Introduction

Excessive emissions of carbon dioxide (CO2) into the atmosphere have disrupted the natural carbon cycle, leading to severe environmental consequences, particularly the exacerbation of the greenhouse effect1,2,3,4,5,6. In response to this urgent global issue, harnessing abundant, clean, and inexhaustible sunlight to convert CO2 into valuable solar fuels has emerged as a promising strategy7,8,9,10,11,12,13. However, the effectiveness of CO2 photoreduction is constrained by the challenging chemisorption and activation of CO2 molecules on catalysts, primarily due to the high dissociation energy of the C=O bond (~750 kJ mol–1)14,15,16,17,18,19. Therefore, the development of advanced photocatalysts proficient in activating CO2 has become a pivotal concern within the realm of photocatalytic CO2 reduction20,21,22,23,24,30,31,32,33. Preliminary density functional theory (DFT) calculations indicate that CO2 molecules adsorbed onto the Nb2O5 undergo changes in both bond lengths and angle compared to free ones. Additionally, the two oxygen atoms of CO2 can form chemical bonds with niobium atoms of Nb2O5, suggesting the potential of Nb2O5 for activating stable CO2 molecules during CO2 photoreduction. Nevertheless, unitary Nb2O5 exhibits poor photocatalytic performance resulting from sluggish electron/hole separation and charge transfer kinetics. Consequently, develo** hybrid heterojunctions involving Nb2O5, capable of activating CO2, promoting charge carrier transfer kinetics, and separating them to improve reduction efficiency, remains a significant yet challenging endeavor.

S-scheme heterojunctions, integrating both reduction and oxidation photocatalysts, have proven effective in spatially separating photogenerated charge carriers with robust redox capabilities34,35,36,37,38,39. Conventionally, constructing S-scheme heterojunctions involves initially acquiring photocatalyst I and subsequently applying photocatalyst II onto I through methods such as in situ growth or electrostatic self-assembly40,41,42. However, these post-hybridization methods cannot ensure the maximum contact area between the two phases at atomic levels, thus impeding the efficient interfacial transport of photogenerated carriers and compromising photocatalytic efficiency (Fig. 1a)43. In this study, we designed an S-scheme heterojunction by coupling Nb2O5 with indium oxide (In2O3), an oxidation photocatalyst with a narrow bandgap (~2.9 eV) and visible light absorption44,45,46,47,48,49,50,51. By mixing the precursors of both phases in the same electrospinning solution, In2O3 and Nb2O5 are simultaneously formed during the high-temperature calcination of the electrospun nanofibers. This “one-pot” preparation method ensures maximum phase contact without any hindrance, providing an unimpeded transport route and promoting interfacial charge transfer between In2O3 and Nb2O5 (Fig. 1a). Analysis using femtosecond transient absorption spectroscopy (fs-TAS) revealed ultrafast photoelectron transfer from the In2O3 CB to the Nb2O5 valence band (VB), inhibiting self-carrier recombination, effectively segregating powerful photoelectrons in the Nb2O5 CB and the holes in the In2O3 VB, as well as prolonging the long-lifetimes of the nanohybrids. Also benefiting from the chemisorption and activation of CO2 molecules on the catalyst, the resulting In2O3/Nb2O5 heterojunctions demonstrated enhanced performance in CO2 photoreduction. This work provides insights into ultrafast charge transfer at the S-scheme heterojunction interface through fs-TAS investigations, offering an essential understanding for the development of heterojunctions and broadening their potential applications in artificial photosynthesis.

Fig. 1: Morphology and structure of In2O3/Nb2O5 heterojunctions.
figure 1

a Schematic and the design concept of this study. OP and RP stand for oxidation photocatalyst and reduction photocatalyst, respectively. b FESEM image and EDX spectrum, (c) TEM image, and (d) HRTEM images of In2O3/Nb2O5 heterojunctions (IN10). e High-angle annular dark-field (HAADF) image and EDX elemental map**s of In and Nb elements in IN10 at different magnifications. f XRD patterns of In2O3, Nb2O5 and INx. g UV-vis spectra of In2O3, Nb2O5, and IN10.

Full size image

Results and discussion

Characterizations and charge separation mechanism of In2O3/Nb2O5 heterojunctions

The In2O3/Nb2O5 heterojunctions, synthesized through a one-step electrospinning procedure, are designated as INx, where I and N represent In2O3 and Nb2O5, respectively, and x signifies the weight percentage of Nb2O5 relative to In2O3. The precise Nb2O5 content of all composites was determined using inductively coupled plasma-atomic emission spectrometry (ICP-AES), and the results are presented in Supplementary Table 1. Field emission scanning electron microscopy (FESEM) images reveal distinctive morphologies: pure In2O3 exhibits a fibrous structure, while pristine Nb2O5 displays a clubbed pattern (Supplementary Fig. 1). The In2O3/Nb2O5 heterojunction (IN10) shows a uniform fibrous morphology with a rough surface and diameters below 100 nm, as observed via both FESEM and transmission electron microscopy (TEM) (Fig. 1b, c). High-resolution TEM (HRTEM) images of IN10 reveal discernible two-phase grain boundaries with lattice fringes corresponding to In2O3 and Nb2O5, respectively (Fig. 1d). The random distribution of In2O3 and Nb2O5 nanoparticles within the nanofibers ensures close contact, facilitating unimpeded interfacial transport and efficient separation of photoexcited charge carriers (as discussed below). Energy-dispersive X-ray (EDX) analysis (inset in Fig. 1b) and elemental map**s of IN10 (Supplementary Fig. 2) unambiguously confirm the presence of In, Nb, and O elements within the nanohybrid, offering compelling evidence for the formation of In2O3/Nb2O5 heterojunctions. Figure 1e displays enlarged elemental map**s targeting a grain-to-grain area within IN10. It is apparent that the distribution of Nb and In elements exhibits a non-overlap** pattern along the interface of the two phases. X-ray diffraction (XRD) patterns of pure In2O3 nanofibers (Fig. 1f) indicate the monoclinic phase (PDF#71-2195), while characteristic peaks corresponding to Nb2O5 (PDF#30-0873) emerge in the INx composites at 20 wt.% Nb2O5 content, signifying the successful synthesis of In2O3/Nb2O5 heterojunctions. UV-vis diffuse reflectance spectroscopy (DRS) delineates the optical properties of In2O3, Nb2O5, and In2O3/Nb2O5 composites (Fig. 1g). The absorption edges of pristine In2O3 and Nb2O5 are positioned at 425 and 390 nm, corresponding to bandgaps of 2.9 and 3.2 eV, respectively. Compared to pure Nb2O5, the slightly improved UV and visible light absorption characteristics of IN10 suggest successful hybridization due to the strong light absorption capacity of In2O3.

X-ray photoelectron spectroscopy (XPS) was utilized to analyze the surface chemical states and compositions of the resulting samples. The survey spectrum of IN10 reveals the presence of In, Nb, and O elements in the hybrid nanofibers (Supplementary Fig. 3). In the high-resolution In 3d XPS spectra (Fig. 2a), two distinct peaks are observed at 444.5 and 452.0 eV, corresponding to the 3d5/2 and 3d3/2 states of trivalent In3+ within In2O3, respectively. The signals associated with Nb 3d5/2 and Nb 3d3/2 appear at 206.9 and 209.7 eV, respectively, confirming the existence of pentavalent Nb5+ in the samples (Fig. 2b)30. The O 1 s XPS spectra of In2O3, IN10, and Nb2O5 (Supplementary Fig. 4a–c) consistently exhibit peaks attributed to lattice oxygen and surface hydroxyls (-OH). Notably, in the IN10 composite, the binding energies (BEs) of In 3d show negative shifts compared to pure In2O3, while the peaks of Nb 3d shift towards higher BEs in comparison with bare Nb2O5. These shifts suggest that electrons are transferred from Nb2O5 to In2O3 upon contact, indicating the creation of a directional interfacial electric field (IEF) from Nb2O5 to In2O3, and simultaneously leading to the bending of the energy bands at the interfaces. To further substantiate the electron transfer process between In2O3 and Nb2O5, the work function (Φ) was determined through DFT simulations by calculating the energy difference between the vacuum and Fermi levels, based on the electrostatic potential of the materials. As illustrated in Fig. 2c, d and Supplementary Figs. 5, 6, the estimated Φ value of In2O3 (111) is larger than that of Nb2O5 (100), with both facets exhibiting the lowest surface energy (Supplementary Tables 2 and 3). Consequently, Nb2O5 possesses a higher Fermi level (EF) than In2O3, promoting the transfer of electrons from Nb2O5 to In2O3 until reaching the same EF at the interface (Fig. 2e). These analyses align with the aforementioned XPS results and contribute to the efficient separation of photogenerated charge carriers (as discussed below).

Fig. 2: Electron transfer between In2O3 and Nb2O5 within the heterojunctions.
figure 2

The high-resolution XPS spectra of (a) In 3d, and (b) Nb 3d of In2O3, Nb2O5, and IN10. Calculated electrostatic potentials of (c) In2O3 (111) and (d) Nb2O5 (100) slabs. The yellow, green, and purple spheres represent In, Nb, and O atoms, respectively. e The formation of In2O3/Nb2O5 S-scheme heterojunction, and the proposed charge transfer and separation mechanism. RR and OR stand for reduction reaction and oxidation reaction, respectively.

Full size image

To investigate the photoinduced charge transfer mechanism of the In2O3/Nb2O5 heterojunctions, the band structure of In2O3 and Nb2O5 was first studied. According to the ultraviolet photoelectron spectroscopy (UPS) spectra (Supplementary Fig. 7a, b), the VB maximum of In2O3 and Nb2O5 is estimated at 2.41 and 2.28 V (vs. standard hydrogen electrode, SHE), respectively. Combined with the bandgap values disclosed in Supplementary Fig. 7c, the CB minimum is established as –0.49 V for In2O3 and –0.96 V for Nb2O5 (Supplementary Fig. 7d)33. Based on the previous discussion involving XPS and DFT results, it is evident that the EF of Nb2O5 is higher than that of In2O3, which induces the migration of electrons from Nb2O5 to In2O3, leading to the creation of an IEF at the interface, as well as band alignment upon contact. Under light irradiation, electrons in the In2O3 and Nb2O5 VBs are initially excited to their respective CBs. Due to the bent energy bands, the IEF with the direction from Nb2O5 to In2O3, and the Coulomb attraction between electrons and holes, the photogenerated electrons in the In2O3 CB tend to transfer to the Nb2O5 VB and recombine with its holes. These consumed photoelectrons and photoholes are characterized by their weak reduction and oxidation capacities. As a result, photogenerated charge carriers with strong redox capabilities within the Nb2O5 CB and the In2O3 VB undergo separation and preservation, actively participating in subsequent photoreactions. This charge transfer pathway implies the formation of an S-scheme heterojunction between In2O3 and Nb2O5, visually depicted in Fig. 2e.

In situ irradiated XPS was conducted to verify the S-scheme charge transfer route within the In2O3/Nb2O5 heterojunctions. As shown in Fig. 2a, b, upon exposure to light, the BEs of In 3d in IN10 display notable positive shifts, while the Nb 3d peaks shift towards lower BEs, with respect to those in the dark. These observed BE shifts provide strong evidence for the transfer of photogenerated electrons from In2O3 to Nb2O5, thus corroborating the proposed S-scheme photocatalytic mechanism. The efficiency of charge separation in the In2O3/Nb2O5 S-scheme heterojunctions was assessed through steady-state photoluminescence (PL) and photoelectrochemical measurements. The PL emission intensity of the IN10 composite is weaker than both pure In2O3 and Nb2O5 (Supplementary Fig. 8), indicating a significant inhibition of the electron/hole recombination within the In2O3/Nb2O5 S-scheme hybrid nanofibers. Moreover, during the long-term photoelectrochemical test, IN10 consistently exhibits the highest and most stable photocurrent density in contrast to pristine In2O3 and Nb2O5 (Supplementary Fig. 9), underscoring the efficient charge separation within the In2O3/Nb2O5 nanohybrids. Electrochemical impedance spectroscopy (EIS) results demonstrate that IN10 displays a smaller arc radius in the Nyquist plot compared to bare In2O3 and Nb2O5 (Supplementary Fig. 10), signifying a lower charge transfer resistance in the In2O3/Nb2O5 composite. These analyses collectively confirm that the hybridization of In2O3 and Nb2O5 to form S-scheme heterojunctions can boost charge transfer and effectively reduce the electron/hole recombination, thus facilitating high-efficiency photocatalytic CO2 reduction52,53,54,55,56.

The accumulation of photogenerated electrons and holes after S-scheme charge separation was explored through electron paramagnetic resonance (EPR) spectroscopy. The reduction potential of 5,5-dimethyl-1-pyrroline N-oxide (DMPO)-•O2− and the oxidation potential of DMPO-•OH are −0.74 and 2.28 V (vs. SHE), respectively. Compared to pristine In2O3 or Nb2O5, the In2O3/Nb2O5 composite shows intensive EPR signals for both •O2− and •OH radicals (Supplementary Fig. 11). This observation signifies the efficient separation and accumulation of energetic photoelectrons in the Nb2O5 CB and photoholes in the In2O3 VB, providing compelling evidence for the S-scheme charge separation mechanism.

Ultrafast electron transfer at the In2O3/Nb2O5 S-scheme heterojunction interface

Fs-TAS was employed to delve deeper into the dynamics of photoelectron transfer at the In2O3/Nb2O5 S-scheme interface. As depicted in Fig. 3a–f, both pristine In2O3 and the In2O3/Nb2O5 heterojunctions (IN5 and IN10) exhibit noticeable negative peaks at ~480 nm when excited at 340 nm, which correspond to the ground state bleaching (GSB) signals of In2O3 and provide insights into the population of photoelectrons in its CB57. This assignment was further supported by an experiment using AgNO3 as an electron scavenger, wherein the signal virtually disappears (Supplementary Fig. 12), indicating that the photogenerated electrons are trapped by the scavenger, leaving no electrons to recombine with the holes. The normalized recovery kinetics of pristine In2O3 at 480 nm, monitored within 50 ps, were fitted with a two-exponential function (Fig. 3g and Supplementary Table 4), assigning to the interband diffusion (Process I) and the trap** by shallow trap states (Process II) of the photogenerated electrons in the In2O3 CB. Upon integrating In2O3 with Nb2O5, an additional ultrafast pathway (<10 ps) emerges for the photoelectrons in the In2O3 CB, namely, their transfer to the Nb2O5 VB (Process III)58. Notably, in an Ar atmosphere, both τ1 and τ2 lifetimes demonstrate a gradual decrease with increasing Nb2O5 content (IN5 and IN10, Fig. 3h, i and Supplementary Table 4). This implies the rapid migration of more photoelectrons from the In2O3 CB to Nb2O5 upon hybridization, resulting in fewer electrons available for diffusion and trap** processes (Fig. 3k, l). Under a CO2 atmosphere (Fig. 3j, m), photoelectrons in the Nb2O5 CB react with CO2 molecules, accelerating the transfer of more photoelectrons from the In2O3 CB to the Nb2O5 VB to recombine with its photoholes, thereby shortening the lifetimes. The fs-TAS analysis of a physically-mixed composite of In2O3 and Nb2O5 reveals longer lifetimes (τ1 and τ2) compared to In2O3/Nb2O5 heterostructures (Supplementary Fig. 13), indicating inefficient electron transfer from In2O3 to Nb2O5 and underscoring the advantages of interfacial phase contact within the S-scheme heterojunction for efficient charge transfer and separation.

Fig. 3: Insights into charge transfer dynamics in pure In2O3 and In2O3/Nb2O5 S-scheme heterojunctions.
figure 3

The pseudocolor plots and transient absorption spectra recorded at indicated delay times measured with 340 nm excitation: (a, d) pure In2O3 in Ar, (b, e) IN10 in Ar, and (c, f) IN10 in CO2. Corresponding kinetic decay curves at 480 nm within 50 ps: g pure In2O3 in Ar, (h) IN5 in Ar, (i) IN10 in Ar, and (j) IN10 in CO2. The decay pathways of photogenerated electrons in (k) pure In2O3, l In2O3/Nb2O5 heterojunctions in Ar, and (m) In2O3/Nb2O5 heterojunctions in CO2.

Full size image

On the other hand, the broad GSB signal of Nb2O5 appears around 500 nm (Fig. 4a, b), corresponding to the population of photoholes in its VB as affirmed by the disappearance of the signal after introducing a hole-trap** agent (lactic acid) (Supplementary Fig. 14). To minimize the interference of In2O3, kinetic decay curves of bare Nb2O5 and the In2O3/Nb2O5 nanohybrids (IN20 and IN10) are derived at 530 nm (Fig. 4c–e), which involve two processes related to photoexcited holes in the Nb2O5 VB, i.e., recombination with self-generated photoelectrons (process 1) and recombination with photoelectrons transferred from In2O3 (process 2) (Fig. 4g). Under an Ar atmosphere, the half-life (τ1/2) of both IN20 and IN10 is shorter than that of pristine Nb2O5, signifying the migration of photoelectrons from the In2O3 CB to Nb2O5 and thereby reducing the number of VB photoholes (Fig. 4h). In addition, the In2O3/Nb2O5 heterojunctions exhibit a composition-dependent half-life, with a shorter value as Nb2O5 content decreases (IN10 < IN20). According to the S-scheme charge separation mechanism, photoelectrons in the In2O3 CB transfer to the Nb2O5 VB and recombine with its photoholes in equal proportions. At an ideal In2O3/Nb2O5 ratio (i.e., IN10), the population of photogenerated charge carriers is approximatively identical in both materials, leaving few excess photoholes in the Nb2O5 VB and consequently shortening the lifetime after S-scheme charge separation. When the Nb2O5 content deviates from its optimal value (i.e., IN20), some excessive photoholes remain in the Nb2O5 VB, leading to a longer lifetime than IN10. Under a CO2 atmosphere, photoelectrons in the Nb2O5 CB actively react with CO2, diminishing their recombination with holes while facilitating the photoelectron transfer from the In2O3 CB to the Nb2O5 VB, thus resulting in no notable alteration in the lifetime (Fig. 4i). The analyses emphasize the ultrafast electron transfer at the In2O3/Nb2O5 S-scheme heterojunction interface for suppressing self-carrier recombination and spatially separating photoelectrons in the Nb2O5 CB and photoholes in the In2O3 VB.

Fig. 4: Insights into charge transfer dynamics in pure Nb2O5 and In2O3/Nb2O5 S-scheme heterojunctions.
figure 4

a The pseudocolor plot, and (b) transient absorption spectra of pure Nb2O5 recorded at indicated delay times measured with 340 nm excitation. Corresponding kinetic decay curves at 530 nm within 100 ps of (c) pure Nb2O5 in Ar, (d) IN20 in Ar, (e) IN10 in Ar, and (f) IN10 in CO2. The decay pathways of photogenerated holes in (g) pure Nb2O5, (h) In2O3/Nb2O5 heterojunctions in Ar, and (i) In2O3/Nb2O5 heterojunctions in CO2.

Full size image

Time-resolved fluorescence spectroscopy (TRPL) was conducted to investigate the long lifetime of the photocatalysts. As presented in Supplementary Fig. 15, under an Ar atmosphere, the IN10 hybrid exhibits a longer average lifetime (τa) with respect to pristine In2O3 and Nb2O5 at an emission wavelength of 470 nm, where the fluorescence signals originate from both In2O3 and Nb2O5. Following the proposed S-scheme mechanism for IN10, photoelectrons in the In2O3 CB migrate to the Nb2O5 VB and recombine with its photoholes, resulting in the accumulation of powerful electrons in the Nb2O5 CB and holes in the In2O3 VB, thereby prolonging the charge carrier lifetimes. The physically-mixed composite of In2O3 and Nb2O5 displays a shorter τa than the In2O3/Nb2O5 heterojunction, highlighting the significance of ultrafast interfacial electron transfer in extending carrier lifetimes. Furthermore, in situ TRPL was employed to explore the relationship between the ultrafast charge transfer-induced carrier lifetimes and the photocatalytic performance. The τa of IN10 recorded under a CO2 atmosphere is shorter than that under an Ar atmosphere (Supplementary Fig. 16a), suggesting that a substantial portion of photogenerated electrons in the Nb2O5 CB is involved in CO2 photoreduction, thereby leaving fewer charge carriers available for recombination. Bare In2O3, Nb2O5, and their physically-mixed composite (Supplementary Fig. 16b–d) reveal almost identical decay curves under both CO2 and Ar atmospheres, indicating their poor photoreaction performance. Overall, the ultrafast interfacial charge transfer within the In2O3/Nb2O5 heterojunctions plays triple roles: preventing the recombination of self-carriers, separating powerful photoelectrons and photoholes, and extending their long lifetimes.

Chemisorption, activation and photoreduction of CO2 over In2O3/Nb2O5 hybrid nanofibers

The adsorption and activation of CO2 molecules on the catalyst, pivotal steps for CO2 photoreaction, were investigated using DFT simulations and CO2-temperature programmed desorption (TPD) analysis. Upon CO2 adsorption on Nb2O5, distinct chemisorption processes occur, as evident from several observations (Fig. 5a–c and Supplementary Fig. 17): (i) a pronounced bending of the O=C=O bond at an angle of 127.4°; (ii) elongation of the bond length compared to free CO2 molecule (1.16 Å); (iii) formation of new bonds between CO2 and Nb2O5; and (iv) transfer of electrons from Nb2O5 to CO2 (Supplementary Table 5). The integrated crystal orbital Hamiltonian population (ICOHP) of the C-O pairs is –18.37 and –13.82 eV in free and adsorbed CO2, respectively, signifying CO2 activation over Nb2O5. Moreover, the formation of new C-O (lattice) bonds with an ICOHP of –12.25 eV provides additional compelling evidence of the robust CO2 chemisorption on Nb2O5 (Fig. 5d, e). The CO2-TPD profiles for Nb2O5 and IN10 (Fig. 5f) demonstrate the desorption of physiosorbed CO2 at 70–150 °C. At a high temperature ranging from 370 to 450 °C, both samples exhibit prominent desorption signals indicative of CO2 chemisorption on the catalysts. The CO2 adsorption energy (Eads) on the Nb atom is more negative than that on the O atom (Supplementary Fig. 18 and Table 6), suggesting that Nb atoms serve as active sites for the chemisorption and activation of CO2 during photoreduction.

Fig. 5: Chemisorption and activation of CO2 over In2O3/Nb2O5 hybrid nanofibers.
figure 5

a, b Optimized structures, and (c) the corresponding charge density difference image of CO2 adsorbed on the Nb2O5; cyan and yellow regions represent electron depletion and accumulation, respectively. The isosurface level is set to 0.002 e Å–3. Projected COHP profiles of (d) free CO2 and (e) adsorbed CO2 molecules. f CO2-TPD spectra of pure Nb2O5 and IN10 composite.

Full size image

The photocatalytic activities for CO2 reduction of In2O3, INx, and Nb2O5 were assessed in an online closed gas-circulation system (OLPCRS-2, Shanghai Boyi Scientific Instrument Co., Ltd., Supplementary Fig. 19) equipped with a glass reaction cell. Blank control experiments affirm that the concurrent presence of photocatalysts, CO2, H2O, and light irradiation is essential for initiating the photoreaction (Supplementary Fig. 20). In the absence of any molecule cocatalyst or scavenger, all the samples yielded CO as the reduction product with nearly 100% selectivity (Fig. 6a). Pure In2O3 and Nb2O5 exhibit poor photocatalytic performance due to the rapid recombination of photogenerated carriers inherent in single photocatalysts. However, the integration of In2O3 with Nb2O5 enhances CO2 photoreduction activities, leading to a maximum CO production yield of 0.21 mmol gactive sites–1 h–1 over the IN10 composite. A comparison of CO2 photoreduction performance was conducted among the In2O3/Nb2O5 hybrid nanofibers, the In2O3/Nb2O5 nanohybrid synthesized via the traditional dip-calcination method, and a physically-mixed composite of In2O3 and Nb2O5. This comparison emphasizes the critical importance of intimate interface contact between the two phases for facilitating ultrafast interfacial electron transfer within the S-scheme heterojunction (Supplementary Fig. 21). Upon introducing tris(2,2′-bipyridyl)ruthenium(II) chloride hexahydrate ([RuII(bpy)3]Cl2·6H2O) and 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) as the molecular catalyst and hole scavenger, respectively, a substantial amount of CO and a minor quantity of H2 were detected, with the highest production yields (109.6 mmol gactive sites–1 h–1 for CO and 3.5 mmol gactive sites–1 h–1 for H2) observed over the In2O3/Nb2O5 nanohybrid (Fig. 6b). To elucidate the origin of the photoreduction product, isotope-labeled carbon dioxide (13CO2) was employed as the substitute source gas for photocatalytic CO2 reduction over IN10. Distinct peaks observed at 1.61 and 2.30 min in the total ion chromatography are assigned to O2/Ar and N2, respectively (Supplementary Fig. 22). Another prominent peak emerges at ~6.55 min, corresponding to CO, which generates the predominant mass spectrometry signal at m/z = 29 (13CO), accompanied by two additional fragments at m/z = 13 and 16 (13C and O) (Fig. 6c). Additionally, thermogravimetric (TGA) analyses of pure In2O3, pristine Nb2O5, and the In2O3/Nb2O5 nanohybrid (IN10) reveal no perceptible weight changes up to 800 °C (Supplementary Fig. 23), suggesting that no carbon residual remains in the samples after a 2-h calcination at 600 °C. These findings confirm that the reduction product originates solely from the input CO2, ruling out other potential carbon sources59. The recyclability and stability of IN10 for CO2 photoreduction are confirmed, demonstrating a negligible decline in production yields over four cycles (Supplementary Fig. S24). The XRD pattern (Supplementary Fig. 25) and the In 3d and Nb 3d XPS spectra (Supplementary Fig. 26) of IN10 after the photoreaction exhibit inconspicuous changes compared to the fresh one, suggesting the photostability of the In2O3/Nb2O5 heterojunctions.

Fig. 6: Performance and mechanism insights into photocatalytic CO2 reduction.
figure 6

The production yields and CO selectivity over In2O3, INx, and Nb2O5 during six-hour experiments conducted under UV-visible light irradiation: a without any molecule cocatalyst or scavenger, (b) with [RuII(bpy)3]Cl2·6H2O and BIH. c The total ion chromatography and the corresponding mass spectra of the products in the photocatalytic reduction of 13CO2 over IN10. d In situ DRIFT spectra for the photocatalytic CO2 reduction over IN10. e Gibbs free energy diagrams of CO2 photoreduction and H2 production over Nb2O5 (100) slab. The influence of H2O volumes on product selectivity over IN10: f without any molecule cocatalyst or scavenger, (g) with [RuII(bpy)3]Cl2·6H2O and BIH. The error bars (mean ± standard deviation) were obtained based on three independent photocatalytic experiments.

Full size image

The reaction mechanism of CO2 photoreduction was explored using both in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS, Supplementary Fig. 27) and DFT calculations. As depicted in Fig. 6d, the presence of bidentate carbonate (b-CO32–) and monodentate carbonate (m-CO32–) after the introduction of CO2 into the system in the dark manifests the chemisorption of CO2 on IN10. Under light irradiation, new adsorption bands of *COOH (1258 and 1507 cm–1), *COO (carboxyl, 900 and 1017 cm–1), *C=O (carbonyl, 1839 cm–1), and *CO (absorbed CO, 1956 cm–1) are detected, which are key intermediates in the conversion of CO2 to CO. In light of these observations, the pathway for CO2 photoreduction over In2O3/Nb2O5 hybrid nanofibers is proposed as follows, where * denotes the reaction active site23,60,61:

$$ \ast+{{{{{{\rm{CO}}}}}}}_{2}\to*{{{{{{\rm{CO}}}}}}}_{2}$$
(1)
$$*{{{{{{\rm{CO}}}}}}}_{2}+{{{{{{\rm{H}}}}}}}^{+}+{{{{{{\rm{e}}}}}}}^{{{{{{-}}}}}}\to*{{{{{\rm{COOH}}}}}}$$
(2)
$$*{{{{{\rm{COOH}}}}}}+{{{{{{\rm{H}}}}}}}^{+}+{{{{{{\rm{e}}}}}}}^{{{{{{-}}}}}}\to*{{{{{\rm{CO}}}}}}+{{{\rm{H}}}_{2}{O}}$$
(3)
$$*{{{{{\rm{CO}}}}}}\to \ast+{{{{{\rm{CO}}}}}}$$
(4)

DFT calculations reveal that the rate-limiting step for CO2 reduction over Nb2O5 is the formation of *COOH intermediate. The adsorption of CO2 molecules displays spontaneity with a decrease in Gibbs free energy, reaffirming the inevitable chemisorption and activation of CO2 on Nb2O5. Based on the aforementioned analyses, the enhanced photocatalytic performance in CO2 reduction achieved by the In2O3/Nb2O5 S-scheme heterojunctions can be attributed to several pivotal factors: (i) the close interconnection between In2O3 and Nb2O5; (ii) the S-scheme-induced ultrafast photoelectron transfer at the heterojunction interfaces; (iii) the effective separation of powerful photoelectrons in the Nb2O5 CB and photoholes in the In2O3 VB; (iv) the prolonged lifetimes of charge carriers within the nanohybrids; and (v) the CO2 chemisorption and activation on Nb2O5.

In the liquid/solid photoreaction system, H2 production from H2O reduction competes with CO generation from CO2 reduction. To explore the influence of H2O content on product selectivity, different volumes of H2O were introduced into the reaction system. The results reveal an initial increase and subsequent decrease in CO production yield with the gradual addition of H2O (Fig. 6f). At a low H2O volume, fewer protons (H+) are available for CO2 reduction, resulting in poor performance. Conversely, excessive H2O in the reaction solvent diminishes the solubility of CO2 and impedes its activation on the catalyst. Notably, significant H2 production is absent in the absence of molecular catalysts and hole scavengers, regardless of H2O volume. Indeed, the high H2-evolution barrier over Nb2O5 poses a challenge for H2 production (Fig. 6e)62. Even if a trace amount of H2 is generated, it can be consumed by the residual O2 within the system, originating from input high-purity CO2 (99.999%), following an exothermic reaction (H2 + 1/2O2 → H2O, ΔG < 0). On the other hand, the CO selectivity is affected by the H2O content in the reaction system involving [RuII(bpy)3]Cl2·6H2O and BIH, exhibiting a noticeable decrease when its volume exceeds 200 μL (Fig. 6g). This suggests that the input CO2/H2O can precisely tune the composition of the photoreduction products.

In a CO2 photoreduction system without hole sacrificial agents, two simultaneous processes govern the overall amount of O2 within the system: O2 generation from H2O photooxidation (H2O + 2 h+ → 1/2O2 + 2H+, OER) and its consumption through photoreduction (O2 + 4e– + 4H+ → 2H2O, ORR). Over time, all the samples manifest a decline in O2 levels (Supplementary Fig. 28), indicating a higher rate of O2 consumption compared to its production. Free energy diagrams revel that the OER is not a spontaneous reaction for both In2O3 and Nb2O5, while the ORR is a spontaneous reaction (Supplementary Figs. 29 and 30), suggesting a preference for O2 consumption over its generation and, consequently, a substantial reduction in O2 quantity in the reaction system.

In summary, S-scheme In2O3/Nb2O5 hybrid nanofibers were synthesized using a facile one-step electrospinning method, establishing intimate phase contact for seamless charge carrier transport. Fs-TAS analyses confirmed the ultrafast electron transfer at the In2O3/Nb2O5 S-scheme heterojunction interface, involving the recombination of feeble photoelectrons in the In2O3 CB and holes in the Nb2O5 VB, while efficiently separating and preserving powerful photoelectrons in the Nb2O5 CB and holes in the In2O3 VB for participation in photoreactions. Both DFT calculations and CO2-TPD results demonstrated the efficient chemisorption and activation of CO2 molecules on the In2O3/Nb2O5 heterojunctions. Benefiting from the prolonged lifetimes driven by the rapid interfacial charge transfer and efficient CO2 activation on the catalyst, the optimized In2O3/Nb2O5 nanofibers exhibited enhanced CO2-reduction performance, yielding CO with an impressive output of up to 0.21 mmol gactive sites–1 h–1 in the absence of any molecule cocatalyst or scavenger. This work highlights the potential of advanced fs-TAS techniques in exploring ultrafast charge transfer at S-scheme heterojunction interfaces.

Methods

Chemicals

All the chemicals are of analytical grade (AR) and were used without further purification. Indium nitrate hydrate (In(NO3)3·xH2O, 99.9%) and ammonium niobium oxalate (V) hydrate (C4H4NNbO9·nH2O, 99.9%) were purchased from Macklin Biochemical Technology Co., Ltd. (Shanghai, China). Polyvinylpyrrolidone (PVP, MW = 1300,000) and tris(2,2′-bipyridyl)ruthenium(II) chloride hexahydrate ([RuII(bpy)3]Cl2·6H2O, 98%) were purchased from Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). N,N-dimethylformamide (DMF, AR) and acetonitrile (AR) were obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) was synthesized according to our previous work34.

Synthesis of In2O3/Nb2O5 (INx) hybrid nanofibers

The In2O3/Nb2O5 hybrid nanofibers were synthesized via a one-step electrospinning method by mixing the precursors of both phases in the same electrospinning solution. Typically, 0.32 g of In(NO3)3·xH2O and 1.50 g of PVP were dissolved in 10 mL of DMF and stirred at room temperature until a clear solution was obtained. Simultaneously, 0.032 g of C4H4NNbO9·nH2O, corresponding to 10 wt.% of Nb2O5 relative to In2O3 (IN10), was dissolved in 1 mL of H2O. This solution was then added to the previously prepared one and stirred for 2 h. Subsequently, the viscous solution was loaded into a syringe equipped with a stainless-steel nozzle, positioned about 10 cm away from the collector. An electric potential of 20 kV was applied, and the solution was fed with a rate of 0.4 mL h–1. The collected nanofibrous mat was calcined at 600 °C for 2 h with a heating rate of 2 °C min–1 to completely remove the PVP, resulting in the formation of the In2O3/Nb2O5 nanofibers with a yield of over 90%. For comparison, In2O3/Nb2O5 heterojunctions with different Nb2O5 contents were synthesized by changing the amount of C4H4NNbO9·nH2O to 0.016 and 0.065 g, resulting in nominal weight percentages of 5 wt.% and 20 wt.% Nb2O5 relative to In2O3 (IN5 and IN20), respectively.

Synthesis of pure In2O3 nanofibers

Pure In2O3 nanofibers were synthesized by dissolving 0.16 g of In(NO3)3·xH2O in 5 mL of DMF. Once the indium nitrate was completely dissolved, 0.75 g of PVP was added, and the mixture was stirred until the solution became clear. This solution was then loaded into a syringe equipped with a stainless-steel nozzle, positioned approximately 10 cm from the collector. A potential of 20 kV was applied, and the solution was fed at a rate of 0.4 mL h–1. The samples were collected and calcined at 600 °C for 2 h with a heating rate of 2 °C min–1 to obtain pure In2O3 nanofibers with a yield of approximately 90%.

Synthesis of pure Nb2O5 nanorods

Typically, 0.08 g of C4H4NNbO9·nH2O was first dissolved in 1 mL of H2O. Once fully dissolved, 5 mL of DMF and 0.75 g of PVP were added while stirring until the solution was clear. This solution was then loaded into a syringe equipped with a stainless-steel nozzle, positioned about 10 cm from the collector. An electric potential of 20 kV was applied, and the solution was fed at a rate of 0.6 mL h–1. After calcining at 600 °C for 2 h with a heating rate of 2 °C min–1, pure Nb2O5 nanorods were obtained with a yield of over 90%.

Photocatalytic CO2 reduction

The CO2 photoreduction was carried out in an online gas-closed system equipped with a gas-circulated pump (OLPCRS-2, Shanghai Boyi Scientific Instrument Co., Ltd.). Typically, 10 mg of photocatalysts, 30 mL of acetonitrile, and 100 μL of H2O were added into a glass reactor connected to the online system. The airtight system underwent complete evacuation using a vacuum pump. Then, ~60 kPa of high-purity CO2 (99.999%) gas was injected. After adsorption equilibrium, a 300 W Xe arc lamp (Microsolar 300 Xenon lamp source, Bei**g Perfectlight, China) was used as the light source without any filter. The reaction system was maintained at 8 °C, controlled by cooling water. The gas chromatograph (GC-2030, Shimadzu Corp., Japan) equipped with barrier discharge ionization detector (BID) and a capillary column (Carboxen 1010 PLOT Capillary, 60 m × 0.53 mm) was employed to analyze the photocatalytic CO2 reduction products. For comparison, 2 mM of [RuII(bpy)3]Cl2·6H2O and 10 mM of BIH were introduced into the photoreaction system, with other parameters unchanged. Regarding active sites, pure In2O3 employs itself as its active sites, while the INx and pure Nb2O5 utilize the Nb2O5 as active sites. The value of mactive sites in all In2O3/Nb2O5 nanohybrids was determined based on the actual weight ratios of Nb2O5 within the composites (Supplementary Table 1). Given that the CB minimum and the VB maximum of the In2O3/Nb2O5 heterojunction are predominantly contributed by Nb 4d and OIn2O3 2p orbitals (Supplementary Fig. 31), it follows that CO2 photoreduction and H2O photooxidation occur specifically over the Nb and OIn2O3 atoms, respectively63.

The isotope-labeling experiment was conducted using 13CO2 (isotope purity, 99%, and chemical purity, 99.9%) as the carbon source. The gas products were analyzed by gas chromatography-mass spectrometry (8890 GC System, 5977B GC/MSD, Agilent Technologies, USA) equipped with the column for detecting the reduction products (HP-MOLESIEVE). Helium was used as carrier gas. The temperatures of the injector and EI source were set to be 150 and 200 °C, respectively.

Statistics and reproducibility

No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were not randomized, and we were not blinded to allocation during experiments and outcome assessment.