pubs.acs.org/Langmuir © 2010 American Chemical Society
Influence of Thermal Treatment on Nanostructured Gold Model Catalysts† Michael Bowker,* Melissa Broughton, Albert Carley, Phil Davies, David Morgan, and Jon Crouch Wolfson Nanoscience Laboratory and Cardiff Catalysis Institute, School of Chemistry, Cardiff University, Cardiff CF10 3AT, United Kingdom
Georgi Lalev, Stefan Dimov, and Duc-Truong Pham Manufacturing Engineering Centre, Cardiff University, Queen’s Buildings, The Parade, Newport Road, Cardiff CF24 3AA, United Kingdom Received April 7, 2010. Revised Manuscript Received May 17, 2010 We fabricated films of Au onto single crystal alumina (Al2O3(0001)) and nanostructured the surface using a high resolution focused ion beam (FIB) to remove specific regions of the film. The nanostructures consist of lines and orthogonal lines cut into the film, resulting in one- and two-dimensional islands of gold. When these films are heated above 300 °C, small nanoparticles of gold form due to the dewetting of the Au film from the alumina surface. The dimensions of these islands are dictated by the nature of the nanopatterning. The isolated islands generally have the smallest nanoparticles after heating, while the unpatterned film has much larger particles. Sintering is reduced within the nanostructured metal domains due to isolation of Au islands from each other. The evaporation rate is higher within these islands, due to the smaller size of nanoparticles and hence the higher effective vapor pressure over the surface (the Kelvin effect).
Introduction An important element of modern nanoscience is the ability to nanostructure surfaces in a controlled manner. A number of different approaches are available to achieve this,1-4 and the field is enormous. In particular, and in relation to this issue of Langmuir, celebrating Gabor Somorjai’s 75th birthday, he has, in the last 20 years, devoted considerable effort to developing methods to make model catalysts and, in relation to the current paper, ordered arrays of nanoparticles as very well-defined models of industrial catalyst surfaces.5-7 He has demonstrated that electron-beam lithography can produce ordered arrays of Pt “posts” on alumina films deposited on a silicon single crystal; this is a so-called “top-down” approach, which has inherent difficulties, particularly with writing arrays of significant areal coverage in a reasonable time scale. He has more recently diversified to a number of bottom-up approaches including colloidal approaches8 and ligand-templating.9 In our laboratory, we are particularly interested in the fabrication of nanoparticle arrays and model catalysts10,11 and in the investigation of their properties. Nanoparticles are important in a wide range of industries, not the least in the chemical industry, where heterogeneous catalysts are employed to influence particular reactions. Such catalysts are often arrays of metal nanoparticles †
Part of the Molecular Surface Chemistry and Its Applications special issue.
(1) Innocenzi, P.; Kidchob, T.; Falcaro, P.; Takahashi, M. Chem. Mater. 2008, 20, 607. (2) Roy, S. J. Phys. D: Appl. Phys. 2007, 40, R413. (3) Xie, X. N.; Chung, H. J.; Sow, C. H.; Wee, A. T. S. Mater. Sci. Eng. 2006, R 54, 1. (4) Henzie, J.; Barton, J. E.; Stender, C. L.; Odom, T. W. Acc. Chem. Res. 2006, 39, 249. (5) Hwang, K.; Yang, M.; Zhu, Ji.; Grunes, J.; Somorjai, G. A. J. Mol. Catal. A: Chem. 2003, 204-205, 499. (6) Epler, A.; Zhu, Ji.; Anderson, E.; Somorjai, G. A. Top. Catal. 2000, 13, 33. (7) Zhu., J.; Somorjai, G. A. Nano Lett. 2001, 1, 8. (8) Joo, S. H.; Park, J. Y.; Tsung, C.-K.; Yamada, Y.; Yang, P.; Somorjai, G. A. Nat. Mater. 2009, 8, 126. (9) Huang, W. Y.; Kuhn, J. N.; Tsung, C. K.; Zhang, Y. W.; Habas, H. E.; Yang, P. D.; Somorjai, G. A. Nano Lett. 2008, 8, 2027. (10) Bowker, M. Chem. Soc. Rev. 2007, 36, 1656. (11) Bowker, M. Phys. Chem. Chem. Phys. 2007, 9, 3514.
Langmuir 2010, 26(21), 16261–16266
anchored to refractory support materials, such as alumina. The metal particles are supported on such oxides to significantly slow down the rate of sintering of particles, by maintaining their integrity and separation in space. Nonetheless, because of the aggressive environments in which they are used (frequently involving high temperatures and high pressures of reactive gases) loss of surface area due to sintering still accounts for a significant loss of activity. Clearly, then, it is important to maintain the concentration of active sites (proportional to the surface area of the nanoparticles), to be as high as possible to give the highest reaction rate. Largescale industrial reactors have to be shut down periodically when the activity becomes reduced below a critical limit, a process which costs significant time and money, and it is often the case that the activity loss is largely due to sintering. Such sintering effects are of significance in many other technological areas, and will be especially important in the new era of nanomaterials manufacturing for a wide variety of applications. Thus, significant effort has been expended toward understanding the nature of sintering. Any approach that can reduce sintering will have beneficial effects on the economics of catalytic processes, and on the environment, in terms of reduced amounts of catalytic materials (due to enhanced average activity) and lengthened replacement cycles for catalytic materials. In relation to these processes, we are particularly interested in the preparation of arrays of nanoparticles on surfaces and in the thermal evolution of these arrays, something which has been relatively infrequently addressed in the literature. Fabrication and structure at approximately ambient temperature have been the primary focus of efforts to date. Thus, we have attempted to investigate the effect of nanostructuring of gold films on the formation of particles and on their morphological evolution during heating. We have used focused ion beams (FIBs) to make ordered arrays. The work is addressed at answering a number of questions, such as: (i) will restriction of the size of gold domains affect the thermal evolution and sintering process; (ii) will single nanoparticles form in the area of a restricted island of gold formed by FIB; and (iii) will such films
Published on Web 05/28/2010
DOI: 10.1021/la101372w
16261
Article
Bowker et al.
simply “forget” the ordering of the nanostructures imposed upon them after heating to elevated temperature? That is, will they diffuse across the boundaries imposed on them by nanostructuring?
Experimental Section The sample used as the base for deposition of gold was an alumina (0001) single crystal of 5 N purity (PiKem U.K.). The preparation of the nanostructured sample first involved the deposition of a thin layer of Au onto the single crystal of alumina held at ambient temperature and with a deposition rate of 0.2 nm/ min. This was achieved by resistive evaporation in which a sample of Au wire was placed in a tungsten evaporation boat and mounted in a thermal resistive evaporator. The pressure in the evaporator was then reduced to 2 10-6 Torr before the evaporation boat was subjected to resistive heating, using a 60 A current source. At this point, the Au source evaporated and was deposited onto the single crystal of alumina from a distance of 30 cm, ensuring that only a thin film of Au was deposited. The film thickness (∼15 nm) was monitored during the deposition process using a precalibrated crystal monitor. The next stage in the crystal preparation involves focused ion beam (FIB, Carl-Zeiss XB 1540) milling in order to fabricate various patterns into the crystal, using gallium ions, which mills away the Au film and then also the alumina support beneath the film by sputtering. The FIB system is controlled externally by using the Eliphy Quantum lithography software package which facilitates not only the design of the pattern to be milled, but also the control of the dwell time of the ion beam at each individual pixel, step size between the pixels, number of loops, and so on. The sample was inspected in situ during the milling process using a high-resolution field emission scanning electron microscope (FE SEM) integrated into the FIB system. In this particular study, trenches were milled into the alumina with a pitch of 400 nm and width of 75 nm. We report the results for three types of structure: (i) the as-deposited gold film (henceforth in the paper, labeled as “film”), (ii) the film with lines milled by FIB as described above, leaving parallel lines of gold on the surface (henceforth “lines”), and (iii) the film with isolated squares of gold film (henceforth “cells”). The sample was calcined to a variety of temperatures and was postimaged with a Veeco Multimode atomic force microscope (AFM) using a contact mode tip; the alumina surfaces were hard enough to withstand the force exerted by this mode of operation. All of the images obtained from the AFM were analyzed, flattened, and/or equalized using the WSxM12 computer program; it was also used to obtain line profiles and particle size distributions at each calcination temperature. The determination of the particle size distribution involved obtaining a map, based on the rate of change of height within each scan line, of the image showing the dispersion of the particles on the surface. This image produced then allowed for the determination of the perimeter and area of the particles, directly from the shape of the particles. The height of the particles then allowed for the calculation of the volume distributions. The X-ray photoelectron spectrometer used was a VG-ESCALAB 200i spectrometer, with an achromatic X-ray source and an Al (KR) anode. The spectrometer runs on 220 W power (20 mA emission, 11 kV potential). Note that although the sample has nanofabricated areas, these only account for about 0.001% of the area which is measured by X-ray photoelectron spectroscopy (XPS), and so the XPS signal is dominated by the unstructured Au film.
Results and Discussion Figure 1 shows images of nonpatterned and patterned areas of the gold film which was deposited on the alumina [0001] single crystal. Figure 1a is an image of the gold film formed by metal vapor deposition as described above in the area outside the nanostructured (12) http://wsxm.software.informer.com/.
16262 DOI: 10.1021/la101372w
Figure 1. Images of the Au/Al2O3(0001) films prepared by metal vapor deposition and FIB milling. (a) AFM image of the unstructured area of the Au film; (b) AFM image of the area patterned with parallel lines by FIB; and (c) AFM image of the cell area fabricated with orthogonal lines by FIB.
Figure 2. XP spectra for (a) the as-prepared sample and (b) after heating to 900 °C and (c) 1100 °C. Note that the peaks in the 7080 eV region in (a) are due to satellite lines from the Au(4f) peaks.
region (the vast majority of the surface area). Although this is a nonstructured area, the film is grainy, with an average apparent grain size of ∼20 nm. However, XPS shows (Figure 2a) that the surface coverage by gold is essentially complete. Since the inelastic mean free path of the Al photoelectrons (kinetic energy ∼ 1400 eV) is only ∼3 nm and the film is ∼15 nm thick, then the Al/Au ratio should be near zero (∼0.007). There is very little extra intensity in the Al(2p) region of the spectrum at 73 eV (note that the doublet at ∼73.5 and 77.5 eV binding energy is due the KR3,4 satellites of the main Au(4f) peaks), thus confirming complete coverage of the surface by gold. Nonetheless, since the growth mode of gold on oxide surfaces is Langmuir 2010, 26(21), 16261–16266
Bowker et al.
Article
Figure 3. AFM images of an unmodified area of the Au/Al2O3 sample. Two different magnifications are shown for the sample shown in Figure 1 after heating in air to the temperatures shown. Note that after heating to 300 °C there was little change from the as-received sample.
generally described as Volmer-Weber type,13 then, growth begins from nuclei of gold on the surface and these nuclei grow to form bigger nanoparticles, which grow to form grain boundaries between adjacent particles. In the equilibrium limit, such a growth mode is favored because metals generally do not “wet” oxides, due to the lower surface energies of the latter. Presumably, these bigger nanoparticles eventually come to merge together as growth continues, but they leave some holes exposing the alumina crystal underneath (the small, black areas in Figure 1a, for instance, near the center-top of the image). This is similar to our previous observations by scanning tunneling microscopy (STM) for Pd films on titania,10,11,14,15 where beautiful films were formed after deposition and annealing, yet holes which exposed the single crystal titania substrate remained, even though the coverage of Pd was ∼7 monolayers. However, from the observed Al/Au ratio (see Figure 2), we can infer that these holes in the gold film represent a very small fraction of the surface. The grains have grown to mostly fill in the holes, leaving grain boundaries between the original nuclei. The Au film was patterned by FIB milling, as shown in Figure 1b and c, with two different nanostructures milled into the film. These consist of sets of 75 nm wide etched lines with a pitch of 400 nm (“lines”) and of intersecting orthogonal sets of lines, leaving isolated square islands (cells) of Au on the alumina surface. Although Ga ions have been used for the milling process, there is no sign of Ga in the XPS spectrum, because the nanostructured areas only account for a very small fraction of the area analyzed by XPS, and the etched area is an even smaller fraction of that area. However, EDAX analysis indicated that even in the etched areas the Ga level in the analyzed region is only ∼0.2 atom %, in agreement with the low levels reported by others after FIB milling.16,17 It is clear from Figure 1b and c that the milling is effective and results in patterns milled into the surface to a depth of about 20 nm. It is noticeable that there is some accumulation of Au on the remaining film close to the edge of the trenches. In fact, this is due to the typical (13) Chatain, D.; Coudurier, L.; Eustathopoulos, N. Rev. Phys. Appl. 1988, 23, 1055. (14) Stone, P.; Poulston, S.; Bennett, R. A.; Bowker, M. Chem. Commun. 1998, 1369. (15) Bennett, R. A.; Stone, P.; Bowker, M. Faraday Discuss. 1999, 114, 267. (16) Heft, A.; Wendler, E. Nucl. Instrum. Methods Phys. Res., Sect. B 1996, 113, 239. (17) McHargue, C. J.; Williams, J. M. Nucl. Instrum. Methods Phys. Res., Sect. B 1993, 80, 889.
Langmuir 2010, 26(21), 16261–16266
swelling effect observed during FIB and results from ion implantation.18 This effect is caused by the Gaussian spatial distribution of the ion beam. The somewhat grainy nature of the gold, which was present in the original film, is clearly seen in this figure as well, though it seems that the cells may be porous, presumably due to the FIB process itself. Figures 3-5 show the effect of heating the two nanostructured areas and the unstructured gold film; the effects are rather marked. In each case, heating to 300 °C has very little effect on the films, but changes begin to occur on heating to higher temperatures. At 500 °C, it is clear the films have separated out to form nanoparticles of distinct shape and identity. There are differences in the size of the nanoparticles between the lines and cells, with the latter being significantly smaller at this temperature, but both of these are smaller than that for the nanoparticles formed from the unstructured film (see Tables 1-3 and Figures 3-5). It is noticeable, however, that there is a bimodal distribution of particles in the unstructured film, with a minority of the gold as small particles of ∼40 nm diameter. The average diameter of the particles for the 700 °C annealed unstructured film is 360 nm (and many of the very small particles have been lost), whereas for the structured films the average diameter is ∼250 nm. Some of these changes are clarified by the line profiles shown in Figure 6, Tables 1-3; and particle size distributions are shown in Figure 7. It is probable that the structures formed at 500 °C are due to the dewetting of alumina from grain boundaries in the Au film, whereas sintering occurs at higher temperatures, resulting in loss of the smaller particles (ripening). It is somewhat surprising to note that there does not appear to be a significant change in particle size distribution between 500 and 900 °C. However, there are some changes in the morphology of the particles. There is evidence in Figure 4 that, after heating, Au accumulates at the edge of the scribed lines on the surface (the line itself is still clearly evident), producing an almost continuous line of Au at the edges of the original linear islands, and leaving a lower height area in the middle of the line. By 700 °C and certainly by 900 °C, this is not evident and only particles remain whose average size is ∼250 nm. In the cells, the particles appear to be somewhat more resistant to sintering, since, although there is some change by heating to 500 °C, it is relatively minor; the (18) Menzel, R.; Bachmann, T.; Machalett, F.; Wesch, W.; Lang, U.; Wendt, M.; Musil, C.; Muhle, R. Appl. Surf. Sci. 1998, 136, 1.
DOI: 10.1021/la101372w
16263
Article
Bowker et al.
Figure 4. AFM images of the Au/Al2O3 sample with lines milled into the surface using FIB. Three different magnifications are shown for the sample as-received, and after heating in air to the temperatures indicated.
Figure 5. AFM images of the Au/Al2O3 sample with orthogonal lines milled into the surface using FIB to create a network of Au cells. Three different magnifications are shown for the sample as-received, and after heating in air to the temperatures indicated. Table 1. Structural Parameters for Nanoparticles on the Unstructured Film anneal temperature (°C)
D (nm)
F (μm-2)
I (nm)
θ (%)
V (μm3)
500 420 4.0 394 42 1.1 700 360 3.5 435 38 1.2 900 330 3.5 423 32 1.3 1100 200 2.5 440 18 0.6 D, mean particle size; F, particle density; I, mean interparticle separation; θ, particle coverage of surface; V, particle volume in the imaged area of 100 μm2.
original structure inside the cells is attempting to separate into very small nanoparticles of ∼50 nm in size, though there still seems to be considerable interconnection between them. Separate particles are definitely seen by 700 °C (average size 260 nm), with little change by 900 °C (230 nm), with each box tending to include one or two nanoparticles. In general, at low temperatures, the particles are rather irregular in shape and become more rounded with annealing. This relates to the original grainy film, which contains somewhat irregular-shaped grains, and these appear to 16264 DOI: 10.1021/la101372w
Table 2. Structural Parameters for Nanoparticles in the Line-Patterned FIB Film anneal temperature (°C)
D (nm)
F (μm-2)
I (nm)
θ (%)
V (μm3)
500 700 900 1100
200 230 270 200
7.6 5.0 6.9 2.2
262 329 285 220
42 36 40 8
0.8 1.0 1.3 0.5
Table 3. Structural Parameters for Nanoparticles in the Cell-Patterned FIB Film anneal temperature (°C)
D (nm)
F (μm-2)
I (nm)
θp (%)
Vp (μm3)
500 700 900 1100
270 260 230 200
10.6 7.9 6.7 1.4
250 240 290 578
42 38 36 4
0.5 0.7 0.7 0.2
separate by dewetting/sintering into irregular shaped nanoparticles at 500 °C, while they convert into more rounded particles at higher temperature. This is due to thermodynamic constraints; Langmuir 2010, 26(21), 16261–16266
Bowker et al.
Article
Figure 6. Line profiles across the cell areas and their evolution with annealing temperature.
Figure 7. Particle size (in terms of perimeter) distribution with annealing temperature for the three types of structures (unstructured and structured).
rounded particles have a lower number of surface atoms and therefore lower energy. This transformation requires thermal treatment to enable the surface diffusion required for minimization of the number of surface atoms. Thus, when such films are heated, it appears that some boundaries merge while others separate as the film tries to minimize its surface area by ultimately forming more rounded particles. Restriction of the dimensions of the Au by prepatterning appears to result in smaller particles on average. This probably occurs because, effectively, the particle size in these basic grains (that is, the dimension of the lines and cells), before annealing, are artificially reduced by the ion beam milling process, effectively “fencing” the gold within these boundaries. This also reduces the intergrain boundaries and reduces the opportunity for fusing at this boundary while heating. Langmuir 2010, 26(21), 16261–16266
At 1100 °C, which is above the bulk melting temperature of gold (1064 °C), it is clear that a significant amount of gold has been lost from the surface. In all cases, there is a much lower number density of particles on the surface and a corresponding loss in volume is calculated (Tables 1-3). It is also clear there has been some patterning of the alumina substrate by the FIB beam below the original Au film, since lines can be seen on the alumina between particles in Figures 4 and 5. An estimate, from the images, is that ∼50% of the gold has been lost from the surface, and that the surface coverage by the particles is much lower, especially for the cells, which, from image analysis, appear to lose nearly 3/4 of the surface coverage of the original layer. In XPS, the Au/Al signal ratio has dropped to ∼0.7 at this point compared with ∼1.0 after heating to 900 °C. Note that this loss of gold is not due to diffusion into the region outside the fabricated area, as no DOI: 10.1021/la101372w
16265
Article
Bowker et al.
evidence of gold build-up there for any of the structured regions was found. The fact that there are fewer particles in the nanostructured area and that they are smaller than those in the nonstructured area after high temperature annealing appears to be due to an enhanced loss rate, not due to sintering, but due to evaporation, analogous to the Kelvin (or Gibbs-Thomson) effect. Due to the smaller size of the nanoparticles, the rate of loss of material from the surface of the particle is high compared with its bulk volume, that is, it runs out of material more rapidly, and hence, such particles disappear before the larger particles do. We can approximate these rates in the following way -d½Au=dt ¼ k e ½Aus ¼ k e 0 2πr2
ð1Þ
where the left-hand side represents the rate of loss of gold from a particle, ke is the evaporation rate constant from the Au particle and this is multiplied by the amount of surface Au available for evaporation. This is, in turn, represented on the right-hand side of eq 1 in terms of surface area and ke0 = keNs, where Ns, the surface concentration of Au atoms, can be expressed in terms of atoms m-2 or monolayers. Substituting for the particle volume (proportional to the total amount of gold), we obtain -dV=dt ¼ k e 0 F2πð3V =2πÞ2=3 ¼ kV 2=3
ð2Þ
where k = ke0 F32/3(2π)1/3 and F is the density of gold. Upon integration, the time evolution of volume is given as V t 1=3 ¼ V 0 1=3 - kt=3
ð3Þ
which can be expressed again in terms of particle radius as rt ¼ r0 - k e 0 Ft
ð4Þ
that is, there is a zero order dependence of radius upon time. If we consider two particles, one twice the size of the other, then the smaller one will have disappeared when the larger one has only decreased in size by a factor of 2. Of course, evaporation is not the only process taking place, since ripening is also, to a degree, involved. A summary of the changes in a number of parameters for the films, both structured and unstructured, is given in Tables 1-3. Note here that the analysis is least accurate for the 500 °C data, since the particle separation is often very small, making the distinction between particles in the image analysis difficult (thus somewhat overestimating the size of the largest particles and underestimating the number of particles). It is interesting to note that diffusion of significant amounts of Au into the channels cut into the film does not occur, and the material remains confined in its original boundary (most obviously in Figure 4 at 500 and 700 °C). This is due to several factors. First, the binding of Au to oxides is generally weak, so any atomic diffusion tends to have an end point at Au regions where atomic binding is higher. The sintering which occurs with heating results in reduced surface coverage by gold and a reduction in gold at the edges of the lines milled into the surface So what is the mechanism of the sintering process we see in Figures 3-5? It appears to be a combination of coalescence between adjacent particles (dominant at high particle densities) and surfacemediated Ostwald ripening. These have been reported by us for small (∼5 nm) Pd nanoparticles on TiO2.10,11,14,15 Coalescence
16266 DOI: 10.1021/la101372w
occurs merely due to contact between particles, but we have shown that, for Pd particles very much smaller than the ones studied here, such particles are immobile at temperatures up to 1000 °C15 and cannot coalesce when separated. However, sintering occurs under low density conditions by detachment of individual atoms from particles, migration across the oxide surface between particles, and capture of the atoms by neighboring particles. In general, large particles get bigger at the expense of smaller particles, and this is simply a reflection of the higher effective 2D vapor pressure in the vicinity of a small particle, due to their lower average surface coordination. Evidence for this is the loss of particles at the small end of the size distribution (Figure 7), between 500 and 900 °C. It is important to stress, though, that these films appear to be reasonably stable over a significant temperature range. Sintering, in terms of loss of particle density, is relatively low for the unstructured film between 500 and 900 °C, presumably because the average separation between particles is fairly large (∼400 nm). Thus, for thermally stable heterogeneous catalysts, although we need to aim for a high density of active sites, we need, counterintuitively, to make sure that this density is not too high (that is, interparticle separation is not too small) or they will be unstable. It appears that the Tamman temperature is an important threshold for the changes in the film; it is considered to represent the approximate temperature at which significant mobility of the bulk atoms begins to occur. The Tamman temperature is approximately 1/3 of the melting point, and for gold is ∼350 °C, and this temperature is between those for which significant changes in the gold film occurs. Below this temperature, there is very little change in the film, whereas above it sintering and particle formation occur. In terms of engineered films, the ones shown here are really rather remarkably thermally stable, and no significant changes occur even up to 300 -500 °C. This is probably a particular property of Au films because other metals (except perhaps Pt) will tend to oxidize when calcined in air at these temperatures. However, in the temperature range examined in this study, the films become noncontiguous and so material properties, such as conductivity and optical blocking, will change severely, but also particle size and surface atom concentration, both of great relevance to catalysis.
Conclusions As stated in the Introduction, this work attempted to answer a number of questions, such as: (i) will restriction of the size of gold domains affect the sintering process; (ii) will single nanoparticles form in the area of a single cell of evaporated gold; (iii) will such films simply “forget” the ordering of the nanostructures imposed upon them at elevated temperature? It is clear that restricting the particle domains by nanostructuring “moats” of alumina around them does indeed restrict the sintering, something which is also then reflected in a higher evaporation rate from those particles when heated to 1100 °C. It is evident that the location of particles is not random but is dictated by the location of the prefabricated structure (for example, Figure 7 at 900 and 1100 °C). The work in this paper represents a way of fabricating ordered arrays of thermally stable nanoparticles anchored to surfaces. Acknowledgment. We are grateful for the support of a number of bodies during this work, including the Higher Education Funding Council for Wales (HEFCW), Cardiff University and the EPSRC for partial funding of a studentship to J.C.
Langmuir 2010, 26(21), 16261–16266
