Targeted search for ultra-hard carbo-nitrides using the AFLOW framework and database

Abstract

A high-memory, combined CPU and GPU cluster is requested for a targeted search of ultra-hard carbo-nitrides. The CPU cluster will be used for the calculation of i. the synthesizability of 5-metal carbide, nitride, and carbo-nitride systems, and ii. the mechanical properties of 5-metal carbides and nitrides for estimation of hardness and ductility. The supercell ensemble approach for modeling these disordered materials requires larger calculations having lower symmetry than the prototypical ones comprising the AFLOW.org repository warranting the high-memory environment. The GPU will be used to construct Property-Labelled-Materials-Fragments-based machine learning models that predict the formation and elastic properties of carbo-nitride systems, facilitating exploration of all transition-metal- composition combinations. Candidates predicted to have high synthesizability, hardness, and ductility will be prioritized for calculation and model retraining. The high-memory environment will also support rapid storage of and access to properties for machine learning model reconstruction as more data becomes available. Data will be formatted and stored in the AFLOW.org repository and programmatically accessible through the AFLUX Search-API.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 15, 2021
Source ID
N000142112132

Entities

People

  • Stefano Curtarolo

Organizations

  • Duke University
  • Office of Naval Research
  • United States Navy

Tags

Readers

  • Materials Science and Engineering.
  • Neural Network Machine Learning.
  • Research Science/Academic Research

Technology Areas

  • AI & ML