Molecular Design of Dispersed Nickel Phthalocyanine@Nanocarbon Hybrid Catalyst for Active and Stable Electroreduction of CO₂

Abstract

The molecular catalyst/nanocarbon hybrid through π–π stacking immobilization is an emerging family of single-atom catalysts with outstanding performance in electrocatalysis, well-defined active site, and tunability at molecular level through functional group substitution. In the present work, we provide a general strategy for the rational design of molecular single-atom catalyst in the form of nickel phthalocyanine@nanocarbon (NiPc@NC) for highly efficient electroreduction of CO2 to CO. We employ density functional theory (DFT) calculations and state-of-the-art electronic structure analysis to explore the mechanism and substituent effects on structural stability, redox chemistry, adsorption properties, and molecule–substrate interactions of the NiPc catalyst. We have revealed that the electron-withdrawing groups facilitate the reductive activation of the catalytic Ni center but weaken the Ni–N bond strength and make the CO desorption sluggish, while the electron-donating groups do the opposite. A substituent-dependent correlation between interaction strength and electron transfer through the interface is also revealed by noncovalent interaction analysis and electron density difference projection. On the basis of the gained insights, we apply semiempirical quantum mechanical (SQM) calculation, machine learning (ML), and genetic algorithm (GA) to screening through the chemical space of ca. 10 trillion substituted NiPc molecules under a descriptor scheme to identify promising molecular candidates for the NiPc@NC hybrid material. The best candidate from GA search outperforms the state-of-the-art catalyst in terms of stability, reduction potential (improved by 110 mV), and interaction with substrate (strengthened by 0.46 eV). Design strategies are proposed based on the top-scoring molecules from computational screening, and the workflow is highly generalizable and transferable to similar molecular systems for other applications.

Inverse molecular design of alkoxides and phenoxides for aqueous direct air capture of CO 2

“…8 million unique molecules (24 substituents, 5 sites), which is beyond the capability of brute force exhaustion. To efficiently explore the vast chemical space, we employ GA, an evolutionary algorithm that has been successfully applied to structural prediction and property optimization of molecular systems ( 15 , 26 ), to search for the substituted phenoxide with minimal on the condition that its is lower than 15.74. To lower the computational cost and speed up the GA search, the SQM method GFN1-xtb, with GBSA implicit solvation, is adopted.…”

“…On the other hand, inverse design is frequently used when there is insufficient insight into the structure–activity relationship; the pool of possible candidates can be screened, or molecular design can be treated as an optimization task where evolutionary algorithms can be applied to efficiently optimize the desired property in a predefined chemical subspace. Both strategies have proven successful in optimizing functional molecules for various applications, including solar cell ( 10 ), redox flow cell ( 11 ), solar heat battery ( 12 ), molecular photocatalysis ( 13 ), and electrocatalysis ( 14 – 17 ).…”

Molecular design of redox carriers for electrochemical CO2 capture and concentration

“…By setting the normal direction of the LSRs as the search goal, a global optimization algorithm can intentionally sample towards that direction and discover chemical principles with a deeper electronic structure origin, that leads to decoupling of the properties. For example, Zhang et al used this technique to propose new design principles to decouple CO 2 binding and redox potential by asymmetric substitution (break degeneracy of the HOMO) 67 and to decouple CO 2 binding and pK a by introducing bulky ortho groups to force CO 2 into a different orientation than that of a proton (so that they bind via different orbitals) to achieve stronger CO 2 binding capacity and mitigate water sensitivity at the same time. 68 However, such discretized and fixed-dimensionality encoding relies entirely on chemical knowledge of the system as in the molecular skeleton, therefore confining the search space in the predefined subspace.…”

Machine Learning for Electrocatalyst and Photocatalyst Design and Discovery

“…Such in silico fitness functions can accelerate the optimization process, analogous to the Baldwin Effect in evolutionary biology (describing how learning in organisms can accelerate evolution). , The use of ML models as fitness functions to replace the need for some experiments in the evolutionary design of materials can reduce the number of experiments required and allow efforts to focus on the material genomes with the highest performance. Genetic algorithms have been applied to a wide range of materials for discovery, property prediction, and optimization. ,− More details of the principle of this algorithm and the corresponding applications can be found in the recent comprehensive review by Le et al…”