Introduction

Hydrogen produced from water using sunlight is an attractive candidate as a next-generation clean fuel. For this reason, the photoelectrochemical (PEC) water splitting reaction has been the subject of significant research following the initial report of the Honda-Fujishima effect1. In order to utilize sunlight effectively during PEC water splitting, various oxynitrides and oxysulfides capable of absorbing a wide range of visible wavelengths have been developed2. At present, Ta-based (oxy)nitrides such as TaON, Ta3N5, LaMg1/3Ta2/3O2N, and BaTaO2N are being intensively investigated both in powder suspensions and PEC systems, and have exhibited relatively high efficiencies3,4,5,6. In contrast, studies regarding Nb-based materials remain limited, even though such compounds actually have longer absorption edge wavelengths than the analogous Ta-based materials. It is also notable that Nb is potentially preferable to Ta both with regard to the lower cost and ready supply of the former, although the efficiencies reported for Nb-based catalysts remain very low7,8.

Among Nb-based oxynitrides, SrNbO2N is a potential candidate for photoanodes in PEC cells. SrNbO2N has a perovskite-type structure and an absorption edge at approximately 690 nm9. It has been reported that SrNbO2N modified with a CoPi cocatalyst functions as a photoanode, generating a current density of 1.5 mA cm−2 at 1.23 V vs. a reversible hydrogen electrode (RHE)10. Typically, SrNbO2N is synthesized from the layered perovskite-type compound Sr2Nb2O7 or from amorphous oxide precursors by heating these materials under an ammonia flow. Sr5Nb4O15, Sr4Nb2O9 and SrNb2O6 are all possible oxide precursors, although the Sr/Nb ratios in these materials are not the ideal value of unity, thus requiring an additional washing step to remove excess Sr or Nb species after nitridation. Recently, Sr1-xNbO3, having a slightly distorted perovskite-type structure, has been reported as a photocatalyst with an absorption edge of 650 nm, and has been found to work as a photoanode11, albeit with very limited PEC performance. Very recently, our own group determined that ANbO2N compounds (A = Ba, Sr) synthesized from ANbO3 (A = Ba, Sr) function as photoanodes in conjunction with a loaded cocatalyst12. As an example, SrNbO2N synthesized from SrNbO3 has exhibited a current density of 0.4 mA cm−2 at 1.2 V vs. RHE under simulated sunlight irradiation. The perovskite-type compound SrNbO3 can therefore be used as the oxide precursor for the synthesis of this catalyst.

One of the key factors associated with improving Nb-based oxynitride photocatalysts is control over the nitridation conditions. Although the nitridation conditions for the oxynitrides have already been largely optimized in each material, our understanding of the nitridation mechanism and the relationship between the nitridation processes and PEC properties remains very limited. In the case of Sr2Nb2O7, the overall nitridation reaction can be summarized according to the theoretical equation:

$${{\rm{Sr}}}_{2}{{\rm{Nb}}}_{2}{{\rm{O}}}_{7}+2{{\rm{NH}}}_{3}\to 2{{\rm{SrNbO}}}_{2}{\rm{N}}+3{{\rm{H}}}_{2}{\rm{O}}$$
(1)

It is considered that the active species during nitridation is NH, NH2 or N radical species13. In this process, at the same time, the decomposition of ammonia partly produces hydrogen as:

$$2{{\rm{NH}}}_{3}\to {{\rm{N}}}_{2}+3{{\rm{H}}}_{2}$$
(2)

meaning that a reductive atmosphere is present during nitridation14. Nitridation typically changes the crystal structure of the oxide due to the exchange of O2− with N3−. As an example, Sr2Nb2O7, initially having a layered perovskite structure, transitions to the perovskite-type material SrNbO2N. On the other hand, SrNbO3 is believed to undergo nitridation to SrNbO2N via the reaction:

$${{\rm{SrNbO}}}_{3}+{{\rm{NH}}}_{3}\to {{\rm{SrNbO}}}_{2}{\rm{N}}+{{\rm{H}}}_{2}{\rm{O}}+1/2{{\rm{H}}}_{2}$$
(3)

In this case, both the oxide precursor and the resulting oxynitride have a perovskite-type structure. Another feature of this process is that Nb4+ is evidently oxidized to Nb5+ in spite of the reductive atmosphere during the nitridation. Although the total nitridation reactions from Sr2Nb2O7 or SrNbO3 to SrNbO2N apparently seems to be relatively simple, as summarized in Eqs (1) and (3), the details of these nitridation processes are not yet known.

In the current work, nitridation of Sr2Nb2O7 and SrNbO3 were investigated for various reaction times. The results demonstrate that the nitridation mechanisms for these two oxide precursors are greatly different, resulting in different optical, morphological and PEC properties for the end products.

Experimental

Preparation of SrNbO2N

Sr2Nb2O7 was synthesized using a flux method10. SrCO3, Nb2O5, and RbCl (as a flux) were mixed in a molar ratio of Sr:Nb:Rb = 1:1:5 and then heated in an alumina crucible for 5 h at 1423 K in air. The mixture was subsequently cooled to 1073 K at a rate of 1 K min−1 and then allowed to cool to room temperature, followed by washing using a copious amount of distilled water.

To obtain SrNbO3, Sr5Nb4O15 was first synthesized by calcining a mixture of SrCO3 and Nb2O5 (at a Sr:Nb molar ratio of 5:4) at 1423 K for 48 h in air. The resulting Sr5Nb4O15 particles were mixed with Nb metal powder at a Sr5Nb4O15:Nb molar ratio of 1:1 in an agate mortar, and then calcined under an Ar flow of 100 mL min−1 at 1773 K for 20 h to synthesize SrNbO311,15. During this calcination, the sample was wrapped in a 5 cm square sheet of Nb foil and placed on an alumina boat.

In a typical nitridation, approximately 0.5 g of the oxide precursor was placed on an alumina boat, heated at 10 K min−1 to 1173 K and then held for 1, 5, 10, 20, 30, or 40 h at 1173 K under a 250 mL min−1 NH3 flow. Finally, samples were allowed to cool naturally to room temperature.

Preparation of photoelectrodes

Photoelectrodes composed of SrNbO2N were prepared by a particle transfer method16. The photocatalyst powder was first suspended in 2-propanol and then dropped onto a glass plate, after which the solution was dried at room temperature. A layer of Nb (approximately 300 nm thick) was then deposited on the dried particles using radio-frequency (RF) magnetron sputtering, followed by the sputtering of a thick Ti layer (>5 μm). The resulting film incorporating the photocatalyst powder was transferred to another glass plate using double sided carbon tape and ultrasonicated in distilled water to remove particles that were not firmly held by the metal film.

Photoelectrochemical measurements

Current-potential curves for the SrNbO2N photoanodes were acquired using a typical three-electrode configuration under intermittent illumination with simulated sunlight provided by a Xe lamp equipped with an air mass (AM) 1.5G filter (San’ei Denki, XES-40S2-CE). Ag/AgCl in a saturated KCl solution and Pt wire were employed as the reference and counter electrodes, respectively. The electrolyte was a 0.2 M aqueous sodium phosphate solution adjusted to a pH of 13 by NaOH addition. In all cases, a CoPi cocatalyst was deposited by electrodeposition at 1.7 V vs. RHE for 200 s17.

Characterization

Samples were characterized using X-ray diffraction (XRD; Rigaku, RINT-Ultima III), UV-vis diffuse reflectance spectroscopy (DRS; JASCO, V-670DS), field-emission scanning electron microscopy (FE-SEM; Hitachi, S-4700, and JEOL, JSM-7001F) and field-emission transmission electron microscopy (FE-TEM; JEOL, JEM-2800). The DRS reflectance (R) data were converted to the Kubelka-Munk function using the equation f(R) = (1-R)2/(2 R). The relative weight (WR) of each specimen was determined from the equation WR = (Wa−W0)/(Wb−W0), where Wa, Wb, and W0 are the total weight of the powder and the alumina boat after nitridation, the total weight of the powder and alumina boat before nitridation, and the weight of the alumina boat, respectively. Elemental analysis was performed using inductively coupled plasma atomic emission spectroscopy (ICP-AES; Shimadzu, ICPS-8100) and oxygen/nitrogen combustion analysis (ON analysis; Horiba, EMGA-620W/C). Nitrogen adsorption isotherms for the samples were acquired at 77 K with a BELSORP-mini instrument (MicrotracBEL). The relative surface areas of the samples were determined using the Brunauer-Emmett-Teller (BET) model.

Results and Discussion

Structural analysis

The Sr2Nb2O7 oxide precursor was synthesized by a flux method and then subsequently nitrided for various nitridation time to investigate the process that generates SrNbO2N. Figure 1A shows the XRD patterns obtained from the Sr2Nb2O7 oxide precursor and the nitridation product SrNbO2N. It is evident that Sr2Nb2O7 with a layered perovskite-type structure was obtained without any impurity phases. As the nitridation time was increased, the diffraction peaks due to the oxide precursor became less intense, while peaks attributed to a perovskite-type SrNbO2N phase appeared. After 10 h, peaks from the oxide precursor completely disappeared. As shown in Table S1, the full width at half maximum (FWHM) values of the (110) diffraction peaks generated by the SrNbO2N did not change after 5 h. This result indicates that the crystallinity was not improved by prolonging the nitridation for more than 5 h.

Figure 1
figure 1

XRD patterns for SrNbO2N obtained from nitridation of (A) Sr2Nb2O7, and (B) SrNbO3. Legend: (a) oxide precursor and (b–f) nitrided samples obtained after nitridation for (b) 1, (c) 5, (d) 10, (e) 20, and (f) 30 h. Closed triangles (▼) and closed circles (●) indicate Sr5Nb4O15 and Sr7Nb6O21, respectively. The inset in (B) is a magnification of the region indicated by the box.

Full size image

Figure 1B presents XRD patterns for the SrNbO3 oxide precursor and the nitridation product SrNbO2N. SrNbO3 was synthesized by a solid state reaction under an Ar flow, and a small amount of a Sr7Nb6O21 phase was evidently present as an impurity phase. It has been reported that stoichiometric SrNbO3 is not readily obtained, while Sr-poor compositions having the formula Sr1-xNbO3 are relatively stable11. Moreover, ICP-AES data demonstrated that the Sr/Nb ratio in the oxide precursor was 0.96 (Table S2), most likely due to the vaporization of Sr species during calcination at 1773 K. Therefore, it is probable that the synthesized oxide was a mixture of Sr1-xNbO3 and a small amount of Sr7Nb6O21. Since SrNbO3 and SrNbO2N both have a perovskite-type structure and their lattice constants are very similar (the lattice mismatch is only 0.7%), it was anticipated that the nitridation would proceed without any change in the crystal structure. Surprisingly, it was found that, following the initial stage of nitridation, a Sr5Nb4O15 phase appeared and then gradually disappeared as SrNbO2N was formed. In addition, the Sr7Nb6O21 phase disappeared after 1 h of nitridation. As shown in Fig. S1, the diffraction peaks attributed to a perovskite-type structure were shifted to lower angles as the nitridation progressed. It is notable that the diffraction peaks associated with SrNbO3 and SrNbO2N could not be separated, indicating that a solid solution of SrNbO3 and SrNbO2N was partly formed in this stage. Figure 2 indicates d-values of (211) diffraction peaks. In the case of SrNbO2N nitrided from Sr2Nb2O7, d-values became constant after nitridation for 5 h. On the other hand, in the case of SrNbO3 series, d-values first became larger then gradually reduced after 10 h. It is expected that the introduction of nitrogen increase the lattice constant since the ionic radius of N3− is larger than that of O2−. Another possibility is the introduction of oxygen vacancies which were produced during nitridation. This effect will be discussed in detail later. The FWHM values for the (110) diffraction peaks resulting from a perovskite structure initially increased and then decreased with increasing nitridation time, although these FWHM values remained larger than those from Sr2Nb2O7 as shown in Table S1. These data demonstrate that the SrNbO2N synthesized from SrNbO3 had a lower degree of crystallinity than that obtained from Sr2Nb2O7. Consequently, nitridation of the perovskite-type material SrNbO3 to the perovskite-type product SrNbO2N proceeded to some extent, but with a phase separation during the initial stage of nitridation.

Figure 2
figure 2

d-values of (211) peaks for SrNbO2N nitrided from different oxide precursors. Closed circles and open diamonds indicate Sr2Nb2O7 and SrNbO3 series, respectively.

Full size image

Morphological properties

The SEM images shown in Fig. 3 reveal that the Sr2Nb2O7 particles were composed of bundles of columnar crystals. The Sr2Nb2O7 particles were typically several micrometers in size but some were also much larger or smaller as shown in Fig. S2. The crystals gradually became more porous as the nitridation time was prolonged (See Fig. 3(b–d)). This trend has also been observed for (oxy)nitrides such as TaON, Ta3N5, and LaTiO2N18,19,20. In most instances of nitridation of oxides to form (oxy)nitrides, a lattice shrinkage occurs. In the case of nitridation of the layered perovskite-type compound Sr2Nb2O7, 2 mol of nitrogen is introduced while 3 mol of oxygen is extracted, resulting in a structural change from a layered perovskite to a perovskite. Since a layered perovskite structure has a lower density than a cubic perovskite, the lattice will shrink. Based on the observation that the secondary particles retained their original sizes after nitridation, it is most likely that porous oxynitrides were obtained after nitridation due to a lattice shrinkage. This porous structure is also supported by the nitrogen adsorption analysis data shown in Table S3, which indicates that the surface area was doubled following nitridation.

Figure 3
figure 3

SEM images of (a) original Sr2Nb2O7, and products obtained after nitridation for (b) 1, (c) 5, and (d) 20 h, (e) original SrNbO3, and products obtained after nitridation for (f) 1, (g) 5, (h) 20 h, and (i) 40 h. The white arrows in (f) indicate parallel cracks.

Full size image

In the case of SrNbO3, large particles with a size of more than 5 μm were observed, and the average particle size was larger than that for Sr2Nb2O7. As shown in Table S3, the surface area for SrNbO3 was much less than that for Sr2Nb2O7 due to the presence of smoother surfaces and larger particles. One distinct difference from Sr2Nb2O7 particles was observed in the sample nitrided for 1 h. The SEM image in Fig. 3(f) demonstrates that parallel cracks were present in the sample nitrided for 1 h, probably suggesting the formation of Sr5Nb4O15 with a layered perovskite related structure. As the nitridation time increased, the surface morphology became porous and cracked, resulting in a steady increase in surface area, as shown in Table S3. In contrast to Sr2Nb2O7, nitridation from SrNbO3 did not involve a change in the crystal structure and so the crystals would be expected to expand as nitrogen was introduced. This effect is believed to have expanded the particles, leading to the formation of cracks and a phase separation.

Figure 4A shows the cross-sectional SEM images of SrNbO2N nitrided from SrNbO3 with various nitridation time. In this measurement, SrNbO2N/Nb/Ti photoelectrodes prepared by a particle transfer method were utilized for suppressing charge up during SEM observation. After nitridation for 1 h, some cracks and pores near surface appeared. In the case of nitridation for 20 h, there are obviously two regions like a “core-shell” structure, in which a “core” has smaller pores while a “shell” has larger pores. Even after nitridation for 40 h, some particles remained a “core-shell” structure while others completely became porous. The cross-sectional TEM image more precisely presented porous structure as shown in Fig. 4B. It was revealed that even “core” was porous although the pores were smaller than pores in “shell.” Although EDS analysis shows both “core” and “shell” contain certain amount of nitrogen, it is notable that the “shell” in Fig. 5 has larger N/O ratio of 0.41 than 0.28 in the “core,” indicating that the “core” was less nitrided.

Figure 4
figure 4

(A) Cross sectional SEM images of (a) SrNbO3/Nb/Ti and subsequently nitrided SrNbO2N/Nb/Ti nitrided for (b) 1 h, (c) 20 h and (d) 40 h. The white arrows in (b) indicate cracks. (B) STEM-DF image of SrNbO2N nitrided from SrNbO3 for 20 h. The N/O ratio by EDS in area (i) and (ii) is 0.28 and 0.41, respectively.

Full size image
Figure 5
figure 5

Diffuse reflectance spectra for SrNbO2N obtained by nitridation of (A) Sr2Nb2O7 and (B) SrNbO3. Legend: (a) oxide precursor and (b–f) samples obtained after nitridation times of (b) 1, (c) 5, (d) 10, (e) 20, and (f) 30 h.

Full size image

Optical properties

Figure 5 shows the diffuse reflectance (DR) spectra of the oxide precursors and the SrNbO2N samples obtained at different nitridation times. The original Sr2Nb2O7 particles had a white color and were only able to absorb light up to 320 nm. However, the absorption edge clearly shifted to longer wavelengths as nitridation proceeded. After nitridation for 1 h, the absorption edge was over 600 nm and reached 690 nm after 5 h. Further nitridation did not affect the optical properties. Absorption beyond the absorption edge as observed here is usually attributed to the formation of reduced Nb species or anion defects such as anion vacancies or ON anti-site defects.

SrNbO3 exhibited strong absorption up to 650 nm and had a blood red coloration, both of which are consistent with previous results11,12. DR spectra of samples nitrided for less than 10 h did not show clear absorption edges and contained intense background signals over 700 nm. After nitridation for 20 h, an absorption edge appeared at 690 nm, although this edge was not well defined. It is notable that the SrNbO3 particles showed relatively minimal absorption beyond the absorption edge region in spite of the reduced valency of the Nb in this material, indicating that reduced Nb species were not responsible for the absorption above 700 nm. Therefore, it is probable that this absorption can be primarily attributed to anion defects. Moreover, the absorption above 700 nm decreased with increasing nitridation time more than 1 h, indicating that the number of anion defects first increased and then gradually decreased with progression of the nitridation.

Relative weight analysis

Nitridation of the oxides to the oxynitride was further examined by following trends in the relative weight of the nitrided particles compared to the oxide precursors, using the equation provided in the Experimental section. This method can readily track the extent of nitridation. In the case of the Sr2Nb2O7 series, it was expected that the relative weight would decrease as a function of the nitridation time, reaching a value of 0.958 at complete nitridation. As shown in Fig. 6, the relative weight rapidly decreased over the initial 5 h, then further decreased up to 10 h, resulting in a plateau value close to the expected value. Thus, these relative weight data provide clear evidence for the progression of the expected nitridation. A similar trend has also been reported for the nitridation of La2Ti2O7-based oxides21. The elemental analysis data in Table S2 also support the above discussions. These values show that the O/Nb ratio steadily decreased while the N/Nb ratio increased, indicating that the introduction of nitrogen and the extraction of oxygen occurred simultaneously based on charge compensation.

Figure 6
figure 6

Relative weight of oxide precursors and nitrided samples as a function of nitridation time. Closed circles (●) and open diamonds (◇) indicate SrNbO2N synthesized from Sr2Nb2O7 and SrNbO3, respectively. Dashed lines indicate the expected value following complete nitridation.

Full size image

Interestingly, a completely different trend was observed in the case of the SrNbO3 series. SrNbO2N has a formula weight (226.53 g mol−1) slightly less than that of SrNbO3 (228.53 g mol−1), and so it was anticipated that a similar trend to Sr2Nb2O7 series in the relative weight would be observed, with a gradual decrease in parallel with the anion exchange. However, as shown in Fig. 6, the relative weight initially increased, followed by a slow decrease. The elemental analysis data in Table S2 also show that the N/Nb ratio increased whereas the O/Nb ratio did not change from that in the oxide precursor after nitridation for 1 h, indicative of an increase in the total anion amount. One unique characteristic of SrNbO3 is that the Nb4+ must be oxidized to Nb5+ to form SrNbO2N during nitridation. In order to compensate for this, the Nb species must be oxidized during this process. Therefore, it appears that nitrogen was introduced during the initial stage of nitridation while oxygen ions remained in the crystals, followed by the subsequent extraction of the oxygen. After nitridation for 1 h, the relative weight became 1.016. When assumed that the increase of relative weight is only due to introduction of N species, the expected composition formula is SrNbO3N0.26. The XRD analysis as shown in Fig. 1 presents that Sr5Nb4O15 has formed at this stage. Therefore, it is more likely that the obtained product was a mixture of Sr5Nb4O15-xNx and Sr1-yNbO3-zNz. To simplify this, considering the mixture is Sr5Nb4O15 and Sr1-xNbO3-yNy, it is roughly estimated that 30–40% of SrNbO3 has changed into Sr5Nb4O15.

Even after nitridation for 30 h, the relative weight did not reach the expected value, indicating that the nitridation of SrNbO3 did not proceed to completion. One possible reason is the larger particle size for the SrNbO3 oxide precursor, although the different crystal structures of SrNbO3 and Sr2Nb2O7 might also be responsible.

The effect of temperature during the nitridation of SrNbO3 was also investigated. As shown in Fig. S3, even at 1123 and 1223 K, similar trends were observed in the relative weight data. The application of higher temperatures during nitridation seems to accelerate the process, although the effect was limited.

Photoelectrochemical performance

PEC water splitting was performed using both the oxides and the oxynitrides. Figure 7A shows current-potential curves obtained from photoanodes made of Sr2Nb2O7 and the nitrided product SrNbO2N, modified using a CoPi cocatalyst. Under simulated sunlight, Sr2Nb2O7 did not show a clear photoresponse because of the minimal photon flux in the UV region. However, the photocurrent at a positive potential increased as the nitridation time was prolonged up to 10 h, reaching 0.77 mA cm−2 at 1.2 V vs. RHE, and then dropping after 20 h. The onset potential did not exhibit any clear dependence on the nitridation time. Moreover, since it was confirmed by the relative weight analysis that a nitridation time of 10 h was sufficient for Sr2Nb2O7, it is likely that the photocurrent reached maximum after completion of crystal structural change and that excess nitridation led to a degradation of the surface of the SrNbO2N. It is thus found that the PEC properties were relevant to the extent of nitridation proceeded.

Figure 7
figure 7

Current-potential curves for CoPi/SrNbO2N/Nb/Ti photoelectrodes synthesized from (A) Sr2Nb2O7 and (B) SrNbO3. Legend: (a) oxides, and samples nitrided for (b) 1, (c) 5, (d) 10, (e) 20, and (f) 30 h. Data were acquired under intermittent simulated sunlight (AM 1.5G) in an aqueous 0.2 M Na3PO4 solution adjusted to pH 13 by NaOH addition. Potentials were swept from positive to negative at a scan rate of 10 mV s−1.

Full size image

As shown in Fig. 7B, an almost negligible photocurrent was observed when using CoPi/SrNbO3/Nb/Ti despite the SrNbO3 was able to absorb light up to 650 nm. This is consistent with our previous results12. A definite photoanodic current was generated even after nitridation for only 1 h. As the formation of SrNbO2N proceeded, the photocurrents steadily increased to reach a maximum of 0.75 mA cm−2 at 1.2 V vs. RHE after nitridation for 20 h, which was the comparable photocurrent to one obtained in Sr2Nb2O7 series. This value is almost twice that previously reported for SrNbO2N synthesized from SrNbO3. It is consistent that the optimum nitridation time was longer than that of Sr2Nb2O7 due to the slow nitridation process and a larger particle size. It is worth noting that SrNbO2N made from SrNbO3 was less crystalline and had a larger concentration of defects compared to the material prepared from Sr2Nb2O7 but exhibited a similar photocurrent.

Proposed nitridation processes

Based on obtained results, we propose the nitridation processes of Sr2Nb2O7 and SrNbO3 as follows. In both cases, the nitridation processes via oxygen vacancy were considered. When assumed that active nitrogen species are NH2, for example, the nitridation process of Sr2Nb2O7 could be explained with equations as:

$${{\rm{O}}}_{{\rm{O}}}+{{\rm{H}}}_{2}\to {{{\rm{V}}}_{{\rm{O}}}}^{\cdot \cdot }+2{\rm{e}}^{\prime} +{{\rm{H}}}_{2}{\rm{O}}$$
(4)
$${{\rm{NH}}}_{2}+3{\rm{e}}^{\prime} +{{{\rm{V}}}_{{\rm{O}}}}^{\cdot \cdot }\to {{\rm{N}}}_{{\rm{O}}}^{\prime} +\,{{\rm{H}}}_{2}$$
(5)
$$2{{\rm{N}}}_{{\rm{O}}}^{\prime} +{{{\rm{V}}}_{{\rm{O}}}}^{\cdot \cdot }\to 2{{\rm{N}}}_{{\rm{N}}}$$
(6)

In the first step, oxygen atoms desorb and oxygen vacancies (VO) form (Eq. (4)). Next, nitrogen atoms are introduced to VO and nitrogen anti sites (NO) are formed (Eq. (5)). Finally, oxygen vacancies disappear by changing the crystal structure (Eq. (6)). In total, 3 mol of oxygen atoms are exchanged by 2 mol of nitrogen. Desorbed O2 reacts with H2 to produce H2O. It is notable that overall nitridation process itself is not an oxidation-reduction reaction. Although the authors did not fully understand the actual active species of nitrogen, above equations indicate that the formation of anion vacancies is inevitable when nitridation process includes a crystal structural change.

As shown in Fig. 8A, the layered perovskite-type material Sr2Nb2O7 has interlayers between the NbO6 layers for nitrogen to penetrate into the crystal. It has been reported that, in the case of an A2B2O7 oxide, nitrogen is introduced through spaces among layers by a so-called ‘zipper’ like mechanism, in which introduced nitrogen connects the interlayers to transform to the perovskite type structure22. Therefore, nitrogen could be exchanged with oxygen concurrently. Moreover, this anion exchange causes a lattice shrinkage. Considering the apparent particle size does not change during nitridation, it is most likely that this process leads the porous structures.

Figure 8
figure 8

Schematic images of nitridation process of (Left) before and (Right) during nitridation of (A) a layered perovskite Sr2Nb2O7 and (B) a perovskite SrNbO3.

Full size image

Contrary, in the case of nitridation of SrNbO3, Eq. (6) should be exchanged by the following equation:

$${{\rm{N}}}_{{\rm{O}}}^{\prime} \to {{\rm{N}}}_{{\rm{N}}}+{\rm{e}}^{\prime} $$
(6&x02032;)

This process does not contain any crystal structural changes. Furthermore, this total nitridation process is an oxidation reaction. As a result, reduced Nb4+ should be oxidized to Nb5+ after nitridation. When step (5) or (6′) are slow, oxygen vacancies stimulate in the crystals. It is reasonable that the oxidation of Nb4+ was challenging under a reductive atmosphere during nitridation. Therefore, it is most likely that considerable number of oxygen vacancies remain in SrNbO3 and obtained products, leading to the large absorption over 700 nm (Fig. 5B(b–f)). Since introduction of oxygen vacancies and nitrogen species both expand the lattice constant (the lattice mismatch of oxide and resulting oxynitride is 0.7%), this causes the crack formation (Fig. 3(f)) to solve the stimulated distortion as shown in Fig. 8B. Our group has reported that A site poor perovskite Ba1-xNbO3-δ was nitrided with maintaining the perovskite type structure12. In this case, the lattice mismatch reduced to only 0.1% at Ba0.84NbO3-δ. Although 0.7% of the lattice mismatch in SrNbO3 system is relatively small in terms of solid solution formation, it was seen that the small lattice mismatch played the critical role in this nitridation process. Formation of cracks requires the increase of the total amount of anion species, resulting in a phase separation of Sr5Nb4O15 and an increase of the relative weight at the initial stage of nitridation (Fig. 6). Additionally, some Nb4+ species are supposed to be oxidized to Nb5+ in this step. As nitridation proceeded, both SrNbO3 and Sr5Nb4O15 are nitrided to the final product, SrNbO2N.

Judging from the lattice constants, in the nitridation of SrNbO3 to SrNbO2N, a lattice shrinkage does not seem to proceed. It is thus likely that the porous structure was inevitably obtained due to exchanging oxygen and nitrogen species regardless of the extent of a lattice mismatch. In contrast to Sr2Nb2O7 oxides, a perovskite-type oxide SrNbO3 does not have interlayers to accelerate the diffusion of nitrogen sources. Therefore, it is more likely that the diffusion of nitrogen is much slower in this material than in Sr2Nb2O7. Although the formation of Sr5Nb4O15 partly contribute to the anion diffusion, SrNbO3 requires more severe conditions, such as the longer nitridation time and higher nitridation temperature than Sr2Nb2O7. These differences in the nitridation process are primarily the result of variations in the crystal structure and the valency of the cations in the oxide precursors. The resulting photocurrents were also affected by the nitridation processes. Photocurrent reached maximum after 10 h in the case of Sr2Nb2O7 series while 20 h of nitridation was optimum for SrNbO3 series. In the case of nitridation of SrNbO3, nitridation speed is dominated by a diffusion of nitrogen species. Therefore, particle size should play a critical role in nitridation time and resulting PEC properties. Since, in this paper, SrNbO3 was synthesized at relatively high temperature, 1773 K, it was challenging to reduce a particle size. Further development of synthesis procedure to control a particle size should improve PEC properties. Moreover, in both nitridation processes, anion vacancies inevitably formed, and this probably lead the limited photocurrent and positive onset potentials. It is therefore a key to develop a nitridation process to reduce the anion vacancies for realizing the solar hydrogen production using oxynitride materials.

Conclusion

In conclusion, the nitridation processes for two oxide precursors, the layered perovskite-type Sr2Nb2O7 and the perovskite-type SrNbO3, to produce the perovskite-type material SrNbO2N were examined. The relationship between the PEC properties of the products and the initial oxide were also investigated. Nitridation without any change in crystal structure was partly achieved in nitridation of SrNbO3, although formation of an impurity phase was detected. The data also suggest that nitridation from SrNbO3 proceeds via a different pathway from Sr2Nb2O7, with variations in lattice expansion and oxidation of Nb cations. Nitridation models based on the crystal structure were presented, in which nitridation processes triggered by formation of oxygen vacancies were considered. The photocurrent obtained from the products was correlated with the extent of nitridation determined from the relative weight analysis. This study demonstrates the importance of understanding the nitridation mechanism for the oxide precursor, focusing on the crystal structure. The knowledge obtained in this study should be widely applicable to other oxynitrides.