The electrochemical reduction of CO₂ represents a promising pathway for the sustainable production of fuels and chemicals while contributing to carbon-neutral energy technologies. However, achieving high activity and selectivity toward multicarbon (C₂⁺) products remains a major challenge due to the complexity of the reaction network and the strong dependence of catalytic performance on catalyst composition, reaction intermediates, and operating conditions. In particular, efficient CO₂ conversion requires the optimization of two fundamental steps: the selective generation of CO and its subsequent transformation into multicarbon products through carbon-carbon bond formation.

Here, we present an integrated computational framework aimed at establishing catalyst design principles for CO₂ electroreduction by connecting these two key stages of the reaction pathway. The framework combines density functional theory (DFT), constant-potential simulations, and machine-learning-assisted materials screening to investigate catalyst behavior under realistic electrochemical conditions and accelerate the discovery of promising electrocatalysts.

We first investigate the origin of the remarkable selectivity of Ag-based electrocatalysts toward CO production.[1] By explicitly accounting for both thermodynamic and kinetic effects as a function of the applied potential, we demonstrate that the competition between CO₂ reduction and hydrogen evolution is governed by a subtle interplay between reaction energetics and activation barriers. The resulting mechanistic picture explains the experimentally observed transition from hydrogen evolution to highly selective CO formation and highlights the conditions required for efficient CO generation, providing a foundation for tandem CO₂-to-C₂⁺ conversion strategies.

Building on these insights, we focus on Cu-based catalysts, which are uniquely capable of promoting carbon-carbon coupling and generating multicarbon products. To explore the vast compositional space of Cu-based alloys, we developed machine-learning models trained on large DFT datasets of CO adsorption energies across different alloy compositions, surface facets, and local atomic environments.[2] The resulting models enable rapid screening of thousands of adsorption configurations while retaining predictive accuracy comparable to first-principles calculations. This approach identifies several promising alloying elements, including Ag, Au, Zn, In, Al, and Ga, capable of tuning CO adsorption toward regimes favorable for C-C coupling.[4]

To validate and rationalize these predictions, constant-potential DFT calculations were employed to investigate CO dimerization on selected CuM surfaces.[3] The simulations reveal that p-block alloying elements, particularly Al and Ga, promote electron donation to adsorbed intermediates, stabilize OCCO species, and lower the activation barrier for C-C bond formation relative to pure Cu.[3] Furthermore, a strong correlation emerges between dimerization energetics and the excess surface charge required to maintain the applied potential, identifying a physically meaningful descriptor that captures both electrostatic and covalent contributions to catalytic activity.[3]

This work provides a unified perspective on the elementary processes governing CO₂ electroreduction, bridging selective CO production and multicarbon product formation. By integrating mechanistic understanding with data-driven catalyst discovery, the proposed framework offers practical guidelines for the rational design of next-generation electrocatalysts for efficient and selective CO₂ conversion.