Interaction of potassium with Fe3O4(111) at elevated temperatures

Applied Surface Science 161 Ž2000. 497–507 www.elsevier.nlrlocaterapsusc

Interaction of potassium with Fe 3 O4 ž111/ at elevated temperatures Sh.K. Shaikhutdinov ) , W. Weiss, R. Schlogl ¨ Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany Received 23 March 2000; accepted 28 April 2000

Abstract The surface structures formed by the annealing of a potassium overlayer on Fe 3 O4Ž111. were investigated by low energy electron diffraction ŽLEED., Auger electron spectroscopy ŽAES. and scanning tunneling microscopy ŽSTM.. Annealing at 600–7008C in vacuum or 10y6 mbar of oxygen resulted in well-ordered surface structures depending on the amount of potassium pre-deposited. As the K coverage increased, the surface transformed gradually from a Ž4 = 4. to a Ž2 = 2., and then, to a Ž1 = 1. structure relative to the original Fe 3 O4Ž111.-Ž1 = 1. surface. At low coverage, the Ž4 = 4. structure was ˚ periodicity, which exhibited an internal 6 A˚ periodicity formed by a long-range modulation of the surface with an ; 24 A characteristic of the Fe 3 O4Ž111.-Ž1 = 1. surface. At mid-coverage, two sorts of Ž2 = 2. domains were observed, which were distinguished by the different diameter of protrusions forming an STM image. They were attributed to the different states of potassium in the top layer. These domains coexisted on the surface and were found in both oxidative and vacuum preparations. At high K coverage, the surface exhibited a Ž1 = 1. structure with a high density of vacancy defects. Auger depth-profile measurements confirmed the diffusion of potassium into the iron oxide bulk at elevated temperatures. The formation of a non-stoichiometric K 2 OrK 2 Fe 22 O 34rFe 3 O4Ž111. interface with a gradually decreasing potassium concentration with depth has been suggested. q 2000 Elsevier Science B.V. All rights reserved. PACS: 68.35.B; 68.35.Bs; 81.65.M; 71.20.D Keywords: Alkali metals; Iron oxide; Potassium; Potassium iron oxide; Scanning tunneling microscopy; Surface reconstruction

1. Introduction The promoting effects of the alkali metals ŽAMs. are well established in catalysis w1,2x. Current progress in understanding AMs’ adsorption on metal surfaces has been recently reviewed by Diehl and Macgrath w3x, which revealed debates so far existing in this field. Less is known about AM interaction

)

Corresponding author. Fax: q49-030-8413-4105. E-mail address: [email protected] ŽS.K. Shaikhutdinov..

with oxide surfaces, since experimental data are relatively scarce Žsee review of Campbell w4x.. Most of the data relates to the adsorption on single crystal TiO 2 Ž110. and NiO surfaces Že.g., Refs. w5–7x and references therein., and only few of these studies have reported on the interaction of AMs with other oxide surfaces such as MgOŽ100. w8x, Cr2 O 3 Ž0001. w9x, FeOŽ111.rPtŽ111. w10x, and quite recently, on Na adsorption on the Fe 3 O4 Ž111.-terminated aFe 2 O 3 Ž0001. surface w11x. At submonolayer coverage, alkali ad-atoms occupy specific sites offered by the oxide lattice, which is frequently a hollow-like site formed by oxygen

0169-4332r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 4 3 3 2 Ž 0 0 . 0 0 3 7 3 - 1

498

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

anions so that AM atoms can maximize their coordination with the oxygen layer w4x. As coverage increases, an AM overlayer grows predominantly by the Stranski–Krastanov mode via the formation of three-dimensional islands on the first layer wetting the oxide surface w4x. Oxidation of thick AM overlayers at room temperature produced various AM oxides, their states were found to depend on the metal coverage, oxygen pressure and exposure w8,12x. Recently, we have investigated the structure and adsorptive properties of the different iron oxide surfaces, FeOŽ111., Fe 3 O4Ž111. and a-Fe 2 O 3 Ž0001., prepared by epitaxial growth onto a PtŽ111. substrate w13–18x. The oxide films were used as model catalysts for the dehydrogenation of ethylbenzene to styrene. This important industrial process is performed on potassium-promoted iron oxide catalysts in the presence of steam at ; 6008C. Potassium was found to increase the catalytic activity by about an order of magnitude w19a,19bx. For the catalysts preparation, the appropriate amounts of Fe 2 O 3 and K 2 CO 3 Ž; 10–30 wt.%. were calcined in air at ; 6008C. Muhler et al. w20,21x have suggested that ¨ the active catalyst phase under reaction conditions corresponds to a thin KFeO 2 film supported on a solid solution of K 2 Fe 22 O 34 in Fe 3 O4 . The K 2 Fe 22 O 34 phase acts as a storage medium from which the surface is continuously supplied with a near monolayer coverage of potassium ions. This potassium migration causes a continuous solid-state transformation of the catalyst during its lifetime. In order to investigate the surface structures of potassium-promoted iron oxides and a structural effect of potassium on a catalytic activity, we first need to prepare a well-defined potassium iron oxide surface for which there are no accepted procedures. Secondly, it is important to study the surface structures formed at reaction temperature, that is at about 6008C in the above-mentioned reaction. Therefore, in the present work, we use a deposition of metallic potassium onto the Fe 3 O4 Ž111. surface, followed by annealing at elevated temperatures. We varied the amount of potassium deposited, annealing temperature and ambient conditions Žoxygen, vacuum.. The resulting structures were investigated by low energy electron diffraction ŽLEED., Auger electron spectroscopy ŽAES. and scanning tunneling microscopy ŽSTM. at room temperature.

2. Experimental The experiments were performed in an ultra-high vacuum chamber Ža base pressure below 1 = 10y1 0 mbar. equipped with AES, LEED, STM and other surface science facilities w22x. Preparation of the Fe 3 O4 Ž111. epitaxial films has been described in detail elsewhere w13,18x. Briefly, metallic iron was evaporated onto a clean PtŽ111. substrate and then oxidized in 10y6 mbar O 2 at ; 7008C. By many deposition–oxidation cycles, an Fe 3 O4Ž111. film up ˚ thick can be prepared. to ; 500 A Potassium was evaporated onto a clean Fe 3 O4Ž111. substrate kept at room temperature, using a commercial ŽSAES. getter source. The pressure did not exceed 2 = 10y9 mbar during K deposition. The sample was placed at about 1.5 cm from the K source. The K coverage, measured by AES, was uniform throughout the sample. Subsequently, the KrFe 3 O4 Ž111. samples were annealed at 600–7008C in 10y6 mbar O 2 or in vacuum. In some preparations, we used repeated deposition–annealing Žor deposition–oxidation. cycles in order to get a high K coverage. AES measurements were performed with a 3 kV primary beam and a modulation amplitude of ; 10 V. STM images presented here were obtained at bias voltage of about 1.4 V applied to the sample and at tunneling current ca. 1 nA. In general, no significant changes were found by switching the bias polarity. 3. Results Fig. 1 displays side Ža. and top Žb. views of the Fe 3 O4 Ž111. surface. Previous dynamical LEED and STM studies combined with ab-initio calculations have proved that this surface is terminated by 1r4 ML of the Fe 3q ions in threefold hollow sites over the close-packed oxygen layer defined as 1 ML w14x. This creates the hexagonal Ž1 = 1. unit cell indicated ˚ lattice constant. The outerin Fig. 1b with a 5.94 A most Fe cations produce a hexagonal lattice of pro˚ periodicity in high resolution trusions with an ; 6 A STM images presented elsewhere w13,14,18x. 3.1. AES and LEED study of the K r Fe3 O4 surface Fig. 2a shows a typical AES spectrum of the KrFe 3 O4 Ž111. surface. The K Ž252 eV. signal is

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

499

the sputtered sample at 7008C in vacuum or in 10y6 mbar of oxygen resulted in a surface segregation of potassium as indicated by arrows Ža. and Žb. in Fig. 2c. Therefore, we cannot determine the amount of K in the top layer precisely, since AES averages the chemical composition of the surface as thick as the

Fig. 1. Slightly tilted side Ža. and top Žb. views of the ironterminated Fe 3 O4 Ž111. surface. Distance between equivalent oxy˚ The Ž1=1. unit cell with lattice gen ŽO1 –O1 . layers is 4.6 A. ˚ is indicated in Žb.. constant 5.94 A

decreased considerably after annealing the K overlayer at 3008C due to thermal desorption of the metallic potassium at ; 2508C w18x. As the temperature is further increased, the K concentration gradually decreases and drops more rapidly at T ) 5008C, but no potassium desorption is observed in this temperature range. This indicates that a significant part of the K atoms diffuses into the bulk. This conclusion is confirmed by Auger depth-profile measurements, as presented in Fig. 2b, showing temporal evolution of the K Ž252 eV., O Ž510 eV. and Fe Ž652 eV. peak intensities, while the surface annealed at 7008C is exposed to 500 eV Arq-ions at a current density of ; 2 mArcm2 . Under the conditions used, 5 min of ion bombardment corresponds to the removal of a few monolayers. Excluding a preferential sputtering effect, if any, the results show that potassium can penetrate into the oxide bulk ; 10 ML in depth depending on the sample pre-history: the more K pre-deposited, the deeper it is found. Annealing of

Fig. 2. Ža. A typical Auger spectrum of the samples studied. Žb, c. A temporal evolution of K Ž252 eV., O Ž510 eV. and Fe Ž652 eV. peak intensities Žb. and I K : IO and I Fe : IO intensity ratio Žc. while the KFe x O y surface is exposed to 500 eV Arq ions at current density ; 2 mArcm2 . Arrows Ža. and Žb. indicate the resulting effect of the annealing in vacuum at 6008C for 2 min Ža. and in 10y6 mbar O 2 at 7008C for 5 min Žb..

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

500

Fig. 3. LEED patterns of the samples prepared by oxidation in 10y6 mbar of O 2 at 7008C, possessing a different amount of potassium on the surface determined by the I K : IO ratio in AES spectra. Ža. Original Fe 3 O4 Ž111.; Žb. I K : IO s 0.45; Žc. I K : IO s 0.7; Žd. I K : IO s1.0. All the patterns were obtained at electron energy Es60 eV, except pattern Žb., which was taken at Es87 eV, in order to better represent the Ž4=4. superstructure. The corresponding unit cells are indicated.

˚ On the Auger electron escape length, i.e., 10–20 A. other hand, this allows us to estimate the chemical composition of the surface region using the general formalism of quantitative AES analysis for homogeneous compounds. Given the Auger sensitivity factor of 0.75, 0.5 and 0.2 for K Ž252 eV., O Ž510 eV. and Fe Ž652 eV. peaks, respectively w23x, the surface composition can be formally written as KFe xO y where x ; 1.1–0.9 and y ; 1.3–2.0 depending on K

coverage. We will use the standard term ‘‘coverage’’ just to indicate the overall amount of K within the surface region. To characterize the K coverage semi-quantitatively, we have used the K-to-O peak intensity ratio Ž I K : IO .. Therefore, we assign the terms ‘‘low’’-, ‘‘mid’’- and ‘‘high’’- coverage as corresponding to I K : IO - 0.5, ; 0.7 and ; 1.0, respectively. Fig. 3 shows the characteristic LEED patterns of samples all treated in 10y6 mbar O 2 at 7008C, but possessing different amounts of K. The LEED patterns from the K-doped surfaces exhibiting additional diffraction spots and a remarkably increased background intensity when compared, with the pattern of the original Fe 3 O4 Ž111. surface shown in Fig. 3a. At low K coverage, these spots appear as satellites around the spots coming from the original Fe 3 O4 Ž111.-Ž1 = 1. structure ŽFig. 3b.. This pattern can be described as a commensurate Ž4 = 4. superstructure relative to Fe 3 O4Ž111.-Ž1 = 1. as indicated by the corresponding unit cells. At K mid-coverage, a commensurate Ž2 = 2. structure is clearly seen in Fig. 3c. As the K coverage increases, the Ž2 = 2. structure vanishes, and the LEED pattern exhibits only the diffuse spots of a Ž1 = 1. structure as shown in Fig. 3d. Thus, the surface undergoes a Ž4 = 4. Ž2 = 2. Ž1 = 1. structural transformation as the K coverage increases. Combined AES and LEED data obtained on the different samples prepared by annealing in oxygen and in vacuum are summarized in Table 1. The table also contains values of the I Fe : IO ratio in order to monitor the stoichiometry of the Fe 3 O4Ž111. phase as a host material. For both preparations, the results show that the I Fe : IO ratio decreases as the I K : IO

™

™

Table 1 AES and LEED data obtained on the KrFe 3 O4 Ž111. samples annealed at 7008C in 10y6 mbar of oxygen and in vacuum Oxygen, 10y6 mbar AES

LEED

I K : IO Ž"0.03.

I Fe : IO Ž" 0.02.

0 0.45 0.7 0.95

0.32 0.27 0.27 0.26

a

Vacuum, 5 = 10y10 mbar a

Ž1 = 1. Ž1 = 1. q Ž2 = 2. w q Ž4 = 4. Ž1 = 1. q Ž2 = 2. s Ž1 = 1. q Ž2 = 2. w

AES

LEED

I K : IO Ž" 0.03.

I Fe : IO Ž"0.02.

0.45 0.5 0.7 1.0

0.28 0.26 0.25 0.26

Ž1 = 1. Ž1 = 1. q Ž2 = 2. w Ž1 = 1. q Ž2 = 2. s Ž1 = 1. q Ž2 = 2. w

The subscripts ‘‘w’’ and ‘‘s’’ indicate the ‘‘weak’’ and ‘‘strong’’ intensity of the corresponding diffraction spots relative to the Ž1 = 1. spots of the Fe 3 O4 Ž111.-Ž1 = 1. structure.

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

˚ 2 STM images of the KFe xO y Fig. 4. Large-scale 5000=5000 A surfaces. Ža. Oxidation temperature Tox s 7008C, I K : IO s 0.45; Žb. Tox s6008C, I K : IO s 0.7; Žc. the same sample as Žb. after further oxidation at 7008C. In Ža. and Žc., the terraces are sepa˚ high running along the main rated by monoatomic steps of ; 5 A crystallographic directions. Numerous islands develop on terraces by oxidation at 6008C, as seen in image Žb., which disappear after subsequent oxidation at 7008C Žc..

501

˚ 2 Žb, c. STM images of Fig. 5. 1000=1000 Ža. and 160=160 A the KFe x O y surface at low K coverage Ž I K : IO s 0.45.. Two different regions labeled a and b are clearly seen. They can be ˚ high, as measured by separated by steps of either 2.5 or 5 A profile lines shown below the image. Region a resolved in image ˚ Žb. exhibits a hexagonal lattice of protrusions with a ;12 A periodicity assigned to the Ž2=2. structure. Region b resolved in image Žc. is characterized by a close-packed structure of triangular ˚ domains, which in turn, is formed by the protrusions with a 6 A periodicity, as shown in the inset. A 2D-FFT obtained from the STM image Ža. is shown below. Three series of six equidistant ‘‘diffraction’’ spots, indicated by arrows, are aligned and corre˚ periodicity on the surface. The spond to 6, 12 and 24"2 A surface dislocations developed on both regions are indicated by the arrows.

502

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

ratio increases. This correlation is also observed during the ion sputtering experiments as shown in Fig. 2c, where the I Fe : IO ratio increases as potassium is removed by sputtering. 3.2. STM study of the K r Fe3 O4 surface STM images of the original Fe 3 O4Ž111. surfaces Žnot presented here for conciseness. revealed the ˚ wide terraces separated by atomically flat ; 1000 A ˚ in height. High-resolumonoatomic steps of ; 5 A tion STM images exhibited a hexagonal lattice of ˚ periodicity assigned to the protrusions with a 6 A outermost iron cations w14x. The large-scale STM image presented in Fig. 4a shows that the mesoscopic roughness of the potassium-doped iron oxide surface at low K coverage is not changed after oxidation at 7008C. Again, smooth large terraces, separated by the atomic steps running along the main crystallographic directions, were observed. At K mid-coverage, the surface oxidized at ˚ in 6008C exhibited numerous islands of ; 100 A size, which were randomly distributed on the terraces ŽFig. 4b.. When this surface was further oxidized at ˚ 7008C, only large triangular terraces of 500–1500 A in size were observed as shown in Fig. 4c. As the K coverage increased, no remarkable changes in the large-scale morphology were found. Close STM inspection of the surfaces at low K coverage revealed two regions labeled a and b in Fig. 5a. These regions were separated by steps of ˚ as measured by profile lines shown either 2.5 or 5 A below the image. There were also some dislocations seen within a-regions as protruding lines, for exam-

˚ 2 Žb. STM images of the Fig. 6. 230=230 Ža. and 490=490 A KFe x O y surface at K mid-coverage Ž I K : IO ; 0.7.. Image Ža. shows the fine structure of the islands observed in the large-scale ˚ STM image in Fig. 4b. Two Ž2=2. domains with a 12 A periodicity are visible in the upper and left-hand parts of the image Ža.. Left-bottom corner of the image Ža. is shown with a higher contrast in order to resolve the Ž2=2. symmetry of this region, but having much narrower protrusions. The half-widths of ˚ respectively. A few different the protrusions are ;6 and ; 4 A Ž2=2. domains neighbor on the same height in the image Žb.. The domain boundaries are indicated by the broken line. The profile line shown below reveals both types of protrusions to be in lateral registry.

ple, in the upper-right part of Fig. 5a. Regions a exhibit a hexagonal lattice of protrusions with an ˚ periodicity that is twice of the lattice con; 12 A ˚ .. stant on the Fe 3 O4 Ž111.-Ž1 = 1. surface Ž5.94 A

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

Therefore, this surface must be assigned to the Ž2 = 2. structure detected by LEED measurements presented in Section 3.1. The Ž1 = 1. sub-lattice cannot be resolved on these regions; therefore, a registry between Ž2 = 2. and Ž1 = 1. lattices is not determined. Fig. 5c shows a high resolution STM image of the b-region. The surface exhibits a long-range periodic ˚ modulation with a corrugation of up to ; 1.5 A, creating triangular domains in turn formed by protru˚ spacing therein. The atomic sions with an ; 6 A structure of the individual domain is shown in the inset of this image. On a large scale, the domains are arranged on the surface in a periodic manner as a nearly close-packed structure. This is evidenced by the two-dimensional fast Fourier transform Ž2D-FFT. of the STM images. One of the 2D-FFT maps obtained from the STM images, where both a and b regions have been resolved, is shown below Fig. 5b. Three series of six equidistant spots indicated by arrows are aligned and correspond to the hexagonal ˚ periodicstructures with an ; 6, ; 12 and 24 " 2 A ity and are therefore assigned to the Ž1 = 1., Ž2 = 2. and Ž4 = 4. surface structures, respectively. Fig. 6a shows the high resolution STM image of the surface, which exhibited the numerous patches in the large-scale STM image presented in Fig. 4b. Again, domains showing a hexagonal structure with ˚ periodicity corresponding to the Ž2 = 2. a 12 A structure are imaged. However, the surface region shown with a higher contrast in the left-bottom corner of the image also manifests the Ž2 = 2. symmetry, but the protrusions forming it are significantly

503

narrower. Such a region is better resolved in Fig. 6b, where two sorts of Ž2 = 2. domains neighbor on the same height level. The dislocation splitting the surface into a few domains is clearly seen in this image as indicated by the broken lines. The half-widths of the protrusions are measured to be of ; 6 and ; 4 ˚ for the surfaces termed ‘‘ball’’ – and ‘‘dot’’ – A surfaces, respectively. The fact that the ‘‘ball’’surface is normally clean while the ‘‘dot’’-surface looks like covered by adsorbates also indicates their distinct chemistry. The profile line measured over a single domain and shown below the image reveals that both types of protrusions are in registry, which means that they occupy the same sites. Another structure which is resolved in the righthand part of Fig. 6a is the corrugated surface, which ˚ periodicity. This structure exhibits an internal ; 6 A resembles the a-regions shown in Fig. 5c, which we assigned to the Ž4 = 4. structure. It seems plausible that these domains, which have been detected after oxidation at lower temperature Ž6008C for 2 min. and have disappeared after further treatment at 7008C for 5 min, correspond to the surface not being completely transformed to the Ž2 = 2. structure. At high K coverage, STM images revealed fewer domains with the Ž2 = 2. structure. Most of the surface exhibits a hexagonal Ž1 = 1. structure with ˚ periodicity, however, it is characterized by an ; 6 A a high density of vacancy defects ŽFig. 7a.. This observation is in line with the LEED pattern shown in Fig. 3d, where the ‘‘Ž2 = 2.’’ spots vanish and the ‘‘Ž1 = 1.’’ spots get diffused due to the highly defec-

˚ 2 Žb, c. STM images of the KFe xO y surface at high K coverage Ž IK : IO s 1.0.. The surface mostly Fig. 7. 165 = 165 Ža. and 220 = 220 A exhibits a Ž1 = 1. structure with a high density of vacancy defects as shown in Ža.. Images Žb. and Žc. show an interface between Ž1 = 1. and Ž2 = 2. ‘‘dot’’-structures indicated.

504

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

˚ 2 Žb. STM images of the potassium iron oxide surface prepared by annealing in vacuum at 7008C and Fig. 8. 210 = 210 Ža. and 100 = 100 A corresponding to Auger ratio I K : IO s 0.7 Ža. and I K : IO s 1.0 Žb.. Again, two Ž2 = 2. surfaces, having different width of protrusions, are ˚ step in image Ža.. At high coverage, the Ž2 = 2. structure vanishes and the surface exhibits a highly defective Ž1 = 1. separated by a 2.5 A structure resolved in image Žb..

tive surface. STM images in Fig. 7b,c display an interface between the Ž1 = 1. and Ž2 = 2. domains, which show that the ‘‘Ž2 = 2.’’-dot protrusions are in registry with those forming the Ž1 = 1. lattice. This means that all the protrusions on these images occupy the same sites. The surfaces prepared by vacuum annealing exhibit structures similar to those observed by preparation in oxidative conditions Žsee Table 1 Two STM images in Fig. 8 illustrate this: the image Ža. presented with a differentiated contrast shows the Ž2 = 2. ‘‘ball’’ — and ‘‘dot’’ — surfaces separated by a ˚ the image Žb. shows the highly step of ; 2.5 A; defective Ž1 = 1. surface formed at high K coverage. 4. Discussion We start preparation of the K-doped iron oxide surface with the deposition of metallic potassium at room temperature. In order to observe any changes in the surface structures of annealed samples, we deposited relatively high amount of potassium, i.e., about 5 ML as estimated from the AES spectra. The corresponding LEED patterns showed very weak and diffuse diffraction spots, if any, coming from the Fe 3 O4 Ž111.-Ž1 = 1. substrate. Therefore, both AES and LEED results indicate a thickness of K overlayer no more than a few monolayers.

Annealing of the K-covered iron oxide surface at 600–7008C strongly decreases the amount of potassium on the surface. This is in part due to the K ad-atoms desorbing at ; 2508C w18x, however, a significant part of the K atoms reacts with the oxide. Auger depth-profile measurements presented in Fig. 2 show that the potassium concentration gradually decreases with increasing depth and can be detected at ; 10 ML deep. Interaction of potassium and iron oxide forms a non-stoichiometric surface phase, which could be formally written as KFe xO y , where x and y vary between 0.5–1 and 1–2, respectively. A search in the structural database w24x resulted in several ternary potassium iron oxide compounds, which show comparable amounts of K and Fe. They are KFeO 2 Žan orthorhombic space group Pbca, lattice parameters ˚ ., K 3 FeO 2 Ža a s 5.60, b s 11.25 and c s 15.90 A tetragonal space group P4 1 2 1 2, a s b s 6.05, c s ˚ ., K 3 FeO4 Žan orthorhombic space group 14.03 A ˚ ., K 6 Fe 2 O6 Pnma, a s 7.70, b s 9.09 and c s 7.84 A Ža monoclinic space group C 1 2rm 1, a s 7.13, ˚ . and K 3 FeO4 Žan orthorb s 11.12 and c s 6.51 A hombic space group Pnma, a s 7.70, b s 5.86 and ˚ .. Note that Fe 3 O4 has a space group c s 10.34 A ˚ Fd3rm and lattice parameters a s b s c s 8.4 A. We could not find any plane in the above-listed compounds, which would exhibit a hexagonal sym-

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

metry and fit the hexagonal Fe 3 O4 Ž111. substrate surface. In addition, a formation of such a compound on the surface must drastically change a LEED pattern. In contrast, Fig. 3 shows no detectable change of Ž1 = 1. lattice parameter at any K coverage studied Žmore precise SPA-LEED measurements could prove this.. It should also be mentioned that, for example, the KFeO 2 compound is normally prepared by calcination of Fe 2 O 3 and K 2 CO 3 at ; 10008C in air. A thin KFeO 2 film was observed by Muhler et ¨ al. w21x on technical K-promoted iron oxide catalysts only under reaction conditions of dehydrogenation of ethylbenzene to styrene in the presence of steam at 6008C and at atmospheric pressure. Both preparation conditions are much more severe than those used in this present work. Therefore, we can rule out the formation of KFeO 2 and the other phases mentioned as being responsible for the surface structures found. Meanwhile, AES results show that the I Fe : IO ratio decreases as the I K : IO ratio increases Žsee Table 1 and Fig. 2c.. In other words, the more potassium, the less iron within the surface region. This fact, together with the conservation of surface symmetry during solid state reaction, could be explained by assuming that potassium substitutes a fraction of the iron ions. It has been previously suggested that the AM adsorption on metals can produce such a ‘‘substitutional’’ phase at high AM coverage Žsee review in Ref. w3x.. On the other hand, ˚ . is much larger the ionic radius of potassium Ž1.33 A ˚ than that of the iron cation Ž0.78 A. w25x. Therefore, this substitutional phase would result in a strong strain, and hence, a large number of dislocations and other structural defects, as were actually found in the STM images presented in Figs. 5–7. Another possible candidate for the surface phase produced is K 2 Fe 22 O 34 or K 2 O P ŽFe 2 O 3 .11 , which has a hexagonal space group P6 3rmmc Žlattice pa˚ . and is formed rameters a s 5.92 and c s 23.79 A from two spinel-like blocks with a layer of potassium oxide in between w26x. The potassium layer is not dense, giving rise to the non-stoichiometry of K:Fe between 1:11 and 1:5 w20x. This surface phase would fit the hexagonal Fe 3 O4Ž111. substrate well, since it has a lattice constant identical to the original ˚ . and would maintain the Fe 3 O4 Ž111. surface Ž5.94 A surface unit cell as judged by LEED. However, the potassium-to-iron ratio in the latter compound is

505

much lower than that measured by AES on the surfaces prepared, i.e., about 1:1, assuming a homogeneous potassium distribution in the near surface region. This indicates that the top layers are enriched with potassium, which can be attributed, for example, to the surface terminating with potassium oxide. Formation of the latter seems quite reasonable during oxidation of the K overlayer in 10y6 mbar of O 2 . Indeed, it has been found that interaction of oxygen with K multilayers on SiŽ111. and MgOŽ100. substrates results in various potassium oxide phases, depending on oxygen pressure and temperature w8,12x. However, the surface structures formed on iron oxide in this work are very similar for both the oxidative and the vacuum preparations. This indicates that potassium reacts with lattice oxygen of iron oxide as well. A similar conclusion has been drawn by Chen et al. w5x based on a HREELS and TPD study of the KrNiOrNiŽ100. surface annealed to 900 K in vacuum. Ion surface scattering experiments performed by Vurens et al. w10x also revealed that potassium deposited on an FeO monolayer grown on PtŽ111. remains in the outermost layer even after heating to 950 K. Thus, we suggest that the interaction of potassium with Fe 3 O4Ž111. at elevated temperature leads to the formation of a non-stoichiometric K 2 OrK 2 Fe 22 O 34rFe 3 O4 interface with a gradually decreasing potassium content from the surface into the oxide bulk. The surface structures formed depend on the K coverage. At low K coverage, the LEED pattern in Fig. 3b and the STM image in Fig. 5c show the Ž4 = 4. superstructure, which is created by a long-range modulation of the surface in turn formed by protru˚ periodicity, which is characteristic sions with a 6 A for the Fe 3 O4 Ž111.-Ž1 = 1. surface. The atomic resolution image in the inset in Fig. 5c can be interpreted as an atom adsorbed on a top site, which modify the electronic structure of the neighboring atoms and hence, result in apparent modulation of the Fe 3 O4 Ž111. surface. Another explanation is that this modulation is caused by the K-induced surface de-relaxation of the original Fe 3 O4 Ž111. surface, which is known to be strongly relaxed under vacuum conditions w27x. It is interesting to note that if we disregard imperfect ordering, the STM images of the Ž4 = 4. struc-

506

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

ture look similar to those obtained on layered compounds, which exhibit a charge density waves ŽCDW. phenomenon, for example, on 1T-TaS 2 w28x. The CDW structure is formed by the so-called ‘‘Star-ofDavid’’ clusters including 13 Ta atoms and manifests itself in STM images as a hexagonal long-range ˚ periodicity of the surmodulation with an ; 12 A face formed by a hexagonal lattice with an internal ˚ periodicity. This similarity can suggest of ; 3 A some interesting analogies, since the charge transfer from potassium to oxide obviously occurs due to the lower ionization potential of potassium, which redistributes the ratio of Fe 3q and Fe 2q states within a surface region as found by XPS w11x. At K mid-coverage, the Ž2 = 2. structure covers almost the entire surface. However, we observe two different Ž2 = 2. structures, which are formed by ˚ protrusions with different widths, ca. 6 and 4 A, respectively ŽFig. 6.. Previous STM studies of the AM adsorption on oxide surfaces show that the AM atoms are imaged as protrusions w6,29x. ŽWith the metal substrates, the STM results are somewhat confused since, for example, potassium on CuŽ110. has been imaged as depression or protrusion depending on the tip conditions and the tunneling parameters w30x.. However, the AMs were always found to form the outermost layer even after heating to elevated temperatures. Therefore, we assigned the STM protrusions to the position of the K atoms in the top layer. Our STM observation of the two different domain structures indicates the different states of potassium in the top layer or the different terminations of the non-stoichiometric potassium iron oxide phase suggested above. As the K coverage increases, the Ž2 = 2. structure vanishes and the surface again manifests the Ž1 = 1. symmetry. This surface can definitely not be attributed to the original Fe 3 O4 Ž111.-Ž1 = 1. surface, because there is too much potassium within the surface region. The Ž1 = 1. structure is probably produced from the Ž2 = 2. structure, as the K coverage gets saturated. The STM images in Fig. 7b,c show that both Ž2 = 2. and Ž1 = 1. protrusions are in registry, which means that they occupy the same ˚ sites. The Ž2 = 2.-‘‘dot’’ protrusions are only 0.1 A higher, this small height difference could be attributed to the different surface relaxation within the Ž2 = 2. and Ž1 = 1. unit cells.

One should also keep in mind that the STM tip probes the local density of electronic states, therefore, an STM image includes not only a topographic, but also an electronic factor. It means, for example, that a ‘‘guest’’ atom embedded into a layer of the ‘‘host’’ material could be seen as a protrusion or a depression atom despite having nearly the same geometrical size. Therefore, the surface composition and structure of the top layers formed during a K– Fe 3 O4 Ž111. interaction at elevated temperatures could not be deduced from STM measurements only. Further experiments using photoelectron spectroscopy and ion scattering spectroscopy are in progress in order to elucidate a detailed structure of potassium iron oxide surfaces. Finally, our results show that the potassium iron oxide surface undergoes a consecutive Ž4 = 4., Ž2 = 2., Ž1 = 1. reconstruction as the K coverage increases. In the case of ‘‘simple’’ adsorption, this type of coverage dependence indicates a repulsive interaction between adsorbed species. But in the present study, the Ž N = N .-structures were formed after heating at elevated temperature only, and not immediately after K deposition at room temperature. However, this analogy seems quite interesting.

5. Conclusions Well-ordered potassium-doped iron oxide surfaces are formed by deposition of metallic potassium on the Fe 3 O4Ž111. surface and subsequent annealing at 600–7008C. The results show a close similarity between the surface structures formed by heating either in 10y6 mbar oxygen or in an ultra-high vacuum. It appears that these structures are thermodynamically stable and independent of the conditions used. Auger depth-profile measurements confirm the diffusion of potassium into the iron oxide bulk at elevated temperatures. We suggest that the interaction of potassium and Fe 3 O4Ž111. leads to the formation of a non-stoichiometric K 2 OrK 2 Fe 22 O 34r Fe 3 O4 interface with a gradually decreasing potassium concentration with depth. The surface structures formed depend on the K coverage. The surface transforms gradually from a Ž4 = 4. to a Ž2 = 2., and then, to a Ž1 = 1. structure relative to the Fe 3 O4 Ž111.-Ž1 = 1. surface as the K

Sh.K. ShaikhutdinoÕ et al.r Applied Surface Science 161 (2000) 497–507

coverage increases. The relative brightness of the diffraction spots in the LEED patterns correlates with the surface area covered by the corresponding structures resolved by STM. The Ž4 = 4. structure is formed by long-range ˚ periodicmodulation of the surface with an ; 24 A ˚ periodicity which, in turn, exhibits an internal 6 A ity characteristic for the Fe 3 O4 Ž111.-Ž1 = 1. surface. ˚ periodicTwo sorts of Ž2 = 2. domains with a 12 A ity are clearly distinguished in the STM images by the width of protrusions. At high K coverage, the surface exhibits a Ž1 = 1. structure with a high density of vacancy defects. The potassium-enriched surface phase suggested in the present work is very similar to that determined as a precursor for the active phase of styrene synthesis on K-promoted iron oxide catalysts. However, our study clearly indicates that an interaction of metallic potassium and iron oxide at elevated temperatures in vacuum or low oxygen pressure does not result in the KFeO 2 phase, which is formed only during catalytic reaction.

Acknowledgements The authors thank Dr. W. Ranke for helpful discussions.

References w1x H.P. Bonzel, A.M. Bradshaw, G. Ertl ŽEds.., Physics and Chemistry of Alkali Metal Adsorption, Elsevier, Amsterdam, 1989. w2x M. Kiskinova, in: B. Delmon, J.T. Yates ŽEds.., Poisoning and Promotion in Catalysis Based on Surface Science Concepts and Experiments, Elsevier, Amsterdam, 1992. w3x R.D. Diehl, R. Mcgrath, J. Phys.: Condens. Matter 9 Ž1997. 951. w4x C.T. Campbell, Surf. Sci. Rep. 27 Ž1997. 1.

507

w5x J.G. Chen, M.D. Weisel, J.H. Hardenbergh, F.M. Hoffman, C.A. Mims, R.B. Hall, J. Vac. Sci. Technol., A 9 Ž1991. 1684. w6x P.W. Murray, N.G. Condon, G. Thornton, Surf. Sci. Lett. 323 Ž1995. L281. w7x A.F. Carley, S.D. Jackson, J.N. O’Shea, M.W. Roberts, Surf. Sci. Lett. 440 Ž1999. L868. w8x H.H. Huang, X. Jiang, Z. Zou, G.Q. Xu, W.L. Dai, K.N. Fan, J.F. Deng, Surf. Sci. 398 Ž1998. 203. w9x M. Wilde, I. Beauport, K. Al-Shamery, H.-J. Freund, Surf. Sci. 390 Ž1997. 186. w10x G.H. Vurens, D.R. Strongin, M. Salmeron, G.A. Somorjai, Surf. Sci. Lett. 199 Ž1988. L387. w11x J. Nerlov, S.V. Hoffmann, M. Shimomura, P. Moller, Surf. Sci. 401 Ž1998. 56. w12x B. Lamontagne, F. Semond, D. Roy, Surf. Sci. 327 Ž1995. 371. w13x W. Weiss, M. Ritter, Phys. Rev. B 50 Ž1999. 5201. w14x Sh.K. Shaikhutdinov, M. Ritter, X.-G. Wang, H. Over, W. Weiss, Phys. Rev. B 60 Ž1999. . w15x Sh.K. Shaikhutdinov, W. Weiss, Surf. Sci. Lett. 432 Ž1999. L627. w16x D. Zscherpel, W. Ranke, W. Weiss, R. Schlogl, ¨ J. Chem. Phys. 108 Ž1998. 9506. w17x W. Ranke, W. Weiss, Surf. Sci. 4141 Ž1998. 238. w18x Sh.K. Shaikhutdinov, Y. Joseph, C. Kuhrs, W. Ranke, W. Weiss, Faraday Discuss. 114 Ž1999. 363. w19ax T. Hirano, Appl. Catal. 26 Ž1986. 65. w19bxT. Hirano, Appl. Catal. 28 Ž1986. 119. w20x M. Muhler, J. Schutze, M. Wesemann, T. Rayment, A. Dent, ¨ ¨ R. Schlogl, ¨ G. Ertl, J. Catal. 126 Ž1990. 339. w21x M. Muhler, R. Schlogl, ¨ ¨ G. Ertl, J. Catal. 138 Ž1992. 413. w22x W. Weiss, M. Ritter, D. Zscherpel, M. Swoboda, R. Schlogl, ¨ J. Vac. Sci. Technol., A 16 Ž1998. 21. w23x L.E. Davis, N.C. MacDonald, P.W. Palmberg, P.E. Riach, R.E. Weber, Handbook of Auger Electron Spectroscopy, Physical Electronics Industries, Eden Prairie, MN, 1976. w24x ICSD Collection Codes 200255, 73215, 65977, 2487. w25x C. Kittel, Introduction to Solid State Physics, Wiley, New York, 1986. w26x JCPDS Card w31 1034x. w27x W. Weiss, M. Ritter, Phys. Rev. B 60 Ž1999. 5201. w28x Sh.K. Shaikhutdinov, V.P. Babenko, D.I. Kochubey, Phys. Low-Dimens. Struct. 11r12 Ž1996. 109. w29x S.J. Murray, D. Abrooks, F.M. Leibsle, R.D. Diehl, R. McGrath, Surf. Sci. 314 Ž1994. 307. w30x R. Shuster, J.V. Barth, G. Ertl, R.J. Behm, Phys. Rev. B 44 Ž1991. 13689.