Aadesh P. Singh*ab,
Camilla Tossic,
Ilkka Tittonenc,
Anders Hellmana and
Björn Wickman*a
aDivision of Chemical Physics, Department of Physics, Chalmers University of Technology, SE-412 96 Göteborg, Sweden. E-mail: bjorn.wickman@chalmers.se; Tel: +46 31 772 51 79
bEngineered Nanosystems Group, School of Science, Aalto University, P. O. Box 13500, 00076 Aalto, Finland. E-mail: aadesh.singh@aalto.fi
cDepartment of Electronics and Nanoengineering, School of Electrical Engineering, Aalto University, P. O. Box 13500, 00076 Aalto, Finland
First published on 9th September 2020
Solar energy induced water splitting in photoelectrochemical (PEC) cells is one of the most sustainable ways of hydrogen production. The challenge is to develop corrosion resistant and chemically stable semiconductors that absorb sunlight in the visible region and, at the same time, have the band edges matching with the redox level of water. In this work, hematite (α-Fe2O3) thin films were prepared onto an indium-doped tin oxide (ITO; In:SnO2) substrate by e-beam evaporation of Fe, followed by air annealing at two different temperatures: 350 and 500 °C. The samples annealed at 500 °C show an in situ diffusion of indium from the ITO substrate to the surface of α-Fe2O3, where it acts as a dopant and enhances the photoelectrochemical properties of hematite. Structural, optical, chemical and photoelectrochemical analysis reveal that the diffusion of In at 500 °C enhances the optical absorption, increases the electrode–electrolyte contact area by changing the surface topology, improves the carrier concentration and shifts the flat band potential in the cathodic direction. Further enhancement in photocurrent density was observed by ex situ diffusion of Ti, deposited in the form of nanodisks, from the top surface to the bulk. The in situ In diffused α-Fe2O3 photoanode exhibits an improved photoelectrochemical performance, with a photocurrent density of 145 μA cm−2 at 1.23 VRHE, compared to 37 μA cm−2 for the photoanode prepared at 350 °C; it also decreases the photocurrent onset potential from 1.13 V to 1.09 V. However, the In/Ti co-doped sample exhibits an even higher photocurrent density of 290 μA cm−2 at 1.23 VRHE and the photocurrent onset potential decreases to 0.93 VRHE, which is attributed to the additional doping and to the surface becoming more favorable to charge separation.
Metal oxide semiconductors are the most promising materials for the PEC reactions thanks to their excellent stability in many electrolytes over a wide pH range, low cost, non-toxic nature and their versatility in terms of fabrication techniques.6–9 Among the various metal oxides, iron oxide (hematite; α-Fe2O3)10 is considered as one of the potential photoanode candidates for solar water oxidation, thanks to its unique properties such as a suitable bandgap (2.0–2.2 eV), low cost, non-toxicity and high stability.4,11–13 However, pristine α-Fe2O3 exhibits poor water splitting efficiency, far below the maximum theoretical efficiency of 12.9%, because of the mismatch between the valence band energy level and the water reduction potential, the short hole diffusion length of 2–4 nm and the low electron mobility.8 Several strategies have been adopted to overcome the intrinsic limitations of hematite:14 for instance, tuning the electrode morphology,15 introduce synergistic interfaces in two-dimensional stacks of semiconductors,16,17 surface activation with co-catalysts,18 and doping with various metal cations to change the electronic structure.19,20 Both nanostructuring and doping in α-Fe2O3 photoanodes have been extensively investigated to improve its PEC properties.13,21 For instance, altering the electrode morphology in the form of nanoribbons,22 nanobelts,23 nanorods,24 mesoporous layers25 can reduce the charge recombination rate by providing a shorter path length for photogenerated holes to reach the surface,26 where they can participate in the water oxidation process thereby enhancing the reaction rate and generating proton (H+) for hydrogen production.27–29 Further, the electrical conductivity of α-Fe2O3 can be improved by cationic doping, which boosts the PEC performance thanks to the increased donor concentration and promoted charge transfer.30,31 Many research groups have reported that co-doping of α-Fe2O3 may significantly improve the PEC performance through various mechanisms.29,32–34 In several studies, a combination of anionic and cationic dopants has been investigated, such as Zn and Ti: co-doped α-Fe2O3 enhances the PEC device performance through increased electrical conductivity and improved charge transport properties.35,36 However, much of the details behind the improvements of co-doping and nanostructuring of α-Fe2O3 is still unknown.
In this work, our goal is to increase the understanding of nanostructuring and co-doping in a α-Fe2O3 model photoelectrode system for solar water splitting, which can then be further integrated into more complicated electrode architectures.37,38 Doping methods and nanofabrication techniques are heavily investigated in regard to the effect on the final performance,38–40 but there is still ample room to investigate the exact mechanisms that cause the improvement: for example, it is well known that annealing improves the hematite water splitting efficiency,22,41 but the diffusion of the dopant into the material from a substrate or a surface decoration as a consequence of the annealing, is investigated more rarely.24,42,43 In parallel, when the dopant is the main subject of interest, controllable yet elaborate methods are employed.11,44,45 In the present study, a relatively simple method leads to a diverse ensemble of effects at once, ranging from the diffusion of different dopants, to the modification of the electrode structure morphology, to the addition of functional groups to the outer surface. The material is doped in situ with indium, forming a gradient, and ex situ with titanium on the surface: gradient doping, as demonstrated by Abdi et al.,46 creates a distributed homojunction that enhances the charge-separation efficiency in water splitting, in comparison to homogeneous doping or to a p–n homojunction. Hematite doped/co-doped by titanium,44 phosphorous,47 zirconium and tin48 has recently been the object of investigation, for example by Srivastav et al.:44 they determined that the deposition of hematite layers with increasingly concentrated titanium doping provided an enhanced performance with respect to homogeneous doping, in terms of photocurrent density and carrier separation efficiency. The mechanism lays in the gradual shift of the Fermi level, which induces a more accentuated band bending than the one present in homogeneous doping: this will cause the formation of an electrical field, as the Fermi level equilibrates, which facilitates the charge movement across the material. Ex situ doping, instead, affects mainly the surface states,18 further encouraging charge separation and transport at the interface, as well as interaction with holes in water oxidation, but without affecting the morphology of the material.49
With respect to nanostructuring, the thickness of the α-Fe2O3 layer is a known important parameter and is herein used as the first modification strategy, i.e., focus is on ultra-thin hematite planar model photoanode with a thickness of ∼25 nm fabricated via oxidation of a ∼10 nm thick iron layer. While with respect to co-doping, in situ indium diffusion from the underlying ITO substrate and titanium doping by post-growth surface modification are used as the second modification strategy. The obtained results demonstrate that a higher annealing temperature, 500 °C instead of 350 °C (the temperature used to manufacture the hematite film), affect the structural, optical and electronic properties of the material, thanks to in situ modifications by diffusion of In from the underlying ITO substrate, and the changes improve the photoelectrochemical performance. The ultra-thin hematite layer was further modified ex situ, by co-doping of Ti from the top of hematite surface. The photocurrent density at 1.23 V vs. reversible hydrogen electrode (RHE) reaches 290 μA cm−2, with an onset potential of 0.93 V vs. RHE. Further analytical results illustrate that oxygen vacancies (VO) are formed after doping, thanks to the high temperature annealing and surface decoration by Ti nano-disks. The presences of VO increase the carrier density, leading to efficient charge separation and transport. The characteristics of In/Ti co-doped ultra-thin hematite photoanodes with high VO enable high PEC performance and introduce a useful strategy to design superstructure-based systems for efficient solar fuel production.50
A 10 nm thick Fe layer, when annealed at 350 °C, transforms into a ∼25 nm thick layer of pure α-Fe2O3, as confirmed by micro-Raman spectroscopy. The annealing at 500 °C results in some chemical and morphological change, marked by a shift in the Raman peak, visible in Fig. 2a. The Raman spectra collected on both the samples (Fig. 2a) exhibit five out of seven spectral signatures for α-Fe2O3 (221 (A1g), 285 (Eg), 396 (Eg), 479 (A1g) and 597 (Eg)).53 The 249 (Eg) and 302 (Eg) spectral signatures are most probably overlapped with the other peaks in Raman spectra, while the peaks associated with either maghemite or magnetite are not observed.54,55 This indicates that both annealing temperatures, 350 and 500 °C for 8 h, are suitable for transforming the Fe films to α-Fe2O3. The inset of Fig. 2a presents the Eg peak at 285 cm−1, where the blue-shifting and FWHM-broadening in the sample annealed at 500 °C are clearly visible. This blue shift in frequency and a corresponding broadening in FWHM of the phonon peaks represents a decreasing crystallite size.55 Further, the intensity of the Raman scattering is higher in the sample annealed at higher temperature (i.e. 500 °C): this is intuitive, as Raman scattering intensity is proportional to the number density of the interacting media.56 Therefore, the increase in the signal intensity in 500 °C annealed sample, as shown in the in-set of Fig. 2a, is an indication of higher film thickness and crystallinity in comparison to the sample annealed at 350 °C.
The optical absorption behavior of hematite samples, as prepared and modified by Ti doping, were characterized by UV-vis absorption in the wavelength range 250–650 nm (Fig. 2b). The absorbance spectra in the samples, annealed at 350 and 500 °C, are similar to each other below 350 nm wavelength, except in the 350–600 nm as the samples annealed at 500 °C exhibit more absorption in this wavelength range. This increased absorption in the visible region is probably due to the slightly higher film thickness in samples annealed at 500 °C, and to the diffusion of substrate impurities (discussed in detail in XPS section);57 such increase is known to happen in In-doped hematite films.45 Further, doping with Ti4+ in both pristine and In-diffused α-Fe2O3 further increases the optical absorption in the wavelength range 350–485 nm, indicating stronger absorption,45,58 and showing an apparent blue-shift of the visible-light absorption peak. This enhancement could be attributed to the addition of the 5 nm thickness over the hematite surface for the incorporation of Ti4+ ions, which might increase the overall thickness of the hematite film after annealing at 350 °C.59 However, Ti doping and In/Ti co-doping in hematite show slightly weaker absorption at a lower wavelength, probably due to the oxidation of Ti to TiO2 which is a wide band gap material and able to absorb UV light. The absorption peak centered around 390–420 nm is assigned to a direct transition O−2(2p) → Fe3+(3d) in hematite, caused by holes generated at oxygen sites, while the broader peak between 500 and 600 nm is assigned to an indirect transition Fe3+(3d) → Fe3+(3d) (spin flip), caused by holes generated at iron sites.60,61 The apparent shift caused by the titanium doping can be justified by an efficiency increase of the first type of transition, while the second type maintains its initial absorption.62,63 All the samples exhibit a gradual decrease in absorbance until 610 nm, corresponding to a 2.02 eV bandgap, which is within the range usually reported for hematite (2.0–2.2 eV).
To analyze the structural and optical results we further characterize the hematite samples for surface morphology by FE-SEM and HR-TEM, and also performed EDX analysis in TEM to see any potential difference in the chemical composition with respect to the annealing temperature. Fig. 3a and b show the FE-SEM images for the samples prepared at 350 °C and 500 °C, respectively. The annealing temperature has a significant effect on surface morphology: the sample annealed at 350 °C shows a flat granular surface with nano-sized grains. In contrast, the sample annealed at 500 °C shows a flaky structure with high surface roughness, which might increase the effective electrode–electrolyte surface area and therefore enhance the total photocurrent density.39 Cross-section TEM micrographs show the formation of dense α-Fe2O3 films, with thicknesses of approximately 25 nm and 27 nm for the samples prepared at 350 and 500 °C, respectively. The samples annealed at 500 °C exhibit high crystallinity and slightly higher thickness (27 nm) as compared to samples annealed at 350 °C. These results support the blue shifting and increase in peak intensity in Raman spectra for the samples annealed at 500 °C. In order to verify the diffusion of substrate impurities into the hematite thin films, EDX analysis installed in TEM was carried out, as shown in ESI Fig. S1(1-c).†
Fig. 3 FE-SEM images of α-Fe2O3 thin films annealed at (a) 350 °C and (b) 500 °C. TEM and HR-TEM images of α-Fe2O3 thin films annealed at (c) 350 °C and (d) 500 °C. |
To further analyze the chemical composition, in relation to both annealing temperature and Ti doping, we characterized the samples by XPS. Fig. 4 shows the XPS plots collected on all four samples. Survey spectra (Fig. S2†) of α-Fe2O3 samples prepared at 350 and 500 °C mainly display the peaks corresponding to Fe, O and C.64 Surprisingly, a few additional peaks, at 444 and 45 eV, were detected, corresponding to In(3d) in the samples prepared at 500 °C, but not in the samples prepared at 350 °C.38,64 This strongly implies that In atoms are diffusing from the ITO substrate into the α-Fe2O3 layer at the higher sintering temperature, where they act as electron donating substitutional impurities. Fig. 4a shows the XPS spectra of the Fe(2p) core level, obtained from both α-Fe2O3 samples: it is evident the Fe(2p) orbital splitting due to spin orbit interaction, resulting in Fe(2p3/2) and Fe(2p1/2) states.65 The Fe(2p3/2) peak in both samples was obtained at a binding energy of ∼711.0 eV (Fig. 4a), consistently with the typical values observed for hematite.66 The energy separation between Fe(2p3/2) and Fe(2p1/2) (Δ = 13.5 eV) in both samples clearly supports the formation of pure Fe(III) oxide, as was already indicated by Raman spectroscopy.66 In the case of Ti doping, the absence of any additional satellite peaks related to Fe2+ at 730 or 715 eV indicates that Fe2+ does not exist at the surface on doped samples after Ti doping.58 The O 1s XPS spectrum (Fig. 4b) exhibits two peaks at around 529.3 and 531.6 eV in all samples: the first peak, can be attributed to Fe–O bonds (lattice O2−) in hematite, while the second peak is clearly variable with respect to the treatment of each sample. The peak can be assigned to O− (Vo), which is developed due to loss of oxygen, creation of oxygen vacancies or presence of absorbed OH− in the samples.58,67 The effect appears in the sample that undergoes only annealing at 500 °C; in the sample annealed at 350 °C and doped with Ti a further growth of the peak is visible, alongside a broadening around 553.5 eV, which is due to adsorbed water on the surface. Finally, the sample annealed at 500 °C and doped with Ti expands on both peaks, suggesting the creation of more oxygen vacancies and the adsorption of more water molecules.68,69 The indium core level XPS measurements are visible in Fig. 4c, showing that a significant amount of In was detected on the surface of hematite samples annealed at 500 °C. However, no signals of these substrate impurities were detected in the hematite samples annealed at 350 °C, which indicates that there was no diffusion of In atoms at the lower temperature. Fig. 4c displays the peaks at binding energies of 451.7 eV and 444.2 eV, which are assigned to In 3d3/2 and In 3d5/2.64 These peaks are especially relevant, because they confirm the existence of In on the surface of hematite, demonstrating that In3+ atoms, acting as a co-dopant, have diffused effectively into the sample, as a consequence of the high-temperature annealing. However, the Ti dopant is in a +4 state, as noted by the peaks at binding energies 463.7 and 458.2 eV (Fig. 4d), allocated respectively to Ti 2p1/2 and Ti 2p3/2,44,58 with no detectable difference between the Ti spectra after annealing at the two different temperatures.
Fig. 4 XPS core level spectra for (a) Fe 2p (b) O 1s (c) In 3d and (d) Ti 2p recorded on α-Fe2O3 annealed at 350 °C and 500 °C and further modified by Ti nanodisks. |
Finally, In/Ti co-doped α-Fe2O3 (annealed at 500 °C) shows the maximum photocurrent density, giving a value of 290 μA cm−2 at 1.23 VRHE and 602 μA cm−2 at 1.50 VRHE, which is around 9-fold increase to that of pure α-Fe2O3 film at 1.23 VRHE and about 2- and 5-fold increase to that of In-doped and Ti-doped α-Fe2O3 photoanode, respectively. It is noteworthy that the onset potential of the photocurrent for In/Ti co-doped α-Fe2O3 also shifted cathodically, which would benefit the PEC water splitting at lower voltage. The negative shift of the onset potentials demonstrates a more efficient interfacial charge transfer that prevents the hole accumulation at the electrode surface and thus decreases surface charge recombination. We also performed the IPCE measurement on pristine and In-doped α-Fe2O3 photoanodes at various wavelengths, as shown in ESI Fig. S4.† Pristine α-Fe2O3 photoanodes exhibit an IPCE value of 8% at 375 nm; however, In3+ doped α-Fe2O3 photoanodes exhibit a higher IPCE, of around 12% at the same wavelength. These IPCE results are consistent with the difference in photocurrent densities as observed in pristine and In-doped α-Fe2O3 photoanodes. From the transient photocurrents measurements at a constant potential (1.23 VRHE) in Fig. 5b, it appears that the current decay (Id) (the difference between initial current (Ii) and final current (If); Id = Ii − If) increased from 5 μA cm−2 pure α-Fe2O3 photoanode to 12 μA cm−2 for Ti/In co-doped photoanode. Beyond that, all samples are prompt in generating a reproducible response to light on–off cycles, indicating that a quick charge transport process can be achieved. When exposing to 1000 s illumination, the In/Ti-co-doped α-Fe2O3 photoanode shows excellent stability with up to 98% retention of its initial photocurrent density. Additionally, in comparison to single element doped α-Fe2O3 photoanodes, the co-doped α-Fe2O3 photoanodes exhibited excellent photochemical stability and photochemical response (Fig. 6b).
Further, we performed Mott–Schottky measurement on all the samples under dark condition and measured the capacitance at the photoanode-electrolyte surface, in order to drive the Mott–Schottky curves to obtain the interface properties between the active material and electrolyte. The obtained Mott–Schottky curves (1/C2 versus VRHE) were used to determine the donor density (Nd) and flat-band potential (Vfb) by using the Mott–Schottky equation: C−2 = (2/qεoεsNd)[V − Vfb − kT/q], where εo is the permittivity of the vacuum, εs is the dielectric constant of the hematite (in case of α-Fe2O3, εs is 80), q is the electronic charge and kT/q is the thermal voltage (26 meV at room temperature).70 The value of Vfb was calculated from the intercept of the straight line to the x-axis in the linear region of the Mott–Schottky plots. The values of Nd were calculated by taking a linear region of the slope between 0.60 and 1.00 VRHE and by considering the geometrical electrode area only by using the equation, Nd = 2/(εoεsq)[d(1/C2)/dV]−1. Fig. 6a shows the Mott–Schottky plots for all the samples; the values of flat-band potential (Vfb) and donor density (Nd) derived from these plots are summarized in Table S1.† From the slope of the Mott–Schottky plots it can be seen that the α-Fe2O3 sample prepared at two different annealing temperature of 350 °C (pristine α-Fe2O3) and 500 °C (In doped α-Fe2O3) have a very similar slope, indicating that Vfb and Nd in both samples correspond to similar values. The pristine hematite sample exhibits a Vfb value of 0.55 VRHE and Nd of 8.78 × 1018 cm−3. However, a shift in Vfb to 0.51 VRHE and a minor increase in Nd up to 9.90 × 1018 cm−3 was calculated for the In doped α-Fe2O3 sample prepared at 500 °C annealing temperature. A slight increase in Nd and change in Vfb in In doped α-Fe2O3 photoanode is probably due to the change in surface morphology and relatively higher thickness of α-Fe2O3 layer annealed at 500 °C. However, unlike the sample annealed at 350 °C, the sample annealed at 500 °C shows a noticeable curvature above the flat-band potential (Fig. 6a). This curvature might originate from the nanostructured morphology caused by the higher annealing temperature, leading to a higher true electrode–electrolyte contact area.60 In general, the Mott–Schottky equation is derived by assuming a perfectly flat electrode surface.60 From the SEM images, we can see the changes in the surface morphology (Fig. 3a and b), where the sample annealed at 500 °C shows the formation of nanostructures on the surface. In the case of Ti4+ doping through making the disks with 40 nm in diameter and 5 nm in thickness following by air annealing at 350 °C, the slope and curvature of the Mott–Schottky plots increases significantly. The increases of slope and curvature in Mott–Schottky plots with Ti4+ doping suggests that both the carrier concentration and surface roughness is increasing. The Ti doping in pristine hematite sample exhibits a Vfb value of 0.52 VRHE and Nd of 9.46 × 1018 cm−3. However, a further shift in Vfb to 0.49 VRHE and increase in Nd up to 9.97 × 1018 cm−3 was calculated for the In/Ti4+ co-doped α-Fe2O3 sample.
To further understand the role of dopant, electrochemical impedance spectroscopy (EIS) analysis was adopted, as shown in Fig. 6b. The equivalent circuit adopted to fit EIS data is given in the inset of Fig. 6b. Rtrap represents the resistance for surface-state trapping of holes and electrons, and Rss describes the surface-state resistance of the charge transfer of holes to reducers in solution. The capacitance of the space charge, as well as a connection of Helmholtz capacitance, are integrated into Cbulk. Besides, Css denotes the surface-state capacitance, and Rs is the series resistance of the electrode–electrolyte setup.71 Although no significant differences could be observed of Rs, Cbulk, and Css, the reduction of Rtrap and Rss is clearly seen upon incorporating the In-dopant into the pristine hematite. This indicates that the surface recombination of holes and electrons is reduced by In-dopant, further by Ti-dopant, and the charge transfer of holes to reducers in solution is fostered. The ameliorated surface reaction may explain the improved PEC performance of Ti doped hematite, and In/Ti co-doped hematite.72,73
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ra04576d |
This journal is © The Royal Society of Chemistry 2020 |