Pulsed Electrolysis with a Nickel Molecular Catalyst Improves Selectivity for Carbon Dioxide Reduction

Pulsed electrolysis can significantly improve carbon dioxide reduction on metal electrodes, but the effect of short (millisecond to seconds) voltage steps on molecular electrocatalysts is largely unstudied. In this work, we investigate the effect pulse electrolysis has on the selectivity and stability of the homogeneous electrocatalyst [Ni(cyclam)]2+ at a carbon electrode. By tuning the potential and pulse duration, we achieve a significant improvement in CO Faradaic efficiencies (85%) after 3 h, double that of the system under potentiostatic conditions. The improved activity is due to in situ catalyst regeneration from an intermediate that occurs as part of the catalyst’s degradation pathway. This study demonstrates the wider opportunity to apply pulsed electrolysis to molecular electrocatalysts to control activity and improve selectivity.


Electrochemistry
Electrochemical measurements were carried out using Biologic SP-200, Biologic VSP and Ivium Vertex potentiostats. All electrochemistry was run in a standard glass half-cell using a glassy carbon electrode (A = 0.071 cm 2 ) (IJ Cambria Scientific Ltd) as the working electrode, a platinum mesh as the counter electrode, separated by a glass sleeve with a Vycor frit and a Ag/AgCl reference electrode (see picture below). The GC working electrode was polished with 1.0 µm and 0.05 µhm Micropolish TM Alumina on 8" mircocloth (Buehler) for 4 minutes before sonicating with Milli-Q. Electrolyte was 0.5 M NaCl from Merck (99.5%) in Milli-Q water (18.2 MΩ) pre-electrolysed (-0.1mA) overnight with a titanium plate (working) and carbon counter. 1 The cell was purged with either N2, CO2 (for cyclic voltammetry and impedance spectroscopy) or CO2 with 1% CH4 (BOC) (for chronoamperometry) for ~30 minutes prior to experiments. All electrochemistry was run at room temperature and pressure. Electrolysis experiments used the same cell and electrode configuration as the CV studies (image below) with the addition of a magnetic stirrer bar. The experiments were run in triplicate, with total current densities and GC measured selectivity the average of all runs with errors calculated as one standard deviation. The error bars given for values derived from these measurements (e.g. partial current densities) were obtained through standard error propagation methods. The current sampling-rate in the long-pulsed electrolysis experiments becomes limited by the large number of data points generated and the intrinsic limitations of the software used (EC lab). For a more detailed kinetic analysis of the chronoamperometric response we carried data collection over a short time period (50 s) with a high current sampling rate (sub-ms data resolution, figure S8). These short, high sampling-rate, data collection periods took place immediately after a period of electrolysis (30 minutes) using the standard data sampling rate.

Product detection
Gaseous products were measured by gas chromatography, by taking manual injections directly from the cell headspace and analysed using an Agilent 6890N with a 5 A molecular sieve column (ValcoPLOT, 30 m length, 0.53 mm ID) and a pulsed discharge detector (D-3-I-HP, Valco Vici). Moles of product were quantified using a calibration curve from known concentrations of H2, CO and CH4. A CH4 internal calibrant of known concentration was also used in the cell to confirm the accuracy of the calibration The faradaic efficiency was calculated using the equation above where the number of electrons for the reaction is 2. Moles of electrons are obtained from charge passed during electrolysis.

Figure S1
Photograph of cell set-up used in electrolysis X-ray photonelectron spectroscopy (XPS) X-ray photoelectron spectroscopy (XPS) was utilised to determine the nature of deposited species on the glassy carbon surface. The samples underwent 3 h standard and pulsed (EA = -1.0 V) electrolysis in 0.1 mM Ni(cyclam) in 0.5 M NaCl and wee thoroughly washed with Milli-Q water and dried with compressed air before inserting in the XPS chamber. Analysis of a powder sample of Ni(cyclam) was used for comparison and is reproduced from Siritanaratkul with permission. 2 The measurements were performed on a Kratos Axis Supra instrument using Al Kα X-ray source. The survey scans were performed at 80 eV pass energy and high-resolution scans were performed at 20 eV pass energy. The energy calibration was performed using O 1s peak at 530.9 eV.

Double Potential Step Chronocoulometry (DPSC) of Ni(cyclam) on GCE
Here double potential step chronocoulometry was performed on 0.1 mM Ni(cyclam) to assess the adsorption of catalyst onto GC working electrode. This was done using methods from literature, 3,4 where the potential of interest was held for 30 s to allow reductive adsorption of [Ni(cyclam)] + to occur, the potential was then jumped to +0.2 VNHE in the first step and then stepped back to the initial potential. The charge passed over 10 ms in the first step was plotted vs t 1/2 while the charge passed over 10 ms in the second step was plotted vs [τ 1/2 + (tτ) 1/2t 1/2 ] where τ is the duration of the first step and t is the time. The difference of the forward and reverse intercepts gives charge of adsorbed species, Qads, from which the surface coverage, Γ, was calculated using: = where F the Faraday constant, n the number of electrons and A the electrode area in cm 2 .

Calculation of Full Cell Voltage Efficiency (VE) and Full Cell Energy Efficiency (EE)
Where Efull cell is the full cell applied potential; ECO = -0.109 V is thermodynamic potential (vs RHE) of CO2 reduction to CO and FECO is the measured CO Faradaic efficiency as a percentage. 5  Here the cell was pulsed 10 times (< 1 min) at a higher time resolution in order to break down the anodic pulse profile (Ea = -0.3VAg/AgCl) into faradaic and non-faradaic regions using methods done by Kimura et al. 10 In Fig S8(c) the natural log of current density is plotted against time, here we see deviation from the linear trend after ~10 ms, suggesting that Faradaic current begins to contribute significantly past this time.    As shown in the Fig S10, the spectrum for the standard sample is different from the Ni(cyclam) powder indicating a change in structure. The limited amount of Ni on the surface resulted in noisy spectrum, which was deconvoluted using literature parameters for Ni, NiO and Ni(OH)2 structures and is shown in Fig S11. In the spectrum, multiplet splitting arises due to the presence of unpaired electrons which upon interaction with the core electrons, can create multiple energy levels. 6 By the deconvolution, it was determined that Ni(OH)2 is the dominating species with 68.5% composition on the surface. 7 With the remainder in NiO (21.5%) and Ni metal forms. O 1s spectrum was analysed for further insight into Ni form remaining on the standard sample ( Fig S13). The presence of single peak of O 1s further corroborates that Ni hydroxide is the dominating phase. 8 For the pulsed sample (red, Fig S3), there is no Ni of any form left on the surface. The absence of corresponding N 1s peak in figure 2, confirms the absence of Ni(cyclam) from the surface.

Differential capacitance curves for calculating PZC
The impedance of the system can be modelled as an ohmic resistance and a capacitance (the double layer capacitance of the glassy carbon electrode) in series, as follows: Where Z is the impedance of the interfacial region between the working and reference electrode, where ZRe and ZIm are the real and imaginary parts respectively; is the angular frequency of the ac perturbation; f is the frequency. Here the capacitance is calculated using the single-frequency impedance over a potential range of +0.6 to -1.2 VAg/AgCl recording every 36 mV with 1-minute equilibration time between each potential with an amplitude (Va) of 10 mV using methods from literature. 9

Figure S15
Normalised differential capacitance curves of GCE in NaCl (aq) Where is all cathodic charge passed, is all anodic charge passed and is all charge passed.

(%) = × 100
Where tc is time spent at the cathodic potential (undergoing eCO2R) and T is total time of cycle.

EC (V) [tC]
EA (V) [ This highlights the benefit of applying a less positive Ea and shorter ta with a of Qc > 97 % for methods highlighted in bold. 11