Undoped Layered Perovskite Oxynitride Li2 LaTa2 O6 N for Photocatalytic CO2 Reduction with Visible Light.

Oxynitrides are promising visible-light-responsive photocatalysts, but their structures are almost confined with three-dimensional (3D) structures such as perovskites. A phase-pure Li2 LaTa2 O6 N with a layered perovskite structure was successfully prepared by thermal ammonolysis of a lithium-rich oxide precursor. Li2 LaTa2 O6 N exhibited high crystallinity and visible-light absorption up to 500 nm. As opposed to well-known 3D oxynitride perovskites, Li2 LaTa2 O6 N supported by a binuclear RuII complex was capable of stably and selectively converting CO2 into formate under visible light (λ>400 nm). Transient absorption spectroscopy indicated that, as compared to 3D oxynitrides, Li2 LaTa2 O6 N possesses a lower density of mid-gap states that work as recombination centers of photogenerated electron/hole pairs, but a higher density of reactive electrons, which is responsible for the higher photocatalytic performance of this layered oxynitride.

Semiconductor materials that can split water into H2 and O2 as photocatalysts have been extensively explored and developed.1 Recently, the research interest is being expanded to CO2 reduction, but it is generally very difficult to achieve the reaction because of the lack of active sites for CO2 reduction on the surface as well as the predominant occurrence of competitive H2 evolution reaction.2 Some layered perovskite oxides consisting of Ti4+, Nb5+, and Ta5+ have been regarded as high‐potential photocatalysts for water splitting and CO2 reduction.3, 4

Since most of metal oxide photocatalysts have large band gaps (>3 eV) and are hence inactive under visible light (λ>400 nm),5 mixed‐anion compounds such as oxynitrides have attracted considerable attention as potential visible‐light‐responsive photocatalysts toward solar energy conversion.6, 7 While nitrogen‐doping into oxides is a conventional and facile method to induce visible light response,8 the charge imbalance between oxide and nitride anions (O2− vs. N3−) inevitably introduces defect states, which act as recombination centers of photoexcited carriers, and lower photocatalytic activity as seen in nitrogen‐doped TiO2.9 Furthermore, the limited amount of doped nitrogen results in insufficient visible light absorption.

Undoped oxynitrides are therefore highly desirable. One such example is a three‐dimensional (3D) perovskite oxynitride CaTaO2N (Figure 1 left), which can serve as an effective semiconductor component in a photocatalytic CO2 reduction system, with the aid of a binuclear RuII complex (RuRu′; Supporting Information, Figure S1).10 The combined system, RuRu′/semiconductor, works as follows; both semiconductor and metal complex are excited by visible light. Holes in the valence band of the semiconductor oxidize an electron donor and the electron in conduction band reductively quenches the excited state of the photosensitizer unit of RuRu′, producing one electron reduced species (OERS) of the photosensitizer unit. Electron transfer occurs from the OERS to the catalyst moiety, finally reducing CO2 (Supporting Information, Figure S2).11 In this hybrid system, low efficiencies of semiconductors and metal complexes for reduction and oxidation reactions can be addressed by utilizing efficient CO2 reduction ability of metal complexes and strong oxidation ability of semiconductors, respectively.

Figure 1

Crystal structures of CaTaO2N (left) and Li2LaTa2O6N (right) with Wyckoff positions of anion sites. The black solid squares indicate unit cells.

Two‐dimensional (2D) layered oxynitrides may improve the catalytic performance under visible light, but synthesis of layered oxynitrides is generally difficult. There are so far only a few reports of undoped layered oxynitrides (for example, Ba2TaO3N, Rb1.8LaNb2O6.3N0.7⋅1.0 H2O, and K1.6Ca2Nb3O9.4N0.6⋅1.1 H2O),12, 13 but accompanied with byproduction of thermally stable 3D perovskite phases in some cases. Additionally, photocatalytic performance of these undoped layered oxynitrides has not been investigated in detail.

Herein we demonstrate that a 2D layered oxynitride Li2LaTa2O6N, composed of double‐layer [LaTa2O6N]2− perovskite slabs that are separated by two Li cations (Figure 1 right), is a promising visible‐light photocatalyst. This material was originally prepared by Fukuda et al.,13 but together with LaTaON2 as a byproduct. Furthermore, photophysical and photocatalytic properties of Li2LaTa2O6N remain unexplored. Herein, we report the synthesis of a phase‐pure Li2LaTa2O6N. The photocatalytic activities for CO2 reduction under visible light irradiation (λ>400 nm) were discussed, particularly in comparison with 3D perovskite analogues (that is, LaTaON2 and CaTaO2N).

Li2LaTa2O6N was synthesized by high‐temperature ammonolysis using an amorphous oxide precursor containing Li, La, and Ta. The precursor was prepared by the polymerized complex method developed by Kakihana,14 a method that allows metal cations in a metal oxide to disperse homogeneously. Large efforts were devoted to identify the synthesis conditions to reduce the quantity of the byproducts. Various parameters (for example, reaction temperature, duration, and NH3 flow rate) need to be optimized (1173 K, 12 h, and 20 mL min−1 of NH3 flow); otherwise both LaTaON2 and LiTaO3 were readily formed (Supporting Information, Figure S3).

We also found that inclusion of an excess amount of Li in the oxide precursor is critical (Supporting Information, Figure S4). Nitridation of a stoichiometric precursor (0 % excess Li) resulted in byproduct formation such as LaTaON2, which is consistent with the previous work.13 Adding 10 % excess of Li resulted in a single‐phase Li2LaTa2O6N. Further addition of Li (20 %) yielded other diffraction peaks assignable to LiTaO3. Because alkali metal species in oxide precursors usually are prone to volatilization at high temperatures,4 it is likely that the use of excess amount of Li compensated the loss during the high temperature ammonolysis. Unless stated otherwise, when we hereafter address Li2LaTa2O6N, it means the one prepared with 10 % excess of Li.

We checked the crystal structure of Li2LaTa2O6N using the Rietveld analysis of the X‐ray powder diffraction data (Supporting Information, Figure S5), assuming the tetragonal (I4/mmm) layered perovskite structure (Figure 1).13 The refined lattice parameters were a,b=3.9533(4) Å and c=18.452(3) Å, with reasonable reliability factors of R wp=0.1076, R B=0.0623, and R F=0.0447. Note that the small contrast in X‐ray scattering between O and N did not allow us to examine the preference at the three anionic sites. We measured neutron diffraction, but could not obtain sufficient data because of high background.15

Quantitative analysis for heavier elements (that is, La and Ta) was carried out using energy‐dispersive X‐ray spectroscopy (EDS), which indicated that the ratio of La/Ta in the synthesized Li2LaTa2O6N was 0.49.16 The nitrogen content of Li2LaTa2O6N, measured using an elemental analyzer, was determined to be 2.07 wt %. These values are fairly close to the ideal values of 0.5 (for the La/Ta molar ratio) and 2.2 wt % (for N), respectively. Furthermore, charge neutrality of this compound led us to (tentatively) conclude that its composition is Li2LaTa2O6N.

A typical scanning electron microscopy (SEM) measurement of Li2LaTa2O6N showed rectangular plate‐like particles with size of a few hundred nanometers were observed, reflecting a layered nature of the structure (Supporting Information, Figure S6). The specific surface area determined by N2 adsorption at 77 K was 2.4 m2 g−1.

We conducted scanning transmission electron microscopy (STEM) to obtain atomic‐resolution images of Li2LaTa2O6N. Figure 2 a shows a typical high‐angle annular dark‐field (HAADF)‐STEM image, indicating the formation of layered structure. As the signal intensity in HAADF imaging is approximately proportional to Z  2 (where Z is the atomic number),17 La and Ta atomic columns can be seen as bright dots and the arrangement of these bright dots agreed well with La and Ta atomic positions expected from the crystal structure of Li2LaTa2O6N (Figure 2 b). The observed high crystallinity of Li2LaTa2O6N can be advantageous as a photocatalyst.

Figure 2

a) HAADF‐STEM image and b) magnified HAADF‐STEM image of Li2LaTa2O6N synthesized at 1173 K for 12 h under 20 mL min−1 of NH3 flow. In (b), the crystal structure of Li2LaTa2O6N is added: Li+ blue, La3+ green, Ta5+ brown, O2−/N3− red.

The yellow color of the as‐synthesized Li2LaTa2O6N (see Figure 3 a) manifests the ability of visible light absorption. Diffuse reflectance spectra (DRS) of Li2LaTa2O6N and its precursor oxide are shown in Figure 3 b. The oxide precursor exhibited an absorption band only in UV light region, attributable to electron transfer from the valence band of O 2p orbitals to the conduction band formed by Ta 5d orbitals. The absorption band of Li2LaTa2O6N was extended to visible light region, with the band gap (estimated from the absorption edge) of ca. 2.5 eV. The red‐shift strongly suggests the upward shift of the valence band owing to the inclusion of N 2p orbitals.

Figure 3

a) Photographs of the Li‐La‐Ta oxide precursor (top) and Li2LaTa2O6N (bottom). b) DRS of Li2LaTa2O6N and its precursor.

We measured the band‐edge potentials of Li2LaTa2O6N by means of an electrochemical technique using a porous Li2LaTa2O6N electrode, which was prepared by an electrophoretic deposition method.18 Mott–Schottky plots (that is, capacitance−2 (C −2) vs. applied potential) of Li2LaTa2O6N with different frequencies are shown in the Supporting Information, Figure S7. In all cases, the C −2 values decreased as the applied potential became more negative, a typical behavior for an n‐type semiconductor. The flat‐band potential of Li2LaTa2O6N was determined, from the potential at which the C −2 value is zero, to be −1.75±0.13 V (vs. Ag/AgNO3). It is known that in an n‐type semiconductor, the potential of the conduction band minimum is 0.1–0.3 V more negative than the flat‐band potential.19 Owing to the uncertainty in conductivity of Li2LaTa2O6N, we assumed the potential to be 0.2. The band structure of Li2LaTa2O6N, along with LaTaON2 and CaTaO2N reported previously, are depicted in the Supporting Information, Figure S8.7, 20 It can be seen that the conduction band potential of Li2LaTa2O6N is more negative than that of LaTaON2, but slightly more positive than CaTaO2N. On the basis of physicochemical characterization demonstrated above, we conclude that Li2LaTa2O6N has a potential to serve as a photocatalyst under visible‐light irradiation.

To clarify the origin of the new visible light absorption, density functional theory (DFT) calculations were performed. There are three different anionic sites, the bridging site (2a), the apical site (4e), and the equatorial site (8g). As we could not experimentally determine the nitrogen site in Li2LaTa2O6N, we examined, for simplicity, three possibilities where nitride anions are selectively located at either 2a, 4e, or 8g site, and found that the structure with nitrogen at 8g gave the lowest energy, consistent with the Pauling's second rule.21

The total and partial density of states of Li2LaTa2O6N in this configuration are shown in the Supporting Information, Figure S9a, where the conduction band was formed by Ta 5d orbital, while the middle‐to‐top of the valence band was occupied by a hybridized O 2p and N 2p orbital. In cases of 2a and 4e (Supporting Information, Figure S9b,c), the general feature found in the valence and conduction bands was almost the same, with only a small difference in the shape of the total DOS in the valence band. This means that regardless of the model, there is a significant contribution of N 2p orbital allowing the upward shift of the valence band maximum, responsible for the visible light absorption.

The result of UV/Vis diffuse reflectance spectroscopy and electrochemical measurements indicated that electron transfer from the conduction band of Li2LaTa2O6N (−1.95 V vs. Ag/AgNO3) to the Ru photosensitizer unit (+0.17 V) of RuRu′ occurs as a thermodynamically favorable process. Thus, we applied the as‐synthesized Li2LaTa2O6N for the CO2 reduction system with the aid of RuRu′. The results of CO2 reduction reactions are summarized in Table 1. After 15 h irradiation with a high‐pressure mercury lamp (λ>400 nm), formate was detected as the major product with 97 % selectivity, with only tiny amount of H2 (entry 1). The turnover number (TON), which was defined as the ratio of the amount of formate generated to that of the adsorbed RuRu′, exceeded 50, confirming that the reaction took place catalytically. In the absence of either RuRu′ (entry 2) or Li2LaTa2O6N (entry 3), no formate production was detected. Without irradiation, no reaction occurred (entry 4). It is notable that no significant change before and after the 15 h of reaction was observed in the XRD pattern and the light absorption profile of RuRu′/Li2LaTa2O6N (Supporting Information, Figure S10), indicating that Li2LaTa2O6N is stable under the given reaction conditions.

Table 1

Photocatalytic performance of CO2 reduction over hybrid catalyst consisting of a semiconductor and RuRu′ binuclear complex.[a]

Entry	Photocatalyst	Products [nmol]		Selectivity toformate [%]
		Formate	H2
1	RuRu′/Li2LaTa2O6N	660	16	97
2	Li2LaTa2O6N	N.D	N.D	–
3	RuRu′/Al2O3	N.D.	N.D.	–
4[b]	RuRu′/Li2LaTa2O6N	N.D	N.D	–
5	RuRu′/Ag/Li2LaTa2O6N	1440	16	99
6	RuRu′/CaTaO2N	N.D	N.D	–
7	RuRu′/LaTaON2	N.D	N.D	–
8[c]	RuRu′/Ag/CaTaO2N	320	N.D.	>99

[a] Reaction conditions: photocatalyst: 4.0 mg, reaction solution: a mixture of MeCN/TEOA (4:1 v/v) 4 mL; reaction vessel, Pyrex test tube with a septum (8 mL capacity); light source, 400 W high‐pressure mercury lamp with a NaNO2 solution filter. Reaction time: 15 h. In each case, RuRu′ of 3 μmol g−1 was adsorbed. ND=not detected. [b] Without irradiation. [c] In DMA/TEOA (4:1 v/v).

It is necessary to check the carbon source of CO2 reduction product(s) because some contaminated organic compounds on semiconductor surface may undergo unidentified photochemical processes to yield carbon‐containing products.22 Thus, we initially performed CO2 reduction under 13CO2 atmosphere, and collected No‐D 1H‐NMR spectra to verify the carbon source of formate. Unfortunately, the low activity/sensitivity of No‐D NMR method did not allow the detection of any signal of products from RuRu′/Li2LaTa2O6N. Accordingly, nanoparticulate Ag (1.5 wt %) was in prior deposited on Li2LaTa2O6N to promote CO2 reduction, as reported previously.10, 11 Deposition of metallic Ag nanoparticles with 20–30 nm in size on Li2LaTa2O6N was confirmed by means of XAFS and TEM observations (Supporting Information, Figures S11 and S12). This modification improved the formate production rate (entry 5, Table 1). After 15 h irradiation, a clear doublet signal (J 13 CH=174 Hz) assignable to 13C of formate10, 11 was observed between 8.54 and 8.05 ppm in the 1H‐NMR spectrum, while only singlet signal derived from H12COOH appeared when 12CO2 was used (Supporting Information, Figure S13). The present results provide firm evidence that RuRu′/Ag/Li2LaTa2O6N reduced CO2 to formate.

We also compared the photocatalytic activity of the present system with those of the CaTaO2N and LaTaON2 counterparts without any modification other than RuRu′.23 Interestingly, Li2LaTa2O6N demonstrated by far the highest performance among them, as listed in Table 1. We previously reported that prior modification of CaTaO2N with Ag nanoparticles improved the formate production in the visible‐light Z‐scheme CO2 reduction with RuRu′.10 However, even without modification of Ag, the activity of RuRu′/Li2LaTa2O6N was two‐fold higher than that of the optimized RuRu′/Ag/CaTaO2N (entry 8). These results clearly indicate great potentials of layered oxynitride materials as a photocatalyst.

To elucidate the origin of different photocatalytic activities between the 2D and 3D oxynitrides (Li2LaTa2O6N, CaTaON/LaTaON2), transient absorption spectroscopy was applied to examine the nature of photogenerated charge carriers. Figure 4 shows transient absorption spectra of these oxynirides after laser pulse excitation at 480 nm under vacuum. The laser excitation induced a broad absorption band over the range from 20 000 to 1000 cm−1. Here, the absorption from 20 000 to 3000 cm−1 is attributed to electrons/holes trapped at energetically deep defects,24 whereas that from 3000 to 1000 cm−1 originates from shallowly trapped and/or free electrons. It is known that the shallowly trapped and/or free electrons are much more reactive and positively influence the photocatalytic activity than deeply trapped electrons. The shapes of the transient absorption spectra for these three materials were different. In particular, absorption bands attributed to the deeply trapped charges (20 000–3000 cm−1) in Li2LaTa2O6N were weaker than those observed for CaTaO2N and LaTaON2. Furthermore, Li2LaTa2O6N exhibited more pronounced signals from shallowly trapped and/or free electrons than CaTaO2N and LaTaON2. These observations could account for the higher photocatalytic activity of Li2LaTa2O6N relative to CaTaO2N and LaTaON2: a larger portion of the photogenerated carriers in Li2LaTa2O6N survived without undergoing recombination. We deduce that the 2D structure of Li2LaTa2O6N as well as high crystallinity contributes to the unique photophysical property.

Figure 4

Transient absorption spectra for a) Li2LaTa2O6N, b) CaTaO2N, and c) LaTaON2 recorded after 480 nm laser pulse excitation under vacuum. d) Enlarged views of each compound in the 3000–1200 cm−1 region.

In conclusion, we synthesized a layered perovskite oxynitride of Li2LaTa2O6N by thermal ammonolysis of a Li‐La‐Ta oxide precursor, and examined as a photocatalyst for visible‐light CO2 reduction in combination with a binuclear RuII complex (RuRu′). The production of single‐phase Li2LaTa2O6N was confirmed by means of XRD and HAADF‐STEM measurements. The key to synthesize Li2LaTa2O6N without producing impurity phases was the use of an oxide precursor that contained a proper (excess) amount of Li. The as‐synthesized Li2LaTa2O6N having a band gap of about 2.5 eV, modified with RuRu′, was capable of photocatalyzing CO2 reduction into formate under visible light (λ>400 nm) with high selectivity (>97 %), while analogues of 3D‐type oxynitride perovskites of CaTaO2N and LaTaON2 were not. Transient absorption spectroscopy indicated that the lower density of trap states and higher density of reactive electrons were responsible for the high activity for CO2 reduction.

Although known oxynitride photocatalysts are comprised of 3D‐type materials, the result of the present study clearly demonstrates the high potential of a 2D layered oxynitride as a visible‐light‐driven photocatalyst. This is the first example of an undoped layered oxynitride that exhibits distinct photocatalytic activity, opening the possibility of new 2D layered oxynitride photocatalysts for artificial photosynthesis.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.