PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
 
Surf Sci. 2009 January 1; 603(1-3): 251–255.
PMCID: PMC2938191

Growth, thermal stability and structure of ultrathin Zn-layers on Pd(1 1 1)

Abstract

Low-energy ion scattering with monolayer sensitivity was applied to investigate ultrathin films of zinc on Pd(1 1 1). Uptake curves taken at 150 K indicate the simultaneous growth of multilayers with negligible interlayer transport. Annealing experiments for two-monolayer films reveal a rapid decrease in the zinc content on the surface layer at temperatures above 300 K, forming a metastable state with a Pd:Zn surface ratio of approx. 1:1 in the temperature region between 400 and 550 K. This state is most easily explained as a slightly buckled p(2 × 1)-PdZn surface alloy, with Zn atoms located approx. 0.25 Å above their Pd counterparts.

Keywords: Low-energy ion scattering, Growth, Compound formation, Palladium, Zinc, Bimetallic surfaces, Methanol steam-reforming

1. Introduction

With fossil fuels becoming rarer everyday, alternative energy sources are on the rise. Hydrogen-based technology is a highly promising contestant in the upcoming battle, offering the additional advantage of environmentally compatible energy supply. One of the main obstacles in the dissemination of hydrogen technology is storage and transport of H2. Here, methanol offers itself as an alternative hydrogen carrier, in particular for mobile applications, as it can be converted to hydrogen via steam reforming (CH3OH + H2O → CO2 + 3H2). Cu/ZnO-catalysts have proven to be highly selective, but their useful properties come at the price of metal sintering at temperatures beyond 600 K, requiring low operating temperatures (and thus low conversion rates) and resulting in deactivation and reduced lifetime of the catalyst [1]. Recently, Pd/ZnO-catalysts have attracted much attention as alternative catalysts with good steam-reforming capabilities [2]. These have been attributed to the formation of a 1:1 PdZn surface alloy, where the electronic states of Pd (d10) and Zn (d10s2) combine, thereby converging to an electronic state similar to that of copper (d10s1). Ultrathin Zn-layers on Pd(1 1 1) surfaces are used as model systems for this new generation of catalysts and were investigated by several groups, however, with partially incompatible conclusions about the possible formation and stability of a PdZn surface alloy [3–6]. While Bayer et al. [3], relying mainly on XPS-data, found alloy formation already slightly above room temperature, Gabasch et al. [4], interpreting CO-titration- and Zn-TPD-experiments, excluded alloy formation below 600 K. Furthermore, the LEED-pattern, which both the groups observed identically at temperatures around 500 K, was interpreted by Bayer et al. as originating from a 3-domain p(2 × 1)-PdZn surface alloy with 50% Zn in the topmost layer, whereas Gabasch et al. suggested formation of a p(2 × 2)-3Zn overlayer with 75% Zn at the surface (note that a 3-domain-p(2 × 1) and a p(2 × 2) superstructure, respectively, yield identical diffraction patterns). To shed more light on this controversial topic, we applied low-energy ion scattering (LEIS) as an independent investigation method. Under suitable scattering conditions, LEIS is sensitive to the top layer only, and can deliver information on both composition and structure of the surface. As in principle LEIS is a destructive method, special care was taken to check for the influence of the ion beam and to minimize the ion dose to a level, where virtually no change in surface composition could be detected in subsequent LEIS energy scans (similar to those shown in Fig. 1). Nevertheless, series of consecutive LEIS measurements on the same sample were abandoned. The only exception to this rule was allowed for the data underlying the annealing experiments in Fig. 3 in order to avoid additional spreading of the data points due to non-ideal reproducibility of the same Zn coverage.

Fig. 1
LEIS-spectra for increasing Zn-coverages. Spectra were taken under conditions, where only top-layer atoms contribute to the scattering yield (5 keV Ne+, scattering angle υ = 160°, angle of incidence Ψ = 45° ...
Fig. 3
Normalized top-layer Zn signal versus annealing temperature.

2. Experiment

Experiments were performed in an ultra-high vacuum chamber with a base pressure below 10−10 mbar. The chamber was equipped with a quadrupole mass filter for temperature-programmed desorption (TPD), a retarding-field analyzer for low-energy electron diffraction (LEED) and an ion gun for sample cleaning. Low-energy ion scattering experiments were performed using a dedicated home-build differentially pumped ion gun and a rotatable hemispherical energy analyzer. All experiments reported in this work were done with 5 keV Ne+ ions at a scattering angle υ = 160°. The Pd(1 1 1) sample (diameter 12 mm, thickness 2 mm) was mounted on a manipulator with three translational and three rotational degrees of freedom [7]. LN2 cooling and resistive heating allowed for sample temperatures between 120 and 1100 K.

Cleaning the sample surface involved sputtering by 600 eV Ar+, annealing to 1080 K, followed by exposure to 10 L of oxygen at T ≈ 350 K. Temperature-programmed desorption was then applied to monitor the desorbing products. The oxygen adsorption/TPD cycle was repeated several times until an intense oxygen desorption peak at T ≈ 800 K was observed, indicating an essentially carbon-free Pd(1 1 1) surface [8]. Zinc was deposited from a home-build Knudsen-type evaporator with integral shutter to enable precise adjustment of the deposition time. During evaporation, the sample was maintained at 150 K.

3. Results and discussion

3.1. Growth mode at 150 K

Fig. 1 shows a series of experiments, where increasing amounts of zinc were deposited onto the clean Pd(1 1 1) surface at 150 K and were analyzed by low-energy ion scattering to reveal information about the growth mode and to simultaneously calibrate the evaporation rate. Experiments were carried out with the ion beam impinging along the [1¯21¯] azimuth at an angle of incidence Ψ = 45° (throughout this work Ψ is given relative to grazing incidence at Ψ = 0°). Under these conditions, second-layer atoms are hidden in the shadow cones cast by top-layer atoms as long as they are located at or close to fcc or hcp positions (compare Fig. 5a). Thus, only the top layer is visible to the ion beam and analyzed by LEIS in these experiments. With increasing deposition time, the Pd-related peak at 2300 eV diminishes, while the Zn-related peak at 1380 eV increases in intensity.

Fig. 5
(a) Polar ICISS scans along the [1¯21¯] and [112¯] azimuths of clean Pd(1 1 1). The intensity increasing at Ψ ≈ 75° and 51°, respectively, results from focusing onto ...

In Fig. 2 the (integrated) intensity of the Zn peak – normalized to the total backscattering yield (IZn/[IZn + IPd]) – is depicted versus deposition time. The total backscattering yield IZn + IPd remains constant (±5%) during the deposition sequence (Fig. 1, inset). This implies that the LEIS-sensitivities (i.e. scattering cross-sections including ion-survival probabilities) are equal for both materials. Accordingly, the value of the normalized Zn-intensity IZn/[IZn + IPd] can be directly interpreted as the Zn content of the top layer. The data points show a smooth increase without any clear break in slope, indicating that growth does not occur in a layer-by-layer (Frank–van der Merwe) mode. In the latter case, a linear increase up to saturation at IZn/[IZn + IPd] = 1 would be expected, as for example observed for growth of Pd and Au on Ru(0 0 1) [9]. In contrast, the present data are well described by a poisson height distribution, as it occurs for simultaneous multilayer growth with negligible interlayer transport. In this case, the top-layer composition is described by the function 1-exp(-ttML), with tML denoting the time required to deposit the equivalent of a fully closed monolayer (ML). The best fit, shown by the solid line in Fig. 2, reveals a value of tML = 11.1 min. In the following, this value forms the basis of our coverage calibration.

Fig. 2
Normalized intensity (IZn/[IZn + IPd]) of the top-layer Zn signal versus deposition time.

An alternative explanation to simultaneous multilayer growth would be alloy formation. However, due the low deposition temperature (150 K), this scenario seems quite unlikely. Furthermore, in this case, one would intuitively expect that the alloy-formation rate scales with either the area of available clean Pd(1 1 1) patches or the length of available clean Pd(1 1 1) steps, depending on the detailed exchange mechanism of substrate and adatoms. In both the cases, this would yield a fast alloying rate in the beginning, ceasing with increasing coverage. Accordingly, the Zn signal should rise only slowly initially and speed up at larger deposition times – just in contrary to what is observed in the experiment. Thus, we conclude from Fig. 2 that at 150 K, zinc grows simultaneously in several layers with only weak transport of material between deposited layers. This finding contrasts somewhat with the conclusions of Bayer et al., who reported essentially layer-by-layer growth for the first two Zn-layers at 105 K [3], although the authors also state that growth is not perfectly two dimensional, i.e. the second layer already starts to grow before the first layer is completed. At the low substrate temperatures used in both Bayer’s and our study the mobility of adatoms across step edges may be kinetically hindered, making simultaneous growth of several layers quite likely.

3.2. Thermal stability

Annealing experiments were performed in order to investigate the thermal stability of the Zn films, which was the most controversial topic in the works of Bayer [3] and Gabasch [4], respectively. Initially, about 2.2 ML monolayers of zinc were deposited at T = 150 K. Then, the sample was consecutively heated to higher temperatures (10 min at each temperature setpoint), and the surface composition was determined by LEIS at each temperature (see Fig. 3). The data can roughly be grouped into four regimes. Already at low temperatures (region I, below 300 K), the Zn signal shows a slight decrease, which may to a small extend be caused by sputtering effects due to the analyzing ion beam. Above 330 K, the Zn content diminishes rapidly (region II), until between 400 and 550 K a plateau region with a nearly constant Pd:Zn ratio of about 1:1 is reached (region III). Beyond 550 K, the Zn signal steadily reduces towards zero (region IV).

The present data show unequivocally that the surface composition changes above 300 K, which is incompatible with the interpretation of Gabasch et al., who suggest that up to 600 K a stable closed Zn monolayer remains on the surface. Generally speaking our data show the same behaviour as reported by Bayer et al., who reported a composition change slightly above room temperature and suggested formation of a (metastable) 1:1 PdZn surface alloy in the temperature range from 400 to 550 K.

Although it is clear from the present data that the surface composition changes above room temperature, there are two different possible interpretations for our data, which, as a matter of principle, cannot be distinguished by LEIS at large scattering angles.

The first possibility is formation of a PdZn surface alloy with 1:1 stoichiometry (at T = 400–550 K), as suggested by Bayer et al. [3]. Alternatively, our data can be explained by emergence of islands with a zinc-rich surface, thus giving birth to “naked” spots of the Pd(1 1 1) surface. However, there is no simple physical reason, why in this case the overall surface composition should be particularly stable at a Pd:Zn ratio close to 1:1, as indicated by the plateau region in the temperature window from 400 to 550 K. Thus, in accordance with the conclusions of Ref. [3] – we favour formation of a (metastable) PdZn surface alloy as the most likely explanation of our experimental data.

Above 550 K, the zinc signal decreases continuously towards zero. This may be caused either by diffusion of zinc into the palladium substrate (as already proposed in Ref. [3]) or by agglomeration of the PdZn alloy into tall islands, thus exposing more of the clean palladium surface. By LEIS, we cannot directly discriminate between both scenarios. However, we noted that desorption of Zn (which would be a third alternative explanation) was not observed in the temperature region up to 800 K.

3.3. Structure of the PdZn surface alloy

After annealing to temperatures around 500 K (region III), LEED shows the p(2 × 1)/p(2 × 2) LEED-pattern already mentioned in the introductory section. The surface structure of this metastable PdZn surface alloy was investigated in more detail using angle-resolved impact-collision ion scattering spectroscopy (ICISS), extensively employing the shadow-cone concept [10,11].

Fig. 4a shows an azimuthal ICISS scan of the clean Pd(1 1 1) surface. The intensity of the Pd-backscattering peak was monitored, while the sample was rotated around the surface normal with the ion beam impinging close to grazing incidence at Ψ = 9°. Under these scattering conditions, surface atoms are completely/partially shadowed by their lateral neighbours along nearest-neighbour/next-nearest neighbour directions, giving rise to pronounced/shallow intensity minima at Φ = 0°, 60° and 120° or Φ = 30°, 90° and 150°, respectively. For the 1:1 PdZn surface alloy prepared by annealing a 2.2 ML zinc film at 450 K, the angular dependence of the Pd and the Zn backscattering intensities (Fig. 4b and c) basically resemble the data of the clean surface. For both the Pd- and the Zn-related intensities, the positions of the minima are identical and the widths of the minima are very similar to those of the clean surface, indicating that for both surface species the lateral surroundings (i.e. nearest and next-nearest neighbour directions) are virtually the same as for the clean Pd(1 1 1).

Fig. 4
Azimuthal ICISS scans taken close to grazing incidence at Ψ = 9°. (a) clean Pd(1 1 1), (b) 1:1 PdZn surface alloy, Zn signal (c) 1:1 PdZn surface alloy, Pd signal.

Information about the vertical arrangement of Pd and Zn surface atoms is best obtained from polar ICISS scans, where the angle of incidence is varied for a fixed azimuthal direction. Fig. 5b shows a polar scan taken in the [1¯21¯] azimuth for backscattering from Pd and Zn, respectively. Most notably the onsets of the Pd and Zn backscattering yield (i.e. the critical angles ΨC) differ for both atoms by approximately ΔΨC = 3°. Below the critical angle Pd or Zn surface atoms, respectively, are located in the shadow cone of their neighbours “to the left”, leading to vanishing backscattering intensities. At the critical angle, the edge of the shadow cone caused by the “left” surface atom (which we will refer to as the shadow-cone “producer”) is just swept across its neighbour to the “right” (“final backscatterer”), thus producing a strong increase in the scattering yield. The critical angle depends on several factors [10]: width of the shadow cone, lateral and vertical distance between shadow-cone producer and final scatterer, and to a weak extent also on the vibration amplitudes of “producer” and final scatterer. The width of the 5 keV Ne+ shadow cone caused by Zn is approx. 0.6° smaller than for a Pd “producer”. Hence, the different onset angles cannot be caused solely by different shadow-cone producers. The vibration amplitudes for both atom types are expected to be similar as inferred from similar bulk Debye temperatures (ΘPd = 275 K and ΘZn = 234 K). Furthermore, as already discussed in connection with Fig. 4 for both Pd and Zn atoms, the lateral arrangement of surrounding atoms is more or less the same. Thus, we attribute the difference ΔΨ = 3° between the critical angles primarily to a buckling of the top layer, i.e. different vertical positions of Pd and Zn atoms, respectively. As the onset of the Zn signal appears at lower angles than that for Pd, this implies that Zn atoms are located slightly higher than the Pd atoms. In a one-domain p(2 × 1) superstructure as drawn in Fig. 6, Pd and Zn atoms alternate along the [1¯21¯] direction. In this case, the angle α between the macroscopic surface plane and the line connecting neighbouring Pd and Zn atoms along [1¯21¯] is just the half of the difference in the critical angles, i.e. α = ΔΨC/2 = 1.5°. With the lateral Pd–Zn distance of 4.76 Å given by the Pd(1 1 1) substrate this translates into a vertical buckling of about 0.13 Å. However, for a three domain (2 × 1) surface, as it occurs in reality, we have to take into account that only two of three domains exhibit the double periodicity and the associated buckling along [1¯21¯], while in the third domain atoms of the same type sit besides each other. In such a multiple domain surface, the critical angle (i.e. onset of the backscattering intensity) is dominated by the domain with the smallest critical angle, which is the domain where the backscattering atom is located highest with respect to the shadow-cone producer. Thus, the onset angle for Zn (which is buckled outwards) is determined by the two domains exhibiting the double periodicity and the buckled Pd–Zn chains along [1¯21¯], while for Pd (which is displaced inwards) that domain dominates, which is not buckled along [1¯21¯] and exhibits Pd–Pd chains along this azimuth. This has two consequences: firstly, the critical angles for backscattering from Pd atoms of clean Pd(1 1 1) and from Pd atoms of a 3-domain p(2 × 1)-PdZn surface alloy should be equal. This is indeed observed in the experiment. Secondly, the angle α now directly equals the difference in the critical angles (α = ΔΨC = 3°), translating into a buckling of  ≈ 0.25 Å. This value agrees perfectly with the corrugation of 0.25 Å with DFT-calculations by Chen et al. [6] for bi-layered PdZn-films on Pd(1 1 1). As in the present experiment, the calculation finds that Zn atoms are displaced slightly out of the surface. Note that in the experiment for both the Pd and the Zn signals, the onset angle ΨC is dominated by those domains, where Pd atoms act as shadow-cone producers. Consequently, the shadow-cone widths cancel out in the difference ΔΨC and do not enter into the extracted value of the buckling.

Fig. 6
Schematic (top view) of a one-domain p(2 × 1)-PdZn/Pd(1 1 1) surface alloy with alternating stripes of Zn (grey circles) and Pd surface atoms (black circles)). Medium-sized and small circles denote second-layer ...

Finally, we note that in principle ICISS is also able to give information about the position of second-layer atoms by evaluation of the critical angles associated with the shadowing of second-layer by top-layer atoms. If atoms are stacked in an fcc sequence, this should result in a second intensity rise roughly 55° above the first onset angle, i.e. around Ψ = 65°–75°. Although this feature is readily observed for clean Pd(1 1 1) (Fig. 5a), the data in Fig. 5b for the PdZn surface alloy exhibit only broad weak structures which cannot be reliably interpreted. This may be caused by two effects: due to the buckling of the first layer and the existence of three domains different angles between first and second-layer atoms exist, yielding different critical angles for focussing from first onto second-layer atoms, thus spreading the second-layer peak into several less intense features. Furthermore we note that the ion beam may also be stronger neutralized upon penetration of the PdZn surface layer (as compared to a pure Pd layer), an effect observed, e.g. for ultrathin films of Pd and Au on Ru(0 0 1) [9].

4. Summary

Ultrathin zinc-films on Pd(1 1 1) were investigated by low-energy ion scattering. At 150 K films grow as simultaneous multilayers. Annealing experiments showed that the surface composition starts to change drastically around room temperature. Therefore, the existence of a closed zinc monolayer film up to 600 K, as proposed by Gabasch et al. [4], can be excluded. As already proposed by Bayer et al. [3], a metastable 1:1 surface composition is found after annealing to temperatures between 400 and 550 K and is attributed to the formation of a p(2 × 1) Pd0.5Zn0.5 surface alloy. Azimuthal scans of this phase show that the palladium and zinc atoms occupy the same lateral positions as the palladium atoms in clean Pd(1 1 1). Polar scans indicate a slight buckling of the topmost layer with Zn atoms located ≈ 0.25 Å above Pd atoms, which is consistent with DFT-calculations by Chen and co-workers [6].

Acknowledgements

We acknowledge excellent technical assistance by Reinhold Pramsoler. This work was financially supported by the Austrian Science Fund through grants S9004-N20 and P20892-N19.

References

1. Palo D.R., Dagle R.A., Holladay J.D. Chem. Rev. 2007;107:3992. A review on various aspects of methanol steam reforming is given by. [PubMed]
2. Iwasa N., Takezawa N. Top. Catal. 2003;22:215.
3. Bayer A., Flechtner K., Denecke R., Steinrück H.P., Neyman K.M., Rösch N. Surf. Sci. 2006;600:78.
4. Gabasch H., Knop-Gericke A., Schlögl R., Penner S., Jenewein B., Hayek K., Klötzer B. J. Phys. Chem. B. 2006;110:11391. [PubMed]
5. Jeroro A., Lebarier V., Datye A., Wang Y., Vohs J.M. Surf. Sci. 2007;601:5546.
6. Chen Z.-X., Neyman K.M., Rösch N. Surf. Sci. 2004;548:291.
7. Dürr H., Fauster Th., Schneider R. J. Vac. Sci. Technol. A. 1990;8:145.
8. Musket R.G., MacLean W., Colmenares C.A., Makowiecki D.M., Siekhaus W.J. Appl. Surf. Sci. 1982;10:143. and references therein.
9. Steltenpohl A., Memmel N., Taglauer E., Fauster Th., Onsgaard J. Surf. Sci. 1997;382:300.
10. Fauster Th. Vacuum. 1988;38:129.
11. Aono M., Oshima C., Zaima S., Otani S., Ishizaewa Y. Jpn. J. Appl. Phys. 1981;20:L829.