Overview

AMReX

People

Publications

Donald Willcox

Contact Information Donald Willcox MS 50A-3111 Lawrence Berkeley National Lab 1 Cyclotron Rd. Berkeley, CA 94720 510-486-6900 (fax) [email protected]

I was a project scientist in the Center for Computational Sciences and Engineering (CCSE) in the Computing Sciences Area at the Lawrence Berkeley National Laboratory. My research interests are in computational astrophysics, specifically in designing large scale hydrodynamics simulations of thermonuclear supernova explosions and progenitors. I am also interested in accelerating these algorithms for GPU-based supercomputers, especially the ODE solvers required for time integration of nuclear reaction networks.

Reaction Network ODE integration on GPUs

I am additionally working on nuclear reaction network integration for GPUs using CUDA. Typical nuclear reaction networks can be quite stiff and are commonly solved with implicit multistep ODE integration methods such as the BDF method implemented in VODE and CVODE. To take advantage of GPU accelerators, I first ported VODE to CUDA Fortran and shown are some early test results for a variety of small reaction networks. My current work involves scaling to larger reaction networks by adapting the CUDA support in CVODE to integrate the ODE systems in a grid of cells together with control of batched GPU operations and optimized GPU linear algebra. This work is in collaboration with the SUNDIALS group at Lawrence Livermore Lab.

Code Generation for Reaction Networks with pynucastro

Related to my work on nuclear reaction networks for GPUs and for science applications, I collaborate with Michael Zingale from Stony Brook University and others to develop pynucastro. pynucastro is a Python code providing an accessible interface to nuclear reaction rate parameterizations, weak rate tabulations, and atomic mass data that allows for easy construction of reaction networks as well as visualization and exploration. pynucastro currently supports JINA Reaclib reaction rates as well as selected weak rate tabulations. pynucastro is also capable of code generation for reaction network ODE systems, supplying the right hand side and Jacobian in ready-to-compile Fortran modules and generated the reaction network I am currently using for the convective Urca process simulations. Recently I added CUDA Fortran support as a step towards GPU integration for large reaction networks.

The Convective Urca Process

The convective Urca process in white dwarf stars near the Chandrasekhar-mass can occur during the late stages of convective simmering in the single degenerate progenitor model for Type Ia (thermonuclear) supernovae. The A=23 convective Urca process couples electron captures onto 23Na to beta decays of 23Ne via convective currents driven by fusion of 12C in the core of such a white dwarf as it slowly approaches thermonuclear runaway. The energy transport and neutrino losses due to these weak nuclear reactions couple to convection and determine the composition structure of the convection zone near runaway and are thought to influence the dynamics of the runaway. Realistic simulations of this process are challenging because of the inherently 3-D nature of convection and the long timescales involved during the convective simmering phase. I am using the low Mach number code Maestro together with a nuclear reaction network consisting of carbon burning and the A=23 weak Urca reactions to simulate the convective Urca process over a range of central conditions.

I completed my PhD in Physics at Stony Brook University in August 2018 working with Alan Calder and Michael Zingale on a variety of projects, some of which I am still involved with. I first explored supernovae explosion simulations from newly proposed "hybrid" carbon-oxygen-neon white dwarfs in the single degenerate progenitor paradigm for Type Ia supernovae. These simulations consisted of a suite of 2-D simulations using the Flash code. Since then I have worked on understanding the convective Urca process using the low Mach hydrodynamics code Maestro, developed by LBNL and Stony Brook. Along the way I've explored interests in reaction network code generation, GPU acceleration for ODE integration, and uncertainty quantification for simulations.

• M. M. Hoffman, D. E. Willcox, M. P. Katz, S. Ferson, F. D. Swesty, A. C. Calder, On the Quantification of Incertitude in Astrophysical Simulation Codes, manuscript in preparation.
• M. Zingale, K. Eiden, Y. Cavecchi, A. Harpole, J. B. Bell, M. Chang, I. Hawke, M. P. Katz, C. M. Malone, A. J. Nonaka, D. E. Willcox, W. Zhang, Toward Resolved Simulations of Burning Fronts in Thermonuclear X-ray Bursts, submitted to the proceedings of the AstroNum 2018 conference, [arXiv].
• A. C. Calder, D. E. Willcox, C. J. DeGrendele, D. Shangase, M. Zingale, D. M. Townsley, Thermonuclear (Type Ia) Supernovae and Progenitor Evolution, submitted to the proceedings of the AstroNum 2018 conference.
• A. C. Calder, M. M. Hoffman, D. E. Willcox, M. P. Katz, F. D. Swesty, S. Ferson, Quantification of Incertitude in Black Box Simulation Codes, 2018, Journal of Physics: Conference Series, 1031, 012016, [doi].
• D. E. Willcox and M. Zingale, pynucastro: an interface to nuclear reaction rates and code generator for reaction network equations, 2018, Journal of Open Source Software, 3(23), 588, [doi].
• M. Zingale, A. S. Almgren, M. G. Barrios Sazo, V. E. Beckner, J. B. Bell, B. Friesen, A. M. Jacobs, M. P. Katz, C. M. Malone, A. J. Nonaka, D. E. Willcox, W. Zhang, Meeting the Challenges of Modeling Astrophysical Thermonuclear Explosions: Castro, Maestro, and the AMReX Astrophysics Suite, 2018, Journal of Physics: Conference Series, 1031, 012024, [doi].
• A. C. Calder, B. K. Krueger, A. P. Jackson, D. E. Willcox, B. J. Miles, D. M. Townsley, Cosmic Chandlery with Thermonuclear Supernovae, 2017, Journal of Physics: Conference Series, 837, 012005, [doi].
• D. E. Willcox, D. M. Townsley, A. C. Calder, P. Denissenkov, F. Herwig, Type Ia Supernova Explosions From Hybrid Carbon-Oxygen-Neon White Dwarf Progenitors 2016, Astrophysical Journal, 832, 13, [doi].

• Stellar Explosion Mechanics: Properties and Physical Processes in White Dwarf Interiors, Student Seminar Series, Institute for Advanced Computational Sciences, Stony Brook University, Stony Brook, NY, 11/15/2017.
• The Dynamics and Origins of Thermonuclear (Type Ia) Supernovae, Interdisciplinary Theoretical and Computational Physical Science meeting, Tokyo Institute of Technology, Tokyo, Japan, 10/05/2017.
• A Brief Tour of the AMReX Astrophysics Suite of Codes, NY Area Computational Hydro Workshop, Flatiron Institute/CCA, New York, NY, 09/29/2017.
• White Dwarfs as Type Ia Supernovae Progenitors, Research Caf� Series, Center for Inclusive Education, Stony Brook University, Stony Brook, NY, 06/28/2017.
• Simulations of Various White Dwarf Progenitor Models for Type Ia Supernovae Current Challenges in the Physics of White Dwarf Stars (invited talk), Santa Fe, NM, 06/16/2017.
• Status of Recent Work for Type Ia Supernovae Progenitors: Hybrid C-O-Ne White Dwarfs, the Convective Urca Process, and Accelerated Reaction Networks, Astrophysics Seminar, Los Alamos National Laboratory, Los Alamos, NM, 06/14/2017.
• Elucidating the Convective Urca Process in Pre-Supernova White Dwarfs Using Three-Dimensional Simulations, Junior Researchers Workshop, JINA-CEE Frontiers in Nuclear Astrophysics Meeting, Michigan State University, East Lansing, MI, 02/05/2017.

• D. E. Willcox, A. Jacobs, X. Li, M. Zingale, pynucastro: Code Generation and Visualization for Nuclear Reaction Networks, Bay Area Scientific Computing Day 2018, Sandia National Laboratories, Livermore, CA, December 7, 2018.
• D. E. Willcox, D. M. Townsley, M. Zingale, A. C. Calder, Three Dimensional Simulations of the Convective Urca Process in White Dwarf Progenitors of Type Ia Supernovae, Current Challenges in the Physics of White Dwarf Stars, Santa Fe, NM, June 12-16, 2017.
• D. E. Willcox, D. M. Townsley, M. Zingale, A. C. Calder, Elucidating the Convective Urca Process in Pre-Supernova White Dwarfs Using Three-Dimensional Simulations, JINA-CEE Frontiers in Nuclear Astrophysics Meeting, Lansing, MI, February 7-9, 2017.
• D. E. Willcox, D. M. Townsley, M. Zingale, A. C. Calder, Three-Dimensional Simulations of the Convective Urca Process in Pre-Supernova White Dwarfs, American Astronomical Society Meeting 229, 244.05, 2017.
• M. M. Hoffman, M. P. Katz, D. E. Willcox, S. Ferson, F. D. Swesty, A. C. Calder, On the Quantification of Incertitude in Astrophysical Simulation Codes, American Astronomical Society Meeting 229, 154.27, 2017.
• D. E. Willcox, D. M. Townsley, A. C. Calder, P. Denissenkov, F. Herwig, Thermonuclear Supernova Explosions From Hybrid White Dwarf Progenitors, American Astronomical Society Meeting 227, 237.17, 2016.
• D. E. Willcox, D. M. Townsley, A. C. Calder, P. Denissenkov, F. Herwig, A Comparison of Type Ia Supernovae with C-O and Hybrid C-O-Ne White Dwarf Progenitors, F.O.E. Fifty-One Erg International Workshop, North Carolina State University, NC, 2015.
• D. E. Willcox, D. M. Townsley, A. C. Calder, A Study of Steady-State Detonation Structures for Hybrid C, O, Ne White Dwarf Models, International Conference (on) Type Ia Supernovae: Progenitors, Explosions, and Cosmology, University of Chicago, IL, 2014.
• D. E. Willcox, M. A. Reber, Y. Chen, K. Halder, T. Allison, Imaging Molecular Structure With High Harmonics, Chemistry Research Day, Stony Brook University, NY, 2013.
• M. A. Reber, Y. Chen, D. E. Willcox, T. Allison, Cavity-Enhanced Transient Absorption Spectroscopy, Chemistry Research Day, Stony Brook University, NY, 2013.