1: \documentclass[aps,prl,twocolumn,showpacs,floatfix]{revtex4}
2: %\documentclass[aps,prl,preprint,showpacs,floatfix]{revtex4}
3:
4: \usepackage{graphicx}
5: \usepackage{color}
6: \usepackage{bm}
7:
8: \setlength{\arraycolsep}{2pt}
9: \renewcommand{\d}{\mathrm{d}\,}
10:
11:
12: \begin{document}
13:
14: \bibliographystyle{apsrev}
15:
16: %\title{Rayleigh-Taylor turbulent spectrum}
17: \title{Rayleigh-Taylor turbulence is nothing like Kolmogorov's in the
18: self similar regime}
19: \author{\surname{Olivier} Poujade}
20:
21: \affiliation{Commissariat \`a l'Energie Atomique\\
22: BP12, Bruy\`eres-le-Ch\^atel, 91168 France}
23: \date{\today}
24:
25:
26: \begin{abstract}
27: An increasing number of numerical simulations and
28: experiments describing the turbulent spectrum of Rayleigh-Taylor (RT)
29: mixing layers came to light over the past few years. Results
30: reported in recent studies allow to rule out a turbulence {\it
31: \`a la Kolmogorov} as a mechanism acting on a self similar RT turbulent
32: mixing layer. A different mechanism is presented, which complies with
33: both numerical and experimental results and relates RT flow to other
34: buoyant flows.
35: \end{abstract}
36:
37: \pacs{52.35.Py, 47.20.Bp, 47.27.eb, 47.27.te} \maketitle
38:
39:
40: A Rayleigh-Taylor (RT) instability \cite{rayl,tayl} occurs whenever a light fluid, $\rho_1$,
41: pushes a heavy fluid, $\rho_2$, or similarly, when a heavy fluid on
42: top of a lighter fluid is subject to a gravitational acceleration. The
43: understanding of such instability in the developed turbulent regime is
44: of primary interest to many fields of physics and technology since it
45: is an important cause of mixing between two fluids of different
46: density. In astrophysics for instance, it is responsible for the
47: outward acceleration of a thermonuclear flame in type Ia supernovae
48: \cite{astro1}, but it also plays an important role in shaping the
49: interstellar medium so that new stars can be born \cite{astro2}. The
50: technology of confinement fusion also relies on a good understanding
51: of RT mixing \cite{icf2} and ways to reduce it \cite{icf1}. The RT
52: flow of two incompressible fluids in the low Atwood limit, ${\cal
53: A}=(\rho_2-\rho_1)/(\rho_2+\rho_1)\ll 1$ (Boussinesq approximation),
54: is governed by a concentration equation (\ref{eqc}), the Navier Stokes
55: equation supplemented with a buoyant source term (\ref{equ}) and the
56: incompressibility constraint (\ref{eqdiv})
57: \begin{eqnarray}
58: \partial_t\,c+\left(\bm{u}\bm{\nabla}\right)c\,&=&\kappa\,\Delta c~,\label{eqc}\\
59: \partial_t\,\bm{u}+\left(\bm{u}\bm{\nabla}\right)\bm{u}\,&=&-\bm{\nabla} P+2\,{\cal A}\bm{g}\,c+\nu\,\Delta\bm{u}~,\label{equ}\\
60: \bm{\nabla}\bm{u} &=&0~,\label{eqdiv}
61: \end{eqnarray} where $\bm{g}$ is a stationary and uniform
62: gravitational acceleration vector field (i.e. planar symmetry is
63: assumed). The coefficient $\kappa$ is the molecular diffusion
64: coefficient and $\nu$ is the kinematic viscosity of the mixture. They
65: are both supposed constant. Without loss of
66: generality, $\bm{g}$ is parallel to the $z$-axis. This
67: is why, for any generic physical value $\Phi$, the average (so
68: defined numerically and experimentally) will be
69: $\langle\Phi\rangle(z)=\frac{1}{S}\int_S dx\,dy\,\Phi$ throughout this
70: paper. The fluctuating part will be denoted with a prime and defined
71: as $\Phi^{\prime}=\Phi-\langle\Phi\rangle$.
72:
73: With the increasing capacity of super computers, many simulations of
74: RT flows in the developed turbulent regime have been performed, which
75: describe the velocity spectrum $E(k)$ \cite{csz,cook2,young},
76: defined in such a way that $\langle{\bm{u}^{\prime}}^2\rangle=\int
77: dk\,E(k)$, or the concentration spectrum $E_c(k)$ \cite{rama1, dim1,
78: cook1, dalz1}, defined as $\langle{c^{\prime}}^2\rangle=\int
79: dk\,E_c(k)$, or both \cite{cab1, cook3}. In the same time, although
80: fewer in number, experimental investigations of $E(k)$ \cite{andr3}
81: and $E_c(k)$ \cite{andr1, andr2, dalz1} have been carried out. A quick
82: inspection of these results shows that no consensus arises concerning
83: the shape of these spectrum. From a theoretical point of view, the
84: situation is not more satisfactory. In \cite{chert1} it is claimed
85: that the Kolmogorov-Obukhov scheme, $E(k)\sim k^{-5/3}$, holds in 3D
86: RT mixing given that the effect of buoyancy on turbulence, although
87: fundamental at the largest scale, becomes irrelevant at smaller
88: scales. In \cite{zhou} the particular RT time scale $1/\sqrt{k g A}$
89: at wave number $k$ has been postulated to vary like the turbulent
90: spectral transfer time scale $\epsilon\,k^{-4}\,E^{-2}(k)$ yielding an
91: RT turbulent spectrum $E(k)\sim k^{-7/4}$. Another troublesome point
92: is that there are no convincing objective criteria to assert whether or
93: not an RT flow has reached the gravitational self similar regime but
94: to plot the mixing zone width $L(t)$
95: versus ${\cal A}\,g\,t^2$ and see if there is a straight line
96: somewhere without any information on the expected slope. Subsequently, suggesting the possibility of any behavior of
97: the velocity spectrum in the self similar regime and comparing it to
98: experiments and simulations can be dubious if it is not known for
99: certain that these experiment or simulations have reached this
100: regime. So, the first aim of this paper will be to describe
101: such an objective criterion. The ultimate goal will be to prove that
102: the Kolmogorov mechanism does not explain the observed numerical
103: and experimental results. A theory, based on a spectral equation, will be
104: presented which shows that a balance mechanism between buoyancy and spectral
105: energy transfer can settle at low wave numbers in the self similar regime.
106:
107: The RT developed turbulent regime, as complex as
108: it may look, is thought to evolve self similarly
109: \cite{rist}. The size of the
110: largest significant turbulent structure, therefore, grows like the size of
111: the turbulent mixing zone which evolves as $L(t)=\alpha\,{\cal A}\,g\,t^2$
112: \cite{dim1} where $\alpha$ is the mixing-zone-width growth rate parameter
113: whose value is around $\sim 0.06$ experimentally and between $0.03$
114: and $0.07$ numerically. It can be assumed that at low wave numbers in the self similar regime the
115: velocity spectrum $E(k,t)=0$ for $k<k_l(t)$, where $k_l(t)$ is the wave number corresponding to
116: the maximum of the velocity spectrum (this point will be called $\lambda$ due to the shape of the
117: idealized spectrum at this location), and $E(k,t)\approx
118: \Psi_l(t)\,k^{-n_l}$ for $k\geq k_l(t)$. Nothing is assumed concerning the behavior of $E(k,t)$ at
119: intermediate and high wave numbers $k\gg k_l(t)$. In the self similar regime, $k_l(t)$
120: decreases like $\sim\left({\cal A} g t^2\right)^{-1}$ and the mean turbulent kinetic energy
121: increases like $\langle {u^{\prime}}^2\rangle\sim\left({\cal A} g t\right)^2$ \cite{rist}. Provided
122: that the spectrum decreases toward high wave numbers up to
123: $k=k_{\eta}(t)$ above which it is zero, it can be found that
124: \begin{equation}
125: \langle
126: {u^{\prime}}^2\rangle=\int_{k_l}^{k_{\eta}}dk\,
127: E(k,t)\sim\frac{\Psi_l(t)}{n_l-1}k_l^{1-n_l}\label{eqinj}
128: \end{equation} if $n_l>1$ and $k_{\eta}\gg k_l$. We could have
129: taken into account an additional and more realistic $\sim k^2$ spectrum
130: in the region $k<k_l(t)$ instead of zero but that would only have
131: changed the unimportant constant coefficient in front of the above
132: result. For the previously depicted self similar
133: evolution to occur, using (\ref{eqinj}), the spectrum level at low
134: wave number must verify $\Psi_l(t)\sim\left({\cal
135: A}g\right)^2\,t^2\,k_l^{n_l-1}\sim \left({\cal
136: A}g\right)^{3-n_l}\,t^{4-2n_l}$. The parametrized trajectory of $\lambda$
137: in the $k$-$E$ plot can then be determined since $k_{\lambda}(t)=k_l(t)$ and
138: $E_{\lambda}(t)=E(k_{\lambda}(t),t)$. After eliminating the variable
139: $t$, it is found that the point $\lambda$ must evolve on the curve
140: \begin{equation}
141: E_{\lambda}\sim\left({\cal A}g\right)\,k_{\lambda}^{-2}~,\label{sign}
142: \end{equation} independently of the slope of the spectrum ($n_l$). This is a universal condition in the
143: sense that both those who believe in a Kolmogorov scenario $n_l=5/3$ and others will
144: agree on. It is objective since it amounts to look for a straight line in the $\log(E)$-$\log(k)$
145: plot whose slope, this time, is unambiguously determined ($-2$). Therefore, such
146: behavior should be checked whenever an RT flow is said to have reach a
147: self similar regime. This regime requires large resolution so that the mixing
148: zone can expand and reach the self
149: similar regime before it collides with the simulation box border. Indeed,
150: large resolution simulations \cite{cook2, cook3, cab1} show
151: the mark (\ref{sign}) of the self similar regime which starts near the
152: end. The only
153: experimental result which shows the evolution of the velocity spectrum
154: \cite{andr3} also complies with this rule.
155:
156: It is now possible to select, among all the simulations cited in the
157: introduction, those which have reached the self similar regime. If it is assumed that RT turbulence follows
158: Kolmogorov's mechanism \cite{rist,chert1}, the
159: turbulent spectrum has the well known form
160: $E(k)=C_K\,\epsilon^{2/3}\,k^{-5/3}$. Using Eq.(\ref{eqinj}) with
161: $n_l=5/3>1$ and the self similar laws for $k_l(t)$ and $\langle
162: {u^{\prime}}^2\rangle$, it is straightforward to conclude
163: \cite{chert1} that
164: $\epsilon\sim t$ and $k_{\eta}(t)=\left(\epsilon/\nu^3\right)^{1/4}\sim
165: t^{1/4}$ and also that $\Psi_l(t)=C_K\,\epsilon(t)^{2/3}\sim
166: t^{2/3}$. This conclusion is refuted by recent DNS/LES
167: simulations \cite{cook2,cook3} showing the velocity spectrum evolution
168: in time. Even though it is not stated in these references, it can
169: clearly be noticed that the level of the velocity spectrum at low
170: wave number, $\Psi_l(t)$, does not grow as time evolves but
171: remains still. This is an important observation which
172: constrained the mechanism of RT turbulence.
173:
174: In order to understand the mechanism, it is worth writing the
175: equation governing the evolution of $E(k,t)$ (a generalisation of Lin's equation \cite{sbp})
176: out of the averaged second moment equation of
177: (\ref{equ}) to make the buoyancy production term appear. It can be achieved by multiplying (\ref{equ}) by
178: $\bm{u}$. The spectral equation can then be derived by
179: applying a Fourier transform, $\widetilde{.}$, in the $xy$ plan to the
180: resulting equation and
181: retaining the zero mode contribution. In addition to the rhs of Lin's
182: equation, we find a spectral buoyancy production term deriving from
183: $2 {\cal A}g\,\widetilde{u^{\prime}_z c^{\prime}}(0)={\cal A}\,g\,\int
184: d^2k\,\left[\widetilde{u}_z(k) \widetilde{c}^*(k)+\widetilde{u}^*_z(k)
185: \widetilde{c}(k)\right]$ which, assuming homogeneity and isotropy in
186: the $xy$ plan and phase coherence between $\widetilde{u}_z(k)$ and
187: $\widetilde{c}(k)$, reduces to $\int dk\,{\cal
188: A}g\,E_c^{1/2}(k)E^{1/2}(k)$ since
189: $\left|\widetilde{u}_z(k)\right|\sim k^{-1/2}E^{1/2}(k)/\sqrt{2\pi}$
190: and $\left|\widetilde{c}(k)\right|\sim k^{-1/2}E_c^{1/2}(k)/\sqrt{2\pi}$. The resulting generalisation of Lin's spectral equation is therefore
191: \begin{equation}
192: \partial_t E=T(k)-2\nu k^2 E+\beta{\cal A}g E_c^{1/2}E^{1/2}~,\label{eqspec}
193: \end{equation} where $\beta$ accounts for phase incoherence between
194: $\widetilde{c}$ and $\widetilde{u}_z$. It
195: depends on $k$ but remains of order unity. The first term in the rhs ($T$
196: term) is the so called spectral energy transfer. In the case of a forced
197: turbulence (Kolmogorov mechanism), this term is negative in the low wave number, to balance
198: with the production, it is approximately zero throughout the inertial
199: range and become positive in the dissipative range to balance with
200: dissipation.
201: It accounts for the non linear triad interaction responsible for the
202: forward cascade in forced
203: turbulence and must verify $\int_0^{\infty}dk\,T(k)=0$. The second term is the exact contribution of viscosity in the spectral
204: evolution equation and it is responsible for
205: dissipation at high wave number in forced turbulence. The last term, on
206: the other hand, has never been written
207: to the author's knowledge. It is this term which makes RT flows and
208: buoyant flows in general so different. As expected intuitively, it depends on the concentration
209: spectrum because it is a concentration heterogeneity that induces motion through buoyancy. In order to distinguish the influence
210: of all three terms, knowledge of the order of magnitude of
211: $T(k)$ is required. This term has contribution from
212: $u^{\prime}_j{u^{\prime}_i}^2$ which in spectral language means
213: $E^{3/2}$ (power counting of $u^{\prime}$ is $3$ which means $3/2$ in
214: term of $E$) and more precisely
215: $k^{3/2}E^{3/2}$ for homogeneity reason. The same sort of
216: argument is used with the pressure term $u^{\prime}_j p^{\prime}$ which
217: is a non local term with two sources : pure advection and
218: buoyancy. Thus, the pressure contribution brings another
219: $k^{3/2}E^{3/2}$ due to pure advection and a ${\cal A}g
220: E_c^{1/2}E^{1/2}$ for buoyancy (power counting of
221: $u^{\prime}$ and $c^{\prime}$ in buoyancy yields $1/2$ for $E$ and $E_c$). And finally,
222: $\nu\,\partial_j{u^{\prime}_i}^2$ is responsible for a $\nu k^2 E$
223: contribution. In a nutshell, $T(k)$ has
224: three spectral contributions whose order of magnitude are
225: (i) $k^{3/2}E^{3/2}$, (ii) $\nu k^2 E$ and (iii) ${\cal A}\,g
226: E_c^{1/2}E^{1/2}$. Thus, the evolution of the velocity spectrum
227: density in Eq.(\ref{eqspec}) is also controlled by these three
228: contributions. It is enlightening to draw on a plot (Fig.\ref{zone}) the
229: predominance domains of the non linearity
230: (i), viscosity (ii) and buoyancy
231: (iii) with $k$ on the horizontal axis and $E(k)$ on the vertical axis.
232: It is done by equating the three terms two by two.
233: It gives three boundary lines in a log-log plot which will
234: be referred to as the non-linearity-buoyancy (NLg), the non-linearity-viscosity
235: (NL$\nu$) and the buoyancy-viscosity (g$\nu$) boarders respectively
236: described by
237: \begin{eqnarray}
238: E_{\mathrm{NLg}}(k,t)&\sim&{\cal A}g\,k^{-3/2}\,E_c(k,t)^{1/2}~,\label{nlg}\\
239: E_{\mathrm{NL}\nu}(k)&\sim&\nu^2\,k~,\label{nlnu}\\
240: E_{\mathrm{g}\nu}(k,t)&\sim&\left({\cal A}g\right)^2\nu^{-2}\,k^{-4}\,E_c(k,t)\label{gnu}~.
241: \end{eqnarray}
242: \begin{figure}
243: \begin{center}
244: \input{zone.tex}
245: \end{center}
246: \caption{\label{zone}Predominance diagram of various terms in the
247: spectral equation (\ref{eqspec}). In this diagram, the point
248: $\lambda$ evolves on the $\lambda$ line following
249: $k_l(t)$. The gray curve represents the velocity spectrum
250: in the RT self similar regime.}
251: \end{figure}
252: Firstly, $E_c(k,t)$ may not have the same power law at
253: low wave number (NLg) in (\ref{nlg}) and at high wave number (NL$\nu$) in (\ref{gnu}). In
254: the self similar regime $\langle{c^{\prime}}^2\rangle$ tends to be a constant \cite{rist}
255: which will be denoted $c_0^2$. At low wave numbers, by applying Eq.(\ref{eqinj}) to
256: $\langle{c^{\prime}}^2\rangle$, it is found that $E_c(k,t)$ must vary
257: like $c_0^2\,k_l^{{n_c}_l-1}(t)\,k^{-{n_c}_l}\sim c_0^2\left({\cal A}g\right)^{1-{n_c}_l}\,t^{2(1-{n_c}_l)}\,k^{-{n_c}_l}$, where ${n_c}_l$ is undetermined yet.
258: That is why, in the self similar regime and at
259: low wave number, Eq. (\ref{nlg}) may be refined as
260: \begin{equation}
261: E_{\mathrm{NLg}}(k)\sim c_0\left({\cal
262: A}g\right)^{\frac{3-{n_c}_l}{2}}\,t^{1-{n_c}_l}\,k^{-\frac{3+{n_c}_l}{2}}~.\label{nlg2}
263: \end{equation} It is to
264: be noticed that when ${n_c}_l=1$, the previous time evolution is
265: changed to $\log(t)$. Secondly, it is worth noticing that $E_{\mathrm{NL}\nu}(k)\sim\nu^2\,k$
266: is independant of the concentration spectrum $E_c(k,t)$, that it does not depend on time and that it is the exact
267: line where the inertial cascade vanishes and where
268: dissipation starts acting in the Kolmogorov mechanism.
269:
270: It is now possible to describe the evolution of the
271: velocity spectrum in an RT mixing layer. Let us imagine that a peaked initial condition is chosen at a wave
272: number below $\left({\cal A}g/\nu^2\right)^{1/3}$. Spontaneous RT
273: can then occur. The closest to $\left({\cal
274: A}g/\nu^2\right)^{1/3}$, the fastest is the linear growth. The spectrum
275: grows until it reaches the NL domain where mode
276: coupling can start. At the beginning, since both $E(k)$ and $E_c(k)$
277: are peaked around the initial wave number, so is the buoyancy
278: production which then acts like a narrow band forcing. A Kolmogorov
279: spectrum can then settle between the initial wave number and
280: NL$\nu$ (where energy is dissipated). This is precisely what is
281: observed after the linear growth regime in the mixing phase. As time
282: goes by, both spectrums spread over a wider range of wave numbers and when
283: mixing is established, production becomes broad band and spectral energy transfer must
284: balance with buoyancy at the lowest wave number. That is why, the velocity spectrum
285: lies alongside $E_{\mathrm{NLg}}(k)$ (see Fig.\ref{zone}), given by (\ref{nlg}) and also
286: (\ref{nlg2}) in the self similar RT case.
287:
288: This balance
289: mechanism is observed for Rayleigh-B\'enard (RB) flows for it
290: explains its velocity and concentration spectrum layout. RT and RB flows
291: are both governed by the
292: same set of equation (concentration is replaced by temperature in RB)
293: but boundary conditions in RB are independent of time and so are the
294: velocity and concentration spectrum. At low wave numbers,
295: $\partial_t{c^{\prime}}^2\sim
296: -\partial_j\left(u^{\prime}_j{c^{\prime}}^2\right)$ which in
297: spectral language approximately means $\partial_t
298: E_c\sim\partial_k\left(k^{5/2}E^{1/2}E_c\right)$. If $E_c$ is
299: independent of time at low wave numbers in the self similar regime,
300: that means the spectral transfer of concentration $\sim
301: k^{5/2}E^{1/2}E_c$ must be independent of $k$, that is to say
302: $\frac{5}{2}-\frac{n_l}{2}-{n_c}_l=0$ for RB flows (power counting of
303: $k$). Moreover, since (\ref{nlg}) must be
304: valid in RB flows (Eq. (\ref{nlg2}) is not valid in this case
305: because RT boundary conditions have been used to derive it) the relation
306: $-n_l=-\frac{3}{2}-\frac{{n_c}_l}{2}$ must also be true. Together
307: with the previous relation it is then found that
308: $n_l=\frac{11}{5}=2.2$ and ${n_c}_l=\frac{7}{5}=1.4$, which is exactly
309: the Bolgiano-Obukov (BO) \cite{bolg} scaling found in RB flows \cite{rb, nature}.
310: This mechanism is also corroborated by experiment \cite{andr3} where
311: velocity spectrum level at low wave numbers can be seen to decrease
312: in time although the turbulent kinetic energy increases. That means,
313: (\ref{nlg2}), that ${n_c}_l\gtrsim 1$, which is confirmed by all
314: numerical and experimental results, and that, once again,
315: Kolmogorov mechanism is ruled out since it predicts a velocity
316: spectrum level increase which would require ${n_c}_l=1/3<1$.
317:
318: In the high wave numbers, depending on the value of the Schmidt number, two
319: things can happen. $Sc\lesssim 1$ means that buoyancy production
320: cannot exist below $\nu^2\,k$ and cannot balance with viscosity to
321: create $g\nu$ (see Fig.\ref{zone}). In this case, the velocity spectrum exponentially decreases
322: after it goes through $NL\nu$. Otherwise, if $Sc\gg 1$ it is possible to prove
323: that an equilibrium between buoyancy and viscosity exists by performing a
324: linear analysis of (\ref{eqc},\ref{equ},\ref{eqdiv}) with
325: $c(\bm{x},t)=c_0+c_1(\bm{k},t)\,e^{-i\bm{k}\bm{x}}$ and $\bm{u}(\bm{x},t)=\bm{u}_0+\bm{u}_1(\bm{k},t)\,e^{-i\bm{k}\bm{x}}$. This solution can
326: be thought of as the low wave number contributions of $\bm{u}$ and
327: $c$ with perturbation corresponding to higher wave number
328: turbulent fluctuations. They can be plugged in equations
329: (\ref{eqc},\ref{equ},\ref{eqdiv}) and after some algebra it is found
330: that
331: \begin{eqnarray}
332: c_1(\bm{k},t)&=&c_1^0(\bm{k})\,e^{(i\bm{u}_0\bm{k}-\kappa k^2)t},\\
333: {u_1}_i(\bm{k},t)&\approx&\frac{g_j}{\nu\,k^2}\left(\delta_{ij}-\frac{k_i
334: k_j}{k^2}\right)
335: c_1^0(\bm{k})\,e^{(i\bm{u}_0\bm{k}-\kappa
336: k^2)t},\label{lin2}
337: \end{eqnarray} in the high Schmidt number limit. As a result and by definition of
338: $E(k)$ and $E_c(k)$ it is then found
339: \begin{equation}
340: E(k,t)=\frac{8 \left({\cal A}\,g\right)^2}{\nu^2 k^4}E_c(k,t)~,
341: \end{equation}in agreement with Eq.(\ref{gnu}). Therefore, the velocity
342: spectrum coincide with $E_{g\nu}(k)$ in the high Schmidt number limit
343: (see Fig.\ref{zone}).
344:
345: As a conclusion, it can be stated that self
346: similarity hypothesis together with equilibrium of spectral energy
347: transfer with buoyancy at low wave numbers constrained velocity
348: spectrum in a way incompatible with Kolmogorov mechanism. It does not
349: provide the exact slope of the concentration spectrum ${n_c}_l$ at low
350: wave numbers but assuming that it lies between $1$ and $2$ we obtain
351: $2\leq n_l\leq 2.5$. Recent high resolution simulations, show that the
352: velocity spectrum level remains close to the $\lambda$ line which
353: would be in favor of ${n_c}_l\approx 1$ and $n_l\approx 2$ as a
354: result. The most important result, here, is that velocity spectrum at
355: low wave numbers is shown to be very sensitive to concentration
356: spectrum. Nevertheless, a Kolmogorov mechanism is not ruled out at
357: intermediate wave numbers. It can be a transition process from low wave
358: numbers to high wave numbers as in the BO mechanism. Most of the gravitational energy is injected through buoyancy at low wave numbers up to a critical wave number. Above it, injection of energy is negligible and a Kolmogorov mechanism is possible. In this case,
359: it does not involve that the turbulent dissipation $\epsilon=2\nu\int
360: dk\,k^2 E(k)$ should vary like $\epsilon\sim t$. As mentioned in
361: \cite{rist}, this common belief that ``the cascade rate is
362: [...] equal to dissipation'' in the self similar regime is always
363: presupposed although it is a strong hypothesis for buoyant flows. It
364: does not have to hold considering the importance of the buoyant
365: production. It would not be in contracdiction with turbulent kinetic
366: energy conservation
367: $\partial_t\langle{u^{\prime}_i}^2\rangle=\Pi-\epsilon$. Indeed, since buoyancy
368: production $\Pi\sim t$ and
369: $\partial_t\langle{u^{\prime}_i}^2\rangle\sim t$, dissipation could
370: vary slowlier without threatening energy conservation in the self similar regime.
371:
372:
373: \begin{thebibliography}{99}
374: \bibitem{rayl} Lord~Rayleigh, Proc. R. Math. Soc. {\bf 14}, 170 (1883).
375: \bibitem{tayl} G.~I.~Taylor, Proc. R. Soc. Lond. A {\bf 201}, 192
376: (1950).
377: \bibitem{astro1} M.~Zingale {\it et al.}, Astrophys.J. {\bf 632}, 1021 (2005).
378: \bibitem{astro2} R.~M.~Hueckstaedt {\it et al.}, to be published in MNRAS, astro-ph/0603807 (2006).
379: \bibitem{icf2} S.~P.~Regan {\it et al.}, Phys. Rev. Lett. {\bf 89},
380: 085003 (2002).
381: \bibitem{icf1} S.~Fujioka {\it et al.}, Phys. Rev. Lett. {\bf 92}, 195001 (2004).
382: \bibitem{young} Y.~N.~Young, H.~Tufo, A.~Dubey \& R.~Rosner, J. Fluid
383: Mech. {\bf 443}, 69 (2001).
384: \bibitem{cook2} A.~W.~Cook \& Y.~Zhou, Phys. Rev. E {\bf 66}, 026312 (2002).
385: \bibitem{csz} W.~Cabot, O.~Schilling \& Y.~Zhou, Phys. Fluids {\bf 16}, 495 (2004).
386: \bibitem{dalz1} S.~Dalziel, P.~Linden \& D.~Youngs, J. Fluid
387: Mech. {\bf 399}, 1 (1999).
388: \bibitem{cook1} A.~W.~Cook \& P.~E.~Dimotakis, J. Fluid Mech. {\bf
389: 443}, 69 (2001).
390: \bibitem{dim1} G.~Dimonte {\it et al.}, Phys. Fluids {\bf 16}, 1668 (2004).
391: \bibitem{rama1} P.~Ramaprabhu, G.~Dimonte \& M.~J.~Andrews, J. Fluid
392: Mech. {\bf 536}, 285 (2005).
393: \bibitem{cook3} A.~W.~Cook, W.~Cabot \& P.~L.~Miller, J. Fluid
394: Mech. {\bf 511}, 333 (2004).
395: \bibitem{cab1} W.~Cabot, Phys. Fluids {\bf 18}, 045101 (2006).
396: \bibitem{andr3} P.~Ramaprabhu \& M.~Andrews, J. Fluid Mech.{\bf 502}, 233 (2004).
397: \bibitem{andr1} P.~Wilson, M.~Andrews \& F.~Harlow, Phys. Fluids {\bf 11}, 2425 (1999).
398: \bibitem{andr2} P.~Wilson \& M.~Andrews, Phys. Fluids {\bf 14}, 938 (2002).
399: \bibitem{chert1} M.~Chertkov, Phys. Rev. Lett. {\bf 91}, 115001
400: (2003).
401: \bibitem{zhou} Y.~Zhou, Phys. Fluids {\bf 13}, 538 (2001).
402: \bibitem{rist} J.~R.~Ristorcelli \& T.~T.~Clark, J. Fluid Mech. {\bf
403: 507}, 213 (2004).
404: \bibitem{sbp} S~.B.~Pope, {\it Turbulent Flows}, Cambridge University
405: Press (2000).
406: \bibitem{bolg} R. Bolgiano, J. Geophys. Res. {\bf 64}, 2226 (1959).
407: \bibitem{rb} X.~Shang, K.~Xia, Phys. Rev. E {\bf 64}, 065301 (2001).
408: \bibitem{nature} J.~J.~Niemela, L.~Skrbek, K.~R.~Sreenivasan \&
409: R.~J.~Donnelly, Nature {\bf 404}, 837 (2000)
410: \end{thebibliography}
411: \end{document}
412: