Thanh Tam Thi
Tran
and
Jeongsuk
Seo
*
Department of Chemistry, College of Natural Sciences, Chonnam National University, 77 Yongbong-ro, Buk-gu, Gwangju 61186, Republic of Korea. E-mail: j_seo@chonnam.ac.kr
First published on 17th June 2024
Perovskite-type LaTiO2N with a bandgap energy of 2.1 eV is a promising semiconductor for solar (sea)water splitting to produce renewable hydrogen. However, the synthesis of an oxynitride with low defect density to achieve strong charge separation for high solar-to-hydrogen energy conversion is still challenging. Herein, we report less-defective, two-dimensional (2D) LaTiO2N crystals prepared from layered perovskite BaLa4Ti4O15 and their highly improved photoelectrochemical (PEC) seawater-splitting activity at neutral pH. The (111)-type layered perovskite BaLa4Ti4O15 has structural similarity to LaTiO2N and excess Ba species with a strong basicity, thus guaranteeing complete and fast nitridation accompanied by suppressed Ti4+ reduction. The prepared oxynitride, capable of absorbing visible light up to 610 nm, was highly crystalline and less-defective in both bulk and surface states, which is favorable for efficient photoreactions. Oxynitrides were also prepared from other layered perovskite oxides, namely La2Ti2O7 and La4Ti3O12, for comparison. Photoanodes using LaTiO2N were fabricated by spin-coating and then employed for driving PEC seawater splitting including chloride oxidation at neutral pH. The less-defective LaTiO2N exhibited a photocurrent density of approximately 2.4 mA cm−2 at 1.23 VRHE (4.9 mA cm−2 at 1.36 VRHE) in 0.5 M NaCl electrolyte at pH 6.4 under AM 1.5G simulated sunlight, indicating hitherto unreported remarkable activity. This work presents the first demonstration of the complete conversion of layered perovskite BaLa4Ti4O15 into perovskite LaTiO2N as well as the types of starting oxides that reduce the defect density of the resulting oxynitride, thereby improving the hydrogen evolution rate at neutral pH.
The synthesis of LaTiO2N, commonly using La2Ti2O7 with a stoichiometric La/Ti ratio of unity, results in high defect density, including reduced Ti3+, excess O2−, and deficient N3− species, because highly electronegative Ti4+ in the oxide is readily reduced during nitridation in a reducing NH3 flow.18,19 The presence of defects, acting as the recombination sites of photogenerated charge carriers, leads to a reduction in the crystallinity of LaTiO2N and surface reconstruction of the amorphous oxide layer, thus hindering the water-splitting activity of the oxynitride under visible-light irradiation. Several strategies to reduce the defect concentration, e.g., doping foreign elements, treating surface defects, and reducing the grain boundaries of LaTiO2N, have been reported for enhancing the water-splitting activity of the oxynitride. Lin et al. attempted to reduce the defect density of LaTiO2N by Mg doping, which improved charge separation and photocatalytic activity with an apparent quantum efficiency as high as 13.02% at 420 ± 20 nm for the oxygen evolution reaction (OER).14 Akiyama et al. reported a photoactive LaTiO2N powder prepared by the nitridation of La2Ti2O7 and the uniform etching of its surface defects by acid treatment in an aqueous poly(4-styrene sulfonic acid) solution.20 An electrode made of the modified LaTiO2N powder served as a highly active photoanode for water splitting, producing a photocurrent of 8.9 mA cm−2 at 1.23 VRHE in 0.1 M NaOH electrolyte at pH 13 under AM 1.5G simulated irradiation. Feng et al. reported crystalline LaTiO2N with high interparticle networking prepared from La2Ti2O7via a solid-state reaction (SSR).21 The starting oxide was crystallized at a high temperature of 1523 K to form continuous interparticle connections. This preparation largely reduced grain boundaries and potential defects between the oxynitrides after nitridation. Co3O4-modified LaTiO2N particles caused fast charge transport during water splitting and produced a high photocurrent of 4 mA cm−2 at 1.23 VRHE in 1 M NaOH electrolyte at pH 13.6 under AM 1.5G simulated sunlight. However, despite several remedies for the synthesis of less-defective LaTiO2N, its water-splitting activity has rarely been reported and is still limited in neutral electrolytes, in which water-splitting systems such as solar panels and PEC cells usually operate (Table S1†).
During nitridation, the starting oxides play a vital role in reducing the defect density of the resulting oxynitrides. A-rich starting oxides with an A/B ratio of 1.25 or even higher have been found to be effective for preparing less-defective AB(O,N)3, thus enhancing PEC water-splitting activity.10,11,22 Layered perovskite A5B4O15 (A = Sr, Ba; B = Nb, Ta) was completely transformed into perovskite ABO2N without any impurity traces during nitridation.11,22 The excess Sr or Ba species remaining after the complete nitridation was readily removed by washing with distilled water.22 The Lewis-basic, A-rich concentrations in A5B4O15 significantly suppressed the reductions of B-site cations during the nitridation.23 Consequently, the less-defective ABO2N photoanodes produced a high water-splitting photocurrent at the milliampere level, which has never been achieved previously in perovskite oxynitrides. La-rich oxide mixtures with La/Nb ratios of more than unity have also been found to be favorable for the synthesis of less-defective, small LaNbON2 particles, although the excess La species is difficult to remove because of the slow reaction with water, leading to impurity traces of La3NbO7.10 The incomplete nitridation of the La-rich mixture causes limited improvement in the PEC water-splitting activity of the resulting LaNbON2. Thus, the aforementioned results indicate that both the A-rich condition in the starting oxide and strong basicity of the A species are necessary to obtain less-defective, single-crystalline oxynitrides during nitridation.
On the basis of the abovementioned background, layered perovskite BaLa4Ti4O15 can be proposed as a starting oxide for the synthesis of less-defective, crystalline LaTiO2N. Lewis-base Ba cations as well as La jointly occupy the A sites in the oxide, leading to a (Ba + La)/Ti ratio greater than unity, which may lower the reduction of Ti4+ during nitridation. In addition, the basic Ba species remaining after the nitridation is de-intercalated to the exterior of LaTiO2N, which may easily dissolve in water.22 Crystallographically, (111)-type layered BaLa4Ti4O15 has structural distortion, i.e., tilting of the TiO6 octahedron in the perovskite block, which is similar to that of the Ti(O,N)6 octahedron in LaTiO2N with an orthorhombic perovskite structure.24 The structural similarity between the starting oxide and resulting oxynitride should guarantee complete and fast nitridation to obtain highly crystalline LaTiO2N. In a study, layered BaLa4Ti4O15 served as an oxide semiconductor itself to drive the photocatalytic reduction of nitrate ions to dinitrogen.25 N-doped BaLa4Ti4O15 prepared via mild, short nitridation was responsive to visible light; however, it showed limited photocatalytic OER.26 Because there are no reports yet on the conversion of layered perovskite BaLa4Ti4O15 into LaTiO2N, a demonstration of the improvement in the hydrogen evolution rate by the resulting oxynitride at neutral pH is a scientifically meaningful endeavor for introducing a new synthesis route.
In this study, we first attempted the complete conversion of layered perovskite BaLa4Ti4O15 as a starting oxide to yield less-defective, single-phase LaTiO2N for sunlight-driven seawater splitting. BaLa4Ti4O15 was crystallized by the flux-assisted calcination of BaCO3, La2O3, and TiO2, followed by nitridation at 1123 K for 20 h under an NH3 flow and subsequent surface-annealing treatment, finally producing LaTiO2N. The oxynitrides were also prepared from other layered perovskite oxides, namely La2Ti2O7 and La5Ti4Ox, for comparison. The prepared LaTiO2N photoanodes were fabricated by spin-coating and necking treatment and then employed for driving PEC seawater splitting including chloride oxidation as well as pure water splitting at neural pH. The oxynitride photoanode exhibited a photocurrent density of approximately 2.4 mA cm−2 at 1.23 VRHE (4.9 mA cm−2 at 1.36 VRHE) in 0.5 M NaCl (electrolyte) at pH 6.4 under AM 1.5G simulated sunlight. The oxynitride prepared from BaLa4Ti4O15 exhibited much higher neutral photoactivity than those of oxynitrides prepared from La2Ti2O7 and La5Ti4O15. Its higher photocurrent also considerably surpassed our previous result for SrNbO2N crystals grown on a Nb substrate in the same electrolyte.27 Various physical and surface properties of the LaTiO2N particles prepared from BaLa4Ti4O15 were investigated to determine the reason behind the enhancement in the PEC seawater splitting by the oxynitride. In particular, the conversion of BaLa4Ti4O15 was compared with those of other starting oxides, and the surface and bulk defect densities of the resulting oxynitrides were estimated by X-ray photoelectron spectroscopy (XPS) and photoluminescence (PL) analyses. The analysis results clearly demonstrate the types of starting oxide that reduce the defect densities of the corresponding oxynitride as well as enhance its PEC oxidation reaction activity.
LaTiO2N was synthesized by the nitridation of the prepared BaLa4Ti4O15. BaLa4Ti4O15 (0.3 g) was transferred to an alumina boat and heated in an NH3 flow (6 N grade) of 250 mL min−1 at 1123 K for 20 h at a ramping rate of 10 K min−1 with intermediate grinding. Then, the resulting LaTiO2N was washed with a large amount of deionized water for 1 h to remove excess Ba species, followed by acid treatment with a 10 mM hydrochloric acid (Samchun Chemicals) aqueous solution for 5 min to eliminate impurity traces. The surface-treated oxynitride was dried naturally and annealed in an Ar flow (5 N grade) of 100 mL min−1 at 973 K for 1 h with a ramping rate of 10 K min−1. The nitridation of the starting oxides La2Ti2O7 and La5Ti4Ox was also performed under the same conditions as those mentioned above.
The seawater-splitting activity of Co(OH)x/LaTiO2N/FTO photoanodes was measured in a typical three-electrode system in conjunction with a potentiostat (WPG100e, WonATech Co., Ltd) using a two-compartment PEC cell separated with a 115 Nafion membrane (NARA Cell-Tech Corporation) under AM 1.5G simulated sunlight (XES-50S2-TT, San-El Electric). The prepared photoanode, an Ag/AgCl electrode, and a Pt wire were utilized as the working, reference, and counter electrodes, respectively. The photoanode and Ag/AgCl electrode were mounted in one compartment, and the Pt wire was placed in another compartment of the PEC cell. An aqueous 0.2 M potassium phosphate (KPi) buffer electrolyte was prepared in both compartments, and 0.5 M NaCl was added into the photoanode side to produce artificial seawater. The electrode potentials were estimated as values versus a reversible hydrogen electrode (RHE) by using the Nernst equation, ERHE = EAg/AgCl + 0.059 × pH + 0.197. Linear sweep voltammograms (LSVs) were recorded during seawater splitting by cathodically scanning the potential range from 1.4 to 0.4 VRHE at a scan rate of 10 mV s−1 at 298 K in an Ar-saturated 0.5 M NaCl electrolyte buffered with 0.2 M KPi at pH 6.4 and 13. The bulk charge separation efficiency, ηbulk, can be determined using the equation8
ηbulk (%) = Jsulfite/Jabs × 100 | (1) |
![]() | (2) |
LHE(λ) = 1 − 10−A(λ) | (3) |
ηsurface (%) = Jwater/Jsulfite × 100 | (4) |
αhν = A(hν − Eg)n/2 | (5) |
![]() | ||
Fig. 1 Schematic illustration showing the conversion of (A) layered perovskite La2Ti2O7, (B) BaLa4Ti4O15, and (C) La4Ti3O12 into (D) perovskite LaTiO2N. |
Fig. 2 shows the XRD patterns for LaTiO2N and the intermediate derivatives generated during the nitridation of different layered perovskite oxides at 1123 K. The nitridation of La2Ti2O7 into LaTiO2N (PDF card no. 4002463) was completed in 5 h, indicating a rapid conversion of the oxide. Prolonged nitridation of the oxynitride for 20 h was performed to improve its surface and bulk crystallinities.30 By contrast, BaLa4Ti4O15 was gradually transformed into LaTiO2N during nitridation for the first 10 h, with segregated crystals such as Ba2TiO4 (PDF card no. 2310344), La2O3, and TiO2 appearing in 10 h. LaTiO2N was further crystallized during subsequent nitridation for 10 h, accompanied by exclusive La2O3 impurity. Then, the acid treatment of the resulting oxynitride removed the La2O3 traces and finally left perovskite-type LaTiO2N. Meanwhile, the nitridation of La5Ti4Ox for 20 h led to the incomplete conversion of primary-phase La4Ti3O12, although the secondary-phase La2Ti2O7 was converted into LaTiO2N even in 1 h. The crystalline La4Ti3O12 was fully transformed into LaTiO2N along with La2O3 traces during nitridation for 50 h, indicating a relatively slow nitridation (Fig. S3†). The impurity traces were removed by the acid treatment as well as LaTiO2N prepared from BaLa4Ti4O15. These results demonstrate for the first time that perovskite LaTiO2N was successfully synthesized from layered perovskite BaLa4Ti4O15 including unnecessary Ba ions and La-rich La4Ti3O12. It also reveals that the rate of nitridation to LaTiO2N at 1123 K is dependent on the different types of layered perovskite oxides. Both La4Ti3O12 and BaLa4Ti4O15 adopt a (111)-type layered perovskite (An+1BnO3n+3) crystal structure; however, their perovskite blocks (n = 3 or 4) and intruding AO layers (A = La or Ba/La) are different as shown in Fig. 1.31 The subtle difference in the crystal structure might be a critical factor determining the nitridation rate of the oxides. According to the full-width-at-half-maximum (FWHM) values for the (121) peak of LaTiO2N, the relatively fast nitridation using BaLa4Ti4O15 caused the high crystallinity of the resulting oxynitride.
The elemental compositions of LaTiO2N particles prepared from the nitridation of different oxides at 1123 K for 20 h were estimated by ICP-MS analysis (Table S2†). The La/Ti atomic ratio of LaTiO2N prepared from La2Ti2O7 was close to its theoretical value of unity, while those of LaTiO2N prepared from BaLa4Ti4O15 and La5Ti4Ox deviated from unity. In the case of La5Ti4Ox, La4Ti3O12 undoubtedly remained even after nitridation contributed to a La/Ti ratio of more than unity. As shown in Fig. 2B, the XRD pattern for the oxynitride transformed from BaLa4Ti4O15 is assignable to the single phase of LaTiO2N. The excess amorphous Ba species remaining after the nitridation of Ba5B4O15 (B = Ta, Nb) was highly soluble in water, leading to Ba/B ratios of unity in the corresponding BaBO2N.11,22 In this study, the washing of the oxynitride produced after nitridation of BaLa4Ti4O15 with deionized water caused highly alkaline waste water, which was gradually changed to neutral water via the repetition of the washing procedure (Fig. S4†). This result proves that the excess Ba species, amorphous BaO, was removed by washing with deionized water due to its strong basicity. Nevertheless, a significant ratio of the Ba species, leading to the La/Ti ratio of less than unity, was detected in the oxynitride prepared from BaLa4Ti4O15. The Ba2TiO4 temporarily appearing in 10 h nitridation (as discussed in Fig. 2B) could result in amorphous BaTiO3 because it reacts with TiO2 to produce BaTiO3.32 The nitridation of perovskite BaTiO3 was attempted for a comparison and clearly showed the generation of segregated phases such as Ba2TiO4 and TiO2 (Fig. S5†). Therefore, although reaction (6) is assumable for the nitridation of BaLa4Ti4O15, LaTiO2N could be prepared in this study by chemical reaction (8)via reaction (7):
BaLa4Ti4O15 + 4NH3 (g) → 4LaTiO2N + BaO + 6H2O | (6) |
Ba2TiO4 + TiO2 → 2BaTiO3 | (7) |
3BaLa4Ti4O15 + 6NH3 (g) → 6LaTiO2N + BaO + 9H2O + 3La2O3 + 4TiO2 + 2BaTiO3 | (8) |
Because the amorphous BaO was washed with distilled water and the segregated La2O3 and TiO2 were removed by acid treatment, the Ba concentration detected in the ICP analysis was attributable to amorphous BaTiO3 traces. However, the nitridation of BaLa4Ti4O15 produced exclusively crystalline LaTiO2N, which is feasible for sunlight-driven water splitting. The unit-cell parameters of the oxynitride were estimated by Rietveld refinement using its XRD results (Fig. S6 and Table S3†). The experimentally acquired pattern for the prepared LaTiO2N well matched with its calculated pattern, indicating an orthorhombic system, and it was also consistent with published results.33 Thus, the abovementioned results prove that layered perovskite BaLa4Ti4O15 was successfully converted into single-crystalline LaTiO2N via the complete separation of Ba species during nitridation.
Fig. 3A presents the DRS spectra of different layered perovskite oxides and the corresponding LaTiO2N prepared by nitridation and subsequent annealing in an Ar flow. All the starting oxides were responsive to UV light below the wavelength of 350 nm. Interestingly, the light-absorption edges of the oxides were easily red-shifted to visible-light wavelengths even in 1 h nitridation, regardless of the types of layered perovskites (Fig. S7†). After complete nitridation, all the oxides were transformed into LaTiO2N capable of absorbing visible-light wavelengths up to approximately 610 nm. However, the background absorption at wavelengths longer than 610 nm were distinguishable for the respective LaTiO2N. The apparent colors of LaTiO2N prepared from La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox are brown, red-orange, and reddish brown, respectively. The difference in color may be attributed to the different levels of background absorption. The high background absorption typically indicates the presence of defect density in the oxynitrides, originating from the reduction of highly electronegative Ti4+ to Ti3+ during nitridation.6 According to the DRS spectra results, the bulk defect density in LaTiO2N prepared from BaLa4Ti4O15 was very low compared to those prepared from other oxides. Excess Lewis-base Ba species locally relieved a reducing atmosphere at BaBO2N surfaces during nitridation, thus reducing the generation of defects.23 In this study, the presence of Ba ions in BaLa4Ti4O15 may have been helpful in reducing the defect concentration in LaTiO2N during nitridation. It was also more effective than the presence of excess La ions in La5Ti4Ox owing to its stronger basicity. The Eg of the prepared LaTiO2N estimated by Tauc plots converted from the DRS spectra is presented in Fig. 3B. The values of La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox were approximately 2.17, 2.13, and 2.12 eV for a direct allowed transition, respectively, which was in good agreement with reported results.6 In addition, the flat-band potentials, Efb, of the LaTiO2N particles were measured by MS analysis (Fig. S8†). The Efb values of LaTiO2N prepared from La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox were determined to be −0.22, −0.21, and −0.28 VRHE, respectively. By assuming that the Efb of an n-type semiconductor is commonly positioned immediately below its conduction band minimum potential, band structure diagrams for LaTiO2N prepared from La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox were constructed as shown in Fig. 3C. The band structures of the oxynitrides were similar and spanned the water redox and CER potentials. Thus, these optical results demonstrate that LaTiO2N prepared from the different layered perovskite oxides was visible-light-responsive up to 610 nm, which is theoretically feasible for realizing sunlight-driven (sea)water splitting. In particular, BaLa4Ti4O15 as a starting oxide afforded a less-defective oxynitride at a bulk level.
Fig. 4 depicts the SEM images of different starting oxides, namely La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox, and the corresponding LaTiO2N particles. La2Ti2O7 particles in the 100–200 nm size range were uniform and rod-like. BaLa4Ti4O15 particles were composed of 2D truncated octahedrons with widths of 1–6 μm in a thickness of approximately 500 nm and small particles of nanometer sizes. The 2D octahedral shape was a typical structure of layered perovskite BaLa4Ti4O15, similar to that of Sr5Nb4O15 in our previous work.28 Although the small particles failed to grow the 2D structure, they were also assignable to BaLa4Ti4O15 according to an XRD result. The La5Ti4Ox particles were rod-like or polyhedron-type, indicating that La2Ti2O7 of nanometer size and La4Ti3O12 of ∼10 μm size were crystallized simultaneously, consistent with an XRD pattern result. After nitridation, the surface morphologies, including the size and shape, of the resulting LaTiO2N were unchanged from those of the corresponding oxides. The difference in the nitridation rate of La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox, discussed in the XRD results above, could be partly attributed to their different particle sizes. In particular, the small La2Ti2O7 particles completed nitridation the earliest because the nitrogen diffused quickly inside the particle. Moreover, the smooth surfaces of the oxides became porous, which originated from the exchange of three O2− with two N3− ions during nitridation.12 The porosity of LaTiO2N prepared from BaLa4Ti4O15 was much higher than that prepared from La4Ti3O12, even though both the particles were relatively large (Fig. S9†). The BET surface area of the oxynitride was also larger than that prepared from La4Ti3O12 and its pore size distribution was relatively shifted to larger pore diameter, which are consistent with the SEM results (Table S4 and Fig. S10†). The origin for the different porosities is still unclear; however, it may be due to the faster nitridation of BaLa4Ti4O15 excluding sintering for 20 h. Consequently, the nitridation of BaLa4Ti4O15 produced 2D truncated octahedral LaTiO2N particles with high porosity, leading to numerous reaction sites, which is favorable for PEC (sea)water splitting.
The surface morphology of LaTiO2N particles prepared from BaLa4Ti4O15 was further analyzed by FE-TEM and EDS mapping analyses as shown in Fig. 5. Both the large truncated octahedron and small particles were highly porous, consistent with SEM images. Fig. 5C displays clear lattice fringes with the d(121) spacing (PDF card no. 4002463) corresponding to a single LaTiO2N phase, demonstrating the high crystallinity of LaTiO2N. In Fig. 5D, the elemental mapping shows that a LaTiO2N particle is composed of elemental La, Ti, O, and N, and Ba species is not detected in the single particle. It reveals that the excess Ba species detected in the ICP analysis resulted from amorphous BaTiO3 impurity traces. Thus, this result proves that the nitridation of BaLa4Ti4O15 produced single-crystalline LaTiO2N particles.
The surface and bulk defect densities of LaTiO2N prepared from layered perovskite oxides with different ratios of La/Ti were estimated by XPS, EPR, and PL analyses. Fig. 6A displays the surface cation defect densities of LaTiO2N particles generated during the nitridation of BaLa4Ti4O15 for 1, 5, 10, 15, and 20 h and after acid treatment. Those arising during the nitridation of La2Ti2O7 and La5Ti4Ox were also measured for comparison. The significantly low-electronegativity La species was excluded from the surface cation defects. The chemical states of Ti species on LaTiO2N were deconvoluted from narrow-scanned Ti 2p XPS spectra with two oxidation states of Ti4+ and Ti3+ positioned at binding energies of 458.0 and 456.7 eV for the Ti 2p3/2 peak, respectively (Fig. S11†).34 The fractions of reduced Ti species on the oxynitride series are summarized in Table S5† and also plotted in Fig. 6A. The surface defects of Ti3+ gradually increased during the nitridation of the different oxides, indicating that long-duration nitridation at high temperatures led to an increase in the surface defect density of LaTiO2N. In addition, the A-site-rich starting oxides BaLa4Ti4O15 and La5Ti4Ox lowered the surface defect density during nitridation compared to La2Ti2O7 with the La/Ti ratio of unity. These phenomena were analogous to those during the nitridation of A-site-rich oxides to obtain Nb- and Ta-based perovskite AB(O,N)3.9–11 In particular, the defect densities on LaTiO2N prepared from BaLa4Ti4O15 were the lowest during the overall nitridation. The excess Ba species in the oxide was more effective at suppressing the generation of defects than La-rich La5Ti4Ox. The surface defect of Ti3+ temporarily decreased during the first 10 h of nitridation, which probably originated from the segregation of intermediate Ba2TiO4 and TiO2 (having Ti4+) during the crystallization of LaTiO2N as illustrated in Fig. 2. The acid treatment to remove the La2O3 and TiO2 impurities also reduced the cation defect density on LaTiO2N.
Fig. 6B shows the EPR spectra of LaTiO2N particles prepared by nitridation of different layered perovskite oxides. The broad EPR signal at g = 2.005 was observed in all LaTiO2N, which has previously been ascribed to unpaired electrons resulting from oxygen and/or nitrogen vacancies in the oxynitride bulk.35,36 The normalized signal of LaTiO2N, prepared from BaLa4Ti4O15, was relatively large as compared with those of other oxynitrides. The large signal may be mainly due to oxygen vacancies in amorphous BaTiO3 traces segregated during nitridation, judging by the lowest background absorbance of the oxynitride observed in DRS spectra. Meanwhile, another broad EPR signal was observed in LaTiO2N prepared from La2Ti2O7 (at g = 1.998) and La5Ti4Ox (at g = 1.986), respectively, which has previously been attributed to free electrons originating from the Ti3+ defect species.35,36 However, it was absent in the oxynitride prepared from BaLa4Ti4O15. Therefore, this result indicates that the synthesis of LaTiO2N using BaLa4Ti4O15 suppressed the generation of bulk cation defects as well as the surface cation defects shown in XPS data.
Fig. 6C depicts the PL spectra of LaTiO2N particles prepared from different layered perovskite oxides. Emission peaks at 640 nm corresponding to the band gap of 1.94 eV were observed in all LaTiO2N, indicating the presence of intermediate defect levels inside the band structure of the oxynitride. The defect levels were assignable to the shallow donor levels of the Ti3+ species, according to its positions close to the conduction band of LaTiO2N.37 Nevertheless, the bulk defect level of LaTiO2N prepared from BaLa4Ti4O15 was the lowest, consistent with DRS spectra results. Fig. 6D shows the time-resolved PL decay spectra of the different oxynitrides acquired at the PL peak position of 640 nm. The LaTiO2N prepared from BaLa4Ti4O15 had a longer average PL decay lifetime of band-edge emission compared to other oxides, definitely leading to a slower recombination rate of photo-carriers. These results demonstrate that BaLa4Ti4O15 is favorable as a starting oxide to synthesize less-defective LaTiO2N both in bulk and surface levels.
For PEC measurements, a Co(OH)x electrocatalyst was loaded on particulate LaTiO2N/FTO photoanodes prepared by the nitridation of different starting oxides (La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox) at 1123 K for 20 h. Fig. 7A displays the LSV curves obtained from the Co(OH)x/LaTiO2N/FTO photoanodes during seawater splitting in 0.5 M NaCl aqueous solution buffered with 0.2 M KPi adjusted to pH 6.4 under chopped AM 1.5G simulated sunlight. The as-prepared oxynitrides were annealed in an Ar flow prior to the fabrication of photoanodes to improve the surface crystallinity of the oxynitrides according to our previous results.22,30 The annealing of LaTiO2N was optimized at 973 K for 1 h (Fig. S12†). The LaTiO2N photoanode prepared from BaLa4Ti4O15 produced a much higher photocurrent density (2.4 mA cm−2) at 1.23 VRHE than those prepared from La2Ti2O7 (0.8 mA cm−2) and La5Ti4Ox (0.6 mA cm−2). Its anodic photocurrent at 1.36 VRHE reached 4.9 mA cm−2. The tendency in the activity was unchanged even during seawater splitting of bare LaTiO2N photoanodes, although the photocurrent was decreased due to the absence of electrocatalyst Co(OH)x boosting the photoreaction (Fig. S13†). The complete nitridation of La5Ti4Ox for 50 h increased the photocurrent to 0.8 mA cm−2 at 1.23 VRHE; however, it was still less than 2.4 mA cm−2 (Fig. S14†). Because amorphous BaTiO3 impurity traces were detected in the LaTiO2N prepared from BaLa4Ti4O15, the PEC seawater-splitting activity of nitrided BaTiO3 was also measured in the same manner for comparison (Fig. S5†). Although the nitridation of the oxide for 20 h allowed it to absorb visible light, no photoresponse was observed during the seawater splitting of the oxide. This result clearly proves that visible-light-responsive LaTiO2N should afford a high seawater-splitting photocurrent. The photocurrent onset of LaTiO2N at 0.90 VRHE was more negative by approximately 0.1 VRHE than that of LaTiO2N prepared from La2Ti2O7. However, the initial photocurrent of the LaTiO2N decreased during long-term seawater splitting for 1 h, probably because of the self-photo-oxidation of the oxynitride surface which was not fully decorated with the Co(OH)x electrocatalyst (Fig. S15†). Nevertheless, the high photoactivity of LaTiO2N at neutral pH, i.e., low onset potential and high photocurrent, was remarkable, considering that the oxynitride has been previously reported for water splitting at a strong alkaline pH 13.20,21
![]() | ||
Fig. 7 (A) LSV curves acquired on particulate Co(OH)x/LaTiO2N/FTO photoanodes prepared by the nitridation of different layered perovskites, namely (a) La2Ti2O7, (b) BaLa4Ti4O15, and (c) La5Ti4Ox, at 1123 K for 20 h and subsequent annealing in an Ar flow at 973 K for 1 h, during seawater splitting under chopped AM 1.5G simulated sunlight. All curves were acquired by sweeping the potential from 1.4 to 0.6 VRHE at a scan rate of 10 mVs−1 in an Ar-saturated 0.5 M NaCl aqueous solution buffered with 0.2 M KPi at pH 6.4. (B) Nyquist plots of the corresponding photoanodes. The EIS measurements were performed in the electrolyte identical to (a) at an applied potential of 1.23 VRHE. The EIS fitting results are summarized in Table S4.† (C) LSV data for the corresponding Co(OH)x/LaTiO2N/FTO photoanodes during seawater oxidation in 0.5 M NaCl aqueous electrolyte (dashed curves) and sulfite oxidation in 0.5 M Na2SO3 aqueous electrolyte (solid curves) under chopped AM 1.5G sunlight. Both solutions were buffered with 0.2 M KPi and adjusted to pH 13. (D) Bulk and (E) surface charge separation efficiency, ηbulk (%) and ηsurface (%), of the Co(OH)x/LaTiO2N/FTO photoanodes presented in (C). |
The surface morphology, such as porosity and size, of LaTiO2N particles prepared from different perovskite oxides was significantly distinguishable, as revealed by SEM images and BET analyses. The electrochemical surface areas of the different oxynitrides were estimated by double-layer capacitance Cdl computed with cyclic voltammograms acquired under darkness at non-faradaic potentials, because the electrochemical properties reflect the surface properties of material particles (Fig. S16†). The Cdl value of the LaTiO2N photoanode prepared from BaLa4Ti4O15 was similar to that from La2Ti2O7, while it was more than four times larger than that from La5Ti4Ox. This implies that the former LaTiO2N photoanodes possess relatively large electrochemical reaction sites compared with those from La5Ti4Ox. However, the trend in the photoactivity of different LaTiO2N, discussed in Fig. 7A, was not in agreement with that in the electrochemical surface area of the oxynitrides. Fig. 7B shows the Nyquist plots of the LaTiO2N photoanodes prepared from different oxides, which were measured by EIS in a neutral NaCl aqueous solution under simulated sunlight at an applied potential of 1.23 VRHE. The Nyquist plots were best-fitted by an equivalent circuit model; the fitting results are summarized in Table S6.†R1, R2, and R3 denote electrolyte resistance (Rs), charge transfer resistance (Rct) between LaTiO2N and the electrolyte, and electron transport resistance (Ret) from LaTiO2N to the FTO substrate, respectively. Ret could originate from the loose adhesion between the oxynitride and substrate prepared via spin-coating despite the subsequent necking treatment. The results revealed that the Rct of the LaTiO2N prepared from BaLa4Ti4O15 was much smaller than those of LaTiO2N prepared from La2Ti2O7 and La5Ti4Ox. In addition, its Ret was relatively reduced because of the high crystallinity of the oxynitride. Thus, the high seawater-splitting activity of LaTiO2N at neutral pH is highly attributable to the semiconducting properties of the less-defective oxynitride prepared from BaLa4Ti4O15, which increases the conductivity of photo-carriers and accelerates the photoreaction.
The thermodynamic potential of 1.36 VNHE for the CER at pH 0 is more oxidative than the 1.23 VNHE for the OER; however, the CER driven by a two-electron pathway is kinetically faster than the OER driven via four-electron transfer. The LaTiO2N photoanode prepared from BaLa4Ti4O15 produced a photocurrent density (1.3 mA cm−2) at 1.23 VRHE during water splitting in 0.2 M KPi aqueous solution adjusted to pH 6.7 (Fig. S17†). The OER activity of the oxynitride was lower than the CER activity discussed in Fig. 7A, consistent with our previous result.27 This obviously indicates that the CER and OER were competitive in the neutral NaCl aqueous electrolyte evaluated in this study. The photoactivity of LaTiO2N prepared from BaLa4Ti4O15 was accordingly evaluated during seawater splitting at pH 13 to determine the effect of the less-defective oxynitride on the kinetically boosted OER in a strongly alkaline electrolyte. Its anodic photocurrent was further increased to 3.3 mA cm−2 at 1.23 VRHE (5.2 mA cm−2 at 1.36 VRHE) during the alkaline seawater splitting (Fig. S18†). The photocurrent of oxynitrides prepared from other oxides was enhanced as well, although their photoactivity order remained unchanged. This result demonstrates that the high photoactivity at neutral pH is mainly attributable to the enhanced semiconducting properties of less-defective LaTiO2N from BaLa4Ti4O15, other than the faster CER kinetics. In addition, the sulfite oxidations of the LaTiO2N photoanodes prepared from different starting oxides were evaluated in 0.5 M Na2SO3 aqueous solutions at pH 13 as shown in Fig. 7C. The PEC behaviors of the oxynitrides for the sulfite oxidation were much more activated, and the trend between the oxynitrides was the same as that during the seawater splitting at pH 13. This reveals that LaTiO2N prepared from BaLa4Ti4O15 was highly photoactive regardless of oxidation reactions.
The separation efficiency (ηbulk) of charges generated in the bulk of LaTiO2N prepared from different oxides was evaluated, as presented in Fig. 7D. The photon absorption rate (Jabs) of the different oxynitrides prepared from La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox was determined to be 11.0, 10.0, and 10.0 mA cm−2, respectively (Fig. S19†). The ηbulk of the oxynitrides was approximately 22%, 37%, and 11% at 1.23 VRHE, respectively, calculated using the photocurrent density for sulfite oxidation. This result was consistent with the degree of crystallinity and bulk defect densities of the oxynitrides, discussed in XRD patterns and PL spectra. Undisputedly, the highest crystallinity and low bulk defect density of LaTiO2N prepared from BaLa4Ti4O15 led to the highest efficiency. Fig. 7E displays the separation efficiency (ηsurface) of charges generated at the surfaces of LaTiO2N prepared from different oxides, estimated by the difference in photoactivities between fast sulfite and slow seawater oxidations. The ηsurface of oxynitrides prepared from La2Ti2O7, BaLa4Ti4O15, and La5Ti4Ox was approximately 41%, 90%, and 61% at 1.23 VRHE, respectively. Remarkably, an efficiency of almost 100% was achieved at 1.36 VRHE during the water splitting by LaTiO2N prepared from BaLa4Ti4O15. This definitely indicates that the water-splitting activity driven by the oxynitride was efficient and sufficiently high. The low defect density and high crystallinity of LaTiO2N suppressed the recombination of photogenerated holes and electrons, thus causing the consumption of the holes entirely for the slow water oxidation. Although the PEC activity of LaTiO2N prepared from La5Ti4Ox was lower than that of LaTiO2N prepared from La2Ti2O7, its separation efficiency became higher. The higher efficiency could be attributed to the 2D-type surface structure of LaTiO2N possible for the short diffusion of charges inside a narrow thickness because a similar effect was observed in 2D octahedral SrNbO2N doped with Zr.28 As a result, the low defect density and 2D surface structure of LaTiO2N prepared from BaLa4Ti4O15 strongly promoted the separation of photo-charges, enhancing PEC seawater-splitting activity. Therefore, as illustrated in Fig. 8, this study clearly proves that the complete conversion of layered perovskite BaLa4Ti4O15 including excess Ba species with high basicity was effective at preparing less-defective, 2D LaTiO2N crystals in both bulk and surface states, realizing efficient neutral seawater splitting under sunlight.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ta01901f |
This journal is © The Royal Society of Chemistry 2024 |