Multifunctional electrocatalysis on transition metal-doped biphenylene: a first-principles and machine learning study of single-atom catalysts for HER, OER, and ORR

Abstract

Single-atom catalysts (SACs) anchored on two-dimensional substrates have emerged as a frontier in electrocatalysis due to their atomic-level efficiency and tunable reactivity. Herein, we present a comprehensive theoretical investigation of 3d, 4d, and 5d transition metal (TM) atoms embedded in single-vacancy biphenylene (SV-BPN), a recently synthesized carbon allotrope with a unique nonbenzenoid topology. By integrating density functional theory (DFT) and machine learning (ML), we evaluated the structural stability, electronic characteristics, and catalytic activity of TM@SV-BPN systems toward the hydrogen evolution reaction (HER), oxygen evolution reaction (OER), and oxygen reduction reaction (ORR). Our screening reveals several high-performing SACs with low overpotentials, including Mo@SV-BPN (η^HER = 0.006 V), Pd@SV-BPN (η^OER = 0.43 V), and Ag@SV-BPN (η^ORR = 0.67 V). Notably, Au@SV-BPN exhibits trifunctional catalytic activity across all three reactions. Electronic descriptors such as the d-band center and integrated crystal orbital Hamilton population (ICOHP) are correlated with intermediate adsorption energetics. A gradient boosting regression model accurately predicts adsorption energies (R² = 0.98), highlighting charge transfer as the most influential feature. This work not only identifies a novel class of trifunctional SACs but also establishes a data-driven paradigm for rational catalyst design based on biphenylene supports.