Formation of Carbon-Induced Nitrogen-Centered Radicals on Titanium Dioxide under Illumination

Titanium dioxide is the most studied photocatalytic material and has been reported to be active for a wide range of reactions, including the oxidation of hydrocarbons and the reduction of nitrogen. However, the molecular-scale interactions between the titania photocatalyst and dinitrogen are still debated, particularly in the presence of hydrocarbons. Here, we used several spectroscopic and computational techniques to identify interactions among nitrogen, methanol, and titania under illumination. Electron paramagnetic resonance spectroscopy (EPR) allowed us to observe the formation of carbon radicals upon exposure to ultraviolet radiation. These carbon radicals are observed to transform into diazo- and nitrogen-centered radicals (e.g., CHxN2• and CHxNHy•) during photoreaction in nitrogen environment. In situ infrared (IR) spectroscopy under the same conditions revealed C–N stretching on titania. Furthermore, density functional theory (DFT) calculations revealed that nitrogen adsorption and the thermodynamic barrier to photocatalytic nitrogen fixation are significantly more favorable in the presence of hydroxymethyl or surface carbon. These results provide compelling evidence that carbon radicals formed from the photooxidation of hydrocarbons interact with dinitrogen and suggest that the role of carbon-based “hole scavengers” and the inertness of nitrogen atmospheres should be reevaluated in the field of photocatalysis.

High QE (HQE) cavity resonator (ER 4122SHQE).g tensors were calibrated for accuracy using known BDPA and Mn 2+ in SrO standards.The acquisition parameters such as the receiver gain, modulation amplitude, and microwave power were all optimized and then kept constant to enable comparison of multiple samples under the same conditions.The modulation amplitude was set to 5G and the microwave power was 0.5 mW unless otherwise stated.10 mg of P25 titania were added into 0.1 mL of DI water.5 vol% of methanol were added as a presence of carbon source.Argon or nitrogen are saturated into the tube for 40 minutes and a 300 W Xenon lamp was used as the illumination source for the light samples.
The samples were sealed under He and the EPR spectra were recorded at 4.2 K.All spectra are simulated using EasySpin software package Release 6.0.0-dev.26(2020-10-12).( StefanStoll, Arthur Schweiger, EasySpin, a comprehensive software package for spectral simulation and analysis in EPR. 1

In-Situ Infared Spectrascopy
In situ ATR measurements were performed using a Thermo Fisher iS20 equipped with an MCT/A detector.The Pike Technologies HATR accessory was used to investigate the reaction on the catalyst surface.The top of the HATR cell was modified in house with the addition of a quartz window to be able to perform photocatalytic experiments.In a typical experiment, a slurry was prepared by adding 2 ml of water to 15 mg of P25 and vigorously mixed until well dispersed.The slurry was then deposited drop-wise on a Ge refractive element (IRE) and the water was evaporated with a heat-gun.The coated IRE was inserted into the modified HATR module.A second IR spectrometer (Thermo Fisher iS5) equipped with a HATR cell using a ZnSe crystal was used to analyze the effluent of the iS20 to account for the contribution of the liquid phase.The solution of 5 vol% of MeOH was saturated with the gas of interest (Ar or N 2 ) for 40 min before being fed into the system.The solution was continuously purged with the gas for the length of the experiment to maintain saturation.
The methanol solution was fed into the system at 0.5 ml min −1 for 30 minutes to stabilize the flow.Afterwards, the Xenon lamp was turned on at 300 W. The acquisition of IR spectra was started on both the iS20 and the iS5 and the evolution of the spetra was followed for 350 minutes collecting a spectrum every 5 minutes by using a resolution of 4 cm −1 and 64 scans.The last spectrum collected in the absence of irradiation was used as background, and subtracted from the spectra obtained in the presence of irradiation to highlight the vibrations of the molecules bound to the catalyst surface.

Density Functional Theory
All stationary geometry optimization and generalized gradient approximation (GGA) functional calculations were performed in the Quantum ESPRESSO software package 2 together with the Atomic Simulation Package (ASE). 3The Bayesian error estimation functional with van der Waals correlation (BEEF-vdW) 4 was used to describe electronic interactions.The plane-wave cutoff energy was set at 400 eV for all calculations; and a Monkhorst-Pack kpoint grid spacing of 4 × 4 × 1 was used for calculations.The NNIN/C pseudopotential 5 was chosen to treat the core-electron interactions.A dipole correction was applied for all slab calculations.All intermediate states were spin-paired except CH*.Adsorbate configurations were determined by taking the most stable binding site from several high-symmetry configurations.When possible, this was achieved by using the most stable site from prior work, while for new adsorbates, configurations were manually enumerated and optimized with the BEEF-vdW functional to determine the most stable site.All geometries were optimized using the BFGS line search algorithm.The convergence threshold was set at 10 -6 eV.Gas phase calculations were performed at the Γ point in a unit cell with 6 Å vacuum.The model surface was a 4 layers rutile (110) slab with one oxygen vacancy as the adsorption site. 6Geometries of adsorbed surfaces were determined by trying adsorbates of multiple orientations and taking the lowest energy.HSE06 7 hybrid functional calculations were performed in the Simulation Package for Ab-initio Real-space Calculations (SPARC) software package. 8The Monkhorst-Pack k-point grid spacing of 4 × 4 × 1 and a mesh spacing of 0.1 Å were used for calculations.For slab calculations, periodic and Dirichlet boundary conditions were prescribed in the plane and perpendicular to the plane of the slab, respectively.For gas phase molecules calculations, Dirichlet boundary conditions are employed in all three coordinate directions.All adsorbed intermediate states were calculated with spin polarization.The convergence tolerance on the normalized residual of electron density of the SCF iteration was set at 10 -6 .The convergence tolerance on the Fock energy was set at 10 -4 Hartree.The maximum number of Fock iterations was 20.The hybrid range screening parameter was set at 0.106. 7 utilized geometries reported by Xie et al 9 as the initial guesses of adsorbate geometries in the OH-assisted pathway on a periodic rutile TiO 2 (110) slab with oxygen vacancy (Perslab) in Fig. S11 and S13.These states were relaxed with BEEF-vdW prior to a single-point calculation with HSE06.The relaxed geometries are shown in Table S5.The term Per-slab is used to distinguish from the oxygen vacancy site on a rutile TiO 2 (110) slab reported by Comer et al. 10 The electronic energies for the reaction energies to form NH 3 , N 2 H 2 , N 2 H 4 and H 2 O from N 2 , H 2 , and O 2 are compared for the BEEF-vdW and HSE functionals (Fig. S2).The "Mixed Ratio (compound)" methods use a linear combination of BEEF-vdW and HSE energies to yield the experimental reaction energy of the compound.
Surprisingly, the error for HSE is the largest, with an error of 0.707 eV for the NH 3 .The BEEF-vdW functional has a lower magnitude error of 0.238 eV, which is also in the opposite direction, suggesting that BEEF-vdW tends to systematically under-estimate N-H bond strength, while HSE tends to systematically over-estimate it.Given the importance of N-H bonding on all intermediates in the mechanism, we ensure that the NH 3 formation energy is correct by using a linear combination of BEEF-vdW and HSE energies for all states in the system.Conceptually this is equivalent to adjusting the ratio of Fock exchange, which is a common practice in hybrid functional design.A ratio of 0.748 BEEF-vdW and 0.252 HSE yields the correct NH 3 reaction energy, and reduces the mean absolute error of other relevant reaction energies to 0.303 eV from 0.433 eV (BEEF-vdW) or 0.393 eV (HSE).Thus, we expect that this combined energy provides the best estimate of intermediate states, and have used these energies in the main text.However, the qualitative conclusions reached do not depend on which functional is used, and the raw energies and free energy diagrams for both BEEF-vdW and HSE are provided below.
To include zero point energy (ZPE) and thermal contributions, vibrational frequency calculations and statistical mechanics corrections were performed using the ASE implementation.Then ground state electronic energies (E ele ) calculated by DFT were converted to free energies (G o i ) using the following equation: where E ZP E is the zero point energy, ∆H and T ∆S are thermal contributions.Gas phase molecules were treated as ideal gas; and adsorbates were treated with the harmonic approximation with a low-frequency cutoff of 30 cm −1 .Relative free energies were computed with respect to reference states using the formula: where G i is the free energy of species i, G o i is the total energy computed from DFT and free energy corrections, n i is the number of atoms j in species i, and µ j is the reference chemical potential.The reference for nitrogen was N 2 (µ N = 1 2 G o N 2 ), the reference for hydrogen was            8.5, 12.9, 38.1   123.5, 132.1, 142.9   167.5, 181.0, 183.6   185.4,365.7, 372.2   380.4,473.4   N 2 gas 0.0, 0.0, 300.5 NH  4.9, 9.6, 22.5   36.4,40.6, 63.4   70.3, 73.1, 143.0   146.2, 268.4,381.4   CH at Oxygen Vacancy  16.3, 52.9, 60.4   69.0, 71.2, 380.8  Table S5: Calculated equilibrium geometries for all species.

Figure S1 :
Figure S1: Schematic diagram of the in situ IR experiment.

Figure S2 :
Figure S2: The scatter plot showing the mean absolute errors of reaction energy calculated from each method.The MAE values of each method are represented by stars.The x-axis represents the methods used, and the y-axis represents the MAE in electron volts (eV).The "Mixed Ratio (compound)" methods use a linear combination of BEEF-vdW and HSE energies to yield the experimental reaction energy of the compound.
FigureS3: EPR spectra of titania with 5 vol% methanol solution at 4.2 K during illumination under 15 N 2 nitrogen environment.Isotope 15 N has nuclear spin of 1/2 and therefore in 15 N 2 environment the signal for nitrogen centered radical has 2 lines with the same intensity and A 15N = 17G and he one for diazo centered radical is composed of 3 lines with line intensity 1:2:1 and hyperfine of A 15N = 17G.

Figure S4 :Figure S5 :
Figure S4: IR spectra of titania in water without methanol in (a) argon and (b) nitrogen environments during illumination.

Figure S6 :
Figure S6: Free energy diagram of nitrogen fixation starting from methanol at the oxygenvacancy site calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase CH 3 OH, H 2 , N 2 and H 2 O in the initial state.

Figure S7 :
Figure S7: Free energy diagram of nitrogen fixation starting from CH at the oxygen-vacancy site calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase H 2 , N 2 and H 2 O in the initial state.

Figure S8 :
Figure S8: Free energy diagram of nitrogen fixation starting from CH 2 at the oxygen-vacancy site calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase H 2 , N 2 and H 2 O in the initial state.

Figure S9 :
Figure S9: Free energy diagram of nitrogen fixation starting from CH 3 at the oxygen-vacancy site calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase CH 3 OH, H 2 , N 2 and H 2 O in the initial state.

Figure S10 :
Figure S10: Free energy diagram of nitrogen fixation starting from CH 3 and methanol at the oxygen-vacancy site at an overpotential (η) of 0.15V.calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase CH 3 OH, H 2 , N 2 and H 2 O in the initial state.

Figure S11 :
Figure S11: Free energy diagram of nitrogen fixation pathways without carbon assistance starting from Per-slab (black) and the oxygen-vacancy site (grey), calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase H 2 , N 2 and H 2 O in the initial state.

Figure S12 :
Figure S12: Free energy diagram of nitrogen fixation pathways without carbon assistance starting at the oxygen-vacancy site, calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase H 2 , N 2 and H 2 O in the initial state.

Figure S13 :
Figure S13: Free energy diagram of nitrogen fixation pathways without carbon assistance starting at the Per-slab, calculated by HSE06 functional (a) and BEEF-vdW functional (b).Both diagrams include gas phase H 2 , N 2 and H 2 O in the initial state.

Table S1 :
Energies of all species used relative to the slab with Oxygen vacancy at 0K(ZPE  included)

Table S3 :
Energies of all species used relative to the Per-slab with Oxygen vacancy at 0K(ZPE included)