Characterization: Structural Descriptors, Data-Intensive Techniques, and Uncertainty Quantification: Grain Boundary Descriptors
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS Extraction and Processing Division, TMS: Computational Materials Science and Engineering Committee, TMS: Materials Characterization Committee
Program Organizers: Shawn Coleman, DEVCOM Army Research Laboratory; Tomoko Sano, U.S. Army Research Laboratory; James Hogan, University of Alberta; Srikanth Patala, SASA Institute; Oliver Johnson, Brigham Young University; Francesca Tavazza, National Institute of Standards and Technology
Tuesday 8:30 AM
February 25, 2020
Room: Theater A-3
Location: San Diego Convention Ctr
Session Chair: Oliver Johnson, Brigham Young University; Srikanth Patala, North Carolina State University
8:30 AM Introductory Comments
8:35 AM Invited
Characterizing GB Atomic Structures at Multiple Scales: Eric Homer1; Derek Hensley1; Conrad Rosenbrock1; Andrew Nguyen1; Jonathan Priedeman1; Gus Hart1; 1Brigham Young University
The atomic structure of grain boundaries plays a defining but poorly understood role in the properties they exhibit. Due to the complex nature of these structures, machine learning is a natural tool for extracting meaningful relationships and new physical insight. We present efforts to characterize the GB atomic structure at different length-scales using a few different methods. These include the smooth overlap of atomic positions (SOAP), local environment representation (LER) and a new structural representation, called the scattering transform. The scattering transform uses wavelet-based convolutional neural networks to characterize the complete three-dimensional atomic structure of a grain boundary. The success of the various metrics is evaluated by machine learning to predict GB energy, mobility, and shear coupling. A discussion of the advantages and disadvantages of the various methods is discussed.
9:05 AM
Basis Functions for Quantifying Grain Boundary Texture in Polycrystalline Microstructures: Srikanth Patala1; Jeremy Mason2; 1North Carolina State University; 2University of California, Davis
The statistical distributions of different grain boundary types play an important role in governing the mechanical and functional properties of polycrystalline materials. However, even for simple microstructures, the capability of representing the distributions of GB character, as a function of the five macroscopic degrees of freedom, has not been established. As the GB character distributions directly influence the interfacial network connectivity, developing a framework for quantifying the statistics in the five-parameter space is a crucial missing step in the inverse-design of interface-dominated phenomena in polycrystalline systems. In this talk, I will present symmetrized functions, using the familiar hyperspherical harmonics, for representing grain boundary texture in the complete five-parameter space. The basis functions will also allow for the quantification of interfacial statistics in experimental microstructures and the interpolation of structure-property relationships of grain boundaries.
9:25 AM
Microstructural Evolution Along Geodesics: Ian Chesser1; Toby Francis2; Marc DeGraef1; Elizabeth Holm1; 1Carnegie Mellon University; 2University of California Santa Barbara
We develop a method to visualize interface properties that accounts for anisotropy in the full 5-D space of macroscopic crystallographic parameters. This method leverages the recently developed octonion metric to define distances between grain boundaries. Multidimensional scaling is used to learn the structure of grain boundary space from a matrix of pairwise octonion distances computed from a list of N grain boundaries. A low dimensional representation of grain boundary space is proposed and its connectivity is found to be consistent with existing grain boundary literature. Grain boundary energies and mobilities computed from molecular dynamics simulations are visualized in 3D for a wide range of grain boundaries, including general boundaries. Energy and mobility are found to be smooth almost everywhere in the reduced grain boundary space. Geodesics in orientation and grain boundary space are used to characterize several microstructural evolution processes: shrinkage of a cylindrical grain and compression of a nano-composite material.
9:45 AM
The Grain Boundary Octonion: Metrics, Paths, and Fundamental Zones: Toby Francis1; Ian Chesser2; Saransh Singh3; Tresa Pollock1; Elizabeth Holm2; Marc De Graef2; 1University of California, Santa Barbara; 2Carnegie Mellon University; 3Lawrence Livermore National Laboratory
With the growing ability to experimentally engineer and determine the five macroscopic degrees of freedom of grain boundaries, it has become important to present statistics on grain boundary configurations in real materials. The lack of a consistent reference frame and fundamental zone given a misorientation and boundary normal makes collecting such statistics challenging. Thus, we present an alternative framework in which a grain boundary is identified by two rotations from a reference frame attached to the boundary plane. We develop an analytical approximation for the natural metric on this space, which defines the distance between two grain boundary configurations by the smallest rotations necessary to transform one configuration to the other. This framework allows us to clearly define symmetries and the shortest paths between boundaries. Examples of microstructural statistics derived from this metric are presented, and a set of open-source code packages is provided for ease of use.
10:05 AM Break
10:25 AM
GB Property Localization: Inference and Uncertainty Quantification of Grain Boundary Structure-property Models: Oliver Johnson1; Brandon Snow1; Sterling Baird1; Christian Kurniawan2; David Fullwood1; Eric Homer1; 1Brigham Young University; 2Carnegie Mellon University
We have been developing a method for inferring grain boundary (GB) structure-property models from measurements (and/or simulations) made on polycrystals, which we call GB Property Localization. One of the major challenges of developing quantitative structure-property models for GBs is that the quantity of available data is small compared to the size of the GB character space, so that the problem is almost always severely underdetermined (i.e. there are far fewer measurements than the number of discrete “bins” in any reasonable discretization of the space). In this talk we present a new formulation of the GB Property Localization procedure that solves the problem of indeterminacy and enables the inference of continuous functions (as opposed to discretized approximations). We describe the probabilistic framework on which GB Property Localization is built, and how it naturally facilitates Uncertainty Quantification (UQ) for the resulting constitutive models.
10:45 AM
Higher Order Spectral Terms in Grain Boundary Networks: Christopher Adair1; Oliver Johnson1; 1Brigham Young University
Grain boundary networks (GBNs) create a highly complex structure in crystalline materials that significantly influence their macroscopic properties. One recent method used to characterize these complex networks is Spectral Graph Theory. In this approach, GBNs are represented by a matrix of nodes and weighted edges, the eigenvalues and eigenvectors of which give a unique description of the GBN configuration. Using a single eigenvalue and eigenvector we identified the dominant microstructural features that govern macroscopic transport properties in polycrystals. In this talk we show how higher order terms provide additional information that can be used to identify and rank the contributions of a variety of microstructural features and their contributions to the macroscopic material properties.
11:05 AM
Investigating the Atomistic Nature of Grain Boundary Failure: Jacob Tavenner1; Christopher Weinberger2; Shawn Coleman3; Garritt Tucker1; 1Colorado School of Mines; 2Colorado State University; 3Army Research Laboratory
Grain Boundaries (GBs) (i.e., the interfaces between misoriented crystalline grains) are a significant factor in determining the microstructural response of a material. Analyzing GB response during critical flaw formation, such as during incipient decohesion, allows for development of key understandings of the propensity of specific boundaries to undergo various failure modes. Although unique decohesion behaviors are identified across the sampled GB set, no clear correlations with commonly used GB metrics exist. Under induced decohesion, GBs vary widely in both their propensity for failure and the underlying fracture mode, indicating that the inhomogeneities specific to individual GBs, which are not captured by current GB descriptor techniques, drive the propensity for the failure of individual boundaries. Since most GB descriptors do not capture atomic information, the problem of transmitting valuable data between these length scales is understandable, as failure modes or mechanisms are driven by critical flaws in the material.
11:25 AM
Characterizing the Energetics and Structural Configurations of Silicon Carbide Grain Boundaries Using High-throughput Atomistic Techniques: Matthew Guziewski1; Dennis Trujillo2; Srikanth Patala3; Shawn Coleman1; 1US Army Research Laboratory; 2University of Connecticut; 3North Carolina State University
High throughput atomistic models are used to probe the uncertainty in the energetics and structure of silicon carbide at grain boundaries. In this work, a Metropolis style Monte Carlo algorithm is applied to a variety of tilt, twist and mixed character grain boundaries. This approach seeks to capture the variance in structure and energy beyond simply the five macroscopic degrees of freedom that describe the boundary geometry, though the consideration of factors such as metastable states, non-stoichiometric interfaces, and chemical environment near the interface. This allows for the analysis of not just pristine interfaces, but also interfaces that more closely resemble those which are seen experimentally. The observed distributions of energy are used to construct simple relations that relate the energetics to both misorientation/disorientation and to the expected metastable states, and allow for the construction of higher fidelity mesoscale models.