Introduction

By development of environmental protection regulations, less CO emission is demanded. Most of the CO emission occurs when the engine is still cold [1]. Gold is the most active catalyst for CO oxidation at room temperature which makes Au catalysts very interesting for this application [2,3,4,5,6]. However, when the engine becomes hot, it reaches the temperatures as high as 700 °C, and high-temperature treatment often causes growth of Au nanoparticles [7,8,9,10,11,12,13,14]. For bulk Au, the melting point is 1064 °C, and the Hüttig temperature at which surface atoms become mobile is 319 °C. For Au nanoparticles, the melting point is even lower [15]. Hence, one can expect the growth of Au nanoparticles at elevated temperatures. Deactivation due to particle growth is irreversible and detrimental to the long-term activity of the catalyst. Different state-of-the-art strategies [16,17,18,19] have been employed to combat the particle growth. For instance, a high thermal stability of Au nanoparticles on combinatory TiO2/SiO2 support [20], hydroxyapatite/TiO2 support [21], nanorods of TiO2 [22], and perovskite support [23] has been reported.

Supports stabilize Au nanoparticles by enhancing inter-particle spacing and sometimes by metal support interaction [24, 25]. For instance, Au nanoparticles on Al2O3 have shown excellent stability upon treatment at 650 °C under oxidizing conditions [2, 26, 27]. However, Au/TiO2 catalyst, which is the most studied type of Au catalyst, is known to grow during CO oxidation [28,29,30] as well as during high-temperature treatment under oxidizing atmosphere [7], though the exact conditions that cause the growth are under discussion [29, 30]. Au/TiO2 catalysts are known to deteriorate during storage, as they are light and moisture sensitive [31, 32]. The stability of Au/SiO2 is under debate [33,34,35,36,37,38,39,40]. Though there are some examples of stability and/or instability of supported Au nanoparticles, the influence of the nature of the support and reaction conditions on the stability is not well understood.

More importantly, the mechanisms that are involved in Au particle growth under different conditions are unclear. Particle growth can take place via two major mechanisms [41]. The first one involves nanoparticle diffusion over the support surface and coalescence to form larger particles (particle diffusion and coalescence). The diffusion is more likely for smaller particles, and it is accelerated when the temperature increases, roughly following Arrhenius behavior [42]. Near the Tamman temperature, at which the metal atoms acquire sufficient energy for their bulk mobility, particle diffusion becomes more dominant.

Secondly, larger particles grow at the expense of smaller particles (Ostwald ripening) [11, 41]. The driving force is that the higher surface energy of low coordinated metal atoms at the surface of small particles destabilizes the small particles and makes the larger particles, compared with smaller particles, energetically more favorable. Hence, metal species detach from small particles, diffuse over the support or through the vapor phase, and can attach to larger particles with a lower chemical potential. This leads to the growth of larger particles at the expense of smaller particles.

According to fundamental studies by Wynblatt and Gjostein [42], Ostwald ripening can have two possible rate-limiting steps: detachment of metal atom from small particles in form of mobile species, the so-called interface-controlled regime, and diffusion of mobile species from smaller particles to the larger one, the so-called diffusion-controlled regime. Mobile species can be metal atoms or metal ions of metal-containing molecules. In the latter, the metal or metal ions are complexed and stabilized by surface groups and/or species from the gas phase that act as ligands. Ligands decrease the activation energy for the detachment of a metal (ion) from a nanoparticle. Hence, the formation of mobile species depends on the presence of reactive gases, certain surface groups, and/or oxidizing or reducing conditions [25]. For example, fast growth of Pt/Al2O3 under O2 atmosphere was attributed to the formation of volatile PtO2 species under oxidizing atmosphere as well as a high diffusivity of PtO2 over the support surface [43]. The growth of Ni nanoparticles during the methanation reaction was ascribed to the formation of gas-phase Ni(CO)5 as mobile species [44]. The most likely mobile species in different reactive conditions can be predicted by calculating thermodynamics and energy barriers for the diffusion of the different possible complexes. For instance, based on DFT calculations, CuCO species were assumed to be the mobile species in Cu/ZnO catalysts under methanol synthesis conditions and during water gas shift reaction [45].

To the best of our knowledge, there is no systematic study on the growth of Au nanoparticles on different supports that are prepared with the same method and have similar composition, support porosity, and initial Au particle size. Catalyst preparation method affects the stability of supported nanoparticles by influencing catalyst properties such as distribution of Au particles over the support and the concentration of contaminants like Cl that can promote particle growth [46]. A uniform spatial distribution of particles, with maximum inter-particle distances of supported nanoparticles, and a very narrow particle size distribution play an important role in minimizing particle growth [44, 47].

In this study, we report on the growth of Au nanoparticles on supports with different properties using non-reducible ones (SiO2 and γ-alumina) and a reducible one (TiO2). We employed a modified incipient wetness impregnation method to prepare 3 to 4 nm Au nanoparticles on TiO2, Al2O3, and SiO2 supports [48]. The effect of different reactive gases on the particle growth was investigated. The activation energies of particle growth for Au on different supports and in different reactive gases were experimentally obtained as well.

Experimental

Sample preparation

Commercially available supports, TiO2 (rutile, BET surface area of 30 m2 g−1, pore volume of 0.12 mL g−1, Sigma-Aldrich), SiO2 (Aerosil, BET surface area of 50 m2 g−1, pore volume of 0.12 mL g−1, Evonik), and Al2O3 (gamma phase, BET surface area of 120 m2 g−1, pore volume of 0.46 mL g−1, Alfa-Aesar) were used. Not the most commonly used P25, TiO2 with mixed phases of anatase and rutile, but instead pure rutile was used because at 500 °C and above, anatase phase transforms to rutile phase [49], and this change in support phase could contribute to particle growth.

Gold was deposited on the supports by a modified incipient wetness impregnation method developed by Delannoy et al. [48]. In a typical preparation, the support (1 g) was dried under vacuum at 200 °C and was impregnated with an aqueous Au solution (appropriate concentration of HAuCl4·3H2O, Sigma-Aldrich) to prepare 1 wt% Au on TiO2 and SiO2 and 4 wt% Au on Al2O3. The sample was aged at room temperature under vacuum for 1 h and then washed twice with ammonia solution (30 mL each time, 1 M) to remove Cl and twice with water (30 mL each time) at RT. Each time, the solid was recovered by centrifugation. The Au/SiO2 sample was washed with diluted ammonia solution at lower pH (pH = 8) to remove Cl and to avoid dissolution of SiO2 as well. The sample was then dried under vacuum at room temperature for 48 h or was dried in a freeze drier at − 20 °C under 0.1 mbar vacuum for 17 h. The dried samples were further treated in air (100 mL min−1) from RT to 300 °C (ramp 2 °C min−1) and kept at 300 °C for 4 h before cooling down and stored in a desiccator in the dark.

Another sample of Au/TiO2 was prepared with the method of deposition-precipitation with urea [46] to promote preparation of a Cl-free sample. In a typical preparation, the support (1 g, rutile, Sigma-Aldrich) was dispersed in water (100 mL) in a 500-mL polyethylene bottle, and the reaction mixture was stabilized in an oil bath at 80 °C. An aqueous Au solution (2 mL, 0.027 M, HAuCl4·3H2O precursor, Sigma-Aldrich) and urea (300 mg) were added. The reaction mixture was stirred for 16 h in a closed container at 80 °C in the dark. Then, the solid was recovered by centrifugation and was washed eight times with water (40 mL each time) at RT. The sample was then dried in a freeze drier at − 20 °C under 0.1 mbar vacuum for 17 h. The dried samples were further reduced in H2 (100 mL min−1) from RT to 300 °C (ramp 2 °C min−1) and kept at 300 °C for 2 h before cooling down and stored in a desiccator in the dark.

Characterization

Elemental analysis was performed on an inductively coupled plasma-mass spectrometer (Mikroanalytisches Laboratorium Kolbe, Germany) after destruction of the samples at high temperature and pressure. Chloride content of the samples was determined by ion chromatography (Mikroanalytisches Laboratorium Kolbe). Transmission electron microscopy (TEM) imaging was performed on a Tecnai 12 (FEI) microscope operated at 120 kV. Particle sizes were determined from the micrographs as ∑nidi/∑ni, where di is the diameter of typically 200–300 individual particles on different areas of the sample. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was performed on a Talos F200X microscope operated at 200 kV. STEM image processing was carried out using Tecnai Imaging Analysis (TIA).

Thermal treatment

The thermal treatment of the samples was performed either under static condition or in a gas flow. The experiments under static condition were performed in a temperature-calibrated muffle furnace. Typically, 40 mg of the sample was heated in air from RT to 500, 600, or 700 °C (ramp 5 °C min−1) and kept at the specified temperature for 4 h. The experiments in flows of different gases, compressed air (wet air), synthetic air (dry air), H2, N2, or N2 that was bubbled through water at RT (100 mL min−1), were performed on 200 mg of the sample in a plug flow reactor with diameter of 2 cm at 500 °C with the same heating program. The experiments in flow of H2 at 600 and 700 °C were done in a crucible with a horizontal tubular furnace under the flow of 25 mL min−1 to imitate the static condition in air. The experiments and analytical measurements were performed at least twice to ensure reproducibility of the results.

Activation energy of particle growth was calculated from the Arrhenius plots (ln (particle growth rate) versus 1/T), where particle growth rate is estimated from the changes in particle sizes upon 4 h of thermal treatment at different temperatures (T).

Results

Sample characteristics

Figure 1 shows transmission electron micrographs of Au nanoparticles supported on three different supports. Particle sizes for these prepared samples were between 2.5 and 4.0 nm (Table 1). Determination of particle sizes relied on TEM micrographs, as crystallite sizes could not be derived from X-ray diffraction patterns owing to low Au loading and overlap of diffraction patterns. With bright-field TEM, particle sizes were difficult to determine for the Au/TiO2 sample due to a low contrast. Therefore, in this case, additional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was employed. HAADF-STEM micrographs (Figure S.1) showed particle size of 2.6 ± 0.6 nm for the Au/TiO2 which confirms the accuracy of the particle size obtained by bright-field TEM.

Fig. 1
figure 1

Transmission electron micrographs for the 0.9 wt% Au/TiO2 (a), 1.0 wt% Au/SiO2 (b), and 4.0 wt% Au/Al2O3 (c)

Full size image
Table 1 Structural properties of the samples
Full size table

Table 1 presents the structural properties of the samples. The Au loading for each sample was chosen based on the BET surface areas of the corresponding supports aiming for similar inter-particle distances for all samples. For instance, the surface area of the commercially available γ-Al2O3 is four times higher than that of TiO2; hence, a four times higher Au loading for the Au/Al2O3 was targeted. The average inter-particle distances were between 30 and 50 nm for all the supported Au nanoparticles. Elemental analysis showed that targeted Au loadings were achieved and the Cl content of all samples is between 0.3 and 0.4 wt%. For comparison, an Au/TiO2 sample with negligible Cl content was prepared by the method of deposition-precipitation with urea. Hence, all the supported Au samples had similar properties, though the as-prepared Au nanoparticles on TiO2 had slightly smaller sizes than the two other samples.

Impact of the support on the thermal stability

Figure 2 shows TEM micrographs of the Au/TiO2 sample upon treatment to temperatures up 700 °C in air. Additionally, Figure S.2 shows the ones for Au/SiO2 and Au/Al2O3. These figures demonstrated that all supported Au nanoparticles have grown upon thermal treatment. However, growth for Au nanoparticles on TiO2 was more pronounced (Fig. 2).

Fig. 2
figure 2

Transmission electron micrographs of 0.9 wt% Au/TiO2 that has been treated, from left to right, upon treatment at 500, 600, and 700 °C in static air for 4 h

Full size image

Figure 3 shows, in summary, the evolution of particle size distributions and shows that the particles grew with increasing temperature. The particles for the Au/TiO2 sample were initially smaller than those for the Au/SiO2 and Au/Al2O3 samples, but they grew more upon thermal treatments: 2.1-fold increase in diameter at 500 °C, 2.4-fold at 600 °C, and 5.0-fold at 700 °C. Furthermore, the particle size distribution was broadening. The Au nanoparticles on SiO2 and Al2O3 grew too, but to a limited extent: for the Au/SiO2: no significant growth at 500 °C, 1.1-fold at 600 °C, and 1.8-fold at 700 °C. The particle size for the Au/TiO2-DPU that contained no Cl increased too: 1.7-fold at 500 °C, 2.4-fold at 600 °C, and 4.3-fold at 700 °C. This particle growth was less than for the Au/TiO2 with some Cl, in agreement with the literature that suggests Cl residue enhances growth of Au nanoparticles on TiO2 [50]. Nevertheless, Au nanoparticles on TiO2, even with negligible Cl content, were thermally less stable than the ones on SiO2 and Al2O3 in air.

Fig. 3
figure 3

Evolution of particle size distributions and log normal fits for the as-prepared Au on TiO2, SiO2, and Al2O3 as well as for the samples that have been treated at 500, 600, and 700 °C in static air for 4 h

Full size image

Effect of atmosphere

Figure 4 compares the thermal stability of Au nanoparticles on TiO2 and SiO2 under different atmospheres, varying from reducing (a flow of H2) via inert (a flow of N2) towards oxidizing (a flow of wet air). First, for the Au/TiO2, the particle growth was negligible in either H2 or N2, but it became more pronounced in the presence of water vapor (~ 2 mol% water in N2, 1.6-fold increase) and in the presence of dry air (3.1-fold increase). It means that presence of O2 and/or water accelerated the thermal growth of Au nanoparticles on TiO2 supports. In contrast, no significant difference in particle growth, and in both cases a high thermal stability, was observed either in inert atmosphere or in the presence of H2. This observation excluded strong metal support interaction between Au and the reducible TiO2 support, which typically occurs under reducing conditions, as a reason for the high stability of Au nanoparticles on TiO2 in the flow of H2. Hence, the particle growth occurred for the Au/TiO2 only in the presence of O2 and/or water.

Fig. 4
figure 4

Effect of gas composition on the size of Au nanoparticles on TiO2 or SiO2 that have been treated at 500 °C in different flows of gases (100 mL min−1) for 4 h

Full size image

In contrast, Fig. 4 shows that for the Au/SiO2, the particle growth was negligible under all atmospheres tested. In summary, the growth of Au nanoparticles was influenced by the atmosphere only when they were supported on the TiO2, but not when they were supported on SiO2, and particle growth was more pronounced in an oxidizing atmosphere.

Figure 5 shows the Arrhenius plots for rates of Au nanoparticles growth (dR/dt, where R is the particle diameter and t is time) on the TiO2 and SiO2 under different atmospheres. The dR/dt was extracted from plots in Figure S.3. The activation energies of particle growth (Ea) estimated from the plots are given in Fig. 5 as well. In air, Ea for Au/TiO2 (36 ± 10 kJ mol−1) was lower than the one for Au/SiO2 (86 ± 14 kJ mol−1). However, in H2, the Ea was similar for Au nanoparticles on both TiO2 and SiO2 (80 ± 6 vs 81 ± 13 kJ mol−1). Notably, Ea for the Au/SiO2 under different atmospheres was also similar. This suggests that the rate-limiting step in particle growth for the Au/TiO2 and Au/SiO2 in H2 and for the Au/SiO2 in air might be similar whereas rate of Au nanoparticle growth on TiO2 is apparently not or much less limited by this step in air. This will be discussed in more detail in the “Discussion” section.

Fig. 5
figure 5

The Arrhenius plots of the rates of Au nanoparticles growth on TiO2 and SiO2 in static air and in a flow of H2 (25 mL min−1, in a tube furnace) after 4 h. R is particle size and t is time

Full size image

Discussion

There are a few reports on the growth of Au nanoparticles supported on TiO2 upon high-temperature treatment and under oxidizing atmosphere [7]. Akita et al. reported that Au nanoparticles of 2.1 nm on TiO2 grew to 9.7-nm particles upon calcination in air at 600 °C for 4 h [32]. A high thermal stability of Au nanoparticles on Al2O3 upon treatment at 650 °C in three-way catalysis conditions [2, 27] and in a flow of O2 [26] was previously reported. There is no consensus on the thermal stability of Au nanoparticles on SiO2. Some reports suggested that they grow [33,34,35,36,37,38]. For instance, Bore et al. [37] showed that Au nanoparticles of around 1 nm on mesoporous SiO2 grew to in average 6 nm upon H2 exposure at 200 °C. In contrast, some other reports claimed that if the Au/SiO2 contains little or no Cl, the Au nanoparticles are actually stable upon O2 exposure at 500 °C [39, 40]. In the latter case, the high stability of Au nanoparticles on SiO2 was attributed to a strong bond between Au and defects on the surface of SiO2. It was proposed that the defects on the surface were made during deposition of Au on the SiO2 by magnetron sputtering [40]. Barret et al. showed as well that enhanced defect sites on the TiO2 surface make the supported Au nanoparticles thermally stable [22]. All mentioned reports are examples obtained under catalysis conditions on powder catalysts. On 2D model systems in vacuum, the interaction of Au with reducible supports like TiO2 is suggested to provide stable Au nanoparticles [28, 51]. Accordingly, Au on non-reducible Al2O3 and SiO2 supports is suggested to be less stable [16]. Our study clearly proves a higher thermal stability of Au nanoparticles on the non-reducible supports SiO2 and Al2O3 than on TiO2.

It is known that gas atmosphere can affect the growth of metal nanoparticles. Oxidizing atmospheres typically induce faster rates of particle growth than reducing atmospheres [43, 52]. Particularly for Au/TiO2, it was reported that Au particle sizes are smaller if the sample is prepared in H2 or Ar than when it is prepared in O2 [31]. However, to the best of our knowledge, this is the first report that shows that the situation is different for different supports and that the Au, regardless of the atmosphere, is thermally stable on the SiO2 support.

This is also the first time that the activation energies of particle growth under conditions closer to catalytic reaction conditions for Au on different supports and in different reactive gases are obtained. For comparison, the activation energy of particle growth for Ni nanoparticle catalysts under steam reforming condition was estimated to be 46 ± 8 kJ mol−1 [53]. The activation energy of particle growth for Au particles supported on single-crystal TiO2 (110) under ultra-high vacuum was estimated as high as 280 kJ mol−1 [54, 55]. Notably, this higher value was obtained under UHV conditions rather than relevant catalytic conditions and on atomically flat clean surfaces rather than powder supports with surface groups and adsorbed molecules [54]. Hence, one can conclude that under conditions closer to the catalytic reaction conditions, the activation energy of particle growth is much lower and often particle growth is much faster than for 2D model systems in vacuum.

As mentioned in the “Introduction” section, particle growth can take place via two mechanisms: particle diffusion and coalescence or Ostwald ripening. Our results suggest three main conclusions on involved particle growth mechanisms: First, particle diffusion and coalescence as a major particle growth mechanism is unlikely. The distance between the particles is estimated to be around 50 nm for all the samples. Our previous reports on the growth of Cu nanoparticles during methanol synthesis showed a very strong dependence of particle diffusion and coalescence on inter-particle size with diffusion dominating at such large distances unlikely [47, 56]. Furthermore, DFT calculations showed that Au nanoparticles interact much stronger with TiO2 surfaces than with SiO2 surfaces [40]. If high-temperature treatments do not induce particle diffusion and coalescence for the Au nanoparticles on SiO2, it is unlikely that particle diffusion and coalescence take place for the TiO2 surfaces under similar conditions. Therefore, particle growth by Ostwald ripening most likely is the main particle growth mechanism at least under oxidizing atmospheres. It has been suggested that particle growth by Ostwald ripening is also the main growth mechanism for the Au on single-crystal TiO2 (110) under ultra-high vacuum upon a high-temperature treatment [28].

Identifying the mobile species is not trivial [57]. Because of the large influence of the chemical nature of the support on particle growth observed in this study, a second conclusion would be that the mobile species are at the support surface rather than in gas phase. At temperatures of 177 °C and above, Au does not form any stable oxides [58] or volatile compounds like Pt does in the form of PtO2. The energy required to remove a metal atom from a metal particle is high, close to the heat of sublimation of metals [53], which is 368 kJ mol−1 for Au [54, 55]. The fact that the experimentally observed Ea for growth of Au nanoparticles are an order of magnitude lower leads to a third conclusion that Au is unlikely to detach from a nanoparticle in metallic form. Furthermore, using the Kelvin equation, saturation pressures for Au nanoparticles with different sizes at different temperatures were calculated (supporting information, estimation of the diffusion of atomic Au in the gas phase at different temperatures). From the difference in the saturation pressures of the Au vapor around the small and large Au nanoparticles, the diffusion rate for atomic Au in the gas phase was estimated. The calculation showed that Ostwald ripening via the atomic Au in the gas phase is not a main factor in explaining the observed particle growth. The observation that the formation and/or diffusivity of mobile species is accelerated on TiO2 and under oxidizing atmosphere as well as in the presence of Cl− ions strongly suggests adsorbed complexed and oxidized Au species being the mobile species.

It is remarkable that the Ea for particle growth in air on Au/TiO2 is smaller than that on Au/SiO2. This means that the formation and/or the diffusivity of the Au-complex mobile species is less favorable on SiO2. It is known, for instance, from CO oxidation studies [59,60,61,62,63,64,65,66,67,68,69], that Au and TiO2 have a very specific interaction, which leads to very active and possibly cationic gold species at the interface between the two, which might also play a role in diminishing the thermal stability of the TiO2-supported particles. It is remarkable that the Ea under other conditions, whether on SiO2 in either reducing or oxidizing conditions, or on TiO2 under reducing conditions, are all similar and in the range of 80–86 kJ/mol. This suggests that in the absence of this specific Au-TiO2 interaction in oxidizing atmosphere, particle growth follows a similar mechanism. Tentatively, we propose that under oxidizing conditions cationic Au species are mobile species that are stabilized on TiO2 to a larger extent that on SiO2 in line with the lower activation energy for growth on the former (Fig. 5). In any case, the fact that Au catalysts on non-reducible supports show such high thermal stability, even though they often show lower activities than on reducible supports like TiO2 [70], is a very relevant consideration for the potential application of Au nanoparticles in gas-phase catalysis.

Conclusion

For the first time, the thermal growth of Au nanoparticles on different oxidic supports and under different reactive atmospheres was studied. Similar initial particle sizes, in the 2.5–4-nm range, allowed direct comparison for the different supports. All supported Au nanoparticles grew upon thermal treatment. Particle growth on TiO2 was much more pronounced than that on either SiO2 or Al2O3. Particle growth on TiO2 was particularly enhanced by an oxidizing atmosphere, the presence of water, and/or the presence of Cl−. Particle growth by Ostwald ripening involving cationic gold species complexed by ligands was the most likely dominating growth mechanism. On non-reducible supports and in non-oxidizing atmosphere, the supported Au nanoparticles were remarkably stable. These results provide a better understanding of the growth of supported Au nanoparticles, and tools for a rational choice of a support for high-temperature gas-phase reactions involving gold catalysts.