Fast and selective reduction of nitroarenes under visible light with an earth-abundant plasmonic photocatalyst

Reduction of nitroaromatics to the corresponding amines is a key process in the fine and bulk chemicals industry to produce polymers, pharmaceuticals, agrochemicals and dyes. However, their effective and selective reduction requires high temperatures and pressurized hydrogen and involves noble metal-based catalysts. Here we report on an earth-abundant, plasmonic nano-photocatalyst, with an excellent reaction rate towards the selective hydrogenation of nitroaromatics. With solar light as the only energy input, the chalcopyrite catalyst operates through the combined action of hot holes and photothermal effects. Ultrafast laser transient absorption and light-induced electron paramagnetic resonance spectroscopies have unveiled the energy matching of the hot holes in the valence band of the catalyst with the frontier orbitals of the hydrogen and electron donor, via a transient coordination intermediate. Consequently, the reusable and sustainable copper-iron-sulfide (CuFeS2) catalyst delivers previously unattainable turnover frequencies, even in large-scale reactions, while the cost-normalized production rate stands an order of magnitude above the state of the art.


Materials and methods
Synthesis of CuFeS2 NCs. In a typical synthesis, 1 mmol (0.192 g) of CuI and 1 mmol (0.162 g) of FeCl3 were dissolved in 27 mL of oleylamine and stirred in a three-necked round bottom flask.
The mixture was heated to 100 °C under vacuum and maintained for 2 h to remove oxygen and moisture. The reaction atmosphere was then switched to nitrogen and then raised to 180 °C and stirred until the solution obtained a clear orange color. 2 mmol of hexamethyldisilathiane (or 2 mmol of sulfur powder) dissolved into 3 mL of 1-octadecene (ODE) was swiftly injected into the above mentioned mixture under vigorous stirring, and the temperature was kept for 10 min at 180 °C for the growth of the CuFeS2 NCs. Then, the solution was cooled to room temperature under ambient conditions. Toluene and ethanol were added to the reaction mixture for at 9418 rcf for 5 min. The precipitate was then washed twice with an 1:1 ethanol:toluene mixture for two times. Finally, the oleylamine-capped nanocrystals were collected and dried at 60 °C under reduced pressure.

Ligand exchange of the CuFeS2 NCs
For exchanging the oleylamine capping ligands of the CuFeS2 nanocrystals a procedure was adapted from prior literature with slight modifications 1,2 . 0.1 g of Na2S.9H2O was dissolved in 5 mL of ultrapure water and then 15 mL of dimethylformamide was added, to form the sulfide ligand-exchange solution. 0.2 g of CuFeS2 NCs were dispersed in 10 mL cyclohexane and then mixed with ligand-exchange solution and stirred for 30 min at room temperature. The mixture was centrifuged at 9418 rcf for 5 min and the supernatant was discarded. The precipitate was washed 3 times with ethanol and the ligand-exchanged nanocrystals were dried under vacuum at 60 °C.
The final solid catalyst was grinded well in a mortar with a pestle before use.

Characterization
Transmission electron microscopy (TEM) images were recorded on JEOL JEM-2100 TEM equipped with a LaB6 type emission gun operating at 200 kV. High-resolution transmission electron microscopy (HR-TEM) and scanning transmission electron microscopy (STEM) in highangle angular dark field (HAADF) mode for elemental mapping were performed with an FEI TITAN G2 60-300 HRTEM microscope with an X-FEG type emission gun, operating at 300 kV, objective-lens image spherical aberration corrector and ChemiSTEM EDS detector. X-ray diffraction (XRD) patterns were recorded with a PANalytical X'Pert PRO MPD (PANalytical, The Netherlands) diffractometer in the Bragg-Brentano geometry, Co-Kα radiation (40 kV, 30 mA, λ = 0.1789 nm) equipped with an X'Celerator detector and programmable divergence and diffracted beam antiscatter slits. The measurement range was 2θ:0 5°-105°, with a step size of 0.033°. The identification of the crystalline phases was performed using the High Score Plus software (PANalytical) that includes the PDF-4+ database. X-ray photoelectron spectroscopy (XPS) was carried out with a PHI VersaProbe II (Physical Electronics) spectrometer using an Al Kα source (15 kV, 50 W). The obtained data were evaluated with the MultiPak (Ulvac -PHI, Inc.) software package. UV-Vis absorption spectra were collected on a Cary 50 UV-Vis spectrophotometer (Varian). Fourier transform infra-red (FT-IR) spectra were recorded on an iS5 FTIR spectrometer (Thermo Nicolet) using the Smart Orbit ZnSe ATR accessory. Briefly, a droplet of chloroform/ethanol dispersion of the relevant material was placed on a ZnSe crystal and left to dry and form a film. Spectra were acquired by summing 64 scans recorded under a nitrogen gas flow through the ATR accessory. ATR and baseline correction were applied to the collected spectra.
Raman spectra were recorded on a DXR Raman microscope using the 613 nm excitation line of a diode laser EPR spectra were recorded on a JEOL JES-X-320 spectrometer operating at the X-band frequency (∼9.0-9.1 GHz) equipped with a variable-temperature controller (He, N2) ES-CT470 apparatus.
The cavity quality factor (Q) was kept above 6000 in all measurements. Highly pure quartz tubes were employed (Suprasil, Wilmad, ≤0.5 OD), and accuracy on g-values was obtained against the Mn II /MgO standard (JEOL standard). The spectra were acquired by careful monitoring that signal saturation from the applied microwave power did not occur during signal's acquisition. In situ light excitations EPR experiments (LEPR) were performed using a HeCd laser source operating @325 nm (max CW power of 200 mW) from Kimmon Koha Co. Ltd. The light was shined directly onto the sample, kept inside the cavity EPR resonator, through its dedicated optical window and the light-off to light-on process was operated by an on-off mechanical light-shutter mechanism.
The band gap of the CuFeS2 NCs was calculated using the Tauc equation and was found to be 0.85 eV for the direct band gap and 1.72 eV for the indirect band gap, corresponding to transitions from the valence band to the intermediate band and from the intermediate band to the conduction band, respectively. A Lambda 1050 UV/Vis/NIR spectrophotometer (PerkinElmer) was used.

Catalytic nitrobenzene reduction
2 and up to 10 mg of CuFeS2 catalyst was added in a glass vial with a teflon-coated cap (5 mL), adding 0.1 and up to 5 mmol of nitrobenzene, 0.05 mL to 1 mL of hydrazine hydrate (50 % solution) and 0 to 3 mL of ethanol. After mixing by stirring (60 s) and sonication (30 s), the closed vial was irradiated in a photoreactor equipped with a light source (EvoluChem PhotoRedOx box, attached with 34 W Kessil 150N LED having total irradiance of 22 mW/cm 2 from 400-500 nm region, peaking at 450 nm) for 0 to 4 h under stirring at room temperature.
For a larger-scale reaction, 40 mg of catalyst and 4 mL of hydrazine hydrate were mixed well and sonicated for 30 s in a closed reaction vessel (60 mL capacity), and 20 mmol (2.46 g, 2 mL) of nitrobenzene was added. The reaction vessel was irradiated in the same photoreactor chamber for 4 hours under stirring. After the reaction, the products were extracted by ethyl acetate and the catalyst separated by centrifugation. The ethyl acetate solution of the product was further diluted and analyzed in GC and NMR.
Reactions were also performed using a solar simulator of 1 Sun (100 mWcm 2 ) intensity (Sciencetech Light Line A4-C250 equipped with an AM 1.5G filter, class CAA). After the reaction, the catalyst was separated by centrifugation and the product was obtained in ethanol or The moles of the catalyst were calculated by considering the dry weight and molecular formula (CuFeS2) of the catalyst. For example, 10 mg of the CuFeS2 catalyst was equal to 54.49 µmol, considering 183.52 as molecular weight.
As an example, the recalculation method of the TOF value for entry σ (Pd3Au0.5/SiC; ref 21) is also described. In the article a TOF value of 900 h -1 is reported based on the metal content only, at 300 mW/cm 2 of irradiation intensity. However, here, we consider the whole catalyst system for calculating the TOF, since the support is an integral part of the catalyst defining its catalytic activity (e.g. with different supports the catalyst displayed different activity; actually, certain conditions SiC is catalytically active, Fig. S16 ref 21). The total catalyst used for converting 6 mmol of nitroarene was 25 mg (609.15 µmol), from which 24.125 mg (601.47 µmol) corresponded to SiC, 0.125 mg (0.63 µmol) to Au and 0.75 mg (7.04 µmol) Pd. Therefore, the molar-based TOF value was equal to 7.9 h -1 (for 6 mmol reactant, 609.15 µmol catalyst, 0.5 h reaction time, and 80 % product yield).
The calculations for the cost-normalized TOF values (TOF $ -1 mmol) were performed considering the commercially available precursors for the synthesis of the catalyst and that the catalyst was obtained at 100 % yield. Metal precursors and other reagents necessary for the synthesis of the catalyst were considered, except common solvents. In all cases, the cost of the catalyst corresponding to the mass used for the conversion of 1 mmol of substrate was considered: Cost-normalized TOF = TOF (ℎ −1 ) cost of catalyst per 1 mmol of product ($ × mmol −1 ) , ℎ −1 $ −1 A detailed excel file, where all the cost calculations are described in detail, is available as a Supplementary data file.

Recyclability and substrate study
The recyclability of the catalyst was tested in the reaction of 1 mmol of nitrobenzene, in 1 mL of hydrazine hydrate, using 2 mg of the catalyst. The above regents were mixed well and sonicated for 30 s in a closed reaction vessel and then irradiated with for 4 hours under stirring at room temperature. After the reaction, the products were extracted by ethyl acetate and the catalyst separated by centrifugation. The ethyl acetate solution of the product was further diluted and analyzed in GC. The recovered catalyst was washed several times with ethanol and dichloromethane to remove any adsorbed molecules and finally dried under vacuum oven overnight at 60 °C before using it for the next cycle.

Transient absorption spectroscopy (TAS) study
Transient absorption spectroscopy (TAS) measurements were performed on a Newport (TAS-1)

Supplementary Tables and Figures
Supplementary Fig. 1: Schematic representation of transient absorption spectroscopy (TAS) experimental setup (Newport, TAS-1). -Entries a,b refer to nanoparticles embedded or wrapped by N-doped graphitic layers.
-Entry c refers to single metal atoms embedded in N-doped carbons -Entry e refers to nanoparticles encapsulated in carbon nanotubes -Entry η refers to the Zn atoms embedded in metal organic framework -Entry σ refers to Pd/Au nanoparticles whose catalytic activity strongly depends on the type of the support.
-Entry h refers to 0.25 % Pt/a-ΜοC single Pt atom catalyst, where the MoC matrix played an active and independant catalytic role for the dissociation of the nitrogen-oxygen bonds in nitrobenzene.
-Entry φ refers to Pd3Au0.5/SiC, where under certain conditions SiC is catalytically active (Fig. S16 in the respective article). The catalytic activity was highly dependent on the light intensity; at 300 mW/cm 2 it was half than that at 800 W/cm 2 (Fig. S8 in the respective article).
In these examples, the metal centers are structurally attainable, stable, or active only within the frame of their support. For this reason, in order to obtain an unambiguous calculation of the TOF, the mass/moles of the whole system was considered. In the case of nanoparticulate catalysts, only the surface atoms are catalytically active and all the inner materials act as support. The calculation of TOF based only on surface atoms (i.e. a 1 Å periphery) would result in a sharp increase of the TOF. For example, in a 9-nm nanoparticle (as the present CuFeS2), a few-atom-thick exposed surface of thickness ~1 Å occupies 30 times less volume and mass than the whole particle. Thus, the TOF in the present system would be close to a value of 700 h -1 . Supplementary Fig. 2: Particle size distribution of CuFeS2 NCs based on the TEM. Supplementary Fig. 3: (a,b) HR-TEM images of the CuFeS2 NCs. Comments: The surface chemical states of the CuFeS2 NCs before ( Supplementary Fig. 5a-d) and after ( Supplementary Fig. 5e-h) the reaction were also probed with X-ray photoelectron spectroscopy. The Cu 2p high-resolution XPS spectra before (Supplementary Fig. 5b) and after ( Supplementary Fig. 5f) the reaction were identical, reflecting the typical spin-orbit splitting of the Cu atoms, resulting in Cu 2p3/2 and Cu 2p1/2 peaks with a separation of 19.8 eV, and no satellite peak, confirming the Cu (I) oxidation state of Cu and absence of Cu (II) species. The Fe 2p spectrum (Supplementary Fig. 5c&g) also showed the characteristic doublet due to spin-orbit splitting. The Fe 2p3/2 envelope centered at 710.9 eV, corresponding to Fe (III) oxidation state, in accordance with previous results 35 . The S 2p core level spectrum ( Supplementary Fig. 5d&h) showed a spin orbit splitting (two main doublets), S 2p3/2 and S 2p1/2, corresponding to the metal-sulfide (sulfide and disulfide) bonding states of sulfur.  Fig. 10: GC graph of nitrobenzene (5 mmol) reduction using catalyst (10 mg) without hydrazine. A major peak of nitrobenzene (11.61 min) and a minor product peak (6.05 min) were visible.
Supplementary Fig. 11: GC graph of nitrobenzene (5 mmol) reduction using hydrazine (16 mmol) without catalyst. A major peak of nitrobenzene (11.19 min) and a minor side products were visible. EPR Section.

2.1.Reactivity of hydrazine under UV light.
The chemical stability of the hydrazine/water solution under UV light (@325 nm) has been followed by monitoring in situ, by light induced EPR technique (LEPR), the electronic changes associated to its spin state, which corresponds in the initial form to the closed-shell S=0 specie.
Results from the measurements are shown in Supplementary Fig. 15a-c. As soon as UV light is applied to the sample kept frozen in the cavity resonator (see Supplementary Fig. 16b, in particular), a strong symmetric signal develops around g = 2.000 ( Supplementary Fig. 16c), which is associated with the formation of a radical specie. The observed signal lacks any resolved 14

2.2.Reactivity of nitrobenzene and nitrobenzene/hydrazine mixture under UV light.
While neat nitrobenzene does not show any detectable EPR signal, both in dark and under application of UV light (@325 nm), (Supplementary Fig. 17a, green spectrum), combination of 1 to 1 mixture (vol/vol) of reducing agent (hydrazine) and reactant (nitrobenzene) shows, on the contrary, that even under ambient light a radical specie is detected by EPR in the frozen solution ( Supplementary Fig. 17b, blue spectrum). This specie is different from the radical signal detected from the neat hydrazine solution shown in Supplementary Fig. 16 (only seen under UV light). In particular, as highlighted in Supplementary Fig. 16b, the resonance signal evidenced a clear ganisotropy, which can be linked to the 14 N nuclear hyperfine component acting on the electron spin moment. Therefore, an electron transfer occurs between hydrazine and nitrobenzene, leading to formation of a radical intermediate, most probably centered on the nitrobenzene. After in situ application of UV light (@325 nm) to the frozen sample kept inside the cavity resonator, a more complex EPR signal appears ( Supplementary Fig. 19a, red-spectrum), which can be interpreted as combination of doublet (S=1/2) and triplet species (S=1, D 17 mT). Overall, it is found that hydrazine molecules do react with the nitrobenzene substrate even in absence of catalysts, but such reaction leads to the formation of various spin active intermediates/species, especially when high energy (h) is provided from the light source, and likely results into an admixture of variously reduced aromatic by-products. Further analysis of the interaction between hydrazine and nitrobenzene without catalyst goes beyond the scope of the current investigation. Importantly these species, as shown from the gas chromatography of a control reaction between nitrobenzene and hydrazine results only into a 5 % conversion of a mixture of products ( Supplementary Fig. 11).

2.3.Reactivity of neat nitrobenzene, neat hydrazine and nitrobenzene/hydrazine mixture in presence of the CuFeS2 NCs catalysts.
Further insights on the mechanism of nitroarene reduction into aniline promoted by CuFeS2 NCs  Fig. 16), it does exhibit some differences. Closer inspection of this signal was obtained by recording its EPR signatures by signal accumulation, in a narrow magnetic field-sweep range, using weaker applied microwave power and smaller modulation-width ( Supplementary Fig.   21d). This spin containing specie clearly belongs to a photoexcited spin state generated under UV light formed by hydrazine and CuFeS2 NCs system. However, differing from the case of the detected EPR signal of neat hydrazine under UV, here the EPR signal associated to the photoexcited spin active specie shows several additional resonances, whose contribution may arise from both the effective interaction of the electron spin moment with nuclear hyperfine terms (e.g. from 1 H centres). It is excluded that these resolved resonances come from part of Cu(I) oxidized to Cu(II), because the observed splitting ( 1.0 mT) is too small to arise from the copper nuclear hyperfine term (I= 3/2). We may suggest that the observed resonance is the overlapped combination of S=1/2 signals from trapped HOO  specie, as seen earlier in Supplementary Fig. 16, and an S=1/2 signals generated by various forms of the hydrazine bound/CuFeS2 NCs system, with tentative structures given in the drawings shown in Supplementary Fig. 21e.
Upon dispersion of CuFeS2 NCs in water/nitrobenzene/hydrazine mixture ( 2 mg, 1 mL, ratios 1/1/1, vol/vol/vol), it was noticed that even under normal ambient light, evolution of N2 from the vial containing the catalyst and the water solvent plus reactants (nitrobenzene and hydrazine) occurred very fast; the same effect has been observed earlier to occur in the hydrazine/nitrobenzene mixture (see Supplementary Fig. 17). Therefore, CuFeS2 NCs were pre-suspended in water/nitrobenzene (2 mg/mL), transferred to an EPR tube and an aliquot of hydrazine was then added inside the EPR tube, followed by fast freeze quenching into liquid nitrogen bath. The fastquenched EPR spectrum of CuFeS2 NCs in water/nitrobenzene/hydrazine recorded without in situ application of UV-light is shown in Supplementary Fig. 22a (black line), and magnification of the magnetic field region around g = 2.000 is given in Supplementary Fig. 22b. A very strong signal appears in this region, which increases in intensity as soon as UV light is applied to the frozen sample inside the cavity resonator ( Supplementary Fig. 22a-b, orange line). See also the EPR spectra shown in Supplementary Fig. 23a-c, obtained by fast LEPR acquisition procedure at T= 77K. This type of signal is highly reminiscent of those expressed by nitroxide-based radicals (-N-O  ) and can be associated to the three-electron reduced intermediate form of the nitrobenzene substrate forming N-phenylhydroxylamine radical, as shown in the Supplementary Fig. 23d (inset).
The spin-Hamiltonian simulation (2 nd order perturbation theory) of this signal is given by the bluespectrum in Supplementary Fig. 23d Fig. 24 and Supplementary   Fig. 25). Therefore, within the catalytic cycle ( Supplementary Fig. 26 and Supplementary Fig. 27), which overall involves transfer of 6 electrons (6e -) and 6 protons (6H + ) from the reducing agent (hydrazine), the preferred reaction pathway appears to proceed through pathway A, rather than via pathway B (Supplementary Fig. 27).

2.4.Spin trap experiments as monitored by the CW-EPR technique.
The spin-trap molecule, -4-pyridyl-1-oxide-N-tert-butylnitrone (POBN) was used to validate the presence of radical species arising from the catalytic conversion of nitrobenzene to aniline under UV-light irradiation. Experimentally, solid POBN was added to a water suspension containing the CuFeS2 NCs catalyst, hydrazine, and nitrobenzene (concentration of POBN = 9 mg/ mL) ( Supplementary Fig. 24a), and the solution was stirred at room temperature for 5 min under constant UV light irradiation (at 325 nm). Then, 0.2 mL of this mixture were transferred into an EPR tube and fast frozen at 77 K in liquid nitrogen prior to the EPR measurement. A blank experiment, under the same experimental conditions reported above, using only hydrazine in water, nitrobenzene and POBN without the presence of a catalyst was also performed as control sample; the EPR results are shown in the Supplementary Fig. 24b. The molecular structure depicted in blue color ( Supplementary Fig. 24a) indicates the trapped N-phenylhydroxylamine radical species using the POBN probe. Supplementary Fig. 24: (a) Description of the spin trap experiment. (b) X-band (9.08-9.09 GHz) CW-EPR signals of the trapped spin specie (blue EPR trace) observed at 77 K (0.2 mW applied power) as detected from the admixture POBN/NH2-NH2/nitrobenzene in presence of CuFeS2 catalyst exposed for 5 min under UV-light. The structure of the POBN spin trapped radical adduct is shown in blue color by the chemical drawing. The EPR plot shows, for comparison, the resonance features (recorded at 77 K) of the radical species (N-phenylhydroxylamine radical) detected during in situ UV-light irradiation of the NH2-NH2/nitrobenzene solution in the presence of CuFeS2 (molecule shown in red, also in red color is shown its EPR signal). The EPR signal in green color shows the result obtained from the blank experiment, performed under identical conditions, using the admixture of POBN/hydrazine/nitrobenzene without the catalyst. The g-tensor parameters of the trapped POBN radical adduct are similar to those derived from the in situ detected N-phenylhydroxylamine radical but expresses a substantial alteration in the anisotropic N hyperfine splitting terms (hfc, Azz of 3.4 mT, Axx and Ayy of 0.46 mT). This experimental result is also validated by theoretical calculations (DFT/UB3LYP/6-31G*) via the analysis of the spin density distribution (theoretical models drawn on the right). b Supplementary Fig: 27: Details of the reaction mechanism for nitrobenzenereduction using the CuFeS2 catalyst in presence of hydrazine hydrate under light with key intermediates experimentally observed enclosed by blue boxes.

Identification of intermediates by GC before complete conversion of nitrobenzene
Gas chromatographs (GC) during the progress of the photocatalytic reaction, revealing the presence of phenylhydroxylamine as the only intermediate ( Supplementary Fig. 28).
Phenylhydroxylamine's identification, being a major intermediate