Explosion Models of Massive Stars

Transcription

1 SFB-TR7 Fifty-One Ergs NC State University, Raleigh, North Carolina, May 13th 17th, 2013 Explosion Models of Massive Stars Hans-Thomas Janka Max Planck Institute for Astrophysics, Garching

2 Heinzi-Ado Arnolds (MPA, 2012) Supernova and neutron star merger research is team effort Students & postdocs: Alexandra Gessner Robert Bollig Andreas Voth Chang Liu Thomas Ertl Tobias Melson Florian Hanke Lorenz Hüdepohl Janina von Groote Else Pllumbi Andreas Bauswein Oliver Just Pedro Montero Bernhard Müller Marcella Ugliano Annop Wongwathanarat Collaborators: Ewald Müller Martin Obergaulinger Almudena Arcones Achim Schwenk Andreas Marek Stephane Goriely Nick Stergioulas Thomas Baumgarte Georg Raffelt Victor Utrobin Shinya Wanajo

3 Outline Introduction: The neutrino-driven mechanism Status of self-consistent models in two dimensions The question of dimensions: How does 3D differ from 2D? Seven arguments for neutrino-driven explosions

4 Neutrinos & SN Explosion Mechanism Paradigm: Explosions by the neutrino-heating mechanism, supported by hydrodynamic instabilities in the postshock layer Rs ~ 200 km Neutrino-heating mechanism : Neutrinos `revive' stalled shock by energy deposition (Colgate & White 1966, Wilson 1982, Bethe & Wilson 1985); Convective processes & hydrodynamic instabilities support the heating mechanism (Herant et al. 1992, 1994; Burrows et al. 1995, Janka & Müller 1994, 1996; Mezzacappa et al. 1998, Fryer & Warren 2002, 2004; Blondin et al. 2003; Scheck et al. 2004,06,08, Iwakami et al. 2008, 2009, Ohnishi et al. 2006).

5 Shock revival : Stalled shock wave must receive energy to start reexpansion against ram pressure of infalling stellar core. O Accretion Si ν Shock can receive fresh energy from neutrinos! ν n, p Shock wave Si Proto-neutron star ν ν

6 Neutrino heating: Neutrino cooling: See, e.g., Janka, ARNPS 62 (2012) 407 Neutrino Heating and Cooling Hydrodynamic instabilities

7 1D-2D Differences in Parametric Explosion Models Nordhaus et al. (ApJ 720 (2010) 694) and Murphy & Burrows (2008) performed 1D & 2D simulations with simple neutrino- heating and cooling terms (no neutrino transport but lightbulb) and found up to ~30% improvement in 2D for 15 Msun progenitor star. (ApJ 720 (2010) 694)

8 But: Is neutrino heating strong enough to initiate the explosion? Most sophisticated, self-consistent numerical simulations of the explosion mechanism in 2D and 3D are necessary!

9 General-Relativistic 2D 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)

10 Neutrino Reactions in Supernovae Beta processes: Neutrino scattering: Thermal pair processes: Neutrino-neutrino reactions:

11 The Curse and Challenge of the Dimensions Boltzmann equation determines neutrino distribution function in 6D phase space and time Θ ϵ f (r, θ, ϕ,θ, Φ, ϵ,t ) Φ θ Integration over 3D momentum space yields source terms for hydrodynamics r ϕ Q (r, θ, ϕ, t), Y e ( r,θ, ϕ, t) Solution approach 3D hydro + 6D direct discretization of Boltzmann Eq. (code development by Sumiyoshi & Yamada '12) 3D hydro + two-moment closure of Boltzmann Eq. (next feasible step to full 3D; cf. Kuroda et al. 2012) 3D hydro + ''ray-by-ray-plus'' variable Eddington factor method (method used at MPA/Garching) 2D hydro + ''ray-by-ray-plus'' variable Eddington factor method (method used at MPA/Garching) Required resources PFlops/s (sustained!) 1 10 Pflops/s, TBytes PFlops/s, Tbytes Tflops/s, < 1 TByte

12 "Ray-by-Ray" Approximation for Neutrino Transport in 2D and 3D Geometry Solve large number of spherical transport problems on radial rays associated with angular zones of polar coordinate grid Suggests efficient parallization over the rays radia l ray

13 Computing Requirements for 2D & 3D Supernova Modeling Time-dependent simulations: t ~ 1 second, ~ 106 time steps! CPU-time requirements for one model run: In 2D with 600 radial zones, 1 degree lateral resolution: ~ 3*1018 Flops, need ~106 processor-core hours. In 3D with 600 radial zones, 1.5 degrees angular resolution: ~ 3*1020 Flops, need ~108 processor-core hours.

14 Performance and Portability of our Supernova Code Prometheus-Vertex Code has been ported to different computer platforms by Andreas Marek, High Level Application Support, Rechenzentrum Garching (RZG). Code shows excellent parallel efficiency, which will be fully exploited in 3D. Strong Scaling Andreas Marek, RZG (2012) Code employs hybrid MPI/OpenMP programming model (collaborative development with Katharina Benkert, HLRS).

15 3D Supernova Models PRACE grant of million core hours allows us to do the first 3D simulations on cores.

16 Relativistic 2D CCSN Explosion Models 8.8 Msun 8.1 Msun 9.6 Msun 11.2 Msun 15 Msun 27 Msun radius North Color coded: entropy time after bounce South 25 Msun Bernhard Müller, THJ, et al. (ApJ 756, ApJ 761, arxiv: Basic confirmation of previous explosion models for 11.2 and 15 Msun stars by Marek & THJ (2009)

17 Relativistic 2D CCSN Explosion Models "Diagnostic energy" of explosion Maximum shock radius

18 Growing set of 2D CCSN Explosion Models Florian Hanke (PhD project) Mass accretion rate Average shock radius Progenitor models: Woosley et al. RMP (2002)

19 Growing set of 2D CCSN Explosion Models Florian Hanke (PhD project) Mass accretion rate Average shock radius Progenitor models: Woosley et al. RMP (2002)

20 2D explosions for 11.2 and 15 Msun progenitors of Woosley et al. (2002, 1995) Suwa et al., arxiv: D explosions for 13 Msun progenitor of Nomoto & Hashimoto (1988) Suwa et al. (PASJ 62 (2010) L49 Support for 2D CCSN Explosion Models

21 Support for 2D CCSN Explosion Models 2D explosions for 12, 15, 20, 25 Msun progenitors of Woosley & Heger (2007) Bruenn et al., ApJL 767, 6 (2013) More details in Bronson Messer's talk!

22 Comparison of 1D & 2D Explosion Models F. Hanke F. Hanke 2D 2D 1D and 2D simulations for 12, 15, 20, 25 Msun progenitors of Woosley & Heger (2007) by F. Hanke (Newtonian with GR gravity corrections) and by B. Müller (GR) agree well with each other. Garching models show significant differences compared to Bruenn et al. (2012) Garching 1D & 2D models are not as close to explosion. 1D GR B. Müller

23 Comparison of 1D & 2D Explosion Models F. Hanke F. Hanke 2D 2D 1D and 2D simulations for 12, 15, 20, 25 Msun progenitors of Woosley & Heger (2007) by F. Hanke (Newtonian with GR gravity corrections) and by B. Müller (GR) agree well with each other. Garching models show significant differences compared to Bruenn et al. (2012) Garching 1D & 2D models are not as close to explosion.

24 2D SN Explosion Models Basic confirmation of viability of the neutrino-driven mechanism Confirmation of reduction of the critical neutrino luminosity for an explosion in self-consistent 2D treatments compared to 1D Explosions in 2D simulations were recently also obtained by Suwa et al. (2010, 2012) and Bruenn et al. (2012), but quantitatively differ from Garching models. Many numerical aspects are different; 2D code comparisons are needed!

25 Challenge and Goal: 3D 2D explosions seem to be marginal, at least for some progenitor models and in some of the most sophisticated simulations. Nature is three dimensional, but 2D models impose the constraint of axisymmetry. Turbulent cascade in 3D transports energy from large to small scales, which is opposite to 2D.

26 Full-Scale 3D Core-Collapse Supernova Models with Detailed Neutrino Transport

27 3D Core-Collapse Models 27 Msun progenitor (WHW 2002) 27 Msun SN model with neutrino transport develops spiral SASI as seen in idealized, adiabatic simulations by Blondin & Mezzacappa (Nature 2007) F. Hanke et al., arxiv: ms p.b. 245 ms p.b. 240 ms p.b. 278 ms p.b.

28 F. Hanke et al., arxiv: Msun progenitor (WHW 2002) 3D Core-Collapse Models

29 F. Hanke et al., arxiv: D Core-Collapse Models 27 Msun progenitor (WHW 2002): Spiral mode axis

30 3D Core-Collapse Models 25 Msun progenitor (WHW 2002) F. Hanke et al., arxiv:

31 3D Explosions?

32 2D-3D Differences in Parametric Explosion Models Nordhaus et al. (ApJ 720 (2010) 694) performed 2D & 3D simulations with simple neutrino- heating and cooling terms (no neutrino transport but lightbulb) and found 15 25% improvement in 3D for 15 Msun progenitor star (ApJ 720 (2010) 694)

33 2D-3D Differences in Parametric Explosion Models F. Hanke (Diploma Thesis, MPA, 2010) in agreement with L. Scheck (PhD Thesis, MPA, 2007) could not confirm the findings by Nordhaus et al. (2010)! 2D and 3D simulations for 11.2 Msun and 15 Msun progenitors are very similar but results depend on numerical grid resolution: 2D with higher resolution explodes easier, 3D shows opposite trend! Hanke et al., ApJ 755 (2012) 138, arxiv: D & 3D slices for 11.2 Msun model, L = 1.0*1052 erg/s

34 Growing "Diversity" of 3D Results Dolence et al. (arxiv: ) find much smaller 2D/3D difference of critical luminosity, but still slightly earlier explosion in 3D. Takiwaki et al. (ApJ 749:98, 2012) obtain explosion for an 11.2 Msun progenitor in 3D later than in 2D. Find a bit faster 3D explosion with higher resolution. Couch (arxiv: ) finds also later explosions in 3D than in 2D and higher critical luminosity in 3D! Ott et al. (arxiv: ) reject relevance of SASI in 3D and conclude that neutrino-driven convection dominates evolution. Reasons for 2D/3D differences and different results by different groups are not understood!

35 Growing "Diversity" of 3D Results These results do not yield a clear picture of 3D effects. But: The simulations were performed with different grids (cartesian+amr, polar), different codes (CASTRO, ZEUS, FLASH, Cactus, Prometheus), and different treatments of input physics for EOS and neutrinos, some with simplified, not fully self-consistent set-ups. Resolution differences are difficult to assess and are likely to strongly depend on spatial region and coordinate direction. Partially compensating effects of opposite influence might be responsible for the seemingly conflicting results. Convergence tests with much higher resolution and detailed code comparisons for clean, well defined problems are urgently needed!

36 Explosion Mechanism: Most Sophisticated Current 3D Models

37 3D Core-Collapse Models 27 Msun progenitor (WHW 2002) Neutrino luminosities Shock position (max., min., avg.) 3D 2D 2D 3D 2D Time scale ratio Florian Hanke, PhD project 3D

38 3D Core-Collapse Models 11.2 Msun progenitor (WHW 2002) Florian Hanke, PhD project

39 3D Core-Collapse Models 11.2 Msun progenitor (WHW 2002) 3D Neutrino luminosities Shock position (max., min., avg.) 2D 2D 2D 3D 3D Florian Hanke, PhD project Time scale ratio

40 3D Core-Collapse Models 20 Msun progenitor (WH 2007) 3D 2D Shock position (max., min., avg.) Neutrino luminosities 2D 2D 3D Florian Hanke, PhD project Time scale ratio 3D

41 Summary 2D models with relativistic effects (2D GR and approximate GR) yield explosions for soft EoSs, but explosion energy may tend to be on low side. Considerable quantitative differences compared to Bruenn et al. (arxiv: ) demand detailed comparison. 3D modeling has only begun. No clear picture of 3D effects yet. But SASI can dominate (certain phases) also in 3D models! 3D models still need higher resolution for convergence. Progenitors are 1D, shell structure and convective perturbations may be important! How important is slow rotation for SASI growth?

42 Data Interpretation Physicists:

43 Data Interpretation Astronomers:

44 Data Interpretation Astrophysicists:

45 7 (at least) Arguments for Neutrino-Driven Explosions

46 1 Mechanism based on self-consistent, "ab initio" model with ascertained (though potentially incomplete) ingredients. No advocation of questionably aspects, e.g. extremely rapid rotation, ad hoc jet production mechanism. Basic understanding of dynamics ===> Predictions are/will be possible.

47 2 Successful neutrino-driven explosion models of O-Ne-Mg core SNe and low-mass Fe-core SNe; Crab-like properties.

48 Neutrino-driven Explosions 8.8 Msun O-Ne-Mg core SN 9.6 Msun (z=0) Fe core SN 2D modeling by B. Müller

49 3 2D simulations show neutrino-driven explosions of Fe-core SNe. The neutrino mechanism is at least close to success!

50 4 If successful, the mechanism will be self-regulated because the heated mass will expand away from heating region close to the neutron star. Accretion supply will be quenched by accelerating explosion. Energy scale of explosion roughly set by gravitational binding energy of shells above mass cut (~ grav. binding energy of hot, white-dwarf like core) ~ ergs.

51 5 Successful neutrino-driven explosion models are compatible with nucleosynthesis constraints. No overproduction of N=50 nuclei.

52 Nucleosynthesis in NeutrinoHeated SN Ejecta 9.6 Msun (z=0) Fe core SN Convectively ejected n-rich matter makes ONeMg-core and low-mass Fe-core supernovae an interesting source of nuclei between the iron group and N = 50 (from Zn to Zr), possibly also of weak r-process nuclei. (Wanajo, THJ, Müller, ApJL 726, L15 (2011)) y p o r ent Ye n-rich matter 2D model 8.8 Msun O-Ne-Mg core SN

53 Nucleosynthesis in Neutrino-Heated SN Ejecta 15 Msun (z=solar) Fe core SN, 2D explosion model by Buras et al., A&A (2006) M(Ye < 0.47) < 10 4 Msun Pruet, Woosley, Buras, THJ, ApJ 623 (2005) 325 Fulfills requirement according to Hoffman & Woosley (ApJ, 1996)!

54 6 Neutrino-driven explosions are very promising for explaining measured NS kicks and spins.

55 Neutron Star Recoil by "Gravitational Tug-Boat" Mechanism (Wongwathanarat, Janka, Müller, ApJL 725 (2010) 106; A&A 552 (2013) A126

56 7 Neutrino-driven explosions are very promising for explaining observed SN mixing and asymmetries.

57 Mixing Instabilities in 3D SN Models L15: 15 Msun (Wongwathanarat, Müller, Janka, in preparation) B15: 15 Msun W15: 15 Msun N20: 20 Msun

58 Supernova 1987A: Bolometric Lightcurves N20 B15-1 W15 B15-2 W2-2 N20 W2-2 B15-1 W15 B15-2 (Utrobin, Wongwathanarat, Janka, Müller, in preparation)