Synthesis strategy and characterization of BOH nanorods. The Sillenite-structured BOH NRs were prepared by a hydrothermal method, with surfactants added to regulate the hydrolysis of Bi(NO3)3 precursor. Specifically, Bi(NO3)3·5H2O, PVP (polyvinyl pyrrolidone-8K) and mannitol were dissolved in deionized water under stirring, and the resulting mixture was subjected to hydrothermal treatment at 160 °C for 24 h. As shown in the scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images (Fig. 1a and the inset), the product BOH has a morphology of NRs with high aspect ratios (10–30 nm in diameter, 1–6 µm in length); most NRs aggregate into bundles by connecting at the middle or one end. The introduction of PVP and mannitol proved critical for the formation of NRs: as shown in Fig. S1 (in Supporting Information), without mannitol, the product has a rod-like morphology, but the NRs are shorter in length and not uniform in diameter; without PVP, the product has a sheet-like morphology. These results indicate that PVP can regulate the growth of BOH along specific directions, thus exposing specific facets and leading to the formation of NRs; mannitol can promote the dissolution of Bi(NO3)3·5H2O in water, leading to uniform nucleation of BOH and the resulting NRs with high uniformity and aspect ratios28.
The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image and the energy-dispersive X-ray (EDX) mapping results (Fig. 1b–d) confirm that in the BOH NRs, the Bi and O elements are evenly distributed. The selected-area electron diffraction (SAED) pattern (Fig. 1e) and high-resolution TEM (HRTEM) image (Fig. 1f) reveal that each BOH NR is of single crystallinity and has well-ordered lattice fringes. The hydrolysis of Bi(NO3)3·5H2O can yield multiple products including Bi5O7(NO3) or basic bismuth nitrates (Bi6O6(OH)2(NO3)4·2H2O,Bi6O5(OH)3(NO3)5·3H2O, Bi2O2(OH)(NO3) etc. ). All of these products are made of backbones of [Bi–O] layers, with anion layers intercalated in between; the main differences are the arrangement of Bi and O atoms in [Bi–O] layers, and the types and amounts of anions (such as NO3− and OH−) 29–31. It is such differences that result in the variation in the X-ray diffraction (XRD) patterns of the hydrolysis products of Bi(NO3)3. By comparing with the XRD pattern (Fig. S2) for the hydrothermal product without PVP (denoted as BOH-nPVP), we confirmed that PVP only plays a role of regulating the nucleation and growth of products. For our BOH NRs, the XRD pattern (Fig. 1h) does not match well with any standard XRD pattern available; yet still, considering the common hydrolysis products of Bi(NO3)3, it could be inferred that our BOH has a similar basic structure with alternating [Bi–O] and anion layers. In this regard, we performed aberration-corrected HAADF-STEM (AC HAADF-STEM) with atomic resolution. Fig. S3 shows that the BOH NR is composed of ~ 10 ordered layers arranged in parallel, each layer with a thickness of ~ 1 nm. In Fig. 1g, it can be clearly observed that each layer has two arrays of Bi atoms (Bi atoms appear as bright dots in the image, whereas O atoms are barely observable owing to the small atomic number), and the inter-array spacing is 0.27 nm, which is identical to that within the [Bi2O2]2+ layer of the known compound Bi2O2(OH)(NO3) (that is, the distance between two neighboring red balls in Fig. 1g) (for the XRD patterns of the two compounds, see Fig. S4; for the two-dimensional structure of Bi2O2(OH)(NO3), see Fig. S5). The results above confirm that our BOH NRs have a layered [Bi2O2]2+ structure similar to that in Bi2O2(OH)(NO3), and the interlayer channels are openly exposed. The alternating [Bi2O2]2+ and anion layers are stacked via van der Waals interaction, forming a layered Sillenite structure; between the neighboring [Bi2O2]2+ and anion layers exists a perpendicular IEF, which could facilitate the carrier separation32,33. In addition, the IEF is also perpendicular to the normal of the exposed facets of BOH, and thus shortens the migration path for photocarriers, which is conducive to the transport and separation of charges34.
In order to probe the chemical identity of the intercalating anions, we performed X-ray photoelectron spectroscopy on the Bi(NO3)3·5H2O precursor, the PVP surfactant and the product BOH NRs. The N 1 s spectra (Fig. 1i) show that the binding energies of N (399.3 eV and 406.5 eV) in the BOH sample are similar to those for PVP, and different with that for NO3− (407.1 eV). These results indicate that there barely exists any NO3− within the 10 nm subsurface region (which constitutes almost the entire volume) of the BOH NR. Figure 1j shows the O 1 s peaks at 529.0 eV, 530.5 eV and 531.9 eV for BOH, corresponding to the binding energies of O atoms in Bi–O bonds in the [Bi2O2]2+ layer, the C = O groups in PVP, and OH− anions, respectively. These results also explain our finding that the XRD patterns for BOH and Bi2O2(OH)(NO3) do not ideally match: NO3− anion has a larger radius (2.00 Å) than OH− (0.89 Å), so when OH− anions are intercalated between the [Bi2O2]2+ layers (in the case of BOH) instead of NO3− (in the case of Bi2O2(OH)(NO3)), the bridging effect of the anions would become less pronounced, leading to an increased spacing between neighboring [Bi2O2]2+ layers owing to coulombic repulsion35. This is in good consistency with the fact that the first primary XRD peak for BOH (2Theta = 7.7º) shifts to lower angles; the altered symmetry for BOH also results in diffraction peaks located at different angles from those for Bi2O2(OH)(NO3), as well as more diffraction peaks. Therefore, we infer that the BOH NRs have a backbone structure similar to that for Bi2O2(OH)(NO3), only with OH− anions intercalated between the [Bi2O2]2+ layers.
Owing to the alternating arrangement of [Bi2O2]2+ and anion layers, the hydrolysis products of Bi(NO3)3·5H2O are usually Bi5O7(NO3) or basic bismuth nitrates with sheet-like morphologies; yet in this work, by introducing PVP and mannitol during the hydrolysis, we obtained rod-like structures with barely any NO3− and with interlayer channels openly exposed. Compared with conventional sheet-like products, this rod-like structure can shorten the paths for photocarriers to migrate from the interior to the surface, and the holes accumulated at the surface-exposed anion layers can be utilized to activate substrate molecules. In addition, by taking advantage of the openly exposed channels, the anion layers may be modified in both composition and structure.
Synthesis strategy and characterization of BOX nanorods (X = Cl, Br, I). We used the BOH NRs as precursors for subsequent anion exchange experiments. Specifically, BOH NRs were dispersed in deionized water, and a proper amount of KX (X = Cl, Br, I) was added. The mixture was sonicated, sealed in an autoclave and then heated at 60 ºC for 12 h. As shown in Fig. 2a, during the reaction process, the OH− anions between the [Bi2O2]2+ layers could partially exchange with halide anions, and the resulting products could well preserve the rod-like morphology and the backbone structure of unmodified BOH NRs. The products after exchange with KI, KBr and KCl are hereafter denoted as BOH-I, BOH-Br and BOH-Cl, respectively (and collectively as BOH-X).
Similar to the pristine BOH NRs, all three BOH-X samples have a rod-like morphology with diameters of 10–30 nm and lengths of 1–6 µm (Fig. 2b–d and Fig. S6). EDX mapping (Fig. 2e–g and Fig. S7–9) showed uniform distributions of halogen atoms over the NRs after anion exchange; HRTEM-STEM images and the corresponding SAED patterns (Fig. 2h–j) revealed well-preserved single crystallinity for the BOH-X NRs. The XRD patterns for the three BOH-X samples (Fig. S10) are almost identical; yet the primary peaks (at 2 Theta = 7.7º for BOH) shift toward lower angles to different extents, indicating that the intercalation of halide anions has altered the interlayer spacings. In addition, the XPS spectra (Fig. S11a) revealed that the binding energies of Bi 4f in the BOH-X samples were elevated by less than 1 eV with respect to that for BOH, confirming that the highly electronegative halide anions had been introduced successfully. By comparing peak areas of the Bi–O bond and O–H bond in the XPS spectra (Fig. S12), we found that the ratios of the O–H bond became lower for BOH-X, again confirming the success of halide exchange and intercalation. Semi-quantitative analyses based on the XPS data revealed that after anion exchange, the X−/Bi3+ ratios are 0.12 (for I), 0.09 (for Br), and 0.11 (for Cl), indicating similar activities of ion exchange for the halide anions. To unveil the distribution of halogen atoms, we selected BOH-I as a representative and performed High-resolution XPS experiments with Ar+ sputtering at different depths (Fig. 3 and Fig. S13). As the sputtering depth (14 nm) is nearly equal to (or larger than) the radii of the NRs, the results suggest that I− anions are evenly distributed within the entire volume, rather than merely at the surface. In addition, no prominent peaks corresponding to N were observed at the subsurface region, again implying that the N atoms come of PVP, and are distributed primarily at the surface. All the above results confirmed the efficacy of modifying the bulk anion layers of BOH via hydrothermal halide-anion exchange.
Spectroscopic characterization of BOH and BOX nanorods. The photoluminescence (PL) spectra (Fig. 4a) of BOH and BOH-X samples show that all four samples give an emission peak at 552 nm. The PL intensities for the BOH-X are all lower than that for BOH, and the intensity decreases with the atomic number of halide anions. The decrease in PL intensity indicates suppression on the recombination of photocarriers, which we believe can be explained as follows: after X− exchange to replace OH−, the charge distribution at the anion layers becomes more uneven, leading to strengthen the internal electric field resulting from the polarization of atoms and orbitals in vicinity21. The results of transient PL spectroscopy (Fig. 4b and Table S1) further validated our inference above. The average luminescence lifetimes are 3.68 ns for BOH, 3.17 ns for BOH-Cl, 2.63 ns for BOH-Br and 1.96 ns for BOH-I. A shorter PL lifetime indicates that the photo-excitons are more prone to separate rather than to recombine36–39.
The introduction of halide anions may modify the band structure of the photocatalyst. As shown in the UV–vis diffuse reflectance spectra (DRS) (Fig. 4c), the exchange with Cl− and Br− did not affect the absorption edge for the pristine BOH (359 nm), so BOH, BOH-Cl and BOH-Br have similar light absorption ranges and similar bandgaps (3.45 eV). By contrast, the exchange with I− induced a redshift of the absorption edge from 359 nm to 420 nm, indicating a narrower bandgap for BOH-I (2.95 eV), and an enhanced absorption in the visible region.
The positive slopes in the Mott–Schottky plots (Fig. S14) revealed a character of n-type semiconductor for all four samples, and a flat band potential ranging within − 0.55 ~ − 0.34 eV. As the conduction band position is near the flat band potential for n-type semiconductor40, BOH-I has the lowest CBM, and the CBM for BOH、BOH-Cl、BOH-Br are higher by 0.13 eV, 0.19 eV, 0.21 eV, respectively. In combination with the UV–vis DRS results, now we can draw the diagrams of the band structures of the four samples (Fig. 4d). Compared with pristine BOH, all BOH-X samples have higher valence band maxima (VBM), and BOH-I has the highest VBM; yet the alterations in VBM are not quite significant. To sum up, the anion exchange induces a moderate change in light absorption only for BOH-I, whereas its influences on the band structures of BOH-X are rather limited.
Photocatalytic performances. Imide derivatives are of major importance for the industries of fine chemicals and pharmaceuticals41–43. We selected the reaction of visible-light-driven photocatalytic oxidative coupling of benzylamine (Fig. 5a) as the model reaction to assess the effect on catalytic performances induced by halide exchange. Fig. 5b shows that with a low ratio (~10%, as mentioned above) of OH– in pristine BOH replaced by Cl–, Br– and I–, the conversion of benzylamine was elevated from 44.1% up to 71.0%, 78.3% and 88.3%, respectively, and the selectivity was between 96.3% and 99.0%. The samples of BOH, BOH-Br and BOH-Cl have almost identical absorption edges, yet their catalytic performances are rather different, indicating that in this case the band structure is not a key influencing factor for the catalytic performance. Since the conversion of benzylamine and the efficiency of carrier separation both follow the same order (BOH < BOH-Cl < BOH-Br < BOH-I), we speculate that the difference in carrier separation and utilization may be attributed to the variation in IEF. As the atomic number of the introduced halide species goes higher, the ionic radius becomes larger, and the charge distribution between the layers become more uneven; the larger electrostatic potential difference between the layers intensifies the interlayer IEF, and thus promotes the carrier separation and utilization. In addition, we also tested the BOH-nPVP nanosheets (that is, the hydrothermal product obtained without PVP) under identical catalytic conditions (Fig. S15), and the sample gave a benzylamine conversion of only 6.5%, far lower than that of BOH. This result indicates that the NRs have a superior photocatalytic activity than the nanosheets, probably because the high aspect ratio of the NRs is conducive to the migration and separation of charge carriers. Subsequently, we assessed the durability of the champion sample BOH-I; after five cycles, the benzylamine conversion was well retained at 77.9%, with selectivity of 98.8% (Fig. 5c). The morphology and microstructure of the catalyst recycled after five runs were also well preserved (as shown in Fig. S16).
Photocatalytic mechanism. We carried out a series of comparative experiments as well as quenching experiments on the possible active species (Fig. 5d). For example, in the case of BOH-I, the conversion of benzylamine is rather low in dark or without the photocatalyst, indicating that both the catalyst and light are essential for this reaction. The conversion in Ar atmosphere was also only marginal, manifesting the essentialness of oxygen. The semiconductor photocatalysts utilize photogenerated electrons and holes to participate in the reaction; the holes can directly oxidize the substrate molecules, and the electrons may reduce molecular oxygen into superoxide radical (žO2−) to oxidize the substrate. In order to unveil the reaction mechanism, we added K2S2O8 or triethanolamine (TEOA) as the scavenging agent for electrons or holes, respectively. The results show that with either agent introduced, the catalytic performances over all four samples declined. Compared with K2S2O8, TEOA would lead to a much more pronounced decline in conversion, particularly for the halide-modified catalysts. These results suggest that both the electrons and holes function as the active species to participate in the catalytic conversion, the latter playing a major role; moreover, after halide-anion exchange, the role of holes (which were mainly collected in anionic layers) becomes even more pronounced. To investigate the influence of žO2− on the reaction pathways, we conducted comparative experiments on radical quenching (Fig. S17). After the introduction of SOD (superoxide dismutase), the selectivity over BOH-I showed a minor decline, from 88.3–74.1%. In addition, the electron paramagnetic resonance (EPR) data on žO2− (Fig. S18) also revealed that, the halide-modified samples are more active in activating O2 into žO2− under visible light; the three BOH-X samples give žO2− signals of similar intensity, implying that žO2− is not responsible for the difference in their catalytic activities. To sum up, it can be concluded that the photogenerated holes are the major active species for oxidizing the substrate molecules, and the promotion in catalytic performances after halide exchange is due to the enhanced IEF and in turn the elevated efficiency of carrier separation.
Theoretical insights on the modulating effect of halide anions. Three bulk BOH-X models were adopted (Fig. 6a–d), with the (001), (100) and (010) facets highlighted for surface cleavage. Owing to the different atomic radii of halides (0.97 Å for Cl, 1.12 Å for Br, 1.32 Å for I), the interlayer spacings of BOH-X are altered (13.6 Å for BOH-Cl, 13.7 Å for BOH-Br, and 13.8 Å for BOH-I), which is in good consistency with the XRD results above.
Furthermore, the calculated DOS (Density of State) confirmed that halide anions in the (001) facet would induce an altered the local electronic structure. Compared with Cl− and Br−, the introduction of I− would greatly promote the uneven charge distribution in the cation and anion layers. As shown in the calculated DOS (Fig. 6e and Fig. S19–22), the p orbital of the introduced halide anion hybridized with the p states of both Bi and O. The introduction of I− would lead to a more pronounced alteration in the bandgap, in contrast to the cases for Br− and Cl−, which is in good accordance with the UV–vis DRS data. However, as the energy levels of valence-shell orbitals are different for holes,as well as the utilization of holes2. The localization of valence electrons and the altered interlayer spacing collectively induce a change in the IEF between the cation and anion layers. Our calculations revealed that the electrostatic potential differences in the halide-modified samples are 0.47 eV (for BOH-Cl), 0.49 eV (for BOH-Br) and 0.51 eV (for BOH-I), all higher than the 0.42 eV for pristine BOH. This trend is in good accordance with the trend for the catalytic conversion of benzylamine (Fig. 6f–g). As a result, we believe that the elevation in the efficiency of carrier separation and utilization is primarily attributed to the enhanced IEF intensity between the layers.