III. RESULTS AND DISCUSSION

Nanostructured iron oxide films were deposited by plasma-assisted chemical vapor deposition of iron penta-carbonyl [Fe(CO)₅] used as source for iron in oxygen plasma to produce homogeneous coatings of amorphous iron oxides. Scanning electron micrographs (Fig. 1) recorded to comparatively analyze the effect of post treatments (oxygen bombardment and rapid annealing treatment) on morphology evolution showed significant changes in the pristine and oxygen plasma-treated films. An iron oxide film synthesized by PECVD exhibits an island structure consisting of ∼100 nm sized primary particles agglomerated, which results in a porous structure [see Fig. 1(a)]. When the oxygen plasma is applied on this island structured film, there is remarkable reduction of spacing between agglomerates accompanying particle coarsening. Microscopic analyses showed that the additional oxygen plasma treatment promoted crystallization and growth of hematite particles [Figs. 1(a)–1(d)] possibly driven by the tendency to reduce the grain boundaries. However, upon rapid annealing, no significant change was observed in the morphology, presumably due to the short annealing steps that did not allow any pronounced bulk diffusion and film densification.

FIG. 1.

Upper row: Hematite films treated with oxygen plasma for (a) 0 min, (b) 10 min, (c) 20 min, and (d) 30 min, Lower row: Hematite films that were shortly annealed (at 750 °C for 10 min) on the oxygen plasma-treated samples for (e) 0 min, (f) 10 min, (g) 20 min, and (h) 30 min. Scale bars are 500 nm.

Nevertheless, annealing of oxygen plasma-treated samples showed an increased surface roughness and coarsening of grains indicating that the bombardment of oxygen species enhanced the coalescence of smaller particles to form larger ones, whereby the grain growth is controlled by the energetics associated with the grain boundaries, free energy of the interface and the film–substrate interface (see Table I and Fig. 2). The pristine film treated with oxygen plasma for 10 min followed by short annealing shows the highest roughness value of 20.708 nm (R _q, root-mean-square) and 16.390 nm (R _a, average) that gradually decreased with the increasing oxygen plasma exposure time. This surface roughening is attributed to agglomeration of particles induced by easy diffusion of oxygen in the grain boundaries to improve the Fe:O stoichiometry in favor of hematite (Fe₂O₃) as the plasma-CVD grown films are generally oxygen-deficient. In addition, oxygen plasma treatments remove any surface contaminants and change the nature of functional groups into the O-rich surface due to the reactive nature of oxygen species present in the oxygen plasma. Consequently, the asorbed oxygen species preferentially migrate inwardly upon the heat treatment manifested in the flattening of the surface and abnormal grain growth (Table I).

TABLE I.

Quantified surface roughness of the pristine and the treated hematite films.

FIG. 2.

AFM images and diagonal line profiles of the prepared samples; (a) pristine hematite, oxygen plasma treated for (b) 10 min, (c) 20 min, and (d) 30 min followed by short annealing. (e) Schematic representation of the changes in morphological features as a function of post-synthesis treatments.

Evaluation of the crystallographic data provided insight in the crystallization of iron oxide films as a consequence of the post-synthesis treatments. The X-ray diffractogram of the pristine sample exhibited a pure hematite phase (PDF 87-1164) with corresponding diffraction peaks that could be indexed as (012), (104), (110), and (024) planes (Fig. 3). In the oxygen plasma-treated films the signal due to the (024) plane was found to be dimished indicating the influence of oxygen diffusion on crystallographic orientation of the grains. However, all samples exhibited XRD patterns corresponding to the hematite phase without the coexistence of any other iron oxides, which indicates that post treatments procedure did not influence the bulk crystallinity as evident in the comparable intensities and broadness of the signals (Fig. 3).

FIG. 3.

XRD patterns of the pristine and the post-treated hematite (20 min O₂ plasma followed by the SA) photoanodes.

The J–V curves of the pristine and post-treated hematite photoanodes (Fig. 4) revealed that onset potentials of the hematite photoanodes treated with oxygen plasma exhibited a significant anodic shift. The photoanode treated with oxygen plasma for 10 min exhibited the strongest anodic shift of 360 mV compared to the pristine sample and the lowest photocurrent density of 0.035 mA/cm² at 1.23 V versus RHE. However, the following short annealing steps recovered the onset potentials and increased photocurrent density when compared to the pristine hematite photoanode. For instance, the highest photocurrent density of 1.306 mA/cm² at 1.23 V versus RHE was obtained from the sample treated with oxygen plasma for 20 min followed by short annealing [see Fig. 4(b)]. Moreover, J–V curves of the pristine hematite photoanode and the short-annealed hematite without preceding oxygen plasma treatment are comparable [Fig. 4(c)]; the photocurrent density of the short-annealed sample was lower in the whole range of applied potentials, and its onset potential was almost the same as that of the pristine photoanode. This observation suggested that the oxygen plasma treatment prior to rapid annealing is crucial to fabricate efficient hematite photoanodes. The values of onset potential and the photocurrent density at 1.23 V versus RHE are listed and depicted in the insets of Fig. 4 and in Fig. 4(d), respectively.

FIG. 4.

J–V curves of the prepared hematite photoanodes (a) only treated with O₂ plasma, (b) treated with the oxygen plasma followed by short annealing (SA); (c) comparison between a pristine and short-annealed samples and (d) a graphical demonstration of onset potential and photocurrent density change at 1.23 V versus RHE based on insets in (a) and (b).

As pointed out by Grätzel et al. ^{Reference Sivula, Le Formal and Grätzel28} and other related reports, the onset potential of hematite photoanode can be improved by chemical surface modification. ^{Reference Zandi and Hamann29–Reference Iandolo, Wickman, Zoric and Hellman31} We observed that the physical treatment of hematite films in oxygen plasma changed the surface states and chemical environment around Fe centers that was supported by XPS studies. The Fe 2p spectrum (Fig. 5) of pristine hematite showed two distinct peaks, Fe 2p _3/2 at 710.95 eV and Fe 2p _1/2 at 724.40 eV with a shakeup satellite located at 718.65 eV, which is a typical feature of Fe₂O₃. ^{Reference Grosvenor, Kobe, Biesinger and McIntyre32,Reference Yamashita and Hayes33} After the oxygen plasma treatment, the shakeup satellite peak which represents the oxidation state of Fe³⁺ became stronger indicating higher compositional uniformity. ^{Reference Yamashita and Hayes33} The XPS studies suggested that the surface of the pristine hematite is intrinsically oxygen deficient (Fe₂O_3−x ) and the oxygen plasma saturates the surface defects via preferential adsorption of oxygen species. Moreover, the signals of Fe 2p _1/2 (724.40 eV) and Fe 2p _3/2 (710.95 eV) in the pristine sample showed a shift toward lower binding energies (724.05 and 710.15 eV) possibly due to additional electronic charges and changed chemical environment around Fe centers, which presumably resulted in higher shielding of core electrons. After further short annealing step, the binding energies of the Fe 2p core levels recovered to 724.30 eV (Fe 2p _1/2) and 710.65 eV (Fe 2p _3/2), which is often the evidence for formation of oxygen vacancies. ^{Reference Zhu, Li, Zheng and Delaunay34} A similar trend in binding energy shift was observed in the O 1s peak at ∼530 eV (Fe–O), along with a shoulder peak at ∼531 eV generally accepted as a contribution of surface adsorbed oxygen species (e.g., O_ads, O⁻, OH⁻). ^{Reference Wandelt35} These surface oxygen species were still evident after the short annealing step, while the main O 1s peak was found to shift back to a higher binding energy. Furthermore, the shift in O 1s peak from 530.10 to 529.45 eV upon oxygen plasma treatment is in good agreement with a previously reported value for lattice oxygen (O²⁻). ^{Reference Hu, Boudoire, Hermann-Geppert, Bogdanoff, Tsekouras, Mun, Fortunato, Graetzel and Braun36} The peak area of the O 1s signals at ∼530 eV significantly increased after the oxygen plasma treatment (from 965.0 to 1194.4), confirming an increased amount of surface-attached Fe–O groups, whereas it decreases upon the short annealing from 1194.4 to 896, indicating either the loss of oxygen or inward migration of oxygen [Fig. 5(e)].

FIG. 5.

Comparisons of XPS spectra between the pristine and the post-treated hematite (20 min O₂ plasma followed by the SA); (a) spectra of Fe 2p peaks, (b, c, and d) spectra of O 1s peaks of the pristine, O₂ plasma treated and O₂ plasma followed by SA treated samples, respectively, and (e) a proposed scheme of hematite surface upon the post-treatments.

We presume that the hematite surface (Fe₂O_3−x ) that is intrinsically oxygen deficient becomes saturated with an excess of oxygen species (O_ads, O⁻, OH⁻) upon oxygen plasma treatment, leading to an increased concentration of Fe³⁺ species [Fig. 5(a)] and an increased area of the shoulder peak at ∼531 eV [compare Figs. 5(b) and 5(c)]. This oxygen-saturated surface with filled up oxygen vacancies impedes water oxidation by reducing surface occupancy of iron centers that was revealed as a faster water oxidation route than oxygen surface states. ^{Reference Hu, Boudoire, Hermann-Geppert, Bogdanoff, Tsekouras, Mun, Fortunato, Graetzel and Braun36} During the thermal treatment, the saturated hematite surface with an excess amount of oxygen species can be reconstructed, inducing an optimized amount of oxygen vacancies for an efficient oxygen evolution reaction (OER). ^{Reference Zhang, Klaver, van Santen, van de Sanden and Bieberle-Hütter37} It corresponds to the earlier studies dealing with the role of oxygen vacancies ^{Reference Zhu, Li, Zheng and Delaunay34–Reference Zhang, Klaver, van Santen, van de Sanden and Bieberle-Hütter37} and oxygen deficient hematite as reported by Cowan et al. ^{Reference Bora, Braun, Erat, Löhnert, Ariffin, Manzke, Sivula, Graule, Grätzel and Constable38} Based on the surface analyses, we propose the surface stoichiometry of crystalline hematite as Fe₂O_3−x for the pristine samples, Fe₂O_3−y for the oxygen plasma treated, and Fe₂O_3−z for the oxygen plasma followed by short annealing sample, where y < x < z.

To verify the proposed oxygen-deficient surface structure of pristine hematite and to investigate the transfer direction of photogenerated electrons upon illumination with monochromatic incident light, (in situ) X-ray absorption spectroscopy (XAS) was performed. In contrast to XPS, XAS probes unoccupied states using soft X-ray synchrotron radiation (<5000 eV). Thereby, core electrons of an atom absorb X-ray photons and get promoted to higher-lying unoccupied states forming short-lived excited states. Moreover, K-edge XAS is an effective tool to approximate the electronic and structural properties affecting OER catalysis. ^{Reference Forster, Potter, Ling, Yang, Klug, Li and Cowan39} In hematite, the electronic transitions from O 1s to hybridized O 2p-Fe 3d states and from Fe 2p to Fe 3d states can be verified by the O K-edge (photon energies of 520–570 eV) and the Fe L_2,3-edge (700–730 eV) spectra, respectively. Since the intensity of the signals in the XAS is directly proportional to the amount of unoccupied states, changes in the electronic structure triggered, e.g., by doping or the formation of composite materials can be unambiguously detected. ^{Reference van Oversteeg, Doan, de Groot and Cuk40,Reference Shen, Zhou, Dong, Hu, Tseng, Guo, Guo and Mao41} Fu et al., for instance, showed that Ta-doping of α-Fe₂O₃ nanorods films did not affect the Fe 3d but the O 2p site. Upon doping, the signal corresponding to the O 2p-Fe 3d t_2g state decreased in intensity which implies that electrons get transferred to this oxygen site. This has been reported to lower the conduction band and thus to enhance the electron conductivity. ^{Reference Suntivich, Hong, Lee, Rondinelli, Yang, Goodenough, Dabrowski, Freeland and Shao-Horn42} As XAS exhibits very high sensitivity toward the chemical environment of an atom, this method was used herein as a sensitive local structure probe to clarify the surface structure of the pristine hematite sample. Both the Fe L-edge and O K-edge spectra were recorded in surface sensitive total electron yield. The oxidation state of iron in pristine hematite films was first investigated by Fe L-edge spectra [Fig. 6(b)]. Due to Fe 2p spin–orbit coupling the spectra are divided into 2 parts, the Fe L₃ and L₂ edges, located at ∼710 and ∼723 eV, which have been assigned to Fe 2p _3/2 and Fe 2p _1/2 transitions to an unoccupied 3d orbital, respectively. ^{Reference Fu, Dong, Zhou, Lee, Chen, Guo, Zhao and Shen43} Both of these edges are further split into two peaks with intensity ratios that are sensitive toward the oxidation state of iron, resulting from an interplay of the crystal field and electronic interactions. ^{Reference Fu, Dong, Zhou, Lee, Chen, Guo, Zhao and Shen43,Reference Crocombette, Pollak, Jollet, Thromat and Gautier-Soyer44} In a simplified model, the splitting was solely ascribed to crystal field interactions with the signals at the L₃ edge corresponding to t_2g and e_g states. As displayed in Fig. 6(b), the e_g states exhibit a higher intensity than the t_2g states. This is in agreement with literature reported spectra for α-Fe₂O₃. ^{Reference van Oversteeg, Doan, de Groot and Cuk40,Reference Suntivich, Hong, Lee, Rondinelli, Yang, Goodenough, Dabrowski, Freeland and Shao-Horn42} In contrast, the L₃ edge of Fe²⁺ shows a reversed intensity ratio with the main peak located at ∼707 eV and a shoulder toward higher energy (∼709 eV). ^{Reference Laan and Kirkman45} The oxygen K-edge spectrum [Fig. 6(a)] shows four distinct components at around 528, 539, 547, and 560 eV. As a result of the crystal field effect, the signal at around 527–532 eV is further split into two peaks of equal intensity corresponding to the lower energy 3d t_2g and higher energy 3d e_g states reflecting the octahedral environment of Fe³⁺ in α-Fe₂O₃. ^{Reference Suntivich, Hong, Lee, Rondinelli, Yang, Goodenough, Dabrowski, Freeland and Shao-Horn42,Reference Jiménez-Villacorta, Prieto, Huttel, Telling and van der Laan46} The broad absorption feature at around 532–544 eV is attributed to O 2p-Fe (4s, 4p) hybridized states, and the high energy structure at about 550–570 eV has been reported to result from resonance scattering. ^{Reference Shen, Zhou, Dong, Hu, Tseng, Guo, Guo and Mao41,Reference Pollak, Gautier, Thromat, Gota, Mackrodt and Saunders47} The signal at 545–549 eV, however, shows a higher intensity than reported for α-Fe₂O₃. ^{Reference Pollak, Gautier, Thromat, Gota, Mackrodt and Saunders47} This has been likewise reported for α-Fe₂O₃ single crystals, which were first subjected to a reductive treatment under ultra-high vacuum at 950 °C and afterward reoxidized via an additional heating step under a fixed oxygen pressure of 10⁻⁶ mbar for 90 min. ^{Reference Jiménez-Villacorta, Prieto, Huttel, Telling and van der Laan46,Reference Pollak, Gautier, Thromat, Gota, Mackrodt and Saunders47} Thus, this feature could derive from the oxygen deficient Fe₂O_3−x surface as a result of the distortion of octahedra by oxygen vacancies. However, the origin of the change in the fine structure of the XAS spectrum requires further investigation. It was reported that the intensity ratio of the O 2p-Fe 3d signal relative to the broad O 2p-Fe (4s, 4p) signal directly reflects the extent of covalent bonding in α-Fe₂O₃ and that an increased O 2p-Fe 3d hybridization results in accelerated surface oxidation reaction kinetics. ^{Reference Shen, Zhou, Dong, Hu, Tseng, Guo, Guo and Mao41} Thus, along with the results gained from XPS analysis and PEC measurements, oxygen plasma treatment reduced the degree of O 2p-Fe 3d hybridization resulting in an anodic shift of the onset potential.

FIG. 6.

In situ XAS measurements of pristine hematite (a) at the oxygen K-edge and (b) at the iron L-edge measured in the dark and under illuminated conditions. The insets show the enlarged spectra in (a) the O 2p-Fe 3d and (b) the Fe 2p _3/2-Fe 3d region.

In case of in situ XAS the sample is additionally illuminated by monochromatic incident light (350–750 nm) which permits to elucidate which orbitals are photocatalytically active as well as the transfer of photogenerated electrons in composite materials. Herein, both the spectra at the Fe L-edge and the O K-edge were recorded in the dark (black graphs) and under illumination (blue, red graphs). The insets in Fig. 6 show the enlarged spectra in (a) the O 2p-Fe 3d and (b) the Fe 2p _3/2-Fe 3d region. No spectral differences (both at the O K-edge and the Fe L-edge) were observed in the dark and under light illumination. This can be ascribed to a rapid electron–hole recombination as a result of the intrinsically oxygen-deficient Fe₂O_3−x structure.

Submillisecond to second timescale TAS analysis was used to elaborate the charge carrier dynamics in detail and to further corroborate PEC performances influenced by surface chemical states of the hematite photoelectrodes. The transient absorption spectrum of hematite in this timescale is composed of two distinct spectral features. The most pronounced spectral feature is a narrow negative transient absorption band (bleach) centered at 570–580 nm that forms in the nanosecond to microsecond timescale, ^{Reference Ruoko, Kaunisto, Bärtsch, Pohjola, Hiltunen, Niederberger, Tkachenko and Lemmetyinen48} which was observed originally by Durrant et al., ^{Reference Barroso, Pendlebury, Cowan and Durrant49,Reference Pendlebury, Wang, Le Formal, Cornuz, Kafizas, Tilley, Gratzel and Durrant50} later by Lemmetyinen et al., ^{Reference Ruoko, Kaunisto, Bärtsch, Pohjola, Hiltunen, Niederberger, Tkachenko and Lemmetyinen48} and assigned to electron trapping into localized energy states near the conduction band edge with a lifetime in the millisecond timescale. The second spectral feature is a broad positive absorption band peaking at 650 nm that has a lifetime in the hundreds of milliseconds timescale. ^{Reference Ruoko, Kaunisto, Bärtsch, Pohjola, Hiltunen, Niederberger, Tkachenko and Lemmetyinen48} This feature is only visible under anodic bias and the transient absorption amplitude of the feature correlates extremely well with the photocurrent density of the photoanode. ^{Reference Barroso, Pendlebury, Cowan and Durrant49,Reference Pendlebury, Cowan, Barroso, Sivula, Ye, Gratzel, Klug, Tang and Durrant51} Since the lifetime of this feature is reduced drastically when hole scavengers are added to the electrolyte, it is reasonable to assign the feature to long-lived photogenerated holes in the depletion region. ^{Reference Forster, Potter, Ling, Yang, Klug, Li and Cowan39,Reference Ruoko, Kaunisto, Bärtsch, Pohjola, Hiltunen, Niederberger, Tkachenko and Lemmetyinen48} The transient absorption decays at 575 and 650 nm of pristine, 20 min oxygen plasma-treated, and 20 min oxygen plasma-treated and short-annealed samples under 1.50 V versus RHE bias voltage are presented in Fig. 7 along with the respective biphasic fits. The absorption spectra (Fig. S2) and transient absorption decays at bias voltages of 0.90 and 1.23 V versus RHE (Figs. S3 and S4) along with the complete fit results (Tables II and III) can be found in the Supplementary Material.

FIG. 7.

Transient absorption decays of the samples under 1.50 V versus RHE bias, (a) at 575 nm and (b) at 650 nm.

TABLE II.

Stretched exponential lifetimes and the associated stretching factors obtained from the biphasic fits to raw data.

TABLE III.

Mean lifetimes calculated from the lifetimes and stretching factors along with the raw and relative amplitudes of the decay components at 650 nm from the fits.

The function used for fitting the TAS decays herein is of the form $\Delta A = {A_1}\exp \left[ {{{\left( {{{ - t} / {{\tau _1}}}} \right)}^{{\beta _1}}}} \right] + {A_2}\exp \left[ {{{\left( {{{ - t} / {{\tau _2}}}} \right)}^{{\beta _2}}}} \right]$ , where A _i are the amplitudes, τ _i are the lifetimes, and β _i are the stretching factors of the stretched exponential decay components. The stretched exponential function is now used to describe relaxation in a disordered system, and both the lifetimes and stretching factors are reported in Table II. The stretching factor was limited to values between 0 and 1, with 1 giving the usual exponential function. The obtained lifetimes and stretching factors were used to calculate the mean lifetimes using ${\tau _i} = {{{\tau _i}} \over {{\beta _i}}}\Gamma \left( {{1 \over {{\beta _i}}}} \right)$ , where τ _i is the mean lifetime and Γ is the gamma function. The mean lifetimes are presented in Table III along with the absolute and relative amplitudes of the decay components at 650 nm.

The most noticeable feature of the decays is that the transient absorption amplitude at 100 μs increases almost 7-fold at 575 nm (−2.6 × 10 μOD for pristine versus −1.73 × 10² μOD for O₂ plasma) and 2-fold at 650 nm (5.9 × 10 μOD for pristine versus 1.13 × 10² μOD for O₂ plasma) after 20 min of oxygen plasma treatment (see Fig. 7). This is a striking difference when taking into account that the same excitation density was used for all samples and the absorption of the pristine sample is the highest at the excitation wave length (Fig. S2). After the short annealing, the transient absorption amplitude at 575 nm recovers slightly to 4-fold that of the pristine sample (−1.03 × 10² μOD), whereas the amplitude at 650 nm recovers completely to the same level of the pristine sample (5.7 × 10 μOD). The increased transient absorption amplitude at 575 nm indicates a greatly increased amount of electron trap states after the oxygen plasma treatment, whereas at 650 nm the increase in amplitude suggests an increase in the yield of long-lived photoholes. ^{Reference Forster, Potter, Ling, Yang, Klug, Li and Cowan39} The smaller amplitude of the bleach and the full recovery of the photohole yield for the short-annealed sample indicate a greatly reduced number of electron trap states when compared with the oxygen plasma sample.

The electron traps are presumed to originate from the plasma treatment of the surface, possibly by generation of oxygen vacancies or the formation of trapping oxygen states on the surface of the photoanode. Furthermore, increased electron trapping has been suggested to be a pathway that competes with recombination, ^{Reference Ruoko, Kaunisto, Bärtsch, Pohjola, Hiltunen, Niederberger, Tkachenko and Lemmetyinen48,Reference Pendlebury, Wang, Le Formal, Cornuz, Kafizas, Tilley, Gratzel and Durrant50} thus explaining the increased photohole density as probed at 650 nm. The bleach signal is clearly visible for the oxygen plasma-treated sample even at 0.90 V versus RHE (Fig. S3), 0.18 V below the photocurrent onset voltage, suggesting that the increased electron trapping is due to newly formed surface states instead of the previously reported intraband trap states within the depletion region. These trap states seem to also act as efficient recombination centers, as indicated by the fast recombination component of the photoholes, being much more pronounced for the oxygen plasma-treated sample (see Table II). The increased amount of recombination centers explains the drastically higher photocurrent onset voltage of the oxygen plasma treated sample, whereas the onset voltage is the same for the pristine and short-annealed samples [see Fig. 4(c)]. This indicates that even though the newly formed electron trap states suppress the sub-microsecond recombination, the millisecond timescale recombination with surface photoholes is increased. This negatively affects the OER, leading to an increased photocurrent onset voltage.

At 1.50 V versus RHE the water oxidation reaction dominates the long-lived decay of the photohole signal at both 575 and 650 nm. Both pristine and oxygen plasma-treated samples have approximately the same water oxidation rate of 2.0–2.5 s⁻¹ (mean lifetimes 510 and 390 ms for pristine and O₂ plasma samples, respectively, see Table II). The long-lived hole decay at 1.23 V remains unchanged for the pristine sample (mean lifetime 470 ms), whereas the lifetime is three times lower for the oxygen plasma-treated sample (mean lifetime 140 ms). The drastic drop in the long-lived hole lifetime is due to the increased back electron hole recombination, indicating that the electron trap states formed by the oxygen plasma treatment also increase long-lived recombination in addition to increasing the millisecond timescale recombination. However, the water oxidation rate of 7.1 s⁻¹ for the short-annealed sample is in the faster end of the rates previously reported for hematite photoanodes (mean lifetime 140 ms). ^{Reference Forster, Potter, Ling, Yang, Klug, Li and Cowan39,Reference Ruoko, Kaunisto, Bärtsch, Pohjola, Hiltunen, Niederberger, Tkachenko and Lemmetyinen48} Since the yield of long-lived holes for the short-annealed sample is approximately the same as for the pristine sample (48 and 45 μOD, respectively), the much faster water oxidation dynamics indicates increased the catalytic activity after oxygen plasma treatment and short annealing. Interestingly, the long-lived hole lifetime remains approximately the same at 1.23 and 0.90 V (see Table II) with only the amplitude of the long-lived component being lower. This indicates that the increased catalytic activity of the photoanode treated with oxygen plasma followed by short annealing is due to a reduced amount of back electron recombination at all measured bias voltages, supporting enhanced PEC performance.

We have additionally conducted investigations on the effect of argon (Ar) plasma onto the PEC activity in comparison to the oxygen plasma treated samples. Figure 8 shows J–V curves of hematite photoanodes post-treated with Ar plasma and followed by short annealing. Significant anodic shifts of onset potentials were observed for the Ar plasma-treated samples (green curves), which recovered upon further short annealing regardless of the duration of the plasma treatment, which was already observed in Fig. 4. It seems that the anodic shift of the onset potential resulted from plasma bombardment regardless of the type of plasma forming gas. It is surprising that the chemical states at the hematite surface do not seem to be affected by the Ar plasma. High resolution XPS spectra of the Fe 2p states given in Fig. S5 support this speculation as there is no evolution of a shakeup satellite peak at ∼718 eV, which proves the non-oxidizing nature of Ar plasma in contrast to the oxidizing nature of oxygen plasma. Thus, the chemical states at the hematite surface are highly dependent on the type of plasma. The increased amount of electron trapping states upon oxygen plasma treatment detected at 575 nm [see Fig. 7(a)] is ascribed to surface defects induced by a physically disturbed surface that can be efficient recombination sites. However, the mechanism of anodic shift in the onset potential observed upon Ar plasma needs to be further investigated carefully since chemical reduction of surface yielding recombination centers such as interstitial Fe ions could occur, resulting in anodic onset shift. In addition, the recovery of the onset potential upon further short annealing indicates that the defective surface structure becomes relaxed or reconstructed by the thermal treatment. However, in case of Ar plasma treatment, no superior PEC performance is observed upon the short annealing step. This implies that the intentionally saturated surface with oxygen species contributes critically to the enhancement in PEC performance of hematite. Further investigations on the hematite surface were carried out by water contact angle (WCA) measurements on prepared films. It is well known that the surface oxygen provides hydrophilicity to various surfaces. ^{Reference Khan, Azimi, Yildiz and Varanasi52} In the WCA measurement, the pristine hematite film exhibited an average WCA of ∼25°, whereas a complete wetting (0°) of the water droplet occurred on the oxygen plasma treated hematite surface. After the additional short annealing step, the WCA value was determined to be ∼20°. These results strongly evidence the incorporation of surface oxygen species upon oxygen plasma treatment and the removal of these species upon the short annealing step. On the other hand, the WCA of the Ar plasma treated sample was ∼17°, which is slightly lower than the WCA value of pristine hematite. However, after short annealing, the hematite surface showed complete wetting, which was likewise observed for the oxygen plasma treated film. It is likely either due to decreased amount of carbon contents on the surface upon Ar plasma treatment (from 41.20 to 30.74 at.%) or due to roughened surface, resulted from plasma etching and then, atmospheric oxygen species adsorbed during the short annealing would induce hydrophilic nature. ^{Reference Li, Wang, Kozbial, Shenoy, Zhou, McGinley, Ireland, Morganstein, Kunkel, Surwade, Li and Liu53} To gain a deeper insight on the activation energy needed for oxygen incorporation at surface and subsurface levels, future studies will deal with the modeling of oxygen deficient hematite systems for PEC water splitting.

FIG. 8.

J–V curves of (blue) the pristine hematite, (green) treated only with the Ar plasma, and (blue) treated with the Ar plasma followed by the SA; The Ar plasma for (a) 10 min, (b) 20 min, and (c) 30 min.