Electroreduction of CO2 in a Non-aqueous Electrolyte—The Generic Role of Acetonitrile

Transition metal carbides, especially Mo2C, are praised to be efficient electrocatalysts to reduce CO2 to valuable hydrocarbons. However, on Mo2C in an aqueous electrolyte, exclusively the competing hydrogen evolution reaction takes place, and this discrepancy to theory was traced back to the formation of a thin oxide layer at the electrode surface. Here, we study the CO2 reduction activity at Mo2C in a non-aqueous electrolyte to avoid such passivation and to determine products and the CO2 reduction reaction pathway. We find a tendency of CO2 to reduce to carbon monoxide. This process is inevitably coupled with the decomposition of acetonitrile to a 3-aminocrotonitrile anion. Furthermore, a unique behavior of the non-aqueous acetonitrile electrolyte is found, where the electrolyte, instead of the electrocatalyst, governs the catalytic selectivity of the CO2 reduction. This is evidenced by in situ electrochemical infrared spectroscopy on different electrocatalysts as well as by density functional theory calculations.


Supplementary Note 1 Electrochemistry and electrochemical infrared spectroscopy Cyclic voltammetry
To electrochemically investigate the oxide free electrodes in an acetonitrile (99.8 %, anhydrous, Sigma Aldrich) electrolyte, the Mo 2 C electrodes were prepared 1  To determine the potential of the quasi-reference electrode, the half-wave potential (E 1/2 ) of the ferrocene/ferrocenium couple in acetonitrile was determined as 0.182 V ( Figure S1). All potentials were shifted against this value of

Electrochemical infrared reflection absorption spectroscopy (EC-IRRAS) studies
The EC-IRRAS studies were carried out using a VERTEX 70v spectrometer (Bruker) with an additional external chamber (XSA, Bruker), equipped with a liquid N 2 cooled mercury cadmium telluride (MCT) photodetector. A linear polarizer (Edmund Optics) was introduced in the optical pathway, before the IR beam entered the spectro-electrochemical cell with the use of a movable gold-coated mirror. A thin-layer cell configuration was employed to minimize the contribution of the electrolyte solution: after passing a CaF 2 hemisphere constituting the bottom of the homemade spectro-electrochemical cell, the IR beam was reflected from the surface of the WE that was pressed against the hemisphere and enters the detector with the use of a second gold-coated mirror. The cell was equipped with a carbon rod (Ultra carbon corporation) counter and the above-mentioned PTFE bound activated carbon quasi-reference electrode. To account for the strong absorption of the thin electrolyte film, the single spectra, recorded at a specific applied potential, are subtracted and normalized with the spectra recorded at a specific reference potential (E = -1.0 V Fc/Fc+ ), at which no reaction should occur, according to with R(E S ) being the reflectance sample single spectra and R(E R ) being the reflectance reference single spectra.
After normalization, each upward or downward facing band corresponds to consumed/disappeared or formed/accumulated species at the electrode surface, respectively.
EC-IRRA spectra are taken either by alternating potential modulation, where the reference spectra are recorded between each sample spectrum, or by step potential modulation, in which the reference spectrum is taken once at the beginning of the experiment. The latter option was usually employed if not stated otherwise.
To ensure that the electrodes are oxide free, the spectro-electrochemical cell was assembled in an Ar-filled glovebox and transported under air exclusion to the IR spectrometer. Before each series of measurements, the electrolyte was purged for 20 minutes, either with Ar or CO 2 , followed by immersion of the WE at a potential of -1.0 V Fc/Fc+ and an approach of the WE surface to the optical window.

Ar-purged electrolyte
To determine the bands originating from the electrolyte (acetonitrile with the conducting salt), measurements in deaerated electrolyte were performed ( Figure S2). The bands in the wavenumber region of 3002-2880 cm -1 and 1500-1376 cm -1 are assigned to the stretching and bending modes of the conducting salt and of acetonitrile. Additionally, acetonitrile shows two strong bands at 2292 cm -1 and 2251 cm -1 and five small bands, which is in perfect agreement with the literature. 5 These bands from the electrolyte are not considered in the interpretation and discussion of the CO 2 RR results determined with EC-IRRAS. Figure S2: EC-IRRA spectra of Mo 2 C recorded in Ar-purged acetonitrile with 0.1M TBAPF 6 . The spectra show the stretching (3000-2880 cm -1 ) and bending (1500-1376 cm -1 ) vibrations for acetonitrile (green boxes) and the conducting salt (yellow boxes) and two distinct (2292 and 2251 cm -1 ) as well as five smaller bands of acetonitrile (green boxes). The reference spectrum was recorded at -1.0V Fc/Fc+ .

S-and p-polarized light
Measurements with p-and s-polarized light were performed to determine if any adsorbed species is formed during the reduction reaction. The different light polarizations were implemented through manual rotation of the linear polarizer (Edmund optics) by 90°. Ppolarized light probes species in solution and species adsorbed at the surface, while s-polarized light probes molecules in solution only, which is due to the destructive interference of incident and reflected beam at the electrode surface. 6 EC-IRRA spectra of Mo 2 C in acetonitrile with 0.1M TBAPF 6 at -2.4 V Fc/Fc+ for both p-and slight polarizations are shown in Figure S3. Both spectra show exactly the same bands, which indicates that in the case of the CO 2 reduction at Mo 2 C no adsorbed species is present.

Acetonitrile decomposition
The decomposition of acetonitrile starts at a potential of -2.4 V Fc/Fc+ with the formation of two distinct bands at 2118 cm -1 and 1517 cm -1 , along with less intense bands ( Figure S4, orange boxes). The two dominant signals are associated with the formation of the 3-aminocrotonitrile anion, as known from the literature 7 . The proposed reaction route is the formation of a hydride ion, by reduction of a hydrogen atom in the metal lattice, which is reacting with acetonitrile to Figure S3: EC-IRRA spectra with s-and p-polarized light for the CO 2 reduction at Mo 2 C in acetonitrile with 0.1M TBAPF 6 . The reference potential was taken at -1.0V Fc/Fc+ . its anion. 7 This acetonitrile anion nucleophilically attacks a second acetonitrile molecule and forms the 3-aminocrotonitrile anion, as shown in the inset of Figure S4.
Foley et al. 7 demonstrated that application of a more anodic potential, after the formation of the anion, leads to a protonation of the anion and formation of 3-aminocrotonitrile. In Figure S5, a potential of -2.8 V Fc/Fc+ was applied to form the anion before increasing the potential back to its reference value of -1.0 V Fc/Fc+ . This leads to a shift of the band from 2118 cm -1 to 2180 cm -1 and to the formation of additional less intense bands at around 3400 cm -1 and 1600 cm -1 ( Figure   S5), which perfectly agrees with the literature. 7 Thus, we conclude that acetonitrile starts to decompose at E ≤ -2.4 V Fc/Fc+ to the 3-aminocrotonitrile anion.

IR transmission studies
As a reference measurement, and to unambiguously determine the nature of the band at 2138 cm -1 , interpreted as dissolved CO, transmission cell IR studies were performed with a classical transmission cell setup (Bruker). In this setup, the electrolyte is placed between two CaF 2 plates, and the difference between CO-purged and CO-free acetonitrile IR spectra was measured. The difference spectra is shown in Figure S6 and it reveals one distinct band at 2138 cm -1 as reference signal for dissolved CO. The formation of dissolved CO with a corresponding band at 2138 cm -1 in the same acetonitrile electrolyte has been earlier shown by Figueiredo et al. 8 .

Supplementary Note 2
Computational details.

Solvation effects of acetonitrile
Solvation effects of acetonitrile were considered within an implicit solvation approach using the Environ module 9,10 provided with QE. Parameter settings are detailed in Table S1. The adsorption free energies (G ad ) are calculated as: where is the adsorption energy, is the vibration correction, and is the correction of solvation effects. Figure S6: Transmission cell IR difference spectrum of CO-purged acetonitrile with 0.1M TBAPF 6 . A distinct band at 2138 cm -1 , corresponding to dissolved CO is clearly visible. The difference spectrum is calculated by subtraction of the single spectra of CO with that in Ar-purged acetonitrile with 0.1M TBAPF 6 .
The Gibbs free energies of CO 2 and H 2 O molecules are calculated as: where is the total energy of an isolated molecule, is the vibration correction, and is the concentration correction (with k B the Boltzmann ( ) constant and T the temperature). The vibration correction is calculated via DFT. The thermodynamic activity a at a low concentration is approximately equal to the molar concentration c, which for CO 2 at 100 kPa and 298 K is 279 mol•m -3 in acetonitrile 11 , and for H 2 O is 10 -6 molar fraction (ppm level) in experimental high-purity acetonitrile.

Benchmark of vibrational frequencies on Cu(211)
Benchmark work is first done for the well-studied system Cu/acetonitrile 8 . In that work, Figueiredo et al. 8 assigned IR spectroscopy data to CO, CO 3 , HCO 3 and HCO 2 surface species.
Vibrational frequencies are correspondingly computed for these and a wide range of other potential C,O species (CO, CO 2 , CO 3 , C 2 O 2 , C 2 O 4 ) and C,O,H species (COH, CHO, COOH, HCO 2 , HCO 3 ) as summarized in Figure S7. These calculations confirm the assignments made by Figueiredo et al. 8 , but exhibit an expected systematic offset in the computed absolute frequencies. Figure S8 shows a corresponding parity plot, revealing the excellent linear relation between experimental and computed data. This linear relationship extends over both data computed in vacuum and in implicit solvent, with the latter in fact leading only to a small reduction of the offset. This suggests that the offset arises predominantly from the approximate DFT functional. Henceforth the linear relationship obtained here is employed to correct for this systematic error. Table S2 compiles the thus corrected absolute frequencies and demonstrates that the remaining error in absolute frequencies after this correction is on the order of a few percent.

Adsorption of CO 2 reduction species on Mo 2 C(110)
The same C-containing adsorbates as in the benchmark are calculated on the C-rich Mo 2 C(110) surface in an acetonitrile environment. A variety of adsorption sites and configurations is considered to find the most stable adsorption mode (in the employed sign convention here reflected by a most negative G ad ). The corresponding results are summarized in Table S3. For almost all adsorbates there is one energetically preferred adsorption configuration. The only exceptions are C 2 O 2 and CO 3 , where the second most stable configuration is only by 40 meV and 70 meV weaker bound, respectively. We consider the latter small energetic difference within the uncertainty of our approach and correspondingly also keep this second most stable configuration in mind for the assignment to the experimental fingerprints.

Vibrational frequency calculation of the products
The vibrational frequencies of the products (3-aminocrotonitrile anion, 3-aminocrotonitrile, carboxylated acetonitrile and carboxylated 3-aminocrotonitrile anion) were calculated in each supercell with a side length of 20 Å (Table S4). The calculated frequencies (ν DFT ) are offset corrected according to following equations ν Theo = s • ν DFT (S1) s = ν Exp (I) / ν DFT (I) (S2) where ν DFT is the calculated vibrational frequency, ν Theo the offset corrected vibrational frequency and s the factor for the offset correction.