Nitrogen-doped, proton-exchanged Dion-Jacobson layered niobate perovskites for photocatalytic hydrogen generation in solar light

Wide band gap semiconductor niobate photocatalysts with Dion-Jacobson layered perovskite structure were nitrogen-doped via simple gas–solid reaction to extend their absorption into the visible light range. Nitrogen doping was performed using ammonia as precursor, resulting in decreased band gaps of doped AB2Nb3O10 compounds (A = Cs, Rb, K; B = Ca, Sr) down to 2.5 eV. The resulting materials were investigated concerning their chemical and electronic structures. Nitrogen-doped AB2Nb3O10 crystals showed a clear red shift in absorption. Photocatalytic performance tests for the doped materials evaluated the capability of H2 production under simulated solar irradiation. The addition of carbonates to the gas–solid reaction turned out to be advantageous for the reduction of defects and the preservation of photocatalytic activity of nitrogen-doped layered niobates AB2Nb3O10.


Introduction
Photocatalysis with semiconductors as light absorbers has been studied intensively because of the potential applications in clean energy and environmental remediation, for example in water splitting for storage of solar energy to chemical energy, [1] and photocatalytic degradation of pollutants in water and air [2]. In order to achieve efficient and sustainable solar energy conversion, the photocatalyst should have a small band gap (< 3 eV). Such visible light activity still remains a key challenge in the area of photocatalysis. Most of the active photocatalysts including many highly stable d 0 oxide semiconductor materials exhibit band gaps in the UV range, [3] such as TiO 2 (3.0 eV and 3.2 eV for rutile and anatase, respectively). This allows such photocatalysts to absorb solely UV light, which only accounts for around 5% of the total solar energy. Extending the band gap of photocatalysts into the wide visible light region seems a logical solution to this problem, since this would enable the photocatalyst to utilize the much larger portion of visible light energy.
Therefore, Asahi et al. introduced nitrogen doping in 2001 to reduce the band gap of oxide semiconductors [4]. Due to the insertion of N 2p orbitals, mid-gap states were introduced, effectively shifting the absorption of N-doped TiO 2 into the visible light range.
Wang et al. later showed that the crystal structure has an influence on the dopant distribution [5]. Treating layered titanates in ammonia gas, they achieved homogeneous nitrogen doping throughout the material due to the open crystal structure, enabling band-to-band visible light photon excitation by mixing of O 2p and N 2p orbitals forming a new valence band [6]. Following up on this, comparable examples of nitrogen-doped materials with layered or tunnel-like open crystal structure were reported to exhibit homogeneous nitrogen doping and visible light activity in photocatalysis [7][8][9][10][11][12].
Interestingly, very little examples of N-doped layered niobates have been reported [13,14]. The reason might be that in many cases, Nb 4+ and oxygen defects are formed leading to low photocatalytic activity [15], sometimes also referred to as Nb self-doping [16]. To reduce the amount of self-doping, mixing niobium oxides with carbonate salt was reported to reduce the extent of self-doping, leading to yellow N-doped KCa 2 Nb 3 O 10 [15].
Recently, we showed that proton exchange of KCa 2 Nb 3 O 10 can lead to a strong enhancement in photocatalytic activity for hydrogen evolution [17]. Here, we combine the latter two strategies, presenting nitrogendoped proton-exchanged Dion-Jacobson layered niobate perovskites for solar hydrogen generation. We present two strategies to reduce defects formed upon ammonolysis, namely oxygen post-treatment and carbonate addition. The latter one-step preparation leads to lower amount of defects in the doped materials, and the highest photocatalytic activities among the nitrogen-doped layered niobate nanosheets studied.
Bulk material AB 2 Nb 3 O 10 compounds (A = Cs, Rb, K; B = Ca, Sr) were prepared via the molten salt synthesis described previously [18]. The material will be denoted AB, according to their respective cations.
For the nitrogen doping with ammonia gas, a tube furnace equipped with a quartz tube was used. Software and mass flow controllers regulated the temperature and gas flow. 300 mg of the bulk material was placed in a ceramic boat and exposed to 75 mL/min NH 3 gas flow at 800 °C for 1 h, resulting in black powders. Samples will be denoted comparable to the untreated samples, AB-NH 3 .
When treated under NH 3 gas flow and re-oxidized with oxygen (5 mL/min at 500 °C for 5 min), this resulted in light yellow and lemon green coloured powders. Samples will be denoted comparable to the untreated samples, AB-NH 3 reox.
When bulk material was mixed with 20% excess of A 2 CO 3 (A = Cs, Rb, K) before NH 3 gas treatment, the ammonolysis resulted in light yellow and lemon green coloured powders. Samples will be denoted comparable to the untreated samples, AB-NH 3 -CO 3 Scheme 1 gives an overview on the synthesis procedure.
For photocatalysis, all of the nitrogen doped materials were proton exchanged through the method described previously [17].
X-ray diffraction patterns were measured on a PANalytical MPD diffractometer using Cu-Kα radiation (λ = 0.1541 nm) in the 2θ range from 5 to 55. The phase Scheme 1 Synthesis procedure for N-doped AB 2 Nb 3 O 10 layered perovskites purity was confirmed using the HighScore Plus software version 3.0e (3.0.5) and the ICSD database.
UV-vis was performed on a PerkinElmer Lambda 750 UV-vis-NIR spectrometer, using a Praying Mantis mirror construction from Harrick. X-ray photoelectron spectroscopy (XPS) was measured with a VersaProbe III Scanning XPS Microprobe from Physical Electronics (III), with a monochromatic Al K α source and a beam diameter of 100 µm. Survey spectra were measured with step size of 0.4 eV and a pass energy of 224 eV, high-resolution spectra were measured with a step size 0.1 eV and a pass energy of 26 eV; the time per step was 50 ms in both cases. Slow moving electrons and Ar ions were used for sample charge neutralisation. Sputter-cleaned beforehand was omitted to avoid surface reduction by preferential sputtering. Data were analysed with CasaXPS using Shirley backgrounds and Gaussian-Lorentzian (GL30) peak shapes. The C 1 s signal was set to 284.8 eV for charge correction. The Nb 3d signal was fitted using a constrained area ratio of 2:3 for the 3d 3/2 and 3d 5/2 doublet, a peak separation of 2.72 eV, [19] and equal FWHM for the peaks of the doublet.
The morphologies of the compounds were investigated by field emission scanning electron microscopy (FESEM, Zeiss LEO 1530). An acceleration voltage of 3 kV was used together with a working distance of 8 mm and an aperture set to 30 µm. The energy dispersive X-ray analysis EDX) were performed with the same device though an acceleration voltage of 20 kV and an aperture of 60 µm were applied.
Hydrogen production experiments were conducted in a 120 mL volume double-walled glass reactor with quartz top. To avoid any thermal catalytic effect, the reactor was cooled down to 20 °C using a double walled jacket, through which cooling water was circulated from a thermostat (Lauda). High purity argon was used as the carrier gas for the reaction products. The flow rate was set at 25 mL/min and controlled by Bronkhorst mass flow controller. The evolved hydrogen was detected using a Shimadzu GC-2014 gas chromatograph equipped with a thermal conductivity (TCD) detector and RESTEK Shin-Carbon ST 100/120 column. Gas separation was achieved isothermal at 35 °C, afterwards the column was heated to remove adsorbed water. In a typical experiment, 30 mg photocatalyst were dispersed in 108 mL H 2 O and 12 mL MeOH. Prior to irradiation, the system was purged with argon at 100 mL/min to ensure complete air removal. 0.05 wt.−% Rh co-catalyst were in-situ photodeposited on the catalyst from Na 3 RhCl 6 introduced in the system with a syringe through a reactor inlet rubber sealing. The solution was stirred constantly. As light source, a 150 W Xe solar light simulator (Newport Sol 1A) was used.

Results and discussion
X-ray powder diffraction patterns of the pristine AB 2 Nb 3 O 10 compounds are presented in Fig. 1. All samples show high crystallinity and reflections characteristic for the respective single-phase Dion-Jacobson layered perovskites [18]. The XRD analysis shows that the nitrogen doping with ammonia gas has no influence on the material structure, hardly any changes can be observed in the XRD patterns after ammonolysis. This confirms that no transformation to oxynitrides has taken place.
SEM images after ammonolysis (Fig. 2) show sheetlike material with small and larger crystallites, as known from the bulk material, even after the ammonia treatment at 800 °C. For comparable images of the bulk materials, we refer to our earlier work [18]. Figure 3 shows absorption spectra for AB 2 Nb 3 O 10 compounds and of nitrogen-doped counterparts, respectively. The amount of nitrogen doping from XPS ranges from 0.1 at.−% (KCa) to 6 at.−% (RbSr) under the same reaction conditions. It is clear that nitrogen doping results in a change of the absorption from band gaps of around 340-350 nm to a broad visible light absorption. Utilizing visible light for photocatalytic hydrogen generation is a key challenge, and visible light absorption of a material is a prerequisite for visible light activity. In contrast to the pristine AB 2 Nb 3 O 10 compounds, the large red-shift by N-doping is in most of the cases very pronounced and comparable to other nitrogen-doped layered titanates or tantalates [5,8].
However, as the increased baseline suggests, strong amounts of defects were generated upon ammonia treatment. All samples exhibited grey to black colour after ammonolysis ( Figure S1) indicating niobium and oxygen defects [20].
To reduce the amount of defects, especially on the surface to reduce the extent of surface recombination, nitrogen-doped samples were treated additionally for a short re-oxidation process in pure oxygen. As a result, greenyellowish samples were gained ( Figure S2).
The respective XRD patterns show no structural changes, as expected, upon this treatment (Fig. 1). The morphology of the materials also hardly changes, as can be seen in the SEM images ( Figure S3), and the elemental distribution of all elements is highly homogeneous. ( Figure S4).
Interestingly, the absorption spectra show distinct changes. As can also be seen in Fig. 3, the baseline of the absorption spectra is recovered to nearly zero baseline, indicating healed surface defects upon treatment in oxygen.
Unfortunately, the reaction seems to go in line with reduced visible light absorption, since the absorption shift 1 3 into the visible light region is less pronounced, in some cases only resulting in absorption shoulders left. This indicates that the absorption shifts observed after ammonia treatment are a superposition of extended light absorption due to mixing of O 2p and N 2p orbitals, and due to defect formation.
The black materials, before proton exchange, are known in literature to contain too many oxygen defects in the crystal structure, and exhibit catalytic activity for hydrogen production only under UV irradiation [15]. Using a solar simulator (AM1.5G), we could confirm that none of the doped bulk materials, prior to H-exchange, show any activity for hydrogen evolution under solar simulation (not shown). Only after proton exchange, the doped compounds show activity under the desired conditions.
The hydrogen production experiments in aqueous suspension with methanol as a sacrificial agent (electron donor) of the proton-exchanged, nitrogen-doped dark compounds reveal that hydrogen can be generated after Rh decoration (Fig. 4). The experiments were performed under simulated sunlight (AM1.5G, 1000 W m −2 ) containing small amount of UV light (< 5%), the hydrogen evolution rate trajectories are given in the Supporting Information ( Figure S5). As can be seen, for the most active compounds, the evolution rates are very stable over time. The highest activity can be observed for HCa 2 Nb 3 O 10-x N x and HSr 2 Nb 3 O 10-x N x , the latter one also being the most active material as proton-exchanged undoped oxide under UV light [17]. Without proton exchange, no activity was observed, as expected [15]. As for the undoped materials, due to their band gap no photocatalytic activity under sun simulation can be observed (not shown).
The re-oxidated materials show, in some cases, considerably higher H 2 production activity (the development of hydrogen evolution rates over time shown in Figure S6). The best activity was observed for NH 3 -treated, re-oxidized and proton-exchanged RbCa 2 Nb 3 O 10-x N x with up to 32 µmol/h steady-state hydrogen evolution rate. The short re-oxidation process for the black-coloured layered perovskite sample resulted in a nearly 300% increase in photocatalytic efficiency compared to its black-coloured counterpart under similar reaction conditions.  In order to explain this behaviour, XPS was performed on RbCa 2 Nb 3 O 10 to unveil the defects present in the layered perovskite samples (Fig. 5). High-resolution XP spectra of Nb 3d region are very complex and reveal three different signals, which are split in 3d 3/2 and 3d 5/2 , respectively. The most prominent one ranges from 206.1 eV to 206.5 eV (all values for 3d 5/2 ); we ascribe this 3d 5/2 signal to oxidic Nb(V) species of mixed oxides. It is commonly observed that binding energies of ternary, or quaternary metal oxides are slightly lower in binding energies compared to the simple binary oxides [19]. Reported binding energy values for ternary Nb oxides are only slightly higher than the energies observed herein [19,21]. Thus, we assign the second signal in the high-resolution Nb spectra centred around 207.0 eV more likely to be due to Nb(IV) defects, which fits close to reported values for CsNbTeO 6 [22]. Additionally, there is a third, minor signal in the range from 204.7 eV to 204.9 eV, which is similar to binding energies reported for NbON [23][24][25].  The O1s spectra also show three different signals: the most intense one is located at 529.3 eV-529.6 eV and can be attributed to lattice oxygen, this value is slightly lower than reported for Nb 2 O 5 [23,26]. The second peak (530.2 eV-530.4 eV) can be ascribed to defective oxygenthat is oxygen located near defects-which is commonly observed for metal oxides. The low intensity of the defective O1s peak confirms the Nb peak assignment, resulting in less Nb defects than oxidic Nb(V) oxide species. The third peak in the O1s spectrum can be assigned to carbon-oxygen species belonging to adventitious carbon.
Comparing the spectra before and after re-oxidation, two details can be concluded. First, the ammonia treatment does generate oxygen and niobium lattice defects. However, both are getting reduced in number by the re-oxidation process, explaining the lowered baseline in the absorption spectra. However, on the other hand the intensity for the NbON assignment is also reduced, indicating the release of nitrogen from the lattice. From survey scans, the amounts of nitrogen are 3.2 and 1.8 at.−%, respectively.
In order to perform nitrogen doping but reducing the defect formation in one step, addition of alkali metal carbonates to the ammonolysis step was suggested in literature for KCa 2 Nb 3 O 10 [15]. Here, we adapt this approach for all our Dion-Jacobson layered perovskite niobates.
As can be seen in Fig. 1, the addition of carbonates to the ammonolysis has in most cases no influence on the crystal structure of the materials, too. The exceptions are CsCa 2 Nb 3 O 10 and KSr 2 Nb 3 O 10 . The ammonia treatment in the presence of their carbonates leads to structural changes and a different catalyst altogether.
Interestingly, however, the absorption spectra indicate much less defects generated upon carbonate additions. Figure 6 shows the absorption spectra after ammonolysis in the presence of carbonates. They appear very similar to the absorption spectra given in Fig. 3 for the re-oxidised materials, as can be seen also in the photographs of the samples ( Figure S7) having yellowish colors rather than black-grey. Only CsCa 2 Nb 3 O 10 and KCa 2 Nb 3 O 10 appear very different. Since CsCa 2 Nb 3 O 10 also exhibited structural changes upon ammonolysis, as observed in XRD patterns, these spectra cannot be compared. In the case of KCa 2 Nb 3 O 10 , the shift in absorption into the visible light region is very prominent and far more pronounced compared to ammonolysis without carbonate and re-oxidation, however this cannot be converted into photocatalytic activity. SEM images of the respective samples show clear changes in the morphologies of CsCa 2 Nb 3 O 10 and KCa 2 Nb 3 O 10 in line with the XRD observations, besides no distinct morphology changes upon ammonolysis in the presence of carbonates ( Figure S8) are observed for the other four materials.
Comparing the high resolution XP spectra after ammonolysis in the presence of carbonates for RbCa 2 Nb 3 O 10 ( Fig. 7) with XPS data shown in Fig. 5, it can clearly be seen Fig. 6 Absorption spectra of bulk AB 2 Nb 3 O 10 (black) compared to the respective spectra after nitrogen doping via ammonolysis in the presence of A 2 CO 3 (AB-NH 3 -CO 3 , green) that the one step approach leads to lower amounts of lattice defects compared to sole ammonia treatment, hence the yellow colour. Moreover, nevertheless, the amount of NbON species is also lower, thus a trade-off between defect generation and nitrogen incorporation for the band gap reduction has to be considered.
The hydrogen production experiments in aqueous suspension with methanol as a sacrificial agent (electron donor) of the H-exchanged, N-doped (in the presence of carbonate) compounds reveal that hydrogen can be generated after proton exchange and Rh decoration (Fig. 8). The experiments were performed under simulated sunlight (AM1.5G) again. The hydrogen evolution rate trajectories are shown in Figure S9. The highest activity can be observed again for HCa 2 Nb 3 O 10-x N x derived from RbCa 2 Nb 3 O 10-x N x , comparable to a sample after re-oxidation. Nevertheless, the activity is even higher now, with up to 40 µmol/h steady-state hydrogen production rate under solar light illumination. This value is 100% higher than for the second best material, HSr 2 Nb 3 O 10-x N x derived from CsSr 2 Nb 3 O 10-x N x , and even higher than the best hydrogen production rates in Fig. 4. Therefore, ammonolysis in the presence of A-cation carbonates can be considered a viable strategy for complex niobates to improve visible light absorption by nitrogen doping with reduced defect formation in one step, additionally preserving photocatalytic activity for hydrogen generation under sunlight illumination.

Conclusion
Defects generated upon ammonolysis of layered niobates AB 2 Nb 3 O 10 (A = Cs, Rb, K; B = Ca, Sr) can be reduced by two strategies, either subsequent re-oxidation with oxygen gas for several minutes, or by addition of A-cation carbonates to the pristine oxides during ammonolysis. Both pathways lead to band gap reductions into the visible light region for all six investigated layered niobates, and photocatalytic hydrogen generation upon sunlight illumination. The pathway using carbonates however leads to slightly higher activities, while at the same time enables a one-step preparation