Brian W. O'Shea - Research

join us! research
cv / bio

My research focuses on theoretical and numerical studies of galaxy formation and evolution, as well as both astrophysical and terrestrial plasma phenomena. I do this primarily using large-scale simulations run with one of several open source simulation tools, many of which are contributed to by members of my research group (including the Enzo, Enzo-E, K-Athena, and Athena-PK simulation codes and the yt data analysis and visualization package; see the software page). More recently, I have become interested in the use of data science methods for accelerating large-scale multiphysics simulations and model interpretation, and I am also interested in learning about how undergraduate science students learn to use computing. My current research interests are detailed below. Some images from my simulations can be found in the image gallery, and information on recent publications can be found on the news page and in my CV. None of this research would be possible without my research group, of course!

I am actively looking for undergraduate students, graduate students, and postdoctoral researchers to join my group! If you are interested in working on the subjects discussed below (or a related topic), please take a look at the Join the group! page and then email me. I am interested in working with (and have research projects suitable for) astrophysics, physics, CMSE, computer science, and applied mathematics undergraduate students, graduate students, and postdoctoral researchers.

Low-redshift galaxies and the circumgalactic medium

Overview: Galaxies are composed of vastly more than the stars that are easily seen by optical telescopes. They are surrounded by halos of gas, which serve as a reservoir for outflows from the central regions of the galaxy that are driven by supernova and supermassivel black hole feedback. In addition, this gas acts a mediator between the central part of the galaxy and its environment. It is crucial to consider the entire galaxy and its environs - stars, gaseous halo, and intergalactic medium, and the interplay between them - when attempting to understand the formation of galaxies like our own Milky Way. We have a tremendous amount of observational information about our own galaxy and about neighboring galaxies, which probe the chemical and dynamical behavior of these galaxies. This work is an essential complement to studies of the most distant galaxies in the universe, since it allows us to probe very different aspects of the structure formation process.

My group's research focus: High-fidelity cosmological and idealized simulations of galaxy formation focusing on the self-regulation of galactic systems; Chemical evolution, including evolving stellar populations; Improved models for star formation and feedback and AGN feedback in cosmological simulations; The use of metal-poor galactic halo stars as probes of Milky Way chemical and dynamical evolution; The development of statistical tools to compare models with large-scale observational datasets; Quantification of the uncertainties associated with modern techniques for studying chemical and galaxy evolution.

Galaxy clusters and diffuse astrophysical plasmas

Overview: Galaxy clusters, composed of tens or hundreds of galaxies orbiting within a single common dark matter halo, are the largest gravitationally-bound objects in the Universe, often weighing more than 1014 times the mass of our Sun (or 100 times the mass of the Milky Way). As the largest bound objects, galaxy clusters are useful probes of cosmology and are very interesting astrophysical laboratories in that they are essentially "closed box" systems. 90% of the baryonic matter in galaxy clusters is outside of the galaxies themselves, residing in a hot, diffuse plasma called the "intracluster medium," or ICM, which is extremely bright at X-ray wavelengths but invisible to the naked eye. This plasma is threaded with magnetic fields and relativistic protons and electrons, which are crucial to controlling its behavior. Understanding the ICM and its interactions with the galaxies contained within it is crucial to gaining a complete understanding of galaxy clusters as a whole, the life cycles of these objects, and to assessing their utility as cosmological probes. One aspect of this understanding relates to the plasma processes that govern the ICM, including magnetic field generation, turbulent transport of energy, and the acceleration of particles to relativistic speeds.

My group's research focus: Non-thermal evolution of the intracluster medium, including the effects of magnetic fields and cosmic rays; The effect of feedback from stellar populations and active galactic nuclei, particularly on radio, gamma-ray and x-ray observables; The effects of non-thermal plasma processes in the ICM on galaxy clusters as cosmological probes; Alternative methods of simulating galaxies within clusters.

Development of software tools for extremely large-scale simulations

Overview: Many problems in modern astrophysics and plasma physics rely on numerical simulations to make significant theoretical progress. As the questions we ask become more detailed and difficult, so too must the calculations that we undertake in our attempts to answer them. Cosmological structure formation is particularly challenging - in order to be useful for probing the evolution of populations of galaxies, we need to simulate large volumes of the Universe at high spatial and mass resolution and with many different kinds of physics (dark matter dynamics, hydrodynamics, magnetic fields, radiation, cooling, ISM chemistry, star formation and feedback, cosmic rays, and so on). Such calculations are incredibly computationally demanding, and push us to use ever-larger computers. Modern supercomputers are composed of hundreds of thousands or millions of computing cores, often with specialized accelerator hardware, that typically are connected together using networks with complex topologies. The current generation of astrophysical simulation codes needs to be extensively modified and upgraded to take advantage of these machines, and strategies for dealing with the petabytes of resulting data need to be devised. Similar challenges exist in simulations of both terrestrial and astrophysical plasma phenomena due to the wide range of spatial and temporal scales required in their study. An additional challenge in plasma modeling is the potential for a system of interest to span multiple plasma regimes as it varies across space and time, which requires careful consideration of the numerical methods that are used to simulate those regimes.

My group's research focus: Scaling of AMR CFD+gravity codes to exascale computers, including memory, solver, parallelism, performance portability, and reliability issues; Large-scale data analysis, including complex and time-ordered datasets; Visualization of petabyte-scale datasets; Extracting statistically useful information from massive, inhomogeneous, time-ordered numerical and observational datasets; Development of algorithms for exascale supercomputers; Development of performance-portable software.

Machine learning and physical systems

Overview: Most astrophysical systems (such as galaxies) and terrestrial plasma experiments (such as Z-pinch or inertial confinement fusion devices) have a similar set of characteristics. Their evolution is characterized by physical phenomena that vary over a large range of spatial and temporal scales, with different physics being relevant at different scales. In many circumstances, this is particularly challenging because the large- and small-scale phenomena interact in complex ways that can't be ignored. As a consequence, direct simulations of these systems may require a range of scales that is computationally infeasible, or potentially require simulation of a system whose fundamental governing equations change either as the system evolves, or which vary from place to place within the system. Machine learning (ML) and artificial intelligence (AI) methods are potentially a promising way to approach this system, either through the physically-constrained emulation of small-scale phenomena in a larger simulation or by dynamically identifying and modifying the governing equations in simulations as they evolve. In addition, ML/AI techniques can be used when one is trying to understand the underlying (astro)physical system using the limited observational or experimental diagnostic tools that are currently available.

My group's research focus: Discovery of the governing equations within a physical system; Automation of model selection; Emulation of small-scale phenomena in multiphysics astrophysical models; Physics-informed Bayesian inference as applied to astronomical observations and experimental diagnostics; data-driven surrogate model approaches appplied to multiscale and/or hierarchical modeling.

Computational Science Education

Overview: The ability to make models of systems - physical, biological, financial, social, or otherwise - is a critical skill that is widely used in the sciences, engineering, and in business. Similarly, the ability to manipulate and visualize data is critical. However, most students learn how to do these things informally, which often means that they have critical gaps in their understanding of model-making and data manipulation. As part of MSU's Department of Computational Mathematics, Science and Engineering, the Computational Education Research Laboratory (CERL) is interested in exploring how students learn to do computational work and their attitudes about doing that work. In particular, we explore how students learn to integrate computation into their education outside of traditional computer science courses.

My group's research focus: I am not currently working directly with students that are interested in this research area; however, my colleagues in CERL are pursuing research projects in a wide range of topics. Please consult the CERL Group Member page and contact the person/people doing work that you find interesting!

The research done by my research group is supported in part by the National Science Foundation, the National Aeronautics and Space Administration, the U.S. Department of Energy, the Department of Defense, and Michigan State University.