Virtual Database Construction and Machine-Learning-Assisted High-Throughput Evaluation of Amorphous Porous Carbon...

Area of Science:

• Materials Science
• Computational Chemistry
• Nuclear Engineering

Background:

• Effective iodine-125 (I2) sorption is critical for nuclear industry applications.
• Amorphous porous carbon (APC) materials show promise as I2 sorbents.
• High-throughput screening methods are needed to identify optimal APC structures.

Purpose of the Study:

• To develop a comprehensive approach for high-throughput screening and analysis of APC materials for I2 sorption.
• To establish a virtual database of APC models and evaluate their I2 adsorption capacities.
• To identify key structural and chemical features influencing APC performance using machine learning.

Main Methods:

• Generation of a virtual database of 19,599 APC models using liquid quenching molecular dynamics simulations.
• Large-scale grand canonical Monte Carlo simulations to determine I2 adsorption capacities at various concentrations.
• Machine learning and SHapley Additive exPlanations (SHAP) analysis to correlate material features with adsorption behavior.

Main Results:

• Identified influential factors for APC development, including surface area and pore size distribution, which vary with I2 concentration.
• Generated an array of I2 adsorption capacities for sampled APCs.
• Established a framework for understanding structure-property relationships in APC materials.

Conclusions:

• The developed approach accelerates the screening and development of APC materials for I2 sorption.
• Fundamental databases and research frameworks are provided to enhance understanding of APC materials for nuclear applications.
• Key material features influencing I2 adsorption were identified, guiding future material design.