Application of High-Performance Graphics Supercomputers and Communication to Provide Improved Interfaces to Scanning Probe Microscopes

Principal Investigators: Russell M Taylor II and Frederick P. Brooks Jr.
Funding Agency: National Science Foundation
Agency Number: ACI-9527192

Abstract
This project will investigate the most powerful and efficient systems, methods and algorithms for applying the high-performance graphics supercomputer, PixelFlow, to scientific visualization. As a specific application, we will create a virtual-reality intuitive-feedback interface to commercial scanning probe microscopes (SPMs). The physicists on the team provide driving problems against which our visualization and control systems will be tested. These problems include the fabrication of nanometer-scale quantum effect devices and the investigation and manipulation of biological samples such as fruit-fly chromosomes and Tobacco Mosaic Virus (TMV) particles.

We have chosen the area of SPM control because of the known improvements that visualization systems can make in this area. Our prior work with the Pixel-Planes 5 graphics computer has shown that providing real-time, user-guided exploration of surface data provides insights into that data which are not available from conventional viewing techniques. [Taylor93] We have also shown that interactive control over the instrument during an experiment enables new, more useful types of experiments. [Taylor94] Most recently, we have shown that providing force feedback to the user during modification enables the precise positioning of colloidal gold particles, which cannot be done without such feedback. [Finch95]

Displaying samples scanned at (scientifically preferred) resolutions of 256x256 requires a 10-fold improvement in the polygon display capability. PixelFlow will be scalable, supporting configurations that offer more than 10 times the performance of today's systems. We propose to augment the PixelFlow system that will be in place at UNC to support this higher-resolution scanning.

In addition to pushing the upper scale of performance, we will pursue methods of providing availability of the visualization to scientists in their laboratories. The first approach to this is to leave the graphics engine in the computer science department and run high-speed communications to the SPM controller located in the scientist's laboratory. The second approach is to investigate a smaller-scale version of PixelFlow that could be afforded by a scientist and installed near the SPM itself.