10978
J. Phys. Chem. B 2005, 109, 10978-10985
Growth, Structure, and Stability of Ceria Films on Si(111) and the Application of CaF2 Buffer Layers Jeannette Zarraga-Colina,† Roger M. Nix,*,† and Helmut Weiss‡ Department of Chemistry, Queen Mary, UniVersity of London, Mile End Road, London E1 4NS, United Kingdom, and Institute of Chemistry, Otto Von Guericke UniVersity Magdeburg, UniVersita¨tsplatz 2, D-39106 Magdeburg, Germany ReceiVed: February 17, 2005; In Final Form: April 4, 2005
The growth of ceria (CeO2) films by oxidation of evaporated Ce metal on Si(111) and on CaF2(111) epilayers on Si(111) is compared. By use of XPS, UPS, and LEED, it has been demonstrated that the application of a CaF2 buffer layer between the ceria and Si substrate prevents the formation of an amorphous oxidized Si layer at the interface and permits the growth of a well-defined epitaxial ceria layer of (111) surface orientation. The thermal stability of the CeO2/CaF2/Si(111) interface structure is limited by the solid-state reaction between CaF2 and ceria. This leads to gradual migration of fluorine into the oxide at elevated temperatures to give a solid-state solution of fluorine in the partially reduced oxide. An analysis of the composition observed after extensive annealing in a vacuum suggests that, with initial layers of CaF2 and CeO2 of similar thickness, the ultimate product may be CeOF. The onset of this solid-state reaction can, however, be significantly delayed by annealing under an oxygen atmosphere.
1. Introduction The heteroepitaxial growth of cerium oxide (ceria) on silicon is of interest because of the potential applications of ceria as a dielectric in silicon-based microelectronic devices and as a buffer layer in high-Tc superconductor/Si systems [refs 1 and 2, and references therein]. Si(111) also appears to be an ideal substrate (with very small lattice mismatch) for the growth of epitaxial ceria films for use in fundamental studies of the catalytic properties of ceria, which is a key component in three-way catalysts for the control of vehicle emissions.3 However, it is clear from our own work and that of other researchers1,4-9 that the oxidation of cerium metal deposited onto Si(111) generally results in oxidation of the silicon surface and the formation of an amorphous interfacial layer. Some success in circumventing this problem has been achieved by passivating the silicon surface by hydrogen termination and by pulsed laser deposition of the oxide.10,11 More recently, we have communicated an alternative technique which involves the use of a CaF2 film as an intermediate layer between ceria and Si(111) to avoid the formation of disordered silicon oxides.12 This approach is based on the very similar lattice parameters of the three materials involved: Si (Fd3hm, a ) 5.431 Å), CaF2 (Fm3hm, a ) 5.463 Å), and CeO2 (Fm3hm, a ) 5.411 Å).13 The CaF2 forms a welldefined epitaxial film on the Si(111) surface, and can itself then act as a template for the epitaxial growth of ceria with a (111) orientation. In this paper, we report on a more extensive study of the epitaxial growth of ceria on CaF2/Si(111) substrates using surface analytical techniques, with emphasis on the growth of films of varying thickness of ceria and their thermal stability under different environments. For comparative purposes, we include reference to similar studies on the ceria/Si(111) system. The characterization of the growth process and the ceria films * Corresponding author. Telephone: +44-20-7882-3273. Fax: +44-207882-7427. E-mail:
[email protected]. † University of London. ‡ Otto von Guericke University Magdeburg.
was carried out by X-ray photoelectron spectroscopy (XPS), ultraviolet photoelectron spectroscopy (UPS), and low energy electron diffraction (LEED). 2. Experimental Section The experiments were carried out in a stainless steel ultrahigh vacuum (UHV) system, equipped with XPS (VG CLAM-2 analyzer; unmonochromated twin-anode Mg KR/Al KR X-ray source), UPS (He discharge source; hν ) 21.2 eV), LEED, and temperature-programmed desorption (TPD). The Si(111) substrate was cleaved from n-type wafer (Asdoped; 0.06-0.08 Ω cm) and the temperature was monitored using a chromel-alumel thermocouple, welded to a Ta foil attached to the rear of the crystal. The native surface oxide was removed by initial outgassing at 650 °C and then heating to 800 °C, while keeping the pressure below 5 × 10-9 Torr. Subsequent cleaning was carried out by 1 keV Ar+ sputtering, after which the crystal was annealed at ca. 770 °C in a vacuum. These procedures gave a Si(111) surface that showed no contamination by XPS and gave a high-quality (7 × 7) LEED pattern. CaF2 was deposited by sublimation from high-purity CaF2 held in a resistively heated tungsten wire basket. After deposition onto the Si substrate at room temperature, the CaF2 film was annealed at 750 °C for 3 min. For the work described here, we have used CaF2 films of ca. 15 ML (monolayer) thickness, which exhibit a sharp (1 × 1) LEED pattern, characteristic of a welldeveloped epitaxial layer. Ce metal evaporation was achieved using a source of similar design. The deposition of cerium metal (∼3 ML) was carried out at room temperature; this was then followed by oxidation (generally 10 langmuirs (L) O2) at room temperature and finally the sample was annealed at 300 or 350 °C in 1 × 10-8 Torr O2 for 5 min. The growth of thick CeO2 films was carried out using several cycles of the above procedure.
10.1021/jp0508296 CCC: $30.25 © 2005 American Chemical Society Published on Web 05/06/2005
Growth of Ceria Films on Si(111) Studies of the thermal stability of the ceria films were carried out by annealing in both vacuum and various pressures of oxygen, and the surfaces were kept at the indicated temperatures for 5 min. When the ceria films were annealed under the highest pressures studied (1 × 10-7 Torr O2), the sample was cooled to just below 200 °C before the oxygen was removed from the chamber. The electron energy analyzer was calibrated using the Cu 2p, Cu LMM, Cu 3p, and Fermi level photoemission signals from a Cu(111) single crystal, and all XPS and UPS spectra from the sample under study have been acquired with the Si(111) crystal grounded. Binding energy shifts were very noticeable in certain circumstances when measuring spectra of samples employing CaF2 buffer layers, in particular, for thin (e3 ML) oxide films grown on CaF2 and after high-temperature annealing of thicker oxide films grown on CaF2 in UHV. These shifts affected all the signals from both the CaF2 and ceria layers. In these particular cases the binding energies that we report (and any displayed spectra) have been corrected by referencing to particular peaks in the spectrum. Having compared a very large number of spectra, we have used the highest binding energy (u′′′) component of the Ce 3d spectrum of the Ce4+ of the oxide as a primary reference, assigning it a binding energy value of 916.9 eV, in accord with previous work on metallic substrates14 where such problems were not apparent. In those cases where this peak was absent we have corrected the binding energies on the basis of shifts observed in the other peaks (generally in the Ca 2p3/2 and F 1s peaks). XPS data were routinely acquired using Mg KR radiation (10 kV, 120 W), a pass energy of 50 eV, and an emission angle of 25° to the surface normal, but in some cases spectra were also acquired at grazing emission angles (ca. 75° to the normal). Film thicknesses have been estimated from the attenuation of the XPS signals from the underlying Si substrate (for Ce/Si, CeO2-x/Si, and CaF2/Si) and from the intermediate CaF2 film (for Ce/CaF2/Si and CeO2-x/CaF2/Si), using inelastic mean free path (IMFP) values estimated by the TPP-2M algorithm,15 ignoring the effects of elastic scattering and any changes to the IMFP that may arise from nonstoichiometry of the oxide. The cerium oxide film thicknesses calculated using this procedure are given in monolayers (ML), where the monolayer thickness has been taken to be 3.1 Å (the layer spacing in the 〈111〉 direction of the bulk oxide). Estimates of the oxide stoichiometry were made by determining the Ce3+/Ce4+ ratio from a constrained deconvolution of the Ce 3d emission profile using previously optimized fits to the pure Ce3+ and Ce4+ spectra, as described in previous work by the group.16 3. Results and Discussion 3.1. Oxidation of Thin Cerium Metal Films on Si(111) and CaF2/Si(111). To enable a direct comparison, experiments were carried out on the oxidation of thin films (ca. 3 ML) of cerium metal deposited directly on Si(111) and deposited on CaF2/ Si(111). In each case the initial, stepwise (0.5-240 L O2) oxidation was carried out at room temperature, but the samples were then annealed at 300 °C in 1 × 10-8 Torr O2 for 5 min, before cooling back to ambient under vacuum. In the case of 3 ML Ce on Si(111) the data of Figure 1A show that the cerium metal layer is characterized by single Ce 3d5/2 and Ce 3d3/2 peaks around 884.1 and 902.6 eV, respectively, while the O 1s spectra confirm that only trace levels of oxygen are present. During the early stages of oxygen uptake there is development of signals in the Ce 3d XPS spectra that are characteristic of Ce3+, but the lack of definition of the
J. Phys. Chem. B, Vol. 109, No. 21, 2005 10979
Figure 1. Ce 3d XPS spectra for the oxidation of thin cerium metal films: (A) 3 ML of Ce on Si(111) and (B) 2.5 ML of Ce on CaF2/ Si(111). (a) Freshly deposited Ce metal, subjected to oxygen exposures of (b) 0.5, (c) 1.5, (d) 4, (e) 40, (f) 90, and (g) 240 L and (h) after annealing at 300 °C in a further 15 L O2. [Exposures at room temperature unless otherwise specified; the location of the four main peak components for Ce3+ and six main peak components for Ce4+, which arise from a combination of spin-orbit splitting and valence configuration effects in the final state, are shown in part B.]
doublet associated with each spin-orbit final state indicates that there is some remaining contribution from cerium metal peaks. Increasing the oxygen exposure leads to a better definition of the four main Ce3+ peaks in the spectra, while for exposures of 4 L and greater there is evidence for the presence of some Ce4+ in addition to the majority Ce3+ species, as indicated by the onset of the growth of a peak at ∼917 eV. After 240 L exposure, an analysis of the near-normal-emission Ce 3d profile would suggest an oxide stoichiometry of around CeO1.59, while the corresponding analysis of the grazing emission data would suggest a fractionally higher value of CeO1.64. After annealing at 300 °C the proportion of cerium present as Ce4+ was significantly reduced, and the spectra are again similar to those expected for a Ce2O3 sample. Immediately after cerium metal deposition, the Si 2s XPS spectra showed the single peak at ca. 150.8 eV that would be expected for the underlying elemental Si substrate. However, after even low oxygen exposures, a shoulder attributable to oxidized silicon was evident at higher binding energy (ca. 153
10980 J. Phys. Chem. B, Vol. 109, No. 21, 2005 eV).1,11 The presence of oxidized silicon is also apparent in the O 1s spectra for such films which are characterized by a rather broad and frequently asymmetric photoemission peak that is clearly composed of contributions from oxygen atoms in at least two distinct chemical environments (a peak attributable to oxidized cerium with a binding energy of around 530 eV, and a peak with binding energy of ∼532 eV). In such cases the ratio of the total integrated O 1s to Ce 3d intensity is also anomalously high compared to that observed for CeO2 itself, again suggesting that there is additional oxygen directly associated with the silicon. In addition, a peak at ca. 8.5 eV is observed in the UPS spectrum, at an energy similar to that seen during the initial oxidation of the clean Si(111) surface.17 In LEED, the (7 × 7) pattern of the clean Si(111) surface was replaced after Ce deposition by a faint (1 × 1) pattern, just visible above a high background. After exposure to low levels of oxygen no LEED pattern was evident. The characteristic reactivity of Ce/Si(111) described above may be compared with observations made during the deposition and oxidation of 2.5 ML of Ce on CaF2/Si(111), for which the corresponding Ce 3d XPS spectra are shown in Figure 1B. In this instance, the lowest exposure of O2 results in almost complete conversion to Ce3+ (i.e., a film of Ce2O3 stoichiometry). After a nominal exposure of 1.5 L O2 the Ce 3d spectral profile shows the presence of substantial amounts of Ce4+, in addition to Ce3+, and for exposures greater than ca. 20 L the spectrum is dominated by the Ce4+ photoemission signals, with more detailed analysis suggesting a stoichiometry of around CeO1.92 (with no significant difference evident between the normal emission and grazing emission data). No new features were evident in the Si 2s, Ca 2p, and F 1s XPS spectra as a result of the room-temperature oxidation of the cerium. In this particular experiment studying the Ce oxidation, the peak in the O 1s spectrum associated with the oxide lattice was accompanied by a broad shoulder located ca. 2.0-2.3 eV to higher binding energy. This peak was not normally observed in other experiments of oxide growth on CaF2/Si(111), and there was no indication from the Ca 2p spectrum that it is associated with oxidation of the CaF2 layer. Instead we believe that it is most likely to be attributable to hydroxyl groups formed by the adventitious adsorption of water vapor from the background gases. The observation of this peak was generally limited to experiments where either incompletely oxidized Ce metal was exposed to the residual gases for an extended period (as in this case where measurements were made in many small increments of oxygen exposure), or when oxide films were extensively annealed at high temperature (in which case a partially reduced oxide surface can, during cooling, come into contact with water vapor liberated by outgassing of the sample holder). Support for this assignment to hydroxyl groups comes from other measurements on these ceria surfaces in which the appearance of features at ca. 6.0-6.5 and 10.0-10.5 eV was noted in UPS data; similar features have been seen in the valence band spectra when other rare earth oxides are deliberately exposed to water vapor.18 This interpretation would be consistent with the conclusion reached by several groups studying cerium oxide surfaces and the oxidation of cerium metal.19-21 However, other authors have favored alternative explanations22,23 and assigned this peak to oxygen ions with unusual coordination in a defective ceria structure, and we cannot exclude the possibility that the O 1s shoulder that we sometimes observe is also due, at least in part, to such species. The (1 × 1) LEED pattern that characterizes the epitaxial layer of CaF2 on Si(111) is essentially unaffected by the
Zarraga-Colina et al.
Figure 2. LEED patterns observed during preparation of ceria films on CaF2/Si(111) substrate: (A) clean CaF2/Si(111) at 82 eV, (B) following deposition of 2.5 ML of Ce and oxidation with 240 L O2 at room temperature at 77 eV, and (C) for a 15 ML oxide layer, annealed to 600 °C in 1 × 10-8 Torr O2 at 70 eV.
deposition of cerium and its subsequent oxidation, with no detectable deterioration in quality (Figure 2A,B). This suggests that the thin film of ceria is growing in an epitaxial fashion on the CaF2(111) surface, even at ambient temperature. After annealing at 300 °C in oxygen (again cooling back to ambient under vacuum) there was a noticeable reduction in the intensity of the aforementioned O 1s shoulder, which might be indicative of desorption of some water or hydrogen, while in the Ce 3d spectrum a contribution from Ce3+ becomes evident at ca. 886 eV, and analysis confirms a reduction to a stoichiometry of around CeO1.83. In summary, clear and distinct differences are evident in the oxidation behavior of a thin (several monolayers) Ce film deposited on CaF2(111)//Si(111) as opposed to Si(111). On Si(111) itself, the cerium is rapidly oxidized to Ce3+, but extensive conversion to Ce4+ cannot be achieved using the type of oxidation conditions that are readily attainable in our system. Instead, oxidation of the Si is observed, which is consistent with the redox reaction of CeO2 with Si (eq 1) being thermodynami-
Growth of Ceria Films on Si(111)
J. Phys. Chem. B, Vol. 109, No. 21, 2005 10981
cally favored (∆GQ298 ) -170 kJ mol-1, based on the data of ref 24).
4CeO2 + Si f 2Ce2O3 + SiO2
(1)
The cerium oxide film that is obtained is disordered, which is attributed to the formation of an amorphous SiOx interface between Si(111) and CeO2 that destroys the potential for epitaxial growth of the ceria on the silicon substrate. We cannot exclude the possibility of some cerium inclusion in the oxidized silicon phase. By contrast, the use of a CaF2 buffer layer permits the growth of an ordered film of cerium oxide possessing a stoichiometry approaching that of CeO2. Furthermore, this can be achieved using mild experimental conditions (just 40 L O2 was sufficient to fully oxidize the cerium in the case studied). This difference is because the CaF2 surface is much more resistant to oxidation, and yet can still act as a template for the growth of the ordered oxide (the lattice mismatch between the ideal CaF2 and CeO2 (111) surfaces is
