Kinetics of the Decomposition and Hydrogen Reduction of Nitric Oxide on Niobium, Nickel and Platinum Filaments at High Temperatures and Low Pressures BY JOHN A. MORGAN AND ANDREW J. B. ROBERTSON* Department of Chemistry, King’s College, Strand, London WC2R 2LS Received 2nd June, 1977 NO decomposition kinetics were studied with ultra-high vacuum techniques on niobium, nickel and platinum wire filaments at high temperatures with pressures between -1 and 100 p N m-2. Niobium readily decomposed NO with first-order kinetics, nitrogen and oxygen being deposited on the surface, resulting in a progressive loss of filament activity. Nitrogen was desorbed from nickel, which otherwise behaved similarly to niobium. Excess H2 affected the kinetics of NO reaction on these metals mainly by reaction with and subsequent removal of surface contaminants rather than by a direct reaction with NO.Platinum did not observably decompose NO at low pressures except in the presence of excess Hz. Above -1300 K the rate appeared to be determined by Hz atomisation, but at lower temperatures a reaction between H2 and NO may have been occurring. The rate-limiting step for this reaction may be the decomposition of H2 and addition of H atoms to adsorbed NO. The pre-exponential term for this reaction indicated that the transition state was mobile on the surface. The decomposition of NO to N2 and O2 and its reduction by H2 to form N2 and M20 are both thermodynamically favourable processes, even at 1000 K and for pressures in the region of 1 , ~ N r n - ~ . However, these reactions are slow in the absence of a catalyst.Owing to the occurrence of NO as an important pollutant in automobile exhausts, in recent years a great deal of interest has been directed at research into the catalytic decomposition and reduction of NO by a variety of catalysts. Reviews by Shelef and Kummer and Dwyer have described such work and its applicability to the control of automobile emissions. The scale of interest in NO decompositions may be judged by the fact that Winter has studied the decomposition on forty metal oxides. However, apart from a few papers such as the adsorption studies by Yates and made^,^ work in this field has been performed in conventional high-vacuum conditions with NO pressures usually above 10 N m-? Research of a more fundamental nature on the interaction of NO with clean surfaces, involving the use of lower pressures and ultra-high vacuum (u.h.v.) techniques, is lacking.The present paper describes the investigation of the kinetics of NO decomposition and reduction by H2 on niobium, nickel and platinum-wire filaments at high tempera- tures. Ultra-high vacuum techniques were used and NO pressures were typically between 1 and 100 ,uN m-2. EXPERIMENTAL APPARATUS A Pyrex glass flow system was constructed, incorporating greaseless stopcocks and mercury cut-offs. The reactant gases passed from separate gas-handling systems into the reaction vessel through calibrated, fine, capillary leaks. The reaction vessel was a 500 cm3 Pyrex bulb in which the filament was suspended in the form of a coiled loop, lashed to 21 12 12 NITRIC OXIDE ON NIOBIUM, NICKEL A N D PLATINUM tungsten leads by fine copper wire.Residual u.1i.v. pressures were measured with a Bayard- Alpert type gauge (I.O.G. 12), while partial pressures of reacting gases were monitored with a small 180" mass spectrometer (V.G. Micromass 2) incorporating an LaB6 coated filament. The m a s spectrometer was calibrated with a h4cLeod gauge, which could be removed from the system when u.h.v. experiments were performed. H20 pressures were derived from relative sensitivity data supplied by the manufacturers. The reaction vessel was pumped by a large mercury diffusion pump, 50nm diameter mouth, which was preceded by two large refrigerated traps. The reaction vessel, inlet leaks and pressure measuring devices were baked by an oven at -6600 I(.The traps and tubing up to the mouth of the pump were baked with heating tapes. The system was normally baked overnight and then cooled in sections, starting at the pump. Residual pressures below 10 nN m-2 could be attained. MATERIALS NO (Air Products) had purity >99 %. Apart froin traces of N2, no impurities were detected by mass spectrometry. The NO was purified by a freeze-and-thaw method. Cylinder hydrogen was purified by diffusion through palladium. In studies with nickel and platinum, the gases before entry into the reaction vessel were passed through a trap cooled with solid C02 to remove mercury vapour, which might have had a contaminating effect on these catalysts. The 130 pm diameter nickel and 70 pm diameter platinum wires (Johnson-Matthey) were of spectroscopic purity.The 150 pm diameter niobium wire contained various bulk impurities listed previ~usly.~ Nickel and platinum were cleaned by degassing at high temperatures, followed by cycles of oxidation in % 100 puN n r 2 of O2 at GZ 1000 K and reduction at the same temperature in % 500 N m-2 of pure HL. Oxidation was omitted from the cleaning procedure for niobium. All filaments were thoroughly degassed at high temperatures prior to use in experiments. The filaments were z 30 cm long. REACTION RATES When reactant gases reached steady pressures in the reaction vessel, reaction rates at a given filament temperature could be calculated, with a knowledge of the appropriate pumping speeds, from the decreases in reactant gas pressures or increases in product gas pressures which had occurred through reaction.6 This assumed that rates were negligible with the filament at room temperature compared with rates at the elevated filament temperatures studied.Reaction orders in NO or H2 were determined by the differential method and by the isolation technique. First-order rates were expressed in terms of the probability P of reaction of a molecule at a single collision with the catalytic surface.6 The catalytic areas were assumed to be equal to the geometric areas of the filaments. P was expressed as Y = B exp (-EIRT). FILAMENT TEMPERATURES Temperatures above red heat were determined by optical pyrometry, using average values of the spectral emissivities, at I. = 0.65 pm, obtained by Jain et al.7* for nickel and platinurn. The spectral emissivity of niobium (0.36) had been determined experimentally, as previously described5 Temperatures below red heat were interpolated from plots of the variation of filament resistance or current with filament temperatures, obtained by using values at incandescent temperatures, 373 K and room temperature.Corrections to filament teinpera- tures for heat losses, including conduction of heat to the leads, were calmlated to be of negligible importance. RESULTS NIOBIUM DECOMPOSITION Clean new filaments were very reactive to NO. No gaseous products of decoin- position were observed at temperatures below ~ 2 0 0 0 K. At ~ 2 0 0 0 I< N2 and NJ. A. MORGAN AND A . J. B. ROBERTSON 213 atoms were desorbed.Although NO appeared to be readily chemisorbed, probably ciissociatively, the product surface species were too strongly bound to niobium to be desorbed, except at the highest temperatures. Reaction rates for NO disappearance from the gas phase were calculated from changes in NO molecular-ion peak heights and were first order in NO for pressures up to xl mNm-2. At higher pressures and for low temperatures, rates of NO reaction decreased with time until the filament eventually became inactive. Decays of activity were more rapid for higher pressures and lower temperatures. Filament activity could be restored, to some extent, by heating the filament at a higher temperature in VQCUO or in NO, or by reducing the NO pressure. Even after prolonged heating at temperatures above 2200 K in uacuo, filaments became less active with repeated use.Very probably the products of reaction not removed by desorption diffused into the filament interior and the bulk and surface concentration of these Y 4 6 8 10 12 10" KIT FIG. 1.-Series of Arrhenius plots for NO decomposition on a niobium filament. Pressures before reaction occurred were, : run 1, 53 pN m-2 (0) ; run 2, 40 pN m-' (0) ; run 5, 2.6 pN m-2 (a) ; run 6, 4 p N m-2 (a) ; run 8, 4 pN m-2 ( x ). Runs 3, 4 and 7 are omitted for clarity. contaminants increased with use. As values of P for NO decomposition on clean niobium were close to unity, P values on used filaments probably reflected the uncontaminated surface areas of these filaments. Fig. 1 shows sets of Arrhenius plots for a filament, illustrating the loss of activity with use in consecutive runs.Results on clean filaments could be expressed as P = (1.8f0.6) exp (- 14f4 kJ mol-l/RT). The results show that all the filaments were extremely reactive towards NO. The rapid increase of P with temperature, on used filaments for temperatures above = 1900 K, corresponded with the occurrence of nitrogen desorption. This suggested that the nitrogen desorption increased the uncontaminated niobium surface area available for reaction. An Arrhenius plot for N, desorption, using rates calculated from m/e 14 peaks heights to avoid complications from possible CO production, gave a desorption energy of 420 & 40 kJ mol-I.214 NITRIC OXIDE ON NIOBIUM, NICKEL AND PLATINUM 0.0 -0.5 Qi t4 0 - - 1.0- REDUCTION The atomisation reaction of H2 was important at temperatures where H2 affected the reaction rate of NO, and rates of reaction of H2 could not be satisfactorily determined by monitoring the m/e 2 peak.Because of the ready adsorption on the reaction vessel walls of H20 produced during reaction, rates calculated from H 2 0 production were unreliable unless the reaction was allowed to continue for consider- able lengths of time until the H20 pressure stabilised. Therefore, NO reaction rates were measured using changes in the mJe 30 peak height in the presence of excess M2. On new filaments excess of H2 had virtually no discernible effect on the probability of NO reaction. If the probability of reaction in the absence of H2 is the probability of chemisorption of NO on niobium, then this would be the maximum probability of NO reaction, and the existence of a bimolecular surface reaction between chemi- sorbed NO and H2 or H atoms would not increase the value of P for NO.The lack of increase in P in the presence of excess H2 also indicated that a possible reaction on the walls of the vessel between adsorbed NO and H atoms desorbed from the filament was unimportant compared with reaction rates on the filament. - - 0 I I I I I 1 4 6 8 10 12 Ik FIG. 2.-Arrhenius plots illustrating the effect of excess H2 on NO reaction probabilities on a used niobium filament : NO pressures 4 p N rnd2 before reaction. H2 pressures were : no H2 (0) ; 2 mN m-2 (0) ; 14 mN m-2 (0). On used filaments the presence of H2 caused a marked enhancement in activity towards NO for temperatures between z I100 and 1400 K compared with rates in the absence of H2 (fig.2). Above 1400 K, P for NO in the presence of H2 became almost constant with increasing temperature at a value of about 0.6, and the activity was comparable to that observed on a clean unused filament. For relatively new filaments the reaction products when H2 was present were mixtures of NH3, H20 and some N2, and some N atoms at higher temperatures. NH3 was the predominant product, but its importance decreased with the contamination of a filament. The more used, and therefore contaminated, filaments produced H20 as the major product. Possibly, surface nitrogen was more easily removed from the surface of these filaments by diffusion to the interior.J .A . MORGAN AND A . J . B . ROBERTSON 21 5 NICKEL DECOMPOSITION On admission of NO to the reaction vessel containing a cleaned nickel filament at room temperature, no NO was at first observed by mass spectrometry. However, a large m/e28 peak due to N, was observed, which decreased in height to a steady value after about an hour, while the m/e peak due to NO had then appeared and reached a steady value. The mass spectrum of the gas in the reaction vessel then corresponded approximately to that of NO. It seems likely that NO readily decom- posed on clean nickeI at room temperature, producing N2 as a gaseous product and oxidising the nickel surface (see also Onchi and Farn~worth),~ The nickel probably became less active fur NO decomposition as the surface coverage of the inhibiting oxygen increased and so the nickel eventually became inactive after an hour or so.The nickel film thrown on the walls of the reaction vessel daring high temperature degassing was also able to decompose NO at room temperature. - 1.0 - 4 5 M - -1.5 - -2.0 - c I I I 10 12 14 16 104 KIT FIG. 3.-Arrhenius plots for NO decomposition on nickel : pressures before reaction 80 (0) ; 26 (0) ; 11 (0) : 4 PN m-2 to). Because of the above effects, it was not surprising that NO decomposition rates were difficult to reproduce unless low NO pressures in the region of 1 to 10 pN m-2 were used. Decays of activity, similar to those observed on niobium, were observed for nickel filaments. To overcome these problems, a special procedure for experi- ments with nickel filaments was developed.NO was admitted to the reaction vessel at a pressure higher than that required and the room temperature oxidation of the nickel surfaces allowed to continue until a steady NO pressure resulted. The NO pressure was then reduced to the pressure required for study and the filament was heated to 1100 K for = 10 or 15 min to allow surface oxygen on the filament to diffuse into the interior. After this treatment, the filament was ready for kinetic studies, whereas the still oxidised thin nickel film on the walls was inactive. When steady rates were observed at very low NO pressures the reaction was first order in NO and the only gaseous product detected was N2, with nearly one molecule of N2 produced per two molecules of NO decomposed (at 960 K).Arrhenius plots for NO decomposition are shown in fig. 3, indicating the reproducibility at lower NO21 6 NITRIC OXIDE ON NIOBIUM, NICKEL AND PLATINUM pressures and the loss of activity at higher pressures. Rates in a run were measured at increasing filament temperatures. The most reliable results for NO decomposition on a clean nickel surface, obtained at the lowest NO pressures, corresponded to p = 101.3fO.1 exp (- 34 k 2 kJ mol-l/RT) for temperatures between x 570 and 890 K. Above 890 K the curves became flattened to give lower E and B values. REDUCTION When H2 was present during NO decomposition H20 was produced. Using the isolation technique with H2 in excess, the rate of NO reaction in molecules cm-2 s-l was found to be proportional to NO pressure at 1040 and 960 K.With NO in at least a hundred-fold excess, the reaction of NO became first-order in HZ, rates being determined by following the H20 production. However, under these conditions the nickel surface was almost certainly heavily oxidised and the rate of NO reaction was similar to that observed in the absence of H2. For H2 pressures of the same order of magnitude as the reacting NO pressure the order in H2 was zero. 8 10 12 14 10' K/T FIG. 4.-Arrhenius plots illustrating the effect of H2 on NO reaction probabilities on an oxidised nickel filament : NO pressures 26 p N m-' before reaction. Hz pressures were : 3 mN m-' for increasing (0) and decreasing temperatures (e) ; no Hz present (a). Except at the most elevated temperatures, excess H2 did not appear to enhance the rate of NO reaction on clean nickel filaments when low NO pressures were used.For higher NO pressures or on much used and oxidised filaments, P values were not steady with time, tending to increase slowly as the reaction continued. Rates were calculated from NO pressures and results depended greatly on the previous history of the filament, being less reproducible for the more oxidised filaments. In fig. 4 a typical set of results for an oxidised filament is compared with results on the same filament in the absence of H2. For temperatures between M 750 and 900 K, P for NO reaction in the presence of H2 increased rapidly with increasing temperature, while above 900 K, P became almost constant at ~ 0 . 5 . This represented an increase in filament activity over that observed in the absence of H2 on clean filaments for temperatures above 900 K.A similar increase was also observed for clean filamentsJ . A . MORGAN A N D A . J . B. ROBERTSON 217 in the presence of excess H2 and indicated that P values were not only increased in the presence of H2 by the removal of surface oxygen by H20 production, but possibly also by a surface reaction between NO and H2 or H atoms. PLATINUM DECOMPOSITION There was no detectable reaction of NO on platinum for filament temperatures below x 1650 K. At higher temperatures, a reaction involving disappearance of NO was observed which was approximately half-order in NO. No products of this reaction could be detected. Arrhenius plots were linear and rates could be expressed as v = 1047*2p&0 exp (-SS0540 kJ mol-l/RT) molecule m-2 s-l, where u is the reaction rate and pNO the NO pressure in N m-2. This reaction was unlikely to be occurring on the filament as the apparent activation energy was greater than the N-0 bond dissociation energy (626 kJ mol-f).10 Homogeneous decomposition of NO in the gas phase, after collision with the heated filament, should be negligible with such a high bond dissociation energy. Calculations also indicated that the probability of reaction by collisions of platinum atoms evaporated from the filament and NO molecules in the gas phase was negligible.Therefore, it seems possible that a trapping reaction of NO molecules by evaporated platinum was occurring on the reaction vessel walls. REDUCTION The reaction between NO and H2 was studied at temperatures below 1650 K.At 1230K a reaction occurred that was first-order in NO and zero-order in H2, unless the NO pressure was vastly in excess of the H2 pressure when half-order dependence on H2 was observed. With NO in excess the rates were not reproducible, and so kinetics were only studied for conditions with H2 in excess and of zero-order dependence in H2. The only reaction products detected were N2 and H20. Reaction probabilities were normally calculated from changes in the NO pressures. Arrhenius 104 K/T FIG. 5.-Typical Arrhenius plot of NO reaction on platinum in the presence of H2 : NO pressure 26 pN m-,, H, pressure 1.3 mN m-,.21 8 NITRIC OXIDE ON NIOBIUM, NICKEL AND PLATINUM plots were linear for temperatures below M 1250 K.A typical plot is shown in fig. 5. E below 1250 K was 45+2 kJ mol-l and B varied between ~ 0 . 5 and 1.5. These results were in reasonable agreement with rates calculated from rates of H 2 0 production, which gave E = 47 kJ mol-1 and B = 2.7. Above 1250 K, NO reaction probabilities calculated from changes in NO pressures corresponded to P = 107*0k2*2 exp (-213 & 12 kJ mol-l/RT). This change in slope of Arrhenius plots corresponded to the temperatures at which H2 atomisation was noticeable, and results above 1250 K may relate to a reaction between NO and H atoms on the filament or on the reaction vessel walls. Cooling the reaction vessel walls to * 193 K did not produce any noticeable changes in the rate of NO reaction. This might indicate that the reaction was occurring on the filament rather than the walls. However, the reaction H + NO + M -+ HNO + M has an activation energy very close to zero in the gaseous phase.A surface reaction between H and NO on the vessel walls would also be likely to have no pronounced temperature dependence; such a reaction may contribute to the observed reaction rate. The observed E for NO reaction above 1250 K is in good agreement with the E for H2 atomisation on platinum (214 kJ rnol-l).l2 In view of these considerations the results for temperatures below 1250 K were probably representative of a reaction between NO and H2 molecules rather than atoms. DISCUSSION Yates and Madey have reported that the chemisorption of NO on tungsten is non-dissociative at room temperature. Decomposition of NO was observed at higher temperatures by the detection of gaseous N2.It is likely that a similar process occurs on niobium and nickel, although desorption of N2 was only observed by us at the highest temperatures on niobium. The observed probabilities of reaction of NO on niobium and, possibly, nickel may, therefore, correspond to the probability of chemisorption of NO on the metal surface in its state during reaction. The absence of N2 desorption from niobium indicates the greater strength of the surface metal-nitrogen bond on niobium than on nickel. When a decomposition reaction occurs on niobium or nickel, the surface becomes increasingly contaminated with nitrogen or oxygen, or just oxygen, and thus the probability of reaction decreases with time and may eventually become too small to measure as the filament becomes completely covered by surface species.On used filaments, the rates of decomposition are limited by the relative rates of deposition of contaminants, and removal of these by desorption, diffusion from the filament surface, or by reaction with a reducing agent, such as Ha, which may produce desorbable products. In contrast to niobium and nickel, platinum did not measurably decompose NO at low pressures. P was therefore less than x Other workers have observed measurable rates of decomposition. However, such studies have been performed at NO pressures between w300 N and 66.5 kN m-2, and reaction orders have varied between zero, for saturated coverage, and second order for a bimolecular decomposition of NO.At the very low pressures, and hence very low surface coverages used in the present work, the probability of a bimolecular surface reaction of NO would be very low. Therefore, it appears that the unimolecular decomposition of NO by a dissociative adsorption is an unfavourable process on platinum. Hydrogen may enhance the probability of NO reaction on contaminated niobium by two processes : by reaction with surface contaminants, or by direct reaction with adsorbed NO. The absence of an increase in P for NO reaction on clean new filaments in the presence of excess €3, would tend to eliminate the latter reaction. However, if P for reaction on clean niobium corresponds to probabilities of NO chemisorption,J. A . MORGAN AND A .J. B . ROBERTSON 219 there would not be an increase in P with excess H, present. Owing to inaccuracies in the measurement of rates of product formation, because of their ready adsorption on the reaction vessel walls, it was not possible to determine whether the increase in P on contaminated niobium filaments in the presence of H2 was due, in part, to a reaction of H2 with NO. If the enhancement of activity of contaminated niobium is entirely due to the removal of surface contaminants, P should increase with time of reaction as the surface becomes less and less contaminated. In fact P was relatively constant with time, which suggests that a nearly steady state occurs and the observed P is determined by the relative rates of removal of surface contaminants by reaction with M2, and their replenishment by deposition and diffusion from the interior to the surface.The same considerations apply in the discussion of the results obtained on nickel. P on clean nickel was increased in the presence of excess H2, indicating that an H,+NO reaction was occurring to a small extent. Because of these complicated considerations, kinetic parameters for the reaction of NO on niobium and nickel in the presence of excess H2 may have no ready interpretation and serve mainly to illustrate trends in P with changing filament temperatures. As platinum did not dissociate NO molecules in the absence of Hz, the observed NO reactions in the presence of H, are probably bimolecular surface reactions. At temperatures above = 1250 K the reaction rate seems to be determined by the rate of atomisation of H2 at the filament.For lower temperatures the reaction is probably between NO and H2 molecules and has an activation energy nearly equal to that observed for the reduction of O2 by H2 on platinum by ourselves (49-1- 1 kJ mo1-l)13 and by Gentry et al. (49+3 kJ mol-l).l4 The rate-determining step may be the same for both reactions, possibly the rupture of the H--H bond, with addition of pi atoms to NO or Q2, and the reduced species then decomposing rapidly to smaller fragments. Such a mechanism would explain the rate dependence on H2 atomisation at temperatures above 1250 K. An interesting aspect of the NO reduction by H, at platinum temperatures below 1250 K is that the term B has a value close to unity, which is in accordance with the transition state theory prediction for a mobile transition state.l This result contradicts the view of some workers, notably Laidler,16 who has stated that bimolecular surface reactions of all types may be satisfactorily explained in terms of an immobile layer. We thank the S.R.C. for their support. M. Shelef and J. T. Kummer, Amer. Znst. Chem. Eng., Chem. Eng. Progr. Symp. Series, 1971, 67, no. 115, 74. F. G. Dwyer, Catalyst Rev., 1972, 6, 261. E. R. S. Winter, J. Catalysis, 1971, 22, 158. ' J. T. Yates and T. E. Madey, J. Chern. Phys., 1966, 45, 1623. J. A. Morgan and A. J. B. Robertson, J.C.S. Furuday I, 1974, 70, 936. D. J. Fabian and A. J. B. Robertson, Proc. Roy. SOC. A, 1956, 237, 1. S. C. Jain and T. C. Goel, Brit. J. Appl. Phys., 1968, 1,573. S. C. Jab, T. C. Goel and V. Narayan, Brit. J. Appl. Phys., 1969,2,109. M. Onchi and H. E. Fansworth, Surface Sci., 1969,13,425. lo A. B. Callear and I. W. M. Smith, Disc. Faraday Soc., 1964,37,96. M. A. A. Clyne and B. A. Thrush, Disc. Faraduy Soc., 1962,33, 139. l2 D. Brennan, Adv. CataZysis, 1964, 15, 1. l 3 J. A. Morgan and A. J. B. Robertson, unpublished results. l4 S. J. Gentry, J. G. Firth and A. Jones, J.C.S. Faraday I, 1974, 70, 600. l 6 K. J. Laidler, Chemical Kinetics (McGraw-Hill, London, 1965), p. 292. A. J. B. Robertson, J. Colloid Sci., 1956, 11, 308. (PAPER 7/946)