The physics of "human" scale systems far from equilibrium is vital to our every day lives, but is not thoroughly understood. Much of physics is historically focused on systems in equilibrium where the math has clean analytic solutions, but these situations represent only a subset of the interesting useful phenomena we encounter in the world around us.

There are still many human scale physics problems that until now have been inaccessible to science due to their small size or high speed. More of these problems are becoming accessible as instrumentation takes full advantage of technological advances. The improvement in instrumentation has brought with it the problem of dealing with massive data sets, true staggering amounts of information can be generated very easily. To understand the physics it is necessary to develop new and novel methods of data analysis and handling.

The nature of glassy and amorphous solid's present one of the last unsolved problems in condensed matter physics. As a material is quenched from the liquid to glass state the atomic structure does not change. Despite no structural change, the macroscopic properties change drastically. The viscosity diverges and the material will support a shear stress. This is contrasted to crystallization where both the changes in macroscopic properties are accompanied by major microscopic changes. Understanding the liquid-glass phase transition will significantly expand our understanding of condensed matter systems. I study glass systems by using a model system made from thermosensitive colloids observed using a high speed confocal microscope.

Confocal microscopes can generate an amazing amount of raw data quickly, however interpretation of the images requires that individual particles be identified and tracked. While this is fairly easy to do by eye for a few particles, but quickly becomes intractable for large numbers of particles. The algorithms for finding and tracking particles were developed by Crocker and Greir and have been widely used by the community.

There are a number of implementations of the algorithm in IDL and MATLAB, however the execution time of these implementations becomes unreasonably long for extremely large data sets. Building on the work of Peter Lu, who reimplemented the particle location algorithm in c++ ( sourceforge project), I have been working on reimplementing the single particle tracking algorithm in c++. Currently the code has minimal functionality, but a framework exists to move almost all analysis from MATLAB or IDL in to c++ which will result in significant speed improvements. Simplistic testing suggests the c++ code is approximately 8-10 times faster than Dan Blair's MATLAB code for identical data and parameters. Much of the execution time is building and then destroying the internal data structure, as more analysis is done in the frame work the importance of this overhead will decrease. Documentation is available at http://jfi.uchicago.edu/~tcaswell/track_doc and the code is available via subversion from svn://innoue.uchicago.edu/tracking/trunk. If you are interested in using this code please feel free to contact me.

At Cornell I worked for Professor Sol Gruner with the detector development group on Pixel Array Detectors (PADs), a class of high-speed direct detection x-ray detectors. I primarily worked on the control systems and experimental applications of PADs, participating in several experiments conducted at the Cornell High Energy Synchrotron Source (CHESS) and adapting a PAD for use in an electron microscope.

In the summer of 2005, I helped prepare for and execute a two week experimental run at CHESS. In collaboration with Jin Wang's group from Argonne National Lab and Visteon Corporation we conducted two radiography experiments. The first experiment was time resolved tomography of the spray cone of a low pressure injection nozzle for gasoline engines. In the second experiment, we studied shock waves from fuel injection in a simulated diesel engine.

For my undergraduate senior thesis I finished adapting a 16x16 PAD for use in scanning transmission electron microscopy (STEM) in collaboration with Peter Ercius from Professor David Muller's group. In STEM, a small electron probe is rastered across the surface of a sample and a full diffraction pattern is formed at each raster position. Commercial STEM systems use point detectors to image the transmitted electrons reducing the information to one or two channels. Point detectors sum intensity across their active area, losing almost all of the spatial information. However, commercially available available area detectors are not practical for use in STEM due to low frame rates. The PAD used for this work is capable of framing at 1.1kHz, making it's use in STEM practical. A practical area detector for STEM could revolutionize the field of electron microscopy by allowing access to orders of magnitude more information from each sample with comparable collection times to current point detectors.