Regeneration of Nickel Catalysts Deactivated by Filamentous Carbon

After coking, the catalysts were regenerated by carbon gasification in H2 at 800 °C. TPO of the carbon deposits was carried out, and the filaments we...
1 downloads 0 Views 226KB Size
3180

Ind. Eng. Chem. Res. 1997, 36, 3180-3187

Regeneration of Nickel Catalysts Deactivated by Filamentous Carbon D. Duprez,* K. Fadili, and J. Barbier Laboratoire de Catalyse en Chimie Organique, URA CNRS 350, 40 Avenue du Recteur Pineau, 86022 Poitiers Cedex, France

Unpromoted and K-promoted Ni/Al2O3 catalysts were prepared, calcined at 400 or 700 °C, and subsequently reduced at 500 or 800 °C. Cyclopentane hydrogenolysis was carried out within the 370-500 °C temperature range to obtain a carbon deposit. After coking, the catalysts were regenerated by carbon gasification in H2 at 800 °C. TPO of the carbon deposits was carried out, and the filaments were examined by SEM and TEM. The structural parameters (particle size, degree of nickel reduction, promotion by K) as well as the hydrogen flow rate showed marked effects on both the coking and the regeneration. The catalysts coked at high temperature (≈480 °C) form very stable carbon filaments whose gasification requires both a high temperature (800 °C) and a high H2 flow rate. On the contrary, those which are coked at a lower temperature (420-460 °C) form filaments much more readily gasified by H2. Introduction For mechanical reasons, the formation of filamentous carbon (whiskers) on nickel catalysts is a serious problem which can affect dramatically industrial processes such as steam reforming (Rostrup-Nielsen, 1975; Bartholomew, 1982) and CO2 reforming (Bhattacharyya and Chang, 1994; Bradford and Vannice, 1996). Many sources of carbon have been employed in the investigation of the formation of these filaments, viz., carbon monoxide and methane (Rostrup-Nielsen, 1975), ethylene (McCarthy et al., 1982; Kim et al., 1991), propylene (McAllister and Wolf, 1992), acetylene (Pen˜a et al., 1996), and n-butane (Tracz et al., 1990). Similar phenomena, probably due to a wall effect in the metallic reactor tubes, have been observed in noncatalytic processes such as steam cracking (Kopinke et al., 1993; Albright and Marek, 1988). We investigated the variables affecting the kinetics of carbon formation in the cyclopentane hydrogenolysis on nickel (Duprez et al., 1990) and on nickel-potassium catalysts (Demicheli et al., 1994) in the 300-500 °C temperature range. We showed that, at low temperature, only bidimensional carbon was formed at the surface of the nickel particles. Above a certain temperature, designated as TB, the filaments started to grow and the rate of carbon formation increased sharply. The value of TB depended on many factors (H2/cyclopentane ratio, particle size of nickel, metal-support interaction, ...). Apparently, a minimum carbon coverage at the nickel surface (C/Nis ∼ 7-8) was required for the filament growth to begin. The present paper deals with a study of the regeneration of nickel catalysts by carbon gasification in hydrogen. Several factors were investigated: catalyst composition and pretreatment, the morphology and the initial amount of carbon, and, lastly, the rate of hydrogen flow which proved to have a dramatic effect on the catalyst regeneration. Experimental Section 1. Catalyst Preparation. The support was a RhoˆnePoulenc SCS79 δ-alumina pretreated at 850 °C (92 m2 g-1; main impurities: Na2O, 640 ppm; CaO, 1130 ppm). The beads were crushed and sieved to 0.1-0.2 mm. The * Author to whom correspondence is addressed. Telephone: (33) 5 49 45 39 98. Fax: (33) 5 49 45 34 99. E-mail: duprez@ cri.univ-poitiers.fr. S0888-5885(96)00649-5 CCC: $14.00

catalysts were prepared by impregnation of the support with an aqueous solution of nickel nitrate, dried at 120 °C, and calcined at 400 °C in an air flow for 6 h (NiAC catalyst). An aliquot portion of the solid was subsequently treated in an air flow at 700 °C for 16 h in order to generate the NiO/Al2O3 interactions (NiAC700 catalyst). The intensity of the XRD peaks corresponding to NiAl2O4 increased, while the characteristic peaks of NiO decreased (though remaining visible). K-doped catalysts were prepared by a similar procedure. The NiO/Al2O3 sample, calcined at 400 °C, was reimpregnated with a 2 M solution of K2CO3, dried, and calcined again at 400 °C (NiKAC sample). Part of this catalyst was treated in air at 700 °C to obtain the NiKAC700 catalyst. 2. Catalyst Reduction. In previous works (Duprez et al., 1990; Demicheli et al., 1994), the catalysts were reduced at 500 °C. As we planned to regenerate coked catalysts by carbon gasification under a flow of H2 at 800 °C, the fresh samples were reduced at that temperature. For comparison, preliminary experiments were carried out, first on catalysts reduced at 500 °C and then on catalysts reduced at 800 °C. The catalyst samples (0.3 g) were heated at 4 °C min-1 under a hydrogen flow (40 cm3 min-1) up to 500 or 800 °C. They were maintained at these temperatures for 12 h and then cooled down to the reaction temperature. 3. Catalyst Characterization. The reduced catalysts were characterized by H2 chemisorption at room temperature (nickel dispersion) and by O2 uptake at 500 °C (degree of nickel reduction). These characterizations were carried out in a dynamic chromatographic reactor described elsewhere (Duprez et al., 1986). We showed that the oxygen uptake at 500 °C, corresponding to the reoxidation of Ni0 to NiO, led to an overestimation of the degree of nickel reduction by about 10% (oxygen excess in NiO and/or oxygen spillover on alumina). Besides, owing to the low pressure of H2 used in this pulsed technique, hydrogen coverage corresponded to about H/NiS ≈ 0.5. The catalyst characteristics were therefore calculated with the equations

NiR 0.9OC % R ) 100 ) 100 NiT NiT

(1)

NiS 2HC % D ) 100 ) 100 NiR NiR

(2)

where % R and % D are the percentages of nickel © 1997 American Chemical Society

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 3181 Table 1. Characteristics of the Fresh Catalysts calcined catalyst

Ni (wt %)

NiAC NiAC700 NiKAC NiKAC700

5.92 5.93 7.28 7.71

reduced at 500 °C

K (wt %)

SBET (m2 g-1)

%D

4.19 4.23

85 71 58 51

19.3 10.5 19.3 7.3

%R

dc (nm)

dx (nm)

%D

%R

Am (m2 Ni g-1)

dc (nm)

dx (nm)

95 32 99 45

7.3 1.4 13.3 1.8

5.1 9.3 3.7 13.3

5.5 7.5 5.4 5.7

7.2 8.9 3.4 2.6

94 99 100 95

2.6 3.6 1.9 1.3

13.6 11.0 28.8 38.2

12.5 9.5 27.4 35.7

reduction and of nickel dispersion, and NiT, NiR, and NiS, the amounts of nickel (total, reduced, and surface, respectively). The values of OC, oxygen uptake at 500 °C, and of HC, hydrogen chemisorbed at room temperature, were taken in µmol atom g-1. In accordance with Bartholomew and Pannell (1980), the metal surface area of nickel Am was calculated assuming a surface area of 6.77 Å2 per nickel atom, while the mean particle size was obtained by

dc (Å) )

971 %D

reduced at 800 °C

Am (m2 Ni g-1)

(3)

The values of dc were compared to those of dx, estimated by XRD for the 〈012〉 direction of NiO. XRD spectra were obtained in a D500 Siemens spectrometer after reoxidation at 500 °C of the reduced catalysts. 4. Coking and Test Reaction. The catalyst sample, reduced at TR (500 or 800 °C), was cooled down to the coking temperature, TC. Pure hydrogen was then replaced by a cyclopentane/H2 mixture for 1 h, after which the cyclopentane flow was interrupted. The catalyst was cooled down to a much lower temperature (200 < T < 250 °C), and cyclopentane (CPA) was again injected for 3 h. CPA hydrogenolysis at low temperature was used as a test reaction for the metallic activity. No carbon deposit was formed within this temperature range. The products (methane, ethane, ethylene, propane, isobutane, butane, n-pentane, cyclopentene, and cyclopentadiene) were analyzed by GC on a Al2O3 capillary column (25 m, i.d. ) 0.32 mm) with H2 as a carrier gas. After the test reaction, the catalyst was cooled down to room temperature under a flow of hydrogen and a portion of the sample was transferred into the TPO apparatus for coke analysis. The standard conditions of this study (reduction, coking, and hydrogenolysis) were as follows: catalyst weight, m ) 0.3 g; hydrogen flow rate, FH ) 0.0425 mol h-1; CPA flow rate, FC ) 0.0243 mol h-1. In further experiments, we showed that the sequence of hydrogenolysis at low temperature did not alter the amount and the reactivity of the coke deposited at TC. 5. Regeneration. After reduction and coking, the catalyst was regenerated in a flow of H2. The sample was heated up to 800 °C at 4 °C min-1 and maintained at that temperature for a time varying from 15 to 310 min. Certain experiments were performed with a prolonged time of gasification (24 h) and at a higher temperature (900 °C). Just after having substituted pure H2 for the CPA/H2 mixture, the cyclopentane disappeared rapidly from the effluents. Methane was the only product analyzed during the temperature rise. The regeneration was investigated at different rates of H2 flow: 0.0425 mol h-1 (low flow rate, the same rate as for the reduction and the coking), 0.129 mol h-1 (medium flow rate), and 0.258 mol h-1 (high flow rate). 6. Characterization of Coked and Regenerated Catalysts. Coke TPO was performed throughout this study to determine the amount and the reactivity of the different carbon species deposited on coked and regen-

Table 2. Activity of the Fresh Catalysts for Cyclopentane Hydrogenolysis at 250 °C (CPA/H2 Molar Ratio ) 1.75; FC ) 0.0243 mol h-1; mcat ) 0.3 g) a0 (mmol h-1 g-1)

a0 (mmol h-1 m-2)

catalyst

TR ) 500 °C

TR ) 800 °C

TR ) 500 °C

TR ) 800 °C

NiAC NiAC700 NiKAC NiKAC700

72 20 12 1.1

181 127 40 13

9.9 14.3 0.9 0.7

70 35 20 10

erated catalysts. The TPO apparatus and the procedure have been described elsewhere (Barbier et al., 1985). In this study a 1% O2/He mixture was used and the temperature ramp was 6 °C min-1. The sample weight was routinely 0.025 g. Most experiments were duplicated with different weights depending on the amount of coke in the catalyst. The formation of carbon filaments was checked by SEM in a JEOL 35 CF electron microscope (Duprez et al., 1990). TEM pictures and EDX spectra of the catalysts at different stages of regeneration were also obtained in a Philips CM 120 electron microscope. Results 1. Characteristics of the Fresh Catalysts. The results are reported in Table 1. For the catalysts reduced at 500 °C, a significant effect of the calcination at 700 °C can be observed: the degree of nickel reduction decreases by a factor 2-3, accompanied by a sintering of the reducible phase of nickel. The presence of potassium improves the NiO reducibility but cannot prevent its sintering at 700 °C. Reduction at 800 °C leads to profound changes in the catalyst characteristics: the major part of the nickel aluminate can be reduced into Ni0, which it is not so at 500 °C. Nevertheless, reduction at 800 °C leads to a significant sintering of the nickel particles, particularly marked on K-containing catalysts. The activity of the fresh catalysts for cyclopentane hydrogenolysis at 250 °C is given in Table 2. As stated in a previous work (Duprez et al., 1991), potassium appears to be a poison for the hydrogenolysis reaction. This effect is specific of a high potassium content (K/Ni > 0.5). The contrary occurred for low potassium contents (K/Ni < 0.1) (Demicheli et al., 1994). By reducing the catalysts at 800 °C instead of at 500 °C, the hydrogenolysis activity increases by a factor somewhere between 2.5 and 12. As demonstrated by the change of the activity of NiAC and NiAC700 with TR, this is mainly due to an increase of the degree of nickel reduction and to a lesser extent to a sintering of nickel particles. 2. Kinetics of Coking. The amount of carbon deposited during the first hour was measured at different temperatures for each catalyst reduced at 500 or 800 °C. The results are presented in Arrhenius coordinates in Figure 1. The two kinetic regimes can be observed for all the catalysts: bidimensional carbon is

3182 Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997

than the activation energy of filament growth (≈140 kJ mol-1; Figuereido et al., 1988). The weight of carbon QC can be calculated by QC ) χnl, where χ is the specific weight of carbon in a filament (per unit of length), n, the number of filaments, and l, the mean length of the filaments. The rate of coking is then

RC )

Figure 1. Temperature dependence of coke formation on nickel catalysts reduced at 500 °C or at 800 °C. Coking conditions: H2/ cyclopentane molar ratio ) 1.75; molar flow rate of cyclopentane ) 0.0243 mol h-1; mcat ) 0.3 g.

deposited on the nickel particles at T < TB and filamentary carbon at T > TB. The appearance of filaments above TB was confirmed by SEM. Micrographs like those shown in our previous papers (Duprez et al., 1990; Demicheli et al., 1994) were obtained for each catalyst. Reducing the catalysts at 800 °C instead of at 500 °C increases the rate of carbon filament formation, this effect being more marked on catalysts calcined at 700 °C. This is due to two conjugated factors: the catalysts reduced at 800 °C are better reduced (particularly in the AC700 series) and have larger nickel particles. Kock et al. (1985) mentioned that all the carbon filaments observed by electron microscopy had a diameter of over 10 nm, while magnetic methods indicated that a fraction of the nickel was present in smaller particles. Recently, we showed that a 2% Ni/Al2O3 catalyst with a mean particle size of 3 nm could form only a very small amount of filamentous carbon (Duprez et al., 1990). It seems therefore that a minimum diameter of about 5 nm is required for the nickel particles to generate carbon filaments. The degree of nickel reduction can also play a decisive role in filament growth. The case of NiAC700 is quite instructive. When reduced at 800 °C instead of at 500 °C, this catalyst has a percentage of Ni reduction of 99% (instead of 32%) but the particle size remains practically unchanged (dc ) 10 ( 1 nm). In both cases, filaments appear at very similar temperatures (421 °C for NiAl700 reduced at 800 °C and 426 °C for the sample reduced at 500 °C), but the rate of filament growth on the wellreduced sample increases more rapidly above TB. The apparent activation energy Ea calculated between TB and 480 °C is 310 kJ mol-1 for TR ) 800 °C instead of 230 kJ mol-1 for TR ) 500 °C so that the carbon filaments formed at 480 °C are 7 times more abundant on the well-reduced sample than on the catalyst reduced at 500 °C. These values of Ea are significantly higher

dQC dn dl )χn +l dt dt dt

[

]

(4)

Filament growth (term in dl/dt) obeys a mechanism in which the diffusion of carbon through the metal particles is the rate-determining step with Ea ≈ 140 kJ mol-1. When TC increases, new filaments are formed so that RC increases more rapidly than the rate of filament growth (additional term in dn/dt). This explains why Ea can be higher than 140 kJ mol-1. The presence of unreduced species of nickel can retard the appearance of new filaments, which results in a lower activation energy for the catalyst reduced at 500 °C. Figure 1 shows that the amount of filamentous carbon reaches a plateau at TA ≈ 480 °C for most catalysts. Above TA, the rate of coking increases extremely slowly compared to what can be observed in the (TB, TA) temperature range. This confers a S-shape on the curves of Figure 1. We shall see that the filaments formed at T g TA and those formed between TB and TA have a very different regeneration behavior. 3. Characteristics of the Coked Catalysts. The catalysts were characterized by their rate of cyclopentane hydrogenolysis at 250 °C. Four temperatures of coking were selected: 420 °C, a temperature at which only bidimensional carbon was formed, 440 °C, 460 °C, and 480 °C, with increasing density and size of filaments. Table 3 gives the amount of coke deposited on each catalyst as well as the normalized activity for hydrogenolysis (a/a0, a and a0 being the activities of the coked and of the fresh catalyst, respectively). The coke deposited at 420 °C is extremely toxic for the metallic function. Potassium decreases the toxicity of the carbon deposit, the catalysts being deactivated by a factor comprised between 8 (NiKAC700) and 140 (NiAC700). The coke deposited at the highest temperature (480 °C) is constituted essentially of carbon filaments. At this temperature, the rate of filament growth ceases practically to increase and the catalyst recovers most of its initial activity in hydrogenolysis: this is the TA point mentioned in section 2. There are some exceptions for the NiAC700 catalyst reduced at 500 °C and for NiKAC700 whatever its reduction temperature. Apparently, these catalysts reach their A point above 480 °C (Figure 1). These particular cases will not be examined any further in this paper. The catalyst behavior at the mean coking temperatures, 440 and 460 °C, deserves attention: the recovering of the metallic activity cannot be observed when the carbon filaments are already formed. This behavior is particularly evident with the unpromoted catalysts. Owing to the presence of potassium, the breakpoints in the variations of a/a0 vs Tc tend to be smoothed off. The coke content of the catalysts was determined by TPO. This technique also allows one to evaluate the reactivity in O2 of the different carbon species deposited on the catalyst. Figure 2 shows the TPO profiles of NiAC and of NiKAC700 for different temperatures of coking. Three types of carbon are shown on the profiles of NiAC: type I (Tmax ) 360 °C) is assigned to bidimensional carbon, while type III (Tmax ) 620-700 °C) is attributed to graphitic carbon layers forming the

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 3183 Table 3. Amount of Coke (wt %) and Normalized Activity of the Coked Catalysts for Cyclopentane Hydrogenolysis at 250 °C (1 h Coking at TC) TC ) 420 °C catalyst NiAC NiAC700 NiKAC NiKAC700

TC ) 440 °C

TC ) 460 °C

TC ) 480 °C

TR (°C)

%C

a/a0

%C

a/a0

%C

a/a0

%C

a/a0

500 800 500 800 500 800 500 800

2.0 1.1 0.3 0.3 1.5 1.2 0.1 0.4

0.054 0.049 0.011 0.007 0.11 0.16 0.077 0.12

2.7 5.5 1.0 1.8 1.6 3.5 0.9 3.8

0.059 0.055 0.013 0.008 0.13 0.18 0.089 0.15

30 40 5.5 13 13.5 33 4.5 19

0.19 0.15 0.017 0.017 0.27 0.22 0.17 0.17

40 54 15 45 30 52 13 27

0.53 0.41 0.31 0.45 0.47 0.45 0.23 0.31

Table 4. Amount of Coke (wt %) and Normalized Activity for Cyclopentane Hydrogenolysis at 250 °C of Coked and Regenerated Catalysts (800 °C; FH ) 0.0425 mol H2 h-1) catalysta NiAC NiAC700 NiKAC NiKAC700 a

TC (°C)

coked % C a/a0

440 460 440 460 440 460 440 460

5.5 40 1.8 13 3.5 33 3.8 19

0.05 0.15 0.01 0.02 0.18 0.22 0.15 0.17

reg. for 15 min %C a/a0 1.2 0.9 0.8 5.3 1.7 1.5 1.0 8.7

0.64 0.95 0.72 0.67 0.50 1.5 1.5 7.4

reg. for 30 min %C a/a0 0.20 0.10 0.05 0.06 0.14 0.45 0.49 1.4

0.95 1.0 0.86 0.96 2.4 2.2 4.3 7.4

The catalysts were reduced at 800 °C before coking.

Figure 2. TPO profiles of the coke deposited on NiAC (top) and on NiKAC700 (bottom) at (a) 420, (b) 460, and (c) 480 °C.

filaments. Carbon deposited at 420 °C is constituted of type I carbon only, while coking at 460 and 480 °C leads essentially to type III carbon. We could not find out the origin of type II carbon (Tmax ≈ 430 °C). Most likely, this species comes from secondary carbon formed at the filament surface by cyclopentane cracking on the graphitic tube or on nickel atoms detached from the parent particle. The presence of potassium changes the TPO profiles of coke. Filamentous carbon oxidizes at a much lower temperature (Tmax ≈ 460 °C) so that type II and III carbons can no longer be discriminated. A similar behavior has been previously reported (Demicheli et al., 1994) and linked to the fact that some potassium has been inserted between the nickel particle and the filament, thus modifying the rate of carbon oxidation at the nickel/carbon interface. 4. Regeneration of Catalysts Coked at T < 480 °C. The carbon deposit formed at 420 °C is extremely toxic for the metallic function. However, this carbon is easily eliminated by hydrogenation at a moderate temperature (500-600 °C) and the hydrogenolysis activity is then fully regenerated. In what follows, we shall only consider the regeneration behavior of the catalysts which have a significant density of carbon

Figure 3. TPO profiles of the coke deposited on NiAC700: (a) catalyst coked at 460 °C for 1 h; (b) catalyst regenerated under a flow of H2 at 800 °C for 15 min; (c) id. regenerated for 30 min.

filaments (TC ) 440 and 460 °C). The percentage of the remaining carbon and the normalized activity for CPA hydrogenolysis were determined (Table 4). A significant part of the carbon can be gasified by H2 in 15 min at 800 °C and the quasi-totality after a 30 min regeneration. Correlatively, the catalysts recover their initial hydrogenolysis activity and even surpass this activity for K-promoted catalysts. This could be attributed to a loss of potassium during the regeneration (about 50% of the initial content). It can be concluded that the coke formed within the (TB, TA) temperature range is easily gasified by H2 and that unpromoted catalysts recover their initial properties. The comparison of the TPO profiles of coked and regenerated catalysts (e.g., see Figure 3 for NiAC700) shows that the three types of carbon are gasified by H2. Type I and II carbons should be hydrogenated first. They remain however present on the catalyst as long as the graphitic carbon (type III) has not been fully gasified. This result suggests that some of type I or of type II carbon species can be reformed at 800 °C during

3184 Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997

Figure 4. SEM picture of the NiAC700 catalyst coked at 480 °C.

Figure 5. Kinetics of carbon gasification (H2, 800 °C) of the filaments deposited on NiAC at 480 °C. The legend gives the rates of H2 flow (mol h-1) used for the regeneration.

the gasification of the filaments. However, these species, resulting from methane cracking, are not as toxic as the carbon formed by coking at 420 °C in a CPA/H2 atmosphere. 5. Regeneration of the Catalysts Coked at 480 °C. At 480 °C, a very high density of carbon filaments can be observed (see SEM micrograph of NiAC700 on Figure 4). Magnified SEM pictures reveal that most of the filaments had lengths over 2 µm. We found that these carbon filaments formed at 480 °C could not be gasified as easily by H2 as those formed at lower temperatures. Furthermore, the rate of hydrogen flow was proven to be a critical parameter for the regeneration. The % C vs t curves recorded during the experiment carried out with NiAC (Figure 5) show there is a rapid loss of carbon during the first 15 min of regeneration, with a significant fraction of this carbon being gasified during heating up to 800 °C. At a low hydrogen flow rate, the rate of carbon gasification decreases rapidly and tends to zero after t ≈ 60 min, while there remains a high density of filament in the catalyst (Figure 6a). Prolonging the regeneration up to 1440 min or increasing the temperature to 900 °C did not change the residual carbon content. The situation is quite different for the regeneration of NiAC at a high H2 flow rate, which leads to a deep

Figure 6. TEM pictures of NiAC after coking at 480 °C and regeneration at 800 °C for 310 min under a flow of H2: (a) low flow rate; (b) high flow rate. The arrows indicate Ni particles detected by EDX.

gasification of the filaments (Figure 6b). A similar behavior can be observed with the three other catalysts (Table 5). Some slight differences are, however, to be noted: (i) the quasi-complete gasification of carbon is obtained at a lower flow rate with these catalysts than with NiAC; (ii) there exists an optimal rate of H2 flow in the 700 series: increasing QH may decrease the efficiency of the regeneration. Extremely high activities for cyclopentane hydrogenolysis are obtained when the catalysts are regenerated at a high rate of H2 flow (Table 5). They become more active than the fresh catalysts long before the carbon deposit is deeply gasified. Except

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 3185 Table 5. Amount of Coke (wt %) and Normalized Activity for Cyclopentane Hydrogenolysis at 250 °C of the Catalysts Coked at 480 °C and Regenerated at 800 °C under a Flow of H2 (Molar Flow Rate: QH) regeneration time (min) catalyst (initial % C) NiAC (54%) NiAC700 (45%) NiKAC (52%) NiKAC700 (27%)

15

QH (mol h-1)

%C

a/a0

0.0425 0.129 0.258 0.0425 0.129 0.258 0.0425 0.129 0.258 0.0425 0.129 0.258

29 35 27 27 22 25 31 22 29 22 14 18

1.25 1.00 1.04 0.76 1.06 1.06 1.27 1.10 1.10 1.15 1.30 1.15

120 % C a/a0 22 24 10 24 5 10 23 3 5 17 2 5

0.84 1.06 1.21 1.18 1.57 1.43 0.72 1.8 1.9 8.7 3.2 2.7

310 % C a/a0 21 24 3 24 1 7 23 2 2 16 1 4

0.79 1.12 1.37 1.11 2.48 2.16 0.62 3.0 3.0 8.7 6.7 5.4

for NiKAC700, the regeneration of the catalysts at low H2 flow rate leads to solids much less active in hydrogenolysis. Discussion Normalized Activity of Coked and Regenerated Catalysts. It is generally accepted that hydrogenolysis reactions are as structure-sensitive on Ni as they are on most group 8 metals (Rostrup-Nielsen, 1973; Nazimek and Ryczkowski, 1986). Nevertheless, only moderate effects can be observed for particles greater than 2-3 nm (Burke and Ko, 1989). Martin (1979) reported that ethane hydrogenolysis could be considered as a structure-insensitive reaction, on the condition that 〈111〉 planes (inactive in hydrogenolysis) were not considered. The results given in Table 2 show that the intrinsic rates of cyclopentane hydrogenolysis depend more on the degree of nickel reduction and on the alkali content than on the particle size of nickel. During coking and regeneration the restructuration of Ni particles and the presence of bidimensional carbon deposits are important parameters which determine the hydrogenolysis activity. A four-step mechanism explaining filament growth on Ni was proposed by Baker et al. (1972): (i) adsorption and decomposition of the hydrocarbon molecule at the nickel surface, (ii) dissolution and diffusion of carbon through the metal particle, (iii) surface diffusion of carbon species, and (iv) precipitation of the carbon species at the rear face of the metal particle. According to Baker et al., the driving force for carbon diffusion would be a temperature gradient through the metal particle due to the endothermicity of the carbon deposition (+40.5 kJ mol-1). Many refinements have been brought to this model, and it is now accepted that (i) bulk or subsurface diffusion of carbon in the nickel particles is the main pathway for carbon transport and (ii) the concentration gradient can be the driving force for carbon diffusion (Rostrup-Nielsen, 1984; Kock et al., 1985; Alstrup, 1988). All TEM studies have shown that nickel particles reconstruct at the very beginning of filament growth. “Pear-shaped” or “coneshaped” particles were observed (Baker, 1972; Audier et al., 1980, 1981; Boellaard et al. 1985; Yang and Chen, 1989; Demicheli et al., 1994). Yang and Chen (1989) also showed that, most likely, (111) and (311) planes existed at the nickel/graphite interface and (002) planes at the gas/metal interface. From the above discussion, we can assume that, during filament growth, (100) and (110) basal planes of

Ni are exposed to gas while (111) planes are in contact with graphite. This configuration is favorable for a good activity for hydrogenolysis provided that the gas/metal interface remains free of any carbon deposit. A relevant finding of this study is the slow evolution of the normalized activity for hydrogenolysis when the coking temperature is increased from 420 to 480 °C. At 420 °C the metallic activity is killed by dimensional amorphous carbon, while at 480 °C all the catalysts, to a large extent, recover their initial activity. The results presented in section 3 show that the catalysts have formed carbon filaments at 440 and 460 °C, and yet they have not recovered their initial activity for hydrogenolysis. In parallel, regeneration studies (section 4) show that these filaments formed at 440 and 460 °C are easily gasified by H2 at 800 °C. All these results support the two-phase model of Baker et al. (1972). The filaments would be composed of an easily oxidizable core with a relatively resistant “skin”. In our case, the filaments formed at 440 and 460 °C would contain less graphitic carbon (the core in Baker’s model), while those formed at 480 °C would have a thick, highly graphitic skin. The surface state of nickel depends on the temperature of coking: at TC ) 440 and 460 °C, bidimensional, toxic carbon stays adsorbed on the metal during filament growth, while at TC ) 480 °C, the nickel surface is practically free of any carbon deposit. Effect of H2 Flow Rate on Carbon Gasification. We showed that a high H2 flow rate was required for the gasification of the filaments formed at 480 °C. To explain this behavior, two hypotheses can be proposed: (a) Chemical Effects. The gasification (eq 5) is a reversible reaction

C/Ni + 2H2 S CH4 (+CxHy) + Ni0

(5)

so that methane (or the small amounts of heavy products also formed) can reform carbon. This hypothesis is quite acceptable since K, the equilibrium constant for methane decomposition (reverse reaction of eq 5), is close to 60 at 800 °C. Assuming there is a mixture of 1 mol of H2 for a mol of CH4, K may be expressed as a function of the fraction R of methane decomposed into C and H2: 2

K)

(1 + 2aR) P [1 + a(1 + R)] a(1 - R)

(6)

Under our experimental conditions, the total pressure P ) 1 atm and a , 1, so

R ) 1 - 1/aK

(7)

There is virtually no carbon reformation if a < 1/K, i.e., a < 1.7%, while for a ) 3.4%, half the methane will be decomposed into carbon and H2. The increase of the H2 flow rate decreases the methane content in H2, which avoids the risk of reforming carbon. (b) Diffusional Effects. It is accepted that the gasification reaction is catalyzed by the nickel particle itself. Figuereido et al. (1988) concluded that growth and gasification occurred by similar mechanisms in which one of the steps involved the diffusion of carbon through the metal. Certain results of Figuereido et al. suggested that the rate-determining step in the nickel-catalyzed hydrogasification of carbon filaments could be the reaction between surface carbon and adsorbed hydrogen. This implies that there remains a significant carbon coverage at the nickel surface during gasification, which

3186 Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997

is not in agreement with the values of the normalized activities found for the regenerated catalysts (Table 5). In the cases where carbon filaments were not fully gasified at 800 °C, the activation energy was very low (no significant increase of the rate of gasification between 800 and 900 °C). Under these conditions, the reaction seems to be governed by a diffusional process. In the model of Figuereido et al., this process would be the carbon diffusion through the metal particle. However, another possibility would be the direct formation of methane at the nickel/filament interface. To explain the uniformity of the filament diameter, a morphological model with slippage of graphitic layers over one another has been reported (Boellaard et al., 1985). Numerous edge dislocations are then produced in the graphite stacking, which opens the carbon structure to gas diffusion. Though H2 can diffuse easily in the dislocations at the nickel/graphite interface, methane would remain confined close to its production sites, which reinforces the chemical effect discussed above. Increasing the hydrogen flow rate decreases the partial pressure of the methane surrounding the metal particles, which increases the flux of diffusion of CH4 from the metal/carbon interface to the gas phase. Conclusions Carbon filaments were formed on nickel catalysts during cyclopentane hydrogenolysis above 400 °C. Two critical temperatures, designated as TB and TA, were found in the kinetics of carbon formation: at TB (≈420 °C) the filaments started to grow, while at TA (≈480 °C) the rate of filament growth ceased to increase. At T < TB, only bidimensional carbon was formed at the nickel surface. This carbon species was very toxic for the metallic function: the activity for cyclopentane hydrogenolysis at 250 °C was 1 or 2 orders of magnitude lower in the coked than in the fresh catalysts. However, this toxic carbon was easily gasified by H2 between 500 and 600 °C and the catalysts were then fully regenerated. At TB < T < TA, the carbon deposit was composed essentially of filaments. However, the nickel particles located at the tip of these filaments remained covered with bidimensional carbon which conferred to the catalyst a low activity for hydrogenolysis at 250 °C. Filamentous carbon deposited within this temperature range was gasified easily by H2 at 800 °C. At T g TA, there was virtually no longer any carbon deposit at the nickel surface: the activity for hydrogenolysis at 250 °C was extremely high (sometimes higher than with the fresh catalysts). By contrast, there was a dense deposit of filamentous carbon, much more difficult to be gasified by H2 at 800 °C. The gasification of these well-graphitized filaments required a high H2 flow rate to be effective. Chemical effects (carbon reformation by CH4 decomposition at 800 °C) and diffusional effects (CH4 confinement in the graphite dislocations at the metal/filament interface) have been proposed to explain these results. Acknowledgment We are grateful to CNRS-Ecotech for its financial support (PICS Deactivation and Regeneration of Metal Catalysts).

Literature Cited Albright, L. F.; Marek, J. C. Coke Formation during Pyrolysis: Roles of Residence Time, Reactor Geometry, and Time Operation. Ind. Eng. Chem. Res. 1988, 27, 743. Alstrup, I. A New Model Explaining Carbon Filament on Nickel, Iron, and Ni-Cu Alloy Catalysts. J. Catal. 1988, 109, 241. Audier, M.; Coulon, M.; Oberlin, A. Relative Crystallographic Orientations of Carbon and Metal in a Filamentous Catalytic Carbon. Carbon 1980, 18, 73. Audier, M.; Oberlin, A.; Coulon, M. Crystallographic Orientations of Catalytic Particles in Filamentous Carbon. Case of Simple Conical Particles. J. Cryst. Growth 1981, 55, 549. Baker, R. T. K.; Barber, M. A.; Harris, P. S.; Feates, F. S.; Waite, R. J. Nucleation and Growth of Carbon Deposits from the Nickel-Catalyzed Decomposition of Acetylene. J. Catal. 1972, 26, 51. Barbier, J.; Churin, E.; Parera, J. M.; Rivie`re, J. Characterization of Coke by Hydrogen and Carbon Analysis. React. Kinet. Catal. Lett. 1985, 29, 323. Bartholomew, C. H. Carbon Deposition in Steam Reforming and Methanation. Catal. Rev.sSci. Eng. 1982, 24 (1), 67. Bartholomew, C. H.; Pannell, R. B. The Stoichiometry of Hydrogen and Carbon Monoxide Chemisorption on Alumina- and SilicaSupported Nickel. J. Catal. 1980, 65, 390. Bhattacharyya, A.; Chang, V. W. CO2 Reforming of Methane to Syngas: Deactivation Behavior of Nickel Aluminate Spinel Catalysts. In Catalyst Deactivation 1994; Delmon, B., Froment, G. F., Eds.; Studies in Surface Science and Catalysis; Elsevier: Amsterdam, 1994; Vol. 88, p 207. Boellaard, E.; De Bokx, P. K.; Kock, A. J. H. M.; Geus, J. W. The formation of Filamentous Carbon on Iron and Nickel Catalysts III. Morphology. J. Catal. 1985, 96, 481. Bradford, M. C. J.; Vannice, M. A. Catalytic Reforming of Methane with Carbon Dioxide over Nickel Catalysts I. Catalyst Characterization and Activity. Appl. Catal. A 1996, 142, 73. Burke, P. A.; Ko, E. I. Propane Hydrogenolysis over Supported Nickel Catalysts: Structural and Support Effects. J. Catal. 1989, 116, 230. Demicheli, M. C.; Duprez, D.; Barbier, J.; Ferretti, O. A.; Ponzi, E. N. Deactivation of Steam-Reforming Model Catalysts by Coke Formation. II. Promotion with Potassium and Effect of Water. J. Catal. 1994, 145, 437. Duprez, D.; Mendez, M.; Dalmon, J. A. Characterization of Nickel Catalysts by Dynamic Volumetry and Magnetic Methods. Appl. Catal. 1986, 21, 1. Duprez, D.; Demicheli, M. C.; Mare´cot, P.; Barbier, J.; Ferretti, O. A.; Ponzi, E.N. Deactivation of Steam-Reforming Model Catalysts by Coke Formation. I. Kinetics of the Formation of Filamentous Carbon in the Hydrogenolysis of Cyclopentane on Ni/Al2O3. J. Catal. 1990, 124, 324. Duprez, D.; Demicheli, M. C.; Barbier, J.; Mare´cot, P.; Feretti, O. A.; Ponzi, E. N. Effect of Potassium on the Catalytic Behavior of Coked Nickel Catalysts in Hydrogenation and Hydrogenolysis Reactions. In Catalyst Deactivation 1991; Bartholomew, C. H., Butt, J. B., Eds.; Studies in Surface Science and Catalysis; Elsevier: Amsterdam, 1991; Vol. 68, p 195. Figuereido, J. L.; Bernardo, C. A.; Chludzinski, J. J., Jr.; Baker, R. T. K. The Reversibility of Filamentous Growth and Gasification. J. Catal. 1988, 110, 127. Kim, M. S.; Rodriguez, N. M.; Baker, R. T. K. The Interaction of Hydrocarbons with Copper-Nickel and Nickel in the Formation of Carbon Filaments. J. Catal. 1991, 131, 60. Kock, A. J. H. M.; De Bokx, P. K.; Boelloard, E.; Klop, W.; Geus, J. W. The Formation of Filamentous Carbon on Iron and Nickel Catalysts. II. Mechanism. J. Catal. 1985, 96, 468. Kopinke, F.-D.; Zimmermann, G.; Reyniers, G. C.; Froment, G. F. Relative Rates of Coke Formation from Hydrocarbons in Steam Cracking of Naphtha. 2. Paraffins, Naphthenes, Mono-, Di-, and Cycloolefins, and Acetylene. Ind. Eng. Chem. Res. 1993, 32, 56. Martin, G. A. Influence of the Surface Structure on the Kinetics of Ethane Hydrogenolysis over Ni/SiO2 catalysts. J. Catal. 1979, 60, 452. McAllister, P.; Wolf, E. E. An Activation-Deactivation Model for Catalytic Deposition of Carbon. J. Catal. 1992, 138, 129. McCarthy, J. G.; Hou, P. Y.; Sheridan, D.; Wise, H. Reactivity of Surface Carbon on Nickel catalysts: Temperature-Programmed Surface Reaction with Hydrogen and Steam. ACS Symp. Ser. 1982, 202, 253.

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 3187 Nazimek, D.; Ryczkowski, J. Influence of the Crystallite Size of Nickel on the Course of the Hydrogenolysis of Propane and n-Butane over Ni/Al2O3 Catalysts. Appl. Catal. 1986, 26, 47. Pen˜a, J. A.; Herguido, J.; Guimon, C.; Monzon, A.; Santamaria, J. Hydrogenation of Acetylene over Ni/NiAl2O4 catalysts: Characterization, Coking and Reaction Studies. J. Catal. 1996, 159, 313. Rostrup-Nielsen, J. R. Activity of Nickel Catalysts for Steam Reforming of Hydrocarbons. J. Catal. 1973, 31, 173. Rostrup-Nielsen, J. R. Steam Reforming Catalysts; Teknisk Forlag: Copenhagen, 1975. Rostrup-Nielsen, J. R. Catalytic Steam Reforming. In Catalysis, Science and Technology; Anderson, J. R., Boudard M., Eds.; Springer-Verlag: Berlin, 1984; Vol. 5, p 1.

Tracz, E.; Scholz, R.; Borowiecki, T. High-resolution electron microscopy study of the carbon deposit morphology on nickel catalysts. Appl. Catal. 1990, 66, 133. Yang, R. T.; Chen, J. P. Mechanism of Carbon Filament Growth on Metal Catalysts. J. Catal. 1989, 115, 52.

Received for review October 15, 1996 Revised manuscript received February 3, 1997 Accepted February 10, 1997X IE9606496 X Abstract published in Advance ACS Abstracts, June 15, 1997.