# Design of Single‐Atom Catalysts Anchored in N‐Doped Biphenylene Using Symbolic Regression for Electrocatalytic Nitrate Reduction to Ammonia

## TL;DR

This paper proposes a new design for single-atom catalysts to efficiently convert nitrate into ammonia using symbolic regression and nitrogen-doped biphenylene networks.

## Contribution

A symbolic regression method is introduced to identify hidden descriptors for nitrate reduction electrocatalysts with high geometric variability.

## Key findings

- NO3RR performance of single-atom catalysts is strongly influenced by their local coordination environments.
- Hybridization between TM-3d and NO3− orbitals enhances nitrate activation through d-π* orbital formation.
- Symbolic regression outperforms traditional descriptors in capturing complex electrocatalytic behavior.

## Abstract

Electrocatalytic nitrate reduction (NO3RR) presents a synergistic strategy, achieving dual benefits in energy transformation and environmental remediation through a single process. However, the complexity of its reaction pathway and the lack of descriptors impede the rational design of high‐performance NO3RR electrocatalysts. Herein, employing a series of transition metal doped and nitrogen decorated biphenylene network (TM‐CxNy@BPN), an inclusive recipe toward the stability, reaction mechanism, and activity trend of single atomic catalysts (SACs) for NO3RR is proposed by integrating multidimensional insights from coordination environment, thermodynamic and electrochemical stability, Gibbs free energy profiles, electronic properties, and activity descriptors. It is revealed that the NO3RR performance of SACs is highly correlated with their local environments. Furthermore, the hybridization between the TM‐3d orbitals and 2π* orbitals of NO3
− gives rise to the formation of d‐π* orbitals, thus promoting the NO3
− activation. A symbolic regression is designed to capture the hidden descriptors of NO3RR, outperforming than using single adsorbate or electronic descriptors for hyper dimensional system entailing large geometric variability.

A high‐throughput in silico calculations for NO3RR based on transition metal doped and nitrogen decorated biphenylene network are performed. A symbolic regression is designed to capture the hidden descriptors of NO3RR for hyper dimensional system entailing large geometric variability. The study sheds light on a new design principle for developing SACs supported on carbon materials with excellent performance toward NO3RR.