I t has been demonstrated that high pressure equipment can be fabricated for studying chemical reactions in fluid systems of at least 500-cc. volume a t pressures u p to 200,000 p.s.i. and temperatures u p to 800’ F.
548 544 552
,K!, 5 4 0 L 2 5362 E
E 524520 516
The authors are indebted to the National Science Foundation for providing substantial financial support (NSF Grant GP-358) during the course of this work. Financial support was also received from Autoclave Engineers, Inc. The authors thank Continental Oil Co. and the University of Michigan for supplying some of the high pressure equipment required for this work.
47 46 36
Figure 1 1.
Cool flame phenomenon
Pressure-time curve for run JLL-254
I n the oxidation of methane, pressure has a strong influence on the product distribution. I n general, an increase in pressure increases the yield of methanol and other oxygenated organic liquid products. At a given pressure the maximum yield of methanol occurs a t an initial temperature corresponding to a relatively short residence time. The cool flame phenomenon was observed a t high pressures for the first time. The nature and surface condition of the vessel affect the kinetics of the oxidation process.
(1) Continental Oil Co., private communication. (2) Furman, M. S., Khim. Prom. 1946,Nos. 1-2, p. 24. (3) Furman, M. S., Shestakova, A. D., Radle-Desyatnik, I. Sh. Tr. Gosudarstvennyi Znstitut Azotnoi Promyshlennosti, No. 1, 100 (1953). (4) Gonikberg, M. G., “Chemical Equilibria and Reaction Rates at High Pressures,” Israel Program for Scientific Translations, Jerusalem, 177 (1963). (5) Zbid., 179 (1963). (6) Hoare. D. E.. T r a m . Faradav SOC.49. 628 (1953). (7) Lobo, ’P. A.,’Sliepcevich, 6. M., Research and Engineering 11, 33-5 (October 1956). ( 8 ) Lobo, P. A., Sliepcevich, C. M., White, R. R., Ind. Eng. Chem. 48, 906 (1956). (9) Lott, J. L., Ph.D. thesis, “Selective Oxidation of Methane a t High Pressures,” University of Oklahoma, 1965. (10) Newitt, D. M., Haffner, A. E., Proc. Roy. SOC. A134, 591 (1932). (11) Newitt, D. M., Szego, P., Zbid., A147, 555 (1943). (12) Pan, L.-S., Andersen, T. N., Eyring, H., IND.ENG.CHEM. PROCESS DESIGN DEVELOP. 5 , 242 (1966). (13) Paris, A,, Chim. Znd. (Paris) 1934, p. 411; C A 58066 (1934). (14) Shtern, V. Ya., “The Gas-Phase Oxidation of Hydrocarbon,” Pergamon Press, Oxford, 1964. (15) Skinner, J. L., Daniels, R. D., Sliepcevich, C. M., Brit. Chem. Eng. J . 8 , 245 (1963). (16) Vanpte, M., Grard, F., Fuel 34,A33 (1955). (17) Walling, C. T., Pellon, J., J . A m . Chem. SOC.79, 4782 (1957). (18) Wiezevich, P. J., Frolich, P. K., Znd. Eng. Chem. 26, 267 (1934). RECEIVED for review December 16, 1965
ACCEPTED August 26, 1966
PYROLYSIS OF COAL PARTICLES IN PULVERIZED FUEL FLAMES J A C K B . H 0 W A R D , Department of Chemical Engineering, Massachusetts Institute R0
. ESS ENH IG H ,
Department of Fuel Science, The Pennsylvania State University, University Park, P a .
converts coal into volatile products and solid The rate of volatiles evolution may be controlled by either chemical decomposition of the coal or gaseous flow through the solid matrix. Knowledge of the kinetics of pyrolysis is required for understanding the combustion of coal particles in pulverized-fuel flames. Volatiles evolution is known to influence the combustion of solid particles ( I I ) , and the influence of preignition volatiles generation on flame speed is subject to clarification. YROLYSIS
l & E C PROCESS D E S I G N A N D D E V E L O P M E N T
of Technology, Cambridge, M a s s .
This investigation was designed to obtain pyrolysis information directly applicable to pulverized-fuel flames. T h e flame itself was used as the experimental system, but the results are believed to provide general information on coal pyrolysis. The object of this work is illustrated by its connection with other research in pulverized-fuel combustion. Considerable progress was made, beginning about 1959, by performing measurements in one-dimensional (flat) flames ( 3 ) . However, such studies had not been extended into the region of
Information on the pyrolysis of coal particles in pulverized fuel flames was obtained by using water-cooled instruments to probe a one-dimensional flame stabilized in a parallel-sided duct with a new type of "burner." The composition of the solid material and the flame temperature were measured along the axis of propagation. While providing a basic description of the kinetics of devolatilization, the study also produced several resullts contrary to initial expectations. Particles attain a temperature of about 1 100" C. before ignition or CI significant amount of devolatilization occurs, and the rate of devolatilization does not become rapid until a short time after ignition. W e conclude that ignition occurs on the surfaces of solid particles, since the amount of volatiles liberated before ignition is too small to form a flammable gaseous mixture. Devolatilization of particles in the range 0 to 200 microns seems to be a volumetric reaction, independent of particle size. The value obtained for the activation energy of devolatilization was 6 kcal. per mole in the preignition zone and 28 kcal. per mole in the postignition zone of the flame.
the flame between the ignition zone and the location of essentially complete pyrolyk. Since regions adjacent to it had been analyzed (7, 5 ) , this unexplored pyrolysis zone constituted a gap in the knowledge of the flame. The present investigation utilized the one-dimensional flame technique and was designed to fill this gap. Background
T h e amount of volartile material produced by pyrolysis of a given coal depends upon the temperature and duration of the process. Particle size is also an important variable if the particles are so large that the heating time required for the center of the particle is large in comparison with the duration of the process. Most investigators of coal pyrolysis fall into one of two broad categories. One group supports the proposition that the ratecontrolling step is the decomposition of the coal (78, ZZ), while the other advocates physical control, saying that decomposition is faster than diffusion of the volatiles to the surface of the particle ( 4 ) . A unifying irheory has not been produced. T h e lack of unification might be caused by the influence of experimental conditions on pyrolysis behavior, since various types of retorts and flow systems, various particle sizes, and various heating rates, have been employed. Considerable discretion must accompany the use of experimental results obtained under conditions differing markedly from those of the application. The pulverized coal flame differs enormously from systems in which pyrolysis has been studied. Instead of about 5 mm., the flame employs an average particle size of about 0.02 mm.; the time in the flame is about 1 second instead of about 30 minutes; and, instead of a heating rate of around 1' C. per second, the rate in the flame is as high as 104' C. per second. Therefore, results from previous experiments cannot be extrapolated with confidence to conditions found in the flame.
surface or volume reaction in the case of particle sizes in the pulverized-fuel range (0 to 200 microns) ; and whether the rate of pyrolysis determines the rate of ignition, and thus the rate of flame propagation. Experimental
A stoichiometric cloud of pulverized bituminous coal and air was passed down through a vertical plug flow reactor in which the particles experienced a known temperature profile established and maintained by the combustion of the coal itself. Upon entering the reactor, the particles were heated rapidly and then pyrolyzed as they passed down through the flame. The behavior of particles was observed by withdrawing samples of solid material from points spaced along the flame axis, determining volatile matter, fixed carbon, and ash content, and measuring the temperature profile along the flame axis. The data thus obtained consist of corresponding values of volatile content, fixed carbon content, temperature, and time. Equipment. The equipment for producing and maintaining the flame, including the furnace, the fuel and air supply system, and monitoring apparatus, is shown schematically in Figure 1. Further detail is presented elsewhere (9).
Contrary to our expectation, preliminary measurements revealed that heterogeneous combustion and rapid pyrolysis occur simultaneously, so that the flame cannot be separated into two distinct zones, each exhibiting only one of the two types of weight loss. Therefore, in addition to pyrolysis, the study was designed to include ignition and heterogeneous combustion, because of the expected interdependence of these three processes. T h e immediate goal was to measure the rates of both pyrolysis and heterogeneous combustion in the flame and to employ the results to determine characteristics of the kinetics of pyrolysis, such a*activation energy; whether pyrolysis is a
Figure 1. Schematic arrangement of experimental equipment for producing and maintaining flame V O L 6 NO. 1
FURNACE.The furnace (Figure 2) consists of a one-dimensional pulverized-fuel burner, a vertical combustion chamber of square cross section, and a flue connection. The furnace walls consist of insulating firebrick lining encased with Transite sheet. Observation and sampling parts are located along one side of the furnace. The most closely spaced sampling points occur just below the burner, where the rates of pyrolysis and temperature increase are most rapid. The burner, whose internal geometry is a truncated pyramid of square cross section, sits on top of the combustion chamber and, receiving the supply of fuel and air a t its truncation, expands the cloud so as to maintain a uniform velocity profile. At the base of the burner is a two-row bank of water-cooled tubes staggered to form a radiation trap; the burner and the I"
E perimental furnace
mixture flowing therein are thus maintained cold, and the cloud moves a finite distance into the combustion chamber before reaching ignition temperature. The flame is thus forced to stabilize a finite distance below the tube bank. If the tubes were absent, the flame would stabilize in the pyramidal section at a point where the cloud velocity equals the flame speed. Since the velocity profile in the cloud leaving the burner is uniform, the flame stabilizes with a flat front situated perpendicular to the axis of propagation, and thus assumes onedimensionality. Recirculation currents are absent, as verified by helium tracer texts (9) ; therefore, the history of material collected a t a given point can be specified with accuracy. I n this respect, the furnace is similar to one used a t the University of Sheffield ( Z ) , the principal difference being that, in the Sheffield furnace, the flame was stabilized in the diverging section (as described above). In our system, the flame is stabilized in a parallel-sided duct and is thus much closer to the idealized model assumed for mathematical analysis. FUEL AND AIR SUPPLYSYSTEM.The feed system (Figures 1 and 3) supplies fuel and air at adequately constant and controllable rates and mixture ratios. A metered stream of compressed air (primary air) collects the coal from a small funnel fitted into the throat of an asymmetric Venturi and injects it up against the truncation of the burner, where turbulent mixing occurs with a second stream of metered air (secondary air). The rate of coal feed is controlled by means of a vibratory feeder enclosed with the Venturi funnel in an air-tight box. A small hopper serves as a self-regulating valve between the primary (stirred) hopper and the rest of the feed system. The average variation in the rate of coal feed was measured to be 2.8%; the rate of air supply was essentially constant. MONITORING APPARATUS.Constant experimental conditions were ensured by monitoring the flame with three different instruments (Figures 1 and 2): The temperature along the inside surface of one wall was monitored by sheathed Pt/Pt 10% R h thermocouples and a 24-point potentiometer-recorder; the carbon dioxide content of the flue gases was monitored by a thermal conductivity instrument; and the pressure a t each end of the combustion chamber was monitored by an oil-type, inverted-bell draft gage. Solid-Sampling Probe, The water-cooled suction instrument (Figure 4) collects a representative sample of particles a t a point and quenches it a t a sufficiently rapid rate. The requirement that the probe be small enough to exert no significant influence on flame conditions resulted in locating the filter chamber outside the flame. Smaller probes could not be used because of clogging by sticky particles.
Figure 3. 76
I&EC PROCESS DESIGN A N D DEVELOPMENT
Fuel supply system
WATER O U 1
C O O L I N G 'WATER CHAMBER
L E A D W I R E TO
When the probe is in sampling position, the nozzle points vertically upward. T o collect a representative sample, flow disturbance is minimized by using a suction velocity of about five times the main stream velocity (72). A rotameter and valve in the suction line control suction velocity. A substantial amount of the sample does not reach the filter chamber but instead collects on the cold walls of the probe. The three individually cooled units of the probe are easily disassembled to facilitate saving the whole sample. The quenching ability of the probe was determined by measuring the temperature profile of sampled material flowing through the probe with a narrow-shield suction pyrometer. In the hottest region of the flame, samples are cooled from 1520' to 420' C. in 0.01 second, and to 25' C. in 0.1 second. Consideration of the pyrolysis rate at 420' C. and the speed of events in the flame indicates that this quenching rate is adequate. Suction Pyrometer, A single-shield suction pyrometer (Figure 5 ) was built for temperature measurement. The instrument uses a Pt/Pt 10% R h thermocouple, a ceramic radiation shield, and thermocouple protection sheath, and is composed of two individually cooled units and a filter chamber.
The suction pyromei.er measurements are taken as particle temperatures; although such instruments measure a temperature between that of the particles and that of the gas (73),our calculations indicate that the temperature difference between the two phases is negligible for particles smaller than about 200 microns. Probe Supporter a n d Positioner. The apparatus (Figure 6) consists of a frame which slides vertically in a graduated track along the front of the furnace. Sampling points are located with an error of less than 0.5 mm. in all three directions. ANALYSIS EQUIPMENT.The equipment is basically the standard ASTM apparatus for the proximate analysis of coal, modified to suit small samples (down to 50 mg.): Smaller crucibles are used to improve weighing accuracy; the residue from the volatile matter test is used in the ash test to eliminate the need for two different samples; and an atmosphere of dry nitrogen is used in the volatile-matter furnace to prevent com-
Probe supporter and positioner
bustion of samples which produce insufficient volatiles to displace ambient air. Measurements. EXPERIMENTAL CONDITIONS. The following conditions were employed: coal feed rate, 10 pounds per hour; fuel-air ratio, 0.135 ounce per cu. foot (at 1 atm. and 0' C.). VOL 6
JANUARY 1 9 6 7
The coal used (Grade E-5) was purchased in pulverized form from the Penn-Rilton Co. I t came from the Mathies mine, where it is taken from the Pittsburgh seam and ground in a Raymond hammer-screen mill. The particle size distribution (Figure 7 ) was measured with a Coulter counter. The proximate analysis of the coal is: ash, 3.86% of dry coal; moisture, 0.97% of raw coal; volatile matter, 37.35% of dry, ash-free coal; and fixed carbon, 62.65% of dry, ash-free coal. Results. In the following presentation, the data from a single experimental run are used for illustration. Data from the other runs (seven 16-hour runs were conducted) agree with those presented here (9). The basic data-Le., composition of solid material and flame temperature, both measured along the flame axis-are presented in Figures 8 to 11. Unexpected observations are that the particles attain a temperature of about llOOo C. before either ignition or rapid pyrolysis occurs; ignition precedes rapid pyrolysis; and a significant amount of volatile matter is carried in the unburned residue leaving the flame. This volatile residue persists over a substantial length of the flame and therefore appears to be a relatively stable component of the original volatile matter. Since the percentages of volatile matter and fixed carbon on a total weight basis are interdependent and time is more useful than distance as a kinetics parameter, the data are presented in Figures 12 and 13 with compositions expressed on the basis of original values and as functions of time. The conversion from distance to time employs the cold flow rate, the temperature profile, and the assumption that the number of moles of gas remains constant during combustion. The fractions of original volatile matter and fixed carbon are calculated using ash as a tracer.
accompanying loss of fixed carbon (Figures 12 and 13). After about 0.05 second, heterogeneous combustion, and hence the loss of fixed carbon, begins abruptly with a rapid rate a t a point corresponding to the location of the visually observed flames front (Figure 13). Soon after ignition, the rate of volatile loss becomes rapid, since volatile matter, in addition to being lost by pure evolution into the gaseous state, is also burned in the solid state along with fixed carbon. Because of the small decrease in volatile-matter content ahead of the flame front, the rate of combustion of “solid” volatile matter immediately behind the flame front (and before the principal decomposition) would be expected to be as fast as-the rate of combustion of the fixed carbon, since they are burning together. Furthermore, since the temperature of the particles reaches a substantial
> 0 IL
4 L W c
Analysis and Interpretation
Qualitative Analysis. BASICPYROLYSIS BEHAVIOR.The data presented above furnish a general picture of pyrolysis in the flame. Slow pyrolysis begins when the particles enter the hot combustion chamber a t the lower surface of the watercooled tube bank, and continues for a short time with no
D I S T A N C E BELOW B U R N E R , C E N T I M E T E R S
Figure 8. Volatile matter content during passage through flame
of solid material
z c 70 Y L
m u 4 0
Figure 7. runs 78
D I A M E T E R , MICRONS
Size distribution of coal used in experimental
I b E C P R O C E S S DESIGN A N D D E V E L O P M E N T
Figure 9. Fixed carbon content of solid material during passage through flame
Figure 12. Relationship between volatile matter content of solid material, temperature, and time
Figure 10. Ash content of solid material during passage through name I
E 3 Iu
D I S T A N C E BELOW B U R N E R , C E N T I M E T E R S
Figure 1 1. Suction-pyrometer temperature of flame and temperature of inside surface of combustion chamber
value behind the flarne front, the rate of evolution increases considerably in that region. After a period of rapid devolatilization which lasts for about 0.1 second, the rate of loss of volatile matter becomes so slow, although about 10% of the original volatile matter still remains in the particles, that about 5% of the original vo1 atile matter is carried from the combustion chamber in the unburned solid residue. T h e variation of the volatile matter-fixed carbon ratio, V/C, with degree of burnout (represented here by the fraction of the original fixed carbon remaining in the solids, C/C,) is
B U R N E R , SECONDS
Figure 13. Relationship between fixed carbon content of solid material, temperature, and time
indicative of the rate of volatile matter loss by gaseous evolution relative to the rate of loss by heterogeneous combustion (Figure 14). During the first 0.05 second in the chamber a small amount of volatile matter is evolved without heterogeneous combustion ; therefore V / C drops. Then, heterogeneous combustion begins at such a fast rate that the combustion of solid volatile matter is much faster than the volatile loss by gaseous evolution ; therefore, V / C remains essentially constant for about 0.03 second while around 10% of the original fixed carbon is burned. Meanwhile, the temperature of the particles becomes so high that the principal pyrolysis reaction, when it finally sets in, is very fast, and the loss of volatile matter by gaseous evolution is much faster than the loss by heteroVOL. 6
0 W P W
I J W
F I X E 0 C A R B O N C O N T E N T , P E R C E N T OF O R I G I N A L V A L U E
Figure 14. Variation of composition of solid material with degree of burn-out
geneous combustion. Thus, V/C drops rather rapidly for about 0.07 second while another 30% of the original fixed carbon is burned. At this point, only about 10% of the original volatile matter is left in the solid material, and the rate of evolution for this last portion is so low that the loss by heterogeneous combustion is again relatively fast and V/C decreases very little in the rest of the flame. Thus, volatile matter in the last 80% of the length of the flame is lost mainly by heterogeneous combustion. QUALITATIVE MODELS. A general, qualitative model, idealizing the progress of pyrolysis, which satisfies these data was selected by eliminating all but two alternative models because of obvious disagreements with the data. T h e difficulty of choosing between these two survivors required the more detailed analysis outlined below. Cenosphere formation is not included in these models, since the occurrence of cenospheres under the conditions of rapid combustion found in the flame is unlikely (75, 77,20,27). Model A. Devolatilization of each particle is assumed to occur uniformly throughout its volume; and heterogeneous combustion is assumed to occur a t the surface of the particle. Since the volatile material remaining in the particle at any time is uniformly distributed, the rate of combustion of solid volatile matter is always a certain fraction of the rate of combustion of fixed carbon, being equal to the ratio of volatile matter to fixed carbon, and varying through the flame as shown in Figure 14. Model B, Devolatilization is assumed to occur in a thin reaction zone located initially at the surface of the particle and then moving inward toward the center, leaving behind a porous matrix containing fixed carbon and a component of the volatile matter which is relatively slow to evolve. T h e unreacted core inside the reaction zone is identical in composition to the original coal. Heterogeneous combustion occurs a t the surface of the particle. The proportion of volatile matter burned in the solid state depends upon the relative speeds of pyrolysis and heterogeneous combustion : If pyrolysis is faster, 80
l&EC PROCESS DESIGN A N D DEVELOPMENT
the pyrolysis zone leaves the heterogeneous reaction behind, and the rate of combustion of solid volatile matter is a constant fraction of the rate of combustion of fixed carbon, the fraction being the ratio V / C in the porous, reacted matrix; if heterogeneous combustion is faster, both reactions occur on the surface of the particle and the loss of volatile matter by heterogeneous combustion can be taken as the total loss of volatiles. AGREEMENT OF MODELSWITH EXPERIMENTAL BEHAVIOR. Model A. In the region of the combustion chamber between the lower surface of the water-cooled tubes and the flame front, pyrolysis occurs uniformly throughout the particle and, since the surface of the particle has not yet ignited, no fixed carbon burns; therefore, the ratio of volatile matter to fixed carbon in the particle drops. Then, the solid particle begins to burn rapidly while pyrolysis continues at a relatively slow rate; the ratio of volatile matter to fixed carbon remains essentially constant, since both materials are burned in approximately the proportion in which they existed a t the onset of heterogeneous combustion. Next, the rate of pyrolysis increases to such an extent that the fraction of volatile matter in the burning particle drops very rapidly while the heterogeneous reaction continues relatively slowly. When only about 10% of the original volatile matters remains in the particle, the rate of pyrolysis becomes so slow that the main source of volatile matter loss is heterogeneous combustion. Therefore, since volatile matter is not completely eliminated from a particle until the solid material is completely burned away, particles which are too large to burn completely in the combustion chamber carry some volatile matter with them as they leave the chamber. Model B. The pyrolysis zone starts moving inward as soon as the particle enters the chamber. Until the flame front is encountered, no heterogeneous combustion occurs, so the ratio of volatile matter to fixed carbon drops. By the time the surface of the particle begins to burn, the pyrolysis zone is located a certain distance inward from the surface, so that heterogeneous combustion occurs at the surface of a porous material in which the ratio V / C is small and approximately constant. For a short time after ignition, the regression rates of the pyrolysis zone and the surface of the particle are such that V / C remains approximately constant for the particle as a whole ; then, because of the temperature increase, the pyrolysis zone regresses faster than the surface, so V / C decreases. When the pyrolysis zone reaches the center of the particle, the only volatile matter remaining is the component which is slow to evolve, so the main source of volatile loss in the rest of the combustion process is the heterogeneous reaction in which volatile matter and fixed carbon burn together in approximately the proportion in which they were left by the passing of the pyrolysis zone. As in the case of Model A, volatile matter is not completely eliminated from a particle until the solid material is burned away, so large particles carry volatile matter from the combustion chamber. Qualitatively, therefore, both models are identically comparable with observation. TESTS OF MODELS. According to Model A, pyrolysis occurs uniformly throughout the particle, so the rate of reaction at any time is proportional to the mass (or volume) of unreacted material (volatile matter) remaining in the particle-Le., pyrolysis is a volumetric reaction. Since the total mass (or volume) of material is independent of the size of the particles in which it is contained, this model indicates that pyrolysis should be independent of particle size; therefore, Model A can be eliminated by showing that the rate of pyrolysis is dependent upon particle size. According to Model B, pyrolysis occurs in a thin zone surrounding a core of undecomposed material. Since the mass
of volatile matter present in such a zone is proportional to the surface area of the particle, this model indicates that pyrolysis is a surface reaction whose rate is dependent upon particle size. Therefore, Model B can be eliminated by showing that pyrolysis is not a surface reaction, or that its rate is independent of particle size. Model B is eliminated by the following points: (1) T h e observed time required for pyrolysis is much larger than that predicted from Model B; (2) evidence from the literature (74) indicates there is no influence of particle size on the rate of pyrolysis for particles less than about 60 microns; and (3) calculation of the temperature distribution inside a coal particle in a flame indicates that pyrolysis could not be a surface reaction due to a temperature gradient inside the particle. Point 1. Essenhigh ( 6 ) found experimentally that the time required to pyrolyze burning coal particles in the size range 0.3 to 5 mm. is proportional to the square of the original particle diameter; he showed this result to be in agreement with the proposition that, in the. case of particles in the above size range, pyrolysis can be represented by a model in which the rate of devolatilization is governed by diffusional escape of the volatiles from a shrinking reaction zone to the surface of the particle. Model B is effectively identical to Essenhigh's model with regard to the diffusional escape of volatiles, and is therefore acceptable for the size range 0.3 to 5 mm. Assuming that the pyrolysis of particles in the pulverized fuel range (0 to 200 microns) also obeys Model B, two cases are possible: Either (A) Model B is obeyed and the rate of devolatilization is controlled by the rate of pyrolysis (chemical control), or (B) Model B is obeyed and the rate of devolatilization is controlled by the rate of diffusion of volatiles to the surface of the particle (physical control). The weight fraction of the original volatile matter, VIVO,remaining in the particle at time t is given by the following equations (Appendix A) :
(V/V,) = (1 - t/t$
- 3(V/V0)213+ 1 = t/t,
where t o = Kld, for Case A and t o = K2dO2for Case B; t , is the time required to devolatilize completely a particle of initial diameter do, and K1 and K2 are constants. We tested Model B by employing Equations 1 and 2 along with the particle-size distribution (Figure 7) to calculate the fraction of the original volatile material remaining in the particles a t any time after ignition. This was accomplished by separating the particles into several size fractions, analyzing each fraction individually, and then combining the results to construct the total volatile decay curve. Pyrolysis was assumed to begin a t the flame front, an assumption which is not far from correct (Figure 12). Essenhigh's (6) experimental value was taken for K z , and K I was estimated by assuming that chemical rate control is impending in the case of the smallest particles (0.3 mm.) studied by Essenhigh (6), thus providing the boundary condition Kid, = KzdO2at do = 0.3 mm. T h e results obtained are shown in Figure 15. The volatile decay rate predicted from Model B is far too large, with the predicted pyrolysis time being about one ninetieth of that expected. This conclusion holds whether physical or chemical control is accepted, and indicates that the dependence of pyrolysis rate on particle size, which is known to occur in the case of particles above 0.3 mm., is not obeyed by particles in the pulverized fuel size range. Therefore, Model B is unacceptable for particles in the pulverized fuel range.
E2 23 00 ~ PREDICTION FROM MODEL l lY M I C E AASSUMING L C O
l ( I I CHEMICAL CONTROL (21 PHYSlCaL CONTROL
II Il Il ' II ' 0 0 2 004 006 008 010 012 014 016 TIME AFTER IGNITION, SECONDS l
Figure 15. Comparison of experimental volatile decay rate with prediction from surface reaction model
Point 2. The above conclusion agrees with Ishihama's (74)observation that the influence of particle size on the lowerlimit concentration required for flammability of mixtures of bituminous coal dust and air diminishes as particle size decreases; below about 60 microns, the effect of particle size disappears. Since pyrolysis would be expected to play approximately the same role in flammability measurements as in a stabilized flame, this observation indicates that pyrolysis rate is independent of particle size for particles less than about 60 microns. About 75% (by weight) of the particles in the present investigation were less than this size; therefore, Ishihama's observation is applicable here. Point 3. An indication of the distribution of pyrolysis rate in a coal particle is obtained by estimating the temperature profile in the particle during heating, the assumption being that a steep profile would imply a thin pyrolysis zone owing to the strong temperature dependence of pyrolysis rate, while a flat profile would imply a uniform pyrolysis rate throughout the particle. T o avoid mathematical difficulties, a simplified case is considered which yields useful results, particularly of limiting conditions that overestimate the magnitude of the temperature gradient. The casz considered is a spherical bituminous coal particle, initially with a uniform temperature profile; the time required for the temperature of the center to rise by a certain fraction of a sudden increment experienced by the surface temperature was calculated as being indicative of the magnitude of the temperature gradient existing in the particle under real conditions. Whether the time required for the temperature of the center to approach closely to that of the surface is large or small on the time scale of events in the flame indicates whether the temperature profile in a particle being heated in the flame is steep or flat, respectively. General mathematical solutions in the literature (24) are used in this calculation along with the necessary physical properties of bituminous coal. T h e heat absorbed (or generated) by pyrolysis is neglected, since it has been observed to be small (76). The results (Figure 16) indicate that a significant temperature gradient probably could be neither created nor maintained in particles in the pulverized fuel flames. For example, if the surface temperature of a particle initially a t 100' C. should suddenly reach l l O O o C., about 0.02 second would be required for the center to reach 1090' C. in the case of a 100-micron particle, and less than 0.001 second would be '401. 6 NO. 1
in the solids is C and that in the original coal is C., The fraction V / V , is given in Figure 12, and the above integral is given by the area under the curve in Figure 14. The two volatile losses, calculated from Equations 3 and 4, are shown in Figure 17. About 70y0 of the original volatile matter is lost by gaseous evolution and about 25% by heterogeneous combustion; the other 5y0 remains in the solid residue. KINETICSOF VOLATILEEVOLUTION.Since we have found pyrolysis to be first-order with respect to the amount of undecomposed volatile matter, the rate of pyrolysis is expressed as
where k , is the rate constant, assumed to be given by the Arrhenius equation
NUMBERS ON L I N E S REPRESENT TEHPERb INCREASE OF CENTER OF PelRTICLE, EXPRESSED 4S PERCENT OF I N I T l 4 L TEMPER4TURE OIFFERENCE BETWEEN SURFICE 4ND CENTER
0 001 0
PART I C L E
61A M E TE R ,MICRONS
Figure 16. Influence of particle size on time required for center of spherical coal particles to heat by a certain amount
required if the particle size was 30 microns. Both these times are small, especially since the occurrence of such a drastic temperature gradient is unlikely and most particles in pulverized fuel are smaller than 100 microns. The conclusion is therefore that pyrolysis is probably not a surface reaction created by a nonuniform temperature distribution. Data from this and other investigations, as well as the above heat transfer analysis, therefore disprove Model B. The inference is that Model A, with the proposition that pyrolysis is a volumetric reaction, is therefore acceptable, as there are no experimental grounds for discarding it. Quantitative Analysis. MODESOF VOLATILELoss. A measure of the progress of pyrolysis is the amount of pyrolysis products evolved. This quantity is calculated as described below by separating the total volatile loss into two parts: that lost by gaseous evolution and that lost by heterogeneous combustion. At any stage in the combustion process, the weight of volatile matter, expressed as a fraction of the original volatile matter, V,, lost by gaseous evolution is
where V i s the weight of undecomposed volatile matter remaining in the solids and AVs is the weight of volatile matter consumed in situ by heterogeneous combustion. According to the pyrolysis model adopted above, AV, is given by the equation (4) where y is the ratio C/Co; the weight of fixed carbon remaining 82
l&EC PROCESS DESIGN A N D DEVELOPMENT
k, exp ( - E , / R T )
where k, is the frequency factor, E , is the activation energy, R is the ideal gas constant, and T is the absolute temperature, Pyrolysis in the flame occurs as though volatile matter in coal exists in two components-I and 11. Component I decomposes fast enough to be evolved in the flame; the devolatilization loss shown in Figure 17 is mainly the decay of this component, Component I1 constitutes about 15% of the total volatile matter in the coal and is mainly either lost by heterogeneous combdstion or passed from the combustion chamber in the unburned solid material. Division of the volatile matter into two components is in general agreement with the literature (7),and the postulation that the slow evolution of component I1 results in a significant amount of volatile matter leaving the combustion chamber in the solid residue agrees not only with our data but also with those of Saji (79). The composition of the solid material when component I has just been evolved corresponds roughly with the composition at Van Krevelen’s carbonization pole (23). Equations 5 and 6 were used to evaluate the activation energy for the evolution of component I. The quantity d(V/Vo)/dt is given by the slope of the devolatilization curve in Figure 17, and the quantity V / V , is obtained by deducting the contribution of component 11 from the volatile content shown in Figure 12. The activation energy, obtained from the slope of the lines in Figure 18, is different in the preignition and postignition regions of the flame. The average values found for the different experimental runs are as follows:
E , in the preignition region = 6 kcal./mole E , in the postignition region = 28 kcal./mole These values show that preignition pyrolysis is much less temperature-dependent than postignition pyrolysis. The major part of the material evolved before ignition probably consists mainly of carbon dioxide and water molecules which were held very loosely in the pores of the original coal. T h e activation energy found in the postignition region is about half as large as values reported (22) for coal carbonization, thus suggesting a difference between pyrolysis occurring in the flame and that occurring in carbonization tests. One possible explanation for this difference is that component I1 of the volatile matter, because of its high activation energy, decomposes slowly enough to be essentially inert with respect to pyrolysis in the flame and thus does not affect the experimental value of activation energy. In the slower carbonization process, however, component I1 has time to react and therefore could be responsible for the relatively high activation energy.
T I M E BELOW B U R N E R , S E C O N D S
Figure 17. Loss of volatile matter by heterogeneous combiistion and b y devolatilization
10001T , * K - l
Figure 18. matter
a concentration of combustible gases of only 0.02% by volume was found by using gas chromatography to analyze samples collected from the preignition zone of the flame with a watercooled probe, the existence of a combustible gaseous mixture a t the flame front appears unlikely. Assuming complete mixing, an analysis of the data a t the flame front shows that the maximum volatile concentration a t ignition is around 0.0026 gram per gram of air, a value which is short of a combustible mixture by a factor of 10 even if carbon dioxide and water are neglected and all the volatiles are assumed to be methane. If the gas analyses mentioned above are included in the analysis, again assuming complete mixing, the combustible concentration at ignition is found to be only about one three-hundredth of a flammable mixture. However, in spite of this evidence against gaseous ignition, the possibility is explored below that mixing of volatiles into the ambient air is slow enough to establish a flammable mixture near the surface of the particles prior to ignition. REALMIXINGCONSIDERED. Turbulent mixing is neglected, since little relative motion is expected between the particles and air in the preignition zone of the flame (Reynolds number in duct around 2000). Assuming spherical particles, the concentration profile of volatiles around the surface of the particle is described by the following steady-state diffusion equation :
Arrlhenius plot for evolution of volatile
Influence of Pyrolysis on Ignition ani Flame Propagation
Proposed Hypotheses. The role of pyrolysis in ignition and flame propagation is evaluated below by determining the phase (gas or solid) in which ignition originates. If ignition originates in volatiles, the rate of flame propagation is governed by the rate of pyrolysis ahead of the flame front; alternatively, if ignition originates on the surface of solid particles, the flame speed is independent of the kinetics of pyrolysis (70). These two modes of ignition constitute testable, alternative hypotheses. Test of Hypotheses. COMPLETE MIXINGASSUMED.The balance of the circumstantial evidence is unfavorable to the gaseous ignition hypothesis, thus implying that ignition originates on the solid surfaces of particles. Since only a small amount of pyrolysis occurs before ignition (Figure 8), and since
where K = (RTfwa2/3 DPM)d(V/VO)/dt;p 7 = mole fraction of volatiles at any distance, r , from the center of the particle; R = ideal gas constan&; T = absolute temperature; w = density of the coal; a = particle radius; D = coefficient of diffusion of volatiles into air; p = total pressure; M = average molecular weight of volatiles; f = weight fraction of volatile matter in the original coal; VIVO = weight fraction of the original volatile matter remaining in the coal; and t = time. Equation 7 indicates that the concentration of volatiles around each particle increases with increasing particle size. This condition is enhanced, but not created, by the pyrolysis model adopted, since, other things being equal, the flux of volatiles leaving the surface of a particle decomposing by a volumetric reaction increases roughly in proportion to particle size. T h e implication is that there exists a critical particle size below which the surrounding concentration of volatiles is too low to support ignition, and above which a flammable mixture surrounds the particle. For particles much larger than the critical size, the flammable mixture surrounding a particle may be displaced from the surface, with the surface conwntration of volatiles being above the upper limit of flammability. The experimental data were used as follows to estimate this critical size. The critical radius, a,, is given by Equation 7 after the condition is imposed that r = a = a, where p7 = p L , where p L is the mole fraction of volatiles in air a t the lower limit of flammability. Assuming the preignition volatiles to be chiefly methane, an assumption grossly in favor of the gaseous ignition hypothesis, p L is 0.05. The other terms in Equation 7 are evaluated as follows: d(V/Vo)/dt is the slope of the devolatilization curve (Figure 17) at the point of ignition; T i s the ignition temperature (Figures 12 and 13) ; and, since the volatiles are assumed to be methane, M is 16 grams per mole and D is 0.2(T/273' K.)'.'5 sq. cm./sec. ( 8 ) . T h e average critical size, 2a,, calculated for the different experimental runs is 130 microns. Owing to our neglect of turbulent mixing and the presence of carbon dioxide and water vapor in the preignition volatiles, this value, though VOL. 6
JANUARY 1 9 6 7
serving as a lower limit, is probably too low by a factor of 2 or 3. Since very few particles in pulverized fuel are as large as 130 microns, the critical size can be regarded as above the pulverized fuel range, thus ruling out the existence of particles surrounded by a combustible gaseous mixture. Accepted Picture. The conclusion from the above analysis is that the rate of transport of volatiles away from the surface of the particles in the preignition zone of the flame is fast enough to prevent the accumulation of a flammable gaseous mixture around the particle surfaces. Hence, ignition a t the flame front cannot occur in gaseous material, and the hypothesis of gaseous ignition must be excluded. The strong inference is, therefore, that ignition originates on the solid surfaces of particles, and that the rate of flame propagation is independent of the rate of pyrolysis. Conclusions
The particles attain a temperature of about l l O O o C. before either ignition or a significant amount of pyrolysis occurs. Ignition precedes rapid devolatilization and originates on the surfaces of solid particles instead of in gaseous volatiles; hence, the rate of flame propagation is independent of the rate of pyrolysis. Pyrolysis and heterogeneous combustion occur simultaneously. Volatile matter is thus lost from the particles by both gaseous evolution and heterogeneous combustion ; the former process accounts for about 3 / 4 of the total loss. Pyrolysis is slow enough in the tail of the flame so that about 5% of the original volatile matter leaves the combustion chamber in the unburned residue. Pyrolysis of particles in the pulverized-fuel size range (0 to 200 microns) is evidently a volumetric reaction occurring uniformly throughout each particle ; the rate of pyrolysis in the flame is therefore independent of particle size. The activation of pyrolysis is about 6 kcal. per mole in the preignition zone and about 28 kcal. per mole in the postignition zone of the flame. Pyrolysis occurs as though volatile matter exists in coal as two different components, one evolved very rapidly and the other very slowly. The slowly evolving component appears to represent about 1570 of the total volatile matter. Appendix
According to the surface reaction model of pyrolysis, a pyrolyzing spherical coal particle of initial radius r, can be visualized as a spherical core (radius = r ) of unreacted material surrounded by a spherical shell of thickness (Y, - r ) . Pyrolysis occurs at the surface of the core, producing volatiles which diffuse through the surrounding porous matrix to the surface of the particle. The following analysis produces two equations describing the relationship between the weight fraction, V/Vo, of the original volatile material left in the particle and time t : One equation describes the case in which the rate of devolatilization is controlled by the rate of diffusion of volatiles out of the particles (physical control), and the other describes the case in which the rate of pyrolysis is the limiting step (chemical control). Case I. Physical Control. According to Essenhigh ( 6 ) , the equation describing the above model under conditions of physical control should be of the form (7
-.(rO2/6t o ) dt
I & E C P R O C E S S DESIGN A N D DEVELOPMENT
where t o is the time required for complete devolatilization. Integration of Equation A-I between the limits r = Y, a t time t = 0 and Y = Y at t = t , together with the substitution ( r / r J 3 = (V/V,), produces the following results :
~(V/VO) ~ ( V / V O )f~ ’1~ = t / t o
Case 11. Chemical Control. At a given temperature, the rate of devolatilization is proportional to the surface area of the unreacted core, and is expressed as follows:
d(V/V,)/dt = K(r/Yo)2
where K is a constant. Since (V/V,) = ( r / r J 3 , Equation A-3 can be integrated between the limits (V/V,) = 1 at t = 0 and (V/V,) = (V/V,) at t = t . When a value is obtained for K by imposing the condition (V/V,) = 0 a t t = t,, the resulting expression is found to be: (A-5)
(1) Betr, J. M., Ph.D. thesis, University of Sheffield, Sheffield, England, 1960. (2) Betr, J. M., Thring, M. W., Mineral Industries Experiment Station, Pennsylvania State University, Bull. 75, 25 (1961). (3) Betr, J. M., Thring, M. W., Essenhigh, R. H., Combust. Flume 3, 557 (1959). (4) Berkowitz, N., Fuel 39, 47 (1960). ( 5 ) Csaba, J., Ph.D. thesis, University of Sheffield, Shefield, England, 1963. (6) Essenhigh, R. H., J . Eng. Power 85, 183 (1963). (7) Essenhigh, R. H., Howard, J. B., American Conference on Coal Science, Pennsylvania State University, 1964, Paper 44. (8) Fristrom, R. M., Westenberg, A. A., “Flame Structure,” p. 259, McGraw-Hill, New York, 1965. (9) Howard, J. B., Ph.D. thesis, Pennsylvania State University, University Park, Pa., 1965. (IO) Howard, J. B., Essenhigh, R. H., Combust. Flume 9, 337 (1965). (11) Howard, J. B., Essenhigh, R. H., Eleventh Symposium (International) on Combustion, Pittsburgh, Combustion Institute, 1966, Paper 38. (12) Howells, T. J., Be&, J. M., Fells, I., J. Znst. Fuel 33, 512 (1960). (13) Hubbard, E. H., Pengelly, A. E., Symposium on Flames and Industry, London, 1957, p. A-9, Institute of Fuel, London, 1957. (14) Ishihama, W., Eleventh International Conference of Director of Safety in Mines Research, Warsaw, Poland, 1961. (15) McCulloch, A., Newall, H. E., Sinnatt, F. S., J . SOC.Chem. Znd. 46,331T (1927). (16) Millard, D. J., J. Znst. Fuel 28, 345 (1955). (17) Newall, H. E., Sinnatt, F. S., Fgel 3, 424 (1924); 5 , 335 (1926); 6, 118 (1927). (18) Pitt, G. J., Zbid., 41, 267 (1962). (19) Saji, K., Fifth Symposium (International) on Combustion, p. 252, Reinhold, New York, 1954. (20) Sinnatt, F. S., Fuel 8, 362 (1929). (21) Slater, L., Sinnatt, F. S., Zbid., 1, 2 (1,822). (22) Van Krevelen, D. W., Schuyer, J., Coal Science,” p. 295, Elsevier, New York, 1957. (23) Van Krevelen, D. W., Van Heerden, C., Huntjens, F. J., Fuel 30,253 (1951). (24) Williams, E. D., Adams, H. L., Phys. Reu. 14, 99 (1919). RECEIVED for review January 7, 1966 ACCEPTED October 26, 1966 Division of Fuel Chemistry, 151st Meeting, ACS, Pittsburgh, Pa., March 1966. Work carried out in the Combustion Laboratory, Department of Fuel Science, The Pennsylvania State University, University Park, Pa. Joint financial support, and permission to publish, received from the Babcock and Wilcox Co. (Alliance Research Station) and the State Funds of the Commonwealth of Pennsylvania for Coal Research.