1: %\documentstyle[preprint,eqsecnum,aps]{revtex}
2: \documentstyle[aaspp4]{article}
3:
4: \def\k{{\bf k}}
5: \def\r{{\bf r}}
6: \def\rhok{{\rho_{\bf k}}}
7: \def\u{{\bf u}}
8: \def\uzk{u_{\bf k}}
9: \def\x{{\bf x}}
10: \def\z{{\bf z}}
11:
12: \begin{document}
13: \title{Emissivity Statistics in Turbulent, Compressible MHD Flows and
14: the Density-Velocity Correlation}
15: \author {Alex Lazarian$^1$, Dmitri Pogosyan$^2$, Enrique V\'azquez-Semadeni$^3$, and B\'arbara Pichardo$^4$}
16:
17: \affil{$^1$Dept.\ of Astronomy, University of Wisconsin, Madison, USA}
18: \affil{$^2$Canadian Institute for Theoretical Astrophysics, University
19: of Toronto, CANADA}
20: \affil{$^3$Instituto de Astronom\'\i a, UNAM, Campus Morelia,
21: Apdo. Postal 3-72, Xangari, 58089, Morelia, Mich., MEXICO}
22: \affil{$^4$Instituto de Astronom\'\i a, UNAM, Apdo.\ Postal 70-264,
23: M\'exico D.F.\ 04510, MEXICO}
24:
25: \begin{abstract}
26:
27: In this paper we test the results of a recent analytical study by
28: Lazarian and Pogosyan, on the statistics of emissivity in velocity
29: channel maps, in the case of realistic density and velocity fields
30: obtained from numerical simulations of MHD turbulence in the
31: interstellar medium (ISM).
32: To compensate for the lack of
33: well-developed inertial ranges in the simulations due to the limited
34: resolution, we apply a procedure for modifying the spectral slopes of
35: the fields while preserving the spatial structures.
36: We find that the
37: density and velocity are moderately correlated in space
38: and prove
39: that the analytical results by Lazarian and Pogosyan
40: hold in the case when these fields obey the fluid conservation
41: equations. Our results imply that the spectra of velocity and density can
42: be safely recovered from the position-position-velocity (PPV) data cubes
43: available through observations, and confirm that the relative
44: contributions of the velocity and density fluctuations to those of the
45: emissivity depend on the velocity resolution used and on the steepness
46: of the density spectral index. Furthermore, this paper supports
47: previous reports that an
48: interpretation of the features in the PPV data cubes as simple density
49: enhancements (i.e., ``clouds'') can be often erroneous, as we observe that
50: changes in the velocity statistics substantially modify the statistics
51: of emissivity within the velocity data cubes.
52:
53: \end{abstract}
54:
55: \keywords{interstellar medium: general, structure-turbulence-radio lines;
56: atomic hydrogen}
57:
58:
59:
60: \section{Introduction}
61:
62: It is generally accepted that interstellar turbulence plays a crucial
63: role in many astrophysical processes, including molecular cloud and
64: star formation, mass
65: and energy transfer in accretion disks, acceleration of cosmic rays,
66: etc. At present we
67: are still groping for the basic facts related to this complex
68: phenomenon and one of the reasons for such an unsatisfactory state of
69: affairs is that the interstellar turbulence statistics are not
70: directly available. For instance, in studies of the neutral medium,
71: indirect measures, such as spectral line widths and centroids of
72: velocity, are employed (e.g. Miesch \& Bally 1994; see also the review
73: by Scalo 1987), while potentially more valuable sources of information
74: available in velocity-channel maps remain mostly untapped (although
75: see Heyer \& Schloerb 1997; Rosolowsky et al.\ 1999; Brunt \& Heyer
76: 2000) because of the difficulty of relating the
77: two-dimensional (2D) statistics available through observations to the
78: underlying three-dimensional (3D) statistics. A discussion of various
79: approaches to the problem of turbulence study using spectral line data
80: can be found in a recent review by Lazarian (1999).
81:
82: In particular, the problem has been recently addressed by Lazarian
83: \& Pogosyan (2000, herafter LP00), who derived the index (i.e., the
84: logarithmic slope)
85: of the power spectrum\footnote{Note that throughout this paper we
86: refer exclusively to the {\it spatial} power spectrum of the various
87: fields, i.e., the Fourier transform of their auto-correlation
88: function, as is common in turbulence studies. Such spectra should not
89: be confused with, for example, spectral line profiles, or the spectra
90: of time series of data, which we do not consider here. Also,
91: throughout this paper, we stick to the convention that the spectral index of an
92: $N$-dimensional field does {\it not} include the $k^{N-1}$ factor
93: corresponding to a summation over a shell of wavenumbers of radius
94: $k$. With this notation, the well-known Kolmogorov $-5/3$ law corresponds
95: to an index of $-11/3$.} of the
96: intensity in velocity channel maps as a function of
97: the corresponding indices for the 3D density and velocity fields of
98: the emitting medium. Their work effectively provides a means of
99: inverting the problem, so that the power spectra of the medium's
100: density and velocity can be retrieved from the power spectrum of the
101: emissivity. Note that this procedure involves considering channels of
102: various velocity widths or ``velocity slices of different thickness'',
103: in the terminology of LP00.
104: Otherwise there is an indetermination due to the fact
105: that for sufficiently shallow density spectra the emissivity
106: in slices is given by a linear combination of the velocity and
107: density indices. Recent measurements by Stanimirivic (2000) and
108: Stanimirivic \& Lazarian (2000) have confirmed
109: a variation of the spectral index of HI intensity maps as the slice
110: width is varied, in accordance with the predictions of LP00.
111: The approach suggested by LP00 allows, in principle, to use the
112: wealth of existing spectroscopic data for deriving interstellar turbulence
113: statistics, thus possibly permitting new levels of understanding of
114: this phenomenon.
115: However, the derivation of LP00 assumes that the statistics of
116: velocity and density are independent. Although those authors showed that for
117: a particular case their results hold even when maximal correlation
118: of velocity and density is assumed, testing the scheme on realistic fields
119: from compressible magneto-hydrodynamic (MHD) simulations is
120: essential.
121:
122: In this paper we assess the degree of correlation between the
123: density and velocity fields from MHD, compressible turbulence
124: simulations, and whether the theoretical results from LP00 apply in
125: this case, our goal being to find out to what extent the spectra of
126: velocity and density can be recovered from the observed emissivity
127: statistics. Note that
128: if the interdependence of velocity and density is important
129: for the recovery of their spectral indices, the inversion
130: must be modified to account for the velocity-density
131: correlations.
132:
133: Numerous measurements of the emissivity within actual data cubes
134: suggest that the spectrum follows a power law (Green 1993, Stanimirovic et
135: al.\ 1999), as is the case in classical high-Reynolds number
136: incompressible turbulence (e.g., Lesieur 1990), and high-resolution
137: numerical simulations of highly compressible MHD turbulence in less
138: than three dimensions (e.g., Passot, V\'azquez-Semadeni \& Pouquet
139: 1995; Gammie \& Ostriker 1996). A power-law assumption
140: was also used by LP00. Numerical simulations of the ISM in 3D, however, usually
141: do not have a large enough inertial range to produce good power laws
142: and this can complicate our interpretation of the results. To
143: deal with this issue, below we describe a procedure for ``correcting''
144: the spectral indices of the numerically-simulated density and velocity
145: fields while preserving their underlying spatial correlations.
146:
147: We describe the way how our numerical data were produced, and the
148: correlation between density and line-of-sight velocity in section
149: \ref{sec:num_exp}, compare numerical results with the predictions of
150: LP00 in section \ref{sec:stats}, and provide a general discussion in
151: section \ref{sec:disc}. Finally, our
152: conclusions are summarized in section \ref{sec:sum}.
153:
154:
155: \section{Numerical Data} \label{sec:num_exp}
156:
157: \subsection{Simulations and Spectrum Modification} \label{sec:sims}
158:
159: In order to test whether the results from the analytical study of
160: LP00 carry over to the case
161: when the density and velocity fields are correlated according to
162: self-consistent fluid evolution, in the following sections we explicitly
163: calculate the spectral slopes for velocity channel-map data in
164: position-position-velocity (PPV) cubes obtained
165: from a three-dimensional simulation of the ISM at intermediate size
166: scales (3--300 pc). The simulation is nearly identical to the one presented by
167: Pichardo et al.\ (2000), except at slightly lower resolution. Both the
168: simulation and the procedure for obtaining the velocity channel data
169: have been discussed at length in that paper, and we refer the reader
170: to it for details. In this section we just outline the information
171: necessary for the purposes of the present paper. Channel
172: maps are essentially maps of column density within a given
173: line-of-sight (LOS) velocity interval. Throughout this paper we refer
174: to such column density as ``emissivity'' or ``map intensity'', in
175: analogy to the
176: observational situation. For the sake of simplicity,
177: thermal broadening, whose effects on the statistical analysis are
178: discussed by LP00, is not considered.
179:
180: The simulation represents a cubic box of 300 pc on a side on the
181: Galactic plane, centered at the solar Galactocentric distance. The
182: magneto-hydrodynamic equations are solved on a $100^3$ Cartesian grid
183: with the $x$, $y$ and $z$ directions respectively representing the
184: radial, tangential and vertical directions in the Galactic disk. A
185: pseudospectral scheme with periodic boundary conditions is used, which
186: includes self-gravity, parameterized
187: heating and cooling, and modeled star formation, such that a ``star''
188: (a local heating source which causes its surroundig gas to expand) is
189: turned on wherever the density exceeds a
190: threshold $\rho_{\rm c}$ and $\nabla \cdot {\bf u} < 0$. The ``stars''
191: remain on for 3.7 Myr after the criterion is met. This procedure is
192: intended to mimic HII region expansion from OB star ionization
193: heating, and provides an energy source for maintaining the turbulence
194: in the simulation. Energy injection by supernovae is not included due
195: to limitations of the numerical scheme, but the total energy injected
196: per stellar event over the stellar lifetime is of the same order of
197: magnitude as that which would be injected by a supernova event. The
198: simulations also include the Coriolis force corresponding to a
199: rotation of the frame around the Galactic center every $2 \times 10^8$
200: yr, and a shear in the $x$-$y$ plane of the form $u_0 = 2.4
201: ~\sin (2 \pi y/300~{\rm pc})$ km s$^{-1}$, where $u$ is the
202: $x$-component of
203: the velocity field. The turbulent motions occur on top of this
204: shearing velocity. This sinusoidal shear is not
205: highly realistic, but is the only
206: one possible with periodic boundary conditions, and was introduced
207: by Passot et al.\ (1995) as a
208: crude approximation of the galactic shear. Due to the periodic
209: boundary conditions, no stratification is present in the
210: $z$-direction, but this is not too strong a concern given the range of
211: scales represented by the simulation.
212:
213: Because the star formation prescription prevents the density from
214: reaching values significantly larger than $\rho_{\rm c}$, we turn it
215: off after the turbulence is fully developed, and focus on a snapshot
216: of the simulation shortly after that time. At the time of the
217: snapshot, the maximum and minimum values of the (number) density are 109 and
218: 0.43 cm$^{-3}$, respectively. A uniform magnetic field of 1.5 $\mu$G
219: parallel to the $x$-direction is included, on top of which turbulent
220: magnetic fluctuations induced by the stellar energy injection
221: occur. The maximum and minimum values of the magnetic field strength
222: are 12.5 and $4.5 \times 10^{-2} \mu$G. An image of the density field
223: is shown in fig.\ \ref{fig:den_uz}a. We adopt the $z$-direction in
224: the simulation (perpendicular to the Galactic plane) as the LOS
225: direction, in order to prevent the shear from introducing power
226: unrelated to the turbulent fluctuations into the velocity
227: spectrum. Other than that, since the simulation has no
228: vertical stratification, the $z$ direction is statistically equivalent
229: to the $y$ direction. Only the $x$-direction, parallel to the mean
230: magnetic field, is special. Figure
231: \ref{fig:den_uz}b shows an image of the LOS-component of the velocity
232: field. The structures in
233: the simulations have been discussed at length by Pichardo et al.\
234: (2000).
235:
236: However, neither the density nor the LOS-velocity fields have spectra
237: well suited for studying the emissivity spectrum since, due
238: to the low
239: resolution of the simulation, the spectra are not clear power laws,
240: while most turbulence theories, including LP00's, consider power-law spectra.
241: To circumvent this problem, in what follows we use modified density
242: and velocity fields, such that their spectra are indeed power laws,
243: but the phase coherence of the original fields is preserved.
244: A partial justification for this procedure stems from the fact
245: that the spatial information is contained in the phases of the
246: Fourier decomposition of a given field, while the spectrum is related
247: exclusively to the relative amplitudes of the various modes (Armi \&
248: Flament 1985). Unfortunately, this justification is not complete, because
249: the velocity-density coherence does change to some extent as the spectral
250: slope is modified. We discuss this issue in some more detail
251: in \S \ref{sec:correl}.
252:
253: As an indication of what would be realistic indices for the fields, we
254: note that incompressible MHD simulations
255: that resolve the inertial range (Cho \& Vishniac 2000a,b) tend to
256: result in a Kolmogorov-type spectrum with slope $-11/3$,
257: as theoretically predicted
258: by Goldreich \& Shridhar (1995), while highly compressible simulations
259: in less than 3D (e.g., Passot et al.\ 1995; Scalo et al.\ 1998 [2D];
260: Gammie \& Ostriker 1996 [1+2/3 D])
261: tend to give shock-spectra for the velocity with slopes near $-4$ and
262: density spectra with slopes near $-2$. Thus, one may expect
263: spectral indices close to these values
264: also in 3D highly compressible turbulence, so that
265: the spatial correlations from our simulations should probably be most
266: appropriate for those ranges of values.
267: On the other hand, we do not know beforehand
268: what sort of phase coherence should be present for either
269: shallower or steeper spectra, and therefore in these cases our present
270: study is limited to testing whether the formulae by LP00 are correct
271: in the presence of the particular sort of correlations produced by our
272: simulations.
273:
274: The spectral index modification procedure is as follows. We perform a
275: Fourier decomposition of the density and LOS velocity as
276: %
277: \begin{eqnarray}
278: &&
279: \rho(\x) = \sum_\k \rho_\k \exp \big(i \k \cdot \x \big) \\
280: &&
281: u_z(\x) = \sum_\k \uzk \exp \big(i \k \cdot \x \big),
282: \label{eq:Fourier_dec}
283: \end{eqnarray}
284: %
285: where \x\ is the position vector, $\rho_{\bf k}$ and $u_{\bf k}$ are
286: the Fourier
287: amplitudes of the fields (we drop the subindex $z$ in the latter for
288: convenience) and \k\ is the (vector) wavenumber. In general, the
289: amplitudes are of the form $X_\k = |X_\k| e^{i \phi}$,
290: where $|X_\k|$ and $\phi$ are respectively the magnitude and phase of
291: $X_\k$. Thus, we replace the magnitudes of both fields by new ones
292: (labeled by primes) satisfying $|{X_\k}^\prime|^2 \propto k^n$, where
293: $n$ is the desired power-law index. The phases are
294: unchanged. The modified amplitudes are then transformed back to
295: configuration space, rendering new fields which preserve the spatial
296: structure of the original fields but with the desired power-law
297: spectrum. We avoid aliasing by including only modes with wavenumbers
298: lying within a circle of radius equal to half the resolution in
299: wavenumber space in the inverse transform.
300: As an example, in fig.\ \ref{fig:old_new_spec} we show the
301: original power spectrum of the density field together with the
302: resulting modified spectrum for $n=-4$, while fig.\ \ref{fig:old_new_dens}
303: shows cuts through the original and modified density fields, to
304: appreciate the kind of structural changes arising from the
305: spectrum-modification procedure. In the remainder of this paper, we reserve
306: the symbol $n$ for the index of the density power spectrum, and denote
307: the velocity power (or energy) spectrum index by $\mu$.
308:
309: Using this procedure, we construct a suite of density fields with
310: spectral indices $-2.5$, $-4$, and $-5$, and of velocity fields with spectral
311: indices $-3.2$, $-3.7$ (the Kolmogorov value), and $-4.5$. From these,
312: we then construct the corresponding PPV cubes for each
313: density-velocity pair, which are used in the next section to
314: obtain the projected intensity and velocity-channel maps.
315:
316:
317: \subsection{Density-Velocity correlation} \label{sec:correl}
318:
319:
320: To characterize the correlation between velocity and density one
321: can use the following function
322: %
323: \begin{equation}
324: C(\r)=\frac{\langle \rho(\x)\ \u(\x+\r)\cdot\r/|r|\rangle}{\sigma_\rho
325: \sigma_{\u}},
326: \label{eq:LP00_corr}
327: \end{equation}
328: %
329: where \r\ is the spatial separation (or ``lag''), \x\ is the spatial
330: position, $\rho$ is the density field, \u\ is the total
331: three-dimensional velocity vector, and
332: $\sigma_\rho$ and $\sigma_{\u}$ are respectively the
333: standard deviations of the density and of the velocity.
334: The averaging is performed over all \x-space and over all angles of
335: the lag vector \r.
336: This function is similar to the function $F({\bf r})$ introduced
337: in LP00 to calculate the effect of maximal allowable velocity-density
338: correlations. The difference between eq.\
339: (\ref{eq:LP00_corr}) and $F({\bf r})$ amounts
340: to the normalization and the $\cos\theta$ dependence, where $\theta$ is the
341: angle between ${\bf r}$ and the $z$-axis.
342:
343: However, since the projection along \r\ precludes computing this
344: correlation using Fourier transform techniques, and computing the full
345: correlation numerically in
346: physical space would take prohibitely long times, in this paper we
347: compute a related
348: correlation involving only the LOS components of both the lags and of
349: the velocity vector. This
350: correlation, given by
351: %
352: \begin{equation}
353: C(|z|)=\frac{\langle \rho(\x)\ {\bf u}(\x+\z)\cdot\z/|z|\rangle}{\sigma_\rho
354: \sigma_{u_z}},
355: \label{eq:cross_corr}
356: \end{equation}
357: %
358: is much more economical numerically, but contains the same
359: essential information as eq.\ (\ref{eq:LP00_corr}). In
360: eq.\ (\ref{eq:cross_corr}), $\z=(0,0,z)$ is the separation along the
361: LOS and the
362: averaging is performed over all positions $\x$ in the simulation
363: box and over separations $z$ and $-z$. This correlation is shown in
364: fig.\ \ref{fig:correl}a for the original density and LOS-velocity fields.
365:
366: As mentioned in \S \ref{sec:sims}, the correlations do change to some
367: extent upon the change in spectral indices. We have observed that the
368: correlation increases as the slopes of either the density or the
369: velocity spectra are made steeper. Since a steeper (resp.\ shallower)
370: slope means that the small scales are depressed (resp.\ enhanced) with
371: respect to the large scales, the observation implies that the density
372: and velocity fields are more similar to each other at large scales
373: than at small
374: ones. The correlation between the modified fields most resembles that
375: between the original fields when the modified slopes are closest to
376: those of the original spectra.
377:
378: The correlations observed are moderate,
379: having a maximum $\sim 0.25$ (in absolute value) for the original fields, and
380: $\sim 0.4$ for the modified fields with steepest slopes.
381: In retrospect, this result can be understood in terms of the fact that
382: the fluid equations link one field with the spatial derivatives of the
383: other: in the continuity equation, the density evolution is determined
384: by the {\it divergence} of the velocity field, and in the momentum
385: equation, the velocity evolution is linked to the pressure {\it
386: gradient}, which at best can be related to the density gradient, if
387: the flow behaves in an approximately barotropic way (e.g.,
388: V\'azquez-Semadeni, Passot \& Pouquet 1996). Thus, stronger correlations
389: may be expected between the density and the velocity divergence, and
390: between the velocity and the density gradient, but not so much between the
391: plain density and LOS-velocity fields. Indeed, fig.\ \ref{fig:correl}b
392: shows the correlation between the original density field and the divergence of
393: the original 3D velocity field. A larger (negative) correlation, of
394: up to $-0.53$ is seen, supporting this interpretation.
395:
396: Whether or not a density-velocity correlation of the observed strength
397: can alter the results of LP00 is not clear {\it a priori}, and this is the
398: motivation for our study below.
399:
400: \section{Statistics of Velocity-Channel Maps} \label{sec:stats}
401:
402: Unlike the observational situation, in numerical simulations
403: the density and velocity fields are available directly. This
404: allows us to simulate observations and to control the accuracy
405: with which the individual 3D statistics are recovered.
406:
407: Spectroscopic observations usually deal with velocity channel maps
408: in which the velocity resolution is determined by the instrument.
409: In the case of our numerical simulation, velocity resolution
410: is not an issue, and we can produce as many velocity channels as desired
411: from the original density and LOS-velocity data cubes. Surely
412: the statistics within an individual slice
413: are degraded as the slice gets thinner, producing progressively higher
414: levels of shot noise, but we compensate for this
415: effect by averaging over the whole set of slices available.
416:
417: To compare our results with the theoretical predictions of LP00 we recall
418: that two possible regimes were found there, depending on the width
419: of the velocity channels. They were termed
420: ``thin'' and ``thick'' slicings, and their emergence is well motivated
421: physically. If we consider turbulence
422: at a scale $l$, the expected squared velocity difference (the
423: second-order structure function) between points
424: separated a distance $l$ scales as
425: $Cl^{m}$ , where $m$ is $2/3$ for
426: Kolmogorov-type turbulence,
427: and $C$ is a constant that depends on the intensity of
428: turbulence (see, e.g., Lesieur 1990). The structure function
429: index $m$ is related to the velocity power spectrum's index $\mu$ by
430: $m=-\mu-3$. Thus, when this velocity difference is larger
431: than the width
432: of the velocity slice, the slice is considered {\it thin}, and
433: {\it thick} otherwise.
434: LP00 showed that different slopes for the emissivity spectrum
435: are expected in the two regimes mentioned above. It is obvious that
436: whether the slice is either thick or thin with respect to the
437: characteristic turbulent velocity difference depends on the scale
438: being considered, and
439: therefore, since the thin and thick asymptotics differ in general,
440: we expect to see a change of the slope of the emissivity spectrum at
441: some transition scale which depends on the channel width.
442:
443: To test the predictions of LP00, in fig.\ \ref{fig:xi} we plot the
444: spectra of the
445: emissivity fluctuations in velocity slices of the PPV data cubes (``channel
446: maps'') ({\it dash-dotted lines}), the spectra of the 3D
447: density fields ({\it solid lines}), and the spectra of the density in
448: thin spatial slices of the 3D density field ({\it dashed lines}), for
449: various combinations of the density and velocity spectral slopes.
450: These plots confirm the theoretical expectations.
451: First of all, the transition from thick to thin slices depends
452: on the spectral index of the velocity.
453: For relatively small
454: numbers of slices we have observed that the slices are still thin
455: for shallow indices and show signatures of becoming effectively thicker
456: for steep indices. The width of the slice is given by
457: $\Delta V/N$, where $\Delta V=V_{\rm max}-V_{\rm min}$ and $N$ is the
458: number of slices. The value of $\Delta V$
459: along the LOS will be a factor of a few higher than the dispersion
460: $\sigma$ along the same LOS. Conservatively, we shall assume this
461: factor to be 5, which corresponds to $2.5\sigma$
462: positive and negative excursions. The squared dispersion
463: over the box size $L$, $\sigma^2$, equals $CL^m$ for power-law statistics.
464: Therefore, in order for the slice to be thin, it is required that
465: $Cr^m>(2.5)^2 CL^m/N^2$ , where $r$ is the scale
466: under consideration in the plane of the slice, and is essentially the
467: inverse of wavenumber in the emissivity spectrum.
468: This means that $N>2.5(L/r)^{m/2}$.
469: Another expected feature is the flattening of the
470: spectrum at large $k$ (small separations $r$) due to shot noise.
471: In our case we derive the slope
472: considering scales from $r\approx L$ down to $r\approx 0.1 L$, where $L$
473: is the length of the integration box. Figure
474: \ref{fig:xi} shows that for $r$ smaller than $0.1 L$ the emissivity
475: spectrum indeed flattens.
476:
477: The considerations above mean
478: that for the steepest velocity index of $-4.5$ (i.e. $m=1.5$),
479: the number of slices should be
480: larger than $30$ in order for them to be thin. On the other hand,
481: an increase in the number of slices results in higher shot noise due
482: to the discrete sampling of the velocity along the line of sight (we
483: only have 100 samples along each line of sight).
484: Thus, we have adopted a slice width
485: of $0.03 \Delta V$, which we have found to correspond to
486: thin slicing for the entire parameter space that we explore, but does
487: not introduce too high a level of shot noise.
488:
489: For thin slices and sufficiently {\it steep}
490: density spectra ($n<-3$), LP00 predicted that the measured emissivity
491: spectrum {\it does not} depend on the density spectral slope,
492: and has an index
493: %
494: \begin{equation}
495: ({\it thin~slice~index~if~density~spectrum~is~steep})=-3+m/2.
496: \label{steep}
497: \end{equation}
498: %
499: For example, for Kolmogorov
500: turbulence, $m=2/3$, and therefore an emissivity index of $-2.66$ is
501: expected. Conversely, for {\it shallow} density spectra, i.e. when
502: $n>-3$, the
503: spectral index in thin slices is expected to be (LP00)
504: %
505: \begin{equation}
506: ({\it thin~slice~index~if~density~spectrum~is~shallow})=n+m/2.
507: \label{shallow}
508: \end{equation}
509: %
510: Table~1 provides a quantitative comparison between the theoretical
511: predictions of LP00 and our numerical results, showing the spectral
512: indices of the emissivity in the velocity channels as a function of
513: the indices of the underlying density and velocity fields. The
514: theoretical predictions of LP00 (given by
515: equations (\ref{steep}) and (\ref{shallow}))
516: are indicated in parentheses, while the
517: measured values in the simulated fields are shown in bold type. It is seen
518: that in all cases the agreement is within 10\%, for both ``shallow''
519: and ``steep'' density spectral indices. We stress that the spectra of
520: the 3D density and
521: velocity data cubes, and of the 2D velocity slices, are obtained by
522: integrating with the appropriate configuration space factors, which are
523: $r^2dr$ and $rdr$ respectively.
524:
525: The main prediction of LP00 is the importance of the
526: velocity statistics in the determination of the emissivity
527: statistics, specifically, its power spectrum. Our numerical results confirm
528: this conclusion. Indeed, Table~1 shows that as the velocity spectral
529: index is varied (downwards along the columns in Table~1), significant
530: variations occur in the emissivity spectral index. Moreover, for steep
531: density spectra, the emissivity statistics are roughly independent of
532: the density spectral index.
533: For example, consider the cases with velocity indices $-3.2$ and $-11/3$.
534: It is seen that as the underlying density spectral
535: index is changed from $-4$ to $-5$, the emissivity spectral index in thin
536: channels remains constant within the precision of our
537: measurements. For the case with velocity index of
538: $-4.5$, the emissivity index changes slightly (from $-2.1$ to $-2.3$)
539: for the same variation of the density index as above, but such slight
540: change is within the uncertainty of 0.2, and brackets the value
541: predicted by the LP00 theory.
542: %In fact, we have actually
543: %considered density slopes starting at $-3.2$
544: %and, in particular, we do not observe a systematic
545: %change of the emissivity slope
546: %as the density spectrum becomes steeper.
547: {\it By itself, this result sends
548: a message of
549: warning against attempts to interpret features in the velocity
550: channel maps as actual density enhancements.}
551:
552:
553: \begin{table}[h]
554: \begin{displaymath}
555: \begin{array}{lrrr} \hline\hline\\
556: {\rm Density~index} \rightarrow& \multicolumn{1}{c}{-2.5~({\rm shallow})} &
557: \multicolumn{1}{c}{-4~({\rm steep})} & \multicolumn{1}{c}{-5~({\rm steep})}
558: \\[1mm]
559: \hline\\
560: {\rm Velocity~index}\\
561: \hfil\downarrow\hfil\\
562: \hline\hline\\
563: -3.2 & {\bf -2.2}~(-2.40)& {\bf -2.9}~(-2.90) & {\bf -2.9}~(-2.90)\\[1mm]
564: \hline \\[1mm]
565: -11/3 &{\bf -1.9}~(-2.17) & {\bf -2.6}~(-2.67) & {\bf -2.6}~(-2.67)\\[1mm] \hline\\
566: -4.5 & {\bf -1.8}~(-1.75) & {\bf -2.1}~(-2.25) & {\bf -2.3}(-2.25)\\[1mm] \hline
567: \end{array}
568: \end{displaymath}
569: \caption{Spectral index of the emissivity fluctuations in {\it thin}
570: velocity channels as a function of the density (top row) and velocity
571: (left column) fields' spectral indices. The actual measured values are given
572: in bold font, while the theoretical predictions are given in parentheses.
573: The differences between the theoretical and observed values are within
574: the uncertainties of determining the spectral slope. The uncertainty
575: of the slope fittings is $\sim 0.2$, due to the effects of shot noise
576: and slight imperfections in the spectrum modification procedure, $\sim
577: \pm 0.1$ in the imposed spectral index.}
578: \label{tab:2Dspk_asymp}
579: \end{table}
580:
581: %An analysis of the results corresponding to other velocity spectral
582: %indices (not shown) also gives good correspondence between the theoretical
583: %predictions of LP00 and
584: %the numerical simulations.
585: The above discussion suggests that, whatever the velocity
586: spectrum, it can be restored from observations. A possible ambiguity
587: that arises in the case of shallow density spectra, when
588: both the velocity and density contribute to the channel map spectrum,
589: can be resolved by considering several slice widths. For example,
590: it is obvious that if integration is performed over the whole
591: velocity range, the resulting statistics can depend on the
592: density only. The corresponding density inversion is discussed
593: by Lazarian (1995). For power-law statistics, the spectral index for
594: {\it very thick} velocity-integrated slices
595: equals the spectral index of the underlying 3D density.
596: We have verified this result by integrating over the velocity (the
597: plots corresponding to the integrated 2D statistics
598: coincide with those of the 3D spectra in fig.\ \ref{fig:xi}).
599:
600: We have also confirmed other predictions from LP00, such as the change
601: of the density spectral index from $n$ in volume to $n+1$ in a thin {\it spatial} density
602: slice, as shown in fig.\ \ref{fig:xi} by the dashed lines.
603: Another important prediction of LP00 was the
604: gradual steepening of the emissivity spectrum in the velocity slices
605: as their width increases. This prediction was confirmed by
606: Stanimirovic (2000) and Stanimirivic \& Lazarian (2000) on the basis
607: of 21~cm Small Magellanic Cloud data
608: and fig.\ \ref{fig:slice_var} shows that
609: this indeed happens in our simulation data. In particular, the maximum
610: possible thickening occurs when the velocity information is averaged
611: out. In our case this corresponds to integration through the velocity
612: box, or, equivalently, from an integration throughout the full line
613: of sight in
614: the original 3D
615: density data. It is evident that in this case the velocity
616: fluctuations should not matter and only density statistics should
617: influence the results. We have verified that both a pure density
618: integration along the LOS and a velocity integration of the emissivity
619: data cubes give the same spectral index.
620:
621:
622: \section{Discussion} \label{sec:disc}
623:
624: In \S \ref{sec:correl} we have shown that the density and LOS-velocity
625: fields in our ISM simulation are moderately correlated.
626: An important question is whether there are interstellar
627: situations in which the velocity-density correlations are much
628: stronger than in our simulations and whether the results by LP00 are
629: applicable to those situations. The answer to this question may
630: be obtained by analyzing simulations of regions with physical
631: conditions different from those assumed here.
632: However, even in the case of
633: strong correlations, the analytical results of LP00 may hold,
634: as it was shown by LP00 for a particular case of maximal possible
635: velocity-density correlations. Further research should clarify the
636: issue.
637:
638: Astrophysical implications of the asymptotic relations obtained in
639: LP00 are discussed at length in Lazarian (1999). Here we will stress two
640: points. First, the agreement between our measured emissivity spectral
641: indices and the
642: predictions of LP00 makes us more optimistic about the application of
643: the LP00 technique to emission-line studies of actual interstellar
644: turbulence. Second, it also gives us confidence that the
645: interpretation by LP00 of the HI intensity spectrum
646: (Green 1993; Stanimirovic et al.\ 1999) as arising from velocity
647: fluctuations with a Kolmogorov (i.e., $-11/3$ spectral index) spectrum
648: is correct.
649:
650: It is worth pointing out that, even though the notion of
651: Kolmogorov turbulence has been mentioned frequently throughout the paper, we
652: do not wish to imply that the ISM in general exhibits Kolmogorov
653: turbulence. Arguments why
654: the Kolmogorov (1941) description is not likely to be adequate for the ISM
655: have been given by several authors (e.g., Scalo 1987; Passot, Pouquet
656: \& Woodward 1988; Lazarian 1995;
657: V\'azquez-Semadeni 1999; V\'azquez-Semadeni et al.\ 2000). Briefly
658: speaking, the ISM flow, unlike
659: that in the Kolmogorov description, is magnetized and compressible.
660: One would think that this should change the scaling. Indeed, strong
661: compressibility may give rise to a
662: spectrum of shocks, with a slope of $-4$ (in the notation convention
663: used in this paper). On the other hand,
664: a theory of magnetized, incompressible turbulence recently
665: put forward by Goldreich and Sridhar (1995) predicts the same
666: $k^{-11/3}$ Kolmogorov scaling
667: for motions perpendicular to magnetic field lines. Such motions would
668: dominate small-scale asymptotics and therefore the measured spectrum
669: may stay Kolmogorov-like, although the nature of the cascade is influenced
670: by magnetic field. Moreover, the interpretation of the observational
671: data given in LP00, and supported in this paper,
672: is suggestive that the Kolmogorov scaling may ultimately survive
673: compressibility or, alternatively, that the HI data considered there
674: sample gas that is only weakly compressible. Further work, both theoretical and
675: observational, is needed to resolve this issue. In particular,
676: application of the techniques used here to molecular-line data should prove
677: of great interest.
678:
679: To conclude we should stress that for Kolmogorov turbulence there
680: exists a {\it coincidence} between the spectral index of a thin,
681: 2D spatial slice of the
682: density, which is:
683: %
684: \begin{equation}
685: ({\it spectral~index~of~thin~2D~density~slice})=n+1
686: \label{2Ddensity}
687: \end{equation}
688: %
689: and the spectral index of the emissivity fluctuations in a thin
690: velocity channel when the density spectrum is steep
691: (e.g. Kolmogorov) and the velocity spectrum follows a Kolmogorov law.
692: Indeed, the former,
693: calculated using eq.~(\ref{2Ddensity}), is $-11/3+1=-8/3$
694: while the latter, calculated using
695: eq.~(\ref{steep}), takes the same value $-3+1/3=-8/3$. Although
696: the accidental character of this coincidence was mentioned in LP00,
697: our experience shows that it does cause confusion.
698: For instance, we have had to confront the point of view
699: that the observed spectrum of HI fluctuations is a simple consequence
700: of eq.~(\ref{2Ddensity}). This is a {\it fallacy} and, while we do not wish to
701: repeat the arguments of LP00, Table~1
702: shows that the naive ``rule'' given by eq.~(\ref{2Ddensity})
703: does not return the correct emissivity spectral indices
704: for velocity spectra other than Kolmogorov's, and that the coincidence is
705: accidental for the Kolmogorov index. The correct
706: formula for calculating the spectral index of the emissivity fluctuations in
707: thin slices when the density spectrum is steep
708: is given by eq.~(\ref{steep}).
709:
710: \section{Summary and implications} \label{sec:sum}
711:
712: In this paper we have used data from a numerical simulation of
713: compressible MHD turbulence in the ISM to study the correlation
714: between the density and LOS-velocity field, and the dependence
715: of the spectral index of the velocity-channel emissivity on the indices of
716: the original three-dimensional velocity and density fields. To do so,
717: we have {\it a)} computed the cross-correlation between the density and
718: LOS-velocity fields, {\it b)} modified their spectral slopes to
719: pre-determined values, and {\it c)} produced PPV data from them.
720:
721:
722: From the PPV data, we computed in turn the emissivity spectrum in velocity
723: slices of those cubes (``channel maps''), allowing us to directly test the
724: results from the analytical study by LP00. We have found
725: that its predictions hold also for the case of fields obeying
726: the fluid conservation equations.
727: In particular, we have found that for steep
728: density spectra with power-law indices $n<-3$, the emissivity spectral index
729: does not depend on the actual value of the density index and is
730: determined exclusively by the velocity fluctuations. This indicates
731: that only the velocity field is
732: responsible for the structure existing in thin slices of the
733: PPV cubes if the density spectrum is steep enough, in agreement with the
734: previous result by Pichardo et al.\ (2000) that the morphology in the
735: channel maps is more resemblant of that of the LOS-velocity field than
736: that of the density field, and with previous suggestions that
737: ``objects'' identified in position-velocity maps may not correspond to
738: actual density features (Adler \& Roberts 1992; Pichardo et al.\ 2000;
739: Ostriker, Stone \& Gammie 2000).
740:
741: The spectral index measured for the emissivity fluctuations in the
742: numerical simulation
743: was within 10\% of the value given by relation (\ref{steep}),
744: derived by LP00, in which $m$ is
745: the exponent of the second-order structure
746: function. For a shallow density spectrum, namely
747: $n>-3$, we found that the emissivity index depends on both velocity and
748: density. Again the LP00 expression for the resulting power-law index,
749: relation (\ref{shallow}), provides a satisfactory description of
750: our measurements.
751: We also estimated the accuracy of our measurements both by integrating out the
752: velocity within PPV data cubes and by taking thin slices of the density
753: data cubes. These cases have straightforward analytical solutions (see
754: LP00) and served as benchmarks.
755: We also verified that variations of the emissivity spectrum
756: resulting from the change of the velocity slice width (see fig.\
757: \ref{fig:slice_var}) are very
758: similar to those reported in the studies by Stanimirovic (2000) and
759: Stanimirovic \& Lazarian (2000),
760: in which the width of observational HI data slices was varied to
761: test the predictions from LP00.
762:
763: %Our main result can be formulated as follows:\\
764: %The density and LOS-velocity fields in a compressible, MHD simulation
765: %of interstellar turbulence are moderately correlated, and
766: %the statistics of the channel-map emissivity fluctuations resulting
767: %from those fields are consistent with the
768: %analytical predictions of LP00.
769: %This suggests that those predictions can be safely used to
770: %interpret observational spectroscopic data in terms of the
771: %density and velocity statistics of interstellar turbulence.
772:
773: In brief, our results imply that:\\
774: 1. Velocity creates small scale structure within slices of PPV data cubes
775: and therefore the interpretation of features seen in PPV data cubes
776: as density structures (``clouds'') may be misleading.\\
777: 2. For sufficiently shallow density spectra, velocity dominates the
778: statistics of intensity fluctuations within PPV slices.\\
779: 3. Emissivity spectra become steeper as the width of the velocity
780: slices increases
781: and, for sufficiently thick slices, the spectra are dominated by
782: density enhancements only.
783:
784: \acknowledgements
785:
786: Part of this work was completed while two of us (A.L.\ and E.V.-S.)
787: participated in the Astrophysical Turbulence Program of the Institute
788: for Theoretical Physics at the University of California at Santa
789: Barbara. The numerical simulation was performed on the Cray Y-MP 4/64
790: of DGSCA, UNAM. This work has received partial funding from CONACYT, M\'exico,
791: through grant 27752-E to E.V.-S.\ and from NSF, USA, through grant PHY94-07194.
792:
793:
794:
795: \begin{thebibliography}{99}
796:
797: \bibitem {} Adler, D. D., \& Roberts, W. W. 1992, ApJ, 384, 95
798: \bibitem{} Armi, L. \& Flament, P. 1985, J. Geophys. Research,
799: 90, no.\ C6, 11779
800: \bibitem{} Brunt, C. \& Heyer, M. H. 2000, preprint (astro-ph/0011200)
801: \bibitem{} Cho \& Vishniac 2000a, ApJ, 539, 253
802: \bibitem{} Cho \& Vishniac 2000b, ApJ, 538, 217
803: \bibitem{} Gammie, C.F. and Ostriker, E.C. 1996, ApJ 466, 814
804: \bibitem{} Green, D.A. 1993 MNRAS, 262, 328
805: \bibitem{} Goldreich, P. \& Sridhar, S. 1995, ApJ, 438, 763
806: \bibitem{} Kolmogorov, A. 1941, Compt. Rend. Acad. Sci. USSR, 30, 301
807: \bibitem{} Heyer, M. H. \& Schloerb. F.P 1997, APJ, 475, 173
808: \bibitem{} Lazarian, A. 1995, A\&A, 293, 507
809: \bibitem{} Lazarian, A. 1999, in Plasma Turbulence and Energetic
810: Particles, ed. M. Ostrowski and R. Schlickeiser, Cracow, 28, astro-ph/0001001
811: \bibitem{} Lazarian \& Pogosyan 2000, ApJ, 537, 720
812: \bibitem{} Lesieur, M. 1990, Turbulence in Fluids, 2nd ed.\
813: (Dordrecht: Kluwer)
814: %\bibitem{} Mac Low, M.-M. \& Ossenkopf, V. 2000, A\&A, 353, 339
815: \bibitem{} Miesch, M.S., \& Bally, J. 1994, ApJ, 429, 645
816: \bibitem{} Ostriker, E.C., Stone, J.M. \& Gammie, C.F. 2000 (preprint:
817: astro-ph/0008454)
818: \bibitem{} Passot, T., Pouquet, A. \& Woodward, P. 1988, A\&A
819: \bibitem{} Passot, T., V\'azquez-Semadeni, E., \& Pouquet, A. 1995
820: \bibitem{} Pichardo, B., V\'azquez-Semadeni, E., Gazol, A., Passot,
821: T. \& Ballesteros-Paredes, J. 2000, ApJ, 532, 353
822: \bibitem{} Rosolowsky E.W., Goodman, A.A., Wilner, D.J., \& Williams, J.P.
823: \ 1999, ApJ, 524, 887
824: \bibitem{} Scalo, J.M. 1987, in Interstellar Processes,
825: eds. D.F. Hollenbach \& H.A. Thronson, Reidel, Dordreicht, 349
826: \bibitem{} Scalo, J.M., V\'azquez-Semadeni, E., Chappell, D. \&
827: Passot, T. 1998, ApJ, 504, 835
828: \bibitem{} Stanimirovic, S. 2000, in ``The Evolution of
829: Galaxies: Observational Clues", eds. J.M. Vilchez,
830: G. Stasinska \& E. Perez (in press)
831: \bibitem{} Stanimirovic, S. \& Lazarian, A. 2000, ApJ, submitted
832: \bibitem{} Stanimirovic, S., Staveley-Smith, L., Dickey, J.M.,
833: Sault, R.J., \& Snowden, S.L. 1999, MNRAS, 302, 417
834: \bibitem{} V\'azquez-Semadeni, E., Passot, T. \& Pouquet, A. 1996,
835: ApJ, 473, 881
836: \bibitem{} V\'azquez-Semadeni, E. 1999, in Millimeter \& Submillimeter
837: Astronomy: Chemistry and Physics in Molecular Clouds'', eds. W. F.
838: Wall, A. Carrami\~nana, L. Carrasco, and P. F. Goldsmith (Dordrecht:
839: Kluwer), 161
840: \bibitem{} V\'azquez-Semadeni, E., Ostriker, E. C., Passot, T., Gammie, C.
841: \& Stone, J. 2000, in ``Protostars \& Planets IV'', eds. V. Mannings,
842: A. Boss \& S. Russell (Tucson: Univ.\ of Arizona Press), 3
843:
844:
845: \end{thebibliography}
846: \clearpage
847:
848: \begin{figure}
849: \plottwo{fig1a.ps}{fig1b.ps}
850: %\plotone{fig01.ps}
851: \caption{{\it a) (Left Panel)} Three-dimensional (3D) logarithmic image of the
852: original density field. The maximum and minimum density values are
853: $109$ and 0.43 cm$^{-3}$. {\it b) (Right panel)} 3D image of the
854: LOS-velocity field, i.e., the $z$-component of the velocity vector. In
855: both cases, whiter color means larger values.}
856: \label{fig:den_uz}
857: \end{figure}
858:
859: \begin{figure}
860: \plotone{fig2.ps}
861: %\plotone{fig02.ps}
862: \caption{Spectra of the density field in original form ({\it solid line}) and
863: after modification to an index of $-4$ ({\it dash-dotted line}). The
864: wavenumber $k$ here is defined as $L/\lambda$, where $\lambda$ is the
865: wavelength associated with $k$, and $L$ is the box size.}
866: \label{fig:old_new_spec}
867: \end{figure}
868:
869: \begin{figure}
870: \plottwo{fig3a.ps}{fig3b.ps}
871: %\plotone{fig03.ps}
872: \caption{Two-dimensional cuts through the 3D density field in original
873: form ({\it left panel}) and after modification to a spectral index of
874: $-5$. Note that the modified-spectrum image appears smoother, because the
875: relative importance of the small scales has been reduced in this case.}
876: \label{fig:old_new_dens}
877: \end{figure}
878:
879: \begin{figure}
880: \plottwo{fig4a.ps}{fig4b.ps}
881: \caption{Cross-correlation between the density and the line-of-sight
882: projection
883: velocity fields ({\it a), left panel}), and between the density and the
884: divergence of the total 3D velocity field ({\it b), right panel}). The
885: density is significantly correlated with the velocity divergence, while
886: the density-velocity correlation is weaker. The non-monotonous character
887: of the correlation defines a range of scales at which
888: the density-velocity correlation is maximal and the LP00 analysis
889: requires testing.}
890: \label{fig:correl}
891: \end{figure}
892:
893:
894: \begin{figure}[ht]
895: {\centering \leavevmode
896: \epsfxsize=.25\columnwidth \epsfbox{fig5_1.ps} \hfil
897: \epsfxsize=.25\columnwidth \epsfbox{fig5_2.ps} \hfil
898: \epsfxsize=.25\columnwidth \epsfbox{fig5_3.ps}
899: }
900: {\centering \leavevmode
901: \epsfxsize=.25\columnwidth \epsfbox{fig5_4.ps} \hfil
902: \epsfxsize=.25\columnwidth \epsfbox{fig5_5.ps} \hfil
903: \epsfxsize=.25\columnwidth \epsfbox{fig5_6.ps}
904: }
905: {\centering \leavevmode
906: \epsfxsize=.25\columnwidth \epsfbox{fig5_7.ps} \hfil
907: \epsfxsize=.25\columnwidth \epsfbox{fig5_8.ps} \hfil
908: \epsfxsize=.25\columnwidth \epsfbox{fig5_9.ps}
909: }
910: \caption{
911: Spectra of the 3D density fields ({\it bold lines}) and of the
912: emissivity in the 2D channel maps ({\it dash-dotted lines}) for all nine
913: combinations of density and velocity spectral indices reported in
914: Table 1. For comparison, the spectra of 2D {\it thin,
915: spatial} slices of the density are also shown ({\it dashed
916: lines}). The wavenumber $k$ here is defined as $2 \pi L/\lambda$,
917: where $L$ and $\lambda$ are as in the caption to fig.\
918: \ref{fig:old_new_spec}.
919: }
920: \label{fig:xi}
921: \end{figure}
922:
923:
924: \begin{figure}[ht]
925: {\centering \leavevmode
926: \epsfxsize=.45\columnwidth \epsfysize=.45\columnwidth
927: \epsfbox{fig6a.ps} \hfil
928: \epsfxsize=.45\columnwidth \epsfbox{fig6b.ps} \hfil
929: }
930: \caption{{\it Left Panel:} {\it Thin dash-dotted lines:} emissivity
931: spectra for velocity slices of
932: various widths, for velocity spectral index $-4.5$ and density
933: index $-4$. From top to bottom, the curves correspond to
934: slices of width $0.01$, $0.02$, $0.04$,
935: $0.06$, $0.1$ $0.25$, and $0.5$ times $\Delta V$. {\it Thick solid
936: line:} power spectrum of the LOS-projected density field (i.e., column
937: density). The emissivity spectra in the velocity slices are seen to
938: approach the spectrum of column density as they get wider. {\it Thick
939: dashed line:} power spectrum of a {\it thin} (1 grid cell) spatial density
940: slice. The emissivity spectra in thin velocity slices are seen to be
941: strongly influenced by the velocity field, as they have different
942: slopes than the spectra of either the column density or the thin
943: density spatial slice. {\it Right Panel:} Variation
944: of the emissivity spectral index as a function of the velocity slice
945: width for the same data cube.}
946: \label{fig:slice_var}
947: \end{figure}
948:
949: \end{document}
950:
951: