Available online at www.sciencedirect.com
Surface Science 602 (2008) 1101–1113 www.elsevier.com/locate/susc
Mechanism of growth and structure of titanium oxide ultrathin films deposited on Cu(0 0 1) P. Finetti a, M. Caffio a, B. Cortigiani a, A. Atrei b,*, G. Rovida a a
Dipartimento di Chimica, Universita` di Firenze, Polo Scientifico di Sesto Fiorentino, 55019 Sesto, Fiorentino (Firenze), Italy b Dipartimento di Scienze e Tecnologie Chimiche e dei Biosistemi, Universita` di Siena, 53100 Siena, Italy Received 21 September 2007; accepted for publication 7 January 2008 Available online 26 January 2008
Abstract The growth mechanism, composition and structure of ultrathin films of titanium oxide deposited on the Cu(0 0 1) surface were investigated by means of XPS, LEIS, LEED and STM. Titanium oxide films were deposited on the Cu(0 0 1) surface previously saturated with p p a ( 2 2 2)R45° structure of chemisorbed oxygen. The oxide films were prepared by evaporation of titanium in O2 atmosphere (pO2 in 6 the 10 mbar range) while the substrate temperature was kept at 573 K. The Cu LEIS signal versus the amount of deposited titanium (as determined by means of XPS) indicates the growth of 2D islands in the early stages of deposition. Upon increasing the amount of deposited titanium multilayer islands begin to grow. The XPS results indicate that the oxide phases formed for Ti coverages above 1 ML have a TiO2 stoichiometry. At very low coverages, a LEED pattern with a centred rectangular unit cell is observed. STM measurements show that at this stage of the growth the oxide islands are incorporated in the outermost layer of the substrate and the removed copper atoms form islands around the oxide regions. The very early stages of titanium oxide growth corresponding to the formation of this rectangular phase were also investigated by Ti deposition on the oxygen chemisorbed phase under UHV conditions at 573 K. In this way it is possible to study the reaction of Ti with chemisorbed oxygen. Upon increasing the Ti coverage above 0.5 ML, an oxide film with a slightly distorted hexagonal unit cell begins to grow. The quasi-hexagonal phase of titanium oxide can also be formed by annealing at 773 K the rectangular phase. At higher coverages, when the substrate surface is completely covered by the oxide, the film exhibits a LEED pattern with a regular hexagonal unit cell. Ó 2008 Elsevier B.V. All rights reserved. Keywords: Titanium oxide; Copper; Epitaxial growth; XPS; LEIS; LEED; STM
1. Introduction The composition, growth mechanism and structure of ultrathin films of titanium oxide deposited on various metal substrates have been investigated by surface science methods. Most of the studies reported in the literature concern titanium oxide films deposited on platinum single crystal surfaces [1–5] or on surfaces of refractory metals such as Ru [6] and Mo [7,8]. Only few studies are reported concerning titanium oxide films on Cu surfaces [9,10]. In a recent paper ultrathin films of titania deposited on the (1 1 0) sur*
Corresponding author. Tel.: +39 0577 234371; fax: +39 0577 234177. E-mail address:
[email protected] (A. Atrei).
0039-6028/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.susc.2008.01.016
face of nickel (a metal with a lattice parameter close to that of copper) were investigated by means of surface science methods [11]. The use of copper surfaces as substrates for the growth of titanium oxide films is motivated by several reasons. The significantly different chemical properties and lattice parameter of copper compared to platinum and to the metals mentioned above should help to understand the effect of the chemical reactivity and of the epitaxial relationships on the structure and composition of the titanium oxide films. If the oxide films are prepared by reactive evaporation, the presence of a chemisorbed layer is expected to influence the growth mechanism and the structure of the interface. In order to investigate these effects,
1102
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
the (0 0 1) surface of copper is particularly suitable since a stable layer of chemisorbed oxygen atoms p can be p formed on this surface. Moreover, the Cu(0 0 1) ( 2 2 2)R45°O phase is well characterised from the structural point of view. On the other hand, if the evaporation of Ti is carried out in vacuum and the oxide is formed by post oxidation, the interaction of metallic titanium with the metallic substrate may play an important role in the structure of the oxide films, at least in the very early stage of growth. Both platinum and copper form surface alloys with titanium [12,13] and the formation of this intermetallic phases or solid solutions should be taken into account when the oxide films are prepared by oxidation of deposited metallic titanium. The growth, structure and chemical state of titanium oxide films deposited on Cu(0 0 1) had been previously investigated by X-ray photoelectron spectroscopy (XPS) and low-energy electron diffraction (LEED) [9]. In that work, the formation of a TiO2 film exhibiting a hexagonal LEED pattern is reported. The authors propose a qualitative model for this phase consisting in an O–Ti–O trilayer. This model is analogous to the one proposed for the quasihexagonal structure formed by titanium oxide on the oxidized (1 1 0) surface of the NiTi alloy [14]. In the present work we investigated the growth mechanism of titanium oxide on Cu(0 0 1) by XPS and LowEnergy Ion Scattering (LEIS). Particular attention is dedicated to the early stages of titanium oxide ultrathin films formation that we studied mainly by Scanning Tunnelling Microscopy (STM) and LEED. We observe two oxide phases in addition to the previously reported hexagonal structure. 2. Experimental details The experiments were carried out in an ultra-high vacuum (UHV) apparatus with a base pressure in the low 1010 mbar range. The chamber was equipped with an X-ray source, a hemispherical electron/ion energy analyser, a focused ion gun for the LEIS measurements and a 3-grid rear view LEED optics. Non-monochromatic Al Ka radiation was used for the XPS measurements. The angle between the analyser axis and the X-ray source was 55°. The XPS spectra were measured with a constant pass energy of 44 eV. The binding energy (BE) scale was calibrated setting the Cu2p3/2 peak to 932.7 eV [15]. The inelastic background in the spectra was subtracted by means of the Shirley method [16]. For the LEIS measurements we used a He ion beam with 1 keV energy impinging on the surface at an angle of 45°, along the [1 1 0] direction. The scattering angle was 135°. The sample was a plate of 10 mm 10 mm 2 mm. cut and polished along the (0 0 1) surface with an accuracy of ±0.1°. The temperature was measured by means of a chromel–alumel thermocouple placed near to the sample. The surface was prepared by cycles of argon ion sputtering (600 eV) and annealing (700 K for 30 min) until no con-
tamination was detectable by means of XPS and LEIS and a sharp (1 1) LEED was visible. Ti (99.999% purity) was evaporated using an electron beam evaporator. When evaporated in vacuum, the pressure during evaporation remained in the low 1010 mbar range. All titanium oxide films reported in this paper were deposited on a Cu(0 0 1) surface previously saturated with chemisorbed oxygen. The oxygen covered surface was prepared by exposing the Cu(0 0 1) substrate at 573 K to an O2 pressure p of 1 p 106 mbar for 10 min. This procedure yields a ( 2 2 2)R45° LEED pattern with a Cu missing-row reconstruction [17–19]. The substrate temperature during Ti evaporation was 573 K. As a default the titanium evaporation was carried out in O2 at a pressure of 1 106 mbar. However the early stages of the oxide growth were also investigated by depositing Ti in vacuum, so that the oxide could only form by reaction with chemisorbed oxygen. Results obtained with this particular type of sample preparation are going to be clearly signalled in the text and figure caption. The amount of the deposited titanium was estimated by measuring the Ti2p on Cu2p3/2 XPS intensity ratio. The number of Ti atoms per unit area is then gauged from this XPS intensity ratio by normalisation to a reference XPS ratio measured on a sample for which: (i) the fraction of surface covered by the oxide was determined by STM (ii) the oxide has a homogeneous structure (as observed by LEED and STM) for which a structural model is proposed. This structure will be discussed in more detail in the LEED and STM sections and it will be referred to as the ‘‘quasi-hexagonal” structure. The number of Ti atoms is given here in monolayer (ML), where 1 ML corresponds to the number atoms per unit area of the Cu(0 0 1) surface. The evaporation rate was typically of the order of 0.1–0.2 ML of titanium per minute. STM measurements were performed in a different vessel connected to the analytical chamber so that the sample could be transferred between the two chambers under UHV conditions. The STM images were collected at room temperature in constant current mode using tungsten or platinum–iridium tips chemically etched and prepared in situ by Ar+ ion sputtering. 3. Results 3.1. LEIS and XPS results LEIS spectra measured for the clean Cu surface and after deposition of two different amounts of titanium oxide are shown in Fig. 1. Information about the growth mechanism of the titanium oxide films can be derived from the plot of the copper and titanium LEIS signals (see Fig. 2a and b, respectively) as a function of the Ti coverage (hTi). The results shown in Fig. 2 are relative to the reactive deposition of titanium oxide. The Cu LEIS signal is normalised to the intensity measured for the clean surface. Due to presence of chemisorbed oxygen the initial value of the Cu LEIS signal is ca. 0.30 times that of the clean surface. Since
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
1103
Fig. 1. LEIS spectra measured for the clean Cu(0 0 1) surface and after deposition of increasing amounts of titanium oxide. E0 is the energy of the primary He ion beam. E is the energy of the scattered He ions.
LEIS is sensitive to the composition of the outermost layer only, the Cu LEIS signal (normalised to the initial value) is proportional to the fraction of substrate surface not covered by the oxide film. A linear decrease of the Cu LEIS signal as a function of the amount of deposited Ti is observed for hTi below ca. 0.5 ML. The linear decrease of the Cu LEIS signal upon increasing the amount of titanium oxide indicates that the oxide film grows forming islands of uniform thickness. Above this value of hTi, there is a deviation from the initial linear behaviour which suggests a change in the growth mechanism and/or the formation of a new phase. The titanium oxide film covers completely the substrate surface for hTi larger than 2 ML. The Ti LEIS intensity vs. hTi curve is consistent with the data of copper (Fig. 2b). Here, the LEIS signal of Ti is normalised to the value measured when the whole substrate is covered by the oxide film. In the initial stages, the Ti LEIS signal increases linearly with the oxide coverage. A deviation from the initial linear behaviour is observed above hTi around 0.5 ML as in the case of the Cu LEIS signal. Then the Ti LEIS signal reaches a constant value when the titanium oxide film covers completely the substrate surface. As it will be discussed later, LEED and STM results indicate that the change in the curve of the Cu (and Ti) LEIS signal versus the amount of deposited titanium oxide at hTi = 0.5 ML corresponds to a change in the structure of the titanium oxide film.
Fig. 2. Cu (a) and Ti (b) LEIS intensities as a function of hTi, the amount of titanium deposited on the surface. The LEIS signal of Cu is normalised to the value measured for the clean Cu(0 0 1) surface. The LEIS signal of Ti is normalised to the intensity measured when the whole surface is covered by the oxide film. The titanium oxide films were prepared by evaporation of Ti in O2 atmosphere (pO2 1 106 mbar) with the sample at 573 K. Before evaporation of Ti the Cu(0 0 1) surface was saturated with p p chemisorbed oxygen which forms a ( 2 2 2)R45° structure.
By extrapolating the initial linear behaviour of the Cu LEIS signal we can estimate the Ti coverage needed to cover the whole substrate surface with a film growing as in the very early stages. The titanium coverage corresponding to this point is equal to 0.9 ± 0.1 ML. The Ti2p and O1s XPS spectra collected as a function of the titanium coverage are shown in Fig. 3a and b, respectively, while Fig. 4 shows the hTi dependent plot of the core levels binding energy values. At the very early stages of the oxide film growth (hTi < 0.5 ML) the Ti2p3/2 BE values are centred around 457.3 eV. This BE value is close to the one reported in the literature for bulk Ti2O3 or more in general for bulk Ti(I I I) species [20,21]. At a coverage of ca. 1 ML the position of the Ti2p3/2 peak moves towards a BE value which is about 1 eV higher. This Ti2p3/2 BE value (458.3 eV) is somewhat lower than the most frequently reported value for bulk TiO2 (459.0 eV) but well within the range of BE values reported for Ti(IV) in ultrathin films of comparable thickness [22,11]. Upon further increasing the oxide coverage, the Ti2p3/2 BE reaches the typical values
1104
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
Fig. 3. Ti2p (a) and O1s (b) XPS spectra measured for various amounts of deposited titanium oxide prepared by evaporation of Ti in oxygen atmosphere. The intensities are normalised and an offset has been added for clarity.
Fig. 4. Ti2p3/2 (top) and O1s (bottom) binding energies as a function of the Ti coverage.
reported for surfaces of bulk TiO2 or more in general Ti(IV) compounds [20,21]. The O1s BE variation with coverage follows a similar trend to that of the Ti2p3/2 core level, although in this case the increase is smaller. All the Ti2p XPS spectra of Fig. 3a, including those at the lowest oxide coverage, clearly exhibit satellites at 13.0 eV from the main peaks. Such satellites, so far, have only been reported for Ti(IV) compounds [23–25]. Another spectroscopic parameter which can be related to the stoichiometry of the Ti oxide is the BE separation between the Ti2p1/2 and Ti2p3/2. The splitting we measure is the same for all the spectra of Fig. 3a. and is equal to 5.6 eV. This value is consistent with those reported for TiO2 [15,20]. For hTi > 1 ML the O1s (after removal of the contribution due chemisorbed oxygen) to Ti2p peak area ratio does not vary significantly with the Ti coverage, as shown in Fig. 5, although the Ti2p BE does vary upon increasing the amount of deposited oxide. Therefore, the BE energy shift observed in this range of coverage is probably due to final state effects which vary with the thickness of the oxide film. At very low oxide coverage, there is a large scattering of the data and no clear conclusion can be derived from the analysis of the intensity ratio. The early stages of the titanium oxide growth (hTi < 0.3 ML) were also p investigated by evaporating titanium p in UHV onto the ( 2 2 2)R45°-O phase. In this case an increase of the Cu LEIS signal is observed after Ti deposition. The Cu LEIS signal decreases after the film is exposed to O2 under the same conditions used for the
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
Fig. 5. Ratio of the O1s to Ti2p peak areas as a function of hTi. The ratios are normalized to the value measured for the thickest film. The data are obtained after subtraction from the total O1s peak areas the contribution of chemisorbed oxygen (rescaled to take into account the decrease of the fraction of substrate surface covered by chemisorbed oxygen).
reactive evaporation. The Cu LEIS signal after oxygen exposure reaches a value very close to that measured when a similar amount of titanium oxide was deposited by evaporation in the presence of O2. For equal amounts of titanium oxide, the Ti2p BE values are very similar independently of the preparation procedure. Moreover, the Ti2p BE of the oxide films at very low coverage p(457.3 p eV) formed by evaporation in vacuum on the ( 2 2 2)R45° chemisorbed phase of oxygen does not vary after exposure to O2, at a pressure of 1x106 mbar and the sample held at ca. 600 K. The O1s BE difference between chemisorbed oxygen and oxygen in the low coverage oxide (Ti2p BE 457.3 eV) is about 0.3 eV This BE shift could be determined rather accurately by p p depositing an amount of Ti in UHV onto a ( 2 2 2)R45°-O Cu(0 0 1) precovered surface sufficient to interact with nearly the whole oxygen adlayer (that is about 0.25 ML of Ti), as measured by the increase of the O1s peak area after subsequent exposure to O2. These results indicate that titanium atoms react with oxygen atoms chemisorbed on Cu(0 0 1) leaving part or even most of the substrate surface uncovered. The subsequent exposure to O2 leads to the adsorption of oxygen on the bare substrate surface. 3.2. LEED In the very early stages of titanium oxide deposition (hTi < 0.5 ML), new diffraction p pspots are observed in addition to those of the ( 2 2 2)R45°-O phase (Fig. 6a). The extra beams appear elongated along the diagonals of the substrate unit cell and partially overlap with those of the chemisorbed oxygen phase. The new diffraction spots can be described in terms of a centred rectangular unit cell
1105
p p indicated as c( 2 2 2)R45°. Since a coincidence mesh was not observed, the size of the unit cell was derived by measuring the distance between the spots on the diffraction diagram. Hence, there is an uncertainty of ±0.1 on the 3.2 factor and the oxide structure might be incommensurate along one diagonal direction of the substrate unit cell. A schematic representation of the LEED pattern is reported in Fig. 6a (right) where the primitive rhombic cell is indicated. The transformation matrices in the reciprocal and direct space for this cell are reported in Table 1. All the spots of the pattern are reproduced if one takes into account the existence of equivalent domains. This diffraction pattern (which will be indicated in the following as ‘‘rectangular”) is observed for low titanium oxide coverage (hTi < 0.5 ML), independently of the preparation procedure (either evaporation of Ti in the presence of O2 or evaporation in vacuum on the chemisorbed oxygen layer). However, when the oxide is prepared by evaporation of Ti in vacuum on the oxygen chemisorbed phase, it is possible to observe the spots of the rectangular p p phase with extremely weak reflexes of the ( 2 2 2)R45°-O. LEED superstructures relative to sub saturation oxygen adlayers [19] were p neverpobserved in the present case. These spots of the ( 2 2 2)R45°-O are again visible after exposing the surface to O2, due to the re-adsorption of oxygen on the Cu(0 0 1) substrate. When the oxide is prepared by evaporation of Ti in O2 atmosphere, at hTi close to 0.5 ML, new spots start to be visible in addition those of the rectangular pattern and p to p those of the ( 2 2 2)R45°-O phase (Fig. 6b). At first glance, this pattern could be interpreted as two domains of an hexagonal unit cell as observed by Maeda et al. [9] but a closer inspection shows that the hexagon is actually distorted. The distortion is detected by systematic inequalities in the distance measured between the 12 spots associated to the two domains of hexagonal superstructure. An example of this type of measurement is given in Fig. 7 (see also the figure caption). Briefly, a profile of the intensity is measured from the LEED diagram of the hexagonal structure (Fig. 7a). The segmented path used to obtain the intensity profile through the diffraction spots is indicated in Fig. 7b. The actual profile (Fig. 7c) is then calculated by averaging the intensity across a wide stripe. Finally the peak position is determined by means of a fitting procedure (Fig. 7c). It can be seen that the distance between adjacent spots pertaining uniquely to the hexagonal structure is systematically shorter than the distance between a purely hexagonal spot and an integer diffraction spot (Fig. 7c). The unit cell parameters resulting from this analysis of the pattern (including also the double diffraction spots to increase the accuracy of the measurements) are as follows: the side of this unit cell is ˚ and the angle between the basis vectors is 119.6°. 2.95 A This cell is only slightly distorted from a hexagonal mesh and we call it ‘‘quasi-hexagonal”. The unit cell of the ‘‘quasi-hexagonal” structure is marked in Fig. 6b and the matrix is reported in Table 1.
1106
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
p p Fig. 6. (a) Left: LEED pattern observed for hTi below 0.5. The diffraction spots of the ( 2 2 2)R45° phase of chemisorbed oxygen are visible. Electron energy: 78 eV. Right: schematic representation of the LEED pattern. The unit cell of one domain of the structure is marked by white lines. (b) Left: LEED p p pattern of the quasi-hexagonal phase. In the pattern also the diffraction spots of the rectangular phase and of ( 2 2 2)R45°-O structure are visible. Right: schematic representation of the LEED pattern of the quasi-hexagonal phase. The cell of one domain of this structure is marked by white lines. (c) Left: LEED pattern of the regular hexagonal phase observed when the whole substrate surface is covered by the titanium oxide film (hTi = 2 ML). Electron energy: 115 eV. Right: schematic representation of the hexagonal LEED pattern. The cell of one domain is marked by white lines.
The rectangular and the quasi-hexagonal phases coexist in a limited coverage range around hTi = 0.5 ML. Above
this coverage the diffraction beams of the rectangular phase gradually disappear.
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
The formation of the quasi-hexagonal phase depends on the temperature of the substrate in addition to the oxide coverage. The quasi-hexagonal phase can be formed at Ti
1107
coverages well below 0.5 ML by annealing at 800 K (either in UHV or in the presence of oxygen) the rectangular phase.
Table 1 Matrices describing the unit cells (in the reciprocal and real space) of the various phases of the titanium oxide films deposited on Cu(0 0 1) Structure ‘‘Rectangular” phase Quasi-hexagonal phase Hexagonal phase
Transformation matrix (reciprocal space) 0:65 0:34 0:34 0:65 1 0:86 2 1 2 0:86 pffiffi ! 1 2
12
3 p2ffiffi 3 2
Transformation matrix (real space) 2:1 1:1 1:1 2:1 1:0 0:58 1:0 0:58 pffiffi ! 3 1 p3ffiffi 1 33
Length of the unit cell ˚) side (A
Angle between the sides of the unit cell (°)
a = 6.04
145.3
a = 2.95
119.6
a = 2.94
120.0
For each structure, the length of the unit cell side in the real space, the angle between the sides of unit cell are reported. For the ‘‘rectangular” phase we reported the lattice parameters of the primitive rhombic cell.
Fig. 7. (a) LEED diagram of the quasi-hexagonal structure from which the profile of the intensity was measured. (b) contour intensity plot of the LEED diagram. The path used to obtain the intensity profile through the diffraction spots is indicated by the broken line. (c) (red) broken line showing the raw intensity profile measured through the path shown in (b). The profile is calculated by averaging the intensity across a wide stripe (approximately the width of the diffraction spot itself) perpendicular to the path. In this way the choice of the path centre is not critical in determining the position of the maxima and the noise in the profile is minimised; (blue) full line showing a fitting of the intensity profile. The numbers between the peaks indicate their relative distance. The total image size in (a) is 1300 pixels 1300 pixels. The reciprocal cell unit size (measurement path not shown) is 545 pixels. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1108
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
At higher oxide coverage, diffraction spots corresponding to a regular hexagonal unit cell are observed in the LEED pattern (Fig. 6c). This pattern corresponds to that observed by Maeda et al. [9]. The side of the hexagonal unit ˚. cell is 2.94 A When the substrate is completely covered by the titanium oxide film, the spots of the hexagonal LEED pattern reach the maximum intensity. Above this coverage, the quality of the LEED pattern decreases. For the thickest ˚ ), prepared at 570 K only a diffilms studied here (ca. 35 A fuse background was observed in the LEED pattern. 3.3. STM The STM images collected for the titanium oxide film exhibiting the rectangular LEED pattern are shown in Fig. 8a–d. At very low titanium oxide coverage (Fig. 8a– b) the STM images are characterised by the presence of dark narrow stripes running along the [1 0 0] directions. p Wide regions of Cu(0 0 1) covered by the ( 2 p 2 2)R45°-O phase are also visible in the images. These zones are characterised by the presence of dark rows with ˚ corresponding to the 2p 2 periodicity a spacing of ca. 7 A due to the missing-row reconstruction of the substrate [17–19]. Domains of nanometric size rotated by 90° to one the other are visible in these regions. In some cases (see the inset of Fig. 8a showing a marked portion of Fig. 8a) the two domains are imaged differently giving the oxygen covered Cu terrace a patchy grey scale appearance. Such contrast in the imaging of oxygen domains was previously reported in the literature and attributed to tip effects [17]. Bright irregular strings (marked with an S in Fig. 8a and c) on the oxygen covered surface seem to connect the titanium oxide islands. These lines, which p p are observed also when only the ( 2 2 2)R45°-O phase is present, were previously attributed to rows of alternating oxygen and copper atoms [18]. These dark stripes can be attributed to titanium oxide since the area corresponding to these zones is in agreement with fractional coverage of oxide determined on the basis of the LEIS results. Moreover, features with a periodicity corresponding to the centred rectangular unit cell of the titanium film can be seen on the dark regions (Fig. 9). The preferential growth of the oxide islands along the missing-row directions explains the elongation of the spots of the rectangular phase in the LEED pattern. p pThe dark stripes never cross the domains of the ( 2 2 2)R45°-O phase. They begin or terminate where two perpendicular domains of the chemisorbed oxygen phase encounter. The oxide islands appear as p depressions with respect to p the terraces covered by the ( 2 2 2)R45°-O phase at all bias voltages used here. Most of the oxide islands are surrounded by a bright edge. The bright edges are present preferentially around the oxide islands located in the middle of a terrace (Fig. 8a–c), relatively far from step edges. Oxide islands growing near step edges (Fig. 8a) or islands interconnected by the string structures observed upon oxy-
gen chemisorption (Fig. 8c) are preferentially not decorated. Upon increasing the amount of titanium oxide, the dark stripes become larger and bends are formed producing ‘‘J” shaped islands (Fig. 8c). The maximum fraction of substrate covered by the rectangular phase prior to the nucleation of the quasi-hexagonal phase is estimated to be around 20–30%. The image shown in Fig. 8d was measured near the maximum coverage of the rectangular phase. At this coverage both the dark and bright rectangular regions are less elongated than at low coverage. The wider extension of the bright regions allows measuring the height of this type of islands. Regardless of the bias voltage or tip condition, it is found that their height, relative to the Cu substrate middle grey region, coincides with the height of the oxygen covered Cu step edges. These results can be explained if the oxide islands are embedded within the outermost layer of the substrate. According to this interpretation the bright regions correspond to islands of copper atoms covered by chemisorbed oxygen. During the growth of titanium oxide, copper atoms are expelled from the outermost layer of the substrate and migrate at the edges of the oxide island. When relatively large islands of the rectangular phase are present (Fig. 8d), we found that the area of the dark regions (oxide islands) is equal to the area of the bright zones (copper islands). For oxide islands located at step edges, copper atoms are free to diffuse along the step. In this case the edges of the oxide islands are not decorated by copper atoms, as shown in Fig. 8a. The bright frames around the titanium oxide regions are not observed when the oxide is formed by evaporation of titanium under UHV condition on the chemisorbed oxygen layer. This observation suggests that when the deposition is carried out in vacuum the copper atoms produced by the embedding of the titanium oxide films in the surface are mobile at 600 K and can diffuse at the step edges. On the other hand, the presence of O2 seems to reduce significantly the mobility of copper atoms which remain around the titanium oxide islands. The image in Fig. 8e was measured for the quasi-hexagonal phase prepared by annealing at ca. 800 K for 10 min the rectangular phase. The islands of the quasihexagonal phase exhibit a polygonal shape. Note that also in this case the oxide islands appear to be interconnected by the same string structures observed upon oxygen chemisorption [18] and upon formation of the rectangular structure (Fig. 8a and c). The islands of the quasi-hexagonal phase prepared under these conditions are flat and appear to have a uniform thickness. The height of these oxide islands with respect to the substrate is lower than one monoatomic step of Cu(0 0 1). The fraction of surface covered by the oxide islands remains constant (within the accuracy of the measurements) after the transformation from the rectangular to the quasi-hexagonal phase due to the annealing. The STM images measured for the flat islands of the quasi-hexagonal phase were used to calibrate the amount
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
1109
Fig. 8. STM images measured for the various phases of the titanium oxide films deposited on Cu(0 0 1). The main crystallographic directions of the substrate are indicated. For each image the size, the tunnelling current and the sample bias are reported. (a) Rectangular phase (hTi = 0.02 ML). 100 100 nm2. It = 0.2 nA. V = 1.0 V. In the inset an enlargement of the zone marked by the dashed square is reported to show that the orthogonal p p domains of the ( 2 2 2)R45°-O are imaged differently. The bright irregular strings observed on the O chemisorbed zones are marked with an S in (a) and (c). (b) Rectangular phase (hTi = 0.11 ML). 50 50 nm2. It = 0.2 nA. V = 0.8 V. (c) Rectangular phase (hTi = 0.11 ML). 70 70 nm2. It = 0.2 nA. V = 0.5 V. (d) Rectangular phase (hTi = 0.28 ML). 100 100 nm2. It = 0.2 nA. V = 0.5 V. (e) Quasi-hexagonal phase prepared by annealing the rectangular phase (shown in Fig. 6d) to 773 K for 10 min. 450 450 nm2. It = 0.14 nA. V = 30 mV. (f) Quasi-hexagonal and hexagonal phases. This image was measured for an oxide film prepared by Ti evaporation in O2 atmosphere with the sample at 573 K. (hTi = 1.32 ML). 100 100 nm2. It = 0.20 nA. V = 20 mV.
1110
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
of deposited titanium with the procedure described in the experimental section. The STM images measured for the film prepared by deposition at 573 K indicate the presence of islands with various heights upon increasing the oxide coverage (Fig. 8f). Stable tunnelling conditions were more difficult to achieve in the case of thicker films with the hexagonal structure. The STM images measured when the hexagonal phase is present show the formation of multilayer islands with irregular borders. 4. Discussion The results of the present work are consistent with those of the Maeda et al. [9] as far as the variation of the Ti2p and O1s BE with the titanium oxide coverage is concerned. The main discrepancies consist in the different growth mechanism that we determined and the observation of various LEED patterns depending on the oxide coverage. The results reported in Ref. [9] indicate a two-dimensional island growth of the titanium oxide film. In addition to that, only the hexagonal phase of the titanium oxide film was observed in the whole range of coverage. Above the completion of the O–Ti–O trilayer they reported that the LEED pattern gradually faded upon increasing the oxide coverage and no additional long-range ordered phase of the TiO2 were observed. The differences in the results reported above may be due to the fact that Maeda et al. [9] used a different preparation procedure, that is deposition of Ti in UHV and subsequent oxidation. As discussed below, the growth of multilayer islands cannot be explained in terms of a stacking of O–Ti–O trilayers.
Fig. 9. Zoom in of the rectangular phase. 17 17 nm2. It = 0.2 nA. p p V = 30 mV. The unit cells of the ( 2 2 2)R45°-O structure and of the p p c( 2 2 2)R45° phase of titanium oxide are drawn on the image.
As shown by the LEIS vs. titanium coverage data as well as by the STM images, the rectangular phase which is observed in the very early stages of deposition forms islands of uniform thickness. The growth of the rectangular phase is characterized by a strong interaction between the oxide film and the substrate. The STM results show that the titanium oxide film does not grow over the surface but it is embedded in surface region, probably in the outermost layer of the substrate. Such complex behaviour was observed also in the case of nickel oxide deposited on Ag(0 0 1) [26]. In the case of nickel on Ag(0 0 1), embedding atom model calculations indicate that nickel atoms are incorporated within the first layer of the substrate before they can diffuse over the surface as adatoms and react with the oxygen molecules impinging onto the surface. Nickel oxide islands are produced by the reaction of embedded nickel atoms with O2 molecules reaching the surface. In the present case, titanium atoms react with the oxygen atoms adsorbed on Cu(0 0 1) to form the oxide. The missing-row reconstruction seems to favour the growth of the oxide within the outermost layer of the substrate. The interaction of the borders of the titanium oxide islands with the step edges of the substrate may play a role in stabilizing the embedding of the titanium oxide. Theoretical calculations for other oxides on metal substrate suggest that these kind of interactions are very important in determining the embedding of the islands and the orientation of their borders [27]. The investigation of the rectangular phase by means of STM confirmed the LEED results about the size of unit cell but did not allow us to obtain information about the position of the atoms in the unit cell. Considering the closepacked planes of the crystallographic phases of bulk titania, we found that the sides of the centred rectangular unit ˚ and 11.54 A ˚ ) are very close to the lattice cell (3.61 A ˚ and parameters of the (101) plane of anatase (3.785 A ˚ [28]) which also has a centred rectangular unit 10.239 A cell. The area of the unit cell of the rectangular phase is expanded by ca. 7% with respect to that of the cell of anatase (1 0 1). Incidentally, the (1 0 1) surface is the thermodynamically most stable surface of anatase [29]. Unfortunately, a study by LEED intensity analysis is hampered by the fact that it is not possible to prepare a complete layer of the rectangular phase. Prior to the completion of the rectangular phase layer, the quasi-hexagonal phase begins to form. The interpretation of the structural results in terms of a TiO2 phase may appear in contrast with the relatively low Ti2p3/2 BE value measured for the rectangular phase compared to the BEs of Ti(IV) compounds. Ti2p BE measurements are commonly used to identify the titanium oxidation state at the surface of bulk compounds, this task being facilitated by the large BE shift observed (e.g. over 5 eV BE increase on going from Ti metal to TiO2). However, in the present case we are considering oxide films with a thickness of a few atomic layers. Both the interaction with the metallic substrate as well as nonlocal final state effects may affect the BEs measured for
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
these ultrathin films. In addition to that, the database of Ti2p BE values available in the literature allows to associate each oxidation state, even those related to a given compound, to a range of BE values rather than to a well defined value [20,21]. In the case of the stoichiometric oxide, discrepancies in the Ti2p BE determination may originate, for instance, from charging effects [20]. In the case of the sub-stoichiometric compounds the main difficulty is given by the high reactivity of their surfaces, so that pure sub-stoichiometric chemical states are hard to observe and BE values may have to be determined by means of a fitting procedure. A further evidence for the formation of a TiO2 phase even in the very early stages of growth is the presence of the satellites at 13.0 eV from the main peaks in the Ti2p spectra which are characteristic of Ti(IV) compounds [23–25]. Satellites in Ti2p XPS spectra were also observed in bulk Ti2O3 but at a lower energy of 12.3 eV [30]. Such energy variation of the satellite could well be detectable in our spectra. When the growth of the oxide film is carried out at 573 K by reactive deposition, above a critical coverage of the oxide, an oxide phase with a quasi-hexagonal unit cell begins to grow. The transition between the two oxide phases characterised by a different long-range order is continuous and there is a range of oxide coverage in which the two oxide phases coexist. The formation of multilayer islands of the (quasi-)hexagonal phase explains the change of the slope of Cu LEIS signal vs. the oxide coverage curve.
1111
The formation of a new phase is indicated also by the variation of the Ti2p3/2 BE in the range of coverage at which the transition from the rectangular to the quasi-hexagonal phase is observed. The XPS results indicate that the quasi-hexagonal and hexagonal phases have a TiO2 composition. The observed shift of Ti2p BE with oxide coverage for hTi > 1 ML could then be attributed to final state effects, i.e. a more effective non-local screening of the Ti2p core hole due to the substrate at lower coverage. This was suggested by Maeda et al. and by Finetti et al. [9,22]. As pointed out in Ref. [22] this interpretation is also supported by the O1s BE increase with the oxide coverage. The gradual variation of the BEs with the oxide coverage is consistent with the growth of multilayer islands of the quasi-hexagonal and hexagonal phase when the deposition is carried out at 573 K. The morphology and growth mechanism of the oxide islands is influenced by the temperature. While the deposition at 573 K, above a given coverage, is characterized by the formation of oxide islands with a distribution of heights, the annealing at 773 K (either in UHV or in the presence of O2) of the film favours the growth of flat islands. In particular, the annealing of the rectangular phase leads to the formation of large and flat islands of oxide with the quasi-hexagonal unit cell. The effect of the thermal treatment suggests that the rectangular phase may be stabilized by kinetic reasons and the quasi-hexagonal phase can be formed if the temperature is high enough or if the film is annealed for a long time at lower temperatures.
[110]
[110]
Fig. 10. Schematic drawing showing the orientation of the quasi-hexagonal unit cell (red line) with respect to the substrate. The (2 7) cell (black line) resulting from the coincidence of the periodicity of the oxide layer with that of the Cu(0 0 1) surface is shown. Black line circles: Cu atoms. Blue line circles: Oxygen atoms arranged with the quasi-hexagonal periodicity. Only the oxygen layer of the oxide film in contact with the substrate is shown. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1112
P. Finetti et al. / Surface Science 602 (2008) 1101–1113
Considering the dimensions of the unit cell and the results of previous studies, we can propose a structural model for the quasi-hexagonal phase. The average length of the ˚ ) is close to sides of the quasi-hexagonal unit cell (2.96 A the oxygen–oxygen nearest neighbour distance in titanium oxides [27]. Hence, the observed periodicity can be attributed to that of close-packed layers of oxygen atoms. The sizes of the two-dimensional unit cell are such that it cannot contain more than one atom. Therefore, the quasi-hexagonal phase can be explained in terms of one layer of titanium atoms between two layers of oxygen atoms. This model was originally proposed by Atrei et al. on the basis of the X-ray photoelectron diffraction (XPD) results for the quasi-hexagonal phase of titania prepared by oxidation of the Ni94Ti6(1 1 0) surface [14]. The O–Ti–O trilayer is basically the same model proposed by Maeda et al. to interpret the hexagonal LEED pattern of the TiO2 film on Cu(0 0 1) [9]. Along one the [1 1 0] directions of Cu(0 0 1), the distance between the rows of the oxide coincides with that between the rows of the of the substrate (Fig. 10). The coincidence between the spacings of the rows is probably the main driving force for the formation of the (quasi-) hexagonal phase with this epitaxial orientation. The small distortion from the regular hexagonal unit cell allows the oxide phase to find a match with the periodicity of the substrate. The unit cell of the oxide determined from the LEED pattern would result in a (2 7) coincidence mesh (Fig. 10). However, the extra beams of the coincidence mesh are not observed in the LEED pattern and the quasi-hexagonal phase may actually be incommensurate or the long-range order is not well developed inside the quasi-hexagonal islands. At higher oxide coverage, the oxide film grows forming multilayer islands. The evidences for the multilayer growth come from the STM images and from the Ti coverage needed to cover completely the substrate surface which exceeds significantly the amount corresponding to a whole O–Ti–O trilayer. For multilayer islands the interaction with the substrate is expected to be less important than for the flat islands and the distortion of the unit cell should decrease or disappear. Indeed, a LEED pattern with a regular hexagonal unit cell is observed for the multilayer islands. However, the quasi-hexagonal and hexagonal phases should have, besides different lattice parameters, also a different stacking of layers since it is not possible to build a sequence of alternating close-packed planes of Ti and O atoms based on the initial O–Ti–O trilayer unit and to maintain a TiO2 stoichiometry. It is likely that the growth of the multilayer islands involves a reconstruction of the simple hexagonal structure. The new phases may be poorly ordered when formed at 573 K and do not produce additional diffraction features. 5. Conclusions Titanium oxide ultrathin films grow on Cu(0 0 1) with different kinds of long-range order depending on the oxide
coverage and on the preparation procedures. The oxide phase with a centred rectangular unit cell which forms in the very early stages of growth is embedded within the outermost layer of the surface. The missing-row reconstruction of the substrate seems to be important for the growth of this oxide phase within the outermost layer of the substrate. When prepared by reactive deposition at 573 K, the oxide film exhibits a LEED pattern with a quasi-hexagonal unit cell for hTi larger than ca. 0.5 ML. Flat islands of uniform thickness can be obtained by annealing the rectangular phase to 773 K. The hexagonal phase formed upon increasing the amount of oxide consists of islands of various thickness (multilayer growth). The structure of the quasi-hexagonal phase is similar to that determined for the titanium oxide film prepared by oxidation of the Ni94Ti6(1 1 0) surface and consists of an O–Ti–O trilayer of close-packed atomic planes. Acknowledgments This work was supported by MIUR through the fund ‘‘PRIN2005” and by Europa Metalli. References [1] F. Sedona, G.A. Rizzi, S. Agnoli, F.X. Liabres i Xamena, A. Papageorgiu, D. Osterman, M. Sambi, P. Finetti, K. Schierbaum, G. Granozzi, J. Phys. Chem. B 109 (2005) 24411. [2] T. Matsumoto, M. Batzill, S. Hsieh, B. Koel, Surf. Sci. 572 (2004) 127. [3] T. Matsumoto, M. Batzill, S. Hsieh, B. Koel, Surf. Sci. 572 (2004) 146. [4] A.B. Boffa, H.C. Galloway, P.W. Jakobs, J.J. Benitez, J.D. Batteas, M. Salmeron, A.T. Bell, G.A. Somorjai, Surf. Sci. 326 (1995) 80. [5] T. Orzali, M. Casarin, G. Granozzi, M. Sambi, A. Vittadini, Phys. Rev. Lett. 97 (2006) 156101. [6] A. Ma¨nnig, Z. Zhao, D. Rosenthal, K. Christiann, H. Hoster, H. Rausher, R.J. Behm, Surf. Sci. 576 (2005) 29. [7] Q. Guo, W.S. Oh, D.W. Goodman, Surf. Sci. 437 (1999) 49. [8] M.S. Chen, W.T. Wallace, D. Kumar, Z. Yan, K.K. Gath, Y. Cai, Y. Kuroda, D.W. Goodman, Surf. Sci. 581 (2005) L115. [9] T. Maeda, Y. Kobayashi, K. Kishi, Surf. Sci. 436 (1999) 249. [10] H. Arita, A. Arita, K. Kishi, Surf. Sci. 516 (2002) 191. [11] A.C. Papageorgiou, G. Cabailh, Q. Chen, A. Rest, E. Lundgren, J.N. Andersen, G. Thornton, J. Phys. Chem. C 111 (2007) 7704. [12] I. Kurzina, V. Shevlyuga, A. Atrei, B. Cortigiani, G. Rovida, U. Bardi, Surf. Rev. Lett. 10 (2003) 861. [13] M. Caffio, G. Rovida, A. Atrei, Surf. Sci. 601 (2007) 528. [14] A. Atrei, U. Bardi, G. Rovida, Surf. Sci. 391 (1997) 216. [15] J. Chastain (Ed.), Handbook of X-ray Photoelectron Spectroscopy, Perkin–Elmer, Eden Prairie MN, 1992. [16] D. Briggs, M.P. Seah (Eds.), Practical Surface Analysis, Wiley, New York, 1985. [17] H.C. Zeng, R.A. McFarlane, K.A.R. Mitchell, Surf. Sci. 207 (1989) L7. [18] F.M. Leibsle, Surf. Sci. 337 (1995) 51. [19] T. Fujita, Y. Okawa, Y. Matsumoto, K. Tanaka, Phys. Rev. B 54 (1996) 2167. [20] J.T. Mayer, U. Diebold, T.E. Madey, E. Garfunkel, J. Electron Spectrosc. Relat. Phenom. 73 (1995) 1. [21] V.V. Atuchin, V.G. Kesler, N.V. Pervukhina, Z. Zhang, J. Electron Spectrosc. Relat. Phenom. 152 (2006) 18.
P. Finetti et al. / Surface Science 602 (2008) 1101–1113 [22] P. Finetti, F. Sedona, G.A. Rizzi, U. Mick, F. Sutara, M. Svec, V. Matolin, K. Schierbaum, G. Granozzi, J. Phys. Chem. C 111 (2007) 869. [23] E.L. Bullock, L. Patthey, S.G-. Steinemann, Surf. Sci. 352–354 (1996) 504. [24] G. van deer Laan, Phys. Rev. B 41 (1990) 12366. [25] M. Ohno, P. Decleva, Phys. Rev. B 49 (1994) 818.
1113
[26] M. Caffio, A. Atrei, B. Cortigiani, G. Rovida, J. Phys:. Condens. Mat. 18 (2006) 2379. [27] A.M. Ferrari, S. Casassa, C. Pisani, Phys. Rev. B 71 (2005) 155404. [28] C.J. Howard, T.M. Sabine, F. Dickson, Acta Cryst. B 41 (1991) 462. [29] U. Diebold, Surf. Sci. Rep. 48 (2003) 53. [30] T. Uozumi, K. Okada, A. Kotani, Y. Tezuka, S. Shin, J. Phys. Soc. Japan 63 (1996) 1150.