astro-ph0211629/ms.tex
1: %\documentclass{aastex}
2: % revised version as of: July, 27th
3: % finally revised on Aug. 15th
4: % and again on August 30th
5: % New Revision as of Nov. 27th 2001
6: % And re-revised on April 7th 2002
7: % And re-revised on April 10th 2002
8: % And re-revised on May 27th 2002
9: % And re-revised see below
10: 
11: %% preprint produces a one-column, single-spaced document:
12: 
13: \documentclass[preprint]{aastex}
14: 
15: %% preprint2 produces a double-column, single-spaced document:
16: 
17: %\documentclass[preprint2]{aastex}
18: 
19: %% If you want to create your own macros, you can do so
20: %% using \newcommand. Your macros should appear before
21: %% the \begin{document} command.
22: %%
23: %% If you are submitting to a journal that translates manuscripts
24: %% into SGML, you need to follow certain guidelines when preparing
25: %% your macros. See the AASTeX v5.0 Author Guide
26: %% for information.
27: 
28: \newcommand{\vdag}{(v)^\dagger}
29: \newcommand{\myemail}{klahr@ucolick.org}
30: %\email{VERSION: 5.0 Tuesday, Sep. 10th 2002}
31: \email{STATUS: in press//
32: ApJ v582 n2 January 10, 2003}
33: 
34: 
35: %% If you wish, you may supply running head information, although
36: %% this information may be modified by the editorial offices.
37: %% The left head contains a list of authors,
38: %% usually a maximum of three (otherwise use et al.).  The right
39: %% head is a modified title of up to roughly 44 characters.  Running heads
40: %% will not print in the manuscript style.
41: 
42: \shorttitle{The Global Baroclinic Instability in Accretion Disks.}
43: \shortauthors{Klahr and Bodenheimer}
44: 
45: 
46: %% This is the end of the preamble.  Indicate the beginning of the
47: %% paper itself with \begin{document}.
48: 
49: \begin{document}
50: 
51: %% LaTeX will automatically break titles if they run longer than
52: %% one line. However, you may use \\ to force a line break if
53: %% you desire.
54: 
55: \title{Turbulence in Accretion Disks.\\ 
56: Vorticity Generation and Angular Momentum Transport\\ 
57: via the Global Baroclinic Instability$^\dagger$}
58: 
59: 
60: %% Use \author, \affil, and the \and command to format
61: %% author and affiliation information.
62: %% Note that \email has replaced the old \authoremail command
63: %% from AASTeX v4.0. You can use \email to mark an email address
64: %% anywhere in the paper, not just in the front matter.
65: %% As in the title, you can use \\ to force line breaks.
66: 
67: \author{H.H.\ Klahr \altaffilmark{1}}
68: \email{klahr@ucolick.org}
69: 
70: \and
71: 
72: \author{P.\ Bodenheimer}
73: 
74: 
75: 
76: \affil{UCO/Lick Observatory, University of California,
77:     Santa Cruz, CA 95064}
78: 
79: %% Notice that each of these authors has alternate affiliations, which
80: %% are identified by the \altaffilmark after each name.  Specify alternate
81: %% affiliation information with \altaffiltext, with one command per each
82: %% affiliation.
83: 
84: \altaffiltext{1}{present address: Computational Physics, Institut f\"ur Astronomie und Astrophysik, Universit\"at T\"ubingen, Germany}
85: 
86: %% Mark off your abstract in the ``abstract'' environment. In the manuscript
87: %% style, abstract will output a Received/Accepted line after the
88: %% title and affiliation information. No date will appear since the author
89: %% does not have this information. The dates will be filled in by the
90: %% editorial office after submission.
91: \newcommand{\tramp}{{\sc Tramp\,\,}}
92: \newcommand{\okappa}{{\tilde{\kappa}}}
93: 
94: \begin{abstract}
95: In this paper we present the global baroclinic instability as a
96: source for vigorous turbulence leading to angular momentum transport 
97: in Keplerian accretion disks.
98: We show by analytical considerations and three-dimensional  radiation hydro
99: simulations that, in particular, protoplanetary disks have a negative radial 
100: entropy gradient, which makes
101: them baroclinic. 
102: Two-dimensional numerical simulations show that 
103: a baroclinic flow is unstable and produces turbulence. These findings are 
104: tested for numerical effects by performing a
105: simulation with a barotropic initial condition which shows that imposed turbulence rapidly decays. 
106: The turbulence in baroclinic disks transports angular momentum outward and  creates a radially  
107: inward bound accretion of matter. Potential energy is released and excess kinetic energy is dissipated.
108: Finally the reheating of the gas supports the radial entropy gradient, forming a self consistent process.
109: We measure accretion rates in our 2D and 3D simulations of $\dot M= - 10^{-9}~{\rm to~}-10^{-7}~\mbox{M}_{\sun}~
110:  \mbox{yr}^{-1}$ 
111: and viscosity parameters of $\alpha = 10^{-4} - 10^{-2}$, which
112: fit perfectly together and agree reasonably with observations.
113: The turbulence creates
114: pressure waves, Rossby waves, and vortices in the ($R-\phi$) plane of the disk.
115: We demonstrate in a global simulation that these vortices tend 
116: to form out of little background noise and to 
117: be long-lasting features, which have already been suggested to lead to 
118: the formation of planets. 
119: \end{abstract}
120: 
121: %% Keywords should appear after the \end{abstract} command. The uncommented
122: %% example has been keyed in ApJ style. See the instructions to authors
123: %% for the journal to which you are submitting your paper to determine
124: %% what keyword punctuation is appropriate.
125: 
126: \keywords{accretion, accretion disks --- circumstellar matter --- hydrodynamics ---
127:  instabilities --- turbulence --- methods: numerical --- solar system: formation ---
128:  planetary systems}
129: \noindent
130: $^\dagger$UCO/Lick Observatory Bulletin, No..........
131: 
132: %% From the front matter, we move on to the body of the paper.
133: %% In the first two sections, notice the use of the natbib \citep
134: %% and \citet commands to identify citations.  The citations are
135: %% tied to the reference list via symbolic KEYs. The KEY corresponds
136: %% to the KEY in the \bibitem in the reference list below. We have
137: %% chosen the first three characters of the first author's name plus
138: %% the last two numeral of the year of publication as our KEY for
139: %% each reference.
140: 
141: \section{Introduction}
142: Protoplanetary disks appear to be a common feature
143: around young stars (Beckwith \& Sargent 1993; Strom, Edwards, 
144: \& Skrutskie 1993). 
145: They are thought to provide the material and the environment
146: for the formation of planets (e.g.\ Lissauer 1993).
147: Thus one needs to know the internal properties of such
148: disks, such as the  density, temperature and turbulence,  in order
149: to estimate the time scales of the formation process. 
150: These quantities are not directly accessible via
151: observation, and so one needs a model for these disks
152: to derive observable quantities like line emission
153: and scattering efficiency for the light from the 
154: central star (Bell, Cassen, Klahr, \& Henning 1997; 
155: Hartmann, Calvet, Gullbring, \& D'Alessio 1998; Bell 1999).
156: 
157:    The basic idea of most models is that there
158: is a process in these disks that transfers angular momentum
159: radially outward, so that mass will flow radially inward (L\"ust 1952).
160: Such a process might be turbulence (hydrodynamical or
161: magneto-hydrodynamical) or self gravity (e.g.\ Larson 1989; Stone, 
162: Gammie, Balbus,  \&  Hawley 2000). Independent
163: of the source for the angular momentum transport, it
164: can be parameterized by an effective viscosity $\nu$, which
165: usually is scaled to  the local sound speed $c_s$ and the
166: pressure scale height $H_D$ of the disk by  a dimensionless
167: number called $\alpha$ (Shakura \& Sunyaev 1973):
168: \begin{equation}
169: \nu = \alpha c_s H_D~.
170: \end{equation}
171: Using this simple description, it was possible to calculate
172: the time evolution of disks (Lynden-Bell \& Pringle 1974; Lin \& Papaloizou
173: 1980),
174: the density and temperature
175: structure, and the turbulent background as well as the laminar 
176: sub Keplerian mean component of the gas flow. The $\alpha$ models
177: have influenced our understanding of the planetary formation process
178: through their implications, for example, regarding the spatial distribution and collision rate of dust 
179: grains (Markiewicz, Mizuno,  \& Voelk 1991; Weidenschilling \& Cuzzi 1993;
180:  Dubrulle, Morfill,  \& Sterzik 1995; Klahr \& Henning 1997), the opening of gaps by giant planets 
181:  (Bryden, Chen, Lin, Nelson, \& Papaloizou 1999;
182: Kley 1999) or the radial drift rate 
183: of planets during their formation phas	e (Goldreich \& Tremaine 1980; Ward 1997).
184: 
185: They are also used to explain the generation 
186: of FU-Orionis outbursts (Bell \& Lin 1994; Kley \& Lin 1999) as well as the measured accretion rates
187: of T-Tauri stars (Hartmann et al.\ 1998).
188: 
189:    Despite the success of these models, $\alpha$  is still
190: a free parameter for protoplanetary accretion disks.
191: A barotropic Keplerian shear flow is in principle stable (Rayleigh stable) 
192: and does not develop turbulence per se despite the high Reynolds numbers.
193: Dubrulle (1993), Richard \& Zahn (1999), and Duschl, Strittmatter, \& Biermann (2000)
194: claim that the high 
195: Reynolds numbers will lead to turbulence in barotropic disks, but
196: numerical investigations so far contradict this idea (Balbus, Hawley,  \& Stone 1996;
197: Godon \& Livio 1999a).  Hydrodynamical turbulence has shown
198: to transport angular momentum outward (Hawley, Balbus, \& Winters 1999; R\"udiger \& Drecker 2001), but it is not able to sustain itself 
199: and usually rapidly decays in barotropic simulations.
200: For a long time thermal convection in the vertical direction was thought to provide
201: a turbulent transport of angular momentum in the outward direction (e.g. Cameron 1978; Lin \&
202: Papaloizou 1985). However analytical investigations of convection 
203:  indicate inward transport of angular momentum (Ryu \& Goodman 1992), a result that was
204: later confirmed by numerical simulations (see below). 
205: For disks around black holes or in cataclysmic variables, 
206: magneto-hydrodynamical instabilities (e.g. magneto-rotational instability)
207: seem to provide a reasonable amount of viscosity (Balbus \& Hawley 1991). 
208: Relatively massive protostellar disks around young stars  ($> 0.1 M_\star$) 
209: also have a proper mechanism to transport
210: angular momentum via self gravity (Toomre 1964; Laughlin \& Bodenheimer 1994).
211: 
212:    In the less massive protoplanetary accretion disks ($\approx 0.01 M_\star$)
213: self gravity is only important at the outermost radii (Bell et al.\ 1997).
214: At the same time the disk is so cool and contaminated
215: by tiny dust grains,  which capture most  of the few free electrons,
216: that magnetic fields have negligible influence on the bulk of
217: the disk matter in the regions where planets are expected 
218: to form (Gammie 1996).
219: The sub-micron sized
220: dust grains reduce the free electron density
221: efficiently enough to 
222: prevent the possibility of the magneto-rotational
223: instability, as the field lines can diffuse faster than
224: the shear can tangle the field.
225: Only in the vicinity of the star, where dust has evaporated
226: and no planets can form, is there  sufficient ionization for 
227: the development of  magnetic turbulence, which can generate 
228:  $\alpha \approx  0.01$ (Stone et al.\ 2000). 
229: In addition, if the optical depth above the disk is sufficiently small,
230: cosmic rays and X-rays (Glassgold, Najita, \& Igea 1997)
231: could ionize the uppermost thin surface layers of 
232: the disk in the hypothetical absence of dust, which has 
233: led some authors to the speculation that between about 0.1 and 30 AU
234: a disk consists of a non-accreting `dead zone'
235: sandwiched between
236: two active layers at the surface (Gammie 1996).
237: 
238:    Indeed, if there is no self gravity working, basically no
239: ionization present and no shear instability possible, the
240: disk would develop no turbulence, transport no angular momentum,
241: release no accretion energy and would basically cool down to 
242: the ambient temperature, while orbiting the star on a time-constant
243: orbit. In such a disk, time scales for the growth of planetesimals
244: would be much longer than in the standard scenarios, as the Brownian
245: motion of the micron-sized dust is smaller and there is  no
246: turbulence acting on the dust in the cm- to m-size range.
247: Sedimentation of the grains to the midplane could create a
248: shear instability (Cuzzi, Dobrovolskis, \& Champney 1993) which then generates
249: turbulence. Some authors (Cuzzi, Dobrovolskis, \& Hogan 1996) argue
250: that the solar nebula at one time must have been turbulent in order
251: to explain the size segregation effects during the formation of chondrites.
252: In any case we want to stress that it is of
253: major relevance for planet formation to know whether protoplanetary disks are
254: turbulent or not. 
255: 
256:    In this paper we present numerical simulations of a purely 
257: hydrodynamical instability that works
258: in accretion disks, namely 
259: a baroclinic instability, similar to the one
260: responsible for turbulent patterns on planets, for example, 
261: Jupiter's red spot and the weather patterns of cyclones 
262: and anti-cyclones on earth. 
263:    Baroclinic instabilities arise in rotating fluids when
264: surfaces of constant density are inclined with respect to the
265: surfaces of constant pressure (e.g.\ Tritton \& Davies 1985).
266: Vortensity, defined  as vorticity per unit surface density,
267:  is not conserved as is the 
268: case in barotropic two-dimensional flows
269: (Kelvin's circulation theorem, see e.g. Pedlosky 1987), and 
270: vortices can be generated.
271: 
272: Kippenhahn \& Thomas (1982) studied 
273: the compatibility of thermal and hydrostatic equilibrium in thin radiative accretion
274: disks but only checked axisymmetric flows. Cabot (1984) investigated the possibility and efficiency
275: of a baroclinic instability for cataclysmic variables (CV) in a local 
276: fashion, concentrating on local vertical fluctuations. He found
277: that this instability  produces insufficient viscosity to explain CVs, 
278: but nevertheless enough viscosity ($\alpha \approx 10^{-2}$)
279:  for protoplanetary disks.
280: Knobloch \& Spruit (1986) also investigated the baroclinic instability
281: due to the vertical stratification in the disk but found that
282: it is not a reliable mechanism for angular momentum transport
283: in thin disks. 
284: Ryu \& Goodman (1992) considered linear growth of non-axisymmetric
285: disturbances in convectively unstable disks. They used the shearing-sheet
286: approximation in a uniform disk and found that the flux of angular
287: momentum was inwards.  Lin, Papaloizou, \& Kley    (1993)
288: performed a linear stability analysis of 
289: non-axisymmetric convective instabilities
290: in disks but allowed for some disk 
291: structure in the radial direction. Their
292: disk model includes a small radial interval in which the 
293: entropy has a small local maximum of 7$\%$ above the background
294: but with a steep drop with an average slope of $K\sim R^{-3}$ ($K$ is the 
295: polytropic constant; see below). 
296: Such a situation could be
297: baroclinically unstable, and in
298: fact that region is found to be associated with outward transport of
299: angular momentum. 
300: Lovelace, Li, Colgate, \& Nelson (1999) and Li, Finn, Lovelace, \& Colgate (2000)
301: investigated
302: the stability of a strong local entropy maximum (a factor 3 above the 
303: background with a slope  
304: $K \sim R^{-11.5}$) 
305: in a thin Keplerian
306: disk and found the situation 
307: to be unstable to the formation of Rossby waves, which transported
308: angular momentum outward and ultimately formed vortices (Li, Colgate, Wendroff, Liska 2001).
309: Sheehan, Davies, Cuzzi, \& Estberg (1999) studied the
310: propagation and generation of Rossby waves in the
311: protoplanetary nebula in great detail, but they had to 
312: assume some turbulence as a prerequisite. 
313: 
314:    In contrast, here we investigate a baroclinic instability
315: that arises from the general
316:  radial (global) stratification of the gas flow in accretion disks.
317: Please keep in mind that global refers to baroclinic and does not necessarily
318: imply that the instability has a global character, i.e.\ depends on
319: the boundary conditions.
320: Our motivation is based on the observation of positive
321: Reynolds stresses in radiation-hydrodynamical 
322: three-dimensional (3D) simulations of thermal convection 
323: in protoplanetary accretion disks (Klahr \& Bodenheimer 2000a). 
324: As convection is known to have  the property of transporting 
325: angular momentum inward (Kley, Papaloizou, \& Lin 1993; Cabot 1996, 
326: hereafter referred to as C96;  Stone \& Balbus 1996, hereafter referred to 
327: as SB96),
328: we were surprised by this result and started an investigation to identify 
329: the special ingredient that would explain this contradictory      
330: result. We found that the size of the simulation domain, especially the 
331: azimuthal extent, influences the sign of the Reynolds stresses 
332: (Klahr \& Bodenheimer 2000a,b). 
333: In order to accomplish a sufficient number of tests to isolate  the 
334: crucial ingredient in our model and to show that artificial boundary effects
335: are not contributing, 
336: we stripped the 3D radiation-hydro simulation
337: down to a flat 2D ($r-\phi$) disk calculation. The 
338: radially varying initial temperature ($\approx$ Entropy $\sim K$) in these calculations approximately reproduces
339: the temperature distribution found in the 3D simulations. These tests
340: show that indeed there is angular momentum transport outwards and that
341: the lower-order azimuthal modes give the fastest growth rates in baroclinic
342: simulations.
343: 
344: In \S 2 we make an argument to show 
345: that protoplanetary accretion disks can be unstable
346: to a non-axisymmetric global baroclinic instability as long as the source term
347: of vorticity (baroclinic term) does not vanish. In \S  3 we explain
348: the changes to the TRAMP code (Klahr, Henning, \& Kley  1999)
349: that became necessary for the  simulations presented here. 
350: Two results from 3D radiation hydro simulations are shown in \S 4.
351: One model 
352: adopts an artificial heating term similar to that in 
353: C96 and SB96, while the other one is self-consistent in the
354: sense that the only possible heating is a pure effect of compression and
355: shock dissipation.
356: In \S 5 we describe tests with the 2D flat 
357: approximation which prove numerically the existence of the 
358: instability and show the strength of the turbulence that 
359: develops as well as the Reynolds stresses that are produced.
360: An initially isothermal (= globally barotropic) calculation which shows decaying turbulence
361: demonstrates the reliability of our numerical investigation
362: and proves that the instabilities observed in baroclinic models
363: are not an effect of the shearing disk boundary conditions, or our numerical scheme.
364: %We briefly compare models of various resolutions and different
365: %sizes of the computational domain.
366: In \S 6 we discuss the first global (2D)  simulation of
367: an accretion disk allowing for the baroclinic instability
368: and find the interesting result that large vortices form. 
369: These vortices are long-lived high-pressure 
370: anti-cyclones with an over-density by a factor of up to four.
371: It has been suggested that such vortices could be 
372:  direct precursors of proto-planets, whose formation could
373: be initiated by concentration of dust toward their centers
374: (Barge \& Sommeria 1995; Tanga, Babiano, Dubrulle, \&  Provenzale 1996; Godon \& Livio 1999b).
375: There is also the possibility  that the over-dense regions could eventually
376: undergo
377: gravitational collapse (see Adams \& Watkins 1995), but our 
378: present code does not allow for this process.
379: In the last section {(\S 7)} we discuss our results.
380: 
381: \section{Stability considerations}
382: In this section we show that a baroclinic instability is plausible in a disk. 
383: The instability can only arise if there is an inclination
384: between the density and 
385: pressure gradients (baroclinic term): 
386: \begin{equation}
387: \nabla \rho \times \nabla p \neq 0.
388: \label{baro_term.ref}
389: \end{equation}
390: Here we analyze the situation using the standard polytropic equation  
391:  
392: \begin{equation}
393: p = K \rho^{\gamma} 
394: \end{equation}
395: where $p$ is the pressure and $\rho$ is the mass density. 
396: We choose a value of $\gamma = 1.43$ which is representative for a mixture
397: of molecular hydrogen and neutral helium with
398: typical abundances for protoplanetary accretion disks.
399: If $K$ is constant as a function of $R$, 
400: as used by various authors (e.g.\ Adams \& Watkins 1995,
401: 1996;
402:  Goldreich, Goodman,  \& Narayan 1986; Nautha \& Toth 1998;
403: Godon \& Livio 1999a,b),
404: the baroclinic term vanishes and vorticity is conserved in plane parallel flows.
405: 
406: In contrast to previous numerical work we
407: choose a radially  varying entropy $K(R)$ (\S 5.2) for the initial state of the 2D simulations.
408: This is the best way to mimic the density
409: and temperature
410: distribution that arises in the 3D radiation-hydro calculation.
411: In other words any realistic disk has a radial entropy ($S$) gradient
412: ($p/{\rho^{\gamma}} \neq {\rm const}$).
413: This can be seen from a simple argument, where we only assume that
414: the aspect ratio of the disk ($H_d/R$) is constant for a range of radii. 
415: The pressure is never only a function of local density but is also
416: a function of local gravity, both of which change with radius: 
417: \begin{equation}
418: p \sim \Sigma H_d \Omega^2 \sim \rho \Omega^2 R^2 \sim \rho R^{-1} 
419: \sim  \Sigma R^{-2}.
420: \end{equation}
421: This result holds for any accretion disk, as long as $H_d/R$ is constant. 
422: The exact dependence of $K$ on $R$
423:  nevertheless depends on the specifics of a given simulation.
424: For example in a thermally convective region a density distribution of
425: $\rho \propto  R^{-1}$ and a 
426: temperature distribution of  $T\propto  R^{-1}$ are  typical. Thus it follows
427: \begin{equation}
428: p \propto  T \rho \propto  K(R) \rho^{\gamma} \,\,\, => \,\,\,
429:  K(R) \propto  R^{-2 + \gamma}.
430: \end{equation}
431: %If we assume $\gamma = 2$,  as in  the shallow water
432: %equation, $K$ would be constant and no vorticity would be generated.
433: 
434: In a more general fashion, if  $\rho \propto  R^{-\beta_\rho}$,
435: $T \propto  R^{-\beta_T}$, and $K \propto  R^{-\beta_K}$ this equation reads:
436: \begin{equation}
437: \beta_K = \beta_T + \beta_\rho \left(1 - \gamma\right).
438: \label{beta_K.ref}
439: \end{equation}
440: It obvious, that there are certain stable
441: profiles. For if $\beta_T = \beta_\rho \left(\gamma - 1\right)$
442: then $\beta_K = 0$ (e.g.\ for $\beta_{\rho} = 1 => \beta_T = 0.43$).
443: But our 3D radiation-hydro calculations do not show
444: this particular  profile (see \S \ 4).
445: Thus  $\beta_K = 0.57$ for the initial state in our baroclinic simulations. 
446: 
447: It follows from equations (\ref{baro_term.ref}) and (\ref{beta_K.ref})
448: that 
449: \begin{equation}
450: \nabla \rho \times \nabla (K(R) \rho ^\gamma) \sim -  \beta_K \left(\partial_\phi \rho\right), 
451: \label{eq:unstable} 
452: \end{equation}
453: which is always non-zero if there is an azimuthal density fluctuation and 
454: a non-zero $\beta_K$.
455: On the other hand, isothermal disks are always barotropic by definition.
456: 
457: Baroclinic instabilities have been widely studied in the context
458: of meteorology and oceanography (e.g.\ Tritton \& Davies 1985, Pedlosky 1987). There are some theoretical models
459: as well as some laboratory experiments which can help to understand
460: the basic mechanism of this instability (Sakai, Iizawa, \& Aramaki 1997).
461: The onset of a baroclinic instability lies in the radial
462: entropy gradient. This entropy gradient can result from the 
463: temperature gradient between the equator and north pole of
464: the earth.
465: It can also be realized in an experiment in which a fluid is
466: contained between two concentric cylinders - one is heated and the other one is cooled down.
467: In the absence of rotation
468: thermal convection will occur in the radial direction.
469: This would lead to one large scale convection cell between
470: equator and north pole (Hadley cell, Hadley 1735) as well as to thermal
471: convection between the cylinders.
472: But the rotation of the earth (or the rotation of the cylinders) 
473: inhibits this convection as it allows only
474: for concentric flows and no convective heat 
475: exchange occurs in the radial direction.
476: But if the entropy gradient is strong enough, 
477: the concentric flow becomes unstable and begins to meander. 
478: This is the baroclinic instability. 
479: As the water meanders, it is once again able to 
480: transport heat radially.
481: The same happens in the earth's  atmosphere, where the baroclinic
482: instability leads to meteorological effects at mid-latitudes
483: like the meandering jet stream and the high and low pressure systems 
484: (anti-cyclones and cyclones),  determining our daily weather.
485: 
486:    Protoplanetary accretion disks are believed to have thermal convection
487: perpendicular to their midplane (Cameron 1978; Lin \&
488: Papaloizou 1985). The disks do not fulfill the Schwarzschild
489: criterion for vertical stability at a wide range of radii. 
490: Also for the radial stratification 
491: a Schwarzschild criterion can be formulated
492: which indicates that if rotational effects were
493: negligible there would be radial thermal
494: convection occurring (e.g.\ convective zones in stars).
495: 
496: The Schwarzschild criterion for stability of the radial stratification 
497: is actually never fulfilled in our baroclinic disks,  as the radial 
498: component of the adiabatic temperature gradient (see Kippenhahn
499: \& Weigert 1990) is
500: \begin{equation}
501: \nabla_{ad} = \frac{P}{T}\frac{\partial T}{\partial P} = 1- \frac{1}{\gamma} = 0.3
502: \end{equation}
503: while the absolute temperature gradient in the radial direction is
504: \begin{equation}
505: \nabla = \frac{P}{T}\frac{dT}{dP} = \frac{\beta_T}{\beta_T + \beta_\rho} = 0.5.
506: \end{equation}
507: Thus $\nabla_{ad} < \nabla$. 
508: Furthermore the Brunt-V\"ais\"al\"a frequency
509: in the radial direction can be calculated via (see Kippenhahn
510: \& Weigert 1990):
511: \begin{equation}
512: N^2(r) =  - \frac{g}{H_d}\left(\nabla_{ad} - \nabla\right).
513: \end{equation}
514: Here $g$ is the  radial component of gravity (not its absolute value as in 
515: Kippenhahn \& Weigert). 
516: At a local radius $R_0$ centrifugal force and gravity may 
517: cancel each other. But if a parcel of gas is radially perturbed
518: in a viscosity free disk, then it conserves its given
519: angular momentum $\Omega_0 R_0^2$. It will therefore feel an
520: effective gravity of
521: \begin{equation}
522: g = \Omega_0^2 \left(R_0^4 R^{-3} - R_0^3 R^{-2}\right),
523: \end{equation}
524: which can be expanded for small deviations $a = R - R_0$ into
525: \begin{equation}
526: g = - \Omega_0^2 a.
527: \end{equation}
528: Radial gravity follows locally  basically the same dependence as
529: the vertical component of gravity.
530: Thus it follows
531: \begin{equation}
532: N^2(r) = - 0.2 \Omega^2,
533: \end{equation}
534: which is obviously negative and hereby convectively unstable.
535: Interestingly, a  positive gradient in entropy also leads to this
536: situation, but in this case the flow would be unstable to perturbations
537: with $a<0$.
538: 
539: Nevertheless, the rotation prevents the disk from being 
540: convectively unstable in the radial direction,  which can
541: be seen from the baroclinic version of the Rayleigh criterion
542: (e.g.\ Solberg-H{\o}iland criterion)
543: \begin{equation}
544: \frac{1}{R^3}\frac{\partial R^4 \Omega^2}{\partial R} - g\frac{\partial ln P}
545: {\partial R}\left(\nabla - \nabla_{ad}\right) > 0
546: \end{equation}
547: which in the case of a Keplerian protoplanetary accretion disk is:
548: \begin{equation}
549: \Omega^2 - N^2 H_D \frac{1}{P}\frac{\partial P}{\partial R}>0.
550: \end{equation}
551: With $p \sim \rho T \sim R^{-2}$ it follows
552: \begin{equation}
553: \Omega^2 \left( 1 - 0.4 \frac{H_D}{R}\right) > 0,
554: \end{equation}
555: which gives a value
556: of about $0.96 \Omega^2$, which is a sufficient condition for stability
557: if only axisymmetric perturbations are permitted. If non-axisymmetric
558: perturbations are allowed it is only a necessary condition for stability.
559: Such a situation is a typical onset to the formation of non-axisymmetric
560: baroclinic instabilities (meanders, vortices) 
561: as we discussed before (see Tritton \& Davies 1985).
562: 
563: A detailed linear or non-linear,  possibly local,  stability 
564: analysis of the global baroclinic
565: instability in the context of accretion disks 
566: still has to be performed to get a sufficient criterion 
567: on the instability.  Such an analysis should not be confused with the
568: global stability analysis of a local baroclinic instability
569: as in Lovelace, Li, Colgate \& Nelson (1999) and Li, Finn, Lovelace, \& Colgate (2000).
570: 
571: 
572: \section{Changes in TRAMP}
573: The code {\tramp} ({\bf T}hree--dimensional 
574: {\bf RA}diation--hydro\-dy\-na\-mic\-al {\bf M}odelling {\bf P}roject)
575: was introduced and extensively discussed by Klahr et al.\ (1999).
576: The equations used here are the same as in that paper.
577: However, two additional features have been added to the code, which
578: are crucial for the work presented in this paper. 
579: \subsection{Shearing Disk}
580: As already mentioned in Klahr et al.\ (1999), rigid wall boundaries 
581: in the radial direction can affect the flow-pattern and are not the
582: perfect choice for simulations of disks. SB96
583: use the ZEUS code described in Stone \&  Norman (1992) in the
584: shearing sheet approximation (Hawley et al.\ 1995).
585: This approximation is only possible for pseudo-Cartesian coordinate
586: systems which themselves strongly limit the scale of the computational
587: domain, i.e.\ the box size has to be small in comparison to the local
588: radius. A direct consequence of the pseudo-Cartesian coordinates is
589: that there can be no global radial gradient of the density, temperature
590: and entropy. Thus, local studies of the global baroclinic instability are
591: impossible.
592: 
593: In order to overcome this problem, \tramp calculates on a
594: spherical ($R, \theta, \phi$) grid; thus large portions of the disk can be simulated, 
595: and one is only restricted by the computational effort required. 
596: We put forward a new set of radial boundary conditions that we
597: call {\em shearing disk} boundaries, which in fact correspond to 
598: a shearing sheet for spherical coordinates and disks.
599: In the shearing sheet approximation one uses simple periodic boundary
600: conditions, i.e.\ the values for the ghost-cells are determined 
601: from the value of cells on the opposite side of the computational
602: domain in a simple fashion [e.g. for the density $\rho(0)=\rho(NX-1)$, where
603: $NX$ is the  maximum number of zones in a given direction]. 
604: In the {\em shearing disk} approximation one makes two assumptions: first, 
605: that the mean values
606: of each quantity follow a power law scaling 
607: with the radius $R$, which is usually
608: the case in steady-state accretion disks for a certain range of
609: radii, and second,  that the fluctuations
610: are proportional to the mean value. Thus it follows
611: for the two inner and two outer ghost cells:
612: \begin{eqnarray}
613: \rho(0,J,K)&=&\rho^*(NX-1,J,K)(R(0)/R(NX-1))^{-\beta_\rho}, \\
614: \rho(1,J,K)&=&\rho^*(NX,J,K)(R(1)/R(NX))^{-\beta_\rho}, \\
615: \rho(NX+1,J,K)&=&\rho^*(2,J,K)(R(NX+1)/R(2))^{-\beta_\rho}, \\
616: \rho(NX+2,J,K)&=&\rho^*(3,J,K)(R(NX+2)/R(3))^{-\beta_\rho}.\label{shearbou}
617: \end{eqnarray}
618: The ($*$) indicates the implementation of the global shear in a manner similar to that given in SB96 and C96.
619: With the assumption that the inner and outer boundaries each move with the 
620: local Keplerian speed, it follows that there is a mean angular offset 
621: between inner and outer grid at time $t$ of 
622: $\delta \phi = (\Omega_{R(1)}+\Omega_{R(2)} - \Omega_{R(NX)} - \Omega_{R(NX+1)}) t/2$.
623: This offset translates into an integer offset $dK = ABS(\delta \phi / d\phi)$ 
624: defined via the angular width of a grid cell $d\phi$ 
625: and two interpolation factors 
626: $C_1 = mod(\delta \phi, d\phi)/d\phi$ and $C_2 = 1 - C_1$:
627: \begin{equation}
628:    \rho^*(K) = C_2 \rho(K+dK ) + C_1 \rho(K + dK - 1).
629: \end{equation}
630:  
631: The value for $\beta_\rho$ is $1$ in all 3D simulations  as we are 
632: performing calculations around $R=5AU$ where the surface density
633: is almost constant ($\beta_\Sigma = 0$). For the azimuthal frequency ($g = \dot\phi = \omega$) $\beta_\omega = 1.5$
634: is obvious. $\beta_{T} = 1.0$ in the 3D simulations for the temperature follows from 
635: the assumption of a constant relative pressure scale height
636:  $H_p/R={\rm const}$, 
637: which is necessary to make inner and outer 
638: vertical stratification fit in polar coordinates. 
639: The 2D simulations use $\beta_{T} = 0.0$ in the isentropic case (model 2)
640: but $\beta_{T} = 1.0$ in the baroclinic unstable case (models 3,4 and 5).
641: Model 6 uses open boundaries rather than shearing-disk conditions. 
642: Finally $\beta_g = 1.5$ for the polar velocity component ($g = \dot\theta$) assumes the 
643: fluctuations to be proportional to the local speed of sound.
644: For the radial velocity $u_r$, $\beta_u = -2 - \beta_\rho$ is 
645: chosen to conserve the mass flux
646: over the radial boundaries in a spherical coordinate system where 
647: $\rho u_r R^2 = {\rm const}$. It was tested that kinetic
648: energy flux  and momentum flux are not artificially amplified
649: by the boundary condition (see model 2). We damp the radial velocity 
650: component of the inner and outer four
651: grid cells by a factor of $f_d = 0.05$ each
652: time step ($u_r := u_r (1.0 - f_d)$) in order to prevent the model from creating artificial
653: resonant oscillations in the radial direction.
654: These radial waves can even occur in 1D radial models if there
655: is no damping. Model 2 (see below) is our basic test that the treatment
656: of our boundary conditions does not lead to a numerical instability. 
657: Angular momentum is not conserved, which allows the possibility of a
658: net transport of angular momentum into or out of the computational domain by turbulence, 
659: and thus of driving an accretion process or suppressing it, depending 
660: on the direction of the transport.
661: 
662: 
663: This new method was extensively tested
664: and will be the subject of an upcoming paper where its properties will
665: be compared to other choices of boundary conditions under
666: various physical disk situations.% (Klahr \& Bodenheimer 2000b).
667:  
668: Using this new numerical method we combine the virtues of local 
669: simulations (good resolution and moderately large time-steps) with
670: the possibility of treating global systematic gradients, previously 
671: exclusively reserved for global simulations.
672: 
673: %{\em \bf
674: %We also perform simulations (see model B) that uses radially
675: %open outflow boundaries with 
676: %\begin{eqnarray}
677: %\rho(0,J,K)=\rho(1,J,K)&=&\rho(2,J,K), \\
678: %T(0,J,K)=T(1,J,K)&=&T(2,J,K), \\
679: %u_r(0,J,K)=u_r(1,J,K)&=&min(0,u(2,J,K)), \\
680: %g(0,J,K)=g(1,J,K)&=&g(2,J,K)), \\
681: %         h(1,J,K)&=&\Omega_{Kepler}(R(1)), \\
682: %         h(0,J,K)&=&\Omega_{Kepler}(R(0)).\label{openbou}
683: %\end{eqnarray}
684: %The values for the outer ghostcells are determined in the same
685: %5fashion. These boundaries alow us to test the results
686: %%of model A, which used the shearing disk boundaries.
687: %5The drawback of these boundaries is, that mass is lost
688: %5in radial direction and the physical problem changes
689: %5from a disk simulation to the simulation of a torus
690: %%with physical edges.}
691: 
692: \subsection{Reynolds Averaging}
693: As an addition to the earlier \tramp version there is now the possibility
694: to measure the transport properties of the turbulence, i.e.\ the
695: correlations of the fluctuations especially for the radial transport
696: of angular momentum. A similar method was already used in C96
697: as well as in SB96. This turbulent stress tensor $T$
698: is then scaled by the square of the sound-speed,  representing the
699: pressure at the midplane. The $T_{r\phi}$ component corresponds
700: to the angular momentum transport which corresponds to the classical
701: $\alpha$ value, or in other words,
702: an $\alpha$ is a measure of how much viscosity would be needed in order to
703: simulate the turbulent angular momentum transport by viscous 
704: diffusion of angular momentum in a laminar flow.
705: As we also investigate other components of the stress
706: tensor we define:
707: \begin{equation}
708: \alpha = A_{r\phi} = \frac{<u_\phi' (\rho u_r)'>}{\bar\rho c_s^2}
709: \label{alpha.ref}
710: \end{equation}
711: with
712: \begin{equation}
713: <u_\phi' (\rho u_r)'> = <\rho u_\phi u_r> - \bar u_\phi <\rho u_r>,
714: \end{equation}
715: where $< >$ or a bar represents a spatial and time average.
716: We use the symbol $A_{ij}$ because
717: the symbol $\alpha_{r\phi}$ is generally used  for
718: the helicity of the velocity field.
719: 
720: A detailed discussion of the measurements of the Reynolds stresses
721: including $A_{r\theta}$ and $A_{\theta\phi}$
722: for compressible turbulence shall be discussed in an upcoming paper
723: (Klahr \& R\"udiger in preparation).
724: In a similar manner we measure the strength and isotropy of the 
725: turbulence via the determination of the on-diagonal terms of
726: the stress tensor $T_{rr}$, $T_{\theta\theta}$ and $T_{\phi\phi}$ which, scaled
727: by the square of the sound speed,  are three different indicators of the 
728: square of the turbulent Mach number: 
729: \begin{equation}
730: A_{rr} = \frac{<{u_r'}^{2}>}{{c_s}^2},
731: \label{alpha_rr.ref}
732: \end{equation}
733: \begin{equation}
734: A_{\theta\theta} = \frac{<{u_\theta'}^{2}>}{{c_s}^2}
735: \label{alpha_tt.ref}
736: \end{equation}
737: and 
738: \begin{equation}
739: A_{\phi\phi} = \frac{<{u_\phi'}^{2}>}{c_s^2}.
740: \label{alpha_pp.ref}
741: \end{equation}
742: In the same way we determine the mean density fluctuation in  the
743: flow by $\frac{<{\rho'}^{2}>}{\bar\rho^2}.$
744: We define $M = \sqrt{A_{rr}+A_{\theta\theta}+A_{\phi\phi}}$ to be
745: the Mach number in our models. 
746: 
747: \section{3D Radiation-Hydrodynamical Simulations}
748: The inviscid 3D simulations of thermal convection in disks 
749: (Klahr \& Bodenheimer 2000)
750: showed radially outward directed angular 
751: momentum transport if the section of the 
752: disk was large enough. In the following we
753: show that the angular momentum transport is not 
754: dominated by the vertical
755: thermal convection but by the hydrodynamical
756: turbulence in the $R-\phi$ plane of the disk
757: which results from the radial entropy gradient
758: self-consistently introduced by solving the
759: energy equation in our radiation hydrodynamical
760: code.
761: 
762: As a continuation of our previous work (Klahr et al.\ 1999),  we have 
763: performed  simulations of 3D chunks of protoplanetary
764: accretion disks. We have tested  the influence of viscosity,
765: artificial heating and boundary conditions extensively.
766: As highly viscous flows tend to become axisymmetric
767: (C96; Klahr et al.\ 1999), we were especially interested
768: in simulations with high Reynolds numbers as in SB96.
769: In model 1 we switch off the application of the viscous 
770: forces in the momentum equation but keep the heating, corresponding
771: to  $\alpha = 0.01$,  from the local 
772: viscous dissipation function ($\Phi$) in the energy equation
773: \begin{equation}
774: \Phi = \left(\nabla{{\bf T}}\right){\vec{u}},
775: \label{Eq_Dissi}
776: \end{equation}
777: where ${\bf T}$ is the viscous stress tensor.
778: For the viscosity $\nu$ we adopt $\nu = \alpha c_s^2 / \Omega_{Kepler}$.
779: For more details we refer to Klahr et al.\ (1999). 
780: In an additional simulation (model 1b) we
781: forgo the artificial heating and show that a self-consistent heating
782: of the disk via the dissipation of compression waves,  especially shocks,
783: and the release of gravitational energy 
784: can lead to the same baroclinic effects as the artificial heating in model 1.
785: Artificial heating simplifies the numerical effort, as it stabilises the
786: vertical and radial structure of the disk and thus allows for longer integration times for
787: the mean values (see Table 1).
788: 
789: 
790: Artificial von Neumann \& Richtmyer (1950) viscosity (see also Stone \& Norman 1992) for the proper treatment shocks is included, as
791: is flux-limited radiative transport with a simple dust opacity  ($\kappa =   2 \times
792: 10^{-4} T^2$ cm$^2$ g$^{-1}$). 
793: The goal was to investigate the influence of our more global 
794: approach, especially a wider azimuthal range, on the simulation results.
795: 
796: The particular models we present here 
797: (model 1 \& 1B; see Table  1) cover a radial range from
798: $3.5$  AU to $6.5$ AU, a vertical opening angle of $\pm 7\arcdeg$, and 
799: an azimuthal extent of $90\arcdeg$. They use only 20 grid cells in
800: the vertical direction but 51 in radius and 60 in azimuth.
801: A follow-up paper will deal entirely with 3D models at various
802: resolutions. At the moment it is sufficient to say that the results
803: of model 1 are typical and remain the same at much higher resolution.
804: The benefit of model 1 is that it  can easily
805: be evolved for 68 orbits at the outer edge of the simulation
806: within sixteen CPU hours on a Cray T90. In future papers
807: we will discuss the influence of artificial heating in contrast
808: to self-consistent heating due to shock dissipation in greater detail. 
809: Calculations as model 1B indicate at least that numerical dissipation 
810: might be a problem as it removes kinetic energy from
811: the system without
812: heating the gas. Thus there is less heating
813: to maintain the initial vertical and radial temperature structure of
814: the disk.
815: Higher resolution models will have to investigate the
816: effect of numerical dissipation. 
817: 
818:    Figure \ref{fig1.ref} shows the three-dimensional rendering of the
819: vertical mass flux distribution after model 1 approached 
820: a quasi-steady state. The azimuthally extended convection cells
821: are slightly twisted and not completely axisymmetric.
822: In Figure \ref{fig2.ref} the normalised Reynolds 
823: stresses and corresponding Mach numbers of the turbulence 
824: for model 1 are given, time-averaged
825: over 68 outer orbits.
826: The radial mean value for the angular momentum transport
827: corresponds to $\alpha = A_{r\phi} = 2 \times 10^{-3}$, 
828: a quite high value, and the tendency towards positive values is obvious.
829: The radially  varying Reynolds stresses reflect their
830: connection to locally and temporarily confined flow structures
831: and events. Only a much longer time for averaging,  as is done 
832: in the 2D simulations, can produce a more homogeneous distribution.
833: Based on the measured stresses one can estimate a viscous time-scale
834: of $\approx \alpha^{-1} = 500$ orbits, while for technical
835: reasons only 68 are possible in one run.
836: The on-diagonal terms of $A$, which characterise the strength and
837: isotropy of the turbulence, show that the fluctuations in radial and 
838: azimuthal velocity are
839: almost sonic at some radii,  and that the fluctuations in 
840: the vertical (polar)  velocities are more than an order of magnitude
841: smaller.  This difference reflects the fact that (1) the radial and azimuthal
842: velocity fluctuations are not driven by the vertical thermal convection, and (2)
843: there is no isotropic (3-D) turbulence present.
844:    The resulting
845: Mach number of the turbulence, the combination of the
846: strengths  in three dimensions,  is close to one. The relative density
847: fluctuations are in the mean as strong as $10\%$.
848: 
849: 
850: 
851: In contrast to this result, with positive values of the Reynolds
852: stress, our corresponding  axisymmetric 
853: 2D simulations, with the same physical assumptions and parameters as in 
854: model 1, always showed clearly negative 
855: Reynolds stresses.
856: The dissipation-free axisymmetric Euler equation would
857: predict precisely zero average stresses,  which is a consequence
858: of strict angular momentum conservation, but radiation hydro
859: calculations are by nature not dissipation free even if one has
860: no viscous forces. Thus the negative $\alpha$ values are
861: a result of the heating and cooling of the gas.
862: 
863: Looking at the flow-pattern and density distribution in the
864: $R-\phi$-plane (Fig.\ \ref{fig3.ref}), 
865: we do not see the convection cells from Figure 
866: \ref{fig1.ref}. 
867: What we see are vortices, vorticity waves (e.g.\ Rossby waves, baroclinic waves), 
868: and pressure waves propagating in the flow. 
869: These are the sources of the positive
870: Reynolds stress. They are counteracting the negative Reynolds stresses
871: generated by the thermal convection. In the turbulence terminology
872: one would possibly identify geostrophic turbulence (irregular waves; 
873: see e.g.\ Tritton \& Davies 1985). This can be seen in
874: the spectral density distribution of the flow in the azimuthal
875: direction (Fig.\ \ref{fig4.ref}). Only at the
876: smallest resolved scales ({\em i.e.\ meso scale}) the slope approaches 
877: the Kolmogorov spectrum ($k^{-5/3}$).
878: At larger scales the slope is even steeper than for earth atmospheric 
879: geostrophic turbulence ($k^{-3}$) which indicates how strong small-scale
880: turbulence is being pumped into  the smallest possible wave numbers.
881: In this calculation the minimal $k$ is four as we calculate only
882: a quarter of the disk. However 360$\arcdeg$ full circle simulations (see model 6) 
883:  indicate that
884: $k=1$ is the mode where the energy will pile up. The dissipation in
885: this calculation thus does not only occur at the smallest scales
886: as in 3D incompressible turbulence
887: but at all scales, especially at the large scales where shocks form.
888: The idea to drive the accretion process via shocks was already
889: suggested by Spruit (1987), but in his barotropic models the
890: shocks could not form without the presence of an external perturber.
891: 
892: The Mach number of the flow is the smallest in the midplane 
893: (compare  $M_{max} = 0.6$ in Fig.\ \ref{fig3.ref}  with 
894: $M=0.5$ in Fig.\ \ref{fig2.ref}). Velocities rise,  
895: eventually beyond the sound speed, with height above the midplane.
896: 
897: From this result several questions arise.
898: (A): Why did neither Stone \& Balbus (1996) 
899: nor Balbus, Hawley \& Stone (1996)
900: observe these positive Reynolds stresses
901: and violent turbulence?
902: (B): What is the source for the vorticity and
903: turbulence in the $R-\phi$ direction?
904: (C): Especially, how can we 
905: be sure that we are not observing  boundary effects?
906: We will explicitly answer these questions in \S 7.
907: 
908: The strategy in order to identify the source for
909: turbulence and angular momentum transport was to
910: simplify our model step by step by removing
911: physics from the simulation and to wait for the turbulence 
912: and Reynolds stresses to disappear. The first effect we
913: found was that whenever we decreased the azimuthal extent
914: of the computational domain below a critical value of
915: $\approx 15$ degrees, the Reynolds stresses switched to negative values, as 
916: in 2D axisymmetric calculations (Klahr \& Bodenheimer 2000b). 
917: As the SB96 calculation only covers roughly 1 degree,  this was a first
918: hint that our findings might result from the more global
919: simulation.
920: 
921: \subsection{No artificial heating}
922:    The artificial heating in model 1  has the benefit that
923: there is a well defined laminar equilibrium state for
924: the disk. This is ideal for any stability 
925: investigation as well as for the setup of a numerical model.
926: 
927:    But one can argue that this artificial heating is 
928: inconsistent with the lack of dissipation of kinetic energy.
929: Consequently, the findings with model 1 could result purely 
930: from the artificial heating, or,  even worse,  from the specific
931: shape of the dissipation function (see Eq.\ \ref{Eq_Dissi}).
932: Thus, we repeated the simulation of model 1, but switched
933: off the heating completely. This simulation is numerically more complicated 
934: and more expensive than model 1 as the initial state is 
935: far from equilibrium and the disk undergoes strong fluctuations
936: in density and temperature. Only 14 orbits could be performed during a
937: 10-hour run on the Cray T90. 
938: The disk vertically shrinks by 20 percent
939: before it reaches a new self-consistent state with a 
940: midplane temperature of 40 K, down from  150 K  in the initial state.
941: 
942: If the temperature changes by a factor of almost 4 (150 K $\rightarrow$ 40K) 
943: one could expect the disk height ($\sim$ pressure scale height ) 
944: to change by about a factor of 2.  
945: But the factor of 2 in scale height is only true for vertically isothermal disks.
946: Our disks have a vertical temperature profile that becomes flatter as the temperatures
947: decrease because  the opacities also  decrease (proportional to $\kappa
948: \propto T^2$). This means that the high-temperature disk has
949: temperature, pressure and density profiles that fall off much more steeply  than the pressure
950: scale height estimated for the midplane temperature might suggest.
951: The cooler disk on the other hand is closer to an isothermal disk and
952: thus does not shrink in proportion to the root of the midplane temperature.
953: 
954: 
955: The vertical grid structure was readjusted during this cooling:
956: the initial vertical opening angle was $\pm 7\arcdeg$, but 
957: during the vertical shrinking the gradients above the disk became
958: too  steep for the numerical technique to work. Thus we constructed a new vertical 
959: grid  $\pm 5.5\arcdeg$ also with  20 grid points and interpolated the
960: values linearly from the old onto the new grid.
961: 
962: The newly gained state of 40 K was then maintained for
963: more than 100 orbits (see Fig.\ \ref{fig5.ref}, which is  similar to 
964: Fig.\ \ref{fig3.ref}). 
965: The disk turbulence
966: was transonic (see Fig.\ \ref{fig6.ref}) and the heating 
967: was highly non-homogeneous in space and time, in contrast to 
968: model 1 where it was assumed to be smoothly distributed.
969:  But nevertheless the general result was the same:
970: the disk develops a radial entropy gradient, turbulence and
971: vorticity. In addition, the accretion of matter which results
972: from the positive Reynolds stresses feeds energy into the system
973: which is radiated away after the turbulence is dissipated in shocks.
974: 
975:    These self-consistent simulations will have to be continued in 
976: our future work. The simulations are numerically much more
977: unstable than the heated or polytropic models. But in the end
978: only they can provide realistic predictions for density, temperature, 
979: turbulence,  and the evolution of accretion disks.
980: 
981: \subsection{Radial entropy gradient}
982:    Another glance at Figure \ref{fig3.ref} 
983: shows that temperature (contour-lines)
984: and density (colors) are not perfectly aligned. Thus also pressure
985: and density gradients do not point in the same direction, a clear
986: indication of a baroclinic flow, where the entropy decreases with
987: radius. To test the relevance of
988: this possible instability for the generation of the observed turbulence 
989: we removed the vertical structure of the disk and got rid
990: of the radiation transport. 
991: Thus the 3D density $\rho$ was replaced by the surface density $\Sigma$
992: and the 3D pressure $p$ by the vertically integrated pressure $P = \Sigma T$
993: (in dimensionless units).
994: In Figure \ref {fig7.ref} 
995: we plot the $K \sim T \rho ^ {1-\gamma}$ values in the midplane (before
996: model 1 became turbulent) as $K/K_{max}$. If one measures
997: this $K$ value in the turbulent state of the disk,
998: the fluctuations of $K$ are too strong, which hides the general
999: trend. The next section describes the results 
1000: of using such an entropy distribution in a two-dimensional ($R,\phi$) disk. 
1001: 
1002: \section{2D Simulations}
1003: The flow characteristics in the 3D simulations, e.g.\ the turbulence
1004: cascade and the small vertical velocity fluctuation, indicated
1005: that the disk instability must be a 2D effect. In order to verify
1006: this thesis we removed the vertical structure and performed
1007: 2D $(R-\phi)$ simulations.
1008: 
1009: Both of the following models were first developed as 1D radial
1010: axisymmetric models with density slope $\rho \sim R^{-1}$
1011: ($\rho[1AU] = 10^{-10}$ g cm$^{-3}$), 
1012: which corresponds, in the 2-D flat disk, to $\Sigma= $ const ($\approx 300$ g cm$^{-2}$). 
1013: The initial temperature was then chosen to be either constant 
1014:               ($T = T_0$) as a function of radius  or to follow a power law ($T = T_0
1015:               \left (R/R_0\right)^{-1}$). The value for $T_0$ was
1016:               adjusted to create a local pressure scale height of $H/R = 0.055$
1017:               at the mid-radius $R_0=5$AU, to match that in the 3D radiation 
1018:               hydro calculations. For obvious reasons this pressure scale height 
1019:               condition is only fulfilled at $R_0$ for the constant temperature
1020:               case but for all radii in the $T \propto R^{-1}$ case.
1021: In the constant temperature  model 
1022: $H/R$ varies as $R^{0.215}$, which
1023: means that the relative pressure scale height increases slightly
1024: with radius.  Both models use 
1025: an identical computational domain with radii between 
1026: 4.0 and  6.0 AU and an azimuthal extent of 
1027: $30\arcdeg$, with a moderate resolution at $64^2$ grid cells. The boundary
1028: conditions are identical to those of model 1: a shearing disk plus  damping of the
1029: radial velocities in 4 grid cells 
1030: near the inner and outer boundaries. 
1031: The artificial viscosity is the same as in model 1.    
1032: Both models are initially kicked
1033: by a random density perturbation with 1$\%$ amplitude: $\rho := \rho * (0.99 + 0.02 * RAND)$, where $RAND$ is
1034: a uniformly distributed random number between 0 and 1.
1035: Afterwards the system was allowed to evolve
1036: freely.
1037:    We advect the internal energy and calculate the $PdV$ work arising from
1038: compression and shock dissipation in the same manner as in the 3D calculations.
1039: For the time being there is no radiative cooling, because cooling would
1040: immediately change the temperature profile and we are interested in the
1041: behavior of given temperature distributions. Models with radiative
1042: cooling shall be dealt with in future work.
1043: 
1044: \subsection{Model 2: radially constant entropy: ($T = const$)}
1045: This model developed no turbulence. The initially induced turbulence 
1046: decayed rapidly with an e-folding time for the kinetic energy in $R$ of 
1047: $\approx 4.3$ orbits (see Figs.\ \ref{fig8.ref} and
1048:  \ref{fig9.ref}).
1049: Nevertheless, the initially introduced turbulence 
1050: always transported angular momentum outward, never inward (see Model 2B).
1051: Figure \ \ref{fig8.ref} shows the surface density distribution and flow field after about 90 orbits.
1052: The surface density is almost constant ($299.955<\Sigma <300.125 $ g cm$^{-2}$). The arrows indicate that there
1053: is still small noise in the velocity field with a Mach number of about  $10^{-4}$. In Figure 
1054: \ \ref{fig9.ref} the evolution of the
1055: total kinetic energy in both coordinates, and the enstrophy of the flow are plotted, in units
1056: of the initial values. 
1057: The overall mean mass accretion rate is given in solar masses per year. 
1058: Here and in the following models, a negative value of \.M indicates
1059: accretion towards the star. 
1060: Enstrophy is the integral square
1061: of the local vorticity. One clearly sees that the kinetic energy  as well as
1062: enstrophy are 
1063: decaying  as is expected for
1064: a barotropic disk. Kinetic energy in $\phi$ and enstrophy are calculated 
1065: from the deviations from the Keplerian profile.
1066: The residual finite values of these quantities
1067: result from the laminar (not strictly Keplerian) steady state flow.
1068: The resulting mean turbulence, measured by the strength 
1069: of the components of the stress tensor (see Fig.\ {\ref{fig10.ref}}) is also very low.
1070: 
1071: We also tested the effect of stronger initial perturbations.
1072: For model 2B we gave model 2 an initial random density fluctuation 
1073: of $\pm  50 \%$.
1074: Still the turbulence rapidly decayed as expected. But we have to
1075: stress that the turbulence in this barotropic simulation
1076: had already the property of effectively transporting angular momentum outward.
1077: The Reynolds stresses during the first orbit (see Fig.\ {\ref{fig11.ref}}) are positive ($<\alpha = A_{r\phi}> \approx  3.0 \times 10^{-4}$), nevertheless rapidly decreasing afterwards.
1078: 
1079: We conclude that positive Reynolds stresses
1080: are a common feature of hydrodynamical turbulence in disks,
1081: and not exclusively a result of baroclinic simulations.
1082: %This confirms the findings of Drecker \& R\"udiger (2000).
1083: This is also a result of Hawley et al.\ (1999),  which shows 
1084: that positive Reynolds stresses are correlated to the decay
1085: of turbulence.
1086: Only thermal convection, which must not be confused with
1087: isotropic turbulence, has a tendency of transporting angular
1088: momentum inward. Nevertheless, in barotropic disks hydrodynamical
1089: turbulence can not sustain itself and decays.
1090: 
1091: 
1092: \subsection{Model 3:  radially  varying entropy: ($T \sim R^{-1}$)}
1093: Using a different initial state leads to a completely different result.
1094: The flow becomes turbulent within a few orbits, with kinetic energy rising
1095: exponentially until saturation due to shock dissipation occurs.
1096: Model 3 (Fig.\ \ref{fig12.ref}) shows some similarities to 
1097: model 1 (Fig.\ \ref{fig3.ref}) even though we are
1098: comparing a 3D and a 2D simulation. After 100 orbits the 
1099: surface density shows deviations from
1100: axisymmetry ($286<\Sigma <318$  g cm$^{-2}$).
1101: %  The 
1102: %pressure gradients resulting from the concentration are balanced by the deviations
1103: %of the azimuthal velocity from the Keplerian profile; the density concentration is
1104: %a result of the tendency for local regions of the disk to approach constant 
1105: %angular momentum with radius. 
1106: %These over-dense azimuthal streams are the disk equivalent to Jupiter's
1107: %zonal winds and the jet stream on earth.
1108: %even though we started with the same small perturbation as in the case of model 2.
1109: %The flow-pattern also shows some vortices. 
1110: The velocities are $10^2$ times stronger than
1111: the ones in model 2. 
1112: 
1113: The initial instability grows quickly  (Fig.\ \ref{fig13.ref}). The kinetic energy
1114: in the radial direction grows by a factor of $10^3$ within 40 orbits.
1115: This translates into a characteristic growth time of $\approx 5.0$ orbits.
1116: The other components, especially the enstrophy,  grow more slowly 
1117: but
1118: do not saturate as quickly 
1119:  as the radial kinetic energy (which is also slightly
1120: damped for numerical reasons [see \S 3.1]). 
1121: The strong rise of enstrophy ($\approx$ vorticity generation) is an
1122: indication that a baroclinic instability is at work.
1123: 
1124: The angular momentum transport and also the on-diagonal components of the
1125:  stress tensor (Fig.\ \ref{fig14.ref})
1126: are orders of magnitude stronger than in model 2.
1127:  They are weaker than in models 1 and 1B, 
1128: which may result from the smaller computational domain, 
1129: especially in the azimuthal direction, artificially limiting
1130: the wave numbers of the instability.
1131: The angular momentum transport ($\alpha = A_{r\phi}$) is in the mean as strong as 
1132: $1.5 \times 10^{-4}$. This value should be 
1133: contrasted  to the strength of the turbulence itself. An
1134:  $A_{rr}$ and $A_{\phi\phi}$
1135: of $\approx 10^{-3}$ correspond to a turbulent Mach number as strong as 0.05.
1136: In isotropic turbulence this value could be used for a mixing
1137: length model and would
1138: then predict an alpha-viscosity of $\alpha = 2.5 \times 10^{-3} - 5\times 10^{-2}$,
1139:  depending on the estimates
1140: for the ``typical eddy size''. The lower value of $\alpha$  
1141: that we obtain indicates that even when strong turbulence  develops
1142: it does not automatically generate strong Reynolds stresses, which again
1143: is evidence
1144: for non-isotropic turbulence. The energy of this turbulence is drawn from the entropy
1145: background which can maintain its profile due to the accretion of mass. 
1146: The measured mass accretion rate of $\dot M \approx 
1147: - 2.0 \times 10^{-9}~\mbox{M}_{\sun}~\mbox{yr}^{-1}$ is in rough agreement with the 
1148: expectation from a viscous accretion disk model with a surface density
1149: of $\approx 300$ g cm$^{-2}$ and an $\alpha$ parameter of $1.5 \times10^{-4}$.
1150: 
1151: The turbulence saturates due to dissipation from shocks and stays at 
1152: a high, almost sonic, level. This level is on the long term quite
1153: variable. 
1154: 
1155: Our simulation demonstrates  that non-magnetic turbulence can drive outward angular 
1156: momentum transport
1157: and is maintained itself by the resulting accretion process.
1158: Such a conclusion has been in doubt for a long time,  as no 
1159: instability mechanism has been convincingly demonstrated 
1160: to work in Rayleigh-stable and non-ionized disks.
1161: 
1162:    We checked our findings on model 3 by redoing the simulation at 
1163: twice the resolution (model 4: $128 \times 128$
1164: grid cells) and obtained general agreement (see Fig. \ref{fig15.ref})
1165: with a mean $\alpha = A_{r\phi}$  slightly larger than that of model 3. 
1166: A detailed study of the influence of resolution on our results
1167: is a current subject of our investigations and will be
1168: part of a future paper.
1169: 
1170: 
1171:    For model 5 we used also a resolution of $128 \times 128$
1172: grid cells, but this time we used a computational domain which
1173: was essentially twice as wide in both directions. It spans radii from
1174: 3 AU to 7 AU and  $60\arcdeg$ in the $\phi-$direction. The numerical
1175: resolution is thus about the same as in model 3. 
1176: In Fig.\ \ref{fig16.ref} we plot the Reynolds stresses,
1177: which are stronger than in the previous models with $\alpha = A_{r\phi} \approx
1178:  6 \times 10^{-4}$.
1179: This larger scale simulation showed also the formation of a vortex (see Fig. \ref{fig17.ref}).
1180: It is an anti-cyclonic prograde vortex with higher density and pressure than the back ground.
1181: The vortex emits spiral waves in the ambient accretion disk.
1182: %The time evolution of the turbulence (see Fig.\ \ref{fig17.ref}) reveals that the model goes into an ``outburst'' during the calculations and
1183: %the accretion rate  at one point reaches values of
1184: % $\dot M= - 10^{-5}~\mbox{M}_{\sun}~\mbox{yr}^{-1}$.
1185: %The outburst is associated with the turbulent destruction of a torus with nearly constant
1186: %angular momentum that had previously formed in the flow.
1187: %We can speculate that this is the result of a secondary
1188: %instability (e.g.\ Papaloizou \& Pringle 1987) of a situation that
1189: %was probably generated by the initial geostrophic turbulence.
1190: 
1191:  %The overall integrated accretion 
1192: %%rate  over the time plotted in Fig.\ 
1193: %%\ref{fig17.ref}
1194: %was $\dot M= - 6\times10^{-8}~\mbox{M}_{\sun}~\mbox{yr}^{-1}$, 
1195: %which agrees pretty well with the measured Reynolds stresses and corresponds to observed accretion rates for T-Tauri stars.
1196: %The highly time-variable accretion rate is 
1197: %another good example of the fact that baroclinic disks are difficult to
1198: %approximate with $\alpha$-disks. 
1199: %But on the other hand it might
1200: %be much easier to explain episodic accretion events like FU-Orionis
1201: %outbursts.
1202: 
1203: As model 5 is so far the best compromise between resolution and
1204: angular width, it is well suited for the examination of the  spectral density
1205: distribution  (see Fig.\ \ref{fig18.ref}) and 
1206: comparison with model 1 (see Fig.\ \ref{fig4.ref}).
1207: Here one clearly sees the transition between 2D and ``3D" characteristics (even though
1208: the calculations are actually 2-D)
1209: of the spectrum at a wave number of about $60$, a value which was
1210: so far observed
1211: in all simulations. This transitional wavenumber is known as ``Rhines
1212: blocking wavenumber''; roughly speaking, 
1213:  for $m = k > 60$  energy only cascades down to smaller 
1214: scales while for $k < 60$ it inversely cascades
1215: up to larger scales. The wave number $k=60$ corresponds to about 6 degrees in azimuth or
1216: a tenth of the radius, which is almost two pressure scale heights.
1217: Structures that are smaller than this are not influenced by the rotation and 
1218: apparently behave like 3D turbulence ($k^{-5/3}$), i.e.\ energy 
1219: cascades down to smaller scales to be ultimately dissipated.
1220: % The analytic slope for 2-D 
1221: % incompressible turbulence is not appropriate for a comparison because it is based on
1222: %the assumption of constant vorticity, which here is clearly not the case.
1223: Structures bigger than this feel the Coriolis force and thus the rotation
1224: and shear. Consequently they are forced to have 2D characteristics and an inverse
1225: energy cascade with $k^{-3}$. The same behavior can be observed
1226: in the earth's  atmosphere,
1227: where small structures up to 500 km in extent (i.e. local winds) 
1228: behave according $k^{-5/3}$, while larger structures 
1229: (like hurricanes) follow a $k^{-3}$ distribution.
1230: 
1231: The studies on the influence of the computational domain
1232: will have to be continued in future investigations.
1233: The same is true for parameter studies on the initial
1234: distribution of density (as well as $\gamma$ and temperature) 
1235: in the disk models and the influence on $\alpha$.
1236: The fact that we found a working hydrodynamical instability for disks which generates
1237: Reynolds stresses with the right sign and reasonable $\alpha$-values,
1238: seems to support the paradigm of an viscously driven accretion process. Considering 
1239: the limitations of our numerical method, we could argue that a better method or higher
1240: resolution could also lead to even stronger turbulent viscosity, as only the
1241: sound speed seems to be a natural upper limit for the velocity fluctuations.
1242: 
1243: Other flow characteristics besides the Reynolds stress
1244: are very strongly developed. First, the 2D-turbulence itself  produces a
1245: significant turbulent (r.m.s.) pressure which is not accounted for 
1246: in $\alpha-$models. Second, the vorticity
1247: and the deviation from the mean Keplerian profile are significant 
1248: features, which also
1249: cannot be handled in diffusion models. And finally even though $\alpha-$
1250: models might assume that
1251: there is some underlying non-axisymmetric turbulence, 
1252: they never would predict the
1253: formation of large scale flow features like long-lived and growing vortices of 
1254: several scale heights diameter. Even if such features were assumed 
1255: to be present initially, $\alpha-$ models would smear
1256: them out and not amplify them 
1257: (Godon \& Livio 1999b).
1258: 
1259: We have now demonstrated the capability of the global baroclinic instability
1260: to form non-axisymmetric structures and vortices in the local simulations
1261: with periodic boundaries. The next question to be investigated concerns
1262: what would happen if 
1263: the global shear were not fixed by the radial boundary conditions but were
1264: free to evolve.  Thus
1265: we set up a global baroclinic simulation to see how a real disk would 
1266: behave and to what extent it could still be described by an $\alpha-$model.
1267: 
1268: \section{A global 2D simulation}
1269: Our global simulation shows that all the findings from the
1270: local models (model 3,4 and 5) can be confirmed. Actually the
1271: Reynolds-stresses are even stronger than in the local simulations,
1272: as smaller wave numbers in the azimuthal direction can be resolved.
1273: The wave number $ m = 1 $ seems to be the preferential mode of
1274: the instability. A further discovery was the self-consistent 
1275: formation of long-lived anti-cyclonic vortices as a direct result
1276: of the global baroclinic instability. This may have major relevance
1277: for the formation process of planets.
1278: 
1279: The simulation (model 6) covers the entire $360\arcdeg$ of the circumference of the
1280: disk and a radial section between 1 AU and 10 AU. The grid measures
1281: $128\times128$ grid-cells, which are radially logarithmically distributed. 
1282: We can thus resolve azimuthal wave numbers between 1 and 64. 
1283: The boundary conditions in the radial direction were
1284: changed from periodic to simple non-reflecting outflow conditions (vanishing 
1285: gradients),  not allowing for inflow. 
1286:  As a result of this change we also could drop the damping of the radial
1287:    component of the velocity close to radial boundaries.
1288: The density distribution 
1289: again was $\rho \propto  R^{-1}$ (constant $\Sigma \approx 300$  g cm$^{-2}$), 
1290:  and the temperature distribution was $T \propto  R^{-1}$, thus we have a
1291:  baroclinically unstable situation  as in model 3, which results from $H/R = 0.055$.
1292: The model was first run into a stable 1D axisymmetric
1293: state, where the residual velocities were less than 
1294: $10^{-4}$ cm s$^{-1}$. 
1295: Without a symmetry-breaking instability and turbulence generation, this disk 
1296: can not evolve and would stay perfectly laminar forever, as in 
1297:  the ``dead zone'' described by  Gammie (1996).
1298: 
1299: The initial density distribution was then 
1300: perturbed by  random noise of amplitude only 0.1$\%$. 
1301: The initial state is
1302: practically axisymmetric.
1303: Figure \ \ref{fig19.ref} illustrates the evolution of the flow in two
1304: space dimensions, over the full 360$\arcdeg$.  
1305: After the first 1 orbit (30 yr at 10 AU; Fig.\ \ref{fig19.ref}a) only little
1306: structure has evolved.  But with time a prominent anticyclonic vortex forms,
1307: which reflects the assumption that $m=1$ is the preferred mode.
1308: Intermittently a  second vortex also forms,  and we assume that
1309: their number is just limited by the narrowness of our disk and 
1310: a lack of matter.
1311: The vortex grows in mass and
1312: propagates radially outward,
1313:  possibly as a result of the gradient of background
1314: vorticity and the fact that anticyclonic vortices 
1315: are a local sink for angular momentum in the global vorticity field.
1316: Anyway, as there was never a radial drift of
1317: vortices in barotropic flows reported, this effect might
1318: also be linked to the baroclinic features of the flow.
1319: A detailed investigation of this effect still has to 
1320: be performed.
1321: 
1322: 
1323: Figure \ \ref{fig20.ref} shows the Reynolds stress and 
1324: turbulent Mach number averaged over the orbits from 430 to 500. 
1325: We see that the Reynolds stress does not
1326: disappear as we remove the ``shearing disk'' boundaries. 
1327: The stresses are comparable in the main part of the disk to those in 
1328: models 3, 4, and 5  and only deviate at the physical edge of the disk where 
1329: density and sound speed drop by orders of magnitude. 
1330: We can conclude that the ``shearing disk'' 
1331: boundary condition does not significantly affect the results. 
1332: 
1333: Figure \ \ref{fig21.ref} shows the situation after about $ 10^4$ years
1334: in real Cartesian coordinates to give an impression of the global
1335: nature of the simulation.
1336: Figure \ \ref{fig22.ref}  shows the flow pattern in more detail 
1337: in ($R,\varphi$) coordinates at the end of the simulation. 
1338: A huge vortex has formed which has a factor 4 over-density with respect to the
1339: ambient disk and a factor 2 over-density with respect to the initial local
1340: surface density. It is a high-pressure anticyclone that has the property of
1341: collecting solid material in its center (Tanga et al.\ 1996; Godon \& Livio 1999b).
1342: At the same time the over-dense blob inherits a substantial fraction of the 
1343: disk gas, which is not confined  by self gravity but only by the pressure
1344: gradient generated by the anti-cyclonic (less pro-grade) rotation. 
1345: While the initial nebula (from $1-10$ AU) had a mass of about 
1346: $10^{-2} M_\sun$, there are only $8 \times 10^{-3} M_\sun$
1347: left after $10^4$ years which corresponds to a mass loss (radially inward and
1348: outward) of $2.0 \times 10^{-7}~ \mbox{M}_{\sun}~\mbox{yr}^{-1}$.
1349: We cannot tell in this simulation how fast mass is being accreted onto the star
1350:  as our computational domain ends at 1 AU. 
1351: The red blob collects a total of $\approx 10^{30} $  g (170 M$_\earth$ 
1352: or 0.5 Jupiter masses [M$_{\rm J}$]).
1353:  Without further addition of mass or cooling 
1354: this object is not gravitationally unstable, as the Toomre $Q$ in the
1355: disk is about 10 and the local Jeans mass in the condensation is 
1356: about 2.5 M$_{\rm J}$. 
1357: 
1358: We see that even in a disk which 
1359: is not massive enough to fragment into planets or brown dwarfs (Boss 1998),
1360: a kind of pre-protoplanet can form simply as a result of baroclinicity and
1361: the resulting vorticity. The object, which is not yet gravitationally
1362: bound, could evolve into a planet in one of two ways: 
1363: (1) it could  efficiently collect solid particles in the center and
1364: wait until the critical mass for gas accretion is reached, or (2) it could 
1365: concentrate enough gas and cool down efficiently to become
1366: gravitationally unstable. In either scenario the time scale for planet 
1367: formation  will
1368: be shorter than in cases without the vortices in the disk.
1369: Additionally the vortex scenario can explain why there could be a solid
1370: core in a
1371: planet even it formed by gravitational instability rather than by accretion
1372: of solid material to form a core followed by gas capture. 
1373: 
1374: \section{Conclusions}
1375: The global baroclinic instability is found to generate
1376: turbulence in disks and drive an accretion process.
1377: 
1378: In general we found numerically that
1379: isotropic turbulence in the $R-\phi$ plane of the disk has
1380: the property of transporting angular momentum radially outward.
1381: But only the global baroclinic instability seems to provide
1382: a reliable source for this turbulence in the first place.
1383: Thermal convection in the vertical direction of the disk
1384: is not necessary for this effect.
1385: 
1386: These results seem to be unaffected by the type of simulation.
1387: Whether 2D or 3D, whether polytropic, artificially heated or
1388: not heated, whether open boundaries or shearing disk, all
1389: simulations lead to the same conclusions.
1390: Nevertheless, only global 3D non-heated models will have
1391: the credibility to predict exact $\alpha$-values.
1392: 
1393: We showed 
1394: that a protoplanetary disk  with a
1395: density and temperature distribution which cannot be described
1396: by a single polytropic $K$ is not barotropic and is possibly 
1397: unstable against non-axisymmetric
1398: perturbations. 
1399: This non-isentropic  situation (with a radial entropy gradient) is ultimately
1400: a consequence of the radially decreasing gravitational force.      
1401: The vertical component of gravity pushes the disk together and determines
1402: the pressure. Thus it is natural for non-isothermal  disks to
1403: be baroclinic.
1404: 
1405: We demonstrated numerically in a simple 2D simulation that 
1406: a baroclinic disk is unstable and develops strong geostrophic
1407: turbulence while a barotropic disk is perfectly stable.
1408: 
1409: Now we can answer the questions raised in \S 4.
1410: (A): Balbus et al. (1996; see also Hawley et al.\ 1999) could not observe 
1411: this kind of instability
1412: as their simulations were barotropic. The simulations in SB96 allowed
1413: for local baroclinic effects but no global entropy gradient was present.
1414: They used the shearing-box 
1415: approximation; thus $\beta_\rho, ~\beta_T = 0$ and $\beta_K$ is 
1416: automatically zero. 
1417: (B): The instability is purely hydrodynamical with an initial e-folding 
1418: time for
1419: the growth of about
1420: 5 orbital periods for an entropy gradient with $\beta_K = 0.57$. It occurs naturally
1421: in rotational shear flows
1422: when surfaces of constant density are inclined versus surfaces
1423: of constant pressure. A detailed stability analysis does not
1424: exist yet, but this is true for a lot of turbulent
1425: situations. Indications are, however, that the lowest-order
1426: modes have the fastest growth rates.  It is also not known yet whether we
1427: are  observing a linear or a non-linear instability operating in the disks. 
1428: (C): The barotropic  shearing-disk simulations as well as the baroclinic 
1429: open-boundary simulations  indicate that
1430: the ``shearing disk'' boundary conditions are not responsible
1431: for the turbulence. 
1432: 
1433:    Thermal convection is only indirectly related to the baroclinic
1434: instability,  as convective and radiative transport are responsible 
1435: for the radial temperature distribution and therefore the 
1436:  radial entropy gradient. 
1437: An isothermal disk would always be barotropic,  but
1438: we are considering only the situation 
1439: where the optical depth is larger than one, so the disks are not
1440: isothermal. 
1441: The optical depth greater than one is also necessary
1442: for thermal convection, but a vertically
1443: convectively stable purely radiative disk 
1444: can also establish a radial entropy gradient.
1445: If convection is present,  then it can be important in generating 
1446: the initial non-axisymmetric perturbations induced in the Keplerian
1447: disk which are necessary to set off the baroclinic instability. 
1448:  On the other hand convection produces negative Reynolds
1449: stresses;  however these turn out to be 
1450:  orders of magnitude weaker than the ones created by
1451: the baroclinic instability.
1452: 
1453: The simulations also generate vorticity, as can be made plausible by 
1454: a simple argument. Equation (\ref{eq:unstable}) shows how a density fluctuation in
1455: the $\varphi$ direction leads to a pressure gradient that does not
1456: exactly line up with the density gradient. Thus 
1457: imagine two parcels of gas on the same orbit around
1458: the central object. When they get pushed apart in the azimuthal direction,
1459: then the pressure will try to restore the previous state by pushing them
1460: together again. In a barotropic disk they  perform a damped oscillation,
1461:  and the perturbation decays. 
1462: But in a baroclinic disk, assuming that the perturbation occurred
1463: only along the azimuthal direction, the gas pressure will now not only push
1464: the parcels azimuthally together again, but also will drive the gas parcels
1465: slightly radially outward. For two reasons they are then driven 
1466:  back to the lower radius.
1467: First, there is  the gas pressure at the larger radius,  and,  second, 
1468:  they do not have enough angular momentum to stay on that higher orbit.
1469: When they  return to their lower orbit, they will fall behind their
1470: original azimuthal position, as the gas on a lower orbit has a larger
1471: angular velocity then the one on the outer orbit. And thus it is clear that 
1472: a non-axisymmetric radial velocity distribution is created ---
1473: $\partial v_r/\partial \phi \neq 0$ --- and thus vorticity generated.
1474: This leads
1475: first to a ``meandering'' flow which eventually
1476: becomes completely chaotic,  characterized by irregular waves.
1477: Furthermore
1478: a radially local perturbation spreads quickly in the radial direction,
1479: as it always affects and perturbs neighboring  radii.
1480:  Other works (e.g. Adams \& Watkins 1995) 
1481:  have considered the behavior of vortices and their interaction 
1482: with viscosity before, but none of them  have created them. We show here  for the first
1483: time that they form necessarily in a realistic disk in the absence
1484: of an underlying viscosity.
1485: 
1486: 
1487: Finally,  the baroclinic instability generates turbulence with velocities close to the sound speed.
1488: Dissipation does not occur (homogeneously in time and space) on the small 
1489: scales as in $\alpha$ models but occurs in       
1490: large-scale shock structures. 
1491: The calculations strongly suggest that this baroclinic instability is a feasible
1492: way to maintain turbulence  and outward angular momentum 
1493: transport in protoplanetary disks,  even if the physical
1494: conditions do not allow for MHD  turbulence.
1495: It would follow that there is no such thing as a ``dead zone'' in protoplanetary disks.
1496: Evidently then also the paradigm of layered accretion has to be revised. 
1497: One also has
1498: to take into account a 
1499: transition zone in  a disk, which separates regions where
1500: MHD turbulence or hydro-turbulence dominate, and in which the two
1501: processes may coexist.  
1502: It also would be fruitful and interesting to study 
1503: the transition zone and in general the interaction between hydro
1504: turbulence and self gravitational instabilities, as such a process
1505: could be crucial for planet formation. These combined effects would be important
1506: at the earliest stages of evolution of a protoplanetary disk, when
1507: the disk is still relatively massive and 
1508: material is still accreting onto the disk from the surrounding molecular
1509: cloud. Once the disk becomes optically thick and radiative 
1510: transfer effects become important, one can expect baroclinic effects to
1511: occur. 
1512: 
1513:  Gravitational and baroclinic instabilities seem to have certain 
1514: properties in common. Each of them leads to non-axisymmetric modes
1515: and an angular momentum transport that is not necessarily describable 
1516: by an $\alpha-$ formalism.  In baroclinic disks we are able to
1517: measure quite reasonable, and even more significant, all positive values
1518: for the Reynolds stresses, but it could  be dangerous to use these values
1519: in the viscous description of an accretion disk. Our global model
1520: shows dramatically how far a real disk can depart from the 
1521: idealized axisymmetric laminar disk that 
1522: evolves in quasi-steady state on viscous time-scales.
1523: Thus, results from $\alpha-$disk models can only reflect the 
1524: longterm average properties of disks.
1525: 
1526:  The trans-sonic  turbulence would be expected to be  
1527: very influential on the passive dust contaminant in generating
1528: collisions and mixing as well as concentrating the dust grains.
1529: Nevertheless the  Mach number associated with mixing of angular momentum is almost two
1530: orders of magnitude less than that of the turbulence, which reflects the non-isotropic
1531: 2D character of the turbulence. 
1532: The spectral density (see Fig.\ \ref{fig4.ref} and Fig.\ \ref{fig18.ref}) indicates that there is also turbulence on the
1533: small scales, but only resolution studies can show how much energy
1534: really decays towards the Kolmogorov scale.
1535: In short, a baroclinic disk is  much more 
1536: turbulent than `viscous'.
1537: 
1538: 
1539:    Our simple global 2D model already shows azimuthal density fluctuations
1540: by a factor of four. But most up-to-date models which try to interpret observational
1541: data use axisymmetric disks. Our global simulation could be a first step
1542: in showing how the spectral 
1543: (radiation) energy distribution (SED) of a baroclinically unstable disk
1544: would look like in contrast to an axisymmetric one.
1545: 
1546: The anti-cyclonic rotating gas parcels are vortices, that
1547: could be the precursors of planetary formation. 
1548: They can be thought of as pre-protoplanets.  
1549: In this connection, 
1550: more realistic 3D models with radiation transport
1551: could allow for higher mass concentration, as the cooling can locally decrease
1552: the pressure. The planets could form either by concentration of dust in the centers of
1553: the vortices, as was suggested by Tanga et al. (1996) and Godon \& Livio (1999b),
1554: or by sufficient gas accretion onto a vortex so that it undergoes 
1555: gravitational collapse (Adams \& Watkins 1995). This second process would
1556: happen in a similar fashion to the model by Boss (1998) with the big difference
1557: that the vorticity takes care of the local mass enhancement even after the disk
1558: or even parts of the disk have become gravitationally stable.  It furthermore is an easy way
1559: to explain a solid core for Jupiter and Saturn, as the pre-planetary embryo, the 
1560: rotating slightly over-dense vortex, will already have accumulated planetesimals.
1561: 
1562:  In a final speculation  we suggest that there is  a connection
1563: between UXor events (Natta  et al. 1999) and the baroclinic instabilities.
1564: It would be logical to expect over-dense vortices to form
1565: in disks around Ae/Be-stars. In a nearly edge-on disk, a transit of such a cloud, which has a 
1566:  higher scale height than the material in the surrounding disk,
1567: could obscure the stellar
1568: light. In a viscous $\alpha-$disk these over-dense regions would smear
1569: out in the azimuthal direction on a dynamical time-scale, which is an orbital period, but 
1570: in a viscosity free disk, they do not only persist, they can form.
1571: 
1572: Further 
1573: 2D and 3D simulations are necessary to determine 
1574: the proper role for the global baroclinic instability 
1575: in accretion disk theory.
1576: Questions to be addressed are abundant, for example 
1577: what is the influence of the aspect ratio $H/R$ of the disk
1578: on the instability? What are the critical values for $\beta_K$?
1579: Can one measure the growth rates for idealized perturbations
1580: with a single wave number $m$?
1581: 
1582: Apparently  the dissipation and conservation properties of different numerical
1583: hydro schemes affect the development of the instability. We already noticed
1584: that not all available codes show the same results if one does baroclinic
1585: simulations. We will continue with these tests and suggest that other
1586: owners of hydro codes should try to do so as well, in order to prove or disprove our findings.
1587: 
1588: The next step in our investigations of this new instability will be a 
1589: detailed stability analysis, which has to be the ultimate proof for
1590: our numerical findings.
1591: 
1592: \acknowledgments
1593: We want to thank 
1594: Steve Balbus,
1595: Robbins Bell,
1596: J\"urgen Blum,
1597: Axel Brandenburg,
1598: Geoff Bryden,
1599: Andi Burkert,
1600: William Cabot,
1601: Pat Cassen,
1602: Jeff Cuzzi,
1603: Sandy Davies,
1604: Steve Desch,
1605: Christian Fendt,
1606: Charles Gammie,
1607: Patrick Godon,
1608: John Hawley,
1609: Thomas Henning,
1610: Sascha Kempf,
1611: Willy Kley,
1612: Greg Laughlin,
1613: Doug Lin,
1614: Mordecai-Mark Mac Low,
1615: Andy Nelson,
1616: Laura Nett,
1617: G\"unther R\"udiger,
1618: Michal R\'o\.zyczka,
1619: Scott Seagroves,
1620: Frank Shu,
1621: Jim Stone,
1622: Paolo Tanga,
1623: Caroline Terquem,
1624: David Trilling,
1625: Dotty Woolum, and 
1626: Richard Young
1627: for support, fruitful criticism and inspirational discussions. 
1628: The results  
1629: are based on calculations
1630: performed at the HLRZ (Hochleistungsrechenzentrum) in  J\"ulich, Germany 
1631: and on the NPACI grant CSC217 and CSC222 in San Diego. 
1632: This research
1633: has been supported in part by the NSF through grant AST-9618548,
1634: by NASA through grant NAG5-4610,  and by a
1635: special NASA astrophysics theory program which supports a joint
1636: Center for Star Formation Studies at NASA-Ames Research Center, UC
1637: Berkeley, and UC Santa Cruz.
1638: 
1639: %% Appendix material should be preceded with a single \appendix command.
1640: %% There should be a \section command for each appendix. Mark appendix
1641: %% subsections with the same markup you use in the main body of the paper.
1642: 
1643: %% Each Appendix (indicated with \section) will be lettered A, B, C, etc.
1644: %% The equation counter will reset when it ncounters the \appendix
1645: %% command and will number appendix equations (A1), (A2), etc.
1646: 
1647: 
1648: %% The reference list follows the main body and any appendices.
1649: %% Use LaTeX's thebibliography environment to mark up your reference list.
1650: %% Note \begin{thebibliography} is followed by an empty set of
1651: %% curly braces.  If you forget this, LaTeX will generate the error
1652: %% "Perhaps a missing \item?".
1653: %%
1654: %% thebibliography produces citations in the text using \bibitem-\cite
1655: %% cross-referencing. Each reference is preceded by a
1656: %% \bibitem command that defines in curly braces the KEY that corresponds
1657: %% to the KEY in the \cite commands (see the first section above).
1658: %% Make sure that you provide a unique KEY for every \bibitem or else the
1659: %% paper will not LaTeX. The square brackets should contain
1660: %% the citation text that LaTeX will insert in
1661: %% place of the \cite commands.
1662: 
1663: %% We have used macros to produce journal name abbreviations.
1664: %% AASTeX provides a number of these for the more frequently-cited journals.
1665: %% See the Author Guide for a list of them.
1666: 
1667: %% Note that the style of the \bibitem labels (in []) is slightly
1668: %% different from previous examples.  The natbib system solves a host
1669: %% of citation expression problems, but it is necessary to clearly
1670: %% delimit the year from the author name used in the citation.
1671: %% See the natbib documentation for more details and options.
1672: 
1673: \begin{thebibliography}{}
1674: \bibitem[]{} Adams, F.C., \& Watkins, R.\  1995, \apj, 451, 314
1675: \bibitem[]{} Adams, F.C., \& Watkins, R.\  1996, Rev. Mexicana, Serie de
1676: Conferencias, 1, 257 
1677: \bibitem[]{} Balbus, S.A., \& Hawley, J.F.\  1991, \apj, 376, 214
1678: \bibitem[]{} Balbus, S.A., Hawley, J.F., \&  Stone, J.M.\ 1996, \apj, 467, 76
1679: \bibitem[]{} Barge, P., \& Sommeria, J.\ 1995, \aap, 295, L1
1680: \bibitem[]{bec93} Beckwith, S.V.W., \& Sargent, A.I.  1993, in 
1681: Protostars and Planets III, ed. E.\ H.\ Levy \& J.\ I.\ Lunine (Tucson: University of Arizona Press), 521
1682: \bibitem[]{bell97} Bell, K.R., Cassen, P.M., Klahr, H., \& Henning, Th.\ 1997, \apj, 486, 372
1683: \bibitem[]{bel94} Bell, K.R., \& Lin, D.N.C.  1994, \apj, 427, 987
1684: \bibitem[]{bel94} Bell, K.R.\  1999, \apj, 526, 411
1685: \bibitem[]{} Boss, A.P.\  1998, \apj, 503, 923
1686: %\bibitem[]{ber89} Bertout, C.G., \& Bouvier, J.  1989, {ESO Workshop on Low Mass Star Formation and Pre-Main Sequence Objects}, (B. Reipurth, Ed.). ESO, Garching, 215
1687: \bibitem[]{}Bryden, G., Chen, X., Lin, D.N.C., Nelson, R.P., \& Papaloizou, J.C.B.\ 1999, \apj, 514, 344
1688: %\bibitem[Cameron 1978]{cam78} Cameron, A.G.W.\ 1978, Moon Planets, 18, 5 
1689: \bibitem[Cabot 1984]{cab84} Cabot, W.\ 1984, \apj, 277, 806
1690: \bibitem[Cabot 1996]{cab96} Cabot, W.\ 1996, \apj, 465, 874, C96
1691: \bibitem[Cameron 1978]{cam78} Cameron, A.G.W.\ 1978, Moon Planets, 18, 5 
1692: %\bibitem[Cassen 1981]{ca81} Cassen, P. M., Smith, B. F., Miller, R. H., 
1693: %\& Reynolds, R. T. 1981, Icarus, 48, 377 
1694: \bibitem[]{}Cuzzi, J.N., Dobrovolskis, A.R.,  \& Champney, J.M.\ 1993, Icarus, 106, 102
1695: \bibitem[]{}Cuzzi, J.N., Dobrovolskis, A.R., \& Hogan, R.C.\ 1996, in Chondrules and the Protoplanetary Disk, ed. R.\  Hewins, R.\  Jones, \& E.R.D.\  Scott (Cambridge: Cambridge Univ.\ Press), 35
1696: %\bibitem[]{}Desch, S.\ 2000,  ApJ, submitted
1697: \bibitem[]{}Drecker, A., \& R\"udiger, G.\ 2000, Geophysical and Astrophysical Fluid Dynamics, submitted
1698: \bibitem[]{}Dubrulle, B.\ 1993, Icarus, 106, 59
1699: \bibitem[]{}Dubrulle, B., Morfill, G., \& Sterzik, M.\ 1995, Icarus, 114, 237
1700: \bibitem[]{}Duschl, W.J., Strittmatter, P.A., \& Biermann, P.L.\ 2000, \aap, 357, 1123
1701: \bibitem[]{}Gammie, C.F.\ 1996, \apj, 457, 355
1702: \bibitem[]{}Glassgold, A.E., Najita, J., \& Igea, I.\ 1997,  \apj, 480, 344
1703: \bibitem[]{}Godon, P., \& Livio, M.\ 1999a, \apj, 521, 319
1704: \bibitem[]{}Godon, P., \& Livio, M.\ 1999b, \apj, 523, 530
1705: \bibitem[]{}Goldreich, P., Goodman, J.\ \& Narayan, R.\ 1986, \mnras, 221, 339
1706: \bibitem[]{}Goldreich, P., \& Tremaine, S. 1980,  \apj,  241, 425 
1707: \bibitem[]{}Hadley, G.\ 1735, Philosophical Transactions, 39 
1708: \bibitem[]{}Hartmann, L., Calvet, N., Gullbring, E., \& D'Alessio, P.\ 1998, \apj, 495, 385
1709: %\bibitem[]{} Hunter, C.\ 1972, Ann.\ Rev.\ Fluid Mech., 4, 219
1710: \bibitem[]{}Hawley, J., Balbus, S.A., \& Winters, W.F.\  1999, \apj, 518, 394 
1711: \bibitem[]{}Hawley, J., Gammie, C.F., \& Balbus, S.A.\  1995, \apj, 440, 742 
1712: \bibitem[]{}Kippenhahn, R., \& Thomas, H.C.\ 1982, \aap, 114, 77
1713: \bibitem[]{}Kippenhahn, R., \& Weigert, A..\ 1990, Stellar Structure and Evolution (Berlin: 
1714: Springer)
1715: %\bibitem[Klahr 1998]{kla98} Klahr, H.H.\ 1998, {Dissertation, Friedrich Schiller University, Jena}
1716: %\bibitem[]{} Klahr, H.H.\ 1999, Proceedings of: Two Decades of Numerical Astrophysics, submitted
1717: \bibitem[]{} Klahr, H.H.,  \& Bodenheimer, P.\ 2000a, 
1718: Proceedings of: Disks, Planetesimals and Planets, ASP Conference Series, eds. F. Garzón, C. Eiroa, D. de Winter, \& T. J. Mahoney (Astronomical Society of the Pacific), 63
1719: \bibitem[]{} Klahr, H.H., \& Bodenheimer, P.\ 2000b, Proceedings of: Planetary Systems in the Universe: Observation, Formation and Evolution, ASP Conference Series, eds. A. Penny, P. Artymowicz, A.-M. Lagrange, \& S. Russell (Astronomical Society of the Pacific), in press
1720: %\bibitem[]{} Klahr, H.H., Bodenheimer, P., \& R\"udiger, G.\ 2002,  in preparation
1721: \bibitem[]{} Klahr, H.H., \& Henning, Th.\  1997, Icarus, 128, 213
1722: \bibitem[]{} Klahr, H.H., Henning, Th., \& Kley, W.\  1999, \apj, 514, 325
1723: \bibitem[]{} Klahr, H.H., \& R\"udiger, G. \ 2002,  in preparation
1724: \bibitem[]{} Kley, W.\ 1999, \mnras, 303, 696
1725: \bibitem[]{} Kley, W., \& Lin, D.N.C.\ 1999, \apj, 518, 833
1726: \bibitem[]{} Kley, W., Papaloizou, J.C.B., \& Lin, D.N.C.\  1993, \apj, 416, 679
1727: \bibitem[]{} Knobloch, E., \& Spruit, H.C.\ 1986, \aap, 166, 359
1728: \bibitem[]{} Larson, R.B.\  1989, in The formation and evolution of 
1729: planetary systems, eds. H. A. Weaver \& L. Danly (Cambridge: Cambridge
1730:  University Press), 31
1731: \bibitem[]{} Laughlin, G., \& Bodenheimer, P. 1994, \apj, 436, 335
1732: \bibitem[]{} Li, H., Finn, J.M., Lovelace, R.V.E., \& Colgate, S.A.\ 2000, \apj, 533, 1023
1733: \bibitem[]{} Li, H., Colgate, S.A., Wendroff,B., \& Liska, R.\ 2001, \apj, 551, 874 
1734: \bibitem[]{} Lin, D.N.C., \&  Papaloizou, J.C.B. 1980, \mnras, 191, 37 
1735: \bibitem[]{} Lin, D.N.C., Papaloizou, J.C.B., \& Kley, W. 1993, \apj, 416, 689
1736: \bibitem[]{} Lin, D.N.C., \&  Papaloizou, J.C.B. 1985, in 
1737: Protostars and Planets II, eds.\ D.C.\ Black \& M.S.\ Matthews (Tucson: University of Arizona Press), 981
1738: \bibitem[]{} Lissauer, J.J.\  1993, \araa, 31, 129
1739: \bibitem[]{}Lovelace, R.V.E., Li, H., Colgate, S.A., \& Nelson, A.F.\ 1999, \apj, 513, 805
1740: \bibitem[]{} L\"ust, R.\ 1952, Z.\ Naturforschung, 7a, 87
1741: %\bibitem[]{lin80} Lin, D.N.C., \& Papaloizou, J. 1980, \mnras, 191, 37
1742: \bibitem[]{}Lynden-Bell, D., \& Pringle, J.E.\  1974, \mnras, 168, 603
1743: \bibitem[]{}Markiewicz, W.J., Mizuno, H., \& Voelk, H.J.\ 1991, \aap, 242, 286
1744: \bibitem[]{} Natta, A., Prusti, T., Neri, R., Thi, W.F., Grinin, V.P., \& 
1745: Mannings, V.\ 1999, \aap, 350, 541
1746: \bibitem[]{} Nautha, M.D., \& Toth, G.\ 1998, \aap, 336, 791
1747: %\bibitem[]{} Nelson, A.F., Benz, W., Adams, F.C., \& Arnett, D.\ 1998, \apj, 502, 342
1748: %\bibitem[]{} Papaloizou, J.C.B., \& Pringle, J.E. 1987, \mnras, 225, 267 
1749: \bibitem[]{} Pedlosky, J.P.\ 1987, Geophysical Fluid Dynamics, 2nd ed. (New York: 
1750: Springer-Verlag)
1751: \bibitem[]{} Richard, D., \& Zahn, J.P.\ 1999, \aap, 347, 734
1752: \bibitem[]{}R\"udiger, G., \& Drecker, A.\ 2001, Astron.\ Nachr., 322, 179%\bibitem[]{} Ruden, S., \& Lin, D.N.C.\ 1986, \aap, 308, 883
1753: \bibitem[]{} Ryu, D., \& Goodman, J.\ 1992, \apj, 388, 438
1754: \bibitem[]{} Sakai, S., Iizawa, I., \& Aramaki, E.\ 1997, http://www.gfd-dennou.org/library/gfd\_exp/
1755: \bibitem[]{sha73} Shakura, N.I., \& Sunyaev, R.A.  1973, \aap, 24, 337
1756: \bibitem[]{} Sheehan, D.P.,  Davies, S.S.,  Cuzzi, J.N., \& Estberg,
1757: G.N.\ 1999, Icarus, 142, 238
1758: \bibitem[]{}Spruit, H.C.\ 1987, \aap, 184, 173
1759: \bibitem[]{sto96} Stone, J.M., \& Balbus, S.A.\ 1996, \apj, 464, 364, SB96
1760: \bibitem[]{} Stone, J.M., Gammie, C.F., Balbus, S.A., \&  Hawley, J.F.\  2000,
1761:  in Protostars and Planets IV,  ed. V. Mannings, A. P. Boss, \& S. Russell 
1762: (Tucson: Univ. Arizona Press), 589      
1763: \bibitem[]{} Stone, J.M., \& Norman, M.L.\ 1992, \apjs, 80, 753
1764: \bibitem[]{str93} Strom, S.E., Edwards, S., \& Skrutskie, M.F.\  1993, in 
1765: Protostars and Planets III, ed. E.\ H.\ Levy \& J.\ I.\ Lunine (Tucson: University of Arizona Press), 837 
1766: \bibitem[]{} Tanga, P.,  Babiano, A.,  Dubrulle, B., \& Provenzale, A.\ 1996, Icarus, 121, 158
1767: \bibitem[]{} Tritton, D.J., \& Davies, P.A.\  1985,  in 
1768: Hydrodynamical Instabilities and the Transition to Turbulence, ed. 
1769: H.L.\ Swiney \& J.P.\ Gollub (Berlin: Springer-Verlag), 229 
1770: %\bibitem[]{tsc79} Tscharnuter, W.M., \& Winkler, K.H.  1979, {Comp. Phys. Comm.}, 18, 171
1771: \bibitem[]{}Toomre, A.\ 1964, \apj, 139, 1217
1772: \bibitem[von Neumann \& Richtmyer 1950]{vNeu50} von Neumann, J., \& Richtmyer, R.D.\ 1950, J.\ Appl.\ Phys., 21, 232
1773: \bibitem[]{}Ward, W. R. 1997, \apj, 482, L211 
1774: \bibitem[]{} Weidenschilling, S.J., \& Cuzzi, J.N.\ 1993, in 
1775: Protostars and Planets III, ed. E.\ H.\ Levy \& J.\ I.\ Lunine (Tucson: University of Arizona Press), 1031
1776: \end{thebibliography}
1777: 
1778: %% Generally speaking, only the figure captions, and not the figures
1779: %% themselves, are included in electronic manuscript submissions.
1780: %% Use \figcaption to format your figure captions. They should begin on a
1781: %% new page.
1782: 
1783: \clearpage
1784: 
1785: %% No more than seven \figcaption commands are allowed per page,
1786: %% so if you have more than seven captions, insert a \clearpage
1787: %% after every seventh one.
1788: 
1789: %% There must be a \figcaption command for each legend. Key the text of the
1790: %% legend and the optional \label in curly braces. If you wish, you may
1791: %% include the name of the corresponding figure file in square brackets.
1792: %% The label is for identification purposes only. It will not insert the
1793: %% figures themselves into the document.
1794: %% If you want to include your art in the paper, use \plotone.
1795: %% Refer to the on-line documentation for details.
1796: 
1797: 
1798: %% Tables should be submitted one per page, so put a \clearpage before
1799: %% each one.
1800: 
1801: %% Three table samples follow, two marked up in the deluxetable environment,
1802: %% one marked up as a LaTeX table.
1803: \begin{table}%
1804: %\footnotesize
1805: \begin{tabular}{l|l|r|r|r|l|l|l|l}
1806:  {name} & {grid} & {$R_{i,o}[{ AU}]$}  &  
1807: {$\delta\varphi$} & {$\beta_K$} & {$\alpha = A_{r\phi}$} & {$M$} &  
1808: {$t_{int}/t_{ dyn}$}   \\
1809: \hline
1810: %\startdata
1811: {Model 1} &$51\times20\times60$&3.5\,--\,6.5&90$\arcdeg$&--&$2\times 10^{-3}$ & $0.5$&68\\
1812: {Model 1B} &$51\times20\times60$&3.5\,--\,6.5&90$\arcdeg$&--&$2\times 10^{-2}$ & $1.5$&14\\
1813: {Model 2} &64$^2$    &4.0\,--\,6.0 &30$\arcdeg$ &  0.0   &$2\times 10^{-9}$&$10^{-4}$&110\\
1814: {Model 2B} &64$^2$    &4.0\,--\,6.0 &30$\arcdeg$ &  0.0   &$3\times 10^{-4}$&$0.04  $&1\\
1815: {Model 3} &64$^2$    &4.0\,--\,6.0 &30$\arcdeg$ &  0.57  &$1.5\times 10^{-4}$ &$ 0.03   $&110\\
1816: {Model 4} &128$^2$   &4.0\,--\,6.0 &30$\arcdeg$ &  0.57  &$2\times 10^{-4}$ &$ 0.04    $&130\\
1817: {Model 5} &128$^2$   &3.0\,--\,7.0 &60$\arcdeg$ &  0.57  &$2.5\times 10^{-4}$ &$ 0.05   $&63\\
1818: {Model 6} &128$^2$   &1.0\,--\,10.0&360$\arcdeg$ &  0.57  &$8\times 10^{-3}$ &$ 0.3    $&50
1819:  %\enddata
1820: \end{tabular}
1821: \caption{\label{tab1} Parameters chosen in the different
1822: simulations. These parameters are the dimensioning of the grid ($n_r, n_\vartheta,n_\varphi$) or 
1823: ($n_r, n_\varphi$),
1824:  the radial extent  of the disk  $R_{i}$  to $R_{o}$,
1825: the azimuthal extent {$\delta\varphi$}, the parameter $\beta_K$
1826:  which describes the variation of
1827: the polytropic $K$ with radius, 
1828: the strength of  the measured Reynolds stresses $\alpha = A_{r\phi}$,
1829: the Mach number ($M$) of the turbulence, and the  ratio of the time
1830: over which the averaging was performed to the orbital period at the
1831: outer edge of the disk. }
1832: \end{table}
1833: \clearpage
1834: 
1835: 
1836: 
1837: 
1838: 
1839: 
1840: 
1841: %% In this first example, note that the \footnotesize command has been
1842: %% used to shrink the table so it will fit on one page. Note also that
1843: %% the \label command needs to be placed inside the \tablecaption.
1844: 
1845: 
1846: 
1847: % mod3.5-6.5_V0_90_block.eps
1848: % /u/klahr/Hydro/SD/work/ux451572/mod3.5-6.5D_V0_90 
1849: \figcaption[fig1.eps]{\label{fig1.ref}Three-dimensional rendering of the
1850: vertical mass flux in model 1, at a time near the end of the 
1851: run.  Red denotes positive velocities,
1852: and blue denotes negative. Compare this figure to Fig.\ 3 in Stone \& Balbus (1996).} 
1853: 
1854: 
1855: 
1856: % mod3.5-6.5_V0_90_alpha3D.ps
1857: \figcaption[fig2.eps]{\label{fig2.ref} Turbulence in model 1: Vertically-, 
1858: azimuthally- 
1859: and time-averaged stresses over 68 orbits, 
1860: measured in units of an effective $A$ (see eq.[ \ref{alpha.ref}])
1861:  and plotted as a function of radius (in AU). 
1862: The mean value for the angular momentum transport ({\it upper left}: $\alpha = A_{r\phi} 
1863: \approx  2 \times 10^{-3}$) is given by the dotted line. Other frames 
1864: give the averages of the relative density fluctuation,
1865: the strength of the turbulence in terms of velocity fluctuations
1866: (eqs.[\ref{alpha_rr.ref}], [\ref{alpha_tt.ref}], and [\ref{alpha_pp.ref}]),
1867:   and Mach number.
1868: See \S 3.2 for explanation.} 
1869: 
1870: 
1871: 
1872: 
1873: % mod3.5-6.5_V0_90_ttxzsv.ps
1874: \figcaption[fig3.eps]
1875: {\label{fig3.ref} Model 1 at a time near the end of the 
1876: run: Surface density (
1877: colors: $96$ [violet] to $335$ [red] g/cm$^2$),
1878: velocities (vectors:  $v_{max} = 0.6 \times$ the local sound speed) and iso-temperature contours in the midplane.}
1879: 
1880: 
1881: 
1882: % mod3.5-6.5_V0_90_spectrum.eps
1883: \figcaption[fig4.eps]
1884: {\label{fig4.ref} Model 1 at the same time as 
1885: in Fig.\ \ref{fig3.ref}:
1886: Spectral density distribution of the velocities at the midplane computed along the {$\varphi$}-direction and averaged over radius. The slope for isotropic, incompressible turbulence (i.e. a Kolmogorov spectrum) is indicated by the dashed line and
1887: the spectrum for 2D geostrophic flows by the dotted line.}
1888: 
1889: 
1890: % \plotone{mod3.5_6.5_A0_ttxzsv.ps}
1891: %/u/klahr/Hydro/Jue/mod3.5-6.5/modD_A0/job2
1892: \figcaption[fig5.eps]
1893: {\label{fig5.ref} Model 1B: Surface density 52 orbits after the initial state (
1894: colors: $20$ [violet] to $523$ [red] g/cm$^2$),
1895: velocities (vectors:  $v_{max} = 0.75 \times$ the local sound speed) and iso-temperature contours in the midplane.}
1896: 
1897: % \plotone{mod3.5_6.5_A0_alpha3D.ps}
1898: \figcaption[fig6.eps]
1899: {\label{fig6.ref} Turbulence in model 1B: Vertically-, 
1900: azimuthally- 
1901: and time-averaged stresses over 14 orbits, 
1902: taken 52 orbits after the initial perturbation,
1903: measured in units of an effective $A$ (see eq.[ \ref{alpha.ref}])
1904:  and plotted as a function of radius (in AU). 
1905: The mean value for the angular momentum transport ({\it upper left}: $\alpha=A_{r\phi} 
1906: \approx  2 \times 10^{-2}$) is given by the dotted line. Other frames 
1907: give the averages of the relative density fluctuation,
1908: the strength of the turbulence in terms of velocity fluctuations
1909: (eqs.[\ref{alpha_rr.ref}], [\ref{alpha_tt.ref}], and [\ref{alpha_pp.ref}]),
1910:   and Mach number.
1911: See \S 3.2 for explanation.} 
1912: 
1913: % mod3.5-6.5D_V0_polyrad.ps
1914: \figcaption[fig7.eps]
1915: {\label{fig7.ref} Model 1: Radial
1916: distribution of the polytropic $K$ (normalized to the biggest value) 
1917: as measured from the temperature
1918: and density in the midplane in the thermal convective flow before the system 
1919: became turbulent, i.e.\ it was axisymmetric ({\it dotted lines}: snapshots at 10 different times; {\it solid line}:
1920: theoretical value for $\beta_{K} = 0.57$).}
1921: 
1922: 
1923: 
1924: % modD_55_ini_ttxz.ps
1925: \figcaption[fig8.eps]
1926: {\label{fig8.ref} Model 2: Surface density (colors: $277$
1927: [violet], $300$ [green]
1928:  to $305$ [red] g/cm$^2$ [same color coding as in Fig.\ \ref {fig12.ref}])
1929: after 90 orbits,
1930: velocities (vectors: $v_{max} = 1 \times 10^{-4} \times$ sound speed) 
1931: and iso-temperature contours in the midplane. 
1932: Barotropic simulation with  no growing non-axisymmetric instability.}
1933: 
1934: 
1935: % modD_55_ini_erhr2D.ps
1936: \figcaption[fig9.eps]
1937: {\label{fig9.ref} Model 2: Time development of 
1938: kinetic energy in the r-direction ({\it upper left}), in the $\phi$-direction 
1939: ({\it upper right}), spatially mean accretion rate in 
1940: M$_\odot$ yr$^{-1}$ averaged over five orbits  ({\it lower  left}), 
1941: and enstrophy (integral square vorticity; {\it lower  right}).  The units for 
1942: kinetic energy and enstrophy are normalized to the first value occurring 
1943: in the data set. As predicted for barotropic
1944: flows, no instability growth can be observed within the first 100 orbits.
1945: No vorticity is generated. 
1946: }
1947: 
1948: 
1949: % modD_55_ini_alpha2D.ps
1950: \figcaption[fig10.eps]
1951: {\label{fig10.ref} Model 2: Azimuthally- 
1952: and time-averaged Reynolds stress over 10 orbits ({\it upper left})
1953: taken 100 orbits after the initial perturbation, 
1954: measured in units of an effective $A_{r\phi}$ (see eq. [\ref{alpha.ref}]) and plotted 
1955: as a function of radius (in AU). 
1956: The mean value is $\alpha=A_{r\phi} \approx  2.0 \times 10^{-9}$, which is practically zero.
1957: Other frames give the overall Mach number,           
1958: and the  strength of the turbulence in the radial direction ($A_{rr}$; 
1959: see eq. [\ref{alpha_rr.ref}]) and
1960: the azimuthal direction ($A_{\phi\phi}$; see eq. [\ref{alpha_pp.ref}]).}
1961: 
1962: 
1963: 
1964: 
1965: % modD_55_B_alpha2D.ps
1966: \figcaption[fig11.eps]
1967: {\label{fig11.ref} Model 2B: Azimuthally- 
1968: and time-averaged Reynolds stress over 1 orbit ({\it upper left})
1969: right after a strong initial density perturbation, 
1970: measured in units of an effective $A_{r\phi}$ (see eq. [\ref{alpha.ref}]) and plotted 
1971: as a function of radius (in AU). 
1972: The mean value is $\alpha=A_{r\phi} \approx  3 \times 10^{-4}$.
1973: Other frames give the overall Mach number,           
1974: and the  strength of the turbulence in the radial direction ($A_{rr}$; 
1975: see eq. [\ref{alpha_rr.ref}]) and
1976: the azimuthal direction ($A_{\phi\phi}$; see eq. [\ref{alpha_pp.ref}]).}
1977: 
1978: 
1979: 
1980: 
1981: 
1982: % modD_ttxz.ps
1983: %/u/klahr/Hydro/Jue/modD_V0
1984: \figcaption[fig12.eps]
1985: {\label{fig12.ref} Model 3: Surface density (colors: $286$
1986: [violet] to $318$ [red] g/cm$^2$),
1987: velocities (vectors: $v_{max} =  0.03 \times$ sound speed) and iso-temperature contours in the midplane after 600 orbits.
1988: Baroclinic simulation with a still growing non-axisymmetric instability.}
1989: 
1990: 
1991: 
1992: % modD_erhr2D.ps
1993: \figcaption[fig13.eps]
1994: {\label{fig13.ref} Model 3: The quantities plotted have the same
1995: meaning as those in Fig.\ \ref{fig9.ref} and are directly comparable.
1996: In baroclinic
1997: flows, vorticity and enstrophy are created and instability 
1998: growth can be observed within the first 100 orbits.
1999: }
2000: 
2001: 
2002: 
2003: 
2004: 
2005: % modD_alpha2D.ps
2006: \figcaption[fig14.eps]
2007: {\label{fig14.ref} Model 3:  Azimuthally- 
2008: and time-averaged Reynolds stress over 380 orbits (starting with 2030 orbits after the initial perturbation),
2009: measured in units of an effective $A_{r\phi}$ (see eq. [\ref{alpha.ref}]) and plotted 
2010: as a function of radius (in AU). 
2011: The mean value ({\it upper left}) is $\alpha=A_{r\phi} \approx  1.5 \times 10^{-4}$. 
2012:  Other frames show the overall Mach number and 
2013: the strength of the turbulence in the radial direction
2014: ($A_{rr}$; see eq. [\ref{alpha_rr.ref}])
2015:  and in the  azimuthal direction ($A_{\phi\phi}$; see eq. [\ref{alpha_pp.ref}]).} 
2016: 
2017: 
2018: 
2019: % modD2_V0_alpha2D.ps
2020: % /u/klahr/Hydro/Jue/modD2_V0
2021: \figcaption[fig15.eps]
2022: {\label{fig15.ref} Model 4: Stresses averaged 
2023: over 380 orbits (starting with 2030 orbits after the initial perturbation).
2024: The quantities plotted have the same meaning
2025: as in Fig.\ \ref{fig14.ref}.
2026: The mean value of $\alpha=A_{r\phi} \approx 2\times 10^{-4}$.} 
2027: 
2028: 
2029: 
2030: % mod2D_V0_alpha2D.ps
2031: % /u/klahr/Hydro/Jue/mod2D_V0/job9_ApJ
2032: \figcaption[fig16.eps]
2033: {\label{fig16.ref} Model 5: Stresses averaged 
2034:  over 180 orbits (starting with 600 orbits after the initial perturbation).
2035: The quantities plotted have the same
2036: meaning as in Fig.\ \ref{fig14.ref}. 
2037: The mean value of $\alpha=A_{r\phi} \approx  5.0 \times 10^{-3}$.} 
2038: 
2039: 
2040: 
2041: % mod2D_V0_erhr2D.ps
2042: \figcaption[fig17.eps]
2043: {\label{fig17.ref} Model 5: The quantities plotted have the same
2044: meaning as those in Fig.\ \ref{fig12.ref}.
2045: This calculation was run over 230 orbits.}
2046: 
2047: 
2048: 
2049: 
2050: % mod2D_V0_spectrum.eps
2051: \figcaption[fig18.eps]
2052: {\label{fig18.ref} Model 5: Spectral density distribution of the velocities at the midplane computed along the {$\varphi$}-direction and averaged over radius. The slope for isotropic, incompressible turbulence (i.e. a Kolmogorov spectrum) is indicated by the dashed line and
2053: the spectrum for 2D geostrophic flows by the dotted line.}
2054: 
2055: 
2056: 
2057: % mod128a_ttxzS.eps
2058: \figcaption[fig19.eps]
2059: {\label{fig19.ref} Model 6: Four snapshots of the evolution of surface density (colors: $650$ [red], $550$ [yellow], $450$ [green], $250$ [blue] to  $<100$ [black] to g/cm$^2$)
2060: in the global model in polar ($r-\phi$) coordinates. 
2061: The times  are  a): 1, b): 95, c): 190 and d): 320 orbits at the outer radius. Contours
2062: of equal pressure are also shown. } 
2063: 
2064: 
2065: 
2066: 
2067: % mod128a_alpha2D.ps
2068: % /u/klahr/Hydro/Jue/mod128A/job1
2069: \figcaption[fig20.eps]
2070: {\label{fig20.ref} Model 6: Stresses averaged 
2071:  over 50 orbits (starting with 190 orbits after the initial perturbation).  Quantities plotted have the same meaning 
2072: as in Fig.\ \ref{fig10.ref}.  The mean value of $A_{r\phi} \approx  8.0 \times 10^{-3}$.} 
2073: 
2074: 
2075: 
2076: % mod128a_rund_SU.ps
2077: \figcaption[fig21.eps]
2078: {\label{fig21.ref} The ``pre-protoplanet'' in Model 6:
2079:  Surface density (colors: $650$ [red], $550$ [yellow], $450$ [green], $250$ [blue] to  $<100$ [black] to g/cm$^2$) in the global model is projected in a cartesian frame after 
2080: 320 orbits at the outer radius, which corresponds to $10^{4}$ yr.
2081:  Note that the condensation  is partially artificially smeared
2082: out in the $\varphi$-direction, which is a result of the low order advection scheme.
2083:  In reality one could expect the ``pre-protoplanet'' to be more strongly confined.} 
2084: 
2085: 
2086: 
2087: % mod128a_ttxzsv.eps
2088: \figcaption[fig22.eps]
2089: {\label{fig22.ref} Model 6: Surface density (colors: $650$ [red], $550$ [yellow], $450$ [green], $250$ [blue] to  $<100$ [black] to g/cm$^2$) and velocity (vectors:  $v_{max} \approx 1.5 \times$ sound speed)
2090:  in the global model in polar ($r-\varphi$) coordinates after 
2091: 320 orbits at the outer radius, which corresponds to $10^{4}$ yr. 
2092: Plotted velocities  in the $\varphi$ direction 
2093: are obtained by subtracting the mean azimuthal velocity at each radius, which explains
2094: why the vortex center seems to be displaced from the density maximum. }
2095: 
2096: 
2097: 
2098: \end{document}
2099: 
2100: 
2101: \begin{figure}
2102: \plotone{fig1.eps}
2103: % /u/klahr/Hydro/SD/work/ux451572/mod3.5-6.5D_V0_90 
2104: % \plotone{mod3.5-6.5_V0_90_block.eps}
2105: \caption{\label{fig1.ref}Three-dimensional rendering of the
2106: vertical mass flux in model 1, at a time near the end of the 
2107: run.  Red denotes positive velocities,
2108: and blue denotes negative. Compare this figure to Fig.\ 3 in Stone \& Balbus (1996).} 
2109: \end{figure}
2110: \clearpage
2111: 
2112: \begin{figure}
2113: \plotone{fig2.eps}
2114: % \plotone{mod3.5-6.5_V0_90_alpha3D.ps}
2115: \caption{\label{fig2.ref} Turbulence in model 1: Vertically-, 
2116: azimuthally- 
2117: and time-averaged stresses over 68 orbits, 
2118: measured in units of an effective $A$ (see eq.[ \ref{alpha.ref}])
2119:  and plotted as a function of radius (in AU). 
2120: The mean value for the angular momentum transport ({\it upper left}: $\alpha=A_{r\phi} 
2121: \approx  2 \times 10^{-3}$) is given by the dotted line. Other frames 
2122: give the averages of the relative density fluctuation,
2123: the strength of the turbulence in terms of velocity fluctuations
2124: (eqs.[\ref{alpha_rr.ref}], [\ref{alpha_tt.ref}], and [\ref{alpha_pp.ref}]),
2125:   and Mach number.
2126: See \S 3.2 for explanation.} 
2127: \end{figure}
2128: \clearpage
2129: 
2130: 
2131: \begin{figure}
2132: \plotone{fig3.eps}
2133: % \plotone{mod3.5-6.5_V0_90_ttxzsv.ps}
2134: \caption{\label{fig3.ref} Model 1 at a time near the end of the 
2135: run: Surface density (
2136: colors: $96$ [violet] to $335$ [red] g/cm$^2$),
2137: velocities (vectors:  $v_{max} = 0.6 \times$ the local sound speed) and iso-temperature contours in the midplane.}
2138: \end{figure}
2139: \clearpage
2140: 
2141: \begin{figure}
2142: \plotone{fig4.eps}
2143: % \plotone{mod3.5-6.5_V0_90_spectrum.eps}
2144: \caption{\label{fig4.ref} Model 1 at the same time as 
2145: in Fig.\ 3:
2146: Spectral density distribution of the velocities at the midplane computed along the {$\varphi$}-direction and averaged over radius. The slope for isotropic, incompressible turbulence (i.e. a Kolmogorov spectrum) is indicated by the dashed line and
2147: the spectrum for 2D geostrophic flows by the dotted line.}
2148: \end{figure}
2149: \clearpage
2150: 
2151: \begin{figure}
2152: \plotone{fig5.eps}
2153: % \plotone{mod3.5_6.5_A0_ttxzsv.ps}
2154: %/u/klahr/Hydro/Jue/mod3.5-6.5/modD_A0/job2
2155: \caption{\label{fig5.ref} Model 1B: Surface density 52 orbits after the initial state (
2156: colors: $20$ [violet] to $523$ [red] g/cm$^2$),
2157: velocities (vectors:  $v_{max} = 0.75 \times$ the local sound speed) and iso-temperature contours in the midplane.}
2158: \end{figure}
2159: \clearpage
2160: 
2161: \begin{figure}
2162: \plotone{fig6.eps}
2163: % \plotone{mod3.5_6.5_A0_alpha3D.ps}
2164: \caption{\label{fig6.ref} Turbulence in model 1B: Vertically-, 
2165: azimuthally- 
2166: and time-averaged stresses over 14 orbits, 
2167: taken 52 orbits after the initial perturbation,
2168: measured in units of an effective $A$ (see eq.[ \ref{alpha.ref}])
2169:  and plotted as a function of radius (in AU). 
2170: The mean value for the angular momentum transport ({\it upper left}: $\alpha=A_{r\phi} 
2171: \approx  2 \times 10^{-2}$) is given by the dotted line. Other frames 
2172: give the averages of the relative density fluctuation,
2173: the strength of the turbulence in terms of velocity fluctuations
2174: (eqs.[\ref{alpha_rr.ref}], [\ref{alpha_tt.ref}], and [\ref{alpha_pp.ref}]),
2175:   and Mach number.
2176: See \S 3.2 for explanation.} 
2177: \end{figure}
2178: \clearpage
2179: 
2180: 
2181: 
2182: 
2183: \clearpage
2184: \begin{figure}
2185: \plotone{fig7.eps}
2186: % \plotone{mod3.5-6.5D_V0_polyrad.ps}
2187: \caption{\label{fig7.ref} Model 1: Radial
2188: distribution of the polytropic $K$ (normalized to the biggest value) 
2189: as measured from the temperature
2190: and density in the midplane in the thermal convective flow before the system 
2191: became turbulent, i.e.\ it was axisymmetric ({\it dotted lines}: snapshots at 10 different times; {\it solid line}:
2192: theoretical value for $\beta_{K} = 0.57$)}.
2193: \end{figure}
2194: 
2195: \clearpage
2196: \begin{figure}
2197: \plotone{fig8.eps}
2198: % \plotone{modD_55_ini_ttxz.ps}
2199: % /l1/klahr/mod_flat/modD_V0_5
2200: \caption{\label{fig8.ref} Model 2: Surface density (colors: $277$
2201: [violet], $300$ [green]
2202:  to $305$ [red] g/cm$^2$ [same color coding as in Fig.\ \ref {fig12.ref}])
2203: after 90 orbits,
2204: velocities (vectors: $v_{max} = 1 \times 10^{-4} \times$ sound speed) 
2205: and iso-temperature contours in the midplane. 
2206: Barotropic simulation with  no growing non-axisymmetric instability.}
2207: \end{figure}
2208: \clearpage
2209: \begin{figure}
2210: \plotone{fig9.eps}
2211: % \plotone{modD_55_ini_erhr2D.ps}
2212: \caption{\label{fig9.ref} Model 2: Time development of 
2213: kinetic energy in the r-direction ({\it upper left}), in the $\phi$-direction 
2214: ({\it upper right}), spatially mean accretion rate in 
2215: M$_\odot$ yr$^{-1}$ averaged over 5 orbits  ({\it lower  left}), 
2216: and enstrophy (integral square vorticity; {\it lower  right}).  The units for 
2217: kinetic energy and enstrophy are normalized to the first value occurring 
2218: in the data set. As predicted for barotropic
2219: flows, no instability growth can be observed within the first 100 orbits.
2220: No vorticity is generated. 
2221: }
2222: \end{figure}
2223: \clearpage
2224: \begin{figure}
2225: \plotone{fig10.eps}
2226: % \plotone{modD_55_ini_alpha2D.ps}
2227: \caption{\label{fig10.ref} Model 2: Azimuthally- 
2228: and time-averaged Reynolds stress over 10 orbits ({\it upper left})
2229: taken 100 orbits after the initial perturbation, 
2230: measured in units of an effective $A_{r\phi}$ (see eq. [\ref{alpha.ref}]) and plotted 
2231: as a function of radius (in AU). 
2232: The mean value is $\alpha=A_{r\phi} \approx  2.0 \times 10^{-9}$, which is practically zero.
2233: Other frames give the overall Mach number,           
2234: and the  strength of the turbulence in the radial direction ($A_{rr}$; 
2235: see eq. [\ref{alpha_rr.ref}]) and
2236: the azimuthal direction ($A_{\phi\phi}$; see eq. [\ref{alpha_pp.ref}]).}
2237: \end{figure}
2238: 
2239: 
2240: \clearpage
2241: \begin{figure}
2242: \plotone{fig11.eps}
2243: % \plotone{modD_55_B_alpha2D.ps}
2244: \caption{\label{fig11.ref} Model 2B: Azimuthally- 
2245: and time-averaged Reynolds stress over 1 orbit ({\it upper left})
2246: right after a strong initial density perturbation, 
2247: measured in units of an effective $A_{r\phi}$ (see eq. [\ref{alpha.ref}]) and plotted 
2248: as a function of radius (in AU). 
2249: The mean value is $\alpha=A_{r\phi} \approx  3 \times 10^{-4}$.
2250: Other frames give the overall Mach number,           
2251: and the  strength of the turbulence in the radial direction ($A_{rr}$; 
2252: see eq. [\ref{alpha_rr.ref}]) and
2253: the azimuthal direction ($A_{\phi\phi}$; see eq. [\ref{alpha_pp.ref}]).}
2254: \end{figure}
2255: 
2256: 
2257: 
2258: \clearpage
2259: \begin{figure}
2260: \plotone{fig12.eps}
2261: % \plotone{modD_ttxz.ps}
2262: \caption{\label{fig12.ref} Model 3: Surface density (colors: $286$
2263: [violet] to $318$ [red] g/cm$^2$),
2264: velocities (vectors: $v_{max} =  0.03 \times$ sound speed) and iso-temperature contours in the midplane after 600 orbits.
2265: Baroclinic simulation with a still growing non-axisymmetric instability.}
2266: \end{figure}
2267: 
2268: \clearpage
2269: \begin{figure}
2270: \plotone{fig13.eps}
2271: % \plotone{modD_erhr2D.ps}
2272: \caption{\label{fig13.ref} Model 3: The quantities plotted have the same
2273: meaning as those in Fig.\ \ref{fig9.ref} and are directly comparable.
2274: In baroclinic
2275: flows, vorticity and enstrophy are created and instability 
2276: growth can be observed within the first 100 orbits.
2277: }
2278: \end{figure}
2279: 
2280: 
2281: 
2282: \clearpage
2283: \begin{figure}
2284: \plotone{fig14.eps}
2285: % \plotone{modD_alpha2D.ps}
2286: \caption{\label{fig14.ref} Model 3:  Azimuthally- 
2287: and time-averaged Reynolds stress over 380 orbits (starting with 2030 orbits after the initial perturbation),
2288: measured in units of an effective $A_{r\phi}$ (see eq. [\ref{alpha.ref}]) and plotted 
2289: as a function of radius (in AU). 
2290: The mean value ({\it upper left}) is $\alpha=A_{r\phi} \approx  1.5 \times 10^{-4}$. 
2291:  Other frames show the overall Mach number and 
2292: the strength of the turbulence in the radial direction
2293: ($A_{rr}$; see eq. [\ref{alpha_rr.ref}])
2294:  and in the  azimuthal direction ($A_{\phi\phi}$; see eq. [\ref{alpha_pp.ref}]).} 
2295: \end{figure}
2296: 
2297: \clearpage
2298: \begin{figure}
2299: \plotone{fig15.eps}
2300: % \plotone{modD2_R_V0_alpha2D.ps}
2301: \caption{\label{fig15.ref} Model 4: Stresses averaged 
2302: over 380 orbits (starting with 2030 orbits after the initial perturbation).
2303: The quantities plotted have the same meaning
2304: as in Fig.\ \ref{fig14.ref}.
2305: The mean value of $\alpha = A_{r\phi} \approx 2\times 10^{-4}$.} 
2306: \end{figure}
2307: 
2308: \clearpage
2309: \begin{figure}
2310: \plotone{fig16.eps}
2311: % \plotone{mod2D_V0_alpha2D.ps}
2312: \caption{\label{fig16.ref} Model 5: Stresses averaged 
2313:  over 180 orbits (starting with 600 orbits after the initial perturbation).
2314: The quantities plotted have the same
2315: meaning as in Fig.\ \ref{fig14.ref}. 
2316: The mean value of $\alpha = A_{r\phi} \approx  5.0 \times 10^{-3}$.} 
2317: \end{figure}
2318: 
2319: \clearpage
2320: \begin{figure}
2321: \plotone{fig17.eps}
2322: % \plotone{mod2D_V0_erhr2D.ps}
2323: \caption{\label{fig17.ref} Model 5: The quantities plotted have the same
2324: meaning as those in Fig.\ \ref{fig12.ref}.
2325: This calculation was run over 230 orbits.}
2326: \end{figure}
2327: 
2328: 
2329: \clearpage
2330: \begin{figure}
2331: \plotone{fig18.eps}
2332: % \plotone{mod2D_V0_spectrum.eps}
2333: \caption{\label{fig18.ref} Model 5: Spectral density distribution of the velocities at the midplane computed along the {$\varphi$}-direction and averaged over radius. The slope for isotropic, incompressible turbulence (i.e. a Kolmogorov spectrum) is indicated by the dashed line and
2334: the spectrum for 2D geostrophic flows by the dotted line.}
2335: \end{figure}
2336: 
2337: \clearpage
2338: \begin{figure}
2339: \plotone{fig19.eps}
2340: % \plotone{mod128a_ttxzS.eps}
2341: \caption{\label{fig19.ref} Model 6: Four snapshots of the evolution of surface density (colors: $650$ [red], $550$ [yellow], $450$ [green], $250$ [blue] to  $<100$ [black] to g/cm$^2$)
2342: in the global model in polar ($r-\phi$) coordinates. 
2343: The times  are  a): 1, b): 95, c): 190 and d): 320 orbits at the outer radius. Contours
2344: of equal pressure are also shown. } 
2345: \end{figure}
2346: 
2347: 
2348: \clearpage
2349: \begin{figure}
2350: \plotone{fig20.eps}
2351: % \plotone{mod128_alpha2D.ps}
2352: \caption{\label{fig20.ref} Model 6: Stresses averaged 
2353:  over 50 orbits (starting with 190 orbits after the initial perturbation).  Quantities plotted have the same meaning 
2354: as in Fig.\ \ref{fig10.ref}.  The mean value of $\alpha = A_{r\phi} \approx  8.0 \times 10^{-3}$.} 
2355: \end{figure}
2356: 
2357: \clearpage
2358: \begin{figure}
2359: \plotone{fig21.jpg}
2360: % \plotone{mod128a_rund_SU.ps}
2361: \caption{\label{fig21.ref} The ``pre-protoplanet'' in Model 6:
2362:  Surface density (colors: $650$ [red], $550$ [yellow], $450$ [green], $250$ [blue] to  $<100$ [black] to g/cm$^2$) in the global model is projected in a cartesian frame after 
2363: 320 orbits at the outer radius, which corresponds to $10^{4}$ yr.
2364:  Note that the condensation  is partially artificially smeared
2365: out in the $\varphi$-direction, which is a result of the low order advection scheme.
2366:  In reality one could expect the ``pre-protoplanet'' to be more strongly confined.} 
2367: \end{figure}
2368: 
2369: \clearpage
2370: \begin{figure}
2371: \plotone{fig22.eps}
2372: % \plotone{mod128a_ttxzsv.eps}
2373: \caption{\label{fig22.ref} Model 6: Surface density (colors: $650$ [red], $550$ [yellow], $450$ [green], $250$ [blue] to  $<100$ [black] to g/cm$^2$) and velocity (vectors:  $v_{max} \approx 1.5 \times$ sound speed)
2374:  in the global model in polar ($r-\varphi$) coordinates after 
2375: 320 orbits at the outer radius, which corresponds to $10^{4}$ yr. 
2376: Plotted velocities  in the $\varphi$ direction 
2377: are obtained by subtracting the mean azimuthal velocity at each radius, which explains
2378: why the vortex center seems to be displaced from the density maximum. }
2379: \end{figure}
2380: 
2381: 
2382: \end{document}
2383: 
2384: 
2385: 
2386: