1: %%%%%%%%%%%%%%%%%%%%%%%%
2: % for submission use:
3: %\documentclass[12pt,preprint]{aastex}
4: %%%%%%%%%%%%%%%%%%%%%%%%
5: % for preprints use:
6: \documentclass[apj]{emulateapj}
7: \usepackage{apjfonts}
8: %%%%%%%%%%%%%%%%%%%%%%%%
9:
10: \usepackage{amscd}
11: \usepackage{amsmath}
12: \usepackage{amssymb}
13: % \usepackage[T1]{fontenc}
14: % \usepackage[varg]{txfonts}
15: \usepackage{verbatim}
16: \usepackage{array}
17: \usepackage{psfrag}
18:
19: \newcommand{\myemail}{fritz@mpa-garching.mpg.de}
20:
21: \newcommand{\mbf}[1]{\ensuremath\mbox{\boldmath{$#1$}}}
22: \newcommand{\sn}{SN~Ia}
23: \newcommand{\sne}{SNe~Ia}
24: \newcommand{\gcc}{\ensuremath \, \mathrm{g} \, \mathrm{cm}^{-3}}
25:
26: \shorttitle{3D deflagration SNe Ia model}
27: \shortauthors{F.~K.~R{\"o}pke et al.}
28:
29: \begin{document}
30:
31: \title{A three-dimensional deflagration model for Type Ia supernovae
32: confronted with observations}
33:
34: \author{F. K. R{\"o}pke\altaffilmark{1,2,3}, W. Hillebrandt\altaffilmark{1,3},
35: W. Schmidt\altaffilmark{4}, J. C. Niemeyer\altaffilmark{4}, S. I.
36: Blinnikov\altaffilmark{1,5}, and
37: P. A. Mazzali\altaffilmark{1,3,6}}
38: \altaffiltext{1}{Max-Planck-Institut f\"ur Astrophysik,
39: Karl-Schwarzschild-Str. 1, D-85741 Garching, Germany}
40: \altaffiltext{2}{Department of Astronomy and Astrophysics, University
41: of California Santa Cruz, 1156 High Street, Santa Cruz, CA 95064,
42: U.S.A.}
43: \altaffiltext{3}{Kavli Institute for Theoretical Physics, University of
44: California, Santa Barbara, CA 93106}
45: \altaffiltext{4}{Lehrstuhl f{\"u}r Astronomie und Astrophysik,
46: Universit{\"a}t W{\"u}rzburg, Am Hubland, D-97074 W{\"u}rzburg, Germany}
47: \altaffiltext{5}{SSC RF ITEP, Bolshaya Cheremushkinskaya 25, 117218
48: Moscow, Russia}
49: \altaffiltext{6}{Istituto Nazionale di Astrofisica, OATs, via Tiepolo,
50: 11, 34131 Trieste, Italy}
51:
52: \begin{abstract}
53: A simulation of the thermonuclear explosion of a Chandrasekhar-mass C+O white dwarf,
54: the most popular scenario of a type Ia supernova (SN~Ia), is presented.
55: The underlying modeling is pursued in a self-consistent way, treating the combustion wave as
56: a turbulent deflagration using well tested methods developed for laboratory
57: combustion and based on the concept of `large eddy simulations' (LES).
58: Such consistency
59: requires to capture the onset of the turbulent cascade on resolved
60: scales.
61: This is achieved by computing the dynamical evolution
62: on a 1024$^3$ moving grid, which resulted in the best-resolved
63: three-dimensional SN~Ia simulation
64: carried out thus far, reaching the limits of what can be done on present
65: supercomputers.
66: Consequently, the model has no free parameters other than the
67: initial conditions at the onset
68: of the explosion, and therefore it has considerable predictive power.
69: Our main
70: objective is to determine to which extent such a simulation can account
71: for the observations of normal SNe~Ia. Guided by previous simulations with less
72: resolution and a less sophisticated flame model, initial conditions were chosen
73: that yield a reasonably strong explosion and a sufficient amount of radioactive
74: nickel for a bright display.
75: We show that observables are indeed matched to a reasonable degree.
76: In particular, good agreement is found with the light curves of normal SNe Ia.
77: Moreover, the model reproduces the general features of the abundance
78: stratification as inferred from the analysis of spectra. This indicates that
79: it captures the main features of the explosion mechanism of SNe Ia.
80: However, we also show that even a seemingly best-choice pure deflagration model
81: has shortcomings that indicate the need for a different mode of nuclear burning
82: at late times, perhaps the transition to a detonation at low density.
83: \end{abstract}
84:
85: \keywords{Stars: supernovae: general -- Hydrodynamics -- Instabilities
86: -- Turbulence -- Methods: numerical}
87:
88: \maketitle
89:
90: % begin main
91: \section{Introduction}
92: \label{sect:intro}
93:
94: Understanding the physics of Type Ia supernova (SN Ia) explosions is a necessary
95: premise for using them as cosmic distance indicators with
96: confidence. Consequently, the development of three-dimensional explosion models
97: has been a major challenge in astrophysics over the past decade. Only
98: recently, however, a self-consistent three-dimensional treatment of
99: the turbulent thermonuclear flame in a white dwarf (WD) star avoiding
100: free parameters became available.
101: The ultimate goal of
102: these efforts is to describe the full observational set of SNe Ia and even to
103: predict their observed properties, such as their luminosities, light curves,
104: and spectra, from first principles, with sufficient accuracy to guide the
105: empirical calibration of their peak luminosity.
106:
107: While this stage of sophistication is clearly not yet reached with current
108: simulations, at least one (perhaps in some cases the dominant) part of the mechanism can already
109: be treated in an non-parameterized approach. In the first stage of the supernova
110: explosion the flame propagates subsonically and is accelerated by turbulence
111: creating a problem of turbulent combustion physics. The knowledge on this
112: subject acquired for laboratory combustion in engineering sciences can be
113: transferred to astrophysics, facilitating self-consistent simulations
114: with particular emphasis on the correct representation of the flame/turbulence interaction
115: \citep{niemeyer1995b,reinecke1999a,reinecke1999b,reinecke2002d,reinecke2002b,roepke2005c,roepke2005b,roepke2006a,schmidt2006a,schmidt2006c}.
116: In fact, most of the properties of turbulent chemical flames
117: are recovered in SNe~Ia, and the largely different relevant length scales,
118: the simpler reaction kinetics, and the absence of complicated boundaries, make
119: computations of thermonuclear supernovae perhaps easier than designing a new
120: car engine, provided the initial conditions of the explosion are known.
121:
122: Recent results from several three-dimensional simulations
123: \citep{reinecke2002d,gamezo2003a,travaglio2004a,roepke2005b,bravo2006a,roepke2006a,schmidt2006a}
124: indicate that a turbulent thermonuclear deflagration---a subsonic
125: flame propagation---alone can
126: lead to powerful explosions of the WD star.
127: In combination with the elimination of free parameters in the model
128: the question arises to what extent the results of such simulations match observational
129: constraints. This question motivated several previous studies which
130: addressed initial parameters of the exploding WD
131: \citep{roepke2004c,travaglio2005a,roepke2006b}, the flame ignition
132: configuration \citep{reinecke2002d,roepke2006a,schmidt2006a}, symmetry constraints
133: in the simulation setup \citep{roepke2005b}, and the particular choice of the turbulence
134: model \citep{schmidt2006c}, as well as the impact of some initial condiditions
135: on observational results \citep{blinnikov2006a}.
136:
137: One question left out until now, however, was whether the results would
138: change significantly with higher resolution. The flame/turbulence
139: interaction determines the propagation of the burning front and thus
140: the results of the model with respect to nucleosynthesis and
141: energetics. Given the expected strong turbulence effects and the
142: associated vast range in relevant length scales this endeavor may
143: seem hopeless. However, as discussed below, with the numerical
144: modeling strategy pursued here, it turns out to be sufficient to
145: resolve the onset of turbulence on the computational grid in order to
146: recover a definitive result. The necessary resolution can be
147: reached with current computational resources and
148: marks the aim of the present study.
149: A direct comparison with observational data is possible and
150: is indeed pursued in this publication for one particular simulation in
151: the framework of the pure deflagration model.
152: With what is so far the best
153: resolved three-dimensional simulation of the deflagration scenario for SNe Ia
154: we assess its consistency and its compatibility with observations.
155: As we show in this paper, such a state-of-the-art deflagration model
156: matches observational data surprisingly well, although not all details
157: are reproduced perfectly. As is argued below, the simulation reaches a state
158: where the physical effects are resolved so that its predictions are
159: expected to be robust and to reflect generic properties of the deflagration
160: scenario.
161:
162: Of course, a single simulation cannot provide the basis for general
163: conclusions on the validity of the deflagration model. Here, we
164: present a simulation performed within the range of realistic
165: initial conditions (as far as known) but explicitly set up the initial flame
166: configuration in a favorable way with regard to energy and
167: $^{56}$Ni production.
168: Combining this single realization of the deflagration model with the
169: exploration of the parameter space carried out
170: previously (and partially with simpler modeling approaches and less resolved
171: simulations), successes and limitations of the deflagration model
172: become visible and will be discussed accordingly (see Sect.~\ref{sect:limits}).
173: In particular, some persisting shortcoming of the deflagration model
174: can be taken as an indication that the deflagration stage alone cannot
175: explain all aspects of SNe Ia.
176: In some cases models beyond the deflagration regime are supported by observational results
177: \citep[e.g.][]{mazzali2007a,wang2007a}, but uncertain or even hypothetical
178: physical mechanisms generally introduce free parameters into the models and
179: make a face-to-face comparison with observations more difficult than
180: in the case of the pure deflagration model.
181:
182:
183: \section{Astrophysical model}
184: \label{sect:model}
185:
186: Recent observations \citep[e.g.][]{mazzali2007a} support the scenario that the
187: majority of SNe~Ia originate from a carbon-oxygen WD disrupted by a
188: thermonuclear explosion \citep{hoyle1960a} when it approaches the
189: Chandrasekhar mass limit by
190: mass accretion from a binary companion which may perhaps be a
191: non-degenerate star (\citealp{ruiz-lapuente2004a}; for alternative
192: progenitor scenarios see e.g.\ \citealp{hillebrandt2000a}). Caused by
193: temperature fluctuations inside the convective core of the electron-degenerate
194: WD, small regions near its center are thought to commence a thermonuclear
195: runaway with rapidly increasing nuclear energy release from burning the
196: carbon-oxygen fuel to heavier elements. This process marks the birth of the
197: thermonuclear flame. The exact number and distribution of ignition kernels is
198: not well determined yet, implying that the largest uncertainty lies in the
199: initial parameters of the explosion. The virtue of three-dimensional
200: explosion models is that they directly relate the ignition conditions to the
201: outcome of the explosion without introducing additional
202: parameters. Thus a comparison with observations
203: sheds light on two aspects of the supernova problem---the progenitor evolution
204: giving rise to these initial conditions and the explosion mechanism itself.
205:
206: Here, we are concerned with the latter, for which several possibilities have
207: been suggested. Hydrodynamically, two fundamentally distinct modes of flame
208: propagation are admissible: a subsonic deflagration and a supersonic
209: detonation. A prompt detonation would incinerate the entire star at the high
210: initial densities, leading to conditions where nuclear reactions produce almost
211: exclusively iron group elements \citep{arnett1971a}. These are found to be
212: abundant in the ejecta, predominantly produced as radioactive $^{56}$Ni.
213: In its decay to $^{56}$Co (and later $^{56}$Fe) gamma-rays and positrons are
214: emitted which down-scatter to optical wavelengths in the SN ejecta and
215: give rise to the observed optical display. At later times, in the
216: so-called ``nebular'' spectra, this iron can be directly observed. However,
217: failure to produce the intermediate mass elements (e.g.\ silicon, calcium, and
218: sulfur) observed with appreciable abundances in the early spectra of SNe~Ia
219: \citep[for a review see][]{filippenko1997a} rules out the prompt
220: detonation mechanism.
221: In contrast, a subsonic deflagration leaves enough time for the stellar material
222: to expand before it is reached by the thermonuclear flame
223: \citep{nomoto1976a}. Therefore, some
224: burning takes place at densities low enough such that mostly intermediate mass
225: elements are synthesized, while most $^{56}$Ni is produced during
226: the early deflagration phase.
227:
228: A laminar
229: deflagration wave, mediated by microphysical heat transport, would propagate
230: far too slowly to explode the WD. However, as burning ignites near the
231: WD's center and propagates outward, a stratification of low-density ashes
232: below denser fuel emerges which is
233: buoyancy-unstable in the gravitational field of the star. This
234: configuration eventually leads to the formation of ascending burning
235: bubbles. Their interfaces are subject to shear. Since typical Reynolds
236: numbers of the flow are of the order of 10$^{14}$, strong turbulent eddies must
237: form which decay to smaller scales in a turbulent cascade.
238: This has tremendous impact on the flame propagation. The flame surface is
239: corrugated by turbulent velocity
240: fluctuations and its area increases dramatically, which enhances the
241: burning rate. This acceleration of the flame speed and the
242: fuel consumption rate enable
243: explosions of the WD \citep[e.g.][]{reinecke2002d}.
244:
245: The energy input into turbulent eddies is due to the
246: Rayleigh-Taylor instability on the largest scales
247: ($\sim$$100\,\mathrm{km}$). However, the flame
248: interacts with turbulent velocity fluctuations down to much smaller
249: scales. As demonstrated in Sect.~\ref{sect:turb} below, these velocity
250: fluctuations originate from an inertia-driven turbulent cascade. They
251: do not obey a simple buoyancy-dominated scaling and therefore require
252: special modeling effort.
253:
254:
255: \section{Numerical implementation}
256: \label{sect:num}
257:
258: Numerical schemes to model the propagation of a deflagration front in a SN~Ia
259: have been developed over the past years and have reached a high level of
260: sophistication.
261: The major challenge in the numerical implementation of the concept
262: outlined in Sect.~\ref{sect:model} is posed by the vast
263: range of relevant scales. It spans about
264: 11 orders of magnitude, from the radius of the WD ($\sim 2000 \,
265: \mathrm{km}$ and expanding in the explosion) down to the Kolmogorov scale
266: (less than a millimeter) where turbulent energy is
267: dissipated. The flame interacts with
268: turbulence down to the Gibson scale at which the laminar flame
269: propagation speed becomes comparable to the turbulent velocity
270: fluctuations. For most parts of the supernova explosion this scale is
271: considerably larger than the flame width ($\lesssim 1\,
272: \mathrm{mm}$). Therefore, the interaction of the flame with turbulence is
273: purely kinematic and burning takes place in the so-called \emph{flamelet
274: regime of turbulent combustion} \citep[e.g.][]{peters2000a}. Here, the flame propagation speed
275: completely decouples from the microphysics of the burning and is
276: entirely determined by turbulence effects. This also implies that the
277: flame propagation does not scale with the laminar burning speed.
278:
279: The challenge of scales can be tackled in an elegant way by employing
280: a \emph{large eddy simulation} (LES) strategy. This approach resolves only
281: the largest scales of the
282: turbulent cascade and models the unresolved scales, so that it becomes
283: possible to accommodate the entire WD on the computational
284: domain. When seen
285: from the scale of the WD, the flame appears as a sharp discontinuity
286: separating the fuel from the ashes. A suitable numerical scheme to
287: track the propagation of such an interface is the level set method
288: \citep{osher1988a,smiljanovski1997a,reinecke1999a}, where the flame
289: front is associated with the zero level set of a signed
290: distance function $G$ set up to be negative in the fuel regions and positive in
291: the ashes. The propagation velocity of this interface is
292: given by the effective turbulent burning speed at the grid scale which
293: in the flamelet regime is completely determined by turbulent velocity fluctuations on that
294: scale. These, in turn, are derived from a subgrid-scale turbulence model
295: \citep[for details see][]{schmidt2006b,schmidt2006c} which follows the properties of
296: turbulence on
297: unresolved scales based on filtering the velocity fluctuations on the
298: resolved scales. This way, a localized treatment of the turbulent
299: properties of the flow becomes feasible and the flame/turbulence
300: interaction is modeled self-consistently.
301:
302: The hydrodynamics is described by the reactive Euler
303: equations (with gravity as external force) and
304: numerically---discretized on a Cartesian computational grid---treated
305: in a finite
306: volume approach \citep{fryxell1989a}. Of course, the tremendous
307: expansion of the WD during the
308: explosion has to be taken into account. This was achieved by employing
309: two nested moving grids---a fine uniform inner grid tracking the
310: flame and a coarse outer grid capturing the WD
311: \citep{roepke2005c,roepke2006a}.
312:
313: With this implementation we performed what is thus far the largest simulation
314: of a thermonuclear supernova explosion. It was set up on more than a
315: billion computational grid cells (a $1024^3$ cells grid). This allowed for
316: a very fine resolution (down to $\sim 870\, \mathrm{m}$) of the central
317: part of the WD at the beginning of the explosion facilitating an
318: exceptionally
319: fine-structured initial flame configuration that was composed of 1600
320: spherical kernels of radius $2.6 \, \mathrm{km}$ distributed inside a
321: sphere extending $180 \, \mathrm{km}$ around the center of the WD (see
322: Fig.~\ref{fig:evo}, upper left panel). Although
323: the details of the ignition process are yet unknown, analytical and
324: numerical studies support such a multi-spot ignition configuration
325: \citep[e.g.][]{garcia1995a,woosley2004a} and the flame ignition kernels are
326: expected to be of the order of a kilometer in diameter \citep{iapichino2006a}.
327: The configuration chosen here is certainly one that
328: favors burning near the center and \citep[as multi-spot ignition scenarios
329: do in general,][]{roepke2006a} enhances the burning and energy release over that found
330: in centrally or sparsely ignited scenarios.
331:
332: The nuclear reactions were treated in the approximative approach outlined by
333: \citet{reinecke2002b}. Only five species were taken into account: the
334: fuel consisted of a mixture of carbon and oxygen in equal amounts by mass,
335: intermediate mass elements were represented by magnesium, and a mixture
336: of $^{56}$Ni and alpha-particles represented nuclear statistical
337: equilibrium (NSE) material. Depending on the fuel density ahead of the
338: flame, $\rho_\mathrm{u}$, the material was converted to NSE (if
339: $\rho_\mathrm{u} > 5.25 \times 10^{7} \, \mathrm{g} \,
340: \mathrm{cm}^{-3}$), in which case the nickel/alpha ratio was adjusted
341: according to temperature and density, to intermediate mass elements (if
342: $5.25 \times 10^{7} \, \mathrm{g} \,
343: \mathrm{cm}^{-3} > \rho_\mathrm{u} > 5.0 \times 10^{5} \, \mathrm{g} \,
344: \mathrm{cm}^{-3}$), or left unburned (for lower fuel densities). The
345: threshold for burning to intermediate mass elements was chosen lower
346: than in previous simulations. As argued by \citet{schmidt2007a}, the
347: level-set description of the flame propagation may be extended beyond the
348: onset of distributed burning at which turbulent eddies penetrate the
349: internal structure (the ``thin reaction zones regime''). This
350: commences as soon as the flame width becomes
351: larger than the Gibson scale---an effect that is expected to occur at
352: fuel densities $\sim$$10^{7} \, \mathrm{g} \,
353: \mathrm{cm}^{-3}$. However, at densities below $\sim$$10^{6}
354: \, \mathrm{g} \, \mathrm{cm}^{-3}$ nuclear reactions will be slow
355: compared to the expansion time scales and, moreover, turbulence will
356: interact with the burning zone inside the flame structure giving rise
357: to the completely distributed (``broken reaction zones'')
358: regime. Therefore, the burning description adopted here may be too
359: optimistic at low densities and over-predict the total amount of
360: intermediate mass elements synthesized in the explosion. On the other
361: hand, at the
362: late stages where low fuel densities are reached, the amplitude of turbulent
363: velocity fluctuations decreases rapidly. Therefore not much material
364: is burned here, such that the fuel consumption at very low densities
365: becomes insignificant and the potential error introduced by
366: uncertainties is expected to be small.
367:
368: \section{Evolution of the explosion phase}
369: \label{sect:evo}
370:
371: \begin{figure*}
372: \plotone{f1.eps}
373: \caption{Evolution of the thermonuclear supernova explosion
374: simulation. The zero level set associated with the thermonuclear
375: flame is shown as a blue isosurface and the extent of the WD is
376: indicated by the volume rendering of the density. The upper left
377: panel shows the initial set up and the close-up illustrates the
378: chosen flame ignition configuration. The subsequent two panels illustrate
379: the propagation of the turbulent flame through the WD and the
380: density structure of the remnant is shown in the lower right panel. \label{fig:evo}}
381: \end{figure*}
382:
383: The evolution of the explosion stage proceeds in agreement with the
384: expectations outlined in Sect.~\ref{sect:intro}. After ignition in a
385: multitude of only partially overlapping flame kernels (see upper left
386: panel of Fig.~\ref{fig:evo}), the
387: ash region grows due to burning and starts floating towards the
388: surface. The individual kernels deform from their initially spherical
389: shape into ``mushroom cap''-like structures---typical for the
390: non-linear stage of the Rayleigh-Taylor instability---and grow. This causes
391: them to merge, so that a
392: connected flame structure forms. This structure evolves dominated by
393: buoyancy instability on large scales and driven by the turbulent
394: cascade on smaller lengths. The
395: continued development of substructure and the merger of features create a
396: deflagration structure with a complex pattern (see upper right panel of
397: Fig.~\ref{fig:evo}). Burning and flotation drive
398: the flame towards the surface of the WD. The fuel density drops as the
399: flame moves outwards and due to radial stratification and the overall expansion of the WD caused
400: by the energy deposit from nuclear burning. Once the fuel density
401: falls below the threshold for the production of intermediate mass
402: elements, nuclear burning ceases. This occurs first
403: at the leading features of the flame and subsequently in more central
404: flame regions. Finally, no burning takes place anymore. At this point
405: the outer ash features have reached the surface layers of the
406: ejecta (lower left panel of Fig.~\ref{fig:evo}). The following
407: seconds in the evolution are characterized by the hydrodynamical
408: relaxation towards homologous expansion. This stage is reached to good
409: approximation at about $10 \, \mathrm{s}$ after ignition \citep{roepke2005c}.
410:
411: This three-dimensional evolution leads to a remnant of the explosion
412: with characteristic properties. The density structure has patterns from
413: unstable and turbulent flame propagation imprinted on it (cf.\ lower
414: right panel of Fig.~\ref{fig:evo}) and ash regions extend to
415: the outermost layers of the expanding cloud of gas.
416:
417: The chemical composition of the ejecta of the explosion as followed in the
418: hydrodynamical simulation is shown in Fig.~\ref{fig:comp3d}. The
419: outer parts consist mainly of unburned carbon and oxygen, but some
420: ``bubbles'' of intermediate mass elements are also present near the
421: surface (Fig.~\ref{fig:comp3d} left). The central part of the ejecta
422: is dominated by iron group
423: elements (Fig.~\ref{fig:comp3d} right). Inside the ash bubbles some
424: mixing of different species is
425: found, but still regions dominated by either iron group elements or
426: intermediate mass elements can be identified. The large-scale bubbles
427: of ash are separated by a few thin but dense fingers of unburned
428: material so that the spherically averaged composition of the ejecta
429: appears strongly mixed.
430:
431: \begin{figure*}
432: \plotone{f2.eps}
433: \caption{Composition of the remnant after the explosion
434: ($t=10\,\mathrm{s}$). The products of mass fractions of three of the species followed in the
435: hydrodynamical explosion simulation with density are volume rendered in
436: different colors. The right panel shows a cut-away to illustrate
437: the composition of the central parts. \label{fig:comp3d}}
438: \end{figure*}
439:
440: \section{Consistency of the subgrid-scale modeling}
441: \label{sect:turb}
442:
443: \begin{figure}
444: \plotone{f3.eps}
445: \caption{Compensated energy spectrum functions derived from our
446: simulation at different instants in time. \label{fig:turb}}
447: \end{figure}
448:
449: The subgrid-scale model employed in the simulation couples the
450: dynamics at unresolved scales to the numerically computed flow. The
451: model is based on the balance equation for the subgrid-scale
452: turbulent energy, where the energy production rate follows from a
453: similarity assumption linking the smallest resolved to the largest
454: unresolved scales \citep{schmidt2006b}. Although the
455: calculation of the rate of turbulence energy production is localized
456: and, consequently, does not rely on stationarity or large-scale
457: homogeneity, asymptotic isotropy towards smaller length scales is
458: required for consistency. In the case of incompressible turbulence,
459: asymptotic isotropy becomes manifest in the Kolmogorov $k^{-5/3}$
460: scaling for sufficiently large wavenumbers $k$. This scaling has been
461: found in simulations of Rayleigh-Taylor unstable stratifications
462: \citep{cabot2006a} as well as in buoyancy unstable flames in
463: degenerate WD matter \citet{zingale2005a}.
464:
465: Fig.~\ref{fig:turb} shows compensated energy spectrum functions,
466: $k^{5/3}E(k)$, derived from our simulation for several instants of time
467: ranging from $0.5$ to
468: $1.0\,\mathrm{s}$ after ignition. For turbulent velocity fields
469: obeying Kolmogorov
470: scaling, the compensated spectrum functions are expected to be
471: constant. Wavenumbers are normalized by $\pi/(512\Delta_{0})$, where
472: $\Delta_{0}$ is the cell size in the uniform part of the grid. Since
473: the grid is inhomogeneous and does not satisfy periodic boundary
474: conditions, the velocity fields obtained from the simulation were
475: zero-padded in the outer regions with Gaussian windowing functions in
476: order to compute Fourier transforms. This is why the compensated
477: spectrum functions are plotted for normalized wavenumbers greater than
478: $16$ only. At all times considered here, it can be seen that the
479: spectrum functions are well approximated by the Kolmogorov power law
480: for intermediate wave numbers, thus indicating the presence of an
481: inertial subrange. Therefore, the simulation is as well resolved as
482: the modeling approach requires. Higher resolution is not expected to
483: change the characteristics of the turbulent flame propagation.
484:
485: \section{Results and comparison with observations}
486: \label{sect:results}
487:
488: The present deflagration simulation of a thermonuclear supernova makes
489: distinct predictions for observables that can be tested against
490: real SNe Ia. With certain observables like polarimetry spectra and
491: spectra of the nebular phase it may be possible to check details of
492: the results of pure deflagration models against observations and
493: potentially assess its validity. This certainly warrants further
494: attention and will be addressed in future work. Here, we are concerned
495: with some more basic global properties which test whether our
496: particular simulation falls into the range of observations.
497:
498: The asymptotic kinetic energy amounted to $8.1 \times
499: 10^{50}\,\mathrm{erg}$. In the hydrodynamical simulation, $0.606\, M_\odot$ of
500: iron group elements and $0.431\, M_\odot$ of intermediate mass
501: elements were produced, leaving
502: behind 0.368 $\, M_\odot$ of unburned carbon and oxygen.
503:
504: \begin{figure}
505: \plotone{f4.eps}
506: \caption{Spherically averaged composition resulting from the
507: hydrodynamical explosion simulation (solid lines) compared to the
508: findings of the abundance tomography of SN~2002bo (dotted lines). Iron group
509: element abundances are shown in red, intermediate mass elements in
510: green and unburned material in blue. \label{fig:at}}
511: \end{figure}
512:
513: The composition of the ejecta can be directly compared with that of observed
514: supernovae obtained from abundance tomography \citep{stehle2005a} using a
515: spherical average over the entire ejecta of the model in homologous expansion.
516: Such a comparison with the abundance tomography of SN~2002bo is shown in
517: Fig.~\ref{fig:at}, where the velocity is used as radial
518: coordinate. Two cautionary remarks are necessary here. First, SN~2002bo was a
519: much more energetic and luminous event than that expected to result from
520: the simulation. Second, as pointed out in Sect.~\ref{sect:evo}, there
521: exist features with high concentrations of particular species (see
522: Fig.~\ref{fig:comp3d}) large enough to dominate spectral
523: observations. The procedure of spherically averaging the mass fractions
524: washes out these structures and artificially increases the mixing of species.
525: Nonetheless, the chemical composition of the
526: ejecta of the simulated explosion and the abundance tomography of
527: SN~2002bo compare surprisingly well. In both cases, the central parts are
528: clearly dominated by iron group elements and the overall distribution of the
529: main species in our simulation generally agrees with that obtained from the
530: abundance tomography.
531: There is good agreement up to velocities of $\sim$$10,000 \, \mathrm{km}\,
532: \mathrm{s}^{-1}$, if we keep in mind that SN~2002bo had a $^{56}$Ni content
533: of $\sim 0.5\, M_\odot$, which is significantly larger than in the simulation.
534: The slope of the decline of the iron group abundance
535: towards larger radii is also well-reproduced. These elements are mixed out
536: to the same radii in the simulation and in the observation. Moreover,
537: both agree in intermediate
538: mass elements being more abundant than unburned material below
539: $\sim$$10,000 \, \mathrm{km}\, \mathrm{s}^{-1}$ and mixed compositions
540: over a wide range of velocities. The abundance of unburned material in
541: the simulation is low in this velocity range, in particular near the
542: center. As previously pointed out by
543: \citet{kozma2005a}, the presence of large amounts of unburned fuel at
544: low velocities gives rise to a pronounced oxygen line in the late-time
545: ("nebular") spectra which is inconsistent with observations. Owing to
546: the more complete burning of the center with the chosen flame ignition
547: condition, this problem is alleviated in the current simulation. The
548: abundance of fuel material near the center is lower by a factor of 10 to 50
549: than the iron group mass fraction here.
550:
551: Still, the abundance
552: of carbon and oxygen seems rather high in comparison with the
553: observation. This may in part be an
554: artifact of the spherical averaging procedure for determining the
555: composition of the ejecta in the simulation.
556: overestimating the contribution of the dense but narrow fuel regions
557: (see Fig.~\ref{fig:comp3d}). As the volume of ejecta that influences
558: the spectra taken form a certain line of sight varies with time, a
559: careful analysis of this effect would require the derivation of spectra
560: from our simulation in a three-dimensional approach.
561:
562: At velocities above $\sim$$10,000 \, \mathrm{km}\,\mathrm{s}^{-1}$,
563: however, there is a significant discrepancy between the abundance
564: tomography data of SN 2002bo and the simulated composition. While in the
565: former the intermediate mass elements dominate out to high velocities,
566: the ejecta in the simulation quickly become dominated by unburned material. This
567: discrepancy, however, is only found in the outer $0.25\,M_\odot$.
568:
569: In order to derive light curves from the simulation, more detailed
570: information about the chemical composition of the ejecta is necessary
571: than the data obtained directly from the hydrodynamical explosion
572: simulation. This information can be derived in a nucleosynthetic
573: postprocessing step \citep{travaglio2004a} on the basis of
574: $53^3$ tracer particles that have been
575: advected in the flow. The mass of the star was equally distributed
576: between those tracers and they add a "Lagrangian component" to the
577: otherwise Eulerian approach. Recording representative temperature and
578: density profiles they facilitate the \emph{a posteriori}
579: reconstruction of the nuclear reactions with a large reaction network
580: (which would be impractical to employ concurrently with the explosion
581: simulation). This way, the chemical
582: composition of the ejecta as needed for the radiative transfer is
583: determined. In particular, it is necessary to separate the
584: $^{56}$Ni, which by radioactive decay provides the energy for the
585: optical event, from the stable iron group isotopes that do not
586: contribute to the luminosity but are important for absorption. The
587: postprocessing yields $0.56 \, M_\odot$ of iron group elements ($0.05
588: \, M_\odot$ less than the estimate from the hydrodynamical simulation)
589: and $0.33 \, M_\odot$ of $^{56}$Ni. This indicates that the simulated
590: supernova would range among the lower end of observed normal
591: SNe~Ia. For events like SN~1991M, SN~1993L, SN~2004eo, and SN~2001el
592: similar masses of $^{56}$Ni and total iron group elements were
593: inferred by \citet{mazzali2007a}.
594:
595: On the basis of the ejecta composition as derived from the
596: postprocessing step and spherically averaged over the entire star,
597: bolometric light curves (true bolometric as well as
598: ``UVOIR-bolometric'') have been derived with the radiative transport
599: code \textsc{Stella}, as described in detail by
600: \citet{blinnikov2006a}. The result is shown in Fig.~\ref{fig:lc_bol}
601: and compared to the bolometric light curves of standard SNe Ia as
602: given by \citet{stritzinger2006a}. Evidently, the synthetic light
603: curve derived from the simulation is a normal SN Ia light curve both in
604: terms of the peak luminosity and the shape. It is slightly dimmer than
605: SN 1992A, as expected from the lower $^{56}$Ni mass
606: (\citealt{stritzinger2006a} report $0.4\, M_\odot$
607: of $^{56}$Ni for SN~1992A), but has a very similar rise and decline
608: rate.
609:
610: \begin{figure}
611: \plotone{f5.eps}
612: \caption{Bolometric light curve derived for our model (black curves;
613: solid is the ``UVOIR-bolometric'' light curve and the complete
614: bolometric light curve is dotted.) The blue dotted curves correspond
615: to observed bolometric light curves \citep{stritzinger2006a}. \label{fig:lc_bol}}
616: \end{figure}
617:
618: \section{Limits of the Pure Deflagration model}
619: \label{sect:limits}
620:
621: The limited success of the presented simulation in reproducing the
622: observables of SNe Ia raises the question of whether it is possible to
623: assess generic limits of the pure deflagration model on this basis.
624:
625: As a realization in a single
626: simulation does not cover the full parameter space, this has to be
627: discussed in combination with simulations performed
628: previously.
629: \citet{reinecke2002d,roepke2004c,travaglio2005a,roepke2006b,roepke2006a},
630: and \citet{schmidt2006a}
631: explored the impact of various initial conditions on the explosion
632: process. These simulations were not performed with the same
633: resolution as the one presented here and are partially based on the
634: simpler subgrid-scale turbulence model proposed by
635: \cite{niemeyer1995b}. Still these results can guide an assessment of
636: the deflagration model since the changes between the two subgrid-scale
637: models are moderate \citep[at most $~$30\% in the iron group element
638: production and the energy release;][]{schmidt2006c}.
639:
640: The two main shortcomings identified in Sect.~\ref{sect:results} were
641: (i) the overall low explosion strength resulting in a low $^{56}$Ni
642: production and a low explosion energy and (ii) the underproduction of
643: intermediate mass elements in the outer layers of the ejecta.
644:
645: Problem (i) has been noted in all deflagration simulations based on a
646: consistent approach of modeling the flame propagation. There are,
647: however, several initial parameters that may remedy the problem. These
648: have been tested in previous simulations.
649:
650: Clearly, a major parameter
651: is the geometrical configuration of the flame ignition. With simple
652: central ignitions (``C3''-setups of
653: \citealp{reinecke2002d,roepke2005b}) the asymptotic kinetic energy of
654: the ejecta falls around $4.5 \times 10^{50}\, \mathrm{erg}$ and the
655: masses of produced iron group elements hardly exceed $0.5\,M_\odot$
656: with $\sim 0.3\, M_\odot$ of it being $^{56}$Ni. This would account
657: for the dimmest examples among the ``normal'' SNe Ia. Ignition in
658: multiple spots around the center evidently increases the explosion
659: strength. For a model ignited in 30 bubbles per octant,
660: \citet{travaglio2004a} report an asymptotic kinetic energy of about
661: $6.5 \times 10^{50} \,\mathrm{erg}$ and a $^{56}$Ni production of
662: $0.418\,M_\odot$. \cite{roepke2006a} explored the impact of multi-spot
663: ignitions in a systematic approach and found an optimum of such
664: configurations resulting in an asymptotic kinetic energy of $6.8\times
665: 10^{50}\,\mathrm{erg}$ and $0.67\,M_\odot$ of iron group elements. The
666: latter value may increase in modeling the ignition as a stochastic
667: process in time to $\sim 0.75 \, M_\odot$ \citep{schmidt2006a}. Given these values, the
668: choice of the ignition configuration in the present model was
669: moderately optimistic producing $0.606\,M_\odot$ of iron
670: group elements, about $0.33\,M_\odot$ of which is $^{56}$Ni.
671: A large ratio of produced stable to radioactive iron
672: group elements (here it amounts to 0.59) seems to be a general feature
673: of turbulent deflagration
674: models ignited in a multitude of ignition kernels since most of the
675: iron group synthesis proceeds here at the early stages where high
676: densities prevail such that electron capture reactions are
677: significant.
678:
679: Other parameters that have an impact on the $^{56}$Ni production and
680: the energy release are the central density of the WD at ignition and
681: its metallicity. A higher central density naturally increases the
682: amount of material available that potentially can burn to iron group
683: elements. Moreover, the increased gravitational acceleration will
684: enhance the buoyancy-induced turbulence and accelerate the flame. On
685: the other hand, with increased densities electron captures become more
686: important and will favor the production of stable iron group
687: isotopes over $^{56}$Ni. The effect of varying the central density has
688: been addressed by \citet{roepke2006b}. In a density range where
689: electron captures are still moderate, increasing the central
690: density by a factor of 2.6 changes the $^{56}$Ni yield only by less
691: than 10\% although the total iron group element synthesis was enhanced
692: by $\sim$30\%. Therefore, although energetic results and enhanced production
693: of the iron group are expected with higher central densities of the
694: exploding WD star, as electron capture reactions become
695: important the fraction of $^{56}$Ni in the NSE material and thus the
696: brightness of the corresponding events will decrease. An
697: overproduction of stable iron group nuclei is anticipated here.
698: Turning this argument around, a way to obtain higher $^{56}$Ni masses
699: (although the total mass of iron group and intermediate mass elements
700: would not increase) would be to consider a WD with lower central
701: density ignited in a multi-spot scenario. This issue will be addressed in a forthcoming study, but as
702: the results of \citet{roepke2006b} indicate the effect is not expected to be
703: sufficient to account for the brighter observed SN Ia events.
704:
705: The metallicity of the WD material will have little effect on the iron
706: group production and the energy release, but evidently affects the
707: ratio of stable to radioactive iron group element masses. Increasing
708: the metallicity favors the production of neutron-rich stable nuclei
709: over $^{56}$Ni \citep{timmes2003a,travaglio2005a,roepke2006a} and
710: makes the corresponding event dimmer. The $^{56}$Ni masses decrease
711: linearly with metallicity and the values given above were derived
712: under the assumption of solar metallicity. Thus, metallicity effects
713: have no potential of increasing the $^{56}$Ni yields of the
714: deflagration model significantly.
715:
716: Issue (ii) is equivalent with the incomplete burning of the outer
717: layers of the WD star. Here, the problem of modeling the flame
718: propagation at conditions where the flamelet picture breaks down
719: becomes important. It has beens shown that a flame acceleration in
720: this phase would significantly enhance the production of intermediate
721: mass elements and the energy release \cite{roepke2005d}, but
722: \citet{schmidt2007a} argues that such an acceleration does not
723: occur and that a continuation of modeling the flame propagation as in
724: the flamelet regime is possible. This approach was followed here,
725: too. Therefore, within the pure turbulent deflagration regime it
726: seems unrealistic to enhance burning at low densities beyond what was
727: achieved in the current simulation.
728:
729: We thus conclude that the initial parameters of the model
730: are unlikely to leave enough room to adjust the simulations in order
731: to account for the more powerful SN Ia events and to reproduce the
732: chemical composition in the outer layers of the ejecta. As the model
733: of flame propagation itself leaves no free parameters, only such initial
734: parameters of the WD and the flame ignition come into consideration
735: here. These have been tested and adjusting them in a favorable way
736: allows to enter the range of the weaker ``normal'' SNe
737: Ia in terms of $^{56}$Ni production, explosion energy, and composition
738: of the inner part of the ejecta. The more powerful observed events,
739: producing close to $1 \, M_\odot$ of $^{56}$Ni
740: \citep[e.g.][]{mazzali2007a} are clearly out of reach for the pure
741: deflagration scenario of SNe Ia.
742:
743: \section{Conclusions}
744:
745: We presented a three-dimensional deflagration simulation of a thermonuclear
746: supernova explosion with the astrophysical modeling and the numerical
747: implementation designed in a way to allow for self-consistent
748: simulation. The model contains no tunable parameters except for the
749: flame ignition condition, which is physically not well determined yet,
750: and the initial conditions of the WD which may vary from event to
751: event in nature.
752: An additional uncertainty of the model is due to burning at the lowest
753: fuel densities which needs
754: further attention, but is expected to lead to only minor changes in the
755: conclusions about the gross properties of the simulation which we present
756: here.
757: Clearly, the most significant parameter is the flame ignition configuration
758: deciding even about the success of the
759: deflagration burning to explode the white dwarf star
760: \cite[e.g.][]{calder2004a,plewa2004a,roepke2007a}. Here, it was set up
761: in a favorable way in order to burn the fuel material at the center
762: of the star as completely as possible and to maximize the energy and
763: $^{56}$Ni production.
764:
765: The goal of the simulation was to reach a regime in which the
766: flame/turbulence interaction modeling is fully self-consistent which
767: requires to resolve the onset of the turbulent cascade on the
768: computational grid. This was achieved by performing the simulation
769: with unprecedented resolution. Such a consistent implementation of the
770: burning model allows the direct confrontation with observables. The
771: comparison with bolometric light curves and the chemical composition
772: of the ejecta derived from observed spectra showed that the simulated
773: explosion indeed reproduces the characteristics
774: of the observed fainter examples of normal SNe~Ia. Gross properties
775: of the simulations such as energy the
776: energy release and the $^{56}$Ni production were also consistent with
777: expectations from SN~Ia observations.
778: One has to note, however, that even choosing favorable
779: ignition conditions \citep[a slight improvement is possible with a
780: time-dependent stochastic distribution of ignition
781: kernels,][]{schmidt2006a}, the explosion energy and the brightness of
782: the model event fall towards the lowest values observed for normal SNe
783: Ia.
784:
785: Thus, upon comparison with previous results from lower resolved
786: simulations exploring the range of the initial parameters, we conclude that
787: within the
788: astrophysical model employed here, it seems impossible to reproduce
789: the more energetic events of the observed SN~Ia sample. Also, a smooth
790: shell of intermediate
791: mass elements surrounding the iron-rich core of the ejecta cloud and
792: the anti-correlation between the iron-group and the intermediate mass
793: elements in the ejecta \citep{mazzali2007a} are difficult to explain
794: within this framework. Therefore an extension of the scenario seems
795: necessary. The shortcomings of the model in
796: reproducing the chemical composition of the ejecta in the outer parts while
797: providing good agreement with the observations in central regions may
798: taken as an indication that the deflagration phase is
799: only the first phase of the explosion mechanism. While the reasonable
800: agreement with observables of weaker SN~Ia events indicates that it
801: may be the dominant part of the explosion mechanism here, the
802: discrepancy is obvious for the brighter parts of the SN~Ia
803: sample. Possible extensions of the model
804: include an improved description of burning in the distributed burning
805: regime at low fuel densities as well as a transition to a supersonic
806: detonation mode of flame propagation at the onset of this regime.
807: Recently, \cite{roepke2007b} modeled delayed detonation scenarios in
808: three-dimensions under the assumption that the
809: deflagration-to-detonation transition occurs as the flame enters the
810: distributed burning regime. Depending on the configuration of the
811: deflagration flame ignition, a wide range of $^{56}$Ni masses was
812: recovered which would reproduce observations of normal to bright
813: SNe~Ia. In this context, the fainter examples correspond to events
814: with a pronounced deflagration phase. For these, most of the energy
815: and $^{56}$Ni
816: production originates from that phase and the character of the events is
817: similar to pure deflagration models. Thus, our results may be
818: interpreted as an indication that the fainter sub-sample of normal
819: SNe~Ia expose the hallmark of the deflagration phase. Since the
820: strength of deflagration phase determined the overall characteristics
821: of the delayed detonation model implemented this way, a careful and
822: self-consistent treatment of the burning physics in the deflagration
823: phases is a premise for credible modeling here as well.
824:
825:
826:
827: \begin{acknowledgements}
828: The simulation presented here was carried out at the Computer Center
829: of the Max Planck Society, Garching, Germany. The authors gratefully
830: acknowledge the hospitality and the inspiring atmosphere at the Kavli
831: Institute for Theoretical Physics at the University of California
832: Santa Barbara, where the paper was written. This work was supported by
833: the European Research Training Network ``The Physics of Type Ia
834: Supernovae'' under contract HPRN-CT-2002-00303 and the Deutsche
835: Forschungsgemeinschaft via the Transregional Collaborative Research
836: Centre TRR 33 ``The Dark Universe''. F.K.R. received support from the NASA
837: Theory Program (NNG05GG08G) and from the SciDAC program of the DOE
838: (DE-FC02-01ER41176).
839: \end{acknowledgements}
840:
841: \begin{thebibliography}{44}
842: %\expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
843:
844: \bibitem[{{Arnett} {et~al.}(1971){Arnett}, {Truran}, \&
845: {Woosley}}]{arnett1971a}
846: {Arnett}, W.~D., {Truran}, J.~W., \& {Woosley}, S.~E. 1971, \apj, 165, 87
847:
848: \bibitem[{{Blinnikov} {et~al.}(2006){Blinnikov}, {R{\"o}pke}, {Sorokina},
849: {Gieseler}, {Reinecke}, {Travaglio}, {Hillebrandt}, \&
850: {Stritzinger}}]{blinnikov2006a}
851: {Blinnikov}, S.~I., {R{\"o}pke}, F.~K., {Sorokina}, E.~I., {Gieseler}, M.,
852: {Reinecke}, M., {Travaglio}, C., {Hillebrandt}, W., \& {Stritzinger}, M.
853: 2006, \aap, 453, 229
854:
855: \bibitem[{{Bravo} \& {Garc{\'{\i}}a-Senz}(2006)}]{bravo2006a}
856: {Bravo}, E. \& {Garc{\'{\i}}a-Senz}, D. 2006, \apjl, 642, L157
857:
858: \bibitem[{{Cabot} \& {Cook}(2006)}]{cabot2006a}
859: {Cabot}, W.~H. \& {Cook}, A.~W. 2006, Nature Physics, 2, 562
860:
861: \bibitem[{{Calder} {et~al.}(2004){Calder}, {Plewa}, {Vladimirova}, {Lamb}, \&
862: {Truran}}]{calder2004a}
863: {Calder}, A.~C., {Plewa}, T., {Vladimirova}, N., {Lamb}, D.~Q., \& {Truran},
864: J.~W. 2004, astro-ph/0405126
865:
866: \bibitem[{{Filippenko}(1997)}]{filippenko1997a}
867: {Filippenko}, A.~V. 1997, \araa, 35, 309
868:
869: \bibitem[{{Fryxell} {et~al.}(1989){Fryxell}, {M{\"u}ller}, \&
870: {Arnett}}]{fryxell1989a}
871: {Fryxell}, B.~A., {M{\"u}ller}, E., \& {Arnett}, W.~D. 1989, Hydro\-dynamics
872: and nuclear burning, MPA Green Report 449, Max-Planck-Institut f\"ur
873: Astrophysik, Garching
874:
875: \bibitem[{{Gamezo} {et~al.}(2003){Gamezo}, {Khokhlov}, {Oran}, {Chtchelkanova},
876: \& {Rosenberg}}]{gamezo2003a}
877: {Gamezo}, V.~N., {Khokhlov}, A.~M., {Oran}, E.~S., {Chtchelkanova}, A.~Y., \&
878: {Rosenberg}, R.~O. 2003, Science, 299, 77
879:
880: \bibitem[{{Garcia-Senz} \& {Woosley}(1995)}]{garcia1995a}
881: {Garcia-Senz}, D. \& {Woosley}, S.~E. 1995, \apj, 454, 895
882:
883: \bibitem[{{Hillebrandt} \& {Niemeyer}(2000)}]{hillebrandt2000a}
884: {Hillebrandt}, W. \& {Niemeyer}, J.~C. 2000, \araa, 38, 191
885:
886: \bibitem[{{Hoyle} \& {Fowler}(1960)}]{hoyle1960a}
887: {Hoyle}, F. \& {Fowler}, W.~A. 1960, \apj, 132, 565
888:
889: \bibitem[{{Iapichino} {et~al.}(2006){Iapichino}, {Br{\"u}ggen}, {Hillebrandt},
890: \& {Niemeyer}}]{iapichino2006a}
891: {Iapichino}, L., {Br{\"u}ggen}, M., {Hillebrandt}, W., \& {Niemeyer}, J.~C.
892: 2006, \aap, 450, 655
893:
894: \bibitem[{{Kozma} {et~al.}(2005){Kozma}, {Fransson}, {Hillebrandt},
895: {Travaglio}, {Sollerman}, {Reinecke}, {R{\"o}pke}, \&
896: {Spyromilio}}]{kozma2005a}
897: {Kozma}, C., {Fransson}, C., {Hillebrandt}, W., {Travaglio}, C., {Sollerman},
898: J., {Reinecke}, M., {R{\"o}pke}, F.~K., \& {Spyromilio}, J. 2005, \aap, 437,
899: 983
900:
901: \bibitem[{{Mazzali} {et~al.}(2007){Mazzali}, {R{\"o}pke}, {Benetti}, \&
902: {Hillebrandt}}]{mazzali2007a}
903: {Mazzali}, P.~A., {R{\"o}pke}, F.~K., {Benetti}, S., \& {Hillebrandt}, W. 2007,
904: Science, 315, 825
905:
906: \bibitem[{{Niemeyer} \& {Hillebrandt}(1995)}]{niemeyer1995b}
907: {Niemeyer}, J.~C. \& {Hillebrandt}, W. 1995, \apj, 452, 769
908:
909: \bibitem[{{Nomoto} {et~al.}(1976){Nomoto}, {Sugimoto}, \& {Neo}}]{nomoto1976a}
910: {Nomoto}, K., {Sugimoto}, D., \& {Neo}, S. 1976, \apss, 39, L37
911:
912: \bibitem[{{Osher} \& {Sethian}(1988)}]{osher1988a}
913: {Osher}, S. \& {Sethian}, J.~A. 1988, J. Comp. Phys., 79, 12
914:
915: \bibitem[{{Peters}(2000)}]{peters2000a}
916: {Peters}, N. 2000, Turbulent Combustion (Cambridge: Cambridge University Press)
917:
918: \bibitem[{{Plewa} {et~al.}(2004){Plewa}, {Calder}, \& {Lamb}}]{plewa2004a}
919: {Plewa}, T., {Calder}, A.~C., \& {Lamb}, D.~Q. 2004, \apjl, 612, L37
920:
921: \bibitem[{{Reinecke} {et~al.}(1999{\natexlab{a}}){Reinecke}, {Hillebrandt}, \&
922: {Niemeyer}}]{reinecke1999b}
923: {Reinecke}, M., {Hillebrandt}, W., \& {Niemeyer}, J.~C. 1999{\natexlab{a}},
924: \aap, 347, 739
925:
926: \bibitem[{{Reinecke} {et~al.}(2002{\natexlab{a}}){Reinecke}, {Hillebrandt}, \&
927: {Niemeyer}}]{reinecke2002b}
928: ---. 2002{\natexlab{a}}, \aap, 386, 936
929:
930: \bibitem[{{Reinecke} {et~al.}(2002{\natexlab{b}}){Reinecke}, {Hillebrandt}, \&
931: {Niemeyer}}]{reinecke2002d}
932: ---. 2002{\natexlab{b}}, \aap, 391, 1167
933:
934: \bibitem[{{Reinecke} {et~al.}(1999{\natexlab{b}}){Reinecke}, {Hillebrandt},
935: {Niemeyer}, {Klein}, \& {Gr{\" o}bl}}]{reinecke1999a}
936: {Reinecke}, M., {Hillebrandt}, W., {Niemeyer}, J.~C., {Klein}, R., \& {Gr{\"
937: o}bl}, A. 1999{\natexlab{b}}, \aap, 347, 724
938:
939: \bibitem[{{R{\"o}pke}(2005)}]{roepke2005c}
940: {R{\"o}pke}, F.~K. 2005, \aap, 432, 969
941:
942: \bibitem[{{R{\"o}pke} {et~al.}(2006{\natexlab{a}}){R{\"o}pke}, {Gieseler},
943: {Reinecke}, {Travaglio}, \& {Hillebrandt}}]{roepke2006b}
944: {R{\"o}pke}, F.~K., {Gieseler}, M., {Reinecke}, M., {Travaglio}, C., \&
945: {Hillebrandt}, W. 2006{\natexlab{a}}, \aap, 453, 203
946:
947: \bibitem[{{R{\"o}pke} \& {Hillebrandt}(2004)}]{roepke2004c}
948: {R{\"o}pke}, F.~K. \& {Hillebrandt}, W. 2004, \aap, 420, L1
949:
950: \bibitem[{{R{\"o}pke} \& {Hillebrandt}(2005{\natexlab{a}})}]{roepke2005b}
951: ---. 2005, \aap, 431, 635
952:
953: \bibitem[{{R{\"o}pke} \& {Hillebrandt}(2005{\natexlab{b}})}]{roepke2005d}
954: ---. 2005, \aap, 429, L629
955:
956: \bibitem[{{R{\"o}pke} {et~al.}(2006{\natexlab{b}}){R{\"o}pke}, {Hillebrandt},
957: {Niemeyer}, \& {Woosley}}]{roepke2006a}
958: {R{\"o}pke}, F.~K., {Hillebrandt}, W., {Niemeyer}, J.~C., \& {Woosley}, S.~E.
959: 2006{\natexlab{b}}, \aap, 448, 1
960:
961: \bibitem[{{R{\"o}pke} \& {Niemeyer}(2007)}]{roepke2007b}
962: {R{\"o}pke}, F.~K. \& {Niemeyer}, J.~C. 2007, \aap, 464, 683
963:
964: \bibitem[{{R{\"o}pke} {et~al.}(2007){R{\"o}pke}, {Woosley}, \&
965: {Hillebrandt}}]{roepke2007a}
966: {R{\"o}pke}, F.~K., {Woosley}, S.~E., \& {Hillebrandt}, W. 2007, \apj, 660,
967: 1344
968:
969: \bibitem[{{Ruiz-Lapuente} {et~al.}(2004){Ruiz-Lapuente}, {Comeron},
970: {M{\'e}ndez}, {Canal}, {Smartt}, {Filippenko}, {Kurucz}, {Chornock}, {Foley},
971: {Stanishev}, \& {Ibata}}]{ruiz-lapuente2004a}
972: {Ruiz-Lapuente}, P., {Comeron}, F., {M{\'e}ndez}, J., {Canal}, R., {Smartt},
973: S.~J., {Filippenko}, A.~V., {Kurucz}, R.~L., {Chornock}, R., {Foley}, R.~J.,
974: {Stanishev}, V., \& {Ibata}, R. 2004, \nat, 431, 1069
975:
976: \bibitem[{{Schmidt}(2007)}]{schmidt2007a}
977: {Schmidt}, W. 2007, \aap, 465, 263
978:
979: \bibitem[{{Schmidt} \& {Niemeyer}(2006)}]{schmidt2006a}
980: {Schmidt}, W. \& {Niemeyer}, J.~C. 2006, \aap, 446, 627
981:
982: \bibitem[{{Schmidt} {et~al.}(2006{\natexlab{a}}){Schmidt}, {Niemeyer}, \&
983: {Hillebrandt}}]{schmidt2006b}
984: {Schmidt}, W., {Niemeyer}, J.~C., \& {Hillebrandt}, W. 2006{\natexlab{a}},
985: \aap, 450, 265
986:
987: \bibitem[{{Schmidt} {et~al.}(2006{\natexlab{b}}){Schmidt}, {Niemeyer},
988: {Hillebrandt}, \& {R{\"o}pke}}]{schmidt2006c}
989: {Schmidt}, W., {Niemeyer}, J.~C., {Hillebrandt}, W., \& {R{\"o}pke}, F.~K.
990: 2006{\natexlab{b}}, \aap, 450, 283
991:
992: \bibitem[{{Smiljanovski} {et~al.}(1997){Smiljanovski}, {Moser}, \&
993: {Klein}}]{smiljanovski1997a}
994: {Smiljanovski}, V., {Moser}, V., \& {Klein}, R. 1997, Combustion Theory and
995: Modelling, 1, 183
996:
997: \bibitem[{{Stehle} {et~al.}(2005){Stehle}, {Mazzali}, {Benetti}, \&
998: {Hillebrandt}}]{stehle2005a}
999: {Stehle}, M., {Mazzali}, P.~A., {Benetti}, S., \& {Hillebrandt}, W. 2005,
1000: \mnras, 360, 1231
1001:
1002: \bibitem[{{Stritzinger} {et~al.}(2006){Stritzinger}, {Leibundgut}, {Walch}, \&
1003: {Contardo}}]{stritzinger2006a}
1004: {Stritzinger}, M., {Leibundgut}, B., {Walch}, S., \& {Contardo}, G. 2006, \aap,
1005: 450, 241
1006:
1007: \bibitem[{{Timmes} {et~al.}(2003){Timmes}, {Brown}, \& {Truran}}]{timmes2003a}
1008: {Timmes}, F.~X., {Brown}, E.~F., \& {Truran}, J.~W. 2003, \apjl, 590, L83
1009:
1010: \bibitem[{{Travaglio} {et~al.}(2005){Travaglio}, {Hillebrandt}, \&
1011: {Reinecke}}]{travaglio2005a}
1012: {Travaglio}, C., {Hillebrandt}, W., \& {Reinecke}, M. 2005, \aap, 443, 1007
1013:
1014: \bibitem[{{Travaglio} {et~al.}(2004){Travaglio}, {Hillebrandt}, {Reinecke}, \&
1015: {Thielemann}}]{travaglio2004a}
1016: {Travaglio}, C., {Hillebrandt}, W., {Reinecke}, M., \& {Thielemann}, F.-K.
1017: 2004, \aap, 425, 1029
1018:
1019: \bibitem[{{Wang} {et~al.}(2007){Wang}, {Baade}, \& {Patat}}]{wang2007a}
1020: {Wang}, L., {Baade}, D., \& {Patat}, F. 2007, Science, 315, 212
1021:
1022: \bibitem[{{Woosley} {et~al.}(2004){Woosley}, {Wunsch}, \&
1023: {Kuhlen}}]{woosley2004a}
1024: {Woosley}, S.~E., {Wunsch}, S., \& {Kuhlen}, M. 2004, \apj, 607, 921
1025:
1026: \bibitem[{{Zingale} {et~al.}(2005){Zingale}, {Woosley}, {Rendleman}, {Day}, \&
1027: {Bell}}]{zingale2005a}
1028: {Zingale}, M., {Woosley}, S.~E., {Rendleman}, C.~A., {Day}, M.~S., \& {Bell},
1029: J.~B. 2005, \apj, 632, 1021
1030:
1031: \end{thebibliography}
1032:
1033:
1034: \end{document}
1035: