Moussab
Harb
*,
Luigi
Cavallo
and
Jean-Marie
Basset
KAUST Catalysis Center (KCC), Physical Sciences and Engineering Division (PSE), King Abdullah University of Science and Technology (KAUST), Thuwal 23955-6900, Saudi Arabia. E-mail: moussab.harb@kaust.edu.sa; Tel: +966-012-8080788
First published on 16th April 2020
The effects of native defects and exposed facets on the thermodynamic stability and photocatalytic characteristics of Ta3N5 for water splitting are studied by applying accurate quantum computations on the basis of density functional theory (DFT) with the range-separated hybrid functional (HSE06). Among the three explored potential candidates for O-enriched bulk Ta3N5 structures with substituted O at N sites and accompanied by interstitial O or Ta-vacancies, the first and third structures are relevant. The four possible (001), (010), (100) and (110) low Miller index exposed facets of Ta(3−x)N(5−y)Oy (y = 7x) are also explored, which show lower formation energies than those of Ta3N5. This highlights O occupation at N sites together with Ta vacancies as native defects in the prepared samples. The most appropriate facets for HER and OER are predicted based on the redox and transport characteristics. Our work predicts (001) and (110) facets only for HER, whereas the (010) facet is predicted for OER. Our findings indicate the importance of understanding the significance of various facets when preparing and testing new material photocatalysts for water splitting reactions.
Tantalum nitride (Ta3N5) is a widely used photocatalyst for solar water splitting, due to its band gap energy of 2.1 eV.18–23 By nitrating Ta2O5 at high temperatures, Ta3N5 powder samples were obtained and the orthorhombic crystal structure was revealed by characterization techniques.18–23 As invoked in the literature, the material could separately generate oxygen or hydrogen from water by the use of sacrificial reagents.18,23 The overall water splitting needs to be carefully examined when using this material in order to overcome the limited efficiency. Photocatalytic measurements are related to the method of sample preparation.20,23 This discrepancy was justified by the presence of unavoidable O impurities in Ta3N5 samples, which originated from Ta2O5 nitridation methods,20,23 leading to defective or non-stoichiometric compounds. One should always be reminded that preparing electrodes using powder materials is not at all easy and leads to a lot of difficulties.19 Ta3N5 flat band potentials were found to be dependent on the photoelectrode fabrication method.23 Because Ta3N5 surfaces tend to be oxidized and the surface oxidation rate may be modified during photocataytic reactions, the expected effects of morphology (nature/type of exposed facets) together with intrinsic defects (such as O impurities and Ta vacancies) on Ta3N5 photoredox reactions need to be well understood. These aspects will be provided in this work via density functional theory (DFT).
In previous theoretical works,24–30 we have studied various fundamental characteristics for solar energy conversion to chemicals with a series of materials commonly utilized in photocatalysis and reproduced with high accuracy the experimental results. This was achieved by performing first-principles computations based on density functional theory (DFT) along with the range-separated hybrid Heyd–Scuseria–Ernzerhof (HSE06)31,32 functional. The choice of the level of theory, in particular, the chosen HSE06 functional, was previously justified based on a systematic comparison with the results obtained using the common standard GGA-PBE functional.24–30 Using this scheme, we recently identified the most appropriate exposed facets of TiO2 (in anatase and rutile phases) for photocatalytic hydrogen and oxygen evolution reactions by combining their optoelectronic and redox properties.33 Our predicted results could successfully give rational insights into the fundamental origin behind the better activity of anatase for HER compared to rutile as found in experiments.33–36
In this paper, we introduce a systematic study on the impact of intrinsic defects and exposed facets on the optoelectronic and redox features of a Ta3N5 water splitting photocatalyst by applying the DFT/HSE06 method. We explored three potential candidates of O-enriched bulk Ta3N5 structures. They are associated with substituted O at N sites (unbalanced charges) and accompanied by interstitial O or Ta-vacancies for the charge balance. By selecting the most relevant bulk structures, we then carried out calculations on the thermodynamic stability, and optoelectronic and redox characteristics for the four possible (001), (010), (100) and (110) low Miler indexes of O-enriched Ta3N5 exposed facets. Moreover, we identified the most appropriate facets for OER and HER reactions and made a comparison with the three principal (023), (310) and (113) facets of prepared Ta3N5 samples. We believe that the concepts resulting from this theoretical study will guide experimentalists in selecting the most suitable exposed facets where the two HER and OER co-catalysts should be anchored for improving reduction of protons and oxidation of water.
Regarding O-enriched bulk Ta3N5 structure simulation, three potential candidates were explored starting from the (3 × 3 × 1) pure Ta3N5 (Ta36N60) supercell. Two substituted O at N sites (Ta36N58O2), two substituted O at N sites together with one interstitial O (Ta36N58O3) and five substituted O at N sites accompanied by one Ta-vacancy (Ta35N55O5). The corresponding stoichiometry Ta(3−x)N(5−y)Oy was adopted with x = 0; y = 0.17 for the first model, x = 0; y = 0.25 for the second one, and x = 0.1; y = 0.5 for the third one, respectively. For each bulk structure, many structural configurations were explored to find the most preferential O location inside the Ta3N5 lattice as well as with respect to the Ta-vacancies.
The thermodynamics of the most relevant Ta(3−x)N(5−y)Oy bulk structures was investigated using the following reaction:
(1) |
(2) |
(3) |
Regarding the partially O-enriched Ta3N5, (001), (010), (100) and (110) facets were modeled using the obtained optimal thicknesses of pure Ta3N5. A (110) slab model of 14 Å (Ta47N73O7), (100) slab model of 13 Å (Ta41N63O7), (010) slab model of 15 Å (Ta35N53O7) and (001) slab model of 15 Å (Ta35N53O7) were selected on the basis of the number of new sub-coordinated surface N generated upon crystal cleavage together with the most relevant sub-coordinated bulk N species generated after releasing the Ta-vacancy. These four models led to the stoichiometry Ta(3−x)N(5−y)Oy with x = 0.072; y = 0.437 for (110), x = 0.072; y = 0.50 for (100), x = 0.084; y = 0.583 for (010) and x = 0.084; y = 0.583 for (001). For each slab structure, key structural configurations were explored by replacing with O some of the new sub-coordinated 2-fold or 3-fold coordinated surface N that appeared after cleavage and the 3-fold coordinated bulk N sites when Ta-vacancies are present inside the lattice to find the most favorable location of O.
Using reaction (1), we investigated the thermodynamics of the most relevant Ta(3−x)N(5−y)Oy surface structures. We calculated the surface formation energy relative to Ta3N5 as follows:
(4) |
Material | Structure | Formation energy | Lattice constants | |||||
---|---|---|---|---|---|---|---|---|
a | b | c | α | β | γ | |||
Ta3N5 | Expt20,23,26,55,56 | — | 3.89 | 10.22 | 10.28 | 90 | 90 | 90 |
Ta3N5 | 1(b) | — | 3.89 | 10.25 | 10.27 | 90 | 90 | 90 |
Ta3N4.83O0.17 | 2(a) | −42.54 | 3.91 | 10.19 | 10.28 | 90 | 90 | 90 |
Ta3N4.83O0.25 | 2(b) | −57.62 | 3.89 | 10.23 | 10.25 | 89.68 | 90 | 90 |
Ta2.9N4.6O0.4 | 2(c) | −98.60 | 3.89 | 10.23 | 10.27 | 90 | 90 | 90 |
Considering the Ta3N4.83O0.17 material, the most favorable structure is found when the two 3-fold-coordinated N sites are replaced with O, as shown in Fig. 3a. For this structure, a singlet state associated with two extra paired up electrons in a delocalized state throughout the crystal is found to be the most favorable spin situation. The computed lattice constants are very similar to those of Ta3N5 (Table 1). The structure associated with two O occupying two 4-fold coordinated N was found to be 4.83 kJ mol−1 less stable than that of the previous one.
Regarding Ta3N4.83O0.25, the geometrical structure is particularly stable when the three O atoms are gathered between two Ta, forming two TaO3N4 polyhedra, one TaO2N5 polyhedron and one TaON5 octahedron (see Fig. 3b). The computed lattice parameters obtained for this structure are also similar to those of pure Ta3N5, except for a small decrease in the angle α, by 0.32° (see Table 1). In this structure, the interstitial O is located nearby the two other substitutional O, occupying two three-coordinated N bridging two Ta. The most favorable spin configuration is a singlet state, leading formally to three O2− defects replacing two N3− species. Another structural configuration derived from two O substituting two separated 4-coordinated N from the interstitial one was found to be 3.86 kJ mol−1 higher than the lowest-energy structure.
With respect to Ta2.9N4.6O0.4, the strong interaction between the Ta-vacancy and the five O was at the origin of its good stability. In the most stable structural configuration, the five O atoms occupying five N sites surround the Ta-vacancy, as displayed in Fig. 3c. The obtained lattice constants are again very similar to those of Ta3N5 (Table 1). For this structure, the most favorable spin situation is found to be a singlet resulting from five O2− replacing five N3− in conjunction with the five holes released from the created Ta vacancy. Note that the geometrical configuration displaying the well-separated five O and the Ta-vacancy was less stable by 9.16 kJ mol−1.
The relative formation energies of the pure materials of Ta3N4.83O0.17, Ta3N4.83O0.25 and Ta2.9N4.6O0.4 were computed at 1200 K (typical annealing temperature for these materials) and values of −42.54, −57.62 and −98.60 kJ mol−1 were obtained, respectively. As their formation energies are much lower than that of pure Ta3N5, it can be clearly concluded that the three potential partially O-enriched bulk structures are thermodynamically more stable than the pure material. This result clearly explains the presence of unavoidable O at N sites with Ta vacancies as intrinsic defects in Ta3N5 samples remaining from Ta2O5 nitridation methods.20,23
For pure Ta3N5, the computed 2.2 eV band gap well reproduced the measured value (2.1 eV)23,26,55,56 (Fig. S2, ESI†).
Considering Ta3N4.83O0.17, the electronic structure analysis shows an n-type conductivity as observed experimentally,23,26 which is represented by pinning up the Fermi level near the CBM of Ta3N5, as shown in Fig. 4a. This phenomenon originates from the newly created donor states made by Ta4+ (5d1) orbitals strongly delocalized over the crystal and located in the 0.7 eV range just below the CBM of Ta3N5. The CBM state is composed of empty Ta5+ (5d0) orbitals and the VBM state consists of filled N3− (2p6) orbitals.
For Ta3N4.83O0.25, a very similar band gap and similar absolute VBM/CBM energy positions together with orbital contributions dominated by N3− (2p6)/Ta5+ (5d0) were found to those obtained in the case of pure Ta3N5 (Fig. 4b and Fig. S2, ESI†).
With respect to Ta2.9N4.6O0.4, the electronic DOS analysis reveals a slightly broader band gap of 2.3 eV (Fig. 4c) compared with that of the pure material (2.2 eV). Similar to pure Ta3N5, the CBM state is mainly made by Ta5+ (5d0) orbitals and the VBM state is also dominated by occupied N3− (2p6) orbitals due to the minor contributions from O2− (2p6) orbitals in the deeper energy zone of VBM. However, its absolute VBM energy position is shifted downward with respect to that of the pure material one (Fig. 4c and Fig. S2, ESI†) and this is expected to lead to modifications in the redox properties of this material.
Based on the electronic structure results, we focused in what follows on the two relevant partially O-enriched bulk Ta3N5 structures with substituted O at N sites together with Ta vacancies for deeper surface investigation.
Inside the Ta3N5 crystalline lattice, four possible crystallographic directions, 〈110〉, 〈100〉, 〈010〉 and 〈001〉, are present (Fig. 1). An anisotropic nature is highlighted in the mobility of electrons and holes over the four crystalline orientations of Ta3N4.83O0.17 and Ta2.9N4.6O0.4 lattices. The obtained values are reported in Table 2. For electron effective masses, small values of 0.9, 0.7, and 0.7 for Ta3N4.83O0.17 and 0.9, 0.8, and 0.8 for Ta2.9N4.6O0.4 are found in the 〈110〉, 〈010〉 and 〈001〉 directions, while greater values of 2.0 and 2.2 are obtained in the 〈100〉 orientation, respectively. Consequently, a good mobility for electrons on the (110), (010), and (001) facets and a low mobility on the (100) surface are expected. Small effective mass values of holes are obtained in the 〈110〉, 〈010〉 and 〈001〉 orientations of 0.7, 0.9, and 0.9 for Ta3N4.83O0.17 and 0.8, 1.0, and 1.0 for Ta2.9N4.6O0.4, while greater values of 3.6 and 3.9 are found in the 〈100〉 orientation, respectively. Consequently, a good mobility for holes on the (110), (010) and (001) facets and a low mobility on the (100) surface are expected.
Direction | Ta3N4.83O0.17 | Ta2.9N4.6O0.4 | ||
---|---|---|---|---|
(me*/m0)ij | (mh*/m0)ij | (me*/m0)ij | (mh*/m0)ij | |
〈110〉 | 0.9 | 0.7 | 0.9 | 0.8 |
〈100〉 | 2.0 | 3.6 | 2.2 | 3.9 |
〈010〉 | 0.7 | 0.9 | 0.8 | 1.0 |
〈001〉 | 0.7 | 0.9 | 0.8 | 1.0 |
The similar trends in the transport properties along the 〈001〉 and 〈010〉 crystal directions may be understood by the similar N and Ta arrangements in the (001)- and (010)-oriented slab structures with similar Ta–Ta and N–N distances (see Fig. 2c and d for more details). The slight discrepancy in the effective masses along the 〈110〉 direction might come from the N and Ta arrangements in the (110)-oriented material structure with longer Ta–Ta and shorter N–N distances (see Fig. 2a). The clear difference in the properties along the 〈100〉 lattice direction with greater effective masses might come from the appearance of parallel planes to the (100) surface. They are positioned according to the longer N–N and Ta–Ta distances in the (100)-oriented structure (see Fig. 2b).
Regarding the partially O-enriched Ta3N5, different surfaces were built by replacing with O some of the new 2- or 3-fold sub-coordinated generated surface N and the 3-fold coordinated bulk N sites when Ta vacancies are present inside the lattice to find the most favorable location of O. These four models led to the stoichiometry Ta(3−x)N(5−y)Oy with x = 0.072; y = 0.437 for (110), x = 0.072; y = 0.50 for (100), x = 0.084; y = 0.583 for (010) and x = 0.084; y = 0.583 for (001). The most favorable structure of the (110) Ta2.928N4.563O0.437 slab is found when the 3-fold and 2-fold coordinated O exists in the bulk and surrounding the Ta-vacancy (Fig. 5a). For the (100) Ta2.928N4.5O0.5 slab, the most favorable surface structure is partially covered by 3-coordinated O with some 3- and 2-coordinated O inside the bulk located around the Ta-vacancy (Fig. 5b). The most favorable slab structure of (010) Ta2.916N4.417O0.583 is fully covered by 2-coordinated O with some minor O content present in the lattice close to the Ta-vacancy (Fig. 5c). For the (001) Ta2.916N4.417O0.583 slab, the most favorable structure reveals a partially covered surface by 2-coordinated O together with 2- and 3-coordinated O inside the material located around the Ta-vacancy (Fig. 5d). The metastable structure is given in the ESI† (Fig. S3). For the four slab structures, the lowest-energy spin situation is found to be singlet. Bader charge analysis highlighted the appearance of Ta4+ that originated from released electrons after the substitution of O for N, in line with the experimental results.23,26
The surface formation energy of Ta(3−x)N(5−y)Oy relative to the pure material was computed at 1200 K, which is the typical annealing temperature for this material, and values of −0.18, −0.13, −0.08, −0.07 J m−2 were found for the (001), (010), (100), (110) facets. Indeed, the relative difference is very small, only in the 0.01–0.05 J m−2 range, which highlights that these surfaces might exist in prepared Ta3N5 samples and must be considered for deep analysis. As the obtained relative formation energy values of the different surfaces are negative and even lower than those obtained without Ta-vacancies,57 this confirms, once again from the thermodynamic point of view, the presence of O impurities together with Ta-vacancies in the prepared Ta3N5 samples.20,23
Based on these results, Ta(3−x)N(5−y)Oy facets should be less sensitive than those of Ta3N5 for oxidation. Having mixed O and N or even only O on the surfaces should lead to greater stability when compared to pure material surfaces.
The original band gaps of the (110), (100), (010) and (001) Ta(3−x)N(5−y)Oy slabs were computed and identical values or 0.1 eV broader or 0.6 eV narrower (Fig. 6) were found in comparison with those obtained for the bulk materials (Fig. 4). The VBM/CBM states consist of N3− (2p6)/Ta5+ (5d0) orbitals as found in Ta3N5, while O2− (2p6) orbitals appear lower in the VB states (see Fig. 6). The analysis of their electronic structures highlights the appearance of donor states made by Ta4+ (5d1) orbitals located in the range of 0.4–1.0 eV lower than the CBM of the original material (see Fig. 6). This result leads to a material with n-type conductivity as obtained in experiments.23,26
The absolute VBM/CBM energy levels of Ta(3−x)N(5−y)Oy slabs are −1.1/+0.6 eV, −1.7/+0.6 eV, −1.1/+1.4 eV, and −1.1/+1.2 eV, for (001), (010), (100), and (110), respectively, as shown in Fig. 6. Their potential profiles give vacuum energies of 3.9 eV, 4.8 eV, 3.9 eV and 4.5 eV for (001), (010), (100) and (110), respectively, as shown in Fig. 7. The band energy levels of Ta(3−x)N(5−y)Oy slabs are then positioned relative to the vacuum (see the ESI† for more details). An anisotropic nature is highlighted on the basis of exposed facet nature. The VBM energy levels of (001), (100) and (110) slabs are found to be 0.73, 0.73 and 0.13 eV higher than the O2/H2O level (see Fig. 8a, b and d). Their CBM energy level is 1.2, 2.0 and 1.2 eV higher than the H+/H2 level (see Fig. 8a, b and d). In contrast, the (010) VBM level is 0.77 eV below the O2/H2O level and the CBM position is 0.3 eV higher than the H+/H2 potential (Fig. 8c). Due to the inappropriate VBM energy levels of (110), (100) and (001) slabs relative to the O2/H2O level, the generated holes on these surfaces will lose the required driving force for water oxidation. Consequently, these surfaces are predicted as suitable candidates only for HER due to their appropriate CBM level with respect to the H+/H2 level. In contrast, as both CBM and VBM levels of the (010) slab are correctly positioned with respect to the H+/H2 and O2/H2O potentials, respectively, this surface is a suitable candidate for HER and OER. Our computed data for the (010) surface are in good agreement with the measurements achieved recently revealing overall water splitting on Ta3N5 nanorod single crystals grown on the edges of KTaO3 through the (010) surface.58 Following this result, the ability to oxidize water using this material can be maintained by putting an OER co-catalyst on top of the (010) surface. The similar trend in the redox properties between (100) and (001) surfaces comes from the fact that 4- or 5-coordinated Ta and 3-coordinated N with 2- or 3-coordinated O are present on the two surfaces. Interestingly, the small downward shift in the (110) VBM position can be correlated with the three-fold coordinated N species present on this surface. However, the appearance of only 2-fold coordinated O on the (010) surface may explain the drastic change in its redox features with an important downward shift.
One should be reminded here that very similar electronic and redox properties are obtained for the two (310) and (110) facets and for the two (023) and (010) facets due to the similarities in N and Ta structural distributions on each surface together with the coordination number of surface species.57 The results presented here highlight the impact of each surface on proton reduction or water oxidation depending on the chemical element structural distribution and the surface species coordination number.
As defined in the introduction, the co-catalysts for H2 and O2 evolution reactions must be, in principle, deposited over promising surfaces to improve the kinetics of electron and hole transfer to water solution. Obtaining such an anisotropic nature in the optoelectronic and redox features makes water splitting more complex.59,60 Consequently, by merging the optoelectronic and redox properties together, our work indicates the (010) surface as being the only appropriate candidate for OER, whereas the two (110) and (001) Ta(3−x)N(5−y)Oy surfaces are the most appropriate candidates for HER. The 5- and 4-coordinated Ta exposed species would be the active sites for proton reduction towards HER on the two (110) and (001) surfaces, respectively. The 2-coordinated O exposed species would be the active site for water oxidation towards OER on the (010) surface. The increase or decrease in the number of defects is expected to minimally affect the number of active sites and, therefore, the overall activity of the photocatalyst.
Having suitable flat band potentials of Ta3N5 powder samples relative to water redox limits cannot ensure an active photocatalytic material for water splitting reactions because of the anisotropic behavior of its redox and optoelectronic features and their direct relation with the type/nature of exposed facets. The low efficiency for water splitting obtained so far using this material can be supported by the obtained results from this work. The most effective way to overcome the water splitting efficiency limitation requires a deposition of the co-catalyst for OER on the (010) surface and that for HER either on the (110) or (001) surface.
Our study predicted the two (110) and (001) Ta(3−x)N(5−y)Oy surfaces as being the most appropriate candidates for HER due to appropriate CBM levels with respect to the H+/H2 level together with good accumulation of photogenerated electrons, whereas the (010) surface was the only appropriate candidate for OER due to its appropriate VBM level with respect to the O2/H2O level along with good accumulation of photogenerated holes.
In summary, our work indicates that different facets can have remarkably different photophysico-chemical properties, with an impact on their potential performance in HER and OER. This concept will help experimentalists in choosing the appropriate exposed facets where the HER and OER co-catalysts should be anchored.
Footnote |
† Electronic supplementary information (ESI) available: Computational details for the different methods used in this work. Convergence tests of the electronic properties of (001), (010), (100) and (110) pure Ta3N5 slabs with crystal thickness based on the DFT/PBE method. Electronic structure of bulk Ta3N5 crystal using the DFT/HSE06 method. DFT/PBE-based metastable slab atomic structures of (110), (100), (010) and (001) partially O-enriched Ta3N5 facets. See DOI: 10.1039/d0cp01394c |
This journal is © the Owner Societies 2020 |