Machine-Learned Interatomic Potential Developed for Tungsten-Boron Surface Sputtering via tabGAP Framework, Guiding MXene Defect Engineering

August 23, 2026

arXiv International

Overview

A machine-learned interatomic potential (MLIP) has been developed using the tabGAP framework to simulate sputtering behavior of tungsten-boron (W-B) surfaces at large scales. Trained with density functional theory (DFT) data, this MLIP maintains DFT-level accuracy while significantly reducing computational costs for high-throughput sputtering and deposition simulations. This technology specifically guides defect engineering in MXenes, a promising class of 2D materials, thereby enhancing their functional properties and accelerating their application development.

In Depth

Key Findings

A significant advancement in computational materials science has been achieved with the development of a machine-learned interatomic potential (MLIP) for simulating the sputtering behavior of tungsten-boron (W-B) surfaces. Utilizing the tabGAP framework, this MLIP is trained on extensive density functional theory (DFT) data, enabling high-throughput and large-scale simulations of sputtering and deposition processes with DFT-comparable accuracy at substantially lower computational costs.

Technical / Clinical Details

The developed MLIP is specifically optimized to model atomic interactions and phenomena at material surfaces, such as sputtering and film deposition. While traditional ab initio methods like DFT offer high accuracy, their computational expense has limited their application to smaller systems and shorter simulation times. This MLIP overcomes these limitations by efficiently learning the intricate physical interactions from DFT-generated data. It accurately represents the complex interatomic forces between tungsten and boron atoms, enabling precise predictions of sputtering yields, deposition dynamics, and surface structural evolution. Crucially, this potential provides critical guidance for defect engineering in MXene materials, paving new avenues for controlling and optimizing their electrical, mechanical, and chemical properties for advanced applications.

Background & Context

Tungsten and boron-containing materials are vital for various advanced applications, including plasma-facing components in fusion reactors, high-performance coatings, and advanced electronic devices. Understanding their surface behavior, particularly damage from ion irradiation and deposition processes, is paramount as it directly impacts device performance and longevity. MXenes, with their unique two-dimensional structures and exceptional properties, are emerging as promising materials for energy storage, catalysis, and sensors. The integration of MLIPs accelerates R&D for these materials, potentially revolutionizing the design and manufacturing processes for high-performance materials in industry.

Strategic Significance & Outlook

The development of this MLIP represents a crucial step forward in computational materials science, with future applications expected to extend to more complex multi-element systems and heterogeneous interfaces. Its ability to provide specific guidance for MXene defect engineering significantly advances their practical implementation. This synergy between machine learning and atomic simulations further enhances the ‘inverse design’ approach in materials science, where desired properties dictate material structure, contributing to the overall progression of the materials informatics field.

Source: https://arxiv.org/html/2608.13038v1

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.