Bulletin of the American Physical Society
2018 Annual Meeting of the APS Four Corners Section
Volume 63, Number 16
Friday–Saturday, October 12–13, 2018; University of Utah, Salt Lake City, Utah
Session E04: Computational Physics 2 |
Hide Abstracts |
Chair: Ryan Wixom, Sandia National Lab Room: CSC 10/12 |
Friday, October 12, 2018 1:30PM - 1:54PM |
E04.00001: Modeling materials degradation in nanostructures: An example of the intersection of computational material science, experimental characterization Invited Speaker: Remi Dingreville Nanoscience is highly interdisciplinary; greater insight into (nanostructured) materials-properties relationships requires the development of multiscale, multiphysics models and comparably advanced experimental design, instruments, and analysis. Such expanding scope of research motivates synergies between the data analytics, experimental and computational communities. |
Friday, October 12, 2018 1:54PM - 2:06PM |
E04.00002: Restricted Boltzmann Machines for Learning Multiple Observables Parker Hamilton, Chandramouli Nyshadham, Gus L.W. Hart A common use of machine learning in materials research is to use existing experimental and computational data to train a model that predicts one property of the system, like thermal stability. We explore a machine learning method using restricted Boltzmann machines that can be used to calculate multiple physical observables. Our data set consists of a 2D Ising model of spins generated in a Monte Carlo simulation. We will use this model to generate spin states from which we calculate physical observables and compare them to the spin states that were generated by Monte Carlo methods. |
Friday, October 12, 2018 2:06PM - 2:18PM |
E04.00003: Investigating superalloys using machine learning Hayden Oliver, Chandramouli Nyshadham, Gus L.W. Hart In 2006, Sato et al., reported a new cobalt-based superalloy (Al-Co-W) that demonstrated superior mechanical properties compared to conventional nickel-based superalloys. We explore this ternary alloy system in detail to understand the phase stability of the Co-alloy. We perform this study using a class of systematically improvable potentials called Moment Tensor Potentials (MTP) which is a machine learning interatomic potential for approximating quantum-mechanical energies. We detail this process on the aforementioned Al-Co-W system and show how it can be broadly applied to any binary or ternary alloy systems. |
Friday, October 12, 2018 2:18PM - 2:30PM |
E04.00004: Modeling Grain Boundaries using Scattering Transformations for use in Machine Learning Derek Hensley, Gus L.W. Hart, Conrad W Rosenbrock Grain boundaries of crystalline materials have a major impact upon their physical properties. Therefore the key to truly understanding materials is to understand grain boundaries. Only within the last year has the materials community successfully been able to model grain boundaries and predict material properties by adopting techniques from machine learning. Due to the sparseness of grain-boundary data, the representation (i.e., how the data is represented for the machine), of the grain boundary is critical to successfully modeling grain boundaries. I will discuss a novel grain boundary representation, the Scattering Transformation, and use it to understand grain boundary properties such as energy, mobility, and shear coupling. |
Friday, October 12, 2018 2:30PM - 2:42PM |
E04.00005: Machine Learning Directed Search for Ultraincompressible, Superhard Materials Marcus Parry, Aria Mansouri Tehrani, Anton O. Oliynyk, Zeshan Rizvi, Samantha Couper, Feng Lin, Lowell Miyagi, Jakoah Brgoch, Boris Kiefer, Taylor D. Sparks Currently, superhard materials with widespread commercial application (i.e. diamond, c-BN) require extreme temperatures and pressures to produce. Of interest is a separate class of superhard materials which combine transition metals with light main group elements and can be synthesized via common high-temperature metallurgical techniques. To expedite the discovery process, a machine-learning (ML) model is developed to predict bulk and shear moduli; mechanical properties which scale with hardness. From the model, a rhenium tungsten carbide and molybdenum tungsten borocarbide are selected and synthesized at ambient pressure via arc melting. Bulk modulus of each compound is determined experimentally through high-pressure diamond anvil cell measurements, supporting the ML predictions with less than 10% error. Vickers hardness is measured, indicating each composition surpasses the superhard threshold of HV = 40 GPa at low loads (0.49 N). Furthermore, DFT calculations are employed on compositions of intermediate predicted hardness to corroborate the ML model. These results demonstrate the promise of machine-learning techniques for the identification of novel materials with desirable mechanical properties. |
Friday, October 12, 2018 2:42PM - 2:54PM |
E04.00006: The best features to know when you first meet a material Kennedy Lincoln, Chandramouli Nyshadham, Gus L.W. Hart A new focus in material science pairs computational material databases with machine learning (ML) techniques to predict new materials. One of the key challenges in predicting properties of materials, such as formation enthalpy, is identifying important features to include in the ML model. We aim to build an ML model capable of reproducing the formation enthalpy of all materials in the AFLOWlib repository. In regards to this, we analyze the important input features that enable successful learning. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700