2.1.

Au Cluster Attachment

To control the size of ONF sources, we used gold clusters. To arrange the gold clusters in the vicinity of the metal complex, first, the gold clusters were deposited on an alumina sphere, and then the metal complex was deposited on the alumina sphere (see Sec. 2.2). We used the alumina sphere [AO-502 (Admatechs)] with a median particle diameter in the range 200 to 300 nm. Glutathione-protected ${\mathrm{Au}}_{25}$ $[{\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}]$ was used as the precursor for the gold clusters.¹⁸ This ${\mathrm{Au}}_{25}$ was attached to the alumina sphere after stirring in water in a manner similar to that described in previous studies.^18–20 In particular, ${\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}$ clusters were adsorbed on the alumina sphere by mixing an aqueous solution containing ${\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}$ with an aqueous solution of alumina sphere (600 mg) for 2 h at room temperature ($\sim 20°\mathrm{C}$). The total volume of the aqueous solution was fixed at 200 mL, and the mixing ratio of the ${\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}$ clusters to the alumina sphere was fixed at 0.05, 0.1, or 0.2 wt. % Au. Because glutathione was simply used as a linker between alumina and gold clusters, it was eliminated by firing at 300°C for 2 h. Here, we expected to observe the ONF effect without the discrete electric structures and plasmon effect. We selected ${\mathrm{Au}}_{25}$ because the number of gold atoms is sufficiently small to ensure there is no plasmonic resonant peak, which is observed when the number of Au atoms exceeds 144.²¹ In addition, the attached clusters were fired so that the discrete electric structures disappeared.²¹

To confirm the attachment of gold clusters on the alumina sphere, we compared the absorbance of gold clusters between before and after attachment on the alumina sphere [Fig. 1(a)]. A comparison of the UV-visible spectra of gold clusters [black solid curve in Fig. 1(a) taken before attachment] with the diffuse reflectance (DR) spectra of gold clusters taken after attachment on alumina sphere [before firing, blue solid curve in Fig. 1(a)] showed that the discrete peaks disappeared after firing in the DR spectra [red solid curve in Fig. 1(a)]. This indicates that the atomicity was destroyed after firing.

Fig. 1

(a) Comparison of absorbance. Black (before attachment): UV-visible spectra of gold clusters $[{\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}]$ only, blue (after attachment): DR spectra of gold clusters attached on alumina sphere, and red (after attachment): DR spectra of gold clusters attached on alumina sphere after firing. (b) TEM image of the gold clusters on alumina sphere. Black dots indicated by white arrows are gold clusters. Size distribution of gold clusters with glutathione-protected ${\mathrm{Au}}_{25}$ with density of (c) 0.05 wt. %, (d) 0.1 wt. %, and (e) 0.2 wt. %.

We also conducted inductively coupled plasma mass spectrometry (ICP-MS) analysis.^19,22 Bi was used as an internal standard. The ICP-MS analysis of the solutions after stirring was conducted to estimate the quantity of Au that was not adsorbed onto the alumina sphere. We found that the gold clusters were almost adsorbed on the alumina sphere with an average adsorption efficiency of 92.1%. Glutathione molecules contain multiple polar carboxyl and amino functional groups. In addition, when metal oxide is added to water, hydroxyl groups ($─\mathrm{OH}$) are normally formed on the surface. Therefore, it can be considered that ${\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}$ was efficiently adsorbed onto the alumina owing to the formation of hydrogen bonds between the $─\mathrm{OH}$ groups on the surface of the alumina and the glutathione functional groups, as often observed in previous studies.^19,22 This high adsorption efficiency allowed us to obtain a high reproducibility of the adsorption of ${\mathrm{Au}}_{25}{(\mathrm{SG})}_{18}$ on the surface of the alumina.

Figure 1(b) shows the typical transmission electron microscope (TEM) images of the gold clusters on an alumina sphere. The sizes of the gold clusters were evaluated from the TEM images. As shown in Figs. 1(c)–1(e), the sizes of the gold clusters were in the range 1.0 to 1.3 nm, with monodispersed size. The size range was below the threshold diameter of 1.6 nm for the plasmon peak.²¹

2.2.

Re-Complex Attachment

We used $\mathrm{Re}(\mathrm{bpy}\text{-}\mathrm{COOH}){(\mathrm{CO})}_3\mathrm{Cl}$ ($\mathrm{bpy}\text{-}\mathrm{COOH}={\mathrm{2,2}}^{\prime }$-bipyridine-4,4′-dicarboxylic acid), abbreviated as Re(bpy–COOH),²³ as an Re-complex (metal complex) to investigate the ONF effect on ${\mathrm{CO}}_2$ reduction. As described in Ref. 24, first, $\mathrm{Re}{(\mathrm{CO})}_5\mathrm{Cl}$ was dissolved in methanol. An equimolar amount of 2,2′-bipyridine-4,4′-dicarboxylic acid was added to the solution, and the produced mixture was stirred under reflux, producing Re(bpy-COOH) [see Fig. 2(a)]. Second, to attach Re(bpy-COOH) to the alumina sphere with gold clusters, Re(bpy-COOH) was dissolved in acetonitrile. Then, the Re(bpy-COOH) solution was mixed with the alumina-containing solution, and the mixture was allowed to stand still for 8 h in a manner similar to that of a previous study.¹⁷ Through this process, Re(bpy-COOH) could attach to the alumina surface because of the deprotonation of the carboxyl group.

Fig. 2

(a) Schematic of the Re-complex attachment onto alumina sphere with gold clusters. FTIR spectra of (b) Re(bpy-COOH) and (c) Re(bpy-COOH) with alumina sphere. Black (alumina): alumina sphere only, blue [Re(bpy-COOH)+alumina]: Re(bpy-COOH) with alumina sphere, purple [Re(bpy-COOH)+alumina_Au0.05]: Re(bpy-COOH) with alumina sphere with gold density of 0.05 wt. %, green [Re(bpy-COOH)+alumina_Au0.1]: Re(bpy-COOH) with alumina sphere with gold density of 0.1 wt. %, and red [Re(bpy-COOH)+alumina_Au0.2]: Re(bpy-COOH) with alumina sphere with gold density of 0.2 wt. %.

To confirm the attachment of Re(bpy-COOH) on the alumina sphere, we obtained the Fourier transform infrared (FTIR) spectra using the KBr tablet method by Shimadzu IRPrestige-21 [see Figs. 2(b) and 2(c)]. In all measurements, we used 0.02 g of the samples. We focused on the wavenumber range corresponding to the carbonyl vibrational stretching modes: an out-of-phase symmetric mode; an asymmetric mode, primarily, among the two equatorial carbonyls; and an in-phase symmetric mode.²⁵ As shown in Fig. 2(b), the curve of Re(bpy-COOH) has three peaks at $\sim 2034$, 1930, and $1880{\mathrm{cm}}^{-1}$, corresponding to the carbonyl vibrational stretching modes, respectively. Figure 2(c) shows the FTIR spectra of the Re(bpy-COOH) attached to the alumina sphere with different gold densities. Because the amount of Re-complex attached to the alumina sphere was small, the absorbance was considerably smaller than that of Re-complex only, shown in Fig. 2(b). Therefore, the lower two peaks at considerably 1930 and $1880{\mathrm{cm}}^{-1}$ were degenerated, and the curves had a broad peak around $1920{\mathrm{cm}}^{-1}$, in addition to the peak at $\sim 2030{\mathrm{cm}}^{-1}$. From these results, it is confirmed that the Re-complex was attached to the aluminum sphere.

2.3.

Evaluation of the Absorption Spectra Change Following the Gold-Cluster Attachment

To check the effects of gold-cluster attachment in the vicinity of the Re-complex, we obtained the diffuse reflection spectra corresponding to the absorption spectra, using a spectrophotometer (JASCO V-670). Figure 3(a) shows the Kubelka–Munk plots of the diffuse reflection spectra²⁶ of Re(bpy-COOH) (solid black curve) and Re(bpy-COOH) attached to the alumina sphere with different gold-cluster densities. Two main peaks were observed at $\sim 310$ (peak 1) and 410 (peak 2) nm, corresponding to the $\pi -\pi *$ transition and metal-to-ligand charge transfer (MLCT) transition, respectively.²⁷ Figure 3(b) shows the ${\mathrm{Au}}_{25}$ density and peak wavelength dependence of the absorption peak. The peak value of peak 1 decreased following the attachment of gold clusters. The peak value of peak 2 increased as the ${\mathrm{Au}}_{25}$ density increased, which could be caused by the field-enhancement effect of the ONF.¹⁷ As described in our previous paper,¹⁷ if the direction of the incident light polarization is the same as that of the dipole field induced by the electronic polarization of nanoparticles, the dipole field enhances the electric field acting on the Re-complex and increases its absorption intensity [Fig. 3(c)]. This assumption was numerically supported by the first-principle calculations. We expect the density of gold clusters to increase as the ${\mathrm{Au}}_{25}$ density increased, it is reasonable that the peak value of peak 2 increased as the ${\mathrm{Au}}_{25}$ density increased. Based on our previous studies,¹⁷ we consider that the field enhancement can be dominant contributor to the increase in the ${\mathrm{CO}}_2$ reduction rate. However, the contribution of ONF depends on the geometrical arrangement of the Re-complex and the details of the geometrical and electric structures of the gold clusters. The main purpose of this study was to enhance the ${\mathrm{CO}}_2$ reduction rate using subnanometer gold without plasmon resonance. Note that the spectrum of Re(bpy-COOH) on an alumina sphere with a gold density of 0.2 wt. % [red solid curve in Fig. 3(a)] has an additional peak (peak 3) at $\sim 510\mathrm{nm}$. To evaluate the change further, we fitted the curve using three Lorentzian peaks at $\sim 314.1$ (peak 1), 413.3 (peak 2), and 508.6 nm (peak 3). The peak wavelength of peak 3 was similar to the plasmon peak of the gold clusters.²⁸ However, the plasmon peak in the absorption spectra of the gold clusters could be observed only when the diameter exceeded 1.6 nm.²¹ In addition, we observed a similar peak at $\sim 510\mathrm{nm}$ when using ZnO nanocrystalline aggregates, i.e., no gold.¹⁷ Therefore, peak 3 might not have been caused by the plasmon resonance. Several reports have been published on the thermal effects in metallic nanoparticles.^29,30 The temperature increase can be described as follows:³¹

Fig. 3

(a) Diffuse reflection spectra. Black solid curve: Re(bpy-COOH) deposited on alumina sphere, blue solid curve: Re(bpy-COOH) and gold cluster (with density 0.05 wt. %) deposited on alumina sphere, green solid curve: Re(bpy-COOH) and gold cluster (with density 0.1 wt. %) deposited on alumina sphere, and red solid curve: Re(bpy-COOH) and gold cluster (with density 0.2 wt. %) deposited on alumina sphere. (b) ${\mathrm{Au}}_{25}$ density and peak wavelength dependence of the absorption peaks. (c) Field enhancement as shown in the dashed red circles owing to the electronic polarization of nanostructure.¹⁷

2.4.

CO₂-Reduction Experiment

As the gold-cluster attachment enhanced the absorbance, we performed a ${\mathrm{CO}}_2$ reduction experiment [Fig. 4(a)]. ${\mathrm{CO}}_2$ gas was dissolved in dimethylformamide (DMF)/triethanolamine (TEOA) mixed solution ($\mathrm{DMF}/\mathrm{TEOA}=1\mathrm{mL}/0.2\mathrm{mL}$).³⁴ A laser light of wavelength 405 nm (50 mW) was used for the ${\mathrm{CO}}_2$ reduction experiment because this wavelength was closed to peak 2 and the difference of Kubelka–Munk intensities between 405 and 413.3 nm were $\sim 1\%$, where the absorbance increased because of the gold-cluster attachment owing to the field-enhancement effect of the ONF¹⁷ [see Fig. 3(a)]. In all experiments, we used 30 mg of Re-complex with the alumina sphere. We detected the CO generation using the FTIR spectra from Shimadzu IRPrestige-21. Because the CO spectra showed strong absorption in the synthetic silica below $2100{\mathrm{cm}}^{-1}$, we used a cell of anhydrous synthetic quartz (S15-IR-10, GL Sciences). Figure 4(b) shows the normalized absorption spectra of the R branch of Re(bpy-COOH) on an alumina sphere with the gold density of 0.2 wt. %, which was normalized with respect to that of the cell.³⁵ Note that these spectra were not obtained using the cell of synthetic silica.

Fig. 4

(a) Schematic of ${\mathrm{CO}}_2$ reduction experiment. (b) Time dependence of the normalized CO absorption spectra.

To evaluate the ${\mathrm{CO}}_2$ reduction rate, we plotted the absorbance at the spectral peaks. Figure 5 shows the ${\mathrm{Au}}_{25}$ density dependence. In all plots, the absorbance was found to be saturated. The saturation could be attributed to the degradation of Re(bpy-COOH). Because $\mathrm{Re}(\mathrm{bpy}\text{-}\mathrm{COOH}){(\mathrm{CO})}_3\mathrm{Cl}$ has three CO as ligands, Re(bpy-COOH) causes both the degradation of $\mathrm{Re}(\mathrm{bpy}\text{-}\mathrm{COOH}){(\mathrm{CO})}_3\mathrm{Cl}$ and ${\mathrm{CO}}_2$ reduction.³⁶ Regardless of the degradation of Re(bpy-COOH), it was found that the CO generation rate increased following the gold-cluster attachment, compared with the rate without the gold clusters. To evaluate this increase, we fitted the data as³⁶

Fig. 5

Normalized time-dependent absorption peak at $2169{\mathrm{cm}}^{-1}$ for different ${\mathrm{Au}}_{25}$ densities. w/o Au: Re(bpy-COOH) with alumina sphere without gold clusters, w/ Au0.05: Re(bpy-COOH) with alumina sphere with ${\mathrm{Au}}_{25}$ density of 0.05 wt. %, w/ Au0.1: Re(bpy-COOH) with alumina sphere with ${\mathrm{Au}}_{25}$ density of 0.1 wt. %, and w/ Au0.2: Re(bpy-COOH) with alumina sphere with ${\mathrm{Au}}_{25}$ density of 0.2 wt. %.

Table 1

Increased rates of in the CO2 reduction rate kred.

The ${\mathrm{CO}}_2$ reduction rate increased with the introduction of gold clusters. However, the ${\mathrm{CO}}_2$ reduction rate of 0.1 wt. % was the maximum value (in comparison with 0.05 and 0.2 wt. %) in most of absorption peaks except the peak at $2177{\mathrm{cm}}^{-1}$ (see Table 1). In the current experiment, the distance between the gold clusters and Re was random, and the position of Re was not controlled. More detailed investigation can be performed using the thiol-gold interaction by introducing thiol in the Re-complex.