I n d . Eng. Chem. Res. 1989,28, 1795-1801
1795
The Role of Initial Reaction Conditions in Direct Coal Liquefaction John M. Shaw* and Ernest Peters Metals and Materials Engineering Department, The University of British Columbia, Vancouver, British Columbia. Canada V6T 1 W5
A semibatch direct coal liquefaction facility was designed and constructed in order to examine the impact of process variables on the yield and distribution of dissolution products for bituminous and subbituminous coals and lignite. A series of parametric investigations, involving solvent composition, catalyst-to-coal ratio, the intensity of turbulence, the dissolved hydrogen concentration, and the slurry residence time distribution, were performed. These investigations showed that process variables have a significant impact on the rates of the dissolution reactions and that dissolution rates for coal and lignite are affected in a similar manner. Optimal values for process variables were found to depend on the initial reaction conditions. The selection of the initial solvent composition, the initial dissolved hydrogen concentration, the catalyst-to-coal ratio, and the intensity of turbulence was shown to determine the subsequent course of dissolution reactions for coal and lignite. During recent years, there has been a growing interest in short contact time direct coal liquefaction, which reflects the perceived importance of so-called coal dissolution reactions. Previous experimental results have shown that (1)a large fraction of many bituminous and subbituminous coals is converted to at least preasphaltenes within the first minutes of reaction (Han and Wen, 1979), (2) the consumption rate of labile or molecular hydrogen is greatest during the initial moments of reaction (Whitehurst et al., 1980), (3) solvent composition and the presence of added catalyst influence the extent of coal conversion and the distribution of liquefaction products (Derbyshire et al., 1983, Frank et al., 1983),and (4) hydrodynamic effects (the intensity of turbulence and the degree of axial mixing) have been suggested as parameters of importance (Han and Wen 1979; Lee et al., 1978; Gertenbach et al., 1982). These findings suggest that purely physical phenomena, such as mass transfer and liquid-liquid solubility, may ultimately limit the initial rate of desirable hydrogenation and hydrogenolysis reactions in industrial coal liquefaction reactors. The possible existence of such limitations has not been addressed extensively in the literature. Curran et al. (1967) suggested that the rate of coal dissolution may be limited by molecular hydrogen transfer across the gas-liquid interface. This observation may reflect a poor selection of operating conditions, as gas-liquid mass transfer is rapid in severely agitated tank reactors and in bubble columns (Varclav et al., 1981). Petrakis and Grandy (1980) and Guin et al. (1978, 1979) suggest that the rate of coal hydrogenation is limited by hydrogen transfer to coal radicals, as these species have been shown to persist in coal liquefaction reaction environments (Petrakis and Grandy, 1980; Derbyshire et al., 1983). If the latter case is presumed to be more prevalent in coal liquefaction systems, one would expect the distribution and concentration of free radicals to have an impact on the apparent rate of hydrogenation/hydrolysis reactions. Radicals trapped within coal particles or a dispersed organic phase, for example, would be hydrogenated more slowly than radicals dispersed in a continuous liquid phase. There is ample experimental evidence to suggest that a dispersed organic phase other than unreacted coal particles can persist in coal liquefaction environments. Asphaltene precipitation occurs in preheaters, under certain conditions. An "antisolvent" is added to product liquids in the
* Present address: Department of Chemical Engineering and Applied Chemistry, University of Toronto, Toronto, Ontario, Canada M5S 1A4. 0888-5885/89/2628-1795$01.50/0
Table I. Proximate and Ultimate Analyses of t h e Test Coals
prox anal. (dry basis) ash fixed C volatile matter ult anal.
C H N
0 S
Forestburg coal
Bryon Creek coal
Saskatchewan lignite
16.19 46.52 37.29
16.26 62.25 21.49
19.91 43.27 36.82
66.63 4.87 1.46 26.38 0.66
83.84 4.63 1.23 9.97 0.32
69.21 4.91 1.24 23.97 0.68
Lummus modification of the SRC process to separate ash and molecular species containing the majority of the heteroatoms from the remainder of the product liquids (Pelso and Ogren, 1979). The rapid decrease in solvency of a solvent as the critical point is approached in the principal method of operation of the critical solvent deashing process (Adams et al., 1979). More recently, Shaw et al. 1988) showed that pyrene-tetralin mixtures exhibit phase splitting under coal liquefaction conditions and are able to relate the phase splitting phenomenon to synergistic fluctuations in coal conversion with solvent composition. In this paper, a number of important physical interactions consistent with the existence of a dispersed organic phase other than unreacted coal are identified by using a unique laboratory-scale reactor system. Two additional papers (Shaw and Peters, 1986,1989) examine the impact of these findings on the development of coal dissolution kinetic models and on the design of optimal reactors for both single- and two-stage direct coal liquefaction processes.
Experimental Section Materials. The liquefaction characteristics of Byron Creek coal, a partially oxidized bituminous coal, Forestburg subbituminous coal, and Saskatchewan lignite were investigated, under a variety of experimental conditions. A -90-pm fraction of commercially pulverized samples of Byron Creek and Forestburg coals was employed in the liquefaction trials. Saskatchewan lignite (1-2-cm-diameter lumps) was ground in an automatic mortar. The -90-pm fraction was retained for experimentation. These screened samples were stored in sealed containers at -10 "C and were dried in a vacuum oven at 50 "C, 2 Pa for 24 h immediately prior to use. Proximate and ultimate analyses of the dried coals are listed in Table I. 0 1989 American Chemical Society
1796
Ind. Eng. Chem. Res., Vol. 28, No. 12, 1989
Table 11. Coal Conversion Statistics for Reactor and Preheater Simulations total coal conversion and gas yield, wt %
preheater (5 min)
simulation + plug flow reactor (25 min)
plug flow preheater (648 K) + reactor (698 K) axially mixed preheater ( 5 min) + plug flow reactor (25 min) preheater (5 min) + plug flow reactor (25 min) with continuous hydrogen addition axially mixed reactor (30 min) plug flow preheater (5 min) axially mixed preheater ( 5 min) (I
Forestburg subbituminous coal solvent 1 solvent 2 (89.64, 89.94), (6.26, 6.26)
90.94, 6.37
(90.10, 89.97), (6.34, 7.03) 84.69, 5.40 89.91, 5.8
~~
~
Byron Creekn bituminous coal solvent 1 solvent 2 45.81, 3.09
49.95, 3.32
51.17, 3.53
47.83, 3.74
91.42, 8.18
two trials failed due to coke formation 64.13, 4.96 64.70, 5.20
84.35, 7.30 62.05, 4.83 59.63, 5.30
Total coal conversion and gas yield comparisons are made a t optimum catalyst loadings.
A Co-Mo catalyst, comprised of 12 wt % MOO, and 3 wt % COO,supported on a-alumina, was employed. This catalyst, supplied by Alfa Products Corp. as extruded pellets (3-mm x 3-mm long), was ground to -3 Wm prior to use and was rapidly sulfidized in situ, by sulfur contained in the solvent. Two solvents were employed. SRC oil, obtained from the SRC-I facility at Wilsonville, AL, was combined with technical-grade tetralin (30 wt % ) to form solvent 1. Solvent 2 was comprised of 100 wt % SRC oil. The SRC oil was distilled at 120 “C to remove water. A detailed description of the SRC oil is given by Southern Services Inc. (1975). Procedure. Liquefaction experiments were performed in a versatile semibatch reactor network. A detailed description of this apparatus and operating procedures has been given previously (Shaw, 1985). The major design features are as follows: (1)the coal slurry can be rapidly injected into a preheated reactor; (2) the coal slurry residence time distribution can be varied; (3) the reactor can be operated at constant pressure or with a fixed initial quantity of gas; (4) reaction time and the reactor temperature can be controlled precisely, as the reactor is equipped with a heating mantle and internal cooling coils; and (5) the coal slurry can be presaturated with N2 or H2 at room temperature, at pressures less than 5 MPa, prior to injection. At the start of an experiment, the reactor, a 2-L Pressure Products Industries autoclave with a magnetically driven stirrer, was charged with 150 g of solvent (seed oil) and catalyst. The reactor was then placed under vacuum, to remove air, sealed, and pressurized to 5 MPa with H2 The reactor was preheated to 25 K above the desired operating temperature for the experiment, and the interior was subcooled to the reaction temperature, using the internal cooling coils. Seven hundred grams of coal slurry, at ambient temperature, with a 2.5:l solvent-to-coal ratio, was then injected. The cooling water flow rate was reduced, during injection, to maintain the operating temperature. The slurry residence time distribution was varied by adjusting the slurry injection rate. The slurry injection time was varied from 250 to 3600 s. Both axially mixed and plug flow reactors were simulated by controlling the injection rate. Experiments with a slurry mean residence time as short as 150 s were performed. The reaction products were divided into component gas yields (CO, CO,, CH,, C2H6) and liquid yield. The total coal conversion, defined by the ash content of the tetrahydrofuran-insoluble residue, and the gas yield were measured directly. The liquid yield was obtained by difference. Mass balances were performed routinely to within 98-99%. More precise balances must await im-
proved techniques for recovering C02and C2H6 from coal oil slurries. The mass balance closures on these gas components are only 70-SO%, due to their high solubilities in coal liquids (Timmermans, 1959; IUPAC, 1982). As much as 50 wt % of each of these gas components is dissolved in the coal slurry at room temperature, and the dissolved fraction of these gases is only partially recoverable. The soluble fraction of sparingly soluble components (CO, H,, etc.) can be estimated accurately by using an argon tracer technique (Shaw, 1985). Experimental results obtained with this apparatus were repeatable to within fine tolerances. The total coal conversion and the gas yield are repeatable to within 0.3-0.5 wt %, a precision at least an order of magnitude better than frequently reported in the literature for data obtained from similar types of experiments. Conversion statistics for two sets of duplicate trials are listed on Table 11. The results from other duplicate sets, e.g., runs 221 and 223 and 224 and 226, are shown in Tables 111-V.
Results and Discussion The impact of process variables on the dissolution characteristics of coal and lignite was evaluated under a broad range of conditions. The examination of effects with possible implications for the design of industrial direct coal liquefaction reactors was emphasized. Consequently,dense coal slurries and complex liquefaction solvents were employed. Solvent 1,comprised of 70 wt % SRC oil and 30 wt % tetralin, was used to simulate a hydrogenated distillate coal oil, the preferred solvent type for two-stage coal liquefaction processes. Solvent 2, comprised of 100 w t % SRC oil, resembles the solvents encountered in single-stage processes. Illustrative results are discussed below, and extensive data tables are included for reference. Reactor and Preheater Simulations. Role of Preheaters. A series of experiments was performed in order to simulate axial mixing patterns, for slurries, in a continuous flow apparatus-Table 11. Two extreme axial mixing patterns were considered: “plug flow”, where all of the slurry had a residence time within 2 min of the mean, and “axially mixed”, where the slurry was injected as a square wave and had a residence time within the interval 0-2t. Preheaters were simulated by isothermal experiments with 5-min mean residence times. The assumption that the rapid heating of the coal slurry on injection, followed by a short isothermal reaction time, simulates the complex heat transfer and mixing patterns in preheaters is a t best crude. Trials with 30-min mean residence times were used to simulate various reactor + preheater combinations. The need to simulate axial mixing patterns in preheaters, as well as reactors, is evident from the liquefaction results shown in Figure 1. Greater than
Ind. Eng. Chem. Res., Vol. 28, No. 12, 1989 1797 Table 111. Trial Summary for Foreetburg Coal
run 2010 205 208 209 210 213 214 215 216 217 218 219 220 221 222 223 224 225 226 229 230 231 232 233 234 235 236 237 238 239 24 1 242 243 245 246
T,O C
rpm lo00 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 lo00 1000 1000 500 500 1000 1000 500 1000 1000 1000 1000 1000 1000 1000 2000
425 425 425 425 425 425 375 375 375/425b 425 425c 425 425 440 400 400 425 425 425 425 425 425 425 425 425 425 425 425 425 425 425 425 425 425 425
operating conditions catalyst, g solvent MRT, min 1 20.0 30.0 1 20.0 30.0 2 20.0 30.0 2 20.0 30.0 2 20.0 30.0 1 20.0 30.0 2 20.0 30.0 1 20.0 30.0 1 20.0 30.0 2 20.0 30.0 2 20.0 30.0 1 20.0 5.0 2 20.0 5.0 2 20.0 30.0 1 20.0 30.0 2 20.0 30.0 2 40.0 30.0 2 80.0 30.0 2 40.0 30.0 1 80.0 5.0 2 0.0 30.0 0.0 2 5.0 2 20.0 30.0 1 20.0 30.0 2 20.0 5.0 1 20.0 5.0 2 20.0 5.0 1 20.0 15.0 2 20.0 30.0 1 80.0 30.0 2 2.44 20.0 1 45.0 20.0 2 30.0 20.0 2 30.0 20.0 5.0 2 20.0
simd PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF 2 2 PF PF 2 PF 2 PF PF PF PF
gas production (consumption), g/kg MAF coal (Hz) CH4 CO CzH6 C02 20.2 10.5 6.6 5.1 35.2 31.6 12.0 8.3 5.1 40.9 33.4 12.0 5.0 5.9 40.5 5.2 46.5 35.4 14.1 4.7 34.3 10.4 5.0 5.7 37.0 11.1 7.5 30.8 5.7 38.4 0.8 30.1 17.7 1.7 2.9 15.9 1.8 2.8 0.5 29.5 28.4 8.0 7.6 3.9 34.6 7.1 48.4 38.2 13.5 5.40 39.1 16.4 8.4 13.6 43.5 24.1 5.0 5.1 3.0 36.5 3.5 33.4 30.0 6.4 5.1 30.3 5.7 3.8 3.1 36.6 27.1 5.1 6.7 2.7 35.3 2.8 39.4 28.3 5.3 5.3 39.1 13.7 4.1 7.1 45.0 7.7 45.1 39.9 14.2 1.7 37.7 13.8 2.9 6.4 44.9 33.4 8.0 4.3 5.2 30.8 7.9 52.3 17.7 14.1 7.8 6.90 5.5 5.7 2.9 44.3 35.1 13.6 5.3 7.6 48.4 35.3 14.1 5.5 8.1 46.4 26.4 5.4 6.0 2.7 38.9 24.5 5.4 6.3 3.0 37.3 27.3 6.3 3.9 3.5 39.3 29.4 8.9 7.1 4.5 39.0 38.1 14.9 4.0 7.7 46.4 11.1 3.9 36.7 5.5 37.2 17.3 4.6 6.0 2.2 33.1 36.6 14.2 5.7 6.2 40.7 37.0 14.2 5.3 6.2 44.1 36.0 12.1 5.2 9.4 70.0
MAF coal input, g 168.31 155.8 163.6 157.5 159.8 152.7 156.5 138.6 133.9 163.0 148.0 121.3 167.5 173.7 128.0 161.4 157.8 154.4 161.1 125.7 147.4 152.3 127.4 160.6 157.8 152.7 160.8 159.5 160.9 161.8 158.5 157.6 160.68
% conversion
gas 5.74 6.62 6.30 7.03 5.81 6.26 3.55 3.46 5.41 7.44 8.19 4.96 4.83 4.91 4.97 5.28 6.98 6.87 6.80 4.83 8.21 5.83 7.48 7.40 5.30 5.19 5.29 5.94 7.30 5.77 4.59 6.68 6.99 9.66
liq 70.9 83.0 83.8 83.0 84.1 83.6 36.1 40.8 79.3 82.1 83.2 59.2 57.3 65.5 68.4 64.4 79.2 77.4 79.6 44.0 61.5 30.4 78.1 75.5 54.3 59.5 52.0 72.0 77.1 74.0 45.4 81.2 80.0 71.0
total 76.7 89.6 90.1 90.0 89.9 89.9 39.6 44.2 84.7 89.5 91.4 64.1 62.1 70.4 73.4 69.7 86.2 84.3 86.4 48.8 69.7 36.2 85.6 82.9 59.6 64.7 57.3 77.9 84.4 79.8 50.0 87.9 87.0 80.7 55.0
a Whole catalyst pellets were employed. bPreheater operated at 375 OC, reactor at 425 O C . CReactormaintained at constant pressure. d P F = plug flow, 2 = axially mixed.
loo
r
8
6
.
40
B
60 x
=+ U
0
20
40
60
0
REACmTIME MINUTES
20
40
60
REACTIONTIME minutes
Figure 2. Molecular hydrogen consumption, a t 698 K, under a variety of reaction conditions. Refer to Figure 1 for reaction conditions.
01
0
20
40
60
REACTIONTIME minutes
Figure 1. Data summary for liquefaction trials performed a t 698 K, with a stirring frequency of 16.7 Hz. Reaction conditions: (A,0,0, m) solvent 1 with lo-, 20-, 20-, and 80-g catalyst charges; (0, 0 , X) solvent 2 with 0-, 20-, and 20-g catalyst charges.
50% of the Forestburg coal and 15% of the Byron Creek coal is liquefied within 2.5 min at 698 K, a time period normally spent in a preheater. Greater than 50% of the molecular hydrogen consumption occurs simultaneously-Figure 2. Importance of Axial Mixing. The results obtained from reactor and preheater simulations, Table 11, illustrate the importance of axial mixing in direct coal liquefaction (DCL) reactor networks. These findings also illustrate inadequacies in current kinetic and process models for direct coal liquefaction (Mohan and Silla, 1981; Singh et al., 1982; Lee et al., 1978). Process and kinetic models are based on the assumption that coal dissolution reactions are controlled kinetically and cannot account for the di-
1798 Ind. Eng. Chem. Res., Vol. 28, No. 12, 1989 Table IV. Trial Summary for Byron Creek Coal
run 312 313 314 315 316 317” 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
T , “C 425 425 425 425 425 425 425 425 425 425 425 400 425 378 425 425 425 425 425 425 425 400 425 425 425
rpm 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 2000 500 1000 1000 2000 500 1000 1000 1000 1000 1000
operating conditions catalyst, g solvent MRT, min 20.0 1 2.44 0.0 2 30.0 20.0 2 30.0 40.0 2 30.0 80.0 2 30.0 20.0 2 2.44 20.0 2 2.44 10.0 2 30.0 0.0 1 30.0 10.0 1 30.0 20.0 1 30.0 1090 1 30.0 20.0 2 60.0 20.0 2 30.0 20.0 2 30.0 20.0 2 30.0 20.0 2 15 5.0 1 30.0 10.0 1 30.0 10.0 1 30.0 10.0 1 15 20.0 2 30.0 20.0 2 38 10.0 1 38 10.0 1 60
sim 2 PF PF PF PF 2 2 PF PF PF PF PF PF PF PF PF PF PF PF PF PF PF 2 2 PF
gas production (consumption). g/kg . -, MAF coal ( H J CH, CO C,H, CO, 12.7 3.3 1.31 2.49 8.88 7.7 12.2 2.61 5.85 29.5 24.4 15.4 2.56 7.49 7.79 25.2 13.5 2.78 7.41 6.06 25.5 15.2 1.28 12.8 7.10 14.5 19.6 4.4 15.0 17.8 12.7 29.3 11.8 25.8 23.2 20.3 14.2 11.7 14.6 12.0 16.8 25.2 15.2 15.1
3.9 13.0 9.7 10.6 10.0 2.97 21.1 1.4
13.5 13.0 9.6 10.4 8.7 10.0 5.0 4.6 16.2 13.5 10.3
1.58 3.66 2.75 3.91 2.58 1.46 2.42 0.96 3.28 2.73 2.29 3.08 2.60 2.85 2.44 1.74 3.43 4.02 3.33
3.16 7.26 4.10 5.67 5.29 1.29 11.7 0.56 10.5 7.80 9.53 5.39 4.91 6.92 1.87 2.21 8.41 7.21 4.75
8.67 10.5 11.2
10.7 8.65 7.33 12.8 6.48 12.5 8.82 13.6 9.19 11.1
12.8 6.95 6.91 9.39 10.6 8.23
MAF coal inwt, E 166.7 170.9 171.1 156.4 166.4 169.6 168.6 171.4 165.4 169.2 161.3 167.2 174.8 163.2 160.6 165.7 170.5 163.1 166.9 162.4 169.6 165.9 162.9 162.2 167.2
% conversion
gas
lia
total
1.60 5.02 3.32 2.97 3.64
15.1 30.7 46.6 44.2 34.9
1.73 3.44 2.78 3.09 2.65 1.31 4.80 0.94 3.94 3.24 3.51 2.80 2.74 3.25 1.63 1.55 3.74 3.53 2.66
5.59 39.0 26.3 42.7 29.4 18.6 48.5 8.60 41.4 43.2 27.9 36.6 37.4 34.0 32.8 24.1 44.1 48.0 41.0
16.7 35.8 50.0 47.2 38.6 13.9 7.3 42.5 29.1 45.8 32.0 19.9 53.3 9.50 45.3 46.4 31.4 39.4 40.1 37.3 34.4 25.7 47.8 51.6 43.7
Slurry presaturated with hydrogen.
Table V. Trial Summary for Saskatchewan Lignite
run 401 402 403 404
T , “C 425 425 425 425
rpm 1000 1000 1000 1000
operating conditions catalyst, g solvent MRT, min 0.0 2 15 20.0 2 15 2 15 40.0 80.0 2 15
sim PF PF PF PF
gas production (consumption), g/kg MAF coal MAF coal (H,) CHI CO CzHB COP input, g 10.3 8.88 6.04 4.72 58.8 153.4 29.4 8.92 4.730 5.38 55.2 154.4 33.2 8.93 4.39 6.23 64.8 146.7 34.3 9.59 3.74 6.37 51.0 150.7
vergent behavior of the two similar solvents when liquefaction results from various simulations are compared. Byron Creek coal conversion in solvent 1, for example, increases 5% when an axially mixed system is replaced with a plug flow one, whereas a reverse trend in conversion is noted when solvent 2 is employed. In addition, these liquefaction models do not predict the excessive “coke” formation which led to the failure of axially mixed reactor simulations with Forestburg coal and solvent 1, where effectively no conversion was realized. (When liquefied in solvent 2 under otherwise similar reaction conditions, the T H F conversion is in excess of 84%.) Only if the underlying physical phenomena, which act as hidden variables in these trials, are considered, can the role of axial mixing be resolved. Apparent Reaction Rates. Coal conversion to products soluble in an arbitrary solvent is, at best, a primitive measure of the extent of coal reaction. The Forestburg coal, which “coked” and plugged the slurry inlet port during axially mixed reactor simulations, with solvent 1, clearly reacted, even though it did not report as T H F soluble or “converted” material. Similar though less dramatic differences between reacted and converted material are observed in other reactor and preheater simulations. Coal conversion statistics only include that fraction of the coal that undergoes hydrogenation or hydrogenolysis reactions. Coal constituents that undergo polymerization, “coking”, or decompose into species that are not soluble in the carrier solvent are lumped together with the truly unreacted material. Thus, the conversion differences to be explained are differences arising from the distribution
% conversion gas liq total 7985 42.9 50.8 7.42 60.9 68.4 8.43 57.2 65.6 7.07 58.1 65.2
of the reactions that the coal undergoes. The apparent synergistism between coal dissolution rate and the extent of axial mixing is attributed to shifts in bulk solvent composition arising from catalytic hydrogenation/hydrogenolysis. The solubility of hydrogen- and coal-derived species in a solvent is sensitive to solvent composition. These latter two variables can have a direct bearing on the observed rates of coal dissolution reactions. At 700 K, for example, hydrogen is 8 times more soluble in tetralin than in pyrene (Shaw, 1987). It is also well-known that tetralin is a poor physical solvent for coal liquids. Pyrene, for example, has only limited solubility in tetralin at 700 K (Shaw et al., 1988). The Role of Molecular Hydrogen. The initial molecular hydrogen consumption rate and the similarity of this rate, for coal dissolution, in both solvents, Figure 2, suggest that the transfer of molecular hydrogen to or into a dispersed phase, comprised of coal particles and/or initial decomposition products, may limit the initial rate of coal dissolution reactions, even in the presence of a labile hydrogen source (solvent 1 comprised of 30 wt 7% tetralin). This effect was demonstrated with Byron Creek coal liquefied in solvent 2. Two experiments with 2.5-min slurry mean residence times were performed at 698 K. The slurry was not presaturated with hydrogen prior to injection in the first experiment and the initial hydrogen concentration in the reacting slurry was much less than the saturated concentration of the seed oil. A total coal conversion of 7.32 wt % was realized. For the second experiment, the slurry was saturated with hydrogen at 290 K, 4 MPa, prior to injection and the total coal conversion increased to 13.93
Ind. Eng. Chem. Res., Vol. 28, No. 12, 1989 1799
0
20
40
60
REACTION TIME minutes
Figure 3. Apparent solubilities of hydrogen in product liquids, measured at 290 K, following reaction a t 698 K, and 16.7 Hz. Reaction conditions: (A,m) solvent 1with 10- and 20-g catalyst charges; ( X ) solvent 2 with a 10-g catalyst charge.
wt 90. Clearly, the initial rate of coal dissolution reactions,
at 698 K, is controlled by molecular hydrogen transfer to or into a dispersed condensed phase. This mass-transfer limitation on the observed rate of coal dissolution does not persist. Hydrogen consumption rates drop rapidly, and excess hydrogen leads to enhanced gas yields a t longer mean residence times. This effect is frequently noted in the literature and is confirmed in Table 11. The solvent dispersed-phase mass-transfer limitation on the initial rate of coal liquefaction reactions is also supported by the apparent solubility data for hydrogen in the two solvents, as the reactions progress. These data, presented on Figure 3, were obtained once the reactor was cooled to room temperature. Hydrogen solubilities at the reaction temperature, 698 K, are greater, and the solubility differences would be amplified (Lin et al., 1981; IUPAC, 1981; Shaw, 1987). The hydrogen solubility in solvent 1, measured at 290 K, drops by a factor of 2 during the first 5 min of reaction a t 698 K, but by increasing the extent of slurry-phase axial mixing, the peak hydrogen demand is reduced and the total conversion in the preheater increases slightly. The hydrogen solubility in solvent 2 drops below the detection limit within the same time interval, as liquefaction reactions progress. So, even though the peak hydrogen demand is reduced, in an axially mixed preheater, the last coal entering the preheater encounters a low dissolved hydrogen concentration, and a net decrease in coal conversion results. The difference in T H F conversion between the two solvents exceeds 5 wt % in axially mixed preheaters. Apparent hydrogen solubility differences can also account, in part, for the observed differences in coal conversion obtained from axially mixed and plug flow reactors. Byron Creek coal, liquefied in an axially mixed reactor with solvent 1, encounters a hydrogen concentration that is slightly greater than the initial hydrogen concentration encountered during a plug flow reactor/preheater simulation. Consequently, the initial coal conversion increases. The hydrogen concentrations, encountered by Forestburg and Byron Creek coal during an axially mixed reactor simulation, are much lower than the initial hydrogen concentrations encountered during plug flow reactor simulations, and the initial coal conversion is reduced. Hydrodynamic effects are superimposed on the differences in the initial hydrogen concentration, as conversion in plug flow reactors is expected to be greater than conversion in axially mixed reactors, if simple nth order reaction kinetics are applicable (Levenspiel, 1972). Solvent Composition. Large changes in hydrogen solubility reflect significant changes in the solvent com-
position (Shaw, 1987). The liquid, encountered by the Forestburg coal during axially mixed reactor simulation with solvent 1,had an extremely high hydrogen solubility, suggesting that it had both a reduced mean molar mass and heteroatom content. The initial dissolution products comprise complex molecules with an appreciable heteroatom content. Such species, particularly those containing a high heteroatom content, are unlikely to be miscible with such a solvent, and a separate liquid or solid phase was probably formed. The segregation of these species, in separate phases, can easily lead to retrogressive reactions and “coke“ formation. It is worth noting that even pyrene-tetralin mixtures undergo phase splitting at 700 K (Shaw et al., 1988). Some coal-derived products also appear to condense near the end of plug flow reactor simulations with solvent 1,if slurry-phase mean residence time is sufficiently long-Figure 1. These two effects are related, as cobalt-molybdenum catalysts catalyze hydrogenation and hydrogenolysis reactions for midrange compounds selectively (Weigold, 1982). Large molecules, which contain most of the heteroatoms present in a solvent (Longanbach, 1984), are either introduced into or remain in a pregressively less amenable solvent, and a greater fraction of these species becomes insoluble with time. The Role of Cobalt Molybdate Catalysts i n DCL Reaction Environments. Catalysts are ubiquitous actors in direct coal liquefaction reaction environments. It is difficult to perform “catalyst free” or even “added catalyst free” experiments, when some mineral matter constituents and catecols, for example, act as catalysts (Yoshii et al., 1982; Tarrer et al., 1977). In addition, one must contend, experimentally, with the “memory effect” when employing Co-Mo catalysts. If a batch experiment is performed with this type of catalyst, and the catalyst is then removed, the catalytic effect persists for three or four additional trials even if no more catalyst is added (Tarrer et al., 1977). Figure 4 illustrates the sensitivity of coal and lignite conversion to the presence of added catalyst. The trials presented in this figure were of 30-min duration for Forestburg and Byron Creek coal and 15-min duration for Saskatchewan lignite. In the absence of added catalyst, the total coal conversion is radically reduced. The difference can be as great as 20 wt 7%. If excess catalyst is present, a comparable reduction in total coal conversion results. Gas yields are also affected by added catalyst, but only when the coal or lignite is liquefied in a solvent with a low labile hydrogen content. In this case, the gas yield drops 1-2 wt 7% in the presence of added catalyst. Excess catalyst appears to have no further effect on gas yield a t 698 K. The sensitivity of total conversion, to the presence of added catalyst and particularly to the presence of excess catalyst, appears to be greatest for bituminous coal. Subbituminous coal is less sensitive, and lignite is least sensitive. Solvent composition also contributes to the sensitivity of coal conversion to the level of catalysis. Coal conversion is more adversely affected by nonoptimal levels of catalysis when liquefied in solvent 1 than in solvent 2. This effect can be attributed to differences in the mean molar mass of the two solvents, as bituminous coals tend to generate larger molecular fragments than subbituminous coals and lignites. Larger molecules are less likely to be soluble in solvents undergoing catalyzed hydrogenation1 hydrogenolysis reactions, as these reactions lead to a reduction in the mean molar mass and heteroatom content of the solvent. The reduced solubility of coal-derived molecular species in the solvent results in lower coal conversions, e.g., greater coke formation. This effect is ac-
1800 Ind. Eng. Chem. Res., Vol. 28, No. 12, 1989 100 r
;
i
E
o
20
::
100
200
300
400
500
600 0
CATALYST s/kg
_z
..
C
100
z
mafcoal
3
200
SASK. LIGNITE
300
400
500
20
30
40
50
SnRRlffi FREGUENCY Hz Figure 5. Impact of turbulence on coal conversion at 698 K. A six-bladed, high shear impeller with a 5-cm diameter was employed. Refer to Figure 3 for reaction conditions.
x
CATALYST 9%
10
600
*af COS,
Figure 4. Impact of catalysis on coal and lignite conversion, at 698 K. The trials are of 30-min duration, except as noted in the text. Reaction conditions: (0, 0)solvent 1; ( X , 0 ) solvent 2.
centuated in solvents with a lower, initial mean molar mass. The impact of catalysis on coal conversion is also time dependent-Figure 1. Conversion differences arising from insufficient or excess added catalyst diminish at longer reaction times. Forestberg coal conversion, in solvent 2, for example, is reduced by 25.6 w t 70at 698 K, after 5 min, if no catalyst is charged with the coal instead of a typical charge of 20.0 g. This difference drops to 21.3 w t 90 after 30 min. A similar time dependence is noted for Forestberg coal conversion in solvent 1 with 20.0 vs 80.0 g charges of catalyst. After 5 min, there is a 15.3 w t 70 difference in coal conversion, which reduces to 10.5 w t % after 30 min. Not only is there an optimum level of catalysis for each coalsolvent system but one defined by initial coalsolvent interactions. Intensity of Turbulence. The intensity of turbulence, which for autoclaves is defined by the stirring rate and impeller geometry, is not considered to be a variable of consequence for direct coal liquefaction reactions. The literature asserts that it is only necessary to assure that reactors are "well mixed" so that adequate gasslurry mass transfer occurs. Stirring rates of 8.33 Hz are frequently employed for this purpose. The impact of stirring on the destruction of coal particles has only been treated in a qualitative manner. Whitehurst et al. (1980) reviewed some of their own work and the work of others and showed that coal particles remained intact with little or no apparent shrinkage, up to 80 wt Ti conversion, in the absence of agitation, but break up rapidly when reacted in an agitated autoclave. They also showed that the stirring rate had no influence on coal conversion to pyridine-soluble material, after 2 min of reaction, at 698 K. The results presented on Figure 5 differ with these previous findings. The coal conversion to tetrahydrofuran-soluble material and the gas yield are both affected
by changes in the stirring rate. For any given level of catalysis, coal conversion increases to a maximum and then declines as the stirring rate is increased. The total coal conversion can vary by as much as 8 wt %. There is no general trend for the dependence of the gas yield on stirring rate. The apparent contradiction between these results and previous findings is readily explained. Macerals present in raw coal are soluble in pyridine but not in tetrahydrofuran (Shalabi et al., 1979; Guin et al., 1979). Therefore, one would not expect pyridine solubility to reflect the degree of conversion of coal-derived species. Coal-derived molecular species must undergo molar mass and/or heteroatom content reduction before they are soluble in tetrahydrofuran. Thus, THF solubility is sensitive to the degree of coal conversion. The results, presented on Figure 5, reflect this sensitivity. These results also highlight the interdependence of stirring rate (the intensity of turbulence) and the optimum level of catalysis. An optimum level of catalysis is only optimal at a single stirring rate. Further, the impact of nonoptimum stirring rate catalyst combinations, on coal conversion, persists from the initial sequence of dissolution reactions. Figure 5 illustrates this effect for Forestberg coal when liquefied in solvent 2, at 698 K, with a 20-g charge of catalyst. After 5 min, the difference in coal conversion between 8.33 and 16.67 Hz is 4.8 wt %; after 30 min, the difference is approximately the same (4.5 wt %). One can only hypothesize that the intensity of the turbulence influences the extent of coal particle break-up during the initial decomposition reactions. This would alter the coal-solvent interfacial area and affect the initial rates of dissolution reactions. At low stirring rates, little particle destruction occurs. Excess catalyst may be present, and the extent of coal conversion is reduced. At high stirring rates, particles are fragmented. The rates of initial reactions are much higher, and if the amount of catalyst is insufficient, inadequate catalysis also leads to a reduction in coal conversion. At an intermediate stirring rate, the initial rate of coal dissolution and the amount of catalyst are well matched and a coal conversion optimum is observed. This hypothesis was tested by examining catalyst- and ash-free residue particle size distributions, at near constant conversion, as a function of stirring rate. Particle size distributions for residue particles extracted from 5-min liquefaction trials with Forestburg coal at 8.33,16.67, and 33.3 Hz were obtained by using a Leitz image analyzer. The total coal conversions for these trials were 57.3, 62.1, and 55.0 wt 90,respectively. The results indicate that the mean diameter of organic residue particles decrease from
Ind. Eng. Chem. Res., Vol. 28, No. 12, 1989 1801 52 to 39 to 34 pm as the stirring rate was increased from 8.33 to 16.67 to 33.3 Hz.The initial mean particle diameter was 52 pm. These results support the proposed relationship between stirring rate and the level of catalysis.
Conclusions The experimental findings depict a novel and detailed description of the direct coal liquefaction reaction environment and highlight the importance of process variables and associated physical phenomena in the overall kinetic scheme for coal liquefaction. Optimal values for process variables were found to depend on the initial reaction conditions. The selection of the initial solvent composition, the initial dissolved hydrogen concentration, the catalyst-to-coal ratio, and the intensity of turbulence was shown to determine the subsequent course of liquefaction reactions for coal and lignite. These findings provide a coherent basis for the development of kinetic models for coal liquefaction reactions and for the optimization of single- and two-stage coal liquefaction reactor designs. None of the physical phenomena described above are included in existing DCL reaction or process models, and consequently, these models are poor predictive tools. The results also show that optimization of coal-solvent-catalyst interactions, by manipulating the intensity of turbulence, the level of catalysis, and solvent composition, may have a greater impact on the space time yields of desirable liquefaction products and the spectrum of products produced than the variation of the extent of axial mixing per se. This latter parameter is frequently discussed without consideration of the many variables affected simultaneously. Acknowledgment The authors thank P. Kemp and G. Roemer for their assistance with instrumental analysis. The financial support provided by British Columbia Research Ltd., in the form of a B.C. Research Fellowship for J. M. Shaw, and NSERC, through strategic Grant 0164, is gratefully acknowledged. &&try NO.CO,7440-48-4; Mo, 7439-987; CO, 630-080; COZ, 124-38-9; CH4, 74-82-8; C2H6, 74-84-0.
Literature Cited Adams, R. M.; Knebel, A. H.; Rhodes, D. E. Critical Solvent Deashing. Chem. Eng. Prog. 1979, 75(6), 44-48. Curran, G. P.; Struck, R. T.; Gorin, G. Mechanism of the Hydrogen-Transfer Process to Coal and Coal Extract. Ind. Eng. Chem. Process Des. Dev. 1967, 6(2), 166-173. Derbyshire, F.; Varghese, P.; Whitehunt, D. Two-Stage Liquefaction of Subbituminous Coal. Fuel 1983,62, 78-80. Gertenbach, D. D.; Baldwin, R. M.; Bain, R. L. Modelling of Bench-Scale Coal Liquefaction Systems. Ind. Eng. Chem. Process Des. Deu. 1982, 21, 490-500. Guin, J. A.; Tarrer, A. R.; Prather, J. W.; Johnson, D. R.; Lee, J. M. Effects of Coal Minerals on the Hydrogenation, Desulfurization and Solvent Extraction of Coal. Ind. Eng. Chem. Process Des. Deu. 1978, 27(2), 118-126. Guin, J. A.; Tarrer, A. R.; Lee, J. M.; Van Backle, H. F.; Curtis, C. W. Further Studies of Catalytic Activity of Coal Minerals in Coal
Liquefaction. Ind. Eng. Chem. Process Des. Deu. 1979, 18, 631-637. Han, K. W.; Wen, C. Y. Initial Stage (Short Residence Time) Coal Dissolution. Fuel 1979,58, 779-782. IUPAC. Hydrogen and Deuterium. Solubility Data Series: Pergamon Press: Oxford, England, 1981; Vol. 5/6. IUPAC. Ethane. Solubilitv Data Series; Pergamon Press: Oxford, England 1982, Vol. 9. Lee. M. H.: Guin. J. A.: Tarrer. A. R. A DisDersion Model for the SRC Process. ‘Ind. Eng. Chem. Process Des. Dev. 1978, 17(2), 127-134. Levenspiel, 0. Chemical Reaction Engineering, 2nd, ed.; Wiley: New York, 1972. Lin, H.; Sebastian, H. M.; Simmik, J. J.; Chao, K. Solubilities of Hydrogen and Methane in Coal Liquids. Ind. Eng. Chem. Process Des. Deu. 1981, 20, 253-256. Longanbach, J. R. The Effects of Heavy Recycle Solvents on Coal Liquefaction. Chem. Eng. Prog. 1984, 80(11), 29-32. Mohan, G.; Silla, H. Kinetics of Donor Solvent Liquefaction of Bituminous Coals in Nonisothermal Experiments. lnd. Eng. Chem. Process Des. Deu. 1981, 20, 349-358. Pelso, M.; Ogren, D. F. Anti-Solvent Deashing. Chem. Eng. Prog. 1979, 75(6), 41-44. Petrakis, L.; Grandy, D. W. Free Radicals in Coals and Coal Conversion 2. Effect of Liquefaction Processing Conditions on the Formation and Quenching of Coal Free Radicals. Fuel 1980,59, 227-232. Shalabi, M. A.; Baldwin, R. M.; Bain, R. L.; Gary, J. H.; Golden, J. D. Non Catalytic Coal Liquefaction in a Donor Solvent. Rate of Formation of Oil, Asphaltenes, and Preasphaltenes. Ind. Eng. Chem. Process Des. Dev. 1979,18(3), 474-479. Singh, C. P. P.; Shah, Y. T.; Carr, N. L.; Prudich, M. E. Liquefaction of Coal by SRC-I1 Process (I). Can. J. Chem. Eng. 1982, 60, 248-260. Shaw, J. M. Novel Design Criteria for Direct Coal Liquefaction Reactors. Ph.D. Thesis, University of British Columbia, 1985. Shaw, J. M. A Correlation for Hydrogen Solubility in Alicyclic and Aromatic Solvents. Can. J . Chem. Eng. 1987, 65, 293-298. Shaw, J. M.; Peters, E. Optimum Reactor Designs for Single and Two Stage Direct Coal Liquefaction Processes. Proc. 36th Can. Chem. Eng. Conf. Sarnia, Oct 1986. Shaw, J. M.; Peters, E. A General Model for Direct Coal Liquefaction Reactions. Ind. Eng. Chem. Res. 1989,28,976-982. Shaw, J. M.; Gaikwad, R. P.; Stowe, D. A. Phase Splitting of Pyrene Tetralin Mixtures. Fuel 1988, 67, 1554-1559. Southern Services Inc. S.R.C. Tech. Report 7, Analysis of Operations, Solvent Refined Coal Pilot Plant, Wilsonville, AL, 1975. Tarrer, A.; Guin, J.; Pitts, W.; Henley, J.; Prther, J.; Styles, G . Liquid Fuels from Coal. Effect of Coal Minerals on Reaction Rates During Coal Liquefaction; Ellington, R. T., Ed.; Academic Press: New York, 1977; pp 45-61. Timmermans, D. Physico-Chemical Constants of Binary Systems; Interscience: New York, 1959; Vol. 1. Varclav, P.; Kubicek, I.; Kucler, J. Chem. Prum. 1981,31(2), 61-66; Chem. Abstr. 1981,94, 177157-F. Weigold, H. Behaviour of Co-Mo-A1,03 Catalysts in the Hydrodeoxygenation of Phenols. Fuel 1982,61, 1021-1026. Whitehurst, D. D.; Mitchell, T. 0.; Farcasiu, M. Coal Liquefaction: The Chemistry and Technology of Thermal Processes; Academic Press: New York, 1980. Yoshii, T.; Yaginuma, R.; Yoshikaura, H.; Utoh, Sooich. Unusual Effects of Catechol upon the Hydro-Liquefaction of Coal. Fuel 1982, 61, 865-866.
Received for review May 9, 1988 Revised manuscript received January 13, 1989 Accepted September 7, 1989
