Modeling hyperenergetic and superluminous supernovae
Philipp Mösta Einstein fellow @ UC Berkeley [email protected]
Roland Haas, Goni Halevi, Christian Ott, Sherwood Richers, Luke Roberts, Erik Schnetter
BlueWaters symposium 2017 May 17, 2017 Astrophysics of core-collapse supernovae
M82/Chandra/NASA ~ Galaxy evolution/feedback Heavy element nucleosynthesis
Birth sites of black holes / neutron stars 2 Neutrinos New era of transient science • Current (PTF, DeCAM, ASAS-SN) and upcoming wide-field time domain astronomy (ZTF, LSST, …) -> wealth of data • adv LIGO / gravitational waves detected • Computational tools at dawn of new exascale era
Image: PTF/ZTF/COO Image: LSST 3 New era of transient science • Current (PTF, DeCAM, ASAS-SN) and upcoming wide-field time domain astronomy (ZTF, LSST, …) -> wealth of data • adv LIGO / gravitational waves detected • Computational tools at dawn of new exascale era Transformative years ahead for our understanding of these events
Image: PTF/ZTF/COO Image: LSST 4 Hypernovae & GRBs
• 11 long GRB – core-collapse supernova associations. • All GRB-SNe are stripped envelope, show outflows v~0.1c • But not all stripped-envelope supernovae come with GRBs • Trace low metallicity and low redshift
Neutrino mechanism is inefficient; can’t deliver a hypernova 5 Superluminous supernovae
Some events: stripped envelope no interaction 45 Elum ~ 10 erg 52 Erad up to 10 erg
Gal-Yam+12
6 Superluminous / hyperenergetic supernovae
SLSN Ic lGRBs SN Ic-bl
Common engine?
7 Core collapse basics Iron core Protoneutron star r~30km Nuclear equation of state 2000km stiffens at nuclear density
Inner core (~0.5 M ) -> protoneutron star + shockwave
8 Core collapse basics Iron core Protoneutron star r~30km Nuclear equation of state 2000km stiffens at nuclear density
Inner core (~0.5 M ) -> protoneutron star + shockwave Reviews: Bethe’90 Janka+‘12 L⌫ Outer core accretes onto shock & protoneutron accretion star with O(1) M /s shock Shock stalls at ~ 100 km
9 Core collapse basics Iron core Protoneutron star r~30km Nuclear equation of state 2000km stiffens at nuclear density
Inner core (~0.5 M ) -> protoneutron star + shockwave Reviews: Bethe’90 Janka+‘12 L⌫ Core-collapse accretion supernova problem: shock How to revive the shockwave?
10 Magnetorotational mechanism [LeBlanc & Wilson ‘70, Bisnovatyi-Kogan ’70, Obergaulinger+’06, Burrows+ ‘07, Takiwaki & Kotake ‘11, Winteler+ 12]
Rapid Rotation + B-field amplification (need magnetorotational instability [MRI]; difficult to resolve, but see, e.g, Obergaulinger+’09, PM+15)
2D: Energetic bipolar explosions Energy in rotation up to 1052 erg Results in ms-period proto-magnetar
11 Burrows+’07 A multiphysics challenge
Magneto-Hydrodynamics Gas/plasma dynamics
12 A multiphysics challenge
Magneto-Hydrodynamics Gas/plasma dynamics
General Relativity Gravity
13 A multiphysics challenge
Magneto-Hydrodynamics Gas/plasma dynamics
General Relativity Gravity Nuclear EOS, nuclear Nuclear and Neutrino Physics reactions & ν interactions
14 A multiphysics challenge
Magneto-Hydrodynamics Gas/plasma dynamics
General Relativity Gravity Nuclear EOS, nuclear Nuclear and Neutrino Physics reactions & ν interactions
Boltzmann Transport Theory Neutrino transport
15 A multiphysics challenge
Magneto-Hydrodynamics Gas/plasma dynamics
General Relativity Gravity Nuclear EOS, nuclear Nuclear and Neutrino Physics reactions & ν interactions
Fully coupled! Boltzmann Transport Theory Neutrino transport All four forces!
16 A multiphysics challenge
Magneto-Hydrodynamics Gas/plasma dynamics
General Relativity Gravity Nuclear EOS, nuclear Nuclear and Neutrino Physics reactions & ν interactions
Fully coupled! Boltzmann Transport Theory Neutrino transport All four forces! Additional Complication: Core-Collapse Supernovae are 3D • rotation • fluid and MHD instabilities, multi-D structure, spatial scales Need 21st century tools: • cutting edge numerical algorithms • sophisticated open-source software infrastructure • peta/exa scale computers 17 http://einsteintoolkit.org 18 3D Volume Visualization of
Entropy PM, Richers+ 14
19 How do we form magnetars? Magnetorotational Mechanism
Big uncertainty so far: How do we get the magnetic field amplification?
21 Burrows+’07 First global 3D MHD turbulence simulations
• 10 billion grid points (Millenium simulation used 10 billion particles) • 130 thousand cores on Blue Waters • 2 weeks wall time • 60 million compute hours • 10000 more expensive than any previous simulations
Does the MRI efficiently build up PM+ 15 Nature dynamically relevant global field? 22 3D magnetic field structure
dx=500m dx=200m dx=100m dx=50m
23 PM+ 15 Nature 24 PM+ 15 Nature Growth at Large Scales
7 ( 2.05 + 0.75 ms 1 (t t )) 1033 · map · 32 (t t )/t 6 5 10 e map , t = 3.5 ms k ·= 4 ] 5 k = 6 erg k = 8 33 4 k = 10 10
)[ k = 20 t
( 3 k = 50 k = 100 ,mag k 2 E
1
0 5 10 t t [ms] map
saturation within 60ms 25 PM+ 15 Nature Global Field Structure PM+ 15 Nature dx=500m dx=50m
t=0ms t=10ms t=10ms 26 PM+ 15 Nature Global Field Structure PM+ 15 Nature dx=500m dx=50m
Magnetar formation?
t=0ms t=10ms t=10ms 27 PM+ 15 Nature R-process nucleosynthesis in magnetar-driven explosions Neutron-rich nucleosynthesis in supernovae Creating the heaviest elements Jet-driven explosions proposed as site for r- process
• Low electron fraction • Medium entropy • Low density • High temperature
29 Sneden+ 08 R-process - Basics
PM, Roberts, Halevi+ 17 (in prep) Halevi, PM+ 17 (in prep)
30 R-process in jet-driven supernovae
Winteler+12
31
Nishimura+15 R-process in jet-driven supernovae
B = 1013 G / octant
32
Halevi, PM 17 (in prep) R-process in jet-driven supernovae 100 12 2 B = 10 G / octant 10
4 10
] 6 10 M [ 8 10
ejecta 10 10 = M L 0 51 12 L = 10 erg/s 10 L = 1052 erg/s 10 14 L = 1053 erg/s 16 10 80 100 120 140 160 180 200 mass number A B = 1012 G full 3D
100
2 10
4 10
] 6 10 M [ 8 10
ejecta 10 10 = M L 0 51 12 L = 10 erg/s 33 10 L = 1052 erg/s 10 14 L = 1053 erg/s 16 10 80 100 120 140 160 180 200 PM, Roberts, Halevi+ 17 (in prep) mass number A 3) From simulations to observations From simulations to observations Full star State of the art now:
Detailed simulations full physics Full 3D, full physics 0.1-1s inner core ~10000km Current frontier:
1) Engine model from full-physics simulations 2) Simplified simulations with engine model to shock breakout
35 PM, Tchekhovskoy 17 (in prep) From simulations to observations
State of the art now: Future: Detailed simulations Full-star simulations full physics 0.1-1s full physics inner core ~10000km shock breakout Current frontier:
1) Engine model from detailed light curves full-physics simulations detailed spectra 2) Simplified simulations with engine model to connect observations and shock breakout engines map progenitor params
36 Summary New (hyperenergetic/superluminous) transients challenge our engine models
Need detailed massively parallel 3D GRMHD simulations to interpret observational data
Magnetoturbulence and large-scale dynamo action create conditions for magnetar engine
Robust r-process elements only from iron cores that were magnetized strongly precollapse
High-performance computing and BlueWaters key to solving these puzzles 37 Thank you!
38 Core-collapse supernovae neutrinos turbulence
(Binary) black holes accretion disks EM counterparts Magnetic~ fields in high-energy astro
Binary neutron stars Extreme core-collapse hyperenergetic gravitational waves superluminous EM counterparts lGRBs 39 sGRBs Core-collapse supernovae neutrinos turbulence
(Binary) black holes accretion disks EM counterparts Magnetic~ fields in high-energy astro
Binary neutron stars Extreme core-collapse hyperenergetic gravitational waves superluminous EM counterparts lGRBs 40 sGRBs MRI
41 MRI Basics
MC Mi Mo
42 MRI Basics
• Weak field instability
• Requires negative angular velocity gradient
• Can build up magnetic field exponentially fast
• Extensively researched in accretion disks: ability to modulate angular momentum transport and grow large scale field
43 What’s the situation in core-collapse?
Stability criterion:
d⌦2 8⌦2 < !2 + r < 0 BV dr
[Balbus&Hawley 91,98, Akiyama+03, Obergaulinger+09]
44 Magnetorotational Mechanism
M. Obergaulinger et al.: MRI in core collapse supernovae 259 • MRI works locally Akiyama+03, Shibata+06 • shearing box simulations z 13 3 Fig. 22. Volume rendered magnetic field strength of a model with b0 = 4 10 Gcomputedinaboxof1 2 1km with a resolution of 20 m at t = 21.5ms(left)andt = 37.2ms(right), respectively. The coordinate directions× are indicated as in Fig. 19×. ×
Obergaulinger+09max term avge term log Mϖφ / Mϖφ log Mϖφ / Mϖφ the weakest initial field, because the theoretical growth rate for 1 the fastest growing MRI mode is σMRI = 1.14 ms− for these models (see Sect. 4.2.1). 0 0.75 1.50 2.25 3 −0.25 0.25 0.75 1.25 1.75 2.25 To investigate the stability properties of large-scale channel modes as a function of the box geometry, we simulated mod- els with an initially uniform magnetic field using boxes of dif- ferent size and shape. The models were rotating according to 1 Ω0 = But1900 s− and whatαΩ = 1.25, and about their initial magnetic global field? z 13 − field was b = 4 10 Gwhenapplyingvelocitydamping z 0 × z 13 / L boundaries, and b = 2 10 Gotherwise.Wevariedboth φ 0 × 1 the ratio between the radial and vertical, Lϖ/Lz,andtheradial 45 log L and toroidal size, Lφ/Lϖ,sizeofthebox.Thegridresolution Burrows+’07 max term was 20 m (see Table B.3). Plotting the stress ratios Mϖφ /Mϖφ (Fig. 23;dampingboundaries)and M /Mterm (Fig. 24;non- ⟨ ϖφ⟩ ϖφ damping boundaries) as a function of the aspect ratio of the com- putational box, provides some indication of the range of Mϖφ values prevailing during the post-growth phase. The ratios al- low one to distinguish models with a strong variability due to 1 1 the dominant re-appearance of channel modes from those mod- log Lϖ / Lz log Lϖ / Lz els exhibiting a smooth evolutionwithoutdominantlarge-scale Fig. 23. The left panel shows the ratio of the maximum Maxwell stress coherent structures. max term per unit volume, Mϖφ ,anditsvalueatMRItermination,Mϖφ ,asa We find that models with a radial aspect ratio Lϖ/Lz = 1 function of the toroidal and radial aspect ratios, Lφ/Lz and Lϖ/Lz for and a toroidal aspect ratio Lφ/Lz 2areunstableagainstpar- the models listed in Table B.3. The right panel shows the ratio of Mϖφ asitic instabilities, independent of≥ the grid resolution in toroidal averaged over the saturation phase and the value at MRI termination. direction. Turbulence develops and leads to a flow structure as Each model is represented by a symbol its color reflecting its maximum Maxwell stress. All models are computed imposing velocity damping shown in Fig. 22.Modelshavingthesameradialaspectratio, at the radial grid boundaries. but a smaller toroidal one are stable and evolve similarly as ax- isymmetric models, i.e., parasitic instabilities do not grow and adominantlarge-scalechannelflowdevelops,whichgivesrise off.Hence,ifachannelflowformsatlatetimeswithawave- to a morphology of the type presented in Fig. 22.Thesefind- length equal to the box size in z-direction, Lz,unstableparasitic ings do not depend on how the growth of the MRI ends, i.e., modes must have a toroidal wavelength 2Lz to grow rapidly. whether velocity damping is applied and reconnection between Thus results in the criterion for the channel∼ flow instability we adjacent channels occurs inside the box, or whether no damp- have found in our simulations. ing is imposed and reconnection occurs near the surface of the In accordance with simulations presented recently by Bodo computational box. et al. (2008), we find that models with a radial aspect ratio These results can be understood from the analysis of par- Lϖ/Lz 1experienceasecondexponentialgrowthphaseas ≤ max term asitic instabilities by Goodman & Xu (1994), who argued that described in Sect. 4.2 (note the large ratios of Mϖφ and Mϖφ three-dimensional flows are unstable against parasitic instabili- for the corresponding models in Fig. 23), whereas a larger ra- ties, but these instabilities can be suppressed by the geometry of dial aspect ratio appears to favor a less violent post-growth phase the computational box. According to their analysis, the growth where coherent channel modes can appear but are disrupted after rate of the parasitic instabilities is highest for modes with half ashorttime.Bodo et al. (2008)obtainedthisresultforsimula- the wave number of the unstable MRI modes they are feeding tions performed with a toroidal aspect ratio Lφ/Lz = 4. From simulations to observations
State of the art now: i i Detailed simulations B b full physics i i 0.1-1s U u inner core ~10000km ~j ~b Current frontier: ...⇥ 1) Engine model from full-physics simulations 2) Simplified simulations with engine model to shock breakout
Get mean fields from 3D sim
46 Squire, PM, Lecoanet 16 (in prep) From simulations to observations
State of the art now:
Detailed simulations 20 full physics
0.1-1s 10 2.5×10-6 inner core ~10000km 1.5×10-6
0 Current frontier: 5.0×10-7
-5.0×10-7 1) Engine model from -10 full-physics simulations 2) Simplified simulations -20 with engine model to shock breakout 0 5 10 15 20 25
Get mean fields Daedalus simulation from 3D sim
47 Squire, PM, Lecoanet 16 (in prep) Magnetic field amplification: A 2D view dx=500m dx=200m dx=100m dx=50m
48 Energy Spectra
1036 a t t = 10 ms map Ekin 50 m 1035 36 5/3 5 10 erg k · · 1034
1033 [erg]
) 1032 ( k E 1031
Emag 500 m 30 10 Emag 200 m
Emag 100 m 1029 Emag 50 m
Emag 50 m (t tmap = 0 ms) 1028
1 10 100 k
49 Energy Spectra
36 b 36 5/3 10 5 10 erg k · · Ekin(k) t = 7 ms 1035
1034
1033 [erg]
) 32
( k 10 E 1031
1030 Emag(k) t = 0 ms t = 1 ms t = 6 ms 1029 t = 2 ms t = 8 ms t = 4 ms t = 10 ms 1028 1 10 100 k • Turbulent saturated state after 3ms • inverse cascade afterwards 50