Analytica Chimica Acta 639 (2009) 62–66
Contents lists available at ScienceDirect
Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca
Quantitative ﬂash pyrolysis Fourier transform infrared spectroscopy of organic materials Richard W. Court ∗ , Mark A. Sephton Department of Earth Science and Engineering, Prince Consort Road, Imperial College London, London SW7 2BP, UK
a r t i c l e
i n f o
Article history: Received 8 December 2008 Received in revised form 20 February 2009 Accepted 23 February 2009 Available online 6 March 2009 Keywords: Pyrolysis Infrared spectroscopy Quantitative analysis Organic matter
a b s t r a c t Thermal degradation is a common technique used to investigate the nature of organic materials. However, existing methods for the Fourier transform infrared (FTIR) identiﬁcation and quantiﬁcation of volatile products from the thermal degradation of organic materials are limited to the technique of thermogravimetric analysis (TGA)–FTIR, which utilizes relatively low heating rates. However, the thermal degradation products of organic materials are known to vary depending on the rate of heating, with lower heating rates of biomass associated with increased yields of solid char and decreased yields of volatiles, as well as a greater opportunity for secondary reactions between the residue and the pyrolysis products. Hence, it is difﬁcult to relate the products of organic matter thermally degraded at <100 ◦ C min−1 in TGA to the products of ﬂash pyrolysis at up to 20,000 ◦ C s−1 . We have developed and applied a novel methodology for quantitative ﬂash pyrolysis–FTIR analysis of the volatile pyrolysis products of organic-rich materials. Calibration curves of water, carbon dioxide and methane have been constructed and used to determine absolute volatile release from wood (ash, Lat. Fraxinus). This technique is quicker and simpler than comparable pyrolysis–gas chromatography–mass spectrometry techniques, and avoids errors associated with the lower rates of temperature increase associated with techniques such as thermogravimetric analysis. © 2009 Elsevier B.V. All rights reserved.
1. Introduction Thermal degradation is a common technique used to investigate the nature of organic materials. Heating in an inert atmosphere cleaves bonds, producing smaller molecular fragments that are more amenable to analysis via other detection and characterization techniques. Fourier transform infrared (FTIR) has been used previously in conjunction with thermal degradation to investigate the nature of organic materials. For example, FTIR has been used to investigate the volatile compounds produced by thermogravimetric analysis (TGA) [1,2]. TGA uses low rates of heating, typically less than 100 ◦ C s−1 [2,3], a heating rate often referred to as pyrolysis in its general sense, but also described as carbonisation to reﬂect the relatively gentle rate of heating . TGA temperature programmes are distinct from the ﬂash pyrolysis regimes typically employed by pyrolysis–GC–MS [5–10], where coil heating rates up to 20,000 ◦ C s−1 are used to break bonds as rapidly as possible and to swiftly remove the products of pyrolysis from the sample. However, it is difﬁcult to apply results from thermal degradation experiments involving low rates of heating, such as employed in TGA, to those of ﬂash pyrolysis . The behaviour of organic materials during thermal degradation is known to be dependent on heating rates. For
∗ Corresponding author. Tel.: +44 2075949982; fax: +44 2075947444. E-mail address: [email protected]
(R.W. Court). 0003-2670/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2009.02.042
example, high heating rates are associated with depolymerisation, enhanced loss of aliphatic carbon and high yields of tars, whereas low heating rates favour dehydration, increased yields of insoluble char and greater amounts of water [12–15]. Slow heating of cellulose leads to a primary char below 400 ◦ C, which then undergoes secondary reactions as it is heated beyond 400 ◦ C . In contrast, rapid pyrolysis lessens the time spent by a sample at lower temperatures, thereby limiting the associated reactions; the potential for secondary reactions is restricted by faster rates of heating. Although secondary reactions occurring during the slow pyrolysis of wood – in particular, the extent and nature of reactions between nascent pyrolysis vapours and the thermally degrading parent material – are difﬁcult to understand [16,17], smaller sample sizes also help to avoid the effect of secondary reactions and are associated with smaller amounts of char, enhanced loss of aliphatic carbon and a reduced proportion of water [11,12]. This suggests that char itself is not a primary product but a result of secondary repolymerisation, whereby laevoglucosan produced by depolymerisation of cellulose is trapped within a sample and breaks down exothermally, a mechanism that suggests that a critical parameter is the residence time of volatiles in the pyrolysing cellulose matrix . Although it is difﬁcult to relate this model to samples that do not contain cellulose, it implies that reduced residence times of volatiles in a sample lead to lesser levels of secondary reactions. Rapid pyrolysis therefore restricts the opportunity for secondary reactions between the sample residue and the products of pyrolysis,
R.W. Court, M.A. Sephton / Analytica Chimica Acta 639 (2009) 62–66
because the sample is exposed to elevated temperatures for less time, and faster production of volatiles causes their more efﬁcient removal from the sample. Hence, it cannot be assumed that the products of the slow thermal degradation of an organic-rich sample are representative of those of rapid pyrolysis [9–11]. Although FTIR has been used in conjunction with slow pyrolysis [2,19] and rapid pyrolysis , no report of quantitative FTIR analysis applied to the products of rapid pyrolysis is available. The development of a quantitative FTIR technique for the analysis of volatiles generated by ﬂash pyrolysis is of great interest to investigations of the nature of complex organic materials, such as coal, kerogen and the macromolecular organic material found in carbonaceous meteorites. Here, we establish and apply a methodology for the quantitative FTIR analysis of methane, carbon dioxide and water, produced upon rapid pyrolysis and gasiﬁcation of organic-rich materials. Quantitative analysis is enabled by the production of calibration curves for each compound of interest, which are used to quantify the yields of those compounds produced upon pyrolysis of the samples. 2. Experimental 2.1. Instrumental setup FTIR spectra were acquired using a Thermo-Nicolet 5700 FTIR spectrometer; pyrolysis was performed using a CDS Analytical Pyroprobe 2500 and a CDS Analytical Brill CellTM . Solid samples were placed in quartz pyrolysis tubes and held in position with quartz wool, both previously cleaned by baking in air at 450 ◦ C, then pyrolysed in a helium atmosphere. The temperatures of pyrolysis were attained at 20 ◦ C ms−1 and held for 15 s, with the Brill Cell itself held at 250 ◦ C during pyrolysis. The masses of solid samples were recorded before and after pyrolysis, using a balance accurate to 50 g, allowing the mass lost upon pyrolysis, and therefore product yields, to be calculated. Spectra were accumulated from 256 individual spectra using a deuterated triglycine sulfate (DTGS) detector; the blank spectra were collected from 64 individual spectra. The Brill CellTM was held at 250 ◦ C and was purged with helium during accumulation of the blanks. During pyrolysis, volatiles were allowed to accumulate in the cell to produce stronger IR absorbance bands.
sequently used to correct sample spectra. The absorption of water and carbon dioxide was quantiﬁed by measurement of the areas of the 3853 and 669 cm−1 absorption bands, respectively. A compound amenable to this investigation that would produce methane in a similar fashion to the production of carbon dioxide and water from sodium hydrogencarbonate could not be identiﬁed, so known volumes of methane at atmospheric pressure were instead inject directly into the Brill Cell through a septum. Although pyrolysis and the CDS Analytical Pyroprobe 2500 were not involved here, the instrumental setup and spectral accumulation was otherwise the same as that used for the accumulation of spectra of water and carbon dioxide from sodium hydrogencarbonate. The Brill Cell was again held at 250 ◦ C and was previously purged with helium to eliminate atmospheric components. The direct injection of methane in this fashion raises two areas of consideration, concerning comparability of measurements. Pyrolysis of the sodium hydrogencarbonate produces carbon dioxide and water in the centre of the Brill Cell, immediately below the path of the FTIR laser. However, the methane is injected into the neck of the Brill Cell, without the pyrolysis unit being present. This has two effects. Firstly, the injected methane must diffuse from the neck of the Brill Cell through to the main body, where the IR beam passes, to give a representative signal and, without the pyrolysis unit in place, the internal volume of the Brill Cell is larger, requiring the application of a correction factor, to account for the reduced volume of the Brill Cell when a sample is pyrolysed. Accounting mathematically for the diffusion of methane from the injection site in the cell neck to the cell main body is more difﬁcult, so it has been investigated experimentally, by the injection of methane followed by pauses of varying duration before spectral accumulation to allow diffusion throughout the cell. 3. Results and discussion 3.1. Calibration curves
2.2. Production of calibration curves
The diffusion of methane from its site of injection in the Brill CellTM neck to the cell body is detailed in Fig. 1. Although the accumulation of the 256 spectra necessary for each sample required approximately 190 s, spectra accumulated immediately after injection of 250 l of methane produced a FTIR response that is signiﬁcantly lower than those produced after diffusion of around a minute or longer. Thirty seconds of diffusion produced rather
To create calibration curves, aliquots of gases were introduced into the Brill Cell and their IR absorbance measured. Calibration curves for water and carbon dioxide were produced by the thermal decomposition of known masses of sodium hydrogencarbonate, which decomposes to water, carbon dioxide and sodium carbonate, beginning around 60 ◦ C but occurring rapidly beyond 200 ◦ C. Thermal decomposition was achieved by heating the samples of sodium hydrogencarbonate to 600 ◦ C within a CDS Analytical Pyroprobe 2500, using the same conditions of pyrolysis and same experimental setup as used for the samples of wood described below. The balance accuracy of ±50 g corresponds to errors in the masses of water and carbon dioxide in the calibration curves that range from approximately 10% at the low end of the calibration curves, to about 0.5% at the high end. Since the Brill CellTM must be opened to allow the insertion of a sample, some leakage of atmospheric water and carbon dioxide into the Cell is inevitable. This source of error was minimized by maintaining a slight overpressure of helium inside the Cell as the sample was inserted, hindering inﬂow of atmospheric gases, and by the collection of several blank analyses, where spectra were collected after the Pyroprobe was inserted into the Cell without a sample, giving an estimate of the extent of atmospheric leakage, which was sub-
Fig. 1. The area of the 3015 cm−1 methane absorption band, following the injection of 250 l of methane, after allowing for diffusion throughout the Brill Cell for various periods of time. Five analyses were performed for each time period of diffusion; the error bars shown indicate one standard deviation of the data. Diffusion for less than 60 s before accumulating spectra is associated with a lower measured absorption; 60 s or longer is required for the injected methane to diffuse throughout the cell and give reproducible results.
R.W. Court, M.A. Sephton / Analytica Chimica Acta 639 (2009) 62–66
inconsistent results, as indicated by the relatively large error bars of one standard deviation in Fig. 1, interpreted here as variably efﬁcient diffusion of methane from the relatively cool cell neck, to the main cell body, held at 250 ◦ C. However, diffusion for 60 s or longer gave similar, reproducible results. Consequently, the volumes of methane injected for production of the calibration curve of methane were allowed to diffuse throughout the cell for 60 s before accumulation of the spectra commenced. The calibration curves for water, carbon dioxide and methane are displayed in Fig. 2. The curves are described by quadratic equations and correlation coefﬁcient r−1 in the bottom-right of the graphs, with the match between the data and the equation described by the values of root mean square error (RMSE). Quadratic equations were chosen to describe the data as they are readily solvable using the quadratic formula and provided good ﬁts to the data points. For methane, a single quadratic equation was sufﬁcient to describe the data; for carbon dioxide and water, two quadratic equations for different areas of the data are shown. The nonlinear responses indicate deviation away from Beer–Lambert Law, as expected at higher concentrations, demonstrating that the calibration curves can only be applied reliably to data that falls within the limits set by the existing data sets. 3.2. Quantitative pyrolysis–FTIR of ash wood The calibration curves displayed in Fig. 2 have been applied to the FTIR spectra of the pyrolysis products of a series of samples of ash wood (Lat. Fraxinus). Five samples of wood were analysed, with masses from 4.7 to 8.6 mg, as measured using a balance accurate to 0.05 mg, producing weighing errors of 0.5–1.1%. The samples were pyrolysed using the same experimental conditions as used for the production of the calibration curves for water and carbon dioxide from sodium hydrogencarbonate, except that pyrolysis was performed at 1000 ◦ C, a temperature chosen to promote gasiﬁcation reactions and increase the yield of the volatiles of interest here. An FTIR spectrum of the pyrolysis products of one of these samples is illustrated in Fig. 3. Carbon dioxide is recognised from its characteristic strong absorptions around 2350 and 669 cm−1 , with the area of 669 cm−1 band used for quantiﬁcation here, with four further weak absorption bands barely visible between 3600 and 3750 cm−1 , partially obscured by absorption resulting from water (3500–4000 cm−1 ); water also causes absorption around 1300–1900 cm−1 . The discrete band at 3853 cm−1 has been chosen for the quantiﬁcation of water, because of minimal interference with other absorption bands. Methane has been quantiﬁed using its strong band around 3015 cm−1 ; methane also produces a weak absorption band around 1305 cm−1 , too weak to be readily apparent in Fig. 3. Absorption resulting from aliphatic methyl (–CH3 ) and methylene (–CH2 –) carbon–hydrogen bonds is also apparent as the ill-deﬁned hump around 2900–3000 cm−1 . Elsewhere, the presence of carbon monoxide is indicated by the weak dual band around 2150 cm−1 . The weak band around 1180 cm−1 is likely a result of absorption by C–O bonds. Given the variety of compounds that these compounds can exist in, it is difﬁcult to attribute them to a speciﬁc compound. However, pyrolysis of lignin, a signiﬁcant component of wood, is known to produce methoxyphenols such as guaiacol and syringol [e.g., 21], so some contribution from carbon–oxygen bonds in these species is expected. Table 1 details the areas of the chosen absorption bands of water (3853 cm−1 ), carbon dioxide (669 cm−1 ) and methane (3015 cm−1 ). The masses of volatile required to produce these absorptions have been determined using the calibration curves of Fig. 2. The pyrolysis of ﬁve samples allows the reproducibility of results obtained using this technique to be established. The nature of the samples, as fragments of a heterogeneous natural material, means that some variation is inevitable. This is apparent in the variations in masses
Fig. 2. Calibration curves for water, carbon dioxide and methane, as produced by the decomposition of masses of sodium hydrogencarbonate, as produced using the areas of the 3853, 669 and 3015 cm−1 bands, respectively. The deviations from the Beer–Lambert law are indicated by nonlinear relationships between sample concentration and absorbance. The equations describe the curves formed by the data, along with the correlation coefﬁcient r2 , while the root means square error (RMSE) describes the difference between the measured data points and those predicted by the equations. A single equation could not be found to satisfactorily describe the water and carbon dioxide data, so two quadratic equations are shown, each describing different areas of the data, as indicated by the different data point markings and .
R.W. Court, M.A. Sephton / Analytica Chimica Acta 639 (2009) 62–66
Fig. 3. The FTIR spectrum of the pyrolysis products of 8.2 mg of ash wood pyrolysed at 1000 ◦ C, with no prior desorption.
lost upon pyrolysis, in Table 1, which vary from 70 to 87%—a range much greater than that expected from errors resulting from weighing the samples. However, Table 1 shows that each of the ﬁve samples produced fairly similar yields of water, methane and carbon dioxide, demonstrating the production of reproducible data. Overall, the yield of methane from the samples of ash wood is around 3 wt%. Water is more abundant, comprising around 20 wt%. of the yield, with carbon dioxide being intermediate, at around 14 wt%. Overall, the masses of water, carbon dioxide and methane comprise just under half of the mass of the pyrolysis products (46.6%). The remaining material is likely to include compounds such as methoxyphenols such as syringol and guaiacol, and carbon monoxide and aliphatic hydrocarbons as inferred from the FTIR spectrum displayed in Fig. 3. 3.3. Combustion or pyrolysis? Our ﬂash pyrolysis–FTIR method enables the quick and easy quantiﬁcation of certain volatile species produced upon pyrolysis of organic-rich samples. However, two of the volatiles quantiﬁed here – water and carbon dioxide – can be produced by combustion, as well as by pyrolysis. Combustion is not normally a problem with pyrolysis techniques such as pyrolysis–GC–MS, where more interest is focussed upon larger molecular fragments. Here, however, we are interested in smaller species, such as water and carbon dioxide, that can form via combustion. It is therefore necessary to understand the extent, if any, of combustion occurring in the cell. The cell, of volume 22.5 mL, is sealed then purged with helium for several seconds at a rate of about 20 mL s−1 , following inser-
tion of the sample. This will remove almost all atmospheric oxygen; however, some oxygen will inevitably remain, adsorbed to the sample and the quartz pyrolysis tubes and wool, or simply trapped within the pyrolysis tube. To investigate the extent of combustion occurring in the cell, we have pyrolysed samples of the styrene–isoprene–styrene copolymer, Kraton® D-1107. This polymer is composed solely of hydrogen and carbon, so the production of any water, carbon dioxide or carbon monoxide is the result of the presence of unwanted oxygen in the cell and combustion. Pyrolysis was performed under the same conditions as used for the pyrolysis of the wood samples, except that a temperature of 600 ◦ C was used, a difference not expected to have any signiﬁcant effects on the extent of combustion occurring. A spectrum typical of the pyrolysis products of Kraton® D-1107 is displayed in Fig. 4. The presence of water, carbon dioxide and carbon monoxide indicates that some combustion has occurred. However, application of the calibration curves of Fig. 2 enables the masses of water and carbon dioxide produced to be determined and, hence, the proportion of Kraton® that is combusted to be estimated. Pyrolysis of three samples of Kraton® D-1107 therefore gave yields of water and carbon dioxide calculated to represent the combustion of about 2% of the mass of the sample of Kraton® . The yield of carbon monoxide cannot be determined, because of the absence of a calibration curve, but it is likely to be similar to the yields of water and carbon dioxide. This low degree of combustion, around 2–3%, therefore indicates that the volatiles produced by the wood samples are overwhelmingly the result of pyrolysis, rather than combustion.
Table 1 Application of the quantitative pyrolysis–FTIR technique to samples of ash wood, W1–W5.
Sample mass (mg) Loss upon pyrolysis (mg) Loss upon pyrolysis (%)
5.4 5.9 72.0
4.7 6.1 70.9
6.8 5.9 86.8
8.6 3.7 78.7
8.2 4.4 81.5
– – 78.0
– – 5.9
Methane Absorbance (band area) Projected mass (mg) Yield (% of sample mass) Yield (% of sample mass lost)
0.4031 0.24 2.9 4.1
0.3896 0.23 2.7 3.8
0.3945 0.23 3.4 4.0
0.2523 0.15 3.2 4.0
0.3068 0.18 3.4 4.1
– – 3.1 4.0
– – 0.3 0.1
Water Absorbance (band area) Projected mass (mg) Yield (% of sample mass) Yield (% of sample mass lost)
0.3368 1.40 17.2 23.7
0.3548 1.49 17.4 24.5
0.3432 1.43 21.1 24.3
0.2537 0.99 21.3 26.9
0.2618 1.03 19.2 23.5
– – 19.2 24.6
– – 1.7 1.2
Carbon dioxide Absorbance (band area) Projected mass (mg) Yield (% of sample mass) Yield (% of sample mass lost)
0.956 1.11 13.5 18.9
0.968 1.15 13.1 18.8
0.911 0.99 14.4 16.8
0.732 0.68 14.5 18.3
0.793 0.77 14.1 17.5
– – 14.0 18.1
– – 0.5 0.8
Sum of masses (mg) Sum of masses as % of mass lost
R.W. Court, M.A. Sephton / Analytica Chimica Acta 639 (2009) 62–66
Fig. 4. The pyrolysis products of the styrene–isoprene–styrene copolymer Kraton® D-1107, following pyrolysis at 600 ◦ C. The presence of water, carbon dioxide and carbon monoxide indicates that combustion has occurred; however, quantiﬁcation of the yields of water and carbon monoxide indicates that the degree of combustion that occurs is small.
3.4. Pyrolysis–FTIR as an analytical technique The technique that we describe can rapidly and easily quantify the yields of volatiles upon the pyrolysis of organic-rich materials. It offers numerous advantages over GC–MS-based techniques of quantitative analysis of volatiles produced upon pyrolysis, which can be time-consuming (GC–MS runs often take around an hour or two to complete), can require the use of specialist columns and detectors, such as Porous Layer Open Tubular columns or ﬂame ionisation detectors. Pyrolysis–GC–MS methods are also unsuited to the analysis of inorganic phases, which can produce column-blocking residues . In contrast, the pyrolysis–FTIR technique described here is quick and simple, with analyses taking a just a few minutes. FTIR has been employed to investigate the products of pyrolysis in many investigations [e.g., 20,21]. However, it has been employed in a comparative or qualitative fashion, or as a supplement to other techniques. Our work represents the ﬁrst application of quantitative pyrolysis–FTIR as a technique capable of standing alone. TGA–FTIR has been employed to investigate the thermal degradation of organic materials with rising temperature, with the products identiﬁed using FTIR [e.g., 22]. However, the low rates of heating employed in these experiments are not comparable with the rate of heating employed here during pyrolysis, of 20 ◦ C ms−1 . Consequently, the quantitative pyrolysis–FTIR technique outlined here represents a novel and important methodology. For some species, such as the water, carbon dioxide and methane analysed here, the production of the requisite calibration curves simply requires the injection of known masses of gas into the cell, or the thermal decomposition of suitable source materials, such as sodium hydrogencarbonate. However, other absorption bands of note in the FTIR spectra are more difﬁcult to quantify, such as the aliphatic absorption features around 2900–3000 cm−1 , or the alkene C–H band around 950 cm−1 visible in Fig. 3. It is possible to estimate detection limits using the calibration curves of Fig. 2 and the range of variation of band areas in blank pyrolysis–FTIR spectra. For water, blank spectra typically show a 3853 cm−1 water absorption band of area less than 0.001, corresponding to a mass of water of 3 g. For carbon dioxide, a typical carbon dioxide 669 cm−1 band area in blank spectra is around 0.03, corresponding to a mass of around 2 g. The typical noise around 3015 cm−1 enables the detection of approximately 1 g of methane. 4. Conclusions We have described a newly developed quantitative ﬂash pyrolysis–FTIR method that combines rapid analysis time and with
absolute determinations of volatile species released under analytical ﬂash pyrolysis conditions. Our studies using natural wood samples compared to authentic standards reveal that the technique is an effective means of quantifying the primary reaction products water, methane and carbon dioxide. The rapid heating associated with ﬂash pyrolysis restricts the opportunity for secondary reactions, relative to the much slower rates of heating associated with TGA–FTIR. The contribution of combustion is found experimentally to be almost negligible. The technique has valuable applications in various scientiﬁc ﬁelds where the production of thermally generated volatiles from solids is of interest, including studies related to wood chemistry, fossil fuels and planetary science. Acknowledgements We are grateful to Judith Pillinger for the supply of the ash wood samples, and to an anonymous reviewer whose comments greatly improved this work. This work was supported by STFC. References  T.B. Brill, P.J. Brush, K.J. James, J.E. Shepherd, K.J. Pfeiffer, Appl. Spectrosc. 46 (1992) 900.  M.X. Fang, D.K. Shen, Y.X. Li, C.J. Yu, Z.Y. Luo, K.F. Cen, J. Anal. Appl. Pyrol. 77 (2006) 22.  S. Volker, T. Rieckmann, J. Anal. Appl. Pyrol. 62 (2002) 165.  V. Strezov, M. Patterson, V. Zymla, K. Fisher, T.J. Evans, P.F. Nelson, J. Anal. Appl. Pyrol. 79 (2007) 91.  M.A. Sephton, I. Gilmour, Planet. Space Sci. 49 (2001) 465.  Z. Parsi, N. Hartog, T. Gorecki, J. Poerschmann, J. Anal. Appl. Pyrol. 79 (2007) 9.  W.J. Irwin, Analytical Pyrolysis: A Comprehensive Guide, Marcel Dekker, New York, 1982.  J.C.J. Bart, Polymer/additive analysis by ﬂash pyrolysis techniques, in: 14th International Symposium on Analytical and Applied Pyrolysis, Elsevier Science Bv, Seville, Spain, 2000, p. 3.  W.J. Irwin, J. Anal. Appl. Pyrol. 1 (1979) 89.  W.J. Irwin, J. Anal. Appl. Pyrol. 1 (1979) 3.  M. Lanzetta, C. DiBlasi, F. Buonanno, Ind. Eng. Chem. Res. 36 (1997) 542.  J. Hayashi, H. Takahashi, S. Doi, H. Kumagai, T. Chiba, T. Yoshida, A. Tsutsumi, Energy Fuels 14 (2000) 400.  M. Hajaligol, B. Waymack, D. Kellogg, Fuel 80 (2001) 1799.  J. Gibbinsmatham, R. Kandiyoti, Energy Fuels 2 (1988) 505.  R. Ball, A.C. McIntosh, J. Brindley, Phys. Chem. Chem. Phys. 1 (1999) 5035.  A.G.W. Bradbury, Y. Sakai, F. Shaﬁzadeh, J. Appl. Polym. Sci. 23 (1979) 3271.  M.J. Antal, G. Varhegyi, Ind. Eng. Chem. Res. 34 (1995) 703.  P. Ahuja, S. Kumar, P.C. Singh, Chem. Eng. Technol. 19 (1996) 272.  W. de Jong, A. Pirone, M.A. Wojtowicz, Fuel 82 (2003) 1139.  J. Feng, W.Y. Li, K.C. Xie, Energy Sources A 28 (2006) 167.  J.C. del Rio, A. Gutierrez, I.M. Rodriguez, D. Ibarra, A.T. Martinez, J. Anal. Appl. Pyrol. 79 (2007) 39.  R. Bassilakis, R.M. Carangelo, M.A. Wojtowicz, Fuel 80 (2001) 1765.