1: % TURBULENT CLOUD CORES
2: %
3: \documentstyle[aasms4]{article}
4: \input{psfig}
5:
6: % Latest correction: 21.04.00
7: % --------------------------
8: %\received{4 October 1997}
9: %\accepted{23 October 1997}
10: %\journalid{337}{15 January 1998}
11: %\articleid{11}{14}
12:
13: \lefthead{Burkert \&\ Bodenheimer}
14: \righthead{Turbulent Molecular Cloud Cores: Rotational Properties}
15:
16: \begin{document}
17:
18: % symbols used in text
19: \def\sol{$_\odot$} % solar mass symbol
20: \def\x{{$\times$}}
21: %
22: % new symbols for less than (greater than) and approximately equal to
23: % COMMENTED OUT! USE \la AND \ga INSTEAD
24: %
25: %\def\ltsimeq{\,\raise 0.3 ex\hbox{$ < $}\kern -0.75 em
26: % \lower 0.7 ex\hbox{$\sim$}\,}
27: %\def\gtsimeq{\,\raise 0.3 ex\hbox{$ > $}\kern -0.75 em
28: % \lower 0.7 ex\hbox{$\sim$}\,}
29: %
30: % Start of text
31: %
32: \title{Turbulent Molecular Cloud Cores: Rotational Properties$^\dagger$}
33:
34: \author{Andreas Burkert\altaffilmark{1}}
35: \affil{$^1$Max-Planck-Institut f\"ur Astronomie, K\"onigstuhl 17, D--69117
36: Heidelberg, Germany; burkert@mpia-hd.mpg.de}
37:
38: \author{Peter Bodenheimer\altaffilmark{2}}
39: \affil{$^2$University of California Observatories/Lick Observatory,
40: Board of Studies in Astronomy and Astrophysics,
41: University of California, Santa Cruz, CA 95064;
42: peter@ucolick.org}
43:
44: \begin{abstract}
45: The rotational properties of numerical models of centrally condensed,
46: turbulent molecular cloud cores
47: with velocity fields that are characterized by Gaussian random fields
48: are investigated. It is shown that the observed line width --
49: size relationship can be reproduced if the velocity power spectrum
50: is a power-law with $P(k) \propto k^{n}$ and $n = -3$ to $-4$.
51: The line-of-sight velocity maps of these cores show velocity
52: gradients that can be interpreted as rotation. For $n = -4$,
53: the deduced values of angular
54: velocity $\Omega$ = 1.6 km s$^{-1}$ pc$^{-1} \times $ (R/0.1 pc)$^{-0.5}$
55: and the scaling relations between $\Omega$ and the
56: core radius $R$ are in very good agreement with the observations.
57: As a result of the dominance of long wavelength modes, the cores also have a net specific
58: angular momentum with an average value
59: of $J/M$ = 7 $\times 10^{20} \times$ ($R$/0.1 pc)$^{1.5}$ cm$^2$ s$^{-1}$
60: with a large spread. Their internal dimensionless rotational
61: parameter is
62: $\beta \approx 0.03$, independent of the scale radius $R$.
63: In general, the line-of-sight velocity gradient of an individual
64: turbulent core does not provide a good estimate of its internal specific
65: angular momentum. We find however that the distribution of the specific angular momenta
66: of a large sample of cores which are described by the same power spectrum
67: can be determined very accurately from the distribution of their line-of-sight velocity
68: gradients $\Omega$ using the simple formula $j=p \Omega R^2$ where p depends
69: on the density distribution of the core and has to be determined from
70: a Monte-Carlo study. Our results show that for centrally condensed cores
71: the intrinsic angular momentum is overestimated by a factor of 2-3 if $p=0.4$
72: is used.
73: \end{abstract}
74:
75: \keywords{hydrodynamics -- stars: formation -- ISM: clouds -- infrared sources}
76: \noindent
77: $^\dagger$UCO/Lick Observatory Bulletin, No..........
78:
79: \section{Introduction: Rotating Cloud Cores}
80:
81: %*****************
82: Although the rotation in the dense ($n \sim 10^4- 10^5$ cm$^{-3}$) cores
83: of molecular clouds has small dynamical effects compared with gravity,
84: %*****************
85: it has important consequences once a core collapses
86: to form a single star or binary system with associated disks. The
87: distribution of separations of binary systems, the distribution of
88: disk sizes, and the properties of emerging planetary systems all
89: depend on the range of angular momenta among the different cores as well
90: as on the angular momentum distributions within individual cores.
91: Most theoretical calculations of the collapse
92: of rotating cloud cores (see Bodenheimer et al. 2000 for a review)
93: assume as an initial condition that the core is uniformly rotating;
94: furthermore, the observational determination of rotational properties
95: of cores are based upon a model of uniform rotation (Goodman et al. 1993;
96: Goldsmith \& Arquilla 1985; Menten et al. 1984.) However the material
97: in molecular clouds is observed to have supersonic line widths over
98: a wide range of scales indicating a supersonic, irregular velocity field.
99: The line width correlates with size,
100: providing evidence that has been interpreted in terms of
101: turbulent motions (Larson 1981, Myers \& Gammie 1999; see below), probably
102: associated with a magnetic field (Arons \& Max 1975). Observed line profiles in
103: molecular clouds have been shown to be consistent with Gaussian
104: velocity fields with a Kolmogorov spectrum
105: (Dubinski, Narayan, \& Phillips 1995, Klessen 2000).
106: Even cores on scales of 0.1 pc or less
107: show non-thermal motions whose velocity dispersion is comparable to,
108: but definitely less than, the sound speed (Barranco \& Goodman 1998).
109: Thus the rotational properties of cores may be more complicated than
110: the simple law of uniform rotation would indicate.
111:
112: The evidence for rotation in the cores of molecular clouds
113: (Myers \& Benson 1983; Goldsmith \& Arquilla 1985) consists of observations
114: of gradients in the line-of-sight velocity along cuts across the cores.
115: Goodman et al. (1993; updated by Barranco \& Goodman 1998)
116: have observed cores in the size range 0.06 to 0.6 pc
117: in the NH$_3$ molecule, finding evidence of rotation in 29 out of 43 cases
118: studied and finding velocity gradients $\Omega$ in the
119: range 0.3 to 3 km s$^{-1}$ pc$^{-1}$
120: (corresponding to 10$^{-14}$ --10$^{-13}$ s$^{-1}$).
121: Over this range of scales, $\Omega$ scales roughly as $R^{-0.4}$, and the specific
122: angular momentum $j \equiv J/M$ as inferred from $\Omega$
123: scales roughly as $R^{1.6}$, with a value of
124: $j \approx 10^{21}$ cm$^2$ s$^{-1}$ on the smallest scales measured.
125: The dimensionless quantity $\beta$,
126: defined as the ratio of rotational kinetic energy divided by the absolute value of
127: the gravitational energy, shows no trend with $R$ and has a mean value of about 0.03
128: with a large scatter. It is also found that cores tend to have gradients
129: that are not in the same direction as gradients found on larger scales in
130: the immediate surroundings (Barranco \& Goodman 1998), an effect which
131: again suggests the presence of turbulence.
132:
133: In this paper, following previous suggestions (Goldsmith \& Arquilla 1985; Goodman et al. 1993;
134: Dubinski et al. 1995) regarding the connection between turbulence and rotation,
135: we investigate
136: %**************
137: a possible
138: %*************
139: origin of rotation in turbulent cores
140: %+++++++++++++++++++++++
141: and investigate the relationship between their line-of-sight velocity maps and
142: their intrinsic rotational properties.
143: Unfortunately, a complete and comprehensive theory of turbulence does still not exist.
144: We therefore adopt the standard simple approach to describe the velocity field inside such
145: cores by a Gaussian random field. This model assumes random phase correlations between the different
146: modes. In order to analyze a large statistical sample we also neglect the coupling
147: between the density and velocity field, which is justified as the flow
148: in observed molecular cloud cores is mildly subsonic.
149: In subsequent papers we plan to include this coupling and to
150: analyze in detail how the rotational properties
151: change during the evolution and collapse of turbulent cores with initial Gaussian random fields.
152: %+++++++++++++++++++++++
153:
154: Here, we try to answer the question whether uniform rotation is a reasonable assumption for such cores,
155: i.e. whether the velocity gradients, determined from line-of-sight velocity maps,
156: in combination with the assumption of rigid body
157: rotation, provide a good estimate of the
158: intrinsic specific angular momenta of turbulent cores.
159: We show that even if the motions in cores are completely
160: random, in many cases systematic velocity gradients in the line-of-sight
161: components of the velocity are present with values that are in good agreement with
162: the observations, that the cores can have net intrinsic
163: angular momenta, and that the line-of-sight velocity gradients provide
164: on average a good
165: estimate of their distribution of specific angular momenta.
166: In \S 2 we describe how random velocity distributions can be derived that
167: are consistent with observed line width -- size relations.
168: \S 3 outlines how the projected angular velocity $\Omega$ is determined from line-of-sight
169: velocity maps. \S 4 describes the results of a set of 4000 different realizations
170: of the turbulent velocity field and shows that the model can explain the typical values
171: of $\Omega$, $j$, and $\beta$ observed in molecular cores. That the model also
172: explains the trends with core size is shown in \S 5.
173: \S 6 explores the amount of intrinsic specific angular momentum of turbulent
174: cores and the relationship between the intrinsic angular momentum and the
175: projected velocity gradient. Conclusions are presented in \S 7.
176:
177: \section{Construction of Models }
178:
179:
180: The velocity field $\vec{v}(\vec{x})$ can be characterized by its Fourier modes
181:
182: \begin{equation}
183: \vec{v}( \vec{x}) = \frac{1}{(2\pi)^3} Re \left[ \int \vec{\hat{v}}(\vec{k}) e^{i \vec{k} \vec{x}}
184: d^3k \right]
185: \end{equation}
186:
187: \noindent Dubinski et al. (1995) show that line-of sight velocity
188: profiles in molecular clouds are consistent with a Gaussian random field with
189: a Kolmogorov spectrum $P(k) \propto k^{-11/3}$. The relation between the turbulent
190: spectrum and the line width - size relation has been discussed by Gammie
191: \& Ostriker (1996) and Myers \& Gammie (1999); the latter authors also suggest random relative
192: phases for the spectral components.
193: %+++++++++++++++++++++++++++++
194: Assuming an isotropic velocity field,
195: %++++++++++++++++++++++++++++++
196: the Fourier components $\vec{\hat{v}}(\vec{k})$ are
197: completely specified by the power spectrum $P(k) = \langle \vec{\hat{v}}^2(k) \rangle$
198: where $k=|\vec{k}|$.
199: The power spectrum will depend on the physical properties of the velocity field
200: which characterizes molecular clouds and has to be determined from the observations.
201: Larson (1981) showed that the observed internal velocity dispersion $\sigma$
202: of a molecular cloud region is well
203: correlated with its length scale $\lambda$, following approximately a Kolmogorov law
204:
205: \begin{equation}
206: \sigma (\lambda) \sim \lambda^q.
207: \end{equation}
208:
209: \noindent with $q \approx 0.38$. Later work found that line width scales with clump size
210: roughly according to $\sigma \sim \lambda^{0.5}$ (Leung, Kutner, \& Mead 1982;
211: Scoville, Sanders, \& Clemens 1986; Solomon et al. 1987). Additional studies
212: (Fuller \& Myers 1992; Caselli \& Myers 1995; Myers \& Fuller 1992) measured
213: slopes in the range 0.25 to 0.75.
214: In a more recent investigation, Goodman et al. (1998) find that the power-law
215: slope $q$ depends somewhat on
216: $\lambda$ with virtually constant line widths ($q=0$) for $\lambda <$ 0.1 pc and
217: $q$ = 0.5 for larger cores.
218: Given equation (2), the power spectrum must also follow a power-law
219:
220: %+++++++++++++++++++++++++++++++++++++++++++++++++
221: \begin{equation}
222: P(k) \propto k^n ,
223: \end{equation}
224: %+++++++++++++++++++++++++++++++++++++++++++++++++
225:
226: \noindent if $q>0$ . Its slope
227: depends on the observed value of $q$, and the relation between $q$ and $n$
228: can be determined through filtering the velocity field by passing over it a volume of characteristic
229: size $\lambda$ and filtering out waves with $k < 1/\lambda$.
230: This leads to a variance
231:
232: \begin{equation}
233: \sigma^2({\lambda}) = \langle \vec{v}^2(\vec{x}) \rangle_{\lambda} \sim
234: - \int_{1/\lambda}^{\infty} P(k) k^2dk \sim \lambda^{-(n+3)}
235: \end{equation}
236:
237: \noindent Note that the integral in equation (4) converges only if $n < -3$.
238: Comparing equation (4) with equation (2), we can determine $n$ from the observed
239: line width-size relationship:
240:
241: \begin{equation}
242: n = -3 - 2 q .
243: \end{equation}
244:
245: Typical molecular cloud cores with $q \approx 0.5$ will be
246: characterized by a power-law index $n \approx -4$ (see also Myers \& Gammie 1999).
247: In the following analysis we will explore
248: the turbulent origin of rotation of molecular cores with a velocity power
249: spectrum $-4 \leq n \leq -3$, which
250: seems to cover most of the observed range.
251: %+++++++++++++++++++++++++++++++++++++++++++++++++++++
252: The energy spectrum $E(k)$ which corresponds to a velocity power spectrum
253: $P(k) \propto k^n$ depends on the dimensionality $d$ of the flow and is given
254: by (Myers \& Gammie 1999) $E(k) \propto k^{n+d-1}$. With $d=3$ and
255: $n \approx -4$
256: the corresponding energy spectrum is $E(k) \propto k^{-2}$.
257: %+++++++++++++++++++++++++++++++++++++++++++++++++++++
258:
259: The velocity field is calculated numerically on a Cartesian 3-dimensional
260: grid with $N$ grid cells in each direction
261: %++++++++++++++++++++++++++++++++++++++++++++
262: (Cen 1992).
263: %++++++++++++++++++++++++++++++++++++++++
264: Due to the limited resolution we
265: only include modes with $\lambda > 2R/N$. This is no severe restriction if $N$ is large enough
266: ($N \geq 16$) as waves with small wavelengths do not contribute to
267: global rotational properties of the cores, which are preferentially
268: determined by waves of $\lambda \sim R$. The results presented below
269: use $N=64$ with a cutoff at a wavelength of $\lambda < R/32$. Test
270: calculations with larger $N = 256$ and correspondingly smaller cutoff
271: wavelengths show that $N=64$ gives adequate resolution. After generating the
272: 3-dimensional velocity field $\vec{v}$ using equation (1) we subtract
273: the center--of--mass velocity. We adopt a coordinate system where ($x,z$) defines
274: the plane of the sky and the $y$ direction is along the line of sight.
275: Adopting a density distribution $\rho(\vec{x})$, we now
276: can generate two-dimensional maps, with $N \times N$ pixels,
277: of density weighted averaged line-of-sight velocities $V_{LS}(x,z)$ which can be
278: analysed and compared with the spectral line maps of observed molecular cloud cores:
279:
280: \begin{equation}
281: V_{LS}(x,z) = \frac{\int \rho(\vec{l}) v_y(\vec{l}) d{\vec{l}}}{\int \rho(\vec{l}) dl}
282: \end{equation}
283:
284: \noindent where the integration is done along a given line of sight $\vec{l}$
285: through the entire cube.
286: Observations indicate that cores are in general centrally condensed
287: with roughly Gaussian density distributions
288: (Ward-Thompson et al. 1994; Andr\'e et al. 1996; review by Bodenheimer et al.
289: 2000). In the following we will use a spherically
290: symmetric density distribution of the form
291:
292: \begin{equation}
293: \rho(r) = \rho_c \times \exp \left(-3 \left( \frac{r}{R_{max}} \right)^2 \right).
294: \end{equation}
295:
296: \noindent where $r$ is the distance from the center, and
297: $R_{max}$ is the outer radius of the core, at which the density
298: is assumed to be a factor 20 smaller than the central value $\rho_c$.
299: Additional test calculations with a constant density show that the results
300: do not depend critically on
301: the specific choice of the density distribution.
302: Observed cores are typically analysed within a radius $R$ where the surface density
303: is above half the maximum value. Here
304: we scale all physical quantities to a typical molecular core with a radius
305: $R=0.1$ pc. For a core with a density distribution given by equation (7),
306: $R \approx 0.5 R_{max}$.
307: The data presented by Goodman et al. (1993, their table 1) indicate that these
308: cores have a wide range of masses.
309: A rough average value is $M \approx$ 5 M$_{\odot}$, which we adopt as the typical core mass.
310: Their typical 1-dimensional velocity dispersions are $\sigma_{1d} \approx$ 0.13 km s$^{-1}$
311: (Goodman et al. 1998);
312: %++++++++++++++++++++++++++++++++++++++++++++++++
313: $\sigma_{1d}$ determines the amplitude $A$ of the power spectrum
314: $P(k)=A \times k^n$.
315: %+++++++++++++++++++++++++++++++++++++++++++++++++
316:
317:
318: \section{Determination of the projected angular velocity and the specific
319: angular momentum}
320:
321: In order to compare our models with observations, we analyze the
322: line-of-sight
323: velocity maps $V_{LS}$ (see eq. [6]) using the least-squares method
324: proposed by Goodman et al. (1993), which minimizes the difference between
325: the observed line-of-sight velocity map and the map expected for a
326: rigidly rotating core. To provide an approximate
327: match to the observations, we include only the inner regions of the cores,
328: where the surface density is above the half-maximum value. In the case of rigid body
329: rotation the line--of--sight velocity is given by
330:
331: \begin{equation}
332: V_{LS} = V_0 + \omega_z x - \omega_x z
333: \end{equation}
334:
335: \noindent where $\omega_z$ and $\omega_x$ are the $z$ and $x$ coordinates of the
336: three-dimensional angular velocity vector,
337: %***********************
338: and $V_0$ is the velocity
339: of the center of mass.
340: %***********************
341: This equation assumes that the rotation axis goes through
342: the center of the core, which is the origin of the coordinate system.
343: Note that we cannot obtain any information about $\omega_y$ from
344: the velocity map. To determine $\omega_x$ and $\omega_z$ we minimize
345: the error $\epsilon$:
346:
347: \begin{equation}
348: \epsilon = \sum_i
349: (V_0 + \omega_z x_i - \omega_x z_i - V_{LS,i})^2
350: \end{equation}
351: \noindent where we sum over all pixels $i$ lying within the inner region
352: defined above.
353: Here $x_i$ and $z_i$ are the $x$ and $z$ coordinates of the $i$th pixel,
354: and $V_{LS,i}$ is the
355: measured line-of-sight velocity in that pixel.
356: We then solve the following set of equations:
357: \begin{equation}
358: \frac {\partial \epsilon }{\partial V_0} =
359: \frac {\partial \epsilon} {\partial \omega_z} =
360: \frac {\partial \epsilon} {\partial \omega_x} = 0.
361: \end{equation}
362: Defining the mean value over the ($x,z$) plane of a quantity $q$ as
363: $\langle q \rangle = \frac{1}{K \times K} \sum_{i=1}^{K \times K} q_i$,
364: where $q_i$ is its value in the i-th pixel and $K = N/2$,
365: and noting that $\langle x \rangle$ =
366: $\langle z \rangle$ = $\langle xz \rangle$ = 0, we find
367: \begin{eqnarray}
368: V_0 & = & \langle V_{LS} \rangle = 0 \nonumber \\
369: \omega_x & = & - \frac {\langle z \cdot V_{LS} \rangle} {\langle z^2 \rangle} \\
370: \omega_z & = & \frac {\langle x \cdot V_{LS} \rangle} {\langle x^2 \rangle} \nonumber
371: \end{eqnarray}
372: It can easily be shown that this solution minimizes the error, namely that
373: the second derivatives are all positive.
374: Given $\omega_x$ and $\omega_z$,
375: %*******************
376: which again are mean quantities
377: averaged over the $K~\times~K$ surface,
378: %*****************
379: we define the projected $\Omega =
380: (\omega_x^2 + \omega_z^2)^{1/2}$ and we determine the angle of the
381: projected rotation axis which is defined by tan $\alpha = \omega_z / \omega_x$.
382: As an additional rotational property of the core we determine its
383: total specific angular momentum
384:
385: \begin{equation}
386: j =\left( \sum m \vec{v}\times \vec{x} \right) / \sum m,
387: \end{equation}
388:
389: \noindent where $m$
390: is the mass of a cell, and the sum goes over all
391: cells of the three-dimensional grid which are located within the
392: projected half-maximum region which is used to determine the
393: projected $\Omega$.
394:
395:
396: \section{The projected rotational properties of turbulent cores}
397:
398: Although the velocity fields drawn from the same $P(k)$ are statistically
399: equivalent, each realization results from a different set of random
400: numbers and therefore is unique. As a result we expect that the
401: projected and intrinsic rotational properties of the cores may differ significantly
402: from one case to the next and will also change as a function of
403: the index $n$. Examples are shown in Figure 1, which
404: illustrates 3 different line--of--sight velocity maps for each of the cases
405: %******************
406: $ n = -4, -3, {\rm and} -2$~(from top to bottom).
407: %*****************
408: The frames in the first column correspond to an
409: example with relatively high $j$ and high $\Omega$,
410: the second, to low $j$ and high $\Omega$, and the third, to both low $j$ and
411: $\Omega$. One can clearly see that with increasing $n$ the power on
412: small scales increases, leading to more substructure and less systematic
413: motion.
414: %***************
415: In the remainder of the paper, we consider only the values
416: $n = -3, -4$, which correspond to the range of observed $q$ values
417: (eq. [2]) in cores (0 to 0.5, respectively).
418: %***************
419:
420: The left panels of Figure 1 show examples of a systematic
421: velocity gradient which can easily be interpreted as a rotation.
422: Even for $n= -3$ (second row of Fig. 1) we find cases with relatively
423: well-defined velocity gradients. That there indeed exists a global
424: projected velocity gradient that could be interpreted as rigid body rotation
425: is illustrated in Figure 2, which shows the line-of-sight component of the
426: velocity averaged over slits parallel to the projected rotation axis as a
427: function of distance from the axis for five cases
428: %***********
429: for $n = -4$,
430: %***********
431: similar to that of the upper left panel of Figure 1.
432: The origin of the velocity gradients are dominant long wavelength
433: modes with small phase shifts with respect to the core center,
434: like $v(x)=v_0 \sin (\pi x/R_{max})$. In the inner regions of the projected
435: velocity maps such a wave would indicate a rigid body rotation
436: with $\Omega \approx \pi v_0/R_{max}$. The dashed lines in Figure 2 show
437: the outer (not observed) parts of the cores where $|v|$ reaches a maximum
438: and decreases again.
439:
440: The probability for observing a certain value of a velocity gradient
441: can be determined
442: as the frequency with which such values would be found given a large number
443: of projected velocity maps,
444: constructed with different sets of random numbers and projection angles.
445: The left panels in
446: Figure 3 show the distributions of
447: %**************************%
448: $\Omega$ (upper panel), $j$ (middle panel), and $\beta$ (lower panel),
449: %***********************
450: for both
451: $n = -4$ and $n = -3$, for a set of 4000 random realizations in each case.
452: Following Goodman et al. (1993) the parameter $\beta$ and the specific internal
453: angular momentum $j$ are determined from $\Omega$, adopting rigid body
454: rotation and a constant core density:
455:
456: \begin{eqnarray}
457: j & = & 0.4 \Omega R^2 \nonumber \\
458: \beta & = & \frac{\Omega^2R^3}{3 G M}
459: \end{eqnarray}
460:
461: \noindent Here $R$ is the radius of the inner region.
462: These assumptions are not justified in typical turbulent and centrally
463: condensed cores. Whether $j$ as determined from equation (13) does indeed
464: provide a good estimate for the internal specific angular momentum in
465: turbulent cores will be discussed in greater detail in \S 6.
466:
467: For $n = -4$, corresponding to the standard observed
468: line width-size relation, one can see a broad distribution
469: in the values of $\Omega$, with a peak at 1.4 km s$^{-1}$ pc$^{-1}$
470: and an average dispersion of order 1 km s$^{-1}$ pc$^{-1}$.
471: As expected, the width of the distribution and the value at the peak
472: decrease with increasing $n$.
473: As shown in the second row of Figure 3, the spread of $\Omega$ leads also to
474: a large spread in the $j$ values which are determined from
475: $\Omega$, with $j$ peaking
476: at $1.7 \times 10^{21}$ cm$^2$ s$^{-1}$. Finally, in the last row of
477: Figure 3, the apparent distribution of the
478: dimensionless rotation parameter $\beta$ is shown,
479: derived from $\Omega$, which peaks at $\beta \approx 0.03$.
480: The distributions of these three quantities for the case
481: $n = -3$ (dashed lines in the left panels of Fig. 3)
482: peak at values
483: %**************
484: 1/2 -- 2/3 of
485: those found for the case $n = -4$.
486: %*************
487: Because of the large spread, the average of
488: a large number of observations would be required to determine the average rotational
489: properties of turbulent cores with similar statistical properties.
490: %++++++++++++++++++++++++++++++++
491: For example, the right-hand panels of Figure 3 show the distribution of mean
492: values of $\Omega$, j and $\beta$, with each value being the average over
493: 50 different random realizations. It shows that from a typical observational sample of 50 cores,
494: the average rotational properties can be determined with an accuracy of order 10\%.
495: %+++++++++++++++++++++++++++++++++++++++++++++++++
496:
497: In summary, our simulations lead to the following characteristic rotational
498: quantities for turbulent cores with $n$ in the range $-3$ to $-4$:
499: $\Omega \approx$ 0.5 -- 2.0 km s$^{-1}$ pc$^{-1}$,
500: $j \approx$ 0.5 -- 2.5 $\times 10^{21}$ cm$^2$ s$^{-1}$,
501: and $\beta \approx$ 0.01 -- 0.05, with
502: a large spread in all quantities. These values are in very good agreement with
503: the observations (Goodman et al., 1993) which for cores with radii of 0.1 pc
504: predict $\Omega \approx$ 1.3 km s$^{-1}$ pc$^{-1}$, $\beta \approx 0.03$
505: and $j \approx 1.2 \times 10^{21}$ cm$^2$ s$^{-1}$.
506:
507: \section{Scaling Relations}
508:
509: Turbulent cores, where the relevant units are the radius $R$, mass $M$,
510: and velocity dispersion $\sigma$, have a well-defined relationship between
511: j, $\Omega$, $\beta$, and size:
512:
513: \begin{eqnarray}
514: \Omega & \propto & \sigma /R \nonumber \\
515: j & \propto & \sigma R \\
516: \beta & \propto & (\Omega^2 R^3)/M \propto \sigma^2 R /M. \nonumber
517: \end{eqnarray}
518: %****************************
519: \noindent The first of these relations follows from the method of
520: determining $\Omega$ (eq. [11]) and the other two follow from
521: equation (13) with the assumption of uniform rotation.
522: %****************************
523: Adopting the standard line width--size relation
524: $\sigma \propto R^{0.5}$, we find $\Omega \propto R^{-0.5} $ and
525: $j \propto R^{1.5}$. The scaling relation for $\beta$ depends on the
526: gravitational energy of the core, so an additional mass--radius relation
527: is needed. Observations indicate that the velocity dispersion $\sigma^2
528: \propto M/R$, leading to $M \propto R^2$, and to a $\beta$ which is
529: independent of radius. A similar derivation of the scaling
530: relations has been presented
531: by Goodman et al. (1993). Using these relations and the values for
532: $\Omega$, $j$ and $\beta$ as derived for 0.1 pc cores,
533: we now can predict the average values
534: of rotational properties of cores with different sizes. Figure 4 shows that the
535: calculated average rotational core properties as function of radius
536: in the range $-4 \leq n \leq -3$ as well as the predicted spread
537: in $\beta$, are in very good agreement with the observational data.
538: %**********************
539: In fact, in the size range above 0.1 pc, $n=-4$ agrees much better
540: than $n = -3$, consistent with the line width--size relation.
541: %**********************
542:
543: \section{The correlation between projected velocity gradient and
544: specific angular momentum}
545:
546: The previous section showed that
547: %******************
548: a Gaussian random velocity field with a power
549: spectrum that is in agreement with the line width--size relationship is a
550: possible explanation of the rotational properties of molecular cloud
551: cores, as inferred from line-of-sight velocity maps.
552: %******************
553: Using this model, we now investigate the relationship between the observed
554: projected velocity gradient $\Omega$ and the intrinsic angular momentum $j$ of the
555: cores.
556:
557: It is generally assumed that $\Omega$ provides a good estimate of $j$.
558: This would certainly be expected in the case of solid body rotation.
559: There, the main uncertainty is the angle $i$ between the line-of-sight direction
560: and the rotation axis. For random orientations,
561: the average value of $\sin^2i $= 2/3,
562: and the mean specific angular momentum can be determined accurately
563: for a sample of cores of a given size.
564:
565: However, if cores are characterized by Gaussian random fields the situation is much
566: more complex and the assumption of rigid body rotation is not valid.
567: Now the line-of-sight velocity field does not provide a good estimate of
568: the amplitudes and phases of the various
569: velocity modes in the perpendicular directions.
570: This is demonstrated in Figure 5, which plots $\Omega$ versus
571: $j$ for a large sample of cores, generated with a power-spectrum $P(k) \propto k^{-4}$.
572: $\Omega$ does not correlate
573: with $j$.
574: It might at first seem surprising that cores with small specific angular momentum
575: can show large projected velocity gradients.
576: To illustrate this effect let
577: us consider a very simple velocity field with a dominant long-wavelength mode
578: in the $x$- and $y$-direction and with zero phase shift with respect to the center:
579:
580: \begin{equation}
581: \vec{v} = v_0 \left( \sin (\pi \frac{x}{R_{max}}) \vec{e}_y + \eta \sin (\pi
582: \frac{y}{R_{max}}) \vec{e}_x \right) .
583: \end{equation}
584:
585: \noindent $R_{max}$ is the radius of the core and $\vec{e}_x$ and
586: $\vec{e}_y$ are the unit vectors in the $x$ and $y$ directions, respectively.
587: The case $\eta = -1$ corresponds to a vortex centered at the origin, while
588: the case $\eta = +1$ places the origin at the ``saddle point" between four
589: vortex cells.
590: Suppose that the line--of--sight direction is along the $y$-axis. As
591: $\Omega$ is determined from the velocity field inside a radius $r \leq R_{max}/2$,
592: where $\sin (\pi x/R_{max}) \approx \pi x/R_{max}$, the measured velocity gradient
593: will be $\Omega \approx \pi v_0/R_{max}$, independent of $\eta$.
594: The left panel of Figure 6 shows the velocity field
595: in the case of $\eta = -1$.
596: The core clearly contains a net angular momentum
597: $j$ around the $z$-axis and $\Omega$ provides a good estimate of $j$.
598: However, for $\eta = 1$ (right panel of Fig. 6),
599: the net angular momentum is $j=0$ whereas
600: the value of $\Omega$ has not changed.
601:
602: In summary, turbulent cores
603: are in general not rigid body rotators. Although they could contain a
604: net angular momentum, as shown in the previous sections, their complex
605: velocity field makes it impossible to determine the intrinsic angular
606: momentum of a core, given its line-of-sight velocity map. This effect
607: results partly from the fact that compressional velocity components
608: introduce line-of-sight velocity gradients that are not related to
609: rotation.
610:
611: The situation is however much more promising if one considers the angular
612: momentum distribution of a large sample of cores that are all
613: described by the same power spectrum $P(k)$. As $P$ does not depend on
614: the direction of $\vec{k}$, a set of maps of the line-of-sight velocity
615: contains much more
616: information regarding the internal kinematical properties
617: of the cores.
618: This is shown in Figure 7 which compares the distribution of
619: specific angular momenta
620: $N(j_{pred}$) as inferred from the line-of-sight
621: velocity gradient $\Omega$ (dashed lines)
622: with the intrinsic distribution of specific angular momenta $N(j)$ (solid lines)
623: of cores with exponential (Fig. 7a) or constant (Fig. 7b) density
624: profiles, and
625: with $P(k) \propto k^{-4}$.
626: The values of
627: $j_{pred}$ have been calculated from $\Omega$
628: using the equation $j_{pred} = p \Omega R^2$ with
629: $p = 0.14$ for the exponential density sphere and $p=0.4$ for the constant
630: density distribution.
631: The intrinsic specific angular momentum $j$, and from this $N(j)$, is determined
632: using the full information of the 3-dimensional velocity field (eq. [12]),
633: summed over the ``observed" inner region.
634: The predicted distribution is in excellent agreement
635: with the real distribution if $p$ is chosen carefully taking into
636: account the underlying density profile. For the constant density case
637: the required value of $p=0.4$ is actually consistent with the real value
638: of the moment of inertia. In the centrally condensed core, however, a value
639: smaller than the actual moment of inertia ($p=0.26$) is required to fit the
640: actual j-distribution.
641: The distribution of $j_{pred}$ is slightly wider because this value
642: is determined from random projections.
643: Note that the $\sin i $-correction is not required in this case.
644:
645: \section{Conclusions }
646:
647: Random Gaussian velocity fields with power spectra $P(k) \propto k^{-3}$
648: to $k^{-4}$ can reproduce both the observed line width -- size relationship
649: and the observed projected rotational properties of molecular cloud cores.
650: They therefore can be used in order to investigate their intrinsic velocity
651: fields in detail or to generate initial conditions for simulations of core collapse
652: and single star or binary formation.
653: We have shown that, due to the dominant large-wavelength modes, these cores contain
654: a non-zero specific angular momentum of order
655: $J/M = 7 \times 10^{20} \times (R/0.1 {\rm pc})^{1.5}$ cm$^2$ s$^{-1}$.
656: As a result of the random nature of the velocity field, cores which are statistically
657: identical, that is which are described by the same power spectrum, show a large
658: spread in their rotational properties, which is in qualitative agreement with the
659: large spread in observed binary periods (Duquennoy \& Mayor 1991).
660: However, the
661: median $j$ for pre-main-sequence and main-sequence binaries is about an order of
662: magnitude less (Simon 1992) than the value we derive for the cores.
663:
664: The line--of--sight velocity gradient
665: does in general not provide a good estimate of the specific
666: angular momentum of a given core. However, on a statistical basis, the distribution of
667: projected velocity gradients $\Omega$ can reproduce very well the distribution
668: of the specific angular momenta $j$, assuming $j = p \Omega R^2$,
669: where $p$ has to be chosen properly through a Monte Carlo study as presented in this
670: paper. In general, $p$ seems to be smaller than the actual value for the moment
671: of inertia in centrally condensed cores. As a result of this effect, the specific
672: angular momenta of cores are overestimated by roughly a factor of 3 if equation
673: (13) is used.
674:
675: It is somewhat surprising that the shape of the
676: angular momentum distribution as inferred
677: from the line-of-sight velocity gradients using
678: the simple formula of rigid body rotation is in such a good agreement with that of
679: the intrinsic angular momentum distribution of turbulent cores.
680: One possible explanation might be that the fluctuation spectrum is dominated
681: by the large-scale eddies for the values of $n$ that we consider. The
682: effect of projection would normally result in an expectation value
683: of $j_{pred}$ smaller than the actual $j$. This effect may be compensated
684: through the effects of compressional modes that would contribute to the
685: measured velocity gradient and could explain the similar shape and width of
686: the two distributions. More detailed analysis is necessary to explain this
687: result and to explain the relationship between the intrinsic moment of inertia
688: and the value of $p$ required to fit the intrinsic angular momentum distribution
689: from the distribution of line-of-sight velocity gradients.
690:
691: In this paper we assumed that the velocity field is uncorrelated with the
692: spectral line emission and adopted a simple spherically symmetric
693: density profile. In reality the velocity field will affect the density
694: distribution and vice versa. In the case of supersonic turbulence the
695: initially Gaussian velocity field will evolve into a system of shocks.
696: Numerical models (e.g. Mac Low et al. 1998, Ostriker et al. 1999)
697: demonstrate that in this case the initial velocity field does indeed not provide a good
698: estimate of the typical dynamical state of an evolved turbulent cloud.
699: However, for the scales which are investigated in the present paper, cloud regions are
700: mildly subsonic and no strong shocks are expected to form.
701: One therefore might not expect a strong evolution into a dynamical state
702: which is very different with respect to the initial power spectrum.
703:
704: The interaction and dynamical evolution of the velocity and density field in
705: turbulent cores and its effect on
706: line-of-sight velocity maps as well as the core collapse and fragmentation
707: will be investigated in detail in subsequent papers.
708:
709: \acknowledgements
710:
711: This work was supported in part through National Science Foundation
712: grant AST-9618548,
713: in part through the Deutsche Forschungsgemeinschaft
714: (DFG), and in part through a special NASA astrophysics theory program
715: which supports a joint Center for Star Formation Studies at NASA/Ames
716: Research Center, UC Berkeley, and UC Santa Cruz. We acknowledge
717: helpful conversations with Richard Klein, Robert Fisher, and Chris
718: McKee at a conference in July, 1999, where we learned of their
719: work on the collapse of turbulent molecular clouds.
720: We also would like to thank
721: L. Blitz, A. Goodman, and P. Myers
722: for interesting discussions and the referee, Ralf Klessen, for many
723: important comments. AB thanks the staff of
724: Lick Observatory for the hospitality during his visits and PB thanks the
725: staff of the Max-Planck-Institut f\"ur Astronomie for the hospitality
726: during his visits.
727:
728: \vfill\eject
729: \begin{thebibliography}{}
730: \bibitem[]{}{Andr\'e, P., Ward-Thompson, D., \& Motte, F. 1996, A\&A, 314, 625}
731: \bibitem[]{}{Arons, J., \& Max, C. E. 1975, ApJ, 196, L77}
732: \bibitem[]{}{Barranco, J. A., \& Goodman, A. A. 1998, ApJ, 504, 207}
733: \bibitem[]{}{Bodenheimer, P., Burkert, A., Klein, R. I., \& Boss, A. P. 2000, in Protostars and Planets IV,
734: ed. V. Mannings, A.P. Boss, \& S.S. Russell (Tucson: Univ. Arizona Press), in press}
735: \bibitem[]{}{Caselli, P., \& Myers, P.C. 1995, ApJ, 446, 665}
736: \bibitem[]{}{Cen, R. 1992, ApJS, 78, 341}
737: \bibitem[]{}{Dubinski, J., Narayan, R., \& Phillips, G. 1995, ApJ, 448, 226 }
738: \bibitem[]{}{Duquennoy, A., \& Mayor, M. 1991, A\&A, 248, 485}
739: \bibitem[]{}{Fuller, G.A., \& Myers, P.C. 1992, ApJ, 384, 523}
740: \bibitem[]{}{Gammie, C. F., \& Ostriker, E. C. 1996, ApJ, 466, 814}
741: \bibitem[]{}{Goldsmith, P.F., \& Arquilla, R. 1985, in Protostars and Planets II,
742: ed. D.C. Black \& M.S. Matthews (Tucson: Univ. Arizona Press), 137 }
743: \bibitem[]{}{Goodman, A., Benson, P., Fuller, G., \& Myers, P. 1993,
744: ApJ, 406, 528 }
745: \bibitem[]{}{Goodman, A., Barranco, J. A., Wilner, D. J., \& Heyer, M. H. 1998,
746: ApJ, 504, 223}
747: \bibitem[]{}{Klessen, R.S. 2000, ApJ, in press}
748: \bibitem[]{}{Larson, R. B. 1981, MNRAS, 194, 809}
749: \bibitem[]{}{Leung, C.M., Kutner, M.L., \& Mead, K.N. 1982, ApJ, 262, 583}
750: \bibitem[]{}{Mac Low, M.-M., Klessen, R.S., Burkert, A., \& Smith, M.D. 1998,
751: Phys. Rev. Lett., 80, 2754}
752: \bibitem[]{}{Menten, K. M., Walmsley, C. M., Kr\"ugel, E., \&
753: Ungerechts, H. 1984, A\&A, 137, 108}
754: \bibitem[]{}{Myers, P.C., \& Benson, P. 1983, ApJ, 266, 309}
755: \bibitem[]{}{Myers, P.C., \& Fuller, G.A. 1992, ApJ, 396, 631}
756: \bibitem[]{}{Myers, P.C., \& Gammie, C. F. 1999, ApJ, 522, L141}
757: \bibitem[]{}{Ostriker, E.C., Gammie, C. F., \& Stone, J.M. 1999, ApJ, 513, 259}
758: \bibitem[]{}{Scoville, N.Z., Sanders, D.B., \& Clemens, D.P. 1986, ApJ, 310, L77}
759: \bibitem[]{}{Simon, M. 1992, in ASP Conf. Ser. 32, Complementary Approaches to Double and Multiple
760: Star Research, ed. H. A. McAlister \& W. I. Hartkopf (San Francisco: ASP), 41}
761: \bibitem[]{}{Solomon, P.M., Rivolo, A.R., Barrett, J., \& Yahil, A. 1987, ApJ, 319, 730}
762: \bibitem[]{}{Ward-Thompson, D., Scott, P.F., Hills, R.E., \& Andr\'e, P. 1994,
763: MNRAS, 268, 276}
764: \end{thebibliography}
765:
766: \vfill\eject
767: \centerline{FIGURE CAPTIONS}
768: \begin{figure}[p]
769: \vspace{0.01 in}
770: \caption{Maps of the normalized line-of-sight velocity for $n = -4$ ({\it top row}),
771: $n = -3$ ({\it center row}), and $n = -2$ ({\it bottom row}) as determined
772: from eq. (6).
773: In the top row, from left to
774: right, the values of $\Omega$ in units of km s$^{-1}$ pc$^{-1}$ and the
775: intrinsic specific angular momentum
776: $j$ in units of $10^{21}$ cm$^2$ s$^{-1}$ for cores with radii of 0.1 pc are,
777: respectively, (1.9, 0.9), (0.2, 1.0), and (0.4, 0.6).
778: In the center row, these quantities are
779: (0.7, 0.4), (0.06, 0.5), and (0.26, 0.2). In the bottom row,
780: these quantities are (0.16, 0.1), (0.004, 0.1), and (0.06, 0.06).
781: Blue areas correspond to positive velocities (toward the observer), red corresponds
782: to zero velocity, and yellow corresponds to negative velocity.
783: %****************
784: Each frame shows the inner ``observed" region with dimensions one-half
785: those of the full $N^3$ computational grid.
786: }
787: %***************
788: \end{figure}
789:
790: \begin{figure}[p]
791: \vspace{0.01 in}
792: \caption{Five randomly chosen examples of simulated cores with large
793: projected velocity gradients
794: %*************
795: with $n = -4$,
796: %************
797: similar to the upper left panel of Figure 1. The line-of-sight velocity
798: (averaged over a strip parallel to the projected rotation axis with width of
799: R/32) is plotted as a function of distance to the projected axis as determined by
800: the least-squares method.
801: The dashed lines indicate unobserved regions.
802: }
803: \end{figure}
804:
805: \begin{figure}[p]
806: \vspace{0.01 in}
807: \caption{Results of 4000 random realizations of turbulent cores.
808: Histograms for the projected velocity gradient $\Omega$ ({\it top}), the
809: specific angular momentum $j$
810: as inferred from $\Omega$ ({\it center}) and $\beta$ ({\it bottom})
811: as inferred from $\Omega$ (see eq. [13]).
812: In the left panels {\it solid curves} correspond to a power index $n=-4$ and
813: {\it dashed curves} correspond to $n=-3$. The right panels compare
814: the histograms for $n= -4$ ({\it solid curves}) with the residual distribution
815: with each value being the average over 50 different random realizations ({\it filled areas}).}
816: \end{figure}
817:
818: \begin{figure}[p]
819: \vspace{0.01 in}
820: \caption{Observed values (Goodman et al. 1993; Barranco \& Goodman 1998)
821: of $\beta$ ({\it top, crosses}), velocity gradients
822: ({\it center, triangles}), and specific
823: angular momentum ({\it bottom, stars}) as a function of the size of the core.
824: %************
825: {\it Heavy solid curves:} the trends predicted by the model with $n= -4$.
826: {\it Heavy dashed curves:} the trends predicted by the model with $n= -3$.
827: {\it Light solid curves: } the half-maximum points of the calculated
828: distribution for the case $n= -4$.
829: {\it Light dashed curves: } the half-maximum points of the calculated
830: distribution for the case $n= -3$.}
831: %*****************
832: \end{figure}
833:
834: \begin{figure}[p]
835: \vspace{0.01 in}
836: \caption{The projected $\Omega$ values of simulated
837: cores ($ n = -4$)
838: are plotted as a function of their internal specific
839: angular momentum $j=J/M$. }
840: \end{figure}
841:
842: \begin{figure}[p]
843: \vspace{0.01 in}
844: \caption{The velocity field which corresponds to eq. (15) is shown with
845: $\eta = -1$ ({\it left panel}) and $\eta = +1$ ({\it right panel}). The size of the vectors
846: is linearly proportional to the absolute value of the velocity.}
847: \end{figure}
848:
849: \begin{figure}[p]
850: \vspace{0.01 in}
851: \caption{The distribution of specific angular momenta for simulated cores, as inferred from the projected
852: velocity gradient determined from the inner region ({\it dashed lines}), is compared with the intrinsic distribution of
853: their specific angular
854: momenta, calculated from the 3-dimensional velocity field ({\it solid lines}).
855: {\it a):} centrally condensed core. {\it b):} constant density core.}
856: \end{figure}
857:
858:
859: \end{document}
860:
861:
862: