Site-selective introduction of MnO₂ co-catalyst onto gold nanocubes via plasmon-induced charge separation and galvanic replacement for enhanced photocatalysis

When a nanoparticle exhibiting localized surface plasmon resonance (LSPR) is in contact with a semiconductor having an appropriate band structure, plasmon-induced charge separation (PICS) takes place at the interface.^1,2 PICS has been widely studied since it has a wide variety of applications, including photovoltaics and photocatalytic energy harvesting.² The charge separation mechanism is often explained in terms of a photoelectric effect, and it has been suggested that energetic electron–hole pairs generated through a plasmon decay process³ play an essential role in PICS. Electrons from the pairs are injected into the semiconductor conduction band and then transported to the semiconductor surface or a counter electrode to drive photocatalytic reduction reactions while the remaining holes are relaxed and the potential corresponding to the particle Fermi level shifts positively, resulting in photocatalytic oxidation reactions at the entire nanoparticle surface based on the charge accumulation mechanism.^4,5 Some oxidation reactions are driven by holes or trapped holes before relaxation, preferentially around the resonance sites of the metal nanoparticle. In the case of the latter, namely the hole ejection mechanism, the oxidation reaction site can thus be controlled by selecting the LSPR mode to be excited.⁴ 

Because LSPR confines light to a nanoscale area, which is smaller than the wavelength of light, PICS enables photo-induced nanofabrication beyond the diffraction limit.^6–15 We have reported a variety of plasmonic nanostructures fabricated via site-selective oxidation reactions, including Ag dissolution and PbO₂ deposition.^16–19 In the case of a plasmonic nanocube (NC) on a highly refractive substrate such as TiO₂, LSPR splits into two modes at different wavelengths, namely distal and proximal modes,²⁰ in which electrons oscillate at the top and bottom face of the NC, respectively. Selective excitation of one of those modes of an AgNC on TiO₂ results in the partial oxidation and dissolution of the AgNC preferentially at the corresponding resonance site through PICS.¹⁸ Such site-selectivity was also observed in the case of photo-oxidation of Pb²⁺ to PbO₂ on AuNCs.¹⁹ This leads to the fabrication of chiral plasmonic nanostructures by circularly polarized light in an enantioselective manner.^21,22 Thus, by taking advantage of PICS, the morphology of nanoparticles can be manipulated at the nanoscale, and the nanofabrication allows optical, electronic, and chemical properties of the particles to be modified. If a catalytic material is introduced to a plasmonic photocatalyst as a co-catalyst, its ability for plasmonic energy harvesting would be enhanced. Photocatalytic activity and specificity could be improved if the location of the co-catalyst is optimized in terms of mass transfer, efficient hole ejection and electron injection, and suppressed charge recombination. However, whether the hole ejection mechanism or the charge accumulation mechanism is chosen depends on the reaction involved in PICS.^4,5 Oxidative polymerization of pyrrole and coordinative dissolution of Au nanoparticles do not necessarily proceed site-selectively.⁵ Thus far, oxidative PbO₂ deposition is known as the only deposition reaction in which sufficient site-selectivity is achieved.^5,19,21,22

We, therefore, developed an indirect approach for the site-selective introduction of some other metal oxides that are difficult to introduce directly by the hole ejection mechanism. Because the redox potential of the Pb²⁺/PbO₂ couple is relatively positive (E^◦_Pb2+/PbO2 = +1.46 V vs NHE), PbO₂ can be substituted with a metal oxide (MO_x) spontaneously via galvanic replacement by exposure to Mⁿ⁺ (2x > n), whose redox potential is more negative than E^◦_Pb2+/PbO2. Actually, bulky PbO₂ film (12 µm thick) was replaced with MnO₂, Co₃O₄, and SnO₂ by galvanic replacement, namely simultaneous reductive PbO₂ dissolution and oxidative MO_x deposition.^23–25 However, for PbO₂ in a nanometer scale, it has not yet been reported whether the deposition site is retained or not during the replacement process. We, therefore, examine galvanic replacement of site-selectively deposited PbO₂ with MnO₂. MnO₂ is known as one of the co-catalysts for water oxidation,^26–29 and has been used as that for semiconductor photocatalysts^30–32 and plasmonic photocatalysts based on PICS.^33,34 Manganese oxide is also applicable to multi-electron oxidation reactions because of its positive charge accumulation ability.³⁵ For plasmonic photocatalysts, the optimal position of co-catalysts should be dependent on the locations of the semiconductor and the resonance sites. Therefore, the selective introduction of MnO₂ to one of the different resonance sites of plasmonic nanoparticles would lead to PICS systems with better efficiency and specificity as well as wider applications.

In the present study, we propose the indirect approach based on the site-selective deposition of PbO₂ onto Au nanocubes via PICS and its galvanic replacement with an MnO₂ co-catalyst and fabricate efficient plasmonic photocatalysts (Fig. 1). The relationship between the MnO₂ deposition site and photocatalytic activity is investigated for the novel structural design of PICS photocatalysts. This approach would be applied to the fabrication of plasmonic nanostructures with many different materials and morphologies.

AuNCs with the size of 80 nm [Fig. S1(a)] were deposited onto a TiO₂ thin layer (thickness ∼40 nm) formed on an indium tin oxide (ITO) electrode.¹⁹ The distal and proximal modes in which electrons oscillate around the top and bottom face of the NC were observed at ∼550 and ∼660 nm, respectively, in the extinction (= absorption + scattering) spectrum [Fig. S1(b)]. The extinction of ITO was subtracted from the spectrum. The AuNC-modified TiO₂ electrode as the working electrode (electrode area ∼2.2 cm²) was immersed in 0.05 M Pb(NO₃)₂ and a bare ITO electrode as the counter electrode was dipped in 0.05 M Ag(NO₃)₂. The electrolytes were connected through a salt bridge. The two electrodes were short-circuited and the working electrode was irradiated with 560 nm light (FWHM = 20 nm, ∼10.2 mW cm⁻², front incidence) or 625–800 nm light (∼136 mW cm⁻², back incidence) for 48 h to introduce PbO₂ onto the distal or proximal site, respectively [Figs. 2(a) and 2(b)].¹⁹ We also prepared a sample with AuNCs entirely coated with PbO₂ by exciting both the distal and proximal modes with 480–800 nm light (∼30.1 mW cm⁻², back incidence) [Fig. 2(c)].

The PbO₂-deposited electrode was immersed in an aqueous solution containing 0.02 M Mn(CH₃COO)₂ and 0.05 M CH₃COOH (pH 5.3) for 24 h at room temperature. Although the morphology of the deposits was changed, the deposition site was almost maintained [Figs. 2(d)–2(f)]. We analyzed chemical composition changes after the galvanic replacement treatment of PbO₂ at the distal sites by X-ray photoelectron spectroscopy (XPS). The spectrum obtained before the replacement showed Pb 4f_7/2 and Pb 4f_5/2 peaks [Fig. S2 and Fig. 3(a)]. The former can be deconvoluted into two peaks at 137.4 and 138.5 eV and the latter into those at 142.3 and 143.4 eV. Regarding the Pb 4f_7/2 peaks, it has been reported that PbO₂ and PbO show peaks at 137.2–137.6 eV^37–39 and 137.7–137.9 eV,^36,37,40 respectively, while that for metallic Pb is observed at lower binding energies.³⁶ It is also known that electroconducting PbO₂ shows a plasmon loss satellite at ∼1 eV higher energy than the pristine peak.^41,42 We, therefore, assigned the deconvoluted Pb 4f_7/2 peaks at 137.4 and 138.5 eV to the pristine and satellite peaks, respectively, of PbO₂. After galvanic replacement, we observed a 10-fold lower Pb 4f_7/2 signal at 137.9 eV [Fig. 3(a)], which is assigned to PbO. Also, Mn2p signals appeared at 642.0 and 653.5 eV [Fig. 3(b)]. These peaks can be assigned to MnO₂.⁴³ We, therefore, conclude that most of the introduced PbO₂ was replaced with MnO₂, and the residual PbO₂ was reduced by Mn²⁺ to PbO.

The samples were also subjected to elemental analysis by energy dispersive X-ray spectroscopy (EDS). The atomic ratio of Pb to Au at the electrode surface before the replacement was determined to be 0.35, and that of Mn after the replacement was 0.15. The ratios of Mn before replacement and Pb after replacement were lower than the detection limit (∼0.04). About half of the deposited oxides were lost during the galvanic replacement process, likely due to detachment induced by structural changes. Taking the detachment into account, the initial amount of PbO₂ should be about twice as much as the targeted amount of MnO₂.

The TiO₂ electrode with AuNCs was subjected to a photocurrent measurement, followed by PbO₂ deposition and galvanic replacement with MnO₂, with a photocurrent measurement after each step. Photocurrents based on the oxidation of ethanol (0.5 M) were measured in an aqueous 0.1 M KNO₃ electrolyte using the working electrode irradiated with monochromatic light (FWHM = 20 nm, ∼3.3 × 10¹⁵ photons cm⁻² s⁻¹) for 10 s at each wavelength from 790 to 430 nm with a 30 s interval. A Pt counter electrode was short-circuited with the working electrode. Action spectra for the external quantum efficiency (EQE) of the measured photocurrents are shown in Fig. S3. All the electrodes exhibited anodic photocurrent responses in the wavelength range examined. Those in the >500 and <500 nm ranges are derived from plasmonic absorption and interband transition, respectively. The former photocurrents are explained in terms of PICS: Charge separation occurs by electron injection from AuNCs into the TiO₂ conduction band, resulting in oxidation of ethanol at the NCs and reduction of dissolved oxygen at the counter electrode, whose potential is negatively shifted due to the injected electrons. Photocurrents at around 800 nm or longer were relatively small even though AuNCs absorb light. This phenomenon, which is often observed for PICS, is attributed to a decreased population of sufficiently energetic electrons to overcome the Au–TiO₂ Schottky barrier (energy height ∼0.9 eV⁴⁴) in the long wavelength range.

Here we define the EQE enhancement factor (EF_EQE) as the ratio of EQE after the MnO₂ introduction to that before PbO₂ deposition (AuNC only). The EF_EQE values were plotted as a function of the excitation wavelength in Fig. 4(a). After PbO₂ deposition, EQE decreased in almost all of the wavelength ranges, regardless of the deposition site (Fig. S3). Although PbO₂ is electroconducting, the conductivity is lower than Au, and it has not been known as a good co-catalyst. After replacement of PbO₂ with MnO₂, however, EQE in the 500–700 nm range increased up to two times when MnO₂ was introduced, chiefly to the top part of the AuNC, reflecting that the ethanol oxidation via PICS was enhanced by the co-catalytic effect of MnO₂. In contrast, when MnO₂ was introduced to the bottom part of the AuNC or the entire AuNC surface, the EQE values were lowered, indicating that the co-catalytic effect in PICS is susceptible to the location of co-catalysts.

Here we explain the mechanisms for the susceptibility using the electronic structure and schematic illustrations in Figs. 4(b)–4(e). In order to determine the potential of MnO₂ for hole acceptance, MnO₂ was deposited on an ITO electrode by the electrodeposition of PbO₂ at +1.50 V vs NHE and the galvanic replacement under the same conditions. Its potential for the Mn(IV/V) redox reaction⁴⁵ was determined by cyclic voltammetry to be ∼+1.1 V vs NHE (Fig. S4), and the potential is shown in Fig. 4(b). In the case where MnO₂ is introduced to the top part, electrons excited in the AuNC are injected into the TiO₂ conduction band, while holes or positive charges accumulated via hole relaxation are used for oxidation of Mn(IV) in MnO₂ to Mn(V) [Fig. 4(b)]. Then, Mn(V) oxidizes an electron donor in the solution, such as ethanol or water, and Mn(V) is reduced back to Mn(IV). Namely, MnO₂ serves as a co-catalyst through the accumulation and transfer of positive charges to the donor. Incidentally, the distance between the top part of the nanocube and TiO₂ is comparable to the mean free path length of ballistic electrons in Au that can cross the Schottky barrier at the Au–TiO₂ interface (∼42 nm).⁴⁶ 

Since the MnO₂ moiety is spatially separated from TiO₂, recombination of the positive charges in MnO₂ with the negative ones in TiO₂ is suppressed and the photocatalytic oxidation at MnO₂ is enhanced, resulting in improved PICS efficiency [Fig. 4(d)]. Although Fig. 4(b) illustrates the separation of one electron–hole pair, it is also possible that two electron–hole pairs lead to a single charge separation process (Fig. S5).⁴⁷ Even when MnO₂ is introduced to the bottom part of each AuNC or the entire NC surface, electrons and holes should be injected into TiO₂ and MnO₂, respectively. However, recombination of the separated charges is likely to take place, since MnO₂ is in direct contact with TiO₂ [Fig. 4(e)]. Since the TiO₂ layer thickness (∼40 nm) is much smaller than the space charge layer thickness for bulk TiO₂ (typically ∼200 nm) and electrons accumulate in TiO₂, the energy barrier at the TiO₂–MnO₂ interface, if any, should be low. The recombination could be the reason why the PICS efficiency decreased at >500 nm for the case where MnO₂ is introduced to the bottom part or the entire surface of AuNCs.

Incidentally, EQE was enhanced in the <500 nm range, in particular, in the case where MnO₂ was in contact with TiO₂. This is probably because a partially reduced state of MnO₂, namely Mn₂O₃,⁴⁸ on the AuNC works as a photosensitizer of TiO₂ [Fig. 4(c)]. MnO₂ itself does not play a role as a sensitizer because the conduction band edge potential of MnO₂ is more positive than that of TiO₂,^50–51 whereas that of Mn₂O₃ (−0.35 to 1.1 V vs NHE^52,53) is more negative, so that photoexcited electrons in the Mn₂O₃ conduction band can be injected into the TiO₂ conduction band. Although Mn(III) constituting Mn₂O₃ was not detected by XPS after the galvanic replacement, trace amounts of Mn₂O₃ less than the detection limit or Mn₂O₃ reduced from MnO₂ by photoexcited electrons in the AuNC may contribute to the sensitization. Thus, it was indicated that controlling the co-catalyst site is of critical importance for the efficiency improvement of PICS-based photocatalysis.

In conclusion, MnO₂ co-catalyst was introduced site-selectively to AuNCs on TiO₂ via site-selective PbO₂ deposition based on PICS and subsequent galvanic replacement. Photoelectrochemical oxidation of ethanol through PICS was performed using the electrode with AuNC–MnO₂ nanostructures thus obtained. The PICS efficiency was improved only when MnO₂ was deposited on the top part of AuNCs. In contrast, the efficiency decreased in the case of MnO₂ in contact with TiO₂, for example, MnO₂ introduced to the bottom part of NCs. Thus, we demonstrated that the PICS system is susceptible to the location of co-catalysts and that it is important to control the location. The present approach for controlling the deposition sites is expected to be applicable to plasmonic nanoparticles with different materials and morphologies, leading to the development of various plasmonic nanomaterials and nanodevices.

See the supplementary material for an SEM image, an extinction spectrum, XPS spectra, photocurrent EQE spectra, a cyclic voltammogram, and a schematic band diagram of the samples.

This work was supported, in part, by a Grant-in-Aid for Scientific Research (A) (Grant No. JP20H00325) and a Grant-in-Aid for Scientific Research (B) (Grant No. JP20H02706) from the Japan Society for the Promotion of Science (JSPS).