1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3:
4: \usepackage{verbatim}
5: \newcommand{\unit}[1]{\,{\rm #1}}
6:
7: \begin{document}
8:
9: \title{Density Probability Distribution Functions in Supersonic Hydrodynamic
10: and MHD Turbulence}
11: \author{M. Nicole Lemaster\altaffilmark{1} and James M. Stone}
12: \affil{Department of Astrophysical Sciences, Princeton University,
13: Princeton, NJ 08544}
14: \altaffiltext{1}{\email{Lemaster@astro.princeton.edu}}
15:
16: \shorttitle{Density PDFs in Supersonic Turbulence}
17: \shortauthors{Lemaster \& Stone}
18:
19: \begin{abstract}
20:
21: We study the probability distribution function (PDF)
22: of the mass density in simulations of supersonic turbulence with
23: properties appropriate for molecular clouds. For this study we use
24: Athena, a new higher-order Godunov code.
25: We find there are surprisingly similar relationships between the mean
26: of the time-averaged PDF and the turbulent Mach number for driven
27: hydrodynamic and strong-field MHD turbulence. There is, however, a
28: large scatter about these relations, indicating a high level of
29: temporal and spatial variability in the PDF. Thus, the PDF of the
30: mass density is unlikely to be a good measure of magnetic field strength.
31: We also find that the PDF of decaying MHD turbulence deviates from the
32: mean-Mach relation found in the driven case. This implies that the
33: instantaneous Mach number alone is not enough to determine the
34: statistical properties of turbulence that is out of equilibrium.
35: The scatter about the mean-Mach relation for driven turbulence, along
36: with the large departure of decaying turbulence PDFs from those of
37: driven turbulence, may illuminate one factor contributing to the large
38: observed cloud-to-cloud variation in the star formation rate per solar
39: mass.
40:
41: \end{abstract}
42:
43: \keywords{ISM: clouds ---
44: ISM: magnetic fields --- isothermal --- simulations ---
45: stars: formation --- turbulence}
46:
47: \section{Introduction}
48:
49: %% observed scatter in star formation rate in clouds
50: %% observational evidence for turbulence in molecular clouds
51: %% how turbulence might affect star formation rate
52:
53: The mechanism that determines the star formation rate (SFR) within a
54: molecular cloud (MC) is not well understood. Observations show that
55: the SFR per solar mass, as measured by the ratio of CO to IR
56: luminosity, varies by as much as 2-3 orders of magnitude from region
57: to region; instead Gao \& Solomon (2004) found that HCN
58: emission, which measures molecular gas at much higher density
59: ($\gtrsim 3 \times 10^4 \unit{cm^{-3}}$), is the better indicator of
60: star formation. This suggests that the amount of mass to reach high
61: density is the key factor in determining the SFR per solar mass within
62: a cloud (McKee \& Ostriker 2007). Observed non-thermal line widths in
63: MCs (Falgarone \& Philips 1990) indicate that supersonic turbulence
64: may be responsible for creating the high density contrasts that lead
65: to clump formation.
66:
67: %% observations that can be used to determine magnetic field strength
68: %% why we care about magnetic field strength
69:
70: Although magnetic fields have been shown not to lengthen appreciably
71: the decay timescale of supersonic turbulence (Stone, Ostriker, \&
72: Gammie 1998, hereafter S98; Mac Low 1999), they do create anisotropy
73: in the structures within the medium (Vestuto et al. 2003).
74: Observations using Zeeman splitting, such as those described
75: by Crutcher (1999), have found magnetic fields in some clouds strong
76: enough (e.g. $\beta \approx 0.04$) that they cannot be neglected.
77: The magnetic field within a MC is typically difficult to detect,
78: motivating the determination of new diagnostics (e.g. Heyer et al.
79: 2008).
80:
81: %% simulations focusing on mean-Mach relation
82: %% what else it would be interesting to find out
83:
84: The probability distribution function (PDF) of the logarithm of mass
85: density can be used to quantify the amount of material within a
86: turbulent medium that has a given density. As the density is likely
87: to have a strong impact on star formation, many groups have
88: investigated the properties of such PDFs. Padoan et al.
89: (1997, hereafter P97), for example, conducted simulations of driven
90: hydrodynamic turbulence, determining a relation between the mean of
91: the PDF and the Mach number. Most of the mass in the simulated clouds
92: was found to be in only a small fraction of the volume, with the width
93: of the approximately Gaussian PDF increasing with Mach
94: number in a predictable way. Ostriker, Stone, \& Gammie (2001)
95: investigated the effect of magnetic field strength on the mean-Mach
96: relation for decaying turbulence, finding that the fast magnetosonic
97: Mach number can be used to predict only a lower limit on the width of
98: the distribution.
99:
100: %% contents of this paper
101:
102: Recently, we have undertaken a comprehensive study of the properties
103: of supersonic turbulence (Lemaster \& Stone 2008, hereafter Paper I)
104: with Athena, a new higher-order Godunov
105: method (Stone et al. 2008). This study represents one of the first
106: applications of Godunov methods to the study of supersonic MHD
107: turbulence, and therefore represents an important test of previous results.
108: In this letter, we
109: investigate the effects of magnetic field strength and Mach number on
110: the PDF. We survey a much larger range of Mach number and field
111: strength, and at a higher numerical resolution, than has been used in
112: previous studies. This provides much better statistics and allows us
113: to constrain the form of the relations more tightly. We describe the
114: numerical methods used to conduct our simulations in
115: \S\ref{sec:methods}, explain the focus of our analysis in
116: \S\ref{sec:pdfs}, and present the results in \S\ref{sec:results}.
117: Finally, we summarize in \S\ref{sec:concl}.
118:
119: \section{Numerical Methods}\label{sec:methods}
120:
121: %% equations and their assumptions
122: %% Athena configuration
123: %% bcs, res, box size, etc.
124:
125: The simulations we present here were conducted at a resolution of
126: $512^3$ with Athena (Gardiner \& Stone 2005, 2008; Stone et al.
127: 2008; Stone \& Gardiner 2008), a new higher-order Godunov code that
128: exactly conserves mass, momentum, and magnetic flux. We solve the
129: equations of ideal isothermal MHD,
130: \begin{equation}
131: \frac{\partial \rho}{\partial t} + {\bf \nabla} \cdot (\rho {\bf v}) = 0,
132: \end{equation}
133: \begin{equation}
134: \frac{\partial \rho {\bf v}}{\partial t} +
135: {\bf \nabla} \cdot (\rho {\bf vv} - { \bf BB} + P + B^2/2) = 0,
136: \end{equation}
137: and
138: \begin{equation}
139: \frac{\partial {\bf B}}{\partial t} =
140: {\bf \nabla} \times ({\bf v} \times {\bf B}),
141: \end{equation}
142: where $c_s = 1$ and $P = c_s^2 \rho$ are the isothermal sound speed
143: and pressure, on a three-dimensional periodic Cartesian grid of
144: length $L = 1$. We use an
145: approximate nonlinear Riemann solver (HLLD; Miyoshi \& Kusano 2005) for
146: our MHD runs and an exact nonlinear Riemann solver for our hydrodynamic
147: runs. We integrate our simulations for roughly 4 dynamical times
148: beyond the saturation time using a directionally-unsplit van Leer
149: scheme (Stone \& Gardiner 2008). Here we have defined the dynamical
150: time to be $t_{\rm dyn} = L/(2\mathcal{M})$, where
151: $\mathcal{M} \equiv \sigma_v/c_s$ is the Mach number and
152: $\sigma_v = \langle v^2 \rho/\bar{\rho} \rangle^{1/2}$ is the velocity
153: dispersion of the gas, calculated using a mass-weighted average.
154: For further details on our numerical methods as applied to
155: turbulence, see Paper I.
156:
157: %% driving method/initial conditions
158: %% definitions of parameters and standard values
159:
160: We drive our turbulence here in a manner very similar to S98.
161: We initialize a uniform, stationary ambient medium
162: with density $\bar{\rho} = 1$ and magnetic field parallel to the
163: $x$-axis whose amplitude $B_0$ is fixed by the value of
164: $\beta = 2c_s^2\bar{\rho}/B_0^2$. We apply divergence-free
165: velocity perturbations before every time step, following a Gaussian
166: random distribution peaked at $k_{\rm pk}L/2\pi = 2$. Before applying
167: the perturbations, we shift them such that no net momentum will be
168: added to the grid. We also normalize them to give the desired energy
169: injection rate, $\dot{E}/\bar{\rho}L^2c_s^3$. For our decaying runs,
170: we begin with a snapshot of fully-developed turbulence from a driven
171: run and allow it to evolve without further energy injection.
172:
173: %% conversion of parameters to physical values
174:
175: We investigate strong-field MHD ($\beta = 0.02$) as well as
176: hydrodynamic ($\beta = \infty$) turbulence. The magnetic fields we
177: use in our simulations correspond to physical values of
178: $B = 2.0 \unit{\mu G} \beta^{-1/2} (T/10 \unit{K})^{1/2}
179: (n_{H_2}/10^2 \unit{cm^{-3}})^{1/2}$, where $T$ is the temperature and
180: $n_{H_2}$ is the number density of molecular hydrogen. Our
181: simulations are scale-free, allowing them to be scaled to any set of
182: physical parameters using appropriate choices of $\bar{\rho}$, $c_s$,
183: and $L$. Utilizing the same values given in S98, i.e.
184: $L = 2 \unit{pc}$, $n_{H_2} = 10^3 \unit{cm^{-3}}$, and
185: $T = 10 \unit{K}$, yields an energy injection rate of
186: $\dot{E} = 0.2 L_\odot$ with magnetic field strength
187: $B = 44 \unit{\mu G}$.
188:
189: \section{Probability Distribution Functions}\label{sec:pdfs}
190:
191: %% significance of pdfs
192:
193: Turbulence in molecular clouds causes converging flows where the gas
194: can be compressed to very high densities. The PDF of the density
195: tells us the fraction of the mass or volume within a cloud that
196: obtains a given density. Since self-gravitating clumps can form in
197: the high-density regions, understanding PDFs is critical for gaining
198: insight into the stellar IFM and SFR.
199: If compression and rarefaction events in the turbulent gas within a
200: molecular cloud are spatially and temporally independent, the PDF of
201: density will have a log-normal distribution (Passot \&
202: Vazquez-Semadeni 1998). The PDF of the logarithm of density, then,
203: will have a normal distribution given by
204: \begin{equation}
205: f(y) dy = \frac{1}{\sqrt{2\pi\sigma^2}}
206: \exp \Big[ \frac{-(y-\mu)^2}{2\sigma^2} \Big] dy,
207: \end{equation}
208: where $y \equiv \ln (\rho/\bar{\rho})$, $\mu$ is the mean of the
209: distribution, and $\sigma^2$ is the dispersion, with
210: $|\mu| = \sigma^2/2$.
211:
212: Our goal in this Letter is to analyze the relationship between the
213: mean of the distribution, which represents the median density within
214: the cloud, and the Mach number. We have investigated this
215: mean-Mach relation over a range of Mach numbers
216: $1.2 \le \mathcal{M} \le 7.0$ for both driven hydrodynamic and
217: strong-field MHD turbulence. We found that a resolution of $512^3$
218: gave smoother-looking PDFs that could be fitted more accurately than
219: those from $256^3$ simulations, justifying the computational expense.
220: Higher resolution also allows us to study scatter in the PDF in
221: sub-volumes of the domain. We show in Paper I that the simulations
222: have converged by this resolution; thus, increasing it further is
223: unlikely to affect the results. To minimize the influence of
224: intermittency, we time-average the PDFs obtained from seven snapshots
225: in the saturated state spanning almost 3 dynamical times before
226: fitting them.
227:
228: Since the tails of the PDFs will deviate from normal form due to the
229: effects of intermittency, we fit only bins with values of at least
230: $10\%$ of the peak value. We perform a Levenberg-Marquardt
231: least-squares fit with uniform weighting. Once we have obtained the
232: mean, $\mu$, of the best-fit distribution, we plot it against a
233: function of Mach number,
234: $\xi(\mathcal{M}) = \ln[1+\alpha\mathcal{M}^2]$. With the appropriate
235: choice of $\alpha$, we can obtain a linear relation between the PDF
236: mean and this function $\xi(\mathcal{M})$. We note that Kowal et al.
237: (2007) have recently shown that higher order statistics can also
238: provide insight into the properties of supersonic turbulence, however
239: in this Letter we will focus only on the density PDF.
240:
241: Observations show a wide range of star formation rates for different
242: clouds. To infer from our simulations the level of cloud-to-cloud
243: variation we would likely observe, we also investigate the mean-Mach
244: relation for regions of size comparable to the driving scale. To do
245: this, we divide our computational domain into eight equal sub-domains,
246: each of resolution $256^3$. We compute the PDF in each of these
247: sub-boxes (which we will refer to as sub-PDFs) individually and plot
248: their means against the Mach number within that sub-box.
249: We do not time-average our results, yielding 56 mean-Mach pairs for
250: each run. Although the snapshots are at intervals of just under half
251: a dynamical time on the global scale, the interval between snapshots
252: is closer to a flow crossing time on the scale of the sub-boxes,
253: making the snapshots sufficiently uncorrelated for our analysis.
254: Substantial scatter in the values within a run might help explain the
255: observed cloud-to-cloud variation in the star formation rate in
256: molecular clouds.
257:
258: \section{Results}\label{sec:results}
259:
260: %% what the hydro and mhd pdfs really look like
261: %% comparison between mhd and hydro pdfs
262:
263: The time-averaged PDFs over the full domain are very smooth and
264: approximate Gaussians, particularly in the hydro case.
265: Although we have not plotted such a PDF in this Letter, a similar
266: example can be seen in Figure 3 of Kritsuk et al. (2007). Presented
267: in Table \ref{tab:runs} are our PDF statistics. The first two
268: columns, respectively, specify the magnetic $\beta$ and energy
269: injection rate. The final six columns are determined from the
270: sub-volumes and describe the error bars in the figures to follow.
271: The values $\sigma_{V,S}$ and $\sigma_{M,S}$ in the table are the
272: standard deviation of the means determined from the PDFs, not the
273: width of the PDFs themselves.
274:
275: \begin{deluxetable}{cccccccc}
276: \tablecolumns{8}
277: \tablewidth{0pc}
278: \tablecaption{Driven Turbulence at $512^3$\label{tab:runs}}
279: \tablehead{\colhead{$\beta$} & \colhead{$\dot{E}$}
280: & \colhead{$\mathcal{M}$} & \colhead{$\sigma_{\mathcal{M}}$}
281: & \colhead{$\mu_V$} & \colhead{$\sigma_{V,S}$\tablenotemark{a}}
282: & \colhead{$\mu_M$} & \colhead{$\sigma_{M,S}$\tablenotemark{a}}}
283: \startdata
284: $\infty$ & 500 & 6.8 & 0.5 & -1.1 & 0.13 & 0.94 & 0.09 \\
285: $\infty$ & 187.5 & 5.0 & 0.4 & -0.87 & 0.12 & 0.76 & 0.10 \\
286: $\infty$ & 70 & 3.6 & 0.3 & -0.65 & 0.08 & 0.57 & 0.07 \\
287: $\infty$ & 20 & 2.4 & 0.1 & -0.40 & 0.05 & 0.35 & 0.05 \\
288: $\infty$ & 1.75 & 1.2 & 0.05 & -0.10 & 0.02 & 0.09 & 0.01 \\
289: \hline
290: 0.02 & 500 & 6.7 & 0.5 & -0.89 & 0.12 & 0.84 & 0.11 \\
291: 0.02 & 187.5 & 4.9 & 0.3 & -0.71 & 0.12 & 0.66 & 0.11 \\
292: 0.02 & 70 & 3.6 & 0.2 & -0.56 & 0.09 & 0.50 & 0.09 \\
293: 0.02 & 20 & 2.5 & 0.1 & -0.34 & 0.05 & 0.30 & 0.05 \\
294: 0.02 & 1.75 & 1.2 & 0.06 & -0.12 & 0.03 & 0.11 & 0.03
295: \enddata
296: \tablenotetext{a}{$\sigma_{V,S}$ and $\sigma_{M,S}$ are the standard deviation of $\mu_V$ and $\mu_M$, respectively.}
297: \end{deluxetable}
298:
299: \subsection{Driven Hydrodynamic Turbulence}\label{sec:hydpdf}
300:
301: %% the mean-Mach relation found for time-avgd full domain
302:
303: Figure \ref{fig:pdfreln} shows the mean-Mach relation found from
304: time-averaged PDFs over the full domain for driven hydro and driven MHD
305: turbulence. We find that a value of $\alpha = 0.5$ in the function
306: $\xi(\mathcal{M})$ gives the best linear relations for the hydro case.
307: For the volume fraction we find
308: \begin{equation}
309: \mu_V = -0.36 \ln[1+0.5\mathcal{M}^2] + 0.10,
310: \end{equation}
311: while for the mass fraction we find
312: \begin{equation}
313: \mu_M = 0.32 \ln[1+0.5\mathcal{M}^2] - 0.10.
314: \end{equation}
315: The rms residuals for these fits are $8.9 \times 10^{-3}$ and
316: $6.5 \times 10^{-3}$, respectively.
317: Because the density fluctuations in subsonic turbulence are not
318: produced by shocks, we have no reason to expect these relations to
319: approach zero with Mach number.
320: The mean-Mach pairs from the time-averaged PDFs fall very close to
321: these relations. Over the full range of Mach numbers tested, however,
322: the time-averaged means are smaller than those found by P97
323: ($\mu_{V,M} = \mp0.5 \ln[1+0.25\mathcal{M}^2]$).
324: To determine the magnitude of the effect that the driving may have had
325: on the relations, we also compare values determined from hydrodynamic
326: turbulence with $k_{\rm pk}L/2\pi = 4$ (not shown), finding that these
327: points fall very close to the $k_{\rm pk}L/2\pi = 2$ relations as well.
328:
329: \begin{figure}
330: \epsscale{1.0}
331: \plotone{f1.eps}
332: \figcaption{
333: PDF mean versus $\xi(\mathcal{M})$ for driven hydro (red hexagons)
334: and strong-field MHD (blue squares) turbulence. Only time-averaged
335: PDFs over the full domain are used to fit the mean-Mach relations
336: (hydro: red dashed; MHD: blue dashed). Also shown are the
337: instantaneous values taken from each of the eight driving-scale
338: sub-boxes (hydro: gold triangles; MHD: green crosses) and the
339: 1$\sigma$ error bars on those points (hydro: red; MHD: blue).
340: For comparison, the relation from P97 is also shown (black dotted).
341: \label{fig:pdfreln}}
342: \end{figure}
343:
344: %% the scatter in sub-boxes
345:
346: Also shown in the figure are the instantaneous mean-Mach pairs found
347: from each of the eight sub-boxes. The scatter in these points,
348: illustrated by 1$\sigma$ error bars on the plot, is quite
349: significant. The ensemble average of the 56 points for each run falls
350: at a slightly lower Mach number than the value found from the
351: time-average over the full domain since the Mach number typically
352: decreases on smaller scales. These averages differ slightly from
353: those computed from time-averaged full-domain PDFs, although they
354: still fall close to the relations found. The 1$\sigma$ scatter in the
355: Mach number is $4\%$--$8\%$, while in the sub-PDF means it is
356: $16\%$--$17\%$ for the lowest Mach number run and $10\%$--$14\%$ for the
357: remaining runs. This scatter puts some of the instantaneous sub-PDF
358: values in the vicinity of the P97 relation.
359:
360: The scatter in $\mu_V$ and $\mu_M$ can be converted to scatter in the
361: median density within the medium. For the ratio of the median to mean
362: density within a cloud, $\tilde{\rho}$, we can define $\tilde{\rho}_+$
363: and $\tilde{\rho}_-$ for clouds where $\mu$ is 1$\sigma$ above or below
364: the mean. The ratios
365: $\tilde{\rho}_{V,+}/\tilde{\rho}_{V,-}=\exp(2\sigma_{V,S})$ and
366: $\tilde{\rho}_{M,+}/\tilde{\rho}_{M,-}=\exp(2\sigma_{M,S})$ then
367: generally increase with Mach number, ranging from 1.0 to 1.3.
368:
369: \subsection{Driven MHD Turbulence}\label{sec:mhdpdf}
370:
371: %% the mean-Mach relation found for time-avgd full domain
372: %% the scatter in sub-boxes
373: %% comparison to hydro mean-Mach relation
374:
375: Figure \ref{fig:pdfreln} also includes the mean-Mach relation for
376: driven strong-field MHD turbulence, where we continue to use
377: $\alpha = 0.5$. The mean-Mach pairs from the time-averaged PDFs
378: again fall very close to these relations,
379: \begin{equation}
380: \mu_V = -0.29 \ln[1+0.5\mathcal{M}^2] - 0.06
381: \end{equation}
382: for the volume fraction and
383: \begin{equation}
384: \mu_M = 0.28 \ln[1+0.5\mathcal{M}^2] + 0.07
385: \end{equation}
386: for the mass fraction,
387: still yielding means smaller than those found by P97.
388: The rms residuals for these fits are $1.1 \times 10^{-2}$ and
389: $1.6 \times 10^{-2}$, respectively.
390:
391: The instantaneous mean-Mach pairs found from sub-boxes have more
392: scatter with a strong magnetic field than they did in the purely
393: hydrodynamic case. The 1$\sigma$ scatter in the Mach number is
394: $5\%$--$8\%$, while for the sub-PDF means it is $22\%$--$24\%$ for the
395: lowest Mach number run and $13\%$--$18\%$ for the remaining runs.
396: Converting this to scatter in the median density, we find that the
397: ratios $\tilde{\rho}_{V,+}/\tilde{\rho}_{V,-}$ and
398: $\tilde{\rho}_{M,+}/\tilde{\rho}_{M,-}$ range from 1.1 to 1.3. This
399: again puts some of the instantaneous sub-PDF values in the vicinity of
400: the P97 mean-Mach relation.
401:
402: The time-averaged mean-Mach relations found for hydro and MHD differ,
403: as one should expect due to differences in the shock-jump conditions.
404: However, the sub-PDF values overlap substantially, making them
405: difficult to distinguish observationally. The time-averaged relations
406: fall a bit less than 1$\sigma$ (computed from the sub-PDF values)
407: apart.
408:
409: \subsection{Decaying MHD Turbulence}\label{sec:decpdf}
410:
411: As it seems likely that molecular clouds are transient entities,
412: it may be more appropriate to study the PDF of decaying turbulence.
413: In Figure \ref{fig:decreln}, we show the evolution of the PDF in a
414: decaying strong-field MHD turbulence run, initialized from a snapshot
415: of fully-developed turbulence from our highest Mach number driven run.
416: Although this snapshot has a full-domain PDF mean roughly 1$\sigma$
417: more extreme than the time-averaged driven relations, this should not
418: affect the results. At first the small change in mean as the Mach number
419: decreases causes a shallower slope than that of the driven relation.
420: Once the mean begins to change appreciably, however, the slope becomes
421: much steeper, crossing the driven relation at roughly $\mathcal{M}
422: = 4.5$. Although the slope shallows as low Mach numbers are reached,
423: it does so only as the means become very small.
424:
425: \begin{figure}
426: \epsscale{1.0}
427: \plotone{f2.eps}
428: \figcaption{
429: PDF mean versus $\xi(\mathcal{M})$ for decaying MHD turbulence
430: (orange solid) as it evolves over time (right to left), contrasted with
431: the time-averaged driven values (blue squares) and relation found in
432: \S\ref{sec:mhdpdf} (blue dashed). The decaying run does not follow
433: the relation found for driven MHD, even considering the large scatter
434: (error bars and green crosses) found in the driven sub-PDFs. For
435: comparison, the hydro relation from P97 is also shown (black dotted).
436: \label{fig:decreln}}
437: \end{figure}
438:
439: %% the scatter in sub-boxes
440: %% implications of discrepancy between driven and decaying
441:
442: The evolution of this decaying run does not parallel or
443: asymptotically approach the driven relation at the same magnetic
444: $\beta$. Contrary to what was found by Passot \& Vazquez-Semadeni
445: (1998) for one-dimensional polytropic gas, then, we find that the
446: instantaneous Mach number is insufficient to describe the
447: properties of the turbulent medium when the gas is not in a
448: statistically steady state. This is consistent with the findings of
449: Ostriker et al. (2001) for the PDFs of decaying turbulence. If MCs
450: contain decaying turbulence, which seems likely (Heitsch et
451: al. 2006), the PDF may be ``out of equilibrium'', making relations
452: obtained from steady state (driven) turbulence inapplicable.
453:
454: \section{Summary}\label{sec:concl}
455:
456: %% similarity between hydro and mhd
457: %% significance/implications of large scatter
458: %% implications of discrepancy between driven and decaying
459:
460: For both supersonic hydrodynamic and strong-field MHD turbulence, we
461: have found a one-to-one correspondence between the mean of the
462: time-averaged PDF and the Mach number. The mean-Mach pairs
463: used to fit these relations have very small residuals; however, the
464: mean of the PDF at any given Mach number, in both the hydro and MHD
465: cases, is smaller than was found by P97 for the purely hydrodynamic
466: case. Although there is substantial scatter of the mean-Mach pairs
467: computed from instantaneous sub-volumes, the ensemble average of these
468: values still falls close to the time-averaged global relation. The
469: scatter puts a small fraction of the instantaneous values in the
470: vicinity of the P97 relation.
471:
472: Although the relations found for hydro and MHD differ, the scatter
473: about the mean relation of the instantaneous sub-PDF values creates
474: substantial overlap between the two. Since the MHD relation gives means
475: that are typically smaller than the corresponding hydrodynamic values
476: by only 1$\sigma$, it will be very difficult to distinguish between
477: the two observationally. We have also found that PDFs in decaying MHD
478: turbulence differ from those of driven MHD turbulence at the same
479: magnetic $\beta$. It would seem that the instantaneous Mach number
480: alone does not adequately describe the statistical state of the
481: turbulent gas when not in equilibrium. Again, however, there is
482: substantial overlap between the equilibrium and non-equilibrium values,
483: preventing this diagnostic from being used to distinguish between the
484: two.
485:
486: The scatter we have found about the mean-Mach relation may help
487: explain the large variation in the observed SFR per solar mass in
488: molecular clouds. Since MCs are likely to be transient entities,
489: relations found from driven turbulence may not even be applicable to
490: real clouds. If large departures from the mean-Mach relation are in
491: fact linked to the large variation in the SFR per solar mass, this may
492: indicate that turbulence in MCs is indeed decaying rather than forced.
493:
494: \acknowledgements
495:
496: We thank Eve Ostriker for very productive discussion. Simulations
497: were performed on the IBM Blue Gene at Princeton and on computational
498: facilities supported by NSF grants AST-0216105 and AST-0722479.
499:
500: \begin{references}
501: \reference{C99} Crutcher, R.M. 1999, ApJ 520, 706
502: \reference{FP90} Falgarone, E., \& Philips, T.G. 1990, ApJ 359, 344
503: \reference{GS04} Gao, Y. \& Solomon, P.M. 2004, ApJ 606, 271
504: \reference{GS05} Gardiner, T.A. \& Stone, J.M. 2005, JCoPh 205, 509
505: \reference{GS08} Gardiner, T.A. \& Stone, J.M. 2008, JCoPh 227, 4123.
506: \reference{H06} Heitsch, F. et al. 2006, ApJ 648, 1052
507: \reference{H08} Heyer, M., Gong, H., Ostriker, E., \& Brunt, C. 2008, ApJ 680, 420
508: \reference{KLB06} Kowal, G., Lazarian, A., \& Beresnyak, A. 2007, ApJ 658, 423
509: \reference{K07} Kritsuk, A.G. et al. 2007, ApJ 665, 416
510: \reference{PaperI} Lemaster, M.N. \& Stone, J.M. 2008, ApJ, in press (Paper I)
511: \reference{ML99} Mac Low, M.-M. 1999, ApJ 524, 169
512: \reference{MO07} McKee, C.F. \& Ostriker, E.C. 2007, ARA\&A 45, 565
513: \reference{MK05} Miyoshi, T. \& Kusano, K. 2005, JCoPh 208, 315
514: \reference{O01} Ostriker, E.C., Stone, J.M., \& Gammie, C.F. 2001, ApJ 546, 980
515: \reference{P97} Padoan, P., Jones, B.J.T., \& Nordlund, A.P. 1997, ApJ 474, 730 (P97)
516: \reference{PV98} Passot, T. \& Vazquez-Semadeni, E. 1998, PhRvE, 58, 4501
517: \reference{SG08} Stone, J.M., \& Gardiner, T.A. 2008, NewA, in press
518: \reference{S98} Stone, J.M., Ostriker, E.C., \& Gammie, C.F. 1998, ApJ 508, L99 (S98)
519: \reference{S08} Stone, J.M. et al. 2008, ApJS, in press (arXiv0804.0402)
520: \reference{V03} Vestuto, J.G., Ostriker, E.C., \& Stone, J.M. 2003, ApJ 590, 858
521: \end{references}
522:
523: \end{document}
524: