Potential Dependent Reorientation Controlling Activity of a Molecular Electrocatalyst

The activity of molecular electrocatalysts depends on the interplay of electrolyte composition near the electrode surface, the composition and morphology of the electrode surface, and the electric field at the electrode–electrolyte interface. This interplay is challenging to study and often overlooked when assessing molecular catalyst activity. Here, we use surface specific vibrational sum frequency generation (VSFG) spectroscopy to study the solvent and potential dependent activation of Mo(bpy)(CO)4, a CO2 reduction catalyst, at a polycrystalline Au electrode. We find that the parent complex undergoes potential dependent reorientation at the electrode surface when a small amount of N-methyl-2-pyrrolidone (NMP) is present. This preactivates the complex, resulting in greater yields at less negative potentials, of the active electrocatalyst for CO2 reduction.

The VSFG experiments were performed using a newly constructed spectrometer at The University of Liverpool.90 % of the output PHAROS-PH1-SP (Light Conversion, 1030 nm, 10 kHz, 10 W, 170 fs pulse duration) laser system is used to generate both the nIR and mIR laser pulses required for the VSFG experiments. 1 W of the output is used to generate a time asymmetric narrow-band nIR pulse (1030 nm, 10 Hz, ~1.5 ps, ~13 cm -1 linewidth) via an etalon (SLS Optics), which is directed to the sample through a half-wave plate (Thorlabs, WPH10M-1030) and polariser (ThorLabs, LPVIS050-MP2) to rotate the light to a horizonal polarisation, and p-polarised w.r.t refection from the sample, at an angle of incidence of ~45°.A 950 nm long pass filter (Thorlabs, FEL0950) is also used to filter out second harmonic 515 nm light generated within the halfwave plate.This nIR pulse is focused by a 30 cm lens (Thorlabs LB1779-B), with the sample placed ~25 cm away from this lens, resulting in an approximate beam diameter at the sample of ~400 µm (1/e 2 diameter) with a power of ~15 mW. 8 W of the laser output is used in an IR OPA (Light Conversion, Orpheus-One-HE) to generate the broadband IR beam which can be tuned across the frequency range of interest (10 kHz, 170 fs pulse duration and ~150 cm -1 @1900 cm -1 ).The mIR output passes through a twisted periscope to switch the polarisation from vertical to the desired horizontal polarisation.The purity of the polarisation is checked using a polariser (Thorlabs, LPMIR050-MP2) which is removed from the beam path prior to VSFG measurement.The mIR beam is focussed onto the sample, with an approximate beam diameter of 250 µm, using a Au parabolic mirror (Thorlabs, MPD249H-M01) at an angle of incidence of ~50° and p-polarisation w.r.t reflection from the sample, with a power of ~50 mW.A gas-purge generator largely removes H2O, CO2 and other contaminants in the air from the IR beam path.Resulting p-polarised VSFG light is directed through 950 (Thorlabs, FES0950) and 900 nm (Thorlabs, FES0900) short pass filters to remove the 1030 nm nIR beam.The beam goes through another polariser (Thorlabs, LPVIS050-MP2) before being focussed through 150 µm slits and into the spectrograph (Andor, Kymera) using a 15 mm focal length lens (Thorlabs, LA1540-B), and is detected on a CCD camera (Andor, iDus416).nIR/mIR delay is introduced using a linear stage (ThorLabs, LTS300C) on the nIR beam path; all spectra reported herein have been obtained with the nIR pulse delayed by ~0.5 ps w.r.t. the mIR pulse to suppress the nonresonant contribution to the spectrum. 8Each spectrum was accumulated on the CCD for 2 s before reading out.The spectrograph was calibrated using Ne spectral lines and the rovibrational P, Q and R

CO2 branches).
All VSFG experiments were performed in a custom spectroelectrochemical "cross-cell" as used in our previous work. 7The 1.6 mm diameter Au (IJ Cambria) working electrode was mechanically polished for 10 minutes using 1.0 µm then for a further 10 minutes using 0.05 µm alumina suspension.The electrode was rinsed thoroughly and sonicated in Milli-Q water between each step.The working electrode was then secured into the cross cell with a Ag wire pseudo reference electrode (sanded, rinsed and sonicated before each experiment) and a Pt counter electrode (flame annealed each day).A Teflon spacer (50 µm), was placed between the Au working electrode and window to ensure the a consistent pathlength between measurements.The solution of 1 mM Mo(bpy)(CO)4, 0.1 M TBAPF6, in CH3CN or CH3CN with 10% (vol.)NMP and electrochemical cell were purged for 10 minutes with Ar, before solution was transferred under Ar to the spectroelectrochemical cell.Figure S1 shows the FTIR spectra of Mo(bpy)CO4 in the solvents used.
Owing to the intense absorption of NMP at ~1675 cm -1 (see Figure S1), the addition of greater volumes of NMP begin to obscure the ν(CO) bands of Mo(bpy)(CO)4, leading to the possibility of phantom transitions in the VSFG spectra to be observed. 9Owing to the agreement between the catalytic activity in 10 % (vol.)NMP and pure NMP, 10 we conclude the addition of a small volume of NMP replicates the electrochemical behaviour of pure NMP.This indicates that species at the electrochemical interface is unlikely to reflect that found in the bulk, resulting in relatively small concentrations of additives to have profound effects on catalytic activity.For spectroelectrochemical measurements in pure CH3CN, the potential was scanned from -0.5 → -2.3 V → -0.5 V at 5 mV s -1 .The forward sweep is shown in Figure 2, whereas the full scan range is shown in Figure S1.For CH3CN with 10% (vol.)NMP, the potential was scanned from -0.3 → -1.5 V → -0.3 V at 2.5 mV s -1 again forward sweep is shown in Figure 2, whereas the full scan range is shown in Figure S2.The lack of VSFG bands at potentials more negative than R2 is likely a result of disorder at the electrode-electrolyte interface.For assignments see supplementary note 2. We have shown that Mo(bpy)CO4 -undergoes CO loss and is rapidly reduced to Mo(bpy)CO3 2-on a Au electrode, 7 a potentially non-reversible couple. 10We limit the most negative potential studied to -1.5 V, ~250 mV more negative than the onset of electrocatalysis in the presence of CO2, and ~300 mV more positive than R2 (Figure 1).

Electrochemical Measurements
Preliminary EIS measurements in the range 100 kHz -1 Hz were conducted, and a frequency in the low region (10 Hz) was chosen to reflect the double layer capacitance and avoid convolutions due to electrolyte resistance at low electrolyte concentration 11,12 (0.1 mM TBAPF6).Potentiostatic EIS was then measured at 10 Hz, starting from +1 V to -2 V vs Ag in 20 mV steps, with a 10 mV amplitude and 10 s equilibration time.The double layer capacitance can be approximated as a capacitance in series to the solution resistance, and the differential capacitance can be calculated from C = -1/(2ΠfZIm), where C is the differential capacitance, f is the frequency, and ZIm is the imaginary component of the measured impedance. 13pplementary Note 2: Assignment of Spectral Bands Table S1.Assignment of bands observed in the in situ VSFG spectra.
Peak positions as shown in Figure 3, have been obtained through fitting with Voigt functions; overlapping bands have been fitted using through a simultaneous combination of Voigt functions (see Figure S3), whereas isolated bands were fit using a single Voigt function in a separate fitting step.As shown in Figures 2 and S1, the VSFG spectra obtained at the least negative potentials are remarkably similar in both solvents, and similar to those reported in our previous work. 7VSFG spectra obtained at -0.55 V are shown in Figure S2a and b in CH3CN and 10%(vol) NMP, respectively.The main band at ~1900 cm -1 in CH3CN is clearly asymmetric, and well fit through a combination of two Voigt functions.
The higher wavenumber band shows strong potential dependence (tuning rate ~ 28 cm -1 V -1 ) whereas the most intense, lower wavenumber band shows a far smaller potential dependence (Figure 2c).The wavenumber and tuning rate of the higher wavenumber band is consistent with CO adsorbed at bridged sites on the Au electrode, 14 denoted CO@Au, with the CO produced from reductive dissociation of Mo(bpy)CO4 on the Au surface from CVs recorded prior to spectroelectrochemical measurements to ensure successful construction of the SEC cell.There is little evidence of the CO@Au band in NMP, (Figure S2b) despite CVs also being recorded prior to SEC measurements.A clear shoulder to the 1890 cm -1 band is observed at ~1868 cm -1 when NMP is present (Figure 2b), and another band is observed at ~1825 cm -1 in both solvents.ν(CO) vibrations of Mo(bpy)CO4 in CH3CN are observed at 2115, 1904, 1875 and 1832 cm -1 , Figure S1; the highest wavenumber mode is outside of the spectral window probed in this work, where as the remaining vibrations are in excellent agreement with the bands observed at ~1890, 1868 (NMP only) and 1825 cm -1 at -0.55 V, hence we assign all of these bands to ν(CO) vibrations of Mo(bpy)CO4.

Figure S3
. Fits of the VSFG spectra obtained at -0.55 V in a) CH3CN and b) 10%(vol) NMP.The broad baseline is a result of incomplete suppression of the non-resonant signal.Bands marked by an asterisk (*)are "phantom transitions," which arise as a result of absorption of mIR radiation by atmospheric water vapour through incomplete purging of the mIR beam path.c) Tuning rate of the overlapped vibrations at ~1900 cm -1 in CH3CN between -0.5 and -0.85 V; CO@Au, red circles and ν(CO) of Mo(bpy)CO4, black squares.
As the potential is swept more negatively, the intensity of the 1890 cm -1 ν(CO) of Mo(bpy)CO4 continues to increase until -0.80 V in CH3CN and -0.65 V in 10%(vol) NMP as Mo(bpy)CO4 accumulates at the electrode surface.Between -0.6 V and -1.0 V, the most dominant band in 10%(vol) NMP switches, with the intensity of the ~1868 cm -1 band increasing simultaneously as the intensity of the 1890 cm -1 band decreases, Figures 1 and S2.Notably, the vibrational wavenumber of the ~1890 cm - 1 ν(CO) band decreases over this potential range (Figure 3 ~1866 cm -1 (Figure 2b, Figure S4) between -0.8 and -1.0 V as the intensity of the ~1890 cm -1 ν(CO) of Mo(bpy)CO4 band decreases; again this coincides with a shift to lower wavenumber of the latter vibration (Figure 3).Along with these changes, the ν(CO) band at ~1825 cm -1 disappears.Importantly in CH3CN the 1890 cm -1 is dominant across the potential range of -0.55 to -1.0 V, whilst when NMP is present the dominant band switches from 1890 cm -1 to 1868 cm -1 , Figure S4.
), indicating a change in local environment is occurring.Through close inspection of the VSFG spectra in CH3CN, a weak band begins to form at