Surface plasmon-driven catalytic reactions on a patterned Co 3 O 4 /Au inverse catalyst †

We report hot electron- and surface plasmon-driven catalytic reactions on an inverse catalyst that is composed of a reactive cobalt oxide (Co 3 O 4 ) for the catalyst portion and a metal as the source of localized surface plasmon resonance (LSPR). To achieve this scheme, a patterned Co 3 O 4 /Au nanostructure was fabricated using the nanosphere lithography technique. The enhancement of catalytic activity for CO oxidation on cobalt oxide by the transfer of plasmonic hot carriers generated from the LSPR of Au under light irradiation was observed. We found that the Co 3 O 4 /Au inverse catalyst exhibited 25 – 50% higher catalytic activity when under light irradiation. We attribute this enhancement of catalytic activity to the ﬂ ow of surface plasmon-driven hot carriers that were created by the absorption of photons on the Au metal, followed by injection into the active Co 3 O 4 layers.


Introduction
Plasmonic metal nanoparticles have been widely used for many applications (e.g. catalytic activity, water splitting reactions, biosensors, solar-cell design, and cancer diagnostics). [1][2][3][4][5][6] Surface plasmon resonance (SPR) is a strong electromagnetic eld near the surface of noble metal nanostructures that can occur in plasmonic metal nanostructures by resonant oscillation of free electrons with the same wavelength as the incident photons. [7][8][9] Recently, many studies have shown the generation of hot electrons with high kinetic energy (1-3 eV) by the electromagnetic decay of surface plasmons at a femtosecond timescale in plasmonic nanostructures under light irradiation and SPR excitation. [10][11][12] Mukherjee et al. revealed that H 2 molecules on Au nanoparticles can dissociate at room temperature because of hot electrons created from surface plasmon decay under visible light irradiation. 13 Lee et al. found a steady-state ow of hot electrons generated by localized surface plasmon resonance (LSPR) at a metal-semiconductor junction. 14 Wang et al. reported that a high concentration of hot electrons can tunnel through the insulating layer and generate photocurrent in a metal-insulator-metal device. 15 To fabricate new plasmonic nanostructures with homogeneous size, shape, and interparticle distance of the structures, several nanolithography techniques have been developed (e.g. electron beam lithography, focused ion beam lithography, and nanoimprint lithography). [16][17][18] Furthermore, the nanosphere lithography (NSL) technique is a low-cost fabrication technique that uses 2-dimensional colloidal crystals (e.g. 2D arrays of polystyrene and silica nanospheres) as a mask. The advantage of the NSL technique for plasmonic nanostructures is that the LSPR of noble metal nanostructures can be easily controlled in the visible to near-IR region because the aspect ratio that determines the extinction maximum wavelength can be accurately adjusted by varying the size of the nanospheres used as the mask and the thickness of the deposited noble metal. 19 Conventional catalysts consist of a metal as the catalyst and an oxide or semiconductor as the support for the metal catalyst. It has been demonstrated that the interaction between the metal and the oxide support at their interface (i.e. strong metalsupport interactions (SMSI)) can affect the catalytic activity with a dependence on the oxidation state of the metal catalyst and the doping state or reducibility of the oxide support. [20][21][22][23] In addition, there are some studies that show the inuence of hot electrons from the oxide or semiconductor support on the catalytic activity of metal catalysts when under light irradiation. [24][25][26] To understand the intrinsic relation between catalytic activity and hot electron ows generated across metal-oxide interfaces, the design of a fresh catalyst model is important. 27 In previous studies, the active site was the metal, which is inuenced by hot electrons from an oxide or a semiconductor. The major shortcoming of this conventional catalyst is the short range of the electron mean free path for the metal, which can de-energize the hot electron before its arrival at the metal surface. Kim et al. reported that the change in catalytic activity for CO oxidation on Au/CeO 2 or Pt/GaN catalysts under light irradiation decreased as the size of the Au or Pt nanoparticles increased because of the length of the mean free path of the metal. 24,26 Lee et al. found that smaller Pt nanoparticles lead to higher chemicurrent yield (i.e. the number of hot electrons detected per molecule of reaction product formed on the Au/ TiO 2 nanodiode surface) during hydrogen oxidation, which is associated with the shorter travel length for the hot electrons, compared with their inelastic mean free path. 28 Therefore, the role of the electron mean free path is critical for studying the relationship between catalytic activity and hot electron ow.
Some oxide materials (e.g. cobalt oxide (Co 3 O 4 ), cerium oxide (CeO 2 ) and cuprous oxide (Cu 2 O)) are catalytically reactive because these oxide materials can be easily and reversibly oxidized and reduced. [29][30][31] Among these reactive oxide materials, Co 3 O 4 is specically known as the most catalytically active oxide material for CO oxidation. Haruta and co-workers demonstrated that some base metal oxides (e.g. Co 3 O 4 , MnO 2 , and NiO) are intrinsically reactive for CO oxidation without a metal catalyst and that the catalytic activity of base metal oxides decreases in the following order: 32 Xie et al. found that Co 3 O 4 nanorods with predominantly exposed [110] planes are very reactive and stable for CO oxidation at temperatures as low as À77 C, giving 100% CO conversion because of the presence of active Co 3+ species at the surface. 33 Song et al. reported that mesoporous cobalt oxide nanoparticle aggregates can achieve 100% CO conversion to CO 2 at À60 C under normal conditions. 34 It was found that a small Co 3 O 4 -SiO 2 nanocomposite catalyst prepared using an activated carbon template exhibits very high activity for CO oxidation at À76 C. 35 To overcome the short electron mean free path problem in the metal, we used an inverse catalyst in this study that is composed of a reactive oxide for the catalytic portion and a noble metal for the LSPR source. Triangular noble metal nanostructures show much higher surface-enhanced Raman scattering (SERS) activities than at noble metals because of the large number of conned hot spots between adjacent triangles. 36 Subsequently, we grew Co 3 O 4 on a patterned triangular Au nanostructure to make the patterned Co 3 O 4 /Au inverse catalyst using NSL (Fig. 1a). We studied the catalytic activity for CO oxidation on the patterned Co 3 O 4 /Au inverse catalyst with and without light irradiation to identify the role of hot electrons from LSPR on the catalytic reaction. The energy band diagram of the Co 3 O 4 /Au junction (Fig. 1b) shows the transfer of hot electrons and their effect on catalytic activity. The Schottky barrier height for Au/Co 3 O 4 is 1.6 eV, which was determined by the difference between the work function of the Au (5 eV) and the electron affinity for Co 3 O 4 (3.4 eV). 37 The main merit of our inverse catalyst is the long mean free path in the oxide created by using a reactive oxide material on a plasmonic noble metal. Energetic hot electrons having high kinetic energy (2.0-2.1 eV) formed via LSPR upon the absorption of photons on the plasmonic gold. The hot electrons then travelled over the Schottky barrier and arrived at the Co 3 O 4 surface because of their long mean free path. Thus, these plasmonic hot carriers can affect the catalytic activity of CO oxidation on Co 3 O 4 .

Fabrication of the two-dimensional patterned Co 3 O 4 /Au nanostructures by nanosphere lithography
Chemicals. Tetraethyl orthosilicate (TEOS, Aldrich, >99.0%), sodium dodecyl sulfate (SDS, Aldrich, >99.0%), ammonium hydroxide (NH 4 OH, Dae Jung Chemicals, 25.0-28.0%), chloroform (CHCl 3 , Junsei, 99.0%), ethanol (C 2 H 5 OH, Merck), and methanol (CH 3 OH, Merck) were used. 300 nm SiO 2 nanospheres were synthesized using the Stöber method, as follows. 0.55 mL of NH 4 OH in 3.5 mL of deionized water was added to 10 mL of ethanol and stirred. 2.3 mL of TEOS in 10 mL of ethanol was injected dropwise into the previous solution. Aer stirring the reaction solution at room temperature for 12 h, the mixture was centrifuged at 3000 rpm for 10 min and washed with ethanol. Aer repeating this washing process three times, the 300 nm SiO 2 nanospheres were dispersed in methanol. Selfassembled monolayer arrays of silica nanospheres were prepared on a clean SiO 2 /Si substrate by utilizing the Langmuir-Blodgett (LB) technique using a Nima Technology 311D. To spread the SiO 2 nanospheres onto the deionized water subphase in a trough, 3 mg of SDS was mixed with 2 mL of the previous silica solution dispersed in methanol. Aer ultrasonication of the mixed solution for 40 min with heating, 3 mL of chloroform was added to oat the silica nanospheres on the water and sonicated again for 30 min. This prepared solution was spread and oated on the water surface; the surface pressure was monitored using a Wilhelmy plate. While maintaining the target surface pressure at 15 mN m À1 , the selfassembled packed silica nanospheres were transferred onto the SiO 2 /Si substrate through the monolayer by slowly liing the substrate at a constant speed of 1 mm min À1 . For fabrication of the Co 3 O 4 /Au nanostructures, thin lms consisting of 5 nm of titanium (Ti), 30 nm of gold (Au), and 40 nm of cobalt (Co) were deposited sequentially using electron beam evaporation on the previously closely packed SiO 2 nanospheres on SiO 2 /Si substrate. At room temperature, all the thin lms were deposited using 10 kV DC voltage in a high-vacuum chamber with a base pressure of 5 Â 10 À7 Torr, evacuated using rotary pumps and turbo molecular pumps. The metal deposition rate was 0.1Å s À1 until 2 nm was deposited for adhesion. The deposition rate was then increased to 0.5Å s À1 until reaching the nal metal thickness. The closely packed SiO 2 nanosphere arrays act as a mask, thus all the metal was deposited with the SiO 2 nanospheres creating the patterning. Aer nishing the metal deposition, the cobalt was oxidized to Co 3 O 4 by heat treatment at 553 K for 2 h 30 min. Finally, the 300 nm SiO 2 nanospheres were removed by sonication for 1 minute with deionized water. The patterned Co 3 O 4 -Au islands were thus obtained using NSL. Based on our experience fabricating Au/TiO 2 catalytic nanodiodes, in the case of the Au/Ti interface, the electrical contact is an ohmic contact and is very unlikely to have a Schottky barrier. To form a Schottky barrier, we would need to heat the Ti layer under oxygen to form TiO 2 before depositing the Au layer.

Characterization, chemical modication, and structures of the patterned Co 3 O 4 /Au nanostructures
The crystalline phases of the cobalt oxide and the Au nanostructure before and aer heat treatment were revealed using Xray diffraction patterns (XRD, Rigaku D/MAX-2500) taken at a 2q scan range of 10-70 , scan speed of 4 min À1 , and step size of 0.01Å using Cu Ka radiation. The oxidation state of the cobalt oxide (Co 2p) before and aer CO oxidation was identied using X-ray photoelectron spectroscopy (XPS, Thermo VG Scientic Sigma Probe system with an Al Ka X-ray source (1486.3 eV)) with an energy resolution of 0.47 eV full width at half maximum (FWHM) under ultrahigh vacuum (10 À10 Torr). The thickness and morphology of the patterned Co 3 O 4 /Au nanostructures were characterized using cross-section transmission electron microscopy (TEM, FEI Titan G2 Cubed 60-300), scanning transmission electron microscopy and high-angle annular dark eld (STEM-HAADF, FEI Titan G2 Cubed 60-300), corresponding energy-dispersive X-ray spectroscopy (EDS, FEI Super X (4 SDD EDS)) elemental maps for 300 s, energy-dispersive X-ray spectroscopy (EDS) line proles, and scanning electron microscopy (SEM, Magellan 400). The TEM specimen was prepared using the focused ion beam (FIB, Helios Nanolab 450 F1) preparation technique for cutting cross sections of the Co 3 O 4 /Au nanostructures. Two sizes of triangular Au nanostructures were created by changing the size of the SiO 2 nanosphere mask. The optical properties of these Au nanostructures were characterized using a UV-vis spectrophotometer (Lambda 1050). Triangular Au nanostructures (30 nm high) were prepared on a quartz window (1 mm thick) using the same method described previously. The spectral range was 400-800 nm, and a clean quartz window was used as a reference.

Measurement of the catalytic activity of CO oxidation on patterned Co 3 O 4 /Au inverse catalysts with and without light irradiation
Reaction studies were performed in an ultrahigh vacuum chamber (1 L) with a base pressure of 1 Â 10 À8 Torr, evacuated using rotary pumps and turbo molecular pumps. Aer isolating with a gate valve, the CO oxidation reaction was carried out in the batch reactor system under 40 Torr CO and 100 Torr O 2 as the reactant gases and 620 Torr He as the balancing gas. All the gases were circulated continuously through the reaction line by a Metal Bellows recirculation pump at a rate of 2 L min À1 . Aer 30 min of recirculation at room temperature for gas equilibration, a DS 6200 gas chromatograph (GC) equipped with a thermal conductivity detector and a 6 in. Â 1/8 in. SS molecular sieve 5A was used to separate the products for analysis. The CO conversion was monitored at various temperatures (523-553 K) and the measured reaction rates were reported as turnover frequencies (TOF), which were measured in units of product molecules of CO 2 produced per active Co 3 O 4 surface site per second of reaction time. All the reaction data were obtained by measuring the rate of CO oxidation at low conversion (i.e., conversion below $15%) to check the initial reaction rate within a kinetically controlled regime, assuming different reaction conditions. The number of active sites was calculated from the geometry in the SEM measurements of the surface area of the nanoparticle arrays. A halogen lamp (Dolhan-Janner D150) was used as the light source to measure the enhancement of catalytic activity while under light irradiation. The distance between the sample and halogen lamp was about 8 cm through a sapphire window and the light intensity was measured to be 49.4 mW cm À2 at this distance by an optical power meter (ADCMT 8230).

Finite-difference time domain (FDTD) simulations
The distributions of the electric eld by localized surface plasmon resonance (LSPR) on the patterned Co 3 O 4 -Au nanostructure and the patterned Au nanostructure were calculated using a two-dimensional FDTD simulation with a commercial FDTD simulation package (FDTD-Lumerical™). The optical properties of the gold were obtained from the materials database in the soware using the Drude model. The refractive index of Co 3 O 4 was obtained experimentally using a spectroscopic ellipsometer (Woollam M2000D). Variable angle spectroscopic ellipsometric (VASE) data (i.e. 65.00, 70.00, 75.00 ) were acquired from a Co 3 O 4 thin lm on a SiO 2 wafer (173 nm cobalt oxide/1000 nm SiO 2 ) and modelled in B-spline mode to calculate the refractive index for cobalt oxide (2.8 at 622 nm). The electric eld distribution was acquired with a mesh size of 0.5 nm at an incident light of 622 nm, which is the extinction maximum wavelength (l ex ) of the nanostructures derived from the UV-vis absorbance spectra. The geometric model used in these calculations for the patterned Co 3 O 4 /Au nanostructure and the patterned Au nanostructure was built from the SEM and TEM images.

Characterization of the patterned Co 3 O 4 /Au inverse catalysts
We fabricated reactive oxide/plasmonic noble metal hybrid structures using the NSL technique. The Co 3 O 4 on triangular Au nanostructures were thus fabricated: rst, 300 nm SiO 2 nanospheres were synthesized using the well-known Stöber method, and self-assembled monolayer arrays of silica nanospheres were prepared on a clean SiO 2 /Si substrate using the LB technique. Next, 30 nm of gold as the LSPR source and 40 nm of cobalt as the catalyst were deposited sequentially by electron beam evaporation on the previously fabricated closely packed SiO 2 nanospheres on SiO 2 /Si substrate. The closely packed SiO 2 nanosphere arrays act as a mask so that all the metal was deposited using the SiO 2 nanospheres for patterning. Finally, the cobalt was oxidized to Co 3 O 4 by heat treatment at 553 K for 2 h 30 min and the 300 nm SiO 2 nanospheres were removed by sonication for 1 minute with deionized water, thus obtaining the patterned Co 3 O 4 -Au islands. All the fabrication procedures and SEM images are shown in Fig. 2a-d and e-h, respectively. Fig. 3a shows the crystalline phase of the cobalt oxide according to annealing temperature, as revealed by XRD patterns. Aer annealing at 553 K, diffraction peaks corresponding to a cubic spinel structure were conrmed, indicating that the cobalt was fully oxidized to Co 3 O 4 . Co 3 O 4 has a cubic spinel structure and the Au nanostructure was polycrystalline aer annealing at 553 K, as shown in Fig. 3a and S1 in ESI. † Fig. 3b illustrates the UV-vis extinction spectra of the 30 nm triangular Au nanostructures fabricated by NSL using 200 and 300 nm SiO 2 nanospheres as masks. All the Au cones remained and there were no SiO 2 nanospheres aer sonication, as shown in Fig. S2 in ESI. † It is well known that the optical properties (e.g. the intensity and extinction maximum wavelength) of the LSPR for noble metal nanoparticles are inuenced by shape, size, and dielectric environment. Previous studies show that the extinction maximum wavelength of triangular noble metal nanostructures fabricated by NSL can be tuned by changing the aspect ratio (i.e. the ratio of the length and the height) of the triangular nanostructures. 38,39 In this work, the aspect ratio was controlled by changing the size of the SiO 2 nanospheres used as the mask. The extinction maximum wavelength peak of the triangular Au nanostructures from the 300 nm SiO 2 nanospheres was red-shied from the wavelength peak of the triangular Au nanostructures from the 200 nm SiO 2 nanospheres because of the increased aspect ratio from the large SiO 2 nanospheres, which reveals that the extinction spectrum represents absorption by the LSPR of the triangular Au nanostructures. In addition, there was an ultraweak absorption band about 750 nm away from the extinction maximum wavelength peak. This broad LSPR band came from some defects in the triangular Au nanostructures that were caused by misalignment of the SiO 2 nanospheres or by the size distribution of the SiO 2 nanospheres, as shown in Fig. S2 in ESI. † 39 To identify the thickness and morphology of the patterned Co 3 O 4 /Au nanostructures, a cross-sectional TEM image was taken (Fig. 4a). During fabrication of the Co 3 O 4 /Au nanostructures, vacancies between three closely packed SiO 2 nanosphere arrays created cobalt oxide/Au nanostructures with nanopyramidic shape. Fig. 4d-h shows a STEM-HAADF image and EDS elemental maps for 300 s of the patterned Co 3 O 4 /Au nanostructure. It clearly shows that the top layer is cobalt oxide (i.e. the catalyst portion) and that there is a Au triangle under the cobalt oxide (i.e. the LSPR source). An EDS line prole of the patterned Co 3 O 4 /Au nanostructure is shown in Fig. 4b and c. The heights of the Au triangle and cobalt oxide are 30 and 40 nm, respectively.

Surface-plasmon driven CO oxidation on patterned Co 3 O 4 /Au inverse catalysts
Here, we carried out CO oxidation on the patterned Co 3 O 4 /Au inverse catalyst with and without light to show the intrinsic relation between catalytic activity and hot electron ow from the LSPR on the triangular Au nanostructure during light irradiation. Reaction studies were performed in an ultrahigh vacuum chamber (1 L) with a base pressure of 1 Â 10 À8 Torr, evacuated   using rotary and turbo molecular pumps. Aer isolation using a gate valve, the CO oxidation reaction was carried out in the batch reactor system under 40 Torr CO and 100 Torr O 2 as the reactant gases and 620 Torr He as the balancing gas. CO conversion was monitored at various temperatures (523-553 K) and the measured reaction rates were reported as TOF, which are measured in units of product molecules of CO 2 produced per active Co 3 O 4 surface site per second of reaction time. Fig. 5a and b shows the catalytic activity for CO oxidation measured on the patterned Co 3 O 4 /Au inverse catalyst at four different temperatures (523-553 K) with and without light. The patterned Co 3 O 4 /Au inverse catalyst shows a higher catalytic activity under light irradiation than in the dark. Previous studies show that the enhancement of the activity of the Pt NPs/ p-doped GaN wafer under light irradiation was 11-33% which was measured using the same batch reactor system and halogen lamp as used in this study; 26 thus, the enhancement under light irradiation (25-50%) observed on the Co 3 O 4 /Au inverse catalyst is signicant. It is associated with hot electrons generated from LSPR excitation of the triangular Au nanostructure by light absorption that can affect the catalytic activity for CO oxidation on the reactive Co 3 O 4 . However, the thermal effect of light irradiation can also result in the enhancement of catalytic activity. We measured sample temperature data with a temperature controller with and without light at 523 K under 40 Torr CO, 100 Torr O 2 , and 620 Torr He (Fig. S3 in ESI †). There was no change in the sample temperature when light was irradiated on the sample, demonstrating that any heating effect by light irradiation can be ignored in our experiments. Also some previous studies show a decrease in catalytic activity for CO oxidation on Pt/n-Si and Pt nanoparticle/n-type GaN resulting from a reduced net charge on the Pt nanoparticles when under light irradiation. 25,26,40 CO oxidation was also carried out on the patterned Co 3 O 4 nanostructure with and without light under the same conditions to compare the catalytic activity of the patterned Co 3 O 4 /Au inverse catalyst (Fig. 5c). Under light irradiation, the catalytic activity of the Co 3 O 4 exhibited an insig-nicant difference (i.e. within the error of measurement). Even if a hot electron is generated on the Co 3 O 4 from electron-hole pair excitation when under light irradiation, it will eventually turn into a low-energy electron through inelastic scattering within a time scale of femtoseconds. Hot electrons from LSPR excitation on Au can be transferred when a metal-oxide interface is formed, revealing that a role for Au nanostructures is as an LSPR excitation source, leading to a change in catalytic activity for CO oxidation. The TOF values of the patterned Co 3 O 4 /Au inverse catalyst were similar to the TOF values of the patterned Co 3 O 4 nanostructure in the same reaction temperature range, demonstrating that there was no synergetic effect between the Au and the Co 3 O 4 . We also carried out CO oxidation on the patterned Au nanostructure at 573 K without Co 3 O 4 , but there was no CO 2 product although the triangular Au nanostructure was polycrystalline. Thus, the CO oxidation reaction was occurring on the Co 3 O 4 , and not on the Au nanostructure. While the triangular Au nanostructure is not reactive in this reaction temperature range, it does perform a role as the LSPR source when under light irradiation. Hence, we can conrm that the enhancement of catalytic activity only originated from hot carriers generated by LSPR excitation of the Au that were then transferred to the reactive Co 3 O 4 .
To conrm whether the plasmonic hot carrier effect on catalytic activity was reversible, CO oxidation was carried out continuously on the patterned Co 3 O 4 /Au inverse catalyst as the light was turned on and off at 543 K (Fig. 5d). The TOF value was similar to prior experiments (Fig. 5a): the catalytic activity was enhanced and then returned to the original catalytic activity as the halogen lamp was turned on and off, respectively, showing that the plasmonic hot carrier effect from the triangular Au nanostructure to the Co 3 O 4 was reversible. The small increase in the TOF value in the second cycle was the result of calcination of the sample during the reaction. Aer the reaction, the morphology of the patterned Co 3 O 4 /Au nanostructure had minor changes, but there were no signicant morphology changes, as veried by SEM images taken before and aer the CO oxidation reaction (Fig. S4 in ESI †). XPS measurements showed that the oxidation state of the cobalt oxide was a mixture of Co 2+ and Co 3+ , representing Co 3 O 4 . Aer the reaction, a higher percentage of Co 3+ was observed (Fig. S5 in ESI †).
To understand the electric eld distribution from the LSPR around the patterned Co 3 O 4 -Au nanostructure, we carried out two-dimensional FDTD simulation analysis. The electric eld distribution was obtained with a mesh size of 0.5 nm and incident light of 622 nm, which corresponds to the extinction maximum wavelength of the nanostructures from the UV-vis absorbance spectra. We set the refractive index of the Co 3 O 4 at 622 nm to 2.8, which was acquired experimentally, and the geometric model for the patterned Co 3 O 4 /Au nanostructure was built on experimentally observed structural parameters from SEM and TEM images. The calculated simulation results show an electric eld distribution that indicates an enhancement of light absorption for the top and cross-section views of the nanostructure. Fig. 6a shows that the electric eld was formed around the vertex points of the patterned Au nanostructure, which was oriented along the direction of polarization. In Fig. 6b, the electric eld was conned prominently along the gold edge between two adjacent Au triangles. In particular, a strong electric eld was formed at the interface between the Au and Co 3 O 4 due to the higher refractive index of the Co 3 O 4 . 41 For the patterned Au nanostructure without Co 3 O 4 , there was no electric eld on the top surface of the Au nanostructure and the electric eld was formed only along the side of the Au nanostructure (Fig. S6 in ESI †). This interesting LSPR conned at the Au-Co 3 O 4 interface can support the plausible hot electron transfer mechanism, indicating that the generated hot electrons at the Au edge or the Au-Co 3 O 4 interface were transferred to the Co 3 O 4 over the Schottky barrier and can easily transfer to the Co 3 O 4 surface owing to their long electron mean free path affecting the catalytic activity on the cobalt oxide.
CO molecules are known to be well adsorbed on Co 3 O 4 , especially on the Co 3+ site, and the mechanism for CO oxidation over Co 3 O 4 is believed to the Mars-van Krevelen type in which CO molecules adsorb on the Co 3 O 4 surface and react with lattice oxygen at Co 3 O 4 followed by O 2 lling in the oxygen vacancy of the Co 3 O 4 . 32, 42 We can explain that the enhancement of catalytic activity when under light irradiation is attributed to the Co 3 O 4 surface becoming negatively charged by hot electrons from LSPR excitation of the triangular Au nanostructure. Previous studies used ambient-pressure X-ray photoelectron spectroscopy (APXPS) and near edge X-ray adsorption ne structure spectroscopy (NEXAFS) to show that the main reaction intermediate for CO oxidation on Cu-based catalysts is adsorbed CO 2 dÀ . 43  ), which play important roles for CO 2 production from CO and oxygen via the Mars-van Krevelen type mechanism.
Another possible mechanism for explaining the enhancement of catalytic activity by LSPR of the triangular Au nanostructure under light irradiation is the partial reduction of Co 3+ to Co 2+ . Marimuthu et al. reported that selectivity for propylene oxide in propylene epoxidation increases on the Cu 2 O shellmetallic Cu core structure when under light irradiation. They found that Cu 2 O was reduced to metallic Cu, which is attributed to LSPR of the metallic copper under visible light irradiation using an in situ UV-vis extinction experiment. 47 Kim et al. demonstrated the reduction of Fe 3+ ions bound to a cyanide/Ag nanoparticle system induced by plasmonically generated hot electrons from the Ag nanoparticles. 48 Also, Li et al. reported the selectivity of benzyl alcohol oxidation to benzaldehyde on a gold core@ceria shell catalyst under visible light irradiation that resulted from reducing Ce 4+ to Ce 3+ by hot electron transfer from the LSPR of the gold core to the CeO 2 shell. 30 Shen and coworkers found that Co 3 O 4 nanorods with predominantly exposed [110] planes are very catalytically active for CO oxidation at low temperature, giving 100% CO conversion because of the presence of active Co 3+ species at the surface. This unexpected catalytic reactivity of Co 3 O 4 nanorods for CO oxidation is explained in the following mechanism: Co 3 O 4 has a spinel structure containing Co 3+ in an octahedral coordination and Co 2+ in a tetrahedral coordination. In these surface atomic congurations, there are only Co 2+ cations on the [001] and [111] planes, which are almost inactive, while the [110] plane is mainly composed of Co 3+ cations, which are regarded as the active site for CO oxidation. Thus, CO only adsorbs on the surface Co 3+ site and oxidation of the adsorbed CO then occurs by abstracting the surface oxygen that was coordinated with the Co 3+ cations. The Co 3+ cations are partially reduced to Co 2+ and, nally, the partially reduced cobalt oxide can be re-oxidized by a gas-phase oxygen molecule to become active Co 3+ cations again. 33 The reaction between the adsorbed CO on Co 3+ and the lattice oxygen is the overall rate-determining step in CO oxidation. 42 Broqvist et al. performed a rst-principles density functional theory (DFT) study regarding the CO oxidation mechanism at the Co 3 O 4 (110) surface and found that the reduction of Co 3+ to Co 2+ (Co 3+ + e À / Co 2+ ) in Co 3 O 4 (110) is crucial for the O abstraction step for CO 2 formation. 49 Here, in our patterned Co 3 O 4 -Au inverse catalysts, if a negatively charged hot electron from LSPR excitation on Au is transferred to the cobalt oxide, reduction of the cobalt oxide can occur, which is a decisive step for CO oxidation.
To nd the exact mechanism for the enhancement of catalytic activity at the cobalt oxide by hot electron transfer from the Au, theoretical DFT studies or in situ spectroscopic and microscopic experiments (e.g. ambient-pressure techniques) are needed for these catalysts under light irradiation. But our studies here on these reactive oxide/plasmonic noble metal inverse catalysts can connect the enhancement of catalytic activity to LSPR excitation from the metal, having possible applications for surface plasmon-based oxide/metal hybrid nanocatalysts or catalytic nanodiodes because of the long electron mean free path of the oxide materials.

Conclusions
In summary, Co 3 O 4 is known as a reactive oxide material for CO oxidation. Triangular noble metal nanostructures fabricated using the NSL technique exhibit higher SERS activities than at noble metals. We used cobalt oxide as the catalytic portion and a triangular Au nanostructure as the SPR source for the patterned Co 3 O 4 /Au nanostructure. The patterned Co 3 O 4 /Au inverse catalyst nanostructures were prepared using the NSL technique with SiO 2 nanospheres as the mask for metal deposition by electron beam evaporation. FDTD simulations showed that a particularly strong electric eld was formed at the Au-Co 3 O 4 interface because of the higher refractive index of the Co 3 O 4 . CO oxidation on the patterned Co 3 O 4 /Au nanostructure was carried out with and without light to nd the relationship between hot electron transfer to the reactive oxide from LSPR excitation on the triangular Au nanostructure and the change in catalytic activity for CO oxidation. The patterned Co 3 O 4 -Au inverse catalysts showed 25-50% higher catalytic activity for CO oxidation when under light irradiation than in the dark. The reversibility of the plasmonic hot carrier effect on catalytic activity was conrmed by observation of two catalytic activities with the light turned on and off, respectively. The change in catalytic activity implies that hot carriers generated from LSPR excitation of the Au nanostructure by absorption of photons on the Au metal can transfer to the Co 3 O 4 , and thus reversibly enhance the catalytic activity for CO oxidation.

Conflicts of interest
There are no conicts to declare.