1583
Ind. Eng. Chem. Res. 1990,29, 1583-1588
A Diffusion Barrier Protected Catalyst Applied to the NO/CO Reaction over Ni3N Yasuyuki Egashira* and Hiroshi Komiyama Department of Chemical Engineering, Faculty of Engineering, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113, Japan
In spite of its high activity in the initial period, Ni3N is easily oxidized to the deactivation product during the NO/CO reaction at 200 "C. We propose a novel concept, the "diffusion barrier protected catalyst" (DBPC), by which this active but sensitive catalyst can survive as long as the atmosphere surrounding the DBPC is reductive. Incorporation of gas-to-surface diffusion resistance magnifies the reducing atmosphere by decreasing the component of lower concentration (NO) in the ambient gas phase t o a greater extent than the component of higher concentration (CO) so as to keep the surface metallic and active. The present experimental results demonstrate that the DBPC method applied to the Ni3N catalyst improved not only the lifetime of the catalyst but also the integrated turnover number (the total number of NO molecules reduced at one surface Ni atom until deactivation occurs), achieving a t least a 330-fold increase.
Introduction Compared with most precious metals and some base metal oxides such as Cr203and CuO, nickel and nickel oxide are believed to be less active in catalyzing the reaction of NO with CO (Shelef and Kummer, 1971; Shelef et al., 1968). Available knowledge of the chemistry of NO and CO on the surface of nickel, however, suggests an inherent activity of nickel for that reaction. Firstly, the dissociation of NO limits the overall rate of reaction over the catalysts active for this reaction (Shelef and Kummer, 1971),and NO molecules adsorbed on a metallic Ni surface dissociate even at room temperature. Specifically, the dissociation of NO on Ni(ll1) (Breitschafter et al., 1981), Ni(ll0) (Price et al., 1976), Ni(100) (Price and Baker, 1980), and polycrystalline Ni surfaces (Carley et al., 1979) under ultrahigh vacuum conditions, and on Ni supported on silica under medium-pressure conditions (Morrow and Moran, 1980),has been reported. Secondly, oxygen atoms chemisorbed on a metallic Ni surface are highly reactive in forming C02 (Labohm et al., 1983) with adsorbed CO molecules, unlike oxygen atoms in the surface oxide (Holloway and Hudson, 1974). These facts strongly indicate the activity of a metallic Ni surface, while nickel oxide is reported to be inactive for this reaction. Following up these considerations, we discovered an extremely high activity of Ni ultrafine particles (UFPs) at low temperature (150 "C) under reductive conditions (CNO< Cco) (Egashira and Komiyama, 1987),although it was found only in the short initial period. Oxidation of the metallic surface was found to be responsible for the rapid and severe deactivation. While nitridation of Ni UFPs to Ni3N, which spontaneously occurred during the reaction under specified conditions, suppressed surface oxidation considerably, the longer lifetime achieved thereby was still far from satisfactory. Considering the inherent activity of Ni, a more efficient method than nitridation to maintain the surface in the metallic state should be found. Theory Surface Oxidation Mechanism during the CO/NO Reaction. Because the oxidation of Ni proceeds through adsorbed oxygen dissociated from NO, lowering the concentration of NO in the gas phase is believed to increase the lifetime of Ni as the catalyst. The low concentration of NO, however, may result in a decrease of the overall reaction of NO and CO. As a measure of the durability 0888-5885/90/2629-1583$02.50/0
of Ni as the catalyst in this system, we define the integrated turnover number (ITN) to be the total number of NO molecules reduced at one surface Ni atom during the lifetime: ITN =
1
lifetime
0
(reaction rate per surface Ni atom) dt
Some recent studies revealed that the oxidation of the metallic Ni surface proceeds through nucleation of NiO islands (Holloway and Hudson, 1974) and that an exchange reaction occurs between chemisorbed oxygen on the metallic Ni surface and oxidic oxygen in the NiO islands (Labohm et al., 1983). This mechanism, including nucleation and growth of NiO surface islands, suggests the existence of a threshold coverage of the surface by oxygen. When oxygen coverage exceeds the threshold value, NiO islands tend to grow, forming a complete NiO surface layer. Below the threshold, NiO islands disappear and the metallic surface is maintained. Let us consider the effect of the concentration of NO in the gas phase on the surface state of Ni at a fixed CO concentration. The coverage of oxygen is determined predominantly by the balance between its production by dissociation of chemisorbed NO molecules and its consumption by reaction with chemisorbed CO molecules. At comparable concentrations of NO and CO, CO molecules hardly chemisorb on the surface, because of the stronger chemisorption of NO than of CO (Morrow et al., 1980). Consequently, oxygen coverage exceeds the threshold value to complete an oxide layer, and the ITN is very small. In contrast, at low concentrations of NO, chemisorbed CO molecules scavenge oxygen supplied from dissociation of NO to keep the surface metallic and active. To summarize, the existence of a threshold of oxygen coverage creates a threshold of the gas-phase concentration of NO. Ideally, the ITN becomes infinite below the threshold of NO; above it, the ITN decreases as the concentration of NO increases. Diffusion Barrier Protected Catalyst Theory. The severe deactivation of Ni in the CO/NO reaction suggests a very low threshold concentration of NO. In this paper, we propose a diffusion barrier protected catalyst concept by which active but sensitive catalysts can survive. The concept of DBPC is outlined as follows. When a diffusion barrier is constructed outside a catalyst, the NO concentration at the catalyst surface (CNoS) is lowered. Even if the NO concentration in the reactor (CNOO) is high 0 1990 American Chemical Society
1584 Ind. Eng. Chem. Res., Vol. 29, No. 8, 1990
make the DBPC effective, CNosshould be much smaller than CNoo. From eq 3, this requres the following criterion: l/k
5F5!&o e
60000 mL.min-'.g-'). ments, even the first gas sample taken immediately after the start of the reaction run showed no evidence of conversion of NO. Allowing for unavoidable ambiguity in the
0 100 200 300
400 500 600 Elapsed Time,min
Figure 6. Typical time dependence of the catalytic properties of Ni3N with diffusion barrier layer (I = 30 mm) a t T = 200 "C,XNO = 2.0%, Xco = 3.2%, and F = 11.2 mlnmin-'.
initial operation, it is concluded that the ITN based on the number of surface Ni atoms in this series of experiments was less than 20. With decreasing space velocity, the catalytic activity of Ni3N appeared as shown in Figure 4. Two characteristic features are (a) the drastic deactivation from complete conversion of NO toward zero conversion, that is, on-off behavior without showing any intermediate reaction rate, and (b) the increase in the production rate of N20 in advance of the deactivation. The appearance of activity at low F / W can be understood when the effect of longitudinal diffusion resistance is taken into account. Combination of the high activity of Ni3N and the low flow rate allows a steep longitudinal concentration profile to maintain the catalyst ambience at low NO concentration. On-off deactivation is explained by a mechanism similar to that in the DBPC system in which deactivation causes further deactivation through the increased concentration of NO. This mechanism also suggests a decrease of ITN due to flattening of the concentration profile with increasing F/W, as was expected theoretically and observed experimentally (Figure 5). Reaction with Diffusion Barrier. Figure 6 shows the typical results of this series of experiments. On-off deactivation preceded by N20 production also characterizes this system, while the active state (on state) does not show complete conversion of NO. In this case, conversion of NO can be controlled by the feed gas flow rate. In all
1586 Ind. Eng. Chem. Res., Vol. 29, No. 8, 1990
0 0 302"
40 7mm
oi
c
I A
d0lyLt
L-
I?
~
800 1200 1600 Elapsed Time min
0
400
A
I
2000
Figure 7. Time dependence of the reaction rate of Ni3N with various diffusion resistances a t T = 200 O C , XNO = 2.0%,and XCO= 3.2%.
2
i
~
7
2.0 2.1
2.2
2.3
c.4
1 / ~ ~ ,K-' 1 0 ~
Figure 10. Temperature dependence of the reaction rate constant over Ni,N with barrier layer (1 = 30 mm) a t XN0 = 2.0% and XCO = 3.2%.
7 7 -7 2;-
z
3o:p
k I
-
+. -1 ;2-
-CC
P . 7
0
0 -7
c P
I,"
Figure 8. Integrated turnover number of NO over Ni3N as a function of length of diffusion barrier layer at T = 200 O C , X N O= 2.0%, and Xco = 3.2%. The turnover number obtained for Ni3N without diffusion barrier layer a t high F/W (=64400mL.min-'.g-') is plotted a t 2 = 0 mm.
i,mm
Figure 9. Relationship between the total resistance of the system (R,d) and the length of the diffusion barrier layer (1) a t the same conditions as in the Figure 7 .
the experiments of this series, NO conversion was less than lo%, and consequently, the NO concentration outside the DBPC is reasonably approximated to be constant at the feed gas concentration. Figure 7 shows the time dependence of the total reaction rate on 1 (length of the diffusion barrier). With increasing 1, the lifetime of the catalytic increases to a greater extent than the decrease in reaction rate in the region 1 = 5-40 mm. Consequently, the total amount of NO reduced increases with 1. This tendency is clearly shown in Figure 8. The ITN increases with the increase of 1 and reaches a maximum of 6600 at 1 = 40 mm. This maximum ITN means that 6600 NO molecules were reduced by 1 surface Ni atom of Ni3N. Compared with the ITN achieved without diffusion barrier and sufficiently high value of F/W (>6OOOO ml-min-l-g-'), plotted at I = 0 mm in Figure 8, the DBPC approach increases ITN at least 330-fold. The incorporation of the diffusion barrier not only lengthened the lifetime but also increased the ITN. It is stressed that the nucleation and growth mechanism of Ni surface oxidation is essential to this improvement of durability, as discussed in the section on surface oxidation mechanism. Although the data were less reproducible, the saturation of the ITN at 1 = 80 mm is indicated. Figure 9 shows the relationship between 1 and Rbd defined by eq 6 (=CNOO(J;total resistance for this process). Their linear relationship indicates that the reaction rate is governed by diffusion through the layer. The criterion for the DBPC (eq 7) is satisfied.
3
*
; &O/CNO
Figure 11. Integrated turnover number of NO as function of CO/ NO ratio over Ni,N with diffusion barrier ( 1 = 30 mm) a t T = 200 "C and C N O = 0.52 m01.m'~.
An Arrhenius plot is shown in Figure 10. The activation energy is about 6.7 kJ-mol-', a reasonable magnitude for gas-phase diffusion and a confirmation that the reaction rate is governed by the diffusion. Integrated turnover numbers of NO at various Ccpo/ CNooratios were measured, varying the CO concentration with a constant diffusion resistance (Figure 11). The catalyst behavior under reductive conditions was completely different from that under oxidative conditions, as predicted from theory. Under reductive conditions (CCO;/CNO*> 1) the ITN is large, but under oxidative conditions (Ccoo/CNoo< 1) it is almost zero. Two unexpected features, however, should be pointed out. Firstly, under reductive conditions, the ITN decreases with increasing Ccoo, though the atmosphere becomes more reductive. Secondly, the ITN increases sharply near the stoichiometric but oxidative condition. Figures 7 and 8 show that durability is improved by the DBPC approach but that at 1 = 80 mm the increment of catalytic lifetime is limited and the ITN decreases together with reaction rate. Furthermore, in the region of Ccoo/ CNoo> 1in Figure 11,the ITN decreases as Ccooincreases, though the atmosphere becomes more reductive. These experimental results suggest the existence of another mechanism that causes deactivation, in which CO is directly concerned. The Boudouard reaction is a candidate mechanism, but some ambiguity remains. N20 Reduction by CO. The catalytic properties of Ni3N for N20 reduction are considerably different from those for NO reduction. The most remarkable difference is that Ni3N is not deactivated during N20/C0 reaction, even with an excess of N20. The reaction rate is proportional to CN, (Figure 12a) and inversely proportional to Cco (Figure 12b), and it is not as fast as that of the NO/CO reaction. The reaction rate form as measured suggests that CO is more strongly chemisorbed than N 2 0 to keep the surface from oxidation. The kinetic data taken at 200 "C,C N I O = 0.05-0.5 m o l ~ m -and ~ Cco = 0.13-1.3 m ~ l . m -are ~ summarized by the following equation: r = 5.57 x 10-5Cco-1CN,o (N,O in mo1.s-l.g-l)
Ind. Eng. Chem. Res., Vol. 29, No. 8, 1990 1587
$105 C
._
1
.-b u
i
I@
1 ul
, , ,
/,/,,
, ,, 10 I
100
xco ,eta Figure 12. Dependence of reaction rate for N20/C0 over Ni3N on (a) N20 concentration of Cco = 0.74 mol~m-~ and on CO concentra= 0.13 m ~ l - m - ~ T. = 200 "C. tion at ,C
In the NO/CO reaction experiments, both N2 and N20 are produced and the ratio of NzO increases in advance of the sudden deactivation. This change in product selectivity, observed with and without diffusion resistance (Figure 6 and Figure 4 respectively), is reasonably explained as a decrease in the consecutive reduction of NzO to N2,due to the deactivation of the catalyst. The observed rate of the NzO/CO reaction is high enough to affect the total selectivity, at least at the initial stage of the reaction system. Just before the sudden deactivation, the effluent stream from the catalyst bed or the DBPC system contained almost the same concentrations of N2 and N20 (Figure 4, Figure 6). Under that condition, only a small fraction of the catalyst layer, existing at the tail end, remains active, and therefore, the consecutive reaction of NzO to Nz is not catalyzed. Consequently, this selectivity of N20 and N2 presumably reflects the selectivity of the primary reaction step. The reaction scheme is expressed, therefore, as primary reaction step (1) 2 N 0 2CO = 2C02 + Nz (2) 2N0 + CO COZ NzO consecutive reaction step N2O CO = COZ + Nz According to this reaction scheme, the observed selectivity of N2 and NzO of about unity reflects plausibly that the ratio of the two reactions included in the primary step happens to be almost the same on this catalyst. If Ccoo/CNoo> 3/4 is satisfied, NO molecules can be consumed by the first reaction step, and the NzO produced is not fatal to surface oxidation. Even in the region of Ccoo/CNoo < 1, therefore, the active state could appear near the stoichiometric concentration, as shown in Figure 11. In addition, it should be stated that N20/N2selectivity for reaction 1 depends on the reaction temperature. The catalyst becomes N20 selective at reduced temperature.
+
+
+
Conclusions Though its strong affinity for oxygen gives a Ni surface the ability to dissociate NO, formation of a surface oxide layer inhibits the catalytic cycle of the NO/CO reaction. Surface oxidation proceeds, however, by the formation of chemisorbed oxygen, which is highly reactive with CO, and
there is an exchange reaction between this chemisorbed oxygen and oxidic oxygen in the surface oxide. So at sufficiently low NO concentration, chemisorbed CO scavenges chemisorbed oxygen to keep its coverage lower than the threshold of nucleation and growth of NiO islands. The metallic surface thus maintained catalyzes the NO/ CO reaction efficiently. Experiments using a conventional flow-type reactor system suggest an extremely low threshold concentration of NO for Ni3N, but its high activity in the metallic state makes possible the application .of the diffusion barrier protected catalyst approach, wherein the incorporation of a gas-to-surfacediffusion barrier reduces NO concentration at the surface and maintains the metallic state of Ni3N with practical NO concentration in the reactor as long as the reactor atmosphere is reductive (CNo0 CcoO). Experiments performed to prove this concept demonstrated an increase of ITN of at least 330-fold. Unfortunately, the existence of another mechanism causing deactivation prevented a further increase of ITN. In practice, Ni supported in carrier particles of proper size and with pores fine enough to cause sufficient diffusion resistance would function as a DBPC. Though the outer layer of the catalyst particles would rapidly deactivate, the layer would act as a diffusion barrier to keep the inside core of the catalyst from deactivating. Further, if a functional material with higher resistance to NO diffusion than to CO diffusion is available, the applicability of the DBPC method could be extended to oxidative conditions. This study was intended to demonstrate an example of novel catalytic systems when a surface reaction mechanism taken from the field of surface science is combined with a reaction-diffusion system suitably designed from the viewpoint of chemical reaction engineering. While the lifetime achieved experimentally in this study is insufficient for practical purposes, the concept of a base metal based metallic catalyst for oxidation reactions makes further studies desirable. Nomenclature Cco = concentration of CO in the feed gas, mol~m-~ CNo = concentration of NO in the feed gas, m~lem-~ CN20= concentration of N 2 0 in the feed gas, m ~ l - m - ~ Ccoo = concentration of CO in the reactor, m01.m~~ CNoo = concentration of NO in the reactor, mol~m-~ Ccos = concentration of CO at the surface of the catalyst, m~l.m-~ CNoS= concentration of NO at the surface of the catalyst, m~l.m-~ D = diffusion coefficient in the barrier layer, m2@ F = flow rate of the feed stream based on standard temperature and pressure, mL-min-' ITN = integraged turnover number of NO based on the number of surface Ni atoms J = flux of the reactant, mol*m-2.s-1 k = surface area based reaction rate constant, m-s-' 1 = length of the diffusion resistance layer, m r = reaction rate per unit amount of catalyst, m o l d g - ' Rbd = total reaction resistance for the DBPC, s.m-' T = reaction temperature, "C W = weight of catalyst, g X c o = mole fraction of CO in the feed gas, % XNO= mole fraction of NO in the feed gas, % XNPO= mole fraction of N20 in the feed gas, % Registry No. NO,10102-43-9;CO,630-08-0; NiBN,12033-45-3. Literature Cited Breitschafter,M. J.; Umbach, E., Menzel, D. An Electron Spectroscopic Investigation of the Adsorption of NO on Ni(ll1). Surf. Sci. 1981, 109, 493-511.
1588
Ind. E n g . Chem. R e s . 1990, 29, 1588-1599
Carley, A. F.; Rasaias, S.; Roberts, M. W.; Wang, Tang-Han. Chemisorption of Nitric Oxide by Nickel. Surf. Sci. 1979, 84, L227L230. Egashira, Y.; Komiyama, H. Spontaneous Nitridation of Ultrafine Particles of Ni during NO/CO Reaction. Chem. Lett. 1987, 2413-2416. Holloway, P. H.; Hudson, J. B. Kinetics of the Reaction of Oxygen with Clean Nickel Single Crystal Surface, 1 Ni(100) Surface, 2 Ni(ll1) Surface. Surf. Sci. 1974, 43, 123-149. Kimoto, K.; Kamiya, Y.; Nonoyama, M.; Uyeda, R. An Electron Microscope Study on Fine Metal Particles Prepared by Evaporation in Argon Gas a t Low Pressure. Jpn. J . A p p l . Phys. 1963, 2, 702-713. Labohm, F.; Gijzeman, 0. L. J.; Geus, J. W. The Interaction of Oxygen with Ni(ll1) and the Reduction of the Surface Oxide by Carbon Monoxide and by Hydrogen. Surf. Sci. 1983, 135, 409-427. Morrow, B. A.; Moran, L. G. The Adsorption of NO and NO, on
Silica-Supported Nickel. J. Catal. 1980, 62, 294-303. Morrow, B. A.; Sont, W. N.; Onge, A. St. The Reaction between NO and CO on silica-Supported Nickel. J. Catal. 1980,62, 304-315. Price, G. L.; Baker, E. G. The Chemisorption of Nitric Oxide on (100) Nickel Studied by LEED, AES, UPS and Thermal Desorption. Surf. Sci. 1980, 91, 571-580. Price, G. L.; Sexton, B. A.; Baker, B. G. The Chemisorption of Nitric Oxide on (110) Nickel Studied by LEED, AES, and Thermal Desorption. Surf. Sci. 1976, 60, 506-526. Shelef, M.; Kummer, J. T. The Behavior of Nitric Oxide in Heterogeneous Catalytic Reactions. Chem. Eng. Prog., Symp. Ser. 1971, 67 (115), 74-92. Shelef, M.; Otto, K.; Gandhi, H. The Oxidation of CO and COz and by NO Supported Chromium Oxide and Other Metal Oxide Catalysts. J . Catal. 1968, 12, 361-375.
Receioed for review October 30, 1989 Accepted March 22, 1990
Binder/Support Effects on the Activity and Selectivity of Iron Catalysts in the Fischer-Tropsch Synthesis Dragomir B. Bukur,*ptXiaosu Lang,+Doble Mukesh,? William H. Zimmerman,? Michael P. Rosynek,' and Chiuping LiI Kinetics, Catalysis and Reaction Engineering Laboratory, Department of Chemical Engineering and Department of Chemistry, Texas A&M University, College Station, Texas 77843
The influence of silica and alumina binders (supports) on the activity and product selectivity of precipitated iron catalysts for Fischer-Tropsch synthesis (FTS) was studied in a fixed bed reactor a t 1.5-3.0 MPa and 220-250 " C , using synthesis gas with H 2 / C 0 = 1 M feed ratio. Both the FTS and the water gas shift activity decreased with increasing support content. The catalyst deactivation rate decreased with catalysts containing 24 and 100 g of Si02 per 100 g of Fe. Secondary reactions (olefin hydrogenation and isomerization) increased with increasing silica content. Hydrocarbon selectivities improved (less gaseous hydrocarbons) with the addition of support, except for the catalyst containing 24Si02/ 100Fe. The results were interpreted in terms of interactions between potassium and/or iron with supports.
Introduction The objective of adding a catalyst support is to provide a large surface area for the formation and stabilization of small metal crystallites in the catalyst. The support may also have significant effects on the catalyst activity and selectivity due to strong metal-support interactions (SMSIs) (e.g., Vannice and Garten, 1980; Bartholomew et al., 1980; Reuel and Bartholomew, 1984). Evidence for SMSIs has been found in studies of Fischer-Tropsch synthesis (FTS) over supported nickel, ruthenium, and cobalt catalysts (e.g., Vannice and Garten, 1979, 1980; Goodwin et al., 1984; Reuel and Bartholomew, 1984). In all of these studies, it was found that the catalyst activity and selectivity were influenced by one or more of the following variables: nature of support, metal dispersion, metal loading, and/or preparation method. Several studies have been made with supported iron catalysts (e.g., Anderson, 1956,1984;Guczi, 1981; Dry, 1981; Jung et d., 1982; Egiebor and Cooper, 1985; Berry et al., 1986), but only a few of them dealt with the effect of support type and content on catalyst activity and detailed product distribution. In general, iron catalysts with high support-tometal ratios have been ineffective FTS catalysts. The
* Author to whom correspondence should be addressed. 'Department of Chemical Engineering. Department of Chemistry. 0888-5885/ 9012629- 1588$02.501 0
major findings from studies of direct relevance to the work presented in this paper are summarized below. Dry (1981) reported some results, from comprehensive studies conducted at SASOL (South Africa), on the effect of supports on the catalytic performance of a series of precipitated iron catalysts (Fe/Cu/K20). Unfortunately, only relative concentrations for some components in the catalysts employed were reported, and the catalyst activity and wax selectivity were also reported as relative quantities. Silica was found to be the best support in terms of both activity and selectivity. Catalysts containing a second support material (CaO, Cr203,A120,, V205,Tho2, MgO, or Ti02)in addition to Si02also showed inferior performance relative to a Fe/Cu/K20/Si02 catalyst. Dry emphasizes that a careful balance between promotional (potassium, in the form of K20) and stabilization (support) additives must be achieved to maximize the desired activity and hydrocarbon product selectivities. Potassium is an essential promoter in iron catalysts for the F'TS, since it enhances the formation of both longer chain and olefinic hydrocarbons (Anderson, 1956; Dry, 1981). When acidic oxides are used as supports, they may react with basic alkali and thereby reduce the promotional effect of potassium. Also, the use of high surface area supports tends to reduce direct contact between iron and potassium since the metal covers only a small fraction of the support surface (McVicker and Vannice, 1980). This would also render the potassium promotion less effective. 0 1990 American Chemical Society