Journal of Molecular Catalysis, 25 (1984) 253 - 262 253

THE EFFECT OF LANTHANUM ADDITIVES ON THE CATALYTIC ACTIVITIES OF Ni--Al,Os COPRECIPITATED CATALYSTS FOR THE METHANATION OF CARBON MONOXIDE

MICHAEL R. GELSTHORPE, KAM BUN MOK, JULIAN R. H. ROSS*?

School of Chemistry, University of Bradford, Bradford BD7 1DP (U.K.)

and RODNEY M. SAMBROOK

Dyson Refractories Ltd., 381 Fulwood Road, Sheffield SlO 3GB (U.K.)

Summary

A nickel-alumina coprecipitate, which on calcination and reduction under suitable conditions gave an active and stable catalyst for the methana- tion of carbon monoxide, was used to prepare a series of doped samples by impregnation with the nitrates of various cations. After calcination and reduction, these samples were examined for their activities in the methana- tion of carbon monoxide using differential scanning calorimetry (DSC). The poisoning effect of the Group I alkaline metals increased in the order:

Li< Na< K< Cs,

while the ions of the Group II cations tested had no major effect. Cu had a poisoning effect similar to that of Na, while the decrease in activity with Cr was smaller. However, La and Ce both gave slight increases in activity. The promotion by La was further investigated by preparing several series of cata- lysts with variable La contents, using both impregnation and coprecipitation techniques. An optimum activity for all these series was found at about 1 mol.% La; above this concentration, the activity decreased. This decrease was particularly marked for the impregnated samples, although it was less when the samples had been aged at high temperatures. It is suggested that low levels of La (probably present as La203) promote the active metallic sites on the catalyst surface but that higher levels block these sites; for the impreg- nated samples treated at higher temperatures, some of the ,La can diffuse away from the surface to give more homogeneous materials akin to those derived from the coprecipitates. The possible nature of the active site is discussed.

*Current address: Department of Chemical Technology, Twente University of Technology, P.O. Box 217,750O AE Enschede, The Netherlands

*Author to whom correspondence should be addressed.

0304-5102/84/$3.00 @ Elsevier Sequoia/Printed in The Netherlands

254

Introduction

When the methanation of carbon monoxide-rich gases is carried out in a tubular reactor, there is a large temperature rise in a very small fraction of the bed. In typical plants of the type being tested currently for the methana- tion of such gases produced in coal gasification or in the so-called ‘Adam and Eve Process’, the inlet temperature of the bed is about 350 “C and the temperature rises over a smaIl section of the bed to 600 “C (with circulation of the product gas) or 800 “C (without circulation) [ 13. Much of the re- mainder of the bed is initially unused, but must resist excessive sintering under the hydrothermal reducing conditions of the gas leaving the reaction zone (particularly HZ + H,O); when the reaction zone moves through the bed (due, for example, to sintering of the catalyst or poisoning by S or C) the previously unused catalyst must still have sufficiently high activity to bring about the reaction at the lower temperatures which it now encounters.

The majority of the catalysts which have been tested for use under these reaction conditions are nickel-based. One such catalyst which is partic- ularly active, as well as being stable up to temperatures around 600 “C, is that formed from nickel and aluminium salts by coprecipitation techniques. Previous work in Bradford and Delft, [2 - 41, and parallel work in the laboratories of British Gas [5] and elsewhere, has established that the co- precipitated catalyst owes much of its activity and stability to the nature of the precursor, a mixed hydroxide phase of the pyroaurite structure, one of the so-called Feitknecht compounds [23. It appears that much of the dis- order of the constituent Ni2+ and A13+ ions built into the pyroaurite struc- ture during coprecipitation is retained in the final, reduced form of the catalyst [3,6]. The pyroaurite structure can tolerate a variation in the Ni2*/ A13+ ratio, and a single phase forms as long as this ratio is between two and three.

A number of parameters can affect the activities of the catalysts, partic- ularly the pH of prep~ation, the temperatures of calcination and reduction, and the way in which these steps are carried out [ 5,7, $1. As long as the pH of precipitation is at or above about 7.0, then the pyroaurite phase contains predominantly carbonate ions [ 21. The ‘interlayer’ of its structure contains CO; and OH- ions as well as molecular water. Sodium ions present during precipitation with sodium carbonate at pHs greater than 7.0 may be trapped in the gel-like structure of the precipitate; this sodium, which is detrimental to the activity of the catalyst, can be removed almost completely by wash- ing with water only after destroying the gel-like structure on drying at -120 “C ]9]_

The effect of foreign ions other than sodium on the activities of these catalysts has not been studied systematically. This paper therefore describes the results of a systematic study of the effect on the methanation activities, under standard conditions, of the materials formed by adding a number of metallic ions (- 1 mol%) to samples of two sodium-free precipitates (with Ni/Al ratios of 2.5). The results showed that only lithium and cerium

255

gave any appreciable improvement in the activities. Further work, also de- scribed, was therefore carried out to show the effect on the activities of La content and of the method of addition of the lanthanum. It has been shown elsewhere that the presence of lanthanum is also beneficial to the properties of a series of catalysts prepared from porous a-alumina matrices by an im- pregnation-deposition technique [lo].

Experimental

Catalyst preparation Two distinct methods of catalyst precipitation were used. The first used

sodium carbonate and the so-called ‘constant pH method’ (pH = 7.0), the solutions being maintained at 80 “C ~roughout; pr~ipitation was followed by washing with boiling deionised water, filtration, drying at 100 “C, and then further washing with boiling deionised water and a final drying step. Two solutions were used for the precipitation; the first was 1 M with respect to Ni(NO& and 0.4 M with respect to Al(NOa)s, and the second was 3.2 M with respect to Na&Os. These were added simultaneously at carefully con- trolled rates to a vessel containing water. The lanthanum-free precursor resulting was designated CS and was found to contain 0.08 wt.% NazO on i~ition after the second w~h~g; a number of l~~~~~on~n~g mate- rials (Series I) were prepared in a similar manner but using instead a nitrate- containing solution to which appropriate quantities of La(NO& had been added.

The second method involved the use of ammonium bicarbonate as precipitant. This salt was found to provide an effective buffering action (pH = 7) during the precipitation, and so the preparation was achieved by the addition of the mixed nitrate solution to a solution con~n~g a 10% excess of the bicarbonate. The temp~at~e during precipi~tion was main- tained at 20 “C (since -the ammonium bicarbonate partially decomposes at 80 “C); a small amount of the sparingly soluble nickel bicarbonate was probably formed at this temperature, as the solution has a slight blue colour, but this decomposes with evolution of CO* during heating to 80 “C. The final precipitate formed in this way is indistinguishable by X-ray diffraction from that formed using the first method, and thus we conclude that the nickel ions formed during the decomposition of the bicarbonate are relatively evenly d~tribu~ in the precipitate. The unpromoted precipitate was designated ClO. It contained within its gel-like structure some ammonium bicarbonate but the drying-rewashing procedure adopted for CS was found to be unnecessary, as the bicarbonate was decomposed completely during calcination: washed and unwashed samples gave rise to materials with iden- tical methanation activities. A number of lanthanum-containing samples were made in a similar way by adding appropriate quantities of LafNO& to the nitrate solution (Series II).

256

A series of doped samples was made from sample Cl0 (ammonium bicarbonate precipitation) by the impregnation of the dried precipitate with solutions of the nitrates of the following metals: Li, Na, K, Cs, Mg, Ca, Sr, Ba, Cu, Cr, La and Ce. The ‘incipient wetness’ technique was used, and con- centrations were chosen to give 1.2 mol% of the added cation in each sample. Further series of samples were prepared from C8 and Cl0 in the same manner by the addition of different quantities of La(NO,), (Series III and IV respectively).

Calcination, reduction and activity measurement It has been found that the activity measured for any single sample of

coprecipitated material could vary by a factor of up to two, depending on the conditions under which it was calcined and reduced. However, if several samples of the same preparation (or, indeed, of several repeat preparations of the same composition) were calcined and reduced together, the activities measured differed by less than 10%. Hence, to avoid differences from sample to sample due only to differences in pretreatment, each series of materials was calcined and then reduced and passivated together in a tube furnace. Up to about 20 samples, in the form of pressed disks (- 1 g), could be treated at one time. The decomposition of the precipitate (here loosely termed ‘calcination’) was carried out in argon (120 cm3 s-l, 1 atm); the furnace was heated to a temperature of 450 “C at 2.3 “C min-’ and held at that temper- ature for - 16 h. The furnace was then cooled to 200 “C and hydrogen was added to the gas stream so that the total flow-rate was then 180 cm3 s-l. The furnace was heated at 2.3 ‘C min-’ to 600 “C and maintained at that temper- ature for 4 h; it was then cooled to room temperature, the hydrogen flow was switched off and the argon was diverted through a water-saturator for 1 h, the water being used to passivate the samples. The methanation activi- ties of these samples (- 1.0 mg for each experiment) were then determined, in the temperature range 200 - 350 OC, using a DuPont DSC system (910 cell- base plus high-pressure cell used here at atmospheric pressure) with a gas of composition 12% CO, 36% H, and 52% Ar. (The use of the DSC for the mea- surement of the kinetics of the methanation reaction has been described elsewhere [ll - 131.) Repeat experiments using different samples of the same passivated sample were totally reproducible, indicating that the re- reduction step has no effect on the activity of the catalyst.

Active metallic surface area determination The metallic surface areas of a number of the catalysts studied were

determined by volumetric hydrogen chemisorption [13], but a more rapid procedure was also used for some of the results presented here: a small quantity of the passivated sample was re-reduced in a DuPont 951 Thermo- balance (re-reduction was complete at < 300 “C and the weight change indicated that approximately 10% of the nickel catalyst had been oxidised during passivation, confirming that re-reduction as carried out in the DSC was adequate) followed by chemisorption of CO from an argon + CO

257

f- 14%) stream at 100 “C. The amount adsorbed correlated well with the volume of hydrogen adsorbed in volumetric experiments with several differ- ent catalysts [13] and, although the magnitude was slightly greater, the results are therefore usable in calculations of specific activities, which will, as a consequence, be slightly lower than those based on hydrogen chemisorp- tion measurements. The CO chemisorbed could partially be desorbed on heating to - 300 “C and there was no evidence for the formation of nickel carbonyl with this pressure of CO during the short duration of the experi- ment (- 30 min); this is in contrast with the observation of Doesburg et al. (61 that a substantial quantity (- 90%) of the nickel of such a catalyst can be removed by CO at 100 “C and atmospheric pressure over a period of 10 days.

Mu terids Analytical and reagent grade chemicals, from various sources, were used

throughout and solutions were made up with deionised water. Gases (> 99.9% purity) were supplied by BOG Ltd.

Results and discussion

Effect of the tuition of various cations on ~ethunation activities ~eth~atio~ activity results similar to those shown in ref. 11 were ob-

tained for all samples; the main difference found was that the equipment used here gave a straight base-line [ 131, Apart from a small transitory devia- tion from the base-line at approximately 180 “C in the first experiment on a fresh sample (due presumably to re-reduction), the results of a sequence of equivalent experiments were completely reproducible: there was an expo- nentially increasing deviation from the base-line, starting at about 185 “C and cont~u~g until approx~ately 320 “C, when the rate of reaction was probably no longer chemically controlled. Using appropriate calibration factors, the results were converted to rates of reaction and hence Arrhenius plots such as those shown in Fig. 1 were obtained; we assume that the DSC operates as a differential reactor, and hence, for a constant reactant composi- tion, we are able to assume that the rate constant is directly proportional to the rate of reaction.

It was found that sample C8 was more active than Cl0 and that the activation energy, E, is the same, within experiment error, for both cata- lysts (E = 100 f 5 kJ moT’). The results for Cl0 are shown in Fig. 1. We therefore conclude that the two different preparative methods give slightly different dispersions, that involving Na&Os being superior; this was con- firmed in repeat preparations by both methods. Determinations of the metallic surface of both types of sample also confirmed the superior disper- sion of C&type preparations; the metallic area of CS calcined and reduced under the standard conditions was determined by hydrogen chemisorption to be 27.4 m2 g-’ while that of Cl0 was 17.1 m2 g-‘, and these results can

O,O’,b 21

d K/T

Fig. 1. Arrhenius plots of the methanation activity data for doped samples of Cl0 cined at 450 “C and reduced at 600 “C.

cab

be compared with values of 29.1 and 18.1 m2 g-l respectively, determined by CO adsorption.

Figure 1 also shows the effect on the methanation activity of adding Group I cations to Cl9 The activities of the doped catalysts decrease in the order Li > Na > K > Cs. The activation energies for the reaction do not change, at least within the accuracy of the experiments, following the addi- tion of these ions md also of Mg, Ca, Sr, Ba, Cu or Cr, for which the results we not shown in Fig. 1, for sake of clarity. The results for the samples doped with Mg, Ca, Sr and Ba are essentially identical to those for the undoped catalyst; Cu gives a decrease in activity roughly equivalent to that caused by Na, and Cr gives a smaller decrease, 130th La and Ce give an increase in activity, as shown in Fig. 1, and this is accompanied by the slight increase in the activation energy, to l15(+ 10) kJ mol-’ ~ there being some variation of the linearity of the Arrhenius plot with temperature.

It was shown previously [9] that sodium is a very marked poison for the methanation reaction, and that the activity of a similar coprecipitated material dropped by a factor of - 10 as the sodium content was increased to - 1 wt.% and then less markedly with further increase. It was shown that the sodium which could not be washed from samples prepared at pH - 10 was present in the calcinecl catalysts as NaNOJ, but that the reduced eata- lysts fat least at higher Na contents) contained sodium aluminate. The nickel

259

areas, as determined by hydrogen chemisorption, changed much less than the activities, and the results were thus interpreted as indicating that the methana- tion reaction requires more sites than does hydrogen chemisorption and that the presence of sodium affects the former sites more than the latter. Puxley et al. [5] have recently suggested that the unusual properties of coprecipi- tated catalysts such as those of the Ni-Al,Os system are due exclusively to paracrystallinity caused by the presence of A13” ions in the nickel crystal- lites. It is possible that a portion of the sodium added to the coprecipitate becomes associated with the aluminium ions at the surface of these para- crystals, affecting the methanation activities of the surrounding nickel atoms more than the hydrogen chemisorption properties. This is equivalent to suggesting that the sodium becomes associated with the aluminium ions of the nickel-rich phase of a two phase-system postulated previously by Alzamora et al. [ 31. The essential difference between the two models seems to be the presence in the second of an alumina-rich amorphous phase indistinguishable by X-ray diffraction. This phase is not thought of as being catalytically active, as the traces of nickel ions present would not be reduced under normal reduction conditions, but appears to be an essential feature in stabilising the structure of the catalysts [ 61. The present results indicate that all the Group I cations have an effect on the catalytic activities of sample Cl0 and that the effect increases, for equal numbers of added ions, with increasing ionic radii. It may therefore be that the greater effect of K or Cs compared with that of Na is due to an increase in the screening effect on the nickel atoms surrounding the surface aluminium ions with which the alkali metal cation becomes associated. However, it is also possible that the greater ionic radii cause a decrease in the solubility of the ions in the crystal- lites of the catalysts and that there is a corresponding increase in surface con- centration of the cations, even for constant all-over concentrations.

The effect of copper is to be expected, but for a different reason; the Cu and Ni are likely to be interchangeable in the structure at all stages in the preparation, but the surface of the alloy resulting upon reduction will be enriched in copper [14]. Calcium, strontium and barium are commonly added to steam-reforming catalysts, as they encourage carbon gasification; the present results show that they have little effect on the methanation activities. This could be because they have a greater affinity than do the Group I cations for the alumina-rich phase postulated by Kruissink et al., or that they have themselves some methanation activity. A third possibility, which seems most probable, is that, because of the small radii of these ions compared with those of the Group I cations, they have a more limited effect on the nickel atoms in the region of the aluminium ions in the surface of the nickel-rich phase.

The beneficial effect of lanthanum or cerium might be due either to promotion of the activity (or turnover number) of each active site or simply to an improvement in the dispersion of the catalyst. Further work on the effect of lanthanum was therefore carried out and is described in the next section.

f f

d

261

The addition of lanthanum ions appears to affect the activities in two opposing ways: at low concentrations, the lanthanum promotes the activity, as shown in the previous section, while, at higher concentrations, it poisons it. Doesburg et al. [15] have recently reported on the effect of lanthanum impregnation on the sintering behaviour of y-A1,03. They showed that the y-A1203 sinters by a surface diffusion mechanism, and that the lanthanum prevents this sintering by the formation of a lanthanum aluminate layer on the surface. In the Ni-Al-La materials discussed here, the La is therefore likely to become associated with the aluminium ions of both the Al- and Ni-rich phases. It may have several effects on the Ni-rich phase: (a) it may help to stabilise higher Ni dispersions; (b) it may promote the activity of the Ni sites in proximity to surface Al-La species; and (c) at higher concen- trations it may block the nickel sites. Lanthanum added during coprecipita- tion becomes more evenly distributed through the samples and thus its effect on the surface properties is rather less than that of La incorporated by im- pregnation .

To establish whether the promotional effect of the lanthanum is due to an improvement in nickel area or to an increase in the activity per unit area of Ni surface, the metallic surface areas were determined by both CO and Hz chemisorption for a number of samples. It was shown that the surface areas were generally decreased by the addition of lanthanum [13]. For example, sample Cl0 had a nickel surface area of 17.1 m2 g-l when determined by hydrogen chemisorption and 18.1 m2 g-l when determined by carbon monoxide chemisorption, and the equivalent coprecipitated sample with 0.5 mol% La had corresponding areas of 14.7 and 16.1 m2 g-’ respectively; however, the activity per gram was almost twice as high for the latter sample than for the former (Fig. 2, Series II). Hence we conclude that the lanthanum promotes the activities of the nickel sites (i.e. improves the turnover num- bers) of the types of catalyst studied here rather than stabilising a larger number of sites. It also seems to improve the stabilities of the resultant catalysts; further work is in progress on this point.

Acknowledgements

Thanks are due to Dyson Refractories Ltd. for their sponsorship of this work and for permission to publish this paper.

References

1 B. Hijhlein, R. Menzer and J. Range, Appl. Catal., I (1981) 125. 2 E. C. Kruissink, L. L. van Reijen and J. R. H. Ross, J. Chem. Sot., Faraday Trans. Z,

77 (1981) 649. 3 L. E. Alzamora, J. R. H. Ross, E. C. Kruissink and L. L. van Reijen, J. Chem. Sot.,

Faraday Trans. I, 77 (1981) 665. 4 E. C. Kruissink, L. E. Alzamora, S. Orr, E. B. M. Doesburg, L. L. van Reijen, J. R. H.

5

6

7

8

9

10

11 12 13 14 15

Ross and G. van Veen, in B. Delmon, P. Grange, P. Jacobs and G. Poncelet (eds.), Preparation of Catalysts II, Elsevier, Amsterdam, 1979, p. 143. D. C. Puxley, I. J. Kitchener, C. Komodromos and N. D. Parkyns, in G. Poncelet, P. Grange and P. A. Jacobs (eds.), Preparation of Catalysts IZZ, Elsevier, Amsterdam, 1983, p. 237. E. B. M. Doesburg, G. Hakvoort, H. Schaper and L. L. van Reijen, Appl. Catal., 7 (1983) 85. G. van Veen, E. C. Kruissink, E. B. M. Doesburg, J. R. H. Ross and L. L. van Reijen, React. Kinet. Catal. Lett., 9 (1978) 143. E. B. M. Doesburg, S. Orr, J. R. H. Ross and L. L. van Reijen, J. Chem. Sot., Chem. Commun., (1977) 734; and unpublished results. E. C. Kruissink, H. L. Pelt, J. R. H. Ross and L. L. van Reijen, Appl. Catal., 1 (1981) 23. K. B. Mok, J. R. H. Ross and R. M. Sambrook, in G. Poncelet, P. Grange and P. A. Jacobs (eds.), Preparation of Catalysts ZZZ, Elsevier, Amsterdam, 1983, p. 291. T. Beecroft, A. W. Miller and J. R. H. Ross, J. Catal., 40 (1975) 281. G. Hakvoort and L. L. van Reijen, 7th Znt. Conf. Thermal Anal., Kingston, Ont., 1982. M. R. Gelsthorpe, K. B. Mok and J. R. H. Ross, to be published. See, for example, M. Araki and V. Ponec, J. Catal., 44 (1976) 439. H. Schaper, E. B. M. Doesburg and L. L. van Reijen, Appl. Catal., 7 (1983) 211.