Ind. Eng. Chem. Process Des. Dev. 1901, 20, 53-62
8 and 9. These figures provide reasonable estimates of the relative quantities of the heterocyclics if each of the initial ring saturation reactions in quinoline HDN were at equilibrium. This can actually occur only if the hydrogenation and dehydrogenation reactions among the heterocyclics are fast relative to the PyTHQ and DHQ hydrogenolysis reactions-a condition approached in our studies. The heterocyclic equilibria favor DHQ at lower temperature and high hydrogen pressure, while quinoline is favored at higher temperature and lower hydrogen pressure. This is consistent with the fact that the ring saturation (hydrogenation) reactions are exothermic and consume hydrogen. The behavior of the tetrahydroquinolines is more complex, since they are partially saturated species subject to either hydrogenation or dehydrogenation. Thus at constant hydrogen partial pressure, the equilibrium concentration of PyTHQ or BzTHQ proceeds through a maximum as temperature increases. The effect of hydrogen pressure on the equilibrium concentration of PyTHQ or BzTHQ depends on the temperature. For example, a t 420 O C an increase in hydrogen pressure from 3.53 MPa to 6.98 MPa increases the equilibrium concentration of PyTHQ or BzTHQ, but the opposite effect occurs a t 330 " C (compare Figures 8 and 9). In commercial hydrotreating processes aimed at nitrogen removal, it is desirable to minimize hydrogen consumption for economic reasons. The equilibrium behavior of the heterocyclics has significant implications in this regard. Saturation of the aromatic ring of quinoline to form BzTHQ is thermodynamically (but not necessarily kinetically) slightly more favorable than saturation of the heteroring, to form PyTHQ (see Figures 8 and 9). (This is consistent with the fact that, at HDN temperatures, saturation of benzene to cyclohexane is thermodynamically
53
more favorable than saturation of pyridine to piperidine for a given temperature and hydrogen partial pressure.) For quinoline HDN, particularly at the higher hydrogen pressures required to achieve satisfactory nitrogen removal rates, the equilibria among the heterocyclics permit substantial oversaturation to DHQ (see Figure 9). These results can most likely be extended, at least qualitatively, to other multiring heterocyclic nitrogen compounds such as acridine and carbazole. Thus the burden of selectively hydrogenating only the heterorings is placed solely on the HDN catalyst. Present commercial hydrotreating catalysts, however, have been shown to exhibit little selectivity for HDN reaction pathways of minimum hydrogen consumption (Shih et al., 1978; Satterfield and Cocchetto, 1980).
Literature Cited Benson, S. W.; Cruickshank, F. R.; Golden, D. M.; Haugen, 0. R.; O'Neal, H. E.; Rodgers, A. S.; Shaw, R.; Welsh, R. Chem. Rev. 1969, 69, 279. Cocchetto, J. F. S.M. Thesis, M.I.T., Cambrldw, Mass., 1974. Cocchetto, J. F. Ph.D. Thesls, M.I.T., Cambridge, Mass., 1979. Cocchetto, J. F.; Setterfield, C. N. Ind. Eng. Chem. Process Des. Dev. 1978, 75, 272. Satterfield. C. N.; Cocchetto, J. F. Ind. Eng. Chem. Process Des. Dev. 1981, accompanying paper In this Issue. Satterfield, C. N.; Modell, M.; Hites, R. A.; Declerck, C. J. Ind. Eng. Chem. Process Des. Dev. 1978, 17, 141. Shih, S. S.; Katzer, J. R.; Kwart, H.; Stiles, A. 6. Am. Chem. Soc.Div. Pet. Chem. Repr. 1977, 22, 919. Shih, S.; Relff, E.;Zawadrki, R.; Katzer. J. R. Am. Chem. Soc. Prepr.. Fuekr Dlv. 1976, 99. Stuli, D. R., Westrum, E. F.. Jr.; Sinke. 0. C. "The Chemical Thermodynamics of Organic Compounds", Wiley: New York, 1969; pp 231, 361, 370. van Krevelen, D. W.; Chermin, H. A. G. Chem Eng. Sci. lS51, 7 , 66.
Receiued for reuiew February 19,1980 Accepted July 21,1980 The work was supported in part by funds from the U S . Environmental Protection Agency.
Reaction Network and Kinetics of the Vapor-Phase Catalytic Hydrodenitrogenation of Quinoline Charles N. Satterfleld" and Joseph F. Cocchetto' Department of Chemical Engineering, Massachusetts Institute of Technology, CambMge, Massachusetts 02 739
Studies were made in a continuous-flow microreactor at 3.55 and 7.0 MPa, 330 to 420 O C , and over a range of contact times, on a presulfided commercial NIMo/A120, catalyst. Quinoline and various intermediate products were reacted individually. In all cases the predominant hydrocarbon product was propylcyclohexane. Reactions among quinoline and its hydrogenated heterocyclic derivatives were all found to be reversible. The complex kinetic behavior observed can be well interpreted in terms of equilibrium limitations on the initial ring saturation reactions as well as large variations among the relative adsorptivities of the various Ncontainingspecies present initially and formed in the reaction. The kinetics of the irreversible intermediate steps-denitrogenation of o-propylaniline (OPA), hydrogenolysis of Py-tetrahydroquinoline (PyTHQ) and of decahydroquinoline (DHQpare modeled.
The titled reaction is a good model for hydrodenitrogenation (HDN) of six-membered ring nitrogen heterocyclic compounds in middle distillate fuels having a high organonitrogen content such as those derived from oil shale, coal, and low-grade petroleum. The steps curSchool of Chemical Engineering, Cornell University, Ithaca,
N.Y. 14853. 0196-4305/81/1120-0053$01.00/0
rently believed to be significant in the reaction network for quinoline HDN, based on the results of the present and previous studies, are shown in Figure 1 of the paper accompanying this one (Cocchetto and Satterfield, 1980). This also lists the acronyms used in both papers for the various compounds discussed. Model compound studies of various heterocyclic nitrogen compounds have shown that hydrodenitrogenation proceeds via saturation (hydrogenation) of the heterocyclic 0 1980 American Chemical
Society
54
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 1, 1981 KEY -
GSV
U
U U Figure 1. Schematic of experimental apparatus.
U
ring, followed by hydrogenolysis of C-N bonds first to open the heteroring and then to convert the resulting aliphatic or aromatic amine intermediates to hydrocarbons and ammonia. Under HDN reaction conditions, hydrogenolysis of C-N bonds is essentially irreversible, but saturation of heterocyclic rings is potentially reversible depending on the temperature and hydrogen partial pressure (Cocchetto and Satterfield, 1976). In a separate publication (Cocchetto and Satterfield, 1980) we report on the chemical equilibria among quinoline and its three hydrogenated derivatives, PyTHQ, BzTHQ, and DHQ. A thermodynamic limitation on heterocyclic ring saturation can occur and will adversely affect the overall HDN rate if the kinetics of the various reaction steps are such that heterocyclic ring opening is substantially rate-limiting (Satterfield and Cocchetto, 1975). In general, both hydrogenation and hydrogenolysis rates may be important in determihing the overall HDN rate, though their relative importance is undoubtedly a function of reaction conditions, catalyst employed, and nature of the nitrogen compound. Previous Studies In a recent paper we report on the intermediate reactions in quinoline HDN, as determined in a vapor-phase microreactor packed with presulfided NiMo/A1203catalyst (Satterfield et ai., 1978). This paper also summarizes previous studies. Rapid equilibration between quinoline and PyTHQ occurred at all reaction conditions. At lower temperatures this equilibrium favored PyTHQ, which was then converted to either o-propylaniline (OPA) or decahydroquinoline (DHQ), but a t higher temperatures the conversion rate of quinoline to Bz-tetrahydroquinoline (BzTHQ) and subsequently to DHQ became significant. Shih et al. (1977) investigated the kinetics of quinoline HDN in a batch liquid-phase (slurry) reactor employing a sulfided NiMo/A1203catalyst and a paraffinic white oil carrier liquid. They, too, observed rapid equilibration between quinoline and PyTHQ, but the other hydrogenation reactions as well as the hydrogenolysis reactions were reported to be kinetically controlled. These reaction rates were each modeled by first-order kinetics, and the corresponding rate constants were determined by a computer fitting technique. It was concluded from the kinetic analysis that nitrogen removal occurred primarily through the BzTHQ and DHQ intermediates. They used a nitrogen-specific detector and therefore were not able in general
CHECK V A L V E
HEATED L I N E
G A S SAMPLING VALVE
CHROMATOGRAPH I V t N I
to analyze for hydrocarbon products. The rates of the various steps in HDN reactions have often been described by first-order or pseudo-first-order kinetics (Flinn et al., 1963; Sonnemans et al., 1973; Shih et al., 1977). The first-order rate constant is found to decrease with increased initial nitrogen concentration, as the result of inhibition of the reaction rates by the strongly adsorbed nitrogen compounds. This can be interpreted through a Langmuir-Hinshelwood kinetic model, but to simplify the mathematical analysis the various nitrogen compounds have all been assumed to have the same adsorptivity. However, some HDN reaction studies as well as adsorption studies show that in fact the strength of adsorption of nitrogen compounds may vary significantly. In a study of shale oil HDN, for example, Koros et al. (1967) found that indole-* compounds were less reactive than the quinolines, but the opposite reactivities have been reported for indole and quinoline when studied separately (Flinn et al., 1963; Aboul-Gheit and Abdou, 1973). This apparent discrepancy was attributed to competitive adsorption effects in mixtures, in which the more basic quinoline-type compounds were preferentially adsorbed and thus converted. However, the relatively nonbasic compounds (indoles) are more reactive than the basic compounds once they are adsorbed onto the catalyst. Reaction and adsorption studies also suggest that hydrogen and the nitrogen compounds adsorb on different, perhaps neighboring, catalyst sites, while the adsorption of hydrocarbons is relatively weak and can usually be neglected (Sonnemans et al., 1973). Experimental Apparatus and Procedure The HDN studies were carried out in a continuous flow, fmed-bed catalytic reactor system employing a presulfided commercial NiMo/A1203hydrotreating catalyst (American Cyanamide AERO HDS-3A, containing about 3 wt % NiO and 15 wt % MoOJ. A schematic of the experimental apparatus is shown in Figure 1;the heavy lines indicate the primary flows during steady-state operation. Liquid reactant (quinoline or one of ita hydrogenated derivatives) was metered into the system with a high-pressure pump and was flash-vaporized into a stream of heated hydrogen. The final preheating coil and the reactor were both immersed in a fluidized sand bed heater for controlled isothermal operation. The reactor was a 3.86 mm i.d. stainless steel tube, packed with 1.5 g of 20/24 mesh catalyst particles (0.774
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 1, 1981 55
mm average diameter). A single charge of catalyst was used for all experiments reported here. Heat or mass transfer limitations in the system were insignificant. Reactor pressure was maintained at the desired level by the reactant hydrogen cyclinder regulator, and total flow rate through the system was controlled with a very fine metering valve downstream from the reactor. Soap-film flowmeters were used to measure exhaust gas flow rates, which varied from about 100 to 4000 standard cm3/min. A tar trap in the reactor exit line was maintained at approximately 240 "C, a temperature too high to condense any of the reactants or major reaction products, but sufficiently low to condense very high-boiling "tar" compounds that would otherwise have condensed in cooler downstream portions of the apparatus, particularly in the analytical system. Vapor-phase samples of the reactor effluent were injected into an on-line gas chromatograph by means of a heated gas sampling valve. The gas chromatograph was equipped with a thermal conductivity detector. For separation of reaction products, a 10 f t X 2 mm i.d. glass column packed with 10% SP-2300 on Chromosorb T (a Teflon support) was temperature-programmed from 110 to 190 "C a t 40 "C/min and held at 190 "C until analysis was completed. The relatively inert Teflon support virtually eliminated the "tailing" of basic amines, particularly ammonia, observed with more conventional diatomite supports. With 20 cm3 /min hydrogen carrier gas flow, analysis time was only 20 min. Absolute detector response factors were determined by injecting known quantities of the relevant compounds into the gas chromatograph and measuring the corresponding peak areas, at the same conditions used for analysis of reactor effluent samples. Except for PCHA, all species shown in Figure 1 of Cocchetto and Satterfield (1980) were identified by gas chromatography by comparison with known samples and/or by gas chromatography/mass spectrometry. An experimental run consisted of the determination of steady-state product distributions for the reactant compound under study, a t typically five different feed rateg, for a constant reactor temperature, pressure, and feed composition. A run was started by heating the reactor to the desired temperature (330, 375, or 420 "C) under an argon purge, and then pressurizing with hydrogen to 3.55 or 7.0 MPa. After hydrogen and liquid feed rates were established and steady state had been reached, a set of 3 to 5 samples was taken and analyzed. The hydrogen and liquid feed rates were then reduced in the same proportion, and after a l-h re-equilibration period, another set of steady-state samples was taken. This procedure was repeated for each of the remaining feed rates to be studied, so the sequence of studies for each run was from the highest to the lowest feed rate. Since sulfur compounds were not present (intentionally) in the reactor feeds, the catalyst was resulfided with a 10% H a in H2mixture after each run, to maintain reproducible activity. The averaged component peak areas from each set of steady-state samples were converted to the equivalent number of moles of each component. The product distribution was then calculated for each set of reaction conditions. Results The HDN of quinoline was studied at the following conditions: temperature, 330, 375, and 420 "C; total pressure, 3.55 MPa (500psig or 35 atm) and 7.0 MPa (lo00 psig or 69 atm); reactant partial pressure, 13.3 kPa (0.13 atm) and 26.7 kPa (0.26 atm); space time based on quinoline feed rate only, W / F Q ,41.7 ~ , to 667 h g of cat./g-mol
of Q. Each of the four intermediates PyTHQ, BzTHQ, DHQ, and OPA was also studied separately at 375 "C, 7.0 MPa, and over a range of contact times. Some of the intermediates were studied at other conditions as well. Hydrogen partial pressure was essentially equal to the total pressure. W is the mass of virgin catalyst (before sulfding) in the reador and Pi,-,is the molar feed rate of the reactant nitrogen compound to the reactor. Nitrogen removal, or denitrogenation, is defined here as the mole percentage of the reactant nitrogen compound converted to hydrocarbons. The product distributions are calculated on the basis of organic products only (i.e., on an ammonia-free basis), so the mole percentage of each organic product (including, of course, unconverted reactant) is equivalent to the mole percentage of the reactant converted into that product, if minor side reactions such as coking are neglected. Because of limitations on the quantity of data that can be presented clearly in a single graph, however, only the organic nitrogen compounds are shown in the graphical product distributions. Catalyst Activity. The activity of the presulfided NiMo/A1203catalyst was checked periodically by reacting quinoline at carefully selected standard conditions (375 "C, 7.0 MPa, 13.3 kPa Q, and 667 h g of cat./g-mol of Q). Complete denitrogenation occurred with the catalyst until nearly 120 h of operation, dropping to 65% nitrogen removal after 340 h and a stable level of activity, corresponding to about 45% denitrogenation, after approximately 400 h on stream. After this time the quinoline HDN product distribution at standard conditions also stabilized. In each of two experimental runs at different reaction conditions, data taken at the beginning of the run were reproduced at the end of the same run. We conclude that no significant catalyst deactivation (as by sulfur loss) took place during individual experimental runs,at least not after the catalyst had been used for 340 h. Catalyst activity can safely be eliminated as a variable in the HDN kinetic experiments reported here, essentially all of which were conducted after 400 h of catalyst use. Hydrodenitrogenation of Quinoline and Py-Tetrahydroquinoline. The percent denitrogenation was quite insensitive to initial quinoline partial pressure at constant ~ each temperature, hydrogen pressure and W / F Q , For quinoline feed rate, doubling the hydrogen pressure increased nitrogen removal by a factor of about four at 420 "C, but the effect was less dramatic at lower temperatures. Quinoline denitrogenation was very sensitive to temperature; at 7.0 MPa and 667 h g of cat./g-mol of Q, nitrogen removal increased from only 6% at 330 "C to 42% at 375 "C, and was 100% at 420 "C. Representative product distributions from quinoline HDN are shown in Figures 2, 3 and 4. These were relatively unaffected by changes in initial quinoline partial pressure. As contact time was increased, the concentrations of PyTHQ, BzTHQ, and DHQ intermediates each passed through a maximum, as did the OPA concentration at the most extreme reaction conditions where complete denitrogenation was observed. The Q/PyTHQ product ratio during each of the 420 "C runs was nearly constant, at essentially the equilibrium ratio. This equilibrium is much more favorable toward PyTHQ at the higher hydrogen pressure, which also favored DHQ relative to the tetrahydroquinolines. One would expect lower temperatures to favor the hydrogenated species in each of the initial ring saturation equilibra, since the hydrogenations are all exothermic. Quinoline HDN product distributions at 375 and 420 "C are in general consistent with this ex-
56
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 1, 1981 I00
P
r,T-T~-I ~
r-I-
--I -TT-
80i-
7-1I-T--T-1
I
i
1
I
PyTHO
0
0
I00
200
W I Fo0
300
hi g
400
500 600 700 800
cotolyst / g -mol
W/FQo
a
Figure 2. Product distributions for quinoline HDN at 420 O C , 3.55 MPa, 13.3 kPa Q (run 20).
hi
g
c o t o l y s t I q -mol
Figure 4. Product distributions for quinoline HDN at 375 O C , 7.0 MPa, 13.3 kPa Q (run 23). I
:
I C ~zoo WIFh
hI g
30c
400
500
~ o t o l y i~g t- m o i
6?c
7)
a>-
3
Figure 3. Product distributions for quinoline HDN at 420 "C, 7.0 MPa, 26.7 kPa Q (run 21).
pectation (compare Figures 3 and 4), indicating that the initial ring saturation reactions approached equilibrium. However, only the Q/PyTHQ equilibrium was approached closely at 330 "C. At 375 "C, quinoline was rapidly hydrogenated to an equilibrium amount of PyTHQ, but with increased contact time, the amount of unreacted quinoline levelled off while the amount of PyTHQ in the products decreased quite rapidly (Figure 4). In other words, quinoline and PyTHQ were in equilibrium at the shortest contact times, but did not remain in equilibrium a t longer contact times. Hydrodenitrogenation of either quinoline or PyTHQ at 375 "C and 7.0 MPa resulted in essentially the same product distributions (compare Figures 4 and 5); thus the unusual behavior of the Q/PyTHQ product ratio was observed with PyTHQ feed as well. Hydrodenitrogenation of o-Propylaniline. Figure 6 illustrates the effects of reaction variables on nitrogen removal from OPA a t 7.0 MPa total pressure. Doubling the partial pressure of OPA from 13.3 to 26.7 kPa in the had no effect on perreactor feed, at constant W/FOPAVO,
3
0
100
200
W'FP,TH,~
I
I
I
500 600 700 800 , hr g c a t a l y s t / g - m o l P y T H Q
300 400
Figure 5. Product distributions for Py-tetrahydroquinoline HDN at 375 "C, 7.0 MPa, 13.3 kPa PyTHQ (run 33).
cent nitrogen removal at 375 "C. The rate of denitrogenation of OPA increased significantly with temperature and also with hydrogen pressure (not shown here). No heterocyclic nitrogen compounds were detected in the products from HDN of OPA at the conditions studied. Hydrodenitrogenation of Bz-Tetrahydroquinoline and Decahydroquinoline. The percent denitrogenation of quinoline and of each of several reaction intermediates at 375 "C and 7.0 MPa is shown as a function of contact time in Figure 7. Nitrogen removal from DHQ was significantly easier than from BzTHQ, while the compounds most resistant to denitrogenation were quinoline and PyTHQ. In view of previous reports, the most striking result is the relative ease with which OPA could be denitrogenated compared to the heterocyclic nitrogen compounds, when OPA was reacted in the absence of heterocyclics. The distribution of nitrogen-bearing products from BzTHQ HDN at 375 "C and 7.0 MPa is shown in Figure 8. BzTHQ hydrogenated quite rapidly to DHQ, from which hydrocarbon products were formed. The concen-
Ind. Eng. Chem. Process Des. Dev., Vol. 20,No. 1, 1981 57 100 an
i"(
L
I!
m.
z
0
I
I
60
L
a
z
irl
50
: T
4c
z
1
1
i
W
30
'?/f
C Q Q ~
20 I
, I
I
1
I
0 0
100
I
300
200
WiFopno
POIIIOI P r u s s u r ~ -I
A A 13.3 k P a OPA 0 0 2 6 . 7 k P a OPA
* hr
I 400
I 500
I
I
600
g Catalyst i g - m o l
700
0 800
OPA
I
' Y
"1 / 80
I
I
4 0 0 500 6 0 0 7 0 0 800 g cotalyst i g - m o l RzTHQ
300
200
, hr
Figure 8. Product distributions for Bz-tetrahydroquinoline HDN at 375 O C , 7.0 MPa, 13.3 kPa BzTHQ (run 35).
Figure 6. o-Propylaniline denitrogenation at 7.0 MPa. 100
100
WiFGITHao
I
OPA
60.-
< 20
4
10
0
100 WIF,,
200
300
400
500
% h i g cotolyst iq-mol
600
700
800
I
Figure 7. Denitrogenation of individual nitrogen compounds at 375 "C,7.0 MPa, 13.3 kPa feed partial pressure.
tration of DHQ went through a maximum as contact time was increased, reflecting its role as a reaction intermediate. Some dehydrogenation of BzTHQ or DHQ occurred, as evidenced by the formation of quinoline and PyTHQ. Figure 9 illustrates the product distribution from DHQ HDN at 375 "C and 7.0 MPa. The DHQ was simultaneously converted to hydrocarbon products and dehydrogenated to BzTHQ. Some dehydrogenation of DHQ to PyTHQ occurred to a lesser extent, but no OPA and only traces of quinoline were detected in the reaction products. A distinct maximum in BzTHQ formation was observed as contact time was increased. This most likely reflects an approach to hydrogenation/dehydrogenation equilibrium between BzTHQ and DHQ, since BzTHQ was not converted to any other products. Hydrocarbon Products. A significant observation is that the same hydrocarbons were formed from denitrogenation of OPA or from each individual heterocyclic nitrogen compound, and the same qualitative kinetic behavior was observed. The dominant hydrocarbon product in every case was propylcyclohexane (PCH). Even starting with OPA, PCH typically comprised over 75% of the hy-
_
0
L
'00
W FDHQo , h r q
r 750
i 800
c a ' i l y s t I g-mol DHQ
Figure 9. Product distributions for decahydroquinolineHDN at 375 O C , 7.0 MPa, 13.3 kPa DHQ (run 38).
drocarbon products, though this varied somewhat with reaction conditions. Smaller amounts of propylbenzene (PB) and propylcyclohexene (PCHE) were also formed, in general, while ethylcyclohexane (ECH) and etylbenzene (EB) were detected only at the higher temperatures. With increased contact time, PCHE formation usually went through a maximum, in contrast to the behavior of the other hydrocarbon products. The PCHE/PB and PCHE/PCH product ratios exceeded the corresponding equilibrium ratios, particularly at lower contact times; this is clear evidence that PCHE was an intermediate in the formation of at least some of the PB or PCH. At some reaction conditions the PCH/PB product ratio was less than the equilibrium ratio, but at other conditions the corresponding equilibrium ratio was exceeded. At the same reaction conditions, denitrogenation of BzTHQ or DHQ yielded a somewhat higher PCH/PB product ratio than did denitrogenation of quinoline, PyTHQ, or especially OPA. The implication is that some direct hydrogenolysis of OPA to PB and ammonia did occur to a minor extent.
58
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 1, 1981
Noncatalytic Reaction. Quinoline was exposed to a variety of HDN reaction conditions in the usual equipment but in the absence of the NiMo/A1203catalyst. Even at the most severe reaction conditions (420 "C, 7.0 MPa, and the same quinoline feed rate as that used with catalyst at 667 h g of cat./g-mol of Q), only the hydrogenation of quinoline to PyTHQ was observed to occur at an appreciable rate, but equilibrium was not attained. It is not clear whether this reaction was truly homogeneous (occurring in the gas phase) or was catalyzed by the stainless steel surfaces of the preheater and reactor walls. For the above pressure and quinoline feed rate the noncatalytic reaction resulted in a 41 %, 63%, and 79% approach to Q/PyTHQ equilibrium at 330, 375, and 420 O C , respectively. This is in sharp contrast to the rapid equilibration of quinoline and PyTHQ at all reaction conditions in the presence of the NiMo/Al,O, catalyst. Discussion Quinoline Hydrodenitrogenation Reaction Network. The quinoline HDN reaction network proposed is well supported by the experimental results. Propylcyclohexylamine (PCHA) was not detected in the HDN products, but is a plausible reaction intermediate which presumably denitrogenated very rapidly. Similar compounds, such as cyclohexylamine and ethylcyclohexylamine, have been shown to denitrogenate readily under HDN conditions (Stengler et al., 1964; Stern, 1979). The behavior of the hydrocarbon products indicates that PCHE was formed by elimination of ammonia from PCHA, a reaction known to occur with aliphatic amines. In addition, PCHE was a reaction intermediate, converted to either PB or PCH. Formation of PB from HDN of DHQ must be attributed to dehydrogenation of PCHE, since no OPA was formed (Figure 9). It is unlikely that PB and PCH were reactive in the presence of the strongly adsorbed nitrogen compounds. Instead, PB and PCH were formed by parallel reactions, and their ratio was controlled primarily by kinetics rather than by thermodynamics. Reaction of each individual heterocyclic nitrogen compound resulted in the formation of all the other heterocyclics, confirming the predicted reversiblity of the initial ring saturation reactions. Only small quantities of quinoline were found in the products from HDN of BzTHQ or DHQ at 375 O C and 7.0 MPa, since quinoline formation is thermodynamically unfavorable at these conditions (Cocchetto and Satterfield, 1980). The heterocyclics can equilibrate under HDN conditions only if the hydrogenation/dehydrogenation reactions are fast relative to the PyTHQ and DHQ hydrogenolysis reactions. In this study, the initial ring saturation reactions approached equilibrium at some conditions, but in general their behavior was complex and cannot be explained by either kinetic or thermodynamic considerations alone. Selectivity Changes during Catalyst Deactivation. Significant deactivation of the presulfided NiMo/A1203 catalyst was observed during its first 400 h of use, but the relative changes in hydrogenation and hydrogenolysis activities of the catalyst are not readily apparent from the changes in product distribution that occurred. However, from the quinoline HDN product distribution and a knowledge of the reaction network, one can calculate the amounts of hydrogen consumed in saturation reactions in contrast to hydrogenolysis reactions, as measures of the hydrogenation and hydrogenolysis activities of the catalyst. During catalyst deactivation, hydrogenolysis activity decreased by nearly 60% while there was only a 20% loss in hydrogenation activity (Figure 10). This markedly different behavior of the hydrogenation and hydrogenolysis
I I
MOA mum Hydroqenolys s of C - N Bonds
L I00
L
l
-
I
1
- O
i
,
200 CATALYST
300 TIME
500
400 ON S T R E A M
530
Pours
Figure 10. Hydrogen consumption in quinoline HDN during cablyst deactivation; severe standard conditions.
activities is experimental evidency, albeit indirect, that these functionalities are associated with different catalyst sites. The deactivation of the catalyst was most likely caused by the accumulation of carbonaceous deposita (coke) on the surface. A theory advanced by Beuther and Larson (1965) to explain the deactivation of dual function metal/ acidic support hydrocracking catalysts seems to be applicable to the NiMo/A1203catalyst used here as well. They and others have presented evidence to show that one of the roles of the metal is to keep acidic sites active by hydrogenating and thus removing coke precursors. One may postulate that here similarly during deactivation, coking eventually occurs on all hydrogenolysis sites not in the vicinity of a "protective" hydrogenation site, which through its catalytic activity is much less susceptible to coking. Quinoline Hydrodenitrogenation Kinetics. In quinoline HDN, hydrocarbons and ammonia were formed from the reaction intermediates, DHQ and OPA (PCHA presumably reacted as rapidly as it was formed). As a result, the percent denitrogenation vs. contact time curves for quinoline, PyTHQ, or BzTHQ feed are S-shaped; the denitrogenation rate (slope of the curve) was zero initially, increased with contact time as the concentrations of DHQ and OPA increased, and eventually decreased at long contact times as the concentrations of intermediates, particularly DHQ, decreased and the ammonia concentration became significant (Figures 4, 5, 7, and 8). For HDN of DHQ or OPA, the percent denitrogenation versus contact time curve is concave downward; the denitrogenation rate was highest initially and decreased with contact time as the concentration of DHQ or OPA decreased (Figures 7 and 9). As can be seen from Figure 7, nitrogen removal from DHQ was substantially easier than from the other heterocyclics, since DHQ is already completely saturated and its hydrogenolysis rate was relatively high. For PyTHQ HDN at 375 "Cand 7.0 MPa, the primary reaction pathway for nitrogen removal was through the DHQ (rather than the OPA) intermediate, since hydrogenation of PyTHQ to DHQ was much faster than hydrogenolysis of PyTHQ to OPA (Figure 5). Comparison of Figures 5 and 8 reveals that the rate of hydrogenation of BzTHQ to DHQ was higher than the PyTHQ to DHQ hydrogenation rate. Both of these observations are in agreement with the relative values of rate constants published by Shih et al. (1977) for liquid-phase studies on a NiMo/A1203catalyst at 342 "C and 3.55 MPa. These factors account for the higher rates of nitrogen removal from BzTHQ than from
PyTHQ (Figure 7). For either quinoline or PyTHQ feed a t the same reaction conditions, the same product distribution and percent denitrogenation were observed, as the Q/PyTHQ equilibrium was established very rapidly (compare Figures 4 and 5). This does not necessarily imply, however, that the equilibrium between quinoline and PyTHQ was maintained at longer contact times. At 375 "C and 7.0 MPa, nitrogen removal from OPA occurred much more readily than from any of the heterocyclics, when each compound was studied individually (Figure 7). The dominant hydrocarbon product from HDN of OPA was always PCH, indicating that hydrogenation of OPA (to PCHA) was much faster than its direct hydrogenolysis to PB. Resonance stabilization of the C-N bond in OPA is undoubtedly responsible for the relative difficulty of direct hydrogenolysis. Only low concentrations of OPA were observed in the products from reaction of each heterocyclic, as a result of the relatively slow hydrogenolysis of PyTHQ to OPA. However, in light of the high reactivity of OPA in the absence of the heterocyclics, the survival of even low concentrations of OPA at long contact times is somewhat surprising. Apparently, in the presence of sufficient concentrations of the heterocyclics, the active catalyst sites were not very accessible to OPA due to competitive adsorption effects. Our results (see later) show that PyTHQ and DHQ, being highly basic, actually are much more strongly adsorbed than OPA and therefore in their presence the rate of HDN of OPA is markedly reduced below the rate in their absence. This can also explain why, of the individual reaction steps, the lowest first-order rate constant reported by Shih et al. (1977) was for reaction of OPA. In their study, the kinetics of OPA denitrogenation were derived from quinoline HDN data only, and the first-order kinetic model employed did not allow for potential inequalities among the adsorptivities of the different nitrogen compounds present. Denitrogenation of quinoline occurred primarily through the DHQ intermediate, and PCH was always the major hydrocarbon product. Thus the presulfided NiMo/A1203 catalyst, a widely used commercial hydrotreating catalyst, exhibited little selectivity for the reaction pathway of minimum hydrogen consumption. At lower temperatures the heteroring in quinoline was selectively hydrogenated, at least initially (forming primarily PyTHQ instead of BzTHQ; Figure 4), but the catalyst did not possess sufficient hydrogenolysis activity to readily convert the PyTHQ to OPA and then to PB and ammonia, so significant saturation of the aromatic ring also occurred. At 420 "C, the hydrogenation rates of quinoline to PyTHQ and to BzTHQ were comparable, and DHQ was formed to a significant extent from BzTHQ as well as from PyTHQ Figures 2 and 3). The relatively low concentrations of DHQ in the products from quinoline HDN at 420 "C were due to thermodynamic, rather than kinetic, limitations. In fact, the amounts of hydrocarbons formed via the OPA and DHQ intermediates were comparable only at 420 "C and 3.55 MPa where the concentrations of OPA were significantly higher than the equilibrium-limited DHQ concentrations (Figure 2). Behavior of the Quinoline/Py-Tetrahydroquinoline Product Ratio. In Figure 11 the relative amounts of quinoline and PyTHQ in the products from quinoline HDN are shown as a function of reaction conditions and are compared with the corresponding equilibrium ratios. The departure of the Q/PyTHQ product ratio from equilibrium at longer contact times is unexpected but was not caused by an experimental artifact; the same results were obtained starting with PyTHQ or with quinoline.
-
w---
- E q u l l ~ b r t u m a t 330'C
0
H,
I-
>. a
P a r t , o l P r e s s u r e 3.53 M P o
I
0
:I
j
I
I
I
1
I
I
;
i?: ; : ;u; - ; - - -
4
$
H
P a r t at
~ r p s s u r p @f$*2H,
6.98 M P a
20
I-
>.
(Q'
0
100 WIF-
200 ,hr
300
( PyTHO)
400 WO 600 g cotatys1 ig-mol Q
700 8 0 0
Figure 11. Behavior of Q/PyTHQ product ratio in quinoline HDN (13.3 kPa Q).
The results can be interpreted in terms of competitive adsorption effects, postulating that quinoline is more weakly adsorbed than PyTHQ and a t least some of ita reaction products (such as DHQ or ammonia). The active catalyst sites, for hydrogenation as well as hydrogenolysis, are most likely acidic in nature. Adsorption of nitrogen compounds is believed to occur through interaction of the basic nitrogen group with acidic sites on the catalyst, so the more basic nitrogen compounds are likely to be more strongly adsorbed. In quinoline HDN, the PyTHQ and DHQ reaction intermediates are the most basic species. Thus,after initial equilibrium with PyTHQ, quinoline was prevented from adsorbing and reacting on the catalyst until the concentrations of PyTHQ and DHQ had decreased significantly. The more dramatic departure of the Q/PyTHQ ratio from equilibrium at lower temperatures (Figure 11) most likely reflects a smaller contribution from noncatalytic reaction, as we show here above. In our own previous study in similar equipment (Satterfield et al., 1978), quinoline and PyTHQ were reported to be in equilibrium a t all reaction conditions. However, only relatively short contact times were studied except a t 420 "C and 7.0 MPa where the Q/PyTHQ ratio remained constant over superficial contact times from 2.1 to 8.1 s (equivalent to about 65 to 250 h g of cat./g-mol of Q). At these reaction conditions, quinoline and PyTHQ were observed to be in equilibrium in the present study as well (Figure 11). Kinetic Modeling. Consideration of adsorption phenomena in quinoline HDN suggests that the catalytic reaction rates might best be described by Langmuir-Hinshelwood kinetic expressions, allowing for strong competitive adsorption of the nitrogen compounds. It is assumed that hydrogen and the nitrogen compounds adsorb on different catalyst sites, as suggested by adsorption measurements and other reaction studies (Sonnemans et d., 1973). For OPA feed, the appropriate Langmuir-Hinshelwood rate expression is
where rj is the net rate of formation of j , Kjand Pi are the
60
Ind. Eng. Chem. Process Des. Dev., Vol. 20,No. 1, 1981 1 c; 330
,i 1
-
I
Feed Portlol
L
-c
pres sur^
40.95
A 13.3 k Pa O P A
1
i
0
Y
4
0
*.
z4 0
Y m
$ Y
A
K,,H3/Kopa.325
..
$98
l
’5r I
1395
J090
-48G
;
i
;
w
2-
U
\\
\
t
\
g
\
reed P a r t , o l P r e s s u r e
~ 0 6 C
13.3 k P a O P A 26.7 k P a O P A
-04C
I
-1320
00
0
100
WlFoPAc,h r 5
200
13I
3^.‘
coialyst l g - m o l OPA
140
Figure 12. Comparison of Langmuir-Hinshelwood kinetic models for o-propylaniline denitrogenation at 375 “C and 7.0 MPa.
adsorption equilibrium constant and partial pressure of j , and kl’, which varies with both temperature and hydrogen pressure, is termed the pseudo-rate constant for OPA denitrogenation. The “1”in the denominator of the above expression is assumed to be negligible, and since P o p A + P N H 3 = POpA,O, eq 1 becomes -TOPA
=
llPOPA POPA
+ (KNH3/KOPA)(POPA,0
- POPA)
(2)
Note that if K N H J K O p A = 1 (equal adsorptivities of OPA and ammonia), a pseudo-first order rate expression results, while if KNH$KopA = 0 (negligible adsorption of ammonia), zero-order kientics are obtained. Substitution of eq 2 into the usual equation for plug flow reaction and integration gives
where X o p A is the fractional conversion (disappearance) of OPA, equivalent to the fractional denitrogenation. This equation can be tested against the experimental data by plotting the right-hand side as a function of W/FOpA,O, at constant temperature and hydrogen pressure. The “best” value of K N H J K O p A should result in a linear correlation which passes through the origin; the pseudo-rate constant is determined from the slope of this line. For OPA conversions greater than about 80%, pseudo-first-order kinetics predicts lower conversions than were actually observed (Figure 12), suggesting that the ammonia product inhibited the denitrogenation rate less than the OPA reactant did (KNH,/KOPA C 1). The “best” K N H S / K O p A value for correlating all of the OPA HDN data is 0.25, implying that OPA adsorbed about four times as strongly as ammonia on the active sites. This kinetic model correctly predicts that percent denitrogenation is independent of initial OPA partial pressure, at constant temperature, hydrogen pressure, and W /Fopk0 (eq 3; Figures 6 and 12). Note too, that the rate of denitrogenation of OPA is between zero and first order in OPA
I 1.50
I 1.60
1 170
10001 T I K - l
Figure 13. Arrhenius plot of the pseudo-rate constant for opropylaniline denitrogenation.
partial pressure; the reaction order varies with the relative concentrations of OPA and ammonia (eq 1 and 2). The Arrhenius plot of the pseudo-rate constants a t 7.0 MPa is linear (Figure 13) and the activation energy for denitrogenation of OPA is 79 kJ/g-mol (19 kcal/g-mol). Knowledge of the OPA denitrogenation kinetics permits the kinetics of PyTHQ hydrogenolysis (to OPA) and DHQ hydrogenolysis (to hydrocarbons and ammonia) to be readily determined from HDN data for any of the heterocyclics. Equation 1 for OPA denitrogenation must, of course, be modified to allow for competitive adsorption of the heterocyclics as well as of OPA and ammonia. The secondary amines PyTHQ and DHQ are assumed to adsorb equally strongly, with an adsorption equilibrium constant KsA. Qual adsorptivities of the aromatic amines (Q, BzTHQ, and OPA) are likewise assumed, characterized by the adsorption constant KAA. In the presence of heterocyclics, OPA is formed by hydrogenolysis of PyTHQ but is also converted to hydrocarbons and ammonia, so the net rate of formation of OPA is given by
Here, k i is the pseudo-rate constant for hydrogenolysis of PyTHQ to OPA, and the “1”in the denominator has again been neglected. This equation is more conveniently expressed in terms of dimensionless partial pressures, Yj (= P j / P i , o )
For K N H J K A A = 0.25 and an assumed KSA/KAA value, k i is derived by numerical integration of the above equation at constant temperature and hydrogen pressure, using Yj vs. W / F i , o data provided by the HDN product distributions. Quality of fit can be assessed from the degree of linearity of the Arrhenius plot, and by checking calculated k i values for constancy as the integrations are
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 1, 1981 61 1 :c
1. * C 420
r
Total
-A 0
c 0
“ m
S
.-
‘
0
Total Pressure
F a a d Part101
Pressure 3.55MPa 0 3.55MPo A 7.0 MPC
Pressure 13.3kPoa 26.7kP00 13.3 kPo 0
A 7.0
26.7 k P a 0
MPc
3.55 M P o 0 3.55MPo 0 3.55MPo
m
S
”\
\
163k\
,
/
2
1 . 6
1
A 7.0 MPa A 7 . 0 MPc V 7 . 0 MPa MPo
Feed P a r t , a l Press d r e
a
I 3 3 k Po 26.7 k Po 0 13.3 k P c BzTHQ 13.3kPo 0 26.7kPaQ 133kPoEzTH0 13.3 k P a DHO
1
OI
4
wl
a
-5
1
1
I
I O
1.40
1.50
1.60
\‘ J 1.70
IOOOIT , K-I
1641 1.40
I 1.50
I 1.60
1.70
10001 1 , K-l
Figure 14. Arrhenius plot of the pseudo-rate constant for hydrogenolysis of Py-tetrahydroquinoline to o-propylaniline.
Figure 15. Arrhenius plot of the pseudo-rate constant for hydrogenolysis of decahydroquinoline to hydrocarbons and ammonia.
carried out to higher W/Fi,o.The pseudo-rate constant for DHQ hydrogenolysis is derived in a similar fashion, recognizing that hydrocarbons and ammonia are formed from both OPA and DHQ. Pseudo-first-order kinetics (KNHs/KAA = 1,KSA/KAA = 1)results in very poor correlation of the experimental data, which is not surprising in view of the strong evidence for preferential adsorption of PyTHQ and DHQ. A KS*/KAAvalue of about 6 (with Km/KOpA = 0.25) gives the “best” correlation of the data; thus, the kinetic modeling results are consistent with the qualitative expectations discussed earlier. Arrhenius plots of the pseudo-rate constants for hydrogenolysis of PyTHQ and DHQ are shown in Figures 14 and 15. Note that the values of the pseudo-rate constants for hydrogenolysis of DHQ are essentially the same regardless of the starting heterocyclic nitrogen compound (Figure 15), further demonstrating the adequacy of the kinetic model. The activation energies of the hydrogenolysis reactions are comparable-an average of 155 kJ/gmol (37 kcal/g-mol) for hydrogenolysis of PyTHQ, and about 138 kJ/g-mol(33 kcal/g-mol) for DHQ hydrogenolysis. These activation energies vary somewhat with hydrogen pressure, so the temperature and hydrogen pressure dependencies of the pseudo-rate constants for the hydrogenolysis reactions are to some extent interrelated. This is not surprising, since adsorbed hydrogen was presumably involved in the catalytic reactions, and the fraction of available sites actually occupied by hydrogen depends on temperature as well as hydrogen pressure. The pseudo-rate constants for DHQ hydrogenolysis are about an order of magnitude larger than those for hydrogenolysis of PyTHQ (compare Figure 14 and 15), while those for OPA denitrogenation are the highest of all. Note, however, that the kinetic model correctly predicts a substantially lower OPA denitrogenation rate when significant concentrations of heterocyclics, particularly PyTHQ and DHQ, are present. The activation energies for the two hydrogenolysis reactions are closely comparable to those reported by Shih et all. (1977) (37 and 33 kcal/g-mol here vs. 35 and 31 kcal/g-mol) and are much higher than the activation energy for the OPA denitrogenation reaction.
This is consistent with the conclusion here, that nitrogen removal from OPA occurred primarily by hydrogenation of OPA (to PCHA) rather than by direct hydrogenolysis to PB. It is noteworthy that the Langmuir-Hinshelwood reaction rate expressions used for kinetic modeling here are dependent on the relative concentrations of nitrogen compounds (Y,),but not on the absolute concentrations (Pior Pi,o).This results from the assumption that the “1” in the denominator of the rate expressions is negligible, which is equivalent to postulating that the catalyst surface was always saturated with nitrogen compounds, even at the lower reactant partial pressures. The fact that HDN product distributions were virtually independent of initial reactant partial pressure, at constant W/Fi,o,temperature, and hydrogen pressure, indicates that this assumption is justified. Summary and Conclusions 1. The reactions among quinoline and its hydrogenated heterocyclic derivatives are all reversible over a wide range of HDN conditions. 2. Quinoline (Q) reaches an equilibrium with respect to Py-tetrahydroquinoline (PyTHQ) rapidly in the presence of a NiMo/A1203 catalyst but with longer contact times the ratio PyTHQ/Q can decrease below the equilibrium value, due to the much greater adsorptivity of PyTHQ which reduces access of quinoline to the catalyst. 3. Noncatalytic conversion of quinoline to PflHQ can be significant at higher temperatures and longer contact times, a factor that must be considered in modeling the catalytic HDN reactions of quinoline. 4. Several researchers have previously concluded that the rate-limiting step in the HDN of quinolines or indoles on a NiMo/A1203 or CoMo/A1203catalyst is conversion of the alkyl aniline intermediate, based on studies with the initial heterocyclic compound only. However, we show that by itself, o-propylaniline is more reactive than any of the heterocyclic compounds present here, but is much less reactive in a mixture with them. This is caused by the much greater adsorptivity of the secondary amine intermediates (Py-tetrahydroquinoline and decahydro-
82
Ind. Eng. Chem. Process Des. Dev. 1981, 20,62-68
quinoline) relative to the adsorptivities of other species present. 5. The principal pathway for hydrodenitrogenation of quinoline under industrial processing conditions is via decahydroquinoline. Propylcyclohexene was identified as a significant intermediate, but the principal hydrocarbon product is propylcyclohexane. 6. As a NiMo/A1203catalyst initially ages, the ratio of its activity for hydrogenolysis to that for hydrogenation decreases, attributed to preferential coking on hydrogenolysis sites. 7. The kinetics of denitrogenation of o-propylaniline (OPA), of hydrogenolysis of Py-tetrahydroquinoline (PyTHQ) and of decahydroquinoline (DHQ) have been modeled in terms of Langmuir-Hinshelwood expressions allowing for the markedly different adsorptivities of the various N-containing species present. Acknowledgment The work was supported in part by funds from the US. Environmental Protection Agency. We appreciate helpful
suggestions from Michael Modell during the early portion of the study. Literature Cited AbouUjheit, A. K.; AWou, I. K. J. Inst. Pet. 1979, 59, 188. Beuther, H.; Larson. 0. A. Ind. Eng. Chem. process Des. D e v . 1985, 4 , 177. Cocchetto, J. F.; Satterfield, C. N. Ind. Eng. Chem. process Des. Dev. 1976, 15, 272. Cocchetto, J. F.; Setterfield, C. N. Ind. €ng. Chem. Process Des. Dev. 1981, accompanying paper in this issue. Flinn. R. A.: Larson. 0. A.: Beuther. H. M & h x b o n Process. Pet. Refiner 1989, 42(9), 129. Koros, R. M.; Bank, S.; Hofmann, J. E.; Kay. M. 1. Am. Chem. Soc. Dlv. Pet. Chem. Prepr. 1967, 12(4), 5 1 8 5 . Setterfield, C. N.; Cocchetto, J. F. A I C M J . 1975, 27, 1107. SatterRekl, C. N.; Modell, M.; Hltes, R. A.; Declerck, C. J. Ind. Eng. Chem. Rocess - - - -Des. . - -h -v . 1078. 17. 141 Shih, S. S.; Katzer, J. RiKwart, H.; Stiles. A. B. Am. Chem. Sac. Dlv. Pet. Chem. Rem. 1977. 22. 919. Sonnemans, J ;: van den B&g, 5. H.; Mars, P. J. Catel. 1979, 31, 220. Stengler, W.; Welker, J.; Lelbnb, E. Frelberg. Forschungsh A 1964, 329, 51. Stern,E. W. J . Catel. 1979, 57, 390.
__
Receiued far review February 19,1980 Accepted July 21, 1980
Effect of Hydrogen Sulfide on the Catalytic Hydrodenitrogenation of Qulnoline Charles N. Satterfield' and Selahattln Giiltekln' Department of Chemical Engineering, Massachusetts Institute of Technology, Cambrklge, Massachusetts 02 739
This extension of an earlier study on the reaction network of quinoline hydrodenhogenation(HDN) showed that H2S has a slight inhibiting effect on the intermediate hydrogenation steps but a marked accelerating effect on the overall HDN rate. Rate constants and activation energies for a set of standard conditions are reported for most of the intermediate reaction steps, in the presence and absence of H2S. On a commercial NIMo/A1203catalyst a primary limiting factor on the overall hydrodenhogenationrate is the very strong adsorption onto the catalyst of secondary amines formed as reaction intermediates.
In a recent paper (Satterfield and Cocchetto, 1981) we have reported a study of the reaction network and kinetics of the vapor-phase catalytic hydrodenitrogenation (HDN) of quinoline, on a commercial NiMo/A1203 catalyst and under conditions representative of industrial practice. Quinoline HDN is a good model for the HDN of sixmember ring nitrogen heterocyclic compounds found in middle distillate fuels having a high organonitrogen content. Under industrial conditions organosulfur compounds are invariably present together with the organonitrogen compounds and the former are converted by hydrodesulfurization reactions (HDS) to H2S during the HDN reactions. In previous studies with thiophene and pyridine as model compounds we have shown that the HDS and HDN reactions interact with each other in a complex manner. Under some reaction conditions thiophene enhances the overall HDN of pyridine and it has been shown by our work (Mayer, 1974; Satterfield et al., 1975; Satterfield et al., 1980) and that of Goudriaan et al., (1973), and Goudriaan (1974) that this is caused by the H2S University of Istanbul, Istanbul, Turkey. 0196-4305/81/1120-0062$01.00/0
formed by the HDS of thiophene. In the case of pyridine, the first intermediate product formed is piperidine, and we showed (Satterfield et al., 1980) that the presence of thiophene caused inhibition of pyridine hydrogenation but enhancement of the hydrogenolysis of piperidine, so the net effect can be an enhancement of the overall nitrogen removal rate. Pyridine and other heterocyclic nitrogen compounds inhibit HDS in general, as has been shown in our earlier studies and by others. The present study was an extension of our earlier study of the HDN of quinoline (Satterfield and Cocchetto, 1981), to determine the effects of H2S. The steps believed to be significant in the reaction network in the presence of H2S are shown in Figure 1,including the acronyms used for the various compounds discussed. Previous studies on quinoline HDN and guidance t~ the earlier literature on HDN reactions in general are given in the paper by Satterfield and Cocchetto (1981). Experimental Apparatus and Procedure The apparatus has been previously described. As with the previous study, a single charge of commercial NiMo/A1203 hydrotreating catalyst was used for all the runs reported here (American Cyanamid AERO HDS-3A, 0 1980 Amerlcan Chemlcal Society
