KINETICS OF HYDROCRACKING OF COAL EXTRACT WITH
MOLTEN ZINC CHLORIDE CATALYSTS IN BATCH AND CONTINUOUS SYSTEMS ROBERT W I L L I A M
T . A .
S T R U C K ,
E .
W I L L I A M
R O S E N H O O V E R ,
C L A R K ,
CLYDE
Research Diuision, Consolidation Coal Go., Library, P a .
W .
PHILIP ZIELKE,
J .
D U D T ,
A N D
EVERETT
G O R I N
15129
Rate data are presented on the hydrocracking of coal extract i n both batch autoclaves and a small stirred continuous reactor using molten zinc chloride catalyst over the following range of conditions: HZ pressure 1500 to 3500 p.s.i.a., ZnO/ZnClz mole ratio 0.01 to 0.08, and temperature 700" to 800'F. A semiempirical rate model was devised which fits both sets of data and assumes that extract is composed of a nonconvertible fraction, a fraction which hydrocracks by a rapid first-order reaction and a fraction which hydrocracks by a relatively slow first-order reaction. The batch autoclave data were analyzed to determine the nonconvertible fraction of the extract, the fraction of the convertible extract which hydrocracks by each reaction, and the rate constant of the slow reaction. The nonconvertible fraction increases w i t h increasing temperature (700" to 800' F.) and ZnO/ZnCI* mole ratio, and decreases w i t h increasing pressure. I t varies between 7 and 14% of the extract over the conditions studied. An average of 75% of the convertible extract hydrocracks by the fast reaction. The rate constant for the fast reaction was derived from the continuous unit data and is up to 40 times greater than that of the slow reaction. The rate constant of the slow reaction is relatively independent of Hz pressure but decreases w i t h increasing ZnO/ZnClz.
THEuse of molten zinc chloride as a catalyst for hydrocracking polynuclear hydrocarbons has been described (Zielke et al., 1966). The superiority of zinc chloride over conventional hydrocracking catalysts was shown for pyrene, coal, and coal extract. In batch autoclaves zinc chloride gave more rapid reaction, more complete conversion to gasoline-range naphtha, and a very high octane number without re-forming. The high octane number is due primarily to the ability of the Lewis acid, zinc chloride, to maintain acid catalysis in the presence of considerable amounts of nitrogen in the feed, thus giving a high percentage of branched-chain hydrocarbons with high octane numbers. Since the previous reports, a continuous zinc chloride hydrocracker has been built and operated, confirming the batch work, and providing effluent catalyst for regeneration studies as well as information on corrosion. A continuous catalyst regeneration unit was also operated. A complete report on all of the work is available from the U.S. Department of Interior, Office of Coal Research (1969). This paper reports rate data on the hydrocracking of extract with zinc chloride via a correlation which predicts conversion of extract in both batch and continuous units. Experimental
Feedstock. The feedstock used for all the work described was a coal extract prepared by continuous extraction of 546
I B E C PROCESS D E S I G N A N D DEVELOPMENT
Pittsburgh Seam coal, Ireland Mine, using Tetralin solvent a t 750" F. with a residence time of 40 minutes. Unextracted coal and mineral matter were removed by filtration a t 646°F. The final extract represents 81% of the moistureand ash-free coal (MAF coal). Properties are given in Table I.
Table 1. Properties of Coal and Extract Used
Ireland
Mine Coal
Coal Extract
MF BASIS Volatile matter Fixed carbon FeS FeS2 Other ash
40.32 46.66 0.08 4.19 10.15
MAF BASIS Hydrogen Carbon Sitrogen Oxygen (by diff.) Sulfur Solvent fractions Benzene-insolubles Asphaltenes Oil
5.73 81.90 1.58 8.61 2.18
6.03 84.65 1.53 6.14 1.65 33 47 20
POWDERED METERED HYDROGEN C4TALYST UTAMELT % TANK
--
DLLT
%‘R
WEIGHING CATALYSTTANK
1
F E D PUMP LUMP EXTRACT.
9 HYDROCRACKING
EXTRET MELT T4NK
Figure 1 . Simplified flow diagram of continuous hydrocracker using zinc chloride catalyst
A measure of the asphaltic nature of extract is given by solvent fractionation. “Benzene-insolubles” is the fraction of extract insoluble in benzene a t its atmospheric boiling point. “Asphaltenes” is the fraction soluble in benzene, but insoluble in cyclohexane, determined by mixing 1 part of benzene-soluble extract with 9 parts of benzene and 100 parts of cyclohexane, by weight, and filtering a t room temperature. The fraction soluble in this mixture is termed “oil.” Catalyst. The zinc chloride used was Fisher Scientific Co. certified reagent, 96 to 9 8 5 pure, dried before use in a vacuum a t 110°C. After this treatment, it contained 1 to 1.5 weight Llc water and 1 to 1.8 weight ci; zinc oxide. Before feeding this to the continuous unit, the remaining water was removed by melting the salt under full pump vacuum. For a number of tests, zinc oxide was added to the zinc chloride to bring the total zinc oxide up to that required for stoichiometric reaction with HC1 liberated by sulfur in the extract reacting with zinc chloride (4.65 weight L c of the catalyst). The zinc oxide used was Baker and Adamson certified reagent; the lots used were 99.0 to 99.EiLcpure. Equipment. The basic unit for the batch hydrocracking tests was a 300-ml. rocking autoclave (American Instrument Co., Catalog No. 40-2150). The normal rocking motion of 36 cycles per minute about the axis was revised to give 86 cycles per minute through an angle of 30” on the end of a 12-inch lever arm. The equipment and procedures have been described (Zielke et al , 1966). The continuous hydrocracking unit flow sheet is given in Figure 1. The reactor is a “stirred-tank” reactor, to which extract, catalyst, and hydrogen are fed separately. Constant inventory in the reactor is maintained by a pressure-drop controller operating on a valve in the “spent” catalyst withdrawal line. The reactor is 3 inches in diameter and contains a liquid inventory of 400 to 600 grams. All products were collected and analyzed, and the results consolidated. Elemental balances were then made and the zinc balance was forced by adjusting the residue collected. This avoided errors due to differing bed levels a t beginning and end of the balance period. The carbon balance was then forced by adjusting the amount of extract fed (uncertainty existed as to the density in the feed pump). The nitrogen, oxygen, and sulfur balances
were forced by assuming the unaccounted-for products to be NH3, H 2 0 , and H2S,respectively, which have reacted with the catalyst. Chlorine was forced by assuming that the unaccounted-for chlorine was present in the catalyst as the double salt, ZnCl2.NH4C1.The hydrogen consumption was then obtained by difference between the hydrogen in the products and that in the feedstock (exclusive of H P gas). After forcing all elemental balances, the overall material balance also closes, and the conversions reported in Table IV are reported on this basis. Conversion, as used here, is the conversion of extract to products boiling below 400°C. (752°F.) a t normal atmospheric pressure, calculated by subtracting the percentage of +4OO0C. residue from 100%;. Because of the heat sensitivity of the materials involved, the 400” C. cut point in a distillation is defined as 240°C. in the pot of a simple flask when the pressure is 1 torr. Zinc chloride in the presence of liquid water can cause typical chloride-ion stress corrosion cracking of austenitic stainless steel. For experimental units, such lowtemperature zones as the product condensation train should be made of metals such as Inconel 625, which are not subject to cracking. Type 316 stainless steel was successfully used for the reactor and catalyst feed system, however, since these areas contain only “dry” zinc chloride. Discussion of Results
Results of the continuous unit generally confirmed previous batch results, showing higher conversion a t lower levels of zinc oxide. Temperature has the expected result-more rapid reaction at higher temperatures, but a t temperatures above 750” F., the maximum liquid yield decreases. Straight-run gasolines from zinc chloride hydrocracking show Research octane numbers (clear) of 87 to 91 and a high sensitivity to lead. With 3 ml. of tetraethyllead per gallon, the R O S values rise to 99 to 101. If re-formed, the product would be 85 to 90% aromatics, suitable for a high-octane blending stock for gasolines or a source of benzene via hydrodealkylation. Kinetic Ahlode].The basic model used to correlate hydrocracking results assumes that extract is composed of many different compounds, each hydrocracking independently a t a different rate that can be expressed as a first-order reaction. The situation has been greatly simplified by assuming only three species present: an unconvertible fraction, a fraction which converts rapidly, and a fraction which converts relatively slowly. For simultaneous independent reactions in a batch unit,
where C, - C is the convertible material remaining a t time t , and N1 is the fraction of the convertible material which reacts via the fast reaction. For a single perfectly mixed, continuous reactor, the probability that a microscopic drop of liquid will have a residence time in the reactor between t and t + dt is 1 / r (e-t” d t ) where T = residence time in the continuous unit The average fraction of extract unconverted then would be
where VOL. 8 NO. 4 OCTOBER 1 9 6 9
547
c, - c (7) H
is the fraction unconverted a t time t as determined from batch data. Substituting from Equation 1 for
H
gives:
Evaluation of Constants. The data available consist of 38 runs in the batch autoclave and 14 runs in the continuous unit, covering the following ranges:
Temperature, F. Partial pressure of H,, p.s.i.g. Residence time, minutes Catalyst/extract weight ratio ZnO ZnCL mole ratio Conversion, wt. 1 of extract
Ratch unit
Continuous Unit
700-775 lt500-35O0 15-360 1.o 0.013--0.0566 41-93
2730-3300 22.8-192.6 0.9-1.3 0.01-0.08 62-89
K a
iX
u
z
P v) K
i ,
I
0.04
L
I
I
/
0.08
0.12
R A T E OF CONVERSION OF E X T R A C T , W T %/MINUTE
Figure 2. Extrapolation of batch rate data to determine C, 548
The first equation applies over the hydrogen partial pressure range of 2500 to 3500 p.s.i.g.:
2500) + c, = 90 + (p -600
700-800
The more extensive batch data have been used t o determine constants in the kinetic equations wherever applicable. However, because of the uncertainty in low residence times inherent in batch runs-Le., some conversion takes place as the reaction temperature is approached-the continuous unit data have been used to determine the rate constant for the fast reaction. It is assumed that a certain fraction of the extract fed is not convertible under the conditions of temperature, pressure, and catalyst composition used, even a t infinite residence time, although the amount may vary with operating conditions. The extent of possible conversion, C,, has been determined by plotting conversion against residence time for each set of conditions (batch data). Tangents drawn to each curve yield the slope, or rate of reaction, dC dt. These instantaneous rates were then plotted against conversion, yielding results shown in Figure 2 . The straight lines obtained confirm the fact that the
75
data can be represented as first-order reactions and when extrapolated to zero rate of conversion yield C,. The values of C,so obtained are listed in Table I1 as "observed C,." Variables of temperature, hydrogen pressure, and ZnOiZnCl, ratio all affect C,. The value of C, is independent of whether batch or continuous data are involved, since they are equal a t infinite residence time. T o estimate C, for conditions other than those a t which runs were made, empirical equations have been developed.
I & E C PROCESS D E S I G N A N D DEVELOPMENT
(750 - T ) 25
+
10 (0.08)-
X)
(3)
For the pressure range, p = 1500 to 2500: -p + c, = 90 - 2500 200
(750 - T ) 25
+
10 (0.08- X)
(4)
The calculated values of C, for the batch runs, also listed in Table 11, compare well with the observed values. Equations 3 and 4 were used in calculating conversions for both batch and continuous runs. They illustrate the facts that C, increases a t low levels of ZnO, at higher pressures, and a t lower temperatures (within the 700" to 800" F. range). At long residence times where the fast reaction is substantially complete, and where N,is constant, Equation 1 shows that a plot of log (C, - C) us. residence time should be a straight line:
log (C, - C) = log (1 - N , ) c, - h,t The slope of such a line is hr and the intercept on the ordinate is log [ (1 - N , )C,]. This provides a technique for determining kr and N 1 . Figure 3 is a representative plot of this type. Again, the confirmation of straight lines can be seen if we ignore runs a t low residence times where the fast reaction is incomplete. The values of hZ and N , from plots like Figure 3 are listed in Table 11. I t can be seen that 67 to 8 1 ' ~ of the convertible extract reacts via the fast reaction ( N I ) . To simplify further calculations, a constant value of NI = 0.75 has been taken as representative of all conditions. The reaction rate constants for the slow reaction, h ~ , have been plotted for the hydrogen pressure level of 2500 p.s.i.g. as an Arrhenius plot in Figure 4. Because the ratio of ZnO to ZnClr is an important parameter, only two points are available a t a single condition. A line was drawn through these points and lines a t other ZnO/ ZnClZ ratios were drawn parallel to it. A line a t ZnO/ ZnClz = 0.02 was interpolated for future use with continuous unit data. Based on the relatively mild effect of pressure on k z shown in Table 11, the values in Figure 4 have been used for estimating conversions a t all pressures.
Table II. Values of
Temp., F.
ZnO/ZnC12 Mole Ratio
700
0.013
750
0.03
1500 2500 2500 3500 1500 2500 2500 3500
0.0765 775
C,, NI,and kz
H , Partial Pressure, P.S.I.G.
0.0765
Determined from Batch Runs
C,, W t . % of Extract Obsd." Calcd.* 87 92 91 92 a5 90 88 (90)
NI'
87.7 92.7 90.5 92.1 85.0 90.0 89.0 89.6
k2,dMin.
'
o.oo5a 0.0048 0.020 0.030 0.0062 0.0063 0.021
0.670 0.712 0.736 0.740 0.750 0.773 0.810 0.788
...
"Intercepts on Figure 2 and similar plots. bCalculated from Equations 3 and 4 . 'Determined from intercepts on Figure 3 and similar plots. From slopes of Figure 3 and similar plots.
05
0.4
d
0.3
E
9 a
02
W
I
c
6
010
z
f
008
z
8
E
0.06
I
775
0.05 0.80
RESIDENCE
TIME,
MINUTES
Figure 3. Effect of residence time on residue remaining
RECIPROCAL TEMPERATURE,
TFl
Figure 4. Reaction rate constant for slow reaction vs. reciprocal temperature
I
1
I 750
725
1 7WoF I
0.82
RECIPROCAL
0.84
TEMPERATURE,
0.86
%
Figure 5. Reaction rate constant for fast reaction vs. reciprocal temperature
The remaining constant in Equations 1 and 2 is k l , the reaction rate constant for the fast reaction. This has been determined from the continuous unit data by using the data from one run a t each condition and solving for k l in Equation 2 . Values of C, were calculated from Equation 3 or 4, those for k 2 were taken from Figure 4, and N1 was taken as 0.75. Generally, the run with the lowest conversion in each set was used to calculate h l , since it would be most representative of the fast reaction. The values of hl obtained are listed in Table I V along with other information on the continuous runs. An Arrhenius plot of hl is given in Figure 5 , showing the desired straight line and indicating an activation energy for the fast reaction of 35 kcal. per gram mole. The values of k l are independent of the ZnO/ZnClr ratio, in contrast to those for h2. Since the range of hydrogen pressures was narrow in the continuous data, the effect of hydrogen pressure, if any, on k l could not be determined. The energy of activation of 35 kcal. per gram mole indicates that the rate-determining step involves a chemical reaction rather than being limited by accessibility of hydrogen in the bulk liquid or through a boundary film. VOL. 8 NO. 4 OCTOBER 1 9 6 9
549
Table Ill. Calculated and Observed Conversions for Batch Runs
Temp , F 700
ZnO ZnCL Mole Ratio
H Partiul Pressure, PSIG
0.013
1500
k l , dM i n
I
Residence Time ' Min
0.078
2500
0.03
750
2500
0.24
3500
750
0.0765
1500
0.24
2500
775
0.0765
2500
Wt Obsd.
0.43
3500
Conversion, cc o f Extract Calcd.
15 60 120 180 360
41.2 61.2 71.0 76.6 82.6
46.8 71.0 76.1 79.3 84.3
15 60 120 180 360
49.1 70.6 78.8 83.4 89.0
49.5 73.6 80.5 83.9 89.2
15 60 120 180
73.0 82.6 88.7 90.3
'72.3 84.3 88.9 90.4
15 60 120 180
T8.6 87.1 91.6 92.1
73.1 85.1 89.9 91.4
15 60 120 180 360
52.0 68.8 74.9 80.0 83.4
64.0 70.5 74.9 78.1 82.8
15 60 120 180 360
58.1 76.7 80.5 83.8 88.3
67.8 74.6 79.3 82.7 87.7
15 60 120 180 360
71.3 81.9 86.4 87.6 87.7
71.9 80.2 86.2 87.5 88.0
15 60 120 180 360
71.7 80.4 83.5 86.5 89.9
73.5 82.6 87.2 89.5 90.0
From Figure 5. O Residence time at reaction temperature. ' Calculated from Equation 1
Table IV. Calculation of Rate Constants and Comparison of Experimental and Calculated Conversions in Continuous Unit
Temp., F. 775
750
ZnO;ZnC12 Mole Ratio
H 2 Pressure, P.S.I.G.
Residence Time, Min.
0.07 0.07 0.09 0.08 0.08 0.08
3040 3000 2860 2860 3100 3300 3090 3040 3190 3150 3040 2815 2730 3210
46.7 95.0 96.8 101.0 102.6 22.8 51.2 105.0 24.3 26.6 102.0 102.0 111.7 80.5
0.02 800 776 700
0.08 0.02 0.02
C,"
k2,' Min.-'
k,,'Min.-'
89.5
0.021
0.43
90.0 91.0
0.0063
0.24
92.0
0.051
0.24
88.5 90.0 93.0
0.070 0.16 0.003
0.72 0.43 0.078
Conversion Wt. % of Extract Obsd. Calcd.' 75.0d 83.1 82.0 85.0 87.1 61.7 68.8' 74.1 71.7' 72.2 84.0 86.6 88.5 64.3'
75.0 80.5 80.5 80.7 81.4 60.7 68.6 74.6 71.7 72.0 87.9 85.0 87.5 64.3
'Calculated uia Equation 3. 'From Figure 4 . 'Calculated uia Equation 2 using N , = 0.75. dExperimental point used to calculate k l .
550
I & E C PROCESS D E S I G N A N D D E V E L O P M E N T
represented by two or more simultaneous first-order reactions represents both batch and continuous unit data. An average of 75% of the extract converts via a fast reaction with an activation energy of 35 kcal. per gram mole. The reaction rate constant for this reaction is affected somewhat by hydrogen pressure, but is independent of the ratio of ZnO to ZnCle. The rate constant for the slow reaction, which may be lower by a factor up to one fortieth, is independent of hydrogen pressure (1500 to 3500 p.s.i.g.1, but a strong function of how much ZnO is present. Acknowledgment
5 a
40
Y2
0
v
I
9 m A
I
750 xa)-3300 008 775 a m - 3 1 0 3 007-om 775 2700 0.02
Nomenclature
1
30 0
30
The assistance of the U.S. Office of Coal Research in sponsoring this work is gratefully acknowledged. Many individuals of the Consolidation Coal Co., Research Division, in addition to the authors were involved in the experimental work for this paper. Especially noteworthy were the contributions of C. P. Costanza, J. M. Evans, and J. M. Pazzo.
60
90
RESIDENCE T I M E ,
I20
I50
MINUTES
Figure 6. Variation of conversion with residence time in continuous reactor
I t coincides with the value found by Weller et al. (1951) for hydrogenation of asphaltenes from coal with tin catalyst. This also was a first-order reaction. Comparison of Calculated and Observed Conversions. The values of C,, N 1 , h l , and k Z as determined above were used in Equation 1 to predict the conversion for each . . . . - . . --. . . run in the batch unit. ‘l’able111 compares these calculated conversions with those actually observed. The “fit” is good, particularly a t high conversion levels. The poorest fit occurs a t low conversions and low hydrogen pressures. This suggests that the reaction rate constant for the fast reaction, hl, actually is a function of pressure, although the data do not permit this relationship to be properly defined. Similar calculations for data from the continuous unit via Equation 2 are shown in Table IV. Again, the observed and predicted conversions show good agreement. The experimental data and calculated curves for 3000 p.s.i.a. of hydrogen are shown in Figure 6. The “fit” is good except for a few of the 775°F. points a t 0.07 to 0.08 mole ratio ZnO/ZnCl*, which show 2 t o 6% higher conversions than predicted. 1
Conclusions
The kinetic model which assumes that the hydrocracking of coal extract with zinc chloride catalyst can be
C = conversion of extract to -400” C. products, weight % of MAF extract
C,
= conversion a t which rate of conversion becomes
hi
= reaction rate constant for extract conversion via
zero fast reaction, min. k2 = reaction rate constant for extract conversion via slow reaction, min. Ni = fraction of C, which occurs via fast reaction P = partial pressure of hydrogen, p.s.i.g. T = temperature, F. t = residence time in a batch or plug flow reactor, min. x = ZnO/ ZnCL mole ratio in catalyst 7 = average residence time in continuous reactor, weight of liquid phase in reactor divided by weight rate of liquid product/min. Liquid is “natural” liquid existing a t reactor conditionsLe., catalyst plus unvaporized oil. Literature Cited
U. S. Department of Interior, OCR Research and Development Rept. 39, Vol. 111, Book 1, 1969. Weller, Sol, Pelipetz, M. G., Friedman, Sam, Ind. Eng. Chem. 43, 1572 (1951). Zielke, C. W., Struck, R . T., Evans, J. M., Costanza, DESIGN C. P., Gorin, Everett, IND.ENG.CHEM.PROCESS DEVELOP.5, 131-7, 158-64 (1966). RECEIVED for review November 4,1968 ACCEPTED July 10,1969 Division of Fuel Chemistry, 156th Meeting, ACS, Symposium on Synthetic Fuels from Coal, Atlantic City, K.J., September 1968.
VOL. 8 NO. 4 OCTOBER 1 9 6 9
551
