Machine learnt bond order potential to model metal-organic (Co-C) heterostructures

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Designing advanced functional materials requires understanding complex structure-property relationships in metallic-covalent systems.
• Developing accurate and efficient atomistic models is crucial for simulating these systems.

Purpose of the Study:

• To develop a unified bond order potential (BOP) for cobalt-carbon (Co-C) systems.
• To accurately predict the structural, thermodynamic, and mechanical properties of Co-C materials.

Main Methods:

• Trained a Tersoff-Brenner type BOP using supervised machine learning (genetic algorithms) and local minimization.
• Validated the BOP against density functional theory (DFT) calculations and experimental data.
• Employed large-scale molecular dynamics simulations and unsupervised machine learning (clustering) to analyze self-assembled Co-C nanostructures.

Main Results:

• The developed BOP accurately describes elemental cobalt and cobalt carbide phases.
• Simulations revealed that increasing carbon concentration in self-assembled structures limits cobalt nanoparticle connectivity and electrical conductivity.
• Higher carbon content leads to highly flexible (low stiffness) hybrid structures.

Conclusions:

• The machine-learnt BOP is a valuable tool for investigating Co-C systems at the atomic scale.
• This model facilitates the design of organic-inorganic hybrid structures with tailored properties for applications like flexible electronics.