I n d . E n g . C h e m . Res. 1987, 26, 1831-1838 Patzer, J. R.; Montagna, A. A. Ind. Eng. Chem. Process Des. Deu. 1980, 19, 382. Ruberto, R. G.; Cronauer, D. C.; Jewell, D. M.; Seshadri, J. S. Fuel 1977, 56, 25. Vernon, L. W. Fuel 1980,59, 102. Watanabe, Y.; Yamada, 0.;Fujita, K.; Takesami, Y.; Suzuki, T. Fuel 1984, 63, 752.
1831
Wiser, W. H. Chemistry of Coal Liquefaction: Status and Requirements; Cooper, B. R., Ed.; Scientific Problems of Coal Utilization, US DOE Symposium Series 46; US DOE: Washington, DC, 1978; Vol. 219.
Receiued for reuiew February 28, 1985 Accepted June 18, 1987
Kinetics of Vapor-Phase Secondary Reactions of Prompt Coal Pyrolysis Tars Michael A. Serio,+William A. Peters, and Jack B. Howard* Department of Chemical Engineering and Energy Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139
T h e kinetics of vapor-phase secondary reactions of newly formed coal pyrolysis tars were studied a t temperatures and residence times of 500-900 "C and 0.6-3.9 s, respectively. Tar vapors, generated by heating a helium-swept, shallow packed bed of Pittsburgh No. 8 bituminous coal from room temperature t o 550 "C, a t 3 "C/min, were either rapidly conveyed t o collection traps for subsequent characterization or else passed through an adjacent reactor for controlled thermal treatment prior to trapping. For short-duration (0.6-1.1 s) thermal treatment, tar conversion was insignificant below 600 "C but extensive (30-5070) a t 700-800 "C, with light gases as the major products. Conversion increased with increasing treatment temperature and reached 60% a t 900 "C and 1.1-s residence time. At 0.6-s residence time, conversion increased only modestly between 750 and 800 "C. There was little effect of tar formation temperature on the tar thermal reactivity. Conversion kinetics were well described by a model treating tar as three noninteracting lumps of different thermal stability. Each lump was assumed to react by one independent parallel first-order reaction, and one lump remained thermally inert a t the highest severities here studied. A multiple, independent, parallel, first-order reaction model also performed well but could not predict step-like behavior in conversion a t 0.6-s residence time. A single-reaction first-order decomposition model correlated the conversion data poorly a t 0.6- and 1.1-sresidence times, and use of the resulting best fit parameters gave totally inadequate predictions of conversions measured a t 2.5 and 3.9 s. Secondary reactions in coal technology refer to thermal chemical transformations of prompt (i.e., newly formed), pyrolysis-derived volatiles. Depending on reactor geometry, operating conditions, and coal type, secondary reactions can exert minor to dominant influence on coal conversion and combustion behavior including product distributions between solids, liquids, and gases; volatiles compositions and heating values; char reactivity (by obstruction of pores or active sites); disposition of sulfur, nitrogen, and oxygen among products; formation of soot, polycyclic aromatic hydrocarbons, and other complex organic pollutants; ignition and flame stability; and the magnitude and duration of plasticity. Plasticity impacts a wide range of coal utilization phenomena (Fong et al., 1985), including oven operability and product quality in metallurgical coke making; particle swelling, and agglomeration in moving and fluidized bed reactors; cenosphere formation in pulverized coal combustion; and mass transfer in liquefaction, pyrolysis (Oh, 1985), and hydropyrolysis/ hydrogasification (Schaub et al., 1985a,b). Tar secondary reactions are of particular interest because previous studies (Peters and Bertling, 1965; Suuberg, 1977) show tars to be more susceptible to secondary reactions than other volatiles (gases and light oils) and because tar accounts for the largest fractions of primary volatiles yields and heating value for most bituminous coals. Secondary reactions are complex, being influenced by coal type, heating rate, residence time, temperature, intra- and extraparticle heat and mass transfer, and
* To whom correspondence should be addressed. 'Current address: Advanced Fuel Research, Inc., East Hartford, CT 06118. 0888-5885/87/2626-1831$01.50/0
Table I. Characteristics of Coal Studiedab ultimate anal., wt %, proximate anal., wt %, dry dry
C H 0 N
S ash
73.6 5.2 7.1 1.3 5.3 10.3
moisture volatile manner fixed C ash
0.6d 41.6 48.1' 10.3
heating value,' kJ/g
30.7
"Coal type: Pittsburgh Seam (No. 8) bituminous coal *Analyzed by Huffman Laboratories, Inc. By difference. Dried basis. e l kJ/g = 430 BtU/lb.
physical structure of the reacting coal (porous, softened, swollen, agglomerated, etc.). Further, these reactions can be heterogeneous (vaporsolid, vapor-liquid, or liquid-solid processes) or homogeneous (vapor phase or liquid phase). This paper presents results of systematic studies of the independent effects of temperature (500-900 OC) and residence time (0.6-3.9 s) on the vapor-phase secondary reactions of prompt tar from a Pittsburgh Seam bituminous coal. Vapor-solid secondary reactions of this tar in the presence of activated carbon or coal char have also been studied (Serio, 1984; Serio et al., 1983; Suzuki, 1984).
Experimental Section Coal Characteristics. Pittsburgh No. 8 Seam bituminous coal from the Ireland Mine of the Consolidation Coal Company was studied. This same coal type was used by Anthony (1974) and Suuberg (1977) in investigations of rapid pyrolysis behavior and by Franklin (1980) in a 0 1987 American Chemical Society
1832 Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987 REACTOR -
- -~ ~
I t-
REACTOR - 2
I
I (LIGHT G4S)-1
I
/
Figure 1. Schematic of apparatus.
study of mineral matter effects during pyrolysis. It is typical of highly softening eastern U S . bituminous coals. Table I presents an ultimate analysis for the particular sample studied here. Apparatus and Procedure. The apparatus was designed to subject newly formed coal pyrolysis tars to controlled extents of postpyrolysis thermal treatment in an environment where temperature, residence time, and total pressure could be independently selected and controlled. It consisted (Figure 1)of two independently heated tubular chambers connected in series. A thin bed of about 1 g of coal, diluted with approximately 8 g of sand to prevent agglomeration, was pyrolyzed in the upstream chamber (reactor 1)at a low heating rate (3 "C/min) and simultaneously swept with a carrier gas (usually helium). Fresh tars were generated by heating the coal from room temperature to maximum temperatures below those favorable to secondary reactions (-550 "C). The tars and other volatiles were rapidly removed from reactor 1and either collected for characterization (control runs) or conveyed to an immediately downstream thermal treatment chamber (reactor 2), for secondary reactions, and then to the collection traps. The volumetric flow rate of carrier gas was sufficiently high that volatiles caused little perturbation to the reactor throughput. Temperatures and average, V / F ,residence times in reactor 2 were maintained constant in the ranges 500-900 "C and 0.6-3.9 s, respectively. These conditions provide measurable tar conversions but cause little, if any, thermal cracking of light oils and light hydrocarbon gases or gasification of tar or coke by C 0 2 or steam (Serio, 1984). Formation of additional volatiles or coke in reactor 2 could therefore be attributed primarily to tar secondary reactions. Figure 2 summarizes the global reaction pathways postulated for each reaction chamber. The tar traps (Figure 1) were made of 7.6-cm lengths of 1.27-cm-0.d. X 0.09-cm wall Type 316 stainless steel tubing packed with Teflon fibers (Alltech No. 4082). The tar yields were determined by weight gain of the traps. The results agreed to within 5% (by weight tar) with the amount of tar recovered by extracting the traps with 2:l v/v methylene chloride/methanol at room temperature and then evaporating the solvent under nitrogen at 50 "C. Tars were continuously collected throughout the entire run by replacing traps at known times after startup. Each trap therefore contained tar derived from a distinct, well-defined stage of coal decomposition. Roughly 25% of the total tar from a given run was recovered from the end of reactor 2 and from a three-way valve (used to switch the reactor effluent between the tar traps), by successively rinsing these regions a t ambient conditions, with methylene chloride, methanol, and pyridine and then evaporating the solvents under N2at 50 "C. This contribution was assumed to be distributed among
I
*
I I I
hl-CHAR-1
GAS)-2
I
I I
Figure 2. Global reaction pathways in reactor system.
the collection traps in proportion to their content of tar. The validity of this assumption was supported (A) by a series of measurements showing that for each interval of tar generation the weight of tar deposited upstream amounted to about the same percentage of the corresponding weight of tar in the collection trap and by (B) GPC measurements showing nearly identical molecular weight distributions for tar accumulated upstream throughout a run and the corresponding cumulative tar make obtained by combining the tar from all the collection traps. Gaseous products were sampled intermittently with a heated (-120 "C), multiloop collection valve and analyzed by gas chromatography (Serio, 1984). The resulting concentration data were combined with measured flow rates of the carrier gas (which contained 1-2% Ar tracer) to compute product yields. Light oils were collected downstream of the gas sampling valve in room-temperature 7.6-cm X 0.95-cm-0.d. X 0.09-cm wall Type 304 stainless steel tubes packed with activated coconut charcoal (Analabs No. GCA-005). A gas chromatographic analysis indicated that benzene, toluene, and xylenes were major consitituents in the light oils. The net rates of species (tar, gas, and oil) formation or destruction were determined from the differences in the product yields measured in the secondary reaction and control experiments. The overall material balances were 95-105%. Since product tars were collected over known intervals of generation temperature, kinetic data could be obtained for the whole tar mixture evolved from reactor 1 and also for different fractions of this mixture evolved throughout known ranges of coal devolatilization temperatures. To facilitate kinetic and mechanistic interpretation of the raw conversion data, samples of primary and secondary tar (tar 1 and tar 2, respectively, Figure 2) were characterized by 'H NMR (average molecular structural parameters); gel permeation chromatography, GPC (molecular weight distribution); gravity column elution chromatography (preliminary fractionation by polarity); and C, H, N, 0 and S elemental content using standard methods. Procedures and results are reported by Lennox (1983), Lennox et al. (1987), and Serio (1984).
Results and Discussion Tar Yields from Primary Pyrolysis in Reactor 1. Cumulative tar yields from pyrolysis of the coal from ambient to different final temperatures are shown in Figure 3. To ensure that tar generation attained completion, the highest final temperature (550 "C) was maintained for 30 min. Tar vapor exiting reactor 1is "primary" in the sense
Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987 1833 Table 11. Kinetic Parameters for Formation of Tar model E or E,,, u, V*, std error type k,, l/s kcal/mol kcal/mol w t % coal of estimate ~~
F_Ok
(SCCM)
O
i
15.01
,
250 750 750 75c
single 4.07 reaction multiple 4.38 reaction 3.90
t 22.9
U
c 4 w
24.8 023.4 x 24.1
0
c>
21
8
1
l0.Ot
200
300
400
500
--A HO.D
5%
r Jn
MIN
tions:
(-e)
single reaction model; (-)
multiple reaction model].
that once it has escaped the coal particles, it has undergone minimal extraparticle secondary reactions. Supporting evidence includes the following observations. The cumulative tar yield (24 wt % dry coal) (shown as the right most data point in Figure 3) is very close to the 25 w t % tar make obtained by pyrolyzing thin layers of the same coal a t 1000 "C (1atm of pressure), in an apparatus designed to minimize extraparticle secondary actions (Suuberg, 1977; Suuberg et al., 1979b). The Suuberg and present tar had similar elemental analyses. In the present work the residence times (0.03-0.14 s) and temperatures (2550 "C) of tar vapor in reactor 1 are disfavorable to vapor-phase tar cracking (see Figure 7), and only small yields of gaseous products typically associated with tar secondary cracking (CO,Hz, and CZHJ were evolved from reactor 1 (Serio, 1984; Serio et al., 1987). "Primary" in the present study does not exclude the possibility that the yield and composition of tars evolving from reactor 1 may still reflect some influence of secondary reactions within the coal particles prior to tar entrainment in the carrier. Studies of such "intraparticle" secondary reactions were not the focus of this work but remain an important research area. Similar profiles of tar yield vs. generation temperature were observed a t carrier flow rates of 500 to 1500 sccm (standard cm3 m i d ) (Figure 3), and tar global composition data (Lennox, 1983; Lennox et al., 1987; Serio, 1984) showed little influence of carrier throughput rate. Thus, the tar generation rates were not determined by masstransfer resistance in the coal-carrier boundary layer, and carrier gas flow rate could be varied to change reactor 2 residence times without disturbing the amount of tar vapor delivered to this reactor. At carrier flows of 250 sccm (Figure 3), the tar yields are somewhat different from those a t the higher carrier rates. A possible explanation is that the normal correction procedure for upstream tar condensation (see above) was invalid a t the extensive tar deposition rates (up to 50% of the total catch) observed a t this flow rate. Figure 3 also shows that at this heating rate the maximum rate of tar formation occurs a t relatively low temperatures (approximately 450 "C) and that over 90% of the tar is formed below 500 "C. The repeated experiments a t 750 and 1500 sccm show that the data reproducibility is good (f5%). Kinetic Modeling-Primary Pyrolysis Kinetic expressions for tar yield from primary pyrolysis were needed (1)to provide kinetic data on tar formation in the absence of significant secondary reactions and (2) to model the simultaneous formation and cracking of tar.
lo6
27.7
X X
lo7 1013
34.1 52.8
1.1 2.5
23.27 (24.21)' 23.49 23.63
0.638 0.616 0.628
"Defined as [~.3=1(V~model - VJ,,,,J2/(n - npsrems)]l/z where n = number of data points and nparams = number of parameters in the kinetic model. *Measure (as opposed t o curve-fitted) value of V*.
The primary pyrolysis data were correlated with the following single and multiple first-order reaction kinetic models, described in more detail by Howard (1981):
REACTOR-1 TEMPERATURE ('C)
Figure 3. Comparison of cumulative tar yield data for base-line experiments performed at different carrier gas flow rates [predic-
X
V = V*[ 1 - exp(-koJt
exp(-E/RT) d t ) ]
(1)
and
V = V*[ 1 -
Jm
exp(-koxt exp(-E/RT) dt) f ( E ) dE] (2)
with
The quantities ko, E , V*, Eo, and a are the preexponential factor (SI), the activation energy (kcal/mol), the ultimate tar yield (wt % daf coal), and the mean and standard deviations of the activation energy distribution function (eq 3), respectively. Best fit values for these parameters were obtained by numerically evaluating the integrals in eq 1 and 2 by using the measured temperature-time history of the coal in reactor 1 and a nonlinear least-squares regression code (Sawada, 1982). Results are presented in Table 11. Parameters for formation of other products such as low molecular weight hydrocarbons are reported elsewhere (Serio et al., 1984, 1987). Both models correlate the present data satisfactorily (Figure 3), as evidenced by their low standard errors (Table 11). For the multiple reaction model, two different initial guesses for the fitted parameters caused the regression routine to converge to two different sets of parameter values. Each set correlated the observations with virtually the same statistical quality. Such multiple solutions are due to compensation between the parameters KO and Eo (Howard, 1981). Kinetic data on the effects of heating rate on products yield would provide a basis for statistical differentiation among sets of fitted parameters. The rate of tar evolution predicted by any of the parameter sets (Table 11) at 450 "C is approximately s&, which is in good agreement with previous results for tar formation from bituminous coal pyrolysis (Suuberg et al., 1979a) or from ethylene-bridged polymers (Solomon and King, 1984). The 52.8 kcal/gmol activation energy for the second set of multiple reaction model kinetic parameters is reasonable if one assumes that the molecular precursors of bituminous coal tars are linked by ethylene or benzyl phenyl ether bonds with dissociation energies of about 55 kcal/gmol (Gavalas, 1982). Tar Yields a f t e r Homogeneous Secondary Cracking. The cumulative yields of tar collected from homogeneous cracking at average residence times of 1.1and 0.6 s are shown in parts a and b of Figure 4, respectively, together with the corresponding base-line (no-cracking) data obtained at similar carrier flow rates. The dotted lines in this figure depict the cumulative tar yields predicted by a tar formation and cracking model discussed below. These curves were differentiated graphically to determine
1834 Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987
I Cl
Reactor-1 Temperature
s ,.
t
83
0.4t
2
; 0.3.
"
I
J
" ?
"'y, 0.21
:
t
2
0 0
,
250
25i
35c
450
550
r
Reactor-1 T e m p e r a t u r e I'CI
Figure 4. Cumulative tar yield data for vapor-phase cracking experiments performed a t different reactor 2 temperatures a t (a, top) 1.1-sresidence time and (b, bottom) 0.6-s residence time [predictions: (-) single reaction tar formation model; combined tar formation/tar cracking model]. (-e)
the rate of tar collection as a function of reactor 1 temperature. The results for parts a and b of Figure 4 are shown in parts a and b of Figure 5, respectively. The maximum rate of tar collection in the secondary reactions runs occurs at about the same reactor 1 temperature (-450 " C ) as does the maximum rate of tar generation. This is consistent with our assumptions that tar cracking kinetics are independent of tar formation temperature and firstorder in tar concentration. Both figures show that tar conversion is insignificant up to 600 "C but becomes extensive at 700-800 "C (30-50%). Modeling Secondary Conversion of Tar Vapor. Kinetic parameters for tar secondary reactions were derived from data on tar conversion at different temperatures and residence times. Fractional conversion, x , was defined as x =
350
450
]
wt of tar collected downstream of reactor 2 (4) wt of tar entering reactor 2 l-[ The denominator in eq 4 was calculated by correcting for the amount of tar which deposited between reactor 1 and reactor 2, which normally amounted to about 5 wt % of the total tar evolved from the coal. An isothermal version of the single reaction model previously used to describe tar generation (eq 1)was employed to fit the tar conversion data by substituting x = 1 - V / V * ,where V* is now the cumulative amount of tar that entered the cracking reactor and V is now the cumulative amount of tar remaining after reaction at temperature T for a residence time, T . The integrated form of eq 1 then gives In (1 - x ) = -k07 exp(-E/RT) (5)
650
550
R e a c t o r - 1 Temperature
I Cl
Figure 5. Predicted rates of tar collection for vapor-phase cracking experiments performed at different reactor 2 temperatures a t (a, top) 1.1-sresidence time and (b, bottom) 0.6-s residence time [predictions: (-) derivative of tar formation curve generated in Figure 4 using single reaction model; (--) derivatives of cumulative yield curves generated in Figure 4 using combined model]. Table 111. Kinetic Parameters for Homogeneous Reaction 'of Tar std error E or E,, 0, wt 70 of '0% kcal/mol kcal/mol tar estimate" Single-Reaction Model 1. 5.42 x 104 24.0 lOOb NA 2. 9.96 X 10' 15.4 lOOb 0.006
v*,
':'
3.95 x 109 A.
B.
c.
1.43 X loe 1.39 X 10"
Multiple-Reaction Model 49.5 8.1 lOOb Three-Lump Model 35.3 56.1
0.040
33 27 40 0.032
Defined in Table 11. bFixed.
Dividing by -T and taking the natural logarithm of both sides gives
The expression in brackets is equivalent to the rate constant, k . For ideal first-order behavior, a plot of the left-hand side of eq 6 vs. 1 / T would give a straight line of slope -E/R and intercept In (k,), allowing E and ko to be readily calculated. This plot and the resulting values for E and k , are shown for the present data in Figure 6 and Table I11 (set l), respectively. Departures from ideal first-order behavior are evident in Figure 6. Numerical regression of the conversion data with eq 6, using a nonlinear fitting algorithm and the set 1 values as initial
Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987 1835
-t 3
2
A 0.6 I Residence Time 0 1.1
s Residence Tine
1.0
0.8
1.2
1.4
Reciprocal Absolute Temperature x lo3
500
Figure 6. Arrhenius plot for vapor-phase tar conversion data based on an isothermal single-reaction first-order model [(-) present data, least-squares fit, k,, = 5.42 X lo4 s-l, E = 24.0 kcal/mol; (---) Wen and Cain (1984),KO = 8.18 X lo3 s-l, E = 20.6 kcal/mol]. 1.0 0 1 . 1 s Residence T i m e
0
0.6
0.5 0.4 0.3
$
0.2
o.lj 0.0 500
600
700
800
900
R e a c t o r - 2 Temperature (‘Cl
F i g u r e 7. Tar conversion data for vapor-phase secondary reaction experiments a t 1.1-s and 0.6-8 residence times compared with predictions from a single-reaction first-order model using parameters from set 2,Table 111.
guesses for E and ko, gave revised E and ko values (Table 111, set 2). The conversions predicted with the second set of parameters a t residence times of 0.6 and 1.1 s are shown in Figure 7. As expected from the nonidealities in the single-reaction Arrhenius plot, the fit is at best fair. The magnitudes of the rate constant calculated with the first set of parameters are in reasonable agreement with those derived by Wen and Cain (1984) (dashed line in Figure 6), from experiments on the secondary reactions of a light fraction of tar obtained as a byproduct of fixed bed gasification of Pittsburgh Seam bituminous coal. Both sets of single reaction parameters in the present study are low compared to those expected for typical organic decomposition reactions (Suuberg, 1977; Howard, 1981). A possible explanation is that tar conversion is a consequence of several rate phenomena. Forcing tar conversion data to be fitted by a single-reaction first-order kinetic model then predisposes the derivation of low E and ko values (Jungen and Van Heek, 1970; Howard, 1981). Best fitting the multiple reaction model, eq 2, to the tar conversion data by using eq 4 gave the predictions shown in Figure 8. The magnitudes of the corresponding fitted kinetic parameters are closer to those expected for pyrolysis of organic vapors than are the values from the single reaction fits (Table 111).
600
700
R e a c t o r - 2 Temperature
(K-’)
800
900
L’CI
Figure 8. Tar conversion data for vapor-phase secondary reactions of tar a t 1.1-s and 0.6-sresidence time compared with predictions from a multiple-reaction mode’ eq 2,using parameters from Table 111.
The multiple reaction model is more mathematically complex than the single reaction description. Neither model predicts the insensitivity of conversion to temperature between 750 and 800 OC a t a residence time of 0.6 s (Figures 7 and 8), the steplike evolution of gaseous products of tar cracking with increasing temperature (Serio et al., 1984, 1987), or similar steplike behavior in the fraction of aromatic hydrogen (Figure 9a) and the fraction of aromatic carbon (Figure 9b) in the tar surviving thermal treatment (Lennox et al., 1987). In response, a lumped kinetic model was tested. This description assumes that the whole tar consists of just three, noninteracting fractions (lumps) of distinct chemical reactivity, each of which pyrolyzes by one independent parallel first-order reaction. Here the first lump, “A”, is the most reactive and decomposes at relatively low severities. The second lump, “B”, is moderately reactive and decomposes with a higher global activation energy than lump A. The third lump, “C”, is unreactive at the highest severity studied. The percentages of A, B, and C in the unconverted whole tar, 33,27, and 40 wt % , respectively, were inferred from apparent plateaus in the conversion vs. temperature data (Figure 10). The integrated form of this three-lump model gives the following expression for tar conversion: x = 1 - [0.33 exp(-KAT) + 0.27 exp(-KBT) + 0.401 (7) where KA = KAo exp(-EA/RT,)
(8)
KB = KBo exp(-EB/RT,)
(9)
The Arrhenius parameters, KAo,EA, KBo, and EB (Table 111),were obtained by best fitting the measured conversion data by using a nonlinear least-squares routine. The model predictions of whole tar conversion, and of conversion of lumps A and B, are shown in Figures 10 and 11,respectively. Quantitative insights on the fraction of the whole tar in each lump and on lump reactivity can also be drawn from information on certain global structural features of the tar, inferred from ‘H NMR measurements (Figure 9) (Lennox, 1983; Lennox et al., 1987). It was assumed that tar conversion was directly proportional to the fraction of aromatic hydrogen (Figure 9a), or the fraction of aromatic carbon (Figure 9b), in the tar which survived thermal treatment. The changes in either parameter with temperature were fitted to a three-lump kinetic model similar
1836 Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987
u
3
0 6 -
0 5-
a
: ' :
0.0
0 . 0 500
700
600
800
900
8
,
,
,
I
,
,
:e 0.B5t
a
r
2
c eo1
800
900
Figure 11. Predicted conversion for tar lumps A and B according to three-lump model, eq 7, using parameters listed in Table 111.
YI
l-l
X
.'J*?F
"B"
0 65
0 60 500
700
Reactor-2 Temperature ( ' C )
Reactor-2 Temperature ( 'Cl 1.00,-
600
500
600
700
800
900
Reactor-2 Temperature ( ' C i
Figure 9. Effects of vapor-phase cracking of tar at residence times of 0.6 and 1.1 s, on structural properties of surviving tars: (a, top) fraction of aromatic hydrogen; (b, bottom) fraction of aromatic carbon; (-) predictions of a three-lump kinetic model (see text). Adapted from Lennox (1983).
Figure 12. Comparison of rate constants for vapor-phase tar cracking determined by measuring net tar yields after controlled thermal treatment, with rate constants derived from changes in average structural parameters of the tars which survived thermal treatment under corresponding reaction conditions. (See text.) [Three-lump model used in each case; (-) from tar yield data (Figure 11);(-) from aromatic hydrogen data (Figure 9a); (---) from aromatic carbon data (Figure 9b). Values for Arrhenius kinetic parameters for Figure 9 are given by Lennox et al. (1987)l.
0 1.1 s Residence Time
0.0
500
600
700
Reactor-2 Temperature
10-31 0.6 0.8 1.0 1.2 1.4 R e c i p r o c a l A b s o l u t e Temperature x 103 (K-')
BOO
900
('C)
Figure 10. Conversion data for vapor-phase secondary reactions of tar at 1.1-s and 0.6-s residence times compared with predictions from a three-lump model, eq 7 , using parameters listed in Table 111.
to that described above. The fraction of the original primary tar in each lump was inferred from apparent plateaus in the NMR data. The derived rate constants are in good agreement with those obtained by direct measurements of tar conversion (Figure 12). Extending the three-lump model to higher temperatures would require data a t tar conversions exceeding 60% to provide kinetic parameters for decomposition of lump C. Experiments to this end must recognize that at tempera-
tures somewhat above 900 "C, carbon (soot) formation would be an expected product of tar pyrolysis at the residence times of current interest (0.6-3.9 s) (Nenninger et al., 1983). The statistical quality of the data fit (Table 111) is slightly better for the three-lump model than for the multiple reaction model. The parameters for the threelump model displayed in Table I11 were derived by inferring the fractions of untreated whole tar in lumps A and B from corresponding plateaus in the conversion vs. temperature curves (Figure 10). Thus, the three-lump model required four fitted parameters (KAo,EA, KBo, and EB) vs. three (k,,, Eo, and a) for the multiple reaction model (since V* was taken at 100%). When all model parameters are fitted, the three-lump model requires eight fitted parameters (the above four, plus KCo, EC, and the fractions of the original whole tar in any two lumps) and the multiple reaction model requires four parameters (since V* is also fitted). Factors in selecting a model will include magnitude and quality of the data set, mathematical
Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987 1837 1 . 0 0
o,61 0.5
I
I
I
I
I
I
, !
/
-
1
/
0
;j
6
0.4
';I 0.3
s
0.2
0.0
500
600
700
800
R e a c t o r - 2 Temperature
500
tractability, ability to reproduce distinct structural features in the data (e.g., steplike variations in conversion vs. temperature plots), and reliability in correlating observed behavior over wide ranges of operating conditions and absolute conversion. Mechanisticdy, a three-lump mode would be consistent with the existence in tar of three broad classes of chemical bond strengths. The bonds in any one class are postulated to be of similar strength and to decompose a t about the same rate. However, the bonds in the different classes have quite different strengths and therefore decompose at different rates. Product spectra from thermal cracking of the tar vapor (Serio 1984, Serio et al. 1987), and global tar characterization data (Lennox, 1983; Lennox et al., 1987; Serio, 1984), imply that decomposition of tar lump A involves breaking paraffins, etheric linkages, and side groups and that B lump reactions include dehydrogenation, dehydroxylation, demethylation, and disruption of heterocyclic rings. The C lump includes the resonance-stabilized carbon-carbon bonds constituting the aromatic structures in the tar. These observations suggest that vapor-phase secondary reactions of tar occur primarily at the periphery of aromatic rings for the temperatures (500-900 " C ) and residence times (0.6-3.9 s) studied. Longer Residence Time Results. A few runs were performed at longer residence times, 2.5 s (700, 800 "C) and 3.9 s (600, 700 "C), to assess the range of applicability of the three-lump model. Figure 13 shows that there is reasonable agreement between the measured conversions a t 2.5 and 3.9 s and corresponding model predictions using the kinetic parameters derived from the lower residence time data (0.6-1.1 s) (Table 111). The longer residence time data can also be used to further assess the other models. Figure 14 shows the long residence time data and predicted conversions from all three models using corresponding fitted parameters from the 0.6- and 1.1-s residence time data (Table 111). The figure allows easy rejection of the single reaction model which badly overpredicts observed conversions at both residence times. Differentiation between the multiple reaction and three-lump models would require more data a t these residence times, ideally including ranges of tar conversions well beyond 60%. Combined Model for Tar Generation and Secondary Conversion. Predictions of the cumulative effects of primary tar generation followed by thermal conversion of the resulting tar vapors at 1.1 and 0.6 s residence times
'
'
700
'
'
BOO
R e a c t o r - 2 Temperature
('CI
Figure 13. Tar conversion data for vapor-phase secondary reaction experiments a t residence times from 0.6 to 3.9 s compared with predictions from a three-lump model, eq 7, using parameters listed in Table 111.
' 600
0 . 0 ' V '
900
' 1
'
900
('CI
Figure 14. Comparison of observed and predicted tar conversion from vapor-phase secondary reactions a t long residence times, using single-reaction, multiple-reaction, and three-lump first-order cracking models and parameters listed in Table 111. Set 2 parameters were employed in the single-reaction model.
are shown as dotted lines in parts a and b of Figure 4, respectively. Tar formation in reactor 1was modeled with a single first-order reaction (eq 11, and tar vapor cracking was described with the three-lump model (eq 7). The required equations are temperature-time history of reactor 1 dT _ - 3 "C/min
Tl,o= 25 "C
dt
(10)
rate of tar evolution from reactor 1
Vl,O = 0
where
net rate of tar evolution from reactor 2 dV2 dt =
where
[ -
Vl* - V,
] dt
d\71
[0.33 exp(-KAT)
+
-
0.27 exp(-KBT) v2,o =
0
+ 0.401 (12)
where T1, V1, V1*, V,, and V, are respectively the temperature of reactor 1, the cumulative amount of tars formed in reactor 1 (i.e., by heating reactor 1from room temperature to Tl), the ultimate amount of tar formed in reactor 1, the amount of tar which condenses between reactor 1and reactor 2, and the cumulative amount of tars collected downstream of reactor 2. KA and KB are evaluated at the temperature of reactor 2. The expression in square brackets in eq 12 provides the three-lump model prediction (eq 7) of the fraction of the tar fed to reactor 2 that survives T seconds of thermal treatment in that reactor. This analysis assumes that the thermal reactivity of tar evolving from reactor 1 is independent of tar generation temperature. Equations 10-12 were solved simultaneously by numerical methods. Generally good fits to the laboratory data were obtained (Figure 4), supporting (but not proving) the validity of the uniform tar reactivity assumption.
Conclusions 1. The evolution rate and total yield of tar generated by controlled heating (3 OC/min) of a shallow gas-swept packed bed of Pittsburgh No. 8 coal up to 550 "C were
1838 Ind. Eng. Chem. Res., Vol. 26, No. 9, 1987
independent of carrier flow rate from 750 to 1500 sccm, implying that tar evolution rates were not determined by mass-transfer resistance in the coal carrier gas boundary layer. These rates were well-correlated by either of two first-order chemical kinetic models: a single-reaction model or a multiple, independent, parallel reaction model with a narrow spread of activation energies. 2. Operating conditions and observed yields of tar and gaseous coproducts suggest that tar vapors so produced were "primary" in the sense that once they had escaped the coal particles they had undergone minimal secondary reaction within the generation reactor. 3. At temperatures and residence times of interest in coal processing and combustion, freshly formed coal devolatilization tars exhibit significant thermal reactivity even in the absence of catalysts or solid treating agents. At 700-800 "C, short-duration (0.6-1.1 s) thermal treatment of prompt tar vapor from low heating rate (3 "C/ min) pyrolysis of Pittsburgh Seam bituminous coal resulted in extensive conversion (30-50 %), with light gases as the major products. Although conversion was insignificant below 600 "C, it increased with temperature or residence time and reached 60% at 900 "C and 1.1-sresidence time. At 0.6-s residence time, conversion increased only modestly between 750 and 800 "C. 4. For secondary cracking, a lumped kinetic model treating whole tar as three noninteracting fractions of distinct chemical reactivity, each cracking by one independent, parallel, first-order reaction, provided a good correlation of conversion behavior over the temperature range 500-900 " C , for residence times from 0.6 to 3.9 s. One lump remained thermally inert a t the highest severities studied here. A multiple, independent, parallel reaction model also performed well but could not predict steplike conversion behavior observed between 750 and 800 "C at 0.6-s residence time. A single-reaction first-order decomposition model correlated the conversion data poorly at 0.6- and 1.1-s residence times, and use of the resulting best fit kinetic parameters badly overpredicted observed tar conversions at 2.5 and 3.9 s. 5. Over the temperature range 500-900 "C, tar conversion kinetics derived by measuring changes in net tar yield after controlled thermal treatment correlated well with the kinetics of changes in carbon aromaticity and hydrogen aromaticity of tars which survived thermal treatment. Opportunity to infer thermal chemical kinetic behavior from global structural information on thermally treated tars is implied. Acknowledgment Financial support of this work by the United States Department of Energy, Morgantown Energy Technology Center, under Contracts DE-AC21-82MC-19207 and DERA21-85MC-22049,is gratefully acknowledged. Technical research contributions from R. B. Lennox, K. Sawada, and several MIT undergraduate students are greatly appreciated. Nomenclature E = activation energy, kcal/mol Eo = mean of activation energy distribution, kcal/mol EA, EB = activation energy for A and B tar lumps, respectively, kcal/mol k = rate constant, s-l ko = preexponential factor, s-l KA, KB = rate constant for A and B tar lumps, respectively, s-1
KA,, KB, = preexponential factor for A and B tar lumps, respectively,
R = gas constant, kcal/(mol K) t = residence time, s T = absolute temperature, K T 2 = temperature in cracking reactor, K V = tar yield at time t (fraction of original coal weight) for tar formation models; unreacted tar at time t (fraction of original tar weight) for cracking models V* = ultimate tar yield at time t = m (fraction of original coal weight) for tar formation models; total quantity of tar entering cracking zone (fraction of original tar weight) for tar cracking models x = fraction of tar converted, defined in eq 4 Greek S y m b o l s a = standard deviation of activation energy distribution, T
kcal/mol residence time in cracking reactor, s
=
Subscripts 0 = initial 1 = reactor 1 2 = reactor 2
Literature Cited Anthony, D. B. Sc.D. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1974. Fong, W. S.; Peters, W. A,; Howard, J. B. "Kinetics of Coal Plasticity", Proceedings Second Annual Pittsburgh Coal Conference, Pittsburgh, Sept 1985. Franklin, H. D. Ph.D. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1980. Gavalas, G. R. Coal Pyrolysis; Elsevier: New York, 1982; pp 19-38. Howard, J. B. Chemistry of Coal Utilization-Second Supplementary Volume, Elliott, M. A,, Ed.; Wiley: New York, 1981; Chapter 12. Jungen, H.; Van Heek, K. H. Reaktionablaufe unter nichtisothermen Bedingungen; Springer-Verlag, Berlin, 1970; Vol. 13, pp 601-699 (Translated by Belov and Assoc., Denver, CO, APTICTR-0776). Lennox, R. B. M.S. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1983. Lennox, R. B.; Serio, M. A.; Peters, W. A,; Howard, J. B., to be submitted for publication in Fuel 1987. Nenniger, R. D.; Howard, J. B.; Sarofim, A. F. Proceedings of the International Conference on Coal Science, Pittsburgh, Aug 1983, pp 521-524. Oh, M. S.Sc.D. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1985. Peters, W.; Bertling, H. Fuel 1965,44,317. Sawada, K. M.S. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1982. Schaub, G.; Peters, W. A.; Howard, J. B. AIChE J . 1985a,31, 903. Schaub, G.; Peters, W. A.; Howard, J. B. AIChE J . 1985b,31, 912. Serio, M. A. Ph.D. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1984. Serio, M. A.; Peters, W. A.; Sawada, K.; Howard, J. B. Proceedings of the International Conference of Coal Science, Pittsburgh, Aug 1983, pp 533-536. Serio, M. A.; Peters, W. A.; Sawada, K.; Howard, J. B. Prepr.-Am. Chem. Soc. ACS Diu. Fuel Chem. 1984,29(2), 65. Serio, M. A.; Peters, W. A.; Howard, J. B., to be submitted for publication in Energy Fuels 1987. Solomon, P. R.; King, H.-H. Fuel 1984,63, 1302. Suuberg, E. M. Sc.D. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1977. Suuberg, E. M.; Peters, W. A.; Howard, J. B., Proceedings, Seuenteenth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, 1979a; p 117. Suuberg, E. M.; Peters, W. A.; Howard, J. B. In Thermal Hydrocarbon Chemistry; Advances in Chemistry Series 183; American Chemical Society: Washington, DC, 1979b; pp 239-257. Suzuki, T. M.S. Thesis, Department of Chemical Engineering, MIT, Cambridge, MA, 1984. Wen, W.-Y.; Cain, E. Ind. Eng. Chem. Process Des. Deo. 1984,23, 627-637.
Receiued for reuiecu October 4, 1985 Revised manuscript received April 24, 1987 Accepted May 1, 1987
