Photocatalytic Abstraction of Hydrogen Atoms from Water Using Hydroxylated Graphitic Carbon Nitride for Hydrogenative Coupling Reactions

Abstract Employing pure water, the ultimate green source of hydrogen donor to initiate chemical reactions that involve a hydrogen atom transfer (HAT) step is fascinating but challenging due to its large H−O bond dissociation energy (BDEH‐O=5.1 eV). Many approaches have been explored to stimulate water for hydrogenative reactions, but the efficiency and productivity still require significant enhancement. Here, we show that the surface hydroxylated graphitic carbon nitride (gCN−OH) only requires 2.25 eV to activate H−O bonds in water, enabling abstraction of hydrogen atoms via dehydrogenation of pure water into hydrogen peroxide under visible light irradiation. The gCN−OH presents a stable catalytic performance for hydrogenative N−N coupling, pinacol‐type coupling and dehalogenative C−C coupling, all with high yield and efficiency, even under solar radiation, featuring extensive impacts in using renewable energy for a cleaner process in dye, electronic, and pharmaceutical industries.


Supplementary
. Calculated vibrational modes for 4-melem model with and without OH.
We have prepared the gCN-OH with a series of KOH concentrations and evaluated their photocatalytic performance (Table S3). The catalytic activity of the gCN-OH gradually improved upon increasing the KOH concentration and reached an optimum at 4 M, suggesting that a certain concentration of KOH is required to effectively hydroxylate the surface of gCN.

Supplementary Figures
The binding energy of an OHion onto crystalline gCN was estimated by subtracting the ΔH(gCN-OH model + OHion far away) from the ΔH(gCN-OH hydroxyl ion attached), again using the same model system for the gCN crystalline structure stated in Fig. S1a and 1b. In addition, the binding energy of an overall neutrally bound hydroxyl radical on gCN was estimated in a similar way, using a neutral hydroxyl radical in place of OH -, and was found to be slightly lower at 2.8 eV.
The structures for NMR calculations were adapted from the optimization calculations by MOPAC and the NMR chemical shift calculations were carried out with the gauge-including atomic orbitals (GIAO) 5 methodology implemented in the Gaussian 09 package 6 with the B3LYP exchange-correlation functional and 6-311+G(2d,p) basis set for all atoms. The references for calculating chemical shift from shielding constant were tetramethylsilane (TMS) for H, and NH 3 for N. All 1 H and 15 N NMR calculations are shown in Fig. S1c and 1d. The experimental values are listed for comparison. Chemical composition and oxidation state of elements on the surface region were characterized by an X-ray photoelectron spectrometer equipped with an Al Kα X-ray source (XPS, K-Alpha, Thermo Fisher Scientific, USA). Survey scans were measured from 1100 to -10 eV using a pass energy of 160 eV with a step size of 1 eV and a dwell time of 0.1 s, whereas the region-of-interest spectra of C1s, N1s, O1s and K2p were collected in the desired energy regions using a pass energy of 40 eV with a step size of 0.1 eV and a dwell time of 0.5 s. The C1s binding energy of adventitious carbon (284.6 eV) was used for binding energy calibration. Survey scans reveal that only C and N are observed on pristine gCN, whereas additional O and K exist in the gCN-OH (Fig. S2). We have performed post-mortem analysis of the washed gCN-OH by XPS, as shown in Fig. S3. The oxygen peaks remain unchanged after 3 times of washing with DI water, confirming that the surface OH functional groups are attached to the catalyst with chemical bond rather than physical adsorption. X-ray absorption spectroscopy (XAS) data were collected at the National Synchrotron Radiation Laboratory (NSRL, Beamlines MCD-A and MCD-B (Soochow Beamline for Energy Materials)). For pristine gCN, the distinct peak at ~288 eV is observed in the Kedge of C (Fig. S4a), which match well to the characteristic triazine structure. 7 An obvious shoulder is observed at ~287.2 eV for the gCN-OH that can be assigned to C-OH. In comparison, no obvious change of the K-edge of N is observed for gCN-OH (Fig. S4b), confirming that the surface hydroxyls are indeed bound to C.  The morphology of gCN and gCN-OH was studied using a transmission electron microscope (TEM, Titan Themis Cubed G2 300). The powder samples were dispersed into ethanol and dropped on the Cu grid for analysis. As representatively shown in Figs. S5a and 5b, gCN and gCN-OH present a similar layered morphology. X-ray diffraction (XRD) patterns were recorded on a Bruker D8 Advance diffractometer with a step size of 0.02° and a scan range of 5° to 60° using a Cu-Kα radiation source (40 kV, 40 mA). The XRD patterns of the gCN and gCN-OH show two characteristic peaks at 12.6° and 27.1°, corresponding to (100) and (002) facets of gCN, respectively (Fig. S5c).
Diffuse reflectance spectra (DRS) of the photocatalysts were measured using a Hitachi photospectrometer (UH4150). The DRS data were recorded in the range of 300-800 nm with spectroscopy grade BaSO 4 as the reference. The bandgap of the photocatalysts was derived using the Kubelka-Munk theory. 8  The specific surface areas of gCN and gCN-OH were determined using a Micromeritics ASAP 2460 system, based on nitrogen adsorption/desorption isotherms measured at 77 K, as shown in Figure S5f. The specific surface area of gCN and gCN-OH are 46 and 34 m 2 ꞏg -1 , respectively. Both materials are non-porous.
Solid-state Nuclear Magnetic Resonance (ssNMR). All 1 H, 13 C and 15 N ssNMR spectra were recorded using an AVANCE III HD-600 MHz. The 1 H spectra were taken at a spinning speed of νR = 12.0 kHz, a 2.5 μs excitation pulse, a 10-s relaxation delay, and 32 scans. The CP/MAS 13 C NMR spectra were carried out using 1.4 ms contact time, 5 s relaxation delay, and 512 scans, respectively. The CP/MAS 13 C and 15 N NMR spectra were carried out using 3 ms contact time, 10 s relaxation delay, and 1024 scans, respectively. The 15 N-labeled gCN and gCN-OH sample were uniformly packed in a MAS rotor for ssNMR. The 1 H chemical shifts were referenced to adamantane at 1.91 ppm. The 13 C chemical shifts were referenced to the adamantane CH 2 peak at 38.48 ppm, and the 15 N chemical shifts were referenced to glycine at 32.0 ppm. Photoluminescence spectra (PL) was recorded at RT using a fluorescence spectrometer (FLS1000) with a laser (λ = 320 nm). Attenuated total reflection infrared spectroscopy (ATR-IR) was performed using a Nicolet 6700 spectrometer.
The photochemical reactions were performed in a multichannel reactor (SUNCAT INSTRUMENTS, China) with blue light emitting diodes (LED, λ max = 410 ± 10 nm). The setup consists of three parts: the LED source, a power supply, and a cooling system (Fig. S6). The liquid phase products were analyzed by gas chromatography (GC, Agilent 8860) and gas chromatography-mass spectrometry (GC-MS, Agilent 8860 network GC system coupled with 5977B Network Mass selective Detector). The GC system was equipped with a HP-5 column and an FID detector. The GC and GC-MS analyses were performed to determine the conversion and selectivity of the photocatalytic reactions. All reactions were carried out with solvents and reagents without further purification. For a typical reaction, 10 mg photocatalyst, 16 µmol reactant, 80 µmol KOH, and 20 µl DI water were added into 2 mL solvent at RT. The reactor was deaerated by N 2 purging prior to the photocatalytic reaction.  Figure S7 shows the kinetic analysis of photocatalytic conversion of nitrobenzene, benzaldehyde, and benzyl bromide. All three reactions display pseudo-first-order kinetics with rate constants of 2.15 h -1 , 2.56 h -1 , and 2.81 h -1 , respectively.  in-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFT) were measured in-situ using a Bruker VERTEX 70 FTIR spectrometer with a MCT detector.
The photogenerated H 2 O 2 was quantified by a colorimetric titration method employing Cu (II) based chromogenic agent. 9 The chromogenic agent is prepared by mixing 2,9-dimethyl-1,10-phenanthroline (DMP) ethanol solution (1 gꞏL -1 ) and aqueous CuSO 4 solution (0.01 M) in 1:1 volume ratio. For analysis, 50 μl of aliquots was added into 3 mL of the chromogenic agent under stirring for 10 min. The absorption spectra of the solution were then measured by a UV-vis spectrophotometer (UV 2600, Shimadzu), and the concentration of H 2 O 2 was calculated using the absorbance at 454 nm by the Beer-Lambert law.
The absorption coefficient (α) of the reduced product, Cu (I)-DMP complex, is 1.79*10 7 M -1 ꞏcm -1 according to the calibration curve (Fig. S9a). Note that one part of H2O2 can reduce two part of Cu(II)-DMP complex into Cu (I)-DMP complex. The UV-vis spectra for colorimetric titration of photogenerated H 2 O 2 is shown in Fig. S9b, while the conversion of nitrobenzene determined by GC is shown in Fig. S9c. The evolution of H 2 O 2 and nitrobenzene is shown in Fig. S9d. The Image of the aliquots taken from the photocatalytic reaction suspension mixed with Cu(II) based chromogenic agent is shown in Fig. S9e. Electron spin resonance (ESR). The room temperature X-band CW ESR measurements were performed by an Jeol JES-X320 spectrometer using 100 kHz field modulation frequency and 0.05 mT modulation width. ESR simulations were performed using the Easyspin toolbox. 20 mM of 5,5-dimethyl-1-pyrroline N-oxide (DMPO, spin trap) was added to the reaction suspension prior to irradiation, ensuring immediate capturing of all generated radicals for the formation of stable adducts. It should be noted that the estimated t 1/2 of the DMPO  OOH adduct is ~ 45 s, 10 therefore irradiation times were limited to 30 s. Aliquots were removed from the reaction solution at the indicated times and filtered to remove all traces of photocatalyst prior to transfer to the ESR tube. Notably, this methodology does not enable accurate kinetic information to be extracted from the data (as the ESR signals are an accumulation of all radicals generated throughout the irradiation period), however this does increase the chance of trapping the full variety of radicals generated over the photocatalyst surfaces.
A series of control measurements were performed in the absence of photocatalyst to identify the presence of any radicals in the inert solvent system. The ESR spectra of the 1 vol% water-dioxane solution, both with and without KOH as a base, and under dark and irradiated conditions are shown in Fig. S10, employing DMPO as the spin trap. A negligible quantity of radical species was observed without photocatalyst in all cases. Semi-empirical molecular orbital geometry optimizations and transition state (TS) calculations were done using MOPAC2016, using the PM7 Hamiltonian. 11 The TS calculations for desorption of H 2 O 2 were obtained using the SADDLE procedure outlined in the MOPAC2016 Manual. The bond dissociation energies estimated in this study were obtained by calculating the energy of undissociated water molecules and their resulting OH and H radicals after dissociating on a model gCN crystalline structure consisting of 2 strips of H-bonded 2-melem strips (Fig. S11), using the PM7 Hamiltonian in MOPAC2016. For the dissociation of a water molecule on the pristine gCN model, the BDE was estimated by subtracting the calculated ΔH of the clean model and a water molecule far away from the model surface, from the ΔH of an adsorbed OH on the model (effectively a model of gCN-OH) with a H radical far away from the surface. For the estimation of the BDE arising from a second dissociative attachment of OH onto gCN-OH, the BDE was estimated by correspondingly subtracting the ΔH(gCN-OH model + H 2 O molecule far away) from the ΔH(gCN-OH with 2 nd OH adsorbed + H radical far away). The resulting gCN model with 2 adsorbed adjacent OH radicals formed the start state for the TS calculation of the desorption of H 2 O 2 from this model surface of gCN. All states used to estimate BDE are shown in Figure S12. There is little difference whether the OH is attached to the two different carbon sites. Figure S13 compares a series of typical photocatalysts for the N-N coupling, pinacol-type coupling, and dehalogenative C-C coupling. BiOBr, TiO 2 , AgGaO 2 , and NiFe-LDH show negligible activity for all reactions, whereas the BiVO4 shows poor performance for the synthesis of azoxybenzene and hydrobenzoin. In general, only gCN-OH shows satisfactory performance for all three reactions. The QE of the three photocatalytic reaction was determined using a leak tight reactor that is connected to a stainless-steel vacuum/gas line, as reported previously (Fig. S14). 4 The reactor (A) is linked to the vacuum line (C) via a gas valve (D). The LED lamp (E) is placed on top of the quartz window (B).
An amount of 50 mg of photocatalyst and 8 mM of reactant were added into the 1 vol% H 2 O-solvent solution. The suspension was dispersed by sonication. Before irradiation, the reactor was evacuated and filled with nitrogen three times. Then the suspension was irradiated by a 410 nm LED (12 mWꞏcm -2 ). The solar-driven photocatalytic coupling reaction was performed using a 25 mL round-bottom flask with leak-tight seals (Fig. S15). 75 mg of gCN-OH and 8 mM of reactant were dispersed in 15 mL of 1 vol% H 2 O-solvent solution, which was purged by N 2 for 5 min prior to irradiation. The light intensity is recorded every hour for reference. Detailed product analyses are shown in Supplementary Figures 16-64.