University of Colorado at Boulder
Hilary Egan accidentally landed in astrophysics. As a Michigan State University undergraduate, she tried various physics research, including condensed matter and nuclear science. But “by my senior year I decided I was too clumsy for precise lab work and really liked the computer science side of things,” says Egan, now an astrophysics doctoral candidate at the University of Colorado at Boulder.
She told staff members at Michigan State’s high-performance computing (HPC) center she wanted to do physics research involving computing. They connected her with Brian O’Shea, who does large-scale cosmological structure simulations.
Egan’s work led her to a Department of Energy Computational Science Graduate Fellowship (DOE CSGF) and research with advisor David Brain. She uses HPC systems to model how winds of charged particles emanating from a star interact with an orbiting planet’s atmosphere. Egan’s goal is to calculate whether exoplanets – bodies orbiting distant stars – might have atmospheres, a key condition for sustaining life.
The models must portray not just the atmosphere and the particle wind that can erode it but also the planet’s magnetic field (if it has one) that can deflect the wind. “You have motions of individual ions around magnetic fields all the way up to planetary scales,” Egan says, aspects so different “that being able to put it all together in one model, particularly as you’re trying to change the magnetic field, is very challenging.”
Mars is a useful test for Egan’s methods. Much of the planet’s atmosphere has dissipated over billions of years. The MAVEN (Mars Atmospheric and Volatile EvolutioN) space probe orbits the red planet, gathering magnetic field and particle movement data to study atmosphere loss. Egan compares her results against its findings.
MAVEN found that a plume of oxygen ions is escaping Mars’ already thin atmosphere. The planet has no global magnetic field, so the solar wind can directly interact with ions (charged particles) in the atmosphere. The magnetic field that is dragged along with the solar wind accelerates ions to escape Mars’ atmosphere in a plume.
Egan found that standard single-fluid magnetohydrodynamics codes, which simulate particle movement under magnetic fields as fluids, accurately portray the draping of the solar wind magnetic field around the planet, but aren’t as good at modeling the plume. The technique “makes the assumption that all of the ions have the same velocity.” The code must model different ions at different velocities to accurately calculate acceleration, especially in the plume. Egan concluded that codes portraying ions as multiple fluids or as particles portray the full picture best.
Based on her results, Egan will simulate exoplanet atmospheres with RHybrid, which models ions as particles and electrons as a fluid. Modeling an exoplanet’s atmosphere, however, will be harder than simulating that of Mars. The range of parameters the models start with is huge: planetary composition, atmosphere, magnetic field and more. Egan will focus on exoplanets that resemble Venus or Mars and orbit stars that compare well with conditions her model uses.
Yet, Egan expects her simulations will say little about a particular exoplanet’s atmospheric escape rate. Instead, she hopes to find a relationship between escape rate and factors like planet size and estimated magnetic field. When combined with results from scientists in other fields, her models could help determine which exoplanets are most likely to be habitable – “even if habitable just means bacteria.”
The calculations will first run on the university’s Summit supercomputer, but Egan expects to request computing time later at a DOE computing facility.
Egan’s 2016 practicum at the National Renewable Energy Laboratory focused almost exclusively on the details of HPC systems. With data scientist Caleb Phillips, she used a machine-learning algorithm to analyze information from the lab’s Peregrine supercomputer, seeking clues to predict and manage its power consumption for particular jobs. The goal is to develop scheduling tools that would minimize power use, especially at peak pricing periods. Such tools will help manage energy use in exascale computers coming on line in a few years.
Egan expects to graduate in 2019. In the meantime, she’ll continue spending much of her time outside her research on ultimate Frisbee. Egan played on a club team that finished third at a national tournament and now coaches the university’s women’s B team. She loves working with others, especially women, to excel in a nontraditional sport but also likes the game’s intellectual side. “I’m a nerd about ultimate Frisbee in the same way that I’m a nerd about physics. Looking at patterns and problem-solving is something that I will always bring to any sport I play.”
Image caption: Mars’ lack of a global magnetic field allows hydrogen ions (blue) in the solar wind to directly interact with the ionosphere, driving the escape of oxygen ions (purple). Credit: R. Jarvinen; simulation image created with the yt visualization program.