Theoretical nuclear and computational physics
May 2012 – Present
I currently work with Prof. Steffen Bass in the quantum chromodynamics (QCD) theory group at Duke University. We study the formation and properties of the quark-gluon plasma (QGP), a hot and dense phase of matter that is produced in collisions of heavy nuclei. It is postulated that the universe was one large quark-gluon plasma in the first moments after the Big Bang.
It would be futile to attempt to write complete research descriptions here. The reader should understand that the following information is incomplete and oversimplified. However, feel free to email me with questions or peruse the Wikipedia articles I linked to above.
- Statistical hadronization models. Statistical modeling is one of the simplest ways to approach heavy-ion collisions and the QGP. We assume that hadrons are produced statistically according to their masses and the surrounding temperature. I am using a statistical package called SHARE to determine how well the model applies to both heavy-ion collisions and elementary collisions such as proton-proton and electron-positron.
- Statistical emulation of computer models. Complex computer models can calculate the evolution of a heavy-ion collision. However, the codes are generally very computationally expensive. An emulator interpolates a set of results from a computer model, providing a fast alternative to actually running the model. I use the emulator with the SHARE model mentioned above.
- Fast emulation on GPUs. The emulator involves a matrix inverse operation, a slow linear algebra calculation. GPUs (graphics cards) can execute numerical linear algebra code much faster than convential CPUs (processors). I am working on porting the emulator code to run on Nvidia GPUs using CUDA.
- “Onion” visualization. Visualization is a valuable tool for understanding many aspects of heavy-ion collisions. One of the most important features is the freeze-out surface—where particle interactions cease. There are in fact several such freeze-out surfaces, leading to an onion-like appearance. I'm using the open-source software ParaView to create and study these “onion” visualizations.
- Data management with iRODS. The QCD group's computer models produce very large data files (often in excess of a gigabyte). It is therefore useful to intelligently and automatically manage these data, keeping recent and relevant files on local disks while moving older files to remote storage. I am implementing such a system using iRODS, a flexible data management software package.