Microelectrode-based transient amperometry of O2 adsorption and desorption on a SrTiO3 photocatalyst excited under water

Oxygen evolution at water–solid interfaces is a key reaction for sustainable energy production. Although some intermediate states have been detected in transient absorption spectroscopy, the O2 evolution kinetics after the multi-step, four-electron oxidation of water remain unknown. In this study, transient amperometry with a microelectrode was applied to operando O2 detection over Al-doped SrTiO3 particles doubly loaded with RhCrOx and CoOy cocatalysts, an efficient photocatalyst for the overall water-splitting reaction. Electrochemical O2 detection at intervals of 0.1 s unexpectedly indicated instantaneous O2 adsorption and desorption in addition to steady, photocatalytic O2 evolution on the photocatalyst modified under intense light irradiation. We hypothesized that electrons excited in the conduction band were transferred to O2 in water thorough Ti cations neighboring an oxygen anion vacancy on the modified Al-doped SrTiO3. The negatively charged O2 was then bound to the Ti cations. It was neutralized and released when shaded through electron back-transfer to the conduction band. The hypothesized mechanism for O2 adsorption and desorption was compared with the photoinduced O2 desorption known to occur on anion vacancies of TiO2(110). The microelectrode-based transient amperometry demonstrated in this paper will be applied to many other phenomena at liquid–solid interfaces.

Molecular oxygen released in water is currently accumulated in the gas phase and quantified with gas chromatography once or twice per hour. An improved time resolution on the scale of minutes could be achieved with advanced setups that shorten the gas-sampling intervals. 21 The authors have been developing a method using microelectrode-based transient amperometry to further speed up operando O 2 detection. Our simple idea is to shorten the physical distance from the place of reaction to the electrode for detection, as was conducted in scanning electrochemical microscopy (SECM) suitable for O 2 concentration mapping. [22][23][24][25][26] In our latest study, 27 we used diffusion simulations and electrochemical detection on a microelectrode to determine the absolute O 2 evolution rate with a time resolution of 0.1 s on a SrTiO 3 photocatalyst film immersed in a nitrogenpurged, aqueous KCl solution.
In the present study, the photocatalyst film was excited in an aerobic solution exposed to air. Surprisingly, molecular oxygen having been dissolved in the solution was adsorbed on the irradiated film, and it was released when shaded from the irradiation. We hypothesized that bandgap-excited electrons were injected into the O 2 in the solution. After receiving the electrons, O 2 was negatively charged and bound to the Ti cations neighboring an oxygen anion vacancy on the surface of SrTiO 3 .

Experimental section
The photocatalyst was SrTiO 3 particles doped with Al cations. 28 SrTiO 3 , Al 2 O 3 , and SrCl 2 were mixed at a molar ratio of 100 : 2 : 10 and heated at 1423 K to produce Al-doped particles that were 200-500 nm in size. A RhCrO x cocatalyst (Rh 0.1 wt% and Cr 0.1 wt%) was loaded onto the particles using impregnation from an aqueous solution of Na 3 RhCl 6 and Cr(NO 3 ) 3 . A CoO y cocatalyst (Co 0.1 wt%) was additionally loaded by photodeposition in an aqueous solution of CoCl 2 . The RhCrO x cocatalyst was loaded for assisting electron-driven H 2 evolution, 29 whereas the CoO y cocatalyst was for hole-driven O 2 evolution. 18,30 This photocatalyst was active for the overall water-splitting reaction with an apparent quantum yield of 55%. The photocatalyst particles were suspended in water with SiO 2 nanoparticles (1 : 2 weight ratio), dropped on a frosted glass plate, and dried at 323 K to be fixed. Photocatalyst-coated plates prepared in a similar manner had previously been placed in a panel reactor and tested for large-scale water splitting under natural sunlight. 28 Details of the experimental setup are described in ref. 27 and its ESI. A millimeter-sized photocatalyst-coated glass plate was placed in a KCl aqueous solution (0.1 mol l À1 , pH = 7) that was exposed to air in this study. A platinum wire with a 10 mm radius was coated with glass to expose a metal section as the working electrode for O 2 detection. The electrode was immersed in the solution perpendicular to the photocatalyst film which was electrically isolated in the solution. The distance between the electrode and film was controlled in the 100-200 mm range with 0.1 mm precision. Molecular oxygen in the solution was electrochemically detected on the electrode with a time resolution of 0.1 s, which was biased at À0.5 V relative to a Ag wire as the counter electrode. A four-electron reduction reaction, O 2 + 2H 2 O + 4e À -4OH À and/or O 2 + 4H + + 4e À -2H 2 O, 31 produced an electrode current on the order of nano-amps that was proportional to the O 2 concentration. The proportional relation of O 2 concentration and electrode current was checked and confirmed. 27 The Ag wire served as a quasi Ag/AgCl reference electrode in the KCl solution. On the other hand, the oxidative reaction on the wire, Ag + h + -Ag + , as the counter electrode caused silver contamination in the solution to a limited extent. A freshly polished microelectrode was used in each set of measurements to minimize silver contamination on the microelectrode apex.

Results and discussion
The immersed film was side lit with ultraviolet (UV) light (center wavelength: 280 nm) passing through the solution, as illustrated in the table of contents entry. The light intensity was tuned to 3 W m À2 (4 Â 10 18 photons m À2 s À1 ) to probe lighttriggered O 2 concentration changes without producing O 2 and H 2 bubbles on the irradiated film. Fig. 1(A) presents the electrode current I that responded to light irradiation at an electrode-film distance d = 100 mm. The negative sign of the current indicates that electrons are transferred from the electrode to the solution. Molecular oxygen in the solution produced a background current of À6.5 nA. When the probe light was turned on at time zero (t = 0), the electrode current gradually increased to À9.5 nA; when the light stopped at 100.5 s, the current gradually decreased back to the background.
We had observed gradual current increases and decreases in response to UV light irradiation on the same photocatalyst, 27 where the whole setup was purged with N 2 to remove background O 2 in the solution. The current changes observed in the N 2 atmosphere were quantitatively interpreted with photocatalytic O 2 evolution on the irradiated film followed by O 2 Fig. 1 Microelectrode current in response to UV light irradiation. Irradiation of a photocatalyst film with light for probing (intensity: 3 W m À2 ) at t = 0-100.5 s in a KCl solution resulted in response (A). The film was then irradiated with intense light to modify the photocatalyst (30 W m À2 ) for 4 h in the solution. Immediate irradiation of the modified film with light for probing (3 W m À2 ) at t = 0-100.2 s resulted in response (B). The instantaneous current decrease and increase are marked in the panel. The film was removed from the solution and dried in air for 9 days. The dried film showed response (C) when immersed in the solution and irradiated with light for probing (3 W m À2 ) at t = 0-101.1 s. The electrodefilm distance d was set at 100 mm. diffusion into the bulk solution. The electrode current response shown in Fig. 1(A) increased and then decreased in a similar manner to what was reported in ref. 27 Hence, the response in Fig. 1(A) was ascribed to photocatalytic O 2 evolution on the irradiated film.
The film examined in Fig. 1(A) was then irradiated with UV light at 10 times the intensity (280 nm, 30 W m À2 ) for 4 h in the solution. The film was photocatalytically active enough to produce O 2 and H 2 bubbles during this intense irradiation. The electrode current exceeded the quantification limit, 10.5 nA. When the intense irradiation was suspended, the film was immediately probed with weak light (3 W m À2 ). Response (B) obtained in this way was qualitatively different from response (A). The electrode current was stable at À7.7 nA, indicating that O 2 was dissolved in the air-exposed solution prior to irradiation. When the probe light was switched on, the current instantaneously decreased from À7.7 to À7.5 nA. It then gradually increased, indicating photocatalytic O 2 production. When the probe light was stopped at t = 100.2 s, the electrode current instantaneously increased from À8.3 to À8.4 nA and gradually reduced back to the background level. We ascribed the instantaneous current increase coupled with the decrease to O 2 adsorption and desorption on the film triggered by light irradiation; background O 2 in the solution was adsorbed on the film when irradiated and was desorbed when shaded.
Pretreatment with intense light irradiation (30 W m À2 for 4 h) was required to modify the photocatalyst feasible for probelight-triggered O 2 adsorption and desorption. Immersing the photocatalyst film in the solution for 4 h without intense light irradiation provided no instantaneous current response, as shown in Fig. S1 in the ESI. † Intense irradiation on the microelectrode induced no current response in the absence of the film (Fig. S2, ESI †). In a K 2 SO 4 solution (0.1 mol l À1 ), the film was modified under intense light to present similar instantaneous current responses (Fig. S3, ESI †) being insensitive to the chemical identity of electrolytes. Oxidation of Cl À anions, if any on the UV-irradiated SrTiO 3 photocatalyst, did not matter to make the instantaneous current responses.
These experimental results supported our interpretation that probe-light-triggered O 2 adsorption and desorption occurred on the photocatalyst particles modified during the pretreatment.
After the film displayed response (B), it was removed from the solution and dried in air at room temperature for 9 days. This film was then immersed in the KCl solution and examined with probe light (3 W m À2 ) again. The observed response (C) displayed a gradual current increase and decrease, which were attributed to photocatalytic O 2 evolution; no instantaneous response could be assigned to O 2 adsorption and desorption. The modified photocatalyst particles returned to their original state when dried in air.
In the descriptions above, we qualitatively characterized the electrode current responses according to the presence or absence of instantaneous responses triggered by probe-light irradiation. We then quantitatively interpreted response (B) in Fig. 1, using a numerical simulation including instantaneous O 2 adsorption, photocatalytic O 2 evolution, instantaneous O 2 desorption, and O 2 diffusion in the solution. The simulation space defined in cylindrical coordinates, r and z, is illustrated in Fig. 2(A). The oxygen diffusion driven by a concentration gradient is described by Fick's law: where C is concentration, and D is the diffusion coefficient of O 2 (D = 2.5 Â 10 À5 cm 2 s À1 at 298 K). 32 The details of the finiteelement scheme, which integrates the equation to predict the electrode current, are described with Fig. S4 in ESI. † To fit the experimentally observed response shown in Fig. 1(B), we assumed probe-light-triggered O 2 adsorption at t = 0 À t, photocatalytic O 2 evolution at t = t À 100.2 s, and O 2 desorption at t = 100.2 s À (100.2 s + t), as depicted in Fig. 2(B). Positive and negative O 2 release rates in the unit of mmol m À2 s À1 were used to represent O 2 release and suction on the film, respectively. The time width t was not necessarily identical for the adsorption and desorption phenomena. However, a common t value was assumed to minimize the number of fitting parameters.
The O 2 release rate depicted in Fig. 2(B) represented a net release rate. When light-triggered O 2 absorption or desorption occurred with photocatalytic O 2 evolution in parallel, the net release rate was evaluated by fitting simulated current responses to the experimental result. The simulated current was fitted to the experimental result in two steps. In the first step, R photo and r evolution were optimized to follow the gradual current increase during probe-light irradiation, while t was set to zero. The other quantities (a, R g , d, R solution , Z solution , and Z photo ) were fixed at experimentally determined values, as described in the caption of Fig. 2. The radius of the irradiated photocatalyst film R photo was difficult to quantitatively determine since the light-spot boundary was oval and graded on the film. Thus, R photo and r evolution were optimized to fit the experimentally observed response. The obtained R photo was 800 mm, while r evolution was 0.085 mmol m À2 s À1 . In the second step, r adsorption and t were fitted to the observed response for t = 0-10 s. Fitting to the response in t = 0-10 s was long enough to evaluate the impulsive O 2 suction having occurred at t = 0 À t, since the time delay for diffusion across the electrode-film gap was characterized by d 2 /D = 4 s.
When irradiation was stopped, r desorption and t were optimized to follow the observed response at t = 100-110 s. The current response simulated in this way was compared with the observed response in Fig. 3, in which the background current (À7.7 nA) was subtracted to deduce the current change induced by probe-light irradiation. The simulated response reproduced the observed current in the three time windows, t = 0-250, 0-10, and 95-110 s, and the optimized parameters were r adsorption = À0.35 mmol m À2 s À1 , r desorption = 0.21 mmol m À2 s À1 , and t = 2.0 s. Two more response curves were experimentally observed on the same film upon changing d to 150 and 200 mm (Fig. 4). The two curves were fitted to validate the simulation scheme and the optimized parameters.
A negative release rate of À0.35 mmol m À2 s À1 was evaluated in the first 2.0 s of probe-light irradiation. The integrated quantity of adsorbed O 2 , which was calculated as the product of r adsorption and t, had a value of 4.2 Â 10 17 molecules m À2 . The integrated quantity corresponded to 6% of the surface unit cell density of crystalline SrTiO 3 , 6.6 Â 10 18 m À2 on its (100) truncation. 33 The integrated quantity suggests O 2 adsorption on SrTiO 3 rather than on the RhCrO x and/or CoO y cocatalysts loaded by 0.1 wt%. Partial contribution of the cocatalysts cannot be excluded, though. Since the film was composed of particulate materials, the unit cell density exposed to the solution should be larger than this number. Hence, the integrated quantity of 6% relative to the unit cell density provides the maximum estimate.
The integrated quantity of desorbed O 2 , which was calculated as the product of r desorption and t, had a value of 2.5 Â 10 17 molecules m À2 . The desorbed quantity was 40% smaller than the adsorbed quantity. The decremental difference was ascribed to O 2 desorption during probe-light irradiation or to the conversion of the adsorbed O 2 to resident species on the shaded film.
Finally, we considered the mechanism through which O 2 in the solution was adsorbed on the Al-doped SrTiO 3 particles. Several peroxotitanate complexes having Ti metal centers coordinated by O 2 2À ligands were previously reported. 34 We therefore hypothesize that negatively charged O 2 species, possibly O 2 2À , were bound to Ti cations on the surface of the SrTiO 3 during probe-light irradiation. Bandgap excitation by probe light provided two electrons, which were transferred to an impinging O 2 molecule. By shading the particles, the electrons were back-transferred to the photocatalyst particle, resulting in the release of the neutralized O 2 into the solution. Titanium cations exposed on pristine particles were unable to capture O 2 even under the probe light, as was evidenced by the absence of the instantaneous current response in Fig. 1(A). The surface of the particles was modified during the pretreatment under intense light irradiation (30 W m À2 ) to enable the capture of negatively charged O 2 . The modified surface returned to its original state when the film was dried in air at room temperature. The reversible state changes suggested that a limited number of oxygen anions were removed during the pretreatment and resultant anion vacancies served as active site for O 2 adsorption. The vacancies were healed on the dried film. If metal (Sr, Ti, Al, Cr, Rh or Co) cations were removed under irradiation and induced the light-triggered adsorption, then they could not be restored when the surface was dried in air.  Fig. 1(B) fitted with the finiteelement simulation of O 2 adsorption, evolution, and desorption on the photocatalyst film. The experimentally observed (red) and simulated (blue) current changes are shown in time windows of (A) 0-250, (B) 0-10, and (C) 95-110 s. Electrode-film distance d = 100 mm. UV light (3 W m À2 ) was irradiated for probing on the photocatalyst film at t = 0-100.2 s. Five fitting parameters were optimized: R photo = 800 mm, r evolution = 0.085 mmol m À2 s À1 , r adsorption = À0.35 mmol m À2 s À1 , r desorption = 0.21 mmol m À2 s À1 , and t = 2.0 s. An analogous mechanism for O 2 adsorption and desorption is known to occur on TiO 2 (110) wafers placed in vacuum. Experimental 35,36 and theoretical 37 studies have concluded that O 2 was adsorbed on Ti 3+ cations that neighbored an oxygen anion vacancy when the vacancy was thermally created on the wafer surface. The adsorbed O 2 receives two excess electrons that were originally localized on the Ti 3+ cations. When the wafer is bandgap excited, two holes attached to the negatively charged O 2 and the neutralized O 2 escaped as gas. Electron transfer and back-transfer switched the O 2 adsorption and desorption on the anion vacancy.
We further hypothesized how oxygen anions were removed from Al-doped SrTiO 3 particles during the pretreatment. Upon capture of two holes, an oxygen anion is neutralized and leave the surface. This process is known as photocorrosion; 38 however, SrTiO 3 is recognized as a stable semiconductor against photocorrosion. The neutralization of an oxygen anion leading to vacancy creation thermodynamically and kinetically competes with water oxidation, when a metal oxide particle is excited under water. The electrochemical potential for this phenomenon, i.e., corrosion potential, is known to be 1.42 eV vs. NHE on TiO 2 in water having a pH of 0. On the other hand, the potential for water oxidation is 1.23 eV at pH 0. When we assume an identical corrosion potential on the surface of Aldoped SrTiO 3 , photoexcited holes are thermodynamically transferred to water and not to surface oxygen anions.
A finite number of holes are still feasible for neutralizing oxygen anions when water oxidation is kinetically limited. The Al-doped SrTiO 3 photocatalyst examined in this study was doubly loaded with the CrRhO x cocatalyst for electron-driven H 2 production 29 and the CoO y cocatalyst for hole-driven O 2 production. 28 A finite number of holes could have been present in Al-doped SrTiO 3 particles even modified with the CoO y cocatalyst for water oxidation.
Thus, the authors hypothesized that a small number of oxygen anions were neutralized and removed during the pretreatment under intense light irradiation. Indeed, a previous study 39 reported photoetching of TiO 2 electrodes which is known to be thermodynamically stable oxide against photocorrosion. The selfoxidation of the cocatalysts, if any during the pretreatment, may have created oxygen vacancies on the cocatalysts and contributed to the probe-light induced O 2 adsorption and desorption. 3 LOER has been recently proposed as an electrochemical cycle of the O 2 evolution reaction on perovskite-structured metal oxide electrodes; it is different from the proton-coupled electron transfer 4 that was previously assumed as the mechanism.
To investigate our hypothesis, we compared the photocatalyst films before and after the pretreatment using X-ray photoelectron spectroscopy (Fig. S5, ESI †) and optical absorption (Fig. S6, ESI †). The results show no sign of reduced Ti cations. This is consistent with the proposed scheme (Fig. 5), which involves Ti cations only in the 4+ valence state.
The SrTiO 3 particles were doped with Al cations in this study to enhance water-splitting efficiency by limiting electron-hole recombination. Transient absorption studies 40,41 showed recombination rate in SrTiO 3 particles decreased by Na doping, whereas recombination rate was not yet examined on Al-doped SrTiO 3 . It is possible that the Al cations embedded in SrTiO 3 lattice played a positive or negative role in removing oxygen anions and in the instantaneous O 2 responses. This possibility should be examined in the future.
The light-triggered O 2 adsorption and desorption found in this study open an additional path for electron-hole recombination. Two excited electrons can be located in an O 2 adsorbed on the vacancy. These electrons recombine with two holes, thus releasing the adsorbed O 2 under steady light irradiation. Two electrons and holes are consumed to catch and release an O 2 molecule in the solution without splitting water.
In addition, the hypothesized hole-induced vacancy creation is a side-reaction that is problematic for photocatalyst durability. Durability of the Al-doped SrTiO 3 photocatalyst was enhanced over a span of 1300 h under simulated sunlight irradiation by loading the CoO y cocatalyst. 30 The limited durability in the absence of CoO y was interpreted with hole-induced degradation of the RhCrO x cocatalyst. The holeinduced vacancy creation hypothesized in this study may have been an additional reason of the limited durability.
A semiconductor photocatalyst that efficiently drives the water-splitting reaction is potentially modified by the holeinduced creation of anion vacancies since electron-hole recombination is limited to increased quantum yields. Note that the light intensity for vacancy creation was 30 W m À2 near UV intensity in natural sun light (40 W m À2 for wavelengths shorter than 400 nm). When this occurs on a photocatalyst under development, the tuning of cocatalysts to accelerate waterinduced, hole-consuming reactions can be a beneficial way to optimize the number of holes present in the semiconductor particles under excitation light of a given intensity.

Conclusion
In this study, we found that the O 2 adsorption and desorption on the Al-doped SrTiO 3 photocatalyst particles doubly loaded with RhCrO x and CoO y cocatalysts were triggered by photoexcitation in water. Electrochemical detection on the microelectrode combined with diffusion simulations allowed us to quantitatively determine the rates of O 2 adsorption, O 2 desorption, and photocatalytic O 2 evolution. The time resolution for O 2 detection, 0.1 s, was critical for recognizing the instantaneous O 2 adsorption and desorption that occurred during steady O 2 evolution. The transient amperometry using a microelectrode can be applied to many other phenomena at liquid-solid interfaces.
The light-triggered O 2 adsorption and desorption were interpreted as the excited-electron transfer to and back-transfer from the adsorbed O 2 . Upon receiving electrons, O 2 in the solution was negatively charged and could bind to the surface Ti 4+ cations, while shaded particles could neutralize and release O 2 by back-transferring the electrons. Intense light irradiation (30 W m À2 ) was required to modify the particle to enable O 2 adsorption. We hypothesized that a small number of lattice oxygen anions were neutralized by receiving holes to leave the surface under intense irradiation. Anion vacancies created in this way served as O 2 adsorption sites. Future studies that examine the role of the Al cations, RhCrO x cocatalyst, and CoO y cocatalyst are needed to verify the hypothesized scheme.

Conflicts of interest
There are no conflicts to declare.