CDI Type I: Collaborative Research: Cyber-Enabled Chemical Imaging: From Terascale Data to Chemical Imaging

University Of Texas At Dallas, Richardson TX

Investigators

Abstract

Chemical imaging is widely used in many areas of science and engineering, including chemistry, materials science, forensic science, medicine, art conservation, and archaeometry. A chemical image is a picture in which the colors indicate the position and concentration of different atoms and molecules. These images therefore provide the chemical composition of the surface of a sample, which is crucial information for understanding its history, behavior, and properties. Chemical images can be very complex. Biological tissue samples, forensic evidence, and other materials may be composed of hundreds or even thousands of different chemical compounds, in amounts varying over lengths as small as a few nanometers. Tools that can correctly identify and resolve these components and changes are highly desirable. In this project, new analysis tools and state-of-the-art computing resources will be used to greatly improve the resolution and quality of these chemical images. This improvement is possible because the great bulk of information obtained in chemical imaging techniques is normally not used. The "raw" data produced by high-resolution experiments (imaging mass spectrometry, infrared and Raman microscopy, scanning Auger microscopy, and x-ray photoelectron spectroscopy imaging) can exceed thousands of gigabytes per square millimeter of sample imaged, which up to now has been far too much for individual researchers to even store, let alone completely analyze. The team will provide software that can take full advantage of the power of modern supercomputers, such as those available through NSF's Teragrid, to extract statistically optimal chemical images from these enormous data sets. These tools will also be able to combine many small-area images into large-area chemical images of unprecedented resolution, enabling detailed chemical imaging of much larger samples than previously possible. This research's goal is to dramatically improve the power, applicability, and ease-of-use of a wide range of chemical imaging techniques, significantly advancing many different fields of science, engineering and medicine. This can only be accomplished through "computational thinking", because the data sets themselves are simply too large and too complex to be comprehensively analyzed by even expert human operators. The software, along with documentation and tutorials, will be freely distributed, and installed for general use at national supercomputing centers.