On the role of CoO in CoOx/TiO2 for the photocatalytic hydrogen production from water in the presence of glycerol

Abstract The photocatalytic water splitting activity of nanocomposite photocatalysts of TiO2 with CoOx was studied under UV and visible light, and the catalysts were characterized by XRD, XPS, and UV–vis techniques. The presence of CoOx enhances the hydrogen production activity of TiO2 by five times at an optimal loading of .2 wt. %. To investigate the role of CoOx, the photocatalytic activity was also studied under visible light and with different amounts of sacrificial agent. Our results indicate that the increasing activity was not due to increasing absorption of the visible light but most likely due to the role of CoOx nanoparticles as hole scavengers at the interface with TiO2. XPS Co2p analyses of CoO/TiO2 showed a considerable decrease in their signal after prolonged reaction time (44 h) when compared to that of the fresh catalyst. Because part of Co2+ cations is dissolved in solution, in neutral or acidic pH, the possible increase in the reaction rate upon their addition to TiO2 under UV excitation was investigated. No change in the reaction rate was observed upon, on purpose, addition Co2+ cations to TiO2 under UV excitation. Thus, one may rule out the reduction of Co2+ to Co0 with excited electrons within TiO2. In order to further increase the reaction rate, we have synthesized and tested a hybrid system composed of CoO and Pd nanoparticles (Pd wt. % = 0.1, 0.3, 0.5, and 1 wt. %) where 0.3 wt. % Pd – 2 wt. % CoO/TiO2 showed the highest rate.


Introduction
Hydrogen is poised to play an important role in a sustainable energy system because it is storable, transportable, and can be converted into electricity efficiently using fuel cells when necessary. Moreover, it is an essential reactant in chemical industry and making it from renewables will contribute in recycling of carbon dioxide via chemical processes, such as the Fischer-Tropsch, methanol synthesis, and ammonia synthesis reactions. Photocatalytic water splitting using sunlight is considered a promising route to clean and renewable hydrogen production. The current efficiency of the process is still below what is needed for commercialization. The water splitting reaction is an uphill reaction in which the Gibbs free energy increases by 237 kJ mol −1 . 1,2 Water splitting reaction is a multistep process involving (i) light absorption (ii) charge separation and transfer, and (iii) redox reactions on the surface. [1][2][3] The water splitting process can be envisaged as two-half reactions: water oxidation and proton reduction to hydrogen fuel. Water oxidation is more challenging than hydrogen ion reductions because the generation of one molecule of gaseous oxygen requires four holes and occurs on a timescale approximately five orders of magnitude slower than that of H 2 evolution. [3][4][5][6][7] Various semiconductors such as TiO 2 , CdS, ZnO, C 3 N 4 , and WO 3 have been explored for water splitting. [1][2][3]6,7 TiO 2 remains the leading semiconducting material for water splitting with its good conversion efficiency (of UV light: ca. 4-5% of the solar spectrum) and stability. Improving the light absorption and charge carrier separation in TiO 2 remains the biggest challenge in the efficiency of the water splitting process. Loading of cocatalysts such as metal nanoparticles or secondary semiconductors, acting as either electron or hole acceptors for improved charge separation, is needed. Various noble metals such as Pt, Au, Ag, and Pd have been studied in some details on TiO 2 for their water splitting activity. [8][9][10][11][12] These metal particles act primarily as reduction cocatalysts/electron sinks, therefore preventing electron-hole recombination and improving the H 2 production rates. In order to improve the oxidation half reaction, semiconductors such as PdS, RuO x , IrO x , and CoO x have been investigated. 6,7 Among these, CoO x has been reported to be a very efficient oxidation cocatalyst with variety of oxide and On the role of CoO in CoO x /TiO 2 for the photocatalytic hydrogen production from water in the presence of glycerol oxy-nitride semiconductors. Domen and coworkers reported the use of CoO as an oxygen evolution promoter when used with GaZnInON for photocatalytic water splitting. 13 The as-prepared material was unstable under visible light with continuous production of N 2 yet upon loading CoO, the self-decomposition of the oxynitride catalyst was suppressed and O 2 evolution increased 7-fold. Barroso and coworkers studied the photo-electrochemical activity of α-Fe 2 O 3 /CoO x nanocomposite electrodes for water oxidation. 14 Using transient absorption spectroscopy, the authors observed a 3-order of magnitudes increase of lifetime of photo-generated holes upon addition of CoO x . A possible explanation for this increase in lifetimes was the formation of Schotky-type heterojunction leading to reduced recombination of electrons and holes. Li and coworkers also investigated CoO x and CoPi as cocatalysts for water oxidation together with BiVO 4 and reported improvements in oxygen evolution. 15 Upon using a dual cocatalyst system of Pt/CoPi with yttrium-doped BiVO 4 , the authors were able to achieve overall water splitting with production of both H 2 and O 2 . Domen and coworkers also reported a CoO x -modified LaTiO 2 N photocatalyst for water oxidation. Under visible light illumination, the O 2 evolution dramatically increased from 25 to 736 μmolh −1 . 16 Recently, Li and coworkers also reported record quantum efficiency of ~11.3% (when excited with light having a wave length between 400 and 500 nm) for water oxidation using CoO x /Ta 3 N 5 photocatalyst; with AgNO 3 as an electron scavenger. 17 Cobalt oxide has also been used in conjunction with TiO 2 for photocatalytic water splitting. 18 Doping of Co leads to improvement in photocatalytic activity; however, the mechanism of charge carriers separation and the reason for activity enhancement in these composites is not well understood. It was proposed that possible charge transfer between Co 2+ and TiO 2 lead to improvement in H 2 production. 18 Moreover, the activity under UV light decreased with time due to leaching of Co metal ions into the solution. Du and coworkers also reported the study of CoO x -loaded titanium dioxide/cadmium sulfide (TiO 2 /CdS) semiconductor composites. 19 The purpose of CoO x is to prevent the photocorrosion of CdS. Using sodium sulfide (Na 2 S)/sodium sulfite (Na 2 SO 3 ) as hole scavengers under visible light irradiation, the maximum rate of hydrogen evolution achieved was 660 μmol g −1 h −1 , which was about seven times higher than TiO 2 /CdS and CdS photocatalysts under the same conditions. The mechanism proposed was that electrons excited in CdS are transferred to the conduction band of TiO 2 and subsequently to the CoO x cocatalyst where redox reactions take place to produce hydrogen. 19 However, this mechanism is unlikely as the conduction band edge of CoO x is higher than that of TiO 2 . 20 In this study, we evaluate nanocomposite photocatalysts of TiO 2 and CoO x for water splitting and study the role of CoO x . Catalysts were characterized using UV-vis absorption, BET, and XPS. We evaluate the photocatalytic activity for H 2 /O 2 evolution under UV and visible light. We have also addressed the possible Co 2+ reduction to Co metal upon photoexcitation of TiO 2 . Lastly, we demonstrate an active system consisting of dual cocatalysts of Pd and CoO x nanoparticles on TiO 2 that can lead to efficient charge carrier separation and result in very high H 2 evolution rates of ca. 11,000 μmolg −1 h −1 .

Experimental
CoO x -TiO 2 were prepared by wet impregnation. Anatase TiO 2 from Hombikat was used as the support catalyst. Different loadings of Co (0.5, 1, 2 and 4 wt. %) on TiO 2 support were prepared by adding known amount of Co(NO 3 ) 2 ·6H 2 O salt solution to 500 mg of TiO 2 support. Excess water was evaporated under constant stirring with slow heating at 80 °C. The dried photocatalysts was calcined at 400 °C for 5 h. Photocatalysts with dual cocatalysts of Pd and CoO x were prepared by sequential impregnation of Pd on CoO/TiO 2 , starting from (PdCl 2 ).
UV-vis absorbance spectra of the powdered catalysts were collected over the wavelength range of 250-900 nm on a Thermo Fisher Scientific spectrophotometer equipped with praying mantis diffuse reflectance accessory. Absorbance (A) and reflectance (% R) of the samples were measured. The reflectance (% R) data were used to calculate the band gap of the samples using the Tauc plot (Kubelka-Munk function). XRD spectra were recorded using a Bruker D8 Advance X-ray diffractometer. A 2θ interval between 20 and 90° was used with a step size of 0.010° and a step time of 0.2 s/step. Based on the (1 0 1) diffraction line, the crystallite size is of the order of 8 nm. This was also further confirmed by TEM measurements. XPS was conducted using a Thermo scientific ESCALAB 250 Xi; the base pressure of the chamber was typically in the low 10 −9 to high 10 −10 mbar range. Charge neutralization was used for all samples (1 eV). Spectra were calibrated with respect to C1s at 285.0 eV. Quantitative analyses were conducted using the following sensitivity factors: Co2p (3.8), Ti2p (1.8), and O1s (0.66). Ar ion bombardment was performed with an EX06 ion gun at 1 kV beam energy and 10 mA emission current. The sputtered area of 900 × 900 μm 2 was larger than the analyzed area: 600 × 600 μm 2 . Data acquisition and treatment were done using the Avantage software. BET surface areas of catalysts were measured using Quantachrome Autosorb analyzer by N 2 adsorption with surface areas of 133 m 2 g −1 for pure anatase TiO 2 and 131 m 2 g −1 for 2 wt. % CoO-TiO 2 , both calcined at 400 °C.
Photocatalytic reactions were evaluated in a 135-mL-volume Pyrex glass reactor using 4 mg of catalyst. 30 mL of 5 vol. % glycerol aqueous solution or 30 mL of a 0.05 M AgNO 3 aqueous solution were used to evaluate the H 2 and O 2 evolution, respectively. The final slurry was purged with N 2 gas to remove any O 2 and subjected to constant stirring. The reactor was then exposed to the UV light, a 100 Watt ultraviolet lamp (H-144GC-100, Sylvania par 38) with a flux of ~5 mW cm −2 at a distance of 5 cm. Similarly, to evaluate the UV + visible light activity a Xenon lamp (Asahi spectra MAX-303) with a total flux of 26 mW cm −2 (UV ~ 3.3 mW cm −2 and visible (up to 600 nm) ~ 22.7 mW cm −2 ) was used. Product analyses were performed by gas chromatography (GC) equipped with thermal conductivity detector (TCD) connected to Porapak Q packed column (2 m) at 45 • C and N 2 was used as a carrier gas. O 2 analysis was performed by GC equipped with TCD connected to packed molecular sieve (5A) column and He was used as a carrier gas.

Results and discussion
The band gaps and absorption properties of the photocatalysts were studied using diffuse reflectance UV-vis spectroscopy.
The UV-vis spectra of CoO x -TiO 2 are recorded in the range of 250-900 nm as shown in Fig. 1(a). Spectra show typical absorption from anatase TiO 2 with a band edge around 370-380 nm (E g ~ 3.2 eV) due to the charge transfer from the valence band formed by O2p orbitals to the conduction band formed by 3d t2 g orbitals of the Ti 4+ cations. 18,19 Spectra of CoO x -TiO 2 nanocomposite photocatalysts showed absorption in the visible region. One can see, in particular, for the 0.5 and 1 wt. % of Co an absorption peak in the region of 500 nm (~2.5 eV) which can be attributed to Co 2+ → Ti 4+ charge-transfer interaction, as indicated in earlier reports. 18,19 Another absorption peak near 300 400 500 600 700 800 900 Absorbance (a.u.) Wavelength (nm) The Kubelka-Munk theory is generally used for the analysis of diffuse reflectance spectra obtained from weakly absorbing samples. It provides a correlation between reflectance and concentration. The concentration of an absorbing species can be determined using the Kubelka-Munk equation: Hydrogen production (mol/g(cat))   In order to further analyze the chemical composition of CoO x and electronic state of the composites, the CoO-loaded TiO 2 sample was analyzed by XPS and X-ray valence regions. Fig. 2(c) shows the Co 2p spectra from 1 wt. % CoO-TiO 2 samples calcined at 400 °C for 5 h. The XPS Co 2p of Co, CoO, and Co 3 O 4 has been studied in some details by many workers. 25 Co2p of Co 2+ has its characteristic satellites, reduction of Co 2+ leads to Co 0 which results in a shift in the binding energy by about 2 eV. The binding energy of Co 3+ is very close to that of Co 2+ but Co 3+ satellites are much more attenuated, and therefore, the presence of strong satellites can gauge the extent of Co 2+ contribution. 26 In Fig. 2(c), XPS Co2p before and after Ar ion sputtering is presented. The binding energies for Co 2p 3/2 and Co 2p 1/2 are at 781.4 eV and 797.1 eV. These positions and the spin orbit splitting of 15.5 eV and satellites presence at about 7 eV above the main peaks (ca. 788 and 804 eV) are consistent with those reported for Co +2 of CoO. 18,23,26 Argon ions sputtering results in the preferential removal of oxygen anions and consequently the reduction of metal cations to lower oxidation states. 27 We use to ascertain that our as-prepared sample is mainly composed of Co 2+ and not Co 0 . This can be seen in Fig. 2(c) and (d). In Fig. 2(c), a shoulder at the lower binding energy side is seen at about 778 eV that is attributed to Co 0 . The appearance of Co 0 is associated with the decrease of the signal of Co 2+ . In Fig. 2(d), the valence band region is presented for the fresh and Ar ions sputtered surfaces. The appearance of the lines at about 1 eV below the Fermi level is indicative of 3d electrons due to both Ti cations in reduced states and metallic Co. [28][29][30] The inset in the Fig. presents the Ti3p and O2s for the same samples. The broad structure at the band gap of semiconductors can be determined by plotting (F(R) × E) 1/r against the radiation energy in (eV), using r = 2 for indirect allowed transitions of charge carriers (indirect band gap material) or r = ½ for direct allowed transition (direct band gap material). The resulting plot has a distinct linear regime, which denotes the onset of absorption. Thus, extrapolating this linear region to the abscissa yields the energy of the optical band gap of the material. Tauc plots of our catalysts are shown in Fig. 1(b) (r = 2) which shows a slight decrease in optical bad gap of TiO 2 with increasing Co loading.
The effect of Co loading on crystal structure of TiO 2 support was studied using XRD. Fig. 2(a) shows the X-ray diffraction patterns of TiO 2 with different loadings of Co (in wt. %). The XRD patterns with the characteristic planes of anatase phase at 2θ = 25.5° (1 0 1), 37.7° (0 0 4), and 48.2° (2 0 0) are seen. The XRD pattern does not show any cobalt phase (up to 4 wt. % loading), indicating that cobalt ions are uniformly dispersed on the TiO 2 support. This was also indicated by others where at low loadings the CoO x diffraction peaks could not be det ected. 18,19,23,24 The XRD peaks positions of anatase TiO 2 also do not show any change upon Co loading, confirming there is no change in structure/crystal phase of TiO 2 or doping of Co ions into TiO 2 . A broadening of the TiO 2 diffraction peaks is, however, observed with the addition of Co, larger FWHM. This broadening could be due to either smaller TiO 2 crystallites and/or lattice strain on TiO 2 due to the presence of CoO nanoparticles. Because BET surface areas measurements did not show change upon CoO (within experimental errors) loading this change might be due to lattice strain. A table of the diffraction lines of TiO 2 (anatase), CoO, and Co 3 O 4 is added, Fig. 2(b). 198 pumped into CB of TiO 2 leading to enhancement of activity. 19 Our results indicate that any charge carriers being generated in CoO x from visible light do not participate in the photocatalytic water splitting process. Other reports have indicated that CoO in nanoparticles has activity for hydrogen production yet the catalyst deactivates very fast within 1 h. It is likely that in that reported study, a non-catalytic surface reaction occurred; in other words, this was a stoichiometric and not a catalytic reaction. 22 CoO is a p-type semiconductor with a relatively low work function (~4.4 eV) making it attractive as an oxidation cocatalyst. 44 The experimentally measured valence band edge formed of O 2p orbitals is at 0.7 V vs. NHE while the conduction band edge is at −1.7 V vs. NHE. 44 It is possible that the enhanced photocatalytic activity of our composite catalysts is due to the formation of a Schottky-type heterojunction, leading to efficient charge carrier separation. The high valence band edge in CoO x is ideal for trapping photogenerated holes in TiO 2 . To confirm this hypothesis, we tested the photocatalysts by changing the concentration of the "hole scavenging" sacrificial agent. The photocatalytic activity under the same conditions was conducted with lowering the glycerol concentration from 5 to 1 vol. %. As shown in Fig. 4(d), in pure anatase TiO 2 , the H 2 evolution rate drops by ~45% when glycerol concentration is reduced to 1 vol. %. In contrast, samples with 2 wt. % of CoO x show better activity, a drop of ~10%. This result may indicate that CoO x nanoparticles function similar to the sacrificial agent i.e., as an oxidation cocatalyst/hole trapping agent.
In order to test the possibility of Co 2+ dissolution into the liquid phase, we have conducted long-term photoreaction experiment (44 h) over 4 wt. % CoO/TiO 2 under UV excitation and analyzed the Co content by XPS. We have used a high loading of 4 wt. % CoO in order to decrease the errors associated with computing the total peak area of the Co2p signal. Fig. 5 presents the XPS Co2p region of the fresh and used catalysts. The signal is typical of the Co2p of Co 2+ cations as indicated in Fig. 2. Qualitatively, the XPS Co2p after reaction is identical to that before. However, the Co2p/Ti2p ratio decreased from 0.16 to 0.04 during the reaction time indicating considerable loss of Co cations.
While the reduction to Co metal (similar to that of Cu 2+ to Cu 0 ) 33 by excited electrons is possible, one may refute it based on the redox potential. The redox potential of Cu is within the band gap of TiO 2 (+0.34 V), and therefore, its reduction by conduction band electrons of TiO 2 occurs. The redox potential of Co is −0.28 V, therefore above TiO 2 conduction band 45 . Still, because the difference between the redox potential of Co 2+ / Co and that of TiO 2 conduction band is very small (within the limits of measurements (Fig. 6), alterations due to changes in dipole moments 46 may occur. We have conducted an experiment in which the photoreaction on TiO 2 alone was monitored and where Co 2+ (from Co(NO 3 ) 2 ) cations were introduced during the run. As shown in Fig. 7, no deviation in hydrogen production is seen upon the addition of Co 2+ cations, indicating that their introduction did not alter the activity of TiO 2 . This result may exclude the possibility of Co 2+ reduction by the CB band electrons of TiO 2 . It is also not possible under the reaction conditions to reduce CoO to Co metal by hydrogen produced low binding energy side of the Ti3p is due to the presence of Ti cations in lower oxidation state than +4 due to preferential removal of oxygen anions upon sputtering. Quantitative analyses of the Co2p, Ti2p, O1s indicated that Co is present in about 0.1 at. % on the TiO 2 surface.
The H 2 and O 2 production activities of CoO x -TiO 2 photocatalysts under UV excitation are presented in Fig. 3. The photocatalytic activity was evaluated over 24 h and was stable and reproducible. Pure anatase TiO 2 calcined at 400 °C showed H 2 production rates of ~10 μmolg −1 min −1 . The loading of CoO x resulted in improvement in the H 2 evolution. The highest H 2 production rates of ~47 μmolg −1 min −1 was obtained when the Co metal concentration was 2 wt. % relative to TiO 2 . The H 2 production rates as a function of Co loading is plotted in Fig.  3(b). One can notice that increasing the Co loading above 2 wt. % decreased the photocatalytic activity. The highest photocatalytic activity of 2 wt. % cobalt loaded samples could be due to the optimum dispersion of CoO x particles over TiO 2 photocatalyst. This trend is similar to other systems such as CuO/ TiO 2 , Cu 2 O/TiO 2 , and NiO/ TiO 2 as reported earlier. [31][32][33] Actually, this trend is also seen for noble metals where a maximum efficiency occurs for a narrow range of concentrations depending on the nature of the metal and of the semiconductor. 9,[34][35][36][37] In order to further probe into this trend, we have monitored the O 2 evolution activity of the same catalysts under UV excitation and in the same reactor but with 0.05 M AgNO 3 solutions to scavenge electrons. While this method has its drawback (such as the possible deposition of the metallic Ag that is reduced on the surface of the catalyst), 38,39 it does give a good indication on the potential of the semiconductor for the hole transfer and therefore for comparative reasons the method has its merit. [40][41][42] As shown in Fig. 3(c), the O 2 evolution monitored in a separate experiment with AgNO 3 as sacrificial agent was linear as a function of time and was observed to be in a stoichiometric ratio of 1:2 to the H 2 production seen earlier with glycerol as the sacrificial agent. This may tell that the potential of the catalyst is indeed non negligible and the rates of electron and hole transfer are comparable. The O 2 evolution also showed a similar trend to the H 2 production as shown in Fig. 3(d) with the 2 wt. % Co loading had the highest activity (~21 μmolg −1 min −1 ).
To further investigate the contribution of CoO x in the enhancement of photocatalytic activity, the reaction was also tested under UV + visible light irradiation under a total flux of ~26 mW cm −2 (UV ~ 3.3 mW cm −2 , visible ~ 22.7 mW cm −2 ). This is indeed important, as numerous reports have indicated that the addition of CoO may enhance the reaction due to further absorption of light. 18,19,43 As shown in Fig. 4(a), similar to the trend under UV lamp, the loading of cobalt on TiO 2 resulted in improvement in the H 2 evolution. The highest H 2 production rates were also achieved when the Co concentration was 2 wt. % as seen in Fig. 4(b). Fig. 4

Conclusions
Nanocomposite photocatalysts by impregnating anatase TiO 2 with different amounts of Co salt solutions were prepared, characterized, and tested. The presence of CoO enhances the activity of TiO 2 with optimal loading determined to be ca. 2 wt. %, and the rate of hydrogen evolution was about five times higher than that of TiO 2 alone. The increase in activity was not due to Co 2+ reduction by TiO 2 CB electrons. The increase in activity was also not due to increasing absorption of the visible light. It is most likely due to the role of CoO nanoparticles as hole scavengers at the interface with TiO 2 . The addition of Pd (as hydrogen ion reduction sites) further improved the reaction rate ca. 4 times compared to that of the composite system, to 180 μmolg −1 min −1 . No catalytic deactivation was seen for prolonged reaction time (up to ca. 24 h). A schematic description of the events is given in which electron transfer occur from the VB of CoO to that of TiO 2 (upon photoexcitation), this results in increasing water oxidation upon electron transfer from hydroxyl oxygen anions to the VB of CoO.
since thermodynamics indicate that at room temperature the partial pressure of O 2 needs to be equal to 10 −30 torr. 47 Fig . 8 presents the events diagram, to complement the electronic diagram of Fig. 6. Upon contact between CoO and TiO 2 and photoexcitation with UV light, electrons are transferred from the VB to the CB of TiO 2 (leaving holes in the VB) -step 1. Electrons can then be transferred from the VB of CoO to the empty states of the VB of TiO 2 -step 2. This results in increasing the likelihood of electrons in the CB of TiO 2 to reduce H + to ½ H 2 -step 3. At the same time, O anions of OH (a) (surface hydroxyls) donate electrons to the VB of CoO -step 4. Based on the work's results (in particular, the absence of enhancement of the rate when visible light is added), the rate of reaction can be explained without invoking visible light excitation of CoO, in the presence of UV light equivalent to that provided from the sun. It is worth mentioning that because no change in the reaction rate is seen between excitation under UV and excitation under UV + vis. light, in addition to the absence of hydrogen production under vis. light alone, electron excitation from the VB of CoO to its CB does not contribute into the reaction. Within this context, electron transfer from CoO CB to TiO 2 CB, as often invoked in many work, may have no physical meaning. This is simply because under UV excitation, the TiO 2 CB is populated by excited electrons and therefore less poised to receive excited electrons from the CoO CB.
The role of dual cocatalysts i.e., oxidation and reduction cocatalysts has been investigated by many workers including those reported in Refs. 6,48,49 . To investigate the role of another cocatalyst, we loaded Pd metal as the reduction cocatalyst on top of 2 wt. % Co-TiO 2 which showed the best activity. In order to assess the activity, we also changed the concentration of Pd on top of the catalyst. The H 2 production activity of 2 wt. % CoO x -TiO 2 photocatalysts impregnated with Pd metal is shown in Fig. 9(a). Further improvement of the H 2 evolution reaction was observed. The highest H 2 production rates were achieved when the Pd concentration was 0.3 wt. % as seen in Fig. 9(b), with H 2 production rates of ~180 μmolg −1 min −1 . Thus it seems that a system where a dual semiconductor-based cocatalyst i.e., CoO as an oxidation cocatalyst and Pd as reduction cocatalyst can function and is stable. The best performing Pd/Co ratio is = ca. 0.1.
It is worth extracting a few points from this study. The optimum amount of CoO on top of TiO 2 is found to be 2 wt. %. Because we do not see a change in the slope of hydrogen production and because the reduction potential is above the CB of TiO 2 , the chemical state of the cobalt oxide(s) is plausibly maintained as CoO. The BET surface area of TiO 2 is = 133 m 2 /g Catal. We have calculated the expected surface area of a 2 wt. % of CoO, as an example. It is equal to 3 × 10 −4 mole/g Catal . = 16 m 2 assuming monolayer dispersion (as an upper limit). This would cover a small fraction of the surface; agglomeration would decrease it further. Therefore, the finding of an optimal reaction rate so narrowly dependent on the fraction of CoO may not be linked to a geometric effect (blocking of available sites) but most likely to an electronic effect where electrons transfer are optimized at the interface TiO 2 and CoO.