1: %%%%%%%%%%%%%%%%%%%%%%%%
2: % for submission use:
3: %\documentclass[12pt,preprint]{aastex}
4: %%%%%%%%%%%%%%%%%%%%%%%%
5: % for preprints use:
6: \documentclass[apjl]{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:
20: \newcommand{\myemail}{fritz@mpa-garching.mpg.de}
21:
22: \newcommand{\gcc}{\ensuremath \, \mathrm{g} \, \mathrm{cm}^{-3}}
23:
24: \shorttitle{Flame-driven deflagration-to-detonation transitions in Type Ia supernovae?}
25: \shortauthors{F.~K.~R{\"o}pke}
26:
27: \begin{document}
28: \journalinfo{The Astrophysical Journal, \normalfont{668:1103--1108, 2007 October 20}}
29: \title{Flame-driven deflagration-to-detonation transitions in Type Ia supernovae?}
30:
31: \author{F. K. R{\"o}pke\altaffilmark{1}}
32: \affil{Max-Planck-Institut f\"ur Astrophysik,
33: Karl-Schwarzschild-Str. 1, D-85741 Garching, Germany}
34: \altaffiltext{1}{Also at: Department of Astronomy and Astrophysics, University
35: of California Santa Cruz, 1156 High Street, Santa Cruz, CA 95064,
36: U.S.A.}
37: \begin{abstract}
38: Although delayed detonation models of thermonuclear explosions of white
39: dwarfs seem promising for reproducing Type Ia supernovae, the
40: transition of the flame propagation mode from subsonic deflagration to
41: supersonic detonation remains hypothetical. A potential instant for
42: this transition to occur is the onset of the distributed burning
43: regime, i.e.\ the moment when turbulence first affects the internal
44: flame structure. Some studies of the burning microphysics indicate that
45: a deflagration-to-detonation transition may be possible here, provided
46: the turbulent intensities are strong enough. Consequently, the
47: magnitude of turbulent velocity fluctuations generated by the
48: deflagration flame is analyzed at the onset of the distributed burning regime in
49: several three-dimensional simulations of deflagrations in
50: thermonuclear supernovae. It is shown that the corresponding probability
51: density functions fall off towards high turbulent velocity
52: fluctuations much more slowly than a Gaussian distribution. Thus, values claimed to be
53: necessary for triggering a detonation are likely to be found in
54: sufficiently large patches of the flame. Although the microphysical
55: evolution of the burning is not followed and a successful
56: deflagration-to-detonation transition cannot be guaranteed from
57: simulations presented here, the results still indicate that such
58: events may be possible in Type Ia supernova explosions.
59: \end{abstract}
60:
61: \keywords{Stars: supernovae: general -- Hydrodynamics -- Instabilities
62: -- Turbulence -- Methods: numerical}
63:
64: \maketitle
65:
66: \setcounter{page}{1103}
67:
68: % begin main
69: \section{Introduction}
70: \label{intro_sect}
71:
72: The question of whether deflagration-to-detonation transitions (DDTs) may
73: occur in type Ia supernova (SN~Ia) explosions has been a puzzle since
74: the scenario of delayed detonations was suggested in the first place
75: \citep{khokhlov1991a}. If a supersonic detonation
76: burning mode occurs at all in SNe~Ia, such a transition is inevitable
77: because a prompt detonation fails to produce the observed
78: intermediate mass elements \citep{arnett1969a}. For
79: these to be synthesized, burning must partially take place at low fuel
80: densities.
81: A pre-expansion of the white dwarf (WD) material before burning is
82: only possible for subsonic propagation of the thermonuclear
83: flame. Consequently, the standard picture is that of an accreting WD
84: which getting close to the Chandrasekhar limit becomes unstable to a
85: thermonuclear runaway and ignites a subsonic deflagration flame.
86: In contrast to the shock-driven detonation, flame
87: propagation is mediated by microphysical transport in this mode.
88: However, this scenario is intricate because the energy released in
89: burning may expand the WD
90: so rapidly, that the slow flame is not capable of burning sufficient
91: amounts of material before it has dropped to densities at which
92: burning cannot be sustained any longer. This way, the explosion
93: would be far too weak to be consistent with observations. Therefore,
94: the flame needs to accelerate and two possible mechanisms
95: for this have been suggested.
96:
97: The first is
98: well-founded on known physical principles and is based on the fact
99: that the buoyant rise of the burning bubbles generates strong
100: turbulent motions due to Rayleigh-Taylor and Kelvin-Helmholtz
101: instabilities. The flame interacts with eddies of the resulting
102: turbulent cascade and,
103: by means of surface area increase, accelerates significantly. This turbulent
104: deflagration scenario has been recently studied in detail in three-dimensional
105: simulations \citep[e.g.][]{reinecke2002d,gamezo2003a,roepke2005b}. It
106: has been shown that it may lead
107: to explosions of the WD star meeting the gross properties of
108: observed SNe~Ia \citep{roepke2007c}. However, some issues remain unsolved in this
109: model. First, in its current implementations it can explain only the weaker events. Second, it
110: may in some configurations leave behind unburnt material in the
111: central parts of the ejecta, which disagrees with observations \citep{kozma2005a}. Third,
112: it has problems explaining the high-velocity intermediate mass
113: elements seen in the spectra of SNe~Ia.
114:
115: These weaknesses of the models may possibly be cured if a
116: second way of flame acceleration, a DDT, takes place in later stages of the explosion.
117: One-dimensional parameterizations of this ``delayed detonation model''
118: showed best agreement with observations assuming this transition to
119: occur once the
120: fuel density ahead of the flame has dropped to $\sim$$10^7 \gcc$.
121: Unfortunately, such a transition is hypothetical as of yet, since no
122: convincing mechanism has been identified that would work robustly in a
123: SN~Ia.
124:
125: \citet {niemeyer1997b} pointed out that the only fundamental change in
126: the flame properties is found at the
127: transition from the flamelet regime of turbulent combustion to the
128: distributed regime. In the first, which holds for most parts of the
129: explosion process, the interaction of the flame with turbulence is
130: purely kinematic. The flame front as a whole is corrugated by turbulent
131: eddies, but its internal structure remains unaffected. At a first
132: glance this seems surprising, since Reynolds numbers $\mathit{Re}(L)$ at scales of
133: $L = 10^{7} \, \mathrm{cm}$ typical for the situation in SNe~Ia
134: are of the order
135: of $10^{14}$. The corresponding
136: Kolmogorov scale, $\eta \sim L \,[\mathit{Re}(L)]^{-3/4}$, down to
137: which the turbulent cascade
138: extends, is $\sim 10^{-4} \,
139: \mathrm{cm}$---much smaller than the flame thickness. However, assuming
140: Kolmogorov scaling, the
141: velocity fluctuations (denoted as $v'$ in the following) decrease
142: with the cubic root of the length scale under consideration within the
143: inertial range of the turbulent cascade. This constitutes
144: the so-called Gibson scale, at which the laminar flame speed is
145: comparable to $v'$. Below the Gibson
146: scale, the flame burns through turbulent eddies faster than they can
147: deform it, and therefore they do not affect the flame shape
148: considerably. However, since the laminar burning velocity and the flame
149: thickness depend on the fuel density \citep{timmes1992a}, the
150: Gibson scale decreases in the course of the explosion while the flame
151: structure becomes wider. Therefore, turbulent eddies will
152: eventually penetrate the internal flame structure and mix heated
153: material, fuel, and ashes.
154:
155: It turns out that this onset of the distributed burning
156: regime takes place at fuel densities favored for DDTs in
157: one-dimensional simulations \citep{niemeyer1997b}.
158: Studying the microphysics of burning in this regime,
159: \citet{lisewski2000b} concluded that
160: triggering a detonation may be possible for sufficiently strong
161: turbulence. The velocity fluctuations should be close to $10^8
162: \, \mathrm{cm} \, \mathrm{s}^{-1}$ (corresponding to $\sim$20\% sound
163: speed), but at this time the available two-dimensional supernova
164: simulations seemed to exclude such high turbulent velocities. A
165: re-analysis of this scenario is underway \citep{woosley2007a} and
166: arrives at similar conclusions.
167: If a mechanism providing a deflagration-to-detonation transition
168: driven by the flame itself existed,
169: the parameter deciding on triggering a detonation would be the strength of the
170: mixing of the flame structure, which is determined by the magnitude of
171: $v'$. This value can be estimated from three-dimensional
172: numerical simulations of the deflagration stage of the explosion and
173: is the focus of the present paper.
174:
175: \section{Approach}
176:
177: \begin{table}
178: \begin{center}
179: \caption{Parameters of the three-dimensional deflagration SN~Ia
180: simulations used for the analysis.
181: \label{tab:par}}
182: \setlength{\extrarowheight}{2pt}
183: \begin{tabular}{llll}
184: \tableline\tableline
185: \multicolumn{1}{p{2em}}{name} &
186: \multicolumn{1}{p{4.5em}}{spatial coverage} &
187: \multicolumn{1}{p{6.1em}}{resolution (computational grid cells per dimension)} &
188: \multicolumn{1}{p{5em}}{central density [$10^{9} \, \mathrm{g}\,\mathrm{cm}^{-3}$]}\\
189: \tableline
190: Simulation I & full star & 1024 & 2.9 \\
191: Simulation II & full star & 640 & 2.0 \\
192: Simulation III & octant & 256 & 2.9 \\
193: \tableline
194: \end{tabular}
195: \end{center}
196: \end{table}
197:
198: \begin{figure}
199: \includegraphics[width = \linewidth]{./f1}
200: %\includegraphics[height = 0.94\textheight]{./f1}
201: \caption{Fits to the histogram of $v'$ in simulation I at t = 0.80 s for $1 < \rho
202: [10^7\, \mathrm{g}\,\mathrm{cm}^{-3}] \le 3$ \emph{(top two
203: panels),} and the
204: probability of finding $v'$ larger than a given value according to
205: eq.~(\ref{eq:int}) for different
206: times in the same density interval \emph{(bottom panel).}\label{fig:fits}}
207: \end{figure}
208:
209: \begin{table*}
210: \begin{center}
211: \caption{Maximum turbulent velocities at a scale of $10^6 \,
212: \mathrm{cm}$, fit parameters according to Eq.~(\ref{eq:fit}), probability of finding $v' \ge 10^8 \,
213: \mathrm{cm}\,\mathrm{s}^{-1}$ ($P(10^8)$), estimated flame area
214: $A_{\mathrm{est}}$, size of the patch of the flame where $v' \ge 10^8 \,
215: \mathrm{cm}\,\mathrm{s}^{-1}$ ($A_\mathrm{est} P(10^8)$), and number
216: of computational cells taken into account in the analysis for
217: Simulation I in the density range $1 < \rho
218: [10^7\, \mathrm{g}\,\mathrm{cm}^{-3}] \le 3$.
219: \label{tab:sim1}}
220: \setlength{\extrarowheight}{2pt}
221: \begin{tabular}{llllllllr}
222: \tableline\tableline
223: \multicolumn{1}{p{2em}}{$t$ [s]} &
224: \multicolumn{1}{p{4.5em}}{$v'_{\mathrm{max}}(10^6 \, \mathrm{cm})$
225: [$10^7 \, \mathrm{cm}\, \mathrm{s}^-1$]} &
226: \multicolumn{1}{p{5em}}{$-a_0$ [$10^{-4}$]} &
227: \multicolumn{1}{p{4.67em}}{$a_1$ [$10^{-1}$]} &
228: \multicolumn{1}{p{4.67em}}{$-a_2$}&
229: \multicolumn{1}{p{4em}}{$P(10^8)$}&
230: \multicolumn{1}{p{4.5em}}{$A_\mathrm{est}$ [$\mathrm{cm}^2$]} &
231: \multicolumn{1}{p{4.5em}}{$A_\mathrm{est} P(10^8)$ [$\mathrm{cm}^2$]}&
232: \multicolumn{1}{p{3em}}{cells}\\
233: \tableline
234: 0.70& 6.70 &$ 0.356691\pm 0.0004$&$ 7.12569\pm 0.0006$&$ 12.4931\pm
235: 0.007$&$ 5.062E-07$&$ 7.85E+16$&$ 3.98E+10$& 55,341\\
236: 0.75& 8.38 &$ 2.97325\pm 0.006$&$ 5.98318\pm 0.001$&$ 11.1650\pm
237: 0.02$&$ 1.702E-06$&$ 5.35E+17$&$ 9.11E+11$& 274,015\\
238: 0.80& 11.5 &$ 15.3360\pm 0.010$&$ 5.12493\pm 0.0004$&$ 9.90905\pm
239: 0.007$&$ 2.179E-06$&$ 1.50E+18$&$ 3.28E+12$& 574,769\\
240: 0.85& 10.1 &$ 8.92170\pm 0.02$&$ 5.44507\pm 0.001$&$ 10.3121\pm
241: 0.02$&$ 5.016E-07$&$ 2.52E+18$&$ 1.27E+12$& 736,593\\
242: 0.90& 7.05 &$ 8.92137\pm 0.006$&$ 5.55688\pm 0.0004$&$ 9.47029\pm
243: 0.007$&$ 8.966E-09$&$ 2.98E+18$&$ 2.67E+10$& 614,195\\
244: 0.95& 6.20 &$ 5.18888\pm 0.01$&$ 5.87899\pm 0.001$&$ 9.71264\pm
245: 0.02$&$ 1.691E-09$&$ 2.89E+18$&$ 4.89E+09$& 443,823\\
246: 1.00& 4.24 &$ 2.20126\pm 0.005$&$ 6.39985\pm 0.001$&$ 9.98804\pm
247: 0.02$&$ 6.351E-11$&$ 2.28E+18$&$ 1.45E+08$& 263,740\\
248: \tableline
249: \end{tabular}
250: \end{center}
251: \end{table*}
252:
253: \begin{table}
254: \begin{center}
255: \caption{Selected parameters for different density intervals and
256: simulations. The notation is the same as in Table~\ref{tab:sim1}.
257: \label{tab:all}}
258: \setlength{\extrarowheight}{2pt}
259: \begin{tabular}{lllr}
260: \tableline\tableline
261: \multicolumn{1}{p{2em}}{$t$ [s]} &
262: \multicolumn{1}{p{4.5em}}{$v'_{\mathrm{max}}(10^6 \, \mathrm{cm})$
263: [$10^7 \, \mathrm{cm}\, \mathrm{s}^{-1}$]} &
264: \multicolumn{1}{p{7em}}{$A_\mathrm{est} P(10^8)$ [$\mathrm{cm}^2$]}&
265: \multicolumn{1}{p{3em}}{cells}\\
266: \tableline
267: \multicolumn{3}{l}{Simulation I, $1 < \rho
268: [10^7\, \mathrm{g}\,\mathrm{cm}^{-3}] \le 2$}\\
269: \tableline
270: 0.70& 5.37 &$ 6.72E+05$& 5723\\
271: 0.75& 7.47 &$ 1.05E+11$& 80,438\\
272: 0.80& 11.5 &$ 1.66E+12$& 276,683\\
273: 0.85& 10.1 &$ 1.20E+12$& 443,841\\
274: 0.90& 6.63 &$ 1.82E+10$& 423,654\\
275: 0.95& 6.20 &$ 9.74E+09$& 355,975\\
276: 1.00& 4.24 &$ 2.58E+08$& 243,480\\
277: \tableline
278: \multicolumn{3}{l}{Simulation I, $2 < \rho
279: [10^7\, \mathrm{g}\,\mathrm{cm}^{-3}] \le 3$}\\
280: \tableline
281: 0.70& 6.70 &$ 6.00E+10$& 49,618\\
282: 0.75& 8.38 &$ 4.40E+11$& 193,577\\
283: 0.80& 9.30 &$ 9.83E+11$& 298,086\\
284: 0.85& 7.79 &$ 3.06E+11$& 292,752\\
285: 0.90& 7.05 &$ 1.78E+09$& 190,541\\
286: 0.95& 4.81 &$ 6.96E+08$& 87,848\\
287: 1.00& 3.22 &$ 4.92E+07$& 20,260\\
288: \tableline
289: \multicolumn{3}{l}{Simulation II, $1 < \rho
290: [10^7\, \mathrm{g}\,\mathrm{cm}^{-3}] \le 3$}\\
291: \tableline
292: 0.75& 5.27 &$ 4.52E+10$& 8312\\
293: 0.80& 6.65 &$ 1.95E+11$& 45,902\\
294: 0.85& 8.36 &$ 8.37E+10$& 121,638\\
295: 0.90& 8.46 &$ 1.37E+11$& 200,764\\
296: 0.95& 7.02 &$ 1.80E+11$& 239,816\\
297: 1.00& 6.65 &$ 6.70E+10$& 230,245\\
298: 1.05& 5.79 &$ 2.91E+11$& 187,809\\
299: 1.10& 5.18 &$ 3.38E+11$& 139,548\\
300: 1.15& 4.45 &$ 1.67E+10$& 90,103\\
301: \tableline
302: \multicolumn{3}{l}{Simulation III, $1 < \rho
303: [10^7\, \mathrm{g}\,\mathrm{cm}^{-3}] \le 3$}\\
304: \tableline
305: 0.75& 5.18 &$ 5.11E+12$& 9865\\
306: 0.80& 5.67 &$ 6.17E+10$& 15,073\\
307: 0.85& 6.51 &$ 1.39E+12$& 18,209\\
308: 0.90& 5.75 &$ 3.80E+12$& 17,166\\
309: 0.95& 4.04 &$ 2.61E+11$& 12,589\\
310: 1.00& 3.00 &$ 2.10E+08$& 8098\\
311: \tableline
312: \end{tabular}
313: \end{center}
314: \end{table}
315:
316: \begin{figure*}[t]
317: \includegraphics[width = \linewidth]{./f2}
318: \caption{Simulation I at $t = 0.80 \, \mathrm{s}$ after ignition. The
319: blue opaque, transparent, or wiremesh surfaces correspond to the
320: flame front (implemented via a level-set approach). Volumes of high
321: turbulent velocity fluctuations are rendered in red/orange. In the
322: lower regions, the green arrow indicates the location of the maximum
323: value of $v'$ found in the simulation. For better visibility, white
324: areas correspond to ash regions and fuel regions are shown in black
325: in a plane intersecting with the maximum $v'$-value in the lower
326: right region. \label{fig:loc}}
327: \end{figure*}
328:
329:
330:
331: To determine $v'$, data
332: from several three-dimensional simulations of the deflagration stage
333: in thermonuclear supernovae are analyzed. These simulations apply the methods described in detail by
334: \citet{reinecke1999a,reinecke2002b}, \citet{roepke2005c}, and \citet{schmidt2006c}.
335: The key features include flame tracking via a
336: level-set approach, a turbulent subgrid-scale model, and a moving
337: computational grid to follow the expansion of the WD.
338:
339: Simulation I \citep{roepke2007c} was carried out on $1024^3$ grid cells and
340: comprised the full star, as did simulation II which was set up on
341: $640^3$ grid cells. Only an octant of the WD was accounted for in
342: Simulation III with a $256^3$ cells grid assuming mirror symmetry with
343: the other octants. The exploding WD had an initial central density of
344: $2.9\times 10^9 \gcc$ in simulations I and III, while for simulation II a
345: central density of $2.0 \times 10^9 \gcc$ was chosen. These
346: parameters of the setup are summarized in Table~\ref{tab:par}.
347: All initial WD
348: configurations were composed of a mixture of equal parts, by mass, of
349: carbon and oxygen. The flame was
350: ignited in multiple kernels spherically distributed around the center
351: of the star and partially overlapping, similar to the scenarios
352: presented by \citet{roepke2006a}.
353:
354: The quantity under consideration here is $v'$ experienced by the flame in the distributed burning
355: regime. Of course, this is not directly
356: resolvable in simulations carried out on the scales of the WD
357: star. Therefore, an estimate of the turbulent velocities at the flame
358: is obtained from the subgrid-scale turbulence model. It provides the value of
359: $v'$ at the scale of the computational
360: grid. Since the size of the grid cells is a dynamical quantity in the
361: implementation, it is rescaled to a length scale of $10^6 \,
362: \mathrm{cm}$ for comparison. To this end, a Kolmogorov
363: spectrum was assumed, which need not necessarily apply to the actual scaling of
364: turbulence at the considered spatial range; however, in the
365: simulations presented by \citet{zingale2005a} and \citet{roepke2007c}.
366: Kolmogorov turbulence is indeed recovered for buoyancy-unstable flames. In
367: any case, the
368: potential error introduced by
369: this procedure is expected to be small, since the
370: grid scales at times examined here will not deviate significantly from $10^6 \,
371: \mathrm{cm}$. Another simplification is introduced by reducing the data sets
372: by a factor of 2 in each direction, leaving out every second cell of
373: the computational grid. Although this deteriorates the statistics, it
374: simplifies the analysis of the large data sets from the highly
375: resolved simulations.
376: The distributed burning regime is expected to be reached for fuel
377: densities in the range $1\ldots 3 \times 10^7 \gcc$
378: \citep{woosley2007a, niemeyer1997d}. Therefore, we
379: measure $v'$ determined by the
380: subgrid-scale model in computational grid cells cut by the flame front
381: in this density range. This introduces two uncertainties. First, these
382: cells contain mixed states of fuel and ashes. Consequently, the
383: values we obtain are not equal (but, due to the small density jump over
384: the flame, similar to) the values in pure fuel. Although it may seem
385: compelling to restrict the analysis to cells sufficiently far away
386: from the flame so that they contain pure fuel, this would require a
387: distance of at least three cells, and turbulence will be considerably weaker
388: that far ahead of the flame. Values derived this way would therefore
389: not well represent the
390: $v'$ experienced by the flame front. Second, as we are
391: looking for maxima of $v'$, the
392: subgrid-scale model will provide only a mean value on the size of the
393: grid cells. Of course, there will be a maximum of this quantity, which
394: will be determined in the following, but due to the intermittent
395: nature of turbulence, rare extrema exceeding these values are possible
396: on smaller scales in the real situation.
397:
398: \section{Results and Implications}
399:
400: The following discussion focuses on simulation I (the
401: corresponding values are given in Tables~\ref{tab:sim1}
402: and \ref{tab:all}). Results from the other
403: simulations are included in Table~\ref{tab:all} and corroborate the
404: generality of these results. The possibility of triggering a detonation
405: at a certain turbulence strength depends on the density of
406: the fuel \citep{lisewski2000b, woosley2007a}.
407: To account for this density dependence we consider $v'$ at the
408: flame front in two separate density ranges in simulation I: $(1 <
409: \rho_7 \le 2)\, 10^7 \gcc$ and $(2 < \rho_7 \le 3) \, 10^7 \gcc$.
410:
411:
412: The instantaneous maxima of $v'$ in the snapshots of the simulations
413: analyzed here fall around $10^8 \, \mathrm{cm} \, \mathrm{s}^{-1}$
414: for all three setups (cf.\ Tables~\ref{tab:sim1}
415: and \ref{tab:all}). They show a similar temporal evolution passing
416: through a peak value.
417: The maxima of $v'$ are
418: slightly lower for less resolved simulations.
419: Since the turbulent subgrid-scale model should compensate for
420: decreasing resolutions, this is not likely due to
421: lower turbulence prediction. Instead, a lower resolution implies a
422: less frequent realization of high turbulence
423: intensities on grid.
424:
425: \subsection{Probability of high turbulent velocities}
426:
427: The information gained from determining the maximum turbulent
428: velocities in the simulation is limited since only stochastic
429: realizations are recovered.
430: The robust physical quantity that can be extracted from the simulationss is
431: the probability density function (pdf) of $v'$. This pdf can be
432: approximated with the normalized histogram of $v'$ determined as described
433: above. An example is shown in Figure~\ref{fig:fits} \emph{(top)}.
434:
435: For determining the possibility of
436: detonations, the high-velocity tail is the relevant part of the histogram.
437: An estimate of the corresponding part of the pdf is obtained by fitting
438: the histogram
439: with the exponential of a geometric \emph{Ansatz:}
440: \begin{equation}
441: \label{eq:fit}
442: P(v') = \exp \left[ a_0 (v')^{a_1} +a_2 \right].
443: \end{equation}
444: The fit parameters for simulation I at different times are given in
445: Table~\ref{tab:sim1}.
446: Integrating this expression from $v$ to $\infty$ yields the
447: probability of finding $v'$ greater than a
448: given value $v$,
449: \begin{equation}
450: \label{eq:int}
451: \int_{v}^{\infty} P(v') dv' = \frac{v\, \exp(a_2) \, \Gamma\left( 1/a_1,
452: -a_0 v^{a_1} \right)}{a_1 (- a_0)^{1/a_1} v},
453: \end{equation}
454: where $\Gamma (.,.)$ denotes the upper incomplete Gamma function.
455: The results of this fitting procedure for an exemplary data set are
456: shown in Figure~\ref{fig:fits}. Obviously, Gaussian or lognormal fits
457: fail to reproduce the high-velocity trend of the histogram (see
458: Fig.~\ref{fig:fits}, \emph{middle}).
459:
460: Figure~\ref{fig:fits} \emph{(bottom)} illustrates the time evolution of the
461: probability of finding a turbulent velocity larger than a given
462: value, according to equation~(\ref{eq:int}). At earlier times, the
463: probability of finding $v' \ge 5 \times 10^{7} \, \mathrm{cm} \,
464: \mathrm{s}^{-1}$ is relatively large, but it decreases steeply for
465: higher threshold values. Later on, this decrease flattens out
466: somewhat, but the overall probability becomes lower.
467:
468: To discuss this behavior, we introduce the probability of finding $v'
469: \ge 10^{8} \, \mathrm{cm} \, \mathrm{s}^{-1}$, $P(10^8)$, because
470: according to \citet{lisewski2000b}, this is approximately the threshold
471: for triggering a detonation. Values of $P(10^8)$ are given in
472: Table~\ref{tab:sim1}. For simulation I they follow the temporal trend
473: of $v'_\mathrm{max}$ and decrease steeply towards late times.
474: This evolution might be expected. Turbulence decreases in the late
475: stages of the supernova explosion because the
476: energy injection from buoyancy instabilities becomes lower and due to
477: expansion \citep[the explosion ultimately establishes homologous
478: expansion;][]{roepke2005c}. Thus, the probability of finding high
479: turbulent velocities
480: on any patch of the flame naturally decreases with time.
481: But here we consider only the part of the flame for which the fuel
482: densities are in a certain range.
483: The corresponding flame area increases first and then decreases, since
484: the fuel densities steadily
485: decrease in the explosion process.
486: This area can be
487: estimated from the number of computational cells
488: cut by the flame front and their sizes, providing the value
489: $A_\mathrm{est}$ given in Tables~\ref{tab:sim1} and \ref{tab:all}.
490: This temporal evolution of $A_\mathrm{est}$ makes realizations
491: of a certain high $v'$ more likely at a particular instant given by
492: the convolution of the area effect with decreasing turbulence
493: strengths. A quantification is provided in Tables~\ref{tab:sim1} and
494: \ref{tab:all} with the size of the flame surface $A_\mathrm{est}
495: P(10^8)$ at which $v'
496: > 10^{8} \, \mathrm{cm} \, \mathrm{s}^{-1}$. Obviously,
497: this area needs to exceed a certain size in order to be relevant for a
498: DDT. As seen from Table~\ref{tab:sim1}, the evolution of $A_\mathrm{est}
499: P(10^8)$
500: follows the maxima of $v'$ determined at different times in
501: simulation I. However,
502: for the other simulations
503: the trend is less clear. In simulation II the density range of interest is
504: reached later due to lower initial central density. This delays the
505: generation of turbulence and thus
506: burning, energy release, and the expansion of the WD. The overall
507: lower values of $A_\mathrm{est}
508: P(10^8)$ derived from simulation II may be taken as an indication that
509: lower central densities decrease the chances of a DDT, but this
510: hypothesis needs to be tested in a larger sample of simulations.
511:
512: \subsection{Location of high turbulent velocities}
513:
514:
515: In order to establish a self-sustained detonation wave, a minimum mass
516: of material needs to be burned in the initiation process
517: \citep{niemeyer1997b,dursi2006a,roepke2007a}. Therefore,
518: it is necessary that the high turbulent velocities are concentrated in
519: larger patches and do not distribute all over the flame in tiny
520: regions. Although $A_{\mathrm{est}}P(10^8)$ exceeds
521: $10^{12}\,\mathrm{cm}^2$ in simulations I and III, there is no guarantee that a connected
522: spatial structure with dimensions larger than
523: $10\,\mathrm{km}$ exists.
524: Moreover, the particular location at the flame
525: where the high turbulent velocities are found may be critical for the
526: detonation ignition \citep{woosley2007a}.
527: Do they occur at leading or trailing edges of
528: burning bubbles, or in small channels of fuel in between large ash
529: structures? Such geometrical information cannot be provided by the
530: pdf.
531:
532: To find answers to these questions we consider the situation at $0.8
533: \, \mathrm{s}$ after deflagration ignition in simulation
534: I.
535:
536: Values of $v'$ exceeding
537: $10^{8}\, \mathrm{cm}\, \mathrm{s}^{-1}$ are directly found in the simulation
538: (see Table~\ref{tab:sim1}).
539: The fact that the grid resolution here is
540: $\sim$$10^6\,\mathrm{cm}$ indicates that such high turbulent
541: velocities are localized in finite patches. If the area of high
542: turbulent velocity fluctuations determined from the pdf was
543: distributed in many very small patches, it would not be directly
544: visible in the simulation, due to numerical smoothing.
545:
546: Figure~\ref{fig:loc} shows the location of the regions of high
547: turbulent intensity at the deflagration flame.
548: The upper left region of Figure~\ref{fig:loc} demonstrates that flame
549: regions with high turbulent intensities are present
550: mainly at trailing edges of bubble-like structures. The location of
551: all regions with turbulent velocities larger than $10^{7} \,
552: \mathrm{cm} \, \mathrm{s}^{-1}$ is shown again in the upper right
553: region, where the flame has been made transparent in order to make
554: these regions better visible. At the instant shown in Figure~\ref{fig:loc}, the
555: peak value of $v'$ in the analysis of the simulation is found. Its
556: location is illustrated in the lower regions.
557:
558:
559: \subsection{Possibility of triggering a detonation}
560:
561:
562: According to \citet{lisewski2000b}, values of $v'
563: \gtrsim 10^8 \, \mathrm{cm} \, \mathrm{s}^{-1}$, corresponding to
564: about 1/5 to 1/4 of the sound speed at the considered densities, are necessary for a
565: DDT (with optimistic assumptions on the DDT mechanism). The exact
566: values depend on the fuel density and
567: composition. For a fuel composition of equal parts, by mass, of carbon
568: and oxygen (this case applies to the simulations analyzed here), a
569: threshold of $v' > 5 \times 10^7 \, \mathrm{cm}\,\mathrm{s}^{-1}$ is
570: given for a fuel density of $2.3 \times 10^7 \, \mathrm{g} \,
571: \mathrm{cm}^{-3}$. For lower fuel densities, this threshold becomes
572: larger, and at $8 \times 10^6 \, \mathrm{g} \,
573: \mathrm{cm}^{-3}$ it is given with $v' > 8 \times 10^7 \,
574: \mathrm{cm}\,\mathrm{s}^{-1}$. The maxima found in the simulations (in
575: particular, in simulation I; see Tables~\ref{tab:sim1}
576: and \ref{tab:all}) exceed these thresholds.
577:
578: The estimates for the total size of the flame patches where $v' > 10^8 \,
579: \mathrm{cm}\,\mathrm{s}^{-1}$ and the fact that high values of $v'$ are found at scales resolved
580: in the simulation indicate that
581: the corresponding regions may be sufficiently large to trigger a
582: detonation (compare with the values of detonator sizes given
583: by \citet{niemeyer1997b} and \citet{roepke2007a}).
584:
585: As noted by \citet{lisewski2000b}, the chances of triggering
586: a detonation increase with higher carbon fractions in the fuel and
587: decrease with higher oxygen fractions. \citet{roepke2004c}
588: noted that the energy release in burning does not depend strongly on
589: the carbon mass fraction of the fuel as long as it proceeds to nuclear
590: statistical equilibrium. This, however, does not apply to the
591: situation considered here. The low fuel densities imply incomplete
592: burning to intermediate-mass
593: elements. Therefore, in addition to the effect of altering the
594: thresholds of turbulence strength required for triggering a detonation,
595: the carbon mass fraction may also affect the turbulence strength in
596: the distributed burning regime. Additional studies are required to
597: settle this question.
598:
599: \section{Conclusions}
600: \label{sect:concl}
601:
602: The intensity of turbulence was analyzed at the onset of the distributed
603: burning regime in three-dimensional SN~Ia simulations.
604: Considerably larger values for the maximum
605: turbulent velocity fluctuations than previously anticipated
606: \citep{lisewski2000b} were found for all simulations considered and
607: the conclusion that these values are typical for
608: three-dimensional simulations featuring a state-of-the-art treatment
609: of turbulence on unresolved scales seems compelling. The
610: histogram of $v'$ features a pronounced high-velocity tail. This part
611: is not well fit by a Gaussian pdf but rather by an exponential of a
612: geometric \emph{Ansatz.}
613:
614: According to our results, a deflagration-to-detonation transition as
615: anticipated by \citet{lisewski2000b} is not ruled out. But, lacking a
616: microphysical model, they cannot provide certainty either, and the
617: concerns pointed out by \citet{niemeyer1999a} persist.
618: However, the results presented here encourage detailed studies of the microphysics
619: of the distributed burning regime at high turbulent velocities, such
620: as those presented by \citet{lisewski2000b} and \citet{woosley2007a}.
621:
622: If a DDT occurs in a SN Ia explosion, then a location at the outer
623: parts of the deflagration flame, but at the trailing edge of a
624: bubble-like feature, seems most likely. The rapid decline of the pdf
625: towards high turbulence intensities indicates that such a transition may
626: be a rare event, possibly realized only once or a few times in a
627: supernova. This, however, as well as the question of whether a DDT
628: occurs in every event, depends on the details of the formation of the
629: detonation and warrants further study.
630:
631: It may thus be possible that detonations do not form at all leading
632: features of deflagration flame but only in a few such locations. As a
633: detonation front cannot cross even tiny regions filled
634: with nuclear ash \citep{maier2006a}, it may not reach all patches of
635: unburnt material embedded in the complex deflagration
636: structure. However, in numerical simulations of
637: delayed detonations in WD with parametrized DDTs, \citet{roepke2007b}
638: found that even a
639: detonation front formed in a single spot burns most of the remaining
640: fuel by reaching the dense center of the star. Such
641: large-scale supernova simulations with parametrized DDTs
642: help to address the question of whether
643: the outcome of delayed detonations is consistent with observational data
644: \citep{mazzali2007a}.
645:
646: \begin{acknowledgements}
647: Stan Woosley greatly supported this work with stimulating
648: discussions.
649: The simulations have been performed on facilities of the Max
650: Planck Society, Garching, Germany, the HPCx, Edinburgh, UK (as
651: part of the DEISA project), and at the
652: National Center for Computational Sciences at Oak Ridge National
653: Laboratory, which is supported by the Office of Science of the
654: U.S.\ Department of Energy, under contract DE-AC05-00OR22725. The
655: work was supported by the SciDAC Program of the DOE
656: (DE-FC02-01ER41176) and the
657: NASA theory program (NNG 05-GG08G).
658: \end{acknowledgements}
659:
660: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
661:
662:
663: %%%%%%%%%%%%
664: \begin{thebibliography}{25}
665:
666: \bibitem[{{Arnett}(1969)}]{arnett1969a}
667: {Arnett}, W.~D. 1969, \apss, 5, 180
668:
669: \bibitem[{{Dursi} \& {Timmes}(2006)}]{dursi2006a}
670: {Dursi}, L.~J., \& {Timmes}, F.~X. 2006, \apj, 641, 1071
671:
672: \bibitem[{{Gamezo} {et~al.}(2003){Gamezo}, {Khokhlov}, {Oran}, {Chtchelkanova},
673: \& {Rosenberg}}]{gamezo2003a}
674: {Gamezo}, V.~N., {Khokhlov}, A.~M., {Oran}, E.~S., {Chtchelkanova}, A.~Y., \&
675: {Rosenberg}, R.~O. 2003, Science, 299, 77
676:
677: \bibitem[{{Khokhlov}(1991)}]{khokhlov1991a}
678: {Khokhlov}, A.~M. 1991, \aap, 245, 114
679:
680: \bibitem[{{Kozma} {et~al.}(2005){Kozma}, {Fransson}, {Hillebrandt},
681: {Travaglio}, {Sollerman}, {Reinecke}, {R{\"o}pke}, \&
682: {Spyromilio}}]{kozma2005a}
683: {Kozma}, C., {Fransson}, C., {Hillebrandt}, W., {Travaglio}, C., {Sollerman},
684: J., {Reinecke}, M., {R{\"o}pke}, F.~K., \& {Spyromilio}, J. 2005, \aap, 437,
685: 983
686:
687: \bibitem[{{Lisewski} {et~al.}(2000){Lisewski}, {Hillebrandt}, \&
688: {Woosley}}]{lisewski2000b}
689: {Lisewski}, A.~M., {Hillebrandt}, W., \& {Woosley}, S.~E. 2000, \apj, 538, 831
690:
691: \bibitem[{{Maier} \& {Niemeyer}(2006)}]{maier2006a}
692: {Maier}, A., \& {Niemeyer}, J.~C. 2006, \aap, 451, 207
693:
694: \bibitem[{{Mazzali} {et~al.}(2007){Mazzali}, {R{\"o}pke}, {Benetti}, \&
695: {Hillebrandt}}]{mazzali2007a}
696: {Mazzali}, P.~A., {R{\"o}pke}, F.~K., {Benetti}, S., \& {Hillebrandt}, W. 2007,
697: Science, 315, 825
698:
699: \bibitem[{{Niemeyer}(1999)}]{niemeyer1999a}
700: {Niemeyer}, J.~C. 1999, \apj, 523, L57
701:
702: \bibitem[{{Niemeyer} \& {Kerstein}(1997)}]{niemeyer1997d}
703: {Niemeyer}, J.~C., \& {Kerstein}, A.~R. 1997, New Astronomy, 2, 239
704:
705: \bibitem[{{Niemeyer} \& {Woosley}(1997)}]{niemeyer1997b}
706: {Niemeyer}, J.~C., \& {Woosley}, S.~E. 1997, \apj, 475, 740
707:
708: \bibitem[{{Reinecke} {et~al.}(2002{\natexlab{a}}){Reinecke}, {Hillebrandt}, \&
709: {Niemeyer}}]{reinecke2002b}
710: {Reinecke}, M., {Hillebrandt}, W., \& {Niemeyer}, J.~C. 2002{\natexlab{a}},
711: \aap, 386, 936
712:
713: \bibitem[{{Reinecke} {et~al.}(2002{\natexlab{b}}){Reinecke}, {Hillebrandt}, \&
714: {Niemeyer}}]{reinecke2002d}
715: ---. 2002{\natexlab{b}}, \aap, 391, 1167
716:
717: \bibitem[{{Reinecke} {et~al.}(1999){Reinecke}, {Hillebrandt}, {Niemeyer},
718: {Klein}, \& {Gr{\" o}bl}}]{reinecke1999a}
719: {Reinecke}, M., {Hillebrandt}, W., {Niemeyer}, J.~C., {Klein}, R., \& {Gr{\"
720: o}bl}, A. 1999, \aap, 347, 724
721:
722: \bibitem[{{R{\"o}pke}(2005)}]{roepke2005c}
723: {R{\"o}pke}, F.~K. 2005, \aap, 432, 969
724:
725: \bibitem[{{R{\"o}pke} \& {Hillebrandt}(2004)}]{roepke2004c}
726: {R{\"o}pke}, F.~K., \& {Hillebrandt}, W. 2004, \aap, 420, L1
727:
728: \bibitem[{{R{\"o}pke} \& {Hillebrandt}(2005)}]{roepke2005b}
729: ---. 2005, \aap, 431, 635
730:
731: \bibitem[{{R{\"o}pke} {et~al.}(2006){R{\"o}pke}, {Hillebrandt}, {Niemeyer}, \&
732: {Woosley}}]{roepke2006a}
733: {R{\"o}pke}, F.~K., {Hillebrandt}, W., {Niemeyer}, J.~C., \& {Woosley}, S.~E.
734: 2006, \aap, 448, 1
735:
736: \bibitem[{{R{\"o}pke} {et~al.}(2007{\natexlab{a}}){R{\"o}pke}, Hillebrandt,
737: Schmidt, Niemeyer, Blinnikov, \& Mazzali}]{roepke2007c}
738: {R{\"o}pke}, F.~K., Hillebrandt, W., Schmidt, W., Niemeyer, J.~C., Blinnikov,
739: S.~I., \& Mazzali, P.~A. 2007{\natexlab{a}}, \apj, 668, 1132
740:
741: \bibitem[{{R{\"o}pke} \& {Niemeyer}(2007)}]{roepke2007b}
742: {R{\"o}pke}, F.~K., \& {Niemeyer}, J.~C. 2007, \aap, 464, 683
743:
744: \bibitem[{{R{\"o}pke} {et~al.}(2007{\natexlab{b}}){R{\"o}pke}, {Woosley}, \&
745: {Hillebrandt}}]{roepke2007a}
746: {R{\"o}pke}, F.~K., {Woosley}, S.~E., \& {Hillebrandt}, W. 2007{\natexlab{b}},
747: \apj, 660, 1344
748:
749: \bibitem[{{Schmidt} {et~al.}(2006){Schmidt}, {Niemeyer}, {Hillebrandt}, \&
750: {R{\"o}pke}}]{schmidt2006c}
751: {Schmidt}, W., {Niemeyer}, J.~C., {Hillebrandt}, W., \& {R{\"o}pke}, F.~K.
752: 2006, \aap, 450, 283
753:
754: \bibitem[{{Timmes} \& {Woosley}(1992)}]{timmes1992a}
755: {Timmes}, F.~X., \& {Woosley}, S.~E. 1992, \apj, 396
756:
757: \bibitem[{{Woosley}(2007)}]{woosley2007a}
758: {Woosley}, S.~E. 2007, \apj, 668, 1109
759:
760: \bibitem[{{Zingale} {et~al.}(2005){Zingale}, {Woosley}, {Rendleman}, {Day}, \&
761: {Bell}}]{zingale2005a}
762: {Zingale}, M., {Woosley}, S.~E., {Rendleman}, C.~A., {Day}, M.~S., \& {Bell},
763: J.~B. 2005, \apj, 632, 1021
764:
765: \end{thebibliography}
766: %%%%%%%%%%%%
767:
768: \end{document}
769: