Advances in Fire Debris Analysis

The practice of forensic fire debris analysis and data interpretation in operational (i.e., casework) laboratories has gone without significant change since the widespread adoption of gas chromatography–mass spectrometry (GC-MS) and a consensus standard for interpretation from the American Society for Testing and Materials (ASTM). The industry standard for fire debris analysis is ASTM E1618, “Standard Test Method for Ignitable Liquid Residues in Extracts from Fire Debris Samples by Gas Chromatography-Mass Spectrometry” (GC-MS) [1]. The first version of the ASTM E1618 was approved in 1997. Since then, there has been some minimal reshuffling of the ignitable liquid classes, with a few class additions and deletions, but no significant changes otherwise. The ASTM standard defines classes of ignitable liquids based on several criteria, including product, process, chemistry, etc. Only one ASTM E1618 ignitable liquid class is currently product-based (gasoline), while one is process-based (petroleum distillates), four are chemistry-based (normal alkane, isoparaffinic, naphthenic paraffinic and aromatic products), and one exhibits no class characteristics or characteristics of multiple classes (others-miscellaneous). An additional class, the oxygenated solvents class, may possess the characteristics of any of the other classes, or multiple classes, so long as the solvent contains a major contribution from an oxygenated compound (alcohol, ketone, ether, ester, etc.). However, gasoline samples containing a significant ethanol component (gasohol) are classified as gasoline, not oxygenated solvents. The classification scheme can be confusing, yet the class of a liquid may be helpful to investigators wishing to identify commercial products. Under the ASTM E1618, target compounds are only designated for the gasoline class, as well as the medium and heavy subclasses of petroleum distillates class (i.e., those in the C8–C20+ range). The standard provides for the use of “target compound chromatograms” [2]; however, these are rarely used by analysts. The ASTM standard directs the analyst to visually recognize the presence of class-associated chromatographic patterns in fire debris GC-MS total ion chromatograms and extracted ion profiles as a means of determining the presence of ignitable liquid residues. Complicating this procedure is the persistent and highly diverse background arising from the pyrolysis (thermal breakdown) of building materials and furnishings, the weathering of ignitable liquids (evaporation) and potential biological degradation of ignitable liquid residues in the sample. The laboratory weathering of ignitable liquids for comparison purposes is time and labor intensive, and difficult to accurately reproduce. The forensic analyst must work through a complicated data set and arrive at a categorical decision and statement (report) regarding the presence or absence of an ignitable liquid. Decisions and statements are softened only by qualifying statements that can be biasing (i.e., “[T]the absence of an ignitable liquid residue does not preclude the possibility that ignitable liquids were present at the fire scene. Ignitable liquids are volatile compounds that may have evaporated, been totally consumed in a fire, environmentally altered or removed, or otherwise indistinguishable from background materials” [1]). Current methods in forensic fire debris data interpretation are without validation,


LIST OF FIGURES
Direct property loss due to fires was estimated at $10.7 billion. An estimated 31,500 intentionally set (arson) structure fires resulted in 315 deaths and resulted in $664 million dollars in property damage. 1 Fire investigations are a challenge since the evidence is partially if not totally destroyed during the event. At the scene, fire investigators establish the point of fire origination to begin the process of determining a cause. Fire debris from the point of origin is collected as evidence then analyzed in the laboratory to detect whether an ignitable liquid residue is present. The presence of an ignitable liquid residue is a key factor in establishing the cause of a fire. The volatile nature of an ignitable liquid requires the container for collection and preservation of the physical evidence is contaminant free and air tight. At the laboratory, the ignitable liquid residue must be extracted from the fire debris before detection and analysis.
Several extraction methods exist, all having advantages and disadvantages, but adsorption methods are the most popular. The hydrocarbon components of the extracted ignitable liquid residue are separated then detected by chromatographic-spectrometric combined methods. The resulting data is interpreted to classify the ignitable liquid residue into a group of ignitable liquids with similar chemical and physical characteristics.
The American Society of Testing and Materials standard practice E1412-00 recommends a procedure for extracting ignitable liquid residues from fire debris by adsorption onto activated carbon suspended in the heated headspace above the fire debris sample within the collection container. Subsequent desorption of the ignitable liquid reside from the activated carbon is accomplished by a solvent. ASTM E1412-00 provides guidelines in the extraction process to reduce the possibility of preferential adsorption of the ignitable liquid components. Limitations of extraction by passive headspace sampling have been addressed concerning the effects of adsorption time, temperature, activated carbon size, and sample concentration. The study demonstrates the effects of chromatographic profile distortion when certain parameters in the extraction process are not controlled. If a representative sample of the ignitable liquid is not obtained, analysis of the results is compromised. The study presented here encompassed a more extensive investigation into the adsorption process. Hydrocarbon molecule interactions with the activated carbon are investigated, including a determination of the activated carbon surface area.
Activated carbon size and ignitable liquid volume were demonstrated to affect the chromatographic profiles due to saturation of the activated carbon. A modification to the extraction process incorporating a sub-sampling technique reduced the effects of saturation of the activated carbon.
The collection and preservation of fire debris evidence is crucial in retaining ignitable liquid residues for analysis. An effective fire debris evidence container must be vapor-tight and contaminant free. The possibility of cross-contamination is a common concern in choosing a suitable container for fire debris evidence. Containers recommended for fire debris evidence include metal "paint" cans with compression lids, glass mason jars with standard pressurecanning flats and bands, and special co-polymer bags. 3

CHAPTER TWO: FIRE DEBRIS EVIDENCE
The analysis of fire debris begins at the crime scene where most of the evidence in the form of an ignitable liquid is consumed in the fire leaving only trace amounts of its residue behind. At the fire scene, investigators determine the point of origin in order to collect any possible ignitable liquid residue. Fire debris from the point of origin is analyzed to determine if an ignitable liquid is present. The presence of an ignitable liquid is a key factor in establishing the cause of the fire. Since the ignitable liquid residue is volatile by nature, the collection of the fire debris is important in the preservation of the physical evidence. The next challenge is for the laboratory analyst to extract the ignitable liquid residue from the fire debris collected at the scene. There are several extraction methods published by the American Society for Testing and Materials each having advantages and disadvantages. Once the ignitable liquid residue is extracted, chromatographic methods of separation, usually coupled with spectrometric methods, are utilized in detecting the residue. After detection, the chromatographic and spectrometric data are interpreted to identify the ignitable liquid or classify the liquid residue into a group according to its composition.

Collection of Fire Debris Evidence
Fire debris analysis begins at the fire scene where a fire investigator determines the point of origin. Fire debris is collected from the point of origin then sent to the laboratory to determine whether the fire debris contains ignitable liquid residues. Timely collection and preservation of fire debris evidence is crucial due to the volatile nature of the ignitable liquid residues. The presence of an ignitable liquid is a key factor in the determining the cause of a fire as incendiary.

Point of origin
The point of origin of a fire is the location where the fire started -the place of beginning. 2 Determination of the point of origin involves incorporating the following information; the fire patterns left by the fire, observations reported by witnesses, analysis of the physics and chemistry in the fire initiation, and the development of the fire which would produce the conditions found at the fire scene. Examination of the fire scene usually provides the information needed for a determination of which area corresponds to the point of origin. The examination begins with a systematic procedure of identifying areas with the least amount of damage then moving toward the area of greatest damage. Once the general location of the origin is determined the specific location is identified based on the patterns produced by the movement of heat, flame, and smoke. The specific location of the origin will be where the heat ignited the first fuel. 3

Cause
The cause of a fire is determined by identifying the circumstances and factors which were necessary for the fire to occur. Those circumstances and factors include the device or equipment involved in the ignition, the ignition source, the material first ignited, and the circumstances or actions that brought all of these factors together allowing the fire to occur.

Solvent Extraction
The second standard extraction practice is E1386-00, Standard Practice for Separation Concentration with Solid Phase Microextraction (SPME) is the newest extraction technique. The SPME practice is best suited for the screening of fire debris samples to assess the relative concentration of the ignitable liquid or for aqueous samples. Solid phase microextraction is also considered to be a non destructive extraction technique because it recovers a small amount of the ignitable liquid residue. The required apparatus is a heating system, a temperature measuring device, a SPME fiber with holder, a punch, and septum. The SPME fiber is coated with a polymeric stationary phase which is held within a needle contained inside a holder. A SPME fiber with a 100 µm thickness of polydimethylsiloxane (PDMS) is recommended for ignitable liquids in the C 10 -C 25 range and a fiber with an 85 µm thickness of polyacrylate or a fiber with a 75 µm thickness of Carboxen/PDMS for ignitable liquids in the C 1 -C 10 range. After the evidence container lid is punctured a septum is inserted into the hole. The container is placed within an oven at a temperature between 60ºC to 80ºC for approximately 30 minutes to volatilize the ignitable liquid residues into the headspace. Immediately after removal from the oven the septum in the container lid is punctured with the SPME needle. The SPME fiber is inserted into the headspace allowing the ignitable liquid residues to adsorb onto the fiber. After one exposure of 5 to 15 minutes the SPME fiber is retracted into the needle and the SPME assembly removed from the septum. Upon removal from the heated headspace, the SPME fiber is inserted into the heated injection port of a gas chromatograph. 10

Detection of Ignitable Liquids
Ignitable liquids are petroleum based liquids that are either flammable or combustible.
Ignitable liquids are complex mixtures of hydrocarbons containing normal, branched, and cyclic alkanes as well as aromatics and polynuclear aromatics. An example is gasoline which is composed of over 400 compounds. 11 These petroleum based liquids are isolated from crude oil by a variety of chemical processes. The best methods for fire debris detection include gas chromatography, which separates the hydrocarbons within in the ignitable liquid residues before detection. There many detectors which may be used with the gas chromatograph. The choice of detector depends on the amount of chemical information or sensitivity required in the analytical method. These methods of detection include gas chromatography, gas chromatography-mass spectrometry (GC-MS), two dimensional gas chromatography -mass spectrometry (GC-GC-MS) and gas chromatography -mass spectrometry -mass spectrometry (GC-MS-MS).

Gas Chromatography
Gas chromatography is the basis for the detection of ignitable liquid residues. An analyte is injected into a heated port to be vaporized, then is carried through a column by an inert gas such as helium. The column consists of a liquid phase immobilized on the surface of an inert solid where the analyte is partitioned between the mobile phase (inert carrier gas) and the stationary phase (liquid phase). Chromatographic columns are housed in an oven in which the temperature is controlled. Columns vary in length, internal diameter, type and thickness of liquid phase. Injector temperatures, columns, gas flow rates, and oven temperatures are modified to achieve separation of the analyte compounds from one another before detection. Figure one is a schematic of a gas chromatograph. Common detectors of a gas chromatograph are flame ionization, thermal conductivity, electron capture detectors, and mass spectrometers. Data is presented as a chromatogram, a plot of retention time versus intensity which contains peaks corresponding to the separated compounds from the analyte. 12,13 Gas chromatography is a natural match for ignitable liquid detection because of its capability to separate the complex mixture into its major components. Advances in chromatography such as capillary columns provided additional data for interpretation since better separation of the components was posssible. Gas chromatography led to pattern recognition techniques for interpretation of the data and enabled an analyst to classify ignitable liquids into groups based on their physical properties.

Gas Chromatography -Mass Spectrometry
Gas chromatography -mass spectrometry has superseded gas chromatography as the most widely used method for the detection of ignitable liquids. GC-MS still has the capability of separating the numerous compounds constituting an ignitable liquid, but with the incorporation of a mass spectrometer with electron ionization (EI) for detection it provides additional chemical information about the compounds. Mass spectrometers basically bombard the separated molecules being eluted from the gas chromatographic column with high-energy electrons. The field produced by the high energy electron passing near an analyte molecule can cause ionization of the analyte and impart large amounts of energy to the newly formed ions. The ion dissipates the energy through numerous processes, which may include fragmentation into lower mass ions.
Electron ionization is the preferred ionization method for analysis of ignitable liquids due to its reproducibility. Some of the molecular fragments formed are ions that are accelerated within an electric field to pass into the mass analyzer which separates the ions according to their mass to charge ratio. Then the separated ions are detected by an electron multiplier which counts the number of ions striking it by producing a proportional electrical current. Figure 2 is a schematic of a mass spectrometer where the column enters the ionization source through the heated interface between the gas chromatograph and mass spectrometer. The ions travel to the mass analyzer with only selected ions allowed to proceed to the detector.

Figure 2: Schematic of a Quadrupole Mass Spectrometer
The data consists of perhaps thousands of mass spectra which are summed on a time axis producing a total ion chromatogram. The total ion chromatogram consists of peaks corresponding to the separated compounds plotted as retention time versus summed intensity. A mass spectrum of a peak within the TIC is a plot of ion abundance versus mass to charge (m/z) ratio. The chromatogram still provides data for pattern recognition, but is now combined with the spectral data which provides structural information about each compound within the ignitable liquid. 12,14 Two Dimensional Gas Chromatography with Mass Spectrometry Gas chromatography is a one-dimensional separation technique. However, separation of the components can be improved by employing two dimensional separations techniques such as GC x GC or hyphenated methods such as GC-MS. Gas chromatography -mass spectrometry is considered a two-dimensional technique because it separates components chromatographically then separates and identifies them spectrometrically. However, spectrometric separation in the second dimension can be limited by the chromatographic separation in the first dimension.
Another separation technique employed combines GC x GC with GC -MS to create GC x GC-MS, a three-dimensional separation technique which provides better chromatographic resolution.
The differences in the method compared to GC and GC-MS are the inclusion of a second column and a focusing apparatus. Figure 3 is a schematic of the instrument showing a sample inlet connected to the first column which is connect serially to the second column by a focusing apparatus and terminating with a detector, which in this case is a mass spectrometer. Typically, the first column (dimension 1) is longer and has a larger internal diameter with a thicker film thickness than the second column (dimension 2). Also, the first column is relatively non-polar compared to the second column thus allowing compounds with similar boiling points but different functional groups to be further separated. The function of the focusing apparatus is to accumulate the analyte components after they elute from the first column then transfer them in their entirety onto the second column. There are multiple types of focusing apparatuses each with a different operating principal.

Figure 3: Schematic of a Two Dimensional Gas Chromatograph
One apparatus is the thermal modulator in which a modulator tube is the interface inserted between the two columns. The modulator tube is heated to desorb the analyte from the tube onto the second column. The other apparatus relies on cryogenic focusing of the analyte. The data is commonly plotted with the x-axis (minutes) reflecting the retention time of the first column and the y-axis (seconds) reflecting the retention time of the second column. The mass spectrum is summed to produce the total ion abundance for each point which is plotted in the third dimension (z-axis) 15 , 16

Gas Chromatography -Mass Spectrometry -Mass Spectrometry
Another method of separation and detection of ignitable liquids gaining some popularity is gas chromatography-mass spectrometry -mass spectrometry. This instrument couples a gas chromatograph with a mass spectrometer containing multiple quadrupole mass analyzers or one

Analysis of Ignitable Liquids
Typically, the ignitable liquids extracted from the fire debris are analyzed by gas chromatography or gas chromatography with mass spectrometry. The American Society for Testing and Materials (ASTM) publishes two testing methods for ignitable liquid residues. One testing method is the ASTM E 1387 Standard Test Method for Ignitable Liquid Residues in Extracts from Fire Debris Samples by Gas Chromatography which provides methods for the instrumental analysis and interpretation of the data by pattern recognition to classify the ignitable liquid residue. 18 The other testing method is the ASTM E 1618 Standard Test Method for

Ignitable Liquid Residues in Extracts from Fire Debris Samples by Gas Chromatography-Mass
Spectrometry which provides methods for the instrumental analysis and interpretation of the data by pattern recognition, extracted ion, and target ion analysis for the classification of ignitable liquid residues. 19

Pattern Recognition
Pattern recognition techniques have been developed with the evolution of chromatographic methods of detection for fire debris. Both gas chromatography (GC) and gas chromatography mass spectrometry (GC-MS) produce chromatograms utilized in visual pattern recognition. Each peak within the chromatogram corresponds to a hydrocarbon in the ignitable liquid. The culmination of the peaks produces a pattern particular to a class of ignitable liquids.
Groups of peaks composed of compounds of similar chemical composition and boiling points are examined to determine if the relative retention times and peak ratios are consistent with known ignitable liquids. 20 An example is the grouping of C 2 alkylbenzenes (o, m, p-xylenes) and C3 alkylbenzenes within gasoline. Another example is the ratio of pristine to heptadecane and phytane to octadecane within diesel fuel. The overall pattern is also examined and compared with the pattern of a known ignitable liquid. The criteria for classifying and identifying ignitable liquids by visual pattern recognition techniques are published in both ASTM standard methods E 1387 and E1618.

Extracted Ion Profiling
Extracted ion profiling is the most commonly utilized method of detecting ignitable liquids by mass spectrometry in conjunction with gas chromatography. The mass spectrometer is capable of producing ion profiles by extracting ions from the total ion chromatogram as well as identifying specific compounds from their mass spectrum.. Petroleum derived ignitable liquids are generally comprised of compounds that can be classified into one of five general categories; alkanes, cycloalkanes, aromatics, indanes, and polynuclear aromatics. These classes of compounds with common chemical structures have common ions which are extracted to produce an extracted ion profile. Table 1 summarizes some of the important ions associated with classes of compounds found in ignitable liquids. 19 Alkanes produce many ion fragments which are typically 14 mass to charge units apart corresponding to the loss of a methylene (CH 2 ) group. Alkenes fragment in a similar fashion to the alkanes, but due to the double bonds in their chemical structure, the mass to charge ions are two less than those of the alkanes. Cycloalkanes have a predominant ion of 83 m/z corresponding to cyclohexyl. Smaller cycloalkanes or the fragmentation of larger cycloalkanes produces ions of 55 m/z and 69 m/z, [C 4 H 7 ] + and [C 5 H 9 ] + respectively. Aromatics have a ring structure which is more stable during the fragmentation process than the alkyl chains and therefore the molecular ion is usually seen in the mass spectrum. A common ion found in aromatic spectra has a mass to charge of 91 which is due to the formation of the tropylium ion, [C 7 H 7 ] + The extracted ion profiles are compared to the ion profiles of known ignitable liquids by visual pattern recognition. 19 Extracted ion profiles are also utilized in determining the relative abundance of certain classes of compounds within the ignitable liquid. This information assists in the classification of the ignitable liquid.

Target Ion Analysis
Target compound analysis uses key specific compounds to characterize an ignitable liquid. Instead of separating the hydrocarbons into classes based on their fragmentation, target compound analysis seeks to identify specific analytes present as well as some selected isomers.
A comparison of the relative peak heights for closely eluting aromatic and aliphatic compounds assist in ascertaining which class of ignitable liquid is present. 21 The hydrocarbons chosen for the peak ratios must be within one minute in retention time to minimize the effects of (weathering) evaporation. 21 Another consideration is that the hydrocarbons are solely present in the ignitable liquids and not from other contaminants from the fire debris. Table 2 contains a list of common hydrocarbon target compounds for medium petroleum distillates and gasoline. The relative ratios of the ions of the target compound from known ignitable liquids of particular classes are compared to the relative ratios of the ions of the same target compounds from the unknown ignitable liquid residue. The target compound data can be plotted as a target compound chromatogram that can be visually compared with other target compound chromatograms of known ignitable liquids.  Because some classes can not be distinguished without mass spectrometry ASTM E1618-01, the most common standard testing method used today, will be discussed here. The carbon number range is determined by comparing the chromatogram to a reference or test mixture containing known normal alkanes. Figure 4 is a total ion chromatogram on n-alkanes utilized as a hydrocarbon ruler, it indicates the three carbon ranges described in ASTM E1618. The light range is between butane and nonane with no major peaks after dodecane. The medium range is between octane and tridecane with the majority of the pattern between heptane and tetradecane.

Target Compounds
The heavy range is between nonane and eicosane or a higher n-alkane and must encompass at least five consecutive n-alkanes. It may be necessary to characterize an ignitable liquid as "light to medium" or "medium to heavy" for those ignitable liquid patterns not fitting neatly into one of the previous carbon ranges. Gasoline has a carbon range between butane and dodecane and therefore does not fall into any of the carbon ranges described earlier. The eight classes are gasoline, petroleum distillates, isoparaffinic products, aromatic, products, naphthenic-paraffinic products, n-alkane products, de-aromatized products, and oxygenated products. If an ignitable liquid can not be characterized into one of these classes it is classified as miscellaneous. 19

Figure 4: Total Ion Chromatogram on n-Alkanes from n-Hexane to n-Eicosane
The characteristics and distinguishing features of the chromatographic patterns for each classification are described below along with examples. 19, 20, , 22 23 A Gasoline chromatographic pattern is characterized by an abundance of aromatic compounds whose peaks cluster in specific patterns within a carbon range of C 4 to C 13 as demonstrated in Figure 5. Since the most prevalent species of compounds in gasoline are aromatics, the aromatic ion profile will be the most abundant. The chromatographic pattern will contain C 2 , C 3 , and C 4 alkyl benzenes in approximately the same relative concentrations of a known gasoline. Most gasoline contains naphthalene, 1-and 2-naphthalene, indan, and methyl indans.  De-aromatized distillates are products characterized by the traditional petroleum distillate distribution with a notable absence of aromatic compounds as demonstrated in Figure 11.
Alkanes are the most abundant ion profile. There is a notable reduction in the abundance of aromatics within the aromatic ion profile compared to the abundance of alkanes within the alkane ion profile for a de-aromatized distillate.
Oxygenated products contain a significant amount of an oxygenated product or products. ASTM E1618-01 suggests at least one order of magnitude above the other peaks within the chromatogram. Oxygenated products usually contain a small number of compounds which produce a chromatogram with no particular chromatographic pattern as demonstrated in Figure   12. Oxygenated compounds within the ignitable liquid must be identified with gas chromatography-mass spectrometry. The analysis of fire debris evidence encompasses four major aspects; collection of physical evidence at the point of origin, extraction of the ignitable liquid residue from the fire debris, detection of the ignitable liquid residue, and data interpretation to classify the ignitable liquid residue. Each aspect relies on the previous one to ultimately provide the fire investigator with useful information in determining the cause of the fire. There are challenges within each step of the process from preventing the loss of the physical evidence through evaporation, extracting a representative sample of the ignitable liquid from the fire debris, increasing the selectivity and sensitivity of instrumentation, and providing a robust data analysis method for the identification of an ignitable liquid. The methods of collection, extraction, detection and interpretation of fire debris evidence within this chapter summarize the current practices and published methods utilized in fire debris evidence analysis. Advances on each aspect of fire debris analysis based on scientific principles forms the basis for this research.

Introduction
The American Society of Testing and Materials standard practice E 1412-00 covers the procedure for removing ignitable liquid residues from fire debris by adsorption onto activated carbon suspended in the static headspace above the sample then desorbing the residue from the adsorbent with a solvent. The extraction of ignitable liquid residues from fire debris by passive headspace sampling with activated carbon is the most commonly used method for separating ignitable liquids from fire debris. 24 The effects of adsorption time, temperature, carbon size, and   Identification of the protonated species and deuterated species of both hydrocarbon solutions was accomplished by comparing retention times and spectra with standards.
The van der Waals calculations used to determine the surface area occupied by each hydrocarbon adsorbed onto the activated carbon was performed with Hyperchem 7 molecular modeling software.

Determination of Activated Carbon Surface Area
Hydrocarbon Solution 1 was deposited in volumes of 12µl, 18µl, 24µl, 36µl, 48µl, 96µl 120µl, 500µl, and 720µl which correspond respectively to 1.52 X 10 -5 , 2.28 X 10 -5 , 3.04 X 10 -5 , 4.56 X 10 -5 , 6.09 X 10 -5 , 1.22 X 10 -4 , 1.52 X 10 -4 , 6.34 X 10 -4 , 9.13 X 10 -4 moles of each hydrocarbon into vial insert(s) within each of nine glass jars. The geometric area of the activated carbon disks created with the hole punch was 33.2 mm 2 . The total number of hydrocarbon moles extracted from the activated carbon disks increased significantly from the lowest volume of 12µl up to 18µl, however further significant increases were not observed for the larger volumes of hydrocarbon liquid being deposited into the system.  Figure 14 shows a plot of the surface area corresponding to the volumes of hydrocarbon liquid.

Hydrocarbon Molecule Substitutions on Activated Carbon
Three sets of three activated carbon disks with an area of 33.2 mm 2 were designated as sets A, B and C. Each set of disks was perforated onto a paperclip which was placed in a vial insert of a glass mason jar with 10 µl of hydrocarbon Solution 1 (heptane, toluene, octane, nonane and decane) as shown in Figure 13. The containers were placed in a 66°C oven for 16 hours. After the containers were removed from the oven and allowed to cool, set A was analyzed while sets B and C were each placed into a clean unused jar. Set B had an additional  Table 3 summarizes the experimental conditions for each container. The hydrocarbons extracted from the activated carbon disks in set A had an average number of 9.71 X 10-6 moles. The hydrocarbons extracted from the disks in set C which were exposed twice to 10 µl of hydrocarbon Solution 1 had an average number of 1.82 X 10-5 moles.
The total number of moles of hydrocarbons extracted from the activated carbons doubled upon the second exposure to the hydrocarbon solution as shown in Figure 15. The number of protonated heptane, toluene, and decane moles from set A were the same as those from set C with only the addition of heptane-d 16 , toluene-d 8 , octane, nonane, and decane-d 22      The liquid volumes recovered from the jars with volumes of 120 µl, 500 µl, and 720 µl hydrocarbon solutions with their respective mole fractions are given in Table 4. The

Modifications to the Extraction Method
The first experiment presented here was conducted to determine the effect of activated carbon saturation on the chromatographic profile of gasoline. Two volumes of un-weathered (un-evaporated) gasoline, 12µl and 96µl were deposited into quart glass jars with 33.2 mm 2 activated carbon disks then extracted following the ASTM E 1412-00 standard testing method.
Two diluted neat solutions of the same gasoline were analyzed, one un-weathered and the other 75% weathered (by volume). The chromatographic profile of the recovered hydrocarbons from the 12µl gasoline sample shown in Figure 21 resembles the chromatographic profile of the same un-weathered gasoline in Figure 22. The chromatographic profile of the recovered hydrocarbons from the 96µl gasoline sample shown in Figure 24 resembles the chromatographic profile of the 75% weathered gasoline in Figure 25.   When the volume of gasoline was low, the activated carbon surface did not become saturated and the chromatographic profile was not distorted from that of the neat solution of 0% weathered gasoline. However, the chromatographic profile became distorted, resembling the 75% weathered gasoline, when the activated carbon was saturated by the larger volume of gasoline.
Since the chromatographic effects of gasoline adsorption onto activated carbon recovered from carpet and vial inserts are similar as shown in Figures 23 and 25 it can be assumed the adsorption of the gasoline onto the carpet has no influence on the chromatographic profile. The amount of ignitable liquid within fire debris cannot be controlled and the prohibitive expense of increasing the size of the activated carbon requires an alternative approach to sampling the fire debris. The approach taken in this study was to limit the sample size, thus reducing the ignitable liquid concentration. Two steps preceding the extraction procedure; heating the sample and removing a removing a sub-sample produced a representative fraction of the original sample. The passive headspace concentration extraction procedure was performed on the sub-sample. The ignitable liquid residue concentration of the sub-sample did not saturate the activated carbon therefore the chromatographic profile resembled the original ignitable liquid as seen in Figure 26.

Introduction
An effective fire debris evidence container must be contaminant free and vapor-tight.
The need for a vapor-tight evidence container is critical for retaining the ignitable liquid residues  Calibration curves were created to quantify the hydrocarbons recovered from the activated carbon by an external standard method.

Placement of Activated Carbon
Five activated carbon disks of 33.2mm 2 geometric areas were adhered to a metal rod   Figure 31. The results demonstrate that the ratio of adsorbed hydrocarbon to vapor phase hydrocarbon is independent of the hydrocarbon volume when complete evaporation has occurred and the activated carbon remains unsaturated.

Figure 31: Ratio of Hydrocarbons in the Vapor Phase to Hydrocarbons Adsorbed
To determine container leak rates at 66°C, 5µl of toluene-d 8, 5µl of toluene, and 5µl of pxylene were deposited into a metal quart can, a glass quart jar, and a quart volume DebrisPAK ® bag, respectively. Each type of container with its corresponding hydrocarbon was placed into a secondary container with 3 carbon disks constituting a system. Triplicate systems were placed in the oven for 20, 50, 100, and 150 hours for a total of 12 systems. The amount of hydrocarbon leaking from each container was ascertained based upon the ratios calculated from results in Figure 31. The percent leak in moles per hour for each type of container was determined to be an approximate linear function of time, i.e. zero order kinetic behavior. The observed leak rates for the metal cans, glass jars, and DebrisPAK ® bags were 3.0 x 10 -3 mol %/h, 6.5 x 10 -3 mol %/h, and 2.0 x 10 -4 mol %/h, respectively as shown in Figure 32.

Figure 32: Container Leak Rates
The glass mason jars exhibited the fastest leak followed by the metal paint cans then the DebrisPAK ® bags. When properly heat sealed the DebrisPAK ® bags did not leak significantly, but when sealed incorrectly the leak rate increased substantially. Only one bag out of twelve bags used in the experiments did not seal properly clearly creating anomalous results.

Effects of Container Leaks Effect Container Leaks have on the Hydrocarbon Molecular Distribution
The following experiment was performed to determine if a leaking container effects the distribution of a hydrocarbon remaining inside the container. A 10 µl volume of an equimolar hydrocarbon solution containing heptane, toluene, octane, nonane, and decane was deposited into a jar, can, and bag of quart size volumes then the containers were properly sealed. Each of these containers was placed inside an individual full sized DebrisPAK ® bag containing three activated carbon disks suspended from the top of the bag by a paperclip and dental floss as shown in

Contamination of Containers from the Environment
Since it has been determined that glass mason jars leak, this type of container was used in the following experiment to determine if hydrocarbons from the environment can cause contamination within the container. A full size DebrisPAK ® bag contained two glass mason jars, a set of three activated carbon disks suspended from the heat seal by dental floss and a paperclip and 10 µl of toluene-d 8 deposited into the bag . One of the jars contained three activated carbon disks perforated onto a paperclip and no hydrocarbon liquid and the other jar contained 50 µl of toluene deposited into a vial insert as shown in Figure 28 D. The bag was placed in a 66ºC oven for 163 hours, removed then allowed to cool to room temperature. The analyses of the activated carbon disks from the DebrisPAK ® bag revealed that toluene constituted seventy percent of the hydrocarbons recovered. Furthermore, the activated carbon disks from the jar originally containing no hydrocarbon liquid now contained an equal quantity of both toluene-d 8 from the bag and toluene from the other glass jar as shown in Figure 35.

Cross Contamination of Containers from Adjacent Containers
In the previous experiment, a sealed glass jar containing toluene-d 8

Introduction
Fire debris analysts utilize pattern recognition, extracted ion profiling, and target ion profiling techniques to classify an unknown ignitable liquid extracted from fire debris into a group or type of ignitable liquid by comparing the relative ratios of the components (peaks) observed in the total ion and extracted ion profiles. The data from an unknown sample is compared to similar data obtained from reference ignitable liquids of known classes. ASTM E1618-01 describes how to use these techniques for class determination of the ignitable liquid.
Various methods for improving detection and alternative methods for data analysis and interpretation have been studied. Parallel -column gas chromatography, GC X GC/MS, and GC/MS/MS methods have been introduced to improve resolution of the components within these complex mixtures thus aiding in the identification of the ignitable liquid components. 15, , 16 42 Automated comparisons of GC/MS data of complex mixtures have been utilized through advanced software, but still rely on visual pattern recognition for data interpretation. 43 Statistical methods for comparison of GC/MS data have been applied for classification of ignitable liquids as well as identification of gasoline but have not provided a method of common source estimation with a known statistical certainty. 44

Methods and Materials
The gas chromatographic -mass spectral data sets utilized in the present study on the analysis of  Table 5 were not altered (i.e.  Shell 87 Thirteen ignitable liquids representing six of the nine classes described in the ASTM E1618-01 were chosen for comparison. The classification and carbon range of each ignitable liquid chosen was ascertained from the ILRC database. The database classification information is provided by a committee consisting of several fire debris analysts from local, state and federal crime laboratories. The committee members review the data sets and then determine the classification, predominant ion profile, carbon range, and major peaks for each ignitable liquid.
The sample reference numbers with their associated ASTM classification and carbon range are shown in Table 6  The Agilent Chemstation 3D-Export option was used to export spectral data into comma- The covariance matrix is a table listing the variances (diagonal) and the co-variances (off diagonal) of two or more variables; each ignitable liquid matrix contains 197 variables. 54 The generation of a covariance matrix Z from the data matrix Y is graphically represented in Figure   37. The generated covariance matrix Z is a symmetric matrix with each element z ij representing the similarity between the measured intensities of the variables (abundances of the m/z ratios).
Since the values in the Y matrix were not normalized before the calculation of Z, the values for the Z matrix are weighted in proportion to the absolute magnitude of the ion abundances of Y. 55 The ion abundances of the Y matrix are concentration dependent therefore the elements of Z are also concentration dependent. To remove any sample concentration dependence each covariance matrix Z was normalized such that the sum of all the elements within the matrix equaled a value of 1.0 and was designated Z N . Normalization of the covariance matrices was required for comparison calculations between two ignitable liquids.

Distance between Covariance Matrices
A distance between two Z N matrices was calculated to facilitate an analytical comparison of two ignitable liquids. A Manhattan distance D was calculated on an element to element basis as the absolute difference between the matrix elements. The distance D of each element was summed over all of the elements then divided by two for an absolute difference between the two matrices. The distance D between two matrices designated Z N1(ij) and Z N2(ij), is calculated by The values of D lie between 0 and 1, where two identical Z N matrices have a minimum distance D of 0 and two non related Z N matrices have a maximum distance D of 1. Alternatively, a similarity index, S, can be calculated based on D which is defined by equation 4. D is defined so that D max equals 1 and lies between 0 and 1. Therefore, the similarity equals 1 minus D. (4)

Characterization of Ignitable Liquids Covariance Maps of Ignitable Liquids
Hydrocarbons with specific functional groups within their molecular structure typically fragment in the electron ionization source in a particular manner producing a set of diagnostic ions. Extracted ion profiles of the m/z ratio constituting five various molecular structures aid in the classification of ignitable liquids according to ASTM E 1618-01 and are listed in Table 1.

Distances between Covariance Matrices of Ignitable Liquids
The distances D calculated between all thirteen ignitable liquid samples are given in Table 7. All of the distances D fall over a range of 0 to 1 with a maximum and minimum values observed of 0.975 and 0.084, respectively. The maximum distance observed was between SRN 119 an isoparaffinic product and SRN 303 a 75% weathered gasoline reflects the striking difference between isoparaffins consisting primarily of branched alkanes and gasoline consisting primarily of aromatics. The minimum distance observed was between SRN 301 and SRN 105 which are 0% weathered gasolines.  The distance between the two 0% weathered gasoline samples was 0.084 and the distance between the two 75% weathered gasoline samples was 0.103. However, a comparison of the distances between 75% weathered gasoline samples and 0% weathered gasoline samples gave an average distance of 0.336 ± 0.0262 thus distinguishing a 0% weathered gasoline sample from a 75% weathered gasoline sample. The average distance between the two 0% weathered gasoline samples and the single 25% weathered gasoline sample was 0.125 ± 0.0170 which may indicate that it is possible to distinguish the 25% weathered gasoline from 0% weathered gasoline after more comparisons are completed. A comparison of the distances between the petroleum distillates suggests the light (LPD), medium (MPD), and heavy distillates (HPD) can be differentiated from one another, especially the light petroleum distillate. The distance between the two MPD was 0.105 while the average distance between the two MPD and the single HPD was 0.153. The LPD average distance with the two MPD was 0.426 and the distance between the LPD and the HPD was 0.456. Most of the remaining Z N matrix distance comparisons demonstrate significant differences between the major classes designated by ASTM E 1618-01 of ignitable liquids.

Discrimination of gasoline samples
Each gasoline sample in Table 5    ) was rejected.
A common statistical approach to differentiating two samples with a known level of statistical certainty is to hypothesize that the two samples come from the same population and therefore have the same value for some measurable parameter. The hypothesis is referred to as the null hypothesis where statistical tests are employed to determine whether the null hypothesis should be accepted or rejected. If the null hypothesis is rejected the alternative hypothesis is accepted, i.e. the two samples come from different populations. The statistical test (t-test) is conducted at some significance level α which controls the risk of making an error when accepting or rejecting the null hypothesis. The error of incorrectly rejecting the null hypothesis is a Type I error which is controlled by the significance level α. The error of incorrectly accepting the null hypothesis is a Type II error and the probability of making a Type II error is given by β. Figure 41 depicts two probability distributions with different means and equal standard deviations depicting two populations. The power of a test corresponds to 1-β and represents the probability of making a correct decision when the null hypothesis is false.

Figure 41: Depiction of Two Probability Distributions
For two given distributions, β is determined when α (significance level) and n (number of measurements for each sample) are set. A smaller α leads to a greater β hence a lower power for the test. In practice the significance level is set α = 0.01 when a Type I error is most costly whereas α = 0.05 is commonly used when a Type II error is most costly.
When comparing the individual gasoline samples, the number of replicate measurements must be determined to provide statistically reliable results which protect against Type II errors.
An analysis of the power of the two-sided t-test for the given DS D and SS D with their associated standard deviations was performed for the varying sample sizes n SS and n DS at a significance level α of 0.05. 58 The result of the power analysis is shown in Figure 41indicating for α = 0.05 at least 7 D DS values are required to achieve 99% probability of making a correct decision when the null hypothesis is false. The result suggests that the triplicate GC-MS analyses of each gasoline should be able to discriminate gasoline samples from different sources with a 1% or less chance of making a Type II error.  rule out a common source with a known risk of Type I error, but can not prove the existence of a common source.

Identification of Two Unknown Gasoline Samples
Two blind tests were performed to evaluate whether the covariance method which was able to discriminate the 10 gasoline samples could correctly identify unknown gasoline sample within a set of possible sources. Aliquots of two gasoline samples from Table 5  Unknown B and each of the ten gasoline samples was gasoline 4, which was significantly smaller than the other average distances. Unfortunately, it was significantly different from SS D of gasoline 4 as proven by the t-test.

Discussion
Covariance mapping with subsequent comparisons by distance measurements could be used in rapidly comparing an unknown ignitable liquid to a database of reference ignitable liquids. A database search between Z N matrix of an unknown ignitable liquid and the Z N matrices of ignitable liquids from a database would generate a list of best matches based on the distances calculated. The distance measurement between covariance matrices of ignitable liquids can distinguish between various classes of ignitable liquids as well as the sub-classifications of light, medium, and heavy. The method was able to distinguish the relative amounts of weathering (evaporation) between gasoline samples.
Covariance mapping with subsequent comparisons by distance measurements and t-tests has distinguished 10 gasoline samples as having come from different sources. GC-MS 3D data has been converted to a covariance matrix then sample comparisons have been made by calculating a distance between the normalized matrices of the two samples. To determine if the distance is significant a t-test was performed while keeping a Type II error low. Blind tests were conducted to determine if an unknown gasoline sample could be correctly identified with the method. No Type II errors (incorrectly accepting the null hypothesis) were made, but a Type I error (incorrectly rejecting the null hypothesis) was made for Unknown B.
The distance measurement of sample covariance matrices with a subsequent t-test is an applicable method for comparative analysis of complex mixtures. The study used neat dilute solutions of ignitable liquids analyzed with the same analytical method and performed almost exclusively on the same instrument. Identification of ignitable liquids collected from a fire scene becomes more complicated with the addition of pyrolysis and combustion products from building materials and furnishing within the structure as well as volatiles remaining from the manufacturing process. 59 Other complications in identification of an ignitable liquid arise from weathering, biological degradation, 60 chromatographic distortion due to the extraction procedure, and inter laboratory differences in analytical methods. These complications are valid with current methods of characterization and identification of ignitable liquids by pattern recognition.
The method presented was not tested with fire debris samples, but has the capability of removing a covariance matrix of substrate compounds from the fire debris sample covariance matrix for an improvement in ignitable liquid comparison with the fire debris sample. The covariance mapping with a distance measurement provides a rapid method to characterize an ignitable liquid by class, carbon range and percent evaporation and as a direct comparison between two liquids with a known statistical certainty

CHAPTER SIX: DISCUSSION
There are four major aspects in the analysis of fire debris beginning at the fire scene with the collection of fire debris evidence from the point of origin, the extraction of ignitable liquid residues from the fire debris, the detection of the ignitable liquid residue, and the interpretation of the data leading to the classification or identification of the ignitable liquid. Valuable information is obtained through the analysis of the fire debris allowing the fire investigator to determine the cause of the fire thus assisting in assigning responsibility and culpability. Current A major concern in fire investigations is the collection and preservation of ignitable liquid residues obtained from fire debris at the point of fire origination. The physical evidence container must be vapor-tight due to the volatile nature of ignitable liquid residues and contaminant free. Metal paint cans, glass mason jars and co-polymer bags are typical fire debris evidence containers. Activated carbon positioned throughout the depth of a glass jar container revealed no significant variation in the hydrocarbon molar distribution hence placement of the activated carbon within the container was not critical in the experiment. Leak rates of the three container types were calculated and compared to one another followed by demonstrations of the possible ramifications. After six days, the glass jar lost 98 percent of the hydrocarbons originally deposited into the container. The fastest leak rate was obtained by the glass jars followed by the metal cans with the co-polymer bags not leaking at all when properly sealed. Molar distributions of the recovered hydrocarbons were affected by the closing mechanisms of the glass jar and metal can. Hydrocarbons with smaller collision diameters leaked from the jars preferentially to those with larger collision diameters. However, the hydrocarbons from the metal can leaked the container at equal rates.
Interpretation of GC/MS data by pattern recognition of chromatographic profiles, extracted ion profiles, and target ion analysis rare described in ASTM E1618-01 with a classification scheme designed to group ignitable liquids together based on their chemical and physical properties. The classification scheme relies on comparing the unknown ignitable liquid residue to a known reference ignitable liquid. The method relies heavily upon the analyst's interpretation of the data and the standard practice. Covariance mapping with subsequent comparisons by distance measurements was able to distinguish between various classes of ignitable liquids and sub-classify by boiling point ranges established by the ASTM E1618-01 classification scheme. States of gasoline evaporation could be ascertained by comparison of covariance matrices of known evaporated gasoline. Ten gasoline samples were compared to one another by calculating a distance between the normalized covariance matrices of two gasoline samples. To determine if the distance (difference) between the covariance matrices was significant a t-test was performed. All gasoline samples were determined to have originated from different sources with no Type I or Type II errors occurring. Two blind tests were conducted to determine if an unknown gasoline could be identified with a gasoline from a known source. One gasoline sample was correctly identified from the 10 gasoline samples. However, the other gasoline sample could not be identified with a known statistical certainty and by rejecting the null hypothesis a Type I error occurred. The combination of covariance mapping with a distance measurement and t-test has the potential to characterize, distinguish and possibly identify ignitable liquids from existing GC/MS data with a known statistical certainty.
A covariance mapping method combined with a distance measurement allows for quick searching of a large database of ignitable liquid GC/MS data. It does not require the analyst to perform pattern recognition, ion profiling, nor target ion analysis to compare the ignitable liquid samples thus saving time. The results in the study of the study are compatible with the ASTM classification system and do not rely on the subjective interpretation of the analyst. Comparisons between GC/MS data sets originating from different laboratories may be possible since during the formation of the covariance matrix the time element (scan number) is removed.
Future studies for software development using covariance mapping with distance measurements would enable an analyst to search a database containing GC/MS data collected from multiple laboratories of ignitable liquid samples. Inter-laboratory comparisons involved further refinements to the method. Other complications with developing software for database searching are weathering (evaporation) of samples which alters the composition, and interfering products from the fire scene. The method has proven to be an excellent tool in comparing ignitable liquids from neat solutions. However, most ignitable liquid samples encountered in a crime laboratory have been weathered and contain interfering products all produced from the fire and fire scene.