Jiao
Wang
a,
Letizia
Liccardo
b,
Heydar
Habibimarkani
b,
Ewa
Wierzbicka
c,
Thorsten
Schultz
de,
Norbert
Koch
de,
Elisa
Moretti
*b and
Nicola
Pinna
*a
aDepartment of Chemistry, The Center for the Science of Materials Berlin, Humboldt-Universität zu Berlin, Brook-Taylor-Str. 2, 12489 Berlin, Germany. E-mail: nicola.pinna@hu-berlin.de
bDepartment of Molecular Sciences and Nanosystems, Ca' Foscari University of Venice, Via Torino 155, 30172 Venezia Mestre, Italy. E-mail: elisa.moretti@unive.it
cDepartment of Functional Materials and Hydrogen Technology, Faculty of Advanced Technologies and Chemistry, Military University of Technology, 2 Kaliskiego Street, Warsaw 00908, Poland
dInstitut für Physik and IRIS Adlershof, Humboldt-Universität zu Berlin, Brook-Taylor-Str. 6, 12489 Berlin, Germany
eHelmholtz-Zentrum Berlin für Materialien und Energie GmbH, Hahn-Meitner-Platz 1, 14109 Berlin, Germany
First published on 16th July 2024
The short lifetime of electron–hole pairs and high electron–hole recombination rate at surface states significantly limit the practical applications of hematite (α-Fe2O3) photoanodes in photoelectrochemical (PEC) water splitting. Surface modification with a TiO2 overlayer has been demonstrated to be an efficient way to improve the PEC performance. However, a fine control of the TiO2 overlayer and a deep understanding of the impact of the TiO2 overlayer with variable thickness on the PEC performance, to the best of our knowledge, has yet to be done. Here, a conformal ultrathin TiO2 overlayer is successfully deposited on hydrothermal grown one-dimensional hematite nanorod arrays by atomic layer deposition. The morphology and thickness of the TiO2 overlayer can be precisely controlled. The effect of the thickness of the TiO2 overlayer on the overall water splitting efficiency of hematite photoanodes under visible and UV light has been systematically investigated. The charge excitation and transfer mechanism at the semiconductor–electrolyte interface has also been studied.
Numerous efforts have been made to overcome these issues and improve the PEC efficiency of hematite, such as element doping, constructing heterojunction and surface modification.14–17 Among these strategies, surface modification of semiconductor photoanodes has proven to be an effective strategy to improve the charge separation and enhance the efficiency of water oxidation of hematite photoanodes. TiO2 has attracted considerable interest as a photoanode for water splitting due to its good chemical stability, outstanding optoelectronic properties and nontoxicity.18–20 Recently, many studies have focused on the modification of TiO2 overlayer on hematite for PEC water splitting applications. For instance, Yang et al. showed that an ultrathin TiO2 layer grown on hematite could lead to better charge separation and a 100 mV cathodic shift of the turn on potential because of the enhanced charge separation.12 Feng et al. displayed an effective strategy in promoting the charge transfer by decorating TiO2 at the grain boundaries of hematite.21 However, these reports mainly demonstrated an improved PEC performance of hematite photoanodes by surface treatment with TiO2. The precise control over the morphology and thickness of the TiO2 overlayer remains challenging. Furthermore, the impact of the thickness of the TiO2 overlayer on the PEC performance and the charge transfer mechanisms at the semiconductor–electrolyte interface, to the best of our knowledge, have yet to be investigated.
A TiO2 overlayer with optimized thickness can increase electronic conductivity and suppress surface charge carrier recombination.18,19,22 So far, a variety of techniques have been used to deposit a TiO2 overlayer, including spin coating, atomic layer deposition (ALD), sputtering, chemical bath deposition and dip coating.21,23–26 ALD is a thin film deposition technique based on self-terminating surface reactions, which allows the conformal and homogeneous coating on high aspect ratio substrates, with precise control of thin film thickness at the Ångström scale.27–29 Progress has already been made in the formation of homogeneous TiO2 overlayer on hematite by ALD.27,30,31 Nevertheless, achieving a conformal, pinhole-free and high quality ultrathin TiO2 film is still challenging due to limited understanding of the TiO2 growth at early stages.
In this work, TiO2 continuous ultrathin films with variable thickness are successfully deposited onto 1D hematite nanorod arrays by ALD. The morphology and the thickness of the TiO2 overlayer can be accurately controlled. Fe2O3 can be excited to generate electron–hole charge carriers under both visible and UV light owing to its low band gap, while TiO2 can only be excited under UV light because of its large band gap (3.2 eV for anatase).32,33 To understand the role of a TiO2 overlayer and the underlying charge transfer mechanism in PEC water splitting, the PEC performances of as-synthesized TiO2/Fe2O3 under visible and UV light as a function of the TiO2 thickness were systematically investigated. The morphology and thickness of the resulting films were characterized by high-resolution transmission electron microscopy (HRTEM). The energy level alignment was analyzed by X-ray photoelectron spectroscopy (XPS). This work provides a promising strategy to elucidate the role of a TiO2 overlayer and the charge transfer mechanisms in PEC water splitting of hematite, which can be further applied to other nanostructured hematite photoanodes.
The phase composition and structure of β-FeOOH and α-Fe2O3 were characterized by XRD (Fig. S1, ESI†). The diffraction peak at 35.2° of 2θ can be assigned to the (211) reflection of tetragonal FeOOH (JCPDS 34-1266, red line). The diffraction peaks located at 35.9° and 64.5° can be ascribed to the (110) and (300) reflections of α-Fe2O3, respectively (COD 96-901-5504, black line). In addition, the reflections located at 26.6°, 33.9°, 38.0°, 51.8°, 61.9° and 66.0° and marked with asterisks correspond to the (110), (101), (200), (211), (130) and (301) reflections of the cassiterite phase of SnO2, respectively. Both samples show the characteristic signals of SnO2, which come from the FTO substrate.
The morphology and microstructure of β-FeOOH and hematite nanorods were thoroughly investigated by TEM (Fig. S2, ESI†). The typical solid β-FeOOH nanorods are uniform, with diameters ranging from 40 to 80 nm (Fig. S2a and b, ESI†). The main rings in the selected area electron diffraction (SAED) pattern correspond to the (200), (310) and (211) planes for tetragonal FeOOH (Fig. S2c, ESI†). Fig. S2d and e of the ESI† illustrate TEM images of pristine hematite nanorods. The concentric rings in the SAED corresponding to the (102), (104), (110) and (113) planes can be indexed to hematite, indicating the polycrystalline structure of Fe2O3 (Fig. S2f, ESI†).
The morphology and homogeneity of the pristine and TiO2-coated hematite samples were examined by FESEM. Fig. 1b displays that the pristine hematite nanorods are vertically aligned on the FTO. The diameter of the nanorods is around 60 to 80 nm. The SEM images of TiO2-40/Fe2O3 demonstrate a well-preserved hematite morphology (Fig. 1c). However, it is difficult to distinguish the TiO2 film because of its low thickness. The composition of the hematite before and after ALD-TiO2 coating was confirmed by Raman spectra (Fig. 1d). The Raman peaks at 220, 241, 286, 404, 494 and 606 cm−1 are assigned to the hematite structure. The LO peak center at 658 cm−1 is detected, which is attributed to the disorder-induced breaking of Raman symmetry properties.14,35 After TiO2 deposition, an intense Raman peak located at 144 cm−1 is observed, which can be indexed to the Eg mode of anatase TiO2.19 The Raman peaks of the Fe2O3 nanorods are not affected by ALD of TiO2.
The HRTEM images of TiO2-coated Fe2O3 nanorods with a different number of ALD cycles (from 10 to 150 cycles) is shown in Fig. S3, ESI.† At 10 ALD cycles, a continuous TiO2 film coverage cannot be clearly observed due to the small amount of TiO2 coating (Fig. S3a, ESI†). After 20 ALD cycles, Fe2O3 nanorods are homogeneously and uniformly coated with an ultrathin TiO2 film (Fig. S3b–e, ESI,† red dotted lines). The thicknesses of the TiO2 on Fe2O3 nanorods increase with the further increase of the number of ALD cycles. Fig. S3f of ESI† shows the lattice fringes of TiO2-150/Fe2O3 match well with the interplanar distance of the TiO2 anatase structure. The calculated lattice spacing of 0.35 nm can be assigned to the interplanar distance of the (011) plane of TiO2.
The HAADF-STEM images and the corresponding elemental mappings for all the TiO2-coated hematite samples are shown in Fig. 2. Fig. 2a shows small TiO2 nuclei randomly deposited on hematite nanorods after 10 ALD cycles. As the deposition proceeds, small nuclei grow up, and small nanocrystallites coalesce together to form a very thin, continuous TiO2 layer. The HAADF-STEM images of TiO2-20/Fe2O3, TiO2-40/Fe2O3, TiO2-80/Fe2O3 and TiO2-150/Fe2O3 show a conformal and homogeneous TiO2 deposition on the Fe2O3 nanorods, where the phases of TiO2 can be easily identified because of their different Z-contrast. The average thickness of the TiO2 film is calculated as 1.7, 3.5, 6.8, and 13.2 nm for 20, 40, 80 and 150 ALD cycles, respectively (Fig. 2b–e). This indicates that the shell thickness of the TiO2 coatings versus the number of ALD cycles shows good linearity with a slope corresponding to a growth per cycle (GPC) of 0.88 Å. The thickness of the TiO2 overlayer from TEM images is well-controlled and shows the same trend as the ellipsometry data on Si wafers (Table S1, ESI†). The corresponding elemental maps further elucidate a homogenous distribution of Ti and O elements on the Fe2O3 nanorods. The EDX spectra for the pristine Fe2O3 and TiO2-20/Fe2O3, further indicates the presence of both Fe and Ti elements (Fig. S4, ESI†). The atomic/weight percentage of all the samples with different ALD cycles are summarized in Table S2 of ESI.†
![]() | ||
Fig. 2 HAADF-STEM and corresponding EDX elemental maps for (a) TiO2-10/Fe2O3, (b) TiO2-20/Fe2O3, (c) TiO2-40/Fe2O3, (d) TiO2-80/Fe2O3 and (e) TiO2-150/Fe2O3. |
The optical properties of the pristine TiO2 on FTO substrate, pristine hematite and TiO2-coated hematite were measured by UV-vis-NIR diffuse reflectance spectroscopy (Fig. S5, ESI†). As expected, TiO2 shows a high absorbance in the UV region, while Fe2O3 and TiO2/Fe2O3 show a high absorbance in the visible region (Fig. S5a, ESI†). Fig. S5b–d in the ESI† shows the estimated optical band gap, which can be measured from the plot of (F(R)·hν)1/2versus hν.36 The band gaps obtained from Tauc-plots for pristine TiO2, pristine Fe2O3 and TiO2/Fe2O3 are around 3.25, 2.07 and 2.02 eV, respectively, that match well with the reported values.4,10 The slight change of band gap between Fe2O3 and TiO2/Fe2O3 indicates that the doping of Ti in hematite is negligible.
The surface composition and chemical states of the samples were investigated by high-resolution XPS (Fig. S6, ESI†). The survey spectra of pristine TiO2, pristine hematite and hematite coated with 20 TiO2 ALD cycles are shown in Fig. S6a, ESI,† confirming the presence of all expected elements. Fig. S6b of ESI† shows the Fe 2p peaks of the pristine Fe2O3 and TiO2-coated Fe2O3. The iron peaks are still well visible for the TiO2-20/Fe2O3 sample, indicating that the TiO2 layer is sufficiently thin, which is in good agreement with TEM measurements. The two peaks located at binding energies of 711.9 and 725.1 eV can be assigned to Fe 2p3/2 and Fe 2p1/2 of hematite, respectively, indicating the oxidation state III of iron.37,38 XPS spectrum of O 1s peak for hematite samples is centered at 530.2 eV, which is consistent with literature values (Fig. S6c, ESI†).21,39
To determine the energy level offset at the α-Fe2O3/TiO2 heterostructures, the Kraut's method was used, as a direct measurement of the valence band offsets is typically difficult due to overlapping valence band signals.40,41 Assuming the energy difference between the valence band and the core levels is the same for the pristine materials and the heterostructure, the valence band offset in the heterostructure can be calculated with the following equation:
![]() | (1) |
The corresponding core level and valence spectra are shown in Fig. 3a–d.
While the binding energy of the core levels was determined from the peak position after fitting, the valence band onsets were determined from a linear extrapolation of the leading edges. This results in a valence band offset of 0.83 eV, with the valence band of α-Fe2O3 being closer to the Fermi-level. Considering the band gaps determined by absorption spectroscopy (Fig. S5, ESI†), we find that the α-Fe2O3/TiO2 interface forms a straddling gap (type I heterostructure) with a conduction band offset of 0.35 eV (Fig. 3e).
To further study the impact of the TiO2 overlayer on hematite photoanodes on the PEC performance, the measurement of the photocurrent response under visible light at 0.5 V bias vs. Hg/HgO (1.4 V vs. RHE) was carried out on all the samples (Fig. 4b). As mentioned above, Fe2O3 can be excited to produce photogenerated electron–hole pairs under visible light due to the low band gap, while TiO2 shows almost no photocurrent response because of its large band gap. Therefore, the photocurrent response is attributed only to Fe2O3 under visible light. The photocurrent density of pristine hematite is 0.01 mA cm−2. The low photocurrent response is due to the high recombination rate of photogenerated charge carriers. By the deposition of the TiO2 overlayer, all the composite samples display improved photocurrent densities, which can be attributed to the passivation effect of TiO2.18,42 It is well known that the passivation layer is normally very thin (<100 nm), which can prevent parasitic light absorption and charge-transfer inhibition.43 TiO2 passivates the surface states, reduces the recombination rate of the photogenerated electron–hole pairs, facilitates photogenerated hole transfer and improves the photocurrent response.22 The photocurrent densities of TiO2 coated hematite with 10, 20, 40, 80 and 150 ALD cycles are 0.06, 0.27, 0.13, 0.06 and 0.02 mA cm−2, respectively. The highest photocurrent response is recorded for TiO2-20/Fe2O3 (∼1.7 nm), showing a photocurrent density ca. 27 times higher than that obtained for the pristine hematite. At this optimized thickness, the TiO2 overlayer passivates surface recombination states of hematite and the photogenerated holes can be transferred from hematite to electrolyte. With a further increase of the TiO2 thickness, the TiO2 overlayer can hinder the hole transfer efficiency and block light from entering the inner hematite layer, leading to a dramatic decrease in photocurrent response.
During visible light illumination, photogenerated holes in pristine hematite travel to the semiconductor/liquid junction (SCLJ) and accumulate there, owing to the existence of surface trap states. This leads to the appearance of large anodic transient peaks, which decay until they finally reach an equilibrium between charge carrier transfer and recombination at the interface.44 On the other hand, cathodic transient peaks overshoot and decay back to zero under dark conditions, which are attributed to the electrons diffusing from the external circuit and recombining with holes accumulated at the SCLJ.12,22 However, the transient photocurrent spike decreases obviously after the deposition of TiO2. Moreover, the feature was highly repeatable for numerous ALD cycles. This result further demonstrates the passivation effect of the TiO2 overlayer. Therefore, Fe2O3 essentially acts as a light absorption layer, absorbing the visible light and producing photogenerated electron–hole pairs.
EIS was measured under light illumination to study the interfacial kinetics of the charge transfer process. Fig. 4c displays the Nyquist plots of pristine Fe2O3 and TiO2-20/Fe2O3 that are simulated using the equivalent circuit (EC) shown in Fig. 4c inset. The equivalent circuit elements include the series resistance in the PEC cell (Rs), the charge transfer resistance (Rct) at the Fe2O3/electrolyte interface and a constant phase element (CPE1).39,45 The fitting parameters for the equivalent circuit elements are shown in Table 1. Compared to pristine hematite, TiO2-20/Fe2O3 has a lower charge transfer resistance than pristine hematite, which indicates that the charge transfer across hematite/electrolyte interface is significantly facilitated due to the passivation effect of TiO2. The EIS result confirms that TiO2/Fe2O3 heterostructure can significantly enhance the charge transfer and separation and improve the PEC water splitting performance.
R s (Ω cm2) | R ct (Ω cm2) | |
---|---|---|
Pristine-Fe2O3 | 7.5 | 247.3 |
TiO2-20/Fe2O3 | 13.9 | 1807.0 |
The open circuit voltage (OCV) of pristine hematite and hematite coated with 20 TiO2 ALD cycles were measured in the dark and illumination conditions to study the behavior of the photogenerated carriers (Fig. 4d). Under illumination, the value of pristine Fe2O3 moves to a negative potential, while the potential of TiO2-20/Fe2O3 shifts more negative than pristine Fe2O3. This suggests that both samples are n-type semiconductors, in line with the XPS results.46–48 Under illumination, both samples exhibit a similar trend and create a similar number of photogenerated electrons, which can accumulate on the surface of photoanodes, leading to a potential negative shift and a sudden increase in the OCV. Meanwhile, the recombination of photogenerated electrons and holes occurs, which leads to a decrease of the OCV until a relatively steady state is reached. It can be seen that the potential change for Fe2O3 and TiO2-20/Fe2O3 is comparable. However, compared to pristine Fe2O3, the photoanode of TiO2-20/Fe2O3 is more stable due to the passivation effect of TiO2 and the low photogenerated carrier recombination dynamics. After irradiation for 20 seconds, the light is turned off, and the OCV continues to decline due to the high recombination of photoelectron–hole pairs. The photovoltages of the pristine Fe2O3 and TiO2-20/Fe2O3 photoanodes are 14.7 mV and 36.0 mV, respectively, as obtained from the difference between the OCV under dark and light conditions. Compared to pristine Fe2O3, the photovoltage obtained with TiO2-20/Fe2O3 is much higher due to the fact that more photogenerated electrons are collected and a more efficient separation of the photogenerated electron–hole pairs.
To study the photo-response properties further, the incident photon to converted electron (IPCE) and stability test of pristine and TiO2-coated hematite photoanodes were measured under visible light. The IPCE values have been calculated according to the equation:
![]() | (2) |
A maximum IPCE of 1.35% at 430 nm is achieved by TiO2-20/Fe2O3, which is much higher than the corresponding value of 0.05% for pristine Fe2O3 photoanodes (Fig. S7a, ESI†). The significant improvement of photoconversion efficiency indicates higher utilization efficiency of photoinduced carriers for the TiO2-coated hematite photoanodes, which is consistent with the J–V measurements (Fig. 4a).
Applied bias photon-to-current efficiency (ABPE) was calculated from the photocurrent–potential plots according to the following equation:
![]() | (3) |
The ABPE was plotted as a function of the applied potential. The TiO2-20/Fe2O3 photoanode achieves the maximum ABPE value of 0.01% at 1.02 V, which is significantly higher than that of 0.004% at 0.98 V for pristine hematite electrode (Fig. S7b, ESI†). Clearly, the improvement of PEC water splitting performance can be attributed to the enhanced charge separation caused by the passivation effect of TiO2.
Fig. S8 in ESI† shows the current retention as a function of the time of pristine hematite and hematite coated with 20 TiO2 ALD cycles under continuously visible light irradiation for 60 min. A significant improvement of the photocurrents of pristine hematite electrode upon TiO2 deposition as well as a good stability can be obtained. The morphology and the composition of the synthesized photocatalyst after stability test are well maintained, which are shown in Fig. S9 and S10 in ESI.†
Fig. S11 in ESI† shows the Mott–Schottky plots of the pristine Fe2O3 and TiO2-20/Fe2O3. The slopes determined from the Mott–Schottky plots are used to estimate the carrier densities. The positive slopes display the n-type semiconductor nature of the pristine Fe2O3 and TiO2-20/Fe2O3. Moreover, the plot of TiO2-20/Fe2O3 depicts a suppressed slope compared to pristine Fe2O3, suggesting a higher carrier concentration in the heterostructure than in the pristine Fe2O3. The as-calculated carrier densities for pristine Fe2O3 and TiO2-20/Fe2O3 are 1.6 × 1019 cm−3 and 4.4 × 1019 cm−3, respectively.
Contrary to the visible light system, both Fe2O3 and TiO2 participate in the photocurrent response under UV light irradiation due to the suitable band gap of Fe2O3 and TiO2. As a result, the photocurrent response is not only attributed to Fe2O3 but is also related to the TiO2 overlayer. To verify the role of TiO2 overlayer in the PEC performance of hematite, the photocurrent response for hematite and TiO2-coated hematite samples were evaluated under UV light (Fig. S12, ESI†).
The experimental results show that the photocurrent response varies significantly with the thickness of the TiO2 overlayer. Compared to pristine hematite, the TiO2 overlayer results in a remarkable enhancement of photocurrent response. Before 80 ALD cycles, the variation trend of the photocurrent response of hematite samples with TiO2 thickness under UV irradiation was similar to that under visible light irradiation. The photocurrent densities of pristine hematite and TiO2 coated hematite with 10, 20, 40 and 80 ALD cycles are 0.03, 0.07, 0.41, 0.22 and 0.09 mA cm−2, respectively. The optimized TiO2-20/Fe2O3 (ca. 1.7 nm) shows a photocurrent response of 0.41 mA cm−2, which is 14 times higher than the photocurrent response of the pristine hematite. TiO2 thicknesses up to 1.7–6.8 nm (20 ALD cycles to 80 ALD cycles) lead to a decrease in photocurrent response. At this stage, the TiO2 overlayer mainly acts as a passivation layer and Fe2O3 acts as a light absorption layer. However, the photocurrent response increases with a further increase of the thickness of the TiO2 overlayer (i.e., from 6.8 to 13.2 nm, 80 ALD to 150 ALD), in contrast to what was observed under visible light. The photocurrent density of TiO2-150/Fe2O3 is 0.14 mA cm−2. This change of photocurrent response at a higher thickness suggests that the photocurrent response is not dominated by Fe2O3 anymore, but is restricted to the TiO2 overlayer.
Fig. 5 compares the photocurrent response of hematite samples coated with different thicknesses of the TiO2 overlayer under different illumination. The experimental results reveal a substantial variation of the photocurrent response as a function of the thickness of the TiO2 overlayer. The highest photocurrent response exhibited by the TiO2-20/Fe2O3 photoanodes is recorded under both visible and UV light. After 20ALD cycles, the photocurrent response decreases with a further increase of the TiO2 thickness under visible light, while it first decreases and then increases under UV light.
![]() | ||
Fig. 5 Current densities of hematite photoanodes with different ALD TiO2 cycles at 0.5 V bias vs. Hg/HgO, 1 M KOH, under visible light and UV light irradiation, respectively. |
Under visible light, the TiO2 overlayer acts as a passivation layer. The photogenerated holes from Fe2O3 can cross the passivation layer primarily by tunneling, which gradually vanishes within a few nanometers.43,50,51 Because of this mechanism, the TiO2 overlayer must be of the optimal thickness. It should not be too thin to passivate the surface recombination states enough, and not too thick to allow the photogenerated holes transfer between Fe2O3 and electrolyte to participate in the water oxidation. Under UV light, when the TiO2 thickness is ≤6.8 nm, it does not absorb enough UV light, and therefore only few photogenerated charge pairs are produced in the TiO2. In this case, TiO2 mainly acts as a passivation layer, and the photocurrent response is dominated by Fe2O3. Therefore, the photocurrent response under UV light follows the same trend as under visible light. However, when the thickness of TiO2 is larger than the optimal 1.7 nm (20 ALD cycles), it hinders the transfer of the photogenerated holes. Therefore, the photogenerated holes in Fe2O3 cannot efficiently cross the TiO2 overlayer. Conversely, when the TiO2 overlayer is thick enough, it absorbs more UV light and produces more photogenerated charge pairs that participate in water oxidation. The contribution of the TiO2 overlayer for photocurrent response becomes dominant at 150 ALD cycles (13.2 nm). At this stage, TiO2 mainly behaves as the active layer, and the contribution of Fe2O3 becomes negligeable. This demonstrates that the photoresponse under UV illumination is largely related to the thickness of the TiO2 overlayer.
![]() | ||
Fig. 6 Schematic illustration of the charge-transfer pathways in the TiO2/Fe2O3 system under visible light and UV light. |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3se01633a |
This journal is © The Royal Society of Chemistry 2024 |