Hematite-based photoanodes were produced by spin-coating a polymeric precursor solution onto FTO substrates, followed by thermal treatments. A rigorous cleaning process of the commercial FTO substrates prior to deposition (Fig. 1a) was adopted due to the resizing of as-received slides to 2 cm2 samples, with subsequent thermal treatment to remove impurities that could impact the substrate conductivity. The pristine hematite (H) polymeric precursor was prepared by mixing Fe3+ ions in citric acid and ethylene glycol solution. This mixture was heated to cause polyesterification and enable cations to chelate through the polyester. For Al3+ doped precursor (H-Al), aiming at favoring solid solution with Fe3+ in hematite structure, the cations were mixed with Fe3+ before the polymerization into the gel, enabling a homogeneous distribution of both ions in the liquid phase (dashed blue box from Fig. 1b). For Zr4+ addition, the previously produced gel containing Fe3+ or Al3+/Fe3+ was reduced to 50% of its volume by slow evaporation and, only after the precursor concentration, a Zr4+ solution in ethanol was poured into the thermoplastic gel to produce the so-called H-Zr and H-Al-Zr gels (dashed green box from Fig. 1b). This protocol was chosen for Al3+ and Zr4+ to have different chemical environments in the polymeric precursor, potentially favoring Al3+ to dope hematite lattice and Zr4+ to segregate at the interfaces.
After spin-coating deposition (Fig. 1c), the samples were annealed at 550 ºC for 30 min in air atmosphere to remove organic compounds and promote hematite phase formation, and then placed at 750 ºC for 30 min under controlled N2 flux to activate the film surface for photoelectrochemical application (Fig. 1d). As part of the synthesis design, the optimal Al+ 3 and Zr4+ concentrations were evaluated based on their photocurrent response under sunlight illumination at 100 mW cm− 2 (Fig. S1). Both cations showed a photocurrent increase as a function of concentration followed by a decrease, likely attributed to multipole clustering 14. The maximum photocurrents were obtained for the samples containing 0.5 wt.% Al3+ (named H-0.5Al) and 3.0 wt.% Zr4+ (named H-3.0Zr), and therefore those concentrations were used to design the dual-modified photoanode (named H-0.5Al-3.0Zr).
For a direct comparison and to highlight the impact of the dual modification in contrast to the individual doping, linear voltammetry measurements were performed for H, H-0.5Al, H-3.0Zr, and H-0.5Al-3.0Zr to evaluate the photocurrent response under simulated AM 1.5 G sunlight (100 mW cm− 2), as outlined in Fig. 2a. The photocurrent response in the dark was negligible due to its low catalytic behavior. Under illumination, H and H-Al photoanodes exhibited a photocurrent response at 1.23V versus reversible hydrogen electrode (RHE) of 0.7 mA cm− 2 and 1.35 mA cm− 2, respectively. The H-Zr achieved values of 1.95 mA cm− 2, and the dual-modified H-Al-Zr photoanode led to a maximum photocurrent response of 3.5 mA cm− 2 at 1.23 VRHE and up to 5.0 mA cm− 2 at 1.5 VRHE. To further improve the photoelectrochemical performance of the photoanodes, a commonly known co-catalyst NiFeOx was electrodeposited 37 onto H-Al-Zr sample, leading to a photocurrent enhancement to 4.5 mA cm− 2 at 1.23 VRHE and up to values beyond 6.0 mA cm− 2 at 1.5 VRHE (Fig. 2a).
The addition of dopants did not significantly change the optical properties of hematite, as suggested by the absorbance profiles shown in Fig. 2b. This resulted in similar values of maximum experimental currents from the absorption properties (Jabs) for the photoanodes (Fig. 2c). Therefore, the half-cell overall efficiency (ηoverall), which represents the total yield of current generated in relation to the Jabs, varied according to the dopant(s) introduced (Fig. 2d). As the reduction of polaronic effects hypothesized for Al3+ lattice doping only improves intragrain electronic mobility (i.e., within the crystal lattice), the overall efficiency of H-0.5Al is slightly higher than for pristine hematite (H). The attributed segregation of Zr4+ at the interfaces, on the other hand, can address the bottleneck of bulk electronic conductivity by efficiently decreasing Schottky energy barriers at the grain boundaries 40, therefore allowing electron hopping through hematite grains and consequently enhancing its overall efficiency more prominently.
To elucidate the effect of Zr4+ incorporation when it segregates at the grain boundary, density functional theory (DFT) calculations was performed using a grain boundary (GB) model with 25° between grains (Fig. 3a), following the procedure proposed in the previous report. 40 The chemical substitution of Fe3+ by Zr4+ at the GB promote a reduction of the effective electron potential energy barrier (Fig. 3b) that can be understood by the lower potential energy (Zeff \(\frac{e}{r}\)) of the higher effective atomic number (Zeff) of Zr. A quantitative estimation of the impact of this lower potential energy barrier on the electron transport across the GB, was done by approximating to a quantum tunneling problem of a rectangular barrier potential profile. Herein, the barrier width is a = 2.5 Å, and the height decreases from V0 = 1.88 in pure hematite to V1 = 1.48 eV upon Zr segregation, relative to the bulk mean potential (\(\stackrel{-}{{v}_{b}}\)). The electron transmission was calculated and increases upon Zr doping by ~ 3 times for lower electron kinetic energies E < V1 and ~ 1.5 times for higher electron kinetic energies E > V0 (for details, see Fig. S2). The conductivity for a polycrystalline material with non-percolating grains transport depends directly on the tunneling rate between grains. As such, the photocurrent density will be proportional to the transmission rate of the electrons between grains, making this microscopic model an effective interpretation for the segregation effects, as already shown for Sn4+ incorporation 40.
The synergistic combination of Al3+ and Zr4+ incorporation (H-0.5Al-3.0Zr) delivered a photoanode with 57% of ηoverall, and the addition of NiFeOx co-catalyst enhanced the overall photoanode efficiency up to 73%, indicating a successful optimization that led to materials capable of achieving both cost-effective photocurrent generation and efficient sunlight into current conversion. Indeed, Fig. 2e shows a 2.5-fold increase in the incident photon-to-current efficiency (IPCE) values for H-0.5Al-3.0Zr/NiFeOx compared to pristine hematite at 350 nm, and 63% of absorbed photon-to-current efficiency (APCE) at the same wavelength (Fig. 2f), with spectral signature due to Al-Zr/NiFeOx modifications.
The high efficiencies are translated into high faradaic efficiency of gas evolution, confirmed by the chromatographic gas quantification (Fig. 2g), which indicates that almost all photogenerated charge is used for the overall water splitting reaction. Moreover, chronoamperometry analysis (Fig. 2h) employed for faradaic efficiency calculations highlights the photoelectrochemical stability of the optimized H-Al-Zr/NiFeOx photoanode. This result implies the modifications introduced did not damage α-Fe2O3 well-known stability 41.
Furthermore, the comparative bar graphic (Fig. 2i) shows the catalytic efficiency (ncat.) and charge separation efficiency (nsep.) values calculated from electrochemical measurements with H2O2 used as a hole scavenger (Fig. S3). While H-Al and H-Zr showed a particular improvement for nsep., the dual-modified photoanode showed a synergistic effect that goes beyond the simple summation of the individual contributions, with a remarkable enhancement of catalytic efficiency. In this context, intensity modulated photocurrent spectroscopy (IMPS) was applied as a frequency-resolved optoelectronic characterization technique to elucidate the charge transport mechanism during simulated PEC operation. Figure 4 shows an IMPS overview of charge carrier dynamics for each photoanode with individual or dual doping.
Low frequency semicircles (first quadrant) of IMPS plots can be used to determine the transfer and recombination rate constants that may depend on the applied potential (Fig. S4) 42. The presence of pronounced semicircles in the first quadrant for pristine hematite (Fig. S5) and H-Al (Fig. 4e) indicates that surface processes strongly affect the photoanode efficiency 43. However, a slight overall enhancement of charge separation and transfer efficiencies are observed for H-Al (Fig.s 4i and 4m) throughout the potential range analyzed. This could be attributed to the reduction of polaronic effects at the hematite grains, promoted by a symmetry change of Fe3+ chemical environment upon Al3+ addition. The enhancement of charge transport is limited, though, because the bottleneck of planar hematite conductivity is the high energy barriers at the grain boundaries 40.
Conversely, the IMPS plots of H-Zr and H-Al-Zr display a charge transfer enhancement at potentials more positive than 1.2 V since the first quadrant loops disappeared. This result shows that Zr4+ is most likely segregating at hematite grain-grain interfaces, hence lowering the energy barrier height at the grain boundaries and allowing improved charge transport. Although a shift in the charge transfer efficiency (Fig. 4n) is noticeable after Zr4+ addition, it seems that the simultaneous modification with Al3+ and Zr4+ minimizes that voltage loss, thus achieving the maximal charge transfer efficiency with smaller overpotential when compared to H-Zr (Fig. 4o). Interestingly, charge separation efficiency (CSE) at 1.2 V for H-Al-Zr presents a 3-fold and 1.5-fold increase compared to pristine hematite or H-Al, and H-Zr, respectively. These results indicate the relevant synergistic effect of Al3+ and Zr4+ doping, while particularly indicating the benefits of designed Zr4+ segregation.
The addition of NiFeOx co-catalyst improved the overall j-V curve of H-Al-Zr by showing a cathodic shift in the onset potential (Fig. S6) and increasing the photocurrent response, suggesting that passivation of surface states occurred prior to charge transport enhancement 44. The consequence is not only the minimization of the overpotential necessary for water oxidation (Fig. 4p), but also an increase in charge separation efficiency (Fig. 4l), which results in higher external quantum efficiency (EQE) values (Fig. 4t). These results show that, although charge carrier dynamics analysis can be a powerful tool to unveil the role of the dopants, a thorough microstructural analysis is also needed to objectively establish the photocurrent dependences on dopant distributions.
Transmission electron microscopy (TEM) coupled with energy-dispersive X-ray spectroscopy (EDS) technique was employed to further consolidate the evidence found about the Al3+ and Zr4+ roles (Fig. 5). From the analyzed representative region, it is possible to observe the FTO substrate at the bottom of the HAADF image with a brighter contrast and a fragment of the porous hematite nanostructure on top, where hematite grains are easily distinguishable. The image also highlights some hollow zones due to imperfect adherence on the substrate.
In addition, EDS analyses performed on the sample cross-sections revealed the elemental distribution across the microstructure induced by the PPS method. It is possible to verify that the Al+ 3 is uniformly spread across the grains, consistent with the formation of solid-solution as expected, considering its favorable enthalpy of mixing and the adopted syntehtic method. Conversely, Mott-Schottky analysis showed a charge donor density of 4.87x1019 cm− 3 for H and 8.53x1019 cm− 3 for H-Al (Fig. S7, Table S1). This slight increase in charge donor density observed upon Al3+ addition is due to its same oxidation state as Fe3+, with the increase being attributed to Al3+ more ionic behavior.
On the other hand, Zr+ 4 ions are segregated to regions that, when combined with the HAADF image, indicate an overlap with the grain boundaries, with some limited excess formation also at the surface regions. In agreement, the Mott-Schottky analysis of H-Zr showed a donor density of 1.14x1020 cm− 3 (Fig. S7, Table S1), corroborating that the enhanced charge separation shown by IMPS analysis is due to the excess of Zr4+ content segregated at the interfaces. This demonstrates the effectiveness of PPS to induce the designed segregation, particularly at the grain-grain interfaces, with a posssible cause-effect relationship with the observed photocurrent trends. Finally, one can see from EDS Sn4+ mapping that no significant amount of tin diffused into the sample, which supports negligible interferences in the electrochemical measurements 16.
X-ray photoelectron spectroscopy (XPS) analysis of pristine hematite (H) photoanode shows Fe 2p1/2 (724.3 eV) and Fe 2p3/2 (710.7 eV) peaks characteristic of α-Fe2O3 crystalline phase, confirming the presence of trivalent oxidation state of iron (Fig. S8a). The peaks attributed to the presence of Fe2+ species (Fe 2p1/2 at 722.9 eV and Fe 2p3/2 at 709.5 eV) are reasonable considering that annealing under inert atmosphere (N2 flux for 30 min) is most likely to change the surface chemistry of the photoanodes, either creating or removing defects which may cause the arise of Fe2+ species for charge compensation 45. Indeed, in the O 1s spectra of pristine hematite (Fig. S8b), while the most prominent peak (529.6 eV) is easily assigned to hematite lattice oxygen atoms (Olatt.), the peak centered at 531.4 eV can be attributed to oxygen atoms in the vicinity of an oxygen vacancy or in the coordination sphere of a metal next to an oxygen vacancy 46. This signal arises from the expected lack of core electrons to be emitted by an oxygen vacancy, as assigned (OV). Moreover, the O 1s peak at 532.8 eV was assigned to superficially adsorbed -OH groups (Oads.) for hematite (H) sample.
There is a slight chemical shift of Fe 2p spectra to higher binding energies after doping regardless of its nature (Fig. S9a-d). This behavior is expected for segregation, since the interfacial excess of ions would interfere in the Fe-O superficial bonds, consequently attracting the electron cloud of both Fe and O 8. The similar behavior being observed for Al3+ doping could indicate that a local distortion caused by the different ionic radius might also change the chemistry of Fe3+ at the surface. However, XPS analysis is inconclusive for Al3+, since EDS mapping showed that this dopant is most likely to form a solid solution at hematite bulk rather than forming excess at the surface. Chemical shifts to lower binding energies observed in O 1s spectra (Fig. S9e-h) corroborate that the introduction of dopants gave a more ionic character to metal-oxygen bonds at the surface of the photoanodes 31. Also, Al 2p peaks at 73.4 eV for H-Al and H-Al-Zr photoanodes show the presence of Al3+ (Fig. S8c), while Zr 3d spectra with doublet Zr 3d3/2 (181.6 eV) and Zr 3d5/2 (184.0 eV) shown in Fig. S8d confirm the predominance of tetravalent species for zirconium. Furthermore, an evaluation of H-0.5Al-3.0Zr/NiFeOx surface composition after the 5h operation shown in Fig. 2h revealed the maintenance of dopant amounts introduced to the photoanode (Fig. S10, Table S2), reinforcing the method’s ability to deliver modifications into hematite structure while preserving its chemical stability under photoelectrochemical operation conditions.
An evaluation of the superficial morphology of the photoanodes done by atomic force microscopy (AFM) (first row of Fig. 6) showed that the grain size of H-Al sample is similar to pristine hematite (H), which is consistent with the assumption of lattice doping. However, the topography maps indicate that grain size seems to be considerably smaller for H-Zr. Ion segregation at interfacial regions of oxide systems have been systematically correlated with grain size reduction 47. This is because the interface energy decrease (either surface or grain boundary) associated with interfacial excess reduces the thermodynamic driving force for coarsening. This phenomenon is connected by the Gibbs adsorption isotherm to a local energy lowering 48 and allows greater stability of nanoparticles during synthesis or annealing protocols. Moreover, the grain size refinement due to ion segregation is commonly associated with a diminution of surface roughness since roughness is related somehow to adherence but mainly to the average grain sizes 11. The electrochemical surface area (ECSA) obtained from cyclic voltammetry and calculated according to the procedure detailed in the supplementary material (Fig. S11) showed ECSA values of 5 (H), 6 (H-Al), 11 (H-Zr) and 18 (H-Al-Zr), corroborating the topographic characteristics observed by AFM.
From Kelvin probe (KPFM) measurements, the surface potential distribution maps (second row of Fig. 6) display by high contrast a gradient accumulation of H and H-Al surface potential. Since both samples exhibit non-optimized interfaces, the electronic conductivity is not favored toward the bulk and therefore the electric current does not flow through the material. For H-Zr, the engineered grain-grain chemistry with Zr4+ allows electron hopping through the grain boundaries by lowering Schottky’s energy barriers for charge conductivity. The result is a low contrast KPFM surface potential distribution map because the electric current flows readily through the material. The dual-modified photoanode presented a somewhat intermediate behavior when compared to H-Al and H-Zr samples, indicating that a superficial trade-off of charge distribution at the space charge layer might be responsible for the synergistic role of the dopants together. While H-Zr has improved photoelectrochemical performance in contrast to H-Al, one may speculate that, for H-Al-Zr, because Al3+ is dissolved in the hematite matrix, Zr4+ is further forced to segregate as the crystal lattice shrank due the Fe3+ substitution to less electronegative ions. This would cause the dual-modified photoanodes to have a greater concentration of Zr4+ interfacial excess and, therefore, the most stable grains.
The capacitance gradient (dC/dz) maps displayed in the third row of Fig. 6 also corroborate the dopant distribution hypothesized by the PPS synthesis protocol. The higher contrast observed at the grain-grain interfaces of H and H-Al suggests that the grain boundaries act as charge recombination sites by accumulating and trapping charge carriers. It is also noticeable a reduction of contrast in the grains with the addition of Al3+ as a dopant, indicating that Al3+ is improving the conductivity within hematite grains by reducing polaronic effects due to Fe3+ electron effective mass 49. Moreover, the H-Zr capacitance gradient map shows that the segregation of Zr4+ at the interfaces and its consequent grain size pinning mitigate the hematite grain boundary trapping effect. The combination of Al3+ and Zr4+ effects led the H-Al-Zr capacitance gradient map to exhibit low contrast grains compared to H-Zr. It also presented negligibly contrasted grain boundaries compared to H-Al, once again reinforcing the synergy of the dual modification designed by the PPS method.
Furthermore, electrochemical impedance spectroscopy (EIS) results support the trends observed in KPFM maps. Open circuit potential (OCP) measurements show the small photovoltages required for water oxidation (Fig. S12b, Table S3) and lower charge recombination rates (Fig. S12c) for dual-modified samples. Nyquist plots at 1.23 VRHE under 100 mW cm− 2 illumination (Fig. S13) were fitted with the model circuit [Rs(CbulkRtrapp)(CPEssRss,ct)], where the electrolyte resistance is represented by Rs, (CbulkRtrapp) represents the hematite bulk capacitance and its charge trapping resistance, and (CPEssRss,ct) represents the surface capacitance and the resistance associated with charge carrier transfer from bulk to the electrolyte in the presence of surface states. The CPE element was used due to the non-ideal dielectric behavior of the hematite solid-liquid interface in thin films 50. The EIS fitting (Table S4) showed the effectiveness of Al-Zr addition, improving hematite's charge transport properties. Fitted data shows a tendency of resistance drop (Rss,ct) with the addition of each element, which means that H-Al-Zr/NiFeOx exhibited the lowest Rtrapp and Rss,ct values of all photoanodes, justifying its best photoelectrochemical behavior.
The ultraviolet photoelectron spectroscopy (UPS) electronic structure analysis of the photoanodes exhibits the effects of dopant addition on the band diagrams of the semiconducting hematite-based photoanodes (Fig. 7). A pronounced shift of the Fermi level toward the conduction band (CB) is observed for Al3+ doping, corroborating that aluminum reduces polaronic effects since small polarons are mainly responsible for hematite Fermi level pinning 51. H-3.0Zr exhibited a 0.66 eV energy difference from its CB position and the water reduction thermodynamic potential (-4.5 eV in Evacuum), which was the largest value observed among the studied photoanodes. These results indicate the H-Zr requirement for a higher overpotential to drive the water oxidation reaction, consistent with the behavior previously observed in the IMPS charge transfer efficiency (Fig. 4n). Also, H-Zr more negative valence band (VB) position compared to H-Al shows the greater efficacy of Zr4+ segregation over Al3+ lattice doping on charge transport, also corroborating the trend observed for IMPS charge separation efficiencies in Fig.s 4i and 4j. For H-Al-Zr/NiFeOx, the trade-off of Al3+ and Zr4+ doping characteristics, associated with the co-catalyst behavior, led to a photoanode with reduced overpotential for water oxidation (0.4 eV from CB to water reduction potential), with the Fermi level closer to CB (thus favoring charge transfer efficiency), and the VB position more negative than pristine hematite (indicating superior charge separation efficiency). Taken together, these findings and the performance of the photoanodes show the intricate relation between the band positioning of the semiconducting absorber materials and their photoelectrochemical behavior, once again corroborating the previous results and validating the hypothesis of selective doping with Al3+ and Zr4+ promoted by the polymeric precursor solution (PPS) method.
In summary, the selective modifications introduced onto hematite following the PPS protocol address three main drawbacks of α-Fe2O3: charge recombination intragrain, charge recombination at the grain boundaries, and recombination due to surface trapping states. As shown in Fig. 8a, charge carrier transport in pristine hematite is strongly affected by bulk and surface recombination. Fig. S16b shows that Al3+ lattice doping (brown circles) cannot enhance the overall conductivity, but only reduces intragrain polaronic effects. When Zr4+ segregates at the interfaces (gray circles), as shown in Fig. S16c, a trade-off between increased conductivity at the grain boundaries and the creation of more surface states is observed, resulting in enhanced charge separation efficiency. When both Al3+ and Zr4+ are incorporated (Fig. 8b), a synergistic effect is observed by aggregating the beneficial individual effects of the dopants. The addition of NiFeOx (small light-yellow circles) shown schematically in Fig. 8c further improve the photoelectrochemical performance of the photoanodes by passivating surface states created by Zr4+ addition and therefore contributing to charge separation efficiency. The result of this rational engineering of hematite interfaces is a photoanode with state-of-art performance, as seen in Fig. 8d and Table S5. The photoanode prepared by PPS showed a competitive photocurrent while employing earth-abundant chemical elements and being easily adapted for industrial scale.