1: \documentclass[preprint,showpacs,aps]{revtex4}
2: %\documentclass[multicol,showpacs,aps]{revtex4}
3:
4: %\documentstyle[preprint,aps]{revtex}
5: \usepackage{graphicx}% Include figure files
6: %\usepackage{dcolumn}% Align table columns on decimal point
7: \usepackage{bm}% bold math
8: % \usepackage{feynmf}
9: \usepackage{psfig}
10:
11: %\input{./figs/gnuplot-latex/plotcommand.tex}
12: \input{./eq.tex}
13: \begin{document}
14: \title{Large-Eddy Simulations of Fluid and Magnetohydrodynamic Turbulence Using Renormalized Parameters}
15: \author{Mahendra\ K.\ Verma and Shishir Kumar}
16: \affiliation{Department of Physics, Indian Institute of Technology, Kanpur -- 208016, INDIA}
17: \date{March 7, 2003}
18:
19: \begin{abstract}
20: In this paper a procedure for large-eddy simulation (LES) has been
21: devised for fluid and magnetohydrodynamic turbulence in Fourier
22: space using the renormalized parameters. The parameters calculated
23: using field theory have been taken from recent papers by Verma
24: \cite{MKV:MHD_PRE,MKV:MHDRG}. We have carried out LES on $64^3$
25: grid. These results match quite well with direct numerical
26: simulations of $128^3$. We show that proper choice of parameter is
27: necessary in LES.
28: \end{abstract}
29:
30: \vspace{1cm}
31: \pacs{47.27.Eq, 47.65.+a, 11.10.Gh}
32:
33: \maketitle
34:
35: Turbulence is one of the most difficult and unsolved problems of
36: classical physics. To probe the complex dynamics of turbulence, one
37: often resorts to computer experiments, known as Direct Numerical
38: Simulations (DNS). Since multiple scales are involved in turbulence,
39: DNS of turbulence is a very expensive task in terms of both computer
40: time and memory, even in modern computers. For example, a
41: pseudo-spectral simulation by Gotoh~\cite{Goto:DNS} on $1024^3$ grid
42: using vector parallel Fujitsu VPP5000/56 with 32 processors took 500
43: hours of computer time, and required 8 Gigabytes of memory per
44: processor. To reduce the required computer time and memory space, an
45: ingenious technique called large-eddy simulation (LES) has been
46: developed (see review article by Metais~\cite{Meta:rev} and references
47: therein).
48:
49: Basic idea of LES is to resolve only the large scales of turbulent
50: flow. The effect of smaller scale interactions are modeled
51: appropriately using the existing theories. In turbulence, Fourier
52: modes of different scales interact with each other. Kolmogorov
53: provided an important model of turbulence in which the interactions
54: effectively yield a constant energy flux from large scales to
55: intermediate scales, and then to small scales. When we observe
56: Fourier modes up to certain length scale $l$ in the intermediate
57: range, the modes with scales less than $l$ act as a sink of energy.
58: According to Kolmogorov's theory, the amount of sink should be equal
59: to the energy flux. In LES, the large scales up to $l$ are resolved by
60: using eddy viscosity at cutoff scale $l$, where energy is drained.
61: Analysis of turbulence using renormalization groups (RG) shows that
62: the above modeling is possible. LES uses this idea to analyze
63: large-scale dynamics of turbulence.
64:
65: Renormalization Group (RG) is a popular tool used by physicists to
66: solve problems with multiple scales. Since turbulence involves
67: multiple scales, RG has been applied successfully to turbulence
68: \cite{FNS,YakhOrsz,McCo:book}. In Wilson's Fourier space RG scheme,
69: Fourier space is divided into many shells. The nonlinear interactions
70: among various shells are computed using first-order perturbation
71: theory, that yields an effective viscosity, called renormalized or
72: eddy viscosity, at any scale. The renormalized viscosity is found to
73: be wavenumber ($k$) dependent. McComb and Watt~\cite{McCoWatt}
74: computed the renormalized viscosity using `self-consistent' RG
75: procedure. When the cutoff wavenumber $k_C$ is in the inertial range,
76: the renormalized viscosity is given by
77: \begin{equation}
78: \nu_r(k_C) = (K)^{1/2} \Pi^{1/3} k_C^{-4/3} \nu^*
79: \label{eqn:nu}
80: \end{equation}
81: where $\Pi$ is the energy flux, $K$ is Kolmogorov's constant, and
82: $\nu^*$ a parameter. McComb and Watt~\cite{McCoWatt} found $\nu_*
83: \approx 0.50$ and $K \approx 1.62$. Verma
84: \cite{MKV:MHD_PRE} also computed the above quantities using a refined
85: technique and found $\nu_* \approx.38$ and $K \approx 1.6$.
86:
87: Zhou and Vahala~\cite{ZhouVaha88,ZhouVaha93} developed an alternative
88: recursive-renormalization-group theory for turbulence modeling. In
89: their calculation they find backscatter of energy from small scales to
90: large scales, and a cusp in renormalized viscosity near $k_C$.
91: These features are attributed to triple correlations, which has not been
92: accounted for in McComb and Watt's calculations. Recently Schilling and
93: Zhou~\cite{SchiZhou} have addressed the above problem using
94: eddy-damped quasinormal Markovian (EDQNM) closure model. In the current
95: paper we neglect backscatter.
96:
97: McComb~\cite{McCo:book} had proposed that the renormalized viscosity
98: $\nu(k_C)$ could be used as effective viscosity for LES, however, this
99: calculation had not been done till date. Earlier, the spectral eddy
100: viscosity $\nu_t(k|k_C)$ has been used for LES in EDQNM formalism
101: (see~\cite{Lesi:book}). In this scheme,
102:
103: \begin{equation}
104: \nu_t(k|K_C) = 0.441 K^{-3/2} \left[ \frac{E(k_C)}{k_C} \right]^{1/2}
105: f(k/k_C)
106: \label{eqn:nu_t}
107: \end{equation}
108: where $f(x)$ is a nondimensional function which tends to 1 as $x$
109: approaches 0. Comparing Eqs.~(\ref{eqn:nu},\ref{eqn:nu_t}), we find
110: that their dependence on Kolmogorov's constant is different. In
111: Eq.~(\ref{eqn:nu_t}), if we assume that $E(k_C)$ follows Kolmogorov's
112: spectrum and $K=1.6$, then the constant multiplying $\Pi^{1/3}
113: k_C^{-4/3}$ is 0.27. In contrast, in Eq.~(\ref{eqn:nu}) the same
114: quantity is $\sqrt{K} \nu^* \approx 0.48$. As it will be shown
115: in the later part of the paper, the choice of constant is quite
116: crucial in LES. We find that $\nu_r(k_C)$ of Eq.~(\ref{eqn:nu})
117: yields better numerical results compared to $\nu_t(k|k_C)$ of
118: Eq.~(\ref{eqn:nu_t}). We believe that the calculation of renormalized
119: viscosity is theoretically more sound than the calculation of spectral
120: eddy viscosity using EDQNM approximation, therefore, former is
121: more appropriate for LES than the later.
122:
123: In this paper we perform LES of fluid turbulence using renormalized
124: viscosity. We have been able to apply the same procedure to
125: magnetohydrodynamic (MHD) turbulence also, except that we need two
126: renormalized parameters: renormalized viscosity and renormalized
127: resistivity. The required parameters for MHD have been recently
128: calculated by Verma \cite{MKV:MHD_PRE,MKV:MHDRG,MKV:MHDflux}. The LES
129: calculations have been performed on $64^3$ grid, and they have been
130: compared with DNS results of $64^3$ and $128^3$. As described below,
131: the inertial range in LES is found to be either equal or larger than
132: that in DNS, hence our LES model is working very well.
133:
134: We solve Navier-Stokes equation in Fourier space~\cite{Orsz}:
135: \begin{equation}
136: \frac{\partial u_i ({\bf k})}{\partial t} =
137: -\nu_r(k_C) k^2 u_i ({\bf k}) - FT\left(
138: u_j \frac{\partial u_i}{\partial x_j} \right)- i k_i p({\bf k})
139: \label{eqn:NSk}
140: \end{equation}
141: where $FT$ stands for Fourier transform. We take $\nu^*$ to be equal to 0.38.
142:
143: We adopt pseudo spectral method on grid size $64^3$ with $dt=10^{-4}$.
144: We apply Adam-Bashforth scheme to integrate the nonlinear terms, and
145: Crank-Nicholson's scheme for the viscous term. We apply 2/3 rule to
146: eliminate the aliasing errors \cite{Orsz}. We use Fast Fourier
147: Transform developed by Frigo and Johnson~\cite{fftw} for our
148: calculations. Our initial condition is taken to be unit energy spread
149: out in wavenumber shells from 2 to 13 with an exponentially decreasing
150: distribution. The modes in a shell have equal energy but random
151: phases, and satisfy divergenceless condition. The most important
152: ingredient in our simulation is renormalized viscosity, which is
153: computed using Eq.~(1) with $k_C=32$. Since $\Pi$ changes with time,
154: it is computed every 0.01 dimensionless time unit. We use dissipation
155: rate for $\Pi$. We carry out our simulation up to 50 time units. Our
156: LES simulation takes approximately 60 hours on Athlon 1.7 GHz
157: processor.
158:
159: In Fig.~\ref{fig:et_fluid} we show the energy evolution as a function of
160: time for $\nu^*= 0.25,0.38,0.48$. The $E$ vs.~$t$ plot for all three
161: $\nu^*$ are overlapping. The LES results are also compared with the
162: standard pseudo-spectral DNS results performed on $64^3$ (DNS64) and
163: $128^3$ (DNS128) with identical initial condition and $\nu_0=2 \times
164: 10^{-4} $. In DNS we apply additional hyperviscous term $1/k_{eq}^2
165: k^4 {\bf u(k)}$ with $k_{eq}=9$ to overcome aliasing errors. Clearly
166: the energy evolution for LES matches quite well with DNS128, but
167: differ significantly with DNS64. Hence, our LES on $64^3$ is able to
168: mimic DNS of $128^3$.
169:
170: In Fig.~\ref{fig:ek_fluid}, we plot $E(k) k^{5/3} \Pi^{-2/3}$ vs. $k$
171: for DNS as well as LES. Again the normalized spectrum of LES matches
172: quite well with DNS128 at small and intermediate wavenumbers. Note
173: that $64^3$ LES has much larger inertial range compared to $64^3$ DNS,
174: where it is almost absent. We find that the wavenumber range of
175: inertial wavenumbers (constant with $k$) is maximum for LES with
176: $\nu^*=0.38$; in fact wavenumber range for LES is larger than that for
177: DNS128. The energy spectrum for $\nu^*=0.25$ has a hump for large
178: wavenumbers (underdamped case), implying that actual $\nu^*$ value is
179: higher than 0.25. The spectrum for $\nu^*=0.48$ shows overdamped
180: character \cite{Math:book}. We have done DNS128 for some more
181: parameters. The trend appears to show that $\nu^* \approx 0.38$ is
182: the most appropriate choice for LES. Fortunately, we obtain the above
183: value using renormalization group calculation~\cite{MKV:MHDRG}. It is
184: interesting to note from Fig.~\ref{fig:et_fluid} that the temporal
185: evolution of energy does not clearly tell us which $\nu^*$ is the most
186: appropriate for LES. Hence we should be careful in concluding the
187: appropriateness of $\nu^*$ using energy evolution. The energy
188: spectrum has more information, and can provide us clues on the correct
189: choice of $\nu^*$.
190:
191: From Fig.~\ref{fig:ek_fluid} we obtain the numerical value of $K$ to be
192: $1.7 \pm 0.1$; this value is close to the theoretically
193: calculated value 1.6 \cite{McCoWatt,MKV:MHD_PRE}. Hence, the renormalized
194: viscosity predicted by Verma~\cite{MKV:MHD_PRE} appears to be
195: consistent and provides us a very good scheme for LES.
196:
197: For LES of magnetohydrodynamic (MHD) turbulence, Agullo et
198: al.~\cite{MullCara_pop1}, and M\"{u}ller and Carati
199: \cite{MullCara_pop2,MullCara_cpc} applied dynamic gradient-diffusion
200: subgrid model. The forms of eddy-viscosity and eddy-resistivity are derived
201: using dimensional arguments, but the constants are calculated using
202: dynamical LES procedure. Their results match very well with DNS
203: counterpart. In one of their main models, turbulent viscosity $\nu_t
204: \approx \bar{l}^{4/3} (\epsilon^K)^{1/3}$ and turbulent resistivity
205: $\eta_t \approx \bar{l}^{4/3} (\epsilon^M)^{1/3}$, where $\bar{l}$ is
206: the resolvable length scale on the LES grid, and $\epsilon^K$ and
207: $\epsilon^M$ are kinetic and magnetic energy dissipation applied by
208: the subgrid scale respectively. Zhou et al.~\cite{Zhou:MHD_LES} have
209: studied subgrid scale and backscatter model for MHD turbulence using
210: EDQNM closure scheme. Verma \cite{MKV:MHD_PRE,MKV:MHDRG} has also
211: calculated the above parameters using renormalization group procedure.
212: Simple calculations show that turbulent dissipative parameters of
213: Verma differ significantly from those of Agullo et
214: al.~\cite{MullCara_pop1} and M\"{u}ller and Carati
215: \cite{MullCara_pop2,MullCara_cpc}, as well as from those of Zhou et
216: al.~\cite{Zhou:MHD_LES}. In the following discussions we will compare
217: the LES results from the above three approaches.
218:
219:
220: For MHD turbulence we apply the same LES method as described for fluid
221: turbulence using renormalized parameters. The pseudo-spectral method
222: to solve MHD equations is very similar to that of fluid turbulence.
223: We also confine ourselves to zero cross helicity, i.e., (${\bf u \cdot
224: b}=0$), and zero mean magnetic field. The difference of LES and DNS
225: is in the values of viscosity and resistivity. In DNS we take
226: $\nu_0=0.00015$ and $\eta_0=0.00015$ with hyperviscosity and
227: hyperesistivity parameters $k_{eq}=7$. However, in LES we take
228: $\nu(k_C)=\nu_{r}(k_C)$, and $\eta(k_C)=\eta_{r}(k_C)$, where $k_C$ is
229: the cutoff wavelength. The renormalized viscosity $\nu_r(k_C)$, and
230: renormalized resistivity $\eta(k_C)$ are taken from Verma
231: \cite{MKV:MHD_PRE,MKV:MHDRG} as
232: \begin{eqnarray}
233: \nu_r(k_C) & = & (K^u)^{1/2} \Pi^{1/3} k_C^{-4/3} \nu^*
234: \label{eqn:nuk} \\
235: \eta_r(k_C) & = & (K^u)^{1/2} \Pi^{1/3} k_C^{-4/3} \eta^* .
236: \label{eqn:etak}
237: \end{eqnarray}
238: Here $K^u$ is Kolmogorov's constant for MHD, $\Pi$ is the total energy
239: flux, and $\nu^*, \eta^*$ are renormalized parameters. The parameters
240: $\nu^*, \eta^*$, and $K^u$ depend on the Alfv\'{e}n ratio $r_A$, which
241: is the ratio of kinetic and magnetic energy. In our decaying MHD
242: turbulence simulation, we start with unit total energy and $r_A=8.0$.
243: The ratio of magnetic to kinetic energy grows as a function of time as
244: expected. Therefore, we need to compute the renormalized parameters
245: for various values of $r_A$. The parameters have been calculated
246: using the procedure described in Verma~\cite{MKV:MHDRG}, and they are
247: shown in Table 1. We use the appropriate $\nu^*$ and $\eta^*$
248: given in the table for our simulations. The energy cascade rates are
249: computed using Fast Fourier Transforms~\cite{fftw}. We take
250: $\nu_{r}(k_C)$ and $\eta_{r}(k_C)$ from Eqs.~(\ref{eqn:nuk},
251: \ref{eqn:etak}). The energy flux $\Pi$ changes with time; we compute
252: $\Pi$ dynamically every $0.01$ time-unit.
253: We carried out LES for MHD up to 25 nondimensional time units, and it
254: took approximately 55 hours.
255:
256: The evolution of kinetic and magnetic energies are shown in
257: Fig.~\ref{fig:et_mhd} as a function of time. The evolution of kinetic
258: energy using LES is quite close to that using DNS. However, the
259: evolution of magnetic energy does not match very well. Comparatively,
260: LES of Agullo et al.~\cite{MullCara_pop1} and M\"{u}ller and Carati
261: \cite{MullCara_pop2,MullCara_cpc} yield a better fit to the temporal
262: evolution of energy. Fig.~\ref{fig:ek_mhd} shows the energy
263: spectra of kinetic and magnetic energies for $r_A=0.5$ at 27 time
264: units of DNS and 12 time units of LES. We find that the energy spectra
265: calculated in LES matches quite well with that in DNS. The
266: Kolmogorov's constant as indicated by the straight line in upper part
267: of Fig.~\ref{fig:ek_mhd} is found to be $1.8 \pm 0.2$, which is
268: close to the theoretical value calculated in
269: \cite{MKV:MHD_PRE,MKV:MHDRG}. We conclude that the LES based on
270: renormalized parameters of Verma \cite{MKV:MHD_PRE,MKV:MHDRG} is quite
271: good. Our numerical results are comparable with results of Agullo et
272: al.~\cite{MullCara_pop1} and M\"{u}ller and
273: Carati~\cite{MullCara_pop2,MullCara_cpc}. However, we believe that
274: our parameters, which are based on field-theoretic calculations, are
275: on a somewhat stronger footing as compared to those used in earlier
276: LES methods.
277:
278: To conclude, we have devised a LES procedure for fluid and MHD
279: turbulence in Fourier space using the renormalized parameters. We
280: take renormalized parameters from Verma~\cite{MKV:MHD_PRE,MKV:MHDRG}
281: and carry out LES for $64^3$ grid. When LES results are compared with
282: DNS of size $128^3$ with the same initial conditions, we find that our
283: LES results on energy evolution and spectra match quite well with the
284: DNS results, except for the temporal evolution of magnetic energy.
285: The inertial range of LES is much larger compared to DNS of the same
286: size. Our results shows that substitution of renormalized parameters
287: for eddy viscosity in LES yield excellent results. Hence, we
288: demonstrate the usefulness of renormalized parameters in LES
289: calculations.
290:
291: %\bibliographystyle{prsty}
292: %\bibliography{/home/mkv/research/mypapers/bib/abbrev,/home/mkv/research/mypapers/bib/surf,/home/mkv/research/mypapers/bib/burg,/home/mkv/research/mypapers/bib/mhd,/home/mkv/research/mypapers/bib/fluid,/home/mkv/research/mypapers/bib/interm,/home/mkv/research/mypapers/bib/misc,/home/mkv/research/mypapers/bib/mkv,/home/mkv/research/mypapers/bib/dnsles}
293:
294: \begin{thebibliography}{10}
295:
296: \bibitem{MKV:MHD_PRE}
297: M.~K. Verma, Phys. Rev. E {\bf 64}, 26305 (2001).
298:
299: \bibitem{MKV:MHDRG}
300: M.~K. Verma, Phys. Plasma {\bf 8}, 3945 (2001).
301:
302: \bibitem{Goto:DNS}
303: T. Gotoh, Comput. Phys. Commun. {\bf 147}, 530 (2002).
304:
305: \bibitem{Meta:rev}
306: O. M{\'{e}}tais, in {\em New Trends in Turbulence: Les Houches Lectures},
307: edited by M. Lesieur, A. Yaglom, and F. David (Springer-Verlag, Berlin,
308: 2001), pp.\ 113--139.
309:
310: \bibitem{FNS}
311: D. Forster, D.~R. Nelson, and M.~J. stephen, Phys. Rev. A {\bf 16}, 732
312: (1977).
313:
314: \bibitem{YakhOrsz}
315: V. Yakhot and S.~A. Orszag, J. Sci. Comput. {\bf 1}, 3 (1986).
316:
317: \bibitem{McCo:book}
318: W.~D. McComb, {\em The Physics of Fluid Turbulence} (Oxford University Press,
319: Claredon, 1990).
320:
321: \bibitem{McCoWatt}
322: W.~D. McComb and A.~G. Watt, Phys. Rev. A {\bf 46}, 4797 (1992).
323:
324: \bibitem{ZhouVaha88}
325: Y. Zhou, G. Vahala, and M. Hussain, Phys. Rev. A {\bf 37}, 2590 (1988).
326:
327: \bibitem{ZhouVaha93}
328: Y. Zhou and G. Vahala, Phys. Rev. E {\bf 47}, 2503 (1993).
329:
330: \bibitem{SchiZhou}
331: O. Schilling and Y. Zhou, Phys. Fluids {\bf 14}, 2002 (2002).
332:
333: \bibitem{Lesi:book}
334: M. Lesieur, {\em Turbulence in Fluids - Stochastic and Numerical Modelling}
335: (Kluwer Academic Publishers, Dordrecht, 1990).
336:
337: \bibitem{MKV:MHDflux}
338: M.~K. Verma, Pramana (submitted), nlin.CD/0103033 (2002).
339:
340: \bibitem{Orsz}
341: S.~A. Orszag and G.~S. Patterson, Phys. Rev. Lett. {\bf 28}, 76 (1972).
342:
343: \bibitem{fftw}
344: M. Frigo and S.~G. Jhonson, www.fftw.org .
345:
346: \bibitem{Math:book}
347: J. Mathieu and J. Scott, {\em The Introduction to Turbulent Flow} (Cambridge
348: University Press, Cambridge, 2000).
349:
350: \bibitem{MullCara_pop1}
351: O. Agullo, W.-C.Muller, B. Knaepen, and D. Carati, Phys. Plasma {\bf 8}, 3502
352: (2001).
353:
354: \bibitem{MullCara_pop2}
355: W.-C.Muller and D. Carati, Phys. Plasma {\bf 9}, 824 (2002).
356:
357: \bibitem{MullCara_cpc}
358: W.-C.Muller and D. Carati, Comput. Phys. Commun. {\bf 147}, 544 (2002).
359:
360: \bibitem{Zhou:MHD_LES}
361: Y. Zhou, O. Schilling, and S. Ghosh, Phys. Rev. E {\bf 66}, 26309 (2002).
362:
363: \end{thebibliography}
364:
365: \pagebreak
366:
367: \centerline{\large Figure Captions}
368: \vspace{1cm}
369:
370: \noindent
371: {\bf Fig. 1} \, Temporal evolution of energy in fluid turbulence using DNS
372: and LES. The figure contains Energy($E$) vs time plots for DNS128
373: (solid line), DNS (DNS64), and three LES runs using $\nu^*$ equal to
374: 0.38 (LES1), 0,25 (LES2) and 0.48 (LES3). The evolution in LES for
375: all the three $\nu^*$ is quite close to DNS128, but not to
376: DNS64. \\
377:
378: \noindent
379: {\bf Fig. 2} \, Energy spectrum for fluid turbulence is calculated using DNS
380: and LES. The figure contains plots of normalized energy spectrum
381: $E'(k)=E(k)k^{5/3}\epsilon^{-2/3}$ with wavenumber $k$ after 50 time
382: units for DNS128, DNS64, and three LES runs using $\nu^*$ equal to 0.38 (LES1),
383: 0,25 (LES2) and 0.48 (LES3). We get the best inertial range for
384: $\nu^*=0.38$. The Kolmogorov's constant is found to be $1.7 \pm
385: 0.1$. DNS64 run has hardly any inertial range. \\
386:
387: \noindent
388: {\bf Fig. 3} \, Temporal evolution of total kinetic and magnetic energy in
389: MHD turbulence using DNS and LES. The kinetic energy matches quite
390: well in both the schemes, but magnetic energy evolves somewhat
391: differently. \\
392:
393: \noindent
394: {\bf Fig. 4} \, Plots of normalized spectra
395: $E'(k)=E(k)k^{5/3}\epsilon^{-2/3}$ with wavenumber $k$ for MHD
396: turbulence. The straight line shows the value of $K_o$ for LES run. \\
397:
398: \pagebreak
399:
400: %
401: \begin{table}
402: \caption{\label{tab:table1}The values of renormalized parameters for
403: viscosity ($\nu^*$) and resistivity $\eta^*$ in MHD turbulence at
404: various values of Alfv\'{e}n ratio $r_A$ and zero cross helicity. We also
405: list the Kolmogorov's constant $K^u$ for MHD turbulence.}
406:
407: \begin{ruledtabular}
408: \begin{tabular}{cccc}
409: $r_A$ & $\nu^*$ & $\eta^*$ & $K^u$ \\
410: \hline
411: 0.3 & 7.20 & 0.20 & 0.50 \\
412: 0.4 & 3.15 & 0.38 & 0.53 \\
413: 0.5 & 2.08 & 0.50 & 0.55 \\
414: 0.6 & 1.64 & 0.57 & 0.59 \\
415: 0.7 & 1.38 & 0.61 & 0.63 \\
416: 0.8 & 1.21 & 0.64 & 0.67 \\
417: 0.9 & 1.09 & 0.67 & 0.71 \\
418: 1.0 & 1.00 & 0.69 & 0.75 \\
419: 2.0 & 0.65 & 0.77 & 1.01 \\
420: 3.0 & 0.54 & 0.79 & 1.15 \\
421: 4.0 & 0.49 & 0.81 & 1.23 \\
422: 5.0 & 0.47 & 0.82 & 1.28 \\
423: \end{tabular}
424: \end{ruledtabular}
425: \end{table}
426:
427:
428: \pagebreak
429:
430: % figures. tex/ps/caption files are in ./figs dir
431: %
432: % \plot{dir-name}{vspace}
433: % -- dir-name is dir in ./figs/ where the plot resides, vspace is the height to be left on top.
434: %
435: %\plot{et_fluid}{2.0in}
436: %\plot{ek_fluid}{2.0in}
437: %\plot{et_mhd}{2.0in}
438: %\plot{ek_mhd_pair}{0.0in}
439:
440: \newpage
441:
442: \begin{figure}[h]
443: \centerline{\mbox{\psfig{file=et_fluid.ps,width=1.0\textwidth}}}
444: \caption{Temporal evolution of energy in fluid turbulence using DNS
445: and LES. The figure contains Energy($E$) vs time plots for DNS128
446: (solid line), DNS (DNS64), and three LES runs using $\nu^*$ equal to
447: 0.38 (LES1), 0,25 (LES2) and 0.48 (LES3). The evolution in LES for
448: all the three $\nu^*$ is quite close to DNS128, but not to
449: DNS64.}
450: \label{fig:et_fluid}
451: \end{figure}
452:
453: \newpage
454:
455: \begin{figure}[h]
456: \centerline{\mbox{\psfig{file=ek_fluid.ps,width=1.0\textwidth}}}
457: \caption{Energy spectrum for fluid turbulence is calculated using DNS
458: and LES. The figure contains plots of normalized energy spectrum
459: $E'(k)=E(k)k^{5/3}\epsilon^{-2/3}$ with wavenumber $k$ after 50 time
460: units for DNS128, DNS64, and three LES runs using $\nu^*$ equal to 0.38 (LES1),
461: 0,25 (LES2) and 0.48 (LES3). We get the best inertial range for
462: $\nu^*=0.38$. The Kolmogorov's constant is found to be $1.7 \pm
463: 0.1$. DNS64 run has hardly any inertial range.}
464: \label{fig:ek_fluid}
465: \end{figure}
466:
467: \newpage
468:
469: \begin{figure}[h]
470: \centerline{\mbox{\psfig{file=et_mhd.ps,width=1.0\textwidth}}}
471: \caption{Temporal evolution of total kinetic and magnetic energy in
472: MHD turbulence using DNS and LES. The kinetic energy matches quite
473: well in both the schemes, but magnetic energy evolves somewhat
474: differently.}
475: \label{fig:et_mhd}
476: \end{figure}
477:
478: \newpage
479:
480: \begin{figure}[h]
481: \centerline{\mbox{\psfig{file=ek_mhd.ps,width=1.0\textwidth}}}
482: \caption{Plots of normalized spectra
483: $E'(k)=E(k)k^{5/3}\epsilon^{-2/3}$ with wavenumber $k$ for MHD
484: turbulence. The straight line shows the value of $K_o$ for LES run.}
485: \label{fig:ek_mhd}
486: \end{figure}
487:
488: \end{document}
489: