Application of surface complexation modeling on modification of hematite surface with cobalt cocatalysts: a potential tool for preparing homogeneously distributed catalysts

Abstract

The knowledge required to synthesize homogeneously distributed catalysts on the support surfaces is strongly reliant on understanding the interaction between the catalyst and surface of support, which is identical to the descriptions of interactions between environmental pollutants and adsorbents in the field of environmental chemistry. The connection between catalyst synthesis and environmental chemistry is demonstrated herein by using surface complexation modeling (SCM), whose physical foundation is built on the stoichiometric reaction between one solute and one adsorption site on the surface of adsorbents. As suggested by our simulation and supported by TEM images, the manipulation of either homogeneous distribution or coarse clusters of cobalt cocatalysts over hematite surfaces can be achieved by adjusting the pH and initial Co²⁺ concentration. This is attributed to the conversion of adsorbed Co²⁺ or Co(OH)₂ precipitates on hematite surfaces. Different surface distributions of cobalt cocatalysts further influence the reactivity of photodegradation of organic dyes and photoelectrochemical reactions, which is strongly dependent on the occurrence of the Fenton reactions or charge transport. Our results not only demonstrate the potential of SCM on guiding the preparation of homogeneously distributed catalysts but also indicate that the abundant amount of knowledge in environmental science could be adopted to design an appropriate protocol for catalyst synthesis.

---

Introduction

The idea of single-atom catalysts (SACs) is to greatly enhance the utilization efficiency of noble metals on support surfaces.¹ In comparison with nanoparticles and clusters, the surface free energy of SACs was maximized and stabilized via the formation of metal–O-support bindings.^2,3 Several methods have been proposed to prepare SACs, including impregnation or ion-exchange techniques, coprecipitation, colloidal deposition method, templates, and adsorption of metal complexes followed by thermal treatment.⁴ Among them, the last one has been widely adopted as a major route to prepare SACs due to the simplicity of preparation. However this preparation route strongly relies on the knowledge regarding the interaction between metal complexes and the support surface, which leads to limited success in the literature. Despite the following thermal treatment, another important factor as it often invariably leads to a variety of mixed oxides, knowledge to manipulate homogeneously distributed metal complexes over support surfaces is highly desired in the first place. Although the relevance may not sound obvious, this issue actually shares remarkable similarities with the one of the most important research topics in the environmental science community-understanding the mechanism of pollutant immobilization by adsorbents such as clay, zeolite, iron oxides and others, which also involves the interaction between metal complexes and adsorbents. In these cases, surface complexation modeling (SCM) have been successfully applied to simulate and explain the interactions between adsorbent surfaces and metal complexes of concern.^5–7 The physical basis of SCM is built on the stoichiometric adsorption reaction between one surface adsorption site and one metal complex/ion driven by electrostatic interaction in a manner consistent with mass and charge balances.^8,9

Taking hematite (Fe₂O₃) as an example, it is one of the most important nature occurring minerals regulating the transport of pollutants in aquifer and soil. Hematite is also known as a promising photocatalyst, whose widespread application is currently limited by the slow water oxidation kinetics. By associating with a few weight percentage of cobalt-based cocatalyst, its photocatalytic performance is demonstrated greatly improved as a result of the reduction in the recombination losses and thus more photogenerated holes are able to participate in water oxidation.^10–13 Several methods were proposed to associate cobalt cocatalyst onto hematite surface, including atomic layer deposition (ALD),¹⁴ (photo)electrodeposition,^13,15–18 thermal hydrolysis,^10–12 and post treatment.¹⁹ These preparation methods, except ALD and thermal hydrolysis, are conducted in a solution containing different concentrations of Co²⁺, sometimes along with some structure directing agents. Despite their success in producing efficient hematite photoelectrodes, our preliminary SCM simulation suggests that the surface of hematite on photoelectrodes cannot accommodate such a large amount of Co-cocatalysts if they were associated by adsorption reaction. For example, the maximum surface adsorption site density based on crystallographic calculation of surface hydroxyl configuration of hematite was determined to be 7.5 sites per nm².²⁰ Taking the most widely studied hematite photoelectrode with a working area of 1 cm² as an example, the maximum number of Co²⁺ adsorbed on this surface would be no greater than 7.5 × 10¹⁴ (∼1.25 nmole, assuming a nonporous surface). This value is obviously several orders of magnitude smaller than any reported values (a few weight percent) in the literature. This simulation apparently conflicts with the observations where a thin Co₃O₄ layer (∼tens to hundreds nm thickness) is found on the top of hematite layer in these successful cases. In this configuration, the photogenerated holes that participate in water oxidation at Co₃O₄/water interface are actually those that have to transport through additional thousands of grain (or grain boundary) of in the Co₃O₄ layer. This also reflects the fact that the knowledge regarding to the hematite/Co interface phenomenon is not yet fully understood. Accordingly, an effective method to prepare homogeneously distributed Co₃O₄ over the surface of hematite is highly desired to fulfill the need for the future exploration in hematite/Co interface phenomenon.

In this study, we demonstrate the application of SCM as guidance to manipulate the distribution of Co₃O₄ on hematite surface (Scheme 1). This is possible, as described previously, because the physical foundation of surface complexation modeling is built on the stoichiometric reaction between the solutes and adsorbent surfaces as shown in eqn (1),

SOH + Co²⁺ ↔ SOCo⁺ + H⁺ (1)

where S refers to the adsorption site on hematite surfaces.^7,8,21 Also, a detail discussion toward the mechanism behind those successfully synthesizing high efficiency cobalt decorated hematite electrodes is given from the point of view of surface chemistry. An illustration of the configuration of SCM is shown in the Scheme 1 in the ESI.†

Scheme 1 The illustration of how surface complexation modeling can be applied to design an appropriate synthesis protocol to manipulate the distribution of Co cocatalysts on hematite surfaces.

Experimental

All chemicals were ACS grade and deionized water was used throughout all experiments. For hematite nanoparticle synthesis, two liters 20 mM FeCl₃ aqueous solution was sealed in a serum bottle and aged at 100 °C for 2 days.²² Cobalt adsorption experiments were conducted in triplicate by mixing 1.0–10 g hematite with 100 mL 10–0.01 mM Co²⁺ solutions (prepared by dissolving aliquot amount of Co(NO₃)₂·6H₂O, Sigma-Aldrich) under different pH environments. In the end of 24 hours of adsorption, the mixtures were centrifuged and the concentrations of Co²⁺ remaining in supernatants were determined by ICP-MS (7500a ICP-MS, Agilent). The difference in the Co²⁺ concentration in the supernatant was assumed the amount of Co adsorbed/precipitated onto hematite surfaces. Detailed experimental conditions of each sample are listed in Table S1.† On the other hand, the centrifuged mixtures were first dried at 60 °C and then subjected a two-step calcination treatment (500 °C 2 hours and then 700 °C 20 minutes) and stored under ambient environment before further photocatalytic experiments. The water chemistry and SCM simulation were conducted using MINEQL, an aqueous chemical equilibrium calculation software. Several assumptions were made prior to SCM simulations in order to keep the number of adjustable parameters as low as possible. First, the surface adsorption site density was assumed to be 7.5 sites per nm².²⁰ Given that Co²⁺ adsorption experiments were conducted with a solid/liquid ratio of 1.0 g hematite/100 mL solution, the corresponding adsorption site concentration in the system was 0.71 mmol L⁻¹ (the surface area of hematite particles determined by N₂-BET method is 8.47 m² g⁻¹). Second, as the equilibrium constant is a state function, we adopted the log K_a1 as 6.9 for protonation reaction and log K_a2 as −12.8 for the deprotonation reaction of hematite, respectively (eqn (2) and (3)).⁷ Last, we assumed the hematite surface is homogeneous, although it is suggested that there are roughly 2% of the surface adsorption sites belonging to the strong sorption sites.²¹

SOH + H⁺ ↔ SOH₂⁺ (2)

SOH ↔ SO⁻ + H⁺ (3)

The photocatalytic activity of as-obtained cobalt-loaded hematite samples was evaluated by both rhodamine B (RhB) dye photodegradation (conducted in pH 7.2 ± 0.2) and by water splitting reaction (1.0 M NaOH solution, pH = 13.6) under the illumination of 100 mW cm⁻² simulated sunlight (AM 1.5G). For RhB photodegradation experiments, 0.02 g of hematite samples were dispersed in 100 mL of solution containing 0.02 mM RhB and 50 mM H₂O₂.²⁶ The assembly of cobalt-loaded samples onto FTO substrates was conducted following by our previous proposed DA method.²⁹ All photoelectrochemical measurements including impedance spectroscopy were conducted using a three-electrode configuration: a hematite photoanode as the working electrode, a saturated calomel electrode (SCE) as the reference electrode, and platinum foil as the counter electrode in 1 M NaOH and a CH Instrument 604E electrochemical workstation under 100 mW cm⁻² illumination. Obtained data was fitted by an equivalent circuit model using Zview software.

Results and discussion

We first conducted a series of Co²⁺ adsorption experiments with synthesized hematite nanoparticles and fitted as-obtained experimental sets with SCM to determine the adsorption equilibrium constant.

Based on above assumptions, the best fitting was shown in Fig. 1a, where the best fitting curve (black) is the sum of the fraction of adsorbed Co²⁺ (red) and those of precipitated Co(OH)₂ (blue). It is worth mentioning that although the adsorption edge can be fitted nicely, the fitting curve failed to describe the adsorption behaviors at pH > 7. Specifically, the fitted adsorption plateau occurred at the region of neutral pH with a Co uptake around 0.7 (red, Fig. 1a), while experimentally observed adsorption plateau was near unity. In the framework of SCM, the adsorption plateau physically means that all active adsorption sites are occupied by Co²⁺ cations under equilibrium. The discrepancy between the fitting and experimental results can be explained by the formation of surface precipitation (blue, Fig. 1b). The term surface precipitation is particularly standing for a phenomenon where the adsorption isotherm departs from the linearity. This is because the amount of loaded Co species obviously exceeds the adsorption capacity of hematite, leading to a solid-like precipitate of Co species over hematite surface. Indeed, the abrupt rising of Co²⁺ uptake experimentally observed in Fig. 1a stemmed from the formation of Co(OH)₂ precipitate, as confirmed by our simulation. This means that the observed Co²⁺ uptake at basic pH was a result of the combination of adsorbed Co²⁺ and precipitated Co(OH)₂ (black, Fig. 1a). This is very important as the adsorption experiment only allows us to evaluate the concentration difference in the solution but cannot tell the reaction and mechanism how these Co species are removed by hematite. Additional aqueous chemistry simulations showed that the threshold concentration for cobalt ions to form Co(OH)₂ precipitates was around 1 mM (blue ▽, Fig. 1b). Furthermore, this threshold concentration shifted to lower pH values as initial Co²⁺ concentration increased (blue ○, Fig. 1b). In consistent with results reported by Kobal et al.,²³ this means a high initial Co²⁺ concentration is prone to induce hematite surface covered with a large number Co(OH)₂ precipitates.

Fig. 1 (a) Result of Co adsorption experiments and fitting (initial Co²⁺ concentration = 1 mM; solid/liquid ratio = 1 g/100 mL; fitting parameters: adsorption reaction equilibrium constant pK = 1.5; red: the fraction of adsorbed Co²⁺ species; blue: the fraction of Co(OH)₂ precipitate; black: the sum of the fraction of adsorbed Co²⁺ and Co(OH)₂ precipitate); (b) the simulation of relative Co species distribution in different initial Co²⁺ conditions; red: adsorbed Co²⁺ species, blue: the fraction of Co(OH)₂ precipitates.

The Co association to hematite surface by adsorption (case A, Fig. 1a) and precipitation (case B, Fig. 1b) can be visualized from TEM images as shown in Fig. 2. The hematite sample prepared in a condition with 1 mM Co²⁺ at pH 8 (equilibrated pH) is expected to have Co²⁺ stoichiometrically adsorbing on the hematite surface without any Co(OH)₂ precipitate. These adsorbed Co²⁺ cations were subsequently converted into fine Co₃O₄ nanoparticles in the end of thermal treatment (confirmed by XPS, Fig. S1†) that is barely observable in the TEM image (Fig. 2a). Under the high magnification, a thin layer with thickness ∼2 nm on the surface of hematite is observed (Fig. 2b), indicating Co-cocatalysts prepared through the adsorption would be in favor of developing into a thin Co₃O₄ layer. Similar phenomenon has been reported by Deng et al., where a thin layer of carbon (∼2 nm) on the surface of hematite was observed as a result of pyrolysis of ferrocene. This thin carbon layer further modifies the electronic structure of hematite, leading to a photocurrent about four times higher than the those without carbon layer.²⁵ In contrast, those obtained from the case B condition contained a large amount of visible Co₃O₄ clusters with size approximately in the range around 10–30 nm (Fig. 2c). Again, subsequent thermal treatment may produce mixed oxides of iron and cobalt which is difficult to distinguish by the darkness in Fig. 2c. However, it is certain that the coarsen Co₃O₄ clusters are a result of Co(OH)₂ precipitation while homogeneously distributed Co₃O₄ could be achieved through the adsorption. Based on these observations, we propose two likely scenarios of how surface Co₃O₄ were dispersed on hematite surfaces (inserted panels in Fig. 2a and c). First, homogeneously distributed Co₃O₄ can be prepared by dispersing hematite particles in a solution with neutral pH and a low initial Co²⁺ concentration, i.e., 1 mM (case A, Fig. 1a) to avoid the formation of Co(OH)₂ precipitates. By contrast, Co₃O₄ clusters could be obtained by preparing samples in a solution containing high Co²⁺ concentrations at high pH (case B, Fig. 1a). Our speculation is substantially supported by results from following zeta potential measurements (Fig. S2†). As expected, the zeta potential of hematite is pH sensitive with the point of zero charge (pH_pzc) around pH 8.8. The zeta potential of hematite with a thin layer of Co cocatalyst (case A) resembles that of hematite but its pH_pzc is shifted to ∼pH 7.2. In contrary, the zeta potential of hematite with coarse Co₃O₄ clusters (case B) becomes less pH dependence, possessing a negative surface charge around −10 mV when pH exceeding pH 4.8. The reported pH_pzc of Co₃O₄ using both potentiometric titration and zeta potential measurement was determined around pH 7.5,²⁶ which is very close to that of case A samples. This indicates that the surface chemistry of case A samples is very similar to that of Co₃O₄, which is a complementary evidence besides the observations from TEM image (Fig. 2b). The cause of the surface chemistry of case B samples behaving pH-independent remains unclear; however, it should be closely related to the interactions between hematite and Co₃O₄ clusters on hematite surfaces.

Fig. 2 Representative TEM images and illustrations of hematite covered with (a) a monolayer of Co₃O₄ nanoparticle (case A in Fig. 1a, barely visible), (b) high magnification image of the zone indicated in (a) and (c) with Co₃O₄ clusters (case B, in Fig. 1a). Inserted magenta circles refer to Co₃O₄ cocatalysts.

Based on our results, SCM is indeed a powerful tool to design an appropriate protocol for the dispersion of Co₃O₄ on hematite surfaces. Also, SCM simulation allows us to understand the mechanisms regarding to how Co₃O₄ is associated on hematite surfaces by electrolysis¹⁵ and photodeposition.^13,16,18,27 In these cases, the association of Co₃O₄ to hematite surfaces was conducted in a solution containing 1.0 M phosphate with the initial Co²⁺ concentration of 0.5 mM in the neutral pH environment. Although the effect of pH on ions adsorption to the support is well known for a long time, we herein attempt the address this phenomenon from the point of view of aqueous and surface chemistry considerations based upon the SCM simulations. As demonstrated in Fig. 1b, a low initial Co²⁺ concentration prevents the formation of Co(OH)₂ precipitates, whereas a neutral pH guarantees the deprotonation of hydroxyl groups on hematite surface so that they are available for Co²⁺ accommodation. This condition is in favor of producing fine Co₃O₄ clusters as they are known possessing high reactivity (size smaller than 5 nm).^10,27 This is very important since the charge transfer from hematite to Co₃O₄ clusters is much kinetically favorable when they are located inside the Helmholtz layer of the hematite surfaces.²⁷ In contrast, those Co₃O₄ loaded hematite samples prepared through the hydrothermal reaction are expected to be possessing very limited Co₃O₄ clusters as SCM simulations predict that cobalt adsorption to hematite in an acidic environment is thermodynamically unfavorable (Fig. 1b). There may be a small population of Co²⁺ allowed to be uptake by strong adsorption sites, which accounts for <2% of total adsorption sites on hematite surfaces.⁹ In this case, the amount of loaded Co₃O₄ would be too small to have any significant influence on photocatalytic performance enhancement. Therefore, the possible mechanism responsible for these successful studies through the hydrothermal reaction is by the consideration of the deposition of hydrolysis product of Co(OOH) on hematite surface.^11,12 In these cases, following thermal treatment transforms Co(OOH) precipitates into coarse Co₃O₄ clusters and thus enhanced photocatalytic activity was achieved.

In order to demonstrate the effect of different distributions of surface Co₃O₄ on photocatalytic performance, RhB dye photodegradation and water splitting reactions were conducted under simulated sunlight illumination. Fig. S3† shows the representative photodegradation profiles of RhB dye using different hematite samples. The results show that although Co₃O₄ is an effective catalyst in water oxidation, charge carrier transfer (photogenerated holes) is apparently not the dominant step in the photodegradation of RhB dyes. Instead, it is likely the Fenton reaction is mainly responsible for organic dye photodegradation.²⁸ In this case, the degradation of RhB dyes under visible light illumination is suggested following two successive stages.²⁹ First, dye molecules are photoexcited and the electrons transfers from the excited dye molecules to surface Fe³⁺ ions, which is accompanied with reduction of Fe³⁺ to Fe²⁺. Reduced Fe²⁺ ions then react with H₂O₂ to generate OH radicals which initiate the photocatalytic reaction to mineralize the organic dyes. In this sense, the higher Co₃O₄ loading on the surface of hematite, the less surface Fe³⁺ are available to participate in the Fenton reaction. This explains the decrease in the decay constant k along with increasing surface Co₃O₄ loadings (Fig. 3a). It is worth mentioning that distinguishing the influence of Co₃O₄ that is converted by Co(OH)₂ precipitates from those by adsorption is very difficult as the former always contained a significant fraction of the latter (blue symbols in Fig. 1b). However, it is clear that once hematite surface is covered with Co₃O₄, fewer surface Fe³⁺ ions are available for participating in Fenton reaction and thus slower photodegradation of RhB dye is observed.

Fig. 3 (a) Determined RhB dye photodegradation constants and (b) photocurrent density of hematite particles decorated with different cobalt cocatalysts as a function of surface cobalt cocatalyst loadings.

When assembling these cobalt decorated hematite samples onto FTO substrates, we noted that the surface Co₃O₄ substantially improved the photocurrent density (Fig. S4†). In contrary to the observations noted in the case of RhB dye photodegradation, photocurrent indeed increases with increasing surface Co₃O₄ loadings (Fig. 3b). This means charge transfer reaction is the essential step accounting for the photocurrent enhancement.¹⁹ Importantly, the degree of photocurrent enhancement is strongly dependent on the dispersion of Co₃O₄ on hematite surfaces (Fig. 3b). For instance, observed photocurrent increased steeply as surface Co₃O₄ loading increased to a level around 10⁻⁴ mol g⁻¹. This enhancement is suggested as a result of surface Co₃O₄ effectively retarding the electron/hole recombination.^13,17 In this way, surface Co₃O₄ plays as a surface state at which photogenerated holes are transferred from hematite to Co₃O₄ and then to the electrolyte.³⁰ By contrast, the photocurrent stops significantly increasing when hematite was covered with Co₃O₄ clusters (with surface Co₃O₄ loading exceeding 10⁻⁴ mol g⁻¹). Again, this is in a good agreement with literature that coarse Co₃O₄ is less reactive than the finer ones.^10,24

More piece of evidence regarding to the effect of surface Co₃O₄ distribution on water-splitting performance of hematite can be found by using the method of electronic impedance spectroscopy. The simplified equivalent circuit adopted herein was to emphasize the role of surface states stemming from the surface Co₃O₄ as shown in Fig. S5.† This equivalent circuit contains a charge transfer resistance (R_ct,trap) and a capacitance (C_trap) of trapping states, which is suggested to be a good approximation for the charge transfer from the surface trap state of surface Co₃O₄ to water molecules.³⁰ Fig. S5† shows the representative EIS spectra collected under illumination with an applied potential of 0.4 V vs. SCE. The nice fitting curves suggest that the adopted equivalent circuit can be applied to describe the surface states of these cobalt decorated hematite samples. When plotting as-determined C_trap against applied potential under illumination (Fig. 4a), it was noted that the charging of the surface state consistently shifted towards lower applied potential as increasing surface Co₃O₄ loadings. This could be attributed to the formation of a Schottky-type hetero-junction at the hematite/cobalt oxide/electrolyte interfaces.¹³ Although the charging of surface states was obvious, we did not observe a concurrent decrease in R_ct, as noted in literature²⁴ (Fig. 4b). A possible explanation is that the huge charge transport resistance in hematite/FTO interfaces overwhelms the reduction of R_ct, as these hematite electrodes were prepared through the dipping-annealing method. In this case, the resistance at hematite/FTO interface was found to be in a level approximately several megaohm.³¹ We are currently working on associating Co₃O₄ to hydrothermally grown hematite electrodes and the results will be submitted separately.

Fig. 4 Determined (a) C_trap and (b) R_ct,trap among hematite samples decorated with different amount of surface cobalt cocatalyst loadings.

Conclusions

By adopting the knowledge of surface complexation modeling, we have demonstrated in this study the preparation of homogeneously distributed Co₃O₄ on hematite surfaces. Also, the mechanism behind these preparation processes has addressed from the point of view of aqueous and surface chemistry. First, a neutral or an alkaline environment is very important for the surface hydroxyl groups of hematite to deprotonate and then stoichiometrically accommodate Co²⁺. The concentration of the initial Co²⁺ is another important parameter as a high Co²⁺ concentration could result in the precipitation of Co(OH)₂, which would be converted to coarse Co₃O₄ in the end of thermal treatment. Different surface distributions of Co₃O₄ directly affect the efficiency of water oxidation reactions and photodegradation of organic dyes. Finally, our results successfully bridged the knowledge between environmental sciences and material chemistry.

Acknowledgements

We thank National Science Council, Taiwan (NSC102-2113-M-007-003-MY2) for supporting this study. The authors greatly appreciate the access of facilities at CNMM of NTHU. WW is grateful for the financial support from National Natural Science Foundation of China (Grant No. 11402069) and Shenzhen Peacock Technological Innovation Program (Grant No. KQCX20140521144102503).

Notes and references

1. J. Lin, A. Q. Wang, B. T. Qiao, X. Y. Liu, X. F. Yang, X. D. Wang, J. X. Liang, J. Li, J. Y. Liu and T. Zhang, J. Am. Chem. Soc., 2013, 135, 15314–15317.
2. J. Thomas, Z. Saghi and P. Gai, Top. Catal., 2011, 54, 588–594.
3. X. Yang, A. Wang, B. Qiao, J. Li, J. Liu and T. Zhang, Acc. Chem. Res., 2013, 46, 1740–1748.
4. F. Schuth, Angew. Chem., Int. Ed., 2014, 53, 8599–8604.
5. T. H. Wang, C. J. Hsieh, S. M. Lin, D. C. Wu, M. H. Li and S. P. Teng, Environ. Sci. Technol., 2010, 44, 5142–5147.
6. C. Tiberg, C. Sjostedt, I. Persson and J. P. Gustafsson, Geochim. Cosmochim. Acta, 2013, 120, 140–157.
7. M. Gunnarsson, A.-M. Jakobsson, S. Ekberg, Y. Albinsson and E. Ahlberg, J. Colloid Interface Sci., 2000, 231, 326–336.
8. W. D. Schecher and D. C. McAvoy, MINEQL+: a chemical equilibrium modeling system, version 4.5 for windows, user’s manual, v2.00, Environmental Research Software, Hallowell, Maine, 2003.
9. D. A. Dzombak and F. M. M. Morel, Surface complexation modeling: hydrous ferric oxide, Wiley-Interscience, New York, 1990.
10. L. F. Xi, P. D. Tran, S. Y. Chiam, P. S. Bassi, W. F. Mak, H. K. Mulmudi, S. K. Batabyal, J. Barber, J. S. Loo and L. H. Wong, J. Phys. Chem. C, 2012, 116, 13884–13889.
11. R. Suresh, R. Prabu, A. Vijayaraj, K. Giribabu, A. Stephen and V. Narayanan, Mater. Chem. Phys., 2013, 134, 590–596.
12. T. P. Almeida, M. W. Fay, Y. Q. Zhu and P. D. Brown, J. Mater. Sci., 2012, 47, 5546–5560.
13. M. Barroso, A. J. Cowan, S. R. Pendlebury, M. Gratzel, D. R. Klug and J. R. Durrant, J. Am. Chem. Soc., 2011, 133, 14868–14871.
14. S. C. Riha, B. M. Klahr, E. C. Tyo, S. Seifert, S. Vajda, M. J. Pellin, T. W. Hamann and A. B. F. Martinson, ACS Nano, 2013, 7, 2396–2405.
15. M. W. Kanan and D. G. Nocera, Science, 2008, 321, 1072–1075.
16. D. K. Zhong, J. W. Sun, H. Inumaru and D. R. Gamelin, J. Am. Chem. Soc., 2009, 131, 6086–6087.
17. M. Barroso, C. A. Mesa, S. R. Pendlebury, A. J. Cowan, T. Hisatomi, K. Sivula, M. Grätzel, D. R. Klug and J. R. Durrant, Proc. Natl. Acad. Sci. U. S. A., 2012, 25, 15640–15645.
18. K. J. McDonald and K. S. Choi, Chem. Mater., 2011, 23, 1686–1693.
19. R. Franking, L. S. Li, M. A. Lukowski, F. Meng, Y. Z. Tan, R. J. Hamers and S. Jin, Energy Environ. Sci., 2013, 6, 500–512.
20. V. Barron and J. Torrent, J. Colloid Interface Sci., 1996, 177, 407–410.
21. C. L. Peacock and D. M. Sherman, Geochim. Cosmochim. Acta, 2004, 68, 2623–2637.
22. W. R. Zhao, J. L. Gu, L. X. Zhang, H. G. Chen and L. J. Shi, J. Am. Chem. Soc., 2005, 127, 8916–8917.
23. I. Kobal, P. Hesleitner and E. Matijevic, Colloids Surf., 1988, 33, 167–174.
24. B. Klhar, S. Gimenez, F. Fabregat-Santiago, T. Hamann and H. Bisquert, J. Am. Chem. Soc., 2012, 134, 4294–4302.
25. J. J. Deng, X. X. Lv, J. Gao, A. W. Pu, M. Li, X. H. Sun and J. Zhong, Energy Environ. Sci., 2013, 6, 1965–1970.
26. S. Ardizzon, G. Spinolo and S. Trasatti, Electrochim. Acta, 1995, 40, 2683–2686.
27. L. Vayssieres, N. Beermann, S.-E. Lindquist and A. Hagfeldt, Chem. Mater., 2001, 13, 233–235.
28. X. M. Zhou, H. C. Yang, C. X. Wang, X. B. Mao, Y. S. Wang, Y. L. Yang and G. Liu, J. Phys. Chem. C, 2010, 114, 17051–17061.
29. K. Q. Wu, Y. D. Xie, J. C. Zhao and H. S. Hidaka, J. Mol. Catal. A: Chem., 1999, 144, 77–84.
30. B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert and T. W. Hamann, J. Am. Chem. Soc., 2012, 134, 16693–16700.
31. T. H. Wang, M. C. Huang, F. W. Liu, Y. K. Hsieh, W. S. Chang, J. C. Lin and C. F. Wang, RSC Adv., 2014, 4, 4463–4471.

---

Footnote

† Electronic supplementary information (ESI) available: The amount of surface cobalt loadings (Table S1) and Fig. S1–S5. See DOI: 10.1039/c5ra08588h

---