Rationale for Machine Use

Since 1993 we have used machines at OSC and elsewhere in a continuing project to study the dynamics and radiation properties of relativistic extragalactic jets. Our first results (Duncan & Hughes 1994) constituted the first-published high-resolution axisymmetric simulations of these objects. That work demonstrated the stability of highly relativistic flows, and suggested an interpretation of a long-known difference between two classes of astrophysical flows in terms of flow speed-dependent stability. That work was performed on a combination of computers, including workstations at BGSU, the University of Michigan, and the OSC Cray and OVL Onyx machines. In addition, OSC personnel helped us make movies of the 2D simulations which proved very useful in both the assessment of the science and the communication of the work. That work has subsequently developed along four parallel tracks: The major focus of the present proposal for OSC resources is the 3D relativistic jet simulations using our Adaptive Mesh Refinement code. Over the last two years we have performed numerous tests on the code and have endeavored to make it as efficient as the underlying algorithms permit. Following this extensive development and testing, the code is ready to be used to perform research level simulations on the OSC Cray T94 and the Origin 2000. The computational requirements of such simulations are significant in both memory and wall clock hours, and the 3D simulations requested herein require facilities of the caliber of those at the OSC, principally due to the memory requirements. The fast, large memory machines such as the Cray T94 and the Origin 2000 are ideal for our purpose, because currently available workstations provide insufficient memory for us to achieve anything like adequate resolution while, as noted in §3.4.2, the use of massively parallel technology still presents serious challenges to AMR-type codes.