Catalysis Science & Technology

3 downloads 0 Views 4MB Size Report
Oct 6, 2016 - Roong Jien Wong,a Jason Scott,*a Gary K.-C. Low,a Haifeng Feng,b Yi Du,b. Judy N. Hartc and Rose Amal*a. The potential for applying UV ...
Catalysis Science & Technology

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

PAPER

Cite this: DOI: 10.1039/c6cy01717g

View Article Online View Journal

Investigating the effect of UV light pre-treatment on the oxygen activation capacity of Au/TiO2†‡ Roong Jien Wong,a Jason Scott,*a Gary K.-C. Low,a Haifeng Feng,b Yi Du,b Judy N. Hartc and Rose Amal*a The potential for applying UV light pre-treatment to enhance the oxygen activation capacity of Au/TiO2 under ambient conditions was examined. Catalytic formic acid oxidation in an aqueous environment was employed as the test reaction. Pre-illuminating Au/TiO2 with UV light can amplify the catalytic formic acid oxidation rate by up to four times with the degree of enhancement governed by system parameters such as Au loading, pre-illumination time, and initial formic acid loading. X-ray photoelectron spectroscopy, photoluminescence spectroscopy and electrochemical assessment of the Au/TiO2 indicated light preillumination invokes photoexcited electron transfer from the TiO2 support to the Au deposits. The Au deposits then utilise the additional electrons to catalyse molecular oxygen activation and promote the oxida-

Received 10th August 2016, Accepted 6th October 2016

tion reaction. Scanning tunneling spectroscopy analysis and first principle calculations indicated the Au deposits introduced new electronic states above the TiO2 valence band. The new electronic states were most intense at the Au–TiO2 interface suggesting the Au deposit:TiO2 perimeter may be the key region for oxy-

DOI: 10.1039/c6cy01717g

gen activation. The current study has demonstrated that pre-illuminating Au/TiO2 with light can be used to augment reactions where oxygen activation is a critical component, such as for the oxidation of organic

www.rsc.org/catalysis

pollutants and for the oxygen reduction reaction in fuel cells or energy storage systems.

Introduction Oxygen activation, commonly occurring as the oxygen reduction reaction (ORR), is a key process in many important reaction systems. O2 activation is essential in life processes (biological respiration), in energy conversion (fuel cells), and in many catalysis applications (organic oxidation and hydrogen peroxide production).1–7 The O2 activation step, which involves electron transfer to the O2 molecules, is known to be the rate-limiting step in many of these processes.1 In organic oxidation reactions, the catalytic oxidation capacity or oxidation rate is dependent on the O2 activation capacity or O2 activation rate. By monitoring the oxidation reaction rate under various reaction conditions, the effects of various parameters on the O2 activation capacity can be determined.

a

Particles and Catalysis Research Group, School of Chemical Engineering, The University of New South Wales, Sydney, NSW 2052, Australia. E-mail: [email protected], [email protected] b Institute for Superconducting and Electronic Materials (ISEM), University of Wollongong, Wollongong, New South Wales 2525, Australia c School of Materials Science and Engineering, UNSW Australia, Sydney, NSW, Australia † The authors declare no competing financial interest. ‡ Electronic supplementary information (ESI) available. See DOI: 10.1039/ c6cy01717g

This journal is © The Royal Society of Chemistry 2016

Since the discovery of catalytically active TiO2-supported Au nanoparticles by Haruta et al.,8–10 other methods have been reported in the literature to increase the catalytic oxidation capacity such as changing the synthesis method to manipulate the metal deposit size11–16 or, in some instances, resorting to high temperature annealing during catalysts synthesis.17 An alternative has been to use mixed transition metal oxides as the support to facilitate the O2 activation step.18 The reported findings can be translated as measures to increase the O2 activation capacity due to the close relationship between catalytic oxidation capacity and O2 activation capacity. Among the various catalysts reported, Au has been widely studied for its low temperature activation in CO oxidation.10,14,19 CO oxidation has been used to investigate the O2 activation step on Au due to the perceived simplicity of the reaction. In addition to its low activation energy, metal oxide-supported Au nanoparticles also exhibit high stability against CO deactivation, giving it superiority over its Pt counterpart,20,21 especially in direct methanol fuel cell applications where CO is produced as a by-product.10 While many of the Au-based catalysts investigated used semiconductor supports with photocatalytic properties, exploitation of the photocatalytic properties of the supports to enhance the thermal catalytic activity is not well reported. Previous work by Denny et al. and recent work by Scott et al. showed that pre-illuminating Pt/TiO2 is able to increase the

Catal. Sci. Technol.

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Paper

formic acid oxidation rate by up to seven times, achieving a rate similar to the photocatalytic reaction and demonstrating that light pre-treatment is able to enhance the O2 activation capacity.22,23 Similar effects of pre-illumination were also reported by Einaga et al. and Hwang et al. in their work on CO oxidation by Pt/TiO2.24,25 Both authors found that Pt/TiO2 exhibited enhanced catalytic activity in the dark after UV illumination. Within this article the enhancement in catalytic activity will be referred to as the ‘pre-illumination enhancement effect’. Metal deposits such as Au or Pt are able to inhibit the recombination of electrons and holes on the semiconductor surface by acting as electron sinks. In our present work, the presence of these electron sinks is believed to be one of the contributing factors promoting the preillumination enhancement effect. In this work, we exploit the semiconducting properties of TiO2 as a support and the low activation energy of Au to investigate the pre-illumination enhancement effect on the oxidation reaction in an aqueous system. Formic acid oxidation was used as the probe reaction to assess the oxygen activation rate. The effects of pre-illumination time, relaxation time, formic acid loading, and re-illumination on the rate of formic acid oxidation were investigated and the O2 activation capacity and photoelectron lifetime were also examined. Scanning tunneling spectroscopy (STS), first principle calculations and electrochemical ORR assessment were used to identify regions on the Au/TiO2 which were most likely responsible for oxygen activation as well as the origin of the improved activity invoked by light pre-illumination. The role of dissolved O2 and the deactivation of Au/TiO2 during storage are also considered.

Experimental methods Catalyst preparation and characterisation Materials. Aeroxide® TiO2 P25 (particle size ∼25 nm, surface area ∼50 m2 g−1, anatase to rutile ratio of 4 : 1) was used as received. All chemicals were of analytical grade and were used as supplied in all experiments: formic acid (>98%, Riedel-de Haën), perchloric acid (70%, Frederick Chemical), goldIJIII) chloride trihydrate (>99.9%, Sigma-Aldrich®), sodium hydroxide (>98%, Chem Supply), potassium hydroxide (>98%, Chem Supply), sodium sulphate anhydrous (>99.9%, Sigma-Aldrich®), acetic acid (>99%, Ajax Finechem), absolute ethanol, and chitosan (medium molecular weight, Sigma-Aldrich®). Milli-Q water (18 MΩ cm) was used in all experiments. Au/TiO2 synthesis. Au/TiO2 was prepared by the deposition-precipitation method with a nominal Au loading of 1.0 at%. An Au precursor solution was prepared by adding an appropriate amount of Au precursor into 200 ml of Milli-Q water. The solution was heated to 70 °C in an oil bath and was left for 30 min to reach thermal equilibrium at the set point temperature. The pH of the solution was adjusted to 7 by adding 0.1 M NaOH dropwise. TiO2 (2 g) was dispersed in 50 ml of Milli-Q water and was added into the Au precursor

Catal. Sci. Technol.

Catalysis Science & Technology

solution. The pH of the resulting solution was again adjusted to 7 by adding NaOH solution, and the solution was maintained at 70 °C for an additional 3 hours. The Au/TiO2 was prepared at pH 7 as this pH has been reported as the optimum pH in terms of gold uptake and catalytic activity.26 The particles were separated from the mother liquor by centrifuging at 10 000 rpm for 15 min. The collected particles were redispersed in Milli-Q water for washing, to remove the Cl− ions, before being dried in a vacuum desiccator for 24 hours. The dried particles were ground into a fine powder, and calcined at 250 °C (5 °C min−1 ramping) in 30 ml min−1 of air for 4 hours, followed by reduction at 450 °C (5 °C min−1 ramping) in 33 ml min−1 of 10% H2 (balance N2) for 3 hours. The purpose of the reduction step is to maximise the amount of metallic Au, as stated in the work of Tan et al.27 Au/TiO2 characterisation. The particle size distribution of the Au deposits was determined by high resolution transmission electron microscopy (HRTEM, Philips CM200) at an electron accelerating voltage of 200 kV. The elemental oxidation states of the as-prepared dry Au/TiO2 powder, Au/TiO2 at 0 min pre-illumination, and Au/TiO2 after 30 min preillumination were assessed by X-ray photoelectron spectroscopy (XPS). For the 0 min pre-illumination sample and 30 min pre-illumination sample, particles were recovered from suspension by centrifuging and then dried in a vacuum desiccator overnight. XPS data were collected using a Thermo Scientific™ ESCALAB™ 250Xi (monochromatic Al-Kα, 1486.68 eV) operating at 150 W with a spot size of 500 μm. All XPS spectra were normalised to C 1s peak = 285.0 eV for adventitious carbon. UV-vis-DRS spectra were obtained with a Shimadzu UV-3600 Plus UV-vis-NIR spectrophotometer. Photoluminescence (PL) measurement was performed under ambient conditions with a Horiba FluoroMax Spectrofluorometer. The actual Au loading on TiO2 was determined by an inductively coupled plasma (ICP) technique using a Perkin Elmer OPTIMA 7300 ICP optical emission spectrometer. Aqua regia was used for the microwave digestion of the Au.

Catalyst performance Light pre-treatment and reaction. The catalytic activity of Au/TiO2, with or without light pre-treatment (pre-illumination), was evaluated by formic acid oxidation undertaken in a 70 ml spiral reactor as detailed elsewhere.23 When molecular O2 adsorbed on the Au/TiO2 is activated or reduced to OOHads and OHads,28 it readily oxidises formic acid. The 50% formic acid oxidation rate (R50) was used to define the activity of Au/TiO2 in order to investigate the effect of pre-illumination. The R50 value is defined as the rate by which 50% of the initial formic acid loading is oxidised (see ESI,‡ Fig. S1). In a typical pre-illumination experiment, a 1 g L−1 Au/TiO2 suspension was adjusted to pH 3.0 ± 0.05 with 0.5 M perchloric acid and transferred into the spiral reactor. Pre-illumination was performed on the catalyst suspension with a 20 W NEC black light blue fluorescent tube (UV light) for a predetermined duration in a closed system. Following the

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Catalysis Science & Technology

pre-illumination step, the system was opened to the atmosphere for 10 min to allow for air equilibration. The system was then resealed for a relaxation period of 10 min before 100–1000 μmol of formic acid was injected into the system. The UV light remained off throughout the duration of formic acid oxidation. Pre-illumination reactions are denoted as “Pre-X”, where X refers to the pre-illumination period in minutes. The catalytic activity of Au/TiO2, with or without pre-illumination, was evaluated by monitoring the CO2 generated (with no illumination) during formic acid oxidation. Reactions without pre-illumination followed the same procedure but without turning on the light during the 30 min preillumination step, and are denoted as “Dark”. The photocatalytic oxidation of formic acid was performed by turning on the UV light after the 10 min relaxation period, and is denoted as “Light”. Experiments were also performed in an anoxic environment. Dissolved O2 was removed from the system by purging with N2 (to 0 ppm O2 as detected by a dissolved oxygen probe) prior to the pre-illumination step. The dissolved oxygen level of the catalyst suspension was measured using an Oakton® DO 6+ dissolved oxygen meter. Electrochemical ORR activity. The electrochemical ORR activity of Au/TiO2 was evaluated by preparing Au/TiO2 as a thin film electrode. The as-prepared Au/TiO2 (500 mg) was ground with glacial acetic acid (500 μL), absolute ethanol (500 μL), and chitosan (20 mg). Glacial acetic acid and chitosan were used as binding agents to provide good adhesion of the catalyst paste on the fluorine doped tin oxide (FTO) glass slide. A homogeneous paste was achieved by adding absolute ethanol and Milli-Q water. The Au/TiO2 thin film electrode was prepared by transferring the paste onto the FTO glass slide by the doctor blading method. The ORR activity was assessed using a three-electrode system consisting of the as-prepared Au/TiO2 as the working electrode, a Pt counter electrode, and a Ag/AgCl reference electrode.29 The electrolyte comprised 0.1 M KOH. A 300 W xenon lamp (PECCELL) was used as the light source. A current–voltage profile was obtained using a potentiostat (PGSTAT302N, Autolab) with GPES interface software. Linear sweep voltammetry was performed from 0 V to −1.0 V after an initial 15 min N2 purge as the baseline, followed by a second measurement after 15 min of air purging. Preillumination for 5–15 min was performed after the N2 and air purging step, and the current density was corrected to the corresponding baseline.

Scanning tunneling microscopy (STM) characterisation STM images and scanning tunneling spectroscopy (STS) spectra were obtained at 77 K with an ultrahigh vacuum (UHV) low temperature STM system, equipped with a molecular beam epitaxy (MBE) (SHINKOSHA Co., Ltd.). Rutile TiO2 (110) single crystals were treated by argon ion sputtering (1 kV, 20 min) and subsequent annealing at 850 K (1 h) for several cycles. The deposition of gold on rutile (110) was

This journal is © The Royal Society of Chemistry 2016

Paper

conducted in the MBE chamber before the sample was transferred to the STM for in situ measurements. Gold clusters were prepared by atom deposition from a high-purity Au source (99.995%) at room temperature. The deposition flux of gold was 0.008 monolayers per minute (ML min−1). 1 ML (5.2 × 1014 atoms per cm−2) is defined as the coverage of one Au atom for each 5f-Ti atom of the rutile TiO2 (110) surface. During Au exposure, the background pressure was lower than 1 × 10−10 Torr. Differential conductance dI/dV spectra were acquired by using a standard lock-in technique with 20 mV modulation at 937 Hz. Before STS measurements, the Pt/Ir tip was calibrated on a silver surface.

Density functional theory (DFT) calculations DFT calculations of an Au cluster on the surface of TiO2 were undertaken following the method described in our previous work.30,31 The model used in the calculations was either a slab of anatase TiO2, cut from pre-optimised bulk TiO2 to expose the (101) surface, consisting of two layers of TiO6 octahedra and a total of 72 atoms, or a slab of rutile TiO2, cut to expose the (110) surface and again consisting of two layers of TiO6 octahedra and a total of 72 atoms. This slab thickness was found in previous work to be sufficient for modelling metallic clusters on TiO2.32 To create the initial geometry, a tetrahedral cluster of four gold atoms was placed on the top surface of the slab, following the geometry found by Zhang et al.33 to be most stable for a cluster of four metallic atoms on a TiO2 anatase (101) surface. For the rutile slab, the initial geometry was taken from Jiang et al.34 For the slabs, the lattice parameters were fixed at those for bulk anatase or rutile TiO2; all Au, Ti, and O atoms on the top surface were allowed to relax, while the atoms on the bottom surface were frozen in place to simulate a bulk-like environment. The final optimised geometry of the gold cluster on the TiO2 surfaces is shown in Fig. S2 in ESI.‡ The optimised geometries and densities of states were calculated using the CRYSTAL09 code.35,36 The B3LYP hybrid method37,38 was used but with the amount of Hartree– Fock exchange energy reduced to 13%, as this has been found to accurately reproduce the experimentally measured band gap of TiO2.39 The Monkhorst–Pack grid for k-points sampling was set at 6 × 6 × 1 for the Brillouin zone; convergence with respect to the number of k-points was checked. The basis sets used were a 8-411(d1) basis set for oxygen,40 a 8-6411(d3) basis set for titanium,41,42 and an effective core pseudopotential for gold.43 The thresholds for convergence of the maximum gradient, RMS gradient, maximum displacement and RMS displacement were 4.5 × 10−4, 3.0 × 10−4, 1.8 × 10−3 and 1.8 × 10−3 a.u., respectively. This method has been used previously for calculations of TiO2;44 the calculated lattice parameters and band gaps of bulk anatase and rutile TiO2 are in good agreement with the experimental values (within ∼1% for the lattice parameters and 0.1 eV for the band gap).45–47

Catal. Sci. Technol.

View Article Online

Paper

Results and discussion

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Physical properties of Au/TiO2 A HRTEM image of the as-prepared Au/TiO2 catalyst is shown in Fig. 1(a) with the Au deposit size distribution given in Fig. 1(b). The Au deposits have an average size of 3.0 nm, with the highest percentage of particles in the 2.5–3.0 nm range, similar to those reported by Akita et al.48 The final Au loading of 0.92 at% on TiO2, as determined by ICP, also agrees well with that reported by Moreau et al. for a nominal Au loading of 1.0 at% on TiO2.26 XRD patterns of neat TiO2 and Au/TiO2 (Fig. S3, ESI‡) further confirm the presence of crystalline Au nanoparticles on the TiO2 support. Catalyst performance Au loading. Fig. 2(a) shows the effect of Au loading on oxidation activity; it can be seen that the effect of increasing the Au loading from 0.5 at% to 1.0 at% varies depending on the light exposure condition. The activity exhibited by Au/TiO2 in the dark condition shows that there is a thermal catalytic effect by the Au deposits which is slightly improved with increasing Au loading. In the ‘light’ condition, there is little change in the oxidation rate with increasing Au loading implying some other experimental factor is rate limiting, with a value of 35.9 μmol min−1 for R50 representing the maximum attainable rate of formic acid oxidation for this set of experimental parameters. The effect of Au loading is most pronounced for the pre-illumination condition where increasing the loading from 0.5 at% to 1.0 at% increases the formic acid oxidation rate by a factor of 1.5. Control experiments were undertaken to confirm that the enhancement in performance upon pre-illumination primarily derived from the Au deposits and was not an experimental artefact or attributable to the TiO2. To identify whether there was any contribution of photolysis to the system perfor-

Catalysis Science & Technology

mance, 100 μmol of formic acid in the reaction medium alone was circulated through the reactor (result not shown). Over a 1 h period there was negligible change in the conductivity, indicating formic acid oxidation originated from the Au/TiO2. Similarly, neat TiO2 exhibited negligible catalytic activity when no pre-illumination was used although a low level of activity was evident (R50 = 1.8 μmol min−1) when the neat TiO2 was pre-illuminated (Fig. 2(a)). While pre-illumination invoked some oxidative activity in the TiO2 it was substantially smaller than in the photocatalytic (i.e. ‘light’) case. When Au was present on the TiO2, the oxidation activity was considerably greater than the control case under all lighting conditions. Consequently, it is reasonable to conclude that the R50 values exhibited by Au/TiO2 are predominantly invoked by the presence of the Au deposits and the light preillumination process. Pre-illumination period. The impact of the light preillumination period on the rate of formic acid oxidation was assessed with 100 μmol of initial formic acid loading (Fig. 2(b)): the R50 value increases with the pre-illumination time. The catalyst is active under the dark condition (R50 = 13.3 μmol min−1) with the activity approximately doubling (27.6 μmol min−1) following 30 min pre-illumination. When illuminated for a further 70 min (i.e. 100 min total), there is an additional small increase in activity up to 32.3 μmol min−1. As the pre-illumination period is increased, the catalyst activity approaches that of the photocatalytic case. Formic acid loading. Higher initial formic acid loadings were employed to evaluate the O2 activation capacity of Au/ TiO2 and to investigate the lifetime of photoelectrons generated during pre-illumination (Fig. 2(c)). When the initial formic acid loading is increased from 100 μmol to 200 μmol, a corresponding (although not proportional) increase in both the 30 min light pre-illuminated and photocatalytic oxidation

Fig. 1 (a) HRTEM image of Au/TiO2 prepared by deposition–precipitation. Red arrows indicate a selection of the Au deposits; (b) size distribution of the Au deposits on TiO2 as evaluated from HRTEM. Au deposit count: 971.

Catal. Sci. Technol.

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Catalysis Science & Technology

Paper

Fig. 2 The effect of: (a) Au loading on the formic acid oxidation rate (R50) for Au/TiO2 (dark = no pre-illumination, pre-30 = 30 min UV light pre-illumination, light = photocatalytic); (b) pre-illumination time on the formic acid oxidation rate for 1.0 at% Au/TiO2. The dotted line indicates the R50 of Au/TiO2 under the ‘light’ condition (i.e. UV-illuminated); (c) initial formic acid loading on its oxidation rate for 1.0 at% Au/TiO2 either following light pre-illumination or under the light condition. The dotted line highlights the formic acid oxidation rate under the ‘dark’ condition; (d) instantaneous rate of formic acid oxidation by Au/TiO2 as a function of time for various formic acid loadings. The dotted line represents the ideal oxidation rate profile for a 1000 μmol formic acid loading whereby no decay in activity (i.e. loss of active species) occurs. Catalyst loading = 1.0 g L−1, initial formic acid loading = 100 μmol, light pre-illumination period = 30 min unless otherwise indicated.

rates occurs, suggesting the low initial formic acid loading (100 μmol) being the rate-limiting factor. A further increase in the initial formic acid loading to 1000 μmol sees an additional increase in the photocatalytic reaction rate while there is no additional rate increase for the light pre-illumination case. Cessation of any further increase in the R50 above 200 μmol loading following 30 min pre-illumination suggests that a saturation point in the O2 activation capacity of Au/TiO2 has been reached. That is, the extent of oxygen molecules which are able to be activated during 30 min of preillumination has become rate limiting. The considerable difference between the light pre-illumination and photocatalytic rates for the 1000 μmol formic acid loading (Fig. 2(c)) highlights this point, with the variance in rates attributed to the direct participation of TiO2 as a photocatalyst through photogenerated radical (hydroxyl, peroxide, and superoxide radicals)49 formation. As photogenerated radicals have a short lifetime (in the order of microseconds), it is assumed they do

This journal is © The Royal Society of Chemistry 2016

not contribute to the R50 of the light pre-illuminated condition. Fig. 2 highlights experimental conditions where the reaction rate is governed by the initial formic acid loading or the availability of the active oxygen species. The importance of oxygen, and hence the availability of active oxygen species, to the reaction was confirmed by a control experiment under anoxic conditions (result not shown). The R50 under anoxic conditions was negligible at ∼0.1 μmol min−1. Given there is sufficient molecular oxygen present in the reactor (R50 increases with initial formic acid loading), the availability of the active oxygen species then becomes regulated by the extent of illumination of the sample. Fig. 2(b) indicates that while a 30 min pre-illumination time increases the number of active species by approximately two (over the dark condition), saturation of these species has not yet been achieved, suggesting the activated oxygen availability is rate limiting. However, Fig. 2(c) indicates that the initial formic acid

Catal. Sci. Technol.

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Paper

loading (at 100 μmol) is actually rate limiting, as when it is increased to 200 μmol the rate for the (30 min) light pretreated case also increases. It appears though that the maximum achievable oxidation rate following 30 min light pretreatment is ∼43 μmol min−1 (at least for this experimental array) as a further increase in the initial formic acid loading does not further increase the activity, despite there being an increase in activity under the ‘light’ condition. The instance of initial formic acid concentration restricting activity, following 30 min pre-illumination, can similarly be seen in the instantaneous rate profiles in Fig. 2(d). The maximum instantaneous reaction rate is indicative of the maximum O2 activation capacity. Again, the same maximum rate is attained for the 200 μmol and 1000 μmol formic acid loadings and is almost double that of the 100 μmol loading. What is more interesting in Fig. 2(d) though is the decay in the instantaneous rate for the 1000 μmol loading with time. The dotted line shows the instantaneous rate profile if the maximum rate was maintained over the entire degradation process (the decay in rate at the 13 min mark is based on the decay rate for the 200 μmol loading where the formic acid concentration becomes the rate-limiting factor, i.e. when there is less than 100 μmol of formic acid left in the system) and illustrates there is a significant divergence compared to the actual instantaneous rate. The actual instantaneous rate is seen to consistently decay with time with the deterioration attributed to the gradual loss in available active oxygen species, possibly reflecting: (a) competition between the transfer of electrons from the activated oxygen (liberated during formic acid oxidation) back into the gold deposit and then either (i) into the TiO2 support, i.e. relaxation of the photoexcited charges, or (ii) activation of another oxygen molecule or; (b) a change in the characteristics of the gold deposits over the course of the reaction. However, no physical or chemical changes were observed from TEM, XPS, and ICP characterisations, and catalytic activity results. The oxygen activation mechanism and role of Au. The elevated R50 values exhibited by Au/TiO2 as compared to TiO2 for both the dark and light pre-illumination conditions (Fig. 2(a)) may be attributed to enhanced charge separation in the TiO2 arising from the Au nanoparticle presence. Au has a greater work function (5.1 eV) than TiO2 (4.9 eV) such that when Au nanoparticles are deposited onto TiO2, the difference in work function allows the Au nanoparticles to extract and retain electrons from the TiO2 support.50–54 It has been reported that the work function of an Au nanoparticle is dependent on its charge state.55 As Au nanoparticles extract electrons from TiO2, the charge state of the Au increases and the work function decreases. The net charge transfer from TiO2 to Au will cease when charge equilibrium is achieved. Evidence of the transfer of charge from the TiO2 to the Au (and charge retention) can be seen in Fig. 3. The diminished PL signal at 412 nm when Au deposits are present on the TiO2 surface is attributed to a lower incidence of electron/ hole recombination within the TiO2.56 That is, the Au deposits attract, capture and retain electrons (in some form)

Catal. Sci. Technol.

Catalysis Science & Technology

Fig. 3 Photoluminescence spectra of neat TiO2 and 1.0 at% Au/TiO2 in water. Excitation wavelength: 320 nm. The presence of Au deposits (Au/TiO2 – red profile) on the TiO2 (black profile) decrease the electron–hole recombination rate as shown by the reduction in PL intensity at 412 nm. Light pre-illumination of the Au/TiO2 for 30 min (purple profile) and 60 min (pink profile) further reduces the recombination rate.

from the TiO2. This new state of charge equilibrium, where the Au deposits have become electron rich, may account for the increased catalytic activity of Au/TiO2 in the dark reaction when compared to the neat TiO2 (Fig. 2(a)). XPS results in Fig. 4 show that, in the as-prepared state, the Au is present in metallic form, indicated by the Au4f binding energy of 83.7 eV. The observed Au4f binding energy is slightly lower than that of the bulk Au at 84.0 eV and is attributed to the strong metal–support interaction, which increases the electron density within the Au nanoparticles.57 When Au/TiO2 is suspended in water (at pH 3) prior to preillumination (0 min pre-illumination), the Au4f binding energy undergoes a positive shift (by 0.4 eV, Fig. 4(i) and (ii)), indicating a loss in electron density. It is therefore anticipated that, upon immersion in water, Au donates electrons to surface adsorbed species. XPS also indicated evidence of water chemisorption occurring on the Au/TiO2 (Fig. 4(iv–vi), binding energy = 532.3 eV58) with the water thought to exist primarily on the TiO2 surface in this instance. It is important to note that the difference in the O1s spectra is small, arising from the presence of lattice oxygen in the TiO2 structure. The small difference is also due to the instability of trapped electron–hole states and superoxy O2− states on TiO2 at ambient condition, as highlighted in literature.59,60 Density functional theory calculations for aqueous phase systems have shown that it is unlikely for water molecules to adsorb on Au nanoparticles due to strong hydrogen bonding interactions between the water molecules.28 Consequently, the species adsorbed on the Au deposits are thought to be dissolved O2 molecules. It has been proposed that the first step of O2 activation on Au in aqueous phase systems is the formation of Au–OOH and Au–OH intermediates.7,28 In our case, the Au4f binding energy for the 0 min pre-illuminated

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Catalysis Science & Technology

Paper

Fig. 4 XPS spectra of: Au4f (i–iii), O1s (iv–vi), and Ti2p (vii–ix) for 1.0 at% Au/TiO2 as-prepared, at 0 min light pre-illumination (i.e. immersed in water), and following 30 min light pre-illumination. Binding energies are normalised to adventitious carbon C1s = 285.0 eV.

sample is too low to represent Au+, which is commonly reported to be at 84.9 eV.11,61,62 In turn, we tentatively attribute the shift in the Au4f binding energy (Fig. 4(i–iii)) upon immersion in water to signify the formation of AuOads species, similar to that observed in our earlier work on Pt/TiO2.23 That is, electrons were donated from Au deposits to adsorbed molecular O2 facilitating O2 activation. This may account for the partial quenching of the PL emission for Au/TiO2 compared to TiO2 (Fig. 3), which may be due to electrons from the TiO2 moving into the Au deposits where they activate molecular oxygen, in effect becoming trapped. The capacity for the Au deposits to activate molecular O2 is further demonstrated by the electrochemical ORR activity profile of the Au/ TiO2 without any light pre-illumination (ESI,‡ Fig. S6). The increasing current density with increasing bias (beyond −0.74 V) represents the activation of dissolved molecular oxygen by the Au deposits. Light pre-illumination is observed to have an impact on the Au/TiO2 PL spectra (Fig. 3) and XPS profiles (Fig. 4) relative to the non-illuminated case. Treating the Au/TiO2 with light further diminishes the intensity of the 412 nm PL peak, signifying a further decrease in electron/hole recombination within the TiO2. The decreasing PL signal with preillumination time may reflect the generation of additional Oads species. This idea is supported by the work of Stevanovic et al., which reported a decrease in PL signal upon exposing TiO2 to O2 in vacuum under UV irradiation.63,64 The adsorption of additional molecular O2 during pre-illumination may withdraw more electrons from the Au nanoparticles, hence

This journal is © The Royal Society of Chemistry 2016

further reducing the rate of recombination of the electron– hole pairs. Interestingly, light pre-illumination returns the Au4f (Fig. 4(iii)) binding energy to a value similar to the asprepared sample (i.e. prior to light pre-illumination) and reduces the Ti2p binding energy. Exposing the Au/TiO2 to light will photoexcite electrons in the TiO2 into the conduction band where they can either be trapped at charge deficient sites or move into the Au deposits and be trapped by surfaceadsorbed species (e.g. molecular O2). This also implies that the photogenerated holes may accumulate in the TiO2 where they could potentially be consumed by adsorbed water or surface OH species oxidation.23 Further photoexcitation of TiO2 when the active sites are occupied will lead to charge accumulation in the Au and TiO2 (increasing electron density), hence a decrease in their binding energies. The shift in the Au4f and Ti2p binding energies suggests both electron pathways are in effect. Fig. 5, which shows the ORR activity for different preillumination times (extracted from ESI,‡ Fig. S6), lends support to the idea that a portion of the photoexcited charges move into the Au deposits, where they are captured by adsorbed O2 species. As the light pre-illumination time is increased, the oxygen reduction reaction (ORR) current density increases (Fig. 5(a)) indicating there is an increasing amount of active oxygen species generated as the pre-illumination time is extended. Additionally, the onset potential decreases when the pre-illumination time increases (Fig. 5(b)) demonstrating that light pre-treatment is able to reduce the

Catal. Sci. Technol.

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Paper

Catalysis Science & Technology

Fig. 5 Impact of light pre-illumination period on oxygen reduction reaction electrocatalytic activity (vs. Ag/AgCl) for 1.0 at% Au/TiO2: effect on (a) current density at −1.0 V bias; (b) onset potential.

activation energy for active oxygen formation. Note that the difference in onset potential for no pre-illumination (i.e. 0 min) and five minutes light pre-illumination in Fig. 5(b) was within experimental error. Although no significant change in the onset potential was observed for these two con-

ditions, a lower applied potential was required to achieve the same current density following five min of light preillumination (ESI,‡ Fig. S6). This effect indicates light preillumination was still able to increase the electrochemical ORR efficiency even though no significant change in the

Fig. 6 STM images showing: (a) Au nanoparticle (pink cross) on rutile (110) single crystal; (b) Au-free region (blue cross) on rutile (110) single crystal; and (c) STS of the corresponding spots marked on the images. STS reveal the presence of new valence states above the valence band of TiO2 at the edge of the Au nanoparticle near the Au/TiO2 interface and a shift in the conduction band towards lower energy. (Vsample = 1.5 V, I = 100 pA, lock-in modulation V = 20 mV at 937 Hz).

Catal. Sci. Technol.

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Catalysis Science & Technology

Paper

Fig. 7 Total and projected electron densities of states in the energy region of the band gap for Au/TiO2: (a) anatase (101) surface; (b) rutile (110) surface. The vertical dotted line shows the energy of the highest occupied state (for the ground state). The yellow band in (a) highlights the new valence states from the Au–TiO2 interaction.

activation energy was observed. The increase in ORR current density with increasing light pre-illumination time coincides with the earlier observed effect on activity results (Fig. 2(b)). The effect of Au nanoparticles on the electronic band structures of TiO2 were investigated by STM and STS, which can precisely probe the local density of states (LDOS) at the atomic level. Instead of obtaining dI/dV vs. V spectra via mathematical calculation to reflect the LDOS,65 the present work provides direct experimental evidence of the LDOS via direct measurement of dI/dV vs. V. Here we use rutile (110), a typical TiO2 substrate, as a model surface, with Au nanoparticles deposited by MBE. As shown in Fig. 6(a) and (b), empty-states STM analyses for ∼2 nm Au nanoclusters (similar diameter to the Au nanoparticles prepared by the deposition-precipitation method) on a rutile TiO2 (110) surface were conducted using a positive sample bias voltage (Vsample = 1.5 V). The similarity in Au particle size indicates that the STM/STS results are reliable and supportive to findings in macroscopic catalytic reactions.66 The STS spectra in Fig. 6(c) were acquired for both the TiO2 surface and an Au nanocluster (sites highlighted by blue and pink crosses, respectively, in Fig. 6). The presence of Au nanoparticles on the TiO2 surface results in a high density of states at Vsample < −1.5 V in the STS, consistent with the XPS valence band spectra (Fig. S7‡). These states are most likely introduced by the Au nanoparticles. DFT calculations (Fig. 7) verify the STS finding, indicating that the presence of gold on a TiO2 surface (either anatase or rutile) introduces new electronic states above the TiO2 valence band. The DFT results also show that new electronic states are generated in the valence band as a result of the Au–TiO2 interactions, which act to promote some of the surface oxygen states to higher energies. The energy levels of the Au 5d3/2 valence band states and the O2p non-bonding states (for both lattice oxygen and adsorbed oxygen) are similar, as seen in Fig. S7,‡ which could promote

This journal is © The Royal Society of Chemistry 2016

strong Au–O interactions, in turn aiding O2 activation by weakening and then cleaving the O–O bonds during the aqueous phase reactions. The effect of these new valence states and the similarity in the Au 5d3/2 and O 2p non-bonding valence electron energy levels can be seen in the enhanced catalytic performance of Au/TiO2 under dark conditions (Fig. 2(a)) when compared to bare TiO2, and the catalytic performance of Au/TiO2 being further improved with UV light pre-illumination. The new valence states, or the shift in valence band maximum towards the Fermi level, are found to be most intense at the edge of the Au nanocluster, near the Au–TiO2 interface, and decrease on moving towards the centre of the Au nanoparticle (away from the metal–support interface – Fig. S8‡). The intensity gradient suggests that there are electron-rich sites at the Au: TiO2 perimeter, which could potentially be the key sites for O2 activation.

Scheme 1 Proposed electron pathways in the process of molecular oxygen activation on Au/TiO2 during UV light pre-illumination. Photoexcited electrons are: (i) trapped by sites on/in the TiO2; (ii) transferred into and then through the Au deposit where they are trapped by adsorbed molecular O2.

Catal. Sci. Technol.

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Paper

Besides the new valence states, an energy state 0.2 eV below the conduction band minimum of TiO2 was also observed in the STS spectrum of the Au nanoparticle. Similar unoccupied states below the TiO2 conduction band were also seen in the DFT results for the rutile surface (Fig. 7b) with these states ascribed to the formation of Ti3+ ions at the Au/ TiO2 interface arising from electron transfer from the Au to the Ti. Ti3+ states and the electron transfer from Au to Ti were not observed by XPS, most likely due to a low concentration and being overshadowed by the net charge transfer from TiO2 to Au, respectively. Note that, the presence of Ti3+ states was not observed by STM, although Ti3+ states can be linked to the local oxygen vacancies found on the surface of TiO2 (Fig. S8‡). The decrease in the conduction band energy level due to the presence of Au suggests that, upon UV excitation (i.e. UV pre-illumination), photoexcited charge from the TiO2 may be transferred to the Au nanoparticle (which acts as an electron sink) or to the Au/TiO2 interface. The electron transfer, as well as improving charge separation, may facilitate O2 activation during UV light pre-illumination, resulting in an increased formic acid oxidation rate after UV light preillumination as seen in Fig. 2. The findings imply that light pre-illumination enhances the O2 activation capacity of Au/TiO2 by first photoexciting electrons in the TiO2. The photoexcited electrons can follow one of two pathways (Scheme 1), either: (i) they are trapped by sites on/in the TiO2 formed when the Au/TiO2 is initially immersed in water; (ii) they move into and then through the Au deposits where they are trapped by adsorbed molecular O2. The latter pathway is the mechanism primarily responsible for the oxygen activation process, mainly occurring at the

Catalysis Science & Technology

Au:TiO2 perimeter sites. The resulting active oxygen species are then available to oxidise organics (such as formic acid) in solution. The extent of molecular oxygen activation which occurs is governed by system parameters such as the Au loading and the light pre-illumination period. The level of active oxygen species available decays with time and may arise from the return of electrons from the Au into the TiO2 (i.e. recombination). The present work also provides an insight on the role of Au in photocatalytic oxidation reactions. During UV light preillumination, the Au/TiO2 experiences the same photoexcitation process as in a photocatalytic reaction. However, the contribution of the TiO2 support arising from photoexcitation in a photocatalytic reaction is eliminated in the present work as the formic acid oxidation was performed following the pre-illumination step (i.e. in the dark). In contrast, as the Au is not excited under UV irradiation, the role of Au in the present work is expected to be the same as in a photocatalytic reaction. That is, besides improving charge separation by acting as an electron sink (inhibiting electron–hole recombination), the presence of Au nanoparticles creates the Au:TiO2 perimeter active sites for catalytic oxygen activation. Au deposit stability upon storage. The impact of storing the Au/TiO2 in the dark under ambient conditions over a fivemonth period on light pre-illumination performance was assessed with the findings presented in Fig. 8. Fig. 8(a) shows that the Au/TiO2 experiences a gradual deactivation during storage, losing up to 70% of its catalytic activity after five months of storage. Nanosized Au deposits on TiO2 have been reported to be unstable and may sinter at room temperature over time16 with a similar phenomenon observed in our Au/ TiO2. Fig. 8(b) illustrates the sintering effect with five months

Fig. 8 The effect of storage time in the dark and under ambient conditions on the: (a) formic acid oxidation rate of Au/TiO2 following 30 min preillumination. Catalyst loading = 1.0 g L−1, initial formic acid loading = 1000 μmol; (b) Au deposit size distribution on TiO2 when freshly prepared (top, particle count = 971) and following 5 months of storage (bottom, particle count = 327). The particle size range below 2 nm is highlighted in green.

Catal. Sci. Technol.

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Catalysis Science & Technology

of storage leading to an ∼0.5 nm increase in the average Au deposit diameter (for the as-prepared Au/TiO2) as well as the disappearance of >80% of the Au deposits below 2.0 nm. The increase in Au deposit size will promote a loss in perimeter interfacial area and loss in the fraction of edge atoms,19 with both reported to be detrimental for catalytic activity. Such an effect agrees well with the literature.19,67,68 For instance, Cosandey also reported that an increase of ∼1.0 nm in Au deposit size on their Au/TiO2 which was sufficient to cause up to 50% activity loss for CO oxidation.67 Additionally, Fujitani and Nakamura provided experimental evidence demonstrating a good correlation between the Au particle size on TiO2 and the turnover frequency with respect to perimeter Au atoms and exposed Au atoms.68

Conclusion The present work has demonstrated that light preillumination can enhance the O2 activation capacity of Au/ TiO2. Pre-illuminating the Au/TiO2 with UV light for 30 min enhanced the O2 activation capacity of Au/TiO2 by up to four times, although the extent of enhancement was a function of system parameters such as Au loading, light pre-illumination time and initial formic acid loading. Immersing the Au/TiO2 in water saw a transfer of electrons from the Au deposits to surface adsorbed species, thought primarily to be molecular oxygen. The activated oxygen species on the Au deposit surface then appear to be responsible for the ensuing (formic acid) oxidation reaction. The significance of Au nanoparticles supported on TiO2 was supported by STM studies and DFT calculations. Both STS and DFT results demonstrated that Au nanoparticles create new valence states, and support the theory that Au:TiO2 perimeter sites are the dominant catalytically active sites. Light pre-illumination promotes a greater degree of electron transfer from the TiO2 to the Au deposits, in turn facilitating further oxygen activation and enhancing the oxidation reaction, as was illustrated by the electrochemical ORR study and could also be a key charge transfer step in photocatalytic reactions. It appears that, as the reaction proceeds, the additional photoexcited charge from the light pre-illumination relaxes as was observed at a higher formic acid loading. The current study has demonstrated that pre-illuminating Au/TiO2 with light can be used to enhance reactions where oxygen activation is a critical component, such as the removal of organic pollutants and the ORR in fuel cells or energy storage systems.

Associated content Supporting information • Sample calculation of R50 from reaction curve. • TiO2 slab models used for DFT calculations. • XRD profiles of TiO2 and Au/TiO2. • XPS Au4f core levels showing heat treatment effect. • XPS C1s spectra showing UV light pre-illumination photocleaning effect.

This journal is © The Royal Society of Chemistry 2016

Paper

• I–V curves of Au/TiO2 for electrochemical ORR at 0–15 min pre-illumination. • XPS valence band spectra of TiO2 and Au/TiO2. • STM images and STS of Au/TiO2. • Line profile of Au nanocluster on TiO2. • Stability of catalytic performance of Au/TiO2.

Acknowledgements The work was supported by the Australian Research Council (ARC) under the Laureate Fellowship Scheme-FL140100081 and the Discovery Projects Scheme-DP140102581. The authors acknowledge Dr. Bill Bin Gong of the XPS Facility within the UNSW Mark Wainwright Analytical Centre for XPS support. The authors also acknowledge the use of facilities within the UNSW Mark Wainwright Analytical Centre. This research was also undertaken with the assistance of computational resources provided by the Australian Government through National Computational Infrastructure (NCI) under the National Computational Merit Allocation Scheme.

References 1 C. Song and J. Zhang, in PEM Fuel Cell Electrocatalyst and Catalyst Layers, Fundamentals and Applications, 2008. 2 Z. Suo, C. Ma, M. Jin, T. He and L. An, Catal. Commun., 2008, 9, 2187–2190. 3 N. Dimitratos, A. Villa, C. L. Bianchi, L. Prati and M. Makkee, Appl. Catal., A, 2006, 311, 185–192. 4 I. X. Green, W. Tang, M. Neurock and J. T. Yates, Acc. Chem. Res., 2014, 47, 805–815. 5 G. I. Panov, K. A. Dubkov and E. V. Starokon, Catal. Today, 2006, 117, 148–155. 6 J. Guzman, S. Carrettin, J. C. Fierro-Gonzalez, Y. Hao, B. C. Gates and A. Corma, Angew. Chem., Int. Ed., 2005, 44, 4778–4781. 7 C. Shang and Z. P. Liu, J. Am. Chem. Soc., 2011, 133, 9938–9947. 8 M. Haruta, J. Catal., 1989, 115, 301–309. 9 M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M. J. Genet and B. Delmon, J. Catal., 1993, 144, 175–192. 10 M. Haruta, CATTECH, 2002, 6, 102–115. 11 R. Kydd, J. Scott, W. Y. Teoh, K. Chiang and R. Amal, Langmuir, 2010, 26, 2099–2106. 12 Q. Xu, K. C. C. Kharas and A. K. Datye, Catal. Lett., 2003, 85, 229–235. 13 W.-Y. Yu, C.-P. Yang, J.-N. Lin, C.-N. Kuo and B.-Z. Wan, Chem. Commun., 2005, 354–356. 14 M. Haruta, Chem. Rec., 2003, 3, 75–87. 15 F. Moreau and G. C. Bond, Appl. Catal., A, 2006, 302, 110–117. 16 G. M. Veith, A. R. Lupini and N. J. Dudney, J. Phys. Chem. C, 2009, 113, 269–280. 17 Y.-K. Choi, S.-S. Seo, K.-H. Chjo, Q.-W. Choi and S.-M. Park, J. Electrochem. Soc., 1992, 139, 1803.

Catal. Sci. Technol.

View Article Online

Published on 06 October 2016. Downloaded by UNSW Library on 19/10/2016 05:16:56.

Paper

18 X. Tang, J. Chen, Y. Li, Y. Li, Y. Xu and W. Shen, Chem. Eng. J., 2006, 118, 119–125. 19 M. Haruta, Faraday Discuss., 2011, 152, 11. 20 X. Cheng, Z. Shi, N. Glass, L. Zhang, J. Zhang, D. Song, Z. S. Liu, H. Wang and J. Shen, J. Power Sources, 2007, 165, 739–756. 21 N. M. Markovic, T. J. Schmidt, V. Stamenkovic and P. N. Ross, Fuel Cells, 2001, 1, 105–116. 22 F. Denny, J. Scott, K. Chiang, W. Y. Teoh and R. Amal, J. Mol. Catal. A: Chem., 2007, 263, 93–102. 23 J. Scott, W. Irawaty, G. Low and R. Amal, Appl. Catal., B, 2015, 164, 10–17. 24 H. Einaga, A. Ogata, S. Futamura and T. Ibusuki, Chem. Phys. Lett., 2001, 338, 303–307. 25 S. Hwang, M. C. Lee and W. Choi, Appl. Catal., B, 2003, 46, 49–63. 26 F. Moreau and G. C. Bond, Catal. Today, 2007, 122, 1–6. 27 T. H. Tan, J. Scott, Y. H. Ng, R. A. Taylor, K.-F. Aguey-Zinsou and R. Amal, ACS Catal., 2016, 2, 1870–1879. 28 J. Liu, X.-M. Cao and P. Hu, Phys. Chem. Chem. Phys., 2014, 16, 4176–4785. 29 J. H. Yun, R. J. Wong, Y. H. Ng, A. Du and R. Amal, RSC Adv., 2012, 2, 8164–8171. 30 M. Jung, J. N. Hart, J. Scott, Y. H. Ng, Y. Jiang and R. Amal, Appl. Catal., A, 2016, 521, 190–201. 31 M. Jung, J. N. Hart, D. Boensch, J. Scott, Y. H. Ng and R. Amal, Appl. Catal., A, 2016, 518, 221–230. 32 C. L. Muhich, Y. Zhou, A. M. Holder, A. W. Weimer and C. B. Musgrave, J. Phys. Chem. C, 2012, 116, 10138–10149. 33 J. Zhang, M. Zhang, Y. Han, W. Li, X. Meng and B. Zong, J. Phys. Chem. C, 2008, 112, 19506–19515. 34 D. Jiang, S. H. Overbury and S. Dai, J. Phys. Chem. C, 2012, 116, 21880–21885. 35 R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V. R. Saunders and C. M. Zicovich-Wilson, Z. Kristallogr., 2005, 220, 571–573. 36 R. Dovesi, V. R. Saunders, C. Roetti, R. Orlando, C. M. Zicovich-Wilson, F. Pascale, B. Civalleri, K. Doll, N. M. Harrison, I. J. Bush, P. D'Arco and M. Llunell, CRYSTAL09, CRYSTAL09 User's Manual, University of Torino, Torino, 2009. 37 A. D. Becke, J. Chem. Phys., 1993, 98, 5648. 38 C. Lee, W. Yang and R. G. Parr, Phys. Rev. B: Condens. Matter Mater. Phys., 1988, 37, 785–789. 39 Y. Zhang, W. Lin, Y. Li, K. Ding and J. Li, J. Phys. Chem. B, 2005, 109, 19270–19277. 40 M. Catti, G. Valerio, R. Dovesi and M. Causà, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 49, 14179–14187. 41 J. Muscat, N. Harrison and G. Thornton, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 15457–15463. 42 J. Muscat, The phase stability, surface structure and defect chemistry of titanium dioxide from first principles techniques, University of Manchester, Manchester, 1999. 43 K. Doll, P. Pyykko and H. J. Stoll, Chem. Phys., 1998, 109, 2339–2345.

Catal. Sci. Technol.

Catalysis Science & Technology

44 N. Martsinovich and A. Troisi, Phys. Chem. Chem. Phys., 2012, 14, 13392. 45 J. J. Murcia, J. A. Navío and M. C. Hidalgo, Appl. Catal., B, 2012, 126, 76–85. 46 C. J. Howard, T. M. Sabine and F. Dickson, Acta Crystallogr., Sect. B: Struct. Sci., 1991, 47, 462–468. 47 L. Li, J. Yan, T. Wang, Z.-J. Zhao, J. Zhang, J. Gong and N. Guan, Nat. Commun., 2015, 6, 5881. 48 T. Akita, P. Lu, S. Ichikawa, K. Tanaka and M. Haruta, Surf. Interface Anal., 2001, 73–78. 49 A. Fujishima, X. Zhang and D. Tryk, Surf. Sci. Rep., 2008, 63, 515–582. 50 S. Trasatti, J. Electroanal. Chem. Interfacial Electrochem., 1972, 39, 163–184. 51 H. B. J. Michaelson, Appl. Phys., 1977, 48, 4729–4733. 52 R. Steim, F. R. Kogler and C. J. Brabec, J. Mater. Chem., 2010, 20, 2499. 53 A. Borodin and M. Reichling, Phys. Chem. Chem. Phys., 2011, 13, 15442–15447. 54 N. T. Khoa, S. W. Kim, D. H. Yoo, E. J. Kim and S. H. Hahn, Appl. Catal., A, 2014, 469, 159–164. 55 Y. Zhang, O. Pluchery, L. Caillard, A.-F. Lamic-humblot, S. Casale, Y. J. Chabal and M. Salmeron, Nano Lett., 2015, 15, 51–55. 56 H. L. Tan, X. Wen, R. Amal and Y. H. Ng, J. Phys. Chem. Lett., 2016, 1400–1405. 57 N. Kruse and S. Chenakin, Appl. Catal., A, 2011, 391, 367–376. 58 M. Wakisaka, H. Suzuki, S. Mitsui, H. Uchida and M. Watanabe, Langmuir, 2009, 25, 1897–1900. 59 E. Carter, A. Carley and D. Murphy, J. Phys. Chem. C, 2007, 111, 10630–10638. 60 T. Berger, M. Sterrer, O. Diwald, E. Knözinger, D. Panayotov, T. L. Thompson and J. T. Yates, J. Phys. Chem. B, 2005, 109, 6061–6068. 61 A. Y. Klyushin, T. C. R. Rocha, M. Hävecker, A. Knop-Gericke and R. Schlögl, Phys. Chem. Chem. Phys., 2014, 16, 7881–7886. 62 K. Qian, Z. Jiang and W. Huang, J. Mol. Catal. A: Chem., 2007, 264, 26–32. 63 A. Stevanovic, M. Büttner, Z. Zhang and J. T. Yates, J. Am. Chem. Soc., 2012, 134, 324–332. 64 A. Stevanovic, S. Ma and J. T. Yates, J. Phys. Chem. C, 2014, 118, 21275–21280. 65 M. Valden, X. Lai and D. W. Goodman, Science, 1998, 281, 1647–1650. 66 A. K. Santra and D. W. Goodman, J. Phys.: Condens. Matter, 2002, 15, R31–R62. 67 F. Cosandey and T. E. Madey, Surf. Rev. Lett., 2001, 8, 73. 68 T. Fujitani and I. Nakamura, Angew. Chem., Int. Ed., 2011, 50, 10144–10147.

This journal is © The Royal Society of Chemistry 2016