Since the discovery that gold nanoparticles (NPs) on TiO₂ are exceptionally active for low-temperature oxidation of CO, great efforts have been made to explore the chemical nature of the catalytically active sites on Au/TiO₂ catalysts, which largely depends on the size of gold particles [1-8]. It is generally acknowledged that the active sites locate at the gold-support interface, based on the experimental observation that the activity is often proportional to the total length of the interfacial perimeter [9-13]. The reaction occurs at the Au-oxide interfacial perimeter: CO is adsorbed and activated on the gold nanoparticle while molecular oxygen is activated by the perimeter of Au-oxide interface [8-11]. With this respect, both the size of gold particles and the chemical properties of TiO₂ surfaces played critical roles in determining the Au-TiO₂ interfacial structure and consequently the catalytic activity, involving both electronic and geometric interactions between Au particles and TiO₂ surfaces.

The commercial P25, a mixture of anatase- and rutile-TiO₂, has been commonly used to support gold nanoparticles for examining the size effect of the metal particles [5, 14]. However, the diversity of TiO₂ surfaces makes it difficult to straightforwardly identify the atomic structure of the interfacial perimeters. Tailoring the morphology of TiO₂ provides a possibility to anchor gold nanoparticles on a specific facet of the oxide-support, forming relatively uniform Au-TiO₂ interfaces [15-17]. For example, TiO₂{001} favoured an intimate contact and a stronger interaction with gold particles, as compared with TiO₂{101}. While the intimate contact of Au-TiO₂{001} improved the electron transfer from Au NPs to the adsorbed O₂, thus being in favour of oxygen dissociation and promoting the catalytic activity in CO oxidation [18, 19]. Recently, the stronger bonding of Au nanoparticles on TiO₂{001} than that on TiO₂{101} has been directly visualized by in situ transmission electron microscopy (TEM) [20]. Au NPs of 4–8 nm, immobilized on TiO₂{101}, were sintered to large particles through the Ostwald ripening and particle migration coalescence as being heated to 773 K in the presence of oxygen, but no sintering was observed on the anatase-TiO₂{001} under the identical conditions. It was attributed to the much higher adsorption energy of gold on TiO₂{001} than that on TiO₂{101} [20].

In this work, anatase TiO₂-nanosheets and nanospindles were used to support Au particles of ~2.2 nm, and the resulting Au/TiO₂{001} and Au/TiO₂{101} interfacial perimeters were examined for CO oxidation. It was found that the Au/TiO₂{001} interface is more active for CO oxidation than the Au/TiO₂{101} system under identical reaction conditions.

The anatase TiO₂ nanosheets (TiO₂-P) and nanospindles (TiO₂-S) were synthesized by a hydrothermal method, as we previously described [21-23]. Au colloids of about 2.2 nm (Fig. S1) were prepared by reduction of HAuCl₄ with NaBH₄ in the presence of polyvinyl alcohol (PVA), following a typical procedure [24]. The Au/TiO₂ catalysts were then prepared by immobilizing Au colloids on TiO₂. The oxide-support was mixed with the Au colloidal solution under vigorous stirring, and the pH of the mixture was adjusted to ca. 6.0 by adding 0.1 M HNO₃ aqueous solution. After stirring for 1 h at room temperature, the solid was collected by filtration and washed thoroughly with distilled water. The sample was dried under vacuum at room temperature overnight and calcined at 623 K for 4 h in air. ICP analysis showed the actual loading of Au was 0.80 wt% for Au/TiO₂-nanospindle (Au/TiO₂-P) and 0.81 wt% for Au/TiO₂-nanosheet (Au/TiO₂-S).

X-ray powder diffraction (XRD) patterns were recorded on a Rigaku D/MAX-2500PC with Cu K_α radiation (λ = 1.5418 Å) at 40 kV, 200 mA. TEM images were recorded on a Hitachi 7700 microscope operated at 120 kV. Aberration-corrected STEM (ac-STEM) images were taken on a JEM-ARM200F at 200 kV. The specimen was prepared by ultrasonically dispersing the sample powder into water or ethanol, and the drops of the suspension were deposited on a carbon-coated copper grid and dried at room temperature in air. The lattice spacing and the morphology of Au nanoparticles was analyzed by a Digital Micrograph software. X-ray photoelectron spectra (XPS) were recorded with an ESCALAB 250 Xi spectrometer (Thermofisher) using an Al K_α radiation source operated at an accelerating voltage of 15 kV. The charge effect was corrected by adjusting the binding energy of C 1s to 284.6 eV.

CO oxidation over the Au/TiO₂ catalysts was performed in a U-shape quartz tubular reactor (inner diameter 6 mm) under atmospheric pressure. 150 mg catalyst was loaded between two layers of quartz wool and pretreated with a 20.0 vol% O₂/Ar mixture at 623 K for 4 h. The feed gas of 1.0 vol% CO/20.0 vol% O₂/He (50 mL min^–1) was introduced through a mass-flow controller, and the temperature was gradually increased from 198 to 473 K. The effluent from the reactor was analyzed by an online gas chromatogram (Agilent-GC7890B). The reaction rate was measured at the temperature range of 263–293 K and the conversion of CO was controlled to be 8%–15% by varying the gas flow rate for assuring a differential reactor condition.

XRD patterns of the Au/TiO₂-P and Au/TiO₂-S samples show prominent diffraction lines at 25°, 38°, 48°, 54°, and 55° (Fig. S4), which are assigned to the anatase-TiO₂ facets of (101), (103), (004), (112), and (200) (JCPDS # 21-1272). There were no diffraction lines of Au particles, primarily because of the well dispersion of particles in small sizes or the lower Au loading (ca. 0.8 wt%). Fig. 1 shows XPS of Au 4f in the Au/TiO₂ catalysts. The binding energies (BEs) of Au 4f_7/2 and 4f_5/2 at 83.1 and 86.8 eV indicated the presence of metallic gold, which are the typical values (~83.2 eV of Au 4f_7/2 and ~86.9 eV of Au 4f_5/2) for gold nanoparticles dispersed TiO₂ [25-30]. The slightly lower binding energies as compared with bulk gold might be caused by the size effect gold nanoparticles and the possible electron transfer from gold to TiO₂ in the Au/TiO₂ catalysts.

The spatial and size distributions of TiO₂ nanosheets and nanospindles are analyzed by TEM. As we previously reported [22], TiO₂-P exhibits a spindle shape with an average size of 50 nm (length) and 20 nm (width) and is enclosed by TiO₂{101} (81%) on the isosceles trapezoidal surface and TiO₂{001} (19%) on the top surface. TiO₂-S has a sheet-like morphology with an average particle size of 75 nm length and 7 nm thickness, and predominantly expose TiO₂{001} (84%) and TiO₂{101} (16%) facets. On TiO₂-P, gold NPs are preferentially anchored on the TiO₂{101}. The majority of the Au NPs on TiO₂{101} were in the size range of 2.6–4.2 nm, with a mean diameter of 3.1 nm and a typical height of 2.63 nm (ca. 12 layers of Au atoms, Fig. 2). STEM analysis on the atomic structures of Au NPs and the metal-support interfaces found that the Au NPs adopt a morphology of quasi-truncated octahedron, which is consistent with our previous work [24]. The prominent Au{100} and Au{111} are clearly identified by the lattice spacing of 0.21 and 0.23 nm on the gold crystallites (Fig. 2).

On TiO₂-S, the distribution of gold NPs showed a dual-model pattern; 76% metal particles located on TiO₂{001} while 24% gold particles deposited to TiO₂{101}, based on the statistics for 556 particles in the sample. The Au NPs on TiO₂{101} had a mean size of 3.2 nm (Fig. 3), similar to the situation of the Au/TiO₂-P sample. On TiO₂{001}, the majority of the Au NPs on TiO₂{001} were in the size range of 1.7–2.7 nm, with a mean diameter of 2.2 nm (Fig. 3), which is the same as the starting Au colloids. It is most likely that TiO₂{001} bonded Au NPs more tightly than TiO₂{101}. That is, the strong metal-support interaction stabilized the 2.2 nm Au NPs on TiO₂{001} and retarded the aggregation of the small gold particles under the thermal treatment (623 K in air).

Based on the STEM observations, the atomic configuration of Au particles and the Au-TiO₂ interfacial structures were quantitatively analyzed (Table 1). For Au/TiO₂{101}, the total number of Au atoms per gold particle is 1228, in which 108 atoms are low-coordinated at the edges and corners. While on Au/TiO₂{001}, the gold NPs contain 549 atoms and 78 atoms are low-coordinated. Particularly, the numbers of the gold atom at the interfacial perimeters, on the basis of a single particle, are 27 for Au/TiO₂{101} and 21 for Au/TiO₂{001}.

Fig. 4(a) shows the catalytic activities of the Au/TiO₂ catalysts for CO reaction in the temperature range 198-473 K. The conversion of CO over Au/TiO₂-S was readily 15% at 243 K, and sharply approached to 100% at 280 K. On the Au/TiO₂-P catalyst, the conversion of CO was 14% at 243 K while the full conversion of CO (99%) was achieved at 468 K. This clearly demonstrates that the superior activity of the Au/TiO₂-S to the Au/TiO₂-P. TEM analysis on the spent catalysts confirmed that the size and shape of gold nanoparticles kept almost unchanged (Figs. S8 and S9).

The reaction kinetics of CO oxidation over the Au/TiO₂ catalysts were then investigated by controlling the conversion of CO at 8%–15% (Fig. 4(b)). The apparent activation energy of CO oxidation over the Au/TiO₂-P and Au/TiO₂-S was 30.0 and 28.2 kJ mol^–1, respectively, in the temperature range of 263-293 K. They are similar to the general values for CO oxidation over gold nanocatalysts at this temperature range [31-33]. At room temperature (298 K), the reaction rate on Au/TiO₂-P was 0.17 mmol_CO g_Au^–1 s^–1, while it was 0.41 mmol_CO g_Au^–1 s^–1 over Au/TiO₂-S (Table 2). By taking the size distributions of gold particles on TiO₂{001} and TiO₂{101} in the Au/TiO₂ catalysts into account, the reaction rate was estimated to be 0.17 mmol_CO g_Au^–1 s^–1 on Au/TiO₂{101} and 0.46 mmol_CO g_Au^–1 s^–1 on Au/TiO₂{001}. It indicates that the small sized Au NPs (2.2 nm) on the TiO₂{001} exhibit the higher catalytic activity than the corresponding large sized Au NPs on the TiO₂{101}, consistence with the previous results [34].

Since the interfacial perimeter is associated with the active sites in CO oxidation over Au/TiO₂, the reaction rates further were normalized with the total number of the perimeter Au atoms, estimated according to the mean size of the Au NPs and the Au-TiO₂ interfacial models as determined by TEM/STEM analysis. For the Au-TiO₂-P catalyst, Au-TiO₂{101} is the only interface, and the total number of perimeter gold atoms was 5.38 × 10¹⁷ per gram. For the Au-TiO₂-S sample, both Au-TiO₂{001} and Au-TiO₂{101} interfaces are presented and their proportions were calculated to be 7:3. Accordingly, their perimeter gold atoms were 7.17 × 10¹⁷ and 2.96 × 10¹⁷ per gram, respectively. The turnover frequency (TOF) on the perimeter Au atoms was determined to be 1.52 s^–1 for Au/TiO₂{101} and 2.34 s^–1 for Au/TiO₂{001}. This result evidences that the Au NPs (average 2.2 nm) on TiO₂{001} are intrinsically more active than these (3.1 nm) on TiO₂{101}, which is mainly because of the unique atomic arrangement of TiO₂{001} that bonded gold particles more strongly.

The gold-support interaction in Au/TiO₂ is governed by the surface orientation of TiO₂. TiO₂{001} had a stronger interaction with Au NPs, and thus avoiding the aggregation of Au NPs (2.2 nm) during calcination at 623 K in air. The Au NPs dispersed on TiO₂{001} exhibited a much higher activity than the particles on TiO₂{101} in CO oxidation, primarily because of intrinsic coordination patterns of the interfacial gold atoms on TiO₂ facets.