Solar light-driven selective photoelectrochemical CO2 reduction to CO in aqueous media using Si nanowire arrays decorated with Au and Au-based metal nanoparticles

To address recent energy and environmental issues, such as global warming and resource depletion, significant interest has been shown in carbon dioxide (CO2) fixation based on photoelectrochemical processes under solar light irradiation. The present paper describes the applicability of gold nanoparticles-decorated silicon nanowire arrays (Au/SiNW) as photoelectrodes to promote CO2 reduction. The decoration with Au nanoparticles of SiNW was performed by an electroless plating utilizing surface hydrogen-terminated silicon groups, generated during the nanowire formation process. Au/SiNW exhibits efficient photoelectrochemical performance for CO2 reduction to produce CO selectively in an aqueous medium under simulated solar light irradiation owing to its vertically aligned nanowire structure and Au nanoparticles as cocatalysts. The former provides high specific surface area and light trapping effect, and the latter induces selective interaction with CO2. Moreover, a unique two-steps method for Au decoration that consists of photo-assisted deposition of copper nanoparticles and the following electroless plating to replace Cu atoms to Au ones achieves more uniform decoration of SiNW with highly dispersed core–shell structured Cu@Au nanoparticles. The resulting photoelectrode, termed Cu@Au/SiNW, shows improved selectivity toward CO production and gives a good Faradic efficiency of 72% in an aqueous medium.


Introduction
Recently, conversion of carbon dioxide (CO 2 ) to carbon monoxide (CO) and other valuable carbon-containing compounds, such as methane and methanol, using solar energy has been recognized as one of the most important challenges in the field of green sustainable chemistry [1][2][3][4][5][6][7][8][9][10]. This process enables us to store solar energy in the form of chemical fuels while reducing undesirable greenhouse gas. However, CO 2 is an extremely stable molecule and its reduction is thermodynamically difficult; thus, the catalyst design needs to meet a strong reduction potential for the facilitation of CO 2 reduction as well as visible-light responsivity for the efficient utilization of solar light. Moreover, in the reaction in aqueous media, which have been attracted much attention from a viewpoint of environmental impact owing to safeness and abundance, the realization of CO 2 reduction becomes much more difficult due to the competition with thermodynamicallyfavorable proton reduction [11,12].
Silicon (Si) is the most typically used material in solar cells because of its wide absorption characteristics from visible-to near infrared-region and easy controllability of its semiconducting properties by heteroatom doping. The recent trend is also toward its application to photocatalytic and photoelectrocatalytic materials [13][14][15][16][17]. In this regard, significant efforts have been devoted to construction of nanostructured Si materials, since the surface area per geometric area of film materials is a crucial property to promote catalytic reactions. Among those, there have been intensive interests in Si nanowire arrays (SiNW), which are readily prepared through a metal-assisted chemical etching catalyzed by silver nanoparticles and can provide improved light absorption and charge separation efficiencies based on light trapping effect and shortening the carrier migration length, respectively. [18][19][20][21][22][23][24][25][26] Moreover, surface hydrogen-terminated silicon groups generated by the chemical etching act as effective reduction sites for the deposition of metal nanoparticles. These attractive features of Si-based materials stimulate our interests in the development of novel photoelectocatalysts for CO 2 reduction reaction. Boron doping makes Si materials P-type semiconductors favorable to reduction reaction under light irradiation. Also, the decoration of the SiNW surface with metal nanoparticles that induce selective interaction with CO 2 , such as gold, silver and copper [11,[27][28][29][30][31][32], will improve its CO 2 reduction performances.
In this context, SiNW has been prepared via the metal-assist chemical etching method and decorated with Au nanoparticles as CO 2 reduction cocatalysts by utilizing surface hydrogen-terminated silicon groups for the development of photocathode materials to facilitate photoelectrochemical CO 2 reduction. The resulting photoelectrode, termed Au/SiNW, demonstrated selective CO 2 reduction properties to produce CO in aqueous media under simulated solar illumination. Furthermore, interestingly, a unique two-steps method that consists of copper nanoparticle deposition on SiNW and the following electroless plating to replace Cu atoms to Au ones achieved an improvement of selectivity toward CO production in CO 2 reduction by virtue of uniform decoration of the whole nanowire surface with highly dispersed Au-based nanoparticles.

Preparation of SiNW
SiNW was prepared by a metal-assisted chemical etching method with Ag nanoparticles in accordance with the previous literature with a slight modification [33,34]. A 10 × 10 mm 2 silicon wafer whose backside was masked with kapton tape, which was preliminary washed by ultrasonication in acetone, methanol and water in sequence, was dipped into a mixed aqueous solution of H 2 SO 4 and H 2 O 2 (H 2 SO 4 /H 2 O 2 = 3) for 15 min to remove adsorbed organics. The thus-cleaned Si wafer was dipped into 1.0 M HF aq. for 5 min to remove surface oxide layer and then transferred to a mixed solution of 50 mM AgNO 3 aq. and 3.0 M HF aq. 1 min reaction in this solution led to the deposition of Ag nanoparticles on the Si wafer. Silver-assist etching of the silicon wafer was performed in a mixed solution of 3.0 M HF aq. and 0.15 M H 2 O 2 aq. at 333 K for 3-18 min to form SiNW. Finally, the obtained SiNW was dipped in 50% HNO 3 aq. for 1 h to remove Ag nanoparticles.

Preparation of Au/SiNW
Au decoration of SiNW was performed by an electroless plating method. SiNW was dipped in 1.0 M HF aq. and rinsed with water. The cleaned SiNW were transferred to a mixed solution of 5.0 mM HAuCl 4 aq. and acetone (3:1) and allowed to react for 1 min, yielding Au/SiNW.

Preparation of Cu/SiNW
Cu decoration of SiNW was performed by a photo-assisted electroless plating method. SiNW was dipped in 1.0 M HF aq. and rinsed with water. The cleaned SiNW was transferred to a mixed aqueous solution of Cu 2 P 2 O 7 ·nH 2 O (83 g L −1 ) and K 2 P 2 O 7 (338 g L −1 ) and allowed to react for 2 min under simulated solar light from a solar simulator PEC-L11 (Peccell Technologies, Inc.), yielding Cu/SiNW.

Electroless plating of Au nanoparticles on Cu/SiNW
Au decoration of Cu/SiNW was performed by an electroless plating method. The above-obtained Cu/SiNW was dipped in a mixed solution of 5.0 mM HAuCl 4 aq. and acetone (3:1) and allowed to react for 1 min, yielding Cu@Au/SiNW.

General methods
Field emission scanning electron microscope (FE-SEM) images and elemental mapping images based on energy-dispersive X-ray spectroscopy (EDS) were obtained by using a Hitachi SU8000 equipped with a Bruker XFlash 6-10 silicon drift detector (SDD), operating under 15-30 kV accelerating voltage. Kr adsorption measurements were carried out by using a BELSORP TCV (MicrotracBEL Corp.) at 77 K. All the film samples were pretreated at 383 K overnight under vacuum. Specific surface areas were estimated from the amount of Kr adsorption at 77 K using the Brunauer-Emmett-Teller (BET) equilibrium equation. X-ray diffraction (XRD) patterns were recorded with a Rigaku SmartLab X-ray diffractometer using Cu Kα radiation (λ = 1.5406 Å). Diffuse reflectance UV-visible-NIR spectra were collected with a JASCO V-670 UV-vis-NIR spectrophotometer. X-ray photoelectron spectroscopy (XPS) measurements were performed with a Shimadzu ESCA-3200. Transmission electronic microscope (TEM) images were taken with a JEM-2000FX operating under 200 kV accelerating voltage. The Au L III -edge X-ray absorption fine structure (XAFS) spectra were measured in beamline BL01B1 facility of SPring-8 at the Japan Synchrotron Radiation Research Institute (JASRI). The X-ray was monochromatized with a pair of Si(111) crystals. All spectra were recorded in glazing-incidence fluorescence mode at 298 K using a 19-element Ge solid-state detector (SSD). To avoid the diffraction from SiNW, the film samples were rotated during measurements. The extended X-ray absorption fine structure (EXAFS) data were examined using an analysis program (Rigaku REX2000). In order to obtain the radial structure function, Fourier transformations of k 3 -weighted EXAFS oscillations in the range of 3-14 Å −1 for Au L 3 -edge spectra or 3-11 Å −1 for Cu K-edge spectra were performed.

Photoelectrochemical measurements
All electrochemical measurements were carried out with a homemade threeelectrode system consisting of Ag/AgCl reference electrode, platinum counter electrode, and photocathode at 298 K. The geometric area of photocathodes was fixed at 10 × 10 mm 2 . EGaIn was smeared on the previously polished backside of photocathodes to enable Ohmic contact. The electrolytes employed were 0.5 M H 2 SO 4 aq., 0.1 M KHCO 3 aq. or 0.1 M acetonitrile solution of TBAP containing 3% methanol. These electrolytes were saturated with Ar or CO 2 by bubbling.

3
Solar light-driven selective photoelectrochemical CO 2 … The electrode potential control and data collections were performed by using an electrochemical measurement system HZ-3000 (Hokuto Denko Corp.). A solar simulator PEC-L11 (Peccell Technologies, Inc.) was used as a light source.

Photoelectrochemical CO 2 reduction
Photoelectrocatalytic CO 2 reduction was performed with a closed quartz glass vessel connected to a vacuum line in CO 2 -saturated 0.1 M KHCO 3 aq. at a constant bias of − 1.0 V vs. Ag/AgCl under simulated solar light irradiation. The evolved gases were analyzed by using a gas chromatograph GC2010Plus Tracera equipped with a barrier ionization discharge (BID) detector (Shimadzu Corp.) and a Micropacked ST column.

Results and discussion
SiNWs were prepared by a metal-assisted chemical etching method with Ag nanoparticles for different etching times of 3, 9, and 18 min. The FE-SEM images of these samples shown in Fig. 1 demonstrated successful formation of vertically aligned nanowire structures. The mean wire lengths increased with increasing etching times and were determined to be 3, 10, and 20 μm for 3, 9 and 18 min, respectively. On the other hand, Kr adsorption measurements showed that optimal etching times existed in terms of specific surface areas ( Table 1). The maximum specific surface area of 374 cm 2 cm −2 was obtained by 9 min etching. This is because the excess long-time etching leads to decrease in the number of nanowires due to dislodging the nanowire, which was confirmed from top view FE-SEM images. Figure 2 shows cyclic voltammograms (CVs) of SiNWs prepared by different etching time, obtained in 0.5 M H 2 SO 4 aq. with Ar bubbling under simulated solar light irradiation. Cathodic currents were observed for all samples under simulated solar light irradiation, while dark conditions without light irradiation gave no current in the whole scan range. These results suggest that the SiNWs possess photoelectrocatalytic properties for proton reduction to produce H 2 . Notably, the construction of nanowire structures provoked a significant increase in photocurrent densities and anodic shift of the onset potentials around from − 0.5 to − 0.2 V vs. Ag/AgCl. The order of photoelectrocatalytic activity of SiNWs prepared by different etching times coincided with that of specific surface areas; as such, the SiNW prepared by 9 min etching gave the best photoelectrocatalytic performance for the reduction reaction. Therefore, subsequent investigations were performed by using the SiNW prepared by 9 min etching.  Fig. 3 are shown XRD patterns of SiNWs before and after Au decoration through the electroless plating method. The diffraction pattern of Au/SiNW showed a new peak at 38.2° corresponding to Au (111) plane for the fcc gold lattice [35,36], indicating the successful deposition of Au metal. On the other hand, almost no change in peaks corresponding to Si was observed after nanowire formation and Au nanoparticle decoration processes. The FE-SEM image of Au/SiNW shown in Fig. 4a demonstrated that Au nanoparticles can be deposited on SiNW while maintaining the nanowire structure. In the TEM image of Au/SiNW, mainly two different types of Au nanoparticles concerning their particle sizes were observed dependent on the location deposited, that is, larger particles existed on the top of nanowires and smaller ones with a mean diameter of 11.4 nm were dispersed along the side (Fig. 4b). It was also found from the elemental mapping images that Au nanoparticles were dispersed across the whole nanowires but slightly segregated near the top surface in part (Fig. 4c, d).
The electronic state of deposited Au nanoparticles was next investigated by XPS measurements. After Au decoration of SiNW, the XPS Au4f spectrum displayed distinct two peaks at 83.5 eV and 87.3 eV ( Figure S1), which were attributable to 4f 7/2 and 4f 5/2 peaks for Au 0 , respectively. [37][38][39] This finding suggests that Au nanoparticles deposited on SiNW through electroless plating exist in a metallic state.
UV-vis-NIR diffuse reflectance spectra of Si wafer, SiNW, and Au/SiNW are shown in Figure S2. The spectrum of Si wafer exhibited a weaker absorption than those of SiNW and Au/SiNW, especially in the UV-vis region. This finding was explained by a light harvesting effect derive from vertically aligned nanowire structures, that is, incident light was trapped in the space between nanowires via repeated reflection on the nanowire surface.
Photoelectrocatalytic performance for CO 2 reduction reaction was tested in 0.1 M KHCO 3 aq. with CO 2 bubbling under simulated solar light irradiation. As shown in Fig. 5, all samples showed reduction current in the scan range from 0.0 to − 2.0 V vs. Ag/AgCl. The formation of the nanowire structure associated with the increase in the specific surface area improved photoelectrocatalytic performance, even under CO 2 -saturated conditions. Moreover, significant improvement in the photoelectrocatalytic performance was observed after Au decoration onto the SiNW. This result indicates that deposited Au nanoparticles behave as effective cocatalysts for photoelectrocatalytic CO 2 reduction. On the other hand, Au foil exhibited more negative onset potential and lower current density than Au/SiNW, suggesting the effective photo-induced assistance by SiNW.
However, in this reaction condition, reduction current density involves the contribution of the competitive reaction of proton reduction and CO 2 reduction; thereby, to gain an insight into the faithful role of deposited Au nanoparticles, CV measurements were performed both in CO 2 -saturated 0.1 M TBAP acetonitrile solution containing 3% methanol and in Ar-saturated 0.5 M H 2 SO 4 aq. As shown in Fig. 6, under the former conditions, Si wafer did not show the reduction current in the whole scan range, while a moderate reduction current was observed by SiNW, indicating that the formation of nanowire structure generates new active sites for CO 2 reduction. Moreover, the Au deposition on SiNW achieved considerable increase in reduction currents and positive shift of the onset potential as compared to the bare SiNW, suggesting a promoting effect of the deposited Au nanoparticles on CO 2 reduction. When the CV of Au/SiNW was measured in Ar-saturated 0.1 M TBAP acetonitrile solution containing 3% methanol, only a weak reduction current was seen in the curve ( Figure S3b). Therefore, most photocurrents observed by Au/SiNW under CO 2 -saturated conditions are considered to be associated with CO 2 reduction reaction. Under the latter conditions, that is in Ar-saturated 0.5 M H 2 SO 4 aq., Au/ SiNW showed a similar onset potential and slightly increased photocurrent density as compared to SiNW ( Figure S4), and its degree of improvement was lower than that observed in CO 2 -saturated conditions. These findings revealed that the deposited Au nanoparticles behave as selective cocatalyst for enhancing CO 2 reduction performance.
Subsequently, Cu decoration of SiNW was attempted in view of the fact that Cu metal is also an effective catalyst for CO 2 reduction. Since the deposition rate of Cu nanoparticles through electroless plating was very slow due to a small difference of standard electrode potentials (E 0 ) between Cu/Cu 2+ (0.34 V) and Si-H/Si-O − (− 1.23 V) [21], decoration of SiNW with Cu nanoparticles was performed with the help of light irradiation. Under light irradiation, photo-formed electrons in SiNW would facilitate the reduction of Cu 2+ ions. In fact, it was confirmed from FE-SEM observations that highly and uniformly dispersed Cu nanoparticles were deposited over the entire surface of SiNW from the top to the root in the image of Cu/SiNWs by using the photo-assisted method (Fig. 7A). However, unfortunately, Cu/SiNW did not exhibit large enhancement of photoelectrocatalytic activity compared to bare SiNW (Fig. 7B), and also, the reduction product analyses showed that Cu/SiNW possessed almost no activity for CO 2 reduction to CO.
In the photocatalytic CO 2 reduction promoted by metal particle cocatalysts, activity and selectivity are known to be affected strongly by the size of metal particles [32]. Considering that the bare SiNW showed a weak activity for CO 2 reduction, the dispersion of and surface coverage by metal particles on SiNW are also of great importance to realize enhanced product selectivity. However, it is unlikely to control the sizes of deposited metal particles precisely by using the electroless plating involving the reduction by the hydrogen-terminated silicon because the particle sizes deposited in this method are not dependent on the concentration of the metal salt solution, but on the ratio of E 0 of the metal cation divided by its charge (oxidation number, n), that is E 0 /n value [21]. On the other hand, as mentioned above, Cu nanoparticles were successfully deposited on SiNW with high dispersion through a photo-assist deposition method. Therefore, we anticipated that a two-steps method consisting of the deposition of highly dispersed Cu nanoparticles and the following metal replacement from Cu to Au nanoparticles would realize the development of effective catalysts for selective CO 2 reduction.
Under this consideration, Cu/SiNW was dipped into HAuCl 4 aq. to replace surface Cu atoms with Au ones (Cu@Au/SiNW). As shown in Fig. 8a, the FE-SEM image of Cu@Au/SiNW revealed the retaining of nanowire structure even after the electroless plating to replace Cu with Au. It was found from TEM images displayed in Fig. 8b, c that the deposited metal nanoparticles exist in more high dispersion than Au nanoparticles on Au/SiNW and that a mean diameter of metal nanoparticles along the side of nanowires is 10.6 nm. Since the rate of the photo-assisted deposition of Cu nanoparticles was slow, the reduction of Cu 2+ ions was hard to be a diffusion-limited process. Therefore, Cu nanoparticle deposition occurred on the whole SiNW surface. The replacement of Au atoms was then selectively proceeded on highly dispersed Cu nanoparticles, resulting that the uniform deposition of Au nanoparticles on SiNWs would be achieved. The presence of highly dispersive Au nanoparticles on Cu@Au/SiNW was also able to be observed by EDX mapping (Fig. 8d, e). By contrast, a mapping image corresponding to Cu showed weak signals, and the residual amount of Cu nanoparticles was found to be very small from a point analysis (ca. 0.35 wt%, Figure S5). To gain an insight into the chemical state of metal nanoparticles on Cu@Au/ SiNW, XPS and XAFS analyses were carried out before and after the Au replacement process. The Au4f XPS spectrum of Cu@Au/SiNW shown in Fig. 9A displayed obvious 4f 7/2 and 4f 5/2 peaks at 83.3 and 87.0 eV assigned to an Au 0 state, which were not detected in the Cu/SiNW spectrum. It should also be noted that these peak positions were shifted to lower binding energy side compared to those of Au/SiNW ( Figure S1), indicating that Au species in Cu@Au/SiNW possessed an increased electron density due to Cu species. XAFS analyses also demonstrated that the Au species in Cu@Au/SiNW exist in a metallic state ( Figure S6). The XANES spectrum of Cu@Au/SiNW showed absorption edge position and spectral shape accord with Au foil. In the radial structure function of Cu@Au/SiNW, a peak corresponding to Au-Au bond was observed at 2.7 Å (without peak shift correction) [40]. As regards the information on Cu side, two peaks observed in the Cu2p XPS spectrum of Cu/SiNW almost disappeared after the Au replacement process (Fig. 9B); however, the radial structure function of Cu@Au/SiNW after Fourier transform of the Cu K-edge EXAFS spectrum gave a small peak at a similar peak position to that of Cu foil ( Figure S7). These findings suggest that replacement of Cu atoms with  Au ones was attained successfully, resulting in the formation of a core@shell-like structure with a very small Cu nanoparticle core. Also, it can be said that the present two-steps method is an effective way to design highly and uniformly dispersed Au nanoparticles on the surface of SiNW.
Finally, photoelectrocatalytic CO 2 reduction reactions were performed using Cu@Au/SiNW at a constant bias of − 1.0 V vs. Ag/AgCl in a CO 2 -suturated KHCO 3 aq. under simulated solar light irradiation. Figure 10 summarizes and compares the formation amounts of reduction products after 1 h obtained by using various photoelectrodes. As discussed in the above CV experiments, the construction of nanowire structures enhanced the reduction catalytic activities of Si wafer. However, bare Si surface without cocatalysts facilitates only proton reduction to produce H 2 . On the other hand, CO 2 reduction proceeded on Au/SiNW to produce CO thanks to selective interaction between Au nanoparticles and CO 2 . The reduction product selectivity of Au/SiNW was determined to be 37% to CO and 63% to H 2 . Moreover, Au decoration led to increase in the total amount of reduction products. In the case of Cu@Au/SiNW, further improvement of selectivity in CO 2 reduction was attained. The reduction product selectivity was 72% to CO and 28% to H 2 , and as a result, the Faradic efficiency toward CO production reached 72%. From the above results, CO 2 reduction to CO selectively proceeded on the Au surface, while proton reduction occurred on the surface of SiNWs. Since highly dispersed Au-based metal nanoparticles in Cu@Au/SiNW that were prepared by the electroless plating to replace Cu with Au would cover most SiNW surface, the selectivity toward CO production of Cu@Au/SiNW was considered to be improved as compared to that of Au/ SiNW prepared by the one step method. Furthermore, the increased electron density of Au species in Cu@Au/SiNW, as revealed by the results of XPS measurements (see Figs. 9A and S1), should also have affected the improved selectivity. In general mechanism ( Figure S8), a CO 2 molecule is adsorbed on Au surface as a form of CO 2 •− species through an Au-C coordination. A proton then reacts with the negatively-charged O atom to form CO 2 H species in aqueous media. The thus-formed CO 2 H species is further reduced by another photo-generated electron and reacts with another proton to give CO and water [28]. The increased electron density of Au species in Cu@Au/SiNWs would promote the formation of CO 2 •− species, resulting in the improved selectivity toward CO production. These results suggest that effective and highly dispersed CO 2 reduction sites are formed on Cu@Au/SiNWs through the developed unique two-steps method. In addition, the reduction current density maintained during the photoelectrocatalytic CO 2 reduction reactions, showing the high durability of Cu@Au/SiNW ( Figure S9).

Conclusions
In summary, Au-decorated SiNW (Au/SiNW) was prepared by a metal-assist chemical etching of Si wafer and the following electroless deposition of Au nanoparticles and applied to photoelectrochemical CO 2 reduction in aqueous media under simulated solar light irradiation. The promotion of CO 2 reduction to afford CO was observed on Au/SiNW despite the proton reduction is more thermodynamically favorable than CO 2 reduction. CV measurements in two different reaction media, which enable to evaluate reactivities for CO 2 reduction and proton reduction individually, indicated that selective interaction with CO 2 will take place on the decorated Au nanoparticles. Moreover, a unique two-steps decoration method including a photo-assisted electroless plating of Cu nanoparticles followed by electroless plating of Au nanoparticles (Cu@Au/SiNW) effected the high dispersion of Au nanoparticles, resulting in enhanced selectivity toward CO production. Cu@Au/SiNW realized 72% Faradic efficiency toward CO production in photoelectrochemical CO 2 reduction at a constant bias of − 1.0 V vs. Ag/AgCl in a CO 2 -suturated KHCO 3 aq. under simulated solar light irradiation. Thus, both the utilization of SiNW as a light absorption unit and uniform decoration of metal cocatalysts were found to be effective for solar light-driven photoelectrochemical CO 2 reduction. The uniform decoration technique developed could be applied to other metal nanoparticle formation for not only photoelectrode materials but supported metal nanoparticle catalysts in the future.