PRACE Scientific & Industrial Conference 2016 Orea Hotel Pyramida, Prague, 10‒12 May, 2016 Solving the Mysteries of Supernova Explosions by 3D Simulations Hans-ThomasHans-Thomas JankaJanka Max Planck Institute for Astrophysics, Garching Supernova 1987A Supernova 1987A (SN 1987A) Supernovae ● Spectacularly bright final stages of stellar evolution ● Birth of neutron stars and black holes ● Origin of heavy chemical elements ● Observational signals: neutrinos, gravitational waves How do massive stars explode? Supernova Theory 1933 Walter Baade and Fritz Zwicky: “The core of a massive, evolved star collapses to a neutron star while the star explodes as supernova.” Fritz Zwicky (1898–1974) CRAB Nebula with pulsar is remnant of Supernova 1054 Neutron Stars as Pulsars 20 km SN-remnant Cassiopeia A X-ray (CHANDRA, green-blue); optical (HST, yellow); IR (SST, red) SN-remnant Cassiopeia A Neutron Star X-ray (CHANDRA, green-blue); optical (HST, yellow); IR (SST, red) Neutron Star Kicks How do Supernovae Explode? 1956 Geoffrey Burbidge Are Supernovae the Origin of Silver, Gold, and Platinum? Final Stages of Massive Stars Neutrinos Energize Explosion 1966 David W. Arnett Stirling Colgate HUGE Binding Energy of Neutron Star 2 GM 53 Egrav ~ > 10 erg R = 1046 J ~ 100 x Supernova Energy Supernova 1987A Supernova 1987A (SN 1987A) Supernova 1987A als Teenager Supernova 1987A Delayed Neutrino-Driven Mechanism 1982‒1985 James R. Wilson Hans Bethe The Smallest (and Cheapest) Supernova Model Experiment Neutrino heating builds up pressure behind the shock front until overlying matter is expelled like the lid on a heated pot. Shock Front Lid ν ν Steam Neutrinos Water Heat Hot Neutron Star Heater Does the neutrino-heating mechanism really work? How does the supernova explosion proceed? This is a quantitative problem that needs computer simulations! Predictions of Signals from SNe & NSs hydrodynamics of stellar plasma relativistic gravity (nuclear) EoS neutrino physics progenitor conditions dynamical models neutrinos nucleosynthesis LC, spectra gravitational waves explosion asymmetries, pulsar kicks explosion energies, remnant masses Scales in the Supernova Problem General-Relativistic Supernova Models of the Garching Group (Müller B., PhD Thesis (2009); Müller et al., ApJS, (2010)) GR hydrodynamics (CoCoNuT) CFC metric equations Neutrino transport (VERTEX) Neutrino Reactions in Supernovae Beta processes: Neutrino scattering: Thermal pair processes: Neutrino-neutrino reactions: The Team Students, Postdocs, Collaborators, Support ● Tobias Melson, Robert Bollig, Thomas Ertl, Georg Stockinger, Robert Glas ● Florian Hanke ● Alexander Summa, Michael Gabler, Oliver Just ● Ewald Müller ● Bernhard Müller (Univ. Belfast & Monash, Melbourne) ● Annop Wongwathanarat (RIKEN, Tokyo) ● Irene Tamborra (Niels Bohr Institute, Copenhagen) ● Georg Raffelt (MPP Munich) ● Andreas Marek, Lorenz Hüdepohl (MPCDF) The Simulation Code Prometheus/CoCoNuT ‒ VERTEX: 1D, 2D, 3D ● Hydro modules: Newtonian: Prometheus + effective relativistic grav. potential. General relativistic: CoCoNuT Higher-order Godunov solvers, explicit. ● Neutrino Transport: VERTEX Two-moment closure scheme with variable Eddington factor based on model Boltzmann equation; fully energy-dependent, O(v/c), implicit, ray-by-ray-plus in 2D and 3D. ● Most complete set of neutrino interactions applied to date. ● Different nuclear equations of state. ● Spherical polar grid or axis-free Yin-Yang grid. Parallel-Programming Model ✔ Mixed (hybrid) MPI-OpenMP parallelization: Hydro module fully MPI parallelized Transport module MPI+OpenMP parallelized over radial “rays” ✔ Several OpenMP threads (↔ cores) per MPI task ( ↔ nodes) ✔ GPUs can be used for special-physics “sub-problems” (speed-up by factor 2‒3; A. Marek, MPCDF) ✗ Memory RAM: 1‒2 Gbyte/core needed (tables, transport variables) ✗ Need powerful cores: Microphysics is compute-intense Tabular input would require even more memory/core Performance and Portability of our Supernova Code Prometheus-Vertex Code employs hybrid Strong Scaling MPI/OpenMP programming model (collaborative development with Katharina Benkert, HLRS). Code has been ported to different ) computer platforms, optimized, 3 1 and further parellelized by 0 2 Andreas Marek, High Level G U Application Support, Max C ( . Planck Computing and Data l a t Facility, Garching (MPCDF). e k e Code shows excellent parallel r a efficiency, which can be fully M . A exploited in 3D. 2D vs. 3D Morphology (Images from Markus Rampp, MPCDF) 3D Supernova Simulations EU PRACE and GAUSS Centre grants of ~1 billion core hours allow(ed) us to do the first 3D simulations on 16.000 cores. Mare Nostrum SuperMUC 3D Core-Collapse SN Explosion Models 9.6 Msun (zero-metallicity) progenitor (Heger 2010) Melson et al., ApJL 801 (2015) L24 3D Core-Collapse SN Explosion Models 9.6 Msun (zero-metallicity) progenitor (Heger 2010) Melson et al., ApJL 801 (2015) L24 CRAB Nebula with pulsar, remnant of Supernova 1054 Explosion properties: 50 Eexp ~ 10 erg = 0.1 bethe MNi < 0.01 Msun Low explosion energy and ejecta composition (little Ni, C, O) are compatible with CRAB (SN1054) Might also explain other low- luminosity supernovae (e.g. SN1997D, 2008S, 2008HA) 3D Core-Collapse SN Explosion Models 20 Msun (solar-metallicity) progenitor (Woosley & Heger 2007) Melson et al., ApJL 808 (2015) L42 Observational Consequences and indirect evidence for neutrino heating and hydrodynamic instabilities at the onset of stellar explosions: ● Neutrino signals (characteristic modulations) ● Gravitational-wave signals ● Neutron star kicks ● Asymmetric mass ejection & large-scale radial mixing ● Progenitor – explosion – remnant connection ● Light curve shape, spectral features (electromagnetic emission) ● Nucleosynthesis (= formation of chemical elements by nuclear reactions) Search for Supernova Neutrinos The Super-Kamiokande Detektor in Japan 50.000 tons of ultra pure water, 11.146 photo multiplier Search for Supernova Neutrinos The IceCube Neutrino Teleskope at the South Pole 1 billion tons of ice, 5.160 photo multiplier on 86 strings Search for Gravitational Waves The turbulent processes in the supernova center cause tiny perturbations (“ripples”) of space-time: “gravitational waves” (Einstein 1915) LIGO: two interferometers in the USA (Hanford, Livingston) with 4 km, a 3rd with 2 km arm length VIRGO interferometer near Pisa in Italy with 3 km arm length GEO600 near Hannover with 0.6 km arm length 3D Core-Collapse Models: Neutrino Signals 11.2, 20, 27 Msun progenitors (WHW 2002) SASI produces modulations of neutrino emission and gravitational-wave signal. at 10 kpc Tamborra et al., PRL 111, 121104 (2013); PRD 90, 045032 (2014) Outlook ● Better resolution for assessing convergence (turbulence) ● Study dependence on initial conditions (progenitor stars) ● Explore new phenomena and unknown physics in neutron stars and black-hole formation ● Long-time simulations to link theory to observations of supernovae and remnants Astrophysics with self-consistent 3D supernova models is only just beginning! A New Nonradial 3D Instability Dipole asymmetry of lepton number emission (LESA) Lepton number flux: νe minus anti-νe 11.2 Msun progenitor (WHW 2002) Tamborra, Hanke, Janka, et al., ApJ 792, 96 (2014) A New Nonradial 3D Instability: "LESA" Lepton Emission Self-sustained Asymmetry caused by dipolar convection and accretion asymmetry When does the next Galactic supernova explode? The signals of 1000 events are already on their way to us! 1 billion kilometers Betelgeuse: Red giant star in 600 light years distance.