Samples characterization. Layered BP and WO3 − x (WO) were synthesized by previous reported methods38,39, and their layer and nanowire morphologies were shown in TEM images (Supplementary Figs. 1a-d), respectively. HR-TEM images (Supplementary Figs. 1b and 1d) show the 0.26 and 0.38 nm lattices spacing assigned to {040} facets of BP layers40 and {010} facets of W18O49 nanowires41, respectively. BP/WO heterostructures were synthesized by using simple mixing in solution under ultrasonication, and TEM images (Figs. 1a and b) showed the structures of 10-BP/WO with 10 wt% BP in composition. The 0.38 and 0.26 nm lattice spacing assigned to WO nanowires and layered BP are observed on HR-TEM image (Fig. 1c) of 10-BP/WO heterostructure. The constructed heterostructures were further characterized by the scanning TEM (STEM) coupled with energy-dispersive x-ray spectroscopy (EDX). The high-angle annular dark-field (HAADF) STEM image (Fig. 1d) shows the different contrast on the positions of nanowires and nanosheets in the heterostructures. Meanwhile, the distributions of P, O and W elements are labeled in EDX mapping images (Fig. 1e-h). O and W atoms are mainly located on nanowires, while, P atoms are located on nanosheets, verifying the BP/WO heterostructures. Crystal structures of BP/WO heterostructures were characterized by x-ray diffraction (XRD, Supplementary Fig. 2), and their XRD patterns are consistent with those of BP and WO without any other peaks.
UV-Vis-NIR diffuse reflectance spectra (DRS, Fig. 1i) shows the optical properties of different samples. 2D BP has a narrow band gap and a broad light response in visible region, and WO nanowires exhibit strong SPR absorption in Vis-NIR region. 10-BP/WO heterostructure has a similar SPR band with that of WO, and the slight enhancement UV-Vis absorption is attributed to 10 wt% BP loading. Chemical structures of 10-BP/WO were analyzed by Raman spectra (Supplementary Fig. 3) and Fourier transform infrared (FTIR) spectra (Supplementary Fig. 4). Weak signal of P-O-W bonds was detected at 1116 cm− 1, indicating that a small amount of P-O-W bonds exist on interface to connect WO and BP42,43. Oxygen vacancies in 10-BP/WO heterostrutures and plasmonic WO were detected by electron paramagnetic resonance (EPR) spectra and similar concentrations were observed in Fig. 1j[44]. X-ray photoelectron spectra (XPS) show the chemical state of elements in different samples. For BP, XPS of P 2p (Fig. 1k) are deconvoluted into two 129.5 and 130.4 eV bands, assigned to 2p3/2 and 2p1/2, respectively45, and another bond at 133.5 eV is attributed to the oxidized phosphorus (PxOy)46. While, the bands of P 2p3/2 and 2p1/2 on 10-BP/WO heterostructure have 0.5 eV shift toward high binding energy direction, implying the decreased electronic affinity on BP. XPS of W 4f on WO (Fig. 1l) show two broad bands which can be divided into four bands with binding energy of 35.1, 37.1, 36.1 and 38.1 eV assigned to W5+ 4f7/2, W5+ 4f5/2, W6+ 4f7/2, and W6+ 4f5/2, respectively47. The W5+ is generated from the W6+ reduction during the solvothermal process. For 10-BP/WO, W 4f bands have a 0.1 eV shift toward low binding energy direction, showing the increased electron affinity on WO. Those results indicate the electrons transfer from BP to WO. P = O and P-OH bonds in 10-BP/WO were detected by O 1s XPS (Supplementary Fig. 5), implying the partial oxidation on layered BP, and it provides the possible for BP to connect WO by W-O-P bonds.
The possiblity of charge transfer in BP/WO heterostructures were analyzed by the simulations. The density functional theory (DFT) was used to simulate charge interactions between W18O49 (001) surface and 2D BP to unravel the bonding mechanism. The charge density difference \(\varDelta \rho ={\rho }_{{\text{W}}_{18}{\text{O}}_{49} \left(001\right)/\text{P}}-{\rho }_{{\text{W}}_{18}{\text{O}}_{49}\left(001\right)}-{\rho }_{\text{P}}\) was calculated as shown in Fig. 2a. Results show that the charge density is redistributed by forming electron-rich and hole-rich regions at interfaces with small amount of P atoms bonded to O atoms. The electrons can be transferred from P to W18O49 (001) using an O bridge. To confirm our results, the work functions of the W18O49 (001) surface and black phosphorene were evaluated further. The vacuum levels Ev and Fermi levels Ef wereobtained as Ev(W18O49 (001)) = 2.81 eV, Ev(P) = 1.25 eV, Ef(W18O49 (001)) = -2.80 eV, and Ef(P) = -3.84 eV. The work function W can be obtained by W = Ev - Ef, further we get WW18O49 (001) = 5.61 eV and W(P) = 5.09 eV. The results quantitatively show that electrons can be transferred from BP to W18O49 (001) surface due to increasing work functions. The electron transfer process was verified by DRS, EPR and XPS spectroscopies, as shown in Figs. 2b-d. Under Vis-NIR irradiation, SPR band of 10-BP/WO becomes stronger with a blue shift as irradiation time prolongs, indicating the increasing electron density on plasmonic WO. The increasing electron density is also observed clearly on EPR spectra of 10-BP/WO after 15 min Vis-NIR irradiation. While, the DRS (Supplementary Fig. 6) and EPR spectra of plasmonic WO do not have changes under Vis-NIR-irradiation, confirming the electron transfer from BP to WO in BP/WO heterostructures. Moreover, the electron transfer frocess can also be observed on XPS of W5+ 4f (Fig. 2d) which have a pronounced increase on intensity after Vis-NIR irradiation, and the amount of W5+ is increased from 19.2 to 34.5 mol% in heterostructures.
The kinetic process of electron transfer in different samples under light irradiation was characterized by using ultrafast transient absorption spectroscopy combined with the femtosecond laser excitation. As shown in transient absorption spectra of BP (Fig. 3a), the electron on VB of BP can be excited in a broad visible spectral region, and the highest photobleaching point is observed at 5.0 ps. Then, the excited electron has a decay resulting in a continuous decrease on spectra as time prolongs. At 2.0 ns, ∆A is near zero meaning the completed decay of photoelectrons, and BP is recovered from an excited state to the ground state. So, the total lifetime of photoelectron on BP is about 2.0 ns. While, for plasmonic WO, due to the short lifetime of plasmonic carriers48,49, the fast decay (< 100 ps) process and weak signals on transient absorption spectra are observed in Fig. 3b. Therefore, the transient absorption spectra (Fig. 3c) of 10-BP/WO heterostructures are mainly from BP, and is much weaker than the spectra of pure BP. The photoelectron decay time (1.0 ns) in 10-BP/WO heterostructure is shorter than that in BP. More detailed photoelectron decay process was studied by the transient absorption kinetic curves (Fig. 3d) collected at λpump = 675 nm. Two decay processes of direct decay (τ1) with short time and trapped state decay (τ2) with long time are simulated based on the kinetic curve. τ1 and τ2 are simulated to be 0.2 ns and 1.8 ns for BP, and 0.16 ns and 1.3 ns for 10-BP/WO. The results indicate that a part of photoelectrons on BP directly transfers to WO, leading to a fast decay process of photoelectrons on BP[50]. Consequently, the photo-excited photoelectrons on BP can transfer to plasmonic WO, and the continuous electron injection can increase the electron density for enhanced SPR.
Photocatalytic performance. The photocatalytic CO2 reduction reaction (CO2-RR) performance is determined in a reactor using a gas-solid configuration. Under Vis-NIR (> 420 nm) light irradiation, 10-BP/WO heterostructures generate 78.3 µmol g− 1 CO in 3 hours (Fig. 4a), more than 7- and 17-fold higher than 10.9 and 4.4 µmol g− 1 of plasmonic WO and BP, respectively. CO generation rates (Supplementary Fig. 7) are calculated to be 26.1, 3.6, and 1.5 µmol h− 1 g− 1 with a selectivity of 98, 97, and 84 % for 10-BP/WO, WO, and BP, respectively. Meanwhile, a small amount of CH4 is detected for the three samples (Supplementary Fig. 7). Different heterostructures with 2.5, 5.0, 7.5, 10, 12.5, and 15 wt.% BP in composition were synthesized as photocatalysts, and 10-BP/WO exhibited the optimal performance on CO generation (Supplementary Fig. 8). CO generation rates for 2.5-BP/WO, 5.0-BP/WO, 7.5-BP/WO, 10-BP/WO, 12.5-BP/WO, and 15-BP/WO are calculated to 7.3, 9.7, 19.1, 26.1, 20.5, and 11.3 µmol h− 1 g− 1 (Fig. 4b), with a selectivity of 91, 95, 95, 98, 98, and 94 %, respectively. With pyroelectric effect and narrow band gap, the increased amount of BP will generate more pyroelectric electrons and photoelectrons to transfer to plasmonic WO which can enhance hot electron generation for CO2-RR. However, as the amount of BP beyond 10 wt%, the plasmonic WO nanowires are covered by BP layers which hinder the SPR excitation for photocatalytic CO2-RR. CO generated from CO2 is verified by using 13C-labled 13CO2 for photocatalysis over 10-BP/WO, and 13CO are detected by the mass spectrometry (MS) as main product (Fig. 4c).
Considering the broad light absorption of 10-BP/WO heterostructures in UV-Vis-NIR region, their photocatalytic performance under different light irradiation was investigated as shown in Fig. 4d. UV- and Vis-irradiation only generates 7.2 and 7.9 µmol g− 1 CO in 3 hours, respectively. While, NIR-irradiation promotes CO generation to be 12.5 µmol g− 1 in 3 hours. As Vis-NIR light is irradiated, CO generation is improved to 86.7 µmol g− 1, 7-fold higher than those under NIR irradiation. While, 109.4 and 148.5 µmol g− 1 CO is generated under UV-Vis and full-spectrum (UV-Vis-NIR) light irradiation, respectively. CO generation rates (Fig. 4e) are calculated to 2.4, 2.6, 4.2, 36.5, 28.9, and 49.5 µmol g− 1 h− 1 CO for UV, Vis, NIR, UV-Vis, Vis-NIR, and UV-Vis-NIR light irradiation, with a selectivity of 95, 94, 99, 100, 98, and 97 %, respectively. Only a littel CH4 (< 1 µmol g− 1 h− 1) is detected during CO2-RR (Fig. 4e). The stability of 10-BP/WO heterostructures during photocatalytic CO2-RR were investigated as shown in Fig. 4f. Light irradiation contianing UV can cause the decreased CO generation over 10-BP/WO, and CO generation rates are reduced from 49.5 to 17.8 and 36.5 to 8.9 µmol g− 1 h− 1 for UV-Vis-NIR- and UV-Vis-irradiation, respectively, during three times repeating CO2-RR. The possible reason is the BP decompostion caused by UV-induced oxidation. Fortunately, 10-BP/WO heterostructures have a sustainable performance on CO generation during Vis-NIR-driven CO2-RR. The chemical structure of 10-BP/WO after Vis-NIR-irradated photocatalysis is analyzed by XPS, and no obvious changes on P 2p and O 1s XPS (Supplementary Figs. 9 and 10) demonstrate their stable structures during CO2-RR. The apparent quantum efficiency (AQE) over different samples during CO2-RR was measured under monchrome light irradiation. AQE of 10-BP/WO (Fig. 4g) has a similiar trend with SPR band confirming that photocatalytic CO generation is attributed to SPR excitation. More interesting is that AQE at 900-nm is higher than those in visible region, and it means NIR-irradiation can enhance CO generation. However, AQE (Fig. 6h) of plasmonic WO shows a completely different result, much lower than that of 10-BP/WO in Vis-NIR region. A great difference on AQE indicates that pyroelectric BP can promote CO2-RR for high-selective CO generation. To enclude the influence of photoelectron on BP by Vis-excitation, NIR (> 800) light which can excite SPR of WO but weakly excite BP is used to irradiate BP, WO, 10-BP/WO for CO2-RR (Fig. 4i), and none, 1.1, and 4.2 µmol g− 1 h− 1 CO are generated, respectively. NIR-irradiation make SPR exciation to generate hot electrons and thermal energy simulatously on WO, and the thermal energy can trigger the pyroelectric BP for carriers, facliating electron transfer from BP to WO and hot electron generation for CO2-RR.
The photothermal effect on 10-BP/WO was further investigated by the thermal camera. Thermal images of 10-BP/WO loaded glass plate (Supplementary Fig. 11) with different Vis-NIR light irradiation time are shown in Fig. 5a. Under Vis-NIR light irradiation, the temperature on 10-BP/WO have a rapidly increase from initial 24.3 oC to 57.1 oC in 2 min, and then to 70.8 oC in 4 min. Finally, the temperature is stablized at 86 oC after 10 min irradiation. Thermal images and temperatures of 10-BP/WO under different light irradiation are measured in Fig. 5b and Supplementary Figs. 12–16. The temperature is stablized at 42.5, 62.2, 83, 66.5, 86.3 and 98.5 oC with the ∆T of 15.6, 37.2, 59.7, 40.5, 62.0 and 72.2 oC under 10 min UV, Vis, NIR, UV-Vis, Vis-NIR and UV-Vis-NIR light irradiation, respectively. It is clear that the main contribution on photothermal energy of 10-BP/WO comes from NIR-irradiation, attributing to the strong SPR absorption in NIR region. The photothermal energy generated on plasmonic WO can trigger the pyroelectric BP of heterostructures for carriers and pyroelectric field faciliating the electron transfer to plasmonic WO for CO2-RR. To verify the contribution of pyroelectric effect on photocatalysis, CO2-RR under heating (85 oC), Vis-irradiation (420–780 nm) and heating/Vis-irradiation over 10-BP/WO were tested as shown in Fig. 5c, and 1.4, 7.9 and 13.5 µmol g− 1 CO were obtained in 3 hours, respectively. Only heating on 10-BP/WO do not improve CO2-RR. However, heating can greatly enhance Vis-driven CO2-RR on 10-BP/WO. The possible illustration is that the heating triggers the pyroelectric BP to convert thermal energy for carriers and transfer to WO promoting plasmonic CO2-RR. The influence of heating on CO2-RR over WO and BP as catalysts were also tested in Fig. 5d, and the heating and Vis-irradiation do not play synergistic roles on CO2-RR. Consequently, SPR and pyrolectric effect in BP/WO heterostructures play an synergistic role on Vis snd NIR light harvesritng for continnous eletron injection and hot electron generation for high-selective CO generation.
The detailed reaction pathway during CO2-RR over BP/WO was monitored by in situ FTIR spectra, as shown in Fig. 6a. Before CO2 absorption in the dark, no remarkable signals are observed on the spectra. After CO2 saturated absorption for 30 min in the dark, the bands of bidentate carbonate (b-CO32−, 1635 cm− 1), monodentate carbonate (m-CO32−, 1507 cm− 1), and bicarbonate (HCO3−, 1416 cm− 1) are detected on the surface of 10-BP/WO51,52. When Vis-NIR light is irradiated, the intensity of b-CO32 bonds has a increase as time prolongs, implying b-CO32 as dominant absorption mode on surface. Moreover, the bicarbonate HCO3− bands become stronger implying that HCO3− is intermediate of CO2 reduction. So, the possible reaction pathway of CO generation is that CO2 adsorbed on WO surface forms b-CO32−, and then reacts with hot electron and a proton to form intermediate HCO3−. HCO3− further reacts with electrons and protons to form CO and H2O. No absorbed *CO is detected by in situ FTIR spectra, indicating the fast desorption of CO as product. The DFT calculations were used to simulate the detailed reaction pathway of CO2-RR. The Gibbs reaction free energy change (ΔG) during CO2 reduction over WO3 − x was calculated as shown in Fig. 6b. CO2 is absorbed on the surface of WO3 − x to be *CO2 with ΔG = 0.35 eV. Then, one electron and proton are obtained by *CO2 to be *COOH with ΔG = -0.53 eV. After reaction with the second electron and proton, *COOH becomes *CO with a ΔG = -0.45 eV. *CO reaction for next step has two different pathways, one is to desorb from surface to be free CO with a ΔG = 0.65 eV, and the other one is to reaction with one electron and proton to be *CHO with a ΔG = 0.91 eV which is generally considered as a key step for CH4 generation. In situ FITR and DFT simulations demonstrate that CO generation is the main reaction pathway of CO2-RR over plasmonic WO as photocatalyst, and it is consistent with the experiment results.
Based on above results, the detailed multi-synergetic processes and the possible photocatalytic mechanism are described in schematic diagram (Fig. 6c). With abundant oxygen vacancies, a defect band (DB) formed on the bottom of the conduction band (CB) in WO nanowires. With abundent W5+, the DB is occupied by the electrons generating SPR on WO in Vis-NIR region, and the large energy difference between valence band (VB) and lowest unoccupied CB restricts electron excitation from VB to CB under Vis-NIR irradiation. While, the electrons on DB can be excited by SPR to be hot electron in WO. The fast decay process of hot carriers leads to short lifetime of hot electron (< 200 ps) which has been observed on transient absorption spectra (Fig. 4). The nonradiative decay of hot carriers generates photothermal energy causing the low efficiency of plasmonic photocatalysis. Combination of pyrolectric BP and plasmonic WO forms heterostructures, and the plasmonic photothermal energy on WO nanowires can transfer to BP by the thermal radiation, triggering pyroelectric field for carriers on BP. Moreover, the visible irradiation on BP also generates photo-carriers. For plasmonic WO, SPR-excitation can generate hot electrons at high energy levels which are acitve for photocatalysis. W-O-P bonds connected WO and BP are demonstrated to be channels for electron transfer (ET) from BP to WO. Therefore, the electrons generated by photo-excitation and pyroelectric effects on CB of BP can both transfer to DB of WO with low energy level. The continuous electron injection from BP to plasmonic WO will increase the electron density on DB for stronger SPR which can generate more active hot electrons. Moreover, the continuous electron injection restricts hot electron deacy from high energy states to low energy states, promoting hot eletrons for CO2 reduction. Meanwhile, the holes left on VB of BP have an oxidization reaction. Under UV-irradiation, electron on VB of WO can be excited to DB of WO leading to the increased electron density for stronger SPR, and it is favor for plasmonic CO2-RR. Meanwhile, the holes left on VB of WO are active to decompose BP, causing the unsustainable activity of heterostructures on photocatalytic CO2-RR. Consequently, multi-physical effects including plasmonic hot carriers and photothermal effect of WO, pyrolectric effect and band excitation of BP, and electron transfer through W-O-P bonds play synergetic roles on enhancing SPR and hot electron generation, boosting high-selective CO2-RR for CO generation.