1: \documentclass[]{emulateapj}
2: \usepackage{graphicx}
3: \shorttitle{Density scalings in supersonic MHD}
4: \begin{document}
5: \title{Density scaling and anisotropy
6: in supersonic MHD turbulence}
7: \author{A. Beresnyak, A. Lazarian}
8: \affil{Dept. of Astronomy, Univ. of Wisconsin, Madison, WI 53706}
9: \email{andrey, lazarian@astro.wisc.edu}
10: \and
11: \author{J. Cho}
12: \affil{Chungnam National Univ., Korea}
13: \email{cho@canopus.chungnam.ac.kr}
14:
15: \begin{abstract}
16: We study the statistics of density for supersonic turbulence in a medium
17: with magnetic pressure larger than the gaseous pressure. This study is
18: motivated by molecular cloud research.
19: Our simulations exhibit clumpy density structures, which
20: contrast increases with the Mach number.
21: At 10 Machs densities of some clumps are three orders of magnitude higher than
22: the mean density. These clumps give rise to flat and approximately isotropic density spectrum
23: corresponding to the random distribution of clumps in space.
24: We claim that
25: the clumps originate
26: from our random, isotropic turbulence driving.
27: When the contribution from those clumps is suppressed by studying logarithm of density,
28: the density statistics exhibit scale-dependent anisotropy consistent
29: with the models where density structures arise from shearing by Alfv\'en waves.
30: It is noteworthy that originally such models were advocated for the case
31: of low-Mach, nearly incompressible turbulence.
32:
33: \end{abstract}
34:
35: \keywords{turbulence: compressible, molecular clouds, star formation}
36:
37: \section{Introduction}
38: The paradigm of interstellar medium has undergone substantial changes
39: recently. Instead of quiescent medium with hanging and slowly evolving
40: clouds a turbulent picture emerged (see review by V\'azquez-Semadeni et al., 2000).
41: With magnetic field being dynamically important and dominating the gas
42: pressure in molecular clouds, this calls for studies of compressible
43: magnetohydrodynamic (MHD) turbulence.
44:
45: Recent years have been also marked by a substantial progress in
46: understanding of the MHD turbulence statistics (see review
47: by Cho \& Lazarian 2005 and references therein). This statistics
48: allows to find regularities in turbulence, e.g. power spectrum
49: allows to learn how much energy is at a particular range of scales.
50:
51: A very important insight into the
52: incompressible MHD turbulence by Goldreich \& Shridhar (1995) (henceforth GS95)
53: has been followed by progress in understanding of compressible MHD turbulence
54: (Lithwick \& Goldreich 2001, Cho \& Lazarian 2003, henceforth CL03, Cho, Lazarian \& Vishniac 2003, Vestuto et al. 2003).
55: In particular, simulation in Cho \& Lazarian (2003) showed that Alfv\'enic cascade evolves
56: on its own\footnote{
57: The expression proposed and tested in CL03 shows that the coupling of Alfv\'enic and
58: compressible modes is appreciable at the injection scale if the injection velocity
59: is comparable with the {\it total} Mach number of the turbulence, i.e. with
60: $(V_A^2+C_S^2)^{1/2}$, where $V_A$ and $C_S$ are the Alfv\'en and sound velocities
61: respectively. However, the coupling gets marginal at smaller scales as
62: turbulence cascades and
63: turbulent velocities get smaller.}
64: and it exhibits Kolmogorov type scaling (i.e. $E\sim k^{-5/3}$)
65: and scale-dependent
66: anisotropy of the Goldreich-Shridhar type (i.e. $k_{\|}\sim k_{\perp}^{2/3}$)
67: even for high Mach number turbulence. While slow modes exhibit similar
68: scalings and anisotropy, fast modes show isotropy. The density scaling
69: obtained in Cho \& Lazarian (2003) was somewhat puzzling. At low Mach
70: numbers it was similar to slow modes, while it got isotropic
71: for high Mach numbers.
72:
73: The uncertainties associated with the earlier study motivate our present
74: one. Density statistics is important for understanding the structure
75: of molecular clouds and the associated processes of star formation.
76: Are density perturbations tend to be elongated along magnetic field
77: lines? How does this depend on Mach number of turbulence? These are the
78: questions that we would like to answer.
79:
80: There are a number of observational implications of the density
81: spectra.
82: Shallow power spectrum can result in a lot of small scale absorption
83: (Deshpande 2000) which can account for the mysterious Tiny Scale
84: Structures or TSAS (Heiles 1997).
85: Power spectrum shallower than the Kolmogorov
86: one was reported in a number of observations (Deshpande, Dwarakanath \& Goss 2000,
87: Padoan et al 2003). Can one explain this? Density anisotropies
88: have been observed
89: in scintillation studies at small scales but it is unclear whether we should
90: expect them at all scales.
91:
92: \section{Theoretical Considerations}
93: Subsonic compressible MHD is rather well studied topic today.
94: It is suggestive that there may be an analogy between the subsonic
95: MHD turbulence and its incompressible counterpart, namely, GS95 model.
96: Therefore the correspondence between the the two revealed in CL03 is
97: expected.
98:
99: It could be easily seen, that in the low-beta case density is perturbed
100: mainly due to the slow mode (CL03). Slow modes are sheared by Alfv\'en turbulence,
101: therefore they exhibit
102: Kolmogorov scaling and GS95 anisotropy for low Mach numbers. However, for high
103: Mach numbers we expect shocks to develop. Density will be perturbed
104: mainly by those shocks. However, the relative perturbation of
105: density is likely to be proportional to density itself.
106:
107: One can also approach the problem from the point of view of underlying hydrodynamic equations.
108: It is well known that there is a multiplicative symmetry with respect to density
109: in the ideal flow equations for an isothermal fluid (see e.g. Passot \& V\'azquez-Semadeni, 1998;
110: henceforth PV98). This assume that if there is some stochastic process disturbing the density
111: it should be a multiplicative process with respect to density, rather than
112: additive, and the distribution for density values should be lognormal,
113: rather that normal. 1D numerical simulations of high-Mach hydrodynamics confirmed
114: that the distribution is approximately lognormal, having power-law
115: tails in case of $\gamma\neq 1$ (PV98).
116:
117: In MHD, however, the above described symmetry is broken by the magnetic
118: tension. This could be qualitatively described as the higher density
119: regions having lower Alfv\'en speed, if we assume there is no significant
120: correlation between density and magnetic field. The latter is usually the case
121: with strongly magnetized, low beta fluid ($P_{\mbox{mag}}>P_{\mbox{gas}}$). It is interesting to
122: test whether this causes substantial deviation of distribution from log-normal law.
123:
124: In a low Mach turbulence the processes leading to the perturbation of density
125: are governed by the sound speed. Self evolution of those will be slow
126: in comparison with shearing by Alfv\'en waves.
127:
128: With a high sonic Mach we expect a considerable amount of shocks arise.
129: In a sub-Alfv\'enic case, however, we expect oblique shocks be disrupted by
130: Alfv\'enic shearing, and, as most of the shocks are generated randomly
131: by driving, almost all of them will be sheared to smaller shocks.
132: The evolution of the weak shocks will be again governed by the sonic
133: speed, and structures from shearing as in low Mach case should arise.
134:
135: We also note that shearing will not affect probability density
136: function (PDF) of the density, but have
137: to affect its spectra and structure function scaling. In other words,
138: we deal with two distinct physical processes, one of which, random
139: multiplication or division of density in presence of shocks, affect
140: PDF, while the other, Alfv\'enic shearing has to affect anisotropy
141: and scaling of the structure function of the density.
142:
143: In order to test this we performed direct numerical simulations.
144:
145: \section{The code}
146: We used data cubes from our direct 3-dimensional numerical simulations
147: (see Cho \& Lazarian 2004). As we are interested in high-Mach turbulence,
148: we performed simulations on a periodic $512^3$ Cartesian grid
149: with the average sonic Mach numbers of $\sim 10$ and $\sim 3$.
150: The effects of numerical diffusion are expected to be important
151: at the scales of less than 10 grid points.
152: We observed that, parallel to the magnetic field, velocities
153: stay supersonic down to 8 grid units for Mach 10, and 20 grid units for Mach 3.
154: We used the isothermal equation of state and randomly drove turbulence
155: on the scale about 2.5 times smaller than the box size.
156: The Alfv\'en velocity of the mean magnetic field was roughly the same as the rms velocity,
157: which correspond to Alfv\'enic Mach number of around unity.
158:
159: \section{Results}
160: In Fig.~1 we show the distribution for log-density for various values
161: of Mach numbers for 3D numerical simulations of MHD. In all cases, except
162: subsonic $\beta$ was chosen so that Alfv\'enic Mach number $M_A$ is slightly
163: less then unity. This was motivated by the idea, that in a strongly
164: super-Alfv\'enic fluid, given enough time, magnetic field will grow,
165: approaching equipartition. As soon as we observe scales smaller than a driving
166: scale, we will see mildly super-Alfv\'enic or sub-Alfv\'enic turbulence.
167:
168: \placefigure{rhodist}
169: \begin{figure}
170: \figurenum{1}
171: \plotone{f1.eps}
172: \caption{Probability density function for a density in direct
173: numerical simulation with Alfv\'enic Mach number around unity and various sonic
174: Mach numbers.}
175: \label{rhodist}
176: \end{figure}
177:
178: We see, that the distribution shows significant deviation from a lognormal
179: law. The rms deviation of density for a subsonic case is consistent with
180: prediction $M^2$ for low beta (CL03), and the rms deviation of log-density
181: for supersonic case is around unity regardless of a Mach number. The distributions
182: are notably broader for higher Mach numbers, though. This is an indication that a
183: distribution is not universal.
184:
185: Dimension of the high-density structures was between 1 and 2, being
186: viewed as a flatted filaments or elongated pancakes. There were no evident
187: preferred orientation of these structures along or perpendicular to
188: the mean magnetic field. Maximum density value in a Mach 10
189: data cube was around $3\times10^3\rho_0$.
190:
191: It is obvious that density clumps with values of 3-4 orders of magnitude of mean density
192: can severely distort power spectrum.
193: It is expected that these clumps can hide density structures created by motions at small
194: scales. We see in Fig.~2 that the power-law spectrum of density for high-Mach is very shallow.
195: Randomly distributed high-density clumps will also suppress
196: any anisotropy originating from motions at small scales.
197:
198: \placefigure{spectrum}
199: \begin{figure}
200: \figurenum{2}
201: \plotone{f2.eps}
202: \caption{Mach number is 10, power spectra
203: of: {\it solid} -- density, {\it dashed} -- velocity, {\it dotted}
204: -- logarithm of density.}
205: \label{spectrum}
206: \end{figure}
207:
208: We attempted to overcome this effect and reveal an underlying density scalings by
209: using a log-density instead of density\footnote{density is dimensionless, normalized
210: by the mean density} for spectra and structure functions.
211: We found that this way to suppress
212: the influence of the high peaks to the spectrum or structure function is superior to other
213: filtering procedures\footnote{Note that a nonlinear transformation can not change
214: function from isotropic to anisotropic and vice-versa. However anisotropy can be
215: substantially suppressed and not noticeable.}. Indeed, if we cut off peaks at some level,
216: it give similar results, but structure function looks worse, as the procedure of cutting
217: off, or restricting density to some level, introduces artificial structures in the real space.
218:
219: \placefigure{SF}
220: \begin{figure*}
221: \figurenum{3}
222: \includegraphics[width=0.32\textwidth]{f3a.eps}
223: \hfill
224: \includegraphics[width=0.32\textwidth]{f3b.eps}
225: \hfill
226: \includegraphics[width=0.32\textwidth]{f3c.eps}
227: \caption{ $M_s\sim 10$, second order structure functions, calculated respective to
228: the {\it local} magnetic field for, {\it left panel}: magnetic field,
229: {\it central panel}: density,
230: {\it right panel}: log-density. We see that in left and right panel structures are mostly elongated
231: parallel to the magnetic field. Structure function, e.g. for magnetic field,
232: is $|{\bf B}({\bf r}_0+{\bf r})-{\bf B}({\bf r}_0)|^2$, averaged over ${\bf r}_0$}
233: \label{SF}
234: \end{figure*}
235:
236: The results and the comparison with the scalings of the magnetic field is presented in Fig.~3.
237: Magnetic field, being perturbed mostly by the Alfv\'en mode, shows GS95 anisotropy.
238: Scaling of log-density is analogous to it. We also see, that such a behavior is seen
239: right after the driving scale (which is around 100 grid units) by the magnetic field,
240: but somewhat into smaller scales by the log-density.
241:
242: In Fig.~4 we checked for the scale-dependent anisotropy of the GS95 type
243: ($r_{\|}\sim r_{\perp}^{2/3}$).
244:
245: \placefigure{anis}
246: \begin{figure*}
247: \figurenum{4}
248: \plottwo{f4a.eps}{f4b.eps}
249: \caption{$M_s\sim 10$, values of $r_\|$ and $r_\perp$ with equal structure function
250: for {\it left panel}: magnetic field,
251: {\it right panel}: log-density}
252: \label{anis}
253: \end{figure*}
254:
255: We also checked for correlation between density and magnetic field
256: magnitude which was expected from a models of external compression of an
257: ideal MHD fluid, and in fact observed in many super-Alfv\'enic
258: simulations (see, e.g., Padoan \& Nordlund 1999).
259: We have not found any significant correlation of this type.
260:
261:
262: \section{Discussion of Results}
263: It is a well known notion that a supersonic turbulence consists
264: mostly of shocks and other discontinuities. Our driving
265: is incompressible, but the modes are not decoupled at the
266: injection scale when Alfv\'enic Mach number is of the order of unity
267: (CL03). Therefore we expect that the driving excites an appreciable
268: amount of compressible motions. Indeed,
269: our testing of data showed that
270: the rms velocity, associated
271: with slow mode was
272: of the same order as the velocity of the Alfv\'en
273: mode. We assume that the resulting flat power spectrum of density
274: is associated with very large perturbation of density from compressible
275: motions that naturally arise at the driving scale due to coupling of compressible
276: and incompressible motions quantified in CL03.
277: Shocks in isothermal fluid can have very large density contrasts,
278: up to sonic Mach squared and can act as shocks in the snowplow phase
279: of supernova, namely, they collect matter keeping the total momentum
280: of the shock constant (see, e.g., Spitzer 1978).
281: However, we do not see strong shocks near density clumps.
282: In magnetically dominated medium that we deal with it is reasonable
283: to assume that the corresponding shocks move material along magnetic
284: field lines the same way that the slow modes do in subsonic case.
285: The shocks are randomly oriented and therefore the clumpy structure
286: that we observe does not reveal any noticeable anisotropy.
287: Density perturbations associated with such shocks should not
288: be correlated with the magnetic field strength enhancement similar
289: as in the case of densities induced by slow modes (see CL03).
290: Our analysis of the data confirms this.
291:
292: If we associate the clumps in simulations with interstellar clouds, in
293: ISM with random driving we would expect the clouds not to be particularly
294: oriented in relation to magnetic fields at least until self-gravity does
295: get important. We observed substantial variation of the gas pressure, of three
296: orders of magnitude which is consistent with findings in Jenkins 2002.
297: Flat spectrum observed is roughly consistent with some observational data.
298: Needless to say, that a more systematic analysis of data is required now
299: when we have theoretical expectations to test, e.g. the change of the
300: density spectrum with the Mach number. Testing the anisotropy of
301: density is another interesting project, even though one cannot directly
302: observe log-density and the effects associated
303: with the projection along the line of sight must be considered carefully
304: (see discussion of this in Esquivel et al. 2003).
305:
306: Surely, for real clouds self-gravity can be important. This effect should
307: make the observed spectrum even flatter, as the density peaks will become
308: higher and more delta-function-like.
309: In addition, cooling may make
310: interstellar gas more pliable to compression than the isothermal gas that
311: we used in the simulations.
312: This, could result in more density contrast when
313: the original gas is warm.
314: However, usually interstellar warm gas has Mach
315: number of the order of unity. For molecular clouds for which Mach number
316: can be substantial our isothermal calculations seems to be adequate.
317:
318: The alternative to this paper hypothesis explaining flat spectrum and isotopy,
319: namely that high-Mach large scale driving could affect turbulence to the
320: scales where effective Mach number is much less then unity is rather counter-intuitive
321: and contradicts the notion of the turbulence locality in $k$ space.
322: We understood
323: that this unusual behavior is due to the fact, that density in high Mach simulations is
324: perturbed significantly nonlinearly, therefore making
325: the interpretation of the power spectra more involved.
326: We used filtering to mitigate this effect,
327: and succeeded in showing that density scaling is anisotropic.
328: The range of scales where incompressible turbulent theory is applicable
329: is shortened in numerical simulation with supersonic driving. Between sub- and supersonic
330: scales there is a region where compressible motions cascade in a way that is
331: yet to be
332: understood.
333:
334: There exist a different effect that can make density spectrum flat at small scales.
335: Incompressible fluid with viscosity much larger than magnetic diffusivity, i.e.
336: in high Prandtl number fluid, at the viscosity-dominated scales, shows a peculiar
337: regime cascade reported in Cho, Lazarian \& Vishniac (2002).
338: If the the magnetic pressure associated with anisotropic magnetic filaments
339: formed in this new regime of MHD turbulence is balanced by the thermal pressure,
340: it will create density structures with shallow $k^{-1}$ spectrum.
341: However, this effect is present only at small scales where viscosity
342: by neutrals gets important. This allows to distinguish the two different causes of
343: spectrum flattening.
344:
345: Most incompressible turbulence theories assume constant density,
346: everywhere. As we see, the supersonic driving provide transient, high contrast density
347: structures, as the back-reaction of the fluid is the term
348: $\nabla P/\rho$ which allow for very high density perturbations. This should not be
349: a problem for theories of strong turbulence where wavepackets are cascaded in a time
350: comparable with inverse frequency.
351: Weak turbulent theories, which have a cascading time $t_{\mbox{cas}}$
352: dependent on energy flow in k-space seem to be less applicable, as the Alfv\'en perturbation
353: has to travel $v_A t_{\mbox{cas}}$ distance to fully distorted. On the other hand,
354: $t_{\mbox{cas}}\propto\epsilon^{-1/2}$
355: for three-wave processes (Zakharov, L'vov \& Falkovich, 1992), where $\epsilon$ is the flux of energy in $k$ space.
356: This distance might be pretty long, depending on $\epsilon$ and exceed the lengthscale
357: at which density changes significantly. Therefore, ``normal'' weak turbulence will be pushed
358: on scales much smaller than subsonic scale. On the other hand, we don't expect weak turbulence
359: to be valid on smaller scales due to the increase of $k_\perp/k_\|$ with smaller scales
360: and the onset of strong Alfv\'enic turbulence.
361:
362: \section{Acknowledgments}
363: AB thanks IceCube project for support of his research.
364: AL acknowledges the NSF grant AST-0307869 and the support from
365: the Center for Magnetic Self-Organization in Laboratory and Astrophysical
366: Plasma.
367:
368: \begin{thebibliography}{}
369:
370: \bibitem{CL03} Cho, J., \& Lazarian, A. 2003, MNRAS, 345, 325 (CL03)
371:
372: \bibitem{CLV02}
373: Cho, J., Lazarian, A., \& Vishniac, E. 2002, ApJ, 566, L49
374:
375: \bibitem{ChoLV03a}
376: Cho, J., Lazarian, A., \& Vishniac, E. 2003, in
377: Turbulence and Magnetic Fields in Astrophysics,
378: eds. E. Falgarone, T. Passot (Springer LNP), 56
379: (astro-ph/0205286)
380:
381: \bibitem{CL04}
382: Cho, J., \& Lazarian, A. 2004, in Proc. of the Summer Program 2004,
383: Center for Turbulence Res., Stanford Univ., 75 (astro-ph/0411031)
384:
385: \bibitem{CL05}
386: Cho, J., \& Lazarian, A. 2005, Theoret. Comput. Fluid Dynamics, in press
387: (astro-ph/0301462)
388:
389: \bibitem{De00}
390: Deshpande, A. A. 2000, MNRAS, 317, 199
391:
392: \bibitem{DDG00}
393: Deshpande, A. A., Dwarakanath, K. S., \& Goss, W. M. 2000, ApJ, 543, 227
394:
395: \bibitem{ELP03}
396: Esquivel, A., Lazarian, A., Pogosyan, D., \& Cho, J. 2003, MNRAS, 342, 325
397:
398: \bibitem{He97}
399: Heiles, C. 1997, ApJ, 481, 193
400:
401: \bibitem{GolS95}
402: Goldreich, P., \& Sridhar, S. 1995, ApJ, 438, 763 (GS95)
403:
404: \bibitem{Je02}
405: Jenkins, E. 2002, ApJ, 580, 938
406:
407: \bibitem{Kol41}
408: Kolmogorov, A. 1941, Dokl.~Akad.~Nauk SSSR, 31, 538
409:
410: \bibitem{LitG01}
411: Lithwick, Y., \& Goldreich, P. 2001, ApJ, 562, 279
412:
413: \bibitem{PBLN03}
414: Padoan, P., Boldyrev S., Langer W., \& Nordlund, \AA. 2003, ApJ, 583, 308
415:
416: \bibitem{PN99}
417: Padoan, P., \& Nordlund, \AA. 1999, ApJ, 526, 279
418:
419: \bibitem{PV98}
420: Passot, T., \& V\'azquez-Semadeni, E. 1998, Phys. Rev. E {\bf 58}(4), 4501-4510
421:
422: \bibitem{Sp78}
423: Spitzer, L. 1978, Physical processes in the interstellar medium (New York: Wiley)
424:
425: \bibitem{Vasq00}
426: V\'azquez-Semadeni, E., Ostriker, E. C., Passot, T., Gammie, C. F., \& Stone J.M. 2000,
427: in Protostars and Planets IV (Tucson: University of Arizona Press), 3
428: (astro-ph/9903066)
429:
430: \bibitem{VOS03}
431: Vestuto, J. G., Ostriker, E. C., \& Stone, J. M. 2003, ApJ, 590, 858-873
432:
433: \bibitem{ZLF92}
434: Zakharov, V. E., L'vov, V. S., \& Falkovich, G. 1992, Kolmogorov Spectra of Turbulence I (Berlin: Springer)
435:
436: \end{thebibliography}
437:
438: \end{document}
439: