new insights into type ia supernova explosions from large...
TRANSCRIPT
An Advanced Simulation and Computation (ASC) Academic Strategic Alliances Program (ASAP) Center
at The University of Chicago
The Center for Astrophysical Thermonuclear Flashes
SciDAC 2007Boston, MA, 27 June 2007
Don Q. LambNNSA ASC ASAP Flash Center
University of Chicago
New Insights into Type Ia Supernova Explosions from Large-Scale Simulations
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
NNSA Advanced Simulation and ComputingAcademic Strategic Alliance Program
ASC Academic Strategic AllianceCenters:
Focus on problems that involvecomplex multi-scale, multi-physicssimulations that require integrationof a number of academic disciplinesand software componentsAddress the difficult problems ofverification and validation of thesimulationsEducate and train students andyoung researchers in computationalscience
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
FLASH Capabilities Span a Broad Range…
Cellular detonation
Compressed turbulence
Helium burning on neutron stars
Richtmyer-Meshkov instability
Laser-driven shock instabilitiesNova outbursts on white dwarfs Rayleigh-Taylor instability
Flame-vortex interactions
Gravitational collapse/Jeans instability
Wave breaking on white dwarfs
Shortly: Relativistic accretion onto NS
Orzag/Tang MHDvortex
Type Ia Supernova
Intracluster interactions
MagneticRayleigh-Taylor
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
FLASH Capabilities Span a Broad Range…
Cellular detonation
Compressed turbulence
Helium burning on neutron stars
Richtmyer-Meshkov instability
Laser-driven shock instabilitiesNova outbursts on white dwarfs Rayleigh-Taylor instability
Flame-vortex interactions
Gravitational collapse/Jeans instability
Wave breaking on white dwarfs
Shortly: Relativistic accretion onto NS
Orzag/Tang MHDvortex
Type Ia Supernova
Intracluster interactions
MagneticRayleigh-Taylor
The FLASH code1. Parallel, adaptive-mesh refinement (AMR) code2. Block structured AMR; a block is the unit of
computation3. Designed for compressible reactive flows4. Can solve a broad range of (astro)physical problems5. Portable: runs on many massively-parallel systems6. Scales and performs well7. Fully modular and extensible: components can be
combined to create many different applications
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
FLASH Capabilities Span a Broad Range…
Cellular detonation
Compressed turbulence
Helium burning on neutron stars
Richtmyer-Meshkov instability
Laser-driven shock instabilitiesNova outbursts on white dwarfs Rayleigh-Taylor instability
Flame-vortex interactions
Gravitational collapse/Jeans instability
Wave breaking on white dwarfs
Shortly: Relativistic accretion onto NS
Orzag/Tang MHDvortex
Type Ia Supernova
Intracluster interactions
MagneticRayleigh-Taylor
The FLASH code1. Parallel, adaptive-mesh refinement (AMR) code2. Block structured AMR; a block is the unit of
computation3. Designed for compressible reactive flows4. Can solve a broad range of (astro)physical problems5. Portable: runs on many massively-parallel systems6. Scales and performs well7. Fully modular and extensible: components can be
combined to create many different applications
Development of the FLASH code was
made possible by nearly a decade of
funding and support by NNSA ASC
Academic Strategic Alliance Program
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
FLASH Is Being Used by Groups Throughout the World
US
Canada
US
GermanyItaly
Canada
Germany
Monthly Notices of the Royal Astronomical SocietyVolume 355 Issue 3 Page 995 - December 2004doi:10.1111/j.1365-2966.2004.08381.x
Quenching cluster cooling flows with recurrent hot plasma bubblesClaudio Dalla Vecchia1, Richard G. Bower1, Tom Theuns1,2, Michael L. Balogh1, Pasquale Mazzotta3 and Carlos S. Frenk1
UK
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
Netherlands
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
FLASH Is Being Used by Groups Throughout the World
US
Canada
US
GermanyItaly
Canada
Germany
Monthly Notices of the Royal Astronomical SocietyVolume 355 Issue 3 Page 995 - December 2004doi:10.1111/j.1365-2966.2004.08381.x
Quenching cluster cooling flows with recurrent hot plasma bubblesClaudio Dalla Vecchia1, Richard G. Bower1, Tom Theuns1,2, Michael L. Balogh1, Pasquale Mazzotta3 and Carlos S. Frenk1
UK
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
Netherlands
FLASH is being used by nearly 300
scientists around the worldto do
simulations in computational fluid
dynamics, astrophysics, cosmology,
plasma physics, ...; to
do scaling
studies, software development, etc.
in CS; and is a key acceptance test
for the IBM BG/P at ANL
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
What Are Type Ia Supernovae?
Supernova Cosmology Project
QuickTime™ and aYUV420 codec decompressor
are needed to see this picture.
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
What Are Type Ia Supernovae?
Supernova Cosmology Project
QuickTime™ and aYUV420 codec decompressor
are needed to see this picture.
Peak luminositiesof most Type Ia SNeare similar -- making
them excellent“cosmic yardsticks”
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Joint Dark Energy Mission
SNAP
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Type Ia Supernovae
Accretion• Stellar binary evolution code with accretion
Smoldering• Subsonic convection
in core of white dwarf• Conductive heat
transport
Flame• Initially compressible
deflagration phase• DDT or expansion/recollapse• Conductive heat transport
Lightcurve• Free expansion of star • Multi-group (non-LTE)
radiation transport>108 yr
~seconds
~1000 yr
Ignition
Image copyrighted by Mark A. Garlick Image credit P. Garnavich/CfA
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Type Ia Supernovae
Accretion• Stellar binary evolution code with accretion
Smoldering• Subsonic convection
in core of white dwarf• Conductive heat
transport
Flame• Initially compressible
deflagration phase• DDT or expansion/recollapse• Conductive heat transport
Lightcurve• Free expansion of star • Multi-group (non-LTE)
radiation transport>108 yr
~seconds
~1000 yr
Ignition
Image copyrighted by Mark A. Garlick Image credit P. Garnavich/CfA
Nuclear energy
powers the explosion
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Type Ia Supernovae
Accretion• Stellar binary evolution code with accretion
Smoldering• Subsonic convection
in core of white dwarf• Conductive heat
transport
Flame• Initially compressible
deflagration phase• DDT or expansion/recollapse• Conductive heat transport
Lightcurve• Free expansion of star • Multi-group (non-LTE)
radiation transport>108 yr
~seconds
~1000 yr
Ignition
Image copyrighted by Mark A. Garlick Image credit P. Garnavich/CfA
Nuclear energy
powers the explosion
Radioactive decay of 56 N
heats expanding gas and
makes explosion visible
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Central Ignition(single or multiple
Ignition points)
Off-Center Ignition(single or multiple
Ignition points)
Four Different Explosion Mechanisms Have Been Proposed
PureDeflagration
DeflagrationTo Detonation
Transition (DDT)
GravitationallyConfined
Detonation
Plewa, Calder, Lamb (2004);Townsley et al. (2007);Jordan et al. (2007)
Reineke et al. (1999, 2002);Schmidt et al. (2006)
Khokhlov (1991)
Khokhlov (1991);Gamezo et al.(2004, 2005)
PulsationallyDriven DDT
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Movie of 3-D Simulation of Deflagration Phase of GCD Model
Jordan et al. (2007)
QuickTime™ and aYUV420 codec decompressor
are needed to see this picture.
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Detonation Conditions Are Robustly Produced by Inwardly Directed Jet
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Detonation Conditions Are Robustly Produced by Inwardly Directed Jet
GCD model is only model so far in
which detonation occurs naturally;
i.e., without being put in by hand
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Movie of End-to-End Simulation of GCDModel Including Detonation Phase
QuickTime™ and aYUV420 codec decompressor
are needed to see this picture.
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Computational Demandsof Type Ia SN Simulations
Following turbulent nuclear burning requires ~ 1-10 km resolution, whereas radius of star is 104 km --adaptive mesh refinement is therefore essentialHigh-resolution, 3-D, whole star simulations require ~ 100K-300K cpu-hrs on current machinesEach simulation ~ 200 MB/cpu-hrs; i.e., ~ 20-60 TB of dataWe will have used ~ 2M cpu-hrs on ASC machines and ~ 5 M cpu-hrs on Office of Science machines this year; expect to use 20-30M cpu-hrs next yearWe will have generated ~ 1 PB of data this year; ~ 3-5 PB of data next year
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Importance of Verification and Validationof Type Ia Supernova Simulations
ConvectionDuring
SmolderingPhase
Rayleigh-Taylor -- Driven Turbulent Nuclear Burning
DeflagrationTo Detonation
Transition
GlobalValidation
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Rayleigh-Taylor--Driven Turbulent Nuclear Burning
Zhang et al. (2007)
Nuclear flame is ~ 1 mmthick while resolution of best simulations is 1 km -- a factor of 108 larger!
A subgrid model of flame is therefore essential
The burning rate -- and thusthe result -- depends on model
Correct model to use is controversial -- further verification and validation studies are badly needed
Verification simulations we have done indicate burning rate is dominated by largest scales
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Convergence With Resolution
Jordan et al. (2007)
Convergence with resolutionis excellent through bubble
breakout
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Global Validation of Light Curves and Spectra Predicted by GCD Model
Comparison of U, V, B, R light curves predicted by GCD model and obs. of Type Ia Supernova SN 2001el
Kasen and Plewa (2006)
Comparison of spectra predicted by GCD model and obs. of Type Ia supernova SN 1994D
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Global Validation of Light Curves and Spectra Predicted by GCD Model
Comparison of U, V, B, R light curves predicted by GCD model and obs. of Type Ia Supernova SN 2001el
Kasen and Plewa (2006)
Comparison of spectra predicted by GCD model and obs. of Type Ia supernova SN 1994DLight curves and spectra
agree with observations
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Global Validation of Light Curves and Spectra Predicted by GCD Model
Comparison of U, V, B, R light curves predicted by GCD model and obs. of Type Ia Supernova SN 2001el
Kasen and Plewa (2006)
Comparison of spectra predicted by GCD model and obs. of Type Ia supernova SN 1994DLight curves and spectra
agree with observations
GCD model can produce
bright Type Ia supernovae;
can it produce faint ones?
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Challenges in Getting to Petascale Type Ia Supernova Simulations
Efficiency on multi-core chips -- FLASH achieved 1.9 x speed-up on BG/L; we expect 3.6-3.8 x speed-up on BG/PScaling to 300K-1M cores -- FLASH scales wellto 65K cores (BG/L) and we expect the same to 128K cores, but 1M cores?Memory -- astro memory needs are largeI/O -- high bandwidth parallel I/O crucialData transfer -- 200Mbytes/cpu-hr generatedData storage -- many Petabytes, must be remoteScientific analysis pipeline -- must be remoteData archive -- Petabytes, must “compress”
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
Challenges in Getting to Petascale Type Ia Supernova Simulations
Efficiency on multi-core chips -- FLASH achieved 1.9 x speed-up on BG/L; we expect 3.6-3.8 x speed-up on BG/PScaling to 300K-1M cores -- FLASH scales wellto 65K cores (BG/L) and we expect the same to 128K cores, but 1M cores?Memory -- astro memory needs are largeI/O -- high bandwidth parallel I/O crucialData transfer -- 200Mbytes/cpu-hr generatedData storage -- many Petabytes, must be remoteScientific analysis pipeline -- must be remoteData archive -- Petabytes, must “compress”
Astrophysical simulations are often “discovery
science:” the phenomena are very complicated
and highly non-linear; we therefore don’t know
all of the questions we want to ask before we
see the data, so we must be able to explore it
The ASC/Alliances Center for Astrophysical Thermonuclear FlashesThe University of Chicago
ConclusionsFlash Center has conducted extensive verification and validationstudies of physical processes in Type Ia SNeCenter has carried out a series of high-resolution, 3-D, whole star simulations of Type Ia SNeThese simulations were made possible by a decade of support for development of the FLASH code and of computational resourcesprovided by the NNSA ASC Academic Strategic Alliance ProgramSignificant computational resources were also provided by the Office of Science INCITE programThey led to the discovery of the “gravitationally confined detonation” (GCD) model of Type Ia SNe -- the only model so far that detonates “naturally” (i.e., without being put in by hand)While almost surely not the final story, the GCD model reproduces many key observed properties of Type Ia SNe and is therefore promisingPetascale Type Ia SN simulations will enable us to get control overkey physical processes, but getting there will be challenging!