Exploring thermodynamic stability of plutonium oxycarbide using a machine-learning scheme.

Area of Science:

• Materials Science
• Computational Chemistry
• Nuclear Engineering

Background:

• Plutonium oxycarbide is vital for nuclear fuel fabrication and understanding plutonium corrosion.
• Predicting the thermodynamic stability of PuOC_1- is essential for materials design.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting the thermodynamic stability of plutonium oxycarbide (PuOC_1-).
• To compare the performance of different ML schemes and structural descriptors for this application.

Main Methods:

• Density-functional theory (DFT) with Hubbard correction was used to generate training data.
• Four machine learning models were trained and evaluated using three structural descriptors.
• The accuracy of the optimal ML model was assessed by comparing predicted and DFT-calculated mixing energies and lattice parameters.

Main Results:

• The optimal ML model achieved high accuracy, with average errors of ~3 meV/atom for mixing energy and 0.003 Å for lattice parameter.
• The ML model successfully predicted the convex hull and identified several stable ordered atomic structures for PuOC_1-.
• Enhanced stability in ordered structures is linked to strong hybridization between Pu 5f/6d and C/O 2p orbitals, forming robust Pu-C and Pu-O bonds.

Conclusions:

• Machine learning provides a highly accurate and efficient method for predicting the thermodynamic stability of plutonium oxycarbide.
• The findings offer insights into the formation of stable plutonium oxycarbide phases and their bonding characteristics.
• This approach can accelerate the discovery and design of advanced nuclear materials.