Coincident with the opening of NCSA to its users, Jim Stone entered UIUC's Department of Astronomy in January 1986. With thesis advisor Dimitri Mihalas and astrophysicist Mike Norman and later as an NCSA postdoctoral fellow, Stone began modeling varied astrophysical phenomena. Today he is guiding his own students as an assistant professor of astronomy at the University of Maryland, College Park.
From graduate student to faculty member, Stone's research focus has been the hydrodynamics of magnetized and/or radiation- dominated plasmas. Broadly "hydrodynamics is the study of the evolution of fluids," he explains. "If there is a magnetic field that permeates the fluid, and the fluid contains charged particles, that magnetic field can change the dynamics of the fluid in fundamental ways. It produces much richer physics." A radiation field has a similar effect.
"Magnetic fields are everywhere --from stars to the Galaxy as a whole," Stone says. "Magnetohydrodynamical (MHD) phenomena are probably important in almost every problem in astrophysical fluid dynamics." Prime examples include the evolution of the interstellar medium, accretion disk dynamics, and the emission of winds and jets from these disks.
Radiation hydrodynamics (RHD) "is another story," he continues. "There you need gas that is hot enough to contain enough photons to make a difference. It is a high-energy phenomenon. This [condition] only occurs in compact objects, such as neutron stars, black holes, supernovae, and hot stars. It is a long list but much more limited than those amenable to MHD."
During his NCSA postdoc years (1990-1992), Stone developed the CMHOG (Connection Machine Higher-Order Godunov) code, which implements Colella and Woodward's piecewise parabolic method (PPM) algorithm. As in ZEUS, "the algorithmic methods are based on finite differencing, where you solve the same equations in each zone," Stone says. NCSA Research Scientist Norman has described CMHOG as "the fastest implementation of the best algorithm on one of the largest parallel supercomputers in the world" [see access, Spring 1994]. The code has been instrumental in his and UIUC graduate student Greg Bryan's cosmology simulations on the CM-5.
Based on three or four decades of development, "the PPM technique for hydrodynamics is now widely available," Stone says. "The challenge to MHD and RHD is to develop new methods that can treat accurately all the new physics. Capturing the new things that happen with a magnetized fluid is hard."
Observations, such as those by the radio telescopes of the Berkeley-Illinois-Maryland Array (BIMA), provide some evidence of a disk's shape and interaction with the central object, Stone states. Supercomputer calculations are necessary to investigate their internal structure and understand the complex time-dependent physics and nonlinear evolution, explains University of Virginia astrophysicist John Hawley in a historical review of accretion disk simulations published in a recent issue of Science (September 8, 1995).
The accretion process essentially works by having mass fall from the disk onto the central object. "In order for that to happen, mass must lose angular momentum" so the gas in the disk can move to an increasingly smaller orbit and "get closer and closer" to the object, Stone says. "The great puzzle that has stumped astrophysicists for over 20 years is: What physical mechanism allows mass to lose angular momentum so it can spiral inwards? Otherwise, the material will orbit around the [object] forever and ever."
Using hydrodynamics techniques alone, researchers have tried to find viable transport mechanisms, but such techniques did not work for all accretion disks, Stone says. Hawley, whom Stone considers his closest collaborator, is a pioneer in applying MHD to this area. Through simulations he and Virginia colleague Steve Balbus discovered an unexpected instability that occurs only in the presence of a weak magnetic field. This Balbus-Hawley instability appears to be able to transport angular momentum in all types of accretion disks.
"So, the story seems to be that if there is not a magnetic field -- a purely hydrodynamic disk -- it is very difficult to find a process for angular momentum transport," Stone adds. "With a magnetic field, it becomes very easy." For this discovery and other developments, Hawley received the 1993 Helen B. Warner Prize for Astronomy.
"The Cray is useful for lots of exploratory runs, to learn about the physics," Stone says. "When you think about doing a really big simulation -- one that takes thousands of hours of time -- you need a parallel machine." A Guest Computational Investigator grant from NASA HPCC's Earth and Space Sciences Project has assisted Stone in developing more advanced parallel algorithms, both for the CM-5 and the MasPar MP-2. "The advantage of massively parallel machines . . . is that they are very efficient for finite difference codes," Stone explains. Performance of 7 to 8 gigaflops is typical on 512 nodes of the CM-5.
Such performance has allowed more realistic models of accretion disks. Thus far, the main effect Hawley and Stone have added is stratification, where density is not uniform across the disk. "[We] find that in stratified disks, magnetic buoyancy results in the production of a strongly magnetized envelope in the outer, low-density regions of the disk," Stone says. This envelope "surrounds a weakly magnetized core at the center of the disk. The results are particularly important to understanding the nature of the nonlinear, saturated stage of the Balbus-Hawley instability in real disks and to the dynamics of accretion disks in general."
The next step includes "looking at disks where the fluid is only partially ionized," Stone says. "We're also doing global simulations of disks, where the computational domain is the entire disk rather than just a piece of it. Those problems are very challenging and are why we have the MetaCenter grant."
More information is available at the Laboratory for Computational Astrophysics Web site, or send email to lca@ncsa.uiuc.edu.
The density (color) and magnetic field (red lines) patterns in a 3D simulation of a magnetically unstable accretion disk. The disk midplane is oriented horizontally in the middle of the grid. The magnetic instability produces MHD turbulence in the disk characterized by complex density and magnetic field structures.
Return to the Table of Contents.