1: \documentclass[12pt,preprint]{aastex}
2:
3: \shorttitle{The Parker Instability}
4: \shortauthors{Kim et al.}
5: \slugcomment{draft of \today}
6:
7: \begin{document}
8:
9: \def\etal{{\it et al.~}}
10: \def\eg{{\it e.g.,~}}
11: \def\ie{{\it i.e.,~}}
12:
13: \title{The Effect of the Random Magnetic Field Component\\
14: on the Parker Instability}
15:
16: \author{Jongsoo Kim}
17:
18: \affil{Korea Astronomy Observatory, 61-1, Hwaam-Dong, Yusong-Ku,
19: Taejon 305-348, Korea \& NCSA, University of Illinois at
20: Urbana-Champaign, 405 North Mathews Avenue, Urbana, IL 61801:
21: jskim@ncsa.uiuc.edu}
22:
23: \and
24:
25: \author{Dongsu Ryu}
26:
27: \affil{Department of Astronomy \& Space Science, Chungnam National
28: University, Daejeon 705-764, Korea: ryu@canopus.chungnam.ac.kr}
29:
30: \begin{abstract}
31:
32: The Parker instability is considered to play important roles
33: in the evolution of the interstellar medium.
34: Most studies on the development of the instability so far have been
35: based on an initial equilibrium system with a uniform magnetic field.
36: However, the Galactic magnetic field possesses a random component
37: in addition to the mean uniform component, with comparable strength
38: of the two components. Parker and Jokipii have recently suggested
39: that the random component can suppress the growth of small wavelength
40: perturbations. Here, we extend their analysis by including gas
41: pressure which was ignored in their work, and study the stabilizing
42: effect of the random component in the interstellar gas with finite
43: pressure. Following Parker and Jokipii, the magnetic field is modeled
44: as a mean azimuthal component, $B(z)$, plus a random radial
45: component, $\epsilon(z) B(z)$, where $\epsilon(z)$ is a random
46: function of height from the equatorial plane.
47: We show that for the observationally suggested
48: values of $\langle\epsilon^2\rangle^{1/2}$, the tension due to the
49: random component becomes important, so that the growth of the
50: instability is either significantly reduced or completely suppressed.
51: When the instability still works, the radial wavenumber of the most
52: unstable mode is found to be zero. That is, the instability is reduced
53: to be effectively two-dimensional. We discuss briefly the implications
54: of our finding.
55:
56: \end{abstract}
57:
58: \keywords{instabilities --- ISM: clouds --- ISM: magnetic fields ---
59: magnetohydrodynamics: MHD}
60:
61: \section{Introduction}
62:
63: The magnetostatic equilibrium of the system of the interstellar gas
64: and magnetic field under the vertical gravitational field of the
65: Galaxy has been shown to be unstable \citep{par66,par67}.
66: The physical mechanism for the instability relies on the fact that
67: a light fluid (represented by the magnetic field) supports a heavy
68: fluid (represented by the gas) and the configuration tends to overturn.
69: It has similarities to the Rayleigh-Taylor instability, when a true
70: light fluid supports a heavy fluid. In the Rayleigh-Taylor instability,
71: the fastest growing mode has an infinite perturbation wavenumber.
72: However, taking into account a ``uni-directional'' magnetic
73: field along the azimuthal direction in the Galactic disk,
74: \citet{par66} showed that the magnetic tension
75: stabilizes large wavenumber perturbations and results in a preferred,
76: finite wavenumber. But when the
77: perturbations along the radial direction are allowed, those
78: with an infinite radial wavenumber prevail \citep{par67}.
79: As a result, the structures formed by the Parker instability
80: are expected to be elongated, and \citet{kim98} confirmed it
81: through three-dimensional simulations for the nonlinear evolution of
82: the Parker instability.
83:
84: The Parker instability in the interstellar medium (ISM) has been
85: thought to be a viable mechanism in forming giant molecular clouds
86: (GMCs) in the Galaxy \citep[see, \eg][]{app74,mou74,bli80}.
87: However, the work of \citet{kim98} raised negative points on that.
88: In addition to the fact that sheet-like structures with the smallest
89: scale in the radial direction are formed, they found that the
90: enhancement factor of column density is at most $\sim2$. The
91: second density issue is eased by noting that the interstellar gas
92: can be further susceptible to the thermal instability, as pointed
93: in \citet{par00}, followed by the gravitational instability.
94: However, the first structural issue, which is the direct result of
95: the infinitesimal radial wavenumber, does not easily go away.
96:
97: Several ideas on effects that could suppress the maximally unstable
98: nature of the mode with an infinite wavenumber have been suggested.
99: One of them is to invoke a ``stochastic magnetic field'' \citep{par00},
100: which represents the random component of the Galactic magnetic field.
101: Using a field composed of the usual mean component and a transverse
102: component whose strength is weak and random, they showed
103: that in ``cold plasma'' (without gas pressure) the weak, random
104: component exerts a significant stabilizing effect on the perturbations
105: with small transverse wavelengths. The physical mechanism is the
106: following. Although weak, the tension of the transverse component
107: that is incurred by the vertical gas motions
108: is strong enough to reinstate the gas. They suggested the possibility
109: of preferred modes with finite transverse (radial) wavenumbers.
110: Such modes would result in broadened structures, which would resemble
111: more the morphology of the GMCs.
112: Their stochastic field model is promising in the sense that i) it is
113: consistent with the turbulent picture of the ISM
114: \citep[see, \eg][]{ms96} and ii) it is supported by the observations
115: of magnetic field in our Galaxy and spiral galaxies
116: \citep[see, \eg][]{bbmss96,zh97}.
117: However, their cold plasma approximation needs to be improved.
118:
119: The purpose of this paper is to analyze fully the effects of the
120: random component of the magnetic field on the Parker
121: instability in a medium with finite gas pressure. We find rather
122: surprising results that the random component either reduces
123: the growth of the instability significantly or suppresses it completely.
124: And the most unstable mode has a vanishing radial wavenumber.
125: The plan of the paper is as follows. Linear stability
126: analysis is carried out by analyzing the dispersion relation
127: in \S 2. Summary and discussion follow in \S 3.
128:
129: \section{Linear Stability Analysis}
130:
131: We consider the stability of an equilibrium system where gas is supported
132: by its own and magnetic pressures against a ``uniform'' gravity, $g$,
133: in the negative
134: $z$ (vertical) direction. With realistic gravities different growth rates
135: and wavelengths of unstable modes would result \citep[see, \eg][]{khr97},
136: but they make the analysis much more involved.
137: In addition, we expect the qualitative features of the stability
138: wouldn't be affected by details of gravity. For the magnetic field
139: configuration, the stochastic model suggested by \citet{par00} is
140: adopted. It is composed of a mean component, $B(z)$, in the $y$
141: (azimuthal) direction, and a random component, $\epsilon(z)B(z)$, in the $x$
142: (radial) direction. $\epsilon(z)$ is a random function of $z$ with zero mean.
143: One assumption made on $\epsilon(z)$ is that the correlation length
144: is small compared to the vertical scale height of the system.
145: So in the equations below, the local
146: average is taken by integrating over $z$ for a vertical scale
147: greater than the correlation length of $\epsilon(z)$ but smaller than
148: the scale height. Then, the dispersion $\langle\epsilon^2\rangle$ is taken
149: as a constant, which becomes a free parameter of the analysis.
150: With finite gas pressure, $p$, the magnetohydrostatic equilibrium is
151: governed by
152: \begin{equation}
153: \frac{d}{dz}\left[ p + (1+\langle\epsilon^2\rangle)\frac{B^2}{8\pi} \right]
154: = - \rho g,
155: \end{equation}
156: where $\rho$ is gas density. Two further assumptions are made, which are
157: usual in the analysis of the Parker instability: i) an isothermal
158: equation of state, $p = a_s^2 \rho$, where $a_s$ is the isothermal speed,
159: and ii) a constant ratio of magnetic to gas pressures,
160: $\alpha=(1+\langle\epsilon^2\rangle)B^2/(8\pi p)$.
161: Then exponential distributions of density, gas pressure, and magnetic
162: pressure are obtained
163: \begin{equation}
164: \frac{\rho(z)}{\rho(0)} =
165: \frac{p(z)}{p(0)} =
166: \frac{B^2(z)}{B^2(0)} =
167: \exp \left( - \frac{|z|}{H} \right)_,
168: \end{equation}
169: where the e-folding scale height, $H$, is given by $(1+\alpha)a_s^2/g$.
170:
171: The above equilibrium state is disturbed with an infinitesimal perturbation.
172: The perturbed system is assumed to be isothermal too. Since linearized
173: perturbation equations for the case without gas pressure were
174: already derived \citep{par00}, the detailed derivation is not
175: repeated here. Instead, a reduced form in terms of velocity
176: perturbations, $(v_x,v_y,v_z)$, is written down as follows:
177: \begin{eqnarray} \label{velx}
178: \frac{\partial^2 v_x}{\partial t^2} &=&
179: a_s^2 \frac{\partial}{\partial x}
180: \left( \frac{\partial v_x}{\partial x}
181: +\frac{\partial v_y}{\partial y}
182: +\frac{\partial v_z}{\partial z}
183: -\frac{v_z}{H} \right) \nonumber \\
184: &+& v_A^2 \left[ \frac{\partial^2 v_x}{\partial y^2}
185: +\frac{\partial^2 v_x}{\partial x^2}
186: +\frac{\partial}{\partial x}
187: \left( \frac{\partial v_z}{\partial z}
188: -\frac{v_z}{2H} \right)
189: -\frac{\langle\epsilon^2\rangle}{2H}\frac{\partial v_z}{\partial x}
190: \right]_,
191: \end{eqnarray}
192: \begin{eqnarray} \label{vely}
193: \frac{\partial^2 v_y}{\partial t^2} &=&
194: a_s^2 \frac{\partial}{\partial y}
195: \left( \frac{\partial v_x}{\partial x}
196: +\frac{\partial v_y}{\partial y}
197: +\frac{\partial v_z}{\partial z}
198: -\frac{v_z}{H} \right) \nonumber \\
199: &+& v_A^2 \left\{-\frac{1}{2H}\frac{\partial v_z}{\partial y}
200: +\langle\epsilon^2\rangle
201: \left[ \frac{\partial^2 v_y}{\partial x^2}
202: +\frac{\partial^2 v_y}{\partial y^2}
203: +\frac{\partial}{\partial y}
204: \left( \frac{\partial v_z}{\partial z}
205: -\frac{v_z}{2H}
206: \right)
207: \right]
208: \right\}_,
209: \end{eqnarray}
210: \begin{eqnarray} \label{velz}
211: \frac{\partial^2 v_z}{\partial t^2} &=&
212: a_s^2 \frac{\partial}{\partial z}
213: \left( \frac{\partial v_x}{\partial x}
214: +\frac{\partial v_y}{\partial y}
215: +\frac{\partial v_z}{\partial z}
216: -\frac{v_z}{H} \right) \nonumber \\
217: &+& v_A^2 \left\{ \frac{\partial^2 v_z}{\partial y^2}
218: +\left( \frac{\partial}{\partial z}
219: -\frac{1}{H} \right)
220: \left[ \left( \frac{\partial}{\partial z}
221: -\frac{1}{2H} \right) v_z
222: +\frac{\partial v_x}{\partial x}
223: \right] \right. \nonumber \\
224: &&\left. + \frac{1}{2H}
225: \left( \frac{\partial v_x}{\partial x}
226: +\frac{\partial v_y}{\partial y}
227: +\frac{\partial v_z}{\partial z}
228: -\frac{v_z}{H} \right) \right. \nonumber \\
229: &&\left. + \langle\epsilon^2\rangle
230: \left[\left( \frac{\partial}{\partial z}
231: -\frac{1}{H} \right)
232: \left( \frac{\partial}{\partial z}
233: -\frac{1}{2H} \right) v_z
234: +\left( \frac{\partial}{\partial z}
235: -\frac{1}{H} \right)
236: \frac{\partial v_y}{\partial y}
237: +\frac{\partial^2 v_z}{\partial x^2}
238: \right] \right\}_.
239: \end{eqnarray}
240: Here, $v_A$ is the Alfv\'en speed, $B/\sqrt{4\pi\rho}$,
241: which is constant over $z$. Note that the linearized perturbation
242: equations for the cold plasma (Eqs. [12] - [14] in \citet{par00}) are
243: recovered from the above equations by i) dropping out the terms with $a_s$
244: and ii) noting that the scale height of magnetic field ($\Lambda$ of their
245: notation) is twice larger than that of gas ($H$ of our notation).
246:
247: The normal mode solution takes the following form
248: \begin{equation} \label{sol}
249: (v_x,v_y,v_z) = (D_x,D_y,D_z)
250: \exp \left(
251: \frac{t}{\tau} + ik_x x + ik_y y + ik_z z + \frac{z}{2H}
252: \right)_,
253: \end{equation}
254: where $D_x$, $D_y$, and $D_z$ are constants. Taking $H$ and $H/a_s$ as
255: the normalization units of length and time, respectively, the dimensionless
256: growth rate, $\Omega=H/(a_s\tau)$, and the dimensionless wavenumber,
257: $(q_x,q_y,q_z) = H (k_x,k_y,k_z)$, are defined. Substituting
258: Eq. (\ref{sol}) into Eqs. (\ref{velx})-(\ref{velz}) and imposing the
259: condition of a non-trivial solution, we get the dispersion relation
260: \begin{equation} \label{disprel}
261: \Omega^6 + C_4 \Omega^4 + C_2 \Omega^2 + C_0 = 0,
262: \end{equation}
263: where the coefficients $C_4$, $C_2$ and $C_0$ are given by
264: \begin{equation}
265: C_4 = 2\alpha q_y^2 + (2\alpha+1)(q_x^2+q_y^2+q_z^2+1/4)
266: + 2\alpha\langle\epsilon^2\rangle(2q_x^2+q_y^2+q_z^2+iq_z/2),
267: \end{equation}
268: \begin{eqnarray}
269: C_2 &=& \alpha(\alpha+1) \left[ q_x^2 + 4q_y^2(q_x^2+q_y^2+q_z^2)
270: \right] \nonumber \\
271: &+& \alpha\langle\epsilon^2\rangle\left\{\alpha q_x^2 + (2\alpha+1)q_y^2
272: +4q_x^2\left[(2\alpha+1)(q_x^2+q_z^2)+(4\alpha+1)q_y^2 \right]
273: + 4\alpha q_y^2 \left[ 2(q_y^2+q_z^2) + iq_z/2 \right]
274: \right\} \nonumber \\
275: &+& 4\alpha^2\langle\epsilon^2\rangle^2 q_x^2 (q_x^2+q_y^2+q_z^2+iq_z /2),
276: \end{eqnarray}
277: \begin{eqnarray}
278: C_0 &=& 2\alpha^2 q_y^2 \left[ 2q_y^2(q_x^2+q_y^2+q_z^2+1/4)
279: -(\alpha+1)(q_x^2+q_y^2) \right] \nonumber \\
280: &+& 2\alpha^2\langle\epsilon^2\rangle
281: \left\{
282: \left[ (\alpha+1)q_x^4+3(\alpha+1)q_x^2q_y^2 +(2\alpha+1)q_y^4
283: \right]
284: +4q_y^2 (q_x^2+q_y^2+q_z^2) \left[ q_x^2+\alpha(q_x^2+q_y^2)
285: \right]
286: \right\} \nonumber \\
287: &-& \alpha^2\langle\epsilon^2\rangle^2 q_x^2
288: \left\{ q_x^2+2q_y^2 - 4(q_x^2+q_y^2+q_z^2)
289: \left[q_x^2+2\alpha(q_x^2+q_y^2)
290: \right]
291: \right\}.
292: \end{eqnarray}
293: Eq. (\ref{disprel}) is a cubic equation of $\Omega^2$ with complex
294: coefficients. For the case with vanishing vertical wavenumber
295: ($q_z=0$), all the $C$ coefficients become real, and the dispersion
296: relation can be easily solved. For small vertical wavenumbers,
297: the imaginary terms, $i\langle\epsilon^2\rangle q_z$ and
298: $i\langle\epsilon^2\rangle^2 q_z$
299: in $C_4$ and $C_2$, can be still ignored. This trick doesn't affect
300: the marginal condition of the stability ($\Omega=0$),
301: since $C_0$ doesn't contain any imaginary term.
302: Here, we remind readers of the definition of $\alpha$. It is reserved in
303: this paper for the ratio of magnetic to gas pressures, whereas it was used
304: for the dispersion of $\epsilon$ in \citet{par00}. For the dispersion
305: $\langle\epsilon^2\rangle$ is used in this paper.
306:
307: Two limiting cases can be considered, which enable us to check the validity
308: of the above relation. The formula with $\langle\epsilon^2\rangle=0$
309: reduces to the
310: dispersion relation for the original Parker instability
311: \citep[see, \eg][]{par67,Shu74}. The $C$'s without the terms containing
312: $\langle\epsilon^2\rangle$ match exactly with the coefficients of Eq.~(53) in
313: \citep{Shu74}, after imposing the isothermal condition $\gamma=1$.
314: The other limiting formula is for the cold plasma with $p=0$ \citep{par00}.
315: As shown above, our linearized perturbation equations recover those
316: for the cold plasma.
317:
318: The full stability property can be analyzed by solving the above dispersion
319: relation numerically. Fig. 1 shows the stability diagram for $\alpha=1$.
320: Equi-$\Omega^2$ contours with positive values corresponding to unstable modes
321: are plotted on the $(q_x^2,~q_y^2)$ plane for a few different values
322: of $\langle\epsilon^2\rangle^{1/2}$. $q_z=0$ has been set.
323: Finite $q_z$'s reduce the growth rate \citep[see, \eg][]{par66}.
324: Three interesting points can be made: with increasing
325: strength of the random component, i) the domain of the instability in
326: the $(q_x^2,~q_y^2)$ plane shrinks, ii) the maximum growth rate decreases,
327: and iii) the $q_{x,{\rm max}}$, which gives the maximum growth rate, decreases
328: and reduces to zero eventually. Note that without
329: the random component, $\langle\epsilon^2\rangle^{1/2}=0$,
330: the most unstable mode has
331: the growth rate $\Omega^2=0.172$ and the radial wavenumber
332: $q_{x,{\rm max}} \rightarrow \infty$.
333:
334: The above points can be seen more clearly in Fig. 2, which shows
335: the growth rate and two horizontal wavenumbers of the most unstable
336: modes as a function of $\langle\epsilon^2\rangle^{1/2}$ for three
337: different values of $\alpha$. Again, $q_z=0$ has been set.
338: Note that the scale height $H$ changes with $\alpha$.
339: Hence, both the growth rate and wavenumber in real units
340: scale as $1/(1+\alpha)$. Even after this factor is taken into account,
341: the maximum growth rate increases with $\alpha$,
342: due to enhanced magnetic buoyancy. Two additional points can be made:
343: i) the critical value $\langle\epsilon^2\rangle_c^{1/2}$,
344: above which the Parker instability disappears completely,
345: is independent of $\alpha$, and it is computed as $1/\sqrt{2}=0.707$
346: from the dispersion relation, and ii) the value of
347: $\langle\epsilon^2\rangle^{1/2}$, above
348: which the radial wavenumber of the most unstable mode vanishes, decreases
349: with increasing $\alpha$. These are the consequences of different roles
350: of uniform and random magnetic fields.
351:
352: On the issue of the formation of GMCs, the most interesting
353: range of the values would be $\langle\epsilon^2\rangle^{1/2} \sim 0.1$,
354: which would result in $q_x \sim q_y$ for the most unstable modes
355: (although it would depend on $\alpha$, see Fig. 2).
356: Then, the structures formed
357: as the result of the instability would be round, so mimicking GMCs.
358: However, observations suggest a larger random component. It is
359: generally quoted that in the Galactic plane
360: $0.5\la\langle\epsilon^2\rangle^{1/2}\la1$,
361: with the strength of the total magnetic field $B\sim 3-4 \mu{\rm G}$
362: \citep[see, \eg][]{bbmss96,zh97}. But others such as \citet{ms96} suggest
363: somewhat smaller values such as
364: $\langle\epsilon^2\rangle^{1/2} \sim 1/4 - 1/3$.
365: If $0.5 \la \langle\epsilon^2\rangle^{1/2} \la 1$, the instability disappears
366: completely or the growth rate reduces significantly by more than 80\%.
367: If $\langle\epsilon^2\rangle^{1/2} \sim 1/4 - 1/3$, the growth rate reduces by
368: $30 - 60$\%. But in any case the radial wavenumber of the most unstable mode
369: shrinks to zero $(q_x =0)$. {\it That is, the instability basically
370: becomes two-dimensional in the plane defined by the azimuthal and
371: vertical directions.} This result is opposite to that of
372: $\langle\epsilon^2\rangle^{1/2}=0$, where the dominant mode of the instability
373: has vanishing radial wavelength $(q_x \rightarrow \infty)$.
374:
375: \section{Discussion and Summary}
376:
377: The Parker instability is induced by the magnetic buoyancy of uniform
378: component of magnetic field, while gas pressure and random component
379: exert stabilizing effects. The role of gas pressure is mainly
380: exercising pressure force along the uniform component, and set finite
381: wavenumbers for the instability along the uniform component
382: direction. On the other hand,
383: random component threads rising and sinking slices across
384: the uniform field. Its tension becomes stronger at larger wavenumbers.
385: So the role of the random field is to suppress the growth of
386: perturbations with large wavenumbers perpendicular to the uniform field.
387:
388: Through the linear stability analysis which includes both
389: gas pressure and random magnetic field,
390: we have found that with the observationally favored values for
391: the strength of the random component,
392: $0.5 \la \langle\epsilon^2\rangle^{1/2} \la 1$,
393: the tension of the random component that is incurred
394: by the vertical gas motions is strong enough that the growth of the
395: instability is either significantly reduced or completely suppressed.
396: For smaller values, $\langle\epsilon^2\rangle^{1/2} \sim 1/4 - 1/3$, which are
397: suggested by others, the Parker instability is still operating
398: but with reduced growth rate and vanishing radial wavenumber.
399: With $\langle\epsilon^2\rangle^{1/2} \sim 1/4 - 1/3$,
400: by taking $H=160$ pc and $a_s=6.4$ km/s
401: \citep{fl73}, the growth time scale and the azimuthal wavelength of the
402: most unstable mode are $70-95$ Myrs and $\sim 2.2$ kpc, respectively.
403: They are too large for the Parker instability to be a plausible mechanism
404: for the formation of GMCs. But it is known that realistic gravity
405: would reduce both \citep[see, \eg][]{khr97}. The more serious obstacle
406: in the context of the GMC formation is the fact that the radial
407: wavelength of the most unstable mode is infinity. This indicates the
408: structures formed would be elongated, in this case, along the radial
409: direction. But it is not clear whether such elongated structures would
410: persist in the stage of the nonlinear development of the instability.
411: That should be tested by numerical simulations.
412:
413: \acknowledgments
414:
415: The work was supported in part by KRF through grant KRF-2000-015-DS0046.
416: We thank Dr. R. Jokipii for discussions and Dr. T. W. Jones for comments
417: on the manuscript.
418:
419: \begin{thebibliography}{}
420:
421: \bibitem[Appenzeller(1974)]{app74}
422: Appenzeller, I. 1974, \aap, 36, 99
423: \bibitem[Beck \etal(1996)]{bbmss96}
424: Beck, R., Brandenburg, A., Moss, D., Shukurov, A., \& Sokoloff, D.
425: 1996, \araa, 34, 155
426: \bibitem[Blitz \& Shu(1980)]{bli80}
427: Blitz, L., \& Shu, F. H. 1980, \apj, 238, 148
428: \bibitem[Falgarone \& Lequeux(1973)]{fl73}
429: Falgarone, E., \& Lequeux, J. 1973, \aap, 25, 253.
430: \bibitem[Kim \etal(1997)]{khr97}
431: Kim, J., Hong, S. S., \& Ryu, D.,1997, \apj, 485, 228
432: \bibitem[Kim \etal(1998)]{kim98}
433: Kim, J., Hong, S. S., Ryu, D., \& Jones, T. W. 1998, \apj, 506, L139
434: \bibitem[Minter \& Spangler(1996)]{ms96}
435: Minter, A. H., \& Spangler, S. R., 1996, \apj, 194, 214
436: \bibitem[Mouschovias \etal(1974)]{mou74}
437: Mouschovias, T. Ch., Shu, F. H., \& Woodward, P. R. 1974, \aap, 33, 73
438: \bibitem[Parker(1966)]{par66}
439: Parker, E. N. 1966, \apj, 145, 811
440: \bibitem[Parker(1967)]{par67}
441: Parker, E. N. 1967, \apj, 149, 535
442: \bibitem[Parker \& Jopikii(2000)]{par00}
443: Parker, E. N., \& Jokipii, J. R. 2000, \apj, 536, 331
444: \bibitem[Shu(1974)]{Shu74}
445: Shu, F. H. 1974, \aap, 33, 55
446: \bibitem[Zweibel \& Heiles(1997)]{zh97}
447: Zweibel, E. G. \& Heiles, C., 1997, Nature, 385, 131
448:
449: \end{thebibliography}
450:
451: \begin{figure}
452: \epsscale{0.3562}\plotone{f1.eps}\epsscale{1.}
453: \figcaption{Equi-$\Omega^2$ contours on the $(q_x^2,q_y^2)$ plane
454: for (a) $\langle\epsilon^2\rangle^{1/2}$ = 0.05,
455: (b) $\langle\epsilon^2\rangle^{1/2}$ =0.10,
456: (c) $\langle\epsilon^2\rangle^{1/2}$ =0.15.
457: In all plots, $\alpha=1$ and $q_z=0$ are used.\label{fig1}}
458: \end{figure}
459:
460: \begin{figure}
461: \plotone{f2.eps}
462: \figcaption{Maximum growth rate, $\Omega^2_{\rm max}$, and its two
463: horizontal wavenumbers, $q^2_{x,\rm max}$ and $q^2_{y,\rm max}$,
464: as a function of $\langle\epsilon^2\rangle^{1/2}$ for (a) $\alpha$=0.5,
465: (b) $\alpha=1.0$ and (c) $\alpha=1.5$. $q_z=0$ for all the cases.
466: Note that the growth rate and wavenumber in real units
467: scale as $1/(1+\alpha)$.\label{fig2}}
468: \end{figure}
469:
470: \end{document}
471: