Understanding the role of phosphate in the photoelectrochemical performance of cobalt-phosphate/hematite electrode systems

Abstract

The complex of cobalt-phosphate (CoPi) is known to be an efficient catalyst that can greatly enhance the photoelectrochemical (PEC) performance of hematite electrodes. However, the complicated role that associated Pi plays in the CoPi catalyst is not yet fully understood. In this study, we noted that the photocurrent density–voltage curves between Co and CoPi associated hematite electrodes are rather different, particularly in the tailings and transient spikes. This means that the reduction in the recombination loss by the associated Pi could probably result from its high electronegativity and tendency to withdraw photoexcited electrons from hematite. The results from additional surface complexation modeling and FTIR analyses further indicate that the conformation of the associated CoPi complexes also directly affect the efficiency of the withdrawal. Interactions between Co and the neighboring Pi would, on the other hand, induce the development of the CoPi catalyst to fine particles or continuous CoPi layers, which would indirectly influence their PEC performance due to the size effect. Based on our results, associated Pi in CoPi catalysts mainly influences PEC performance by withdrawing photoexcited electrons and reducing the size of CoPi catalysts.

---

1. Introduction

Raising the photoelectrochemical (PEC) efficiency of hematite (α-Fe₂O₃) remains a challenging goal. To achieve this goal, the short hole diffusion length (2–4 nm), low hole mobility (∼0.01 cm² V⁻¹ s⁻¹), slow kinetics of the oxygen evolution reaction, and fast carrier recombination of hematite have been intensively investigated.¹ For instance, reducing the hematite structure to a low dimensional configuration was demonstrated to effectively mitigate the loss of carriers transporting from hematite to the semiconductor liquid junction (SCLJ). This can be realized due to the nature of the anisotropic conductivity of hematite, such that the resistance within the (001) plane is 4 orders of magnitude lower than the plane parallel to [001].^2,3 Introducing additional mobile charge carriers by doping external elements is known to greatly increase the conductivity of hematite.^4,5 Associating surface catalysts, such as iridium complexes, were proven to significantly enhance the kinetics of oxygen evolution by bending the conduction band near the SCLJ close to the water oxidation potential.^6,7 Increasing the internal oxygen vacancy by a short period of annealing in an oxygen deficient environment was also demonstrated to greatly enhance the PEC performance of hematite electrodes.⁸ This is because the formation of oxygen vacancies would simultaneously inject additional electrons to neighboring Fe³⁺, which remarkably increases the carrier density by approximately two orders of magnitude.⁹

In addition to iridium complexes, cobalt-based catalysts, including cobalt oxides (Co₃O₄) and cobalt phosphate (CoPi) complexes, have drawn the intensive attention of material scientists due to their capability of enhancing PEC performance. For example, a 67% photocurrent enhancement was reported when 5% Co₃O₄ was in situ associated with hematite nanorods during a hydrothermal reaction, which is explained by the high surface roughness, larger Co₃O₄/hematite interfacial area and smaller Co₃O₄ particle size.¹⁰ The association of CoPi complexes was further noted to considerably enhance the photocurrent and oxygen evolution rate by fivefold under the conditions of an additional + 1.0 V vs. RHE applied bias. The authors attributed this remarkable PEC performance enhancement to the improvement in the kinetics of water oxidation by the associated CoPi complexes.¹¹ Photodeposited CoPi complexes were found to increase the photocurrent by facilitating electron–hole pair separation and by improving oxygen evolution kinetics.¹² The authors suggested that the enhanced photocurrent was due to associated CoPi complexes efficiently collecting and storing photogenerated holes from hematite and therefore reducing the surface state recombination that leads to the increased water oxidation efficiency.¹³ This argument is evidentially supported by transient absorption spectroscopy and intensity modulated photocurrent spectroscopy studies, which indicate that the surface CoPi layer could result in the accumulation of charge carriers at the SCLJ due to the reduction in recombination loss.^14,15 The conflicting experimental observations regarding the role of CoPi thickness have recently been explained to be the result of the fundamental differences between structured and planar CoPi/hematite electrodes.¹⁶ However, the role that Pi plays in the CoPi complex is not fully understood. Accordingly, attempts were made to elucidate this matter, particularly focusing on exploring the role of Pi in the PEC performance of the CoPi/hematite system. By comparing the photocurrent density–voltage curves between the Co and CoPi associated hematite electrodes, along with additional surface complexation modeling and FTIR analyses, the unique role that associated Pi plays was comprehensively evaluated and discussed in this study.

2. Experimental section

2.1. Preparation of Co and CoPi associated hematite electrodes

All chemicals used in this study were ACS grade purchased from Sigma-Aldrich and Merck and used as-received without any further purification. The hematite electrodes (2 × 1 cm FTO substrates) were obtained by hydrothermal reaction at 120 °C, 4 hours in the solution containing 0.1 M Fe(III) and 1.0 M NaNO₃ at pH 1.2. Association of Co/CoPi complexes were conducted by immersing hematite electrodes in the solution containing different concentrations of Co(II) (Co associated samples) and 1 mM Co(II) with various Pi concentrations (CoPi associated samples) for 12 hours. In the end of immersion, the samples were left air-dried followed by a two-step heating treatment (500 °C 1 hour than 750 °C 10 minutes) with a heating rate of 10 °C min⁻¹. Six samples were prepared for each condition and the one exhibited the photocurrent density lying closest to the average value (the relative standard deviation, RSD, was about 15%) was chosen for further characterization.

2.2. PEC measurements

For PEC performance measurement, hematite electrodes were first covered by nonconductive Teflon bands to allow a working area (1 × 1 cm) to expose to the illumination of simulated sunlight (AM 1.5 G with the light power density of 100 mW cm⁻²). All PEC measurements were conducted using CHI 608E electrochemical workstation in a three-electrode configuration (hematite as the working electrode, a 1 × 1 cm platinum plate as the counter electrode and a saturated calomel electrode (SCE) as the reference electrode) in the electrolyte containing 1.0 M NaOH (pH 13.6). The measured potential was swept from −0.5 V to 1.0 V vs. SCE with a scan rate of 50 mV s⁻¹. To determine the flat band potential and the charge carrier density of these hematite samples, Mott–Schottky plots were derived from the electrochemical impedance obtained in the potential between −0.5–0.7 V vs. SCE with 1000 Hz frequency in the dark and the capacitance was extracted according to eqn (1):¹⁷

Z′′ = 1/2πfC (1)

where the Z′′ stands for the imaginary impedance, f is the scan frequency and C refers to the capacitance of measured sample. Electrochemical impedance spectroscopy (EIS) measurement was carried out under illumination with +0.1 V applied voltage by the same workstation in the frequency range of 1 Hz to 100 kHz. Obtained data were fitted by an equivalent circuit using the Zview software.

2.3. Characterization

Scanning electron microscopy (SEM) images of hematite nanostructures were obtained using a JSM700F, JEOL scanning electron microscope with an acceleration voltage of 20 kV. Given the difficulty in precisely determining the surface area of hematite on FTO substrates, the distributions of Co/CoPi complexes on the surface of hematite electrodes were simulated by adopting the Co/CoPi adsorption to hematite particle as the alternative analog. The chemical equilibrium modeling system MINEQL+ 4.6 version was used.¹⁸ Several assumptions were made in advance to keep the number of adjustable parameters as low as possible (and to avoid introducing undefined properties, such as charge density, on the surface of hematite). First, the surface adsorption site density was assumed to be 7.5 sites per nm.¹⁹ Because the Co/CoPi adsorption experiments were conducted with a solid/liquid ratio of 1.0 g of hematite/100 mL of solution, the corresponding adsorption site concentration in the system was 0.71 mmol L⁻¹ (with 8.47 m² g⁻¹ as the surface area of hematite particles as determined by the N₂-BET method (77 K liquid nitrogen) using Micromeritics ASAP 2020). Second, as the equilibrium constant is a state function, we adopted the log K_a1 as 6.9 for the protonation reaction (eqn (2)) and log K_a2 as −12.8 for the deprotonation reaction of hematite (eqn (3)) in accordance with our preliminary study.²⁰

SOH + H⁺ ↔ SOH₂⁺ (2)

SOH ↔ SO⁻ + H⁺ (3)

where S refers to the adsorption sites on the surface of hematite.

3. Results and discussion

3.1. Photoelectrochemical behavior of the Co₃O₄–hematite system

Although the exposed surface area of the hematite film on the FTO substrate is difficult to precisely determine, the alternative is to simulate the distribution of Co species on the surface of hematite particles. The stoichiometric reaction between the surface of hematite particles and Co²⁺ (SOH + Co²⁺ ↔ SOCo⁺ + H⁺) was considered.²⁰ As shown in Fig. 1(a) (with 0.1 mM initial Co²⁺), the cobalt adsorption edge appears at pH 4 and no Co(OH)₂ is found in the system. When the initial Co concentration is further increased to 1.0 mM (Fig. 1(b)), the appearance of the Co adsorption edge remains at pH 4, while the formation of Co(OH)₂ becomes significant (∼40%) when the pH is increased to 8. This means that there would be a maximum of 60% adsorption sites on the surface of hematite to be bound with Co²⁺ because Co(OH)₂ is neutrally charged and would not be attracted by an electrostatic interaction. Similarly, increasing the initial Co²⁺ concentration to 10 mM (Fig. 1(c)) would lead >90% of the solution Co²⁺ to precipitate as Co(OH)₂. This implies that the surfaces of the hematite electrodes prepared in this condition would be likely covered with randomly deposited Co(OH)₂ precipitates. These randomly deposited Co(OH)₂ precipitates would be thermally converted to Co₃O₄ catalysts in the subsequent sintering process.²⁰

Fig. 1 Simulations of the distribution of Co species at various pH values with initial Co²⁺ concentrations of (a) 0.1 mM, (b) 1 mM, (c) 10 mM; morphologies of hematite electrodes prepared at pH 8 with initial Co²⁺ concentrations of (d) 0.1 mM, (e) 1 mM, (f) 10 mM; and those prepared at pH 11 with initial Co²⁺ concentrations of (g) 0.1 mM, (h) 1 mM, (i) 10 mM; (j) proposed mechanism of the Co²⁺ association to the hematite surface as a function of pH.

Based on our simulations, we note that the initial Co²⁺ concentration and solution pH have essential influences on the distribution of associated Co catalysts over the surface of hematite electrodes. As shown in Fig. 1(d)–(i), increasing the initial Co²⁺ concentration would gradually lead to the formation of regions of bulk Co₃O₄ on the surface of the hematite electrodes. That is, associated Co₃O₄ catalysts from the solution containing an initial concentration of 0.1 mM Co²⁺ at pH 8 and pH 11 are barely observed in the SEM images (Fig. 1(d) and (g), respectively). This is because the association of Co₃O₄ catalysts was carried out through stoichiometric adsorption, as indicated in Fig. 1(a). In both cases, an ultrathin Co₃O₄ catalyst layer would appear (<5 nm) on the surface of the hematite electrodes,²⁰ and this explains why the surface morphologies in Fig. 1(d) and (g) are rather similar. Conversely, when hematite electrodes are prepared in a solution containing a vast amount of Co(OH)₂ precipitates (Fig. 1(e)), randomly deposited Co(OH)₂ clusters would be converted to a discontinuous Co₃O₄ layer in the subsequent sintering process, as shown in Fig. 1(f) and (i), respectively. Based on these SEM images, a proposed mechanism accounting for the distribution of Co₃O₄ catalysts over the hematite surface is shown in Fig. 1(j). Specifically, in an acidic environment, a very limited amount of surface Co₃O₄ catalyst is expected to be associated with the surface of hematite due to the electrostatic repulsion between Co²⁺ and hematite. On the other hand, although the surface of hematite is negatively charged in an alkaline environment (pH 11, for instance), which favors an accommodating solution of Co²⁺, the formation of Co(OH)₂ precipitates would result in a few Co₃O₄ catalyst regions being associated on the surface of the hematite electrodes. As fine Co₃O₄ catalysts (<5 nm) are known to exhibit higher reactivity than large ones,^10,11 it is therefore expected that the PEC performance of these hematite electrodes would be significantly affected by the distribution and size of the surface Co₃O₄ catalysts.

The influence of the distribution and size of surface Co₃O₄ catalysts on PEC performance is shown in Fig. 2(a)–(c). Because no Co²⁺ adsorption is expected in an acidic environment, as demonstrated by our simulations (Fig. 1(a)), the associations were only conducted at pH 8.2 and pH 11. It is noted that the photocurrent curve of a pristine hematite electrode exhibits a saw-like appearance (measured in the chop mode), while its photocurrent density is approximately 0.02 mA cm⁻² at 0.4 V vs. SCE (pH 8.2 and pH 11). By contrast, the association of surface Co₃O₄ catalysts increases the photocurrent density to 0.10 mA cm⁻² at the identical condition, which is approximately five times higher than that without surface Co₃O₄ catalysts. When further raising the initial Co concentration to 1 mM and 10 mM, the photocurrent density decreased to 0.04 and 0.02 mA cm⁻², respectively, which is only slightly higher than the values for pristine hematite (Fig. 2(c)). When the initial cobalt concentration is 0.1 mM, associated Co₃O₄ catalysts would be uniformly distributed on the surface of hematite (Fig. 1(d) for pH 8.2 and Fig. 1(g) for pH 11). By contrast, once the initial cobalt concentration is increased to 1 and 10 mM, Co(OH)₂ colloids would form in the electrolyte, and they are susceptible to the drag of gravity that limits their access to the hematite electrodes. This explains the SEM image observations that showed that there were only a few Co₃O₄ catalysts associated on the hematite surface and that most of them are represented by less reactive Co₃O₄ clusters (Fig. 1(d)–(i)). Therefore, low PEC performances are observed.

Fig. 2 The photocurrent of hematite electrodes obtained in various Co²⁺ concentration solutions at (a) pH 8 and (b) pH 11; (c) photocurrent density of all of the examined hematite electrodes at 0.4 V vs. SCE; (d) representation of a selected equivalent circuit; the extracted resistance (e) and capacitance (f) of each component in these Co-associated hematite samples.

In addition to the enhanced PEC performance, surface Co₃O₄ catalysts change the shape of the photocurrent by reducing the tailing along with inducing the appearance of transient spikes (indicated by arrows in Fig. 2(a) and (b)). The observed photocurrent tailing when the light is chopped off is likely attributed to the fact that the photoexcited carriers continue to participate in water oxidation, while the appearance of a transient spike implies that a severe recombination loss of charge carriers occurs in hematite.^12–14 Because the determined carrier densities among these samples are rather comparable (they remain in the same order of magnitude and thus the data are not shown), the observed higher photocurrent must be strongly related to the combination of reduction in recombination loss and improved water oxidation. As suggested, surface Co₃O₄ catalysts would bend the conduction band near the semiconductor liquid junction close to the water oxidation potential and thus improve water oxidation.^6,7 More charge carriers are able to transport to the SCLJ and participate in water oxidation before their recombination, and this explains the observed higher photocurrent and disappearance of the tailing behavior. On the other hand, once the water oxidation is improved, the recombination loss in the hematite due to its low conductivity becomes a significant issue of concern, and this accounts for the appearance of the transient spike shown in Fig. 2(a) and (b). These observations therefore lead to the conclusion that the enhanced PEC performance of cobalt associated hematite is due to the improved water oxidation kinetics.^14,21

To understand the effect of the Co₃O₄ catalyst distribution on photocatalytic reactivity, electrochemical impedance spectroscopy was adopted. The as obtained EIS results were further fitted using a simplified equivalent circuit, as shown in Fig. 2(d), where CPE₁ and R₁ were the capacitance and resistance of hematite; CPE₂ and R₂ refer to the capacitance and resistance of the hematite/cobalt oxide interface; and CPE₃ and R₃ were assigned and describe the capacitance and resistance at the cobalt oxide/electrolyte interface. As shown in Fig. 2(e), it is noted that the surface Co₃O₄ catalysts would reduce the resistances of all components, but the resistances increase with the increasing Co concentration. By contrast, C₁ and C₂ are rather constant and insensitive to the increasing Co concentration, while C₃ obviously drops significantly with the increasing Co concentration. Decreased resistance can be realized by considering the improved kinetics of water oxidation. That is, increased water oxidation kinetics reduce the recombination loss, as reflected in the resistance decrease. This argument is supported by the comparable C₁ and C₂ values shown in Fig. 2(f). Another interesting feature noted in Fig. 2(e) is the increasing resistance along with the increasing Co concentration. As indicated by our simulations and SEM images (Fig. 1), the association of Co₃O₄ catalysts in a solution containing a high Co concentration would lead to randomly distributed Co₃O₄ clusters. As Co₃O₄ clusters are less reactive than their complementary fine particles due to the size effect,¹⁰ it is expected that slower water oxidation in the former would result in the accumulation of charge carriers at the hematite/Co₃O₄ interface, which induces the increment in the R₃ value. Identical consideration also explains the decrement in the capacitance at the Co₃O₄/electrolyte interface (C₃), where the slow water oxidation due to the size effect limits the transport of photogenerated holes toward water molecules.

3.2. The orientation of CoPi catalysts on the surface of the hematite electrodes

In the above section, the effect of the preparation conditions (initial Co²⁺ concentration and pH) on the PEC performance of Co₃O₄ associated hematite electrodes is discussed, which leads to a conclusion that is consistent with other reports, that the enhanced PEC performance of Co₃O₄ associated hematite is mainly due to the improved water oxidation kinetics.^10,14 In this section, the role of phosphate (Pi) on the PEC performance of CoPi catalysts associated with the hematite system is examined. Prior to the discussion, adsorption experiments were conducted to investigate the interactions between Co²⁺, Pi, and hematite in this ternary system. The results were then numerically fitted using surface complexation modeling to determine the corresponding equilibrium constants. To maintain the consistency of the obtained equilibrium constants, three stoichiometric reactions, including phosphate adsorption to hematite and the concurrent adsorption of Co²⁺ and phosphate to hematite (the ternary adsorption system), were taken into consideration, while the identical reaction constant (the one determined in Fig. 1) was adopted for the Co²⁺ adsorption to hematite reaction.

As shown in Fig. 3(a), it is noted that the Co²⁺ adsorption edge is unaffected by the presence of phosphate in the ternary system. By contrast, the phosphate adsorption edge shifts to high pH in the ternary system and the ratio of phosphate adsorption to hematite only slightly drops from 100% to 90%, even at pH 12. This implies that in the alkaline environment surface, adsorbed Co²⁺ could either be the bridge to complex negatively charged phosphate or can somehow screen the negatively charged hematite surfaces. The best fittings (solid lines) using surface complexation modeling are shown in Fig. 3(a), where the determined reaction constants are listed in the caption of Fig. 3. Based on the modeling, both the distribution of adsorbed species and the mechanism of how the CoPi catalysts were associated on the surface of the hematite electrodes can be quantitatively interpreted. In the ternary system containing both 0.1 mM of Co²⁺ and phosphate, 15% of surface sites would be occupied by Co²⁺ (SOCo⁺) and another 15% by phosphate (SOH₂PO₄²⁻) on the surface of hematite when the pH > pH 4 (Fig. 3(b)). Importantly, approximately 3% of surface sites would be occupied by CoPi complexes (SOHPO₄Co) at pH 3 to pH 5. In the case of the solution containing 0.1 mM Co²⁺ and 1.0 mM phosphate (Fig. 3(c)), the distribution of surface sites is significantly different. When pH < pH 8, Co²⁺ associated to hematite mainly occurs through the CoPi complexes (∼15% SOHPO₄Co), while Pi occupies another 28% of the surface sites of hematite. While in an alkaline environment (pH > pH 10), they both occupied approximately 15% of the surface sites of hematite. At the same time, the amount of surface SOHPO₄Co species significantly drops from 14% to 0% when the pH is increased from pH 8 to pH 10. Likewise, in the ternary system containing 0.1 mM Co²⁺ and 10 mM phosphate, it is noted that there are approximately 45% of surface adsorption sites occupied by phosphate (∼30% SOH₂PO₄²⁻) and CoPi complexes (∼15% SOHPO₄Co), respectively, when the electrolyte pH < pH 8. This means that further increasing the phosphate concentration would not influence the Co²⁺ adsorption to hematite. Similarly, when the electrolyte pH > pH 10, the surface of hematite accommodates ∼20% of the SOH₂PO₄²⁻ and ∼15% of the SOCo⁺. This means that Co²⁺ and phosphate are associated to the hematite surface separately when no CoPi complexes are present. Based on our simulation, two important features are noted from the point of view of thermodynamics. First, there would only be approximately 15% of the surface sites that are thermodynamically favorable for the association of CoPi catalysts. Second, the orientation of associated CoPi catalysts is very sensitive to the pH of the preparation environment.

Fig. 3 (a) Adsorption of Co/Pi in the 0.1 mM Co, Pi and the binary (both 0.1 mM Co and Pi) systems; the distribution of Co, Pi and CoPi complexes on the surface of hematite with initial 0.1 mM Co and (b) 0.1 mM; (c) 1.0 mM; and (d) 10 mM Pi; the best fitting curves shown in (a) are based on the stoichiometric reaction of SOH + Co²⁺ ↔ SOCo⁺ + H⁺, pK_Co = 3.2; SOH + PO₄³⁻ + H⁺ ↔ SOH₂PO₄²⁻, pK_Pi = 20.6; and SOH + Co²⁺ + PO₄³⁻ + H⁺ ↔ SOHPO₄Co, pK_CoPi = 30.5.

To examine the validity of the proposed stoichiometric reaction, a parallel set of experiments was conducted by immersing hydrothermally grown hematite electrodes in the solution containing 10 mM Pi (control) and 10 mM Pi along with 0.1 mM Co²⁺. The as-obtained wet samples were then characterized using ATR-FTIR to evaluate the orientation of the associated CoPi catalysts on the surface of the hematite electrodes (Fig. 4). The assignment of the IR-active bands of surface phosphate (800–1200 cm⁻¹) is based on the conformation proposed by Arai and Sparks.²² For the control sample (Pi only, Fig. 4(a), black line) obtained at pH 4, three bands centered approximately at 1150, 1065 and 970 cm⁻¹ are noted, indicating that the adsorbed Pi has C_2v symmetry. In the ternary system with an identical pH environment (Fig. 4(a) red line), the intensities of 1150 and 970 cm⁻¹ absorbance significantly decrease, while the absorbance at 1065 cm⁻¹ shifts to 1050 cm⁻¹. This shift might be attributed to the resonance of the protonated CoPi complex, as illustrated in Fig. 4(d). Similarly, the FTIR spectrum of the sample collected at pH 8 also exhibits three distinct absorptions, but the absorption at 1065 cm⁻¹ is significantly shifted to 1025 cm⁻¹ (Fig. 4(b)). The cause of this shift is unclear; however, the concurrent appearance of three absorptions in the region from 1200 to 800 cm⁻¹ suggests that adsorbed phosphate likely remains in the C_2v symmetry. With the presence of Co²⁺, he absorption at 950 cm⁻¹ becomes blurry and the absorptions at 1150 and 1050 cm⁻¹ seem to be overlapped to become a broadened absorption in the range of 1000 to 1200 cm⁻¹. Similar results have been reported, showing that adsorbed Cu²⁺ and Cd²⁺ would interact with adsorbed phosphate in the vicinity on the surface of the hematite. This interaction would slightly alter the symmetry of the adsorbed CoPi complexes.²³ In this case, the symmetry of adsorbed phosphate might exist as a combination of both C_2v and C_3v symmetries, and we speculate that a similar interaction might also occur between the adsorbed Co²⁺ and Pi in this study. Based on our simulations (Fig. 3(d)), however, both the Co²⁺ and Pi are concurrently adsorbed on the hematite surface via the CoPi complexes. Accordingly, the origin of the emerging C_3v symmetry is speculated to stem from the interaction between the adsorbed CoPi complexes in the vicinity. Unlike in an acidic environment, where the adsorbed phosphate is protonated, deprotonation of bridged phosphate might cause an interaction of the complexed Co²⁺ in the vicinity and thus slightly affect the symmetry of CoPi complexes to a certain degree. Undoubtedly, this speculation requires more sophisticated spectroscopic analysis, such as EXAFS, to support it. At pH 10, the obtained FTIR spectra are distinctly different from those observed at pH 7 and pH 4 (Fig. 4(c)). As evidenced by the appearance of absorptions at 1077 and 1001 cm⁻¹, adsorbed phosphate is likely to be in C_3v symmetry. As no significant shift in absorption is noted when Co²⁺ is present, it is apparent that the symmetry of adsorbed phosphate is not affected by the adsorbed Co²⁺ in the vicinity. Based on our SCM simulations and the results from FTIR analysis, the possible orientations of adsorbed Co and Pi are given in Fig. 4(d).

Fig. 4 FTIR spectra of wet hematite paste samples obtained with 0.1 mM Co and 1.0 mM Pi at (a) pH 4; (b) pH 8; and (c) pH 11; the corresponding configuration of the associated Co/Pi species. (d) Representative conformation of Pi and Co associated on the surface of hematite based on FTIR analyses.

Fig. 5 shows the surface morphology evolution of CoPi-associated hematite electrodes obtained from the solution containing 0.1 mM Co²⁺ and varying Pi concentrations in different pH environments. It is noted that the surfaces of CoPi-associated hematite electrodes are rather different, indicating that the pH and phosphate concentrations have a significant influence on the surface morphology of the CoPi catalysts. Generally at a given pH, increasing the Pi concentration would lead surface CoPi catalysts to gradually transforming from globular nodule clusters to a porous structured film-like nodule appearance. This can be understood by considering that more phosphate would be adsorbed on the surface of hematite along with the increasing phosphate concentration. Accordingly, more adsorbed phosphate is able to participate in the formation of a P–O–P–O–P network in the following thermal treatment. Therefore, a film-like CoPi layer is observed. By contrast, when the initial phosphate concentration is fixed, raising the pH would induce the development of surface CoPi catalysts into finer and elongated grains. This is because the interaction between Co and Pi (Fig. 4(d)) would interrupt the formation of the P–O–P–O–P network in the thermal treatment process. As a result, finer CoPi catalysts appear on the surface of the hematite electrodes.

Fig. 5 Surface morphology evolution of hematite samples obtained in the system with 0.1 mM Co and different Pi concentrations (0.1; 1.0; and 10 mM Pi) in different pH environments.

Fig. 6(a)–(c) exhibit the photocurrent density of CoPi-associated hematite electrodes as a function of the applied potential for hematite electrodes prepared at different initial phosphate concentrations and pH values. The photocurrent densities of these samples at 0.4 V vs. SCE are recorded and arranged in Fig. 6(d). From Fig. 6(d), it is noted that, in general, increasing the Pi concentration would gradually enhance the photocurrent density. Without the associated phosphate, the photocurrent density of the sample Fe–Co-0.1 at pH 4 is approximately 0.02 mA cm⁻², which is identical to the photocurrent density of the bare hematite electrode. This is because the electric repulsion hinders the association of Co²⁺ to the surface of hematite, as we mentioned above. When increasing solution pH to pH 8 and pH 11, the surface of hematite bears neutral and negative charges, respectively. This allows Co₃O₄ to associate with the hematite electrode, and therefore, a comparable photocurrent density is observed in the sample Fe–Co-0.1 at pH 8 and Fe–Co-0.1 at pH 11. On the other hand, the sample Fe–CoPi-0.1 at pH 4 exhibits a photocurrent density of 0.11 mA cm⁻². This value is approximately five times higher than that of the sample Fe–Co-0.1 at pH 4 and is very comparable to that of the Co₃O₄-associated hematite electrode (the sample Fe–Co-0.1 at pH 8 and Fe–Co-0.1 at pH 11). Based on our simulations (Fig. 3(b)), this can be explained by the fact that the adsorbed Pi somehow screens the electrostatic repulsion between the hematite surface and Co²⁺ and therefore allows 7% of the surface sites to associate the Co₃O₄ catalysts along with approximately 3% of the CoPi complexes. Furthermore, because there are 15% of surface sites adsorbed by Co²⁺ at pH 8 and pH 11 in this system, the amount of associated Co₃O₄ catalysts on the surface of the hematite electrodes would be similar. This explains why the sample Fe–CoPi-0.1 at pH 8 and Fe–CoPi-0.1 at pH 11 exhibit very similar photocurrent densities.

Fig. 6 Photocurrent of CoPi-associated hematite samples obtained at (a) pH 4; (b) pH 8; and (c) pH 11; (d) photocurrent densities of all of the examined CoPi-associated hematite electrodes at 0.4 V vs. SCE; and the extracted capacitance (e) and (f) resistance values of each component in these CoPi-associated hematite samples.

In the case in which the hematite electrodes were immersed in a solution containing 0.1 mM Co²⁺ along with a high concentration of phosphate (1.0 and 10 mM Pi, sample Fe–CoPi-1 and FeCoPi-10 series), sample Fe–CoPi-1 at pH 4 and Fe–CoPi-10 at pH 4 exhibit photocurrent densities of approximately 0.13 and 0.16 mA cm⁻², respectively. This PEC performance is very similar to that of the Co₃O₄ associated hematite electrode. On the other hand, Fe–CoPi-1 at pH 8 and Fe–CoPi-10 at pH 8 are noted to exhibit the photocurrent densities approximately 0.30 and 0.33 mA cm⁻², respectively, which is approximately 3 times higher than that of Fe–CoPi-0.1. Prepared at pH 11, enhanced photocurrent densities up to 0.26 mA cm⁻² (Fe–CoPi-1 at pH 11) and 0.16 mA cm⁻² (Fe–CoPi-10 at pH 11) are also noted. Based on Fig. 6(d), it is clear that CoPi-associated hematite electrodes exhibit a higher PEC performance than the Co₃O₄-associated ones. Generally, samples prepared at a neutral pH exhibit the highest photocurrent, while those obtained in acidic environments show the lowest PEC performances. A likely explanation would be to consider the electronegativity between P (2.19) and Fe (1.83).²⁴ It has been reported that the associated phosphate would facilitate the withdrawal of photogenerated electrons from hematite due to its larger electronegativity.²⁵ By trapping photogenerated electrons at the associated phosphate, the recombination loss is greatly reduced and the photocurrent density is thus enhanced. This speculation is supported by the absence of transient spikes in the photocurrent of CoPi-associated hematite electrodes, as shown in Fig. 6(a)–(c).

While the electronegativity can account for the higher photocurrent observed in the sample of the CoPi-associated hematite electrodes compared to the Co₃O₄-associated ones, it is not sufficient to explain the observation that the photocurrent is very sensitive to the pH. For example, as evidenced by our simulation (Fig. 3(c)), the fraction of surface CoPi (SOHPO₄Co) is always kept at 15% along with another 25% SOH₂PO₄ when pH < pH 8. However, the photocurrent of samples prepared at pH 8 (sample Fe–CoPi-1 at pH 8) is approximately 3 times higher than those prepared at pH 4 (sample Fe–CoPi-1 at pH 3). Additionally, Fe–CoPi-1 at pH 8 has a comparable photocurrent to that of Fe–CoPi-10 at pH 8, although they both have the same amount of surface SOHPO₄Co. Accordingly, we speculate that, in addition to the electronegativity, the different distribution and texture of the surface CoPi layer on the hematite surface might be responsible for the different PEC performance values. A recent study provided by the Gamelin group has noted the importance of distribution as well as the texture of the surface CoPi layer on the PEC performance of hematite photoanodes.¹⁶ In their study, there is an optimal CoPi thickness (<10 nm) for mesostructured (porous) hematite electrodes, as a thick CoPi layer suffers enhanced electron–hole recombination inside the CoPi layer, known as a kinetic bottleneck.¹¹ This is attributed to the fact that in a mesostructured hematite electrode, photogenerated electrons must travel parallel to the SCLJ, and it is known that the conductivity of hematite is four orders of magnitude higher at the (001) plane than parallel to the (001) plane.²⁶ Comparing the samples prepared at pH 4 with those at pH 8, it is found that the distribution of surface CoPi catalysts becomes distanced, as shown in Fig. 5(b) and (e), despite the fraction of surface CoPi (SOHPO₄Co) always remaining approximately 15% along with another 25% SOH₂PO₄ when the phosphate concentration is greater than 1.0 mM whether at pH 4 or pH 8 (Fig. 3(c) and (d)). We thus speculate that the interactions between the deprotonated P–O– group and Co at SOHPO₄Co at pH 8 might be responsible for the observed distanced CoPi catalysts. This is because the dehydration during thermal treatment would lead to the formation of a successive phosphate network through P–O–P bonding. Interactions between the P–O group and Co at SOHPO₄Co suppress the formation of the phosphate network (P–O–P–O–P) and thus allow the CoPi catalyst to develop into finer particles. Because sparsely distributed CoPi catalysts would reduce the interfacial recombination between conduction-band electrons in hematite and holes within the CoPi layer by sweeping electrons away from the SCLJ, a higher photocurrent is observed in Fe–CoPi-1 at pH 8 and Fe–CoPi-10 at pH 8 than in Fe–CoPi-1 at pH 4 and Fe–CoPi-10 at pH 4.¹¹ The same feature also explains the higher photocurrent observed in samples prepared at a 10 mM Pi concentration, as their surface CoPi catalysts are finer in comparison with those obtained in the 1.0 mM Pi solution. When samples were prepared in alkaline environments (pH 11), surface Co₃O₄ catalysts were severely aggregated, with phosphate in the vicinity at the end of the thermal treatment (Fig. 3(d) and 5(e) and (i)). In this way, unbound P–O bonding would interact with phosphate in the vicinity, leading to the formation of a phosphate network. This explains the observed film-like CoPi catalysts over the hematite surface at the end of the thermal treatment. In this configuration (separated Co and Pi), trapping photogenerated electrons by P is expected to be less efficient than the configuration where CoPi is homogeneously distributed (i.e., sample Fe–CoPi-10 at pH 8). This explains why the photocurrent of samples prepared at pH 11 is observed to be lower than those prepared at pH 8.

To explore the interfacial behaviors of these hematite photoanodes, we collected their EIS spectra under illumination and fitted them with the identical equivalent circuit as mentioned above in Fig. 2(e) and (f). The obtained resistance and capacitance are presented in Fig. 6(e) and (f). A close inspection suggests that the resistance R₁ is relatively unaffected by the surface CoPi catalysts, while the R₂ (square series) gradually increases with the increasing Pi concentration, and R₃ (triangle series) shows a significant opposite trend relative to R₂ (Fig. 6(f)). On the other hand, while C₁ and C₂ are rather constant and unaffected by the increasing Pi concentration at all of the studied pH values (in other words, unaffected by the surface CoPi catalysts), the C₃ values (triangle series, Fig. 6(e)) obviously increase with the increasing Pi concentration in all of the studied pH environments. As expected, consistent R₁ indicates that the resistance of hematite is independent of the surface CoPi catalysts, as no external dopants are introduced that result from the association of CoPi catalysts. The increment of R₂ follows the order of pH 11 > pH 4 > pH 8, which is consistent with the change in the size of CoPi catalysts as shown in the top view SEM image (Fig. 5). This correlation leads to speculation that the interface resistant at hematite/CoPi (R₂) is also affected by the distribution of the CoPi catalysts. Given that its counterpart C₂ is noted to be independent of the Pi concentration, it is likely that the increased R₂ resistance stems from the charge transferring from hematite toward CoPi catalysts due to the anisotropic conductivity of hematite. Importantly, the decreasing R₃ together with the increasing C₃ along with the increasing Pi concentration for samples prepared at pH 8 substantially indicates the charging of surface states under illumination, and this explains the observed higher photocurrent in the sample Fe–CoPi-10 at pH 8.¹²

A recent transient absorption study revealed that CoPi catalyst plays an important role by band bending at the hematite/electrolyte interface, and this feature extends the lifetime of photogenerated holes by suppressing the loss of recombination.¹³ In this way, photogenerated holes are stored in the Co–Pi catalyst from the hematite electrode under illumination, which is evidenced by the charged surface state capacitance values extracted from the EIS spectrum.¹² Our results support this observation, as indicated by the increasing C₃ values (Fig. 6(e)), and further imply that the charging of the surface state would be strongly influenced by the distribution of surface CoPi catalysts. Furthermore, we speculate that the Co and Pi seem to play different roles in the photocatalytic reaction, where the former is likely to be responsible for water oxidation, while the latter appears to facilitate the withdrawal of photogenerated electrons away from the hematite, as demonstrated in Fig. 7. This explains why the hematite samples associated with Pi alone exhibit no increase in photocurrent density compared to the pure hematite one (data not shown). On the other hand, this synergistic effect would be significant when the CoPi catalysts are associated in an aligned fashion (the CoPi complexes, sample Fe–CoPi-1 at pH 8) rather than separately (sample Fe–CoPi-1 pH at 11). We speculate that this might be attributed to the loss of charge carriers being transported between CoPi grains. While the associated Pi would help to withdraw photogenerated electrons, it seems that the size of the CoPi catalysts is another determinant factor that is responsible for PEC performance enhancement (a higher photocurrent is found in the sample Fe–CoPi-1 at pH 8 than in the sample Fe–CoPi-1 at pH 4). In addition, the size of CoPi catalysts is also directly controlled by the interaction between Co and the associated Pi in the vicinity, which disrupts the formation of a continuous phosphate network. This interaction leads to the development of fine CoPi catalysts instead of film-like CoPi ones over the surface of hematite and explains the higher photocurrent observed in the former sample (sample Fe–CoPi-1 at pH 8). These observations clearly suggest the unique role of Pi in the CoPi catalysts that would simultaneously directly and indirectly enhance the PEC performance of hematite electrodes.

Fig. 7 The proposed roles of Pi in the CoPi catalysts on the PEC performance of hematite electrodes. Inserted scale bars refer to 0.1 mA cm⁻².

4. Conclusions

In this study, attempts were made to isolate the role of phosphate (Pi) in CoPi catalysts on the surface of hematite electrodes. Comparing the PEC performance between the Co-associated hematite samples and CoPi-associated ones, it is speculated that the suppression of recombination loss by CoPi catalysts is likely through the withdrawal of photogenerated electrons from hematite to the associated Pi. Additional surface complexation modeling and FTIR analyses suggest that the conformation of associated CoPi complexes would directly affect the efficiency of the withdrawal. In addition, the interaction between Co and neighboring Pi would lead to the CoPi catalyst developing into fine particles or continuous CoPi layers, which would also pronouncedly influence their PEC performance values. These observations indicate that Co and Pi play different roles in the PEC performance of CoPi-associated hematite electrodes.

Acknowledgements

This study is supported by Ministry of Science and Technology, Taiwan (MOST 104-2119-M-007-015). THW is grateful for the financial support by Ministry of Science and Technology, Taiwan (MOST 104-2811-M-007-097).

References

1. K. Sivula, F. L. Formal and M. Grätzel, ChemSusChem, 2011, 4, 432–449.
2. F. J. Morin, Phys. Rev., 1954, 93, 1195–1199.
3. A. Kleiman-Shwarsctein, Y.-S. Hu, A. J. Forman, G. D. Stucky and E. W. McFarland, J. Phys. Chem. C, 2008, 112, 15900–15907.
4. C. H. Miao, T. F. Shi, G. P. Xu, S. L. Ji and C. H. Ye, ACS Appl. Mater. Interfaces, 2013, 5, 1310–1316.
5. S. C. Riha, M. J. D. Vermeer, M. J. Pellin, J. T. Hupp and A. B. F. Martinson, ACS Appl. Mater. Interfaces, 2013, 5, 360–367.
6. L. Badia-Bou, E. Mas-Marza, P. Rodenas, E. M. Barea, F. Fabregat-Santiago, S. Gimenez, E. Peris and J. Bisquert, J. Phys. Chem. C, 2013, 117, 3826–3833.
7. J. R. Swierk, N. S. McCool, T. P. Saunders, G. D. Barber, M. E. Strayer, N. M. Vargas-Barbosa and T. E. Mallouk, J. Phys. Chem. C, 2014, 118, 17046–17053.
8. T. Y. Yang, H. Y. Kang, U. Sim, Y. J. Lee, J. H. Lee, B. J. Koo, K. T. Nam and Y. C. Joo, Phys. Chem. Chem. Phys., 2013, 15, 2117–2124.
9. J. J. Deng, X. X. Lv, J. Gao, A. W. Pu, M. Li, X. H. Sun and J. Zhong, Energy Environ. Sci., 2013, 6, 1965–1970.
10. L. F. Xi, P. D. Tran, S. Y. Chiam, P. S. Bassi, W. F. Mak, H. K. Mulmudi, S. K. Batabyal, J. S. C. Barber, L. H. Loo and L. H. J. Wong, J. Phys. Chem. C, 2012, 116, 13884–13889.
11. D. K. Zhong and D. R. Gamelin, J. Am. Chem. Soc., 2010, 132, 4202–4207.
12. K. J. McDonald and K. S. Choi, Chem. Mater., 2011, 23, 1686–1693.
13. B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert and T. W. Hamann, J. Am. Chem. Soc., 2012, 134, 16693–16700.
14. M. Barroso, A. J. Cowan, S. R. Pendlebury, M. Gratzel, D. R. Klug and J. R. Durrant, J. Am. Chem. Soc., 2011, 133, 14868–14871.
15. L. M. Peter, K. G. Upul Wijayantha and A. A. Tahir, Faraday Discuss., 2012, 155, 309–322.
16. G. M. Carroll, D. K. Zhong and D. R. Gamelin, Energy Environ. Sci., 2015, 8, 577–584.
17. A. W. Bott, Electrochemistry of Semiconductors, Curr. Sep., 1998, 17, 87–91.
18. W. D. Schecher and D. C. McAvoy, MINEQL+: A Chemical Equilibrium Modeling System, version 4.6, Environmental Research Software, Hallowell, ME, 2007.
19. I. A. Mudunkotuwa and V. H. Grassian, J. Am. Chem. Soc., 2010, 132, 14986–14994.
20. T. H. Wang, H. T. Hung, W. Wang, P. C. Li, Y. K. Hsieh and C. F. Wang, RSC Adv., 2015, 5, 67700–67705.
21. T. Hisatomi, F. Le Formal, M. Cornuz, J. Brillet, N. Tetreault, K. Sivula and M. Gratzel, Energy Environ. Sci., 2011, 4, 2512–2515.
22. Y. Arai and D. L. Sparks, J. Colloid Interface Sci., 2001, 241, 317–326.
23. W. Li, S. Z. Zhang, W. Jiang and X. Q. Shan, Chemosphere, 2006, 63, 1235–1241.
24. P. Atkins and J. D. Paula, Physical Chemistry, W. H. Freeman and Company, New York, 8th edn, 2006.
25. L. Fu, H. M. Yua, C. K. Zhang, Z. G. Shao and B. L. Yi, Electrochim. Acta, 2014, 136, 363–369.
26. M. Marelli, A. Naldoni, A. Minguzzi, M. Allieta, T. Virgili, G. Scavia, S. Recchia, R. Psaro and V. Dal Santo, ACS Appl. Mater. Interfaces, 2014, 6, 11997–12004.

---