Characterization of CuFeS2 NCs
Oleylamine-capped CuFeS2 NCs displayed an average size of 8–10 nm, as indicated by transmission electron microscopy (TEM) (Fig. 1a–c and Supplementary Figs 2 and 3). Energy dispersive X-ray analysis (EDS) (Fig. 1a, inset and Supplementary Fig. 3c) and elemental mapping with high-angle annular dark field–scanning TEM (Fig. 1d–h) confirmed the homogeneous distribution of Cu, Fe and S elements throughout the crystal. The selected area electron diffraction (Fig. 1b, inset) and the X-ray diffraction (XRD) pattern (Fig. 1j) showed the characteristic diffraction rings and reflections, respectively, of the (112), (204) and (312) lattice planes of the tetragonal CuFeS2 phase28, confirming the purity of the product. Ultraviolet–visible light (UV–vis) absorption spectra of the NCs (Fig. 1i) demonstrated broad absorption at 520 nm, attributed to the plasmon resonance of the CuFeS2 NCs28.
The oleylamine-capping agents of the CuFeS2 NCs were exchanged with S2− ions to render them more dispersible in polar solvents and improve the interactions with the reactants. UV-vis absorption spectra before and after the ligand exchange (Fig. 1i) indicated that the plasmonic band was only slightly broadened and red-shifted. XRD (Fig. 1j) and Raman spectra (Supplementary Fig. 4a) also confirmed the preservation of the crystal structure. The successful ligand exchange was confirmed with Fourier transform infrared spectroscopy (FTIR) (Fig. 1k), showing the elimination of the oleylamine spectral features at 2,987 and 2,900 cm−1. Similarly, X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. 4b) showed a dramatic reduction—or complete elimination—of the nitrogen peak (circled in red) in the CuFeS2-S2− after removal of oleylamine. More details on the XPS characterization are available in Supplementary Fig. 5.
Photocatalytic performance of CuFeS2 NCs
The photocatalytic activity of the CuFeS2 NCs for the hydrogenation of nitroarenes (Fig. 2a) was evaluated using hydrazine hydrate as a hydrogen and electron donor. Hydrazine is an attractive choice because of the high hydrogen content (8.0 mass%), simply separable by-products (only hydrogen and nitrogen) and scalable synthesis from ammonia. The reaction was optimized under 400–500 nm of light, at a very low flux of 22 mW cm−2 and maximum intensity at 450 nm. Reaction optimization using 10 mg of the CuFeS2 catalyst showed that at 2 h with 0.8 mmol of hydrazine afforded the product (aniline) at 100% yield and selectivity, using 0.1 mmol of the nitrobenzene substrate (Fig. 2b, left part). By increasing the amount of the substrate tenfold (1 mmol) and the amount of hydrazine to 16 mmol (in 1 ml of H2O) similar results were obtained at 4 h of reaction (Fig. 2b, middle part), corresponding to a molar TOF of 4.6 h−1, this being already among the highest reported (Supplementary Table 1; TOF is calculated with respect to the total moles of all components of the catalyst, as explained in the notes of the same Table 1). It was very gratifying to observe that by further challenging the catalyst via increasing the substrate to 5 mmol under the exact same conditions, aniline was again obtained at 100% conversion and selectivity, affording the highest TOF value of 22.8 h−1 (Fig. 2b, right part). Reactions without catalyst or without hydrazine did not yield any aniline, while a control reaction in the dark at 25 °C delivered a yield of 19% (Fig. 2b), suggesting intrinsic catalytic activity of the system. CuFeS2 NCs coated with the oleylamine molecules (Supplementary Fig. 7) showed lower yield than the S2− passivated NCs. The reaction yield and rate depended on the amount of the catalyst (Fig. 2c) reaching a maximum yield of 99.4% and a molar average TOF of 22.8 h−1 with an optimum catalyst to substrate ratio of 10 mg per 5 mmol of nitrobenzene. This TOF is substantially higher than any recently disclosed state of the art thermal catalyst or photocatalyst for nitroarene reduction, as later discussed and described in Supplementary Table 1. Even in a large-scale reaction with 20 mmol (2.5 g) of nitrobenzene, the TOF was retained at 22.2 h−1 (Supplementary Figs. 8–12).
To gain further insights, control reactions were performed in the dark in an oil bath at 25 or 40 °C, affording aniline with a yield of 19.7 and 44.1%, respectively (Fig. 2d), verifying that CuFeS2 NCs are intrinsically active, which is an important feature of an ideal photocatalyst32. Control experiments were performed using Cu, Fe or S elements, as well as Fe2O3, FeCl3 CuI and mixtures thereof, which showed very low activity (Supplementary Fig. 7). CuFeS2 is also a well-known photothermal agent29, thus light irradiation during the catalytic reaction caused a spontaneous temperature increase reaching 58 °C (Fig. 2e, ‘light without fan’). When the same reaction was performed using the cooling fan of the photoreactor, the temperature stabilized at 33 °C, giving a slightly lower yield of 89.7% with 100% nitrobenzene conversion (Fig. 2e, ‘light with fan’). Control reactions in the dark at 40 or 33 °C delivered lower yields (44.1 and 32%, respectively, Fig. 2d,e ‘dark’) than the reaction at 33 °C but under light, indicating that the NCs did not act only through photothermal activation, but also through intermediate photoexcited species. The catalyst was finally challenged using a 1-sun solar-light simulator, delivering a TOF of 20 h−1 (around 84.2% yield) within 4 h (Fig. 2e, 1 sun). The slightly lower selectivities in the presence of light with fan-cooling and with the sun-simulator (Fig. 2e) are probably attributed to the lower temperature and broader irradiation spectrum, respectively.
The importance of these results can be better recognized if evaluated within the state of the art. For instance, a Zn-based metal organic framework15 showed very efficient nitrobenzene photo-reduction (TOF = 13.3 h−1, Supplementary Table 1, entry υ), using very high intensity of light and costly organic ligands (detailed description of costs is available in the Supplementary Information). A Pd3Au0.5/SiC photocatalyst showed excellent nitroaromatic hydrogenation22 (TOF = 7.9 h−1, Supplementary Table 1, entry φ), but with the need of high-cost noble metals, H2 flow and high fluence of light (300 mW cm−2 as opposed to the 22 mW cm−2 in the present case). Semiconductor photo-catalysts, such as CuxS-ZnCdS (TOF = 3.9 h−1) and Zn1 − xCdxS (TOF = 1.1 h−1, Supplementary Table 1, entries τ and π, respectively), also showed good activity, but the toxic heavy metals9,17 rise environmental concerns. Even in some exemplary cases of highly sustainable catalysts of iron and cobalt oxides embedded on nitrogen-doped graphitic layers for the efficient chemo-selective hydrogenation of nitroarenes5,6, harsh conditions were required, such as pressurized H2 (50 bar) at 110–120 °C (Supplementary Table 1, entries a,b). Moreover, single Co atoms in N-doped carbon13 or Co nanoparticles encapsulated in carbon nanotubes12 achieved high activity under relatively benign reaction conditions, but still requiring 2–4 bar of H2 pressure12,13 and temperature of 110 °C (ref. 12). However, the present catalyst delivered higher reaction rates while using low-cost and sustainable metals without any other energy input than the irradiation from solar light.
Recyclability and substrate scope
The recyclability of the CuFeS2 nano-catalyst was investigated for five consecutive reactions with 1 mmol of nitrobenzene and 2 mg of catalyst (that is, at its maximum performance, Fig. 3a and Supplementary Fig. 13). The results indicated that there was marginal loss in the catalytic activity even after the fifth cycle (100% conversion and 86% yield or better at conditions lower than its maximum performance, Supplementary Fig. 14). Moreover, there was no need for increasing the reaction time or the pressure and temperature, as often required6,13. XPS analysis before and after the reaction (Supplementary Fig. 5) confirmed the preservation of its structural features. Besides the high activity of the catalyst, its ability to reduce effectively a wide variety of substrates with high selectivity (Fig. 3b), irrespectively of the presence of other functionalities, is of additional importance. Challenging substrates, with competing reducible groups (that is, 4-nitrobenzonitrile, 4-iodo-nitrobenzoate and 4-ethynylnitrobenzene5,6) were obtained with yields of 99, 93.8 and 86.1%, respectively. Indicatively, previously achieved yields of 4-nitrobenzonitrile and 4-ethynylnitrobenzene were 75 (ref. 5) and 83% (ref. 6), respectively, at high temperature and 50 bar H2 atmosphere.
Benchmarking of the catalyst
To interpret our results within the context of the current state of the art and with respect to the related costs, we collected data on the TOF values as well as on TOF with respect to the cost of the catalyst (Supplementary Table 1 and Fig. 4). For an unambiguous comparison, we included the whole catalyst system for calculating the TOF; regarding the price, we took into account the initial key reagents used in the synthesis of the catalysts, considering 100% yield (details are given in the Supplementary Table 1, in the Experimental section in the Supplementary Information). According to this analysis, the present CuFeS2-S2− plasmonic photocatalyst revealed its high production rate and a transformative performance based on TOF with respect to the catalyst costs (Fig. 4).
Insights into the mechanism of action of the CuFeS2 plasmonic photocatalyst
To better understand the high activity of the catalyst, ultrafast laser time-resolved transient absorption spectroscopy (TAS) and continuous-wave light-induced electron paramagnetic resonance experiments were performed (Fig. 5). In TAS studies, the difference in optical density (ΔOD) at various time delays and wavelengths (Fig. 5a,b) revealed the presence of two main processes: (1) a photo-induced absorption (PIA) and (2) a photobleaching feature in the vicinity of 590 and 750 nm, respectively. The PIA profile is attributed to transitions from temporary occupied states in the intermediate bands to the conduction band28,29, while the simultaneously observed photobleaching feature is attributed to transitions from the depleted valence band to states within the intermediate band28,29.
The decay dynamics of these two relaxation processes unveiled that both PIA and photobleaching exhibited an identical two-step decay profile, with a fast component of few ps, followed by a slower component of several tens of ps (Fig. 5b). The fast time component is related to the nonradiative intraband electron–electron and electron–phonon scattering relaxation processes taking place in the intermediate band and in the conduction band, which results in carrier cooling on transferring the excess energy of the excited electrons to the crystal lattice, ultimately leading to NC heating20,29,30,33. The slower time component is attributed to the heat transfer to the surrounding environment of the nanoparticles20,29,30,33. The very similar fast decay profiles of PIA and photobleaching features indicate that hot electrons and heat are generated in both the conduction band and the intermediate band. Although these timescales are beyond the fs processes of Landau damping (when hot electron–hole pairs are generated33) and therefore cannot be observed, nonradiative plasmonic nanostructures (such as CuFeS2) favour hot electron generation and heating20,30,33. According to the theoretically calculated band structure of CuFeS2 (ref. 34), the intermediate band–conduction band gap is about 2 eV (Fig. 5c), corroborating the PIA feature at the spectral window around 590 nm (2.1 eV). The valence band–intermediate band gap is 0.7–1 eV, matching the photobleaching feature with maximum around 910 nm, which was also verified by the Tauc plots, at around 0.85 eV (Supplementary Fig. 15). The full agreement between the experimental and theoretical data clearly supports the formation of holes in the valence band of CuFeS2 (with maximum energy of around −5.2 eV35) and hot electrons in the intermediate band and conduction band (at around −4 eV and above −2 eV, respectively, Fig. 5c). At the same time, the HOMO of hydrazine is positioned at −5.1 eV (ref. 36), extremely close to the upper valence band energy levels of CuFeS2, where the holes are created. This energy matching promotes a favourable interaction of hydrazine’s antibonding and bonding HOMO electrons with the holes from CuFeS2 valence band (generated during photoexcitation), which leads to weakening the N–H bond, proton and electron abstraction from hydrazine via formation of the intermediate complex, as depicted in the possible structure of Fig. 5d. Through continuous-wave light-induced electron paramagnetic resonance experiments, we observed such an interaction and electron transfer from hydrazine to the catalyst in water before the addition of nitrobenzene, revealing a new photoexcited spin state (Supplementary Figs. 20 and 21) with hyperfine parameters suggesting the structure of Fig. 5d (and Supplementary Fig. 21e). On the addition of nitrobenzene, a new radical species produced a strong signal as time evolved (Fig. 5e and Supplementary Figs. 22 and 23). This type of signal corresponds to N-phenylhydroxylamine radical species (–N•–OH), as verified by the simulated spectrum with the corresponding spin-Hamiltonian parameters (Supplementary Fig. 23d) and by the spin-trap experiments (Supplementary Fig. 24). This radical can be associated with the three-electron reduced intermediate form of nitrobenzene (highlighted in the reaction mechanism in Supplementary Fig. 27), identifying a possible and previously elusive three-electron intermediate in the overall reaction pathway A. Further support for the predominance of pathway A in the presence of light is provided by gas chromatography results, showing hydroxylamine or azoxybenzine as the only stable intermediates in the presence of light or in the dark, respectively (Supplementary Fig. 28). The energy matching of the catalyst’s photogenerated holes with the HOMO of hydrazine could be considered responsible for the excellent performance of the catalyst.
CuFeS2 NCs also use the synergic contribution of the two metal centres, Fe and Cu. The Fe site is responsible for binding and activating hydrazine, forming the transient spin‑active species, [H(FeS2)NH-NH2]•, S = 1/2 system, which delivers the protons and electrons to the neighbouring Cu(I)S2 site. The Cu(I)S2 sites interact with the nitro-substrate, producing the N-phenylhydroxylamine radical, as experimentally trapped in situ (Supplementary Fig. 23). The results take forward the concept that by a judicious combination of metal centres bound to rigid ligand-field environments, a highly effective catalytic system can be conveyed, harnessing the power of cooperative enzymatic catalytic centres37, for example, to effectively transfer H+ and e− to the substrate38. The use of the identified energy flow pair (CuFeS2-H2NNH2) extends beyond this reaction, affecting a broad family of hydrogen transfer and reduction catalytic reactions in valuable processes for biomass valorization39.