Atomically Dispersed Copper Sites in a Metal–Organic Framework for Reduction of Nitrogen Dioxide

Metal–organic framework (MOF) materials provide an excellent platform to fabricate single-atom catalysts due to their structural diversity, intrinsic porosity, and designable functionality. However, the unambiguous identification of atomically dispersed metal sites and the elucidation of their role in catalysis are challenging due to limited methods of characterization and lack of direct structural information. Here, we report a comprehensive investigation of the structure and the role of atomically dispersed copper sites in UiO-66 for the catalytic reduction of NO2 at ambient temperature. The atomic dispersion of copper sites on UiO-66 is confirmed by high-angle annular dark-field scanning transmission electron microscopy, electron paramagnetic resonance spectroscopy, and inelastic neutron scattering, and their location is identified by neutron powder diffraction and solid-state nuclear magnetic resonance spectroscopy. The Cu/UiO-66 catalyst exhibits superior catalytic performance for the reduction of NO2 at 25 °C without the use of reductants. A selectivity of 88% for the formation of N2 at a 97% conversion of NO2 with a lifetime of >50 h and an unprecedented turnover frequency of 6.1 h–1 is achieved under nonthermal plasma activation. In situ and operando infrared, solid-state NMR, and EPR spectroscopy reveal the critical role of copper sites in the adsorption and activation of NO2 molecules, with the formation of {Cu(I)···NO} and {Cu···NO2} adducts promoting the conversion of NO2 to N2. This study will inspire the further design and study of new efficient single-atom catalysts for NO2 abatement via detailed unravelling of their role in catalysis.

transform (FT) magnitude spectra were calculated. Simulation of the EPR spectra was performed with the EasySpin/MATLAB toolbox, which employs the exact diagonalization of the spin Hamiltonian matrix 5 . The difference between the two Cu isotopes ( 63 Cu and 65 Cu) is included in the simulation program but the effect is not resolved at the linewidths observed.
MAS NMR spectra were recorded using two regimes. Moderate-field experiments employed a Bruker 9.4 T (400 MHz 1 H Larmor frequency) AVANCE III spectrometer equipped with a 4 mm HFX MAS probe.
Experiments were acquired at ambient temperature using a MAS frequency of 10 kHz. 1 H-pulses of 100 kHz were used and 13 C spin-locking at ~25 kHz was applied for 2 ms, with corresponding ramped (70-100 %) 1 H spin-locking at ~50 kHz for CP experiments with 100 kHz of SPINAL-64 6 heteronuclear 1 H decoupling used throughout signal acquisition. Samples were treated and packed into 4 mm o.d. zirconia rotors under inert conditions and sealed with a Kel-F rotor cap fitted with an O-ring. For NO2 adsorption/desorption, the rotor packed with Cu/UiO-66 was opened under inert conditions and the sample remained in this container during the applied treatments, before subsequent resealing with the O-ring Kel-F cap. High-field NMR spectra, recorded at the UK High-Field Solid-State NMR Facility, employed a Bruker 20.0 T (850 MHz 1 H Larmor frequency) AVANCE NEO spectrometer equipped with a 1.3 mm HXY MAS probe, and was used in doubleresonance mode. Experiments were acquired at ambient temperature using a MAS frequency of 60 kHz. 1 Hpulses of 100 kHz were used, except during the S3 recoupling sequence 7,8 where 30 kHz-pulses were required.
One full loop of S3 recoupling was used to reintroduce the homonuclear dipolar interaction between 1 H-spins during both excitation and reconversion periods, giving a total mixing time of 267 s. For the 2D DQ-SQ experiments, 32 transients were acquired for each of 32 complex (STATES-TPPI) rotor-synchronised t1 S6 increments (2 rotor periods). Samples were treated and packed into 1.3 mm o.d. zirconia rotors under inert conditions and sealed with a Vespel rotor cap.
Molecular dynamics (MD) simulations were performed using CP2K (http://www.cp2k.org) 9 , based on the mixed Gaussian and plane-wave scheme 10 and the Quickstep module 11 . The calculation used molecularly optimized Double-Zeta-Valence plus Polarization (DZVP) basis set 12 , Goedecker-Teter-Hutter pseudopotentials 13 , and the Perdew-Burke-Ernzerhof (PBE) exchange correlation functional 14 . The plane-wave energy cutoff was 400 Ry, and a DFT-D3 level correction for dispersion interactions, as implemented by Grimme et al 15 , was applied with a cutoff distance of 15 Å. The calculation was performed on Gamma point only, with no symmetry constraint. Structural optimization was performed using the Broyden-Fletcher-Goldfarb-Shannon (BFGS) optimizer, until the maximum force fell below 0.00045 Ry/Bohr (0.011 eV/Å).
Finite displacement method was used for the phonon calculation, with incremental displacement of 0.01 Bohr (0.0053 Å). INS spectra were simulated using the OClimax software 16 .
In order to measure the NO2 adsorption ability of the catalysts, dynamic breakthrough experiments were performed at 25 °C. NO2-TPD experiments were carried out to determine the adsorbate-adsorbent strength of intereaction. Typically, 50 mg of sample was packed in a fixed-bed reactor which was equipped with a Bruker Matrix MG5 FTIR spectrometer (resolution = 0.5 cm -1 ) as the detector. The sample was then pre-treated under helium flow (100 mL min -1 ) at 250 °C for 2 h. After activation, the NO2 breakthrough experiment was carried out at 25 °C. When the outlet concentration of NO2 was equal to that of inlet, 100 mL min -1 of pure helium was used to flush the sample for about 2 h, removing the physically adsorbed NO2. The NO2-TPD experiments were then carried out at a heating rate of 5 °C min -1 from 25 °C to 250 °C.

S7
Powder X-Ray diffraction (PXRD) patterns of the catalysts

Electron paramagnetic resonance (EPR) spectra of Cu/UiO-66
Figure S7. X-band (9.4 GHz) EPR spectra of Cu/UiO-66 before H2 reduction (black) and Cu/UiO-66 after H2 reduction (red) at room temperature. Inset: expansion of low-field region, illustrating absence of characteristic signals for binuclear and/or aggregated Cu species (S ≥ 1). S14 Figure S8. EPR spectra of fresh Cu/UiO-66 (before and after H2 reduction) and used Cu/UiO-66 (after plasma reaction) at 6.4 K (left); Second integrals of EPR spectra (right). The small peak at g = 2.0 in spectra was identified as emanating from the cavity of the EPR instrument in this particular experiment; the signal was not from the sample.

Deconvolution of 1 H MAS NMR spectra
It is worth noting that there are noticeable correlation peaks between 1 H from the ligand and adsorbed solvent 1 H (at DQ{ 1 H}~11 ppm in Figure 2d in the main text), consistent with the 13  S18 Table S3. Integral intensities from fits to 1 H DEPTH NMR spectra (from Figure S11).  Table S4. The local environment of the Cu(II) site was investigated by hyperfine sublevel correlation (HYSCORE) spectroscopy 4 , revealing hyperfine interactions with surrounding 1 H nuclei ( Figure S12). The spectra were modelled considering contributions from the electron-nuclear ( 1 H) dipolar and isotropic hyperfine interactions:

Experimental setup for catalytic testing
The catalytic performance, including NO2 conversion and N2 selectivity, over the prepared catalysts were tested under non-thermal plasma (NTP) activated conditions. The gas flow rate of NO2 was 100 mL min -1 at 25 °C (500 ppm, diluted in helium) controlled by mass flow controllers. The reaction temperature was controlled using a fan and monitored by an infrared thermometer (IRT670, General Tools & Instruments). The same gas feed was used to test the catalytic performance of all the catalysts in this work. Prior to the NTPassisted catalytic reaction, all catalysts were packed in a fixed-bed reactor 23 and pre-treated at 80 °C for 1 h under a flow of helium (100 mL min -1 ) to remove the residual water in the system. A gas mixture of NO2 and He was then allowed to pass through the fixed-bed reactor to test the catalytic performance of each catalyst.
For the NTP-assisted reaction, the reactor comprised of two coaxial quartz tubes; the outside diameter of the outer tube was 6 mm and the outside diameter of the inner tube was 3 mm giving a discharge gap of 0.5 mm 23 . The outer tube was covered by a metal mesh electrode that was connected to a high-voltage output, and a metal wire electrode (ground electrode) was placed inside the inner tube. The catalysts (MOF powders) were packed in the discharge region to ensure that plasma was generated around the catalysts. An alternating current plasma generator (CPT-2000K, 0-25 kV, 10 kHz) was used to ignite the plasma, and an oscilloscope (Tektronix TDS 2022B) was used to monitor the electrical parameters. The discharge power used for reaction was about 0.6 W and the specific energy input (SEI) was calculated to be 0.4 kJ L -1 , with the AC peak-to-peak voltage (Vpk-pk) at around 13±0.5 kV and a frequency of 10 kHz.
S22     26 . The plasma was generated using an alternating current power source (PVM500 model) and the electrical parameters were monitored using an oscilloscope (Tektronix TBS1062) that was connected to the reactor through a high voltage probe (Tektronix, P6015). The applied voltage was 8±0.5 KV at a frequency of 27 kHz. During the DRIFTS measurements, the IR spectra were recorded with a resolution of 4 cm -1 and with an accumulation of 128 scans for every 60 s. Figure S20. Operando DRIFTS spectrum for UiO-66-NH2 exposed to NO2 (500 ppm NO2 in He) under steady-state NTP conditions. The DRIFTS spectrum was recorded at a resolution of 4 cm -1 . The spectrum of bare UiO-66-NH2 has been subtracted as the background. S27 Figure S21. In situ DRIFTS spectra of adsorbed NO2 on Cu/UiO-66 and UiO-66. The DRIFTS spectra were recorded at a resolution of 4 cm -1 . The background spectra of the bare MOF have been subtracted. Figure S22. EPR spectra of Cu/UiO-66 after H2 reduction (red) and Cu/UiO-66 after H2 reduction, then NO2 loading (blue) at 6.4 K. The spin Hamiltonian used for the simulation of the dominant contribution to the spectrum was: H = gzmBSzBz + gx,ymB(SxBx + SyBy) + AzSzIz + Ax,y(SxIx + SyIy), where g is g-tensor, S is the electron spin, I is the Cu nuclear spin, A is the 63/65 Cu hyperfine interaction (HFI) and B is the applied magnetic field. The small peak at g = 2.0 in the spectra was identified as emanating from the cavity of the EPR instrument in this particular experiment; the signal is not from the sample. Table S6. EPR simulation parameters of Cu(II) signals indicated in the spectra of Figure S22.