Statistical Sampling-Driven Design for Supported Bimetallic Nanocatalysts for CO₂ Reduction

Tailoring supported bimetals to alloyed or phase-separated structures is of vital importance while this process is blocked by the support interferences during universal impregnation processes. Conventional trial-and-error approaches rooted in chemical intuition often lack efficiency and generality. Here, we present a design strategy guided by the statistical sampling of comparative ease of alloy formation through the metadynamics-based gas-solid nanoreactor approach, which enables the rational and systematic development of bimetallic nanocatalysts (NCs). Using metal oxide-supported PdAu coupling with model CO₂ reduction as a proof-of-concept system, the integrated theoretical and experimental results not only validated the reliability of simulation results but also successfully predicted and realized the alloy formation or phase separation of supported PdAu NCs. The generated PdAu alloys over CeO₂ weaken the metallicity of supported Pd species and thus the catalytic hydrogenation property, but increase moderate basicity, contributing to activated CO₂ hydrogenation to CO via a formate intermediate. However, the phase separation of Pd and Au over TiO₂ support promotes formic acid production efficiency attributed to increasing weak basicity to accelerate CO₂ activation during a bicarbonate pathway. These findings highlight statistical sampling as a general broadly applicable framework for the rational design of advanced bimetallic NCs.