Session P32: Put Big Data in Your Physics Toolbox; APS-AIP Industrial Physics Forum

Improving Electron Microscopy with Artificial Intelligence and Big Data

In this presentation, we will describe a vision for a future research paradigm, wherein a tight coupling of in-situ and operando experimental methods, data analytics and automated data analysis are coupled with artificial intelligence to direct how we use electron microscopy to characterize the mechanisms by which processing/structure/property relationships are determined. The presentation will be forward looking, and will incorporate research results and ideas culled from a variety of sources and authors. First, we will describe the motivations for working towards this type of research paradigm. These include the desire to speed up the rate of scientific discovery and time to market, as well as a more pedestrian desire to maximize the utilization of expensive instrument time. Second, we will review examples of autonomous research methods, both through data mining of the literature [1], and through the use of real time feedback and artificial intelligence methods to direct experimental outcomes.[2] Specifically, we will discuss how these approaches may be utilized in electron microscopy research in the near future, and the developments needed to bring this to reality. Third, we will describe how this approach can be used to explicitly and efficiently test operative hypotheses, and to efficiently understand the relevant experimental parameter space. This yields insight as to where detailed experimentation is most valuable. In this portion of the talk, the need for operando methods and correlative experimentation will be emphasized.[3]

1. Science Mapping: A Systematic Review of the Literature, Chaomei Chen, Journal of Data and Information Science, 2, 1-40, 2017
2. Autonomy in materials research: a case study in carbon nanotube growth; P. Nikolaev, et al., npj Computational Materials, 2, 16031 (2016)
3. Complex structural dynamics of nanocatalysts revealed in Operando conditions by correlated imaging and spectroscopy probes, Y. Li, et al., Nature Comm., 6, 7583
2015.
  

Quantum Computing at D-Wave

This abstract not available.   

Polymer Discovery Using Big Data and Analytics

Novel materials can transform industries and societies. Approximately 1000 materials science articles are published every day, a number far beyond the human bandwidth to read. At IBM Research -- Almaden, we are exploring AI technologies to help human researchers collect, extract, organize, and explore this wealth of materials science data. In this talk we will focus on polymeric materials, explore the polymeric materials data challenge, and share some of the data science strategies we have used at IBM to speed polymer R&D.   

Combinatorial Experimentation and Machine Learning for Materials Discovery

Throughout the history of mankind, scientists and engineers have relied on the slow and serendipitous trial-and-error approach for materials discovery. In 1990s, the combinatorial approach was pioneered in the pharmaceutical industry in order to dramatically increase the rate at which new medical compounds are identified. The high-throughput concept is now widely implemented in a variety of fields in materials science. We have developed combinatorial thin film synthesis and characterization techniques in order to perform rapid survey of previously unexplored materials phase space in search of new inorganic functional materials with enhanced physical properties. Over the years, the challenges in the high-throughput approach has evolved from synthesis of large number of disparate compounds to developing quantitatively accurate rapid characterization tools to analysis and digestion of large amount of data churned out by the methodology. To address the last challenge, we are increasingly relying on machine learning techniques including pattern recognition within diffraction data to construct phase diagrams and mining experimental databases to look for trends in materials properties for future predictions. I will also discuss our latest effort where active learning is used to design and steer the sequence of experiments in order to maximize attainable knowledge, minimize experimental resources, and as a result further speed up the materials discovery process. This work is performed in collaboration with A. Gilad Kusne, V. Stanev, A. Mehta, B. DeCost, J. Hattrick-Simpers, and Y. Liang.
  

Making Big Data Work for Physicists

Learn how MATLAB’s latest big data capabilities can be applied to your research, and how other scientists have used them to surmount the challenges of handling and interpreting enormous amounts of raw information, while scaling from desktops to clusters using the same MATLAB code.