Accelerating Density Functional Theory Simulations Via Machine Learning: The example of Stress-Corrosion Cracking in Metals
Abstract
New algorithms are developed that speed up molecular dynamics simulations and ensemble-averaged property predictions using atomistic quantum calculations. ABSTRACTThe primary goal of the present proposal is to utilize machine learning based force fields to stu"dy the mechanical behavior of a variety of elemental metals experiencing high stresses, high temperatures and corrosive environments"", in other words, fundamental studies of stress corrosion cracking at time and length scales relevant to this phenomenon. Specifical""ly, Al (an FCC metal), Ti (a HCP metal) and W (a BCC metal) will be studied. An intentionally created crack (i.e., a notch) will be"" placed in each material, which will be subjected to notch-opening stresses at various temperatures, in order to understand the prim""ary differences between the three chosen metals. Subsequently, oxygen will be introduced in the notch to probe the role of corrosion"" on the crack propagation behavior. In order to accomplish these goals, several extensions of this forcefield concept proposed will"" be necessary, including methodological developments to handle multiple elements, strategies to optimally choose the initial trainin""g set, and strategies to recognize a new environment when such is encountered during the course of a simulation. It is believed that"" this work will lead to a fundamental understanding of stress-corrosion cracking, a phenomenon that has enormous aerospace and navy"" relevance, especially in materials such as Ti which has significant technological and application relevance.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 23, 2018
- Source ID
- N000141812113
Entities
People
- Ramamurthy Ramprasad
Organizations
- Georgia Tech Research Corporation
- Office of Naval Research
- United States Navy