J . Phys. Chem. 1986, 90, 47264129
4726
Ammonia Synthesis over Iron Single-Crystal Catalysts: The Effects of Alumina and Potassium Simon R. Bare, D. R. Strongin, and G. A. Somorjai* Materials and Molecular Research Division, Lawrence Berkeley Laboratory, and Department of Chemistry, University of California, Berkeley, California 94720 (Received: January 30, 1986)
We have investigated the effects of potassium and aluminum oxide on the synthesis of ammonia from nitrogen and hydrogen over model iron single-crystal catalysts at 20-atm reactant pressure. Elemental potassium does not remain on a Fe( 100) surface under reaction conditions but a small amount of potassium can be stabilized by coadsorption with oxygen. The presence of the stabilized K 0 has no effect on the rate of ammonia synthesis. Substantially more potassium can be stabilized at higher temperatures on Fe(100), Fe(l1 l), and Fe(ll0) surfaces by coadsorption with aluminum oxide. The cooperative interaction between the potassium and oxidized aluminum seems to be due to compound formation. There is an inverse relationship between the rate of reaction and amount of aluminum oxide and potassium present on all three iron surfaces investigated.
+
Introduction The catalytic synthesis of ammonia from its elements is one of the most important industrial reactions. The modern ammonia synthesis catalyst is similar to the original industrial catalyst in regard to the basic composition.' It is made from iron oxide (Fe,O,), to which a few percent by weight of other oxides are added as promoters. Although there is considerable variation in the bulk composition of the catalysts (dependent upon the particular synthesis conditions employed), the oxides usually employed are those of aluminum, potassium, and calcium. In addition, oxides of silicon and magnesium are now added, whereas they were only present as impurities in the raw materials in the early catalysts. Even though the synthesis of ammonia catalyzed by promoted iron has been studied for over 80 years, a detailed understanding of the effect that the promoters play in the reaction is not known, but there have been many ideas put forward over the years.2 The promoters are generally divided into two classes: (i) structural promoters and (ii) electronic promoters. Since this is not the place for a detailed critical review of the various mechanisms of the promoters that have been proposed, the types of effects will be briefly reviewed without comment. The oxides of aluminum, magnesium, and silicon are thought of as structural modifiers. For example, in the case of aluminum oxide, during the reduction of the Fe304to the active metallic phase (a-Fe), the presence of aluminum oxide inhibits the rate of diffusion of iron in the magnetite lattice. Thus, nucleation of small iron particles proceeds quickly compared with their growth rate. This leads to the formation of small crystallites and thereby a very porous structure with high surface area (10-20 m2 g-I).*J It has also been postulated that alumina inhibits the conversion of crystallographic planes with high catalytic activity (the Fe( 111) planes4) into those less active (( 100) and (1 10) planes4). Potassium oxide is considered as an electronic promoter that can increase the specific activity of the Fe surface (activity per Fe atom).2 Indeed, in a series of chemisorption studies of potassium and nitrogen on Fe single crystals in ultrahigh vacuum (UHV), Ertl was able to show that the preseace of submonolayer amounts of elemental potassium increased the dissociative sticking probability of nitrogen (the rate-limiting step in the ammonia (1) Krabetz, R., Huberich, T. Fen. Sci. Technol. Ser. 1977, 2, 123. (2) (a) Ertl, G . In Robert A. Welch Conferences on Chemical Research XXVHeterogeneous Catalysis, Houston, TX 1981, p 119. (b) Emmett, P. H. In The Physical Basis for Heterogeneous Catalysis; Drauglis, E., Jaffee, R. I. Eds.; Plenum: New York, 1975, p 3. (c) Nielsen, A. Catal. Rev. 1970, 4, 1. (d) Frankenburg, W. G . Catalysis 1955, 3, 171. (3) Nielsen, A. An Investigation on Promored Iron Catalysts for the Synthesis of Ammonia, 3rd ed.; Jul. Gjellerups Forlag: Copenhagen, 1968. (4) Spencer, N. D.; Schoonmaker, R. C.; Somorjai, G. A. J , Cutal. 1982, 74, 129.
0022-3654/86/2090-4726$01.50/0
synthesis reaction2*5g6 by 2 orders of magnitude and that of potassium coadsorbed with oxygen (see later discussion) by up to an order of magnitude.' This effect was explained in terms of an increase of the heat of adsorption of molecularly adsorbed nitrogen together with a lowering of the activation energy for dissociation. These results give rise to the question of the chemical state of potassium under ammonia synthesis conditions, which has also been the subject of much debate.* An alternative opinion on the promoting effect of potassium was given by Frankenburg.u He suggested that the alkali oxide neutralizes weakly acidic centers on the surface of the catalyst, thus preventing the catalyst surface from being blocked by adsorbed N H 3 molecules or species such as N H or NH2. Support for this proposal comes from the observation that potassium improves the catalyst performance better at high pressures (Le., high ammonia concentration) than at lower pressures.' The industrial catalyst is a very complex system. In an attempt to understand the effects of alumina and potassium promoters in more detail, we have studied a simpler model system using Fe single crystals as catalysts. Studies of the industrial catalyst have usually been carried out in reactors operating at pressures greater than 1 atm (1 atm = 101.3 kN m-2),2w while indirect studies of the adsorption of nitrogen, etc., on iron single crystals have been performed under ultrahigh vacuum conditions, where pressures Torr (1 Torr = 133.3 N m-2).2a The former do not exceed approach suffers from a lack of direct information on the nature of the catalyst surface, and the UHV approach does not allow the synthesis of ammonia to occur to any measurable extent, since the thermodynamic equilibrium is unfavorable. The present investigation, using an apparatus that allows both UHV and high-pressure conditions to be obtained within the same chamber, bridges the pressure gap and enables the rate of formation of ammonia to be directly related to the state of the catalyst surface. This approach has been shown to be successful in many catalytic reaction s y ~ t e m s . ~ In this paper we report the results of the effect of potassium and alumina on Fe( 11 l ) , Fe( loo), and Fe( 110) single crystals. We find that elemental potassium is not stable on the iron surface under our reaction conditions but a maximum of 10% of a monolayer of K can be stabilized by coadsorption with oxygen. Substantially more potassium can be stabilized by coadsorption with aluminum oxide. This stabilization may be due to compound formation on the iron surface. However, no promotional effect (5) Bozso, F.; Ertl, G.; Grunze, M.; Weiss, M. J . Catal. 1977, 44, 18. (6) Bozso, F.; Ertl, G.; Weiss, M. J. Catal. 1977, 50, 519. (7) Ertl, G.;Weiss, M.; Lee, S. B. Chem. Phys. Lett. 1979, 60, 391. (8) Paal, Z.; Ertl, G.; Lee, S . B. Appl. Surf. Sci. 1981 8, 231. (9) Somorjai, G. A. Chemistry in Two Dimensions: Surfaces; Cornell University Press: Ithaca, NY, 1981; Chapter 8.
0 1986 American Chemical Society
Ammonia Synthesis over Iron Catalysts is observed with submonolayer amounts of the stabilized K-A1203.
Experimental Section The experiments were performed in a diffusion-pumped stainless steel UHV chamber with a base pressure of 1 X Torr (described in detail elsewhere4). Briefly, the chamber is equipped with a retarding field analyzer for Auger electron spectroscopy (AES) and low-energy electron diffraction (LEED), a mass spectrometer for temperature-programmed desorption (TPD), an ion gun for sample cleaning, and a hydraulically operated highpressure cell that constitutes part of a microbatch reactor. The ionizer of the mass spectrometer is enclosed in a gold-plated tube with an aperture at the end to enhance the sensitivity and to permit sampling of gas desorbed almost exclusively from the front face of the crystal. When the high-pressure cell is enclosed over the crystal, the reactor loop can be pressurized to 20 atm, and the gases can be circulated over the crystal with a positive displacement pump. A typical reaction sequence would be to clean and characterize the crystal under UHV, to close the high-pressure cell, and to pressurize to 20 atm with a 3:l stoichiometric mixture of hydrogen and nitrogen. While the gases are circulating over the crystal the crystal is brought to the desired reaction temperature. Once this temperature is reached gas samples are periodically taken from the loop via a sampling valve and passed into a photoionization d e t e ~ t o r .The ~ detector signal is directly proportional to only the ammonia partial pressure. After a given reaction time the loop is evacuated, the cell is opened, and the composition and structure of the surface are analyzed under UHV. Three single-crystal faces of iron were studied: the (1 1l), (loo), and (1 10). Both sides of the single crystals were oriented and polished by standard techniques. Prior to the insertion of each crystal into the vacuum chamber, the near-surface concentration of the major contaminant, sulfur, was substantially reduced by heating the crystal in a furnace to 923 K in 1 atm hydrogen for 36 h. The crystal was spot-welded to two 0.25-mm platinum support wires and was resistively heated. The temperature was monitored with a chromel-alumel thermocouple spot-welded to one side of the sample. During a reaction the crystal temperature could be controlled to *2 K of the desired temperature. The reactant gases, research-purity nitrogen and hydrogen, were further purified by passing them through molecular sieve and a liquid-nitrogen cooled coil before they entered the reaction loop. A Saes Getters source mounted close to the sample was used for evaporating potassium. The amount of potassium on the surface was measured by AES. The relationship between the potassium Auger signal and the coverage was obtained with potassium uptake curves. After the crystal was cleaned, the desired amount of potassium was deposited by controlling the deposition time. Coadsorption of water (oxygen) with the potassium could not be completely avoided because of the relatively high base Torr. The effect of pressure of our system, about 1 X coadsorbed oxygen is discussed in the text. A Knudsen-type cell was used for evaporation of aluminum. The absolute intensity of the 67-eV A1 LVV Auger peak was unreliable owing to its location on the large secondary electron background of the RFA. A more reliable estimate of the amount of adsorbed aluminum was obtained by using carbon monoxide chemisorption. C O does not adsorb on aluminum oxide at room temperature.2b.10 Therefore the amount of A120,-free iron surface could be estimated by taking the difference in the integral areas of the TPD peaks of C O from a initially clean iron surface and that covered with Al2O3. I3CO was used for this calibration to separate the CO TPD peak from nitrogen, which surface segregates and desorbs from the bulk of the Fe sample in the same temperature range. It is noted that the presence of aluminum oxide or aluminum oxide coadsorbed with potassium did not alter the shape or the peak maxima of the CO desorption peaks. After initial outgassing, both the K and A1 sources proved to be clean and effective sources of the respective metals. Submonolayer amounts of A1 were easily and rapidly oxidized to AI2O3on the (10) Silverman, D. C.; Boudart, M. J . Cutul. 1982, 77, 208.
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4727 Fe surface by heating to 873 K in 5 X lo-* Torr oxygen for 60 s. Following this procedure the minimum of the second derivative of the LVV A1 Auger peak moved from 67 to 53 eV, indicating complete oxidation, in good agreement with AI,O, thin films and No splitting of the 47-eV MVV Fe peak was bulk observed, indicating that the Fe substrate was not being 0~idized.l~ If monolayer amounts of oxygen were present on the Fe surface it is known that under the synthesis conditions the oxygen would rapidly be removed. In fact even a heavily oxidized iron single crystal is readily reduced to metallic Fe under our ammonia synthesis conditions such that the activity is the same as that of the clean ~ u r f a c e .In ~ a separate study, a polycrystalline AI foil was inserted into the UHV chamber and was heavily oxidized (5 X Torr O2at 723 K for 15 min). The same shift in the A1 LVV Auger peak was observed, and moreover the intensity ratio was of the peak-to-peak heights of the LVV A1 peak to the 0520 about 0.2 (A1 peak recorded with 1-V peak-to-peak modulation, 0 peak using 7 V peak-to-peak), which was the same ratio as that observed for oxidized aluminum on the Fe( 100) single crystal.
Results and Discussion Stabilization of Potassium under Reaction Conditions. Potassium on Iron. The synthesis of ammonia over clean Fe( l l l ) , Fe( loo), and Fe(l10) single crystals has previously been ~ t u d i e d . ~ The relative rates of ammonia synthesis activity were found to be 420:25:1 for F e ( l l l ) , Fe(100). and Fe(ll0) samples, respectivelyS4 We repeated these experiments using a stoichiometric mixture of nitrogen and hydrogen (1 :3, respectively) a t 20 atm total pressure and a temperature of 673 K. Very good agreement was reached as to the initial rates of ammonia formation on these clean surfaces, the Fe( 11 1) being the most active by far. Additionally, an apparent activation energy for the reaction of 18.8 0.5 kcal/mol was determined with a Fe(100) crystal in good agreement with our previously determined value of 19.4 k ~ a l / m o l . ~ Once this structure sensitivity of the ammonia synthesis reaction had been corroborated, the next stage was to investigate the role of the potassium promoter. Elemental potassium was evaporated onto both sides of a Fe( 100) crystal under UHV to give coverages between 0.05 and 1 .O monolayer. The crystal was then enclosed in the high-pressure cell, the synthesis reaction was performed at 673 K using 20-atm 3:l hydrogen-to-nitrogen, and the rate of ammonia production was determined. For all coverages of potassium the rate of ammonia formation was the same as the clean Fe(100). Moreover, no potassium signal was detectable by AES at the end of the reaction. It can be concluded that under our reaction conditions (and therefore most likely those of the industrial synthesis) elemental potassium is not stable at any coverage and simply volatilizes from the surface. This conclusion is in agreement with detailed UHV studies characterizing the adsorption of potassium on Fe single crystals, where it was determined that potassium desorbs in the temperature range 450-850 K, depending upon the coverage of potassium present.14 The strongly ionic character of the Fe-K bond (which results in thermal desorption of potassium at temperatures that are higher than that where the sublimation rate of bulk potassium is rapid) is insufficient to hold elemental potassium on the iron surface under our reaction conditions. Potassium and Oxygen on Iron. The promoter that is added to the industrial catalyst, however, is not metallic potassium but potassium oxide. Further, UHV studies of potassium adsorption on Fe single crystals have shown that coadsorption of oxygen with the potassium stabilizes the potassium on the surface to higher temperatures such that even after briefly heating to 1000 K some potassium remains on the s ~ r f a c e . ~The ~ ' ~conditions of these UHV experiments are quite different from the industrial synthesis
*
(11) Madden, H. H.; Goodman, D. W. Surf.Sei. 1985, 150, 39. (12) Citrin, P. H.; Rowe, J. E.; Christmann, S. B. Phys. Reu. B: Solid State 1976, 814, 2642. (13) Langell, M.; Somorjai, G. A. J . Vae. Sei. Technol. 1982, 21, 858. (14) Lee, S. B.; Weiss, M.; Ertl, G. Surf.Sei. 1981, 108, 357. (15) Pirug, G.; Broden, G.; Bonzel, H . P. Surf.Sci. 1980, 94, 323.
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986
4728
Bare et al.
C
8 04 0
2
c
-0 al [L
,
40% AI,O,
+
K/Fe(100)
1
02
i 0
ZOO
400
600
800
Energy of Auger Electron(eV)
Figure 1. Auger electron spectra of promoted Fe(100) surfaces before and after ammonia synthesis reaction. Conditions given in the text.
conditions, where the catalyst is kept at temperatures of 573-873 K in 150-300 atm of reactant gases for periods of 8 years or more. In an attempt to stabilize potassium on the Fe(100) surface under our conditions, coadsorption of potassium with oxygen was examined. Figure 1 summarizes the results in the form of Auger electron spectra before and after the synthesis reaction. An Auger spectrum corresponding to that of a complete monolayer of potassium that had subsequently been exposed to oxygen under UHV is shown together with a spectrum recorded from the same surface after performing the synthesis reaction at 673 K at 20 atm total pressure of a stoichiometric mixture of N, and H2 for 30 min. The potassium signal is greatly reduced after reaction and corresponds to approximately 10% of a monolayer. If the reaction is performed at higher temperatures, up to 873 K, substantially less potassium is stabilized on the surface. Furthermore, the effects of different initial potassium coverages (0.05-1 .O monolayers), oxygen exposures, and the order of exposure (Le., K + 0 or 0 K) were also studied. In no case could more than 10% of a monolayer of potassium be stabilized at 673 K. The initial rate of ammonia synthesis using our standard conditions was also determined for many different K + 0 adlayers on an Fe( 100) sample. In all situations the initial rate of N H 3 formation was unaltered from that of clean Fe(100). That is, no promotional effect on the rate was observed with the stabilized K 0. Ertl et a1.* determined the effect of coadsorbed oxygen and potassium on the adsorption of nitrogen on polycrystalline iron under UHV. They observed two main effects: (i) the presence of oxygen with potassium blocks sites for nitrogen chemisorption and thus decreases the saturation coverage and (ii) the presence of oxygen with potassium decreases the rate of nitrogen chemisorption. Thus, although the addition of oxygen stabilizes the potassium to higher temperatures, the electronic promoter effect deteriorates. It is likely that at the concentrations of potassium and oxygen that we can stabilize the enhancement of the rate of nitrogen chemisorption by the potassium has been nulled by the presence of oxygen. This result alone indicates that the role of potassium in the industrial catalyst is probably not one of simply increasing the rate of nitrogen chemisorption. Potassium and Oxidized Aluminum on Iron. The industrial catalyst is a complex system of many promoter oxides and it is likely that there will be some cooperative interactions between the various oxides. To investigate the possibility of cooperative interactions we extended our model catalyst system to examine
+
+
0
loo0
Figure 2. Relative concentration of potassium on a Fe(100) surface as a function of time at 673 K under UHV for several different oxidized
aluminum coverages. both the stability of potassium in a mixed adlayer of potassium and aluminum oxide and its effect on the reaction rate. The Auger spectrum recorded at the end of a 30-min synthesis reaction at 673 K with 20 atm 3:1 hydrogen:nitrogen, after first depositing under UHV 40% of a monolayer of A1203followed by a complete monolayer of potassium, is shown in Figure 1. Comparison of this spectrum with that of a K 0 adlayer shows that substantially more potassium is stabilized. In fact the potassium AES signal corresponds to 40% of a monolayer of potassium. To further probe this cooperative behavior between the oxidized aluminum and potassium, the following experiment was performed. A known amount of A1203was deposited onto a Fe( 100) crystal, followed by a coverage of potassium corresponding to a complete monolayer. The crystal was then heated to 673 K under UHV and the intensity of the potassium Auger signal recorded periodically. The results are summarized in Figure 2, where the relative concentration of potassium is shown as a function of time at 673 K for several different A1203coverages. A rapid initial decrease in the amount of potassium is observed in all cases but then a steady state is reached, the value of which depends upon the amount of Al,03 present. In fact, it can be seen that there is an approximate 1:l correlation between the amount of potassium that is stabilized and the amount of aluminum oxide present. The correlation is better observed in Figure 3, where the relative concentration of potassium remaining on the surface after heating to 673 K under UHV for 25 min is plotted against the relative concentration of oxidized aluminum. A least-squares fit to the data points shown in the figure yields a correlation of 1.2 f 0.1:l between the amount of aluminum oxide to the amount of stabilized potassium. In all cases the amount of potassium stabilized under UHV is also stable under our reaction conditions (20 atm total pressure at 673-823 K). If the K A120,-covered Fe( 100) is heated above 873 K, potassium is lost from the stabilized surface overlayer. This decomposition temperature agrees well with the temperature at which an industrial catalyst looses potassium from the surface, as monitored by X-ray photoelectron spectroscopy.2a The increased stability of potassium with aluminum oxide in our model system is in agreement with results from the industrial catalyst, where it was found that catalysts singly promoted with KzO or those with a high KzO to A120, ratio lose alkali during use.16
+
+
(16) Nielsen, A. Adu. Cutuf. 1953, 5 , 1
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4729
Ammonia Synthesis over Iron Catalysts 0.8
I
I
I
I
I
0,O.E -
a
* 0 C
.-0 L
0
2 0.4 C W U
c
0
u
> 0
-
0.2
W
E
0 Relative Concentrotion of Stabilized Potossium Figure 3. Relative concentration of potassium stabilized after heating
the crystal at 673 K for 25 min under UHV plotted against the amount of oxidized aluminum present. The dotted line corresponds to a 1:l correlation between the amount of potassium and oxidized aluminum. I
-.-
I
I
I
IO
in
L
c
g
Y
.-in VI
f $ 5 rr)
I 2 6,-
0 0 0
c
a
I
0 20
I
I
60 00 '10 Free Iron S u r f a c e (Determined by CO T D S )
40
Figure 4. Initial rate of ammonia synthesis on a Fe(100) surface with submonolayer amounts of stabilized oxidized aluminum and potassium, shown as the rate of N H 3 synthesis vs. percentage of free iron surface as determined by "CO TPD. Conditions given in the text.
Figure 4 summarizes the results of the initial rate of ammonia synthesis on a Fe( 100) surface with submonolayer amounts of stabilized oxidized aluminum and potassium. The data is shown as the rate of ammonia synthesis vs. the percentage of free iron surface (as determined by 13C0 TPD). The reactions were performed at 723 K with a 20-atm stoichiometric mixture of nitrogen and hydrogen. There is an approximate linear decrease in the rate of NH3 synthesis with decrease in the amount of exposed iron surface (no ammonia production was detected from a surface completely covered with alumina and potassium). The results indicate that there is no promotional effect per exposed surface iron atom (Le., no increase in specific activity) at any of the surface concentrations of the A1203 K layers that were studied. All that is observed is a site-blocking-type effect on the rate. A similar series of experiments was performed with submonolayer amounts of oxidized aluminum alone on a Fe(100) surface,
+
and the same result was obtained; there was no promotional effect. The same is also observed for alumina and potassium on Fe( 110) and F e ( l l 1 ) surfaces. The addition of aluminum oxide and potassium to single-crystal iron surfaces results in poisoning rather than promoting ammonia synthesis. Even though we have simulated industrial synthesis conditions, our model catalyst did not in fact simulate the industrial catalyst so far as the effects of alumina and potassium are concerned. The nature of the stabilized potassium species in the industrial ammonia synthesis has been the cause for much debate for many years. Considering that K 2 0 remains in the industrial catalyst after many years of use, it has been suggested that K 2 0 is present in another chemical form, most likely in combination with other Our combined UHV and high-pressure kinetic results add further insight to the problem. We must infer from the data presented here that there is a strong interaction between aluminum oxide and potassium oxide. The approximate 1:1 ratio of surface potassium to aluminum oxide suggests that this interaction results in compound formation (e.g., potassium aluminate) rather than a nonstoichiometric potassium-aluminum oxide surface layer. Even though potassium can be stabilized with the coadsorption of aluminum oxide, the presence of the additives results in the inhibition of ammonia production. An explanation for this observation is that the amount of oxygen or aluminum oxide needed to bind the potassium under our conditions completely destroys the process of electronic promotion observed by Ert1.7,8 This lack of an electronic effect and the fact that we cannot stabilize any free potassium or substantial amounts of potassium oxide suggest that the effect of potassium in the industrial catalyst comes via a cooperative interaction with the other promoters, resulting in structural promotion. The addition of potassium to an industrial catalyst promoted only with A1203actually decreases the surface area of the ~ a t a l y s t Since . ~ ~ ~it is also known that the ammonia synthesis reaction is structure sensitive, the role of potassium with other promoters might be to create more active iron crystallographic planes. Work is in progress to probe this possibility.
Summary The results of a model system of the effects of potassium and alumina promoters on the ammonia synthesis reaction have been discussed. We find that elemental potassium is not stable on the surface of an iron single crystal under our reaction conditions. However, coadsorption of oxygen with the potassium stabilizes up to 10% of a monolayer of potassium at 20-atm synthesis gas pressure at the lower temperatures of the reaction (673 K), although this amount decreases continually with increase in reaction temperature. Substantially more potassium can be stabilized under our reaction conditons, and at much higher temperatures, by coadsorption with aluminum oxide. The cooperative interaction between the potassium and alumina may be due to compound formation. No electronic promotional effect is observed with submonolayer amounts of the stabilized A1203-K. The nature of the promotional effect in the industrial catalyst is uncertain at this time, but our results indicate that a more intimate interaction of iron with the other catalyst constituents is required, and this is undergoing further study. Acknowledgment. This work was supported by the Director, Office of Energy Research, Office of Basic Energy Sciences, Materials Science Division of the US.Department of Energy under Contract No. DE-AC03-76SF00098. Registry No. Al2O3, 1344-28-1; N,, 7727-37-9; H,,1333-74-0; 02, 7182-44-7; Fe, 7439-89-6; K, 7440-09-7.
(17) Brunauer, S.; Emmett, P. H. J . Am. Chem. SOC.1940, 62, 1732. (18) Nielsen. A. Cutul. Reu. Sci. Ena. 1981, 23. 17.
