Interfacial Effects on the Band Edges of Ta3N5 Photoanodes in an Aqueous Environment: A Theoretical View.

Ta3N5, as a fascinating photoanode for solar hydrogen production, is expected to split water without any bias, because its band edge potentials straddle H2O redox potentials. Unfortunately, Ta3N5 photoanodes can split water only when a bias of at least 0.6-0.9 V is applied. It means that they exhibit an onset potential as high as 0.6-0.9 VRHE (reversible hydrogen electrode). In this study, density functional theory calculations show that the band edge potentials of Ta3N5 have a shift of approximately -0.42 eV relative to vacuum level when exposed to water. The increased ratio of dissociated water at Ta3N5-water interface will further make the band edge potentials shift -0.85 eV relative to vacuum level, implying the anodic shifts of the onset potential for water oxidation. The findings may reveal the mystery of the unexpectedly high onset potential of Ta3N5, as high as 0.6-0.9 VRHE.

Introduction

Because the available solar energy is inexhaustible and pollution-free, energy harvested directly from sunlight is a prospective way to solve the problems arising from fossil fuel consumption. Among all the utilization of solar energy conversion and storage, photoelectrochemical (PEC) water splitting, which mimics natural photosynthesis, is very attractive, because PEC-based tandem cells for water splitting exhibit theoretical solar-to-hydrogen efficiency of 29.7% (Hu et al., 2013, Chen et al., 2018). In general, suitable band gaps and band edge potentials are the primary condition when designing the PEC tandem cell system for solar water splitting. Ta3N5 with band gap of 2.1 eV is a promising photoanode candidate for overall water splitting (Higashi et al., 2011, Li et al., 2013a, Li et al., 2013b, Li et al., 2015). As its band edge potentials straddle H2O redox potentials, it is possible to split water with very small bias or even without any bias (Chun et al., 2003, Wang et al., 2014, Fan et al., 2017). In other words, its onset potential for water oxidation may be close to 0 VRHE (reversible hydrogen electrode) (Chun et al., 2003, Wang et al., 2014, Fan et al., 2017).

Although near-theoretical-limit photocurrent densities for water oxidation on Ta3N5 have already been reported (Liu et al., 2016), a large bias of at least 0.6–0.9 V is required for water splitting over a single Ta3N5 photoanode in the experiments (Liu et al., 2016, Li et al., 2013a, Li et al., 2013b, Seo et al., 2015, Wang et al., 2016, Zhang et al., 2016, Zhong et al., 2017, Chen et al., 2013). The solar-to-hydrogen efficiency of Ta3N5 photoanodes are greatly being hindered by unexpectedly high onset potentials (0.6–0.9 VRHE) (Liu et al., 2016, Li et al., 2013a, Li et al., 2013b, Seo et al., 2015, Wang et al., 2016, Zhang et al., 2016, Zhong et al., 2017, Chen et al., 2013). This problem has been plaguing chemists and materials scientists for many years. Therefore, it is of great significance to understand what causes the high onset potentials of Ta3N5 photoanodes.

Oxygen defects in the bulk are easy to be introduced during the synthesis process of Ta3N5 photoanodes. In our previous study, density functional theory (DFT) calculations have revealed that these oxygen impurities in the bulk Ta3N5 make its band edge potentials relative to vacuum shift approximately −0.2 eV (Fan et al., 2017, Wang et al., 2015a, Wang et al., 2015b). The influence of these oxygen defects in the bulk only partially helps to explain the high onset potential for water oxidation (van de Krol and Grätzel, 2012, Selcuk and Selloni, 2016, Bard and Fox, 1995). However, it cannot interpret why the onset potentials for water oxidation are as high as 0.6–0.9 VRHE even when oxygen evolution reaction (OER) electrocatalysts are loaded on the surface of Ta3N5 photoanodes.

Very recently, catalyst-water interfaces have been found to significantly affect the electronic structures and catalytic performances (He et al., 2016, Onda et al., 2005, Pham et al., 2014, Pham et al., 2017, Kharche et al., 2014, Hu et al., 2017, Carrasco et al., 2012, Selcuk and Selloni, 2016). For instance, the band edges of Si are shifted by approximately 0.5 eV irrespective of the termination, when it is exposed to water (Pham et al., 2014). It enlightens us to unravel the mystery of the unexpectedly high onset potentials for water oxidation over Ta3N5 from the view of the Ta3N5-water interface.

To the best of our knowledge, there have been no theoretical studies on the performances of Ta3N5 semiconductors in an aqueous environment. Here the effect of catalyst-water interface on the band edge potentials of Ta3N5 has been investigated by using DFT calculations. All the DFT and first-principles molecular dynamics (FPMD) simulations were performed by using the Vienna ab initio simulation package code (Kresse and Furthmüller, 1996). It is found that dissociated water can make the band edge potentials relative to vacuum shift −0.85 eV.

Results

Theoretical Models and Theoretical Methods

Figure 1A shows the snapshot of the equilibrated structure for Ta3N5-water model. From Figure 1A, it can be observed that water partially spontaneously dissociates and results in a partial hydroxylated surface. According to previous work, water tends to dissociate to form the hydroxylated surface at certain catalyst-water interface (He et al., 2016, Held et al., 2001, Feibelman, 2002). Hence, we have considered the effect of the ratio of the dissociated water on the band edge potentials of Ta3N5. To illustrate the effect of catalyst-water interface on the band edge potentials of Ta3N5, we have calculated the band edge potentials of Ta3N5 not only in vacuum but also in an aqueous environment. To eliminate the underestimation of band gap by PBE (Perdew-Burke-Ernzerhof) level, hybrid-DFT method has been used; the results of hybrid-DFT illuminated in Transparent Methods and Figure S1, are in good agreement with the experimental values (Chun et al., 2003). More detailed information is given in the Transparent Methods. For the calculations of the DFT-based molecular dynamics simulations, canonical ensemble was used and the functional optB88-vdW (Klimeš et al., 2011, Dion et al., 2004) was added to revise the long-range van der Waals interactions.

Figure 1

Ta3N5-Water Interfaces

(A) Atomic structure of the snapshot of equilibrated Ta3N5-water unit cell. The clay brown, light blue, red, and light pink balls represent Ta, N, O, and H atoms, respectively.

(B) Simplified schematic diagram of the band edge potentials of Ta3N5 relative to vacuum. VBM, V, V, and ΔV are the energy level of the VBM of Ta3N5, electrostatic potential of the bulk Ta3N5, the vacuum energy level, and the energy difference between the electrostatic potential of the Ta3N5 and vacuum level, respectively.

(C) The potential difference between the electrostatic potential of Ta3N5 and water at the Ta3N5-water interface.

(D) The density of water of the snapshot in an equilibrated Ta3N5-water unit cell.

To clarify the calculation methods, a simplified schematic diagram is shown in Figures 1B and 1C. Figure 1B illustrates a two-step method to calculate the energy levels of Ta3N5 in vacuum. First, the bulk Ta3N5 model has been performed, and we can obtain the potential difference between the valence band maximum (VBM) or the conduction band minimum (CBM) and the electrostatic potential of Ta3N5. Second, the slab model of Ta3N5 in vacuum gives the potential difference between electrostatic potential of Ta3N5 and vacuum potential level. Therefore, the VBM or CBM of Ta3N5 relative to vacuum can be obtained. For the band edge potentials of catalyst in water, we have first calculated the band offset between water and catalysis. Then combining the band edge potential of water relative to vacuum, we can calculate the band edge potentials of Ta3N5 in water. The detailed information is shown in Figure S2 and Equation S3. The density of water along z axis in equilibrated structure for Ta3N5-water model is illustrated in Figure 1D. The radial distribution functions of water in Figure S4 imply that water molecules are randomly dispersed. The average density is close to 1 g/cm3. The high mass density value of water at the two sides corresponds to the partial hydroxylated interface.

Although approaching a high photocurrent for solar water splitting (Liu et al., 2016), the low solar-to-hydrogen efficiency of Ta3N5 is still far below the requirements of commercial application (Bard and Fox, 1995). Previous theoretical studies on Ta3N5 prove that oxygen impurities and doping elements have an influence on the band edge potential of Ta3N5 (Fan et al., 2017, Liu et al., 2016; Wang et al., 2017). However, the working environment of Ta3N5 photoanode has not been considered in the previously calculated Ta3N5 models. The surface structure and catalytic activity of catalysts in an aqueous environment may be strikingly different from those of the catalyst in vacuum (He et al., 2016, Onda et al., 2005, Pham et al., 2014, Pham et al., 2017, Kharche et al., 2014). Whether the high onset potential of Ta3N5 is relative to an aqueous working environment is what we want to investigate in this study. The calculated band edge potentials of Ta3N5-water interface are the average values of 13 equilibrated Ta3N5-water unit cells (see Transparent Methods, Figure S3). Thirteen unit cells are enough for the required precision because the variations of the average band edge potential values of 9, 11, and 13 unit cells are smaller than 0.05 eV.

Calculations of Band Edge Positions

Figure 2 displays the band edge potential of Ta3N5 relative to vacuum in different conditions. The HSE (Heyd-Scuseria-Ernzerhof) revised band gap of Ta3N5 is 2.14 eV, which is very close to experimental values (Chun et al., 2003, Liu et al., 2016, Li et al., 2013a, Li et al., 2013b). The VBM and CBM of Ta3N5 relative to vacuum are calculated to be −6.30 and −4.16 eV, respectively. It is well known that the relationship between the absolute potential relative to vacuum and a normal hydrogen electrode (NHE) at 298 K can be shown as follows:Therefore, the VBM and CBM of Ta3N5 are equal to 1.86 and −0.28 V versus NHE, respectively. This implies that the VBM is more positive than the water oxidation potential and the CBM is more negative than water reduction potential versus NHE. Subsequently, the band edge potentials of Ta3N5 in vacuum indicate that Ta3N5 can split water without any external bias. It may be an ideal photoelectrode for water splitting. Nevertheless, water splitting for a single Ta3N5 photoanode without external bias has been not achieved experimentally yet. External bias of 0.6–0.9 V is always required for photoelectrochemical water oxidation over single Ta3N5 photoanodes even if they are loaded by OER electrocatalysts. This may arise from the fact that Ta3N5 behaves differently in an aqueous environment.

Figure 2

Band Edge Potentials

The calculated band edge potentials of the bulk Ta3N5, Ta3N5-water interface, Ta3N5-water interface with 50% dissociated interfacial water, and full hydroxylated Ta3N5-water interface relative to vacuum. The gray dashed lines are the water reduction and oxidation potentials, respectively.

Figure 2 also shows that both aqueous environment and the ratio of dissociated water will shift the band edge potentials negatively relative to vacuum level, which is in agreement with the results from PBE level (see Figure S5). The negative shifts of the band edge potentials are harmful to the onset potential of Ta3N5 photoanodes because the counterelectrode needs more external bias to drive the water reduction reaction. Therefore, the negative shifts of band edge potentials of Ta3N5 can increase the onset potential for water oxidation, thus reducing the solar-to-hydrogen efficiency of Ta3N5. Comparing the band edge potentials of Ta3N5 in vacuum with those at the Ta3N5-H2O interface, we can find that when exposed to water, the band edge potential relative to vacuum shifts −0.42 eV. The ratio of dissociated water also has a great influence on the band edge potentials of Ta3N5. When we intentionally build the Ta3N5 models with 50% and 100% initial hydroxylated interface, their band edge potentials relative to vacuum shift −0.72 and −0.85 eV compared with the Ta3N5-H2O model, respectively. It indicates that the ratio of dissociated water has an important effect on the band edge potentials. From Figure 1A, we can find that water and hydroxide are bonded to Ta atoms, whereas N sites are protonated. Our previous study (Wang et al., 2013) has also suggested that the single dissociated water on the Ta3N5 (1 0 0) surface tends to gain electrons.

Charge Transfer Calculations

To understand the surface charge transfer property, the one-dimensional (1D) charge density difference along z axis shown in Figure 3 is defined as follows:where Δρ, Δρ, and Δρ are the charge densities of Ta3N5-water unit cell, Ta3N5 slab model, and water unit cell, respectively.

Figure 3

Charge Transfer in the Ta3N5-Water Interface

The positive values represent the addition of the electrons, whereas the negative values imply the donation of electrons. The blue area corresponds to the water at the interface. Meanwhile, the orange area corresponds to the surface Ta3N5 layer and the subsurface Ta3N5 layer.

Figure 3 demonstrates that both the water and the dissociated water at the Ta3N5-water interface gain electrons, whereas the surface Ta3N5 layer loses electrons. This means that the electrons transfer from Ta3N5 to water, that is, the Ta3N5 surface layer may form an additional surface dipole. Similar surface dipole has also appeared in Si, GaN, and ZnO semiconductors when exposed to water (Pham et al., 2014, Kharche et al., 2014). To quantify the charge transfer at the surface, the integrals of the 1D charge density have been performed (see Figures S6–S9). Figure S6 suggests that the hydroxylated surface facilitates the electron transfer from Ta3N5 surface to water at the interface, implying that both the ratio of dissociated water at the interface and the aqueous environment have a significant effect on the formation of surface dipole and surface oxidation of Ta3N5 at the Ta3N5-water interface.

Bader Charge Analysis

To further confirm the charge transfer process, bader charge analysis of Ta3N5-water has been performed. The total and average charges of Ta and N atoms in the surface layer are exhibited in Table 1. According to Table 1, the total electrons of the surface Ta3N5 layer decrease when exposed in an aqueous environment and further decrease with the increase of the dissociated water ratio. This clearly demonstrates that when exposed to water surface Ta3N5 layer acts as electron donor and the increase of hydroxylated ratio accelerates the electron transfer from the surface Ta3N5 layer to water at the interface. As well known, for an ideal photoanode, the photoinduced holes instead of electrons should transfer from its surface layer to water at the interface, and then take part in the reaction of water oxidation. Obviously, the electron transfer from the surface Ta3N5 layer to water will lead to recombination of electrons and holes at the interface, which also reduces their PEC performance for water splitting. The surface dipole shown in Figure 3 will lead to a negative shift of band edge potentials of Ta3N5 relative to vacuum.

Table 1

Bader Charge Analysis of Ta3N5-Water Unit Cells

Charge Transfer/Electronsa	Ta3N5-Water	Ta3N5-Water with 50% Dissociated Water	Ta3N5-Water with 100% Dissociated Water
Total charge transfer of Ta	−1.22	−1.52	−1.72
Average charge transfer Ta	−0.10	−0.13	−0.14
Total charge transfer of N	0.12	0.23	0.40
Average charge transfer N	0.01	0.01	0.02

The total charge transfer is the difference between the total bader charges of all the surface Ta or N atoms in water and the total bader charges of the surface Ta or N atoms in vacuum. The average charge transfer is the average electron transfer of per Ta or N atom in the surface. The bader charge of a surface Ta atom and N atom in vacuum are 2.76 and 6.35, respectively. There are 12 surface Ta atoms and 20 surface N atoms in all models.

Table 1 also reveals that the variation of the total charge of N atoms is opposite to that of Ta atoms. Nonetheless, the absolute change of charge transfer of Ta is much greater than that of N atoms, which may favor the oxidation of the Ta3N5 surface layer. A Ta2O5 amorphous nanometer layer on the surface of Ta3N5 photoanodes has been observed after PEC water oxidation (He et al., 2016). In our experiments, the carrier concentration of Ta3N5 photoanodes has been found to become smaller during the process of PEC water oxidation, which may be due to the surface oxidation of Ta3N5. Combining charge density difference and bader charge analysis, we may conclude that the aqueous environment and the hydroxylated surface promote the formation of dipole at the interface and even the surface oxidation of Ta3N5 photoanodes. Facet and co-catalyst play important roles in real water splitting system; Figures S10–S12 show the effects of facet and co-catalyst (or co-catalyst) on the shifts of band edge potentials. According to these figures, the facet, the termination of facet, and different co-catalysts or overlayer have a great influence on the charge transfer at interface, and therefore affect the shift in band edge potentials.

Discussion

As discussed above, Ta3N5-water interface has a great influence on the band edge potentials and stability. The common strategy to overcome these problems is to tune the surface electronic structure by coating an overlayer or loading co-catalysts. Recently, a GaN overlayer has been proved to favor the onset potential for water oxidation and improve photochemical stability of Ta3N5, due to the restriction of the formation of self-oxidation surfaces (Zhong et al., 2017). However, the deposition of OER co-catalysts, such as Co3O4 (Liao et al., 2012), NiFe layered double hydroxide (Wang et al., 2015a, Wang et al., 2015b), Ni(OH)x (Liu et al., 2016), and IrO2 (Higashi et al., 2011), have limited effect on the onset potential performances of Ta3N5 photoanodes. The effect of termination in Figure S11 suggests that it is worthwhile to optimize the methods to deposit the co-catalysts, because the nanostructure of co-catalysts (Seger et al., 2013, Ping et al., 2015) as well as the experimental parameters (Paracchino et al., 2012) will influence the electronic environment and the catalyst-water interface. Therefore, interface engineering is a promising way to enhance the performance of Ta3N5 by regulating the electronic structure at the catalyst-water interface and restricting the oxidation of Ta3N5 surface. The investigation of the mechanism of water oxidation under illumination at Ta3N5-electrolyte interface is meaningful, because it is generally accepted that the photogenerated holes will oxidize the surface Ta3N5 and generate N2 (Liao et al., 2012). However, this dynamic oxidation process at Ta3N5-electrolyte interface is still unclear. Combining theory and in situ experiments (Herlihy et al., 2016) is a promising strategy to investigate the dynamic process of water oxidation at the Ta3N5-water interface under illumination.

In conclusion, the DFT and FPMD simulations were performed to investigate the band edge behavior of Ta3N5 in an aqueous environment. The HSE revised band gaps and band edge potentials of Ta3N5 are close to the experimental values. The band edge potentials of Ta3N5 are found to shift negatively relative to vacuum level when exposed to water. The negative shift is harmful to the onset potential performance of Ta3N5 photoanodes. The negative significant shift of the band edge positions may explain why a bias of at least 0.6–0.7 VREH should be applied for water splitting over Ta3N5 photoanodes. Facet, termination of facet, and co-catalyst will also affect the shift of the band edge positions of Ta3N5. Interface engineering, as well as defect engineering (Feng et al., 2019), is a promising way to improve photoelectrochemical performances for water splitting over photoanodes.

Limitations of the Study

This study demonstrates the effect of water on the shifts of band edge potentials of Ta3N5. We also consider the effect of co-catalyst or overlayer on the shifts of band edge potentials. However, water-overlayer (or co-catalyst)-Ta3N5 in real Ta3N5 photoelectrochemical water splitting system is very complex for DFT calculation. Therefore the development of a more complex model that can accurately describe the real water splitting system will be necessary.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.