Znd. Eng. Chem. Res. 1988,27, 1629-1639
1629
where outliers are expected.
Literature Cited
Acknowledgment The authors thank the Computer Centre, National University of Singapore, for providing computing facilities. G.P.R. is grateful to A. Kalantar of the University of Alberta for many valuable discussions. Nomenclature a = parameter in linear eq 3 A = frequency factor in the Arrhenius expression, s-l b = parameter in linear eq 3 ei = experimental error in the rate coefficient, k i (eq 1) e{ = experimental error in In ki (eq 2) e p = normally distributed random variable with mean zero and unit standard deviation E = activation energy in the Arrhenius expression, cal mol-' E = mean estimated value of activation energy, cal mol-l k i = rate coefficient in the ith experiment, s-' n = number of experiments in a set of data R = universal gas constant, cal mol-' K-' R1 = root mean square error expressed as a percentage of true parameter value R2 = R1 per unit (percent) error in rate coefficients at the highest temperature Ti= temperature in the ith experiment, K W i= weight assigned to ki in weighted nonlinear regression (eq 5 ) x i = independent variable in the ith experiment (eq 3) yi = independent variable in the ith experiment (eq 3)
Atkins, G. L. Comput. Biol. Med. 1982, 12, 201-215. Bevington, P. R. Data Reduction and Error Analysis for the Physical Sciences; McGraw-Hill: New York, 1969; Chapter 1. Box, G. E. P.; Hill, W. J. Technometrics 1974, 16, 385-389. Box, G. E. P.; Hunter, W. G.; Hunter, J. S. Statistics for Experimenters; Wiley: New York, 1978; pp 231-241. Cornish-Bowden, A. Fundamentals of Enzyme Kinetics; Butterworths: London, 1979; Chapter 2. Endrenyi, L.; Tang, H. Comput. Biomed. Res. 1980, 13, 430-436. Hines, W. H.; Montgomery, D. C. Probability and Statistics in Engineering and Management Science; Wiley: New York, 1980; Chapter 9. Hollander, M.; Wolfe, D. A. Nonparametric Statistical Methods; Wiley: New York, 1973; Chapter 9. ZMSL Reference Manual, 9th ed.; IMSL: Houston, TX, 1982. Kalantar, A. H. J . Phys. Chem. 1986, 90, 6301-6303. Kalantar, A. H. Chem. Eng. J. 1987a, 34, 159-164. Kalantar, A. H. Comput. Biol. Med. 1987b, 17, 209-219. Kreyszig, E. Advanced Engineering Mathematics; Wiley: New York, 1979; Chapter 20. Nimmo, I. A.; Atkins, G. L. Anal. Biochem. 1979,94, 270-273. Pritchard, D. J.; Downie, J.; Bacon, D. W. Technometrics 1977,19, 227-236. Rangaiah, G. P. Chem. Eng. J . 1984,29, 159-166. Rousseeuw, P. J.; Leroy, A. M. Robust Regression and Outlier Detection; Wiley: New York, 1987; Chapter 5. Sen, P. K. J . Am. Stat. Assn. 1968,63, 1379-1389. Topping, J. Errors of Observation and their Treatment; Chapman and Hall: London, 1972; Chapter 3. Walpole, R. E.; Myers, R. H. Probability and Statistics for Engineers and Scientists; Collier Macmillan: London, 1985; Chapter 8.
Greek Symbol
4 = parameter in the power transformation weighting method Received for review August 14, 1987 Revised manuscript received March 30, 1988 Accepted May 4, 1988
Superscript * = true value
Hydrodemetalation of Vanadium and Nickel Porphyrins over Sulfided CoMo/A1203 Catalyst Hung-Jean Chen and F. E. Massoth* Department of Fuels Engineering, University of Utah, Salt Lake City, U t a h 84112
Hydrodemetalation (HDM) of vanadium porphyrin (VP) and nickel porphyrin (NP) model compounds with a sulfided CoMo/A1203 catalyst was performed in a batch stirred autoclave a t several temperatures, hydrogen pressures, and initial porphyrin concentrations. A hydrogenated intermediate leading to deposited metal was found for both reactants. The time course of the reaction followed pseudo first order in reactant concentrations above 350 "C, but followed lower order a t lower temperatures. Runs a t different initial concentrations showed that the reaction was inhibited by adsorption of reactant and products. HDM rates increased with temperature and hydrogen pressure and were very low without catalyst or hydrogen present. An apparent activation energy of 24 kcal/mol for the overall disappearance was obtained for both reactants. Kinetic analysis of the HDM of nickel porphyrin showed, in addition to a pathway through the hydrogenated intermediate, an apparent direct pathway to hydrocarbon products. The latter was interpreted in terms of a direct conversion of adsorbed NP to products in a single adsorption step, without desorption of hydrogenated intermediate. Evidence was obtained for a change in mechanism above 350 "C. Catalysts consisting of cobalt molybdena supported on alumina have been successfully used in the petroleum refining industries for many years for hydrodesulfurization of light crudes. Because of higher asphaltenic and metals content in petroleum residuum and heavy crudes, the application has been less successful due to more rapid catalyst deactivation and shorter catalyst active lifetime. During hydroprocessing of heavy residuum, organometallic 0888-5885/88/2627-1629$01.50/0
containing (mainly vanadium and nickel) compounds in the feed leave a deposit of metal sulfide on the catalyst through a catalytic hydrodemetalation (HDM) reaction. Also, multiaromatic compounds (asphaltenes) tend to produce carbonaceous deposits (coke) on the catalyst. Metal and coke deposits on the catalyst can cause loss of reaction sites as well as a change in the pore structure, giving a corresponding loss in catalyst activity. 0 1988 American Chemical Society
1630 Ind. Eng. Chem. Res., Vol. 27, No. 9, 1988
Table I. HDM Run Conditions' VP concn, ppm pressure, atm
temp, "C
NP
concn, ppm
pressure, atm
temp, "C
38.4 38.8 39.5 37.4 38.3 40.1 39.8 38.6 30.7 21.3 38.8 39.5 39.5 39.5
275 298 321 340 349 357 368 379 358 356 363 320 320 320
5 6 7 8 9 10 11 12 13 14 15 16
36.4 35.0 36.0 34.6 34.8 34.6 34.4 34.2 35.2 33.8 26.5 20.1
69.0 69.0 69.0 69.0 69.0 69.0 69.0 69.0 89.4 41.8 69.0 69.0
275 301 320 338 357 380 366 350 358 356 357 359
5 6 7 8 9 10 11 12 13 14 15 16 17 18
69.0 69.0 69.0 69.0 69.0 69.0 69.0 69.0 69.0 69.0 69.0 103.0 35.0 14.6
0.6-0.7 g of catalyst in 400 g of solution.
Heavy crudes and resids can contain relatively high concentrations of metals, sometimes in excess of lo00 ppm (Yen, 1975; Speight, 1981). Among the trace metals, with some exceptions, nickel and vanadium are the most abundant metals in petroleum (Yen, 1975). van Ginneken et al. (1975) have determined that in Middle East residuum the concentration of vanadium is in the range of 50-100 ppm and nickel is in the range of 10-30 ppm. Filby (1975) reported that in two California crudes, which have relatively high concentrations of nickel and low concentrations of vanadium, nickel is present in porphyrin and nonporphyrin forms. Nickel porphyrin was found in the low molecular weight fractions of both asphaltenes and resins, the percentage of nickel as non-porphyrin Ni increasing with increasing molecular weight of the fraction. Sugihara et al. (1970) obtained similar results for nickel and vanadium porphyrins from Boscan asphaltene and concluded that metalloporphyrins (V and Ni) are concentrated in the high molecular weight fraction of the asphaltenes. A number of investigators have studied the kinetics of the metal removal rate over commercial CoMo/Alz03 catalysts with various crudes and resid feedstocks. The kinetic order of hydrodemetalation with respect to the metal concentration has been reported as first order (Larson and Beuther, 1966;Riley, 1978;Pazos et al., 1981) and second order (Mosby et al., 1973; Oleck and Sherry, 1977). Two hypotheses can be used to explain the disagreement in the HDM reaction orders reported in the literature. One of these is that the kinetic order of the HDM reaction is affected by the presence of many different sulfur, nitrogen, metal, and asphaltenic compounds in crudes and resids. The other explanation is due to the different reactivity of the various metal compounds contained in crudes and resids, which have different porphyrin-type tQ non-porphyrin-type metal ratios, the former being generally more reactive than the latter (Silbernagel and Riley, 1980). The skeleton of a porphyrin is a porphine, which consists of four pyrrole groups linked a t the a-positions by four methine groups. Baker and Palmer (1978) have identified the structure of major porphyrin compounds. Three types of porphyrin compounds have been found in petroleum: (1)deoxophylloerythroetioporphyrin (DPEP) type; (2) etio type; and (3) rhodo type (Yen, 1975). A study of the HDM of model nickel porphyrins under typical processing conditions at 350 "C with a CoMo/Al,03 catalyst (Weitcamp et al., 1984) showed by UV spectra that the products produced consisted of partially hydrogenated intermediates containing nickel, dipyrroles, and single-ring aromatics; no Ni-free porphyrins were detected. Rankel (1981) also reported the presence of polypyrroles as major reaction products and no metal-free porphyrin for reaction
of nickel and vanadium porphyrins over a sulfided CoMo/Alz03 catalyst. Hung and Wei (1980a,b) have studied the kinetics of HDM in a batch reactor system over a non-sulfided commercial CoMo/Alz03 catalyst, using pure nickel and vanadium porphyrins. They reported that the HDM reaction order was close to 0.5 but varied with the operating temperature. Agrawal and Wei (1984), in comparison studies in a flow reactor, observed first-order kinetics. Subsequent studies over sulfided CoMo/AlzOs catalysts by Morales et al. (1984) and Ware and Wei (1985b) also found first-order kinetics. The above studies also reported the presence of hydrogenated metal porphyrins (chlorins) as intermediates in the HDM reaction pathway. The object of the present investigation was to determine the kinetics of HDM of vanadium and nickel porphyrins over a sulfided hydrotreating catalyst, which had not been previously reported when this study was undertaken.
Experimental Section Catalysts. A CoMo/AlzOs catalyst (Topsoe A/S) containing 1.9% Co and 6.9% Mo was used. It was received as 0.079-cm extrudates (1/32 in.) and crushed and sieved to 0.021-0.042-cm (35-65-mesh) size for testing. The catalyst had a surface area of 194 m2/g and pore volume of 0.60 cm3/g. Materials. The model compounds used in the hydrodemetalation experiments were vanadium (IV) tetrakis(3-methylpheny1)porphyrin (VP) and nickel tetraphenylporphyrin (NP), purchased from Mid-Century. A white oil, Nujol (Plough Inc.), was employed as the solvent. Dissolution of the model compounds into Nujol was based on a method described by Hung and Wei (1980a). Because the solid compounds have low solubility at room temperature and air-prepared solutions may have different demetalation behavior from that of air-free solutions, heating under helium pressure was employed for preparation of the solutions, which resulted in complete solution of the porphyrins in 8 h. Dimethyl disulfide, a source of H2S, was added to the model compound solution for maintaining the presulfided catalyst in its sulfided form during reaction (Broderich et al., 1978). Helium, nitrogen, and a mixture of 10% H,S/H, were used as received. Hydrogen (99.9% purity) was further purified to remove traces of impurities, oxygen, and water. Procedure. The hydrodemetalation runs were performed in a stirred, 1-L, stainless steel autoclave reactor (AutoclaveEngineers, Inc.). Run conditions are presented in Table I. A catalyst loader unit, used to inject the catalyst into the reactor, was made of a 3.2-mm (l/s-in.) 0.d. tube with quick connectors at both ends, followed by a 12.7-mm (l/*-in.) ball valve. A ceramic filter, located at
Ind. Eng. Chem. Res., Vol. 27, No. 9, 1988 1631 the inlet of the sampling line, was used to prevent loss of catalyst during sampling. Further details are given elsewhere (Chen, 1986). A weighted amount of crushed catalyst (0.6-0.7 g) was sulfided in a separate, fixed-bed microreactor with 10% H2S/H2at 400 "C for 2 h. After cooling in He, the sulfided catalyst was removed from the reactor by dumping into about 4 mL of dimethyl disulfide and 5 mL of Nujol, without exposure to air, and then was quickly introduced into the catalyst loader. This presulfiding procedure results in an essentially completely sulfided state for the cobalt and molybdenum (DeBeer et al., 1976). About 400 g of model compound solution was charged into the autoclave reactor. After sealing, the reactor was purged with helium to remove the air and then was purged with hydrogen. The reactor was pressurized to about 5 atm of H2, the stirrer was operated at 500 rpm, and the heater was turned on. After the reactor reached the desired temperature, the catalyst loader was pressurized with H2 and the ball valve was opened to inject the catalyst into the reactor, after which the ball valve was closed. This injection procedure was repeated 3 times in order to ensure all the catalyst was injected into the reactor. The reactor was then pressurized to the final pressure. -The time required to inject the catalyst and pressurize the reactor was about 1.5 min. Zero reaction time was taken at the time that the catalyst was first injected. About 2-2.5 g of liquid sample was taken at various times via a sampling port. The change in concentration with time was monitored by a Beckman Model 25 UV spectrophotometer. Several solutions of different concentrations of each of the pure model compounds were employed to obtain calibration curves, from which the sample concentrations were determined. The experiments were stopped after the solution was depleted of solute. Duplicate runs gave results within 5%. To account for the effect of decreasing reactor solution volume due to sampling, a corrected reaction "space time", 7,was used, viz., i 7
=
wcC(ti- ti-1) / Ws,i 1
(1)
where W , is the catalyst weight, ti is the sampling time from the start of run, and W,,i is the weight of solution remaining at ti. During the course of the HDM runs, the presence of an intermediate compound was detected in the UV spectra of liquid samples from both VP and NP runs. Attempts were made to isolate the intermediate following the method described by Agrawal and Wei (1984). The product solution, diluted with xylene, was poured into the top of a 2.5-cm vertical glass tube packed with about 8 cm of 80200-mesh activated alumina. Poor separation was achieved with the VP product solution. However, the NP product solution separated in the column into a light green section (intermediate) and a red section (NP) which only partially overlapped. The alumina packing in the green section was taken out of the column, washed with chloroform, and evacuated under vacuum to evaporate the chloroform. The residual solid was green. The separated intermediate was dissolved in Nujol by using the same preparation method as for the NP. The absorption coefficient for the intermediate was then determined and used to measure the concentration of Ni intermediate during reaction time.
Results Overall Disappearance of Vanadium Porphyrin (VP). Figure 1 displays the UV-visible spectra of a sample
w
0
z
4mK
sm 4:
400
450
500
550
WAVELENGTH
600
650
700
nm
Figure 1. UV-visible absorption spectra of pure VP (-) sample collected during an HDM run (- - -).
and a
0.9 1
0.8
\ 0.7
0.6
w 0.5 4:
m a
0
14:
0.4
03
\
0.2
0.1
0 0
1
2
3
4
5
TIME, hr
Figure 2. Absorbance-time profiles for VP ( 0 )and intermediate (A)during an HDM run. Conditions: initial concentration, 40 ppm VP; pressure, 69 atm; temperature, 320 "C.
of pure VP, which shows an intensive absorption peak at a wavelength of 548 nm, and of a sample collected after 1-h reaction time during an HDM run at 320 "C and 69 atm. The latter sample shows a new peak at 633 nm. Figure 2 shows the absorbance-time profiles for this run. The intermediate increased to a maximum and then decreased during reaction time, indicating that it was an intermediate in the HDM reaction. It was not possible to calculate the absolute concentration of the intermediate since a pure sample of the intermediate could not be obtained to determine its absorption coefficient. However,
1632 Ind. Eng. Chem. Res., Vol. 27, No. 9, 1988 10
z 0 m
08
a W
>
50
06
a
F
> -I a
z
O4
I6
1:
a
1
i O O O / T , 1-1
a
.L!
Figure 4. Arrhenius plots for initial pseudo-first-order rate constants for VP (0) and N P (0) runs. Conditions: initial concentration, 40 ppm for VP, 35 ppm for NP; pressure, 69 atm.
02
0 (
1
I
I
I
I
I
I
1
2
3
4
5
6
7
T I M E , hr
Figure 3. Effect of H2pressure on VP reaction. Conditions: initial concentration, 40 ppm; temperature, 320 "C. Symbols: (A)103, (0) 69, ( 0 )35, (0) 14.6 atm.
2
10
Table 11. Dependence of Best Fit Kinetic Order on Temperature for VP Reaction" temp, "C best fit temp, "C best fit 275 298 321 340 349 a
0.1 0.2 0.5 0.4 0.6
357 363 368 379
0.9 1.0 0.9 1.0
Conditions: 39 ppm, initial concentration; 69 atm.
based on NP results (see below), the amount of intermediate appears to be small in comparison with the amount of VP in solution. Similar trends were found for the other HDM runs. Therefore, the HDM activities were subsequently analyzed as VP disappearance rates. Comparative runs a t 69 atm and 320 OC with and without catalyst showed that very little HDM (30
11
O35 ppm, 69 atm. b(ppm.g cat..h)/kg oil. cEstimated from eq 9.
hydrocarbons and ammonia. The deposited nickel rapidly sulfides from H2S generated from the dimethyl disulfide in the feed solution. Analysis showed that the reaction order in porphyrin was less than unity at lower temperatures, increasing to first order above about 350 "C (Table V). Runs at different concentrations showed the reaction to be inhibited by porphyrin (Table VI). Therefore, the demetalation rate can be expressed with an equation of the type ~ N P=
~NPKNPCNP/D
(3)
D = 1 + EKic,
IC
where rNPis the rate of disappearance of nickel porphyrin, kNp is the rate constant, cNP is the concentration, Ki and c, are the adsorption constant and concentration, respectively, of reactant and products, and D is the inhibition term. A preliminary analysis of the data will be made on the basis of a pseudo-first-order rate equation, which in integrated form becomes - In (1 - x) = k'7
(4)
kNpKNp/D
(5)
Data for reaction of N P at various temperatures are shown in Figure 9 in terms of first-order plots. Significant curvature in the plots a t lower temperatures is evident, while above 350 "C the plots follow essentially first order. The latter signifies a negligible change in D with conversion. A t lower temperatures, the major reaction products are dipyrroles. The inhibition term then becomes
D = 1 + KNPCNP + KHNPCHNP + KDPCDP
(6)
The adsorption constants for N P and HNP can be reasonably expected to be essentially identical, as partial hydrogenation of the metal porphyrin should have relatively little effect on its adsorption. Further, since 2 mol of DP is formed from each mole of NP reacted, the concentration of DP is given by cDp = 2(c0Np- cpy), where c o w is the starting concentration of NP and cpy is the sum of cNp and cHNP. Thus, eq 6 becomes
D =1
2K~pC'iqp
(KNp - 2 K ~ p ) C p y
(7)
In this equation, cpy decreases as conversion increases. If
KNp> 2KDP,then D will decrease with conversion, and according to eq 5, k'will increase with conversion. Since k'represents the slope of the plots of Figure 9, the upward curvature can be explained by adsorption of NP greater than 2 times DP. This proposition does not seem unreasonable in view of the larger polyaromatic network in NP than in DP, contributing to a stronger bonding to the catalyst surface.
4
oil
Figure 9. First-order plots of N P disappearance as a function of temperature. Conditions: 35 ppm, 69 atm.
In order to check the conformance of the data with the above analysis, eq 3 together with eq 7 was integrated to yield 7 = -a1 In (cNp/c0Np) t
where x is the conversion of NP. The first-order rate constant is then given by
k'=
2
2010 T ,h - g c o t / k g
"2[ CoNP
"1
=
- cNP -
1 + ~KDPCONP "2
l C H N P / C N P dCNP]
=
KNP
(8)
- 2KDP
~NPKNP ~NPKNP From concentration data for N P and HNP, the integral in the last term was evaluated and then eq 8 was solved by multilinear analysis to give the best values for a1and Due to the excy2. These data are listed in Table VII. pected strong adsorption at low temperatures, it may be presumed that 2 K ~ p c ' ~ >> p 1, and thus (9)
Values of KNp/KDp given in table VI1 show the adsorption of NP (and HNP) to be some 7-14 times greater than DP. Individual values of KNPand KDp could not be obtained from a single run because only two grouping constants (al and a2)are available from analysis. Nevertheless, the goodness of fit of the derived constants to the data can be checked by calculating 7 for the set of CY values. The results, given by the solid lines in Figure 9, show good agreement with the data. It is noted that the ratio of KNp/KDP decreases with increasing temperature. A plot of In (KNp/KDp) versus 1/T gave a positive slope of 10 f 4 kcal/mol. This indicates that NP has a higher heat of adsorption than DP by some 10 kcal/mol. For runs above 350 "C, the data of Figure 9 show good pseudo-first-order plots. It may be expected that the K values will decrease with increasing temperature so that at high temperature the inhibition term approaches unity; i.e., the K values become sufficiently low as to be negligible in the D term. This may be true for the highest temperature investigated. However, it was shown earlier (Table
Ind. Eng. Chem. Res., Vol. 27, NO. 9, 1988 1635 U
4
3 25
,
,
050
0 75
2 C8
0
I-x, Figure 10. Selectivity plot of X, versus 1- XAfor run NP-15. Fit to eq 10 with k3 (-1 and without k3 (- -1.
-
IV) that appreciable adsorption inhibition was still present at 357 "C, viz., KNp = 0.14, and for CON^ = 35, KNpcONp = 4.9, so that D = 5.9 > 1. Therefore, a more likely explanation is that, in view of the temperature response of KNP/KDPin Table VII, at higher temperatures, KNpdecreases faster than KDpand thus KNp approaches 2KDp. Hence, D becomes essentially independent of conversion according to eq 7, with the result that the data follow pseudo first order via eq 4 and 5. 2. Reaction Network for Intermediate. Preliminary kinetic analyses of a number of runs showed that the direct series reaction of Scheme I did not adequately fit the selectivity data. Consequently, the kinetics were modeled according to the reaction network Scheme I1 HNP
NP (A)
Jk2
T ,h - g c o t / k g o i l
Figure 11. Data fit t o a and 6 values of Table VI1 for run NP-6. Table VIII. Constants for Run NP-15 k3 = 1.74 kg oil/(g cat..h) kNp = 16.4 kg oil/(g cat..h) k, = 0.56 kg oil/(g cat..h) K N P= 0.14 ppm-' k2 = 5.91 kg oil/(g cat..h)
Values of P for other runs are listed in Table VII. From eq 10, k2/kl = P2/P1. The k2/kl ratios increase only slightly in the low-temperature range but show larger increases in the high-temperature range. The relatively large ratios of k2/kl are consistent with the low concentrations of HNP found in these runs. In order to check the goodness of fit of the original concentration data, rates of disappearance of A (-rA) and formation of B (rB) were cast in terms of cy and values to give rA = -
products
(C)
Solution of the appropriate differential equations describing this system, assuming the same sites are involved in each step, yields
kl k2 = P2 = where Xi are mole fractions. For XA and XB data for a given run, values of p1and P2 were obtained by nonlinear regression analysis. Figure 10 presents experimental data for run NP-15 as XB versus XA. The data scatter is typical and represents the limits of analytical accuracy in determination of NP and HNP by UV-visible spectroscopy. Since XB (HNP) is relatively small, it has a higher error than XA. A fit of the data of this run via eq 10 gave P1 = 0.244 f 0.103 and P2 = 2.57 f 1.29 (A values are 95% confidence intervals). By use of these values, the solid curve in Figure 10 was derived, which shows a reasonable fit to the experimental data. If the path NP products is ignored (k3 = 0), then P1 in eq 10 becomes unity and P2 = k2/k1. A fit of the data gave P2 = 14.7, and the derived curve is shown by the dotted curve in Figure 10. Clearly, this simplified series reaction pathway gives a poor fit to the data. Similar results were obtained for other runs. P1
-
rB =
(1/cyl)CA 1+
((YZ/al)(CA
+ CB)
(p1/"1)cA - (@2/("1)cB 1 + ( a z / a l ) ( C A + CB)
(11)
(12)
These equations were then solved by numerical integration using the cy and p values of Table VI1 to give CA and CB versus 7. An example of the fit obtained is shown in Figure 11 for run NP-6. In the above analyses, only group parameters cy and 0 could be obtained. However, for the different initial concentration series of runs at 357 OC, it is possible to extract actual rate parameters. Previous analysis of this data by eq 2 gave kNp and KNp(Table IV). According to Scheme 11, kNpKNp= kl kl. Combination of this value with the P1 and P2 values of eq 10 derived from these runs allows calculation of the individual ki's. Finally, since k, = kolKNp,the true ko, rate constants can be obtained. These are listed in Table VIII. The goodness of fit of these values was checked by numerical integration of the appropriate differential equations of Scheme 11. The resulting fits are shown by the solid curves in Figure 1 2 for NP-15. 3. Effect of Hydrogen Pressure. The effect of increasing hydrogen pressure was to greatly accelerate the overall rate of disappearance of NP, as shown in Figure 8 from runs at 357 "C. From first-order plots, the following power equation in pHwas obtained:
+
1636 Ind. Eng. Chem. Res., Vol. 27, No. 9, 1988
'"a 23
A 4 3
Y
c
2
1
I
4
6
I
i
iOOO/T, K-'
Figure 14. Temperature variation of
PI (A) and Bz (B).
r ,h-gcof/kgo~/
Figure 12. Data fit to parameters of Table VI11 for run NP-15.
-3
1
I
1
I 7
1
1
E
1000/ i , t-1
Figure 13. Temperature variation of a l / a z (A) and cy1 (B).
The noninteger order in pH signifies that the overall rate is not a simple function of hydrogen pressure. As shown above, the overall rate for disappearance of NP consists of two paths (Scheme 11) having rate constants k1 and k,, each of which may have a different hydrogen pressure dependence. The 1.57 power in pH can be rationalized if path 1 is assumed to be first order in pH and path 3 is second order. However, the data were insufficient to establish the validity of this assumption with certainty. 4. Effect of Temperature. a. Low-Temperature Region (275-350 "C). For the low-temperature range, eq 11 and 12 express the rates of disappearance of NP (A) and rate of HNP (B) in terms of the parameters a and p. The temperature variations of a1and a2/a1are given in Figure 13 and that of p1and p2 in Figure 14. From these plots, activation energies and preexponential factors were obtained, which are listed in Table IX, together with their ki and Ki equivalences. Incorporation of these values gives the following rate expressions:
Table IX. Temperature DeDendencu of a and B Parameters temp energy, preexponential range, "C uarameter eauivalence kcal/mol" factor 275-350 l/a, (k1 + k&KA 16.2 f 3.9 3.38 X lo5 1 + 2K,coA a2/a1 KA - 2K;-9.6 f 2.9 2.84 X lo-' 1 + 2Kc C0~ P1 kl 4.3 f 1.6 1.76 X 10' kl + k3 P2 k2 4.6 f 1.7 1.38 X lo2 kl + k3 350-380 Pi kl 15.5 f 2.0 1.37 X lo5 kl + k, Pz k2 30.4 f 6.7 1.42 X 10" kl + k , k' (k1 + k 3 ) K ~ 20.8 f 2.0 5.35 X IO6 1 + KAcOA
'f values are standard deviations. to an initial concentration of cOA = 35 ppm. Also, they only apply to a H2 pressure of 69 atm. Equations 14 and 15 predict concentration-time data quite well in this temperature range. However, they fail to fit the higher temperature data adequately, implying a change in mechanism above 350 "C. b. High-Temperature Region (350-380 "C). In the high-temperature region, the reaction of NP is pseudo first order. Hence, the inhibition term is independent of conversion as discussed before but not necessarily equal to unity. This leads to the simplified rate expressions,
rA = -k%A
(16)
r, = k'@lCA- k'&Cg
(17)
where k' is the pseudo-first-order rate constant. The temperature dependencies of k', pl, and p2 in this region are given in Table IX. It is noted that the temperature responses of the p values in this region are different than for the low-temperature region (Figure 14), which explains why eq 14 and 15 do not adequately fit the high-temperature data. The sharp transitions in slope at about 350 "C almost surely indicate a change in mechanism in the reaction. If the cause were due to the K's becoming van3.38 x 105~162001~~ CA rA = small as the temperature increased, a more gradual 1 + 2.84 X 10-5e9600/RT(~A + cB) (14) ishingly transition in slopes would be expected to occur. Incorof the appropriate parameters gives rB = (5.94 x - 4.66 x 1 0 7 ~ - 2 0 ~ 0 / ~ ~ ~ ~ )poration / (1 + 2.84 X 10-5e96001RT(~~ + CB)) (15) rA = -5.35 x 106e-20m/RTCA (18) These equations can be solved by numerical integration to give concentrations of A (NP) and B (HNP) as a function of r at any temperature. However, since the parameters include a (1 + 2KccoA)term, and since a value of K, could not be determined, eq 23 and 24 only apply
rB = 7.33 X 10lle-36300/RTcA- 7 . 3 x 10l7e-512WlRTCB (19) Again, these expressions only apply to COA = 35 ppm and pH= 69 atm. Calculations show that this set of equations
Ind. Eng. Chem. Res., Vol. 27, No. 9, 1988 1637 fits the high-temperature data quite well.
Discussion The noncatalytic demetalation reaction can be regarded as a thermal demetalation reaction. The results indicate that the thermal effect is very small. Thus, VP and NP are thermally stable under the reaction conditions employed here. The HDM runs were carried out with crushed catalyst to obtain intrinsic kinetics. Comparison of our NP data with those reported by Hung and Wei (1980a) under the same conditions showed comparable HDM rates. Since our catalyst size was some 21/2 times smaller, and Hung and Wei showed the absence of diffusional effects in their runs, we conclude that our results represent intrinsic data, uninfluenced by diffusion. The UV-visible absorption spectrum for reaction of VP showed a new peak at 633 nm. This peak is attributed to the presence of an intermediate, as it disappears after several hours of reaction. From the literature (Hung and Wei, 1980b; Weitcamp et al., 1984), it is believed that this peak is due to vanadium chlorin, in which one of the pyrrole groups of V P is partially hydrogenated. On the basis of the same reasoning, the presence of a new peak in the NP runs is likely due to nickel chlorin. The relative concentration of intermediate in the NP runs was somewhat lower than that obtained by Agrawal and Wei (1984). This difference may be due to the fact that their catalyst was not presulfided nor were any sulfur compounds present in their system. Thus, their CoMo catalyst was undoubtedly in a reduced state, which would be expected to have a higher hydrogenation activity than our sulfided catalyst. Reaction of VP in helium was very low compared to reaction in hydrogen, indicating that the hydrogenated intermediate mentioned above is probably necessary in a sequence reaction path leading to deposited vanadium on the catalyst, as proposed by Agrawal and Wei (1984). The deposited vanadium rapidly sulfides under our reaction conditions (Fleish et al., 1984). The small amount of reaction in helium may be due to a small amount of hydrogen transfer between VP molecules adsorbed on the catalyst, forming some intermediate which undergoes further reaction to deposit vanadium. A similar result was obtained in the NP run with helium. In a study of the kinetics of HDM of NP over an oxidized CoMo catalyst, Agrawal and Wei (1984) proposed the following reaction scheme: Scheme I11 NP
kl k-3
HNP
k2
products
They reported an increase in the rate constants of about double when the NP concentration was reduced from 27 to 15 ppm. However, in a later paper, Ware and Wei (1985a) found no effect of NP concentration between 16 and 90 ppm with the same catalyst. Our data, obtained under similar conditions as theirs, but with a sulfided CoMo catalyst, show similarities in the overall reaction sequence, e.g., presence of a hydrogenated intermediate porphyrin. However, significant differences in the reaction scheme are evident. First, the HNP to NP ratio increased with conversion while their ratio rapidly reached a steady-state value. Also, our ratios were lower than they found. This may be attributed to a lower hydrogenation rate for our sulfided versus their reduced catalyst. Second, analysis of our data clearly showed the need for an alternate, direct path, NP products (Scheme
-
-
11). Furthermore, inclusion of the reversible ,path, HNP NP ( k - l ) , gave considerbly poorer fits to the data. Third, our data clearly showed reactant inhibition was important in the kinetic analysis for the low-temperature runs. Selectively analysis of the data showed that a direct path from NP products (path 3 in Scheme 11) was necessary to obtain good data fits. However, this direct path is probably not a real path but is rather due to a direct conversion of adsorbed NP to adsorbed HNP to products in a single adsorption step, without desorption of HNP, whereas paths 1and 2 in Scheme I1 represent that fraction of NP converted through desorption and readsorption of HNP (normal path for adsorption equilibrium). Consider the following surface reaction scheme:
-
Scheme IV A
B
C
JT where A, B, and C represent NP, HNP, and products, and subscripts a, d, and s stand for adsorption, desorption, and surface reaction, respectively. In this scheme, A and B are not in adsorption equilibrium (C can or cannot be in adsorption equilibrium). Applying the steady-state treatment for adsorbed A and B, Clark (1970) has shown that the following rate constants of Scheme I1 are related to those of Scheme IV: kl/k3
=
kd,B/ks,B
(20)
According to eq 20, the ratio of k l to k3 (path 1 to the apparent path 3 of Scheme 11) is actually related to the ratio of the desorption rate constant of B to its surface reaction rate constant. If kd,B > k8,B,then k l > k 3 or, conversely, a significant value of k3/k1 indicates that desorption of B is not very much faster than its surface reaction to products. Thus, the apparent direct reaction path 3 of NP products can be explained as that fraction of reaction that proceeds by a single adsorption step without desorption of B. From previous selectivity analysis, k,/k3 can be obtained from data fit to eq 10, viz.,
-
k d k 3 = 1/61 - 1
(21)
and consequently values of kd,B/k,,Bcalculated via eq 20. Table VI1 gives the latter ratios for runs at different temperatures. It is seen that the rate of desorption of B (kd,B&) is not significantly larger than that for the surface reaction of B to products (ks,BeB)and in fact is lower at lower temperatures. This is in line with the concept that path 3 of Scheme I1 is an artifact and can be explained by a relatively slow desorption rate of B in Scheme IV. It is also noted that the values of kd,B(ks,Bincrease with increasing temperature, signifying a higher activation energy for desorption of B than for its surface reaction. An Arrhenius plot Of the kd,B/ks,Bratios gives E d , B - E s , =~ 10 f 2 kcal/mol. Although unusual in the normal sense that activation energies for desorption are generally lower than for surface reaction, quite high activation energies for desorption have occasionally been noted. For example, Cretanovic and Amenomiya (1967) reported desorption activation energies of 27 and 36 kcal/mol for ethylene adsorbed on two different sites on alumina. The present result can be considered a manifestation of the strong adsorption of NP and can account for its lack of equilibrium adsorption under our conditions.
1638 Ind. Eng. Chem. Res., Vol. 27, No. 9, 1988
In their HDM study with CoMo oxide catalyst, Agrawal and Wei (1984) reported that k , of Scheme I11 was proportional to pH. This implies either a direct reaction between adsorbed NP and gaseous H2 (Rideal mechanism) or a surface reaction between adsorbed NP and adsorbed H2,with sparse coverage of H2. According to our surface reaction, Scheme IV, there is only one reaction path for A, B,. Since the amount of intermediate product B is relatively small and B, C, is relatively rapid, the surface reaction A, B, can be considered relatively slow (k,