Toward Controlled Modification of Nanoporous Gold. A Detailed

*E-mail: [email protected] (A. Schaefer); [email protected] (A. Sandell). Info icon. Your current credentials do not allow ret...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/JPCC

Toward Controlled Modification of Nanoporous Gold. A Detailed Surface Science Study on Cleaning and Oxidation A. Schaefer,*,† D. Ragazzon,† A. Wittstock,‡ L. E. Walle,§ A. Borg,§ M. Baü mer,‡ and A. Sandell*,† †

Uppsala University, Department of Physics and Astronomy, P.O. Box 516, SE-751 20 Uppsala, Sweden University of Bremen, Institute of Applied and Physical Chemistry, P.O. Box 33 04 40, D-28359 Bremen, Germany § Norwegian University of Science and Technology, Department of Physics, NO-7491 Trondheim, Norway ‡

ABSTRACT: Nanostructured metals and especially unsupported nanoporous gold (np-Au) have attracted considerable attention in a variety of fields because of their special surface chemical properties. For applications in catalysis and sensorics, the oxidation of the metal and the availability of oxygen at the very surface are crucial and also are capable of altering structural properties. In this article, we will discuss the state of the np-Au surface after annealing in vacuum. We shed light on the nature of Au-oxide obtained after cleaning the surface from carbon impurities with atomic oxygen provided by ozone decomposition, and we consider the effect of this procedure on silver residues. The results provide new insight into possible oxide species at the np-Au surface and represent a vital step toward controlled modification of the np-Au surface in the future.



INTRODUCTION Nanoporous gold (np-Au) is a fascinating bulk-nanostructured material. It is made up of a three-dimensional mesh of ligaments which can be between some tens of nanometers up to several micrometers depending on the preparation conditions chosen.1 This material is coined by its extended metallic surface and its high surface to volume ratio. Many of the unique properties of np-Au are a direct consequence of its spongelike morphology making it penetrable for fluids and gases and thus providing a high surface area. In recent years, several applications of this material in optics, sensorics, and heterogeneous catalysis have been proposed.2−9 The key to many of these applications is the extended metallic surface being reversibly oxidized by molecular oxygen or ozone. In view of sensors and actuators, the change of the macroscopic dimensions of np-Au by oxidation of the surface by ozone readily decomposing on Au surfaces10,11 is highly interesting.4 By bonding of oxygen to the Au surface, the surface stress is altered inducing a macroscopically detectable strain response owing to the high fraction of surface atoms in np-Au in the order of several percent.12 Recent studies covering the thermal stability of np-Au revealed that surface oxygen plays a crucial role in stabilizing its structure and preventing coarsening13 since the structure of np-Au is intrinsically instable by curvature driven diffusion (Gibbs-Thompson effect). By bonding to, for example, oxygen atoms, the surface self-diffusion rate of Au atoms is reduced, and thus, the structure is stabilized. When npAu samples are annealed under an ozone-containing atmosphere leading to sufficiently high oxygen coverage, thermal stability increases. Besides its impact on the mechanical properties of np-Au, surface-bonded oxygen can also be used as a reactant in heterogeneous catalysis.3,14−17 For example, np© 2012 American Chemical Society

Au was shown to be an excellent catalyst for low-temperature oxidation of CO and for highly selective oxidation of alcohols, such as methanol.3,18,19 These studies demonstrate that these applications are strongly linked to the reversible oxidation of the np-Au surface. Experimental as well as computational studies on oxide-supported Au nanoparticles revealed that the low-coordinated atoms are essential for their catalytic activity.20 However, the nature of the different oxygen species which can be formed on np-Au remains unclear. Au oxides and oxygen species have been observed in surface science studies of gold single crystals by means of scanning tunneling microscopy (STM),21 thermal desorption spectroscopy (TDS),22,23 and photoemission.24,25 The observed structures and stoichiometries were dependent on the preparation method. To shed light on possible oxide species on np-Au and the changes of the surface upon oxidation and reduction, we used high-resolution photoelectron spectroscopy (PES) based on synchrotron radiation (SR) under ultrahighvacuum (UHV) conditions. We investigated a sample after heating and ozone treatment in UHV followed by subsequent annealing. This high-resolution core level spectroscopy (HRCLS) study on np-Au reveals details in, especially, the Au 4f spectra which are of importance for a more detailed understanding of the Au/oxygen interaction. In this article, we discuss the nature of gold oxide generated by ozone treatment and the ratio of the different oxygen species present, which is found to be reversed compared to single crystal surfaces. The high-resolution Au 4f spectra reveal new Received: August 9, 2011 Revised: January 4, 2012 Published: January 17, 2012 4564

dx.doi.org/10.1021/jp207638t | J. Phys. Chem. C 2012, 116, 4564−4571

The Journal of Physical Chemistry C

Article

a gasdoser which was placed in close proximity (∼1 mm distance). The sample temperature was kept at 120 °C. The pressure during dosage was 2.5 × 10−7 mbar. Employing a quadrupol mass spectrometer (MKS Instruments), no contaminations in the gas could be detected, and thus, no further purification was applied. Because of the low efficiency of oxygen dissociation on these gold surfaces,31 the decomposition of ozone is needed to create an oxidized surface.10,11 Scanning electron microscopy (SEM) was conducted using the Zeiss SUPRA 400 instrument of the geology department of the University Bremen, Germany.

features related to the oxide species formed and will be discussed in detail. Moreover, the so far unknown response of the residual Ag to oxidation treatment will be addressed. In summary, the results provide essential information for experiments on controlled modification of np-Au in the future.



EXPERIMENTAL SECTION Disks of np-Au with diameters of 5 mm and thicknesses of approximately 300 μm were prepared by etching of Ag(70 atom %)−Au(30 atom %) alloys in concentrated nitric acid for 48 h (HNO3, 65 wt %, Fluka Chemical Corp.). This procedure is very efficient in removing Ag almost quantitatively to a bulk concentration below 1 atom %.26,27 However, the silver content remains slightly higher at the surface (∼4−7 atom % reported in the literature previously).18,28 The estimated amount of silver can, however, differ slightly from sample to sample because of differences in the sample preparation and the quantification method (cf. below). The photoelectron studies were conducted at the undulator beamline I311 at the MAX-lab synchrotron radiation facility in Lund, Sweden. Heating and ozone experiments were done at the endstation of I311. The sample was mounted and degassed together with the manipulator. The sample temperature reached ∼160 °C for a couple of minutes before it was cooled to 145 °C for the rest of the bakeout (19 h) . The temperature was measured by a type-K thermocouple placed in close proximity to the sample. The endstation is housing a SCIENTA SES200 electron energy analyzer. The spectral resolution obtained was in the range of 100 meV or better as determined from the width of the Fermi edge. The base pressure during analysis was in the low 10−10 mbar range. For data evaluation, the spectra were delineated using the program fitt (Hyun-Jo Kim, Department of Physics, Seoul National University, v1.02 2000). The program uses a convolution of a Gaussian with a Doniach-Šunjić function (DSF) and a Levenberg−Marquardt method. Fitting of the Au 4f spectra for np-Au is a nontrivial task. The np-Au morphology will potentially lead to a multitude of chemically shifted components because of the abundance of steps and kinks in addition to the surface core level shift associated with terraces.29 In addition, the presence of Ag is expected to induce positive shifts versus pure Au that are of a few tens of meV.30 To capture the most essential changes, a minimum number of components were used on the basis of previous studies of single crystal surfaces. As a result, the function describing each of the individual peaks should be viewed as a model that takes the above-mentioned effects into account. Specifically, the lifetime broadening given by the width of the Lorentzian in the DSF was set to 350 meV. The Gaussian contribution was small for the main lines (∼80 meV) but was varied for the oxide signal in the Au 4f spectra in the range of 650−720 meV. The calibration of the binding energy (BE) scale was done for each photon energy (PE) applied using the position of the Fermi edge. Binding energy values as obtained by the fits are accurate by ±50 meV. The oxygen and silver spectra shown are normalized to their low binding energy side. Au 4f spectra are normalized to the bulk peak maximum to better show features on the low binding energy region. Ozone treatments were applied using a simple method which is easy to establish at different facilities, namely, an O2/O3 gas mixture provided by a commercially available ozone generator (TopChem) providing ∼7 vol % ozone in an O2 stream. The sample was exposed to the O2/O3 gas mixture for two hours via



RESULTS We will first present the results obtained after bakeout of the vacuum chamber (see Experimental Section). The bakeout procedure is only sufficient to remove volatile (hydro-) carbon contaminations present after atmosphere exposure of the freshly dealloyed sample. The carbon 1s core level spectrum presented in Figure 1 a shows that mainly reduced carbon

Figure 1. Carbon 1s core level spectra recorded at 390 eV photon energy (PE) (a) after heating the np-Au sample in UHV for 19 h to 145 °C and (b) after subjecting the sample to an oxygen/ozone [7 vol % ozone] gas mixture at 120 °C sample temperature for 2 h.

species remain on the surface (around 284.2 eV BE). A small amount of oxidized carbon species also contributes to the spectrum broadening it toward higher BE values. The Au 4f7/2 spectrum after the bakeout displays a bulk component at 84.01 eV, Figure 2a. In addition, deconvolution of the spectrum reveals two further contributions on the low binding energy side at 83.83 and 83.62 eV. Studies of single crystal Au surfaces have shown that states at these binding energies can be assigned to low-coordinated Au atoms at the surface, that is, at terrace and step (kink) sites.29,32 As the presence of Ag is expected to result in only very small Au 4f shifts (about +40 meV for Au80Ag2030), we assume that the identification of the low BE components for single crystal Au also holds for npAu. From this follows that the pristine sample, as studied after bakeout, exposes clean surface gold atoms. There is evidence of an appreciable amount of residual silver, and the Ag 3d5/2 signal is found at 367.89 [±0.05] eV after bakeout (Figure 2a). This value is 0.1−0.3 eV lower than those reported for pure Ag samples.33 The observed difference, with a shift in opposite direction to that of Au 4f, is most likely due to Ag−Au alloying.34,35 The corresponding O 1s spectrum in 4565

dx.doi.org/10.1021/jp207638t | J. Phys. Chem. C 2012, 116, 4564−4571

The Journal of Physical Chemistry C

Article

Figure 2. O 1s, Ag 3d5/2, and Au 4f7/2 core level spectra obtained after the different sample treatments. O 1s: the ozone treatment generated two new species which are assigned to chemisorbed oxygen (CSO) and a surface oxide (SO); the oxide is not stable in vacuum over time. The spectra are normalized to the low binding energy side. Ag 3d: after heating, the residual silver is metallic. Ozone treatment completely oxidizes the silver, and the signal intensity increases. After three days in vacuum, the silver is reduced again (this spectrum is taken from a survey scan and is thus measured with lower resolution than the others). Au 4f: the spectra are normalized to the peak maximum. Bulk and two surface components from terraces and steps can be discerned. After ozone treatment, the gold is oxidized and a new contribution appears in the spectra at 83.41 eV. Heating to 150 °C for 15 min slightly reduces the gold. The gold oxide has completely vanished after three days in vacuum.

surfaces.10,25,36 The large Gaussian width observed for the oxide signal originates most likely from phonon broadening and disorder in the generated Au-oxide layer. The Au 4f7/2 peak comprising states associated with metallic gold (BE < 84.5 eV) still shows features at binding energies below that of the bulk component. That is, clean surface gold appears to be present also after oxidation. Moreover, the peak appears to be even broader when comparing to the spectrum measured directly after bakeout. Deconvolution of the signal reveals that the broadening can be attributed to an increased Au 4f (step)/Au 4f (terrace) intensity ratio and, in addition, to a new Au 4f7/2 component at a BE of 83.41 eV. This is a remarkably low BE since it is even lower than that associated with steps (kinks).29

Figure 2a displays two weak features that, on the basis of their BE values, can be assigned to oxygen bound to carbon species (lower binding energy) and probably to hydroxyl groups (higher binding energy). In the next step, we subjected the sample to an oxygen/ ozone gas mixture containing ∼7 vol % of ozone (see Experimental Section for details). As can be seen from spectrum 1b, this method is very effective in removing the carbonaceous species almost completely. Only spurious amounts of carbon remain after the cleaning located at slightly higher BE values than the main line in spectrum 1a. Changes are also observed in the Au 4f7/2 spectrum in Figure 2b. A broad and intense signal located 1.25 eV above the bulk signal has appeared. Clearly, the ozone treatment leads to oxidization of the gold in agreement with previous studies on flat 4566

dx.doi.org/10.1021/jp207638t | J. Phys. Chem. C 2012, 116, 4564−4571

The Journal of Physical Chemistry C

Article

The Au 4f spectra have also been normalized to the background level at the low BE edge in order to address absolute changes. This comparison (not shown) yields that the intensity from states associated with clean surface gold atoms is lower after oxidation than for the situation directly after bakeout. We can therefore conclude that oxidation leads to a very efficient removal of carbon surface contaminants and to the formation of an oxide layer that covers a larger part of the surface than that covered by the contamination layer present after bakeout. The residual silver at the surface becomes completely oxidized during ozone treatment. This is observed as a shift of the Ag 3d5/2 signal down to 367.08 eV accompanied by a broadening of the signal in Figure 2b. Reported BE values for the highest oxidized state of Ag, that is, AgO, are in the range of 367.2 eV.33 In the O 1s spectrum, shown in Figure 2b, two new distinct components appear at BEs of 529.08 and 529.94 eV. These two states are usually associated with species denoted chemisorbed oxygen (CSO, low BE component) and surface gold oxide (SO, high BE component)25,37 (cf. Discussion). From the O 1s states observed after bakeout, only a small contribution at 530.6 eV is left while the signal at highest BE is no longer visible. For the discussion of the generated oxide layer, an estimation of the oxide coverage is helpful. The Au 4f signal attenuation indicates a coverage of approximately 0.8 monolayers (MLs) when assuming a flat surface. The same value is obtained by considering the ratio of the Au 4f oxide signal intensity to the Au 4f bulk signal of the untreated sample. Since the surface still contained carbon impurities and a flat surface is assumed, this is only a rough estimation of the oxide coverage but still is in line with results obtained on flat gold surfaces. The core level shift (CLS) for Au oxide reported in literature varies in a range from +2.1 eV to +1.4 eV (going from thicker to thinner films) depending on the preparation method.24,25,36−38 The BE position of the oxide signal in the Au4f spectra in Figure 2 (+1.25 eV relative to the Au4f bulk signal) is thus close to the lowest values reported for oxidized gold.25,36,37 As a general rule for flat Au surfaces, the BE shift of the oxide peak seems to correlate with the layer thickness. Oxide thicknesses limited to a monolayer when using moderate ozone treatments have been reported in earlier studies on flat surfaces,10 while several layers of gold oxide (Au2O3) could be generated when using UV irradiation while ozone is supplied.25 To explore the stability of the oxygen species, we heated the sample to 150 °C for 15 min (pressure during annealing 0.33 ML.41 The two intense signals in the O 1s spectrum observed after oxidation using ozone (Figure 2c) are usually associated with CSO (low BE signal) and SO (high BE signal).37 In the case of flat Au surfaces (e.g., Au(111)) with increasing temperature and oxygen supply, the more ordered SO is strongly favored. For an oxygen coverage of 0.55 ML on Au(111), experiments and calculations result in a CSO/SO ratio of only 0.14−0.35 in the temperature range of 200−500 K. We observe a very different behavior for the np-Au surface: The O 1s spectra show clearly that the CSO is the prevailing species also at higher temperatures and oxygen coverages. We generate a nominal oxygen coverage of 0.8 ML at 400 K for which we obtain a CSO/SO ratio of ∼2.5. Thus, compared to a flat single crystal surface, the ratio between CSO and atomic SO for np-Au is 1 order of magnitude larger. A possible explanation for this observation is that the highly porous structure of np-Au prevents the formation of ordered oxide structures. According to common models, np-Au exposes a high number of low-coordinated atoms at steps, mounts, and pits.26,42 On the basis of a simple geometric consideration (cutting along, e.g., the 100 direction), a large fraction of the atoms are located at step edges (approximately 10−15% with a coordination number