Chemical structure characterization. The structures and morphologies of the as-prepared photocatalysts were observed with the aid of SEM and TEM at various magnifications. As displayed in Supplementary Fig. 1a, the SEM imaging of TiO2 NFs demonstrating the typical electrospun nano-fiber morphology with uniform and smooth. The flexible TiO2 NFs before annealing were continuous with an average diameter about 200 nm centered in the range of 70–310 nm (Supplementary Fig. 1b). After that, the rough surface of m-TiO2 NRs with average diameter about 108 nm were obtained by calcination of the TiO2 NFs in air to remove organic components and then further grinding, as can be seen from Supplementary Fig. 1c. As depicted in high-resolution TEM image of Fig. 1b, the irregular pores of the m-TiO2 NRs (Marked with cyan dotted line) as well as a good crystallinity of the m-TiO2 NRs with a 0.352 nm lattice spacing which matchs (101) diffraction facet of anatase TiO2 could be clearly observed. Simultaneously, the TEM, HRTEM and EDS element mapping were further used for analyzing the structure, composition, and distribution of cocatalysts (e.g., se-Pd NPs etc.) on the TiO2 NRs. It can be evidently seen from the TEM image that a large proportion of se-Pd NPs (~ 12.2 nm) are uniformly distributed on the m-TiO2 NRs support in a semi-encapsulated structure (Fig. 1c and Fig. 2a). Notably, the HRTEM images more accurately and intuitively testified the successful preparation of the semi-encapsulated art configuration between Pd NPs and TiO2 support (Fig. 1d). As indicated by the inset of Fig. 1d, these se-Pd NPs had a lattice spacing of 0.224 nm, corresponding to the (111) facet of metallic Pd as well as confirming that these NPs were well-defined Pd particles14,15. Analogously, the XRD assessments were conducted for demonstrating the phase structure of the as-prepared samples. As displayed in supplementary Fig. 1d, the diffraction patterns of the m-TiO2 NRs samples match well with that of anatase phase TiO2 (PDF#21-1272, Anatase, syn). After se-Pd NPs loaded on the m-TiO2 NRs, the peaks of metallic Pd phase could be found from XRD patterns of the corresponding samples. The very weak peak at around 40.12° could be attributed to the Pd (111) reflection, in good agreement with HRTEM results (Inset of Supplementary Fig. 1d). To verify the presence and distribution of Pd NPs in the as-prepared catalyst, elemental energy-dispersive X-ray spectroscopy (EDS) mapping (Fig. 1e-h) was implemented for analyzing the sample, revealing that Pd mainly distributes the surface of the entire architecture. In addition, the as-prepared m-TiO2 NRs modified with se-Au NPs (~ 10.1 nm), se-Pt NPs (~ 9.6 nm), fe-Pd NPs (~ 4.7 nm), fe-Au NPs (~ 6.2 nm) and fe-Pt NPs (~ 3.5 nm) are clearly identified and visualized by TEM and TEM-EDS mapping (Fig. 2c-f and Supplementary Fig. 2–4).
The high-resolution spectra of XPS regarding Ti 2p, O 1s and Pd 3d were interrogated particularly for determining the chemical state and surface composition of m-TiO2 NRs and se-Pd/m-TiO2 NRs catalysts. About high-resolution XPS spectrum concerning Ti 2p of m-TiO2 NRs (Fig. 3a), two peaks were observed at 458.32 eV (Ti 2p3/2) and 464.02 eV (Ti 2p1/2), and could be earmarked to Ti4+ states16. As for the high-resolution XPS spectrum of O 1s (Fig. 3b), the peaks located at 529.56, 531.42, and 532.7 eV are earmarked to oxygen related to O species in the lattice (OL), O species vicinial the defects or vacancies (OV), and dissociated or chemisorbed (OC) species17,18. Nevertheless, for the se-Pd/m-TiO2 NRs sample, the binding energy of Ti 2p and O 1s was shifted to a greater value, suggesting a potent interaction between se-Pd NPs and m-TiO2 NRs19,20. Figure 3c depicts the Pd 3d spectrum of se-Pd/m-TiO2 NRs. The peaks centered at 336.64 and 341.91 eV demonstrated Pd species with oxide states. The energy of banding at 334.79 and 340.05 eV, relevant to the Pd 3d5/2 and Pd 3d3/2 peaks, were earmarked to metallic Pd21,22. Hence, the initial Pd species over the catalysts were in the shape of the Pd0 state. Additionally, the Low-temperature electron paramagnetic resonance (EPR) analysis was implemented for the evaluation of the presence of OVs. As shown in Fig. 3d, a characteristic oxygen vacancies (OVs) signal with a g factor of 2.003 was observed on the individual m-TiO2 NRs, indicating that OVs could be generated on the surface of m-TiO2 NRs by high-temperature calcination23,24. Meanwhile, compared with m-TiO2 NRs, the se-Pd/m-TiO2 NRs show a sharper and stronger OVs signal after semi-encapsulated Pd NPs assembling, which could be attributed to the existence of massive asymmetric OVs at the TiO2/Pd interface25,26.
To verify the coordination state and electronic structure of Pd in se-Pd/m-TiO2 NRs, the X-ray absorption near-edge structure (XANES) and X-ray absorption fine structure spectra (EXAFS) were conducted at the Pd K-edge. The spectra of XANES display that the absorption edge of Pd K-edge for se-Pd/m-TiO2 NRs placed between that of PdO and Pd foil (Fig. 3e), divulging the valence state of Pd is between 0 and + 2, which may be attribute to the formation of Pd-O or Pd-Ti coordination bond at the TiO2 (101)/Pd (111) interface by Pd atoms on the surface of se-Pd NPs.
Meanwhile, the Fourier-transformed k3-weighted EXAFS spectra (Fig. 3f) confirm the co-existence of Pd-O and Pd-Pd coordination in the se-Pd/m-TiO2 NRs sample with respect to PdO sample and Pd foil. In the current research, the Pd foil was applied for the validation of the se-Pd NPs in the se-Pd/m-TiO2 NRs sample. The nature of metallic NPs or Pd bulk is Pd-Pd coordination. Apparent Pd-Pd coordination in se-Pd/m-TiO2 NRs was detected, demonstrating the presence of Pd NPs; which is consistent with the TEM and HRTEM data. Moreover, the reference of PdO was employed for testifying the contributor of Pd-O coordination. In regard to se-Pd/m-TiO2 NRs, only Pd-O coordination can be observed and no Pd-O-Pd coordination existed. Thus, the Pd-O coordination mode in the sample is different from that of PdO phase. Besides, the bond lengths at 3.14 Å corresponding to Pd-Ti are identified in se-Pd/m-TiO2 NRs (Supplementary Table 2), suggesting the existence of Pd-Ti coordination bond in proximity to Ti atoms at the TiO2 (101)/Pd (111) interface. The plot of wavelet transform of se-Pd/m-TiO2 NRs shows that one notable peak was observed at ~ 6 Å-1 (Fig. 3g), which corresponds to the Pd-O bonding through the comparison of PdO and Pd foil counterparts. Meanwhile, the faintish peaks of Pd-Ti and Pd-Pd with intensity maxima at ca. 7 Å-1 and above 9 Å-1 were also observed (Fig. 3g), which matches well with the EXAFS results. On the basic of previous reports and Supplementary Table 2, the plentiful Pd-O coordination (2.07 Å) is earmarked to Pd-O-Ti species27, contributed through the interfacial between TiO2 (101) and se-Pd NPs28. More importantly, a large number of Pd-O and Pd-Ti bonds were formed at the TiO2 (101)/Pd (111) interface, resulting in the anchoring structure of se-Pd NPs and m-TiO2 NRs has a good stability.
The optical characteristics of the ingeniously constructed catalysts were examined through UV-vis DRS spectra (Fig. 4a). The m-TiO2 NRs exhibit a steep absorption edge placed at ~ 380 nm and are in agreement with the intrinsic bandgap absorption of anatase TiO2 (~ 3.2 eV)29. Loading on m-TiO2 NRs with semi- or full-encapsulated NPs (e.g., Pd, Au, and Pt NPs) will result in a shift to the visible range, which could be attributed to the generation of metallic gray or lattice plasmons in these samples30,31. Notably, decorating m-TiO2 NRs with se-Pd NPs demonstrate a stronger and wider band of absorption in the visiblelight region (from 400 to 800 nm) than other configurations of noble metal decorated-TiO2 NRs.
To deeper understanding the local electric field enhancement of these catalysts by the the localized surface plasmon resonance (LSPR) effect, the finite-difference time-domain (FDTD) simulations were carried out to assess the spatial distribution of local electric-field intensity at the interface between plasmonic metal NPs and m-TiO2 NRs as a function of incident light wavelength. The electric field at the interface of se-Pd, Au, Pt/m-TiO2 NRs were stronger than that of fe-Pd, Au, Pt/m-TiO2 NRs. The interface between the se-Pd NPs and m-TiO2 NRs was “hot” at an excitation wavelength of 650 nm (Fig. 4d-i); this high field intensity suggested more charge-carrier formation32,33. More interestingly, at an excitation wavelength of 550 nm, the electric field intensity at the interface of the se-Pd/m-TiO2 NRs is still comparable to that of se-Au/m-TiO2 NRs as well as far superior to that of fe-Au/m-TiO2 NRs, although the Au NPs possess an excellent SPR characteristic peak at about 550 nm (Fig. 2g-h, and Supplementary Fig. 4h). Therefore, the interface of se-Pd/m-TiO2 would be high-efficient catalytic hotspots of pNRR.
With the intention of exploring in-depth the kinetic behaviors of carriers, the steady-state photoluminescence (PL) measurements were carried out. The PL signals of the catalyts are determined under an excitation light of 365 nm as shown in Fig. 4b. In comparison to the potent emission peak of m-TiO2 NRs, the metallic granular-TiO2 composites demonstrate a rather weaker PL peak, illustrating that the modification plasmonic metal NPs could effectively limit electron-hole recombination34,35. Of theses, the se-Pd/m-TiO2 NRs sample possesses the lowest peak intensity, manifesting the highest separation rate of the photogenerated charge carriers. Relative to the time-resolved photoluminescence decay measurements for the other samples (Supplementary Fig. 5 and Table 3), the se-Pd/m-TiO2 NRs display the longest average decay times of 10.44 ns, indicating that this unique exquisite mode is beneficial for promoting the lifetime of photogenerated electrons.
The Brunauer-Emmett-Teller (BET) surface area and pore structure of the synthesized catalysts were studied by nitrogen adsorption-desorption. As shown in Supplementary Fig. 6, all the samples take possession of a stepwise adsorption/desorption hysteresis, illustrated throughr type IV isotherms, which has the features of mesoporous materials36. The variations in BET surface area and average pore size following the plasmonic metal NPs modification was summarized in Supplementary Table 4. The reduction of BET specific surface area was due to the slight blockage of pores in m-TiO2 NRs after modification of these NPs37. Besides, the average pore size of the prepared catalysts was determined by Barrett–Joyner–Halenda (BJH) approach (Insets of Supplementary Fig. 6, Supplementary Table 4). In general, it can be found that the average pore size of TiO2 NRs modified with larger NPs tends to be enlarged, while smaller NPs tends to be reduced. Such phenomenon could be attribute to the extrusion of the larger NPs on the irregular pores of the substrate TiO2 during the nucleation and growth38,39.
To further understand the pNRR activity of se-Pd/m-TiO2 NRs, the essence of catalytic active site was revealed by temperature-programmed desorption of N2 (N2-TPD) on catalyst surfaces (Fig. 4c). The peak of desorption at about 150°C for N2 physisorption was detected for these catalysts and the integral intensity of peak (i.e., peak area, mean the amount of the physically adsorbed N2) was positively correlated with the catalyst specific surface area (Supplementary Fig. 6 and Table 4). Both se-Pd/m-TiO2 NRs and fe-Pd/m-TiO2 NRs represented a desorption peak at about 420°C for N2 chemisorption, while the se- and fe-Pt/m-TiO2 NRs possess a lower chemisorption desorption peak at about 450 ℃. Besides, the negligible N2 chemisorption of m-TiO2 NRs, se-Au/m-TiO2 NRs and fe-Au/m-TiO2 NRs accounts for their poor adsorption sites. N2 chemisorption is the prerequisite for photocatalytic activation of N2. Accordingly, the se-Pd/m-TiO2 NRs catalyst with more chemically adsorbed N2 is responsible for the higher NH3 production. Thus, the well-designed se-Pd/m-TiO2 NRs presents a good visible light response, greatly promoted energetic charge carriers separation efficiency, effective and abundant N2 adsorption sites, which are beneficial for pNRR.
Photocatalytic performance. The photocatalytic N2 fixation performance of the as-prepared samples appraised by employing water as the solvent and proton source under full spectrum through spectrophotometrically measurement of the generated NH3 with indophenol indicator (Supplementary Fig. 7). To eliminate the interference of environmental adsorption of NH3, the photocatalyst suspension was in succession bubbled with N2 prior to the light radiation for at least 30 min. The se-Pd/m-TiO2 NRs sample shows an excellent NH3 production rate of 635.73 µg gcat.-1 h-1 under full spectrum light irradiation (Fig. 5a), which is about 24.91, 8.42, 1.73, 1.56, 17.04, and 2.49 times higher than that of m-TiO2 NRs (25.52 µg gcat.-1 h-1), se-Au/m-TiO2 NRs (75.51 µg gcat.-1 h-1), se-Pt/m-TiO2 NRs (367.84 µg gcat.-1 h-1), fe-Pd/m-TiO2 NRs (406.42 µg gcat.-1 h-1), fe-Au/m-TiO2 NRs (37.31 µg gcat.-1 h-1), and fe-Pt/m-TiO2 NRs (255.22 µg gcat.-1 h-1), respectively. Moreover, several rigorous contrast experiments were performed for the elimination of the probable contamination from the ambient environment (Fig. 5b). When the photocatalytic procedure was conducted under Ar ambient or in the dark condition, just trace NH3 could be observed, suggesting that the generated NH3 over se-Pd/m-TiO2 NRs indeed originates from the photocatalytic reaction. Besides, no N2H4 was detected during the process (Supplementary Fig. 8c), demonstrating an excellent selectivity for NH3 production.
In addition, the stability of se-Pd/m-TiO2 NRs was investigated through recycling the catalyst (Fig. 5c). Following eight consecutive reaction rounds, approximately 92.37% of its original catalytic activity was retained, indicating remarkable catalytic stability. To appraise the efficiency of light utilization, the apparent quantum efficiency (AQE) of se-Pd/m-TiO2 NRs was estimated under the irradiation of monochromatic light (Supplementary Fig. 10a). In particular, the AQE was implemented to be 0.37% at 375 nm for se-Pd/m-TiO2 NRs, much higher than many reported results (Supplementary Table 5). Hence, the se-Pd/m-TiO2 NRs catalyst presents one of the best performances of photocatalytic N2 fixation reported to date (Supplementary Table 6).
For demonstrating whether photocatalytic N2 fixation on the se-Pd/m-TiO2 NRs was authentic, the photocatalytic N2 fixation under 15N isotope-labeled N2 (with the purity not lower than 99%) was conducted. The generated NH4+ is capable of reacting with phenol and hypochlorite for the formation of 15N-labeled indophenol40,41, which could be evaluated accurately via a liquid chromatography-mass spectrometry (LC-MS). Supplementary Fig. 11a depicts a vigorous mass spectroscopy signal of 14N-labeled indophenol anion at 198 m/z in LC-MS investigation when employing 14N2 as the feeding gas. Of note, the 15N-labeled indophenol negative anion displays a remarkable enhanced mass spectrum signal at approximately 199 m/z in LC-MS analysis (Supplementary Fig. 11b). The signal gives a greater intensity relevant to the 14N:15N natural abundance ratio following 30 min illumination. These data identify that the generate ammonium ion explored inthe current research emanated from N2 photofixation.
To meticulously verify that the pNRR occurred on the surface of se-Pd/m-TiO2 NRs, time-dependent in-situ Fourier-transform infrared (FTIR) assessments were executed to investigate the surface intermediates formed during the pNRR process (See Supplementary Fig. 10b). Whthin the photoreaction for instance under the irradiation of light over se-Pd/m-TiO2 NRs, diverse signals amplified with the time of reaction (Fig. 5d). Peaks at about 3555 cm-1 (i) and 3240 (ii) cm-1 could promptly be earmarked to asymmetric ν(N-H) stretching modes of NH3, while the two absorption bands at 1713 cm-1 (v) and 1667 cm-1 (vii) were allocated to the bending mode of σ(N-H)42,43. Analogously, peaks at ~ 1750 cm-1 (iv) and ~ 1470 cm-1 (viii) are σ(H-N-H) bending vibration of NH344. The signal at about 2873 cm-1 (iii) and 1654 cm-1 (vi) correspond to an antisymmetric deformation vibration of NH4+. The intensity of these signals firstly increased and then stabilized over time, indicating that NH3 or NH4+ were continuously formed on the surface of se-Pd/m-TiO2 NRs, which provides an adequate evidence for pNRR procedure.
The pNRR mechanisms. To gain atomic-level insights into N2 photofixation, the probable coordination geometry of N2 on the surfaces of TiO2 (101), Pd (111), OVs-TiO2 (101)/Pd (111), TiO2 (101)/Pd (111) interface (Supplementary Fig. 12); Au (111), OVs-TiO2 (101)/Au (111) and TiO2 (101)/Au (111) interface (Supplementary Fig. 13); Pt (111), OVs-TiO2 (101)/Pt (111) and TiO2 (101)/Pt (111) interface (Supplementary Fig. 17) was investigated by the density functional theory (DFT) calculations. The configuration of optimized adsorption for a molecule of N2 is a distal end-on mode on TiO2 (101), Pd (111), Au (111), Pt (111), OVs-TiO2 (101)/Pd (111), OVs-TiO2 (101)/Au (111), single Ti sites at the TiO2 (101)/Au (111) interface, OVs-TiO2 (101)/Pt (111), single Pt sites at the TiO2 (101)/Pt (111) interface and a side-on bridging mode on Ti-N2-Pd dual sites at the TiO2 (101)/Pd (111) interface (Fig. 6, Supplementary Fig. 14 and Supplementary Fig. 18). The energy of adsorption for the N2 side-on bridged on bimetallic Ti-Pd centers (-0.64 eV) is excel than that for the N2 end-on bound to a Ti sites of TiO2 (-0.29 eV), Pd site of Pd (111) (-0.48 eV), OVs-TiO2 (101)/Pd (111) (-0.56 eV), OVs-TiO2 (101)/Au (111) (-0.42 eV), Ti sites at TiO2 (101)/Au (111) interface (-0.38 eV), OVs-TiO2 (101)/Pt (111) (-0.43 eV), and Pt sites at TiO2 (101)/Pt (111) interface (-0.51 eV) (Fig. 8a and Supplementary Fig. 21). However, the OVs with respect to a single metal site was thermodynamically desirable for the activation of N ≡ N bond, demonstrating the rationality of general defect engineering strategy45, the bimetallic Ti-Pd center illustrated substantial dominance to the OVs.
To further insight into the differences of two N atoms (N1 and N2) at the various models, the differential charge diagram was employed for reflecting the polarization of adsorbed N2 molecule. The charge discrepancy between two adsorbed N atoms on Ti-Pd dual sites was 0.11e (-0.15|e| vs -0.26|e|), which is greater than 0.02e (0|e| vs -0.02|e|) for N2 adsorbed on Ti sites of TiO2 (101), 0.07e (-0.02|e| vs -0.09|e|) for N2 adsorbed on Pd sites of Pd (111), 0.10e (-0.21|e| vs -0.11|e|) for N2 adsorbed on OVs of TiO2 (101)/Pd (111), 0.05e (-0.06|e| vs -0.01|e|) for N2 adsorbed Ti sites at TiO2 (101)/Au (111) interface, as well as 0.08e (-0.14|e| vs -0.06|e|) Pt sites at TiO2 (101)/Pt (111) interface (Fig. 7, Supplementary Fig. 15 and Supplementary Fig. 19), indicative of more efficacious polarization of the N2 adsorbed on Ti-Pd dual sites.
Meanwhile, Fig. 8b illustrates that the electrons could be preferably agglomerated on the atom of N1 bonded with Ti-end at the bimetallic Ti-Pd active centers. This type of imbalanced charge distribution differentiates the two atoms of N for adsorbed N2, initiating the associative distal pathway at the Ti-end of Ti-Pd dual sites7. For deeper comprehending the polarization effect of Ti-Pd dual sites on adsorbed N2 molecules, the bond length of N ≡ N bond was also used as the descriptor for polarization of N2 molecules. As displayed in Fig. 8c, the bond of N ≡ N for N2 bridged on Ti-Pd dual sites was enhanced to 1.222 Å, analogous to the double bond lengths of azobenzene (PhN = NPh) (1.255 Å) and diazene (HN = NH) (1.201 Å)11. On the contrary, the N-N bond lengths of N-N for N2 adsorbed on single metal sites (e.g., Ti, Pd, Au and Pt) or OVs is equivalent to that of free N2 (1.155 Å), holding a triple-bond order (Fig. 8c and Supplementary Fig. 22). Therefore, the effective N2 adsorption with bimetallic Ti-Pd center of se-Pd/m-TiO2 NRs is beneficial for boosting the activity in pNRR.
The N bridged on Ti-end of Ti-Pd center illustrated a greater electron density and subsequently it was more simply hydrogenated in comparison to that on Pd-end (Fig. 8b). Futhermore, protonation initiating from N on Ti is an obvious process of gradually lengthening the N ≡ N bond, resulting in the formation of NH3, whereas protonation starting from Pd does not (Fig. 8d). Then, the Gibbs free-energy diagrams of pNRR pathways relative to se-Pd/m-TiO2 NRs with Ti-Pd dual sites were unfolded (Fig. 8e). The initial adsorption of N2 bridged on Ti-Pd dual sites releases 0.11 eV of free energy and polarizes N2 molecule with charge redistribution. The Gibbs free energy change (∆G) for the hydrogenation of *N-N to *HN-N (blue line) was only 0.57 eV (from − 0.11 to 0.46 eV), while *N-N was hydrogenated to *N-NH with the ∆G of 0.75 eV (from − 0.11 to 0.64 eV). Next, the subsequent hydrogenation process including *N-HN to *NH-NH (∆G = -0.39 eV), *NH-NH to *NH2-NH (∆G = -0.13 eV) and *NH2-NH to *NH3-NH (∆G = -0.84 eV) steps relative to the optimized pNRR procedure (blue line) is facile because the ∆G values of these process are negative. Inversely, the other possibilities of pNRR paths (gray line) containing *N-HN to *NH2-N, *NH-NH to *NH-NH2, and *NH2-NH to *NH2-NH2 (gray line, ∆G values are 0.27, 0.31 and − 0.22 eV, respectively) require overcoming a certain reaction energy barrier or hold a lower ∆G compare to the optimal pNRR process. Afterwards, the first NH3 molecule formed at the Ti-end of Ti-Pd dual sites absorbs the energy of 0.1 eV and further releases. Eventually the remaining N atom adsorbed on the Pd-end of Ti-Pd dual sites undergos a continuous hydrogenation process such as *NH to *NH2 (∆G = -1.06 eV) and *NH2 to *NH3 (∆G = -0.06 eV) paths then overcome a reaction energy barrier of 0.88 eV to release the second NH3 molecule. Taken together, the unbalanced charge difference of N2 molecule adsorbed at the Ti-Pd dual sites enables the pNRR process develop a distal association path dominated by the Ti-end, which greatly diminishes the barrier of activation energy and improves photocatalytic NH3 synthesis efficiency.