resonant photoemission spectroscopy investigation of fe2o3 tio2 heterojunctions for solar water splitting

Available online at www.sciencedirect.com ScienceDirect Physics Procedia 85 (2016) – 11 EMRS Symposium: In situ studies of functional nano materials at large scale facilities: From model systems to applications,EMRS Spring Meeting Resonant PhotoEmission Spectroscopy investigation of Fe2O3 – TiO2 heterojunctions for solar water splitting Maxime Rioulta, Dana Stanescua, Patrick Le Fèvreb, Antoine Barbiera, Hélène Magnana,* b a SPEC, CEA, CNRS, Université Paris-Saclay, CEA-Saclay, 91191 Gif-sur-Yvette Cedex, France Synchrotron SOLEIL, L’Orme des Merisiers, Saint-Aubin BP 48, 91192 Gif-sur-Yvette Cedex, France Abstract The implementation of semiconductor heterojunctions photo-anodes appears as a very promising way to improve the performances of devices for solar water splitting (sun light assisted hydrogen production from water) However, assembling different materials together results in the existence of interfaces which usually not have the same electronic structure than the simply stacked individual layers The electronic structure of the valence band being a key parameter for water splitting, it is necessary to investigate it for each layer and for the interfaces A very powerful technique to tackle these issues is Resonant PhotoEmissionSpectroscopy (RPES) In this work, we present RPES results of the valence band performed on epitaxial TiO2/Ti-doped Fe2O3 heterojunctions The provided insights concerning the interface and the electronic structure are correlated to the water splitting performances © by Elsevier This is an open access article under the CC BY-NC-ND license © 2016 2016 Published The Authors PublishedB.V by Elsevier B.V (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the organizing committee of the EMRS Spring Meeting 2016 Peer-review under responsibility of the organizing committee of the EMRS Spring Meeting 2016 Keywords: heterojunctions; solar water splitting; hematite; Fe2O3; TiO2; resonant photoemission; absorption spectroscopy Introduction Producing hydrogen through solar water splitting is very appealing as it is an energy carrier of choice which does not lead to any greenhouse gas production when produced with solar light During the process, electron-hole pairs photogenerated in semiconductors in contact with an aqueous solution participate to the water oxidation and reduction reactions (oxygen and hydrogen production respectively) [Fujishima and Honda (1972)] Since the pioneering discovery of water photo-assisted electrolysis using TiO2 [Fujishima and Honda (1972)], several materials were investigated as photoanodes where water oxidation occurs (2H2O + 4h+ Ỉ O2 + 4H+ in acidic medium and 4OH+ 4h+Ỉ O2 + 2H2O in basic medium) [van de Krol et al (2008)] Unfortunately no simple semiconductor fulfils all requirements and new routes need to be explored Combining materials in heterojunctions, to take advantage of their individual properties, appears as a very promising way to improve the photoanode performances [Kronawitter et al (2011)] In this paper we focus on the combination of Ti-doped hematite (Ti:Į-Fe2O3) and TiO2 to build photoanode heterojunctions As a matter of fact, Ti:Į-Fe2O3 and TiO2 are the most promising candidates as photoanodes for solar water splitting devices The main advantage of hematite is its low band gap of ca 2.1 eV Ti-doping of hematite greatly enhances the performances (photocurrent gain > 100) through an increase of the carriers’ diffusion length while conserving the corundum structure [Magnan * Corresponding author Tel.: +33169089404; fax: +33169088446 E-mail address: helene.magnan@cea.fr 1875-3892 © 2016 Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the organizing committee of the EMRS Spring Meeting 2016 doi:10.1016/j.phpro.2016.11.074 Maxime Rioult et al / Physics Procedia 85 (2016) – 11 et al (2012), Rioult et al (2014)] However its conduction band edge position below the water reduction potential requires a large external bias to drive water splitting Besides, TiO2 features very interesting charge transport properties and electronic structure (conduction and valence band positions above water reduction level and below oxidation energy level respectively), despite a large band gap of ca 3.2 eV [van de Krol et al (2008)] Additionally, both are abundant on earth and very stable in aqueous environments Several studies reported the use of Fe2O3 - TiO2 heterostructures [Liou et al (1982), Kuang et al (2009), Zhao et al (2011), Yang et al (2014)] However in these works, the materials are polycrystalline, often in core-shell structure, e.g Fe2O3 nanoparticles deposited on TiO2 nanotubes, making the investigation of the electronic structure extremely challenging Here we have studied systems constituted of nanometric single crystalline epitaxial films with well-defined interfaces The use of resonant photoemission spectroscopy (RPES) allows characterizing the electronic structures of the heterojunction and of the interface between the two materials, which are crucial parameters defining the carriers transport properties In addition photocurrent measurements allow correlating electronic structure and water splitting performances Experimental methods The heterojunctions were deposited on single-crystalline Pt (111) substrates using Atomic Oxygen plasma-assisted Molecular Beam Epitaxy (AO-MBE), a technique that makes possible the deposition of single crystalline layers of controlled composition and thickness [Magnan et al 2012] High purity metals (99.99% grade) were evaporated from dedicated Knudsen cells in the presence of an atomic oxygen plasma (350 W power) in order to obtain well defined oxides under good vacuum conditions (i.e 10-7 mbar working pressure, 10-10 mbar base pressure) During the deposition, the samples were rotated continuously around their normal to ensure a homogeneous deposit, and radiatively heated at a temperature of ca 900 K The oxide deposition rate was about 0.15 nm/min In situ Reflexion High Energy Electron Diffraction (RHEED) patterns were observed and acquired during film growth to monitor the crystal quality and structure of the samples The set of realized samples is detailed on Table The terminology of the samples name was chosen as follows: F stands for Ti:Į-Fe2O3, T stands for TiO2 and the number before F or T gives the thickness of the corresponding layer in nm We chose to keep a constant total thickness of 20 nm This thickness corresponds to a good compromise: it is thick enough to allow acceptable photocurrents in photo-electrochemical characterization and it is thin enough to allow synchrotron radiation measurements without charge build-up issues and using soft X-rays For heterojunctions, we varied the relative thickness of TiO2 and Ti:Į-Fe2O3 We also studied single layers as references Table Set of samples studied For the terminology of the samples name, F stands for Ti:Į-Fe2O3, T stands for TiO2 and the number before F or T gives the thickness of the corresponding layer in nm Sample name Complete architecture 20F 20 nm Ti:Į-Fe2O3 / Pt (111) 5T/15F nm TiO2 / 15 nm Ti:Į-Fe2O3 / Pt (111) 10T/10F 10 nm TiO2 / 10 nm Ti:Į-Fe2O3 / Pt (111) 10F/10T 10 nm Ti:Į-Fe2O3 / 10 nm TiO2/ Pt (111) 20T 20 nm TiO2 In situ X-ray Photoelectron Spectroscopy (XPS) spectra were systematically recorded just after deposition in order to determine the stoichiometry and the electronic structure of the films More precisely, we recorded Fe2p, Ti2p, O1s core levels and the valence band (VB) region using the Al KĮ radiation (1486.7 eV) The electronic structure of the films was investigated using RPES and the experiments were carried out on the CASSIOPEE beamline at synchrotron SOLEIL (Saint-Aubin, FRANCE) RPES conditions are fulfilled when XPS is realized using excitation photons of an energy close to the absorption threshold of a core level previously determined by X-ray Absorption Spectroscopy (XAS), which provides (in a simple picture) chemically selective photoemission It is a useful technique for the investigation of VB features in solids [Magnan et al (2010)] We used this technique to acquire resonant photoemission spectra of the VB region using photons around the L3 absorption edge for Fe and/or Ti The mechanism of resonant photoemission of valence band in transition metal oxides is well known In an atomic picture (see Figure 1), it is caused by the final-state interference between: • direct photoemission from the 3d levels: 3dnĺ 3dn-1 +e • auto-ionization process: 2p63dnė 2p53dn+1ė 2p63dn-1 +e Maxime Rioult et al / Physics Procedia 85 (2016) – 11 Fig Valence band photoemission paths for RPES at the L3 edge: (a) photoemission from the valence band, (b) Auger relaxation (auto-ionization process) Therefore, by choosing the appropriate photons energy, one can expect an increase of the photoemission signal coming from the atoms for which the excitation energy matches the L3 absorption edge, thanks to the coherent interferences phenomena RPES can be seen as a chemically selective photoemission process enhanced by the resonance phenomena However, it should be noted that this technique is limited to nanometric thin film geometries, since the probed depth will be given by the electrons mean free path (< 10 nm at a kinetic energy of 800 eV) The use of a high brilliance tunable photon energy source such as synchrotron radiation is required to perform RPES The photocurrent of our films was measured by acquiring I-V curves in a three electrodes cell previously described [Rioult et al (2016)] with an incident light flux of ca 100 mW/cm² provided by a Xe arc lamp The potential reference is the Ag/AgCl electrode (VAg/AgCl = +0.197 V vs SHE) The photocurrent is defined as the difference between the current recorded under light and the one in the dark Results and discussion 3.1 Growth The RHEED patterns, recorded during growth, for the 20F film (Figure 2.a) correspond to hematite as shown in previous works [Rioult et al (2014)] In the case of the 20T sample (Figure 2.b), large dots on the RHEED patterns are visible, indicating a rough surface and a Volmer-Weber growth mechanism (islands) Lastly for the 10T/10F sample (Figure 2.c), the RHEED patterns are similar to the 20T sample ones For these both 20T and 10T/10F samples, the RHEED patterns correspond to the rutile crystallographic structure of TiO2 on Pt(111) with the epitaxial relationship [010] rutile // [1-10] Pt and the presence of three rotational variants in agreement with previous reports [Artiglia et al (2012)] Fig (Left panel) RHEED patterns over the two lowest index surface diffraction directions and (right panel) the corresponding surface reciprocal lattices The diffraction directions are made explicit, as well as the variants for TiO2 rutile (100) (3 rectangles with colored contours) Samples: (a) 20F, (b) 20T and (c) 10T/10F Maxime Rioult et al / Physics Procedia 85 (2016) – 11 In-situ XPS spectra after the films growth evidenced Fe3+ and Ti4+ ionic states in all samples Fe2+ or Ti3+ species could not be detected within the experimental detection limits The Ti-doping level of hematite was determined to be at.% 3.2 RPES at the Fe L3 edge Figure 3shows the X-ray absorption spectra (XAS) and the corresponding resonant photoemission maps of the valence band (VB) obtained with photons energies around the Fe L3 absorption edge for the 20F sample and the heterojunctions Fig RPES at the Fe L3 edge (Left panel) Fe L3 XAS (Right panel) Corresponding valence band (VB) photoemission maps, the color scale is given at the figure bottom Samples: (a)+(d) 20F, (b)+(e) 5T/15F, (c)+(f) 10T/10F The qualitative analysis of the Fe L3 XAS has demonstrated being able to distinguish the electronic configurations of Fe in the different possible oxide environments [Rioult et al (2015), Jiménez-Villacorta et al (2011)] The L3A (resp L3B) contribution, highlighted on Figure 3.a., is strengthened in the presence of Fe2+ (resp Fe3+) ions Experimentally we found out that the L3A contribution increases when the hematite film is covered by a TiO2 film (Figure 3.b and 3.c) In the case of the 10T/10F sample (Figure 3.c), the L3A contribution is even higher than the L3B contribution, accounting for a high number of detected Fe2+ species in the probed zone However we did not detect by XPS the presence of Fe2+ species It is because in our setup the probed depth is lower in XAS than in XPS for three reasons: (i) the beam incidence is more grazing in XAS (10° from the surface in XAS, 45° from the surface in XPS), (ii) due to the L3 Fe resonance the mean free path of the incident beam in XAS is reduced and (iii) the detection of the signal in XAS is in grazing emergence (34° from normal emergence in XAS, normal emergence in XPS).We evaluate the “apparent mean free paths” for XAS at about 1.4 nm and at about nm for XPS Moreover the eventual Fe2+ detection in XPS is done by investigating the shape of the Fe2p satellite, which is a broad component with rather low intensity, hence with a lower sensitivity than an absorption edge Therefore we can conclude that the Fe2+ species are situated near the interface betweenTiO2 and Ti:Į-Fe2O3 Concerning the VB resonant photoemission maps for heterojunctions (Figure 3.e and 3.f), one can notice an increase of intensity for binding energies below eV (within the band gap, between the two vertical black dotted lines) This zone, which accounts for electronic states within the band gap, widens and intensifies when the TiO2 upper layer thickness increases and is visible only at photons energies matching the Fe L3A XAS contribution Since these states resonate at the photons energy corresponding to Fe2+ like in Fe3O4 [Magnan et al (2010)], we can conclude that they correspond to Fe2+ species which are favored by the presence of a TiO2 upper layer This result confirms the XAS results and highlights a major advantage of RPES, which is its sensitivity to buried species Maxime Rioult et al / Physics Procedia 85 (2016) – 11 3.3 RPES at the Ti L3 edge Figure shows XAS and the corresponding resonant photoemission maps of the VB obtained with photons energies close to the Ti L3 absorption edge for the 20T sample and the heterojunctions No particular resonance was observed for the doped only 20F sample Fig RPES at the Ti L3 edge (Left panel) Ti L3 XAS (Right panel) Corresponding valence band (VB) photoemission maps, the color scale is given at the figure bottom Samples: (a)+(d) 20T, (b)+(e) 5T/15F, (c)+(f) 10T/10F In the case of TiO2, the analysis of the Ti L3 absorption spectra can give insights in the crystallographic structure (anatase or rutile) [Kruger (2010)] More precisely, the Ti L3 XAS in TiO2 shows two major features: a sharp peak around 458 eV and a broader component around 460 eV The latter is also split into two contributions: one around 459.5 eV and the other around 460.5 eV (highlighted by asterisks on Figure 4.c) This splitting is attributed to a non-cubic ligand field due to the distortion of the TiO6 octahedra but also to the long range band structure and the relative intensities of the two components is different in anatase and rutile (in rutile the 459.5 eV contribution is less important than the 460.5 eV one) Hence from Figure 4.a-c we can conclude that the TiO2 structure depends on the substrate: when it is deposited on hematite it seems to be mainly in the rutile crystallographic structure although the Ti L3edge of the 20T sample resembles more anatase Since for all the samples RHEED results demonstrated a long range rutile structure we can conclude that these differences in the XAS data are more likely due to different distortions of the TiO6 octahedra VB maps on Figure present zones of increased intensity within the band gap, like in the case of the Fe L3 edge Those states are likely due to Ti3+ species present in the material, as in the case of oxygen-deficient TiO2 studied by RPES [Le Fèvre et al (2004)] The fact that for our samples Ti3+ was not detected in conventional XPS but detected by RPES is due to the higher sensitivity of RPES We can notice that Ti3+ is detected in all samples and that its quantity increases when the thickness of the TiO2 layer increases Hence the Ti3+ species are probably intrinsically present in our TiO2 films deposited by AO-MBE on Pt (111) or Ti:Į-Fe2O3 (0001) (i.e the appearance of Ti3+ occurs during the growth) Lastly, in the case of the 5T/15F sample (Figure 4.e), a shift of the VB edge toward lower binding energies is observed (see the increase of the signal between and 2.2 eV of binding energy, between the dashed and dash-dotted black lines) This is due to the measure of a photoemission signal corresponding to the VB of the Ti:Į-Fe2O3 underlayer, which we are able to record even with photons energies far from the Fe L3 edge resonance since the TiO2 upper layer for this sample is very thin 3.4 Photocurrent measurements Figure presents the photocurrent density vs voltage curves for single layers and heterojunctions In the case of the 20F sample, a good photocurrent is achieved, but at the cost of a high onset potential (§ 0.2V vs Ag/AgCl) On the opposite, the 20T sample shows a maximum photocurrent almost three times lower, although it is achieved with a low onset potential (§ -0.4V vs Ag/AgCl) For heterojunctions with a TiO2 upper layer (samples 5T/15F and 10T/10F), the photocurrent at high voltage Maxime Rioult et al / Physics Procedia 85 (2016) – 11 decreases when the TiO2 thickness increases, with respect to the 20F sample but the onset potential is slightly reduced by ca 0.1V Lastly, the 10F/10T sample shows very poor photoelectrochemical properties with an onset potential ca 0.7V vs Ag/AgCl and a loss of photocurrent of almost one order of magnitude at 0.8V vs Ag/AgCl with respect to the 20F sample Fig Photocurrent density vs voltage curves for the 20F (black solid line), 5T/15F (orange dotted line), 10T/10F (green dashed line), 10F/10T (blue wide dashed line) and 20T (cyan solid line) samples If photocurrent discrepancies between single layers and heterojunctions could be explained solely by absorption discrepancies due to differences in the effective band gap, a linear combination of the photocurrent of the single layers should simulate the photocurrent of heterojunctions Since the penetration depth of UV-visible photons is in the hundreds of nm scale, our linear combination approach is reasonable considering the thicknesses of our samples (tens of nm) We calculated the photocurrent corresponding to the linear combination of the 20T and 20F samples according to the (TiO2):(Ti:Į-Fe2O3) content ratio in the different heterojunctions (5:15 for the 5T/15F sample and 10:10 for the 10T/10F and 10F/10T samples) [Rioult (2015)] As a matter of fact the photocurrent of the linear combination of single layers is always superior to the photocurrent of the actual heterojunctions At high applied potentials, the differences are less pronounced (except for the 10F/10T sample), meaning that for this range of potentials the differences between single layers and heterojunctions are dominated by absorption discrepancies However for applied potentials lower than 0.2V vs Ag/AgCl, where only TiO2 features a substantial photocurrent, the photocurrent in heterojunctions is almost zero This means that other phenomena, like less efficient surface kinetics or detrimental charge transport due to heterostructuring, are responsible for photocurrent losses Our results are very consistent with the observations reported by Steier et al [Steier et al (2014)], which are that the photocurrent onset potential is reduced through surface charge recombination diminution (the so-called “surface state passivation”) and the increase in the maximum photocurrent value proceeds through an augmentation of the charge carrier concentration The photoelectrochemical characterizations for our samples show that a TiO2 upper layer induces a reduction of surface charge recombination rate (i.e the reduction of the photocurrent onset potential for heterojunctions with respect to a single layer of Ti:Į-Fe2O3) Moreover, Ti:Į-Fe2O3 allows a high photocurrent, thanks to a higher charge carrier concentrations upon photons absorption as compared with TiO2, resulting in a higher maximum photocurrent for heterojunctions with respect to a single layer of TiO2 This highlights a possible use of TiO2 as a surface treatment in order to reduce the onset potential [Yang et al (2014)] A configuration where only few layers (and not few nm) of TiO2 would be deposited on the top of Ti:Į-Fe2O3 would feature the advantageous absorption properties of hematite and the well-suited band structure and surface kinetics properties of TiO2 3.5 Discussion Since the presence of Fe2+ features induced by the deposition of a TiO2 film on Ti-doped hematite could be due to interdiffusion between the two layers at the interface we investigated this possibility in more details To determine the spectroscopic signature of Fe inclusions in TiO2, we studied a 0.5 at.% Fe-doped TiO2 sample (elaborated in the same conditions than the other TiO2 films) The RHEED patterns during the growth of this material and the Ti2p core level XPS spectra were identical to the ones of pure TiO2 (20T sample) The Fe2p core level XPS spectra showed classical features for iron oxide however it was not possible to conclude on the Fe valence state because the satellite position could not be determined with such a small doping level (too poor statistics) The RPES data for this sample are shown on figure The Fe L3 XAS (figure6.a) showed a single broad feature around 709 eV corresponding to the L3A contribution This shape presented a lot a resemblance with the Fe L3 XAS in the FeO environment [Jiménez-Villacorta et al (2011)] where the state of Fe is only Fe2+ We can then conclude that the iron state in our 0.5 at.% Fe-doped TiO2 is mainly Fe2+ Moreover, the corresponding VB map (figure6.c) exhibited a substantial photoemission signal for binding energies within the band gap for photons energies matching the L3A contribution, i.e the same behavior than those measured for heterojunctions at the Fe edge Therefore a possible explanation of the presence of 10 Maxime Rioult et al / Physics Procedia 85 (2016) – 11 Fe2+ is the interdiffusion of Fe into TiO2 upon TiO2 deposition on Ti-doped hematite, which is, besides, expected according to the Ellingham diagrams Fig RPES at Fe and Ti L3 edges for a 20 nm 0.5 at.% Fe-doped TiO2 / Pt (111) sample (a)+(b) Fe and TiL3 XAS (c)+(d) Corresponding valence band (VB) photoemission maps, the color scale is given at the figure bottom The Ti L3 XAS (Figure 6.b) has the same shape than the 20T sample one Since RHEED patterns were also identical, the structure can be considered to be the same (rutile (100) with distorted TiO6 octahedra) The corresponding valence band map (Figure 6.d) shows electronic states within the band gap, but of a smaller amount as compared to the 20T sample, accounting for a smaller Ti3+ concentration in this sample as compared to undoped TiO2 We previously considered that the reduction of Ti4+ to Ti3+ was linked to charge balance restoring due to defects like oxygen vacancies In the case of Fe:TiO2, Fe2+ species particpate in the charge restoration and may even be more efficient than Ti3+ species originating from Ti4+ reduction This can explain the smaller Ti3+ concentration in Fe:TiO2 with respect to undoped TiO2 RPES results demonstrated the creation of a diffuse interface leading to Fe2+ species that induce states within the band gap when combining Ti-doped hematite with TiO2 Those states are likely to act as recombination centers during charge separation or charge transfer between the two materials An illustration of this interpretation is given on Figure7 Also the concentration of these recombination centers increases when the TiO2 upper layer thickness increases Figure also illustrates why the 10T/10F heterojunction shows extremely poor photocurrent; indeed the VB offset for this photoanode is highly detrimental for electron transport from the Ti:Į-Fe2O3 surface layer to the TiO2 layer Fig Illustration of the detrimental effect of a Fe:TiO2-like interface between Ti:Į-Fe2O3 and TiO2 in the form of a band diagram Double-lined, green dotted and red dotted arrows convey respectively the photons absorption, the charge transport and the losses due to charge recombination induced by Fe2+ states in the band gap at the interface The band positions were taken from [Xu and Schoonen, 2000] and the band offsets between TiO2 and Ti:Į-Fe2O3 were determined using in situXPS after thin films growth Conclusions The study of nanometric Ti:Į-Fe2O3 - TiO2 heterojunctions highlighted concomitant effects due to the heterostructuring: (i) the creation of a detrimental interface between the two materials, inducing (bulk) charge recombination which limits the overall photocurrent; and (ii) a beneficial effect in the case of a TiO2 upper layer for which the onset potential is reduced Maxime Rioult et al / Physics Procedia 85 (2016) – 11 For our multilayer systems, the combination of two materials together with the idea of benefiting from the assets of each material turns out to be dominated by surface and interface effects These findings were evidenced thanks to the use of RPES as a chemically selective technique which is sensitive to the surface/interface electronic structure One could extend the use of RPES to the investigation of other systems dedicated to solar water splitting, for instance the use of surface treatments or more complex heterojunctions Such approaches, using model systems and powerful techniques like RPES, are very likely to bring new insights in the field of solar water splitting Acknowledgements This work is supported by a public grant overseen by the French National Research Agency (ANR) as part of the “Investissements d’Avenir” program (LabexNanoSaclay, reference: ANR-10-LABX-0035) and by the Region Ỵle de France in the framework of DIM-OXYMORE under the grant RUMATO References Artiglia, L., Zana, A., Rizzi, G A., Agnoli, S., Bondino, F., Magnano, E., Cavaliere, E., Gavioli, L., Granozzi, G., 2012 Water Adsorption on Different TiO2 Polymorphs Grown as Ultrathin Films on Pt(111) Journal of Physical Chemistry C 116, 12532-12540 Fujishima, A., Honda, K., 1972 Electrochemical Photolysis of Water at a Semiconductor Electrode Nature 238, 37-38 Jiménez-Villacorta, F., Prieto, C., Huttel, Y., Telling, N D., van der Laan, G., 2011 X-Ray Magnetic Circular Dichroism Study of the Blocking Process in Nanostructured Iron-Iron Oxide Core-Shell Systems Physical Review B 84, 172404 Kronawitter, C X., Vayssieres, L., Shen, S., Guo, L., Wheeler, D A., Zhang, J Z., Antoun, B R., Mao, S S., 2011 A perspective on solar-driven water splitting with all-oxide hetero-nanostructure Energy & Environmental Science 4, 3889-3899 Kuang, S., Yang, L., Luo, S., Cai, Q., 2009 Fabrication, Characterization and Photoelectrochemical Properties of Fe2O3 Modified TiO2 Nanotube Arrays Applied Surface Science 255, 7385-7388 Krüger P., 2010 Multichannel multiple scattering calculation of L2,3 –edge spectra of TiO2 and SrTiO3: importance of multiplet coupling and band structure Physical Review B 81, 125121 Le Fèvre, P., Danger, J., Magnan, H., Chandesris, D., Jupille, J., Bourgeois, S., Arrio, M.-A., Gotter, R., Verdini, A., Morgante, A., 2004 Stoichiometry-Related Auger Lineshapes in Titanium Oxides : Influence of Valence-Band Profile and of Coster-Kronig Processes Physical Review B 69, 155421 Liou, F.-T., Yang, C Y., Levine, S N., 1982 Photoelectrolysis at Fe2O3/TiO2 Heterojunction Electrode Journal of the Electrochemical Society 129, 342-345 Magnan, H., Le Fèvre, P., Chandesris, D., Krüger, P., Bourgeois, S., Domenichini, B., Verdini, A., Floreano, L., Morgante, A., Resonant Photoelectron and Photoelectron Diffraction across the Fe L3 edge of Fe3O4 Physical Review B 81, 085121 Magnan, H., Stanescu, D., Rioult, M., Fonda, E., Barbier, A., 2012 Enhanced Photoanode Properties of Epitaxial Ti Doped Į–Fe2O3 (0001) Thin Films Applied Physics Letters 101, 133908 Rioult, M., Magnan, H., Stanescu, D., Barbier, A., 2014 Single Crystalline Hematite Films for Solar Water Splitting: Ti-Doping and Thickness Effects Journal of Physical Chemistry C 118, 3007-3014 Rioult, M., Belkhou, R., Magnan, H., Stanescu, D., Stanescu, S., Maccherozzi, F., Rountree, C., Barbier, A., 2015 Local Electronic Structure and Photoelectrochemical Activity of Partial Chemically Etched Ti-doped Hematite Surface Science 641, 310-313 Rioult, M., 2015 Hematite-Based Epitaxial Thin Films as Photoanodes for Solar Water Splitting PhD Dissertation, Ecole Polytechnique Rioult, M., Stanescu, S., Fonda, E., Barbier, A., Magnan, H., 2016 Oxygen Vacancies Engineering of Iron Oxides Films for Solar Water Splitting Journal of Physical Chemistry C 120, 7482-7490 Steier, L., Herraiz-Cardona, I., Gimenez, S., Fabregat-Santiago, F., Bisquert, J., Tilley, S D., Grätzel, M., 2014 Understanding the Role of Underlayers and Overlayers in Thin Film Hematite Photoanodes Advanced Functional Materials 24, 7681-7688 van de Krol, R., Liang, Y., Schooman, J., 2008 Solar Hydrogen Production with Nanostructured Metal Oxides Journal of Material Chemistry 18, 2311-2320 Xu, Y., Schoonen, M A A., 2000 The Absolute Energy Positions of Conduction and Valence Bands of Selected Semiconducting Materials American Mineralogist 85, 543-556 Yang, X., Liu, R., Du, C., Dai, P., Zheng, Z., Wang, D., 2014 Improving Hematite-based Photoelectrochemical Water Splitting with Ultrathin TiO2 by Atomic Layer Deposition ACS Applied Materials & Interfaces 6, 12005-12011 Zhao, H., Fu, W., Yang, H., Xu, Y., Zhao, W., Zhang, Y., Chen, H., Jing, Q., Qi, X., Cao, J., Zhou, X., Li, Y., 2011 Synthesis and Characterization of TiO2/Fe2O3 Core-Shell Nanocomposition Film and Their Photoelectrochemical Property Applied Surface Science 257, 8778-8783 11