A hydrophobic Cu/Cu2O sheet catalyst for selective electroreduction of CO to ethanol

Electrocatalytic reduction of carbon monoxide into fuels or chemicals with two or more carbons is very attractive due to their high energy density and economic value. Herein we demonstrate the synthesis of a hydrophobic Cu/Cu2O sheet catalyst with hydrophobic n-butylamine layer and its application in CO electroreduction. The CO reduction on this catalyst produces two or more carbon products with a Faradaic efficiency of 93.5% and partial current density of 151 mA cm−2 at the potential of −0.70 V versus a reversible hydrogen electrode. A Faradaic efficiency of 68.8% and partial current density of 111 mA cm−2 for ethanol were reached, which is very high in comparison to all previous reports of CO2/CO electroreduction with a total current density higher than 10 mA cm−2. The as-prepared catalyst also showed impressive stability that the activity and selectivity for two or more carbon products could remain even after 100 operating hours. This work opens a way for efficient electrocatalytic conversion of CO2/CO to liquid fuels.

The formation of ethanol starts by the C-C coupling step leading to *C2O2 species, whose energy is close to the reference (pristine surface and gas phase CO) . This is a limiting step of the reaction on (111) surface.
For several initial adsorption configurations, a spontaneous dissociation of C2O2 into two CO molecules is observed during the simulation. However, if the electron transfer, decoupled from the proton transfer, occurs already at this step, it can be stabilized.
Besides, this species can be stabilized by solvent molecules. Therefore, the applied electrode potential of -0.70 V vs RHE ensures the feasibility of ethanol formation with Cu2O catalyst. All consecutive reaction steps are exergonic (∆ < 0) with gradually decreasing Gibbs free energy. The energy of adsorption on the (111) Cu2O surface is particularly favorable for the hydrogen-rich intermediates, such as *C2OH, *C2H2O, *C2H3O and *C2H5O. In particular, Cucus sites are important for stabilizing these species through the interaction with non-polar carbon C, to which H atoms are progressively attached. The Cucus-C distance in these intermediates is relatively short (1.8-1.9 Å), suggesting a strong interaction between the adsorbate and the surface.
Besides, the Cucus site participates in the stabilization of *C2H5O adsorption via O atom. Therefore, the appropriate surface structure of the employed electrocatalyst predetermines its efficiency in the ethanol formation reaction.

COSMO-RS calculation
In this work, BP functional combined with def2-TZVPD basis set was employed to carry out the quantum chemical COSMO calculations for CO, water and n-butylamine molecules. After obtaining their COSMO files, COSMO-RS calculations were subsequently performed using COSMOtherm C30_1601 program [54,55] to evaluate the macroscopic solubility of CO in water and n-butylamine respectively.

Molecular dynamics simulations
Molecular dynamics simulations were performed to obtain the diffusion coefficients of CO in water and n-butylamine with Gromacs 2019.6 program package [56]. 10 CO molecules and 560 water or n-butylamine molecules were packed into the simulation boxes using Packmol program [57]; it is worth mentioning that the system size has been reported to be enough to gain the reliable CO diffusion coefficients in the molecular solvents [58]. The water molecule adopted the SPC/E (Extended Simple Point Charge) model and the n-butylamine molecule used the classical GAFF force field [59], whereas the parameters of CO were taken from the literature [58]. The restrained electrostatic potential (RESP) method was employed to get partial charges of the systems. The initial systems were energetically minimized with the convergence criteria of 100 kJ⋅mol -1 /nm. Following that, NPT ensembles were used to perform 80 ns simulations to make the systems verge to equilibrium. The temperature was set at 298.15 K with the velocity-rescale heat bath [60], while the pressure was controlled by Berendsen algorithm for the former 30 ns and Parrinello-Rahman scheme for the last 50 ns [61,62]. The long-range coulomb interactions were calculated by particle mesh Ewald (PME) method [63]. The LINCS algorithm [64] was used to constrain all the bonds connecting with hydrogen. The equations of motion were integrated by the leap-frog algorithm and the time step was set to 2 fs.
After the simulation boxes equilibrium reached, the last frame of simulation trajectory was chosen as the initial configuration to carry out another 50 ns NVT production simulation for the diffusion coefficient calculations.

Supplementary
Besides, the presence of hydrophobic n-butylamine on the surface increases the chemical potential of water. This happens both due to the increase of the internal energy of water (E) within the hydrophobic layer originating from the unfavorable hydrophobic interactions between the water molecules and hydrocarbon chains of nbutylamine and due to the restriction of the phase space (Q), accessible for water molecules within the hydrophobic environment of n-butylamine coating (hydrophobic n-butylamine repels water molecules): respectively.
As a consequence, the formation of oxygen vacancies, and thus departure of O atoms and reduction of the Cu2O towards the pristine Cu becomes even more unfavorable, since water molecules being able to approach the Cu2O surface is a prerequisite of the oxygen vacancy formation reaction taking place (in fact, direct reaction 2 = 2 + 1 2 � 2 is much more unfavorable than the watermediated one). The corresponding Gibbs free energy diagrams are demonstrated in the Supplementary Fig.25.
Another factor that plays a role here is that the oxygen vacancy formation process is less energetically favorable than the isoelectronic process of CO reduction at the External method can also be employed to calculate the FEs of the liquid products.
As shown in Supplementary Table 6 and Supplementary Figure   In a full collection protocol, a flask of water for washing the CO off-gas was added to collect the liquid products before CO exhausted and the flow rate of the outlet was monitored with a flow meter (Fig. S29a). To identify the crossover of the liquid-phase products formed during the CORR, the catholyte, anolyte and water in the flask were collected and analyzed using 1 H-NMR. The results for CORR were exhibited in Supplementary Figs. 16-17 and 22. And the results showed that liquid products indeed migrated across GDE and AEM. By adding up all of these detected products, the total FE reached 100±3%, confirming that the liquid products in the CO off-gas and anolyte were the "missing" products. Acetone, acetaldehyde and propionaldehyde were only detected in the CO off-gas but not in the catholyte and anolyte. This should be attributed to the low production rates and high volatility.
To elucidate the carbon balance path, flow meters were used to monitor the inlet and outlet flow out of the reactor (Fig. S29a). Figs. S29b-d show the outlet flow rate as a function of current density. When J=0 mA cm -2 , there's no obvious discrepancy in the flow rate between gas inlet and outlet. As current densities increased, the outlet rate gradually decreased in all the electrolytes, which corresponds to a gradual enhancement in the consumption rate of CO. In addition, as OHcannot react with CO, the outlet flows are approximately equivalent to inlet flows at low current densities in basic solution. And the basic electrolytes cannot change the outlet flow rates, which is different from the CO2RR in basic electrolyte (Energy Environ. Sci., 2020, 13, 977).
For the final carbon balance (Supplementary Table 7), the unreacted CO flow rate after the reactor and consumed CO flow rate for the conversion into products added up to a total of ~20 mL min -1 at various current densities, which was equal to CO inlet flow rate used in the experiment.