A soft-chemistry assisted strong metal–support interaction on a designed plasmonic core–shell photocatalyst for enhanced photocatalytic hydrogen production

Getaneh Diress Gesesse a, Cong Wang a, Bor Kae Chang b, Shih-Hsuan Tai b, Patricia Beaunier c, Robert Wojcieszak d, Hynd Remita a, Christophe Colbeau-Justin a and Mohamed Nawfal Ghazzal *a
aInstitut de chimie Physique, UMR 8000 CNRS, Université Paris-Saclay, 91405, Orsay, France. E-mail: mohamed-nawfal.ghazzal@universite-paris-saclay.fr
bDepartment of Chemical and Materials Engineering National Central University, No. 300, Jhongda Rd., Jhongli District, Taoyuan City 32001, Taiwan
cSorbonne Université, CNRS, Laboratoire de Réactivité de Surface, F-75005 Paris Cedex 05, France
dUniv. Lille, CNRS, Centrale Lille, ENSCL, Univ. Artois, UMR 8181 – UCCS – Unité de Catalyse et Chimie du Solide, F-59000 Lille, France

Received 20th November 2019 , Accepted 27th January 2020

First published on 28th January 2020

Engineering photocatalysts based on gold nanoparticles (AuNPs) has attracted great attention for the solar energy conversion due to their multiple and unique properties. However, boosting the photocatalytic performance of plasmonic materials for H2 generation has some limitations. In this study, we propose a soft-chemistry method for the preparation of a strong metal–support interaction (SMSI) to enhance the photocatalytic production of H2. The TiO2 thin overlayer covering finely dispersed AuNPs (forming an SMSI) boosts the photocatalytic generation of hydrogen, compared to AuNPs deposited at the surface of TiO2 (labelled as a classical system). The pathway of the charge carriers’ dynamics regarding the system configuration is found to be different. The photogenerated electrons are collected by AuNPs in a classical system and act as an active site, while, unconventionally, they are injected back in the titania surface for an SMSI photocatalyst making the system highly efficient. Additionally, the adsorption energy of methanol, theoretically estimated using the density functional theory (DFT) methodology, is lower for the soft-chemistry SMSI photocatalyst accelerating the kinetics of photocatalytic hydrogen production. The SMSI obtained by soft-chemistry is an original concept for highly efficient photocatalytic materials, where the photon-to-energy conversion remains a major challenge.


In the last decade, hydrogen, as a chemical fuel, has become an alternative solution to fossil sources of energy. Hydrogen is usually produced from a variety of feedstocks, including fossil resources, such as natural gas and coal. A large variety of processes are used for its production including chemical, biological, electrolytic, photolytic and thermo-chemical methods. The photocatalytic production of hydrogen from aqueous alcohol solutions, such as methanol and ethanol, appears as a viable and attractive process based on the conversion of sustainable solar energy to chemical fuel. Since the pioneering work of Kawai et al.1 and Courbon et al.,2 many attempts have been made to increase the photocatalytic efficiency of titanium dioxide for hydrogen production. The most promising work involves plasmonic gold nanoparticles (AuNPs), which act as a co-catalyst, charge carrier separator and nanoantenna for visible light harvesting.3–12 Therefore, the size, shape and loading of AuNPs appeared as important parameters to be considered, since they influence the photocatalytic performance.3,4,13–15 The photocatalysts’ configuration also had an impact on the photocatalytic activity. Different designs, such as Au@TiO2 core–shell nanoparticles14,16–19 or Au/TiO2 Janus nanoparticles5,6 or 3D architecture,12,20 were proposed as an efficient strategy to achieve higher photoefficiency. In all cases, the improvement of the photocatalytic performance was attributed in general to the formation of the Schottky contact, crucial for efficient electron transport and charge carrier separation,8,21 or to the enhancement of the localized surface plasmon resonance intensity (LSPR), rising from the oscillation of electrons at the same frequency in the AuNP surface, by either increasing the size of Au or TiO2 nanoparticles.5,6,12,19,22 These studies were carried out regardless combining the effect of the charge carriers’ dynamics, the dissociative adsorption of methanol occurring during the photocatalytic production of hydrogen and more importantly the metal–support interaction effect.

The concept of a strong metal–support interaction (SMSI) was introduced in the 1970s by Tauster et al.23 The fundamentals of the SMSI method used a high-temperature H2 reducing environment, which induces the reduction of a metal oxide and its migration to the top of the metal nanoparticles (encapsulation of the metallic nanoparticles). The method showed some stability issues, since the re-oxidation or receded SMSI overlayer structures during the catalytic reaction induces inactivation of the catalyst and/or reduction of its selectivity.24–26 A revisited SMSI method has emerged claiming higher stability of the overlayer metal oxide and good activity for CO2 reduction.27 The method proposed the use of HCOx adsorbate, which mediated the metal oxide overlayer formation at the surface of Rh nanoparticles. Most of the time, the SMSI was restricted to metals from group VI, VII and VIII supported on reducible oxides such as C3N4, TiO2, V2O3, Nb2O5, Ta2O5, and CeO2.26–31 AuNPs were believed to be excluded from forming an SMSI with titanium dioxide due to its low work function, surface energy and failure to dissociate molecular hydrogen necessary for the oxide migration to the metal surface.32,33 However, recent theoretical and experimental studies demonstrated that gold can dissociate H2 and hence, promote the reduction of oxide supports.34–36 Hence, a classical SMSI33 and more recently wet-chemistry SMSI were proposed for depositing substoichiometric TiOx at the surface of AuNPs for CO oxidation.37

In general, the development around the SMSI concept was restricted to heterogeneous catalytic redox reactions. The use of the SMSI concept in photocatalytic hydrogen generation has not yet been considered for AuNPs. While this method is suitable for catalytic reactions, it seems less applicable in heterogeneous photocatalysis. Indeed, amorphous TiO2 or non-stoichiometric titanium oxide is often obtained under reductive conditions and high temperature treatment, which do not correspond to the preferred anatase photoactive phase. Furthermore, the existence of a consensus regarding the location (on the top of the photocatalyst surface) and the role played (electron collector) by AuNPs meant that the concept was not taken into account.38,39 Additionally, the photocatalytic process occurring during the hydrogen production is managed by several parameters such as photogenerated charge carrier dynamics and lifetime, methanol adsorption/desorption, light harvesting, band-gap energy, TiO2 crystallite phase, etc.

In this work, the soft-chemistry SMSI is considered for the first time for photocatalytic hydrogen production using a plasmonic photocatalyst. The hydrogen production efficiency was assessed for AuNPs encapsulated within a TiO2 overlayer getting a SMSI using the soft-chemistry method (SiO2@Au@TiO2) and compared to AuNPs grown at the surface of a silica–titania core–shell system (SiO2@TiO2@Au, labelled as a conventional system) (Scheme 1). The photoactivity of the core–shell system was found to be enhanced when AuNPs were covered by a thin layer of TiO2, following at the same time the evolution of the LSPR intensity. The dynamics of charge carriers studied by time resolved microwave conductivity indicates variable pathways suggesting different mechanisms of hydrogen production. Finally, the dissociative adsorption energy of methanol at the surface of each system was theoretically simulated by DFT and a mechanism of the hydrogen production was proposed in order to explain the global kinetics of the hydrogen production rates.

image file: c9nr09891g-s1.tif
Scheme 1 Scheme of different steps of the synthetic procedure of plasmonic core–shell nanoparticles. AuNPs were loaded on the synthesized SiO2@TiO2 core–shell or SiO2 core with a weight ratio (Au/TiO2) of 0.25, 0.5, 1, and 1.5 wt%.

Experimental methods


All reagents (analytical grade) were used without further purification. Titanium isopropoxide (TTIP, 97% Sigma-Aldrich), 3-aminopropyltrimethoxy silane (APTMS), tetraethyl ortho silicate (TEOS, Sigma-Aldrich), chloroauric acid trihydrated (HAuCl4·3H2O, Sigma-Aldrich), tetrakis (hydroxylmethyl)phosphonium chloride (THPC, 80% in water, Sigma-Aldrich), and ammonium hydroxide (NH4OH, Sigma-Aldrich). Furthermore, Milli-Q water and absolute ethanol (99%, Aldrich) were used as solvents.

Materials synthesis

The synthesis of core–shell nanoparticles is described in detail. The Stöber method was used to prepare spherical silica nanoparticles. A mixture of 20 mL of Mill-Q water and 15 mL of absolute ethanol was stirred for 10 minutes at room temperature. 5 mL of NH4OH was used to adjust the pH and the temperature was increased to 40 °C under magnetic stirring (10 minutes). Then, TEOS (4.5 mL) was added dropwise to the solution, and aged under magnetic stirring for 1 hour. The white precipitate was obtained and separated by centrifugation. The solid was washed 3 times with Milli-Q water and dried in an oven at 70 °C overnight.

The functionalization of SiO2 nanoparticles was performed as follows: SiO2 (250 mg) nanoparticles were added to ethanol (10 mL) while stirring vigorously, followed by sonication for 30 minutes to completely disperse the microspheres. While the colloidal solution was stirred, 0.1 mL of APTMS was added dropwise and was stirred (500 rpm) for 6 h at room temperature. The APTMS functionalized silica nanoparticles were collected from the colloidal solution by centrifugation and washed 3 times with absolute ethanol.

The synthesis of SiO2@TiO2 core–shell particles was performed as follows: the APTMS functionalized SiO2 nanoparticles were added to a solution of 20 mL absolute ethanol and 0.5 mL of water while stirring vigorously. The nanoparticles were dispersed under sonication for 30 minutes. Concentrated ammonium hydroxide solution (0.3 mL, 25 wt%) was used to adjust the pH, which was recorded (stabilized at 12.4). The solution was stirred for 20 minutes, and a mixture of 50 mL of absolute ethanol and 1.5 mL TTIP was added dropwise and left for 2 hours under stirring (1000 rpm) at room temperature. The concentration of TTIP reported in this section corresponds to the optimal amount leading to an optimal thickness of the TiO2 shell for the best photoactivity. The coated colloidal spheres were separated by centrifugation and washed with absolute ethanol three times. The core–shell nanoparticles were dried in an oven at 80 °C overnight.

Gold nanoparticles (AuNPs hereafter) were loaded on SiO2@TiO2 or embedded between SiO2 and TiO2. Scheme 1 illustrates all the steps of the synthetic procedure of core–shell nanoparticles. AuNPs were loaded on the synthesized SiO2@TiO2 core–shell particles according to the Duff and Baiker method, to obtain a weight ratio of Au/TiO2 of 0.25, 0.5, 1, and 1.5 wt%. First, the synthesized SiO2@TiO2 nanospheres were further functionalized with APTMS, as described above. The APTMS functionalized SiO2@TiO2 core–shell particles were separated by centrifugation followed by washing with ethanol 3 times. Another gold containing solution was prepared separately as follows: 1.5 mL (0.2 mol L−1) NaOH solution was added to 45 mL of water while stirring vigorously. After 5 minutes of stirring, a solution of THPC (1 mL, 0.05 mol L−1) was added to the prepared solution, followed by addition of an adequate amount of HAuCl4·3H2O to the mixture as a gold source to obtain a weight ratio of 0.25, 0.5, 1, and 1.5 wt% respectively. The appearance of red-purple colour indicates the reduction of Au(III) to Au(0). The solution was stirred for 30 minutes. Then, APTMS modified SiO2@TiO2 core–shell particles were dispersed in a solution containing gold and stirred for 4 hours to control the areal density of AuNPs on TiO2. The obtained solid material was collected by centrifugation, washed with water 3 times and dried at 60 °C for 48 hours. The synthesized product was assigned as SiO2@TiO2@Au.

A similar procedure was used to synthesize the SMSI SiO2@Au@TiO2 system, but this time by using silica nanoparticles loaded with an adequate amount of AuNPs (0.25, 0.5, 1 and 1.5 wt%) as the starting material. APTMS functionalized SiO2 core nanoparticles were dispersed in a solution containing the target gold ratio. SiO2@Au nanoparticles with different loading ratios were collected by centrifugation until the supernatant became colorless. The solids were washed 3 times with absolute ethanol and dried at 60 °C for 48 h. APTMS functionalized SiO2@Au nanoparticles were dispersed in a mixture of (20 mL) absolute ethanol and (0.5 mL) Milli-Q water followed by sonication. Then, 0.3 mL of concentrated NaOH was added to adjust the pH at 12.4. The solution was stirred for 20 minutes, and a mixture of (50 mL) absolute ethanol and (1.5 mL) TTiP was added dropwise. The solution was left under agitation for 2 hours (1000 rpm) at room temperature. The nanoparticles were collected by centrifugation, washed 3 times with absolute ethanol and dried at 80 °C overnight. All the plasmonic photocatalysts were calcined at 500 °C for 2 hours under air.


The diffusion reflectance spectra (DRS) of the core–shell nanocomposites samples were collected using a Cary-5000 spectrophotometer (Agilent) equipped with a Cary 4/5 diffuse reflection sphere. The baseline was recorded using a poly(tetrafluoroethylene) reference. Transmission electron microscopy (TEM) observations and EDS X-ray microanalysis were carried out on a JEOL JEM 2100 Plus transmission electron microscope, operating at 200 kV, interfaced to the Oxford Instruments AZtec EDS system with an X-Max T large area (80 mm2) SDD detector. The images were collected with a 4008 × 2672 pixels CCD camera (Gatan Orius SC1000). TEM images were processed by Image J software. The size estimation and their size distribution were estimated by taking into account 150–250 AuNPs.

ICP-OES (Inductively Coupled Plasma Optical Emission Spectrometry) analysis was performed using Agilent 720-ES ICP-OES equipment combined with a Vulcan 42S automated digestion system. Vulcan 42S is an automated digestion system combining the two essential steps in sample preparation prior to analysis by ICP: sample digestion followed by sample work-up. The digestion procedure was as follows: firstly, 10 mg samples were weighed and 2.4 mL of aqua regia and 1 mL of HF were added to each sample by a robot and then heated for 2 hours up to 110 °C (this step is repeated 3 times), followed by 1 hour heating up to 110 °C. Almost all the digester components are made of Teflon to avoid corrosion with the use of acids. Questrom uses a highly efficient fume removal after neutralizing the acid vapour thereby avoiding cross-contamination.

Photocatalytic hydrogen production from an aqueous methanol solution was performed in a closed Pyrex glass reactor under an argon atmosphere and with vigorous stirring. In a typical photocatalytic test, a water/methanol 75/25 v/v (final volume was 20 mL) solution was stirred under a flow of nitrogen for 30 minutes to eliminate dissolved oxygen. Then, 20 mg of the photocatalyst were added to the reaction medium. The solution is subsequently illuminated using a mercury lamp (150 W) light source as an artificial source to simulate the entire solar spectrum. The hydrogen production was simultaneously followed each hour by gas chromatography on a Bruker Scion gas chromatograph, with a thermal conductivity detector (column, molecular sieve 5 A, 75 m × 0.53 mm i.d. oven temperature at 50 °C; flow rate = 22.5 mL min−1; detector temperature = 250 °C; carrier gas, nitrogen). Visible light experiment was performed using a cut-off filter (AM-32603-1, LOT-Oriel; λ > 420 nm).

After UV or visible illumination, the charge carrier lifetimes of plasmonic core–shell nanoparticles were studied by the Time Resolved Microwave Conductivity method (TRMC). For TRMC measurements, the incident microwaves were generated by a Gunn diode of the Kα band at 30 GHz. The pulsed light source was an OPO laser (EKSPLA, NT342B) tunable from 225 to 2000 nm. It delivers 8 ns FWMH pulses with a frequency of 10 Hz. The density of light energy received by the sample was 2.3 mJ cm−2 at 350 nm and 3.7 mJ cm−2 for λ = 550 nm. To minimize the noise, a TRMC signal is obtained by averaging measurements during 200 laser pulses. The TRMC technique is based on the measurement of the relative change of the microwave power reflected by a sample (semiconductor), during its simultaneous irradiation by a laser pulse. The TRMC technique has been described in more details elsewhere.11,40

Computational methodology

Density functional theory (DFT) calculations at the generalized gradient approximation (GGA) level using the CASTEP plane-wave code41 were performed with the Perdew–Burke–Ernzerhof (PBE) density functional42 and ultrasoft pseudopotentials. These are prevalent in the literature and have been recently used in similar bulk TiO2 calculations.43 The plane-wave basis set cutoff energy was set at 620 eV, while the Γ-point was sampled in the k-space because of the large supercell models used. Convergence criteria used during full geometry optimization of the TiO2 unit cell include changes in the energy of less than 10−5 eV per atom, forces on each atom smaller than 0.03 eV Å−1, the final stress in the unit cell of less than 0.05 GPa, and atomic displacements smaller than 0.001 Å. 3 × 3 × 2 supercell models of anatase TiO2 were constructed based on experimental data,44 which were then expanded to 7 × 1 × 7 in order to obtain a big enough structure to create an anatase (101) surface with a 15 Å vacuum layer. The final surface and bulk model contained 96 oxygen atoms and 48 titanium atoms. The silica substrate is not considered in our models, since it does not participate in reactions, thus SMSI SiO2@TiO2@Au and SiO2@Au@TiO2 were modeled as TiO2 with a small gold cluster on the surface or reversed with the gold cluster embedded below. Au3 gold clusters were added in our system of anatase (101) surfaces based on previous literature,44,45 and all atomic positions were fully optimized again without relaxing the cell parameters. The DFT-D3 dispersion correction method by Grimm,46 which has parameters for gold, was implemented in this study after gold clusters and later methanol molecules were included to correct for London dispersion in this adsorption scenario.

Methanol adsorption was considered on both sides of the model described above to simulate the two synthesized materials. During geometry optimization, the bottom layer of the anatase cell was fixed to simulate bulk properties and the top four layers were relaxed. In the case of Au@TiO2, both the bottom layer and with the attached gold cluster were fixed during the calculations. The initial distance between methanol and the anatase (101) surface or gold cluster is 2.22 Å. The adsorption energy of a single methanol molecule is calculated as follows:

Eads = Emethanol+Au@TiO2 − (EAu@TiO2Emethanol)
Eads = Emethanol+TiO2@Au − (ETiO2@AuEmethanol).

Results and discussion

The morphology and microstructure of SMSI SiO2@Au@TiO2 and SiO2@TiO2@Au nanocomposites were investigated by TEM as depicted in Fig. 1. The images exhibit core–shell nanostructures with a distinct high-density silica core, much darker than the low-density titania shell. The Stöber method enables the synthesis of spherical SiO2 nanoparticles with an average diameter of 50–70 nm. The hydrolysis followed by the polycondensation of the TTIP precursor in an alkaline ethanoic mixture (pH = 12) leads to a uniform TiO2 shell covering the silica core. The APTMS introduces amine terminal groups at the silica surface providing nucleation sites where the TiO2 thin shell can homogeneously grow during the sol–gel process. The TiO2 shell thickness, estimated from the HR-TEM images, is approximately in the range of 4–10 nm (Fig. 1a). The high magnification TEM image evidences a crystallized shell with an interplanar spacing of 0.342 nm, consistent with the (101) crystallographic plane of TiO2 anatase (magnification of the dotted square in Fig. 1b). The fast Fourier transform pattern showed a TiO2 anatase thin layer surrounding the (111) face of AuNPs separated with an interplanar spacing of 0.23 nm (magnification of the white square in Fig. 1b). UV-visible spectra and X-ray diffraction patterns (00-021-1272 JCPDS card number), in agreement with the HR-TEM analysis, confirmed the anatase form of the TiO2 shell after calcination at 500 °C for both systems (ESI Fig. S1). The UV-visible spectra of the core–shell nanostructure showed a typical anatase absorbance band initiated at approximately 390 nm, which corresponds to the bandgap energy of 3.2 eV.
image file: c9nr09891g-f1.tif
Fig. 1 HR-TEM image of nanocomposites (a) SiO2@TiO2@Au with 0.5 wt% ratio and (b) SiO2@Au@TiO2 with 1 wt% ratio, included fast Fourier transform of a rectangular region in HR-TEM. TEM images of (c–f) SiO2@TiO2@Au and (g–j) SiO2@Au@TiO2 systems with a variable AuNPs loading: (c and g) 0.25 wt%, (d, h) 0.5 wt%, (e and i) 1 wt%, (f and j) 1.5 wt%. The size distribution of Au nanoparticles calculated from the TEM images is also presented respectively for each Au/TiO2 ratio.

The reduction of the HAuCl4·3H2O precursor occurred efficiently, resulting in the formation of hemispherical AuNPs. Further TEM images of AuNPs deposited at the surface of TiO2 are reported in the ESI Fig. S2 for each gold ratio. The AuNP weight ratio as estimated by ICP analysis was 0.13%, 0.53%, 1.13%, and 1.56% is in agreement with the theoretical ratio, except for a lower ratio. The TEM images reveal that the AuNPs exhibit a narrower size distribution with an average size ranging from 4.5 to 7 nm.

More TEM investigations were carried out evidencing the effective SMSI through the formation of a thin TiO2 overlayer covered on the AuNPs (Fig. 2a–d). The TiO2 shell with a thickness of 4–10 nm enables embedding AuNPs inside of an oxide overlayer of 2–3 nm. The APTMS functionalized AuNPs provide growth sites where the TiO2 thin shell is readily formed.

image file: c9nr09891g-f2.tif
Fig. 2 TEM images of the SMSI SiO2@Au@TiO2 nanostructure with increasing gold ratio: (a) 0.25 wt%, (b) 0.5 wt%, (c) 1 wt% and (d) 1.5 wt%.

To proceed further with the microstructure investigations, HAADF-STEM images and EDS chemical mapping on SiO2@Au@TiO2 with 0.5 wt% ratio were performed to obtain more information on the distribution of Si, O, Ti and Au elements (Fig. 3 and ESI Fig. S3 for 1.5 wt% gold ratio). Despite the low thickness of the TiO2 shell, the Ti element uniformly distributed on the surface could be observed. The obtained images confirmed the core–shell structure. The highly dispersive Au element distribution confirms isolated AuNPs on the surface of silica.

image file: c9nr09891g-f3.tif
Fig. 3 (a and b) HAADF-STEM images and EDS elemental mapping of Si (K), Ti (K), O (K) and Au (L) for the SMSI SiO2@Au@TiO2 nanostructure with 0.5 wt% ratio.

Additionally, X-ray photoelectron spectroscopy analysis was performed to study the surface composition of both samples (ESI Fig. S4). The general spectra of the two samples are very similar. The elements detected at the surfaces are Ti, Si, O, C and Au. The Ti 2p3/2 and Ti 2p1/2 spin–orbit components were well resolved and centered at 458.7 and 464.6 eV. These two peaks are separated by 5.7 eV, in agreement with the presence of Ti4+ in a tetragonal structure suggesting an anatase TiO2 crystalline form.47 The Si 2p peak is assigned to the Si–O bond present in the SiO2 core of the nanostructure. The 4f7/2 orbital of gold was recorded as it provides a convenient and sensitive measure of the electronic state of Au. The 4f7/2 peak falls in the binding energy of 83.5 eV with 3.7 eV separation to the 4f5/2 peak for both samples, indicating that the gold standard is bulk gold (i.e. Au0).

The LSPR properties of SiO2@TiO2@Au and SMSI SiO2@Au@TiO2 nanocomposites with increasing Au loading were studied by steady-state UV-visible diffuse reflectance spectroscopy within the 400–800 nm wavelength range (Fig. 4a and b). Core–shell nanostructures containing AuNPs exhibited LSPR in the visible range. However, the gold-free samples (bare SiO2@TiO2) showed no absorption in the visible region. TiO2 in the anatase form (Eg = 3.2 eV) absorbs only in the UV region. AuNPs deposited at the SiO2 surface show a broadband absorption with the maximum at 520 nm (data not shown). After the TiO2 thin shell was coated, the maximum wavelength of the LSPR band was red-shifted to 541 nm for the lower Au/TiO2 ratio. The LSPR wavelength maximum of AuNPs is related to their shape, size and dielectric constant (refractive index) of the surrounding medium. The size distribution of AuNPs, as estimated from the TEM images, indicates very small changes of the AuNP size distribution regardless of the loading. Therefore, following up on our procedure, the small change of the nanoparticle diameter could not explain the observed red-shift. Indeed, tuning the size of AuNPs needs more sophisticated methods such as the use of stabilizers48 or a complex growth method.49 The AuNP spatial separation is too large avoiding any strong near-field interaction that could rationalize the observed red-shift.19 This red-shift arises due to an overall increase in the refractive index of the dielectric environment surrounding the AuNPs. Comparing both systems, in the case of the classical sample (SiO2@TiO2@Au) the AuNPs are surrounded on the one side by TiO2 with a high refractive index (nTiO2 = 2.1 (ref. 47)) and from the other side by air (nair = 1). In the SMSI SiO2@Au@TiO2 system, the AuNPs are embedded between TiO2 and SiO2 (nSiO2 = 1.4 (ref. 50)). For both systems, the LSPR maximum wavelength is the same, even if the surrounding media are not, probably because of the small difference between the air and SiO2 refractive indexes.

image file: c9nr09891g-f4.tif
Fig. 4 Experimental absorption spectra recorded in the visible range for core–shell nanocomposites for both systems and at a variable AuNP weight ratio. As a control, the SiO2@TiO2 core–shell shows no absorption in the 400–800 nm wavelength range.

Worth mentioning is the change of the LSPR intensity in each system. As the AuNP loading increases, the LSPR intensity shows an increase for the classical SiO2@TiO2@Au system, following the AuNP ratio (Fig. 4a). The same observation was made when AuNPs are embedded between SiO2 and TiO2, but the enhancement of the LSPR absorption band is much higher. The enhancement is 1.7 times higher for 1 wt% AuNP ratio before it drastically decreases (Fig. 4b). The decrease of the LSPR intensity is unexpected since usually the collective oscillation of the electrons after excitation usually tends to increase with the gradual increase of the AuNP loading. The decrease of LSPR intensity could be due to the aggregation of gold nanoparticles.51 The decrease of the LSPR intensity could contribute to the decrease of the photocatalytic activity at a higher gold ratio.

The photocatalytic efficiency of plasmonic core–shell materials was investigated for hydrogen production from 1[thin space (1/6-em)]:[thin space (1/6-em)]3 v/v methanol/water solution, under UV-visible light illumination and is shown in Fig. 5. A SiO2@TiO2 core–shell with no gold exhibited very low activity for hydrogen production compared to nanocomposite materials. Published studies report on the beneficial effect of AuNPs supported on a TiO2 surface on hydrogen production.4–6,52 In agreement with these studies, AuNPs supported on the SiO2@TiO2 system improve the photocatalytic production of hydrogen compared to the metal-free sample. Surprisingly, covering AuNPs by TiO2 drastically boosts the production of hydrogen, and it is much higher compared to the conventional sample (AuNPs grown on the core–shell surface). The H2 production was observed to be 6.7 and 10 times higher for 0.5 and 1 wt% of AuNP ratio (Fig. 5b), respectively, compared to the best classical photocatalyst (SiO2@TiO2@Au) series (Fig. 5a). Under UV-visible illumination, the rate of hydrogen generation was significantly improved and reached 12 mmol h−1 g−1 for 1 wt% of Au ratio. The photocatalytic efficiency was compared to that of Pt/TiO2P25 under the same reaction conditions. The rate of H2 production was 13.6 mmol h−1 g−1 using 1 wt% of Pt and showed efficiency as good as SMSI SiO2@Au@TiO2 with 1 wt%. The photonic efficiency (ηq, see the ESI for details) was determined to be 0.033% for 1 wt% and 0.038% for Pt/TiO2P25. It appears that the photocatalytic performance of the plasmonic SiO2@Au@TiO2 core–shell photocatalyst rather follows the evolution of the LSPR intensity as opposed to the AuNP ratio. In the meantime, no such conclusion could be made for the classical SiO2@TiO2@Au system. Indeed, for the maximum rate of H2 production (SMSI SiO2@Au@TiO2 with 1 wt% of AuNPs loading) the LSPR showed the highest intensity. As the LSPR decreased, the photocatalytic efficiency of hydrogen production also decreased. This suggests that the LSPR acts as a promoter of the H2 production since the reaction has been performed under UV-visible illumination. Radiative (including near-fields and scattering mechanism) and non-radiative phenomena were suggested to describe the photocatalytic mechanism.53

image file: c9nr09891g-f5.tif
Fig. 5 Hydrogen production under UV-visible illumination for different AuNP loadings (0.25, 0.5, 1, and 1.5 wt%) for both core–shell systems: (a) SiO2@TiO2@Au, (b) SiO2@Au@TiO2 under UV-visible illumination, and (c) hydrogen production under visible light illumination (λ > 420 nm).

The first mechanism involves the interaction of the LSPR of AuNPs with incident light resulting in a local enhancement of the electromagnetic near-field in the immediate vicinity of the TiO2 surface. The radiative energy transfer mechanism in the nanocomposite system reaches its maximum when an overlap of the nanoparticles’ LSPR band and the semiconductor electronic band gap is obtained.16 Therefore, the radiative effect should have a small impact since the LSPR band (visible) is spectrally misaligned with the TiO2 absorption (UV). Additionally, AuNPs are too small to consider any light scattering of the electric field enhancement.19 Considering the core–shell configuration of our photocatalysts, the so-called “non-radiative” effect could mostly explain the enhancement of the photocatalytic performance. In the context of hydrogen production, the kinetics of the hydrogen ion reduction at the surface of TiO2 is very low due to its large band gap and fast recombination of charge carriers. Under the UV-visible illumination, AuNPs exhibit an LSPR, i.e. collective oscillations of free-electrons, in response to the incident electromagnetic fields. Indeed, the SMSI SiO2@Au@TiO2 (1 wt%) under visible illumination exhibited very good photoactivity for the hydrogen production, as shown in Fig. 5c, while no hydrogen was detected either for SiO2@TiO2. The H2 produced after 5 hours of irradiation was 240 μmol g−1 with a rate of 52 μmol h−1 g−1. The mechanism behind the H2 production remains unclear, especially the hole generation needed for the methanol dissociative adsorption. The latter likely occurs at the Au/TiO2 interface where holes were reported to be positioned.54 The LSPR-excited energetic electrons (hot electrons) either contributed to the reduction of hydrogen ions, or injected to TiO2 increasing the H2 formation photocatalytic rate. It has been shown that hot electrons can hardly pass through an insulating barrier with a thickness greater than 10 nm.16 Hence, the H2 production rate increase could be mediated by the hot electron transfer from AuNPs to TiO2 considering the visible light illumination. Although the nanocomposite core–shell system exhibits an improved performance compared to the SiO2@TiO2 photocatalyst, it is obvious that the SMSI of Au/TiO2 in the core–shell nanostructure leads to much higher photocatalytic efficiency for hydrogen production. Thus, the charge transfer on both nanocomposite systems should be considered to define the active site where the photocatalytic reaction occurs. As has been reported, the charge transfer of gold coupled to the TiO2 junction is quite complex. Accordingly, tracking the charge carrier dynamics as well as their lifetime under selective wavelength excitation using TRMC tools will enable further understanding of the difference between the two systems.

The TRMC, as a contactless technique, was used to evaluate the AuNP effect on the charge carrier density and dynamics for each system at different AuNP loadings. The TRMC signal mainly arises from photogenerated charge carriers produced or transferred at the surface of the TiO2 semiconductor. Furthermore, the TRMC signal is mainly related to the electrons since holes are heavier. Thus, their mobility is limited and holes are trapped in the bulk in the early time range.40,55 Therefore, the signal observed from the TRMC is correlated with the electron density at the surface of the TiO2 shell after the laser excitation in the early stage, and the decay of free electron number is caused by recombination, trapping and capture by AuNPs.4,27Fig. 6 shows the TRMC signal obtained from both systems at different AuNP loadings, after 350 nm laser pulse excitation (UV illumination). The signal to noise measured on our experiment was about 3 to 1, which is why the thickness of the TiO2 overlayer is reasonable. Core–shell SiO2@TiO2 nanospheres were assumed to have the same signal after UV light excitation, and therefore AuNPs would not affect the absorption of photons in the UV range (no structural modification in TiO2). The gold-free SiO2@TiO2 sample showed a low signal after UV light excitation, which is due to the thin shell layer at the surface (4–10 nm) as shown in Fig. 6a. Given the TRMC results, it seems obvious that the photogenerated charge carrier density relies on the system. Indeed, after UV light excitation the photogenerated charge carrier density undoubtedly increases for SMSI SiO2@Au@TiO2 compared to the AuNP-free sample, while in the case of the classical SiO2@TiO2@Au system, the signal TRMC decreases dramatically. At an early excitation stage, SMSI Au/TiO2 in the core–shell system collects the photogenerated electrons and injects them back at the TiO2 surface leading to an increase in the TRMC signal. This complex pathway explains how the quenching of the photogenerated charge carriers is not observed for the SMSI SiO2@Au@TiO2 system. The transfer of hot electrons from AuNPs is a well-known process compared to the opposite trajectory. Electrons obtained under UV illumination by energy higher than the band gap energy of TiO2 are indeed able to surround the Schottky barrier (SB) and be trapped by AuNPs. The SB in the Au/TiO2 system has been estimated to be 0.23 eV (ref. 56) and the energy of the conduction band of TiO2 was −0.16 eV.57 Accordingly, knowing that the Fermi level for noble metals is around 0 eV on the normal hydrogen electrode scale,8,58 the irradiation of TiO2 with an energy higher than 3.27 eV (considering the anatase phase with a band gap of 3.2 eV (ref. 59)) or a wavelength lower than 379 eV will produce electrons with enough energy to pass through the SB and then be trapped by AuNPs.60 This pathway was observed for the Au/TiO2 system. As expected, under mixed UV-visible light irradiation hot electrons injected from the AuNPs to TiO2 under visible light irradiation surmount the SB, and flow back into the TiO2. While in the SiO2@TiO2@Au system, the TRMC signal decreases dramatically. In the classical SiO2@TiO2@Au system, the weak TRMC signal (low charge carrier density) could be due to: (1) the shield effect caused by AuNPs, (2) surface recombination centers created by the synthesis method, and (3) fast electron scavenging by the metal (<10 ns).4,11 The first two hypotheses could not explain the weak TRMC signal for classical SiO2@TiO2@Au compared to that of SMSI SiO2@Au@TiO2. As we recently showed, AuNPs do not affect the electric field after illumination regardless of where AuNPs are deposited, since their size is small.19 Therefore, the electrons are scavenged by AuNPs drastically reducing the charge carrier density at the surface of TiO2,61 in agreement with the results from the photocatalytic production of hydrogen.

image file: c9nr09891g-f6.tif
Fig. 6 TRMC signal for (a) SiO2@TiO2@Au and (b) SiO2@Au@TiO2 systems for variable AuNP loading obtained before and immediately after 350 nm light excitation (Iex = 2.3 mJ cm−2 for λ = 350 nm).

The TRMC results enabled us to shed light on the effect of the localization of AuNPs on the photogenerated charge carriers’ dynamics, which is correlated with the photocatalytic efficiency. The electron transfer occurring in both systems follows different physical pathways, leading to a drastically variable photocatalytic efficiency. The main consequence of modulating the core–shell system is that the active sites where hydrogen ion reduction reactions occur are different. After trapping the photogenerated electrons from the conduction band of TiO2 in the SiO2@TiO2@Au system, AuNPs reduce hydrogen ions to form H2 molecules. In this case, AuNPs as the active sites are much less efficient for the reduction reaction compared to the TiO2 active sites in the SiO2@Au@TiO2 system. In the latter, AuNPs act as a promoter of the photogenerated electron lifetime via different mechanisms. The electrons are transferred from TiO2 to Au, and then are injected back to TiO2 overpassing the Schottky barrier, and speeding up the kinetics of hydrogen production. The electronic properties of the samples loaded with 1 wt% of AuNPs were studied under visible laser excitation (420–550 nm), and the results are shown in ESI Fig. S5. A TRMC signal was observed for the SMSI SiO2@Au@TiO2 sample, whereas there was no signal for the classical system (results not shown), confirming the efficient electron generation and their transfer in the conduction band of the TiO2 overlayer. This result does not exclude the generation of free electrons at the surface of AuNPs in the classical system, but only confirms the beneficial effect of the SMSI concept toward efficient electron transfer from AuNPs to TiO2 under visible light excitation.

The mechanism of the charge transfer does not address the problem of the minority charge-carrier diffusion length encountered in both systems. The diffusion length could be considered as a detrimental parameter, particularly when the energetic charge carriers are produced far from the active sites. The difference of the hydrogen production rate in both systems cannot be restricted to the photonic effect and/or to the charge transfer. Indeed, considering the electron diffusion length as well as the distance between the adsorbed MeOH and the active sites (reductive species), the kinetics of dissociative adsorption of MeOH and the reduction of H+ over both systems should be addressed to further understand the difference in photoefficiency.

DFT calculations were performed to further understand the difference in adsorption energies resulting from the two core–shell systems. This was performed by simulating the adsorption of a single methanol molecule on the surface of classical SiO2@TiO2@Au and SMSI SiO2@Au@TiO2. Four initial adsorption sites were considered for the former material because of the added complexity arising from the presence of the gold cluster, while two initial configurations were considered for the latter (ESI Fig. S6 and S7). Each of these adsorption scenarios were geometrically-optimized at fixed lattice parameters to find the most stable complex, and adsorption energies were compared to determine the most stable site. For SiO2@TiO2@Au, methanol adsorbs on the five-coordinated titanium (Ti5c) on the surface near the gold cluster interface as expected with an adsorption energy of −1.0788 eV (Fig. 7a), whereas in the SMSI SiO2@Au@TiO2 model, adsorption also occurs on the titanium site. However, in the absence of gold, the adsorption energy is weaker at −0.9479 eV (Fig. 7b).

image file: c9nr09891g-f7.tif
Fig. 7 Most stable calculated adsorption sites and configuration of methanol on (a) TiO2@Au and (b) Au@TiO2 nanocomposite photocatalysts. Close-up on the bottom shows the adsorption energies and measured distance from oxygen in methanol to the nearest five-coordinated titanium site on the anatase (101) surface. Red atoms are representative of oxygen, grey titanium, white hydrogen, and yellow gold. (c) and (d) Scheme of the mechanism proposed for the H2 production over each system: kdes(a′) > kdes(b′). After the surface illumination ❶, the dissociative adsorption of MeOH occurs ❷ leading to the formation of methoxy species and release of the hydrogen ion (hole scavenging) ❸. H+ ions are reduced by electrons available either in the TiO2 surface or AuNPs ❹. To be reduced by AuNPs, hydrogen ions (if released far from AuNP active sites; the case of SiO2@TiO2@Au) have to migrate or diffuse on the surface to get reduced. Methoxy species are oxidized by holes causing the formation of formaldehyde ❺ and the hydrogen ions. The reduced hydrogen atoms recombine at the active surface and for H2 gas ❻. Finally, the photocatalyst completes the cycle getting back to its initial state ❼.

The adsorption distance in both systems is different by only 0.012 Å, which is not enough to cause a relatively large difference in adsorption energies of 0.13 eV. This result demonstrates that the rate of methanol adsorption/desorption at the surface of the SMSI SiO2@Au@TiO2 system would occur faster compared to classical SiO2@TiO2@Au. With a stronger adsorption found in SiO2@TiO2@Au, it is speculated that this nanocomposite photocatalyst exhibits lower activity due to the reduced desorption of methanol from the surface, a crucial step in the hydrogen production from methanol aqueous solution; this is in agreement with the photocatalytic data shown in Fig. 5.

The reaction pathway of the photocatalytic hydrogen production over the nanocomposite system is as follows:1,62

image file: c9nr09891g-t1.tif(1)
image file: c9nr09891g-t2.tif(2)
image file: c9nr09891g-t3.tif(3)

With the overall reaction being

image file: c9nr09891g-t4.tif(4)

The illumination induces a dissociative adsorption of methanol and hole scavenging, followed by ethoxide formation and hydrogen ion (H+) release leading to the formation of a H2 molecule (eqn (1)).63 The hydrogen production kinetics would then also be affected by the dissociative adsorption of MeOH. The next step is the desorption and the water splitting reaction (eqn (2) and (3)). The kinetics of adsorption/desorption of methanol should determine the hole scavenging efficiency and thus the kinetics of the reaction “k1”. Our pathway proposes the reduction of hydrogen obtained from the dissociative adsorption of MeOH after illumination and charge carrier generation. Furthermore, the reduction of hydrogen ions adsorbed at the TiO2 surface to H2 molecules should also be considered. In fact, the active sites cover a larger surface in the SMSI SiO2@Au@TiO2 system compared to the SiO2@TiO2@Au system (since AuNPs are the active sites in the latter case). In order to achieve an efficient reaction, the adsorption site of hydrogen atoms should most likely occur at the Au/TiO2 interface.64 The probability of the hydrogen reduction in the SMSI SiO2@Au@TiO2 system is favorable since a TiO2 overlayer surrounded the AuNPs and provides closer active sites. Thus, the adsorbed H+ ions are reduced by the electrons available at the TiO2 surface where the MeOH is adsorbed (electrons are injected back from AuNPs to TiO2) (Fig. 7d). The mechanism is probably different in the case of the SiO2@TiO2@Au system (Fig. 7c), which has a similar configuration to that reported by Yang et al.64 In the latter configuration, the hydrogen reduction reaction is limited by the distance between the adsorption site and AuNPs as well as the diffusion kinetics of these atoms from the TiO2 surface to reach AuNP active sites (Fig. 7c). To achieve an efficient reduction of hydrogen ions, they have to migrate from the adsorption site of methanol to the active site (AuNPs). However, if the hydrogen atoms are not reduced (produced far from the active site) they would migrate to the bulk. The diffusion of the hydrogen atoms over the surface is unlikely since the migration in the bulk is a thermodynamically spontaneous process.65 This configuration suggests the formation of H2 molecules in TiO2 bulk, from which desorption is kinetically difficult.


In summary, a successful method was applied for the design of a tunable plasmonic core–shell nanostructure. The synthesis method enables the control of the formation of the TiO2 overlayer, the size distribution and the dispersion of AuNPs. The LSPR intensity of AuNPs was found to be very sensitive to the localization of the AuNPs and to the surrounding media. Furthermore, the electronic properties assessed by TRMC suggested different charge carrier dynamics regarding the system. Under UV illumination, the electrons are injected back from AuNPs to the TiO2 overlayer increasing their lifetime for the SMSI SiO2@Au@TiO2 system, while they are collected by AuNPs in the SiO2@TiO2@Au system. Furthermore, the photocatalytic activity for H2 production was strongly affected by the localization of AuNPs. A strong metal–support interaction between AuNPs and the titania shell showed boosted photoactivity, with an optimal Au/TiO2 ratio of about 1 wt%. Finally, we found that the rate of adsorption/desorption of methanol is a crucial step in the hydrogen production.

Conflicts of interest

There are no conflicts of interest to declare.


Cong Wang acknowledges the China Scholarship Council for his research follow position.


  1. T. Kawai and T. Sakata, J. Chem. Soc., Chem. Commun., 1980, 15, 694–695 RSC.
  2. H. Courbon, J. Herrmann and P. Pichat, J. Phys. Chem., 1984, 88, 5210–5214 CrossRef CAS.
  3. M. Murdoch, G. I. N. Waterhouse, M. A. Nadeem, J. B. Metson, M. A. Keane, R. F. Howe, J. Llorca and H. Idriss, Nat. Chem., 2011, 3, 489 CrossRef CAS PubMed.
  4. M. G. Méndez-Medrano, E. Kowalska, A. Lehoux, A. Herissan, B. Ohtani, S. Rau, C. Colbeau-Justin, J. L. Rodríguez-López and H. Remita, J. Phys. Chem. C, 2016, 120, 25010–25022 CrossRef.
  5. J. Zhang, X. Jin, P. I. Morales-Guzman, X. Yu, H. Liu, H. Zhang, L. Razzari and J. P. Claverie, ACS Nano, 2016, 10, 4496–4503 CrossRef CAS PubMed.
  6. Z. W. Seh, S. Liu, M. Low, S. Y. Zhang, Z. Liu, A. Mlayah and M. Y. Han, Adv. Mater., 2012, 24, 2310–2314 CrossRef CAS PubMed.
  7. K. Awazu, M. Fujimaki, C. Rockstuhl, J. Tominaga, H. Murakami, Y. Ohki, N. Yoshida and T. Watanabe, J. Am. Chem. Soc., 2008, 130, 1676–1680 CrossRef CAS PubMed.
  8. S. Linic, P. Christopher and D. B. Ingram, Nat. Mater., 2011, 10, 911–921 CrossRef CAS PubMed.
  9. S. W. Verbruggen, M. Keulemans, M. Filippousi, D. Flahaut, G. Van Tendeloo, S. Lacombe, J. A. Martens and S. Lenaerts, Appl. Catal., B, 2014, 156–157, 116–121 CrossRef CAS.
  10. E. Kowalska, R. Abe and B. Ohtani, Chem. Commun., 2009, 2, 241–243 RSC.
  11. A. L. Luna, E. Novoseltceva, E. Louarn, P. Beaunier, E. Kowalska, B. Ohtani, M. A. Valenzuela, H. Remita and C. Colbeau-Justin, Appl. Catal., B, 2016, 191, 18–28 CrossRef CAS.
  12. G. D. Gesesse, C. Li, E. Paineau, Y. Habibi, H. Remita, C. Colbeau-Justin and M. N. Ghazzal, Chem. Mater., 2019, 31, 4851–4863 CrossRef CAS.
  13. M. Ye, H. Zhou, T. Zhang, Y. Zhang and Y. Shao, Chem. Eng. J., 2013, 226, 209–216 CrossRef CAS.
  14. A. Sousa-Castillo, M. Comesaña-Hermo, B. Rodríguez-González, M. Pérez-Lorenzo, Z. Wang, X. T. Kong, A. O. Govorov and M. A. Correa-Duarte, J. Phys. Chem. C, 2016, 120, 11690–11699 CrossRef CAS.
  15. K. Shoueir, S. Kandil, H. El-hosainy and M. El-Kemary, J. Cleaner Prod., 2019, 230, 383–393 CrossRef CAS.
  16. S. K. Cushing, J. Li, J. Bright, B. T. Yost, P. Zheng, A. D. Bristow and N. Wu, J. Phys. Chem. C, 2015, 119, 16239–16244 CrossRef CAS.
  17. B. Wu, D. Liu, S. Mubeen, T. T. Chuong, M. Moskovits and G. D. Stucky, J. Am. Chem. Soc., 2016, 138, 1114–1117 CrossRef CAS PubMed.
  18. R. Lee, Y. Kumaresan, S. Y. Yoon, S. H. Um, I. K. Kwon and G. Y. Jung, RSC Adv., 2017, 7, 7469–7475 RSC.
  19. G. D. Gesesse, T. Le Neel, Z. Cui, G. Bachelier, H. Remita, C. Colbeau-Justin and M. N. Ghazzal, Nanoscale, 2018, 10, 20140–20146 RSC.
  20. R. Boppella, F. M. Mota, J. W. Lim, S. T. Kochuveedu, S. Ahn, J. Lee, D. Kawaguchi, K. Tanaka and D. H. Kim, ACS Appl. Energy Mater., 2019, 2, 3780–3790 CrossRef CAS.
  21. D. Ding, K. Liu, S. He, C. Gao and Y. Yin, Nano Lett., 2014, 14, 6731–6736 CrossRef CAS PubMed.
  22. C. Jia, X. Li, N. Xin, Y. Gong, J. Guan, L. Meng, S. Meng and X. Guo, Adv. Energy Mater., 2016, 6, 1–11 Search PubMed.
  23. S. J. Tauster, S. C. Fung and R. L. Garten, J. Am. Chem. Soc., 1978, 100, 170–175 CrossRef CAS.
  24. T. Uchijima, Catal. Today, 1996, 28, 105–117 CrossRef CAS.
  25. C. De Leitenburg and A. Trovarelli, J. Catal., 1995, 156, 171–174 CrossRef CAS.
  26. K. Fujiwara, K. Okuyama and S. E. Pratsinis, Environ. Sci.: Nano, 2017, 4, 2076–2092 RSC.
  27. J. C. Matsubu, S. Zhang, L. DeRita, N. S. Marinkovic, J. G. Chen, G. W. Graham, X. Pan and P. Christopher, Nat. Chem., 2017, 9, 120–127 CrossRef CAS.
  28. M. E. Strayer, J. M. Binz, M. Tanase, S. M. Kamali Shahri, R. Sharma, R. M. Rioux and T. E. Mallouk, J. Am. Chem. Soc., 2014, 136, 5687–5696 CrossRef CAS PubMed.
  29. M. Xu, S. Yao, D. Rao, Y. Niu, N. Liu, M. Peng, P. Zhai, Y. Man, L. Zheng, B. Wang, B. Zhang, D. Ma and M. Wei, J. Am. Chem. Soc., 2018, 140, 11241–11251 CrossRef CAS PubMed.
  30. S. Li, Y. Xu, Y. Chen, W. Li, L. Lin, M. Li, Y. Deng, X. Wang, B. Ge, C. Yang, S. Yao, J. Xie, Y. Li, X. Liu and D. Ma, Angew. Chem., Int. Ed., 2017, 56, 10761–10765 CrossRef CAS PubMed.
  31. P. Zhou, F. Lv, N. Li, Y. Zhang, Z. Mu, Y. Tang, J. Lai, Y. Chao, M. Luo, F. Lin, J. Zhou, D. Su and S. Guo, Nano Energy, 2019, 56, 127–137 CrossRef CAS.
  32. A. G. Shastri, A. K. Datye and J. Schwank, J. Catal., 1984, 87, 265–275 CrossRef CAS.
  33. M. Haruta, D. Su, M. A. Isaacs, J. Huang, H. Tang, J. Wang, L. Li, B. Zhang, B. Qiao, X. Liu, T. Zhang, Y. Su, K. Wilson, Y. Ren, C. Jin and A. F. Lee, Sci. Adv., 2017, 3, e1700231 CrossRef PubMed.
  34. H. Dyrbeck, N. Hammer, M. Rønning and E. A. Blekkan, Top. Catal., 2007, 45, 21–24 CrossRef CAS.
  35. D. A. Panayotov, S. P. Burrows, J. T. Yates and J. R. Morris, J. Phys. Chem. C, 2011, 115, 22400–22408 CrossRef CAS.
  36. K. Sun, M. Kohyama, S. Tanaka and S. Takeda, J. Phys. Chem. C, 2014, 118, 1611–1617 CrossRef CAS.
  37. J. Zhang, H. Wang, L. Wang, S. Ali, C. Wang, L. Wang, X. Meng, B. Li and D. S. Su, J. Am. Chem. Soc., 2019, 141, 2975–2983 CrossRef CAS PubMed.
  38. V. Kumaravel, S. Mathew, J. Bartlett and S. C. Pillai, Appl. Catal., B, 2019, 244, 1021–1064 CrossRef CAS.
  39. D. A. Panayotov, A. I. Frenkel and J. R. Morris, ACS Energy Lett., 2017, 2, 1223–1231 CrossRef CAS.
  40. C. Colbeau-Justin, M. Kunst and D. Huguenin, J. Mater. Sci., 2003, 38, 2429–2437 CrossRef CAS.
  41. S. J. Clark, M. D. Segall, C. J. Pickard, P. J. Hasnip, M. I. J. Probert, K. Refson and M. C. Payne, Z. Kristallogr., 2005, 220, 567–570 CAS.
  42. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868 CrossRef CAS PubMed.
  43. H. L. Yeh, S. H. Tai, C. M. Hsieh and B. K. Chang, J. Phys. Chem. C, 2018, 122, 19447–19454 CrossRef CAS.
  44. J. K. Burdett, T. Hughbanks, G. J. Miller, J. V. Smith and J. W. Richardson, J. Am. Chem. Soc., 1987, 109, 3639–3646 CrossRef CAS.
  45. X. Q. Gong, A. Selloni, O. Dulub, P. Jacobson and U. Diebold, J. Am. Chem. Soc., 2008, 130, 370–381 CrossRef CAS PubMed.
  46. S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem. Phys., 2010, 132, 154104 CrossRef PubMed.
  47. M. N. Ghazzal, H. Kebaili, M. Joseph, D. P. Debecker, P. Eloy, J. De Coninck and E. M. Gaigneaux, Appl. Catal., B, 2012, 115–116, 276–284 CrossRef CAS.
  48. J. Kimling, M. Maier, B. Okenve, V. Kotaidis, H. Ballot and A. Plech, J. Phys. Chem. B, 2006, 110, 15700–15707 CrossRef CAS PubMed.
  49. C. Ziegler and A. Eychmüller, J. Phys. Chem. C, 2011, 115, 4502–4506 CrossRef CAS.
  50. M. N. Ghazzal, D. P. Debecker and E. M. Gaigneaux, Thin Solid Films, 2016, 611, 117–124 CrossRef CAS.
  51. R. Pamies, J. G. H. Cifre, V. F. Espín, M. Collado-González, F. G. D. Baños and J. G. De La Torre, J. Nanopart. Res., 2014, 16, 2376–2387 CrossRef.
  52. Z. Hai, N. El Kolli, D. B. Uribe, P. Beaunier, M. José-Yacaman, J. Vigneron, A. Etcheberry, S. Sorgues, C. Colbeau-Justin, J. Chen and H. Remita, J. Mater. Chem. A, 2013, 1, 10829 RSC.
  53. W. R. Erwin, H. F. Zarick, E. M. Talbert and R. Bardhan, Energy Environ. Sci., 2016, 9, 1577–1601 RSC.
  54. S. Wang, Y. Gao, S. Miao, T. Liu, L. Mu, R. Li, F. Fan and C. Li, J. Am. Chem. Soc., 2017, 139, 11771–11778 CrossRef CAS PubMed.
  55. Y. Tamaki, A. Furube, M. Murai, K. Hara, R. Katoh and M. Tachiya, Phys. Chem. Chem. Phys., 2007, 9, 1453–1460 RSC.
  56. M. S. Arshad, Š. Trafela, K. Ž. Rožman, J. Kovač, P. Djinović and A. Pintar, J. Mater. Chem. C, 2017, 5, 10509–10516 RSC.
  57. X. Zhang, L. Zhang, T. Xie and D. Wang, J. Phys. Chem. C, 2009, 113, 7371–7378 CrossRef CAS.
  58. S. I. Naya, M. Teranishi, R. Aoki and H. Tada, J. Phys. Chem. C, 2016, 120, 12440–12445 CrossRef CAS.
  59. M. N. Ghazzal, N. Chaoui, M. Genet, E. M. Gaigneaux and D. Robert, Thin Solid Films, 2011, 520, 1147–1154 CrossRef CAS.
  60. Z. Lin, X. Wang, J. Liu, Z. Tian, L. Dai, B. He, C. Han, Y. Wu, Z. Zeng and Z. Hu, Nanoscale, 2015, 7, 4114–4123 RSC.
  61. V. Subramanian, E. E. Wolf and P. V. Kamat, J. Am. Chem. Soc., 2004, 126, 4943–4950 CrossRef CAS PubMed.
  62. R. Fretwell and P. Douglas, Photochem. Photobiol. Sci., 2002, 1, 793–798 RSC.
  63. M. Chen, Q. Guo, C. Xu, D. Dai, X. Yang and W. Yang, J. Am. Chem. Soc., 2013, 135, 10206–10209 CrossRef PubMed.
  64. Y. Z. Yang, C. H. Chang and H. Idriss, Appl. Catal., B, 2006, 67, 217–222 CrossRef CAS.
  65. M. M. Islam, M. Calatayud and G. Pacchioni, J. Phys. Chem. C, 2011, 115, 6809–6814 CrossRef CAS.


Electronic supplementary information (ESI) available. See DOI: 10.1039/c9nr09891g

This journal is © The Royal Society of Chemistry 2020