Computational Astrophysics

A tiny man working on a microchip.

The LCSE, under the direction of Paul Woodward, is engaged in a wide range of research projects. The lab focuses mainly upon collaborative projects between the government, industry, K-12 schools, and the University's Institute of Technology. Examples of research done by Tom Jones and his students at the Minnesota Supercomputing Institute include simulations of galaxy clusters, evolution of turbulence, radio galaxy dynamics and cosmic ray acceleration and transport.

Computational Astrophysics at the Minnesota Supercomputing Institute (MSI)

Research Topics

Paul R. Woodward

Over the past several years, we have been simulating brief but important events in the evolution of stars that stretch our ability to harness the power of present large computing systems and that will continue to put demanding requirements on future systems. These events are brief, in that they last only a few days, or even less, while typical time scales for stars are very much longer. Nevertheless, they are not as brief as explosive events, and for this reason, we must simulate the behavior of the star for a great many dynamical times. These events also depend critically upon processes that are inherently 3D. The events we simulate occur in deep convection zones that develop above nuclear burning shells. We find that the most important modes of convection in these shells are comparable in size to the entire shells themselves, forcing us to include in our simulation domain the entire interior region of the star, not just a small section of it. Our simulations therefore present the challenge that we must describe the whole stellar interior over a long time interval, which requires millions of time steps. Because we are unwilling to wait the necessary time for these simulations to complete on a small cluster of machines—which would be years—we must find a way to get all this done in a reasonable time by using large numbers of tightly coupled machines in a large, single computing system. We must find ways to have enormous numbers of computational engines simultaneously involved productively in the work, despite the fact that the greatest challenge is the number of time steps needed rather than the number of spatial grid cells. A focus of recent work is the hydrogen ingestion flash. This can occur in a late stage of evolution for an intermediate mass star. In the phase of their evolution after core helium burning, asymptotic giant branch (AGB) stars have periodic outbursts called thermal pulses. Each such thermal pulse begins with helium that is accumulating between the hydrogen-burning shell and the degenerate carbon-oxygen core suddenly igniting in what is called the helium shell flash. The energy generation in the just-ignited helium burning shell rises so rapidly at this point that the energy cannot be carried outward effectively by radiative heat transport, and therefore a convection zone develops above the helium burning shell. This convection zone is called the pulse-driven convection zone (PDCZ). The PDCZ grows in mass by incorporating more and more of the gas above it, and it also grows in radius by lifting the overlying layers upward. This upward lifting of the hydrogen burning shell reduces its temperature by an expansion of the gas, and the hydrogen burning effectively ceases. What is special about this sequence of events in AGB stars of the early universe (as well as in some post-AGB stars of the present universe) is that the PDCZ can grow by incorporating gas from above it right up to the point where it encounters gas where the hydrogen has not yet burned. Once this fresh hydrogen fuel is entrained into the convection flow, even in small concentrations, it can burn violently in a hydrogen ingestion flash.

The simulated interior of a 2 MꙨ AGB star of the early universe.
Two volume-rendered views of the far hemisphere of the simulated interior of a 2 MꙨ AGB star of the early universe, with metallicity Z = 10-5, at time 2,699 min. A violent wave of combustion of entrained, H-rich gas is now seen at the lower right, about to cross the star, pulling down large concentrations of H-rich fuel. At this stage, the presence of our outer bounding sphere becomes strongly felt, and our simulation can no longer be trusted. All the hydrogen within the bounding sphere is quickly consumed after the combustion wave reaches the opposite side of the star. We are now testing a new code that will be able to move the bounding sphere outward a factor of 2 in radius.