. Profile 1 consisted of MG-O, which did not contain any Category 1 compounds, while powders in Profile 2 contained nitroglycerin. Profile 3 included powders that contained nitroglycerin and ethyl centralite, and powders in Profile 4 contained all three Category 1 compounds. The unburned powder extracts were also categorized into chemical profiles based on both Category 1 and Category 2 compounds, which encompassed all nine of the compounds detected by GC-MS and resulted in 14 unique profiles (Table 3.2). 51 Table 3.2 Unburned Smokeless Powder Profiles based on Category 1 and 2 Compounds Profile Powder Identity* NG 2,4-DNT DPA MC EC DBPH 2nDPA AK II 4nDPA 1 2 3 4 5 6 7 8 9 10 11 12 13 14 MG-O 22-O (3) PMC-O (1) AA-O (2) SB-nL AA-O (2) 22-O (2) AA-O (1) Horn-L (3) AA-N Rem-L AEFED-L Win-L (4) Horn-L (2) Win-L (1) SB-N Rem-nL (3) Rem-nL (2) Win-nL 44-N 44-O PMC-L PMC-O (3) BZR-nL MG-nL ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ * Parentheses indicate number of cartridges, if cartridges were chemically different ✓ indicates compound was detected in the sample 52 The addition of Category 2 compounds resulted in further differentiation of the unburned powders. Profile 1 remained unchanged, as MG-O was the only powder that did not contain nitroglycerin. Conversely, 44-N and 44-O were categorized into a separate profile than PMC-L and three cartridges of PMC-O due to the addition of 4-nitrodiphenylamine (Table 3.2) These four powders previously shared the same profile when only Category 1 compounds were included (Table 3.1). The inclusion of Akardite II resulted in the separation of Rem-nL and Win-nL into different profiles (Table 3.2), which initially shared the same profile due to the presence of nitroglycerin, methyl centralite and ethyl centralite (Table 3.1). When only Category 1 compounds were included, twelve of the smokeless powders were classified to Profile 2 (Table 3.1). The addition of Category 2 compounds resulted in the separation of these twelve powders into eight different profiles. For example, BZR-nL and MG-nL were isolated from all other powders because they did not contain diphenylamine. Of the fourteen profiles, seven described powders with unique compositions, indicated by profiles that contained only one type of unburned powder (Table 3.2). For example, all five cartridges of MG-O belonged to Profile 1, which was the only profile that did not contain nitroglycerin, and Profile 8 consists of SB-N, which was the only powder that contained diphenylamine but did not contain 2- or 4-nitrodiphenylamine. Based on the original parameters defined by Goudsmits et al., the absence of these nitrated derivatives of diphenylamine could indicate environmental contamination contributing to the presence of diphenylamine in SB-N.7 The majority of extracts corresponding to cartridges that were obtained from the same original packaging had the same chemical profile. For example, the extracts of all five cartridges of unburned SB-nL ammunition contained the same compounds and thus had the same chemical profile. However, for Horn-L, Rem-nL and Win-L, there was variation even among cartridges 53 obtained from the same original packaging, resulting in different profiles among the cartridges (Table 3.2). For example, the five analyzed cartridges of Horn-L, which were all obtained from the same box of ammunition, displayed differences in chemical composition based on the presence or absence of Akardite II and/or 4-nitrodiphenylamine. Unlike Akardite II, 4-nitrodiphenylamine is not typically added to smokeless powders during production, but rather was present due to the nitration of diphenylamine (Figure 3.4). As such, differences in composition that arise due to the presence or absence of 4-nitrodiphenylamine likely relate more to storage conditions and the aging process rather than differences in original chemical composition of the powders. Decomposition of the powders is accelerated by the presence of nitroglycerin in addition to nitrocellulose, as with double-based smokeless powders. The diphenylamine content is expected to decrease over time as the nitrated products are produced. This was supported by the low abundance of diphenylamine in nearly every powder analyzed. One exception was MG-O, which had a high abundance of diphenylamine (more than ten times more abundant than present in any other powder) and did not contain nitroglycerin. The lack of nitroglycerin would decrease the production of nitric oxide as the powder ages, consequently slowing the rate of diphenylamine nitration. The aged ammunition that was not received in the original packaging had varied composition among the five cartridges analyzed. The differences among the aged ammunition could primarily be attributed to the presence of 2,4-dinitrotoluene and ethyl centralite. For example, cartridges of AA-O were separated into Profiles 2, 4, and 5 (Table 3.2). Cartridges of AA-O belonging to Profile 2 contained higher abundances of 2,4-dinitrotoluene, ethyl centralite and dibutyl phthalate. The cartridge of AA-O in Profile 5 does not contain 2,4-dinitrogoluene, 54 and cartridges belonging to Profile 4 contained 4-nitrodiphenylamine, which was absent in the other cartridges of AA-O. As previously discussed, 2,4-dinitrotoluene can be added to serve as a deterrent and plasticizer but is no longer commonly used in smokeless powder formulations due to toxicity and carcinogenic properties. The differences in organic composition supports the hypothesis that the ‘loose’ ammunition did not originate from the same box of ammunition, which suggests that a single distributor may use different smokeless powders in ammunition production. Considering the organic composition of smokeless powders rather than comparisons based on manufacturer alone provides more information about the similarities between two powders. The profiles of unburned ammunition with leaded primers were also compared to those with non-leaded primers. As expected, there was not a pattern of compositional differences, indicating the primer composition did not impact the formulation of the smokeless powder chosen. 3.3.2 Comparison of Organic Profiles of Unburned Powder: GC-MS and LC-TOFMS Previous analysis of the same 18 unburned powders via liquid chromatography-time of flight mass spectrometry (LC-TOFMS) identified similar compounds (Table A3.3).9 However, LC-TOFMS analysis also identified N-nitrosodiphenylamine in nearly every powder, whereas this compound was not detected by GC-MS. This is due to the high temperatures employed during GC-MS analysis; temperatures above 200ºC are known to cause N-nitrosodiphenylamine to decompose to produce diphenylamine.10,11 LC-TOFMS also detected diaminotoluene, which can act as an explosive or as a stabilizer, in all five cartridges of MG-O. Diaminotoluene was not detected by GC-MS, likely due to low abundance of this compound. 55 When the composition based on GC-MS data differed from the composition determined by LC-TOFMS, the differences were primarily due to compounds present at low abundances. Methyl centralite was not previously detected powders by LC-TOFMS analysis, but was detected at low abundances in cartridges of Rem-L and Win-nL. Analysis by GC-MS identified 2,4-dinitrotoluene, ethyl centralite, dibutyl phthalate and Akardite II in more powders than did LC-TOFMS analysis (Figure 3.5). Conversely, LC-TOFMS analysis detected 4-nitrodiphenylamine in more powders than did GC-MS analysis. s r e d w o P f o r e b m u N 90 80 70 60 50 40 30 20 10 0 84 84 79 79 19 17 89 67 67 45 GC-MS LC-TOFMS 39 30 49 16 NG 2,4-DNT DPA EC DBPH AK II 4nDPA Figure 3.5 Composition of unburned powders as determined by GC-MS, compared to composition determined by LC-TOFMS. Analysis using GC-MS provides advantages over LC-MS for compositional analysis. In LC-TOFMS analyses, Akardite II, N-nitrosodiphenylamine, 4-nitrodiphenylamine, nitroglycerin, and diaminotoluene were only identified after multiplexed collision-induced dissociation (CID) was employed. GC-MS analysis resulted in the identification of Akardite II and 4-nitrodiphenylamine using the NIST/EPA/NIH Mass Spectral Library, without needing further analyses. Additionally, LC-TOFMS negative-ion mode was necessary for the detection of 56 nitroglycerin and 2,4-dinitrotoluene, requiring each sample to be analyzed in both positive-ion and negative-ion mode. With the exception of N-nitrosodiphenylamine and diaminotoluene, GC-MS detected all compounds that LC-TOFMS detected, without a requiring a second analysis. GC-MS is also more widely available in forensic laboratories than LC-TOFMS. The categories proposed by Goudsmits et al. were originally proposed to identify compounds that could be used for the confirmation of GSR, and thus were developed based on the presence of the identified compounds.7 Abundances of the identified compounds were not considered. As seen with the high abundance of ethyl centralite and dibutyl phthalate, differences in abundance of compounds could be used in conjunction with the presence or absence of compounds to investigate association or differentiation of powders, rather than just identification of a residue as GSR. Rather than visually determining trends in composition among the unburned powder extracts, chemometric procedures were used as a more objective method to observe patterns in the data. 3.4 Association and Differentiation of Unburned Powders using Principal Components Analysis As previously discussed in Section 3.2, two peaks provisionally identified as nitroglycerin were detected in all double-based powders. The presence of two peaks was determined to be an artifact of the instrument rather than the samples, and only the second peak (tR 11.0 min) was determined to be chemically relevant. To investigate the impact of only including the second peak attributed to nitroglycerin (tR 11.0 min) on multivariate statistical analyses, PCA was performed on four different data sets. In addition to all other identified variables, data sets including (1) both peaks attributed to nitroglycerin, (2) only the peak at tR 10.6 min, (3) only the peak at tR 11.0 min, and (4) neither peak attributed to nitroglycerin were 57 separately subjected to PCA. The scores plots illustrated similar powder groupings among the data sets, regardless of which nitroglycerin peak was included (or excluded). As excluding the first peak attributed to nitroglycerin (tR of 10.6 min) did not change the overall groupings observed, all further analyses including nitroglycerin were performed using the second peak (tR 11.0 min) to represent the abundance of nitroglycerin. 3.4.1 Principal Components Analysis of Unburned Powders, based on GC-MS data Principal components analysis (PCA) was performed on the normalized abundances of the three Category 1 compounds (nitroglycerin, methyl centralite and ethyl centralite) in the 18 unburned powders (Figure 3.6). The first two principal components (PCs) described 99.9% of the variance in the data set. 58 ) . % 4 9 1 ( 2 C P ) . % 4 9 1 ( 2 C P a) 4 0 -4 -8 Group 2 MG-O Group 1 SB-N MG-nL BZR-nL AA-O (2) -4 44 N BZR nL MG O 44 O MG nL Horn L Group 3 All remaining powders PC1 (80.5%) AA O SB N Rem nL AA N SB nL Rem L 0 4 22 O PMC L Win nL AE-FED L PMC O Win L b) EC 1 0 -1 -1 MC NG 0 PC1 (80.5%) 1 Figure 3.6 Principal components analysis (PCA) of unburned powders using Category 1 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 59 Group 1 consisted of unburned powder extracts of BZR-nL, MG-nL, SB-N and two cartridges of AA-O, which had the highest abundances of ethyl centralite, at approximately three to nine times higher abundance than any of the remaining powders. Ethyl centralite was weighted negatively on PC1, resulting in the negative positioning of these four powders, while the remaining powders were positioned positively on PC1 (Figure 3.6). One exception is a third cartridge of AA-O; this cartridge had an abundance of ethyl centralite that was much higher than the other remaining powders, but still only approximately 30% as abundant as the powders in Group 1. Ethyl centralite was weighted positively on PC2 while nitroglycerin was weighted negatively. BZR-nL and MG-nL had lower abundances of ethyl centralite than SB-N and the two cartridges of AA-O, resulting in less positive positioning on PC2. Additionally, BZR-nL and MG-nL also had higher abundances of nitroglycerin, contributing further to negative positioning on PC2. All powders of Group 1 belonged to the same profile in Table 3.1, due to the presence of both nitroglycerin and ethyl centralite. While Group 1 only contained powders belonging to Profile 3 and Group 2 consisted of all five cartridges of MG-O (Profile 1), Group 3 contained the remaining powders from Profiles 2, 3, and 4. When only Category 1 compounds were considered, all five cartridges of MG-O were positioned almost identically, with the most positive positioning possible on both PC1 and PC2. Regardless of the chemical profile, the remaining powders were positioned positively on PC1 and were spread along PC2. These powders had abundances of ethyl centralite that were less than 10% the abundance of ethyl centralite in the powders positioned negatively on PC1. Because only three compounds were included as variables for PCA, the high abundance of ethyl 60 centralite in Group 1 dominated the variance. When only Category 1 compounds were included, PCA allowed for limited differentiation among the unburned powders. Principal components analysis was then performed on the normalized abundances of the combined Category 1 and 2 compounds, where all nine of the identified compounds were included. The first three PCs accounted for 89.4% of the total variance in the data set. The resultant scores plot and loadings plot for PC1 vs PC2 are shown in Figure 3.7, while the scores plot and loadings plot for PC1 vs PC3 are shown in Figure 3.8. 61 10 a) 0 Group 1 Group 2 Group 4 Group 3 ) . % 9 5 2 ( 2 C P -10 -10 44 N BZR nL MG O 44 O MG nL Horn L 0 PC1 (42.5%) AA O SB N Rem nL AA N SB nL Rem L 10 22 O PMC L Win nL AE-FED L PMC O Win L ) . % 9 5 2 ( 2 C P b) 1 0 -1 -1 EC NG DBPH AK II 4-nDPA MC 2-nDPA 2,4-DNT DPA 0 PC1 (42.5%) 1 Figure 3.7 Principal components analysis (PCA) of unburned powders using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 62 10 a) Group 1 ) % 1 2 ( 3 C P 5 0 -5 -8 1 0 ) % 1 2 ( 3 C P -1 -1 44 N BZR nL MG O b) EC Group 3 Group 2 Group 4 -4 44 O MG nL Horn L 0 PC1 (42.5%) AA O SB N Rem nL AA N SB nL Rem L 4 22 O PMC L Win nL 8 AE-FED L PMC O Win L DBPH DPA 2,4-DNT 2-nDPA 4-nDPA MC AK II NG 0 PC1 (42.5%) 1 Figure 3.8 Principal components analysis (PCA) of unburned powders using Category 1 and 2 compounds (a) scores plot of PC1 vs PC3 and (b) loadings plot of PC1 vs PC3 63 The variance described by the first three PCs was dominated by ethyl centralite, dibutyl phthalate, nitroglycerin, 2,4-dinitrotoluene and diphenylamine. On PC1, ethyl centralite was the most influential contributor, followed closely by dibutyl phthalate. Dibutyl phthalate was weighted most heavily on PC2, followed by ethyl centralite, diphenylamine and 2,4-dinitrotoluene. Ethyl centralite was also the most influential contributor to PC3, followed by diphenylamine, 2,4-dinitrotoluene, nitroglycerin, and dibutyl phthalate. Group 1, which was positioned negatively on PC1 and positively on PC2 (Figure 3.7a), again consisted of powders of MG-nL, BZR-nL, SB-N and two cartridges of AA-O. This was consistent with the grouping observed on the first two PCs when only Category 1 compounds were included (Figure 3.6a) and again can be attributed to high abundances of ethyl centralite in these powders. While the powders in Group 1 all share similarly high abundances of ethyl centralite, they belong to three different profiles (Table 3.2). The differences between the profiles were due to compounds that contributed less to the overall organic composition of each cartridge than ethyl centralite. For example, the two cartridges of AA-O and five cartridges of SB-N all contained diphenylamine and 2-nitrodiphenylamine, while these compounds were absent in MG-nL and BZR-nL. However, the abundance of diphenylamine was less than 0.1% of the total organic composition of each of these powders, while ethyl centralite made up more than 50% of the composition. The two cartridges of AA-O were separated from the rest of Group 1 on PC3 (Figure 3.8a), due to the presence of positively weighted 2,4-dinitrotoluene and dibutyl phthalate on PC3. Dibutyl phthalate and 2,4-dinitrotoluene were either absent or present in low abundances in the other powders in Group 1. Positioning of SB-N on PC1 vs PC3 can be explained in a similar manner. 64 The positioning of Groups 2 and 3 can also be explained using the loadings plots in Figures 3.7b and 3.8b. Group 2 consisted of the 44-N and 44-O cartridges, which were positioned positively on both PC1 and PC2 (Figure 3.7a) due to high abundance of dibutyl phthalate. Group 3 consisted of all five cartridges of MG-O, which were positioned positively on PC1 and negatively on PC2 (Figure 3.7a) due to high abundance of diphenylamine and 2,4-dinitrotoluene, and lack of nitroglycerin. All five cartridges of MG-O had the same chemical profile (Table 3.2). However, the five analyzed cartridges of MG-O differed greatly in abundances of 2,4-dinitrotoluene, diphenylamine and 2-nitrodiphenylamine, resulting in the spread observed on PC2. All of the remaining powders belonged to Group 4 and were positioned around the origin. Though the remaining powders belonged to ten different profiles (Table 3.2), nitroglycerin, diphenylamine, 2-nitrodiphenylamine, and dibutyl phthalate were common to all powders in Group 4, and approximately 85% of the powders in this group also contained ethyl centralite. Additionally, the profiles of the powders belonging to Group 4 differed in the presence of methyl centralite, 4-nitrodiphenylamine, and Akardite II, which did not contribute heavily to the variance on either PC (Figure 3.7b). This helps to explain why powders belonging to many different profiles were positioned together. The powders in Group 4 that contained dibutyl phthalate and ethyl centralite had lower abundances of dibutyl phthalate and ethyl centralite compared to powders in Groups 1 and 2. For example, while extracts of Win-nL, Rem-nL and Horn-L all had higher than average abundance of dibutyl phthalate, the abundances were still only approximately 50% of the abundance of dibutyl phthalate in 44-N and 44-O. Examination of the scores plot of PC1 vs PC2 (Figure 3.7b) 65 did show that Win-nL, Rem-nL and Horn-L were positioned more positively on PC1 than the rest of the powders in Group 4, but these powders were not clearly isolated as a separate group. The majority of extracts corresponding to cartridges that were obtained from the same original packaging were positioned similarly on the scores plot of PC1 vs PC2 (Figure 3.7a). However, the differences in chemical composition among cartridges of Horn-L, Rem-nL, and Win-L resulted in a spread in positioning on the scores plot (Figure 3.7a). The aged ammunition that was not received in the original packaging had varied composition among the five cartridges analyzed, resulting in differences in positioning on the scores plot of PC1 vs PC2 (Figure 3.7a). The most notable example of this varied positioning can be seen with AA-O. Differences in 2,4-dinitrotoluene and ethyl centralite abundance among the five cartridges of AA-O resulted in two cartridges belonging to Group 1 while the final three were positioned in Group 4 (Figure 3.7). The two cartridges of AA-O that were positioned with Group 1 had much higher abundances of 2,4-dinitrotoluene, ethyl centralite, and dibutyl phthalate than the remaining three cartridges. This is consistent with the PCA scores plot of Category 1 compounds (Figure 3.6), where the same two cartridges were positioned with Group 1 due to higher abundance of ethyl centralite, and with Table 3.2, where cartridges of AA-O were divided among Profiles 2, 4, and 5. While Category 1 compounds may be the most significant for the confirmation of GSR, the inclusion of the additional compounds in Category 2 allowed for increased differentiation among unburned powders, primarily due to dibutyl phthalate, 2,4-dinitrotoluene and diphenylamine. 66 3.4.2 Comparison of PCA of Smokeless Powders using GC-MS to PCA using LC-TOFMS data Principal components analysis (PCA) performed using normalized abundances from GC-MS analysis (Figure 3.7) was then compared to previously published PCA using LC-TOFMS data of the same powders (Figure 3.9).9 67 4000 3000 2000 1000 0 -1000 -2000 -3000 ) . % 7 3 2 ( 2 C P a) Group A Group I Group B Group C Group F Group E Group H Group D -4000 -3000 Group G -2000 -1000 0 1000 3000 4000 5000 6000 7000 44 N BZR nL MG O 44 O MG nL Horn L 1 b) 2000 PC1 (49%) AA O SB N Rem nL AA N SB nL Rem L AKII . ) 7 3 2 ( 2 C P 0 -1 -1 N-nitrosoDPA DPA DBPH 0 PC1 (49%) 22 O PMC L Win nL AE-FED L PMC O Win L EC 1 Figure 3.9 Principal components analysis (PCA) of unburned powders using LC-TOFMS data (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC29 68 Nine groups were identified on the scores plot when PCA was performed using LC-TOFMS data (Figure 3.9a). Positioning was based on similarities in organic composition, with ethyl centralite, diphenylamine, Akardite II, dibutyl phthalate and N-nitrosodiphenylamine contributing most to the variance described by the data set (Figure 3.9b). Ethyl centralite contributed the most to positive positioning on PC1, while diphenylamine was weighed most negatively on PC1 (Figure 3.9b). Akardite II contributed the most heavily to positive positioning on PC2, while dibutyl phthalate contributed most heavily to negative positioning on PC2 (Figure 3.9b). Conversely, PCA performed on the normalized abundances from GC-MS analysis resulted in four groups, with the variance primarily due to differences in abundance of ethyl centralite, dibutyl phthalate, 2,4-dinitrotoluene and diphenylamine (Figure 3.7). The scores plot based on LC-TOFMS data (Figure 3.9a) provided increased differentiation of some of the unburned powders that belonged to Group 4 of the scores plot based on GC-MS data (Figure 3.7a). For example, Win-L and Win-nL were positioned in Group 4 on the scores plot based on GC-MS data (Figure 3.7a), while these two powders formed Group I in Figure 3.9a. Unburned Rem-L and AEFED-L were also isolated on Figure 3.9a, forming Group A on the scores plot based on LC-TOFMS data. There were also number of similarities between the scores plot based on LC-TOFMS data (Figure 3.9a) and the scores plot based on GC-MS data (Figure 3.7a). MG-nL and BZR-nL were positioned closely together on the LC-TOFMS scores plot of PC1 vs PC2 and are labeled as Group D (Figure 3.9a).9 While SB-N and two cartridges of AA-O were positioned closely to MG-nL and BZR-nL on PC1 vs PC2 based on the GC-MS data (Group 1), PCA of the LC-TOFMS data resulted in the isolation of SB-N (Group C) as well as the isolation of the two cartridges of AA-O (Group F). When PCA was performed using the GC-MS data, AA-O was 69 separated from MG-nL, BZR-nL and SB-N on the scores plot of PC1 vs PC3 (Figure 3.8a). Unburned AA-O contained 2,4-dinitrotoluene, which was weighted positively on PC3 and was not present in MG-nL, BZR-nL or SB-N. Additionally, 44-N and 44-O formed a distinct group on the PCA scores plot of the previously collected LC-TOFMS data (Group G).9 This was also observed by PCA of GC-MS data (Group 2, Figure 3.7a). In both cases, the isolation of 44-N and 44-O was due to the high abundance of dibutyl phthalate in these powders. The compositional differences among the five cartridges each of AA-O, 22-O and PMC-O were also observed by PCA of LC-TOFMS data, most notably resulting in the positioning of AA-O along PC1.9 However, previous LC-TOFMS analysis did not result in the isolation of MG-O. Unburned MG-O was isolated on the scores plot based on GC-MS data (Group 3, Figure 3.7a) due to the absence of nitroglycerin and high abundance of 2,4-dinitrotoluene. In LC-TOFMS analysis, nitroglycerin and 2,4-dinitrotoluene were detected using negative-ion mode, while the other additives were detected using positive-ion mode. Principal components analysis of the LC-TOFMS data was performed only on the data obtained from positive-ion mode, and thus did not include nitroglycerin or 2,4-dinitrotoluene. Analysis by GC-MS detected nitroglycerin and 2,4-dinitrotoluene in the same analysis as the other additives. The identification of all compounds in a single analysis saves time and resources, both of which are in high demand in forensic laboratories. 70 3.5 Association and Differentiation of Unburned Powders using Hierarchical Cluster Analysis The normalized data set comprised of Category 1 compounds was also subjected to hierarchical cluster analysis (HCA). The resulting dendrogram (Figure 3.10) can be used to visualize groupings based on compositional similarity. HCA resulted in similar groupings as PCA but has the added advantage of providing a metric to quantify the extent of similarity among groups. With only the three Category 1 compounds included (nitroglycerin, methyl centralite, and ethyl centralite), the powders that contained a high abundance of ethyl centralite (Group 1 on Figure 3.6) clustered separately from all other powders (Figure 3.10). 71 ) % ( l e v e L y t i r a l i m S i 0 50 100 Group 3 All remaining powders Group 2 MG-O Horn-L (2) 44-N (1) Group 1 MG-nL BZR-nL SB-N AA-O (2) Figure 3.10 HCA dendrogram of Unburned Powders using Category 1 Compounds. Colors correspond to the legend of Figure 3.6. 72 At a similarity level of 59%, HCA resulted in three clusters that were consistent with the three groups observed by PCA of the Category 1 compounds (Figure 3.6). MG-nL, BZR-nL, SB-N, and two cartridges of AA-O (C3 and C4), which formed Group 1 in the PCA scores plot (Figure 3.7a), clustered at a similarity level of 63% and did not cluster to the rest of the powders until a similarity level of 0% (Figure 3.10). This indicates that the powders in Group 1 were the most dissimilar from all the remaining powders. These powders all had similarly high abundances of ethyl centralite, with abundances greater than any other powder in the sample set. Similar to PCA using Category 1 compounds, BZR-nL and MG-nL are the most similar to each other, clustering together at a similarity level of 84% (Figure 3.10). Interestingly, some of the cartridges of SB-N were more similar to the two cartridges of AA-O than the other cartridges of SB-N. Examination of the chromatographic data showed that the cartridges clustered with AA-O had abundances of ethyl centralite more similar to that observed in the two cartridges of AA-O, while the final two cartridges of SB-N had lower abundances of ethyl centralite. Because MG-O did not contain any of the compounds included in Category 1, it was expected that the cartridges of MG-O would be separated from all other powders. The five cartridges of MG-O did form a cluster with each other first, at a similarity level of 95%, but then this group clustered to two cartridges of Horn-L and one cartridge of 44-N at a similarity level of 80% (Figure 3.10). This was consistent with what was observed by PCA (Figure 3.6). These cartridges contained little to no ethyl centralite, and also had the very low abundances of nitroglycerin, explaining the similarity to MG-O. These eight powders clustered with the rest of the powders at a similarity level of 40% (Figure 3.10). Dibutyl phthalate was not included as a Category 1 compound, resulting in the clustering of 44-N and 44-O with three cartridges of SB-nL, four cartridges of Rem-nL, two cartridges of 73 Horn-L, one cartridge of AEFED-L, and two cartridges of 22-O at a similarity level of 82% (Figure 3.10). This group then clustered to one cartridge of AA-O, at a similarity level of 71%. This cartridge of AA-O had lower abundances of ethyl centralite than the two cartridges included in Group 1, but higher than the final two cartridges of AA-O. HCA was then repeated, using the Category 1 and Category 2 compounds (Figure 3.11). Powders from MG-nL, BZR-nL, SB-N, and two cartridges of AA-O, which formed Group 1 in the PCA scores plot (Figure 3.7a), clustered at a similarity level of 65% on the HCA dendrogram (Figure 3.11). In addition to the previously described similarity of MG-nL and BZR-nL, all analyzed cartridges of SB-N also had a similarly high abundance of ethyl centralite, and also contained nitroglycerin at abundances similar to BZR-nL and MG-nL. However, SB-N also contained low abundance of Akardite II and diphenylamine, resulting in a slightly lower similarity level to BZR-nL and MG-nL (Figure 3.11). The cartridge of SB-N that clustered to the other members of Group 1 at a lower similarity level was due to higher abundances of ethyl centralite and Akardite II, while the other four cartridges analyzed were very similar in abundance. Finally, the two cartridges of AA-O had the highest abundances of ethyl centralite present in the entire set of unburned powders, but also contained 2,4-dinitrotoluene and dibutyl phthalate, resulting in a similarity level of 65% to BZR-nL, MG-nL and SB-N (Figure 3.11). 74 ) % ( l e v e L y t i r a l i m S i 0 50 100 Group 4 All remaining powders Group 1 MG-nL BZR-nL SB-N AA-O (2) Group 2 44-N 44-O Group 3 MG-O Figure 3.11 Hierarchical cluster analysis (HCA) dendrogram of unburned powder extracts using Category 1 and 2 compounds. Colors correspond to the legend of Figure 3.7. 75 Unburned 44-N and 44-O, which formed Group 2 in the PCA scores plot, clustered at a similarity level of 58%, and did not cluster to other powders until a similarity level of 20% (Figure 3.11). The high similarity of 44-N and 44-O was primarily due to the high abundance of dibutyl phthalate in these powders. These powders also did not contain ethyl centralite and had lower abundances of nitroglycerin than the majority of the other double-based powders. All cartridges of 44-N and 44-O shared the same profile, which was distinct from all other powders in the sample set (Table 3.2). The five cartridges of MG-O, which formed Group 3 on the PCA scores plot, clustered at a similarity level of 38% (Figure 3.11). This cluster formed separately from all other powders, indicating the lowest level of similarity to the rest of the powders. All five cartridges of MG-O shared the same profile, distinct from every other powder in the sample set (Table 3.2). As previously discussed, MG-O was the only powder that did not contain nitroglycerin, and also had the highest abundances of 2,4-dinitrotoluene and diphenylamine in the sample set. The range of similarities within this cluster (38-93%) was due to differences in abundance of 2,4-dinitrotoluene and diphenylamine. These differences also caused a spread in positioning on the PCA scores plot (Figure 3.7a). The remaining powders, which formed Group 4 in the PCA scores plot, clustered at a similarity level of 50% (Figure 3.11). The powders in Group 1 clustered to Group 4 at a similarity level of 45%, and Group 2 clustered to the group containing Group 1 and Group 4 at a similarity level of 30%. 76 3.6 Summary of the Association and Differentiation of Unburned Powders Before any conclusions can be drawn about the firing process, the organic composition of the corresponding unburned powder must be characterized. The known composition of the unburned powder will allow conclusions to be drawn regarding any potential burned products that are formed during the firing process or products that may be artifacts of the weapon or cartridge. The unburned powder extracts were analyzed by gas chromatography-mass spectrometry (GC-MS) and the resultant chromatographic abundances were subjected to principal components analysis (PCA) and hierarchical cluster analysis (HCA) to demonstrate association and differentiation among different powders based on the organic composition. First, compound categories based on those proposed by Goudsmits et al. were used to classify the unburned powders into different profiles.7 Category 1 compounds included nitroglycerin, methyl centralite and ethyl centralite. When only these compounds were considered, the unburned powders were separated into four groups. Category 2 compounds included 2,4-dinitrotoluene, Akardite II, diphenylamine, 2-nitrodiphenylamine, and 4-nitrodiphenylamine. Additionally, dibutyl phthalate was included as a Category 2 compound. Classification based on both Category 1 and Category 2 compounds resulted in 14 unique organic profiles. Principal components analysis (PCA) was first performed using the abundances of Category 1 compounds. The powders with the highest abundance of ethyl centralite (Group 1) were isolated on the PCA scores plot. When all compounds were included (Category 1 and Category 2), the PCA scores plot revealed four main groupings based on organic composition. Groups 1 and 2 were composed of powders that had high abundances of ethyl centralite and dibutyl phthalate, respectively. Group 3 consisted of MG-O, which was the only single-based 77 powder analyzed. Group 4 contained the remainder of the powders, which contained moderate abundances of the identified compounds. The same data set was subjected to HCA, and the resultant HCA dendrogram gave similar groupings. Powders that had very high abundance of ethyl centralite or dibutyl phthalate were clustered together, while MG-O formed a group that did not cluster to the rest of the powders. HCA and PCA of chromatographic data were used to associate and differentiate among unburned powders based on organic composition. It was necessary to establish organic profiles of the unburned powders prior to fired cartridge residue analysis, so that any compounds resulting from the firing process could be eliminated, and association of fired cartridges to unburned powders could be investigated. 78 APPENDIX 79 Table A3.1 Composition of Unburned Powders, as determined by GC-MS APPENDIX Powder Identity NG 2,4-DNT DPA MC EC DBPH 2nDPA AK II 4nDPA 22-O (1,3,5)* 22-O (2,4) ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ AA-O (1) AA-O (2,5) AA-O (3,4) AEFED-L BZR-nL Horn-L (1,5) Horn-L (2,3,4) MG-nL MG-O PMC-L Rem-L 44-N 44-O AA-N ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ PMC-O (1,2,4) ✓ ✓ PMC-O (3) ✓ PMC-O (5) ✓ Rem-nL (1,3,5) ✓ ✓ Rem-nL (2,4) ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ *Parentheses indicate specific cartridges ✓ indicates compound was detected in the sample ✓ ✓ ✓ SB-N SB-nL Win-L (2,3,5) Win-L (1,4) Win-nL ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ 80 Table A3.2 Approximate Retention Times of Compounds Approximate Retention Time (tR, min) 11.0 12.2 12.7 13.6 13.8 13.9 14.1 14.2 15.3 Compound Nitroglycerin (NG) 2,4-Dinitrotoluene (2,4-DNT) Diphenylamine (DPA) Methyl Centralite (MC) Ethyl Centralite (EC) Dibutyl Phthalate (DBPH) 2-Nitrodiphenylamine (2nDPA) Akardite II (AKII) 4-Nitrodiphenylamine (4nDPA) 100 46 ) % ( y t i s n e t n I e v i t a l e R 80 60 40 20 0 76 50 100 150 m/z 200 250 300 Figure A3.1 Representative mass spectrum of nitroglycerin 81 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R 89 63 119 50 100 165 182 150 m/z 200 250 300 Figure A3.2 Representative mass spectrum of 2,4-dinitrotoluene 169 84 51 77 50 100 150 m/z 200 250 300 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R Figure A3.3 Representative mass spectrum of diphenylamine 82 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R 134 106 77 240 50 100 150 m/z 200 250 300 Figure A3.4 Representative mass spectrum of methyl centralite 120 148 77 268 50 100 150 m/z 200 250 300 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R Figure A3.5 Representative mass spectrum of ethyl centralite 83 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R 149 50 100 150 m/z 200 250 300 Figure A3.6 Representative mass spectrum of dibutyl phthalate 167 214 180 77 139 50 100 150 m/z 200 250 300 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R Figure A3.7 Representative mass spectrum of 2-nitrodiphenylamine 84 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R 169 77 226 50 100 150 m/z 200 250 300 Figure A3.8 Representative mass spectrum of 1-methyl-3,3-diphenylurea (Akardite II) 100 80 60 40 20 0 ) % ( y t i s n e t n I e v i t a l e R 167 214 184 77 50 100 150 m/z 200 250 300 Figure A3.9 Representative mass spectrum of 4-nitrodiphenylamine 85 UBP_22-O_Cartridge1 4 2 e c n a d n u b A d e z i l a m r o N 0 8 10 12 time (min) 14 16 Figure A3.10 Representative Chromatogram of Unburned Powder 22-O (1) UBP_22-O_Cartridge2 4 2 e c n a d n u b A d e z i l a m r o N 0 8 10 12 time (min) 14 16 Figure A3.11 Representative Chromatogram of Unburned Powder 22-O (2) 86 UBP_44-N_Cartridge1 10 8 6 4 2 0 e c n a d n u b A d e z i l a m r o N 8 10 12 time (min) 14 16 Figure A3.12 Representative Chromatogram of Unburned Powder 44-N UBP_44-O_Cartridge2 10 8 6 4 2 0 e c n a d n u b A d e z i l a m r o N 8 10 12 time (min) 14 16 Figure A3.13 Representative Chromatogram of Unburned Powder 44-O 87 UBP_AA-O_Cartridge1 4 2 e c n a d n u b A d e z i l a m r o N 0 8 10 12 time (min) 14 16 Figure A3.14 Representative Chromatogram of Unburned Powder AA-O (1) UBP_AA-O_Cartridge2 4 2 e c n a d n u b A d e z i l a m r o N 0 8 10 12 time (min) 14 16 Figure A3.15 Representative Chromatogram of Unburned Powder AA-O (2) 88 UBP_AA-O_Cartridge3 8 6 4 2 0 e c n a d n u b A d e z i l a m r o N 8 10 12 time (min) 14 16 Figure A3.16 Representative Chromatogram of Unburned Powder AA-O (3) UBP_BZR-nL_Cartridge1 8 6 4 2 0 e c n a d n u b A d e z i l a m r o N 8 10 12 time (min) 14 16 Figure A3.17 Representative Chromatogram of Unburned Powder BZR-nL 89 UBP_MG-O_Cartridge1 6 4 2 e c n a d n u b A d e z i l a m r o N 0 8 10 12 time (min) 14 16 Figure A3.18 Representative Chromatogram of Unburned Powder MG-O UBP_SB-N_Cartridge1 8 6 4 2 0 e c n a d n u b A d e z i l a m r o N 8 10 12 time (min) 14 16 Figure A3.19 Representative Chromatogram of Unburned Powder SB-N 90 Table A3.3 Chemical Composition of Unburned Smokeless Powders, as determined by LC-TOFMS9 Profile Powders NG DBPH DPA EC AKII N-nitroso DPA ✓ ✓ ✓ ✓ 1 ✓ ✓ ✓ ✓ 4nDPA Diamino- toluene 2,4-DNT ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Win-L Rem-L AEFED-L Horn-L Rem-nL AA-O (1) AA-O (2) AA-O (5) AA-O (3) AA-O (4) PMC-O (3) PMC-O (5) ✓ ✓ Win-nL PMC-L BZR-nL SB-nL MG-nL 44-N 44-O SB-N MG-O ✓ ✓ ✓ ✓ ✓ ✓ ✓ PMC-O (1) ✓ PMC-O (2) PMC-O (4) 22-O (1) 22-O (2) 22-O (4) 22-O (3) 22-O (5) 2 3 4 5 6 7 8 9 10 11 12 13 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ 14 ✓ ✓ *Parentheses indicate specific cartridges ✓ indicates compound was detected in the sample 91 REFERENCES 92 REFERENCES (1) Stenerson, K. K.; Buchanan, M. D. Prevent GC Inlet Problems Before They Cost You Time and Money https://www.sigmaaldrich.com/technical-documents/articles/reporter-us/gc- inlet-problem-prevention.html (accessed Jan 27, 2020). (2) NIST Standard Reference Database Number 69; P.J. Lindstrom, W.G. Mallard, Eds.; NIST Chemistry WebBook; National Institute of Standards and Technology: Gaithersburg MD, 20899. (3) National Research Council. Black and Smokeless Powders: Technologies for Finding Bombs and the Bomb Makers; National Academies Press: Washington, D.C., 1998. https://doi.org/10.17226/6289. (4) Apatoff, J. B.; Norwitz, G. Role of Diphenylamine as a Stabilizer in Propellants; Analytical Chemistry of Diphenylamine in Propellants (A Survey Report). 1973. (5) Bergens, A.; Danielsson, R. Decomposition of Diphenylamine in Nitrocellulose Based Propellants. Talanta 1995, 42 (2), 171–183. (6) Environmental Protection Agency. Technical Fact Sheet – Dinitrotoluene (DNT) https://www.epa.gov/sites/production/files/2017- 10/documents/emerging_contaminant_dinitrotoleune_dnt.pdf (accessed Sep 5, 2019). (7) Goudsmits, E.; Sharples, G. P.; Birkett, J. W. Preliminary Classification of Characteristic Organic Gunshot Residue Compounds. Science & Justice 2016, 56 (6), 421–425. https://doi.org/10.1016/j.scijus.2016.06.007. (8) Joshi, M.; Rigsby, K.; Almirall, J. R. Analysis of the Headspace Composition of Smokeless Powders Using GC–MS, GC-ΜECD and Ion Mobility Spectrometry. Forensic Science International 2011, 208 (1–3), 29–36. https://doi.org/10.1016/j.forsciint.2010.10.024. (9) Reese, K. L.; Jones, A. D.; Waddell Smith, R. Characterization of Smokeless Powders Using Multiplexed Collision-Induced Dissociation Mass Spectrometry and Chemometric Procedures. Forensic Science International 2017, 272, 16–27. (10) Environmental Protection Agency. Method 8070A: Nitrosamines by Gas Chromatography. December 1996. (11) Lennert, E.; Bridge, C. Analysis and Classification of Smokeless Powders by GC–MS and DART-TOFMS. Forensic Science International 2018, 292, 11–22. https://doi.org/10.1016/j.forsciint.2018.09.003. 93 4. Association and Discrimination of Fired Cartridge Residues to Unburned Powders using Multivariate Statistical Analysis of Organic Compounds 4.1 Introduction Four types of ammunition were selected for fired cartridge residue analysis. Following examination of the corresponding subset of unburned powders, fired cartridge residue extracts were analyzed in replicate by gas chromatography-mass spectrometry (GC-MS) to generate organic profiles. Fired cartridge residues from cartridges that were fired without barrel cleaning (“Successive”) were compared to cartridges fired when the barrel of the weapon had been cleaned prior to each round (“Cleaned”). Cleaning the barrel did not impact the composition of the resultant fired cartridge residue. Though the effects of the firing process were proven to be inconsistent and resulted in variable abundances, the most abundant compounds present in the unburned powders were generally maintained in the corresponding fired cartridge residues. Across all ammunitions, ethyl centralite, dibutyl phthalate and/or nitroglycerin were present at the highest abundance in both the unburned powders and fired cartridge residues. The chromatographic abundances of compounds in unburned powders and the fired cartridge residues were subjected to principal components analysis (PCA) and hierarchical cluster analysis (HCA) to investigate association of the fired cartridge residues to the corresponding unburned powders. Association success was highly dependent only the composition of the unburned powder. Unburned ammunition that contained high abundances of ethyl centralite or dibutyl phthalate (e.g. MG-nL and Win-nL, respectively) resulted in fired cartridges that were chemically similar enough for successful association. Meanwhile, fired cartridges that contained lower abundances of ethyl centralite or dibutyl phthalate resulted in limited association success. 94 4.2 Comparison of the Organic Composition of Fired Cartridge Residues to the corresponding Unburned Powders Of the original 18 unburned powders, a subset of four powders was selected to investigate association of fired cartridge residues to the corresponding unburned powders. The unburned subset consisted of the previously analyzed cartridges of MG-nL, Win-nL, Win-L and Rem-L. Details of the chemical composition of these powders are described in detail in Chapter 3. Briefly, nitroglycerin (NG), ethyl centralite (EC) and dibutyl phthalate (DBPH) were present in all four unburned powders. Unburned Win-nL was the only powder that contained methyl centralite (MC), and 4-nitrodiphenylamine (4nDPA) was only present in one cartridge of unburned Win-L. Unburned MG-nL did not contain diphenylamine (DPA), 2-nitrodiphenylamine (2nDPA), or Akardite II (AKII), which were present in the other three powders. A comparison of the composition of the unburned powders in the subset can be found in Appendix Figure A4.1. All fired cartridge residues were collected analyzed in replicate by GC-MS, as described in Chapter 2. Briefly, ten cartridges of each of the four types of ammunition were fired. The barrel of the weapon was cleaned before each cartridge for the first five cartridges (“Cleaned”), while the last five were fired without barrel cleaning between each cartridge (“Successive”). The goal of cleaning the barrel was to have each Cleaned cartridge collected under equivalent conditions, whereas the Successive residues were used to investigate carryover between fired rounds. Fired residue carryover from previously fired cartridges was expected to be present on the exterior of the fired cartridge, due to passage through the barrel. Contamination from prior fired rounds has also been shown to contribute to the composition of subsequent gunshot 95 residue.1 However, it was not known if the carryover would impact the residue inside the fired cartridge. Thus, the residue in the fired cartridges was extracted by rinsing the inside of the cartridge rather than soaking the entire cartridge in solvent. If carryover from previous cartridges impacted the fired cartridge residue, then multivariate statistical analyses would be impacted by carryover rather than true chemical differences among the fired cartridges. Ethyl centralite was present at the highest abundance in the MG-nL fired cartridge residues, while the maximum abundances of dibutyl phthalate and nitroglycerin were both detected in Win-nL fired cartridge residues. Initial examination of the chromatographic abundances of the fired cartridge residues showed that regardless of weapon cleaning, fired Rem-L and Win-L contained lower abundances of all compounds than fired MG-nL and Win-nL. The barrel of the weapon was much dirtier after the leaded (L) cartridges were fired when compared to the non-leaded (nL) cartridges. The lower abundances of the leaded fired cartridge extracts might be due to loss of residue that remained in the barrel of the weapon. Lower abundances were also observed in the unburned Win-L and Rem-L when compared to unburned Win-nL and MG-nL. A comparison of the composition of all fired cartridge residues can be found in Appendix Figure A4.2. Due to the variable nature of the firing process, differences among the cartridges within each ammunition type were expected.2 However, despite differences in abundance, the compositions of the five fired cartridges of each type of ammunition were more similar than anticipated. Additionally, the compositions of the fired cartridges were largely consistent with the compositions of the corresponding unburned powders (Tables 4.1-4.4) but contained lower abundances of all compounds. No additional compounds such as pyrolysis products resulting from the firing process were detected in the fired cartridge residues. Representative 96 chromatograms for the fired cartridge residues and the corresponding unburned powders are shown in Figures 4.1-4.4. The peaks eluting before ten minutes were attributed to chemically uninformative column bleed and were not included in the analyses. Fired MG-nL was mainly composed of nitroglycerin and ethyl centralite (Figure 4.1b). Nitroglycerin and ethyl centralite were also the most abundant components of unburned MG-nL (Figure 4.1a). The abundances varied among the cartridges of fired MG-nL, but all ten cartridges shared the same profile (Table 4.1). There was not a clear pattern to the abundances based on the order fired. Based on the samples collected, cleaning the barrel between fired cartridges did not appear to impact the composition of the remaining fired cartridge residue. Table 4.1 Organic Profiles of MG-nL Unburned and Fired Cartridge Residues, as determined by GC-MS Powder Identity Unburned Unburned MG-nL Fired MG-nL Cleaned (5) MG-nL Successive (5) NG DPA MC EC DBPH 2nDPA AK II 4nDPA ✓ ✓ ✓ ✓ ✓ ✓ * Parentheses indicate number of cartridges ✓ indicates compound was detected in the sample 97 UBP_MG-nL_Cartridge1 EC NG I.S. a) 8 6 4 2 0 e c n a d n u b A d e z i l a m r o N 8 10 12 14 16 time (min) FCR_MG-nL_Successive_Cartridge1 4 b) 2 e c n a d n u b A d e z i l a m r o N 0 8 EC I.S. NG 10 12 time (min) 14 16 Figure 4.1 Representative chromatograms of MG-nL (a) unburned powder and (b) fired cartridge residue 98 The fired cartridges of Rem-L were the most variable and were categorized into six different profiles (Table 4.2). Fired cartridges of Rem-L also displayed the greatest differences in overall abundance among the ten rounds fired, as the first Rem-L cartridge fired for both the Cleaned and Successive sample sets had abundances nearly three times higher than the other fired cartridges of Rem-L (Appendix Figure A4.2). The organic compositions of the first Cleaned and Successive Rem-L fired cartridges were dominated by nitroglycerin and ethyl centralite, while the other Rem-L fired cartridges had higher proportions of diphenylamine. While diphenylamine was present in unburned Rem-L, the proportion of diphenylamine in these fired cartridges (20-75% of the total organic composition) was greater than that observed in the unburned powder (<1%). Nitroglycerin was the most abundant compound in unburned Rem-L, followed by dibutyl phthalate and Akardite II (Figure 4.2a). Despite being the most abundant compound in unburned Rem-L, five of the fired Rem-L cartridges did not contain nitroglycerin (Table 4.2). Though the composition of the Rem-L fired cartridges differed, there was not a consistent difference between Cleaned and Successive fired cartridges. As expected, the majority of compounds detected in the fired cartridge residues were present at lower abundances than in the unburned powders. Two exceptions were the increased abundances of ethyl centralite and diphenylamine present in the fired cartridge residue of Rem-L (Figure 4.2). The increased abundance of ethyl centralite was most pronounced in the first Cleaned and Successive fired cartridges of Rem-L, whereas the remaining Rem-L fired cartridges had decreased abundance of ethyl centralite when compared to the unburned powder. This trend was not observed for the other fired cartridges, which suggests that the higher abundance of ethyl centralite in some of the Rem-L fired cartridges could partially be attributed 99 to inherent homogeneity between cartridges in the same box of ammunition, as discussed in Chapter 3. Another notable difference is the increase diphenylamine abundance observed in the Rem-L fired cartridge residues. Unlike the abundances of ethyl centralite, all fired cartridge residues had higher abundances of diphenylamine than the unburned Rem-L. Table 4.2 Organic Profiles of Rem-L Unburned and Fired Cartridge Residues, as determined by GC-MS Powder Identity Unburned Unburned Rem-L NG DPA MC EC DBPH 2nDPA AK II 4nDPA ✓ ✓ ✓ ✓ ✓ ✓ Rem-L Cleaned (1) Rem-L Successive (3) ✓ ✓ Fired Rem-L Cleaned (1) Rem-L Cleaned (1) Rem-L Cleaned (1) Rem-L Cleaned (1) Rem-L Successive (1) Rem-L Successive (1) ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ * Parentheses indicate number of cartridges ✓ indicates compound was detected in the sample 100 UBP_Rem-L_Cartridge 4 a) NG 2 e c n a d n u b A d e z i l a m r o N 0 8 I.S. DBPH EC AKII 10 12 14 16 FCR_Rem-L_Successive_Cartridge5 time (min) 4 b) 2 e c n a d n u b A d e z i l a m r o N 0 8 DBPH I.S. NG DPA EC AKII 10 12 time (min) 14 16 Figure 4.2 Representative chromatograms of Rem-L (a) unburned powder and (b) fired cartridge residue 101 Fired cartridges of Win-L were divided between two profiles, due to differences in the presence of 4-nitrodiphenylamine (Table 4.3). 4-nitrodiphenylamine could be a product of decomposition, and thus would not necessarily be expected to be consistent among the cartridges. All ten cartridges contained nitroglycerin and dibutyl phthalate as the largest contributors to the overall composition, which was consistent with unburned Win-L (Figure 4.3). There was not a consistent difference between Cleaned and Successive fired cartridges. Table 4.3 Organic Profiles of Win-L Unburned and Fired Cartridge Residues, as determined by GC-MS Powder Identity Unburned Unburned Win-L Win-L Cleaned (3) Fired Win-L Cleaned (2) Win-L Successive (5) NG DPA MC EC DBPH 2nDPA AK II 4nDPA ✓ ✓ ✓ (1) ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ * Parentheses indicate number of cartridges ✓ indicates compound was detected in the sample 102 UBP_Win-L_Cartridge1 a) 4 NG 2 e c n a d n u b A d e z i l a m r o N 0 8 DBPH I.S. AKII EC 10 12 14 16 FCR_Win-L_Cleaned_Cartridge5 time (min) 4 b) 2 e c n a d n u b A d e z i l a m r o N 0 8 I.S. DBPH NG EC AKII 10 12 time (min) 14 16 Figure 4.3 Representative chromatograms of Win-L (a) unburned powder and (b) fired cartridge residue 103 Fired Win-nL was also divided between two profiles, again based on the presence of 4-nitrodiphenylamine (Table 4.4). Both unburned and fired Win-nL were composed mainly of nitroglycerin and dibutyl phthalate (Figure 4.4). Unlike the other three ammunitions, successively fired cartridges of Win-nL displayed a trend of generally increasing with each additional shot, with Successive cartridge 1 having the lowest overall abundance and Successive cartridge 5 having the highest overall abundance. The cleaned cartridges did not display an increase in abundance with subsequent shots, but there was a general trend that cartridges with lower abundances of nitroglycerin contain higher abundances of diphenylamine. Table 4.4 Organic Profiles of Win-nL Unburned and Fired Cartridge Residues, as determined by GC-MS Unburned Fired Powder Identity NG DPA MC EC DBPH 2nDPA AK II 4nDPA ✓ ✓ ✓ ✓ Unburned Win-nL Win-nL Successive (3) ✓ ✓ ✓ ✓ Win-nL Cleaned (2) Win-nL Successive (2) ✓ ✓ ✓ ✓ Win-nL Cleaned (3) ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ * Parentheses indicate number of cartridges ✓ indicates compound was detected in the sample 104 UBP_Win-nL_Cartridge1 6 4 2 e c n a d n u b A d e z i l a m r o N 0 8 a) NG DBPH I.S. EC AKII 10 12 14 16 time (min) FCR_Win-nL_Successive_Cartridge2 4 b) 2 e c n a d n u b A d e z i l a m r o N 0 8 DBPH I.S. NG EC AKII 10 12 time (min) 14 16 Figure 4.4 Representative chromatograms of Win-nL (a) unburned powder and (b) fired cartridge residue 105 With the exception of Win-nL, there was no clear impact of cleaning the weapon before each cartridge was fired. Cleaned and Successive cartridges each varied in abundance, though the overall composition was similar within each ammunition type. However, a larger sample set of fired cartridges would be necessary to investigate if this difference was significant. In addition to differences due to the firing process, the impact of the sample preparation and extraction process could introduce variation. Only ten cartridges of each ammunition type were fired and collected, which did not allow for extensive comparisons. Additionally, as the extent of variability of the firing process is not known, it was difficult to compare different cartridges to determine specifically extraction reproducibility of the fired cartridge residues as compared to differences due to the firing process. As there was no obvious pattern of differences between Cleaned and Successive cartridges, successively fired cartridges were deemed more forensically applicable and further analyses focused only on successively fired cartridges. 4.3 Principal Components Analysis for the Association of Fired Cartridge Residues to Unburned Powders 4.3.1 Principal Components Analysis of a Subset of Unburned Powders Principal components analysis (PCA) was first performed on the unburned powder subset using all Category 1 and 2 compounds to investigate association and differentiation of the subset (Figure 4.5). The first two principal components (PCs) accounted for 90.3% of the variance in the data set. 106 ) . % 1 1 2 ( 2 C P a) 4 0 -4 -6 1 b) ) . % 1 1 2 ( 2 C P EC 0.5 0 -0.5 -1 0 PC1 (69.2%) MG nL Rem L Win L Win nL DBPH 0.05 2,4-DNT 4nDPA -0.05 -0.05 AKII MC 2nDPA DPA 0.05 NG 0 PC1 (69.2%) 6 1 Figure 4.5 Principal components analysis (PCA) of the unburned powder subset using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 107 Unburned MG-nL was the only powder of the subset that was positioned negatively on PC1, while unburned Win-nL, Rem-L and Win-L were positioned positively on PC1 (Figure 4.5a). Examination of the loadings plot (Figure 4.5b) shows that ethyl centralite was weighted negatively on PC1 and dibutyl phthalate was weighted positively on PC1. Though the other powders in the subset also contained ethyl centralite, the abundance of ethyl centralite in unburned MG-nL was 25-55 times higher than in the other unburned powders, resulting in the negative positioning of MG-nL on PC1. Unburned Win-nL had the highest abundance of dibutyl phthalate, resulting in the most positive positioning on PC1 (Figure 4.5a). Unburned Win-nL had abundances of dibutyl phthalate that were 3-8 times more abundant than the abundances of dibutyl phthalate in unburned Win-L and Rem-L, and almost 130 times higher than the abundance of dibutyl phthalate in unburned MG-nL. Unburned MG-nL and Win-nL were both positioned positively on PC2, while unburned Rem-L and Win-L were positioned negatively on PC2 (Figure 4.5a). As previously discussed in Chapter 3, unburned Win-L and Rem-L were too chemically similar to be distinguished, due to having very similar presence and abundances of compounds. Akardite II was present in unburned Win-L, Rem-L, and Win-nL, with abundances approximately 20% higher in Win-L and Rem-L than in Win-nL. Akardite II was weighted slightly positively on PC1 and slightly negatively on PC2, contributing to the positioning of unburned Win-L and Rem-L (Figure 4.5b). 108 4.3.2 Principal Components Analysis of Fired Cartridge Residues Principal components analysis (PCA) was then performed on the successive fired cartridge residues using all Category 1 and 2 compounds (Figure 4.6). The first two principal components (PCs) accounted for 93.7% of the variance in the fired cartridge residue data set. This is comparable to the 90.3% of the variance described by the first two PCs of the unburned powder subset (Figure 4.5). 109 ) . % 7 3 3 ( 2 C P 3 2.5 2 1.5 1 0.5 0 -0.5 -1 -1.5 a) -2 0 PC1 (60.0%) 2 4 Fired MG nL Fired Rem L Fired Win L Fired Win nL ) . % 7 3 3 ( 2 C P 1 b) EC 0.5 0 -0.5 -0.5 NG DBPH 4nDPA 2nDPA AKII MC DPA 0 PC1 (60.0%) 0.5 1 Figure 4.6 Principal components analysis (PCA) of fired cartridge residues using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 110 The overall positioning of the fired cartridge residues was similar to the positioning of the unburned powders seen in Figure 4.5, indicating that despite the chemical changes that occur during the firing process, important chemical differences were maintained. Fired MG-nL was positioned negatively on PC1 and positively on PC2 due to the presence of ethyl centralite at higher abundances than observed in the other fired cartridge residues. The spread in positioning of fired MG-nL along PC2 was due to differences in abundance of nitroglycerin and ethyl centralite. Fired Win-nL was positioned positively on PC1 and PC2 due to higher abundances of dibutyl phthalate (Figure 4.6). The spread of fired Win-nL along PC1 and PC2 was due to differences in abundance of nitroglycerin and dibutyl phthalate (Figure 4.6). The first two cartridges of fired Win-nL were positioned negatively on PC2 due to lower abundances of nitroglycerin, which was weighted positively on both PC1 and PC2 (Figure 4.6b). Fired Rem-L and Win-L were positioned closely together (Figure 4.6), which was consistent with the positioning of unburned Rem-L and Win-L (Figure 4.5). However, one cartridge of fired Rem-L was positioned negatively on PC1 and positively with PC2; this cartridge was the first Rem-L cartridge fired and had higher abundances of ethyl centralite. The increased abundance resulted in the cartridge of fired Rem-L having abundances of ethyl centralite more similar to fired MG-nL than the other Rem-L fired cartridges. Principal components analysis (PCA) was then performed on the combined unburned and fired cartridge residue data set to investigate association of fired cartridges to the corresponding unburned powders (Appendix Figure A4.3). However, the fired cartridge residues were too dissimilar from the unburned powders for successful association. Instead, all the fired cartridge residues were positioned together, separately from the unburned powders, likely because of the decreased abundances of the majority of compounds in the fired cartridge residues. 111 A more forensically relevant approach would be to investigate the association of a single fired cartridge residue to a larger collection of unburned powders. Thus, PCA was performed on the unburned data set, with one fired cartridge residue included at a time. The unburned powders dominate the variance described by PCA, due to having higher abundances. Changing the fired cartridge residue included did not dramatically impact the positioning of the unburned powders. Additionally, changes in the loadings plots following the introduction of different fired cartridge residues were negligible. Thus, the following scores plots (Figures 4.7-4.9) can be interpreted using the loadings plots shown in Figures 4.5. The loadings plots for Figures 4.7-4.9 can be seen in Appendix Figures A4.4-8. Though the composition of the fired residues were similar to the unburned powder counterparts, the decreased abundances observed in the fired cartridge residues limited association of fired cartridges to the correct unburned powder. Unburned MG-nL was positioned negatively on PC1 and positively on PC2. When introduced individually, the fired cartridges of MG-nL that had the highest abundances were positioned negatively on PC1 (Figure 4.7a). The lower abundances of the second fired MG-nL cartridge resulted in positive positioning on PC1 (Figure 4.7b). 112 ) . % 9 1 2 ( 2 C P a) 6 3 0 -3 -6 -6 b) 6 3 0 -3 ) . % 8 2 2 ( 2 C P -6 -6 -3 -2 -1 -5 -4 Unburned MG nL Unburned Win L FCRMGnLSuccessiveCartridge4 PC1 (64.6%) -4 -5 -3 -2 Unburned MG nL Unburned Win L FCRMGnLSuccessiveCartridge2 -1 PC1 (63.0%) 0 1 2 3 4 Unburned Rem L Unburned Win nL 0 1 2 3 4 Unburned Rem L Unburned Win nL Figure 4.7 Principal components analysis (PCA) scores plots of the unburned powder subset and MG-nL fired cartridge residue (a) fired MG-nL cartridge 4 and (b) fired MG-nL cartridge 2 113 Unburned Win-nL was positioned positively on both PC1 and PC2. The fired Win-nL cartridges were all positioned positively on PC1 but negatively on PC2, resulting in positioning close to the unburned Rem-L and Win-L cartridges (Figure 4.8). The positioning can be attributed to the abundances of dibutyl phthalate in the fired Win-nL being more similar to the abundances of dibutyl phthalate found in unburned Rem-L and Win-L, whereas the unburned Win-nL has abundances of dibutyl phthalate that were more than three times higher than what was observed in the fired Win-nL. 6 3 0 -3 -6 -6 ) . % 5 0 2 ( 2 C P -4 -3 -5 -2 Unburned MG nL Unburned Win L FCRWinnLSuccessiveCartridge3 -1 PC1 (65.5%) 0 1 2 3 4 Unburned Rem L Unburned Win nL Figure 4.8 Principal components analysis (PCA) scores plots of unburned powder subset and one Win-nL fired cartridge residue 114 With the exception of the first fired cartridge of Rem-L, all fired Rem-L and Win-L were positioned positively on PC1 and more negatively on PC2 than the unburned Rem-L and Win-L (Figure 4.9a). When PCA was performed using the unburned subset and the first fired cartridge of Rem-L (Figure 4.9b), the fired cartridge was positioned negatively on PC1 instead of positively. This was consistent with the positioning seen on Figure 4.6, indicating that the abundance of ethyl centralite is contributing heavily to the positioning of the cartridges positioned negatively on PC1. 115 ) . % 1 1 2 ( 2 C P ) . % 1 1 2 ( 2 C P a) 6 0 -6 -6 b) 6 0 -6 -6 -4 -3 -2 -5 Unburned MG nL Unburned Win L FCRWinLSuccessiveCartridge1 -1 PC1 (69.2%) -4 -3 -2 -5 Unburned MG nL Unburned Win L FCRRemLSuccessiveCartridge1 -1 PC1 (69.2%) 2 1 0 Unburned Rem L Unburned Win nL 2 1 0 Unburned Rem L Unburned Win nL 3 4 3 4 Figure 4.9 Principal components analysis (PCA) scores plot of the unburned powder subset and one fired cartridge residue using Category 1 and 2 compounds (a) fired Win-L and (b) the first fired Rem-L residue 116 4.3.3 Comparison of PCA of Smokeless Powders using GC-MS to PCA using LC-TOFMS data Separately collected fired cartridge residues of the same 18 powders were also analyzed by LC-TOFMS.3,4 Principal components analysis was performed on the 18 unburned powders, and the scores of five fired cartridges of one type of ammunition were manually calculated to project the fired cartridge residues on the scores plot of the unburned powders.4 As the fired cartridge residues were projected onto the scores plot, the positioning of the unburned powders and the loadings plots for each comparison did not change. The resulting scores plots with the projected fired cartridge residues were used to investigate association of the fired cartridge residues to the corresponding unburned powder. Fired cartridge residues of MG-nL were positioned similarly to the unburned MG-nL on the scores plot.4 However, the fired cartridge residues were not exclusively positioned closely to unburned MG-nL, but rather were positioned with other unburned powders that contained high abundances of ethyl centralite. As previously discussed, four powders in the sample set contained abundances of ethyl centralite that were at least three times larger than in the remaining powders. The positioning of MG-nL in relation to the cartridges with high abundances of ethyl centralite further supports the importance of association based on dominant chemical compounds rather than attempting to associate to a specific brand of ammunition. The similarity of fired MG-nL to unburned MG-nL by LC-TOFMS was consistent with the association observed using GC-MS data (Figure 4.7a). Fired cartridge residues of Rem-nL, Win-L, and Win-nL were all positioned similarly on the scores plot.4 The similarity of fired Rem-L and Win-L to the unburned Rem-L and Win-L was consistent with the association observed using GC-MS data (Figure 4.9). However, examination of the positioning of other fired cartridge residues indicated that many of the other 117 fired residues were positioned in a similar location when PCA was performed using the LC-TOFMS data.4 The similar positioning of many of the fired cartridges may suggest that this positioning has more to do with the lower abundances of all compounds present in the fired cartridge residues than the presence or absence of specific compounds. This is also supported by the observation that the fired cartridges with higher abundances had greater association success overall. A similar result was seen in the positioning of the fired Win-nL on the PCA scores plot based on GC-MS data (Figure 4.8), where the fired cartridges of Win-nL were positioned negatively on PC2 due to having lower abundances than the corresponding unburned powder. Overall, the association achieved using GC-MS data was comparable to using LC-TOFMS data. To further investigate the association of fired cartridges to the corresponding unburned powders, hierarchical cluster analysis (HCA) was performed, using the same data sets that were used for PCA. 4.4 Hierarchical Cluster Analysis for the Association of Fired Cartridge Residues to Unburned Powders 4.4.1 Hierarchical Cluster Analysis of a Subset of Unburned Powders The unburned powder subset was then subjected to HCA (Figure 4.10). Hierarchical cluster analysis resulted in similar groupings as PCA but has the added advantage of providing a metric to quantify the extent of similarity among groups. All five cartridges of unburned MG-nL clustered together at a similarity level of 81% and did not cluster to the remaining powders before a similarity level of 0%, indicating no similarity to the other powders in the subset. This was due to the much higher abundances of ethyl centralite present in unburned MG-nL compared 118 to the remaining unburned powders, as well as a lower abundance of dibutyl phthalate in unburned MG-nL (Figure 4.1). 0 50 100 ) % ( l e v e L y t i r a l i m S i 30% L n - n i W 70% L - n i W L - m e R 62% L n - G M 81% Figure 4.10 Hierarchical cluster analysis (HCA) dendrogram of unburned powder subset, based on GC-MS data The remaining three unburned powders clustered at a similarity level of 30% (Figure 4.10). The five cartridges of unburned Win-nL clustered at a similarity level of 70%, with two cartridges of Win-nL being the most similar powders in the subset, clustering at a similarity level of 97%. As was previously observed on the PCA scores plot of the unburned powder subset (Figure 4.10), unburned Win-L and Rem-L were too chemically similar to be distinguished, and clustered at a similarity level of 62%. 119 4.4.2 Hierarchical Cluster Analysis of Fired Cartridge Residues Hierarchical cluster analysis was performed on the unburned powder subset with the fired cartridge residues introduced. The resulting dendrogram was used to visualize clusters based on compositional similarity. As the fired cartridges had undergone chemical changes as a result of the firing process, similarity levels of the fired cartridges were expected to be lower than those observed within the unburned powders alone. Thus, the clustering order was deemed most significant when examining the relationship between fired cartridges and the unburned powders. When the subset of unburned powders and all successive fired cartridge residues were subjected to HCA, the majority of the fired cartridge residues clustered together before clustering to an unburned powder (Figure 4.11). This indicated that a majority of the fired cartridges were more similar to each other than to an unburned powder, which can be explained by the lower abundances that are shared by the fired cartridge residues. The unburned cartridges of MG-nL were the most dissimilar from the rest of the unburned and fired cartridges (Figure 4.11), which was also observed following PCA of the unburned powders in the subset (Figure 4.5). 120 18% 40% L - n i W L - m e R ) % ( l e v e L y t i r a l i m S i 0 50 100 L n - n i W L n - n i W : R C F L - n i W L - m e R L n - G M : R C F 60% L n - G M Figure 4.11 Hierarchical cluster analysis (HCA) dendrogram of the unburned powder subset with all fired cartridge residues (indicated by black dots) introduced One of the five Win-nL fired cartridges clustered first to a cartridge of unburned Win-nL, at a similarity level of 95%. This fired Win-nL cartridge had higher abundances of all compounds than the other four fired cartridges of Win-nL, resulting in higher similarity to the unburned Win-nL. The remaining four cartridges of fired Win-nL clustered first to unburned Win-L and Rem-L, and did not cluster to unburned Win-nL until a similarity level of 18%. All fired cartridge residues of Win-L, Rem-L and MG-nL clustered together at a similarity level of 60% (Figure 4.11). This is consistent with the PCA scores plot of the unburned subset and fired cartridge residues (Appendix Figure A4.2). Of the fired cartridges, the fired Win-L were the most similar to each other, clustering at a similarity level of 93% to other before clustering to fired Rem-L at a similarity level of 92%. This was consistent with PCA performed on the fired cartridge residues alone, where fired cartridges of Win-L were positioned tightly together on the scores plot (Figure 4.6). 121 At a similarity level of 40%, the cluster of fired cartridge residues clustered to the group containing unburned Win-L, unburned Rem-L, and the remaining four cartridges of fired Win-nL. The clustering of fired Win-nL to unburned Rem-L and Win-L was consistent with the PCA scores plot of the unburned subset with one fired Win-nL cartridge introduced (Figure 4.8), where cartridges of fired Win-nL were positioned closely to the unburned Win-L and Rem-L. Introducing all the fired cartridges to the unburned powder subset only allowed for association of one cartridge of fired Win-nL to the corresponding unburned Win-nL, while the majority of the remaining fired cartridges were more chemically similar to each other than to any of the unburned powders. Additionally, this approach makes less sense for a forensic application. If a fired cartridge had been recovered from a crime scene, that cartridge would be extracted and investigators would be interested in how that single residue compares to either unburned powder from an unfired cartridge recovered from a suspect, or perhaps other fired cartridges collected at a different scene or from a suspect’s possession. It thus makes the most sense to compare the single fired cartridge residue to either a library of unburned powders, or even perhaps other fired cartridges. To investigate the association of a single fired cartridge residue, each fired cartridge residue was introduced individually to the unburned powder subset and HCA was performed. Exemplar dendrograms are shown in Figures 12-14. Despite the addition of a fired cartridge residue, the clustering of the unburned powders was very similar to what was previously shown in Figure 4.10. With the one exception, all fired MG-nL residues successfully clustered first to the unburned MG-nL, at similarity levels ranging from 8-32% (Figure 12a). The association of fired MG-nL to unburned MG-nL was attributed to the high abundances of ethyl centralite present in 122 both unburned and fired MG-nL. The remaining cartridge of MG-nL clustered first to unburned Win-L and Rem-L, at a similarity level of 33% (Figure 4.12b). 123 14% L n - G M 33% 0 50 100 0 50 100 L - n i W L - m e R a) b) L n - n i W L - n i W L - m e R L n - n i W L n - G M Figure 4.12 Hierarchical cluster analysis (HCA) dendrograms of the unburned powder subset with fired MG-nL (a) cartridge 1 and (b) cartridge 4 introduced (indicated by black dot) All cartridges of fired Win-nL were most chemically similar to the unburned Win-L and ) % ( l e v e L y t i r a l i m S i ) % ( l e v e L y t i r a l i m S i Rem-L, clustering first to these unburned powders instead of unburned Win-nL. An exemplar dendrogram is shown in Figure 4.13. The lower abundances of the fired cartridges, along with 124 the lower abundance of dibutyl phthalate, decreased the similarity to the unburned Win-nL. The similarity of the fired Win-nL to the unburned Rem-L and Win-L was also observed by PCA, where the fired cartridges were positioned closely to the unburned Rem-L and Win-L rather than the unburned Win-nL. L - n i W L - m e R L n - n i W L n - G M Figure 4.13 Exemplar HCA dendrogram of the unburned powder subset with one fired Win-nL cartridge introduced (indicated by black dot) With the exception of two Rem-L fired cartridges, all cartridges of fired Win-L and Rem- L clustered first to the larger cluster containing unburned Win-nL, Win-L and Rem-L. The fired cartridges each clustered to the group of unburned powders at a similarity level, ranging from 5-7% (Figure 4.14a). The fired cartridges clustered first to the group of unburned Win-L, Rem- L, and Win-nL because they shared a higher degree of similarity with these powders than with the unburned MG-nL, which contained much higher abundances of all compounds, as well as 125 drastically higher abundances of ethyl centralite. The similarity observed was consistent with the positioning of the Win-L and Rem-L fired cartridges on the PCA scores plots (Figure 4.9). However, one cartridge of fired Rem-L clustered first to unburned MG-nL, at a similarity level of approximately 10% (Figure 4.14b). This was also consistent with the PCA scores plot (Figure 9b) and was due to the first fired cartridge of Rem-L having higher abundances of all compounds than the other cartridges of fired Rem-L and Win-L. As previously discussed, the first fired cartridge also had higher abundances of ethyl centralite, which was most abundant in unburned MG-nL. 126 L - n i W L - m e R 0 50 100 0 50 100 L n - n i W L n - G M L n - n i W L n - G M a) b) L - n i W L - m e R Figure 4.14 Exemplar HCA dendrograms of the unburned powder subset with (a) one fired Win-L cartridge and (b) Rem-L cartridge 1 introduced (indicated by black dot) 127 ) % ( l e v e L y t i r a l i m S i ) % ( l e v e L y t i r a l i m S i 4.4.3 Comparison of HCA performed using GC-MS data to HCA performed using LC-TOFMS data The abundances of the unburned powders and fired cartridge residues following LC-TOFMS analysis were subjected to HCA.4 While all GC-MS comparisons were restricted to the unburned powder subset, LC-TOFMS comparisons included all 18 unburned powders. A different fired cartridge residue was added to the unburned powders each time HCA was performed. The success of association by HCA was again determined by the clustering order of the introduced fired cartridge residue. When HCA was performed on the LC-TOFMS data with one cartridge of fired MG-nL included, fired MG-nL clustered first unburned SB-N, and then clustered to a larger group containing unburned MG-nL and BZR-nL.4 The introduction of fired Win-L, Rem-L and Win-nL resulted in similar trends; while some fired cartridges successfully clustered first to the corresponding unburned powder, many fired cartridges clustered to a larger group containing the corresponding unburned powder. This was also observed on the HCA dendrograms performed using GC-MS data. 128 4.5 Summary Of the 18 unburned powders, four powders were selected for fired cartridge residue analysis. Two powders with leaded primers (Rem-L and Win-L) were selected, as well as two powders with non-leaded primers (MG-nL and Win-nL). Prior to fired cartridge analysis, the normalized abundances of the unburned powders were subjected to principal components analysis (PCA) and hierarchical cluster analysis (HCA). Unburned MG-nL was distinguished from the other powders in the subset by both PCA and HCA. Unburned Rem-L and Win-L were too chemically similar to be distinguished, resulting in similar positioning on the PCA scores plot and clustering within the same group on the HCA dendrogram. Unburned Win-nL was chemically similar to Rem-L and Win-L, but contained higher abundances of dibutyl phthalate that allowed for differentiation from Rem-L and Win-L. For each ammunition type in the subset, ten cartridges were fired; five rounds where the barrel of the weapon was cleaned before each round was fired (“Cleaned”), and five where the barrel of the weapon was not cleaned before each round (“Successive”). Consistent differences between Cleaned and Successive fired cartridges were not observed, and thus further analyses were performed using only the Successive fired cartridge residues. The composition of the fired cartridge residues was largely consistent with the original compositions of the corresponding unburned powders. The compounds that were most abundant in the unburned powder remained the most abundant compounds in the fired cartridge residues. For example, both unburned and fired MG-nL contained ethyl centralite as the most abundant compound. Due to inconsistencies in the firing process, the fired cartridge residues had higher RSDs than the corresponding unburned powders, but the overall composition was consistent. The 129 high abundances of ethyl centralite and compositional similarity of fired MG-nL to unburned MG-nL resulted in successful association when PCA and HCA were performed. However, when compared to the unburned powders, the fired cartridge residues exhibited lower abundances of all compounds, which limited the success of association. For example, unburned Win-nL contained a high abundance of dibutyl phthalate. While the fired cartridge residues of Win-nL also contained dibutyl phthalate as the most dominant contributor to the total organic composition, PCA and HCA of the fired cartridge residues resulted in association first to unburned Rem-L and Win-L rather than to Win-nL. This can be attributed to the lower abundances present in the unburned powders of Rem-L and Win-L when compared to MG-nL and Win-nL. Overall, it has been demonstrated that both GC-MS and LC-TOFMS analysis result in limited association success, where success was highly dependent on the composition of the unburned powder. Ammunition that contained high abundances of ethyl centralite or dibutyl phthalate (MG-nL and Win-nL, respectively) resulted in fired cartridges that were chemically similar enough to result in successful association. Meanwhile, fired cartridges that contained lower abundances of ethyl centralite or dibutyl phthalate resulted in limited association success. 130 APPENDIX 131 t c a r t x E r e d w o P d e n r u b n U UBP Win-nL C5 UBP Win-nL C4 UBP Win-nL C3 UBP Win-nL C2 UBP Win-nL C1 UBP Win-L C5 UBP Win-L C4 UBP Win-L C3 UBP Win-L C2 UBP Win-L C1 UBP Rem-L C5 UBP Rem-L C4 UBP Rem-L C3 UBP Rem-L C2 UBP Rem-L C1 UBP MG-nL C5 UBP MG-nL C4 UBP MG-nL C3 UBP MG-nL C2 UBP MG-nL C1 0 5 APPENDIX 10 15 Total Normalized Abundance NG DPA MC EC DBPH 2nDPA AKII 4nDPA 20 25 Figure A4.1 Total organic composition of the powders included in the subset of unburned powders, as determined by GC-MS 132 t c a r t x E e u d i s e R e g d i r t r a C d e r i F FCR Win-nL Successive C5 FCR Win-nL Successive C4 FCR Win-nL Successive C3 FCR Win-nL Successive C2 FCR Win-nL Successive C1 FCR Win-nL Cleaned C5 FCR Win-nL Cleaned C4 FCR Win-nL Cleaned C3 FCR Win-nL Cleaned C2 FCR Win-nL Cleaned C1 FCR Win-L Successive C5 FCR Win-L Successive C4 FCR Win-L Successive C3 FCR Win-L Successive C2 FCR Win-L Successive C1 FCR Win-L Cleaned C5 FCR Win-L Cleaned C4 FCR Win-L Cleaned C3 FCR Win-L Cleaned C2 FCR Win-L Cleaned C1 FCR Rem-L Successive C5 FCR Rem-L Successive C4 FCR Rem-L Successive C3 FCR Rem-L Successive C2 FCR Rem-L Successive C1 FCR Rem-L Cleaned C5 FCR Rem-L Cleaned C4 FCR Rem-L Cleaned C3 FCR Rem-L Cleaned C2 FCR Rem-L Cleaned C1 FCR MG-nL Successive C5 FCR MG-nL Successive C4 FCR MG-nL Successive C3 FCR MG-nL Successive C2 FCR MG-nL Successive C1 FCR MG-nL Cleaned C5 FCR MG-nL Cleaned C4 FCR MG-nL Cleaned C3 FCR MG-nL Cleaned C2 FCR MG-nL Cleaned C1 NG DPA MC EC DBPH 2nDPA AKII 4nDPA 0 2 4 Total Normalized Abundance 6 Figure A4.2 Total organic composition of all fired cartridge residues, as determined by GC-MS 133 ) . % 3 9 3 ( 2 C P a) 6 3 0 -3 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 Unburned MG nL Unburned Rem L Fired MG-nL Fired Rem-L Unburned Win L Fired Win L PC1 (48.4%) Unburned Win nL Fired Win nL ) . % 3 9 3 ( 2 C P 1 b) 0.5 EC 0 -0.5 -1 NG DBPH AKII 4nDPA MC 2nDPA DPA -0.5 PC1 (48.4%) 0 0.5 134 Figure A4.3 Principal components analysis of the unburned powder subset and successive fired cartridge residues using Category 1 and 2 compounds (a) scores plot of PC1 vs PC2 and (b) loadings plot of PC1 vs PC2 1.0 EC DBPH NG 4nDPA AKII MC 2nDPA DPA 0.0 PC1 (64.6%) 1.0 Figure A4.4 Loadings plots for PCA of fired cartridge residue MG-nL C4 ) . % 9 1 2 ( 2 C P 0.0 -1.0 -1.0 135 ) . % 8 2 2 ( 2 C P 1 0 -1 -1 EC NG DBPH 4nDPA 2nDPA MC AKII DPA 0 PC1 (63.0%) 1 Figure A4.5 Loadings plots for PCA of fired cartridge residue MG-nL C2 ) . % 5 0 2 ( 2 C P 1.0 0.0 EC -1.0 -1.0 DBPH NG MC 4nDPA AKII 2nDPA DPA 0.0 PC1 (65.5%) 1.0 Figure A4.6 Loadings plots for PCA of fired cartridge residue Win-nL C3 136 DBPH NG AKII MC 4nDPA 2nDPA DPA -0.01 0.00 0.01 0.02 0.03 0.04 0.05 0 PC1 (69.2%) Figure A4.7 Loadings plots for PCA of fired cartridge residue Win-L C1 1 1 EC 0.05 0.00 -0.05 EC 0.05 0.00 -0.05 ) . % 1 1 2 ( 2 C P ) . % 1 1 2 ( 2 C P 1 0 -1 -1 1 0 -1 -1 DBPH NG AKII MC 4nDPA 2nDPA DPA -0.01 0.00 0.01 0.02 0.03 0.04 0.05 0 PC1 (69.2%) Figure A4.8 Loadings plots for PCA of fired cartridge residue Rem-L C1 137 REFERENCES 138 REFERENCES (1) MacCrehan, W. A.; Patierno, E. R.; Duewer, D. L.; Reardon, M. R. Investigating the Effect of Changing Ammunition on the Composition of Organic Additives in Gunshot Residue (OGSR). J. Forensic Sci. 2001, 46 (1), 14911J. https://doi.org/10.1520/JFS14911J. (2) Goudsmits, E.; Blakey, L. S.; Chana, K.; Sharples, G. P.; Birkett, J. W. The Analysis of Organic and Inorganic Gunshot Residue from a Single Sample. Forensic Science International 2019, 299, 168–173. https://doi.org/10.1016/j.forsciint.2019.03.049. (3) Reese, K. L.; Jones, A. D.; Waddell Smith, R. Characterization of Smokeless Powders Using Multiplexed Collision-Induced Dissociation Mass Spectrometry and Chemometric Procedures. Forensic Science International 2017, 272, 16–27. (4) Reese, K. L. Association and Differentiation of Smokeless Powders Utilizing Non-Targeted Mass Spectrometry and Multivariate Statistical Analysis, Michigan State University, 2016. 139 5. Conclusions and Future Work 5.1 Conclusions The overall objective of this work was to investigate the association of fired cartridge residues to the corresponding unburned powders based on GC-MS data. Following GC-MS analysis of 18 different unburned powders, the resultant chromatographic abundances were subjected to principal components analysis (PCA) and successful differentiation of the unburned powders based on differences in organic composition was demonstrated. Hierarchical cluster analysis (HCA) resulted in similar groupings as PCA. Comparisons using all nine compounds detected in this work resulted in increased differentiation when compared to using only the compounds proposed by Goudsmits et al.1 Fired cartridge residues collected without barrel cleaning (“Successive”) were compared to those collected when the barrel had been cleaned between each round (“Cleaned”), and the variation observed in the Successive cartridges was comparable to that of the Cleaned cartridges. Based on these data, barrel cleaning did not have an impact on the composition of the fired cartridge residue. While the firing process resulted in decreased abundances of all compounds, the overall compositions of the fired cartridge residues were comparable to those of the corresponding unburned powders. Despite the chemical changes that occur during the firing process, the composition of the fired cartridges was consistent with the composition of the unburned powders, with no additional pyrolysis products or decomposition products detected. Following PCA and HCA, successful association was observed for fired MG-nL, which was the only powder in the subset that contained ethyl centralite in high abundance. Unburned Rem-L and Win-L were chemically indistinguishable, and the resultant fired cartridge residues were positioned in close proximity to Rem-L and Win-L unburned powders on their respective 140 PCA scores plots. The majority of the Rem-L and Win-L fired cartridge residues also clustered first to the group containing unburned Rem-L and Win-L on the HCA dendrogram. Finally, association of fired Win-nL was limited because the fired Win-nL residues had lower abundances of dibutyl phthalate and nitroglycerin than what was detected in unburned Win-nL. This resulted in the fired cartridges being more closely associated to unburned Rem-L and Win-L, which had a similar profile to unburned Win-nL but contained lower abundances of all compounds. If a fired cartridge was recovered from a crime scene, the residue remaining in the cartridge could be solvent extracted and the organic components analyzed by GC-MS. Based on this work, the organic profile of the residue could be compared to organic profiles of unburned powders (either analyzed in-house or retrieved from the NCFS Smokeless Powders Database), using PCA and HCA to investigate association.2 The next step in this research is to investigate organic profiles of gunshot residue, applying similar methods to examine association of GSR to the original unburned powder, followed by the association of GSR to fired cartridge residue. This would allow practitioners to compare fired cartridge residue recovered from a crime scene to GSR connected to a potential suspect, potentially linking the suspect to the crime scene. 5.2 Future Work This work demonstrated the use of GC-MS chromatographic data for the association and differentiation of fired cartridge residues and unburned powders. However, association was limited by the original composition of the powder. This limitation was also observed following LC-TOFMS analysis and subsequent multivariate statistical procedures.3 Additionally, ethyl centralite and dibutyl phthalate were responsible for most of the differentiation observed by 141 PCA. To further investigate the ability to associate a fired cartridge to the corresponding unburned powder, a larger study with additional smokeless powder samples should be carried out to incorporates additional compositional differences, such as including other single-based powders or powders that were known to be recently manufactured. All powders in this study showed evidence of diphenylamine nitration, which occurs as powders age. Variability among the five fired rounds of each ammunition had a limited impact on association success, namely with the association of the first fired cartridge of Rem-L. The first fired cartridge of Rem-L was more chemically similar to MG-O than to the remaining fired Rem-L due to a higher abundance of ethyl centralite detected. This was illustrated by the corresponding PCA scores plot and HCA dendrogram. However, further work needs to be done to examine the extent of variability caused by the firing process. For the work described here, commercially available rounds of ammunition were fired and analyzed. However, it was not possible to analyze the specific unburned powder that was then analyzed as fired cartridge residue. Consequently, it is possible that the differences detected in the fired cartridges were due to differences in composition of the unburned powder. To address this concern, future work could involve buying a cannister of reloading powder and self-loading the rounds of ammunition. This would allow for the correction of the final masses based on known mass of the unburned powder added. Using a cannister of reloader powder would also allow for more samplings to determine heterogeneity of the bulk powder. An interesting discovery in this work was the detection of two peaks attributed to nitroglycerin when the powders were analyzed by GC-MS. Subsequent analyses on separate GC-MS instrumentation indicated that one of the peaks was an artifact of the initial instrument rather than the samples. The additional peak was attributed to contamination of the gold seal at 142 the base of the GC inlet, which can happen over time if the seal is not changed regularly. To the best of this author’s knowledge, this effect has not previously been reported in the analysis of smokeless powders. Future work should investigate if this has been encountered by forensic laboratories, which also have instrumentation that is heavily used and may be subject to contamination of the gold seal. Additionally, other multivariate statistical procedures should continue to be investigated. Gallidabino et al. recently developed a new modelling approach known as quantitative profile-profile relationship (QPPR) modelling to predict the chemical composition of unburned powders based on the composition of the corresponding fired cartridge residue.4 Both unburned smokeless powders and the fired cartridge residues were analyzed using GC-MS. Many compositional inconsistencies were observed when the fired cartridge residues were compared to the corresponding unburned powders.4 A certain degree of difference was expected, due to chemical changes that occur during the firing process. Additionally, the unburned powder and fired cartridge residue samples were not subjected to the same sample preparation methods – the unburned powder was solvent extracted, while the fired cartridge residues were sampled using headspace sorptive extraction.4 Due to these differences, comparisons based on chemical composition alone were difficult. Thus, Gallidabino et al. investigated the application of individual multivariate statistical techniques and noted that the performance of individual methods often was variable among different applications.4 The QPPR model takes into account a combination of many different machine learning techniques, and successfully predicted the chemical composition of unburned powders based on the respective fired cartridge residue composition. 143 This novel modelling approach gave highly promising results but requires the analyst to have an understanding of at least eight different multivariate statistical techniques, as well as different modelling techniques. The development of QPPR modelling is an example of the continued application of multivariate statistical analyses to the study of smokeless powders, rather than considering compositional analysis alone. Using this method, the composition of an unburned powder could be predicted from a fired cartridge collected at a crime scene. The predicted unburned powder composition could then be compared to unburned powder obtained from a suspect or firearm. 144 REFERENCES 145 REFERENCES (1) Goudsmits, E.; Sharples, G. P.; Birkett, J. W. Preliminary Classification of Characteristic Organic Gunshot Residue Compounds. Science & Justice 2016, 56 (6), 421–425. https://doi.org/10.1016/j.scijus.2016.06.007. (2) Smokeless Powder Database. National Center for Forensic Science, University of Central Florida http://www.ilrc.ucf.edu/powders/. (3) Reese, K. L.; Jones, A. D.; Waddell Smith, R. Characterization of Smokeless Powders Using Multiplexed Collision-Induced Dissociation Mass Spectrometry and Chemometric Procedures. Forensic Science International 2017, 272, 16–27. (4) Gallidabino, M. D.; Barron, L. P.; Weyermann, C.; Romolo, F. S. Quantitative Profile– Profile Relationship (QPPR) Modelling: A Novel Machine Learning Approach to Predict and Associate Chemical Characteristics of Unspent Ammunition from Gunshot Residue (GSR). Analyst 2019, 144 (4), 1128–1139. https://doi.org/10.1039/C8AN01841C. 146