Validation: Coevolved Experimentation and Computation at the Nanoscale

Summary

The increased speed and increased size of datasets implemented in modern computational systems has enabled the size scale of models of materials to be increased significantly. In parallel, the increased precision of modern experimental systems has enabled the size scale of measurements of properties of materials to be decreased significantly. Hence, at the nanoscale it is now possible to make direct comparison of experimentally determined and computationally predicted properties for the same structure. The goal of the co-evolved experimentation and computation projects is to provide validation of computational models of materials properties by direct comparison of predicted and measured properties at the same length scale.

(a) and (b) Finite element model of a tungsten (grey) electrical connector extending through silicon, forming a "through silicon via" for advanced semiconductor structures. The contours represent the stress field generated in the silicon on via fabrication. (c) The measured Raman light scattering shift (symbols) measured in the silicon adjacent to the via compared with the shift computed from the stress field. The agreement between the measurement and simulation validates the stress field model and provides confidence that predictions of other stress fields will be accurate. The through silicon via is an example of a micro-scale structure useful for coevolved experimentation and computation. See http://link.aip.org/link/doi/10.1063/1.3644971 for more details.

Description

A central tenet of the Materials Genome Initiative is that advanced computational methods will enable materials with new or superior properties to be discovered so as to improve the performance of components in applications ranging from communications to energy to health care. Such methods will take advantage of the reduced cost, increased speed, and increased data handling ability of modern computational hardware and software to implement models of materials, and in particular to generate models of material structures that can be interrogated to predict material properties. The success of such a scheme relies on the validity of the computational models. The goal of the co-evolved experimentation and computation projects is to provide validation of computational models of materials properties by direct comparison of predicted and measured properties at the same length scale.

The increased speed and increased size of datasets implemented in modern computational systems has enabled the size scale of models of materials to be increased significantly. For example, molecular dynamics simulations can be performed using many millions of virtual atoms to describe and predict the properties of structures many tens of nanometers in scale. In parallel, the increased precision of modern experimental systems has enabled the size scale of measurements of properties of materials to be decreased significantly. For example, scanning probe microscope-based measurements can be performed to measure and map mechanical, thermal, and electrical properties with sub-nanometer resolution. Hence, at the nanoscale it is now possible to make direct comparison of experimentally determined and computationally predicted properties for the same structure. Examples of such structures include nanoparticles with unique chemical properties for biochemical applications, thin films with superior mechanical properties for nanoelectromechanical sensor and actuator applications, and lithographically-defined structures with novel quantum mechanical properties for nanoelectronic applications.

A key element of the model validation process is the deliberate co-evolution of the experimental and computational aspects to advance the Material Genome Initiative goals. For example, a nano-scale measurement might reveal a variation in a property with the shape of a structure. Computational research could then be used to perform atomistic simulations of structures of various shapes to provide guidance on the underlying cause of the property variation. If the agreement between experiment and simulation is very good, structures with new shapes could then be simulated with confidence to search for structures with new or enhanced properties for new applications. This conventional approach is extended here to include the development of new experimental techniques that deliberately seek to verify simulation predictions. These experimental techniques provide both validation of models and key values of quantities required for the simulations. A critical element to this approach is the deliberate a priori selection of nano-scale systems and structures that are amenable to both experimental and computational investigation in a direct and absolute manner. This last point is an essential part of the co-evolution: Both the experimental measurement and simulated prediction must also include an assessment of accuracy such that absolute values of properties can be predicted, not simply trends.

Additional Technical Details

Specific projects that seek to validate models through co-evolved experimentation and computation on nano-scale structures include (in order of increasing length scale):

• Defect Structures and Electronic Properties of Graphene—Density Functional Theory simulations of defect structures and electronic properties of graphene for direct comparison with scanning tunneling microscope measurements of electronic properties
• Development of atomic-scale models for determination of local structure of nanostructures from combined X-ray, electron, and neutron diffraction techniques
• Concurrent Experiments and Simulations of Zinc Oxide Nanowires—Molecular Dynamics simulations of structure and properties of zinc oxide nanowires for direct comparison with atomic force microscope measurements of elastic properties
• Electron Backscatter Diffraction measurements of structure and strain of barium titanate for direct comparison with continuum models of strain
• Strain Mapping and Simulation—Finite Element simulations of deformation, strain, and stress states of silicon and silicon-germanium thin-film structures for direct comparison with electron backscatter diffraction, X-ray diffraction, confocal Raman microscopy, and atomic force microscopy measurements of deformation, strain, and stress.
• Finite element and analytical mechanics models of buckled metallic lines on low dielectric constant thin films for direct comparison with atomic force microscopy measurements of buckling geometry and material properties
• Mapping of Strain Fields and Phase Distributions of Indented Si using Raman-Spectroscopy Enhanced Instrumented Indentation Technique (IIT)—In Situ Raman spectroscopy measurements of strain states in indented silicon for direct comparison with atomic-scale predictions of Raman shift with strain
• Machine Learning for High Throughput Materials Discovery and Optimization Applications—Development of search algorithms to mine large experimental data sets to identify phase and other cluster diagrams with a focus on electronic materials