|Title||Predicting Detailed Product Distributions For Pyrolysis of Diverse Forms of Biomass|
|Spatial Coverage||Kauai, Hawaii|
|Subject||AFRC 2013 Industrial Combustion Symposium|
|Description||Paper from the AFRC 2013 conference titled Predicting Detailed Product Distributions For Pyrolysis of Diverse Forms of Biomass by Stephen Niksa|
|Rights||No copyright issues|
Predicting Detailed Product Distributions For Pyrolysis of Diverse Forms of Biomass Stephen Niksa Niksa Energy Associates LLC, 1745 Terrace Drive, Belmont, CA 94002 USA, firstname.lastname@example.org Prepared for Oral Presentation to the AFRC 2013 Industrial Combustion Symposium, Kauai, Hawaii, Sep. 22-25, 2013 Abstract Biomass pyrolysis technologies often process an assortment of biomass forms determined by availability and cost. To support the screening of diverse biomass forms for utilization in pyrolysis, NEA developed detailed reaction mechanisms to predict the complete distributions of all major products from any biomass form. Separate mechanisms describe primary devolatilization at heating rates fast enough to prevent tar deposition within the fuel, and secondary volatiles pyrolysis of tar, first, into PAH and, ultimately, into soot with simultaneous elimination of heteroatoms as noncondensables. These mechanisms predict the yields of tar, aromatic oils, light oxygenates, C1 - C4 hydrocarbons, soot, CO, CO2, H2O, H2, NH3, and H2S, plus the tar MWD and the elemental compositions of tar, soot, and char. The kinetics cover the domain of temperature (300 - 1200C), pressure (<1 - 50 atm), and transit times in advanced fluidized bed and entrained flow processing technologies. This paper emphasizes the validation of predicted product distributions with test data from lab- and pilot-scale facilities. Our validation database represents 2 cellulose samples, 7 woods and wood wastes, and 4 agricultural residues, and covers nearly the entire domain of elemental composition for biomass. One dataset completely covered the conversion of tar from two cellulose samples, and another determined the complete distributions of all major products from four diverse biomass samples with very good mass and elemental balance closures. Transient and temperature-dependent tar yields from all samples were simulated within the measurement uncertainties, with only a few stray discrepancies. Our kinetic analysis depicts the broad maxima in tar yields with temperature for the wood samples and cellulose, as well as the sharper maxima for corn stover, bagasse, red maple, and wheat straw. For every biomass sample, the correct temperature for maximum yield is apparent in the simulations, except for wheat straw. Reaction Mechanisms The simulations in this paper are based on two distinct reaction mechanisms. Bio-FLASHCHAIN , describes the primary devolatilization of any biomass form under rapid heating conditions, given the fuel's proximate and ultimate analyses. A separate reaction mechanism, introduced here, describes the succeeding decomposition of primary tars into aromatic oils, polynuclear aromatic hydrocarbons (PAH), soot, and additional noncondensable gases in the free stream away from the fuel particles. Together, these mechanisms predict complete distributions of all major products, including tar, aromatic oils, light oxygenates, C1 - C4 hydrocarbons, soot, CO, CO2, H2O, H2, NH3, and H2S, plus the tar molecular weight distribution (MWD) and the elemental compositions of tar, soot, and char. The only restriction is that the heating rate be sufficiently fast to prevent tar decomposition and coking within the fuel particles during primary devolatilization, which are omitted from the analysis. Bio-FLASHCHAIN® represents biomass as a chain copolymer of cellulose and a lignin-like component. Xylans and hemicelluloses are ignored. The actual composition and structure of cellulose are implemented, whereas the composition of the lignin-like component is assigned from the ultimate analysis of the whole biomass. Specifically, nine C atoms and three O atoms are assigned to the lignin monomer unit, then the ultimate analysis is used to determine the number fraction of cellulose and the H number of the lignin monomer. If the first assignments violate constraints on the H/C ratio of the lignin monomer, then the oxygen number is adjusted for subsequent assignments. After the mass fractions and compositions of the copolymers have been specified, distinctive reaction mechanisms are independently applied to each component, and the product distributions are averaged to determine the complete product distribution. In bio-FLASHCHAIN®, the macromolecular structure of biomass is modeled as a mixture of chain fragments ranging in size from a monomer to the nominally infinite chain. The diverse assortment of structural components in real biomass is rendered coarsely with three generic structural components: bridges, dehydrated bridges, and char links. In biomass, bridges connect to other bridges or to char links, and none of the components is refractory throughout devolatilization. A whole bridge is the monomer unit in the chain, which is glucosan for cellulose but is variable for the lignin-like component. Dehydrated bridges are the condensed-phase products of bridge scission; consequently, all chain fragments are capped with dehydrated bridges initially and throughout devolatilization. Char links are unbreakable connections formed through an extensive decomposition of both whole and dehydrated bridges. In the model's reaction mechanism, labile bridges are the key reaction centers. Their conversion governs the evolution rates and yields of both gas and tar. Due to their complex chemical structures, bridge conversions in bio-FLASHCHAIN® are concerted chemical processes involving numerous steps and many reaction species, not unimolecular scissions. The details of this chemistry are not elaborated in the theory; instead two distributed-energy rate expressions represent only the thermal response of bridge conversion. Conversion of a bridge initiates two distinct reaction pathways, called scission and spontaneous condensation. Bridge scissions generate smaller fragments including precursors to tar, and also noncondensable gases of mostly water with minor amounts of CO2, alcohols, and ketones. Spontaneous condensation forms a new refractory char link plus the gaseous products of bridge dehydration plus additional CO, CO2, H2, gaseous hydrocarbons (GHCs), alcohols, aldehydes, and ketones. This pathway to char links depletes the bridge population without inducing fragmentation, thereby suppressing the production of tar precursors. As an analog to crosslinking, additional char links and gases may also form by bimolecular recombination between the ends of smaller (metaplast) fragments. At elevated temperatures, char links release their residual oxygen as CO. According to this mechanism, the influences of thermal history, pressure, and particle size can be understood in terms of only four mechanisms: (1) Biomass macromolecules depolymerize into fragments with a broad size distribution; (2) A phase equilibrium establishes the mole fraction of tar fragments in a gas stream that is convected out of the particle with no transport resistance; (3) The conversion of labile bridges in the fragments into char links suppresses depolymerization and simultaneously generates noncondensable gases; and (4) Fragments also crosslink in the condensed phase to form nonvolatile components of char. According to the flash distillation analogy , the phase equilibrium shifts to retain a larger portion of the lighter fragments in the condensed phase as the pressure is increased. These fragments would constitute the heavy end of the tar MWD at low pressures, but remain in the fuel at elevated pressures. Consequently, tar prepared at higher pressures becomes lighter and the tar yield diminishes. The fragments retained in the char also contain precursors to noncondensable gases which are eventually released, so gas yields increase as the pressure is elevated, but not by enough to compensate for the retention of tar precursors. Finite-rate transport mechanisms are not needed to explain the pressure effect. In fact, the scaling for negligible transport resistances in bio-FLASHCHAIN® is consistent with the lack of a particle size effect for devolatilization of small biomass particles. Thermal history effects are rooted in the chemical heterogeneity of biomass' key reaction centers. A distribution of activation energies for the depolymerization chemistry represents the very broad thermal response of this reaction system, and explains why asymptotic volatiles yields change for different pyrolysis temperatures. But competitive char formation chemistry is needed to explain the proportions of tar and gas from different forms of biomass. Heating rate affects the rate but not the yields and composition of volatiles. As the heating rate is increased, the onset of devolatilization moves to higher temperatures and the devolatilization rate increases in rough proportion to the heating rate. Of course, primary volatiles are usually transformed beyond recognition in any practical biomass utilization technology due to high processing temperatures and the presence of several reactive gases. Secondary volatiles pyrolysis denotes homogeneous chemistry under an inert gas that transforms primary volatiles after they have left the condensed fuel phase. Tar conversion is the dominant aspect of secondary volatiles pyrolysis. In many respects, tar continues to decompose just like similar molecular fragments continue to pyrolyze in the condensed fuel phase. First aliphatic functional groups within labile bridges and then heteroatoms within char links are eliminated, so there is a marked progression toward PAH during secondary volatiles pyrolysis. At moderate temperatures, PAH-like tars will survive but at temperatures above 800 - 900C, PAH re-polymerize with GHCs to form soot. Secondary pyrolysis chemistry sustains tar conversion into noncondensable gases and some oils at low to moderate temperatures, but the process will ultimately produce soot, another solid carbonaceous residue, when conducted at elevated temperatures. Since secondary volatiles chemistry occurs in the gas phase, in commercial processes it is often affected by reactive gases, especially H2, steam, and O2. But none of these interactions with ambient gases are considered here. NEA's mechanism for the secondary pyrolysis of biomass volatiles at high temperatures contains eight independent steps. As during primary devolatilization in the solid phase, bridges in tar fragments are converted by two channels. (1) Bridge scissions reduce the fragment size and release water along with minor amounts of other noncondensables. (2) Bridge condensations introduce refractory char links into the fragment population, thereby inhibiting the succeeding production of smaller fragments, and are a major source of noncondensables. Bridge condensations release all the gaseous products of bridge scissions, along with an extensive distribution of all the noncondensables in biomass conversion, including NH3 and H2S. (3) The remnants of bridge scissions on newly generated fragment ends are released in essentially the same process as bridge condensation, with about half the amounts of noncondensables. (4) Bimolecular recombinations directly increase the fragment size distribution and also incorporate char links. (5) Single-ring aromatic oils are not produced by primary devolatilization, but they can be expelled during the decomposition of isolated monomer fragments at low temperatures and/or early reaction times, along with appreciable amounts of CO2 and smaller amounts of other noncondensables. (6) CO is released from the residual oxygen in char links at elevated temperatures. (7) Any size secondary tar fragment from lignin may nucleate into soot once its composition is within the range for PAH. (8) And, finally, any secondary tar fragment from lignin and any oil from both lignin and cellulose may condense onto soot to accelerate soot production. In addition, the distribution of GHCs is reduced to CH4, C2H2, and lesser amounts of C2H4 while oxygenates are completely eliminated during sooting, in proportion to the extent of conversion of secondary tar and oils from lignin into soot. These mechanisms have been incorporated into a commercial software package called PC Coal Lab that simulates any user-specified thermal history and pressure that satisfies the stipulation on rapid heating. The thermal histories of the fuel suspension and entrainment gases are specified independently, to handle systems with very large temperature gradients. Each simulation takes a few seconds on ordinary PCs. Validation With Reported Product Distributions The validation database in this paper represents 2 cellulose samples, 7 woods and wood wastes, and 4 agricultural residues, and covers nearly the entire domain of elemental composition for biomass. The ultimate analyses and the dry ash contents were reported for all biomass samples, and these were used with bio-FLASHCHAIN to predict the distributions of the primary products. Collectively, they cover broad ranges of C (44 - 51 daf wt. %); H (5.2 - 6.4%); O (37 - 52 %); and ash (0.1 - 11 dry wt. %). Several samples contain sufficient ash to inhibit the production of liquids during primary devolatilization, to the benefit of their gas yields. Unfortunately, no proximate volatile matter (PVM) contents were reported for any of these samples. Whereas PVM values are usually used to automatically calibrate the rate parameters in bio-FLASHCHAIN to describe variations in the split between liquid and gaseous volatiles among different forms of biomass , we needed to develop another calibration procedure for primary devolatilization for this work, as follows: It is important to realize that the thermal history for primary devolatilization is inconsequential for subsequent tar decomposition, provided that the total yield and the proportions of tar, char, and gas are the same as in the tests. Accordingly, we arbitrarily developed thermal histories for primary devolatilization simulations that gave char yields that matched the reported yields at every temperature. Then the primary calibration factor in bio-FLASHCHAIN was adjusted to match the predicted liquids yield to the reported values for the lowest temperature in each dataset. The liquids yield in the simulation was evaluated as the sum of the predicted yields of tars and light oxygenates. The tacit assumption is that the reported liquids yields for the lowest temperature in each dataset were unaffected by secondary volatiles pyrolysis and are, therefore, the yields of primary liquids. In fact, our calibration procedure gave primary liquids yields that matched the reported yields of liquids for all temperatures cooler than 450C, but the liquids yields were a few percent greater than the reported yields for temperatures from 450 to 500C. Consequently, the primary product distributions in our simulations necessarily contain accurate levels of liquids and char, so that discrepancies in the simulations for hotter temperatures can be attributed to the tar decomposition mechanism, as intended. The first case completely covers the conversion of tar from two cellulose samples in the freeboard of a lab-scale fluidized bed . Figure 1 shows the liquids remaining beyond the freeboard at temperatures from 400 to 800C as a function of the transit time. Liquids yields for zero time were measured and simulated for the freeboard inlet, over the fluidized bed surface. Since both forms of the parent biomass are essentially pure cellulose, the same set of kinetic parameters was applied for both series of calculations. The simulation results are accurate throughout the entire ranges of transit time and temperature for both fuels. The minimal tar decomposition at 400C accelerates for progressively hotter temperatures, and becomes complete by the end of the test at 700C. Whereas cellulose leaves no soot residue at even more severe conditions, the silver birch produces about 5 % soot. One minor flaw is that the simulations under-predict the liquids levels at intermediate reaction times at 600 and 700C for cellulose, and at both of the hottest temperatures for silver birch. Figure 1. Remaining liquids levels as a function of transit time through the freeboard for bed temperatures from 400 to 800C from Stiles and Kandiyoti . Complete product distributions are evaluated more formally with the reported distributions from four biomass forms in a pilot-scale fluidized bed in Table 1 . In this tabulation, the sum of the simulated yields of tar, oils, and oxygenates are compared to the reported yields of organic liquids, on the assumption that the chilled 2-stage liquids collector caught all of these products. The tabulation also contains the yields of the major oxygenated products, H2, and GHCs as the sum of the yields of CH4, C2's, and C3's. In this dataset, CH4 is the predominant GHC component in the measurements, and there are minor amounts of C2H4 and C3's. In the simulated distributions, CH4 and C2H4 often make comparable contributions, and there are more C3's. But GHC distributions are difficult to evaluate because chain compounds are extremely labile and very rapidly converted into C1/C2 mixtures. For IEA poplar, the simulated liquids yields are within the measurement uncertainties throughout, and correctly exhibit the maximum around 500C. Similarly, the simulated yields of CO and GHCs are within the measurement uncertainties throughout except, perhaps, for the GHC yield at 625C (although this test was tagged as suspect due to large breaches in several balances). But the simulated yields of CO2 and H2O have larger discrepancies. Under-predicted CO2 yields and over-predicted H2O and CO yields are seen for nearly all biomass forms. These discrepancies do not necessarily reflect problems with the tar decomposition mechanism. Rather, they are mostly due to a systematic bias in the bio- FLASHCHAIN mechanism that predicts much more primary H2O and much less primary CO2 than is reported in this study. Note that the discrepancies are present at the coolest temperature for each of the four fuels in the dataset, and that this temperature is too cool for essentially any tar decomposition. So it is best to evaluate the contributions from tar decomposition with the difference between the maximum reported value minus the reported value at the lowest temperature in each dataset. On this basis, the reported incremental CO2 yield for IEA poplar is 4.9 daf wt. % vs. the simulated increment of 6.4 wt. %. The analogous comparison for H2O is 5.5 vs. 4.1 %. It is difficult to assess the accuracy of the simulated liquids yields for maple because the three tests with nominal temperatures of 530C gave measured values that evenly span 8.3 daf wt. %. We somewhat arbitrarily emphasized the greatest reported value of 69.9 daf wt. % for 530C, because the associated kinetic parameters also gave reasonably accurate predictions at the other test temperatures. The 0 1 2 3 4 0 20 40 60 80 100 800C 700 650 600 500 450 400C Cellulose Liquids Yield, daf wt.% Freeboard Residence Time, s 0 1 2 3 4 800C 700 600 500 450 400C Silver Birch Freeboard Residence Time, s Table 1. Evaluation of the simulated product distributions from the products of pyrolysis of four biomass forms in a pilot-scale fluidized bed . Sample T,C Organic Liq. CO2 CO H2O GHCs H2 M P M P M P M P M P M P IEA Poplar 425 56.1 57.8 3.7 1.5 2.1 2.7 3.8 7.0 0.2 0.1 0.0 0.1 465 67.5 67.3 4.9 1.8 3.3 3.7 5.5 7.6 0.4 0.3 0.0 0.1 500 66.1 72.9 6.3 2.1 5.3 4.6 9.3 7.9 0.8 0.5 0.0 0.1 500 71.5 72.9 5.9 2.1 5.4 4.6 6.6 7.9 - 0.5 - 0.1 541 64.0 67.3 8.6 2.8 10.5 10.1 7.4 8.8 1.9 2.0 0.1 0.3 625 40.5 45.5 8.4 7.9 23.0 21.0 4.1 11.1 4.9 6.6 0.2 0.4 Maple - IsoT 480 66.9 65.8 4.3 2.7 5.3 8.5 6.9 8.9 0.9 1.4 0.0 0.2 500 64.6 68.3 3.9 2.9 7.3 8.5 9.0 9.1 1.1 1.5 0.1 0.2 530 61.6 67.9 4.5 3.1 10.5 9.4 8.2 9.2 1.5 1.8 0.1 0.2 Maple 482 61.0 65.6 4.1 2.7 3.9 8.8 6.3 8.8 0.4 1.4 0.0 0.2 500 65.1 67.3 4.5 2.8 5.2 9.3 8.2 9.0 0.6 1.6 0.0 0.3 532 56.6 68.3 4.0 3.0 7.0 9.3 7.4 9.1 1.0 1.7 0.0 0.2 531 69.9 68.3 3.7 3.0 5.8 9.3 7.8 9.1 0.9 1.7 0.0 0.2 Whole Poplar 450 56.5 58.5 5.3 2.6 3.0 5.0 5.4 7.8 0.3 0.3 0.0 0.0 500 66.6 66.7 5.8 3.3 5.4 7.9 4.4 8.3 0.8 0.7 0.0 0.1 550 62.8 62.0 7.0 5.7 9.7 10.6 5.4 9.5 1.9 2.0 0.0 0.0 Wheat Straw 500 44.7 53.1 11.1 3.4 7.2 8.9 13.7 9.9 1.3 1.0 0.0 0.1 520 59.4 54.5 8.4 4.0 6.5 12.1 14.0 10.8 1.2 1.6 0.0 0.2 575 49.5 49.0 9.7 5.8 11.5 18.4 14.0 12.3 2.8 3.1 0.1 0.4 simulated CO2 and H2O yields are hardly perturbed over the tested temperature range, in accord with the very small changes in the reported values. But the simulated GHC and H2 yields are slightly greater than the reported values for all temperatures. The reported CO yields exhibit the expected tendency for more CO at progressively hotter temperatures, whereas the simulated CO yields are hardly perturbed upward. For whole poplar, the simulated yields of organic liquids, GHCs, H2, and CO are accurate across the entire temperature range, and the flaws in the Figure 2. Liquids yields for all biomass forms from a lab-scale fluidized bed from Scott and Piskorz [5-7]. CO2 and H2O yields should be attributed to the discrepancies for the primary products noted above. For wheat straw, only the simulated GHC yields are within the measurement uncertainties across the entire temperature range, but it is difficult to assess the discrepancies because there are large breaches in the O-balances in the raw test data. A somewhat broader range of biomass composition was tested in a lab-scale fluidized bed [5-7], although this dataset has breaches in the mass and elemental balances that are considerably worse than those from the pilot-scale fluidized bed. As seen in Fig. 2, the simulated organic liquids yields are within the measurement uncertainties across their test temperature ranges, except for the hottest temperature with poplar wood and the coolest temperature with red maple. Our kinetic analysis depicts the broad maxima with temperature for poplar wood, whole poplar, and cellulose, as well as the sharper maxima for corn stover, bagasse, red maple, and wheat straw. For every biomass sample, the correct temperature for maximum yield is apparent in the simulations, except for wheat straw, for which the predicted maximum yield is shifted toward cooler temperatures by about 20C. Clearly, the proposed mechanism can depict the temperature dependence of the liquid yields for a diverse assortment of biomass forms. Unfortunately, the kinetic parameter assignments behind this performance are not entirely consistent with our hypothetical two-component constitution submodel. According to this submodel, the constitution of the cellulose component is essentially the same in all biomass forms, so that only the fractional contribution varies on a mass basis. Consequently, we would hope to use the same kinetic parameters for the cellulose in any biomass form to describe tar conversion. Indeed, the same rate parameters were used to describe bridge conversion in the cellulose contributions, and also in the contribution for the lignin-like component. But the frequency factors for the production of both CO and oils had to be varied over substantial ranges in both components to accurately interpret our validation datasets. As noted earlier, the same parameter set was used to interpret both samples from , because they were both essentially pure cellulose. For the pilot-scale testing , one set interpreted poplar and maple woods and another interpreted whole poplar waste and wheat straw. For [5-7], the frequency factors to produce CO and oils had to be varied for both components in almost every sample. Whereas the temperatures in all these tests were too cool to promote sooting, additional simulations for a 400 450 500 550 600 650 700 750 30 40 50 60 70 80 UWat1 Lab-Scale Corn Stover Whole Poplar Poplar Wood Organic Liquids, daf wt. % Temperature, C 400 450 500 550 600 650 700 750 UWat1 Lab-Scale Bagasse Red Maple Cellulose Wheat Straw Temperature, C Figure 3. Correlation between predicted soot yields and mass fractions of the lignin-like component in the biomass. heating rate of 1000C/s to 1150C characterized soot production from all the biomass forms that were used in the testing. As seen clearly in Fig. 3, there is a good correlation between the ultimate soot yields and the lignin fraction that gives a correlation coefficient approaching 0.9 and a std. dev. of 2.2 daf wt. %. When the soot yield was expressed as a percentage of the predicted primary tar yield, the correlation coefficient increased to 0.91. But the scatter in the yields for similar lignin fractions shows that variations in composition are also important. The predicted yields of CO and GHCs were not correlated at all with the lignin fraction, and instead appeared to be determined by a sample's ultimate analysis. Summary Our primary goal - to accurately interpret the dynamics of tar decomposition throughout secondary volatiles pyrolysis - has been met, in so far as transient and temperature-dependent tar yields from all samples were simulated within the measurement uncertainties, with only a few stray discrepancies. Our kinetic analysis depicts the broad maxima with temperature for the wood samples and cellulose, as well as the sharper maxima for corn stover, bagasse, red maple, and wheat straw. For every biomass sample, the correct temperature for maximum yield is apparent in the simulations, except for wheat straw, for which the predicted maximum yield is shifted toward cooler temperatures by about 20C. Clearly, the proposed mechanism can depict the temperature dependence of the liquid yields for a diverse assortment of biomass forms. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 2 4 6 8 10 12 14 16 r=0.882; =2.20 YSOOT, daf wt.% fLIGNIN The greatest flaws in the distribution of noncondensable products - over-predicted H2O yields and under-predicted CO2 yields - could be attributed to the primary devolatilization mechanism, especially since the incremental additions of these products during tar conversion were largely accurate. The surge in CO yields for the hottest test temperatures and the ultimate, maximum CO yields were accurately simulated, and there were no systematic discrepancies in either the GHC or H2 yields. Hence, the proposed reaction mechanism simulates the distributions of all major products from any biomass form within useful quantitative tolerances. The time required to achieve an ultimate soot yield under fluidized bed conditions is strongly temperature dependent, diminishing from 225 s at 650C to 15 s at 750C to 0.8 s at 850C to 0 s at 950C. Considering that total vapor residence times through bubbling beds and freeboards rarely exceed 5 s, NEA's secondary pyrolysis mechanism would only predict appreciable levels of soot at temperatures hotter than 800C, which is consistent with expectations. For a diverse assortment of different biomass forms, the ultimate soot yields increase in direct proportion to the lignin fraction from bio- FLASHCHAIN's constitution submodel. But the scatter in the yields for similar lignin fractions shows that variations in composition are also important. The predicted yields of CO and GHCs were not correlated at all with the lignin fraction, and instead appeared to be determined by a sample's ultimate analysis. The biggest remaining problem is that the kinetic parameter assignments behind this performance are not entirely consistent with our hypothetical two-component constitution submodel. According to this submodel, the constitution of the cellulose component is essentially the same in all biomass forms, so that only the fractional contribution varies. Consequently, we would hope to use the same kinetic parameters for the cellulose in any biomass form to describe tar conversion. Indeed, the same rate parameters were used to describe bridge conversion in the cellulose contributions, and also in the contribution for the lignin-like component. But the frequency factors for the production of both CO and oils had to be varied over substantial ranges in both components to accurately interpret our validation datasets. Acknowledgement The project behind this work was sponsored by the Department of Chemical Engineering, Gunma University, Kiryu, Japan. References  Niksa, S., "Predicting the rapid devolatilization of diverse forms of biomass with bio- FLASHCHAIN®," Proc. Int. Symp. On Combust., Vol. 28, The Combustion Institute, Pittsburgh, PA, 2000, pp. 2727-2733.  Niksa, S., "Rapid coal devolatilization as an equilibrium flash distillation," AIChE Journal, 34(5): 790-802 (1988).  Stiles, H. N. and Kandiyoti, R., "Secondary reactions of flash pyrolysis tars measured in a fluidized bed pyrolysis reactor with some novel design features," Fuel, 68(3):275 (1989).  Scott, D. S. and Piskorz, J., "The continuous flash pyrolysis of biomass," Can. J. Chem. Eng., 62(6):404-412 (1984).  Scott, D. S. and Piskorz, J., "The flash pyrolysis of aspen-poplar wood," Can. J. Chem. Eng., 60(10):666-74 (1982).  Scott, D. S., Piskorz, J., and Radlein, D., "Liquid products from the continuous flash pyrolysis of biomass," Ind. Eng. Chem. Process. Des. Dev., 24:581-88 (1985).  Liden, A. G., Beruti, F., and Scott, D. S., "A kinetic model for the production of liquids from the flash pyrolysis of biomass," Chem. Eng. Comm., 65:207-21 (1988).