ANALYTICAL SCIENCES NOVEMBER 2010, VOL.

26 1127
2010 © The Japan Society for Analytical Chemistry

Original Papers

Novel Fire Investigation Technique Using Needle Extraction in

Gas Chromatography
Ikuo UETA,* Yoshihiro SAITO,*† Kenta TERAOKA,* Hisashi MATSUURA,** Koji FUJIMURA,** and
Kiyokatsu JINNO*

*Department of Environmental and Life Sciences, Toyohashi University of Technology,

Toyohashi 441–8580, Japan
**Shinwa Chemical Industries, Ltd., Kyoto 612–8307, Japan

A novel fire investigation technique using a needle extraction device was studied. Using a polymer particle-packed needle
device, air samples containing volatile organic compounds (VOCs) generated from fire accelerants, gasoline and kerosene
were extracted effectively, and subsequent gas chromatographic (GC) analyses were successfully carried out. Carpet and
wood samples were spiked with gasoline and kerosene, followed by monitoring of the time-variation profiles of emitted
VOCs up to 48 h. The fire accelerants were also measured for combusted carpet and wood samples, and the applicability
of the proposed method to fire investigations was confirmed. Even at 48 h after spiking, groups of characteristic
compounds were clearly observed in the air environments near the combusted sample. This method was further applied
to the determination of VOCs in simulated fires, strongly suggesting the applicability of the developed method to real fire
investigations.

(Received September 22, 2010; Accepted October 8, 2010; Published November 10, 2010)

require longer sampling/desorption times in addition to the use

Introduction
Among the approximately 60000 fires that occur in Japan every
year, arson represents the worst cause. In the case of arson, fire
accelerants, such as gasoline and kerosene, are often used to
start fires. Because these fire accelerants consist of mixtures of
volatile organic compounds (VOCs), accurate and sensitive
determinations of the VOCs found at the scenes of fires can be
quite important in forensic investigations, especially if arson is
suspected from fire debris.1–3 In general, the objective of
determining the VOCs in a fire investigation is to confirm the
presence of fire accelerants and the origin of a fire. For
determining fire-related VOCs, the analysis in gas
chromatography (GC) is one of the most promising methods,
mainly because of its selectivity and sensitivity, as well as wide
availability.3,4 In fires, fire accelerants are subject to high
temperatures, with the result that a large proportion of VOCs is
evaporated. The concentrations of VOCs emitted from typical
fire debris are often low, and therefore some kind of sample
preconcentration process is necessary before analyses are
performed, as is typically done in GC, taking into consideration
any possible interferences from pyrolysis products.5,6
Several sample preconcentration techniques have been
developed for GC analyses of fire accelerant-related VOCs
emitted from combusted samples, such as those using activated
carbon7 and Tenax TA.4–6 However, these techniques normally
†
To whom correspondence should be addressed.
E-mail: saito@ens.tut.ac.jp
I. U. present address: Department of Applied Chemistry,
Interdisciplinary Graduate School of Medical and Engineering,
University of Yamanashi, Kofu 400–8511, Japan.
of specific instruments. Solid-phase microextraction (SPME)
has several attractive features over traditional sample-preparation
techniques, and this has also been applied to the determination
of fire accelerants.8 Needle-type sample preparation is an
alternative preconcentration technique for analyses of VOCs.9–16
In our previous work, a polymer particle-packed extraction
needle showed both good extraction performance and
satisfactory sample storage capabilities with respect to typical
VOCs,10 and needle extraction devices were successfully applied
to analyses of environmental tobacco smoke12,17 and human
breath.18
In this work, a new method for fire investigations has been
developed using a polymer particle-packed extraction needle
and subsequent GC analysis. The time-variation profiles of
VOCs emitted from fuel-spiked samples were observed up to
48 h. VOCs emitted from combusted wood and carpet samples
were also measured, and the proposed method was further
applied to the determination of fire accelerants in simulated
fires.
Experimental
Materials
All of the reagents and solvents were of analytical reagent
grade and obtained from either Wako Pure Chemical (Osaka,
Japan) or Tokyo Chemical Industries (Tokyo, Japan). All of the
fabric (cotton), wood (untreated Japanese cypress, pine and
plywood) and carpet samples (3 kinds of acrylic carpets) used in
this work were commercially available, and purchased from a
local store (Toyohashi, Japan). Standard gaseous samples were
prepared in Tedlar bags (GL Sciences, Tokyo, Japan). A vacuum
1128
Fig. 1 Schematic representation of the determination procedure for VOCs from fire accelerant-
spiked samples and combusted samples using a needle extraction device.
sampling device for needle-extraction procedures was purchased
from Komyo Rikagaku Kogyo (Tokyo, Japan).
Extraction and desorption of VOCs
As the extraction medium, a polymer of divinylbenzene
(DVB) was synthesized, and a copolymer of methacrylic acid
(MA) and ethylene glycol dimethacrylate (EDMA)10,17 was also
prepared for a comparison. The DVB and MA/EDMA extraction
media were designed and synthesized in order to achieve both
the successful extraction of VOCs that are typically found in
environmental air samples at room temperature, and successful
desorption of preconcentrated VOCs in a conventional GC
injector without special modifications. The synthetic schemes
for these materials were similar to those described in previous
studies.10,17
These copolymer beads with diameters ranging from 150 to
180 μm were packed into specially prepared needles with the
following dimensions: 0.5 mm i.d., 0.7 mm o.d.; length,
85 mm.10 For extractions, needles were attached to a
commercially available vacuum sampling device. Gaseous
samples were introduced into extraction needles by manually
pulling the handle of the vacuum sampling device, and analytes
were adsorbed on the polymeric particles therein. The typical
sampling volume was 50 mL, and it took approximately 5 min
to complete the sampling. In the case that the level of the target
compounds is expected to be higher than the maximum loading
capacity of the extraction device, the sampling volume can be
reduced to 10 mL in order to avoid a possible breakthrough of
the samples.
After extraction, extraction needles were attached to an
injection syringe containing 1.0 mL of N2 gas. Extraction
needles were then inserted into a heated GC injection port, and
the analytes were injected using the N2 gas, where the needles
were pre-heated for 10 s in the injector before injection. It was
confirmed in preliminary experiments that the pre-heating time
could help realize a quantitative desorption of the analytes.
Determination of fire accelerant-related VOCs using needle
extraction devices
In this work, fire accelerant-related VOCs were extracted
using two types of polymer particle-packed needle devices, and
ANALYTICAL SCIENCES NOVEMBER 2010, VOL. 26
determined by GC with a mass spectrometer (GC-MS). Figure 1
shows the scheme for the extraction and desorption process for
VOCs used in this work. A fire accelerant of 1.0 mL was spiked
dropwise onto one side of a wood sample with the dimensions
of 2.5 cm (width) × 10 cm (length) × 1.5 cm (thickness). The
wood sample was then combusted using a gas burner, with the
spiked side being placed downwards to be directly subjected to
the burner flame. After combustion, the wood sample was
placed in a glass vessel with a volume of 250 mL. After
allowing the interior of the vessel to equilibrate for 30 min, an
extraction needle was inserted into the vessel, and 50 mL of the
gaseous sample was extracted. During extraction, pure N2 gas
was continuously supplied from a gas sampling bag, as shown
in Fig. 1.
Five-centimeter squares of carpet were also spiked with fire
accelerants and combusted. When spiked with gasoline and
kerosene, the flame was automatically extinguished after about
1 and 2 min, respectively. Thirty minutes after extinction, air at
a height of approximately 10 mm above the carpet was extracted
using an extraction needle. Fire accelerants (1.0 mL each) were
spiked on wood and carpet pieces of the same size as mentioned
above, and on a 10-cm square of cotton fabric. The time-variation
profiles of VOCs were also monitored in the air above these
samples.
The detection of fire accelerants was carried out for two cases
of simulated fires. The rooms used in these simulations were
prefabricated rooms with volumes of 15.1 m3, 2.7 m (width)
× 2.0 m (height) × 2.8 m (depth), and were made to represent
the typical interior of a small Japanese room in winter, except
for the height of the ceiling. The first case was assumed to be a
fire that resulted from an accident involving a kerosene heater
with a tank capacity of approximately 4 L. The second case
was assumed to be an arson fire in which approximately 4 L of
kerosene was poured onto the floor before ignition. Thirty
minutes after extinction of the fires using excess water,
room  air  was extracted using extraction needles at a height of
approximately 10 cm above the floor, and the extracted VOCs
were determined as described below.
GC-MS measurements
For all GC-MS measurements, either an Agilent 6890N gas
ANALYTICAL SCIENCES NOVEMBER 2010, VOL. 26
Table 1 Desorption efficiencies for typical compounds in fire
accelerants
Desorption efficiency, %
Desorption
Extraction 1,2,4-
temperature/
medium
°Ca Heptane Trimethyl- Dodecane Tridecane
benzene
MA/EDMA 200 99.3 99.4 90.8 87.1
DVB 280 99.4 99.1 99.8 99.9
a. Optimum desorption temperature confirmed in this work.
chromatograph (Yokogawa Analytical Systems, Musashino,
Japan) with a split/splitless injection port and an HP 5972A
mass selective detector (MSD), or a GCMS-QP2010 Plus
(Shimadzu, Kyoto, Japan) were used. All of the injections were
performed using a split mode with a ratio of 15:1. GC
separations were performed on an HR-1 fused-silica capillary
column with dimensions of 30 m × 0.25 mm i.d., and a film
thickness of 0.25 μm (Shinwa Chemical Industries, Kyoto,
Japan). As the carrier gas, He was used at a typical head
pressure of 30 kPa, and the injector temperature was maintained
at 200 or 280° C, as described below. The GC-MS interface
temperature and the ionization voltage were set at 250° C and
70 eV, respectively, and electron impact ionization was used.
The column temperature was started at 40° C for 2.0 min, and
then programmed to 240° C at a rate of 20° C/min. The mass
spectrometer was operated in either a total ion monitoring mode
(TIM) with an m/z range of 45 to 250, or a selected ion
monitoring mode (SIM). Data collection was performed using
Chrom NAV (Jasco, Tokyo, Japan) on a personal computer.
Results and Discussion
Optimization of the method
Although various compounds are found in gasoline and
kerosene, the main components of these fire accelerants are
aromatic compounds and normal alkanes,4 as was also confirmed
in our preliminary experiments, where determinations of
headspace gases were performed using a GC-MS in a TIM
mode. Based on these results, an SIM mode consisting of m/z
43, 57, 71, 78, 91 and 105 was chosen for sensitive determinations
of fire accelerant-related VOCs, and the sum of these ions was
monitored. By introducing an SIM mode, interfering peaks
generated by pyrolysis products were eliminated, allowing for
accurate determinations of target VOCs related to the fire
accelerants.
Successful extractions and desorptions of typical VOCs had
been performed previously using extraction needles packed with
a copolymer of MA and EDMA.10,17 This extraction medium
showed a wider selectivity for both polar and non-polar analytes,
and therefore extraction needles packed with MA/EDMA were
employed as the first choice of the extraction medium in this
study. The desorption temperature was set at 200° C based on
preliminary experiments, and desorption efficiencies were
measured for each compound using a similar way as described
previously,10 where the desorption efficiencies were calculated
based on the ratios of the peak areas obtained in the first
desorption to the total peak areas obtained by adding the peak
areas obtained in the first and second desorptions. The
desorption efficiencies of VOCs typically found in the fire
accelerants are given in Table 1. These results indicated that
1129
Table 2 LOD and LOQ for typical VOCs found in fire
accelerants
Benzene Toluene Octane Decane
LOD/ng L –1
0.1 0.1 0.03 0.03
LOQ/ng L–1 0.3 0.3 0.1 0.1
successful desorptions were achieved for volatile compounds.
However, somewhat unsatisfactory desorption efficiencies for
compounds having a low volatility, such as tridecane, were
observed, probably due to a low desorption temperature.
Based on the above results, a polymer of DVB was introduced
as an alternative extraction medium in this work. Compared
with the MA/EDMA material, the DVB material has a higher
thermal stability, which enables extraction needles to be
subjected to higher desorption temperatures. Taking advantage
of the good thermal stability of the DVB material, the desorption
temperature for DVB-packed needles was set at 280° C. As
summarized in Table 1, the desorption efficiencies were
significantly improved, where the desorption efficiencies for
most of the VOCs found in fire accelerants were higher than
95%. Although the correlation of the polymer structure to the
resulting extraction selectivity/efficiency should be studied in
more detail, the results clearly demonstrate the excellent
performance of the DVB material as an extraction medium for
these fuel-related compounds, and therefore DVB-packed
extraction needles were used in the following experiments.
In order to confirm the sensitivity and repeatability of the
present method, several standard gaseous samples were prepared
using the same procedure as previously described.10 The limits
of detection (LOD) and the limits of quantification (LOQ) for
typical fire-related VOCs are summarized in Table 2. The upper
limits of the calibration ranges were more than 500 μg/L for all
of the investigated VOCs. The RSDs of the peak areas were less
than 3.0% (n = 5) for all of the standard samples. Despite the
short extraction time of approximately 5 min, the results clearly
demonstrated low detection limits and wide linear ranges for
both aromatic and aliphatic compounds typically found in fire
accelerants. In Table 2, quite similar detectability for both
benzene and toluene can be observed, as found in our preliminary
investigation for xylenes and ethylbenzene. Additionally, the
detectability for octane and decane given in Table 2, and that for
hexane and heptane (data not shown) were approximately the
same. Therefore, it can be considered that most of the VOCs
found in the fire accelerants could be quantitative at the ng/L
level, suggesting a good possibility for the successful
determination of typical fire-related VOCs in air samples
obtained during fire investigations.
Time-variation profiles of the fire accelerants
In fire investigations, the detection of fire accelerants from
uncombusted materials at a fire scene is one of the clearest
evidence for the presence of fire accelerants. The time-variation
profiles of VOCs evaporated from fuel-spiked samples of cotton
fabric, carpet and wood, were measured for up to 48 h. Figure 2
illustrates the time-variation profiles for the separations of
groups of relatively low-volatility compounds, demonstrating
the successful detection of these fuel-related compounds, as
well as fingerprint patterns. For a group of relatively volatile
compounds, such as 3-methylhexane in gasoline and octane in
kerosene, measurements of the time-variation profiles were
much easier, but only possible for up to 12 h after spiking in
order to obtain satisfactory detectivity. Compared with the
1130 ANALYTICAL SCIENCES NOVEMBER 2010, VOL. 26

Fig. 2 Time-variation profiles of VOCs from samples spiked with fire accelerants. (A) carpet and
(B)  wood spiked with (a) gasoline and (b) kerosene. Peaks: (i) 1-ethyl-2-methylbenzene, (ii)
1,2,4-trimethylbenzene and (iii) 1,2,3-trimethylbenzene. Conditions: sampling volume, 50 mL; oven
temperature, 40°C (2 min) to 240° C at 20°C/min; monitoring mode, SIM (m/z: 43, 57, 71, 78, 91 and
105). Other conditions can be found in the text.

Detection of fire accelerants from combusted samples

Fig. 3 Chromatograms for the separations of VOCs from combusted
carpet samples. (A) Gasoline-spiked and (B) kerosene-spiked. Peaks:
(a) o-xylene, (b) 1-ethyl-2-methylbenzene, (c) 1,2,4-trimethylbenzene
and (d) 1,2,3-trimethylbenzene. Conditions are the same as in Fig. 2.
evaporation profile of wood and carpet samples, all of the fire
accelerants spiked to cotton fabric sample were more rapidly
evaporated, and VOCs were detected for up to 24 h (data not
shown).
Since the detection of fire accelerants evaporated from various
combusted substances can be regarded as being an important
approach in forensic analysis, VOCs generated from fire debris
were also studied in this work. Figure 3 shows typical
chromatograms of VOCs emitted from combusted carpet
samples. Although the surfaces of the carpet samples and the
accelerants thereon were mostly combusted, the accelerants that
had penetrated into the carpet samples were successfully
detected. Three types of carpet were tested in this combustion
study, and similar sets of VOCs, clearly indicating the existence
of the spiked fuels, were detected in all cases.
For the detection of VOCs from combusted wood samples,
wood samples spiked with fire accelerants were also combusted.
To ensure successful determinations of trace amounts of VOCs,
combusted wood samples were placed in a glass vessel, and
determinations of VOCs in headspace gases were performed.
As shown in Fig. 4, the peak areas of representative gasoline
(1-ethyl-2-methylbenzene) and kerosene (undecane) compounds
were monitored with changes in the degree of combustion of the
wood samples. The representative fire accelerant VOCs were
detected in Japanese cypress up to weight losses of approximately
30%, and approximately 50% in both pine and plywood samples.
In order to confirm the suitability of the method to actual fire
investigations, excess of water was spiked into the combusted
wood samples, and VOCs in the headspace gas were measured.
ANALYTICAL SCIENCES NOVEMBER 2010, VOL. 26 1131

Fig. 4 Effect of the degree of combustion on the observed amounts of (A) 1-ethyl-2-methylbenzene
and (B) undecane as representative compounds of gasoline and kerosene, respectively. (a) Japanese
cypress, (b) pine and (c) plywood. Conditions are the same as in Fig. 2.

The results indicated that almost the same amounts of VOCs

were observed for all wood samples with and without water
spiking. These results suggest that the proposed method is
suitable for the determinations of trace amounts of accelerants
emitted from combusted wood and carpet.

Simulated fire analyses

On the basis of the above successful extractions and subsequent
determinations of fuel-related VOCs from spiked and combusted
spiked samples, the proposed method was further applied to the
determination of fire accelerants in simulated fires using two
prefabricated rooms. The fire experiments were carried out at
the Hyogo Prefectural Emergency Management and Training
Center (Miki, Hyogo, Japan), where professional training
programs for firefighter students are conducted using actual
model rooms. During this experiment, each room was engulfed
in flames, and all of the windows and aluminum window frames Fig. 5 Typical chromatograms for the separation of the air samples
were melted. After the power of each fire had decreased containing kerosene-related VOCs obtained after simulated fires.
somewhat from the maximum, each fire was extinguished by a Simulation of (A) an accident involving a kerosene heater and (B) an
arson fire. Conditions: sampling volume, 100 mL; oven temperature,
group of students under the guidance of an instructor. Room air
50°C (5 min) to 250° C (5 min) at 10° C/min; monitoring mode, TIM
was extracted using extraction needles 30 min after extinction, (m/z: 50 – 250).
and the extracted VOCs were analyzed in GC-MS.
Figure 5 shows typical chromatograms for separations of the
air samples taken from the two types of simulated fire accidents.
These chromatograms were obtained using a TIM mode with an interferences from fire debris. VOCs characteristic of kerosene,
m/z range of 50 and 250 to confirm the absence of any possible such as nonane, decane and undecane, were clearly found as
1132
fingerprint patterns of kerosene in both air samples taken from
the simulated fires, and no significant interferences from
pyrolysis products were observed.
Although further studies should be undertaken involving more
precise investigations of actual fire scenes, the results of this
work demonstrated the excellent detection capabilities of the
proposed method with regard to confirmations of the presence
of kerosene in fire investigations.
Conclusions
A new method for fire investigations was successfully developed
based on a polymer particle-packed extraction needle. The
proposed method allows simple and sensitive determinations of
trace amounts of fire accelerants commonly found at the scenes
of fires, and the method could be a promising technique for the
confirmation of the fire accelerants and the origin of a fire. The
practical applicability of this method has been confirmed using
systematic experiments in this work. In order to obtain suitable
extraction capabilities with respect to other target compounds,
the polymer structure can be further optimized together with
optimization of the corresponding extraction/desorption and
separation conditions. These studies are currently being
continued in our laboratory, and include the development of a
novel separation column19 for these applications.
Acknowledgements
A part of this research was financially supported by a Grant-in-
Aid for Scientific Research C (No. 21550078) from the Japan
Society for the Promotion of Science (JSPS). The authors
would like to acknowledge the valuable cooperation of the
Hyogo Prefectural Emergency Management and Training
Center, Miki, Hyogo, Japan. Technical support from Mr. N.
Yamamoto, Mr. S. Ellwood and Dr. H. Wada of Shinwa
Chemical Industries is also acknowledged. One of the authors,
I. U., was financially supported by a Grant-in-Aid for JSPS
Fellows.
ANALYTICAL SCIENCES NOVEMBER 2010, VOL. 26
References
1. J. Dolan, Anal. Bioanal. Chem., 2003, 376, 1168.
2. J. R. Almirall and K. G. Furton, J. Anal. Appl. Pyrolysis,
2004, 71, 51.
3. S. Baechler, S. Comment, and O. Delémont, Talanta, 2010,
82, 1247.
4. R. Borusiewicz, G. Zadora, and J. Zieba-Palus,
Chromatographia, 2004, 60, S133.
5. M. S. Fernandes, C. M. Lau, and W. C. Wong, Sci. Justice,
2002, 42, 7.
6. R. Borusiewicz, J. Z. Palus, and G. Zadora, Forensic Sci.
Int., 2006, 160, 115.
7. A. Jayatilaka and C. F. Poole, Chromatographia, 1994, 39, 200.
8. K. G. Furton, J. Bruna, and J. R. Almirall, J. High Resolut.
Chromatogr., 1995, 18, 625.
9. J. Lipinski, Fresenius’ J. Anal. Chem., 2001, 369, 57.
10. Y. Saito, I. Ueta, K. Kotera, M. Ogawa, H. Wada, and K.
Jinno, J. Chromatogr., A, 2006, 1106, 190.
11. Y. Saito, I. Ueta, M. Ogawa, and K. Jinno, Anal. Bioanal.
Chem., 2006, 386, 725.
12. Y. Saito, I. Ueta, M. Ogawa, M. Hayashida, and K. Jinno, J.
Pharm. Biomed. Anal., 2007, 44, 1.
13. M. Ogawa, Y. Saito, I. Ueta, and K. Jinno, Anal. Bioanal.
Chem., 2007, 388, 619.
14. Y. Saito, I. Ueta, M. Ogawa, A. Abe, K. Yogo, S. Shirai,
and K. Jinno, Anal. Bioanal. Chem., 2009, 393, 861.
15. I. Ueta, Y. Saito, N. B. A. Ghani, M. Ogawa, K. Yogo, A.
Abe, S. Shirai, and K. Jinno, J. Chromatogr., A, 2009, 1216,
2848.
16. D. W. Lou, X. Lee, and J. Pawliszyn, J. Chromatogr., A,
2008, 1201, 228.
17. I. Ueta, Y. Saito, K. Teraoka, T. Miura, and K. Jinno, Anal.
Sci., 2010, 26, 569.
18. I. Ueta, Y. Saito, M. Hosoe, M. Okamoto, H. Ohkita, S.
Shirai, H. Tamura, and K. Jinno, J. Chromatogr., B, 2009,
877, 2551.
19. M. Inoue, Y. Saito, I. Ueta, T. Miura, H. Ohkita, K.
Fujimura, and K. Jinno, Anal. Sci., 2010, 26, 687.