Dr Amanda S. Barnard

  • Office of the Chief Executive Science Leader
    Commonwealth Scientific and Industrial Research Organisation
    Email: Amanda.barnard@data61.csiro.au


Dr. Amanda Barnard is an Office of the Chief Executive Science Leader within Data61 at CSIRO. She received her Ph.D. (Physics) in from RMIT in 2003, followed by a Distinguished Postdoctoral Fellow in the Center for Nanoscale Materials at Argonne National Laboratory (USA), and the prestigious senior research position as Violette & Samuel Glasstone Fellow at the University of Oxford (UK) with an Extraordinary Research Fellowship at The Queen’s College. She joined CSIRO as an Australian Research Council Queen Elizabeth II Fellow in 2009, and now as an OCE Science Leader she leads research developing structure/property relationships using computational physics and chemistry, machine learning, deep learning and AI. Dr Barnard is a member of the Nature Index Panel (Nature Publishing Group), and has previously served as an Associate Editor for Science Advances (AAAS). She is the Chair of the National Computational Merit Allocation Scheme for Australia (awarding $10 million in resources annually) and a Fellow of the Australian Institute of Physics. For her work she has won the 2009 Young Scientist Prize in Computational Physics from the International Union of Pure and Applied Physics, the 2009 Malcolm McIntosh Award from the Prime Minister of Australia for the Physical Scientist of the Year, the 2010 Frederick White Prize from the Australian Academy of Sciences, the 2014 ACS Nano Lectureship (Asia/Pacific) from the American Chemical Society, and the 2014 Feynman Prize in Nanotechnology (Theory) from the Foresight Institute, being the first woman to do so in the history of the award.


Capta-Driven Materials Design

Even though it is always there, dealing with the complexity of nanomaterials, the diversity of individual samples, and the persistent imperfection of individual structures has been secondary to our search for novel properties and promising applications. However, for our science to translate into technology we will inevitably need to deal with the issue of polydispersivity and integrate this feature into the next generation of more realistic structure/property predictions. Our predictions need a fault tolerance, but uncovering the underlying connection between specific structures and properties is difficult to do experimentally, particularly when many (if not all) of the important design parameters are cross-correlated. Fortunately the strategic use of computational methods and high performance computing can provide these insights, and so much more. A range of reliable statistical and data-driven methods, such as machine learning and deep learning, have become widely available to help us to take greater advantage computational data, or “capta”. Appropriate sampling of our enormous parameter spaces, judicious data cleaning and curation, and selection of the right learning models are essential elements of capta-driven materials design, and differ from approaches applied to experimental data. In this presentation we will explore the development of capta sets and the use of simple statistical and machine learning methods to predict properties and performance. We will also see how to predict structure/property relationships for entire samples of structures, and how we can investigate the impact of different manufacturing processes that restrict the polydispersivity in different ways.