Pyrolysis Products, Substrate Interferents and Background Subtraction

Pyrolysis is the thermal decomposition of organic material in the absence of sufficient oxygen for complete combustion. In a structural fire, the pyrolysis zone exists between the oxygen-rich flame front and the cooler, unburned substrate behind it. This zone produces the volatile organic compounds that feed the flame; it also produces the residues that remain in the debris after the fire is suppressed and cooled.

The pyrolysis products of a given substrate depend on three variables: the chemical composition of the starting material, the temperature reached, and the duration of heating. Most building and furnishing materials are complex polymer or cellulosic materials whose pyrolysis pathways are known but whose precise product distribution varies with the manufacturing formulation, the presence of flame retardants, plasticisers, and dyes, and the local fire conditions. This variability is the reason that substrate background cannot be reliably predicted from first principles; it must be measured from the actual material present in the scene.

General classes of pyrolysis product and their origin:

Alkenes (particularly 1-alkenes with terminal double bonds at C6 to C12) arise from the thermal cracking of polyolefin polymers, especially polypropylene (PP) and polyethylene (PE). These are dominant structural polymers in carpet fibres, carpet backing, and many packaging materials. The GC-MS signature of PP pyrolysis is a series of triplets (n-alkane, terminal alkene, diene or other internal alkene) repeated at each carbon number from C5 to C15, with the pattern most intense around C8 to C12. This pattern overlaps directly with the carbon range of gasoline and light to medium petroleum distillates.

Aromatic compounds arise from pyrolysis of polystyrene, polyurethane foam, ABS plastic, and to a lesser degree from cellulosic materials at high temperatures. Polystyrene produces styrene monomer (m/z 104 and 78), alpha-methylstyrene, and a range of dimers and trimers. Polyurethane foam produces toluene, benzene, 2-butanone (methyl ethyl ketone), and various nitrogen-containing heterocycles. The aromatic pyrolysis products from a polyurethane foam sofa cushion can, in a heavily burned interior, produce an aromatic-fraction chromatogram that resembles a weathered gasoline or an aromatic naphtha.

Chlorinated compounds arise from pyrolysis of polyvinyl chloride (PVC). Floor covering, electrical cable insulation, window frames, and pipe lagging in older residential properties are frequent PVC sources. PVC decomposes at around 240 to 300°C to release hydrochloric acid and produce a complex chlorinated hydrocarbon mixture, including chlorinated benzenes and biphenyls at higher temperatures. These compounds appear at characteristic m/z values (e.g., m/z 112, 146 for dichlorobenzene isomers) and can create confusion if the analyst interprets them as oxygenated solvent or aromatic product residues.

Terpenoids and other natural product pyrolysis fragments arise from wood, paper, and natural-fibre materials. Alpha-pinene, limonene, and terpinen-4-ol are among the compounds that can appear in debris from burned wood or paper. These contribute to the chromatographic complexity but are generally distinguishable from petroleum distillate patterns by their mass spectral fragmentation (m/z 93, 121, 136 for monoterpenes rather than the m/z 91, 105, 119 alkylbenzene series characteristic of gasoline).

Residential and commercial carpet is a composite material with multiple components, each with its own pyrolysis profile. The pile fibre may be nylon (polyamide 6 or polyamide 66), polypropylene, or wool. The primary backing is usually a woven or nonwoven polypropylene fabric. The secondary backing is typically a polypropylene or SB latex (styrene-butadiene) bonded to the primary backing with an adhesive that may contain hydrocarbon plasticisers. The pile is tufted through the primary backing and held in place by the adhesive layer.

Polypropylene pyrolysis produces the hydrocarbon triplet pattern described above, most intense in the C8 to C12 range. Styrene-butadiene rubber (SBR) pyrolysis produces styrene (m/z 104), 4-vinylcyclohexene (m/z 108), and a range of dimers that generate a complex aromatic background. Carpet adhesive, depending on formulation, may contribute aliphatic and aromatic hydrocarbon plasticisers. The combined pyrolysis profile from a burned carpet sample with pad and adhesive can be, in terms of mass spectral pattern and TIC envelope, a credible mimic of a complex petroleum product.

The most studied false-positive substrate in the fire debris literature is polypropylene carpet, documented in Wineman and colleagues (Journal of Forensic Sciences, 1994) and subsequently in DeHaan and Icove's Kirk's Fire Investigation, which established that polypropylene carpet pyrolysis consistently produces GC-MS patterns meeting the criteria then in use for light and medium petroleum distillate identification. This finding prompted the development of the comparison sample requirement and the background subtraction methodology as standard practice, and it was a significant driver of the revision of ASTM E1618 to its current form with more explicit compound-class criteria.

In accredited laboratories worldwide, the current practice is that carpet samples from any burned room should be accompanied by a comparison sample from unburned carpet of the same type from the same room or an adjacent room. The E1618 guidance notes specifically that polypropylene carpet backing is a "Type C interferent" (a substrate that can produce patterns meeting E1618 category criteria) and that its presence must be accounted for by comparison sample analysis before a positive identification opinion can be issued. The scene-side protocol that ensures the comparison sample is collected correctly alongside the debris sample is covered in accelerant detection at the scene.

Polyurethane (PU) foam is the principal cushioning material in upholstered furniture and mattresses in most countries. At pyrolysis temperatures of 200 to 400°C, PU foam degrades through a complex pathway that produces isocyanate fragments, polyol fragments, and a range of small aromatic and aliphatic molecules including toluene, benzene, styrene, 2-butanone (methyl ethyl ketone), and acetonitrile. The aromatic fraction of PU foam pyrolysis, particularly the toluene and C2-benzene signals at m/z 91 and 106, can produce extracted ion profiles that superficially resemble the aromatic fraction of a weathered gasoline or an aromatic naphtha product (E1618 category 6).

The critical distinguishing feature is the absence of the full alkylbenzene series in PU foam pyrolysis. Gasoline and aromatic products show a progression through C1-benzene, C2-benzene, C3-benzene, and C4-benzene homologues. PU foam pyrolysis is dominated by lower-carbon aromatics (benzene, toluene, C2-benzenes) without the higher homologues. An analyst applying the full E1618 criteria and comparing against the comparison sample from unburned foam cushion material will generally distinguish the two, but an analyst pattern-matching the TIC without extracting the full ion profile series may not.

Vinyl flooring (PVC-based resilient flooring) presents a different challenge. Its pyrolysis, dominated by dehydrochlorination, does not produce petroleum-mimic aromatic patterns as its primary product. However, the plasticiser fraction of flexible PVC flooring (typically phthalate esters and adipate esters at 20 to 50 percent by weight of the formulation) can release on pyrolysis to produce aliphatic and aromatic components in the C8 to C18 range. Some phthalate plasticiser thermal decomposition products (particularly dioctyl phthalate and diisononyl phthalate decomposition fragments) appear in the C9 to C13 n-alkane mass spectrum range. The combination of plasticiser and chlorinated aromatic pyrolysis products from PVC floor covering can produce a chromatographic pattern with both aliphatic and aromatic components, and without a careful comparison sample analysis and chlorine-containing ion profile review (m/z 35/37 chlorine isotope pattern), may be incorrectly attributed to a petroleum product.

Asphalt (bitumen) shingles on roofing and asphalt-based underlayment present a well-documented, serious interferent problem for medium petroleum distillate identification. Asphalt is itself a petroleum product: a heavy residue of crude oil distillation with a complex mixture of aromatic, naphthenic, and paraffinic compounds in the C20 to C60+ range. The chemistry of petroleum products and lubricants explains why the asphalt composition overlaps so closely with kerosene-class ILR markers. When asphalt shingles burn and the lighter components volatilise into the debris below, or when asphalt-based roofing falls into the burn debris during collapse, the resulting GC-MS pattern can closely resemble a medium petroleum distillate or a naphthenic-paraffinic product (E1618 categories 3 and 7). DeHaan and Icove report extensively on asphalt shingle interference as one of the most common sources of contested fire debris interpretations in US casework, and the SWGFEX Type C interferent designation specifically includes asphalt-based building products.

Background subtraction in fire debris analysis is the practice of systematically comparing the extracted ion profiles (EIPs) from the fire debris sample against the EIPs from a comparison sample of the same substrate type collected from an unburned area of the same scene. The objective is to identify chromatographic features that are present in the debris but absent or significantly elevated relative to the comparison, and to determine whether those additional features are consistent with a known ignitable liquid category under E1618.

The execution of background subtraction follows a structured sequence. After GC-MS acquisition of both the case sample and the comparison sample extracts, the analyst generates EIPs at the same set of m/z values for both: typically m/z 57, 71 (alkanes), m/z 91, 105, 119, 128, 142 (aromatic series), m/z 55, 69 (naphthenes/alkenes), and m/z 31, 43, 45 (oxygenates if relevant). The EIPs are displayed at the same intensity scale, aligned on retention time, and compared peak by peak.

Features present at similar intensity in both the case sample and the comparison sample are attributed to substrate background. Features present in the case sample but absent or at significantly lower intensity in the comparison sample are candidate accelerant signals. The analyst then evaluates whether the set of candidate signals, taken together, meets the E1618 criteria for any of the eight ignitable liquid categories.

Three outcomes are possible. First, the candidate signals meet E1618 criteria for a specific category: the analyst issues a positive identification opinion. Second, the candidate signals are present but do not meet the full E1618 criteria for any category: the analyst may issue a qualified opinion noting the presence of ignitable liquid-related compounds insufficient for category classification, or may report "possible" accelerant residue with appropriate qualification. Third, all features in the case sample are fully accounted for by the comparison sample background: the analyst issues a no-ignitable-liquid-identified opinion.

1. Generate EIPs for case and comparison samples at identical scale
   Extract ion profiles at m/z 57, 71, 91, 105, 119, 128, 142, 55, 69 for both the debris sample and the comparison (control) sample. Use the same intensity axis scale so that relative peak heights are directly comparable.
2. Overlay and align on retention time
   Display the case and comparison EIPs overlaid on the same retention time axis. Mark the retention windows for known compound classes. Confirm that both samples used the same GC method and that retention times are comparable.
3. Identify background-only features
   Mark peaks present at equivalent intensity in both sample and comparison as substrate background. These are not further considered in the identification assessment.
4. Identify candidate accelerant signals
   Mark peaks present in the case sample but absent or significantly reduced in the comparison sample. List these by retention time, m/z, and candidate compound identification from the MS library.
5. Apply E1618 pattern criteria to candidates
   Evaluate whether the candidate signals, as a set, meet the ASTM E1618 criteria for any of the eight ignitable liquid categories. Do not classify based on the presence of one or two isolated peaks alone.
6. Document and qualify the opinion
   Record the specific EIPs and peaks that support the classification. Note whether the comparison sample was from the same substrate type and location. State in the report whether the identified pattern is consistent with a fresh or weathered form of the classified product.

The SWGFEX best practice guide (Section 6.3) and the UK Forensic Science Regulator's guidance on fire debris interpretation both require documentation of this process, not just the conclusion. A laboratory report that states only "a medium petroleum distillate was identified" without any documentation of the comparison sample analysis, the EIP comparison, or the E1618 criteria applied is not compliant with either framework. In courts across the US, UK, and Australia, defence experts have successfully challenged fire debris opinions that lacked this analytical documentation, and the resulting case outcomes have ranged from acquittals to retrial orders. The origin determination that directs the investigator to the specific floor area where these substrate samples are collected is described in origin and cause determination.

The complexity of fire debris chromatograms has attracted significant interest in statistical and computational approaches to pattern recognition over the past two decades. The most established of these is principal component analysis (PCA) applied to chromatographic data matrices, developed in the fire debris context principally by the groups of Philip Marsden at the Forensic Science Service (UK), and Claude Roux and colleagues at the University of Technology Sydney, Australia.

PCA reduces a high-dimensional dataset (the intensities at hundreds of retention time points across many samples) to a small number of principal components that capture the major sources of variance. When applied to a dataset of reference ignitable liquid chromatograms plus substrate pyrolysis chromatograms, PCA can generate a scores plot where different ignitable liquid categories cluster in distinct regions and substrate pyrolysis products cluster separately. A case sample projected onto the same PCA space will fall near the cluster most similar to its chromatographic pattern.

The practical limitation of PCA for fire debris analysis is the training set dependency. A PCA model built on reference ignitable liquids and a limited set of substrate pyrolysis profiles will perform well for debris that matches the training conditions and poorly for novel substrates, unusual formulations, or products that were not represented in the training data. The asphalt shingle interferent problem demonstrates this limitation: a PCA model trained on common residential substrates but not on asphalt shingles will misclassify asphalt shingle pyrolysis products as petroleum distillate with high confidence, because the model has no asphalt shingle cluster to assign them to.

Machine learning approaches, including support vector machines (SVMs) and, more recently, convolutional neural networks (CNNs) applied to mass spectra, have been explored for automatic E1618 classification and false-positive detection. The work of Chiaberge and colleagues (Forensic Chemistry, 2020) demonstrated that a CNN trained on a corpus of verified fire debris chromatograms could reproduce E1618 classification with greater than 90 percent agreement with expert analysts. However, all published machine learning tools in this domain share the training-set limitation and none has been validated for case use in a manner that meets OSAC or Forensic Science Regulator standards as of 2024.

In current casework these tools function as decision-support aids, not standalone classifiers. An analyst who uses PCA or ML tools to support pattern recognition must still perform the comparison sample analysis, must still apply E1618 criteria, and must still be able to explain the classification in terms of chromatographic features visible in the raw data. Courts evaluating expert opinion reliability under Daubert (US), Criminal Procedure Rules Part 19 (UK), or equivalent standards will not accept a classification unsupported by the underlying chromatographic analysis.

The most systematically studied repository of fire debris wrongful convictions is the US Innocence Project and the related Conviction Integrity Units in US state attorneys' offices. Fire and arson cases account for a disproportionate share of wrongful convictions relative to their frequency in the criminal case portfolio, and among the factors cited in post-conviction reviews, the misinterpretation of fire debris GC-MS data is present in a significant minority of those cases.

The Gerald Wayne Lewis case in Jacksonville, Florida (fire October 1990, charges dropped 1991) involved fire pattern evidence that investigators attributed to accelerant use. Independent review by fire expert John Lentini, which included a full-scale fire recreation in an identical adjacent house, demonstrated that the burn patterns attributed to accelerant could be produced by an accidental couch fire without any added accelerant. No GC-MS accelerant residue was identified. The charges were dropped before trial. The case is the most cited early example of faulty arson-indicator reasoning, though its primary lesson concerns burn-pattern methodology rather than comparison sample absence.

In Australia, the Farah Jama case (wrongful conviction 2008, acquittal 2009) does not involve fire debris directly, but the contamination principles it exposed prompted the Victorian Institute of Forensic Medicine and Forensic Science Victoria to conduct a systematic review of fire debris protocols, resulting in revised comparison sample requirements in state laboratory SOPs.

In India, the Central Forensic Science Laboratory has conducted workshops since 2018 on substrate interferent identification, following concerns raised in post-trial reviews of arson cases in Maharashtra and Gujarat. The CFSL Hyderabad's forensic chemistry division has published internal guidance (CFSL Technical Bulletin TB-FC-2019-04) on the mandatory collection of comparison samples and the application of background subtraction before issuing accelerant identification opinions.

In the UK, the review of historic arson convictions following the 2011 publication of the Forensic Science Regulator's thematic review of fire investigation evidence (FSR-T-03-01) identified seven cases in which fire debris analytical reports lacked adequate comparison sample analysis. The outcomes ranged from sentence reduction to retrial orders. The review directly informed the current Forensic Science Regulator Codes of Practice and Conduct provisions on fire debris interpretation documentation.