Autonomous materials discovery models and algorithms for the AFLOW framework and database

Abstract

Several new descriptors, algorithms, and online applications are proposed all working within the AFLOW Framework for Materials Discovery to investigate and predict materials synthesizability. Thermodynamic analysis tools catering to disorder will be developed: i. online visualization of convex hull descriptors for multi-principle component alloys, ii. Implementation within AFLOW of the new coordination corrected enthalpies (CCE) method and the Lederer-Toher-Vecchio-Curtarolo (LTVC) model for order-disorder transitions, along with corresponding online applications, iii. online module for pressure-temperature phase diagrams, and iv. workflow for calculation of vibrational properties of o-stoichiometric systems. New symmetry and crystal-matching functionality will reduce the cost and number of calculations needed for the analyses, while also identifying existing database entries that may be recycled for these purposes. Several database and online applications will be created as well: i. on-the-fly online machine learning application, ii. interactive visualizations of phonon band structures, thermal conductivity versus temperature plots, Hugoniot shock compression curves, and iii. a new balanced AFLOW.org repository tree infrastructure. Finally, the visualizations and online tools are the focus of future AFLOW School workshops, through which we train the next cadre of DoD researchers.

Document Details

Document Type
DoD Grant Award
Publication Date
May 08, 2020
Source ID
N000142012525

Entities

People

  • Stefano Curtarolo

Organizations

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

Tags

Readers

  • Distributed Systems and Data Platform Development
  • Materials Science and Engineering.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML