ITR: Computational Tools for Multicomponent Materials Design (07/01/2002 – 06/30/2007) Long-Qing Chen Department of Materials Science and Engineering Penn State University
Project Team Members PI: Zikui Liu (Mater. Sci. & Eng., Penn State) Co-PI: Long-Qing Chen (Mater. Sci. & Eng, Penn State) Padma Raghavan (Comp. Sci., Penn State) Qiang Du (Math., Penn State) Collaborators: Steve Langer (Computational physicist in the Div. of Math. and Computational Sci. in the Information Tech. Lab at NIST) Chris Wolverton (Computational physicist at Ford Motor Company)
Research Interests of Individual Team Members Zikui Liu – computational thermodynamics and kinetics, phase equilibria, phase transformations, and database development. Long-Qing Chen – computational phase transformations and microstructure evolution using microscopic and continuum phase-field approaches. Padma Raghavan – parallel scientific computing, development and application of algorithms and software for scalable sparse solvers, and software components and architectures for computational science and engineering applications on heterogeneous, high-performance computing environments.
Qiang Du – development of innovative numerical algorithms and their applications to problems in various scientific disciplines. Christopher Wolverton – theoretical/computational studies of thermodynamics and phase transformations via first-principles electronic structure calculations and statistical thermodynamics. Stephen Langer – the principal author of OOF, a program that uses image (microstructure) data and finite element analysis to predict materials responses under applied fields and was named one of the top 25 Technologies of the Year by Industry Week magazine in December 1999.
Project Objective Develop a set of computational tools for predicting the relationships among the chemistry, microstructures and mechanical properties of multicomponent material systems Processing Structure Properties
Computational Materials Science Tasks (1) Determine thermodynamic properties, lattice parameters, and kinetic data of unary, binary, and ternary alloys using first- principles calculations, linear response theory, cluster expansions, and Monte-Carlo (Wolverton) (2) CALPHAD data optimization to extract thermodynamic properties, lattice parameters, and kinetic data of multicomponent systems combining results from the first- principle calculations and experimental data (Liu) (3) Develop a multicomponent phase-field model to predict microstructures (Chen) (4) Develop a 3D Object-Oriented Finite element analysis tool to predict the materials responses from the simulated microstructure under applied fields (Langer).
Computational Thermodynamics CALPHAD Approach Thermodynamic models Structure of phases Unary and binary experimentally Model parameters for measured thermodynamic unary and binary systems properties and phase equilibrium data in the literature Prediction/evaluation for Theoretically calculated ternary systems thermodynamic properties Prediction for multi- Ternary experimental data in the component systems: literature and well designed new phase diagrams, isopleth, experiments phase fractions, Scheil diagram, thermodynamic data
Phase-Field Modeling of Microstructure Evolution Thermodynamic and kinetic parameters Input or generate initial microstructure Calculate driving forces Microstructure & Integrate microstructure statistics output evolution equations
Object Oriented Finite Element Software for Materials Scientists A collaboration between NIST’s Information Technology Laboratory and its Center for Theoretical and Computational Materials Science. Steve Langer NIST ITL Andrew Reid NIST CTCMS/MIT Andrew Roosen NIST CTCMS Edwin Garcia MIT Materials Science Ed Fuller NIST Materials Science Seung-Ill Haan U. Md, Baltimore County. Mech. Eng. Craig Carter MIT Materials Science
OOF1 Uses real microstructural geometry to construct finite element meshes. Uses microscopic material properties on the meshes to compute effective macroscopic behavior via virtual experiments. Linear elasticity and thermal conductivity. Simple models of fracture. Geometry can come from images or simulations ( eg , phase field). Material properties can be measured or simulated ( eg , first principles calculations, CALPHAD).
OOF2 More powerful and flexible. Designed for easy addition of new materials, physics, and types of finite elements. Linear and non-linear systems. Elasticity, plasticity, thermal & electrical conductivity, chemical diffusion, piezoelectricity, ferroelectricity, etc . Time dependence. Automatic mesh refinement. C++ modules in a Python framework for easy interoperability. Currently under development at NIST. OOF2 will be the basis for OOF3D.
First principles calculations First principles calculations and experiments and experiments Thermodynamic data of unary, Thermodynamic data of unary, Lattice parameters and Lattice parameters and Kinetic data of unary, Kinetic data of unary, binary and ternary systems binary and ternary systems interphase boundary energy interphase boundary energy binary and ternary systems binary and ternary systems CALPHAD approach CALPHAD approach to data optimization to data optimization Thermodynamic database for Thermodynamic database for Database for lattice parameters, elastic Database for lattice parameters, elastic Kinetic database for Kinetic database for multicomponent systems multicomponent systems constants, and interfacial energies constants, and interfacial energies multicomponent systems multicomponent systems A multicomponent A multicomponent phase-field model phase-field model Simulated microstructure Simulated microstructure Elastic constants of Elastic constants of in 1, 2, and 3 dimensions in 1, 2, and 3 dimensions individual phases individual phases OOF: Object-oriented finite element OOF: Object-oriented finite element analysis of material microstructures analysis of material microstructures Mechanical response of Mechanical response of simulated microstructure simulated microstructure Figure 1: An integrated set of computational tools for multicomponent materials design .
Advances in algorithmic design • Coupling of spectral methods with FEM • Adaptive computation in space and time • High order stable schemes • Parallelization, multi-scale resolution • Scalable domain specific software design
IT Research and Development Areas Developing scalable algorithms for sparse linear system solution. A component-based software approach to couple SPMD (Single Program Multiple Data) MPI-based packages Globus grid-services based model to provide complete materials-modeling environment through wide-area asynchronous interconnection of components and databases. Interactive client-server services for design parameter selection, visualization, and computation; servers can be parallel SPMD components. In collaboration with I. Foster's Globus group, B. Smith's PETSc group and L. Freitag's component architecture group at Argonne National Labs.
A schematic chart of software architecture for distributed multicomponent materials design
Education Components • Support 9 graduate students and 4 postdoctors • Integrate with the current NSF supported education program in thermodynamics and kinetics of materials (DMR-0073836) including summer short courses in computational thermodynamics and kinetics of materials • High Performance Computing Graduate Minor – 15 credits in designated computational science and engineering courses Examples of courses o Computational Thermodynamics o Computational Materials Science I: At the Atomic Scale o Computational Materials Science II: At the Meso/Continuum Scale o System Materials Design o Component Based Software Design for Engineering Applications o Numerical linear algebra o Numerical solution of ordinary and partial differential equations o Numerical optimization techniques o Finite element methods
Recommend
More recommend
