1: \documentclass[preprint,Letter]{aastex}

2: %\documentclass[Letter,onecolumn]{emulateapj}

3: \usepackage{epsf}

4: %\usepackage{amsmath}

5: %\usepackage{emulateapj5}

6: %\usepackage{apjfonts}

7: 

8: %\usepackage{natbib}

9: %\citestyle{apj}

10: 

11: 

12: \def\kms{\hbox{\kern 0.20em km\kern 0.20em s$^{-1}$}}

13: \def\cm{\mbox{\kern 0.20em {\rm cm}\kern 0.20em}}

14: \def\K{\mbox{\kern 0.20em {\rm K}\kern 0.20em}}

15: \def\pc{\mbox{\kern 0.20em {\rm pc}\kern 0.20em}}

16: 

17: \def\msun{{\rm M}_\odot\ }

18: 

19: \newcommand {\apgt} {\ {\raise-.5ex\hbox{$\buildrel>\over\sim$}}\ }

20: \newcommand {\aplt} {\ {\raise-.5ex\hbox{$\buildrel<\over\sim$}}\ } 

21: 

22: \def\kB{k_{\rm B}}

23: 

24: \shortauthors{Koyama \& Ostriker}

25: \begin{document}

26: \title{Pressure Relations and Vertical Equilibrium in the Turbulent,

27:   Multiphase ISM}

28: \author{Hiroshi Koyama\altaffilmark{1,2} and Eve C. Ostriker\altaffilmark{1}}

29: 

30: \altaffiltext{1}{Department of Astronomy,

31: University of Maryland, College Park, MD 20742, USA;

32: hkoyama@astro.umd.edu, ostriker@astro.umd.edu} 

33: \altaffiltext{2}{Current address: High-Performance Computing Team,

34: Integrated Simulation of Living Matter Group, RIKEN,

35: 61-1 Ono-cho, Tsurumi, Yokohama, 230-0046 Japan; hkoyama@riken.jp}

36: 

37: 

38: \begin{abstract}

39: 

40: 

41: We use numerical simulations of turbulent, multiphase,

42: self-gravitating gas orbiting in the disks of model galaxies to study

43: the relationships among pressure, the vertical distribution of gas,

44: and the relative proportions of dense and diffuse gas.  A common

45: assumption is that the interstellar medium (ISM) is in

46: vertical hydrostatic equilibrium.  We show that the disk height and

47: mean midplane pressure in our multiphase, turbulent simulations are indeed

48: consistent with effective hydrostatic equilibrium, provided that the

49: turbulent contribution to the vertical velocity dispersion and the

50: gas self-gravity are included. 

51: Although vertical hydrostatic equilibrium gives a

52: good estimate for the mean midplane pressure $\langle P\rangle_{\rm

53: midplane}$, this does not represent the pressure

54: experienced by most of the ISM.  Mass-weighted mean pressures $\langle

55: P \rangle_{\rho}$ are typically an order of magnitude higher than

56: $\langle P\rangle_{\rm midplane}$ because self-gravity concentrates

57: gas and increases the pressure in individual clouds without raising

58: the ambient pressure.

59: 

60: We also investigate the ratio $R_{\rm mol}=M_{\rm

61: H_2}/M_{\rm HI}$ for our hydrodynamic simulations.

62: \citet{2006ApJ...650..933B} showed that $R_{\rm mol}$ is proportional

63: to the estimated midplane pressure in a number of systems.  We find

64: that for model series in which the epicyclic frequency $\kappa$ and

65: gas surface density $\Sigma$ vary together as $\kappa \propto \Sigma$,

66: we recover the empirical relation.  For other model series in which

67: $\kappa$ and $\Sigma$ are varied independently, the midplane pressure

68: (or $\Sigma$) and $R_{\rm mol}$ are not well correlated.  We conclude

69: that the molecular fraction -- and hence the star formation rate -- of a

70: galactic disk inherently depends on its rotational state, not just the local

71: values of $\Sigma$ and the stellar density $\rho_{\ast}$.  The empirical

72: result $R_{\rm mol} \propto \langle P\rangle_{\rm midplane}$ implies

73: that the three ``environmental parameters'' $\kappa$, $\Sigma$, and

74: $\rho_{\ast}$ are interdependent in real galaxies, presumably as a

75: consequence of evolution: real galaxies trend toward states with 

76: Toomre $Q$ parameter near unity.

77: Finally,

78: we note that $R_{\rm mol}$ in static comparison models far exceeds

79: both the values in our turbulent hydrodynamic simulations and observed

80: values of $R_{\rm mol}$, when $\Sigma > 10 \msun \ \pc^{-2}$,

81: indicating that incorporation of turbulence is crucial to obtaining a

82: realistic molecular fraction in numerical models of the ISM.

83: \end{abstract}

84: \keywords{galaxies: ISM --- hydrodynamics --- ISM: general

85: --- method: numerical --- instabilities, turbulence

86: --- stars: formation}

87: 

88: 

89: \section{Introduction}

90: 

91: All phases of the interstellar medium (ISM) are turbulent, and this

92: turbulence has many effects.  In the astrophysical literature,

93: turbulence is often treated as yielding a simple addition to the

94: thermal pressure, $P_{\rm total}= \rho(c_s^2 + v^2_{\rm turb})$, where

95: $v^2_{\rm turb}$ is the dispersion in the (one-dimensional) turbulent

96: velocity, and $c_s^2=P/\rho= f\kB T/\mu$ for gas with a total number

97: density $f n$ and mass density $\mu n$.  This approach is often

98: adopted when analyzing the stratification of interstellar gas clouds

99: and the ISM as a whole, with the combined pressure gradients taken to

100: balance the gravitational force per unit volume such that hydrostatic

101: equilibrium is maintained by the total pressure.  The turbulent

102: pressure is believed to be especially important in the cold components

103: of the ISM, for which observed linewidths far exceed the values of

104: $c_s$ inferred from excitation of atomic and molecular lines.

105: 

106: Models of effective hydrostatic equilibrium in the vertical direction, usually

107: assuming the turbulent and thermal velocity

108: dispersions are constants independent of height, are often applied to

109: observations of the large-scale Galactic ISM, and to observations of

110: the ISM in external galaxies (e.g.

111: \citealt{1991ApJ...382..182L,1994ApJ...433..687M,1995ApJ...448..138M,

112: 1997A&A...326..554C,2000MNRAS.311..361O,

113: 2002A&A...394...89N,2004ApJ...608..189D,

114: 2004ApJ...612L..29B,2006ApJ...650..933B, 2008AstL...34..152K}).  For

115: example, \cite{2002A&A...394...89N} showed that the observed atomic

116: and molecular disk thicknesses in the Milky Way can be fit well by

117: assuming effective hydrostatic equilibrium, and accounting for both the gas

118: self-gravity and the external gravitational potential of stars and

119: dark matter.  \cite{2004ApJ...612L..29B} and

120: \cite{2006ApJ...650..933B} (hereafter BR06) used a simplified approach

121: to hydrostatic equilibrium in order to estimate the midplane gas

122: pressure in a sample of disk galaxies, adopting a single velocity

123: dispersion for the gas, treating the gravitational potential as

124: dominated by the stars, and assuming the stellar disk's scale height

125: is independent of radius.  \cite{2008AstL...34..152K} extended the

126: analysis of BR06 but instead of adopting a constant scale height for

127: the stellar disk, they assumed that the velocity dispersion for the

128: stars is consistent with a state of marginal gravitational instability

129: (with Toomre parameter $Q_{\ast}=1.5$) for the corresponding stellar surface

130: density.  They then assumed hydrostatic equilibrium for all (gaseous

131: and stellar) components separately, and computed the self-consistent

132: midplane pressure, finding differences of order $30-40\%$ from the

133: simplified BR06 approach.  Although widely adopted, the effective

134: hydrostatic equilibrium model for the large-scale ISM has not, to our

135: knowledge, been explicitly verified using actual turbulent flows.  One

136: of the goals of this work is to test this formulation systematically,

137: using the solutions of time-dependent numerical hydrodynamic

138: simulations of turbulent, multiphase gas.  

139: 

140: In addition to providing support against gravity, pressure also

141: affects the phase balance in the ISM.  For a static system at a given

142: mean density $\bar n$, changing the pressure alters the proportions of

143: mass divided between dense clouds and diffuse intercloud medium;

144: e.g. for cold and warm components in pressure equilibrium, the mass

145: ratio of cold to warm gas is 

146: $M_{\rm cold}/M_{\rm warm}=[\bar n/n_{\rm warm} - 1]/[1 - \bar

147:   n/n_{\rm cold}]=

148: [\bar n k T_{\rm warm} - P]/[P - \bar n k T_{\rm cold}]$. The mean

149: density itself, however, depends on pressure through the

150: condition of vertical hydrostatic equilibrium. Turbulent pressure, as

151: it affects the response to external and self-gravity, can be expected

152: to change both the mean density and the mass fractions of dense and

153: diffuse gas.  Here, we investigate these effects quantitatively.

154: 

155: The fraction of ISM mass in dense gas is important from the point of

156: view of galactic evolution, since this component is the immediate

157: precursor to star formation.  A recent observational study of external

158: disk galaxies by BR06 identified a linear relationship between the

159: mean ratio of molecular-to-atomic mass, $R_{\rm mol}$, and an estimate

160: for the total midplane pressure $\propto \sqrt{\rho_{\ast}} \Sigma$, where

161: $\rho_{\ast}$ is the stellar volume density and $\Sigma$ is the total

162: gaseous surface density.  BR06 propose that the molecular fractions in

163: widely-varying types of galaxies -- and hence their respective star

164: formation efficiencies -- are therefore determined essentially by a

165: single parameter, the midplane pressure.  To investigate this

166: proposal, we use multiphase turbulence simulations in which we

167: independently vary the input galactic ``environmental'' parameters.

168: The observational study of BR06 focused on the dependence of $R_{\rm

169: mol}$ on $\rho_{\ast}$ and $\Sigma$, but another important -- and

170: independent -- environmental

171: parameter is the angular rotation rate $\Omega$ (and the associated

172: epicyclic frequency $\kappa^2= R^{-3}d (\Omega^2 R^4)/dR$).  Using our

173: data sets from turbulent simulations, we compare the pressure estimate

174: of BR06 to the true value of the pressure, and also test how $R_{\rm

175: mol}$ relates to the mean pressure measured in two different ways.

176: 

177: We note that a number of recent numerical studies have investigated

178: the formation of ISM structures with internal densities reaching those

179: similar to Giant Molecular Clouds (GMCs).  Some studies 

180: \citep[e.g.][]{2002ApJ...564L..97K,2005ApJ...633L.113H,2006ApJ...648.1052H,2007A&A...465..431H,2007A&A...465..445H,2008A&A...486L..43H,2007ApJ...657..870V}

181: have focused

182: on how this may occur as a consequence of the collision of large-scale

183: high-velocity flows that shock and cool,

184: becoming turbulent at the same time.  Other studies have focused on

185: the ability of self-gravitating instabilities to induce converging

186: flows over sufficiently large scales that massive, high-column density

187: structures similar to observed GMCs are created

188: \citep[e.g.][]{2001ApJ...559...70K,2007ApJ...660.1232K,

189: 2005ApJ...620L..19L,2006ApJ...639..879L}; these models include the

190: galactic shear and rotation that are important on these large scales,

191: and in some cases also include magnetic effects

192: \citep[e.g.][]{2002ApJ...581.1080K,2003ApJ...599.1157K}.  As spiral

193: arms are observed to be strongly associated with high molecular

194: fractions and star formation, some studies have focused on the

195: interaction between large-scale spiral shocks and self-gravity in

196: inducing GMC formation

197: \citep[e.g.][]{2002ApJ...570..132K,2006ApJ...646..213K,2008MNRAS.tmp.1249D}.

198: The details of conversion from diffuse to dense gas by cooling

199: downstream from spiral shock fronts has also recently been studied in

200: the absence of self-gravity

201: \citep[e.g.][]{2008MNRAS.389.1097D,2008ApJ...681.1148K}.  Taken

202: together, these and other recent studies have shown that significant

203: quantities of dense gas form naturally as a result of large-scale ISM

204: dynamical processes.  Of course, dense gas in the ISM 

205: is also returned to the

206: diffuse phases by the energetic inputs from star formation.  In the

207: present work, by incorporating feedback, we are able to evolve our

208: models until a quasi-steady state is reached.  This enables an

209: analysis of the correlations among statistical properties of the

210: system, in terms of their influence on the fraction of dense gas when

211: the system has reach a quasi-steady state of cloud formation and

212: destruction.

213: 

214: 

215: This paper is organized as follows: In \S 2 we briefly summarize our

216: numerical methods.  The specification of model parameters and the

217: results of statistical analysis in comparison to the

218: vertical-equilibrium approximation are presented in \S 3.  In \S 4, we

219: discuss the molecular fraction and investigate how it relates to the

220: ISM pressure in our models.  We summerize our results and discuss

221: implications for ISM structure and evolution in \S 5.

222: 

223: 

224: 

225: 

226: 

227: \section{Numerical Methods}

228: 

229: The analysis in this paper are based on time-dependent numerical

230: hydrodynamic simulations of turbulent, multiphase, interstellar gas.

231: Details of our numerical methods are presented in a companion paper

232: (Koyama \& Ostriker 2008, hereafter Paper I); here, we

233: briefly summarize the model properties and parameterizations.  The

234: models we use are two-dimensional, representing slices through the

235: ISM in radial-vertical ($R-z$) planes.  We include sheared galactic

236: rotation, a radial gravitational force (the centrifugal force and

237: gravity balance in the

238: unperturbed state, which assumes a rotation curve $V_c=const$), and

239: Coriolis forces in the equations of motion, as well as gaseous

240: self-gravity and vertical gravity representing the potential of the

241: stellar disk.  The gas is treated as a single fluid in chemical

242: equilibrium, and we include (volumetric) radiative heating and cooling

243: processes as a function of density and temperature appropriate to the

244: range $10 < T < 10^4$ K.  The thermal processes we incorporate include

245: photoelectric heating from small grains and polycyclic aromatic

246: hydrocarbons, heating and ionization by cosmic rays and X-rays,

247: heating by H$_2$ formation and destruction, atomic line cooling from

248: Hydrogen Lyman $\alpha$, CII, OI, Fe II, and Si II, rovibrational line

249: cooling from H$_2$ and CO, and atomic and molecular collisions with

250: grains.  We adopt shearing-periodic boundary conditions in the radial

251: direction.

252: 

253: To drive turbulence, we also include a model of stellar feedback:

254: within ``HII regions'' (which are defined by contours of the perturbed

255: gravitational potential surrounding regions where the density has

256: exceeded a specified threshold), the gas heating rate is increased by

257: a factor 1,000. As a consequence, gas within these ``HII regions''

258: heats to temperatures $\sim 10^4$K, irrespective of density.  The

259: detailed recipe for the feedback phenomenon is described in

260: \citeauthor{Paper1}. Our aim is not to represent star formation

261: feedback in a fully realistic manner, but to drive turbulence in a way

262: similar to that which occurs within the dense

263: ISM.  In this sense, our feedback prescription is similar

264: in spirit to simulations of giant molecular clouds in which turbulence

265: is applied via arbitrary forcing functions (e.g.

266: \citealt{1998ApJ...508L..99S,1999ApJ...524..169M,

267: 2000ApJ...535..887K}).  Thus, our results should be taken as

268: demonstrating the physical importance of turbulence to setting

269: properties such as the vertical thickness of the disk, not as giving

270: quantitative predictions for what the value of the disk thickness,

271: etc., should be.

272: 

273: 

274: \section{Model Series and Results}

275: 

276: In our local disk models, three free parameters are needed to

277: characterize the ``galactic environment'': the total surface

278: density of the gas $\Sigma$, the local epicyclic frequency $\kappa$, 

279: and the local stellar density $\rho_{\ast}$.  

280: As we assume a flat rotation curve, $\kappa=\sqrt{2}\Omega$ where $\Omega$ is

281: the angular rotation rate at the center of our domain.  The stellar

282: density is used in order to specify the vertical gravity 

283: ${\bf g}_{\ast} =- 4 \pi G \rho_{\ast} z \hat z$.  

284: 

285: Following \citeauthor{Paper1}, we study four Series of models to explore

286: the parameter dependence of our results. 

287: For each Series, we hold two quantities fixed and vary a third

288: quantity, as follows:

289: \begin{itemize}

290: \item Series Q: $\kappa/\Sigma$ and

291:       $\sqrt{\rho_{\ast}}/\Sigma$ are constant while $\Sigma$ varies;

292: \item Series K: $\kappa$ and $\sqrt{\rho_{\ast}}/\Sigma$ are constant

293:   while $\Sigma$ varies;

294: \item Series R: $\kappa/\Sigma$ and $\rho_{\ast}$ are constant while 

295: $\Sigma$ varies;

296: \item Series S: $\Sigma$ and $\rho_{\ast}$ are constant while $\kappa$

297:   (and $\Omega$) varies.

298: \end{itemize}

299: Since Toomre's parameter is proportional to $\kappa/\Sigma$,

300: Series Q and R would have constant gaseous $Q=\kappa c_s/(\pi G \Sigma)$ if

301: the sound speed $c_s$ were constant.  

302: The Q and R series correspond to values of $Q=2.1 (c_s/7\kms)$. 

303: Assuming a constant stellar velocity

304: dispersion, $\Sigma_{\ast} \propto \sqrt{\rho_{\ast}}$, so that 

305: the stellar Toomre parameter 

306: (hereafter $Q_{\ast}$) would also have the same value for all members of 

307: Series Q.  

308: In all members of the R and S Series and in the 

309: $\Sigma=15.0 ~\msun\pc^{-2}$ models

310: of the Q and K series, we take 

311: $\rho_{\ast}=0.14 ~\msun\pc^{-3}$.  

312: In the K Series, we use 

313: $\kappa=62.4 ~{\rm km}~{\rm s}^{-1}{\rm kpc}^{-1}$, 

314:  while in the S Series we use $\Sigma=15.0 ~\msun\pc^{-2}$.

315: 

316: This paper focuses on how turbulence affects the vertical structure of

317: the galactic ISM.  An important aspect of our studies is to understand

318: how the results differ from the situation in which turbulence is

319: absent.  Thus, as baselines for comparison, we have two vertical

320: non-turbulent model Series: one in which the gas and stellar surface

321: densities are proportional (Series HSP), and one in which the stellar

322: surface density is constant (Series HSC).  These correspond to

323: dynamical Series Q and K (for HSP) and Series R (for HSC),

324: respectively.  These models are one-dimensional in the vertical ($z$)

325: direction; each model represents the asymptotic hydrostatic

326: equilibrium state which develops in the absence of any stellar

327: feedback.

328: 

329: 

330: Figure \ref{fig:snapshot} shows a snapshot of the gas 

331: pressure in a dynamical model from Series Q, compared to the

332: hydrostatic model from Series HSP.  The density and temperature are

333: shown for the same snapshot in Figure 1 of Paper I. In the dynamical

334: model, the pressure overall increases toward the midplane, but there

335: are significant variations associated with structure in the gas; for

336: the particular snapshot shown, there is also a high-pressure region

337: near the left of the figure, which is associated with a locally-heated

338: star formation feedback region.  The hydrostatic model shows a secular

339: increase in pressure towards the midplane.

340: 

341: 

342: 

343: 

344: \begin{figure}

345: \vskip 2in

346: 

347: \epsscale{1.5}

348: \plottwo{p2f1acolor.ps}{p2f1bcolor.ps}

349: %\plottwo{p2f1amono.ps}{p2f1bmono.ps}

350:  \caption{ {\it Left}: A snapshot of gas pressure (logarithmic color

351:  scale) from Model Q11 simulation.  For comparison, the {\it right} panel 

352: shows the pressure in the hydrostatic model (HSP11) that has the same

353:  total gas surface density $\Sigma$ and stellar density $\rho_*$ 

354: as Model Q11. }

355: \label{fig:snapshot}

356: 

357: \vskip 4in

358: \end{figure}

359: %\clearpage

360: 

361: 

362: 

363: \subsection{Vertical Scale Height

364: \label{scale_height}

365: }

366: 

367: 

368: 

369: 

370: \begin{figure}

371: \epsscale{1.0}

372: \plottwo{Q2vz.ps}{K2vz.ps}

373: 

374: \plottwo{R2vz.ps}{S2vz.ps}

375:  \caption{Mean vertical velocity dispersion, weighted by mass.  Both

376: the thermal $c_s$ ({\it circles}) and the total (thermal + turbulent)

377: $\sigma_z=\sqrt{c_s^2+v_z^2}$ ({\it triangles}) dispersions are shown

378: for all Series.}

379: \label{fig:vz}

380: \end{figure}

381: 

382: 

383: 

384: We begin by examining the vertical velocity dispersion of gas in all

385: of the model Series.  Figure \ref{fig:vz} shows space- and

386: time-averages (weighted by mass) of both the thermal velocity

387: dispersion $c_s=(P/\rho)^{1/2}$ ({\it circles}) and the combined

388: thermal + turbulent velocity dispersion $\sigma_z=\sqrt{c_s^2+v_z^2}$

389: ({\it triangles}).  The four panels correspond to the Series Q, K, R,

390: and S.  In Series Q and R, the mean thermal velocity dispersion

391: decreases with increasing surface density. The reason for this is that

392: the mass fraction of cold, dense gas increases with $\Sigma$ in all of

393: these models (see \citeauthor{Paper1}). This is because gravity is

394: lower, and gas is less compressed (both vertically, and horizontally

395: by self-gravity), at low $\Sigma$.  In Series K, on the other hand,

396: the mean thermal speed has a local minimum at intermediate $\Sigma$.

397: Again, this can be understood in terms of the mass fraction of warm

398: gas, which is largest at low and high $\Sigma$ (see

399: \citeauthor{Paper1}) in this Series; at high $\Sigma$, the model is

400: extremely active in terms of feedback because (with constant $\kappa$)

401: the disk is quite unstable gravitationally.  For all the series in

402: which $\Sigma$ is the variable parameter (i.e. Q, K, and R), the

403: turbulent part of the total velocity dispersion increases with

404: $\Sigma$; this is because the higher-$\Sigma$ models have higher

405: feedback rates, and therefore increasing (or flat) turbulence levels.

406: For Series S (with constant $\Sigma$), the turbulence decrease as

407: $\Omega$ increases, as high $\kappa$ stabilizes the disk and prevents

408: gravitational collapse and feedback (see Figure 11 in

409: \citeauthor{Paper1}).  For all series, the (mass-weighted) turbulent

410: vertical velocity dispersion approaches or exceeds the (mass-weighted)

411: thermal velocity dispersion for some part of parameter space, so that

412: turbulent support of gas in the vertical gravitational field is

413: expected to be important.

414: 

415: 

416: \begin{figure}

417: \plottwo{Q2Have.ps}{K2Have.ps}

418: 

419: \plottwo{R2Have.ps}{S2Have.ps}

420:  \caption{Disk scale heights, for all hydrodynamic and hydrostatic models.

421: {\it Open boxes} denote the directly-measured scale height (see

422: eq. \ref{Have_eq}) for all hydrodynamic Series.

423: {\it Filled boxes} show the measured scale height for corresponding 

424: hydrostatic models (HSP for Series K and Q, HSC for Series R). 

425: {\it Open circles} and {\it triangles} 

426: show the estimated scale heights (see eq. \ref{Hest_eq}) using thermal and

427:  thermal plus turbulent velocity for $\sigma_z$, respectively. 

428: The bottom part of each panel shows the ratio of estimated scale heights to

429: direct measurements. 

430: }\label{fig:Have}

431: \end{figure}

432: 

433: 

434: 

435: 

436: 

437: Next, we measure (for all Series) the vertical scale height, using 

438: the following averaging:

439: \begin{eqnarray}

440: H_{\rm ave}=\sqrt{\frac{\sum_{\rm all ~ zones} \rho z^2 }

441: {\sum_{\rm all ~ zones} \rho}}

442: \label{Have_eq}

443: \end{eqnarray}

444: where $z$ is the vertical coordinate relative to the midplane.  We

445: further average the values of $H_{\rm ave}$ over time.

446: In order to test whether the velocity dispersion can be used to

447: obtain an accurate measure of the scale height, we also compute 

448: ``estimated'' vertical scale heights defined as:

449: \begin{eqnarray}

450: H_{\rm est}&=&

451: \frac{1}{\sqrt{2\pi}}

452: \frac{\sigma_z^2}

453: {G\Sigma+\left[(G\Sigma)^2+2G\rho_{\ast}\sigma_z^2\right]^{1/2}}

454: =

455: \frac{\sigma_z}{\sqrt{4\pi G \rho_*}}

456: \frac{1}

457: {A+\left[A^2+1\right]^{1/2}}\label{Hest_eq} \\

458: &=&\left\{\begin{array}{ll}

459: \displaystyle \frac{\sigma_z}{\sqrt{4\pi G\rho_{\ast}}} & (\Sigma \to 0)\\

460: \displaystyle \frac{\sigma_z^2}{\sqrt{8\pi}G\Sigma} & (\rho_{\ast} \to 0).

461: \end{array}

462: \right.

463: %&\approx&

464: %\frac{\sigma^2}{2\pi G\Sigma}

465: %=\frac{\lambda_{\rm J}}{2\pi}

466: %=Q\frac{\sqrt{\sigma^2}}{\kappa}

467: %\quad (\rho_{\ast}\ll \Sigma/2H).

468: \end{eqnarray}

469: This formula (see Appendix for derivation) accounts for both gas

470: self-gravity and stellar gravity; the limiting forms are for

471: negligible gaseous and stellar gravity, respectively.  

472: 

473: In equation (\ref{Hest_eq}), 

474: $A$ is a dimensionless factor that measures the relative

475: densities of the gaseous and stellar disks,

476: \begin{eqnarray}

477: A\equiv\sqrt{\frac{G\Sigma^2}{2\rho_{\ast}\sigma_z^2}}

478: =\frac{\Sigma \, c_{\ast,z}}{\Sigma_{\ast} \, \sigma_z \sqrt{\pi}}.

479: \end{eqnarray}

480: The latter expression treats the stellar disk as an isothermal

481: self-gravitating equilibrium, with 

482: $H_{\ast}=c_{\ast,z}^2/(\pi G \Sigma_{\ast})$,

483: and shows that $A\sim Q_{\ast}/Q$ (assuming that vertical and radial

484: velocity dispersions are proportional).  

485: The formula (\ref{Hest_eq}) may be thought of as an extension of

486: the usual non-self-gravitating scale height formula to account for the

487: gravity of the gas.  Since $A>0$, the correction factor depending on

488: $A$ is always $<1$.  

489: If the gas disk is much more gravitationally unstable than

490: the stellar disk ($A \sim Q_{\ast}/Q \gg 1$), the correction factor is

491: large; otherwise the correction factor is order-unity.

492: 

493: Figure \ref{fig:Have} shows the measured ($H_{\rm ave}$) and ``predicted''

494: ($H_{\rm est}$) disk scale heights for all series of hydrodynamic

495: models.  For $H_{\rm est}$, we show results using for $\sigma_z$

496: either the thermal velocity dispersion ($\sigma_z=c_s$; subscript $c$)

497: or the total velocity dispersion ($\sigma_z^2=c_s^2+v_z^2$; subscript

498: $c+v_z$).  To show how turbulence contributes to setting the disk

499: thickness, $H_{\rm ave}$ is also shown for the hydrostatic models.

500: The difference between $H_{\rm ave}$ in hydrostatic and hydrodynamic

501: models can be quite large, up to a factor 10 in some cases.  We note

502: that $H_{\rm ave}$ of the hydrostatic models ({\it filled

503: boxes}) differs from $H_{{\rm est},c}$ ({\it open circles}) because the

504: mass-weighted mean sound speed differs for hydrostatic and

505: hydrodynamic models.

506: 

507: 

508: 

509: Overall, Figure \ref{fig:Have} shows that the estimate for scale

510: height $H_{{\rm est},c+v_z}$ that includes turbulence traces the

511: measured $H_{\rm ave}$ quite well, for all the Series.  The difference

512: between $H_{{\rm est},c}$ and $H_{{\rm est},c+v_z}$ increases with

513: increasing $\Sigma$, with quite large differences for some of the

514: models in Series Q and R.  This indicates that high surface density

515: disks are supported largely by turbulent velocities, in these cases.

516: To facilitate comparisons between estimated and measured value of the scale

517: height, in the lower part of each panel we also show the ratios

518: $H_{{\rm est},c }/H_{\rm ave}$ ({\it circles}) and $H_{{\rm est},c+v_z

519: }/H_{\rm ave}$ ({\it triangles}).  At low values of $\Sigma$ in Series

520: Q, K, and R, both estimates of $H$ exceed the true measured value.  It

521: is notable that where the turbulent contributions are large, at high

522: $\Sigma$ in Series Q, K, and R, the estimated and measured disk

523: thicknesses are in quite good agreement (within $\sim 10-20\%$).

524: Thus, we conclude that if measurements of the vertical

525: velocity dispersion together with the gaseous surface density and

526: stellar surface density can be made observationally, they can be

527: combined to yield an accurate estimate of the gas disk's thickness.

528: 

529: 

530: \subsection{Gas Pressure}

531: 

532: The gaseous pressure, like the scale height, is often difficult to

533: measure directly.  As a consequence, other proxies are often used to

534: obtain an estimate of the value of the pressure, with an assumption

535: that vertical equilibrium is satisfied.  Here, we

536: test how well such pressure estimates agree with the directly-measured

537: pressure, for our multiphase turbulent models.

538: 

539: Figure \ref{fig:Pressure} shows for all models in all hydrodynamic 

540: Series the average gas pressure.

541: We consider two different ways of averaging:  

542: weighting by mass 

543: $\langle P \rangle_{\rho}$ ({\it open box}), and weighting by volume

544: $\langle P\rangle_{\rm midplane}$ ({\it open circle}).  

545: The value $\langle P \rangle_{\rho}$ is interesting because it

546: characterizes the value of pressure experienced by the average atom or

547: molecule, whereas $\langle P\rangle_{\rm midplane}$ is interesting 

548: because it represents the pressure in the diffuse

549: (non-self-gravitating) part of the ISM that is closest to star-forming

550: regions. 

551: 

552: The mass- and volume- weighted averages

553: are defined by the following: 

554: \begin{eqnarray}

555: \langle P \rangle_{\rho}&=&\frac{\int P dm}{\int dm},\\

556: \langle P \rangle_{\rm midplane}&=&\int 

557: \frac{P_{\frac{N_z}{2}}+P_{\frac{N_z}{2}+1}}{2} \frac{dx}{L_x}.

558: \end{eqnarray}

559: For $\langle P \rangle_{\rho}$, all zones in the domain are included,

560: while for $\langle P\rangle_{\rm midplane}$, the  

561: subscripts $\frac{N_z}{2}$ and $\frac{N_z}{2}+1$ indicate

562: that only zones in the two horizontal planes closest to the midplane

563: are included.

564: Time averaging is applied in all models after the above space averaging.

565: We also show the same pressure averages for the 

566: hydrostatic Series ({\it filled box} and {\it filled circle}).

567: Interestingly, in the hydrodynamic models 

568: $\langle P \rangle_{\rho}$ always exceeds 

569: $\langle P\rangle_{\rm midplane}$ by a large factor $\sim 10$.

570: This indicates that self-gravity is important in increasing the

571: pressure above the ``ambient'' value, for much of the gas.

572: Pressures cannot exceed the ambient midplane value without horizontal

573: gradients, which are balanced by the gravity within individual clouds 

574: (see Fig. \ref{fig:snapshot}).  

575: In the hydrostatic models, $\langle P \rangle_{\rho}$

576: ({\it filled boxes})

577: is generally slightly below $\langle P\rangle_{\rm midplane}$

578: ({\it filled circles}), because the pressure at the midplane is the

579: maximum within any system, and weighting by mass includes 

580: lower-pressure gas which reduces the average.  

581: (Note that for the hydrostatic models, there are no horizontal

582: gradients in any quantities; see Figure \ref{fig:snapshot}.)

583: Except in the most

584: active disks, the mass-weighted

585: averages for the hydrostatic models are close to the midplane values

586: for the hydrodynamic models.

587: In Figure \ref{fig:Pressure} we also

588: display the pressure estimate of BR06 ({\it solid line}) defined as: 

589: \begin{equation}

590: P_{\rm BR}=\Sigma v \sqrt{2G\rho_{\ast}},

591: \label{PBR_eq}

592: \end{equation}

593: where $v=8$ km/s is adopted. 

594: This line falls between $\langle P \rangle_{\rho}$ and 

595: $\langle P\rangle_{\rm midplane}$

596: for all the hydrodynamic Series. 

597: 

598: For hydrostatic Series HSP (shown in the Series Q and K panels), the

599: slope of the midplane pressure is close to that predicted by equation

600: (\ref{PBR_eq}), while being offset to lower $P$ by a factor 2-3.  The

601: difference in slope is because the medium has multiple phases, rather

602: than a single phase at a given thermal sound speed.  The offset is

603: because (i) much of the mass in the hydrostatic models is at low

604: temperatures, for which the sound speed is well below $8$ km/s, and

605: (ii) equation (\ref{PBR_eq}) does not include the gaseous vertical

606: gravity, which is comparable to the stellar gravity when vertical

607: velocity dispersion is low and the disk is very thin (see below).

608: These effects push $P$ in opposite directions, and hence partially 

609: compensate each other.  For

610: hydrostatic series HSC (shown in the Series R panel), the prediction

611: of equation (\ref{PBR_eq}) departs significantly from the slope of the

612: midplane pressure results, because in the HSC series (which has

613: $\rho_*$ constant) vertical gravity is strongly 

614: dominated by gas rather than the stellar component at large $\Sigma$.

615: 

616: 

617: \begin{figure}

618: %\epsscale{0.45}

619: \plottwo{Q2Pressure.ps}{K2Pressure.ps}

620: 

621: \plottwo{R2Pressure.ps}{S2Pressure.ps}

622:  \caption{Gas pressure averages for all Series.

623: {\it Open boxes} show mass-weighted averages and {\it open circles}

624: show the midplane pressure, for hydrodynamic models. 

625: {\it Filled boxes} and {\it filled circles} show 

626: the same for hydrostatic models.

627: The pressure estimate of BR06 is also indicated {\it solid line}.}

628: \label{fig:Pressure}

629: \end{figure}

630: 

631: 

632: 

633: \begin{figure}

634: %\epsscale{0.45}

635: \plottwo{Q2Pressure3.ps}{K2Pressure3.ps}

636: 

637: \plottwo{R2Pressure3.ps}{S2Pressure3.ps}

638:  \caption{Measured, fitted, and estimated gas pressures. 

639: {\it Open boxes} and {\it circles} are the same as in Figure

640:  \ref{fig:Pressure}.

641: {\it Solid} and {\it dashed} lines are the corresponding fits to 

642: equation (\ref{Pfit_eq}). Our vertical-equilibrium midplane 

643: thermal pressure

644: estimate (eq. \ref{eq:Pnewth}) is plotted as {\it filled triangles}.}

645: \label{fig:Pressure2}

646: \end{figure}

647: 

648: In \S \ref{scale_height}, we defined an ``average'' vertical

649: equilibrium using the total surface density and the total vertical

650: velocity dispersion, and showed that this could yield an accurate

651: measurement of the disk thickness.  The same model (see Appendix) can

652: be used to estimate a midplane ``effective hydrostatic pressure,'' 

653: which we can compare to

654: measured values.  If $H$ is the scale height, then in equilibrium 

655: the mean midplane gas density is $\rho_0=\Sigma/(\sqrt{2\pi}H)$.

656: Using the total velocity dispersion, the predicted total 

657: gas pressure at the midplane is then given by 

658: $P_{0,{\rm tot}}=\sigma_z^2\rho_0$,

659: which using equation (\ref{Hest_eq}) gives 

660: \begin{eqnarray}

661: P_{0,{\rm tot}}&=&

662: \frac{\sigma_z^2\Sigma}{\sqrt{2\pi}H_{{\rm est},c+v_z}}

663: =\Sigma\left(G\Sigma+\left[(G\Sigma)^2+2G\rho_{\ast}\sigma_z^2\right]^{1/2}\right)\cr

664: &=&\Sigma \sigma_z \sqrt{2 G\rho_{\ast}}

665: (A+\sqrt{A^2+1}).

666: \label{eq:Pnew}

667: \end{eqnarray}

668: The expression (\ref{eq:Pnew}) corresponds to an extension of the

669: pressure estimate formula of BR06 using the inverse of the

670: $A$-dependent correction

671: factor that appears in the scale height estimate (\ref{Hest_eq}).

672: This correction factor is $>1$.

673: 

674: Equation (\ref{eq:Pnew}) gives an estimate of the total midplane

675: pressure, but the thermal pressure should represent only a fraction

676: $\langle c_s^2\rangle/\langle c_s^2 + v_z^2\rangle= 1- \langle

677: v_z^2\rangle/\sigma_z^2$ of $P_{0,tot}$, where $\langle v_z^2\rangle^{1/2}$

678: is the mass-weighted RMS turbulent velocity dispersion in the vertical

679: direction.  Thus, our estimate of the mean thermal pressure at the

680: midplane is

681: \begin{equation}

682: P_{0,{\rm th}}=\langle c_s^2\rangle \frac{\Sigma}{\sqrt{2\pi}H_{{\rm

683:     est},c+v_z}}

684: =\Sigma \frac{\langle c_s^2\rangle}{\sigma_z} \sqrt{2 G\rho_{\ast}}(A+\sqrt{A^2+1}).

685: \label{eq:Pnewth}

686: \end{equation}

687: In Figure \ref{fig:Pressure2}, we compare the pressure estimate from

688: equation (\ref{eq:Pnewth}) ({\it filled triangles}) with the

689: measurements of midplane pressure.  At large

690: $\Sigma$, the agreement is quite good, while at low $\Sigma$ the

691: estimated midplane pressures lie slightly above the measured values.

692: This behavior is similar to our results for estimated scale heights, 

693: which were in best agreement with the measured $H_{\rm ave}$ at large

694: $\Sigma$ (where the dense gas dominates the mass, and the velocity

695: dispersion is turbulence-dominated).  

696: 

697: 

698: For all the Series in which $\Sigma$ is the independent variable,

699: we have fit the measured gas pressure to the formula:

700: \begin{eqnarray}

701: P/\kB

702: =D\sqrt{\frac{\rho_{\ast}}{\msun\pc^{-3}}}

703: \left(\frac{\Sigma}{\msun\pc^{-2}}\right).

704: \label{Pfit_eq}

705: \end{eqnarray} 

706: For $\langle P\rangle_{\rho}$ and $\langle P\rangle_{\rm midplane}$,

707: we find the respective coefficients are $D_{\rho}=1.3\times 10^4 \K

708: \cm^{-3}$ 

709: and

710: $D_{\rm midplane}=1.1\times 10^3 \K

711: \cm^{-3}$, respectively.  The largest and

712: smallest surface density models are excluded in the fits.  The results

713: of the fits are displayed as {\it solid} and {\it dashed} lines,

714: respectively, in Figure \ref{fig:Pressure2}.  To compare with the BR06

715: formula, we also fit $\langle P\rangle_{\rho}$ and $\langle

716: P\rangle_{\rm midplane}$ to $P=C \Sigma v \sqrt{2G\rho_{\ast}}$ with

717: $v=8$ km/s. We find $C_{\rho}=3.6$ and $C_{\rm midplane}=0.3$;

718: i.e. the BR06 formula for pressure yields values that are typically a

719: factor $\sim 3$ larger than our measured midplane pressures, and a

720: factor $\sim 4$ below the mass-weighted mean values of pressure. 

721: As noted above, the mass-weighted average pressures are

722: about ten times larger than the midplane pressures; this is evident in

723: the ratio of the fitting coefficients.

724: 

725: Finally, we note that for most models (except at low $\Sigma$), the measured

726: midplane pressure exceeds the maximum pressure of the warm neutral

727: medium, $P_{w, {\rm max}}/\kB= 5.5\times 10^3 \K \cm^{-3}$ 

728: for our adopted heating and cooling functions.

729: Dense clouds that are externally confined by the warm medium cannot

730: have pressure exceeding $P_{w, {\rm max}}$ unless they are internally

731: stratified (implying they are self-gravitating);  thus, 

732: $P_{w, {\rm max}}$ is the largest the midplane pressure could be in the absence of

733: self-gravity.  

734: Equation (\ref{eq:Pnewth}) can be solved for $\Sigma$ in terms of the

735: midplane value of $P_{0, {\rm  th}}$.  The maximum surface density for an

736: atomic-only 

737: disk without self-gravitating clouds is then obtained by setting 

738: $P_{0,{\rm  th}}\rightarrow P_{w, {\rm max}}$, with the result

739: $\Sigma \rightarrow (P_{w, {\rm max}}/G)^{1/2}\sigma_z/c_s$ times a

740: function of $A$ that varies between 0.3 and 0.6 for $A= 0.1 - 1$.

741: Assuming  $\sigma_z/c_s \sim \sqrt{2}$ and taking 

742: $P_{w, {\rm max}}/\kB=5.5 \times 10^3 \K \cm^{-3}$, the maximum

743: surface density for a pure-atomic disk is 

744: $\sim 10 \msun\pc^{-2}$; this is consistent with the saturation

745: levels for HI gas observed e.g. by \cite{2002ApJ...569..157W}.

746: Since the measured midplane pressure is a

747: volume-weighted sum of the pressures in different phases, a mean value

748: exceeding $P_{w, {\rm max}}$ implies that self-gravitating dense clouds occupy

749: a non-negligible fraction of the midplane volume,

750: $f_V=(M_{\rm dense}/M_{\rm diffuse})(\rho_{\rm diffuse}/\rho_{\rm dense})$, with

751: $\langle P\rangle_{\rm midplane}= (P_{\rm dense}- P_{\rm diffuse})f_V+

752: P_{\rm diffuse}$. 

753: In the next section, we turn to a discussion of the

754: relationship between the dense-to-diffuse mass ratio and global parameters.

755: 

756: 

757: \section{An Application: Molecular Mass/Pressure Relations}

758: 

759: In this section, we explore relationships between the dense gas

760: fraction and ``environmental'' conditions, including the gas pressure

761: and the gas surface density.  We are motivated by observations that

762: show high molecular fractions in environments -- including spiral arms

763: and galactic center regions -- where both the total gas surface

764: density and stellar density are high.  In particular, BR06 found for a

765: number of disk systems that the mean ratio of molecular-to-atomic mass

766: scales nearly linearly with the pressure estimate $P_{\rm BR}$ defined

767: in equation (\ref{PBR_eq}).  Although our turbulent, multiphase

768: simulations show that $P_{\rm BR}$ in fact overestimates the pressure

769: of the typical volume element and underestimates the pressure of the

770: typical mass element, $P_{\rm BR}$ nevertheless systematically

771: increases in a similar way to both $\langle P\rangle_{\rm midplane}$

772: and $\langle P \rangle_{\rho}$.  Thus, it is interesting to test how

773: the dense-to-diffuse gas mass ratio depends on the true values of

774: pressure.  In addition to empirical results suggesting a relation

775: between mass ratio and pressure, there are theoretical reasons that

776: the mass ratio should depend on the mean gaseous surface density.  For

777: example, if

778: atomic gas is converted to molecular clouds through gravitational

779: instabilities on a timescale $t_{\rm form}\sim

780: \sigma_{\rm HI}/(G\Sigma)$, and molecular clouds are destroyed by star

781: formation on a timescale $t_{\rm dest}$, equating cloud formation and

782: destruction rates implies $M_{\rm H_2}/M_{\rm HI}=t_{\rm dest}/t_{\rm form}$,

783: which is $\propto \Sigma$ if the HI velocity dispersion and cloud

784: destruction time are relatively constant.  Thus, it is interesting to

785: explore dependence of $M_{\rm H_2}/M_{\rm HI}$ on the surface density -- which

786: appears in both the effective hydrostatic pressure and the rate of

787: self-gravitating instabilities.

788: 

789: \subsection{Molecular Gas

790: \label{mol_fraction_disc}

791: }

792: 

793: Although our numerical model does not directly include

794: formation/dissociation processes of H$_2$, we can nevertheless relate

795: our results to observed gas phases in an approximate way, using

796: density as a proxy.  Namely, we expect gravitationally

797: bound dense clouds at $n>100$ cm$^{-3}$ to consist primarily of H$_2$,

798: whereas diffuse gas at lower densities consists primarily of HI.

799: We argue for this approximate identification based on the

800: formation/dissociation equilibrium condition for 

801: $H_2$ molecules, which includes photodissociation and cosmic ray

802: dissociation, and formation on dust grains:

803: \begin{eqnarray}

804: (R_{\rm pump}&+&\zeta_{\rm CR}^{\rm H_2})nf_{\rm mol} 

805: = R_f n^2(1-f_{\rm mol})

806: \end{eqnarray}

807: \citep{1985ApJ...291..722T}.

808: Here, $f_{\rm mol}\equiv 2 n({\rm H}_2)/n$ is the molecular fraction, and

809: $1-f_{\rm mol}=n({\rm H I})/n$ is the atomic fraction.  

810: The FUV dissociation rate is limited by shielding, which depends on

811: the optical depth in H$_2$ lines and the extinction.  Formation on grains

812: depends on the sticking probability.  The details of the terms

813: involved are listed in Table \ref{tbl:param}.  We adopt 

814: FUV field strength $G_0=1.7$, 

815: gas and dust temperature $T=10$ K, and

816: cosmic-ray ionization rate of hydrogen atoms

817: $\zeta_{\rm CR}^{\rm H}=1.8\times 10^{-17}$s$^{-1}$.

818: For any total hydrogen column $N_{\rm H}$ and volume density $n$, we can

819: solve to obtain $f_{\rm mol}$, the molecular fraction.  

820: Figure \ref{fig:Molecular} shows, in the $n-N$ plane, the boundary

821: ({\it solid line}) between the predominantly-atomic and

822: predominantly-molecular regimes, which we define by the locus of

823: points for which $f_{\rm mol}=0.5$.   

824: 

825: 

826: \begin{table}

827: \begin{center}

828: \caption{Processes and parameters for H$_2$ formation/dissociation}

829: \label{tbl:param}

830: \begin{tabular}{lllll}

831: \hline\hline

832: Formation of H$_2$ on dust grains

833: &$R_f$&=& $6\times 10^{-17}(T/300)^{0.5}S(T)~{\rm cm^{3}s^{-1}}$ &

834: [1]\\

835: Sticking probability

836: &$S(T)$&=& $[1+0.04(T+T_d)^{0.5}+2\times 10^{-3}T +8\times

837:   10^{-6}T^2]^{-1}$&

838: [1]\\

839: Photo dissociation rate

840: &$R_{\rm pump}$ &=& $3.4\times 10^{-10}G_0\beta(\tau)\exp(-2.5A_v) ~ 

841: {\rm s^{-1}}$ & [1] \\

842: Self-shielding function

843: &$\beta(\tau)$ & & $\cdots$ & [1] \\ 

844: Optical depth

845: &$\tau$ &=& $1.2\times 10^{-14}fN_{\rm H}\delta v_d^{-1}$ & [1] \\

846: Cosmic-ray dissociation

847: &$\zeta_{\rm CR}^{\rm H_2}$&=& $2.29\zeta_{\rm CR}^{\rm H}$ & [2]\\

848: Turbulent line broadening

849: &$\delta v_d$ &=& 

850: $1~{\rm km/s}\left(\frac{N_{\rm H}/n}{1~\pc}\right)^{0.5}$ &

851: [3]

852: \\

853: Visual attenuation

854: &$A_v$ &$\equiv$& $N_{\rm H} /1.5\times 10^{21} \cm^{-2}$ &\\

855: \end{tabular}

856: \tablenotetext{}{[1] \cite{1985ApJ...291..722T}}

857: \tablenotetext{}{[2] \cite{1989ApJ...342..306H}}

858: \tablenotetext{}{[3] \cite{1987ApJ...319..730S}}

859: 

860: \end{center}

861: \end{table}

862: 

863: 

864: \begin{figure}

865: %

866: \epsscale{0.6}

867: \plotone{beta1.ps}

868:  \caption{Phases in the density-column density plane. The 

869: {\it solid line} divides the area into predominantly-HI at low 

870: $n_{\rm H}$ and

871: $N_{\rm H}$, and  predominantly-H$_2$ at high $n_{\rm H}$ and $N_{\rm H}$, adopting

872: molecule formation and destruction processes as described in the text.  

873: Shown as a {\it dashed line} is $N=n L_{\rm J}$, where $L_{\rm J}$ is the

874: Jeans length at $n$ 

875: for $T=10$K gas.  The left

876:  and right hand sides of this line are gravitationally unbound and

877:  bound, respectively. Dotted lines show the typical resolution limit of the

878:  simulations $\Delta x=2$ pc, and a maximum cloud scale of $100$ pc.

879: }

880: \label{fig:Molecular}

881: \end{figure}

882: %

883: 

884: 

885: 

886: At any density, we can also define the Jeans length $L_{\rm J}=c_s

887: (\pi/G\rho)^{1/2}$, where $c_s^2=\kB T/\mu$ (we adopt $T=10$K).  

888: This defines a corresponding total column

889: of gas, $N_{\rm H}=n L_{\rm J}$, that could be expected to be gravitationally

890: bound.  The boundary between gravitationally unbound (low $n$ and $N$)

891: and bound (high $n$ and $N$) gas, based on this criterion, is shown in

892: Figure \ref{fig:Molecular} as a {\it dashed} line with $n\propto N^2$.

893: Note that if instead of $L_{\rm J}$ we had chosen as a length scale the

894: diameter $D$ of sphere containing mass equal to the Bonnor-Ebert

895: \citep{1956MNRAS.116..351B,1957ZA.....42..263E} critical mass,

896: $M_{\rm BE}=1.182 c_s^3 /(G^3\rho)^{1/2}$, then $D=0.74 L_{\rm J}$. This would

897: shift the unbound/bound line in the $\log(N)-\log(n)$ plane to the

898: left by $\log(0.74)=-0.13$.  

899: 

900: We note that the gravitational binding criterion discussed above considers only

901: support by thermal pressure.  Turbulence can lend further support

902: against gravity, and this is particularly important for molecular gas,

903: which is quite cold.  For example, if we considered

904: turbulence-supported clouds with velocity dispersion following the

905: observed linewidth-size relation of Galactic GMCs

906: \citep{1987ApJ...319..730S}, then the column density separating

907: gravitationally bound from unbound regions would have a constant value

908: equal to half of the mean observed GMC column, amounting to $N_{\rm H}=7.5

909: \times 10^{21} \cm^{-2}$.  A higher normalization for the

910: linewidth-size relation (as occurs in galactic center regions; see

911: \citealt{2001ApJ...562..348O}) would further shift the unbound/bound

912: limit to larger $N$.  Thus, moderate-density molecular gas can in

913: principle be gravitationally unbound under conditions of sufficiently high

914: turbulence \citep{1993ApJ...411..170E}.  For our current simulations,

915: however, turbulence levels are not this high (see discussion below).

916: 

917: We have denoted the three different regions in the $\log(N)-\log(n)$

918: plane according to their expected chemical and gravitational

919: properties.  The crossing point of the two separation loci is at

920: $n\approx 125$ cm$^{-3}$ and 

921: $N_{\rm H}\approx 1.2\times 10^{21} \cm^{-2}$

922: (Av $\sim$ 0.75 mag), with corresponding local Jeans length of 

923: $L_{\rm J}=3.1$ pc.  

924: This size is in fact slightly larger than

925: the typical resolution limit of our simulations, $\Delta x =2$pc; we show this

926: limit in Figure \ref{fig:Molecular} as a {\it dotted} line, with

927: regions to the right resolved and those to the left below the

928: resolution limit.  The resolution limit crosses the HI/H$_2$ separation

929: curve at 

930: $n=180~{\rm cm}^{-3}$.

931: Because the resolution limit falls at larger $N$ than the bound/unbound

932: separation nearly everywhere in the molecular domain, all zones at a given

933: density that are resolved and molecular would also be gravitationally

934: bound.  In practice, clouds do not exceed $\sim 100$ pc in crossection; we have

935: marked this limit in the Figure as a {\it dotted} line.  

936: 

937: 

938: According to the limits shown in Figure \ref{fig:Molecular}, any resolved

939: regions in our simulations at $n>100\cm^{-3}$ would be molecular.

940: This is a conservative definition, since it omits some gas between

941: $n\sim 10-100$ with  $N_{\rm H} > 10^{21} \cm^{-2}$

942: that could be molecular.  However, gas at these

943: densities could also be in the cold atomic phase (which extends down

944: to $n_{\rm cold, min}= 8.6 ~\cm^{-3}$ for the cooling curve we

945: adopt); we choose

946: the stricter definition.  We note that when the virial ratio

947: ($\sim$kinetic/gravitational energy; see \citeauthor{Paper1}) is

948: measured for gas in the range $n=1-100\cm^{-3}$ (most of which is at

949: $10\cm^{-3}<n$), the values are well above unity -- implying that gas

950: parcels in this density range are mostly found in non-self-gravitating

951: regions with low surrounding

952: column densities, to the left of the unbound/bound curve.  

953: This suggests that in practice very little

954: high-column density gas that would be molecular is missed when we set

955: the minimum threshold at $100\cm^{-3}$.  From the point of view of

956: dynamics, this is because the density rises whenever any

957: region becomes gravitationally bound, so low-density  regions at

958: high column are rapidly depopulated.  

959: We also note that the H$_2$

960: formation time in dense, cold regions is expected to be short ($\sim

961: 10^6$ years from \citealt{2007ApJ...659.1317G}), because supersonic

962: shocks increase the density above ambient values 

963: and accelerate the molecule formation process, which occurs at a rate

964: $n R_f$.

965: 

966: \subsection{Molecular Mass-Pressure Relation}

967: 

968: 

969: \begin{figure}

970: \epsscale{1.0}

971: \plottwo{Q2H2Pressure.ps}{K2H2Pressure.ps}

972: 

973: \plottwo{R2H2Pressure.ps}{S2H2Pressure.ps}

974:  \caption{Mean molecular-to-atomic mass ratio $R_{\rm mol}$ as a

975:    function of $P_{\rm BR}$ (see eq. \ref{PBR_eq}), shown as {\it open

976:    boxes} for all hydrodynamic models.  {\it Filled boxes} show the

977:    results from hydrostatic models.  The {\it solid line} shows the

978:    empirical fit from BR06 (eq. \ref{eq:Rmol}).}

979: \label{fig:H2P}

980: \end{figure}

981: 

982: Following the discussion in the previous section, we adopt a working

983: definition of molecular gas as that at $n_{\rm H}>100 \cm^{-3}$.  Atomic gas

984: therefore consists of the lower-density complement, including both

985: what would be observable as warm and as cold HI in 21 cm emission.

986: The mass ratio of molecular to atomic hydrogen is

987:  then defined as

988: \begin{eqnarray}

989: R_{\rm mol}\equiv 

990: \frac{M(n>100\, {\rm cm}^{-3})}{M(n<100\, {\rm cm}^{-3})},

991: \end{eqnarray}

992: where we apply space- and time-averages before taking the ratio.

993: 

994: 

995: Figure \ref{fig:H2P} shows $R_{\rm mol}$ as a function of 

996: $P_{\rm BR}=

997: \Sigma v \sqrt{2G\rho_{\ast}}$ for all

998: hydrodynamic and hydrostatic Series (we use $v=8\kms$ as in BR06).  

999: We also show as a {\it solid

1000:   line} the empirical fitting formula from the observational study of

1001: BR06 (see their eq.13):

1002: \begin{equation}

1003: R_{\rm mol}=

1004: \left[\frac{P_{\rm BR}/\kB}{4.3\times 10^4 \K \cm^{-3}} \right]^{0.92}. 

1005: \label{eq:Rmol}

1006: \end{equation}

1007: Interestingly, we find that our results for $R_{\rm mol}$ follow the

1008: empirical result for some but not all series.  In particular, the

1009: models in Series Q and R -- which have values of $\Omega$ that scale

1010: with $\Sigma$ in such a way as to keep the gaseous Toomre parameter

1011: constant -- are close to the BR06 fit.  The models in Series K, which

1012: have constant $\kappa$ and therefore high (or low) values of

1013: $\kappa/\Sigma$ where $P_{\rm BR}$ is low (or high, respectively), do

1014: not follow the empirical result of BR06, but instead show a ratio

1015: $R_{\rm mol}$ that is near unity independent of $P_{\rm BR}$.  This

1016: has two interesting implications. First, our models with $\Omega

1017: \propto \Sigma$ have similar behavior to real galaxies, indicating

1018: that real systems evolve (by converting their gas to stars) in such a

1019: way as to have Toomre parameter fall within a limited range of values.

1020: Second, because the K Series departs from the BR06 result, our models

1021: suggest that the molecular fraction does not have a one-to-one

1022: relationship to the effective pressure parameter $P_{\rm BR}$.  

1023: Comparing series Q and K which have the same $\Sigma$ and

1024: $\rho_{\ast}$, $R_{\rm mol}$ increases with increasing $\kappa$.  For

1025: example, Figure 8 shows that $R_{\rm mol}$ increases by factor 2.4

1026: when $\Omega$ (and $\kappa$) increases by factor $2\sqrt{2}$, for the

1027: highest-$\Sigma$ ($\Sigma= 42 \msun \pc^{-2}$) model. For the 

1028: $\Sigma= 21 \msun \pc^{-2}$ model, the $R_{\rm mol}$ increase is 60\% 

1029: for an $\Omega$ increase by a factor  $\sqrt{2}$, comparing the Q and

1030: K series. 

1031: Series S, which varies $\kappa$ at a given value of

1032: $\Sigma$ and $\rho_{\ast}$, also shows departures from the empirical 

1033: $R_{\rm mol}$ vs. $P_{\rm BR}$ relation. 

1034: 

1035: Given

1036: that molecular gas in our models is primarily found in

1037: gravitationally-bound systems, it in fact makes sense that the

1038: molecular fraction should not have a one-to-one relationship to the

1039: parameter $P_{\rm BR}$, since $P_{\rm BR}$ does not include any

1040: effects of galactic rotation.  Galactic rotation and shear are crucial

1041: for regulating the large-scale gravitational instabilities that create

1042: giant molecular clouds in real galaxies as well as in our models, so

1043: we believe that the molecular-to-atomic ratio must intrinsically be

1044: sensitive to environmental factors that are not captured in $P_{\rm

1045: BR}$.  Thus, if observed galaxies {\it do} show a one-to-one relation between 

1046: $R_{\rm mol}$ and $P_{\rm BR}$, it implies that the environmental

1047: parameters $\kappa$, $\Sigma$, and $\rho_{\ast}$ are not all independent in

1048: real systems. 

1049: 

1050: \begin{figure}

1051: \epsscale{1.0}

1052: \plottwo{Q2H2vsSigma.ps}{K2H2vsSigma.ps}

1053: 

1054: \plottwo{R2H2vsSigma.ps}{S2H2vsSigma.ps}

1055:  \caption{

1056: Molecular-to-atomic mass ratio $R_{\rm mol}$ vs.

1057: total gaseous surface density $\Sigma$. {\it Open boxes} show the

1058: results from hydrodynamic models, and {\it filled boxes} show the 

1059: hydrostatic model results (HSP for Series K and Q,

1060: HSC for Series R). The {\it solid line} indicates the empirical result

1061: from BR (eq. \ref{eq:Rmol} using eq. \ref{PBR_eq}).}

1062: \label{fig:H2S}

1063: \end{figure}

1064: 

1065: 

1066: In Figure \ref{fig:H2S}, we show $R_{\rm mol}$ as a function of the

1067: surface density for all of our model Series.  The behavior is similar

1068: to that shown in Figure \ref{fig:H2P} because $P_{\rm BR}$ depends

1069: monotonically on $\Sigma$ for all our Series: $P_{\rm BR}\propto

1070: \Sigma^2$ for Series Q and K (which have $\rho_{\ast}\propto \Sigma^2$),

1071: while $P_{\rm BR}\propto \Sigma$ for Series R (which has

1072: $\rho_{\ast}=const.$).  From both Figures Figure \ref{fig:H2P} and

1073: \ref{fig:H2S}, it is evident that the hydrostatic models ({\it filled

1074: boxes}) generally have a much larger molecular component than both the

1075: hydrodynamic models and the empirical results, except at low gaseous

1076: surface density.  This indicates that turbulence is essential for

1077: determining the phase balance between diffuse and dense gas in the ISM

1078: as a whole.  If the ISM were a static system, it would be

1079: overwhelmingly molecular even at fairly moderate values of $\Sigma$

1080: and $\rho_{\ast}$.  In real galaxies, turbulence limits gaseous settling

1081: into the midplane and the extreme self-compression that would

1082: otherwise ensue.

1083: 

1084: 

1085: \begin{figure}

1086: \epsscale{1.0}

1087: \plottwo{Q2H2Pr.ps}{K2H2Pr.ps}

1088: 

1089: \plottwo{R2H2Pr.ps}{S2H2Pr.ps}

1090:  \caption{Molecular-to-atomic mass ratio $R_{\rm mol}$ vs. measured

1091:    mass-weighted mean pressure $\langle P\rangle_\rho$ 

1092:    and midplane pressure $\langle P\rangle_{\rm midplane}$ for all  

1093:  hydrodynamic models.  Lines show linear fits (see text).}

1094: \label{fig:H2Pressure}

1095: \end{figure}

1096: 

1097: While we have argued that the molecular content of a galaxy cannot (in

1098: general) be predicted solely from $\Sigma$ and $\rho_{\ast}$ because

1099: self-gravitating horizontal contraction is also responsive to the

1100: local rotation and shear rates, it still is plausible that the

1101: molecular fraction should reflect the {\it true} mean pressure in the

1102: ISM.  If molecular gas is collected in self-gravitating clouds, then

1103: since their internal pressure is higher than ambient levels, an

1104: increase in the molecular fraction should go hand-in-hand with a

1105: higher mass-weighted mean pressure $\langle P\rangle_\rho$.  At the

1106: same time, ambient midplane pressures $\langle P\rangle_{\rm

1107: midplane}$ increase when the total gas surface density increases, and

1108: (provided that $\kappa$ is low enough)

1109: larger $\Sigma$ also renders the disk susceptible to gravitational

1110: instabilities 

1111: that would form dense, bound clouds and increase the

1112: molecular fraction.  

1113: 

1114: We explore these ideas by plotting in Figure

1115: \ref{fig:H2Pressure} the molecular-to-atomic ratio against our two

1116: measures of mean gas pressure, $\langle P\rangle_\rho$ and $\langle

1117: P\rangle_{\rm midplane}$.  We also fit the combined results for all

1118: Series to single linear relations. These fits, overplotted in Figure

1119: \ref{fig:H2Pressure} as {\it solid} and {\it dotted} lines, are

1120: $R_{\rm mol}=\langle P\rangle_\rho/[7.6 \times 10^4 \cm^{-3} \K \kB]$ 

1121: and $R_{\rm mol}=\langle P\rangle_{\rm midplane}/[6.7 \times 10^3

1122: \cm^{-3}\K \kB]$.  

1123: For Series Q and R, the fits using $\langle P\rangle_\rho$ are quite

1124: good, and the fits using $\langle P\rangle_{\rm midplane}$ are also

1125: fairly good (Series Q and R also show better agreement with empirical

1126: results than the other Series).  For Series K, the fit using $\langle

1127: P\rangle_\rho$ is reasonably close to the models results, but the fit

1128: based on $\langle P\rangle_{\rm midplane}$ fails in a similar manner

1129: to that shown in Figure (\ref{fig:H2P}) and discussed above.  The

1130: basic reason for this is that the midplane pressure, either directly

1131: measured or estimated using equation (\ref{eq:Pnewth}), increases with

1132: increasing $\Sigma$. However, the molecular-to-atomic ratio for Series

1133: K does not strongly and secularly increase with $\Sigma$ due to the

1134: differences in rotational effects in this constant-$\kappa$ Series

1135: compared to the other Series.  At high values of $\Sigma$ in Series K,

1136: the disk is extremely gravitationally unstable overall, and as a

1137: consequence is more active in producing feedback than other models at

1138: the same $\Sigma$.  As a consequence, a smaller fraction of the gas

1139: mass ends up being in the dense phase than in Series Q and R.

1140: Overall, we conclude that $R_{\rm mol}$ is indeed well correlated with

1141: the mass-weighted mean pressure, $\langle P\rangle_\rho$, as (almost

1142: definitionally) is expected.  The measured mean midplane pressure, which is 

1143: more closely related to simple vertical-equilibrium

1144: pressure estimates, is less well correlated with $R_{\rm

1145: mol}$ when environmental parameters $\kappa$, $\rho_{\ast}$, and $\Sigma$

1146: are all independent.

1147: 

1148: 

1149: \section{Summary and Discussion}

1150: 

1151: We have used numerical simulations of turbulent, multiphase,

1152: self-gravitating gas orbiting in the disks of model galaxies to study

1153: the relationships among pressure, the vertical distribution of gas,

1154: and the relative proportions of dense and diffuse gas.  In particular,

1155: we compare the results on vertical stratification obtained from

1156: space-time averages of fully-dynamic -- and often turbulence-dominated

1157: -- systems with simple estimates based on single-component effective 

1158: hydrostatic

1159: equilibria.  We also investigate how vertical-equilibrium estimates

1160: for the pressure compare with measured mean values of the pressure in

1161: our models.  Empirical studies by BR06 have identified a linear

1162: relation between the molecular-to-atomic mass ratio $R_{\rm mol}$ and

1163: a midplane ISM pressure estimate, $P_{\rm BR} \propto \Sigma \sqrt{

1164: \rho_{\ast}}$.  We study the origin and implications of this relation by

1165: testing the correlations among $R_{\rm mol}$, $P_{\rm BR}$, and the

1166: directly-measured midplane and mean pressures in our models.

1167: 

1168: 

1169: Our chief conclusions, and their implications, are as follows:

1170: 

1171: 

1172: 1. The average disk scale height is well represented by estimates that

1173: assume hydrostatic equilibrium and an effective total pressure based

1174: on the total (thermal + turbulent) vertical velocity dispersion (see

1175: Fig. \ref{fig:Have} and eq. \ref{Hest_eq}).  Thus, provided that gas

1176: surface densities, vertical velocity dispersions, and stellar density

1177: can be measured, an accurate estimate for the disk thickness can be

1178: obtained.

1179: 

1180: Hydrostatic equilibrium with an effective turbulent pressure is

1181: commonly assumed in both Galactic and extragalactic observational

1182: studies (e.g.

1183: \citealt{1991ApJ...382..182L,1994ApJ...433..687M,1995ApJ...448..138M,

1184: 1997A&A...326..554C,2000MNRAS.311..361O,

1185: 2002A&A...394...89N,2004ApJ...608..189D,

1186: 2004ApJ...612L..29B,2006ApJ...650..933B,

1187: 2008AstL...34..152K}), but to our knowledge the relations that are

1188: generally adopted have not previously been tested with direct

1189: numerical simulations.  Our hydrodynamic studies demonstrate that for

1190: determining the scale height $H$, the effective hydrostatic

1191: equilibrium assumption is indeed sufficient, even when turbulent

1192: support far exceeds thermal support (and provided that magnetic

1193: effects are sub-dominant; see below).  Thus, measured disk thicknesses

1194: in edge-on disk galaxies could in principle be used to determine the

1195: unobservable vertical velocity dispersion, and measured line-of-sight

1196: velocity dispersions in face-on galaxies could be used to determine

1197: the unobservable disk thickness.

1198: 

1199: 

1200: The basic reason the hydrostatic formula can be

1201: used to obtain an accurate measure of $H$ is that what is

1202: really being equated is the total vertical momentum flux 

1203: $\rho (\kB T/\mu + v_z^2)$ averaged over the midplane, and the total

1204: vertical weight of the ISM, $\int dz\, \rho\, g_z \sim \rho 4 \pi G (\rho

1205: + \rho_{\ast}) H^2$, averaged over the horizontal direction.

1206: Provided that the time-averaged value of the momentum per unit volume

1207: in the midplane does not change, momentum conservation including

1208: gravitational source terms demands that the difference between 

1209: vertical momentum flux and vertical weight must be zero, 

1210: independent of details of the dynamics.  The formula

1211: $H^2\approx \sigma_z^2/[4\pi G (\rho_{\ast} + \Sigma/H\sqrt{2\pi})]$ is

1212: therefore fundamentally an expression of momentum conservation.

1213: 

1214: 

1215: 2. Mass-weighted mean pressures $\langle P \rangle_{\rho}$ in our

1216: hydrodynamic models significantly differ from the mean midplane

1217: pressure $\langle P\rangle_{\rm midplane}$, while these quantities

1218: are quite similar to each other in our static comparison models.

1219: Typically, the hydrodynamic models yield values of $\langle P

1220: \rangle_{\rho}$ an order of magnitude larger than $\langle

1221: P\rangle_{\rm midplane}$. The difference can be attributed to

1222: self-gravitating condensation, which makes concentrated clouds 

1223: with high internal pressure

1224: rather

1225: than a horizontally-uniform gas distribution

1226: with more moderate pressure.

1227: 

1228: Simple estimates of the pressure based on vertical hydrostatic

1229: equilibrium fall between mass-weighted and midplane values, with the

1230: formula used by BR06 (see our eq. \ref{PBR_eq}) comparable to the

1231: geometric mean $P_{\rm BR} \sim \sqrt{\langle P \rangle_{\rho} \langle

1232: P\rangle_{\rm midplane}}$. A single-component estimate for the

1233: midplane thermal pressure that accounts for self-gravity and the mean

1234: thermal and turbulent velocity dispersions (see eq. \ref{eq:Pnewth})

1235: follows the measured midplane pressure fairly closely, especially at

1236: high $\Sigma$.  Thus, if turbulent and thermal vertical velocity

1237: dispersions can be measured directly (for face-on galaxies), a good

1238: estimate of the midplane total or thermal pressure can be computed via

1239: equation (\ref{eq:Pnew}) or (\ref{eq:Pnewth}).  For an edge-on system

1240: in which the scale height is measured, the midplane total or thermal

1241: pressure can be estimated as $\Sigma/(H \sqrt{2\pi})\times \langle

1242: \sigma_z^2\rangle \ {\rm or}\ \langle c_s^2\rangle$.  Midplane

1243: pressure estimates based on large-scale observables that assume

1244: hydrostatic equilibrium can be quite accurate, but this depends on an

1245: accurate measure of the vertical velocity dispersion or vertical

1246: thickness.  Even if the velocity dispersion is not known, the {\it

1247: relative} midplane pressures of different regions within a galaxy (or

1248: from one galaxy to another) can be obtained using the hydrostatic

1249: formulae, provided the variation in the (unknown) velocity dispersion

1250: within the observational sample is small compared to the variation in

1251: the stellar volume and gaseous surface densities.  Midplane pressure

1252: estimates made in this way should not, however, be treated as a proxy

1253: for the pressure in the typical mass element, $\langle P

1254: \rangle_{\rho}$, which can be much larger than the pressure in the

1255: typical volume element.

1256: 

1257: 

1258: 

1259: 

1260: 

1261: 

1262: 3. Based on calculations of molecular abundance as a function of

1263: hydrogen volume density $n$ and column density $N$ combined with the

1264: resolution and measured virial ratios in our simulations, we adopt a

1265: working definition of gas at $n\ge 100 \cm^{-3}$ as ``molecular'' and

1266: $n<100 \cm^{-3}$ as ``atomic''.  We then investigate the ratio $R_{\rm

1267:   mol}=M_{\rm H_2}/M_{\rm HI}$ for all our models.  We find that 

1268: Series Q and R, which have rotation rate $\Omega \propto

1269: \Sigma$, show correlations between $R_{\rm mol}$ and $P_{\rm BR}$

1270: (or $R_{\rm mol}$ and $\Sigma$) that are similar to the empirical 

1271: result reported

1272: by BR06, $R_{\rm mol}\propto P_{\rm BR} \propto

1273: \Sigma\sqrt{\rho_{\ast}}$. On the other hand, Series K and S, in which $\Sigma$ and

1274: $\Omega$ do not vary together, depart from the empirical relation

1275: $R_{\rm mol} \propto P_{\rm BR}$.

1276: 

1277: We conclude that (i) the molecule fraction inherently depends on the

1278: rotational state of a galactic disk, not just on the local values of

1279: the stellar volume and gaseous surface densities $\rho_{\ast}$ and

1280: $\Sigma$; and (ii) the empirical relation $R_{\rm mol} \propto P_{\rm

1281: BR}$ identified by BR06 implies that the third ``environmental

1282: parameter,'' the epicyclic frequency $\kappa = \sqrt{2} \Omega$

1283: (assuming a flat rotation curve), is not independent of $\rho_{\ast}$ and

1284: $\Sigma$ in real galaxies.  This dependence can be accomplished by

1285: evolution: for example, disk galaxies may convert gas into stars until

1286: they reach a state in which the Toomre parameter $\propto

1287: \kappa/\Sigma$ approaches a critical value.

1288: 

1289: 4. We have tested the correlation between $R_{\rm mol}$ and the

1290: measured pressures in our models, $\langle P \rangle_{\rm midplane}$

1291: and $\langle P \rangle_\rho$, and find a good correlation in all

1292: Series only for the latter.  The correlation between $\langle P

1293: \rangle_\rho$ and $R_{\rm mol}$ is potentially useful as a way to

1294: estimate the typical internal pressure within gravitationally-bound

1295: regions when only the total molecular-to-atomic mass is easily

1296: accessible, as for low-resolution observations.  This internal

1297: pressure is important in the small-scale aspects of star formation

1298: such as determining the IMF \citep{2007ARA&A..45..565M}, as well as in

1299: molecular chemistry.  The lack of correlation between $R_{\rm mol}$

1300: and $\langle P \rangle_{\rm midplane}$ in Series K implies that the

1301: molecular content cannot in general be predicted solely from $\Sigma$

1302: and $\rho_{\ast}$ (i.e. without knowledge of $\kappa$), as noted above.

1303: This reflects the fact that the formation of self-gravitating clouds

1304: is regulated not just by gravitational processes and pressure, but

1305: also by angular momentum.

1306: 

1307: 

1308: 5.  For our non-turbulent comparison models, we find that $R_{\rm

1309:   mol}$ far exceeds observed values.  This indicates that turbulence

1310: is essential to setting the observed phase balance in the ISM.

1311: Recent theoretical investigations of the origin of Kennicutt-Schmidt

1312: laws have focused on the dependence of star formation rates on the

1313: molecular, rather than total, gas surface density

1314: (e.g. \citealt

1315: {2007arXiv0711.1361N,2008ApJ...680.1083R}).  Since turbulence is crucial in

1316: determining the abundance of dense gas, in simulations that aim to

1317: compute this abundance realistically it is necessary to incorporate

1318: the feedback effects that drive turbulence, and to run on a fine

1319: enough mesh (or with sufficient SPH particles) that the turbulence is well

1320: resolved. While technically challenging in global disk models, local

1321: models may offer a more immediate route to this goal.

1322: 

1323: 

1324: {\it Caveats --} The models analyzed in this paper are subject to a

1325: number of limitations, which could potentially affect some of our

1326: conclusions. The chief limitations of the simulations are that (i)

1327: they are two-dimensional, representing cuts in the $R-z$ plane, rather

1328: than three-dimensional; (ii) we have adopted a very simple model to

1329: implement turbulent driving as a star formation feedback effect from

1330: HII regions, and we have not included other drivers of turbulence such

1331: as supernovae, spiral shocks, and shear instabilities; (iii) we have

1332: not included magnetic fields (or cosmic rays).  

1333: We intend to pursue these extensions in future work.

1334: 

1335: Inclusion of magnetic fields and altered turbulent driving would

1336: certainly affect the specific quantitative findings for $H_{\rm ave}$,

1337: $\langle P \rangle_{\rm midplane}$, $\langle P \rangle_\rho$, and

1338: $R_{\rm mol}$ in our models.  We believe, however, that the results we

1339: have emphasized regarding physical {\it relationships} are robust.  In

1340: particular, with appropriate modifications to include magnetic

1341: stresses, the time-averaged vertical momentum flux through the

1342: midplane must still equal

1343: the time-averaged vertical weight if the mean vertical momentum is

1344: conserved.  This can be used to predict the total

1345: midplane pressure (including the magnetic pressure) and $H$ given the

1346: values of $\Sigma$, $\rho_{\ast}$, and the thermal, turbulent, and

1347: Alfv\'en velocities. Thus, we anticipate that inclusion of magnetic

1348: fields and alternate turbulence sources would not fundamentally alter

1349: the conclusion that reasonable estimates of scale heights can be

1350: made using observable quantities even in highly-dynamic systems.

1351:   

1352: 

1353: 

1354: Further, we expect that our conclusions regarding the presence or

1355: absence of correlations between $R_{\rm mol}$ and $\langle P

1356: \rangle_\rho$ or $\langle P \rangle_{\rm midplane}$ would continue to

1357: hold in models that include additional turbulence sources and magnetic

1358: fields, although the details of correlations might change.  Namely,

1359: angular momentum inherently must be important in permitting or

1360: preventing formation of dense, self-gravitating clouds.  Our present

1361: models account for angular momentum effects, and show that $R_{\rm

1362: mol}$ does not in general have a one-to-one relationship with $\langle

1363: P \rangle_{\rm midplane}$ or $\Sigma \sqrt{\rho_{\ast}}$; we expect this

1364: result would carry over into any model that incorporates sheared

1365: background rotation of the galactic disk.  Thus, if a one-to-one

1366: relationship between $R_{\rm mol}$ and $\Sigma \sqrt{\rho_{\ast}}$ indeed

1367: exists empirically, it implies that $\Sigma$, $\rho_{\ast}$, and $\kappa$

1368: are not all independent quantities in real galaxies.

1369: 

1370: 

1371: 

1372: We are grateful to the referee for a number of comments that have helped 

1373: improve our presentation.  Numerical computations used in this project

1374: were carried out on the OIT High Performance Computing Cluster, and

1375: the CTC cluster in the Department of Astronomy, at the University of

1376: Maryland.  This work was supported by grant NNG05GG43G from NASA.

1377: 

1378: 

1379: \bibliographystyle{apj}

1380: 

1381: \begin{thebibliography}

1382: 

1383: \bibitem[Blitz \& Rosolowsky(2004)]{2004ApJ...612L..29B} 

1384: Blitz, L., \& Rosolowsky, E.\ 2004, \apjl, 612, L29 

1385: 

1386: \bibitem[Blitz \& Rosolowsky(2006)]{2006ApJ...650..933B}

1387: ---. 2006, \apj, 650, 933

1388: 

1389: \bibitem[Bonnor(1956)]{1956MNRAS.116..351B} Bonnor, W.~B.\ 1956, \mnras, 

1390: 116, 351

1391: 

1392: 

1393: \bibitem[Combes 

1394: \& Becquaert(1997)]{1997A&A...326..554C} Combes, F., \& Becquaert, J.-F.\ 1997, \aap, 326, 554 

1395: 

1396: 

1397: \bibitem[Dalcanton et al.(2004)]{2004ApJ...608..189D} Dalcanton, J.~J., 

1398: Yoachim, P., \& Bernstein, R.~A.\ 2004, \apj, 608, 189

1399: 

1400: \bibitem[Dobbs(2008)]{2008MNRAS.tmp.1249D} Dobbs, C.~L.\ 2008, \mnras, 1249 

1401: 

1402: 

1403: 

1404: \bibitem[{Dobbs} et al. (2008)]{2008MNRAS.389.1097D}

1405: {Dobbs}, C.~L., {Glover}, S.~C.~O., {Clark}, P.~C., \& 

1406: 	{Klessen}, R.~S. 2008, \mnras, 389, 1097

1407: 

1408: 

1409: \bibitem[Ebert(1957)]{1957ZA.....42..263E} Ebert, R.\ 1957, Zeitschrift fur 

1410: Astrophysik, 42, 263

1411: 

1412: \bibitem[Elmegreen(1993)]{1993ApJ...411..170E} Elmegreen, B.~G.\ 1993, 

1413: \apj, 411, 170

1414: 

1415: \bibitem[Glover \& Mac Low (2007a)]{2007ApJS..169..239G}

1416: {Glover}, S.~C.~O., \& {Mac Low}, M.-M. 2007a, \apjs, 169, 239

1417: 

1418: \bibitem[Glover \& Mac Low (2007b)]{2007ApJ...659.1317G} 

1419: ---. 2007b, \apj, 659, 1317 

1420: 

1421: \bibitem[{Heitsch} et al. (2005)]{2005ApJ...633L.113H}

1422: {Heitsch}, F., {Burkert}, A., {Hartmann}, L.~W., {Slyz}, A.~D., \& 

1423: 	{Devriendt}, J.~E.~G. 2005, \apjl, 633, L113

1424: 

1425: \bibitem[{Heitsch} et al. (2006)]{2006ApJ...648.1052H}

1426: {Heitsch}, F., {Slyz}, A.~D., {Devriendt}, J.~E.~G., 

1427: 	{Hartmann}, L.~W., \& {Burkert}, A. 2006, \apj, 648, 1052

1428: 

1429: 

1430: \bibitem[{Hennebelle} \& {Audit} (2007)]{2007A&A...465..431H}

1431: {Hennebelle}, P., \& {Audit}, E. 2007, \aap, 465, 431

1432: 

1433: \bibitem[{Hennebelle} et al. (2007)]{2007A&A...465..445H}

1434: {Hennebelle}, P., {Audit}, E.,\& {Miville-Desch{\^e}nes}, M.-A. 2007,

1435:  \aap, 465, 445

1436: 

1437: \bibitem[{Hennebelle} et al. (2008)]{2008A&A...486L..43H}

1438: {Hennebelle}, P., {Banerjee}, R., {V{\'a}zquez-Semadeni}, E., 

1439: 	{Klessen}, R.~S., \& {Audit}, E. 2008, \aap, 486, L43

1440: 

1441: 

1442: \bibitem[{{Hollenbach} \& {McKee}(1989)}]{1989ApJ...342..306H}

1443: {Hollenbach}, D. \& {McKee}, C.~F. 1989, \apj, 342, 306

1444: 

1445: 

1446: 

1447: \bibitem[Kasparova 

1448: \& Zasov(2008)]{2008AstL...34..152K} Kasparova, A.~V., \& Zasov, A.~V.\ 2008, Astronomy Letters, 34, 152 

1449: 

1450: \bibitem[Kim et al.(2008)]{2008ApJ...681.1148K} Kim, C.-G., Kim, W.-T., 

1451: \& Ostriker, E.~C.\ 2008, \apj, 681, 1148 

1452: 

1453: 

1454: 

1455: 

1456: \bibitem[{{Kim} \& {Ostriker}(2001)}]{2001ApJ...559...70K}

1457: {Kim}, W.-T. \& {Ostriker}, E.~C. 2001, \apj, 559, 70

1458: 

1459: 

1460: \bibitem[Kim \& Ostriker(2002)]{2002ApJ...570..132K} Kim, W.-T., \&

1461: Ostriker, E.~C.\ 2002, \apj, 570, 132

1462: 

1463: 

1464: 

1465: \bibitem[{{Kim} \& {Ostriker}(2006)}]{2006ApJ...646..213K}

1466: ---. 2006, \apj, 646, 213

1467: 

1468: \bibitem[{{Kim} \& {Ostriker}(2007)}]{2007ApJ...660.1232K}

1469: ---. 2007, \apj, 660, 1232

1470: 

1471: \bibitem[Kim, Ostriker, \& Stone(2002)]{2002ApJ...581.1080K}

1472: {Kim}, W.-T., {Ostriker}, E.~C., \& {Stone}, J.~M. 2002, \apj, 581, 1080

1473: 

1474: \bibitem[Kim, Ostriker, \& Stone(2003)]{2003ApJ...599.1157K} Kim, W.-T., Ostriker, 

1475: E.~C., \& Stone, J.~M.\ 2003, \apj, 599, 1157 

1476: 

1477: 

1478: 

1479: \bibitem[Klessen et al.(2000)]{2000ApJ...535..887K} Klessen, R.~S., 

1480: Heitsch, F., \& Mac Low, M.-M.\ 2000, \apj, 535, 887 

1481: 

1482: \bibitem[{{Koyama} \& {Inutsuka}(2002)}]{2002ApJ...564L..97K}

1483: {Koyama}, H. \& {Inutsuka}, S. 2002, \apjl, 564, L97

1484: 

1485: 

1486: \bibitem[{Paper I}(2008)]{Paper1} Koyama, H. and Ostriker,

1487: E. C. 2008, submitted (Paper I)

1488: 

1489: \bibitem[{{Li} {et~al.}(2005){Li}, {Mac Low}, \&

1490:   {Klessen}}]{2005ApJ...620L..19L}

1491: {Li}, Y., {Mac Low}, M.-M., \& {Klessen}, R.~S. 2005, \apjl, 620, L19

1492: 

1493: \bibitem[{Li} {et~al.}(2006)]{2006ApJ...639..879L} 

1494: ---. 2006, \apj, 639, 879 

1495: 

1496: 

1497: \bibitem[Lockman 

1498: \& Gehman(1991)]{1991ApJ...382..182L} Lockman, F.~J., \& Gehman, C.~S.\ 1991, \apj, 382, 182 

1499: 

1500: \bibitem[Mac Low(1999)]{1999ApJ...524..169M} Mac Low, M.-M.\ 1999, \apj, 

1501: 524, 169

1502: 

1503: 

1504: \bibitem[Malhotra(1994)]{1994ApJ...433..687M} Malhotra, S.\ 1994, \apj, 

1505: 433, 687

1506: 

1507: 

1508: \bibitem[Malhotra(1995)]{1995ApJ...448..138M} ---.\ 1995, \apj, 

1509: 448, 138

1510: 

1511: \bibitem[McKee 

1512: \& Ostriker(2007)]{2007ARA&A..45..565M} McKee, C.~F., \& Ostriker, E.~C.\ 2007, \araa, 45, 565 

1513: 

1514: 

1515: 

1516: \bibitem[Narayan 

1517: \& Jog(2002)]{2002A&A...394...89N} Narayan, C.~A., \& Jog, C.~J.\ 2002, \aap, 394, 89 

1518: 

1519: \bibitem[Narayanan et al.(2007)]{2007arXiv0711.1361N} Narayanan, D., Cox, 

1520: T.~J., Shirley, Y., Dave, R., Hernquist, L., 

1521: \& Walker, C.~K.\ 2007, ArXiv e-prints, 711, arXiv:0711.1361 

1522: 

1523: 

1524: 

1525: \bibitem[Oka et al.(2001)]{2001ApJ...562..348O} Oka, T., Hasegawa, T., 

1526: Sato, F., Tsuboi, M., Miyazaki, A., \& Sugimoto, M.\ 2001, \apj, 562, 348 

1527: 

1528: 

1529: \bibitem[Olling 

1530: \& Merrifield(2000)]{2000MNRAS.311..361O} Olling, R.~P., \& Merrifield, M.~R.\ 2000, \mnras, 311, 361 

1531: 

1532: 

1533: 

1534: \bibitem[Piontek 

1535: \& Ostriker(2007)]{2007ApJ...663..183P} Piontek, R.~A., \& Ostriker, E.~C.\ 2007, \apj, 663, 183 

1536: 

1537: 

1538: \bibitem[Robertson 

1539: \& Kravtsov(2008)]{2008ApJ...680.1083R} Robertson, B.~E., \& Kravtsov, A.~V.\ 2008, \apj, 680, 1083 

1540: 

1541: 

1542: \bibitem[Solomon et al.(1987)]{1987ApJ...319..730S} Solomon, P.~M., Rivolo, 

1543: A.~R., Barrett, J., \& Yahil, A.\ 1987, \apj, 319, 730 

1544: 

1545: 

1546: \bibitem[Stone et al.(1998)]{1998ApJ...508L..99S} Stone, J.~M., Ostriker, 

1547: E.~C., \& Gammie, C.~F.\ 1998, \apjl, 508, L99

1548: 

1549: 

1550: \bibitem[{{Tielens} \& {Hollenbach}(1985)}]{1985ApJ...291..722T}

1551: {Tielens}, A.~G.~G.~M. \& {Hollenbach}, D. 1985, \apj, 291, 722

1552: 

1553: \bibitem[{V{\'a}zquez-Semadeni} et al. (2007)]{2007ApJ...657..870V}

1554: {V{\'a}zquez-Semadeni}, E., {G{\'o}mez}, G.~C., {Jappsen}, A.~K., 

1555: 	{Ballesteros-Paredes}, J., {Gonz{\'a}lez}, R.~F., \& {Klessen}, R.~S.

1556: 2007, \apj, 657, 870

1557: 

1558: 

1559: 

1560: \bibitem[Wong 

1561: \& Blitz(2002)]{2002ApJ...569..157W} Wong, T., \& Blitz, L.\ 2002, 

1562: \apj, 569, 157 

1563: 

1564: 

1565: \end{thebibliography}

1566: 

1567: 

1568: \clearpage

1569: \appendix

1570: \section{Vertical Equilibrium with Stellar and Gas Gravity}

1571: 

1572: The vertical momentum equation, when averaged over a horizontal plane,

1573: is given by 

1574: \begin{equation}

1575: \frac{\partial}{\partial t} \langle\rho v_z \rangle 

1576: + \frac{\partial   }{\partial z }

1577: \left\langle P + \rho v_z^2 + \frac{

1578:   {\bf B}\cdot{\bf B}}{8\pi} - \frac{ B_z^2 }{4\pi  } \right\rangle

1579: = -\left\langle \rho \frac{\partial \Phi}{\partial z  }\right\rangle

1580: \end{equation}

1581: (see e.g. \citealt{2007ApJ...663..183P}).

1582: Here, $\bf B$ is the magnetic field and $\Phi$ is the total (stellar

1583: plus gaseous) gravitational potential.  In steady state 

1584: $\langle\rho v_z \rangle$ is time-independent, so if we neglect

1585: magnetic fields and assume that $\rho$, $v_z^2$, $c_s^2=P/\rho$, and 

1586: $\partial \Phi/\partial z$ 

1587: are statistically independent quantities, we obtain

1588: \begin{equation}

1589: \frac{ 1 }{\langle \rho\rangle}

1590: \frac{\partial   }{\partial z }\left[\langle c_s^2 + v_z^2 \rangle \langle \rho \rangle\right]

1591: = - \frac{\partial \langle \Phi\rangle}{\partial z }.

1592: \end{equation}

1593: The Poisson equation, also averaged over the horizontal plane and

1594: assuming $R \Omega$ is independent of $R$, is 

1595: \begin{equation}

1596: \frac{\partial^2  \langle\Phi\rangle}{\partial z^2}

1597: = 4 \pi G (\langle\rho\rangle + \rho_{\ast}),

1598: \end{equation}

1599: where $\rho_{\ast}$ is the background stellar density.

1600: 

1601: If we now define $\sigma_z^2 = \langle c_s^2 + v_z^2 \rangle$ and

1602: assume that this total velocity dispersion is independent of height $z$,

1603: we can combine the vertical momentum equation with the Poisson

1604: equation to obtain  

1605: a second-order differential equation for

1606: the density profile $\langle \rho \rangle \rightarrow \rho(z)$:

1607: \begin{eqnarray}

1608: \frac{\partial}{\partial z}\left(

1609: \frac{\sigma_z^2}{\rho(z)}\frac{\partial \rho(z)}{\partial z}

1610: \right)

1611: =-4\pi G\left(\rho_{\ast}+\rho(z)\right),

1612: \label{eq:HSE}

1613: \end{eqnarray}

1614: Henceforth, we assume that $\rho_{\ast}$ is 

1615: uniform within the gas disk, which is a good approximation provided

1616: that the stellar scale height is significantly larger than the gaseous

1617: scale height.  Equation (\ref{eq:HSE}) is the expression for effective

1618: hydrostatic equilibrium in the vertical direction.

1619: 

1620: 

1621: 

1622: Introducing a variable $f(z)=\ln (\rho(z)/\rho_{\ast})$ and 

1623: a constant $h^2=\sigma_z^2/(4\pi G\rho_{\ast})$,

1624: we have

1625: \begin{eqnarray}

1626: f^{\prime\prime}=-\frac{1}{h^2}(1+e^f),

1627: \end{eqnarray}

1628: where the prime denotes a $z$ derivative. 

1629: This can be integrated once as

1630: \begin{eqnarray}

1631: \frac{(f^{\prime})^2}{2}=-\frac{1}{h^2}(f+e^f)+\mbox{const}

1632: =\frac{1}{h^2}\left(f_0-f+e^{f_0}-e^{f}\right),

1633: \end{eqnarray}

1634: where $f_0=\ln(\rho_0/\rho_{\ast})$ 

1635: is the boundary condition at the midplane where $f^{\prime}=0$.

1636: If we Taylor expand and retain only the two lowest order terms, i.e.

1637: $f(z)=f_0 - f_1 z^2$ so that 

1638: $\rho/\rho_0= \exp(-f_1 z^2)$, 

1639: the governing ODE becomes an algebraic equation: 

1640: \begin{eqnarray}

1641: \frac{(2f_1 z)^2}{2}

1642: =\frac{z^2}{h^2}\left(f_1+\frac{\rho_0}{\rho_{\ast}}f_1\right) 

1643: =\frac{4\pi G(\rho_0+\rho_{\ast})}{\sigma_z^2}f_1 z^2.

1644: \end{eqnarray}

1645: The coefficient $f_1$ is 

1646: \begin{eqnarray}

1647: f_1=\frac{1}{2H^2},\quad 

1648: H^2=\frac{\sigma_z^2}{4\pi G(\rho_{\ast}+\rho_0)}.

1649: \label{eq:Height}

1650: \end{eqnarray}

1651: Therefore, the gas density and pressure are 

1652: approximately given by Gaussian profiles

1653: \begin{eqnarray}

1654: \rho(z)=\rho_0 e^{\displaystyle -\frac{z^2}{2H^2}},\quad

1655:  P(z)=P_0e^{\displaystyle -\frac{z^2}{2H^2}},

1656: \end{eqnarray}

1657: where $P_0=\sigma_z^2\rho_0$.

1658: The midplane gas density $\rho_0$ is determined by requiring that 

1659: the profile integrates to the (known) gas surface density,

1660: \begin{eqnarray}

1661: \Sigma=\int_{-\infty}^{\infty}\rho(z)\, dz=\sqrt{2\pi}\rho_0 H.

1662: \end{eqnarray}

1663: Substituting for $\rho_0$ 

1664: in equation (\ref{eq:Height}), the scale height $H$ must satisfy 

1665: \begin{eqnarray}

1666: H^2&=&\frac{\sigma_z^2}{4\pi

1667:  G(\rho_{\ast}+\frac{\Sigma}{\sqrt{2\pi}H})}.

1668: \end{eqnarray}

1669: This yields a quadratic equation for $H$, with solution given

1670: by equation (\ref{Hest_eq}) of the text.

1671: 

1672: \end{document}

1673: