2016-2017 Blue Waters Graduate Fellows

Ten outstanding computational science PhD students from across the country have been selected to receive Blue Waters Graduate Fellowships for 2016-2017. The fellowship program, now in its third year, provides substantial support and the opportunity to leverage the petascale power of National Center for Supercomputing Applications (NCSA) at the University of Illinois's Blue Waters supercomputer to advance their research. The awards are made to outstanding PhD graduate students who have decided to incorporate high performance computing and data analysis into their research.

Elizabeth Agee, University of Michigan, will examine the contributions of species-specific strategies to individual and community drought resilience in the Amazon Basin region. It is well established that the Amazon Basin region is critical to global energy, water, and carbon cycles. The increased frequency and severity of drought events in the region highlight the vulnerability of tropical forests to heat- and drought-induced stress. Despite a number of previous and ongoing studies, the full effects of drought and the essential factors controlling rainforest response are not resolved and thus remain a significant source of uncertainty for land surface models. Leveraging advances in trait-based modeling, this study presents an opportunity to examine the contributions of species-specific strategies to individual and community drought resilience and decrease uncertainties in the current suite of land surface models.

Iryna Butsky, University of Washington, plans to study the effects of cosmic rays on the galactic magnetic field evolution. Using hydrodynamical simulations of isolated spiral disk galaxies, Butsky will study the effects of cosmic rays on the galactic magnetic field evolution. These simulated galaxies will be magnetized solely by the injected magnetic fields of supernova remnants, which are also sources of cosmic rays. Butsky plans to introduce a new prescription for the interaction of cosmic rays with the galactic magnetic field, which she will add to the existing framework of magnetohydrodynamic capabilities within the Enzo code. Due to the coupled interactions between cosmic rays and the magnetic field, Butsky expects to find that cosmic rays amplify the rate magnetic field growth and are more efficient in expanding the resulting field into the surrounding dark matter halo.

August Guang, Brown University, will characterize HIV transmission networks. Improving HIV treatment and prevention efforts is enabled by characterizing the transmission network. Phylogenetic analyses are important for inferring these networks. However, there are no statistically grounded analysis methods for inferring transmission network structure from viral phylogenies, in addition to numerous analysis steps required to infer phylogenies from sequenced read data. Guang's research uses sensitivity analyses and simulations to identify what features in the HIV phylogeny reliably correspond to features in the HIV transmission network, and how that is influenced by uncertainty generated during the analysis steps. This provides a quantitative measure of how well existing approaches approximate the transmission network.

Paul Hime, University of Kentucky, will investigate the deep branches in the Tree of Life. Collecting DNA sequence data is no longer a rate-limiting step in molecular phylogenetics. Yet for all its promise, genome-scale phylogenetics is currently limited to unrealistically simple evolutionary models due to computational constraints. Hime's work will leverage the massive CPU/GPU resources on Blue Waters and newly developed statistical models to explore codon-based Bayesian methods for resolving deep (inter-ordinal) branches in the amphibian phylogeny with hundreds of genes. This research stands to make significant and impactful contributions to our understanding of tetrapod relationships. But more broadly, the methods proposed here have the potential to fundamentally change the ways that phylogeneticists analyze multi-gene data sets.

Michael Howard, Princeton University, will develop a simulation method for designing complex fluids. Engineering the flow of complex fluids is an important challenge in many technologies including biomedical devices, consumer products, and enhanced oil recovery. Complex fluid rheology is controlled on microscopic length scales by molecular structures and interactions, including hydrodynamics, which are difficult to predict from theory or to efficiently probe experimentally. Howard will develop a massively parallel multiscale simulation method for designing complex fluids using graphics processing units. Systematically coarse-grained models for fluid interfaces and solutes will be assessed. A set of principles will be developed to design a complex fluid for enhanced oil recovery that maximizes oil recovery and minimizes environmental impact.

Andrew Kirby, University of Wyoming, will simulate the most accurate calculations of wind farms to date. Today, wind farm simulations use reduced fidelity modeling via actuator discs to represent turbines. However, this technique does not accurately capture the aerodynamics required for engineering purposes. To properly engineer wind energy applications, blade-resolved aerodynamics are needed to capture the complex physics on turbine blades and to capture the unsteady flow interactions between wind turbines. To capture the complex multi-physics at high fidelity, tightly coupled, multi-disciplinary codes are needed. A state-of-the-art adaptive, high-order code called WwAaKE3D has been developed in part by Kirby and his advisor's research team. The WwAaKE3D code will be used to simulate the most accurate calculations of wind farms to date. These calculations will form major milestones in wind farm simulation capabilities.

Sherwood Richers, California Institute of Technology, will use Blue Waters to carry out simulations of 3D core collapse supernova and neutron star merger simulations. Three dimensional (3D) core collapse supernova and neutron star merger simulations suffer from approximate treatments of neutrino transport that cripple their reliability and realism. Richers proposes to use Blue Waters to carry out direct Monte Carlo simulations of the neutrino transport problem to calculate the true closure relation that will perfect the two-moment neutrino transport in 3D GRMHD simulations of these systems. These 3D simulations will be the most realistic to date and will be the first to have concrete error estimates on the accuracy of the neutrino transport method. Richers will also provide the first open-source benchmark with which all neutrino transport methods of any number of dimensions can compare.

Sean Seyler, Arizona State University, will develop a hybrid continuum-particle method for simulating large-scale heterogeneous biomolecular systems. A parallel hybrid numerical algorithm will be implemented that combines the accuracy of all-atom molecular dynamics (AAMD) in a restricted subdomain with the computational frugality of a continuum model external to the region of direct interest. Starting with an existing 3D Navier-Stokes (NS) MPI-based code and a production MD engine like NAMD, Seyler will: (1) augment the NS code to model fluctuating hydrodynamics, (2) implement a parallel coupling scheme to handle flux-/state-based boundary conditions across the particle-continuum interface, and (3) perform a sequence of tests on reference systems—from a Lennard-Jones fluid up to a solvated polymer system—to examine whether the hybrid code agrees with conventional AAMD simulations.

Ronald Stenz, University of North Dakota, will study how to improve the realism of tornado simulations. Current numerical simulations of tornadoes lack centrifuging of precipitation, producing an unrealistic maximum of precipitation in simulated tornado cores that creates an unrealistic source of negative buoyancy in the tornado updraft, limiting the stretching of vertical vorticity. A centrifuging algorithm developed for the non-hydrostatic thunderstorm simulation model, CM1, and trajectory analysis will be used for a statistically significant sample of cases to determine the importance of the inclusion of centrifuging on tornado vorticity budgets. This study will improve the realism of tornado simulations and provide further insights into the dynamical processes occurring within tornadoes. For the first time, impacts that centrifuging of precipitation has on the vorticity budget for numerically simulated, supercell-spawned tornadoes will be quantified.

Erin Teich, University of Michigan, will work to shed light on the physics of glass formation. The physics of glassy behavior is relevant in systems ranging from superconductors to sand. Despite this ubiquity, the thermodynamics underlying the slow kinetics of glass formation remain frustratingly murky. Teich's project will shed light on the physics of glass formation by examining the role that entropy plays during vitrification. In particular, Teich will examine the glassy behavior of faceted hard particle systems. Teich hypothesizes that directional entropic forces between faceted hard particles emerge and induce the formation of a variety of incommensurate locally dense motifs. The hard particle fluid then experiences a structural identity crisis, leading to dynamic arrest and glassy behavior. Teich will use a scalable in-house Monte Carlo software to test our hypothesis by simulating various glass-forming hard particle systems and identifying locally dense motifs.