Gravimetric study of adsorbed intermediates in methanation of carbon

Aug 1, 1981 - Note: In lieu of an abstract, this is the article's first page. Click to increase image size Free first page. View: PDF. Related Content...
0 downloads 10 Views 871KB Size
Ind. Eng. Chem. Fundam. 1981, 20,229-234

s = of the solute w = at the cake or gel surface Literature Cited

229

Heertjes, P. M. “Formation of flker cakes and precoetes”, In “The Sclentlflc Basis of Flltratlon”, Ives, K. J., Ed., Noordhoff Intematbnal PuMlehing: The Netherlands. 1975 D 297. Henry, J. D., Jr. “C&s-fbw filbatbn”, in “Recent Development In Separation Sclence”, Vd. 2 CRC Press: Cleveland, 1972;p 205. Karnis, A.; Wsmlth, H. L.; Mason, S. 0. a n . J. Chem. Eng. Aug 1968,

Ammerfeen, A. C. F.; L d , B. F.; wley, A. J. r8PP/ 198% 52(1), 118. Balky, M. W. W.D. Dissertation, North Carolina State University, Raleigh

181.

1973.

Kraus, K. A. “Cross-Fbw Filtration and Axlai Flltratlon”, paper presented at the 29th Annual Pwdue Industrlal Waste Conference, Lafayette, IN, May

Blxkr, H. J.: Rappe, G. C. U.S. Patent 3 541 006, Nov 17, 1970. Blatt, W. F.; Dravid, A.; Michaels, A. S.; Nelson, L. ”Solute Pelarizetion and Cake F m t b n in Membrane Wtraflltration: Causes, consequences,and control Techniques", in “Membrane Science and Technology”, J. E. Fllnn, Ed.; Plenum Press: New York, 1970; p 47. Banner, H. “HWodynamic Resistance of Particles at Small Reynolds Nunbers”, in “Advances In chemical Englneerlng”, Vd. 8, Drew, T. E.; Hoopes, J. W.; Vermeulen, T., Eds.; Academic Press: New Y W k 1966;P

1974.

b b n , R. E, “Hyperfiltration mamatron in pletka&~rame systems”, Elsevier: New York, 1977. Porter, M. C. Ind. Eng. Chem. prod. Res. Dev. 1972, 17, 234. Rublnow, s.I.; Keller, J. E. J . F h r l d m h . 1061, 11.447. *e, G.; s l l m r g , A. J . Rum M.1982, 74, 138. Trettln, D. R.; Doshi, M. R. Ind. Eng. chem.Fundem. 1980, 79,189. Trettln, D. R.; Doshi, M. R. ACS Symp. Ser. 1981, In press.

287.

Cox. R. G.: Brenner. H. J . FhrklMeCh. 1967. 28.391. Cox; R. G.i Brenner’H. (3”. €q.Sci. 1968, 23, 147. Dd’d, M. R., mpubllshed work, The Institute of Paper Chemistry, 1979. &ace, H. P. Chem. fng. Reg. 1959, 49(6),303.

Received for reuiew May 30, 1980 Accepted March 30,1981

Gravimetric Study of Adsorbed Intermediates in Methanation of Carbon Monoxide David C. Gardner and Calvin H. Bartholomew* BYU &kr&sb

b h m ,o e p e m n t of Chemical Engineering, B

W m Young University, Rovo, Utah 84602

The nature of adsorbed species on nickel during methanation of CO and the kinetics of active carbon hydrogenation were studied using a gravimetric flow system. The amount of CO adsorbed during methanation decreases with temperature indicating the increase of the methanation rate relative to that of adsorption. From adsorption/da sorption experiments it is apparent that three kinds of surface species are present on the nickel surface during methanation: (i)species easily desorbed such as CO, CH,, Hp, etc., (ii) species reactive with H2 such as atomic carbon, and (iii) unreactive species such as polymeric carbon. Carbon deposited during CO disproportionation at 573 K is rapidly ysifi at 473 K with an apparent activation energy of 70 f 8 kJ/mol, a turnover number of Ncn, = 3.5 X lo3 s- and a reaction order for hydrogen of 0.5 f 0.1. Comparison of rates for methanation, carbon formation, and carbon gasification suggests that carbon formation is rate controlling at bw temperatures and carbon gasification rate determining at high temperatures. Sulfur lowers moderately the amount of CO dlssociatively chemisorbed on Ni under reaction conditions but substantially decreases the rate at which active carbon is gasified by hydrogen thereby enhancing the transformation to inactive carbon.

Introduction Mechanisms proposed for nickel-catalyzed CO methanation can be divided into two major categories. The first kind supposes a carbon-hydrogen-oxygen containing species as the principal intermediate, the second an active, hydrogenable carbon produced by disassociative adsorption of CO. A number of recent studies (Wentrcek et al., 1976; Araki and Ponec, 1976; Ponec, 1978; Rabo et al., 1978; McCarty and Wise, 1979; Zagli et al., 1979; Goodman et al., 1980) provide strong evidence that hydrogenation of active carbon is the principal route to formation of methane on nickel. Nevertheless, the precise nature of the rate-determining step remains unresolved. The purpose of this study was to more fully elucidate the adsorbed intermediates and mechanism involved in catalytic methanation of CO on a typical nickel methanation catalyst. Rates of adsorption and desorption of surface species and of gasification of carbon were measured gravimetrically to determine their kinetics and possible roles in methanation. Experimental Section Apparatus. Experiments were performed using a thermogravimetric (Perkin-Elmer TGS-11) flow system. Sample containers were platinum pans, having 25 mg capacity. Catalyst temperature was monitored using a cal0196-4313/81/1020-0229$01.25/0

ibrated K-type thermocouple positioned within 1.5 mm of the catalyst bed. Calibration with magnetic transition standards showed the temperature variation between catalyst sample and the thermocouple read out to not vary more than 2 K. Gases were metered through calibrated capillary rotamers. Hydrogen (99.99%) and N2 (99.96%) were mixed and purified of O2and H20using a Deoxo unit followed by fresh 5A molecular sieve. Carbon monoxide (99.99%)was also passed through a heated 5A molecular sieve to remove iron carbonyl prior to combining with other reactants. Catalyst. The preparation, subsequent initial reduction in batch quantity, and measurement of metal surface area of the 14% Ni/A1203 catalyst is reported elsewhere (Bartholomew and Farrauto, 1976; Batholomew, 1977). Samples were finely crushed to minimize pore diffusional resistance. Calculations indicate that bulk diffusional limitations were minimal at the temperatures and reactant concentrations of the experiments. A partially sulfur poisoned sample was obtained by exposing the finely divided catalyst to 9 ppm of H2S in H2 in a fluidized bed reactor maintained at 723 K until approximately 50% of the surface was poisoned. Indeed, a subsequent surface area determination indicated 54% of the original surface was poisoned for H2chemisorption. A completely sulfur poisoned sample was obtained from in situ poisoning 0 1981 American Chemical Socletv

Ind. Eng. Chem. Fundam., Vol. 20, No. 3, 1981

230

Table I. Gravimetric Measurement of Adsorbed Species during Methanation at 473 K (H,/CO = 3, PCO = 4.5 kPa)

% Ni

41 0

3.0 3.16c

1

14.0

““-----, I

I

373

123

473

523

T

573

1

I

623

673

15.05c 14.0d

I

723

IO

Figure 1. Weight of adsorbed species during initial stages of methanation as a function of temperature (Pm = 4.5 kPa, H2/C0 = 3); 0,14% Ni/A1203(this study); A, 15% Ni/A1203(Farrauto, 1976).

measurements at 525 K, 110 kPa (Bartholomew et al., 1979). Procedure. The 10-15-mg samples of previously reduced and passivated catalyst were placed in the sample pan to a depth of about 1 mm and typically reduced for mol/min of Hz. The weight 4 h at 753 K with 3.6 X of each sample was monitored during reduction, usually reaching a steady-state value within 3 h, and this steadystate weight of the reduced catalyst was used in all subsequent calculations. Each sample was then flushed with N2 to remove H2 and cooled to the desired reaction temperature. Hydrogen and N2 flows were adjusted to the desired flows and the weight was allowed to equilibrate before the CO was introduced.

Results Adsorption and Desorption of Reactants and Products. Carbon monoxide at 3.6 X lo4 mol/min (Pco = 4.5 Wa) was added to a reactant stream of 16% H2and 84% Nz.The resulting mixture with a H2/C0 ratio of 3 was passed over the reduced sample causing a rapid weight increase in the first 1-2 min; the rate of weight gain then decreased significantly, reaching almost zero after 20-40 min. The reactants adsorbed during the first 1-2 min, could be almost completely reversibly desorbed with either N2 or H2/N2(PH, = 14 kPa) at 473-673 K, if the CO flow was ceased at that point. However, if the flow of CO was maintained longer than approximately 2 min, the adsorption was not completely reversible, indicating the presence of species on the catalyst which could not be appreciably removed with either N2or the H2/N2mixture at these temperatures. The weight increase during the initial 1-2 min is shown as a function of temperature in Figure 1. These results are compared with those obtained by Farrauto (1976) using a similar technique and a 15% Ni/A1203catalyst having a metal dispersion and surface area almost identical with our 14% Ni/A1203. The amount adsorbed is a maximum at approximately 473 K and then decreases with increasing temperature for both catalysts. The agreement between the data for the two catalysts at 423-473 K is excellent. The differences in the adsorbed weight between the two catalyst at temperatures greater than 473 K are probably a result of (i) different sample sizes used in these two studies causing differences in diffusional limitations at high temperatures and (ii) slight differences in experimental procedure (particularlyH2/C0 ratio and uptake time) and equipment. Nevertheless, the qualitative agreement is very good. Measured values of H2 adsorption uptake at 298 K and weight increases due to coadsorption of H2and CO during the first 1-2 minutes of reaction at 473 K are summarized

mmol of “complex”/mmol of H, H, uptakea mmol of wt increaseb COHdg of cat.) (mg/gofcat.) H, CO CH, C 0.079 0.076 0.376 0.351 0.376

1.98 2.05 8.76 7.92 13.39

0.84 0.90 0.77 0.75 1.2

0.90 0.96 0.83 0.81 1.3

1.8 1.9 1.7

2.1 2.2 1.9 1.6 1.9 2.5 3.0

a Total H,uptake at 298 K; H, refers to atomic, surface hydrogen. Weight increase measured gravimetrically during the first 1 - 2 min of CO and H, coadsorption in this study; during the first 20-30 min in the study by Sample Farrauto. E Data reported by Farrauto (1976). was run for 8 h before H, and CO were shut off.

in Table I. The mmol of “complex”/mmol of H, is the ratio of the adsorption uptake under reaction conditions, assuming one of the structures shown to be the surface complex, to the chemisorptive uptake of hydrogen atoms [H,] at 298 K, Le., the quantity of active sites. For example, if one were to assume that the most abundant surface methanation intermediates exists as a CO-H2 complex, occupying one nickel site per complex, then 70-90% of these sites would be occupied with such intermediate. The same is true for CO. If carbon were the most abundant surface intermediate, approximately 2 monolayers would need to adsorb to account for this weight. Depending upon reaction conditions, the surface actually contains a number of different species including C, COYCHI, and H20 as discussed below. Since active surface carbon obtained by CO disproportionation is so readily hydrogenated (McCarty and Wise, 1979; Goodman et al., 1980), hydrogenated forms of carbon such as CH2are also undoubtedly present on the surface. Information on the relative abundance of these different species was obtained from desorption experiments. After CO and H2 in N2 had been passed over the sample for about 20 minutes, CO and Hz flows were shut off leaving only N2diluent to flow over the sample. This resulted in a weight loss the rate of which was initially quite rapid but then decreased significantlyso that within 4-5 h a constant weight was obtained. If H2 was then readmitted at its former flow rate, a further rapid and discrete weight loss was observed. The sample at this point weighed more than it did initially, indicating the presence of a nonvolatile, nonhydrogenetable species on the surface. It was observed that in general the weight of adsorbed species not removable by either N2or H2in N2was approximately equal to the weight increase during the 2-20 min period after the addition of CO. Figure 2 shows a schematic of these results. The various weights adsorbed and desorbed by the catalyst using these techniques after exposure to H2 and CO for 20 min are listed in Table 11. The weight increase denoted by A is that accumulated by the catalyst during the rapid uptake (first 1-2 min) and plotted as a function of temperature in Figure 1. It obviously decreases significantlywith increasing temperature. The weight gain represented by B is that accumulated during the 2-20 min period after CO addition and appears to be considerably less dependent on temperature. The weight of volatile species that could be desorbed in just flowing N2is denoted by C. Interestingly, nearly 70% of this weight is desorbed in the first few minutes, but considerably more time was required for the remaining 30% to be removed. Gas chromatographic sampling of the desorbing species under similar conditions showed only

Ind. Eng. Chem. Fundam., Vol. 20, No. 3, 1981

231

Table 111. Gasification of Active Surface Carbon from CO Disproportionationa on 14% Ni/AI,O,

T, K 453 453 47 3

rateb X 10' 2.0 3.0 4.6

f f f

0.1 0.1 0.1

NCH,X

loJc

PH,,kPa

1.5 f 0.1 2.3 t 0.1 3.5 ? 0.1

PCO= 6.4 kPa. Units of (mol/g of cat.-s). of (molecule CHJsite-s).

____-

J

0

L

t C 0 added 0

10

€4

30

90

J

120

150

TIME AFTER ADDITION OF COjmin)

Figure 2. Gravimetricadsorption and desorption during methanation of CO (PCO= 6.4 P a , H2/C0 = 3) on 14% Ni/A120S. (See Table I1 for explanation.) Table 11. Gravimetric Measurementa of Adsorption and Desorption of Surface Species during Methanation of CO on 14% Ni/Al,O, fraction not desorbed not adsorbed desorbed desorbed (c+D)l fraction desorbed T, -K A C B C Cc D c EC (A+B)d E/(A+B)e 473 8.7 2.4 4.8 3.7 573 6.0 1.9 4.1 1.5 673 3.4 1.8 3.0 0.8

2.6 2.4 1.4

0.77 0.70 0.73

0.23 0.30 0.27

a A = initial weight accumulated during the first 1-2 min after exposure to CO; B = weight accumulated during 220 min period of exposure to CO; C = weight desorbed with N, after shutting off CO and H, flow; D = weight removed with HJN, (PH,= 1 4 kPa); E = weight not removed with either N, or H, at reaction temperature. Pco 7 4.5 kPa, HJCO = 3. Units of mg/g of cat. Fraction of adsorbed species removed by either N, or HJN, streams. e Fraction of adsorbed species which could not be desorbed by either N, or HJN, streams.

CO, CHI, and C02to be present in the gas phase. A t 473

K, CO was the predominate gas-phase species from desorption; at 673 K it was CHI. The tailing of desorbing species may indicate either a heterogeneous surface with varying degrees of affinity for reactants and products or the presence of species with varying degrees of reactivity with adsorbed Ha. The weight loss upon readmission of H2to the N2stream is represented by D. It apparently also decreased significantly with temperature. Since both chemisorbed CO and active carbon could be hydrogenated under these conditions, the previously described chromatographic experiments and subsequent gravimetric experiments were designed to differentiate between these two species. The weight of species not removable by either N2 or Hz (PH2= 14 kPa) at reaction temperatures of 473-673 K is represented by E. The values of B and E are fairly well correlated suggesting that most of the species accumulating after approximately 1-2 min of reaction with CO, exists as unreactive carbon or carbide depending on the temperature. No significant change occurred in any of the A , B, C, D , or E values with changes in Hz/CO ratio from 3 to 5. The fraction of adsorbed species which could be desorbed by either Nz or Hz/N2streams, (C+ D ) / ( A + B ) , or could not be desorbed, E / ( A + B ) , is also shown in Table I1 at three different temperatures. As the temperature is increased,slightly less of the total accumulated reaction weight is desorbed with N2 or Hz/Nz. Gasification of Carbon from the Boudouard Reaction with HP Carbon monoxide at 7.2 X lo4 mol/min (Pco = 6.4 kPa) was added to N2 at 1.0 X mol/min

2.5 5.0 2.5 Units

and flowed over a freshly reduced sample of 14% Ni/AlzO3 for approximately 1min at 573 K. The sample was quickly cooled to 473 or 453 K and flushed with N2until no further weight loss occurred. Hydrogen at 2.5 or 5.0 kPa was then added to the N2diluent and flowed over the catalyst. The sample lost weight rapidly, the rate of weight loss achieving a maximum after approximately 2 min. These maximum rates of gasification with H2are tabulated in Table III. At this point the sample had lost approximately 30% of the gasifiable carbon; 2.5-3.0 min after the addition of H2,the rate began to decrease until after approximately 6 min the rate was only 10% of the maximum. After the addition of H2,about 20 min was required for essentially all of the carbon, approximately 1 monolayer, to be removed. The specific activity of the 14% Ni/AlzO3 based on the number of nickel sites measured by Hz chemisorption or turnover number at 473 K, of 3.5 X molecules CHI/ site-s is within a factor of 2 of the value reported by Ho and Harriott (1980) for the gasification of active surface carbon on a 10% Ni/Si02 catalyst at about the same temperature. The apparent activation energy for the gasification of surface carbon on the 14% Ni/Al2O3 at 453-473 K is 70.3 f 8 kJ/mol (based on repeated runs at two temperatures). This value compares well to values of 63 and 71 i 19 kJ, reported for the same temperature range by Ho and Harriott (1980) and McCarty and Wise (1980). respectively. Assuming an equation for the hydrogenation reaction of the form: r = kO&,: where Oc is the fraction of active sites covered with active carbon, x was found to be 0.5 f 0.1 at 453 K (PH2 = 2.5-5.0 P a ) . Ho and Harriott (1980) also report a valaue for x of 0.5 for the hydrogen dependency of active carbon gasification at 498 K on a Ni/SiOz catalyst . Gasification with Hzof the Surface Species from CO/H2 Coadsorption. Other samples were exposed to both CO (Pco = 9.0 kPa) and H2 (H2/C0 = 3) at 573 K for 1 min, after which both CO and H2 flows were discontinued and the sample was quickly cooled to 473 or 453 K and flushed with flowing N2 until the sample weight equilibrated. Hydrogen at 2.5 kPa was then added to the N2 diluent and allowed to contact the sample. Turnover numbers of 1.0 and 3.3 X molecules CH,/site-s were obtained at 453 and 373 K, respectively, which are the same within experimental error as those observed for the gasification of active carbon from the Boudouard reaction (Table 111). Adsorption and Desorption of Reactants and Products on Presulfided Ni. Samples of 14% Ni/Al2O3 having been poisoned with H2Sto coverages of approximately 50% and loo%,respectively,evidenced considerably different behavior toward adsorption/desorption of surface species during methanation compared to a fresh sample, as shown in Table IV. The initial uptake corresponding to CO adsorption was reduced almost 50% by complete sulfur poisoning. The total amount adsorbed in the first 20 min was likewise reduced by about half. The percent of species which could be desorbed in either Nz or Hz/Nzstreams was reduced from 78% to 57% and then

292

Ind. Eng.

Chem. Fundam., Vol. 20, No. 3, 1981

Table IV. Adsorption-Des0 tion of Surface Species for Poisoned and Unpoisoned 14% Ni/Al,O, Samples during Methanations fraction desorbed, adsor bed desorbed not (CtD)/ sample A B C D desorbed, E (A+B) unpoisoned 50% poisoned 100% poisoned

9.5 6.4 5.0

2.1 2.5 1.2

6.2 4.3 0.7

2.9 0.8 0.45

2.5 3.7 5.0

0.78 0.57 0.19

fraction not desorbed, E/(A+B) 0.22 0.43 0.81

A = initial weight accumulated during the first 1-2 min after exposure t o CO; B = weight accumulated during 2-30 min period of exposure t o CO; C = weight desorbed with N, after shutting off CO and H, flow; D = weight removed with HJN, (PH = 30 kPa); E = weight not removed with either N, or H, at reaction temperature. T = 573 K; Pc, = 9.0 Wa; HJCO = 3;A-E have units of mg/g of cat.

to 19% as 50 and then 100% of the nickel sites were poisoned with sulfur; at the same time the percent of unreactive species increased from 22 to 43 and finally to 81% .

Discussion Adsorption and Desorption of Surface Species during Methanation. The gravimetric adsorption measurements during the first 1-2 min after admission of CO provide useful information on the nature of CO adsorption on Ni under reaction conditions. The amount of CO initially adsorbed by the catalyst (species A in Table 11) corresponds to approximately83% of a monolayer at 473 K, decreasing to 32% as the temperature was increased to 673 K. The decreasing adsorption of CO with increasing temperature (Figure 1and Table 111,may be the combined result of two different effects: (i) weaker adsorption of CO at higher temperatures and (ii) a higher rate of CO methanation at higher temperatures, reducing the surface coverage of CO (Farrauto, 1976). The latter effect is probably more important. From the gravimetric adsorption/desorption experiments it is apparent that three kinds of surface species are present on the nickel surface during methanation of CO: (i) volatile species (C) easily removed by inert gas, (ii) species (D) which react readily with hydrogen and subsequently desorb, and (iii) nonvolatile, apparently unreactive species (E). Chromatographic analyses of the gaseous effluent during similar adsorption and desorption experiments in a differential tubular reactor revealed that species C, desorbed in inert gas, are primarily reactants and products from methanation such as H2,CO, CHI, etc. Moreover, the data revealed that the distribution of these species depends upon temperature. The fact that CO is the predominant carbon-containing species observed in the gas phase effluent at 473 K (1atm) suggests that CO is also the most abundant steady-state surface intermediate under these reaction conditions. This conclusion is reasonable in view of the low conversions of CO to CH4observed at 473 K and 1 atm. It is also consistent with pulse reactor data (Rabo et al., 1978) suggesting that CO is the predominant surface species on nickel at 473 K. That the gas-phase CO is probably not a result of association between C(a) and O(a) under these conditions was recently demonstrated in a temperature programmed desorption study of the same 14% Ni/A1203by Zagli et al. (1979). These authors showed that CO adsorbed and desorbed for the most part nondissociatively (Le., without formation of C02) at temperany atures below 473 K and that in the presence of H2, dissociatively adsorbed CO desorbed principally as CHI. The fact that methane is the primary species observed in the gaseous effluent at 673 K implies that either methane is a predominant surface species or that the active surface precursors to methane (presumably active carbon and atomic hydrogen) react rapidly to form methane during

the initial few minutes of desorption. The data of %bo et al. (1978) suggest that active carbon (rather than CO or CHI) is the predominant surface intermediate at this higher temperature (673 K). The supposition of a rapid reaction of competitively adsorbed carbon and atomic hydrogen to form a methane during desorption is further supported by the observation of Ho and Harriott (1980) that during methanation a step decrease in CO concentration resulted in a significant, rapid increase in methane concentration followed by a gradual decrease. Previous pulse reactor, TPD/TPSR, and Auger data (Wentrcek et al., 1976; McCarty and Wise, 1979; Zagli et al., 1979; Goodman et al., 1980) have established that atomic (a)or “carbidic” carbon from CO disproportionation reads wiith hydrogen to form methane and that this is the most favorable route for methanation of CO. In view of this evidence there is little doubt that species D, which was easily removed with hydrogen, consisted primarily of atomic carbon (C,). This is further supported by the data from this study showing that the specific rates for carbon hydrogenation and hydrogenation of species formed by coadsorption of CO and H2 are the same. In other words, both processes proceed through the same intermediate, carbon. Probably two factors account for the decreasing amount of species D (presumably C,) with increasing temperature (see Table 11): (i) the total amount of adsorbed species was observed to decrease with increasing temperature (see Table I1 and Figure 1) presumably as a result of faster reaction rates at the higher temperatures (Farrauto, 1976) and (ii) a larger fraction of active carbon could react with adsorbed hydrogen to methane during the preceding desorption in N2 at the higher temperatures. The first of these explanations is consistent with a recent observation (Kelley and Goodman, 1980) that the rate of methanation correlates inversely with the steady-state coverage of C,. On the basis of evidence from this and previous studies (McCartyand Wise, 1979;Rabo et aL, 1978)the unreactive, nonvolatile species E is presumed to be an amorphous, polymeric form of carbon, sometimes designated as C, (McCartyand Wise, 1979). The gravimetric data in this study indicate that a steady-state coverate of C, of about 20-30% is established gradually within the first 20-30 min of reaction at 473473 K. This observation finds very good agreement with the pulse reactor study of Rabo et al. (1978) in which an unreactive surface residue of typically 10-20% of a monolayer was observed on Ni/Si02 at 573 K. McCarty and Wise (1979) observed that C, and C, formed from CO exposure at 550 K populated the nickel surface in a ratio of a bout 2:1, in excellent agreement with our results. Mechanism and Rate-Determining Steps in Methanation. The gravimetric data from this study provide additional evidence entirely consistent with previous studies showing that carbon is the principal intermediate

Ind. Eng. Chem. Fundam., Vol. 20, No. 3, 1981

._

' . .

Hydrogenatbn d Carbon

\

-1.0

c

. '-0

CO Dirrociclion Ea-92-105 kJ

20

21

22

WT x t ~ 3

Figure 3. Arrhenius plot of turnover numbers ( N C ~ 0 ) ,:Gasification (PH,= 2.6 kPa) of active carbon from CO disproporationation at 673 K;A, methanation of CO (PHI)= 2.5 P a , H2/C0 = 4); 0 , monolayer carbon deposition from CO disproportionation (Goodman et al., 1980).

in methanation of CO. In addition, the data of this study also constitute a basis for suggesting the rate-determining

steps in methanation of CO under various reaction conditions. The rapid initial uptake of CO and desorption of products observed under reaction conditions in this study, qualitatively much faster than formation (Goodman et al., 1980) or gasification (McCarty and Wise, 1979; Ho and Harriot, 1980; Goodman et al., 1980) of active carbon, indicates that surface reactions rather than adsorption or desorption reactions are rate controlling. The role of surface reactions is best illustrated by comparing rate data for the most important steps, CO dissociation,and carbon gasification,with rate data for the overall reaction. Figure 3 (an Arrhenius plot) shows rate data for hydrogenation of active carbon on Ni/AlZO3from this study compared with CO methanation data for the same catalyst from a previous study (Fowler and Bartholomew, 19791, and recently obtained data for CO dissociation to carbon on Ni (100) (Goodman et al., 1980). The data in Figure 3 show that at 450 K (1/T = 2.2 X 10") the rate of hydrogenation of carbon is significantly higher (by about a factor of 10) than the rate of CO dissociation to carbon. The rate of methanation, if extrapolated to the same temperature, is the same within experimental error (* 25%) as for CO dissociation. Thus at low temperatures (450-550 K), CO dissociation is apparently the rate-determining step, a conclusion supported by several previous studies (Dalla Betta and Shelef, 1977; Zagli et al., 1979; Palmer and Vroom, 1977). If the data in Figure 3 are extrapolated to higher temperatures the picture changes significantly. Because the activation energy for hydrogenation of carbon is significantly lower than for CO dissociation (70 kJ compared to 92-105 kJ), the lines on the Arrhenius plot representing these two processea should cross over at a h igher temperature, estimated to be about 600 K (1/T = 1.7 X lo"). Thus, at temperatures above 600 K, the rate of carbon deposition should exceed the rate of carbon hydrogenation and the latter reaction should become rate determining. Accordingly, these data predict a shift above 600 K in the activation energy for methanation from about 100 to 70 kJ/mol. Indeed, this has been observed in a recent kinetic study in this laboratory (Sughrue and Bartholomew, 1981). What these data suggest then is that in steady-state methanation of CO there is a delicate balance between formation and removal by H2 of carbon, a conclusion also

233

reached independently very recently by other workers (Goodman et al., 1980). However, the rate-determining step can shift with changes in temperature and partial pressures of the reactants. This has important implications for both the kinetics of the reaction and the deactivation of the catalyst by carbon deposits. That the kinetics of CO hydrogenation on nickel are consistent with this mechanistic view will be demonstrated in a paper to follow (Sughrue and Bartholomew, 1981). Effects of Sulfur Poisoning on the Rate-Determining Steps. It is well known that nickel methanation catalysts are very sensitive to poisoning by sulfur compounds at very low concentrations. However, the question of how sulfur poisons the reaction has been a matter of some controversy. Some of the previous workers (Ponec, 1978) have speculated that sulfur poisons CO hydrogenation by breaking up the nickel surface ensembles necessary for dissociative adsorption of CO. The data of this study provide new insights and an alternative explanation. The data in Table IV show that sulfur poisoning lowers somewhat the amount of CO adsorbed on nickel under reaction conditions but does not completely prevent the dissociative adsorption. Indeed, even on a completely sulfur poisoned catalyst the amount of CO adsorbed dissociatively is lowered by only 50% (see Table IV). Nevertheless, there is a drastic alteration in the product distribution. Essentially no measurable methane is produced by the poisoned catalyst (Bartholomew et al., 1979). From Table IV it is apparent that the fraction of volatile or hydrogenatable species C, is substantially reduced (from 78 to 19%) and the fraction not desorbed C, is substantially increased (from 22 to 81% ). What accounts for the significantly smaller fraction of volatile and C, species and the substantiallylarger fraction of C, species in the presence of sulfur? We postulate that absorbed sulfur poisons the dissociative adsorption of Hz, thus preventing the gasification of active carbon, C,. Since C, is no longer rapidly removed by reaction with hydrogen, it has sufficient ;esidence time on the surface to react to C,. In addition to C, formation, a relatively small amount of liquid or waxy hydrocarbons are undoubtedly formed (Dalla Betta et al., 1975),accounting for the small amount of species C-and D for the 100% poisoned sample in Table IV. Thus by preventing atomic hydrogen from reaching the active carbon, sulfur modifies the selectivity in favor of hydrogen poor products. This alternative explanation for sulfur poisoning finds support from the work of Dalla Betta et al. (1975) and recent pulse reactor studies by Wentrcek et al. (1980).

Conclusions 1. Upon admission of CO to a H2/N2mixture flowing over Ni/AlzO3, a rapid uptake is observed during the first few minutes of reaction due to adsorption of CO. At 473 K the surface stoichiometry is consistent with CO/Ni, = 0.8-0.9. At higher temperatures the amount of CO adsorbed decreases significantly (CO/Ni8 = 0.3 at 673 K) presumably because the surface species are depleted by fast reaction. 2. Three kinds of adsorbed species are observed in CO methanation: (i) easily desorbed species such as CO, CH,, Hz etc., (ii) species reactive with Hz such as atomic carbon, C,, and (iii) unreactive species such as polymeric carbon, C,. The distribution is quite temperature dependent. At 473 K, CO is probably the most abundant surface species; at 673 K the surface species are mainly C,, C,, or CH,. Thus carbon is a measuratjle, primary intermediate in methanation, the most abundant surface species under some reaction conditions, and present on the surface in

234

Ind. Eng. Chem. Fundam. lS81, 20, 234-239

both active and inactive forms. 3. The close agreement of rate data from this study for hydrogenation of carbon and for hydrogenation of coadsorbed CO and H2provides new evidence that carbon is the primary intermdediate in CO methanation. 4. The data from this study for hydrogenation of carbon combined with rate data from other recent studies suggest that the rate-determining step in CO methanation is: (i) CO dissociation at low temperatures (450-600 K) and low PCOor (ii) hydrogenation of atomic carbon at high temperatures (above 600 K) and low p ~ Thus, ~ . there is a delicate balance between these two reactions which depends upon temperature and reactant composition. 5. Sulfur poisons the dissociation of H2to a greater extent than the dissociation of CO thereby modifying the selectivity in favor of hydrogen-poor products. 6. This study illustrates the potential of gravimetric measurements combined with chromatographic analysis in the study of reaction intermediates and mechanisms. Acknowledgment The authors gratefully acknowledge financial support from the National Science Foundation (ENG 76-81869), encouragement to perform this work by Dr. Robert J. Farrauto, and technical assistance by Mr. Gordon D.

Weatherbee and others of the BYU Catalysis Laboratory. Literature Cited Araki, M.; Ponec, V. J. Catal. 1976, 44. 439. Bartholomew. C. H.; Farrauto, R. J. J. Catel. 1976, 45, 41. Bartholomew, C. H. "Alloy Catalysts with Monolith Supports for Methanatbn of Coal-Derived Gases", Final Technlcal Progress Report FE-1790-9 (ERDA), Sept 6, 1977. Bartholomew, C. H.; Weatherbee, 0. D.; Jarvl, G. A. J. Catal. 1979. 60, 257. Dalla Betta, R. A.; Piken. A. Q.; Shelef, M. J . Catal. 7975, 40, 173. DaUa Betta, R. A.; Shelef, M. J. Catal. 1977, 49, 383. Farrauto, R. J. J. Catal. 1976, 47, 482. Fowler, R. W.; Bartholomew, C. H. Ind. Eng. Chem. prod. Res. Dev. 1979, 78, 339. Goodman, D. W.; Kelley, R. D.; Madey, T. E.; White, J. M. J. Catal. 1980. 64, 479. Ho. S. V.; Harrlott, P. J. Catal. 1980, 64, 272. KeHey, R. D.; Goodman. D.W.; Madey, T. E. Prepr. ACS Dlv. Fuel Chem. 1080, 25(2), 43. McCarty, J. C.; Wise, H. J. Catal. 1979, 57, 406. Palmer, R. L.; Vroom, D. A. J. Catal. 1977, 50, 248. Ponec, V. Catel. Rev. Sci. Eng. 1078, 78, 151. Rabo, J. A.; Rbch, A. P.; Poutsma, M. L. J. Catal. 1978, 53, 295. Sughrue, E.; Bartholomew, C. H., BYU, paper In preparatbn, 1981. Wentrcek, P. R.; Wood, B. J.; Wlse, H. J . Catal., 1976, 43, 363. Wentrcek, P. R.; McCarty, J. 0.; Ablow, C. M.; Wlse, H. J . Catal. 1980, 67, 232. Zagli, A.; Falconer, J. L.; Keenan. C. A. J . Catal. 1979, 56, 453.

Received for reuiew June 13, 1980 Accepted March 13, 1981

Experiences with Dynamic Estimators for Fixed-Bed Reactors P. Henrlk Wallman and Alan S. Foss' D e p a m n t of Chemical Engineedng, Univefsity of California, Berkeley, California 94720

A slmple strategy is proposed for constructing a state and dlsturbance estimator for use In control systems on

fixed-bed reactors when the reactor dynamics are incompletely and inaccurately known. Use Is made of an estlmator of conventional form having a single tuning parameter through which one may conduct estimator poles to locations that are favorable to the retention of system stability when the control loop is closed. The modes of such an estimator are fast relative to the process modes. The practicabkty of this fast estimator is demonstrated by its superior performance over a conventional Kalman filter when tested in a computer control system for an operating laboratory reactor. Computational analyses are made to elucidate the reasons for its superior performance.

It is widely recognized that effective control of a process often requires information about process variables that cannot be measured directly or reliably. In response to this need, there have appeared numerous proposals in the literature for the estimation of unmeasured process variables under dynamic conditions. All approaches rely generally on the combination of available measurements with available process models to yield estimates of the unmeasured variables and possibly an enhancement of the quality of measured variables. However, the approaches advanced so far do little to help the engineer who is uncertain about his process model and who has no hope of improving that model owing to an inability to measure the variables to be estimated. From a practical point of view, a strategy for estimation and control of unmeasured variables is needed in such circumstances. Model inaccuracies can be manifold: structural inconformity, parameter inaccuracies,m o d e l e d modes, and neglected norilinearities. More often than not, the control system designer is limited 0196-4313/81/ 1020-0234$01.25/0

to a particular model without knowledge of its deficiencies and is therefore faced with a trial-and-error tuning task when the estimator is implemented in the process control system. In cases where several measurements are processed, this is a difficult, time-consuming, and frequently unavailing task unless a coordinated strategy to cope with the various inaccuracies has been developed before hand. In the work reported here, we advance a simple strategy for constructing an estimator when knowledge of the process model is shaky, as it is for the fixed-bed reactor with exothermic reaction to which we apply the estimator (Silva et al., 1979; Wallman et al., 1979). A major source of model inaccuracy in this case is neglected fast modes, and an interpretation of the ameliorating features of the estimator is given by examining the migration of fast process modes when the loop is closed around a system incorporating the estimator. The scheme attempts to minimize the migration by partial separation of the estimator modes from the process modes. The approach de@ 1981 American Chemical Society