Anal Bioanal Chem (2009) 395:1853–1865 DOI 10.1007/s00216-009-3071-7

ORIGINAL PAPER

Chemical causes of the typical burnt smell after accidental fires Katharina Heitmann & Hubertus Wichmann & Müfit Bahadir

Received: 30 April 2009 / Revised: 13 August 2009 / Accepted: 14 August 2009 / Published online: 4 September 2009 # Springer-Verlag 2009

Abstract The components responsible for the typical burnt smell that occurs after accidental fires (e.g. in buildings) were identified. For this purpose, samples of odorous materials were taken from different real fire sites. Their volatile fractions were analysed by means of thermal desorption, headspace analysis and solid-phase microextraction (SPME) combined with gas chromatography–mass spectrometry (GC/MS). Measurements performed with SPME gave the highest number of analytes as well as the highest signal intensities. A divinylbenzene/carboxen/ polydimethylsiloxane SPME fibre was found to be the most suitable for this task. To distinguish the odour-active compounds from the ca. 1,400 identified volatiles concentrated by SPME, an olfactory detection port was attached to the GC/MS and the column effluent was assessed by panellists. The results revealed that eleven odorous compounds were present in most of the investigated samples: acetophenone, benzyl alcohol, 4-ethyl-2-methoxyphenol, 2hydroxybenzaldehyde, 2-hydroxy-5-methylbenzldehyde, 2-methoxyphenol, 2-methoxy-4-methylphenol, 2methylphenol, 3-methylphenol, 4-methylphenol and naphthalene. Their odour activities were confirmed in additional olfactory experiments, and the relative ratios of these eleven compounds were determined. Based on these ratios, standard solutions that presented an intense odour with typical characteristics of the burnt smell were produced. K. Heitmann was awarded a Lecture Prize for this work at the 19th Doktoranden–Seminar AK ‘Separation Science’, 11–13 January 2009, Hohenroda, Germany. K. Heitmann : H. Wichmann (*) : M. Bahadir Technische Universitaet Braunschweig, Institute of Ecological Chemistry and Waste Analysis, Hagenring 30, 38106 Braunschweig, Germany e-mail: [email protected]

Keywords Accidental fire . Burnt smell . Olfactory detection . GC/MS-O . SPME

Introduction After accidental fire there is usually a typical intense burnt smell that is perceptible at and around the site of the fire and lingers for a long time after insufficient fire damage restoration. It is often apparent that materials which were visibly unaffected by the fire release the odour. The chemicals responsible for this disturbing odour have not yet been systematically determined. However, due to their odour activity and their presence after a fire, there are presumptions about the chemical characteristics of these compounds; for example, they are considered to have moderate polarities and volatilities. Only a few publications have touched on this subject without solving it [1–4]. In practice, odours at accidental fire sites are still assessed by experts based on their subjective impressions. A great deal of stock and many room interiors are disposed of based on their judgements. Furthermore, the lack of knowledge in this area has meant that suitable restoration techniques for quickly reconstructing buildings, interiors and goods and thus preventing waste have not yet been established. The aim of the analytical approach described in this work was to determine the substances with the most potent odours that are present at almost all sites of accidental fire involving heterogeneous fuels (typically residential or business buildings). The volatile fractions of various odorous sample materials collected at real fire sites were concentrated on adsorbents for thermal desorption, by a headspace technique, and by solid-phase microextraction (SPME), although the main focus was on the latter method, which is often used to analyse volatile compounds [5–7].

1854

K. Heitmann et al.

Separation and identification of the volatile compounds were done by means of gas chromatography–mass spectrometry (GC/MS). To distinguish whether a compound is odour active or just volatile, an olfactory detection port was attached to the GC. Some of the column flow was then assessed by panellists using the human nose as the detector. Olfactometry is a technique that is commonly used to identify odour-active compounds [5, 8, 9]. Based on the results obtained, mixtures of compounds were prepared that presented the typical burnt smell. These compounds were identified from among a pool of 1,400 substances that were detected in the samples.

Materials and methods Odorous samples from real fire sites Samples were taken at diverse accidental fire sites. The odorous sample materials were either burnt or untouched by the fire (they only released a burnt smell). Additionally, wipe samples from sooty surfaces were taken with cotton pads wetted with an ethanolic detergent solution. For each sample, data on the course of combustion, the kind of fuel involved, the oxygen supply, the completeness of combustion and the period between fire and collection were noted. The samples were wrapped in aluminium foil and stored in PE bags at −20°C in the dark. Sample materials No. No. No. No. No. No. No. No. No. No. No.

1 2 3 4 5 6 7 8 9 10 11

Wipe sample, sooty Ash, fuel unknown Textile fabric, camel hair, slightly sooty, unburnt Mainly milk powder, almost completely combusted Plastic household item, sooty, unburnt Cable, sooty, unburnt Plaster drill dust, unburnt Textile fabric, probably cotton, unburnt Wood covering, sooty, surface partly burnt Plastic film, slightly sooty, unburnt Paper tissue, slightly sooty, unburnt

All eleven samples included in the investigations were taken at different fire sites in order to obtain a broad range of materials. Sample preparation Before measurements were taken, the sample materials were allowed to thaw in closed bags in order to stop water from condensing on the materials. They were mechanically divided and homogenised if necessary. For the SPME and headspace

experiments, 20 mL vials (Agilent Technologies, Palo Alto, PA, USA) were maximally filled to one-third for SPME and to one half for the headspace experiments, and then crimp-cap sealed with a butyl rubber septum covered with PTFE (Agilent Technologies). Portions of the same mass were used for repeated measurements with one sample type. Additionally, an artificial sample was required for SPME parameter optimisation because the odorous sample materials differed widely. In 20 mL crimp-cap vials, samples of 2 g silica gel (water content 10%) doped with 100 μL of a standard solution containing 100 ng/μL of benzene, naphthalene, benzaldehyde, benzophenone, phenol, 2-methoxyphenol, nonane and hexadecane each in hexane were prepared. Thermal desorption experiments The sample materials (0.1–3 g) were inserted into heated microchambers (μ-CTE Micro-Chamber/Thermal Extraction System, Markes International, Llantrisant, UK). Each chamber had a volume of 45 cm3 and was heated to 80°C and flushed with warmed synthetic air (100 mL/min). After 5 min of stabilisation, the tubes, filled with 300 mg Tenax TE 60/80 and Tenax TA 20/35 and 30 mg CarboTrap 20/40, were connected to the outlets. Volatiles were trapped for 5 min. Afterwards, the tube ends were detached and the tubes were capped for storage. SPME All of the SPME samplings were performed with a manual fibre holder (Supelco, Bellafonte, PA, USA). Three different types of fibre were used: 65 μm polydimethylsiloxane/ divinylbenzene (PDMS/DVB); 50/30 μm divinylbenzene/ carboxen/polydimethylsiloxane (DVB/CAR/PDMS) and 85 μm polyacrylate (PA) (Supelco). Initially, the optimum extraction temperature was determined for each type of fibre between 30 and 110°C in 10 °C steps. Before extraction, the vial containing the sample was heated to the selected extraction temperature for 30 min in a headspace sampler oven (Headspace Sampler 19395A; Hewlett Packard, Palo Alto, CA, USA). Then the septum was pierced with the SPME needle, and the fibre was exposed to the headspace for 30 min. After extraction, the fibre was withdrawn, removed from the vial, and immediately inserted into the heated GC injection port working in splitless mode. Desorption took place at 250°C for PDMS/ DVB, 260°C for DVB/CAR/PDMS, and 280°C for PA. The split flow (45 mL/min) was started after 1.25 min. The fibre remained in the inlet for 30 min for conditioning. After selecting appropriate temperatures for each fibre type, tests were performed to determine the optimum exposure times at these temperatures. Exposure times of between 15

Chemical causes of typical burnt smell after accidental fires

and 60 min (15 min steps) were tested. The general procedure remained as described above. All optimisation experiments were performed threefold. To determine which of the fibres collected the widest range of substances and the greatest amounts of those substances, real samples were investigated (nos. 1–7, see “Sample materials”) with the three types of fibre, and with the optimised parameters applied (see “SPME: parameter optimisation”). Olfactory experiments were conducted with the DVB/CAR/PDMS fibre using the abovementioned sampling parameters and dry sample materials (nos. 4–11). Gas chromatography–mass spectrometry Thermal desorption and analysis were conducted with a thermal desorber (UltrA TD (Autosampler)/Unity, Markes International) coupled to a gas chromatograph (6890 N Network GC Systems, Agilent Technologies) with a massselective detector (5973 Mass Selective Detector, Agilent Technologies) equipped with a DB 5 MS column (60 m length, 0.25 mm inner diameter (i. d.), 0.25 μm film thickness, Agilent Technologies). The tubes were flushed for 0.5 min with a flow of 40 mL/min and then additionally heated to 290°C. Desorption took place for 12 min, and 25 mL/min of the flow were cryofocussed at −10°C in a glass tube two-thirds filled with Tenax TA 60/80 and one-third with Carbopack B. Thereafter, the trap was heated to 300 °C. The analytes were desorbed in counterflow mode for 10 min. The flow of 16 mL/min was transferred into the GC injection port with a split flow of 15 mL/min, giving a column flow of 1 mL/min. The following temperature programme was applied: 32°C, 0.2°C/min to 34°C, 5°C/min to 150°C, 20°C/min to 300 °C (6 min). The transfer line was kept at 280°C. The ion source was operated in EI mode at an energy of 70 eV. The scan range was set to a molecular weight-to-charge ratio (m/z) of between 20 and 250 for 10 min and then between 35 and 450 for the remaining time. The compounds were identified by comparing the MS data with library spectral data (NIST 98). Headspace experiments were carried out with a headspace sampler (Network Headspace Sampler G1888, Agilent Technologies) coupled to a gas chromatograph (6890 Series GC Systems, Agilent Technologies) with a mass-selective detector (5975C inert MSD with Triple-Axis detector, Agilent Technologies) and a DB-1301 column (60 m length, 0.32 mm i.d., 1 μm film thickness, J&W Agilent Technologies). The 20 mL sample vials were equilibrated for 45 min at 100°C. The sample loop held at 110°C was filled by pressurising the vial for 0.08 min with 27 psi of helium. The sample loop was filled for 0.1 min and equilibrated for 0.05 min. The sample was injected

1855

with a helium flow of 4.7 psi through the transfer line (kept at 115°C) for 1 min into the GC injection port working in split mode with a flow of 11 mL/min. The column flow was set to 1 mL/min. Separation was obtained by the following temperature programme: 40°C (5 min), 5°C/min to 260°C (15 min). The transfer line was kept at 280°C. The ion source was operated in EI mode at an energy of 70 eV. The scan range was set to m/z 45–450. Compounds were identified by comparing the MS data with library spectral data (NIST 98), and partly by comparing the retention times and MS data with those of reference substances. SPME and olfactory measurements were conducted with a gas chromatograph (6890 Series GC System, Agilent Technologies) with a mass-selective detector (5973 Network Mass Selective Detector, Agilent Technologies) equipped with a DB-1301 column (60 m length, 0.32 mm i.d., 1 μm film thickness, J&W Agilent Technologies). The injection port was used in splitless mode for 1.25 min and then purged with a 45 mL/min helium flow. The helium column flow was set to 1 mL/min. Compounds were separated with this GC/MS using the following temperature programme: 40°C (5 min), 10°C/ min to 280°C (15 min). The MS parameters were the same as those described above for headspace experiments. The compounds were identified by comparing the MS data with library spectral data (NIST 98), and partly by comparing the retention times and MS data with those of reference substances. For olfactory assessment, the abovementioned GC/MS was additionally equipped with an olfactory detection port (ODP 3, Gerstel, Muelheim a.d. Ruhr, Germany). The column flow was set to 1.3 mL/min and the temperature programme was: 40°C (5 min), 5°C/min to 280°C. The column flow was split 1:1.27 between MS and ODP via a crosspiece. The detection methods were adjusted to operate simultaneously by using a 0.635 m length of 0.1 mm i.d. splitter capillary for the restrictor to the MS and a 0.987 m length of 0.15 mm i.d. splitter capillary for the restrictor to the ODP. Both transfer lines were kept at 280°C. The auxiliary heating at the outlet of the ODP 3 was set to 200°C. A nitrogen make-up gas flow (50 mL/min) was sent to the splitter column outlet to disperse the analyte-containing effluent in the sniffing funnel and enhance uptake. Furthermore, a humidified nitrogen gas flow (10 mL/min) was sent into the sniffing funnel to avoid desiccation of the nasal mucosa. Using the olfactory detection input device, it was possible to record the time at which each odour was perceptible as well as a voice comment. The odorous analytes were identified by superimposing the recorded olfactogram on the GC chromatogram (MS recording in total ion current (TIC) mode) using software.

1856

Panellists A panel of five participants (two females and three males) aged 17 to 27 carried out the olfactory assessment. The panellists were trained in the procedure prior to the analysis, using standard solutions. They were instructed to push a button on the input device when an odour was perceptible and to give a voice comment about the subjective intensity and odour character. Sniffings were performed for a maximum of 35 min, starting at a retention time of 20 min for the real samples after the elution of lowconcentration organic solvents. The retention time period from 40 min to the end (64 min) was also assessed for each sample material in an additional measurement performed by one participant. Blind values with fibres exposed to empty vials were obtained to determine equipment-related odours and the accuracy of the assessors. Additionally, sniffings with standard mixtures (mixtures 1–5, see “Reference substances”) started at 27 min and ended at 42 min of retention time.

K. Heitmann et al.

Hill, MA, USA), 1-methylnaphthalene (Riedel-de-Haën, Seelze, Germany). Standard solutions were prepared by solvating the reference compounds in hexane and/or ethyl acetate (Merck, Darmstadt, Germany). The following standard mixtures were prepared in hexane for the olfactory assessment: Mixture 1 (10 ng/μL): benzyl alcohol, acetophenone, 2-methylphenol, 4-methylphenol and naphthalene Mixture 2 (10 ng/μL): 2-methoxyphenol, 3-methylphenol, 2-methoxy-4-methylphenol, 4-ethyl-2-methoxyphenol, 1methylnaphthalene Mixture 3 (10 ng/μL): 2-hydroxybenzaldehyde, 2hydroxy-5-methylbenzaldehyde, 2-methylnaphthalene Mixture 4 (20 ng/μL): benzyl alcohol, 1methylnaphthalene Mixture 5 (30 ng/μL): benzyl alcohol; (20 ng/μL): 2hydroxy-5-methylbenzaldehyde.

Reference substances

Results and discussion

The following reference compounds were used for confirmation and standard preparation purposes: acenaphthene, acenaphthylene, anthracene, benzophenone, benzyl alcohol, biphenyl, 1,2-dimethylbenzene, 1,3-dimethylbenzene, 1,4-dimethylbenzene, 2,4-dimethylphenol, ethylbenzene, fluorene, hexadecane, 2-methylnaphthalene, 2-methylphenol, naphthalene, nonane, phenol, pyrene, styrene (Supelco), acetophenone, benzaldehyde, 1-ethyl-2methylbenzene, 1-ethyl-3-methylbenzene, 1-ethyl-4-methylbenzene, 2-hydroxybenzaldehyde, 2-methoxyphenol, 2-methylbenzaldehyde, 3-methylphenol, 3,4-dimethylphenol, 4-methylacetophenone, 4-methylphenol (Fluka, Buchs, Switzerland), aniline, dibenzofurane, 1,2-diethylbenzene, 1,3-diethylbenzene, 2-methylpropylbenzene, 1,2,3,4-tetramethylbenzene, 1,2,4,5-tetramethylbenzene, 1,2,3-trimethylbenzene (Accu Standards, New Haven, CT, USA), caprolactam, 2,6-dimethoxyphenol, 3,5-dimethylphenol, 2-furaldehyde, furfuryl alcohol, 2-hydroxy-3-methylbenzaldehyde, 2hydroxy-5-methylbenzaldehyde, 2-methoxy-4-methylphenol, 3-methoxyphenol, 4-methoxyphenol, 2-methylacetophenone, 3-methylacetophenone, 3-methylbenzaldehyde, 4methylbenzaldehyde, 5-methyl-2-furaldehyde, phenanthrene, (±)-α-pinene, 2,3,5-trimethylnaphthalene (Aldrich, St. Louis, MO, USA), decanal, nonanal (Theta, Newtown Square, PA, USA), 1,2-dimethylnaphthalene (Ultra Scientific, Kingstown, RI, USA), 2,3-dimethylphenol, 2,6-dimethylphenol, toluene (Merck, Darmstadt, Germany), 4-ethyl-2methoxyphenol (Alfa Aesar, Ward Hill, MA, USA), 1,2,3, 5-tetramethylbenzene (Alltech, Rottenburg-Hailfingen, Germany), isoquinoline, quinoline (Lancaster, Ward

In order to analyse the burnt odour in the lab, odorous samples were taken from real fire sites. The preconcentration of volatile and potentially odorous compounds was then required. Many different techniques can be used to achieve this, such as extraction and liquid–liquid partitioning, purge and trap, solid-phase extraction and headspace analysis [5–7]. Unfortunately, such methods are often timeconsuming and costly, and some of them do not produce sufficient concentrations for olfactory assessment [5]. Nevertheless, thermal desorption and headspace experiments were conducted to elucidate the substance classes involved and to enhance the analytical quality by applying at least three different methods for one purpose. The headspace technique was found to be particularly useful because it is conducted without any adsorbent, and adsorbents are always selective. The main extraction method applied was solid-phase microextraction (SPME), which is a simple, efficient and solventfree technique. Analytes are concentrated on a coated fibre when it is exposed to the headspace, and they are then thermally released into a GC injection port for separation. Different fibre coatings, ranging from nonpolar to polar, are available commercially. Hence, this method can easily be combined with gas chromatography–olfactometry (GC-O), which is used for olfactory assessment [8, 9]. The human nose is used as a detector of odorous analytes. The GC column flow is passed into a sniffing funnel, dispersed by the make-up gas flow, and assessed by a panellist. Parallel detection using a typical physical detector (like a MSD) is also possible. Using this technique, odour-active compounds can be detected from among all of the analytes present. After

Chemical causes of typical burnt smell after accidental fires

identifying the odorous analytes, the relative proportions of the odorous analytes present in the sample was determined. The reference substances were then combined according to these proportions in order to recreate the artificial burnt smell.

1857 Table 1 Number of detected substances, maximum abundances, and maximum abundances of typical peaks, as obtained using HS-SPMEGC/MS experiments with different fibre types and sample nos. 1–7 (see also “Sample materials”) Sample no.

Number of detected substances Maximum abundance*

Thermal desorption experiments

Maximum abundance of a typical peak*

The volatile components of the sample materials were first investigated by concentrating them on adsorbents and then conducting thermal desorption (see “Thermal desorption experiments”). Compared to the results from SPME (Table 1), the chromatograms obtained showed fewer analytes (between 70 and 100) and most of the analytes exhibited lower abundances. Most of the detected substances were found in the late retention time area—from 35 min to the end—as were the most abundant of the substances. Measurements were performed in order to elucidate the substance classes present in the volatile fractions of the odorous samples. The substances found were mainly nonpolar, like alkanes, alkenes, arenes, benzenes and polycyclic aromatic hydrocarbons (PAH), although some polar classes (like aromatic carbonyl compounds or phenols) were also noted. These analytes were comparable to those detected after various fuels are combusted [10–20]. The results were useful for choosing substances that would optimise the other techniques to be used. Headspace analysis In the headspace technique, the sample is heated in a vial to enhance the release of volatile compounds. An aliquot of the headspace is directly injected into a GC injection port. In contrast to other extraction methods, such as SPME, no sorbent is used. SPME coatings tend to concentrate analytes with similar polarities, and are thus potentially biased against analytes with different polarities to them [21]. Static headspace experiments were also conducted to check that all of the compound classes present in the sample were detected in the SPME experiments. The same samples that were applied for olfactory assessment (nos. 5–11) were analysed by headspace GC/MS. As expected, the headspace measurements showed lower sensitivities, resulting in fewer analytes and lower abundances than those observed in the SPME measurements. Between 30 and 130 analytes were detected. Abundances were at least fivefold lower than those seen in the SPME experiments (Table 1), although these experiments utilised different GC/MS systems, so the headspace and SPME abundances were not directly comparable. However, the results do give a strong indication that the concentrations obtained by this technique were lower. The highest abundances were observed

1

2

3

4

5

6

7

PDMS/DVB

DVB/CAR/PDMS

PA

156 25,000 5,000 198 21,000 6,000 85 1,800 1,000 75 2,400 1,000 106

196 35,000 15,000 240 24,000 12,000 282 23,000 10,000 194 20,000 8,000 222

Not measured

25,000 5,000 59 1,700 800 136 14,000 2,000

52,000 10,000 173 14,000 5,000 208 24,000 5,000

33,000 8,000 116 40,000 3,000 124 2,000 800

214 36,000 10,000 209 25,000 7,000 206 5,300 2,000 202

* Values should be multiplied by factor of 1,000. The abundance values are device specific and are only used to approximately compare the signals that were obtained.

for analytes with short retention times and low boiling points—a known phenomenon for static headspace [22]. In terms of chromatographic properties, the headspace results showed conformity with the substances detected in SPME analyses; no additional substance classes were found. SPME: parameter optimisation Odour-active compounds are volatile, and so the SPME experiments were performed in headspace extraction mode (HS). This is a known technique from aroma analysis [23– 25]. The extraction temperature and the extraction time are parameters that must be optimised to achieve high concentrations for a wide range of analytes [23]. Due to the lack of a homogeneous and universal burnt sample material,

1858

optimisation experiments were carried out with an artificial sample. Four compound classes were selected because they appeared in both the thermal desorption and the headspace experiments; these compounds had polarities ranging from nonpolar to moderately polar. Two representatives with different boiling points from each class were selected: nonane and hexadecane from the alkanes, benzene and naphthalene from the aromatic compounds, phenol and 2methoxyphenol from the phenols, as well as benzaldehyde and benzophenone from the aromatic carbonyl compounds. This approach was used to ensure that the parameters used for concentration were optimised for a wide range of analytes on the fibre. The peak areas of the eight compounds resulting from GC/MS measurements were observed to optimise the temperature and exposure time. The temperature that yielded the maximum amount of analyte before the predomination of thermal desorption and the time required for equilibration were then identified: 70°C and 30 min for PDMS/DVB, 90°C and 45 min for DVB/CAR/PDMS, and 100°C and 45 min for PA. SPME: fibre selection Three commercially available different fibres were chosen for their probable applicability to volatile and moderately polar compounds. When selecting the appropriate fibre for olfactory assessment, the number of analytes concentrated and their abundances were taken into consideration without focusing on particular compounds. Measurements were conducted with seven odorous samples (nos. 1–7, see “Sample materials”). All of the sample materials released many different volatile compounds. Between 60 and 250 different compounds were concentrated, depending on the type of fibre type used and the sample material (Table 2). Various classes of substances were detected—about 1,400 different substances in all—including those detected in the thermal desorption and headspace experiments. The volatile compounds present in the headspace above the sample material could be concentrated on the fibre, although it was not possible to determine whether the sample material contained these volatiles before the fire or whether they were generated during the fire. A comparison revealed that many compounds were sample specific while others could be found in almost all sample types. Typical samplespecific compounds included phthalates in plastic samples or fragrances in textile fabrics. Table 2 shows the nonsample-specific (recurring) substance classes along with typical compounds. Compounds were usually identified based on a MS data library search. Many of these substances have already been described as combustion products in other publications. These product classes can be found after the combustion of fuels such as polyethylene

K. Heitmann et al.

[10], cellulose [11], polyester fabrics [12], polystyrene [13], wood/biomass [14–17], equipment normally found in homes and businesses [18] and various other fuels [19, 20]. When the three different types of SPME fibres were compared, significant differences in the number of analytes detected and their abundances were noted. Table 1 shows, for each sample, the number of substances detected as well as the absolute maximum abundance and the maximum abundance of a typical peaks. There was no noticeable difference in the substance classes detected. The PDMS/ DVB fibre concentrated the fewest analytes and yielded the lowest amounts, resulting in the lowest abundances. The DVB/CAR/PDMS fibre gave the highest number of analytes for almost all sample materials. The DVB/CAR/ PDMS fibre combines two different porous coverings (DVB and CAR), which are embedded into PDMS. Both coatings are moderately polar, and when used in combination they are able to concentrate a wide range of analytes with different molecular sizes and polarities. This fibre was actually developed for the extraction of odorous compounds [24]. In food chemistry it is often used to sample the whole aroma of, say, orange juice or truffles [21, 26]. Using it minimises any potential bias towards particular analytes [21]. The PA fibre yielded a similar or a slightly lower number of compounds than the DVB/CAR/PDMS fibre, but lower abundances for most of the analytes. Therefore, olfactory assessments were performed with the DVB/CAR/PDMS fibre. Olfactory assessments In order to identify the odour-active compounds among all of the volatiles, GC-O measurements were conducted. Seven sample materials (nos. 5–11, see “Sample materials”), selected based on amount and intensity of odours, plus one blank sample were investigated by means of HS-SPME-GC/ MS-O and assessed by five panellists, all of whom were nonsmokers. Training with standard solutions was carried out in advance to familiarise them with the technique. During the GC run, the assessors triggered a signal using the attached input device immediately upon perceiving an odour. A voice comment providing a subjective description of the odour and its intensity was recorded at the same time. GC-O measurements were used to screen for odour-active compounds. The sniffings lasted a maximum of 35 min to avoid participant fatigue. Olfactory measurements started at a retention time of 20 min and ended after 55 min. Before this retention time, only low-concentration organic solvents were detected. The panellists detected between five and 65 different odours and described the retention time window between 30 and 40 min as the region that presented the most intense odours. After 45 min only a few odours had

Chemical causes of typical burnt smell after accidental fires

1859

Table 2 Non-sample-specific substance classes with selected representatives and retention times Substance class

Retention time (min)

Representatives

11.16 14.01

Linear and branched Starting with C6 Linear and branched Starting with C4 Linear and branched Starting with C4 Benzene* Toluene*

16.42 16.25 18.34 16.80 17.05 18.76 22.96 26.29 26.61 27.09 27.15 27.46 28.79 30.11 30.23 31.65 31.78 32.18

1,3-/1,4-Dimethylbenzene* Ethyl benzene* 1-Ethyl-3/4-methylbenzene* Phenylethyne Styrene* α-Methylstyrene Naphthalene* C12H8 C12H10 C13H12 Acenaphthylene* Acenaphthene* Fluorene* C14H12 C16H18 Phenanthrene* Anthracene* C16H12

33.29, 33.42, 33.75 33.75 36.87 38.23 24.92 24.64 26.02 26.20, 26.44 29.30, 29.38, 29.80, 29.87, 29.89, 29.98 16.47 17.10 18.01 19.17 19.44 20.41 27.90 20.20 21.24 21.68

C15H12 C15H10 C16H10 Pyrene* 1-Methylnaphthalene* 2-Methylnaphthalene* Ethylnaphthalene Dimethylnaphthalene Diisopropylnaphthalene 2-Furaldehyde* Furfuryl alcohol* Acetylfurane 5-Methyl-furanaldehyde* Benzofurane Methylacetylfurane Dibenzofurane* Phenol* 2-Methylphenol* 3-/4-Methylphenol*

21.93 22.50, 23.11 22.65 23.11

2,6-Dimethylphenol* Ethylphenol 2,4-/2,5-Dimethylohenol* 3,5-Dimethylphenol*

Alkanes Alkenes Aldehydes Arenes

Benzenes Alkyl benzenes

Other benzenes

PAH

Alkyl naphthalenes

Furans

Phenols Alkyl phenols

1860

K. Heitmann et al.

Table 2 (continued) Substance class

Retention time (min)

Representatives

23.20 23.52 23.97 24.97 21.39 23.05 24.37 25.55 19.05 20.63 21.33

2,3-Dimethylphenol* 3,4-Dimethylphenol* Trimethylphenol t-Butylphenol 2-Methoxyphenol* 2-Methoxy-4-methylphenol* 4-Ethyl-2-methoxyphenol* 2,6-Dimethoxyphenol* Benzaldehyde* 2-Hydroxybenzaldehyde* 4-Methylbenzaldehyde*

N Compounds

22.06 22.61 23.61 21.03 22.81 23.03 13.89 19.56 19.74 24.01 24.43 24.93 25.14

2-Hydroxy-3-methylbenzaldehyde* 2-Hydroxy-5-methylbenzaldehyde* Dimethylbenzaldehyde Acetophenone* 3-Methylacetophenone* 4-Methylacetophenone* Pyridine Aniline* Benzonitrile Quinoline* Isoquinoline* Caprolactam* C9H7N

Others

25.81 28.07, 28.52 19.41 20.64 23.72

Methyl(iso)quinoline Naphthalenecarbonitrile Limonene* Benzyl alcohol* Phenoxyethanol

24.83 25.71 25.89 27.55 27.61 27.98 27.99 29.16 30.28 31.25 32.88 37.29

2,3-Dihydro-1H-indenone Biphenyl* Diphenylether 2H-1-Benzopyran-2-one Bibenzyl Hydroxybiphenyl Isopropyllaurate 1,1′-(1,3-Propanediyl)bis-benzene Isopropylmyristate 9H-Fluorenone Isopropylpalmitate Terphenyl

Methoxy phenols

Benzaldehydes

Acetophenones

* Confirmed by reference substance

Chemical causes of typical burnt smell after accidental fires

1861

been detected, and there was a low correlation between the participants’ sniffing results. The odour descriptions were subjective and thus showed a low correlation, as expected. Objective odour descriptions and intensity assessments are only possible after intensive and time-consuming training [8, 9]. One of the participants also performed an olfactory assessment of the control, starting at a retention time of 40 min and lasting until the run’s end. For three of the samples no odour was perceived during the runs. In another four of the samples three signals were triggered after 55 min at maximum. A comparison of the olfactograms showed no correlation. This provides a strong indication that no odorous compound was missed. In the blank runs, at most two of the participants detected an odour at the same time; therefore, the occurrence of three or more detections at the same retention time was considered an odour signal. The odorous substances were identified by comparing the olfactograms obtained from the sample measurements. As an example, Fig. 1 shows the most odour-active time window (30–40 min) in a chromatogram obtained for sample no. 10. The arrows point to the peaks where an odour was perceived by three or more panellists, and each of these peaks is labelled with the corresponding detected substance (see also Table 3). It is striking that the odour active peaks correspond to substances with medium or low abundances, not high abundances. This observation applied to all of the olfactory measurements, indicating that the key components of the burnt smell have high odour potentials. In Table 3 the retention times marked in three or more sniffings are listed with the correlating analytes for the assessed samples. By comparison of the results shown in Table 3, it was possible to determinate the odorous substances present in minimum three different samples and those present in only Abundance

3000000

naphthalene

4000000

nonanal acetophenone

2-ethylhexanol

5000000

unsaturated ketone

6000000

4-ethyl-2-methoxyphenol

7000000

3-/4-methylphenol methylbenzonitrile 2-methoxyphenol

8000000

2-methylphenol

9000000

2-hydroxy-5-methylbenzaldehyde

1e+07

2-hydroxybenzaldehyde/ benzyl alcohol

Fig. 1 Chromatogram obtained from sample no. 10 after SPME, with the odour-active substances marked

one or two samples. First odour detected in five sample materials was caused by a siloxane, most likely originating from fibre or column material and therefore not fire effect depending and not further considered. The siloxane was detected in lower concentration in blank runs, too. Apart from organic substances formed during accidental fire, the formation of inorganic acids is possible. These would be released during extractions as well and could damage the fibre or column and contribute to an enhanced release of siloxanes. As observed in blank runs, odours were perceived at 28.80 min and 32.90 min indicating that these substances are not related to the fire event. For 28.80 min the MS data were not sufficient for identification and at 32.90 min nonanal was detected. In three samples 2ethylhexanol is marked as odour active. Both, nonanal and 2-ethylhexanol are typical known indoor air compounds [27] and should not be considered in proving burnt smell. Almost all of the remaining detected substances belong to the class of polar substituted aromatic compounds. Odour active compounds present in nearly all investigated samples were as follows, sorted by retention time: 2-hydroxybenzaldehyde, benzyl alcohol, acetophenone, 2-methylphenol, 2-methoxyphenol, 3methylphenol, 4-methylphenol, naphthalene and 4-ethyl2-methoxyphenol. Even if 2-methoxy-4-methylphenol and 2-hydroxy-5-methylbenzaldehyde fulfilled in only two olfactory assessed samples the criteria for olfactory detection, 2-methoxy-4-methylphenol was detected eight times and 2-hydroxy-5-methylbenzaldehyde seven times in the eleven investigated sample materials. This indicates an impact on burnt smell but only if they are present in higher concentrations. Thus, they are taken into consideration for a component in burnt smell. Further substances with only a

2000000 1000000

Time [min]-->

30.00

35.00

40.00

1862

K. Heitmann et al.

Table 3 Substances marked three or more times in olfactory assessments Retention time (min)

Substance

Sample no. 5 6 7 Number of detections

8

9

10

11

21.15 23.90 24.54 26.37 26.50 26.90 28.80 29.90 29.90 30.70

Hexamethylcyclotrisiloxane Chlorobenzene 2-Furaldehyde* C7H12 α-Pinene* Methylcyclopentenone Unknown Benzofurane C6H6O2 2-Ethylhexanol

3 – – – 3 3 3 – – –

4 – – – – – 4 – – –

3 – – – – – – – – 3

– – 3 – – – 4 – – –

– – – – – – – – – –

4 – – – – – 4 3 – 4

3 3 – 4 – – 3 – 4 3

31.00 31.40 32.02 32.12 32.80 32.90 33.24 33.33 33.55 34.05 34.20 34.55 34.70 35.26 35.25 35.30 35.30 35.35

Butanedioic acid dimethyl ester Phenol* 2-Hydroxybenzaldehyde* Benzyl alcohol* Acetophenone* Nonanal* Acetylpyrrole 2-Methylphenol* 2-Methoxyphenol* Methylbenzonitrile 3-/4-Methylphenol* 2,6-Dimethylphenol* N-Ethylbenzamine Si compound Unsaturated aldehyde Chlorobenzonitrile Unsaturated ketone Unknown

– – 3 3 3 3 – – 4 – – – – 3 – – – –

– 3 3 3 4 3 – 5 3 – 3 – – – – – – –

– – 3 3 3 5 – – 5 – 5 – 4 – – – – 4

– – 3 3 3 5 – 3 4 – 3 4 – – 3 – – –

3 – 4 4 4 – 3 3 4 – 5 – – – – – – –

– – 4 4 5 4 – 3 4 3 5 – – – – – 3 –

– – 4 4 3 – – 3 – – 3 – – – – 4 – –

35.68 35.70 36.00 36.16 36.40 36.70 36.98 37.10 37.20 37.70 39.10

Chlorobenzonitrile 2-Hydroxy-5-methylbenzaldehyde* Benzamide derivate Decanal* Naphthalene* 2-Methoxy-4-methyphenol* Unknown Benzoic acid 2,2-Oxybis-1-propanol Unknown 4-Ethyl-2-methoxyphenol*

– – – – – – – 3 – – –

– – – – 5 3 3 – 3 – 4

– – – – 4 – – – – – 3

– – – 3 4 – – – – –

– 4 – – 5 3 – – – –

– 3 – – 4 – – – – –

4 – 3 – 3 – – – – 3

39.15 40.04 44.00 44.02

Unknown 2-Methylnaphthalene* Eugenol Dimethylnaphthalene

4 – – –

– 3 – 3

– – – –

– – – – –

5 – – – –

3 – – 3 –

– – – – –

* Confirmed by reference substances

Chemical causes of typical burnt smell after accidental fires

1863

slight effect seem to be the different dimethylphenols as well as methylnaphthalenes. They were only marked once or twice in olfactory assessments. The mentioned substances were identified by reference substances and their odour activity was confirmed by additionally conducted GC/MS-O measurements with the same panel. Standard mixtures (mixture 1–5, section “Reference substances”) were prepared avoiding close elution or even overlapping of the analytes. Experiments gave hints to distinguish whether in case of close elution occurring in sample measurements both or just one substance were responsible for the perceived odour. Measurements started with a concentration of 10 ng reference substance/μL. If no confirmation was reached mixtures of 20 ng reference substance/μL respectively 30 ng reference substance/μL were prepared and measured. Under consideration of a 1 μL application and the 100:127 (MS:O) splitting in the cross piece, 55.9% of the analytes reached the sniffing funnel. In Table 4 the observed substances and their applied concentration for detection are given. 2-Hydroxybenzaldehyde and benzyl alcohol were eluted from the column within six seconds. While 2hydroxybenzaldehyde was perceptible at a concentration of 10 ng/μL, benzyl alcohol was not detected at the maximum concentration applied, 30 ng/μL, indicating that the odour was mainly caused by 2-hydroxybenzaldehyde. The results also show that the co-eluting 3-methylphenol and 4-methylphenol are both odour active at low concentrations, and both contribute to the smell. Most of the identified substances are known for their odour activities, mainly from aroma analysis, and are mentioned in odour threshold compilations [28, 29]. Some have been described as being detectable after fire events. In particular, the methoxylated and alkylated phenols are

Table 4 Necessary concentration for olfactory detection of the given compounds

known from wood combustion and pyrolysis [14–17]. They are expected to be responsible for the aroma in liquid smoke, a food flavour made from pyrolysed wood [30, 31]. The 2-methoxy-4-alkyl-substituted phenolic structure that some of the substances show probably originates from lignin [17, 30]. This assumption is supported by sample no. 11. None of the 2,4-substituted phenols were detectable when there was no wood in the fire load. These variations in substance ratios and the information associated with them are the subject of ongoing investigations. After accidental fires there are many different detectable volatile compounds, including known odour-active compounds. However, only the strategy employed in this study has made it possible to identify the compounds that are responsible for the typical burnt smell present at almost every site of accidental fire involving heterogeneous fuels that are characteristic of household or business buildings. Artificial burnt smell In [26], a DVB/CAR/PDMS fibre was exposed to the headspace of orange juice. All the sorbed compounds were then released, and the odour was assessed. This aroma probably resembled the original aroma because the volatiles from orange juice were sorbed onto this type of fibre in the same ratio as they were present in the headspace. Inspired by this experiment, the ratios of eleven components were determined. These components were benzyl alcohol, 2-hydoxybenzaldehyde, acetophenone, 2-methylphenol, 2-methoxyphenol, 3-methylphenol, 4-methylphenol, 2hydroxy-5-methylbenzaldehyde, naphthalene, 2-methoxy4-methylphenol, and 4-ethyl-2-methoxyphenol. A calibration curve was generated for each substance by measuring a dilution series from 0.5 to 100 ng/μL, and the slope of this curve was determined. Analyses were based on the peak areas

Retention time (min)

Substance

Applied concentration(ng/μL)

32.02 32.12 32.80 33.33 33.55 34.20 34.20 35.70 36.40 36.70 39.10 39.56 40.04

2-Hydroxybenzaldehyde Benzyl alcohol Acetophenone 2-Methylphenol 2-Methoxyphenol 3-Methylphenol 4-Methylphenol 2-Hydroxy-5-methylbenzaldehyde Naphthalene 2-Methoxy-4-methyphenol 4-Ethyl-2-methoxyphenol 2-Methylnaphthalene 1-Methylnaphthalene

10 up to 30, no detection 10 10 10 10 10 20 10 10 10 10 20

1864

K. Heitmann et al.

Table 5 Characteristic masses and the calculated ratios with standard deviation for sample nos. 5–11 Substance

Masses

Sample no. 5 6 7 Proportion (%) ± standard deviation

8

9

10

11

5.03±0.47 5.50±0.91 47.35±1.17 7.70±0.99 2.99±0.41 2.06±0.17 2.06±0.17 n.d. 21.65±1.93 3.29±0.20

2-Hydroxy-benzaldehyde Benzyl alcohol Acetophenone 2-Methylphenol 2-Methoxyphenol 3-Methylphenol 4-Methylphenol 2-Hydroxy-5-methyl-benzaldehyde Naphthalene 2-Methoxy-4-methyphenol

122/121 108/107/79 120/105/77 107/108 124//109/81 108/107 108/107 136/135/107 128 138/123/95

4-Ethyl-2-methoxyphenol

152/137/122 2.37±0.19

2.92±0.10 1.02±0.05 25.81±2.70 4.71±0.50 2.71±0.30 1.51±0.11 1.51±0.11 0.41±0.05 55.51±2.77 2.49±0.21

1.39±0.11 81.93±0.75 1.98±0.10 5.64±0.27 2.99±0.13 1.32±0.11 1.32±0.11 0.44±0.07 1.41±0.25 1.21±0.06

2.66±0.34 1.89±0.22 3.15±0.20 35.08±2.93 11.03±1.01 6.68±0.46 6.68±0.46 n. d. 22.37±5.10 7.59±0.47

0.83±0.15 0.96±0.09 1.67±0.36 28.01±1.54 12.93±0.99 12.59±0.85 12.59±0.85 1.15±0.20 4.94±0.86 16.18±1.69

2.12±0.49 0.29±0.13 1.24±0.34 63.10±2.47 0.99±0.26 6.41±0.51 6.41±0.51 11.38±0.57 3.63±0.89 2.79±0.52

18.19±2.60 2.07±0.60 13.12±0.67 13.08±1.50 n.d. 3.38±0.66 3.38±0.66 n.d. 46.78±4.55 n.d.

1.38±0.14

0.38±0.04

2.87±0.33

8.14±1.35

1.64±0.29

n.d.

of up to three characteristic masses for each component; these masses are given in Table 5. In order to avoid any influence of the detector response, the resulting mass peak areas from the sample measurements were divided by their corresponding calibration curve slopes. Afterwards, the proportion of each of the eleven components was calculated for each sample. 3-Methylphenol and 4-methylphenol were not separated chromatographically, have the same characteristic masses, and are both odour active. Therefore, they were dealt with in combination, and the calculated ratio was divided by a factor of two. Table 5 lists these proportions. As expected, a comparison did not elucidate any general ratio. Fire is a complex phenomenon that is influenced by many parameters, such as the fuel composition, the temperature and the oxygen supply. Naturally, changes in combustion parameters lead to different combustion product ratios. As already mentioned, the formation of 2,4-substituted phenols depends on the presence of lignin; if lignin is not present they are not formed and cannot contribute to the burnt smell. From aroma analysis it is known that it is possible to recreate an aroma artificially by recombining the individual compounds [25]. In order to create an artificial burnt smell, standard solutions of the abovementioned eleven compounds were combined in the proportions determined for each sample material. Of course, this is a general approach, using only the main compounds responsible for the burnt smell and likely neglecting the influences of other compounds that are responsible for characteristic nuances. Nevertheless, all of the mixtures presented intense odours with burnt characteristics. In particular, the recreated odours for sample nos. 9 and 10 very closely resembled the original burnt smells. The recreated odour for sample no. 9 had a sweet, phenolic note, and that for sample no. 10 had a smoky note.

Conclusions Currently, the management of accidental fire damage is based on simplified assumptions and models. On the one hand, some knowledge of the smoke distribution has been obtained from experimental work or software simulations [32]. Pollutants have been of minor or no interest. On the other hand, a few markers that are known to occur in fire events and which have well-known properties have been chosen in order to assess fire damage and contamination. These procedures are based on results obtained between 1980 and 1995 [33, 34]. Apart from a few acids, these marker analytes are mainly polycyclic aromatic hydrocarbons and polychlorinated dibenzo-p-dioxines and -furanes, which are predominantly nonpolar organic substances with high boiling points and similar distributions and environmental properties. Particles adsorb them well, and they spread with the particles. The removal of visible deposition is usually regarded as successful decontamination. Nowadays this procedure appears insufficient, because that an intense and long-lasting burnt smell is also present. It is remarkable that materials that are spatially separated from the actual site of the fire and are visibly unaffected by it, and which are not contaminated with marker pollutants, often release an intense odour. This indicates that substance classes with different distribution properties are involved. The work presented here on the main component mixture that produces the burnt smell will be used to develop a practical analytical method of detecting the burnt smell, which can then be used to obtain a spatial impression of a contamination situation. These results should also contribute to the development of new decontamination methods.

Chemical causes of typical burnt smell after accidental fires Acknowledgements The work was supported by a scholarship from Deutsche Bundesstiftung Umwelt (DBU). Our thanks go out to the five panellists: Mario Argentari, Thorben Nawrath, Josefine Ohnesorge, Nancy Paetsch and Thore Reimer.

References 1. Maekawa M, Nohmi T, Inoute T, Higuchi A (2003) Significance of odor sensing and roasted smell of wood. J Mass Spectrom Soc Japan 51:187–195 (in Japanese) 2. Adamczyk Z, Bialecka B (1999) Air pollution in an area of coal waste dumping as illustrated by a burning waste heap. Ochr Powietrza Probl Odpadów 33:193–198 (in Polish) 3. Endo S (1996) Study of odor substances in combustion gas from unvented kerosene space heaters. Taiki Kankyo Gakkaishi 31:30– 42 (in Japanese) 4. Meakawe M, Nohmi T (1994) Burning smell of cellulose and synthetic polymers. Nippon Aji to Noio Gakkaishi 1:448–451 (in Japanese) 5. Sides A, Robards K, Helliwell S (2000) Development in extraction techniques and their application to analysis of volatiles in foods. Trends Anal Chem 19:322–329 6. Richter J, Schellenberg I (2007) Comparison of different extraction methods for the determination of essential oils and related compounds from aromatic plants and optimization of solid-phase microextraction/gas chromatography. Anal Bioanal Chem 387:2207–2217 7. Mallia S, Fernandez-Garcia E, Bosset JO (2005) Comparison of purge and trap and solid phase microextraction techniques for studying the volatile aroma compounds of three European PDO hard cheeses. Int Dairy J 15:741–758 8. van Ruth SM, O’Connor CH (2001) Influence of assessors’ qualities and analytical conditions on gas chromatography– olfactometry analysis. Eur Food Res Technol 213:77–82 9. Delahunty CM, Eyres G, Dufour JP (2006) Gas chromatography– olfactometry. J Sep Sci 29:2107–2125 10. Font R, Aracil I, Fullana A, Conesa JA (2004) Semivolatile and volatile compounds in combustion of polyethylene. Chemosphere 57:615–627 11. Sakuma H, Munakata S, Sugawara S (1981) Volatile products of cellulose pyrolysis. Agric Biol Chem 45:443–451 12. Molto J, Font R, Conesa JA (2006) Study of the organic compounds produced in the pyrolysis and combustion of used polyester fabrics. Energy Fuels 20:1951–1958 13. Hu YH, Han X, Tan R, Li SF (2007) Effects of magnesium hydroxide on combustion products of polystyrene. Chin J Chem Phys 20:185–190 14. Kjällstrand J, Ramnäs O, Petersson G (2000) Methoxyphenols from burning of Scandinavian forest plant materials. Chemosphere 41:735–741 15. Conde FJ, Afonso AM, Gonzales V, Ayala JH (2006) Optimization of an analytical methology for the determination of alkyl- and methoxy-phenolic compounds by HS-SPME in biomass smoke. Anal Bioanal Chem 385:1162–1171

1865 16. Re-Poppi N, Santiago-Silva MR (2002) Identification of polycyclic aromatic hydrocarbons and methoxylated phenols in wood smoke emitted during production of charcoal. Chromatographia 55:475–481 17. Simoneit BRT (2002) Biomass burning—a review of organic tracers for smoke from incomplete combustion. Appl Geochem 17:129–162 18. Almirall JR, Kenneth GF (2004) Characterization of background and pyrolysis products that may interfere with the forensic analysis of fire debris. J Anal Appl Pyrolysis 71:51–67 19. Lemieux PM, Lutes CC, Santoianni DA (2004) Emissions of organic air toxics from open burning. A comprehensive review. Prog Energy Combust Sci 30:1–32 20. Adebona BE, Chawla RC, Martin EJ, Wheeler JW (1998) Organic products of incomplete combustion of colored bags and inks. J Hazard Mater 60:57–72 21. Diaz P, Senorans J, Reglero G, Ibanez E (2002) Truffle aroma analysis by headspace solid phase Microextraction. J Agric Food Chem 50:6468–6472 22. Stevenson RJ, Chen XD, Mills OE (1996) Modern analyses and binding studies of flavour volatiles with particular reference to dairy protein products. Food Res Int 29:265–290 23. Pawliszyn J (1997) Solid phase microextraction: theory and practice. Wiley-VCH, New York 24. Pawliszyn J (1999) Application of solid phase microextraction. Royal Society of Chemistry, Cambridge 25. Belitz HD, Grosch W, Schieberle P (2006) Lehrbuch der Lebensmittelchemie, 6th edn. Springer, Berlin 26. Rega B, Fournier N, Guichard E (2003) Solid phase microextraction (SPME) of orange juice flavour: odor representativeness by direct gas chromatography olfactometry (D-GC-O). J Agric Food Chem 51:7092–7099 27. Salthammer T (1999) Organic indoor air pollutants—occurance, measurement, evaluation. Wiley-VCH, Weinheim 28. Ryschlik M, Schieberle P, Grosch W (1998) Compilation of odor thresholds, odor qualities and retention indices of key food odorants. Deutsche Forschungsanstalt für Lebensmittelchemie, Garching 29. Devos M, Patte F, Rouault J, Laffort P, van Gemert LJ (1990) Standardized human olfactory thresholds. Oxford University Press, Oxford 30. Simon R, de la Calle B, Palme S, Meier D, Anklam E (2005) Composition and analysis of liquid smoke flavouring primary products. J Sep Sci 28:871–882 31. Guillén MD, Manzanos MJ, Zabala L (1995) Study of a commercial liquid smoke flavouring by means of gas chromatography/mass spectrometry and fourier transform infrared spectroscopy. J Agric Food Chem 43:463–468 32. McGratten KB, Baum HR, Rehm RG (1998) Large eddy simulations of smoke movement. Fire Saf J 30:161–178 33. Marklund S, Andersson R, Tysklind M, Rappe C (1989) Emissions of PCDDs and PCDFs from PVC-fire in Holmsund, Sweden. Chemosphere 18:1031–1038 34. Wichmann H, Lorenz W, Bahadir M (1995) Release of PCDD/F and PAH during vehicle fires in traffic tunnels. Chemosphere 31:2755–2766