1: \documentclass[12pt,preprint]{aastex}

2: %\citestyle{aa}

3: %\twocolumn

4: 

5: % added to compile on Mac--remove before resubmission

6: \usepackage{graphicx}

7: \usepackage{epstopdf}

8: \DeclareGraphicsRule{.tif}{png}{.png}{`convert #1 `basename #1 .tif`.png}

9: 

10: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

11: %%%%%     Include some definitions.   %%%%%

12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

13: \def\er{\relax} \def\sr{\relax}

14: \def\ls{{_<\atop^{\sim}}}

15: \def\gs{{_>\atop^{\sim}}}

16: \def\mic{~$\mu$m}

17: \def\um{~$\mu$m}\def\etc{{\it etc.}}

18: \def\cf{{\it cf.}}

19: \def\ie{{\it i.e.}}

20: \def\eg{{\it e.g.}}

21: \def\iras{{\it IRAS}}

22: \def\iue{{\it IUE}}

23: \def\rosat{{\it Rosat}}

24: \def\ginga{{\it Ginga}}

25: \def\et{{et al.~}}

26: 

27: \def\h0{H$_0$}

28: \def\q0{q$_0$}

29: 

30: \def\hf{\mbox{$H_{1.6}$}}

31: \def\Msun{M$_{\odot}$}

32: \def\Lsun{L$_{\odot}$}

33: \def\Myr{\Msun/yr}

34: \def\kp{{\rm K}$^{\prime}$}

35: \def\lya{{\rm Ly}$\alpha$}

36: \def\Ls{{ L$^{*}$}}

37: \def\han {\mbox{{\rm H}$\alpha$}}

38: \def\pa {\mbox{{\rm Pa}$\alpha$}}

39: \def\hb {\mbox{{\rm H}$\beta$}}

40: \def\hgamma {\mbox{{\rm H}$\gamma$}}

41: \def\oiii{\mbox{{\sc [OIII]}}}

42: \def\hii{\mbox{H~{\sc ii}}}

43: \def\ha{\han}

44: \def\xray{\hbox{X-ray}}

45: \def\brg {\mbox{{\rm Br}$\gamma$}}

46: \def\bra {\mbox{{\rm Br}$\alpha$}}

47: \def\kms {\mbox{{\rm km} {\rm s}$^{-1}$}}

48: \def\spose#1{\hbox to 0pt{#1\hss}}

49: \def\simlt{\mathrel{\spose{\lower 3pt\hbox{$\mathchar''218$}}

50:      \raise 2.0pt\hbox{$\mathchar''13C$}}}

51: \def\simgt{\mathrel{\spose{\lower 3pt\hbox{$\mathchar''218$}}

52:      \raise 2.0pt\hbox{$\mathchar''13E$}}}

53: \def\lsim{\rlap{$<$}{\lower 1.0ex\hbox{$\sim$}}}

54: \def\gsim{\rlap{$>$}{\lower 1.0ex\hbox{$\sim$}}}

55: \def\be{\begin{equation}}

56: \def\ee{\end{equation}}

57: \def\av {\mbox{{\rm A}$_{\rm V}$}}

58: \def\ak {\mbox{{\rm A}$_{\rm K}$}}

59: \def\NH {\mbox{{\rm N}$_{\rm H}$}}

60: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

61: 

62: 

63: %%%%%%%%%%%%%%

64: % Title Page %

65: %%%%%%%%%%%%%%

66: 

67: \slugcomment{\it Accepted to ApJ}

68: 

69: \shortauthors{Gilbert \& Graham}

70: \shorttitle{Feedback from Antennae ELCs I.}

71: 

72: %\received{2005 November XX}

73: \begin{document}

74: 

75: \title{Feedback in the Antennae Galaxies (NGC 4038/9): 

76: I. High-Resolution Infrared Spectroscopy of Winds from Super Star

77: Clusters\altaffilmark{1,2}}

78: 

79: \author{Andrea M. Gilbert\altaffilmark{3,4,5} \&

80: James R. Graham\altaffilmark{4}

81: }  

82: 

83: \altaffiltext{1}{Data presented herein were obtained

84: at the W.M. Keck Observatory, which is operated as a scientific

85: partnership among the California Institute of Technology, the

86: University of California and the National Aeronautics and Space

87: Administration.  The Observatory was made possible by the generous

88: financial support of the W.M. Keck Foundation.}

89: 

90: \altaffiltext{2}{Data presented herein were obtained at the European 

91: Southern Observatory Very Large Telescope.}

92: 

93: 

94: \altaffiltext{3}{Institute of Geophysics \& Planetary Physics, Lawrence Livermore National Laboratory, L-413, 7000 East Ave., Livermore, CA 94550}

95: 

96: \altaffiltext{4}{Department of Astronomy,  

97: University of California,

98: 601 Campbell Hall,

99: Berkeley, CA, 94720-3411, USA}

100: 

101: %\altaffiltext{4}{Max-Planck Institut f\"ur extraterrestrische Physik,

102: %Postfach 1312, D-85741 Garching, Germany}

103: 

104: %\altaffiltext{5}{agilbert@mpe.mpg.de}

105: 

106: \altaffiltext{5}{agilbert@igpp.ucllnl.org}

107: 

108: 

109: %%%%%%%%%%%%

110: % Abstract %

111: %%%%%%%%%%%%

112: 

113: \begin{abstract}

114:  We present high-resolution ($R \sim 24,600$) near-IR spectroscopy of

115:  the youngest super star clusters (SSCs) in the prototypical starburst

116:  merger, the Antennae Galaxies.  These SSCs are young ($3-7$ Myr old)

117:  and massive ($10^5 - 10^7$ M$_\odot$ for a Kroupa IMF) and their

118:  spectra are characterized by broad, extended Brackett$\gamma$

119:  emission, so we refer to them as emission-line clusters (ELCs) to

120:  distinguish them from older SSCs.  The Br$\gamma$ lines of most ELCs

121:  have supersonic widths ($60-110$ km s$^{-1}$ FWHM) and non-Gaussian

122:  wings whose velocities exceed the clusters' escape velocities.  This

123:  high-velocity unbound gas is flowing out in winds that are powered by

124:  the clusters' massive O and W-R stars over the course of at least

125:  several crossing times.  The large sizes of some ELCs relative to

126:  those of older SSCs may be due to expansion caused by these outflows;

127:  many of the ELCs may not survive as bound stellar systems, but rather

128:  dissipate rapidly into the field population.  The observed tendency

129:  of older ELCs to be more compact than young ones is consistent with

130:  the preferential survival of the most concentrated clusters at a

131:  given age.

132:  \end{abstract} 

133: 

134: \keywords{galaxies: individual (NGC4038/9, Antennae Galaxies) ---

135: galaxies: ISM --- galaxies: starburst --- galaxies: star clusters ---

136: infrared: galaxies --- HII~regions}

137: 

138: 

139: %%%%%%%%

140: % Body %

141: %%%%%%%%

142: 

143: \section{Introduction}

144: \label{intro}

145: 

146: Star formation in starbursts creates massive young super star clusters

147: (SSCs) that are not often found in more quiescent environments like

148: the Milky Way's disk.  Their inferred masses ($10^5-10^6$ \Msun) and

149: Lyman continuum photon rates ($10^{52}-10^{53}$ s$^{-1}$) dwarf those of their

150: lower-mass analogs in the Local Group.  The most massive local SSC

151: analog is R136 in 30~Doradus, with a mass of at least

152: $10^{4.5}$ \Msun\  \citep{massey98a} and  $10^{51.4}$ s$^{-1}$ \citep{walborn91}.

153: 

154: SSCs are found in all galaxy environments, from dwarf irregular

155: starbursts \citep[e.g. NGC 1569,][]{oconnell94} to the prototypical

156: starburst disk galaxy M82 \citep[e.g.][]{oconnell95}, to

157: merging systems such as the Antennae Galaxies (NGC~4038/39)

158: \citep[e.g.][]{whitmore95,gilbert02}, and some have been found in

159: normal spiral galaxies  \citep[like NGC~6946][]{larsen01}, whose young clusters typically populate a lower mass range, $\sim 10^3 - 10^5$ \Msun\  \citep[e.g.][]{larsen02}.

160: % added citation/comment

161: Most

162: observations of SSCs have focused on optical wavelengths, but

163: near-infrared (near-IR) imaging has revealed young massive clusters

164: that excite luminous H~{\sc ii} regions (HIIRs) that are

165: invisible at optical wavelengths \citep[e.g.][]{gilbert00,turner03}.

166: Narrow-band imaging may also be important to track the faintest

167: youngest members of the population, which may not be detected with

168: broad-band IR imaging although they are bright in narrow filters tuned

169: to the recombination lines of their HIIRs \citep{alonso02}.  Radio

170: and mid-IR observations have uncovered still more heavily embedded

171: regions of massive star formation that are heavily obscured even at

172: 2 $\mu$m \citep[e.g.][]{kobulnicky99,beck00,beck01,vacca02,cabanac05}; these objects

173: are very young, dense, compact HIIRs that are the massive analogs of

174: ultra-compact \hii\ regions (UCHIIs).  Hence they are dubbed

175: ultra-dense HIIRs, or UDHIIs, by \citet[][]{kobulnicky99}, and they are

176: the supposed precursors of optically visible SSCs.

177: 

178: The high masses and stellar densities of SSCs resemble those of

179: globular clusters (GCs), although metal abundances ought to be higher in SSCs than GCs

180: and SSC ages are measured in Myr rather

181: than Gyr.  This suggests an evolutionary sequence in massive

182: star-forming regions: UDHIIs become optically visible SSCs

183: \citep{kobulnicky99} that may finally become (metal-rich) GCs.  The idea of GC

184: formation in mergers is supported by observations of young GCs (ages

185: of a few Gyr) in several merger remnants

186: \citep[e.g.][]{schweizer87,ashman92,fritzevalvensleben94,schweizer96,schweizer98},

187: as well as the multi-modal color distributions and varied spatial

188: distributions of GCs.  However, establishing the relationship between

189: SSCs and GCs requires examination of both the photometric and

190: kinematic properties of the SSC population.  By measuring an

191: individual SSC's age, mass, density profile and initial mass function

192: (IMF), and modeling its stellar and dynamical evolution, one can

193: predict whether it may survive internal and galactic dynamical

194: processes to join a population of GCs

195: \citep[e.g.][]{ho96a,boeker99,smith01,larsen01,mengel02,gilbert02,mccrady03}.

196: The IMF of SSCs is especially important for this question because the

197: light of SSCs is dominated by massive stars, while old GCs are

198: comprised entirely of sub-solar mass stars (and stellar remnants).  Observations of SSC

199: dynamical masses to date suggest that some SSCs do have normal IMFs

200: and could become GCs, while others cannot

201: \citep[e.g.][]{sternberg98,smith01,mengel02,gilbert01b,mccrady03,mccrady05},

202: although these results could be subject to systematic errors in

203: extinction corrections or neglect of effects like spatial mass

204: segregation, and they always rely on the assumption of cluster

205: virialisation.

206: 

207: Another important issue is the evolution of the SSC population as whole: GC

208: populations have kinematic distributions resembling those of bulges

209: (spheroidal with high velocity dispersions), while SSCs form where gas is

210: available, usually in disk-like environments with much lower dispersions; GCs

211: have log-normal luminosity functions \citep{harris91}, while SSCs display

212: power-law luminosity functions \citep[e.g.][]{whitmore99}, as do the giant

213: molecular clouds (GMCs) and \hii\ regions from which they presumably form

214: \citep[e.g.][]{williams97,kennicutt88,mckee97}.  Simulations have shown that

215: power-law cluster LFs can evolve into log-normal ones as the faint end of the

216: LF is eroded by tidal disruption in the galactic

217: potential, dynamical friction, stellar mass loss, and evaporation of stars

218: from clusters \citep[e.g.][]{fall01}.

219: 

220: Whether young SSCs survive to become GCs or disperse completely into

221: the field star population of a galaxy, they have great potential to

222: affect the energetics of its interstellar medium (ISM) because they

223: harbor thousands of massive stars producing ionizing and FUV

224: radiation.  The radiation from OB stars excites HIIRs and

225: photodissociation regions (PDRs), and heats the ISM.  Their winds and

226: supernova (SN) ejecta stir and inject energy into the surrounding ISM.

227: In the most extreme starbursts, the combined effects of a starburst

228: drive large-scale galactic winds \citep[e.g. in M82,][]{shopbell98}

229: that can eject matter into the intergalactic medium \citep[especially

230: in dwarf galaxies, e.g. NGC 1569,][]{heckman95}. SSCs are concentrated

231: power sources for feedback on both star-cluster and galactic scales.

232: 

233: The Antennae (NGC~4038/9) are a nearby pair of disk galaxies in an

234: early stage of merging that are well-known for their numerous SSCs,

235: distributed along their spiral arms and around their interaction

236: region \citep{whitmore99}.  Their molecular gas

237: distribution peaks at both nuclei and in the overlap region

238: \citep{stanford90}, but the gas is not yet undergoing a global

239: starburst typical of more advanced mergers \citep{nikola98}.  The

240: current global star-formation rate of the system is estimated at about

241: 20 \Msun\ yr$^{-1}$ by \citet{zhang01} from extinction-corrected

242: H$\alpha$.  ISO observations of the Antennae show that the mid-IR

243: ($8-15$ \um) flux from warm dust follows the optical distribution of

244: bright blue SSCs, but peaks in the overlap region in a single point

245: source that emits 15\% of the ISO flux \citep{mirabel98}.  This source

246: is an optically faint and unremarkable star cluster \citep{whitmore95}

247: that near-IR observations reveal as one of the youngest ($\sim 4$

248: Myr), most massive ($\sim 10^7$ \Msun) SSCs in the system

249: \citep{gilbert00,mengel01a}.  This cluster powers a dense (average

250: $n_e= 10^4$), large (half-light radius 32 pc) \hii\ region with a

251: Lyman continuum rate of order $Q[H^+]=10^{53}$ s$^{-1}$; it excites even

252: more extended, clumpy PDRs with nearly fluorescent H$_2$ emission; and

253: it is embedded behind $A_V \approx 10$ mag \citep{gilbert00}.

254: 

255: In \S~\ref{sec:obselc} we present near-IR imaging and spectroscopic

256: observations of a sample of young Antennae SSCs in order to identify

257: the emission-line clusters (ELCs) among them.  We derive ages and

258: stellar masses for the ELCs from their \brg\ spectra and magnitudes in

259: \S~\ref{sec:agemass}.  In \S~\ref{sec:linewidths} we compare the

260: widths of the broad, extended \brg\ emission lines of ELCs with the

261: inferred escape velocities of the exciting SSCs, and argue that ELCs

262: drive cluster outflows.  In \S~\ref{sec:densities} we estimate

263: electron densities for ELCs, and we compare ELCs with other types of

264: HIIR.  We discuss the evolution of ELCs and survival of massive clusters undergoing expansion due to mass loss in \S~\ref{sec:evolution}, estimate the ELC contribution to the

265: total star-formation rate in the Antennae in \S~\ref{sec:sfr}, and

266: conclude in \S~\ref{sec:conclusions}.  In a companion paper 

267: \citep[Paper II,][]{gilbert07b}, we present

268: a kinematic model for the cluster outflows, fit it to the data presented here,

269:  and discuss the mass-loss rates and energy budget of ELC outflows

270:  relative to similar systems such as superbubbles and galactic superwinds.

271: 

272: 

273: \section{Observations \& Data Reduction}

274: \label{sec:obselc}

275: 

276: On 2002 January 21 and February 22, we obtained $K$-band spectra of a

277: sample of 17 young SSCs and the nuclei of NGC~4038 and NGC~4039 in the

278: Antennae using NIRSPEC \citep{mclean00}, a facility near-infrared

279: ($0.95 - 5.6 \mu$m) spectrometer for the Keck-II telescope

280: \citep{mclean98}.  Using the cross-dispersed echelle mode and the N7

281: order-sorting filter, we detected Brackett$\gamma$ with little or no

282: continuum in 16 of the targets with $\lambda/\Delta\lambda \simeq 24,600$.  During the

283: course of these observations we recorded N7 images of the fields

284: around the $0.''432$ (3-pixel) slit using NIRSPEC's slit-viewing

285: camera (SCAM); these images were used to create a mosaicked finding

286: chart of the Antennae field from which offsets between the infrared

287: clusters were measured.  Cluster photometry was performed on $K$-band

288: images taken with NIRSPEC's slit-viewing camera (SCAM) in good seeing

289: ($0.''6$ or better) and photometric conditions on 2000 January 17 and

290: 2001 March 10.  We later obtained higher quality VLT ISAAC

291: \citep{moorwood98} narrow-band images at 2.17 and 2.25 \um\ on 11 January 2004 

292: in K-band seeing of $0\farcs4-0\farcs5$;

293: the former filter includes \brg\ emission and is shown in the finding chart of Figure~\ref{fig:ant_hst}.

294: % added ISAAC seeing here

295: 

296: \subsection{Imaging}

297: \label{sec:irlf}

298: 

299: The bright northern and southern sources in Figure~\ref{fig:ant_hst}

300: are the two nuclei in the system, NGC~4038 and NGC~4039.  Overplotted

301: are the positions of the brightest young ($< 30$ Myr old) clusters

302: detected in optical HST images reported by \citet{whitmore99}.  The

303: HST and NIRSPEC data sets were registered using the foreground star at

304: the origin of the figure (Star 4 of Whitmore et al.

305: 1999)\footnote{\citet{whitmore99} report a position for Star 4 of RA =

306:   $12^h 1^m 56.04^s$ and DEC = $-18^\circ 52' 43\farcs66$ (J2000),

307:   although an offset of $1\farcs2$ to the southwest of the HST positions

308:   from radio observations \cite{neff00} was reported by

309:   \citet{whitmore02}.}.  Comparing the relative HST positions of the

310: nine cleanly overlapping sources with their relative IR positions

311: yielded a SCAM plate scale of $0\farcs178$ per pixel.  This is

312: consistent with the value of $0\farcs18$ measured by \citet{figer00}. 

313: % modify opt/ir coincidence...

314: %The coincidence of so few bright optical sources in Figure~\ref{fig:ant_hst} with bright infrared ones reveals the incompleteness of the optical observations due to dust extinction in the region between the nuclei.  This highlights the importance of long-wavelength observations in probing the most recent star formation. 

315: About half of our sample of bright young SSCs does not appear in the optical brightest

316:    cluster list due to dust extinction, although most of these sources are

317:    detected at least in F814W (see Figure~\ref{fig:thumbs}).   SSC B1 highlights the importance of

318:    long-wavelength observations in probing the most recent star

319:    formation: it is deeply embedded and quite faint in the optical,

320:    with a K-band extinction of at least 1 mag \citep{gilbert00},

321:    and it is the most massive young SSC in the system after correcting

322:    for this extinction (Table~\ref{tab:agemass}).

323: 

324: The $K$-band SCAM images were reduced and calibrated relative to the

325: \citet{persson98} standards SJ 9150 and SJ 1952.  Aperture magnitudes for the

326: clusters were calculated using DAOPhot \citep{stetson87} routines in IDL.  The

327: fields containing the clusters are complex and it is evident that the accuracy

328: of our photometry is limited by systematics associated with eliminating

329: contributions from unrelated sources and with estimating the local background.

330: In an attempt to reduce these systematic errors we also computed magnitudes

331: based on fitting a 2-d Gaussian to the resolved cluster light profile.  The

332: mean difference between the $2''$ diameter aperture and fitting magnitudes is

333: $0.12$ mag with an rms of 0.2 mag.  Inspection of the field confirms that the

334: $2''$ apertures represent a compromise between isolating the clusters from

335: their environment while encompassing most of the light  \citep[point-source aperture

336: corrections were applied using Star 4 of][]{whitmore99}. 

337: We therefore adopt these magnitudes in what follows

338: and assume 0.2 mag as the rms error.  Table~\ref{tab:phottable} lists the

339: magnitudes of the selected targets for spectroscopy.  Our photometry agrees

340: with other measurements from the literature: for SSC B1 \citep[known as SSC A by][]{gilbert00}, we find K = 14.7 $\pm$

341: 0.2 mag, \citet{gilbert00} report N7 = 14.6 mag, and \citet{mengel01a} report

342: K = $14.8 \pm 0.2$ mag; for SSC F we find K $=15.9 \pm$ 0.2 mag

343:  and \citet{mengel02} report K$_s = 15.9$ mag.  For SSC S 

344:  we did not have adequate data for photometry, so we use values reported by

345:  \citet{mengel02}.  

346:  

347:  % added detail about isaac sizes here. 

348:  The ISAAC images were reduced, corrected for distortion, and combined.  Sizes of ELCs and Star 4 were measured via 2D Gaussian and Lorentzian fits, which gave comparable results, and the latter were adopted because they better reproduced the observed profiles.  We derived FWHMs from the geometric mean radii of the fits because it is a better indicator than the arithmetic mean of the area in an image that encloses half of the profile's light.  The FWHM of Star 4 was subtracted in quadrature to

349: compute the deconvolved FWHMs that are reported in Table~\ref{tab:phottable}

350: and plotted in Figure~\ref{fig:sizes}.

351:  %

352:  Throughout this paper we adopt a redshift distance for the Antennae of 19.2 Mpc ($1\arcsec = 93$ pc, distance modulus $(m-M)_0=31.41$ mag for $H_0=75$ km s$^{-1}$ Mpc$^{-1}$), although we note that a smaller distance of 13.8 Mpc has been measured by \citet{saviane04} based on the tip of the red giant branch (TRGB) in the southern tail of the Antennae.  If the TRGB distance is correct, all linear dimensions discussed in this paper decrease by a factor of 1.4.

353: 

354: \subsection{Brackett$\gamma$ Spectroscopy}

355: \label{sec:brg}

356: 

357: A finding chart similar to that in Figure~\ref{fig:ant_hst} from

358: prior $K$-band observations permitted the selection of cluster targets

359: for spectroscopy. Targets were chosen on the basis of brightness;

360: whenever feasible we oriented the $24''$ long slit to encompass

361: multiple clusters.

362: 

363: Table~\ref{tab:obselc} lists exposure times for the targets that are

364: identified in the image of the Antennae in Figure~\ref{fig:ant_hst}.  

365: %

366: SSCs are named roughly according to the lettering scheme adopted for \ha-emitting regions in the system by \citet{rubin70}.

367: %

368: Along with the SSCs we observed the two nuclei and detected broad

369: emission in the northern nucleus, NGC 4038, but not in the southern

370: one, NGC 4039, although we did detect broad Br$\gamma$ in a faint cluster near

371: the latter nucleus.  This off-nuclear source is labeled ``A1'' in

372: Figure~\ref{fig:ant_hst} and in Table~\ref{tab:measurements}.

373: 

374: Raw exposures were first corrected for electronic noise which appears

375: as horizontal striping in a few of the 32 analog channels.  The images

376: were then dark-subtracted (because on-minus-off sky subtraction was not

377: possible for a few exposures),

378:  flat-fielded, and corrected for bad pixels

379: and cosmic rays.  A wavelength scale was determined from atmospheric

380: OH lines \citep{rousselot00} and used to rectify the Brackett $\gamma$ echelle order

381: (order 35), in which spatial and spectral directions are curved, onto

382: an orthogonal grid of wavelength and slit position.  In order 35 the

383: spectral resolution measured from OH lines was

384: ${\lambda}/{\delta\lambda}= 24,600$, or about 12 km s$^{-1}$ FWHM.

385: 

386: A primary sky subtraction was performed by differencing pairs of

387: frames where possible, and then a secondary sky subtraction was done

388: by fitting and removing the background in each column.

389: Emission-line spectra were optimally extracted using a Gaussian

390: weighting function matched to the spatial extent of the Br$\gamma$

391: emission, and then an aperture correction was applied to recover the

392: full flux.

393: 

394: A B4IV/V star was observed as an atmospheric standard.  Although most

395: of our Br$\gamma$ spectra were unaffected by atmospheric absorption,

396: at the radial velocities of some objects a small feature absorbs some

397: flux in the wing of the line, so we applied the atmospheric correction

398: in all spectra.  A flux scale was derived by requiring the 2.2 $\mu$m

399: continuum flux of the standard star to equal that corresponding to its K

400: magnitude.  Reduced line spectra for the ELCs from which we

401: detected broad Br$\gamma$ (excepting only D1) are shown in Figure~\ref{fig:brg}, and

402: position-velocity diagrams for the four brightest ELCs are shown in

403: Figure~\ref{fig:elcpv}.  Signal-to-noise ratios per pixel in the fully

404: reduced spectra ranged from around 3.5 in the weakest lines to 75 in

405: the strongest, with continuum levels  low in all but the brightest

406: sources, which have continuum signal-to-noise ratios per pixel of only a few.

407: Sources for which we did not detect \brg, clusters 10, 16, and 1 from 

408: \citet{whitmore99}, are not considered further in this paper.

409: We refer to the SSCs with \brg\ emission as emission-line clusters (ELCs).

410: 

411: 

412: The Br$\gamma$ profiles in Figure~\ref{fig:brg} are well resolved and

413: exhibit high-velocity non-Gaussian wings.  The isothermal sound speed

414: of ionized gas at temperature $T=10^4$ K is $ c =\sqrt{k_B T/\mu m_p}

415: \approx $ 12 km s$^{-1}$ where $\mu$ is the mean molecular weight and

416: $m_p$ is the proton mass, so linewidths FWHM = $2.355 c > 28 $ \kms\ 

417: are supersonic. Nearly all of the line widths, determined from

418: Gaussian fits to the line cores, are supersonic. Cluster D1 (FWHM = 14

419: km s$^{-1}$) is the only Br$\gamma$ source without a supersonic

420: component and cluster F is marginal with FWHM = $30$ \kms.  We refer

421: to these young SSCs with supersonic Br$\gamma$ line widths as broad

422: emission-line clusters (bELCs). 

423: 

424: The broad Br$\gamma$ lines of ELCs may arise from individual hot stars

425: or dilute photoionized gas.  Individual O stars have EW[\brg] of only

426: a few \AA~\citep[e.g.][]{hanson96}, while the Antennae clusters have

427: EW[\brg] from a few up to a few hundred \AA.  The observed Br$\gamma$

428: lines are also too narrow to be associated with individual O-star

429: winds, and as already noted, they are too broad (and non-Gaussian) to

430: simply reflect the virial motion of the constituent stars.  Thus,

431: photospheric and wind emission from O stars alone cannot explain the

432: intensity or width of \brg\ emission from ELCs.  Even though the ELC

433: lines are too narrow to be stellar wind features, WR stars could

434: produce the observed equivalent widths.  \citet{figer97} measure the

435: combined EW of blended He~{\sc i}, He~{\sc ii} and \brg\ lines from

436: low-resolution IR spectra of WR stars, finding values of about 20--130

437: \AA.  However, the line widths are far too narrow to be dominated by

438: WR emission \citep[which has line widths $\sim 10^3$ \kms,

439: e.g.][]{abbott87}.  Thus the ELC line fluxes most likely originate not from gas in photospheres

440: and winds of hot stars, but in the expanding \hii\ regions that they excite.

441: The broad lines must be formed from rapidly expanding ionized gas swept up by

442: the cumulative effect of stellar winds and photoevaporated flows driven by the

443: thousands of hot stars in the clusters.

444: 

445: % add comparison to Whitmore05 Ha radial velocities

446: Table~\ref{tab:measurements} summarizes the measured Br$\gamma$

447: fluxes, equivalent widths, barycentric radial velocities, and line

448: widths for the ELCs.  

449: %

450: Our \brg\ radial velocities are consistent with those reported from HST \ha\ spectroscopy by \citet{whitmore05} for sources in the vicinities of SSCs B, D, and F (a direct cluster-to-cluster comparison is not possible because the slit positions and sizes are different).

451: The listed \brg\ FWHMs are the observed values, not corrected

452: for either the instrumental line-spread function (12 \kms) or

453: thermal broadening (21.4 \kms).

454: We also detected He~{\sc i} 3F$^0-$3G emission at 2.1647 $\mu$m near

455: the blue wing of Br$\gamma$ in the brightest sources (B1, B, C, and D)

456: at a level of $4.4 \pm 0.5$\% of the \brg\ fluxes.  For SSC D2, whose

457: Br$\gamma$ emission includes a very narrow component (FWHM = 23 \kms)

458: as well as a weak broad component (FWHM = 55 \kms), we list the total

459: line flux but only the line width of the broad part.

460: 

461: % change gerssen reference to bottema; add more dispersion discussion

462: We consider the kinematics of the ELC population in the Antennae by

463: computing the rms radial velocity dispersion of the 14 targets in the

464: overlap region.  We find a value of $74$ km s$^{-1}$ rms, which is

465: comparable with a disk galaxy's rms integrated velocity dispersion \citep{bottema93}.

466: %\citep[e.g. NGC~488,][]{gerssen97}.  

467: Some of this velocity dispersion is due to a general gradient in the north-south direction along the overlap region between the two nuclei \citep{amram92,gilbert02}, but the gradient is not monotonic: both nuclei have higher radial velocities of at least 1600 \kms\ \citep[][]{amram92}, and some ELCs lie at minima in \ha\ radial velocity, i.e. in \ha\ superbubbles \citep{amram92,whitmore99,zhang01}.  In the northern overlap region, the ELC rms radial velocity dispersion is 21 \kms\ (for eight ELCs) about a mean of 1573 \kms.  In the southern overlap region, the velocity dispersion of six younger ELCs is 48 \kms, equivalent to that of the giant molecular complexes, \citep[49 km

468: s$^{-1}$,][]{wilson00} about a mean of 1445 \kms, so the kinematics of this group reflect those of the nearby molecular material. 

469: %

470: Velocity dispersions for smaller groupings on scales of $10-20\arcsec$, e.g. the two B, three D, five E, or three F clusters, are 21, 56, 22, and 18 \kms, respectively, which apart from the larger D cluster value are comparable with the raw dispersions for sources over similar spatial scales in \citet{whitmore05}.

471: 

472: High-resolution near-IR spectroscopy has recently revealed broad supersonic \brg\ emission from massive embedded star-forming regions in two other systems:

473: \citet{turner03} observed a deeply embedded radio source in NGC~5253 and \citet{henry07} observed two similar sources in Henize $2-10$, for which they derived Q[H$^+$] values above $10^{52}$ s$^{-1}$ and FWHMs of order $60-80$ km s$^{-1}$.  

474: %

475: While these dwarf galaxies are different in scale from the merging Antennae disks, their starbursts have much in common:  optically visible SSCs that have already disrupted their ISM on large scales (seen e.g. in X-ray and \ha\ superbubbles), bELCs that appear to be blowing out of their natal cocoons, and even younger heavily enshrouded star-forming regions seen as radio-bright UDHIIs. 

476: % refs there? see hiir section below

477: 

478: %their SSCs have blown multiple X-ray superbubbles \citep{strickland99}, and harbors just a handful of SSCs \citep[e.g.][]{calzetti97} and one heavily enshrouded young star-forming region that may be a UDHII and a bELC. 

479: 

480: \section{Ages \& Masses of ELCs} 

481: \label{sec:agemass}

482: 

483: Br$\gamma$ fluxes trace the ionized gas within a cluster and hence

484: the young, massive stellar content.  We assume that this flux arises

485: from the O/WR stars of a population that formed in a short-lived burst

486: and has an initial mass function (IMF) that is usually characterized as a power law, $MdN/dM \propto M^{-\alpha}$.  Employing a stellar population

487: synthesis model such as Starburst99 \citep{leitherer99} permits an

488: inference of ELC age by comparing observed and predicted Br$\gamma$

489: equivalent widths (EWs) for a cluster as a function of age, and an

490: inference of mass follows from scaling the predicted magnitude of the

491: fiducial $10^6$ \Msun\ cluster at the determined age to match the

492: observed magnitude.  We assume solar metallicity for the

493: clusters\footnote{We note that while \citet{mengel02} found a

494:   handful of Antennae SSCs to be consistent with having solar

495:   metallicity, they argue that one (ELC S, their [W99]-2) may have a

496:   supersolar metallicity.} and adopt the Galactic field star average

497: IMF of Kroupa \citep[a broken power law with $\alpha = 2.3 $ for

498: stellar masses $M = 0.5 - 100$ \Msun\ and $\alpha = 1.3 $ for $M = 0.1

499: - 0.5$ \Msun,][]{kroupa01}.  The slope of this IMF at the high-mass

500: end is nearly equivalent to that of the solar neighborhood IMF of

501: Salpeter \citep[a single power law with $\alpha =

502: 2.35$,][]{salpeter55}, and is consistent with that measured by

503: \citet{massey98a} for the massive, dense cluster R136 in the Large

504: Magellanic Cloud.  

505: 

506: Those assumptions yield ELC ages of $3.4-7$ Myr and masses of $7 \times

507: 10^4 - 12 \times 10^6$ \Msun (incorporating an extinction correction

508: where possible), as listed in Table~\ref{tab:agemass}.  The ELC ages

509: have a gradient across the overlap region: the five ELCs that are

510: younger than 6 Myr reside in the southern overlap region where the

511: bulk of the molecular gas is concentrated \citep[e.g.][]{wilson00},

512: and the older sources reside in the northern overlap region.  We

513: estimate fit errors for the ages of less than 0.1 Myr and relative

514: errors in masses of 20\% for most ELCs (see Table~\ref{tab:agemass}).

515: However, more significant than the fit errors may be systematic

516: errors, possible sources of which include the unknown $K$-band

517: extinction for some ELCs, our assumptions that the ELCs are

518: ionization-bounded, and uncertainties in the models.  We discuss these

519: possibilities in turn below, and then compare our ages for two ELCs with literature values.

520: 

521: From an ELC \brg\ line flux we derive 

522: Q[H$^+$] under case B assumptions for an ionized \hii\ region at $10^4$ K

523: \citep[e.g.][]{osterbrock74,hummer87}.  The resulting values  

524: are listed in Table~\ref{tab:agemass} where they are uncorrected for extinction, 

525: whereas in the text we always adopt the listed extinction corrections where available.

526: ELC B1 has Q[H$^+$] $=1.5 \times

527: 10^{53}$ s$^{-1}$, so it emits nearly half of the total Lyman

528: continuum rate from all 17 ELCs observed: $3.3 \times 10^{53}$

529: s$^{-1}$.  ELCs B1

530: and B have Q[H$^+$] at the high end of the range observed for SSCs in

531: other systems \citep[e.g. He $2-10$ region A at $10^{52-53}$

532: s$^{-1}$,][]{johnson00}.  UDHIIs are also observed to have Q[H$^+$]

533: values of order $10^{51-53}$ s$^{-1}$ \citep{beck00}.  The faintest

534: Antennae ELCs in our sample have Q[H$^+$] $= 10^{51}$ s$^{-1}$, which

535: is comparable with the fluxes of the brightest Galactic giant \hii\ 

536: regions (e.g. W43 and NGC~3606).

537: 

538: ELCs can suffer significant extinction even at $K$ band, and where estimates are

539: available we include them in Table~\ref{tab:agemass}.   In the case of ELC B1,

540: \citet{gilbert00} measured an extinction at $K$ band of ${A}_{\rm K} = 1.1 \pm

541: 0.1 $ mag from Br$\gamma$ and 20 Pfund recombination lines.

542: This corresponds to ${A}_{\rm V} = 10$ mag for a standard extinction

543: law \citep{rieke85}.  Other authors have inferred lower extinctions to

544: this source: \citet{mengel01a} derived ${A}_{\rm V} = 4.3$ mag by

545: comparing \brg\ and \ha\ fluxes; \citet{whitmore02} report ${A}_{\rm

546:   V} = 7.62$ mag from optical imaging.  We use radio continuum flux

547: densities for ELC counterparts \citep[from][]{neff00} together with

548: \brg\ fluxes to estimate $K$-band extinctions for a few ELCs (B1, B, D2, and F1) under the

549: assumption that the radio emission is unabsorbed.  For ELC B1, our

550: inference of ${A}_{\rm V} = 10$ mag \citep{gilbert00} from K-band data agrees with

551: the extinction derived using the radio-IR flux ratio, suggesting that

552: % add detail about mixing

553: the emitting nebular gas and obscuring dust in the ELC are probably mixed; 

554: optical measurements only probe part-way into the emitting region so they underestimate the overall extinction and hence the intrinsic ${Q}

555: [{\rm H}^+]$.  \citet{turner03} find the same bias for the compact, heavily

556: obscured source 

557: in NGC 5253: near-IR and radio measurements give a consistent value

558: for the extinction of ${A}_{\rm V} = 18$ mag, a value that is six

559: times greater than the optically-determined one.  We emphasize that

560: infrared-radio determinations of ${Q} [{\rm H}^+]$ are more reliable

561: than optical-infrared ones.  In Tables~\ref{tab:measurements} and

562: \ref{tab:agemass} we present observed fluxes and derived masses (uncorrected

563: for extinction).  Extinction does not affect the ages since

564: they are derived from extinction-independent equivalent widths.

565: 

566: There is evidence that ionizing photons escape from young clusters in

567: starburst HIIRs \citep[e.g. in He $2-10$ and

568: NGC~4214,][]{johnson00,leitherer96a}.  Photoionization models of GHIIRs

569: require $10-73\%$ of the Lyman continuum flux to escape

570: \citep{castellanos02}.  

571: %

572: The escape fraction of Lyman continuum photons from Galactic OB associations is

573: estimated at $6-15\%$ \citep{dove00}.  Observations of diffuse

574: H$\alpha$ emission throughout the Antennae \citep[not only in

575: association with stellar clusters, e.g.][]{whitmore99} may also

576: suggest leakage of Lyman continuum flux from its HIIRs, i.e. that ELCs

577: are density-bounded.

578: %

579: \citet{beckman00} find that a transition between ionization-bounded

580: and density-bounded HIIRs in spiral galaxies occurs at a constant \ha\ 

581: luminosity of $L_{\ha} = 10^{38.6}$ erg s$^{-1}$.  For Case B

582: recombination this corresponds to a \brg\ luminosity $L_{\brg} =

583: 10^{36.6}$ erg s$^{-1}$, which is below the observed values for all of

584: the ELCs. Presumably such a threshold in a merging pair of disk

585: galaxies is higher than in a spiral because it depends on the

586: available gas supply and the strength of the starburst, both of which

587: are stronger in the overlap region of a merger.

588: %

589: If ELCs are density-bounded, then some of their Lyman continuum photons

590: escape to ionize and dissociate the surrounding medium, and our

591: Br$\gamma$ measurements would underestimate $Q[H^+]$.

592: We would thus overestimate ELC ages and expect them to be fainter,

593: which would also lead to overestimated masses.

594: 

595: For two sources in our sample, \citet{mengel02} measured $K$-band

596: magnitudes and dynamical masses based on high-resolution spectroscopy,

597: and used additional diagnostics with EW[\brg], i.e. EW[CO] and

598: EW[CaT], to derive ages for the clusters.  We compare our ages with

599: theirs to give an idea of the significance of IMF choice, metallicity,

600: and the level of systematic errors that may be inherent to the observations and

601: Starburst99 models, which we both employ.

602: For ELC S (their cluster [W99]2), we infer an age of $6.6 \pm 0.1$ Myr

603: using EW[\brg], assuming a full ($0.1-100$ \Msun) Kroupa IMF at solar metallicity.

604: \citet{mengel02} obtain the same age ($6.6 \pm 0.3$ Myr) for a

605: Salpeter IMF ($1-100$ \Msun) at twice solar metallicity.  Our

606: measurements of EW[\brg] differ by a factor of two (ours: $4.8\pm1.6$

607: \AA, theirs: $10.0\pm2.5$ \AA), most likely because we use different apertures:

608:  theirs is a 2\farcs2 box that includes strong diffuse line emission surrounding ELC S

609: (which is not at a peak in \ha\ or \brg), while our aperture  encompasses

610: less  diffuse line emission because it is the best-fit Gaussian

611: spatial profile of the cluster \brg\ line, or 0\farcs9 FWHM for ELC S.  In

612: this case the differences in ages due to these other differences

613: (between IMFs, abundances, and even EW[\brg]) cancels out to be insignificant.

614: However, the typical size of the age offsets due to these factors is a few tenths of a Myr in

615: the sense that the Salpeter IMF gives an older age than the Kroupa IMF for a given EW[\brg], and a 

616: larger EW[\brg] for the same IMF gives a younger age, while the metallicity difference 

617: in age is quite small for this range of EW[\brg].

618: But \citet{mengel02} infer that neither of these IMFs may be appropriate for the

619: cluster based on the dynamical mass measurement ($2.0 \times 10^6$

620: \Msun), which is lower than the photometric mass predicted by the

621: models. This implies that the cluster has less than a full complement

622: of low-mass stars or a shallower IMF slope \citep{mengel02}.

623: 

624: In ELC F (their [W99]15), \citet{mengel02} measure a high EW[CO

625: 2.29$\mu$m] of $17 \pm 0.2$ \AA.  This requires an age $>$ 7 Myr, when the

626: brightest supergiants dominate a cluster's infrared emission.  They

627: estimate an age for ELC F of $8.7 \pm 0.3$ Myr for a Salpeter IMF,

628: while we find $6.5 \pm 0.1$ Myr for a Kroupa IMF or ($6.6 \pm 0.1$ Myr 

629: for the same Salpeter IMF), where we both assume solar metallicity.

630: Again the dynamical mass implies that neither of these IMFs is

631: correct, but in the opposite sense to that of ELC S.  \citet{mengel02}

632: measure a dynamical mass of $3.3 \times 10^6$ \Msun, which suggests a

633: steeper than Salpeter IMF slope (2.5 rather than 2.35).  

634: The discrepancy between ages determined from \brg\ and those

635: determined from photospheric features of supergiants cannot be

636: explained by our different IMF choices or different magnitudes or

637: extinctions (these are the same).  It may be due to different values of EW[\brg] 

638: determined from different types of data: a spectrum in this work versus narrow-band

639: imaging in theirs, which are subject to different difficulties in

640: removing diffuse stellar and nebular backgrounds.  The discrepancy may

641: also be due to diverging model predictions for our differing spectral

642: diagnostics.  Strong \brg\ from O stars and CO absorption in

643: supergiants are not found simultaneously in the spectrum of a

644: single-aged stellar population based on the latest tracks used in Starburst99.

645: However, these are single-star evolutionary tracks, so the models do

646: not yet account for binary systems (in which most stars are thought to

647: reside) or the interactions between binaries and stars in 

648: cluster environments, e.g. mass transfer, wind-wind interactions, etc.

649: %

650: Finally, SSCs may not form instantaneously, but rather over the course of several Myr, and if SSCs observed at large distances are actually unresolved clusters of clusters then they could have larger age spreads--on scales similar to the ELC sizes, 30 Doradus has an age range of about 25 Myr due to propagating star formation: 

651: R136 contains low-mass stars with ages of $4-5$ Myr and massive stars with ages of $1-2$ Myr \citep{massey98}, while nearby Hodge 301 ($\sim 40$ pc away) harbors supergiants with ages of $20-25$ Myr \citep[e.g.][]{grebel00}.

652: %

653: We view ELC F and 30 Dor as a warning that ages determined via single-star, single-aged population synthesis models may have systematic errors of a few Myr at ages near

654: the end of the O star era, and proceed with this caveat in mind.

655: 

656: 

657: \section{Mass Loss from ELCs}

658: \label{sec:linewidths}

659:  

660: With mass estimates for the ELCs in hand, we now

661: assess the ability of the ELCs to gravitationally bind their \hii\

662: gas by comparing the observed \brg\ line widths $\sigma_{HII}$ of ELCs with their

663: escape velocities $v_{esc}$, which depend upon cluster mass and size.  We can write $v_{esc}$ in terms of the 1D stellar velocity dispersion $\sigma_*$ of a virialized cluster as $v_{esc}=\sqrt{2} \sigma_*$.  The virial relation can be expressed as $\sigma_*^2 = GM/\eta r_{hp} $, where $r_{hp}$ is the projected half-light radius and $\eta$ is a constant that depends upon the mass distribution of the cluster \citep[e.g.][]{spitzer87}.  The value of $\eta$ is often taken to be about 10, although for a range of realistic globular cluster concentration parameters, $0.5-2.5$, \citet{mengel02} found that $\eta=  9.7-5.6$ (more concentrated clusters have smaller $\eta$).  \citet{boily05} calculated the time evolution of $\eta$ in SSCs using a gas-dynamical model that incorporates stellar evolution; they found that in low-density clusters, $\eta$ does not change much over the first 10 Myr, but in intermediate- and high-density clusters that resemble M82-F and R136 in 30 Doradus, respectively, $\eta$ increases by factors of about $1.3$ and $1.7$, while the clusters grow more compact and $r_{hp}$ decreases.  An increasing $\eta$ implies a decreasing $v_{esc}$, which makes it progressively easier for \hii\ gas to escape from a cluster.  

664: If the measured $\sigma_{HII}$ equals or exceeds $v_{esc}$, then the cluster cannot gravitationally bind the high-velocity ionized gas in the line wings, which has sufficient energy to escape from the cluster potential in an unbound outflow (in the absence of a confining external pressure). 

665: For the two ELCs (S and F) whose $\sigma_*$ were measured by \citet{mengel02},

666: we can directly compare $v_{esc}$ and $\sigma_{HII}$.  Assuming that the measured $\sigma_*$ represents a virial velocity dispersion, ELC S has $v_{esc}=20.1$ 

667: %$\sigma_* = 14.2$

668: \kms\ and $\sigma_{HII} = 31.8$ \kms\ (corrected for the instrumental line profile, Table~\ref{tab:measurements}), so most of its gas is not bound to the cluster.  If the other ELCs in

669: Table~\ref{tab:measurements} share this typical value for an SSC

670: stellar velocity dispersion, then nearly all of them have

671: predominantly unbound gas.  ELC F has the second-lowest intrinsic nebular line width at

672: $\sigma_{HII} = 11.6$ \kms, and it has $v_{esc} = 28.6$ \kms,

673: %$\sigma_*=20.2$

674: which suggests that most of its gas is bound to the cluster.  If mass segregation is present in these ELCs, then the measured $\sigma_*$ would be an underestimate of the true value because it is dominated by the light of massive, centrally concentrated stars.  In that case, ELCs S and F would be better able to bind their nebular gas.

675: 

676: In order to estimate escape velocities for the other ELCs, we

677: consider the mass of both the cluster's stellar component (implied by

678: Starburst99 for a Kroupa IMF extending from 0.1 to 100 \Msun,

679: \S~\ref{sec:agemass}) and its ionized gas component (whose mass ranges from 0.01 to 2.6 times the stellar mass\footnote{The ionized gas mass is deduced from outflow model fits in Paper II, \S 2.3 and Table 1.}).  We adopt an extinction correction where possible

680: (as in Table~\ref{tab:agemass}), and we assume that stars and gas fill the same volume out to the smaller of the half-light radii measured from the 2.17 and 2.25 \um\

681: FWHMs (see Figure~\ref{fig:sizes} and Table~\ref{tab:phottable}).  These assumptions are conservative in the sense that they place the total amount of mass available in the minimum volume, which leads to an upper limit for $v_{esc}$.   

682: Adopting a value of $\eta=10$, we derive escape velocities for most sources that are well below the observed \hii\ line widths: $v_{esc}/\sigma_{HII}$ ranges from 0.1 to 0.3 for all ELCs except the two clusters with the narrowest \brg\ emission (i.e., all bELCs have $\sigma_{HII} > v_{esc}$).  

683: The exceptions are ELC F, which we have already concluded can bind its gas, and

684: ELC D1, which has the narrowest \brg\ line in the sample, and $v_{esc}/\sigma_{HII} = 2.1$.  

685: %

686: A more conservative estimate of the gas boundedness comes from assuming that most of the ELC mass is actually concentrated in a 4 pc radius \citep[median value for SSC effective radii in][]{whitmore99} that is unresolved in our data; this leads to higher values of $v_{esc}/\sigma_{HII} = 0.2-0.8$  for bELCs, but still suggests that much of the HII gas in the observed profiles is not bound to the clusters.

687: %

688: If we consider both this smaller radius and decrease $\eta$ to the minimum value of 5.6 that corresponds to a high-concentration globular \citep{mengel02}, then the ratios $v_{esc}/\sigma_{HII}$ for ELCs B1, B, and E5 rise to $ 0.9- 1.1$, while those of the remaining bELCs are in the range $0.2- 0.8$.  Thus even with the most compact mass distribution that we imagine they could have, even in the ELCs with the largest $v_{esc}/\sigma_{HII}$ ratios, the high-velocity gas in the broad line wings would be unbound.

689: %

690: This suggests that a significant fraction of the ionized gas in bELCs is unbound and may be outflowing, but that narrow-lined ELCs can gravitationally bind their HIIRs and do not drive outflows.  

691: 

692: An exception to this rule may be the compact, embedded broad-line source in NGC~5253 \citep{turner03} that resembles the Antennae bELCs in terms of line width, mass, and luminosity.  Although its \brg\ line is very broad (76 \kms\ FWHM) and hence supersonic, it appears to be compact enough to bind its \hii\ gas if the most conservative choices for $\eta$ and half-light radius (the smallest radio semimajor axis is 0.5 pc) are adopted for a $10^6$ \Msun\ cluster \citep{turner03}; in this case, $v_{esc}=131$ \kms, which greatly exceeds the measured line width, and leads to the conclusion that the cluster easily binds its UDHII.  However, if we assume $\eta = 10$ and use the $K$-band size of the nebula to delineate the \brg-emitting region \citep[1 pc radius for their H2, ][]{alonsoherrero04} rather than the radio size, then we find $v_{esc}=69$ \kms.  This would suggest that plenty of gas in the \brg\ line wings has adequate energy to escape the cluster's potential well.  Either way, the NGC~5253 cluster is far more compact than the Antennae ELCs; this difference may be due to a higher ISM pressure in NGC~5253 that confines the UDHII at high density and prevents any significant outflow of gas.

693: % mention he 2-10 sources here??

694: 

695: 

696: % radii shift slightly

697: In the foregoing discussion we have assumed that the stars in ELCs are distributed like their infrared light.  The $K$-band half-light radii of ELCs range from about 10 to 50 pc, measured either in the continuum narrow-band filter at 2.25 \um, which is dominated by starlight (Table~\ref{tab:phottable}), or in the 2.17 \um\ filter, which also includes \brg\ nebular emission (Figure~\ref{fig:sizes}).  ELC radii are up to an order of magnitude larger than the median effective radius of optically visible SSCs of all ages in the Antennae (4 pc), measured from HST images \citep[][and see Figure~\ref{fig:ant_hst}]{whitmore99}. \citet{mengel02} measured an effective radius of $3.6$ pc for the

698: apparent optical counterpart of ELC F from the $I$-band HST

699: observations, and they found similar values up to 6 pc for a few other

700: optically revealed clusters.  However, \citet{mengel05} reported that the mean $I$-band effective radius is 16 pc (with a standard deviation of 15 pc) for seven isolated young (age $< 4$ Myr) Antennae clusters.  The 2.25 \um\ half-light radii of ELCs B1, B, and D, whose ages are below 4 Myr, are 29, 49, and 40 pc, respectively (Table~\ref{tab:phottable}, Figure~\ref{fig:sizes}).  Thus young clusters are intrinsically large at visible wavelengths even at the WFPC2 resolution of $0\farcs1$.  Most ELCs are resolved in our near-infrared images (with point source FWHM of $0\farcs4$ in the combined frames, see Fig.~\ref{fig:radprofs}), but what we view as a single source could break up into multiple sources when viewed with better angular resolution.  

701: % 

702: In order to check this, we examined archival optical and near-IR HST images of several ELCs together with ISAAC ones: Figure~\ref{fig:thumbs} shows that while some ELCs consist of multiple sources, even at the finest spatial resolution they tend to be dominated by a single bright one.

703: %

704: ELC B, for example, is comprised of a very bright source in $K$ band surrounded by fainter emission and at least one neighboring source to the southwest, but there are a handful of spatially coincident sources in HST filter F555W \citep[Figure~\ref{fig:ant_hst} inset,][]{whitmore99}; it may be a cluster of clusters.  ELC B1 is a single bright source (like most other ELCs) in $K$ band and a very faint one in $I$.  Caution must be exercised in attempting to match sources observed at different wavelengths and resolutions, especially when extinction can be important.

705: % add ref to hst fig 2, fig:thumbs, 

706: 

707: The position-velocity diagrams in Figure~\ref{fig:elcpv} show that the

708: \brg\ emission in ELCs B1, B, and D has spatial and velocity structure.  ELC C, which

709: appears complex and diffuse in optical and near-IR images, also has

710: extremely extended spatial and velocity structure in \brg\, with a

711: very weak continuum that is centered between its high- and low-velocity 'lobes'; it may comprise several cluster components, or it may be driving a bipolar outflow.  Figure~\ref{fig:sizes} shows that most (but not all) bELCs have slightly more extended emission in the 2.17 \um\ filter, which includes \brg, than in the 2.25 \um\ filter, and that there is a weak tendency for the ratio of 2.17 \um\ to 2.25 \um\ sizes to increase with age.  An HIIR might have more extended line than continuum emission (which

712: is dominated by stars) because Lyman continuum photons are escaping to

713: ionize the surrounding ISM, and this may produce some of the observed

714: extended emission of ELCs.  However, another process that would

715: produce extended line emission is the escape of ionized gas in an

716: outflow, which is suggested by the high velocities of some of the

717: extended \brg\ relative to the line centers, as well as the double-lobed appearance of ELC C in Figure~\ref{fig:elcpv}.  An increase in \brg-to-continuum size ratio with bELC age could be due to expansion of an outflow into the surrounding ISM and/or contraction of the stellar cluster with time (see \S~\ref{sec:evolution}).

718: 

719: 

720: \section{The Nature of ELC HIIRs}

721: \label{sec:densities}

722: \label{sec:lumsig}

723: 

724: It is useful to compare the HIIRs that are excited by ELCs with three familiar catagories of HIIRs 

725: (Galactic compact and ultra-compact HIIRs ((U)CHIIs), Galactic HIIRs, and extragalactic GHIIRs), as well as with the recently discovered class of UDHIIs \citep{kobulnicky99}.  

726: We will discuss four properties that distinguish different types of HIIRs: luminosity (in the form of $Q[H^+]$), line width, size, and electron density $n_e$.

727: 

728: We presented $Q[H^+]$ values for ELCs in \S~\ref{sec:agemass} and discussed

729: the \brg\ line widths of ELCs in  \S~\ref{sec:linewidths}; we now place bELCs (excluding the narrow-lined  ELCs D1 and F) in context with other HIIRs by plotting their $Q[H^+]$ vs.

730:  H~{\sc i} recombination line widths (with instrumental and thermal components, assuming a $10^4$ K HIIR, removed),  together with data for other objects in Figure~\ref{fig:lumsig}. The luminosity of 30 Doradus in the LMC is taken from \citet{walborn91} and its integrated \ha\ line width

731: inferred from the spatially resolved echelle spectroscopy of

732: \citet{chu94}. The luminosity and \brg\ width for the ELC-like

733: source in NGC~5253 are from \citet{turner03}.  

734: %

735: For the UDHIIs in He $2-10$ the $Q[H^+]$ values are derived from the radio data of \citet{johnson03}, and the line width adopted is the typical value of about 55 \kms\ observed for \brg\ (corrected for instrumental and thermal contributions, which resembles bELC line widths) in several slits that overlap with several of the radio sources \citep{henry07}.

736: % added He 2-10 widths

737: The line widths for

738: a sample of (U)CHIIs  from \citet{garay99} are for H~{\sc i} radio recombination lines.

739: The sample of top-ranked  GHIIRs in nearby galaxies is from \citet{arsenault88}, who compiled \ha\ line widths, luminosities, rms $n_e$ values, and diameters from the literature.  We assume case B recombination to determine $Q[H^+]$ values from line fluxes.

740: %

741: Figure~\ref{fig:lumsig}  shows bELCs at the high end of 

742: the range in $Q[H^+]$ and FWHM, where they overlap with the GHIIRs, which are characterized by supersonic linewidths \citep{smith70} and $Q[H^+] \sim 10^{50}-10^{52}$ s$^{-1}$.  

743: %

744: % UDHIIs are not shown in the $Q[H^+]$-FWHM plot because line widths are not available for them, but their $Q[H^+]$ values are very high, $\sim 10^{52}$ s$^{-1}$ \citep[for the UDHIIs in He $2-10$,][]{kobulnicky99}, which would place them among the bELCs and brightest GHIIRs.

745: %

746: Galactic (U)CHIIs have much lower luminosities because they are powered by just one or a few massive stars \citep[\protect{$Q[H^+] \sim 10^{44}-10^{50}$ s$^{-1}$},][]{garay99}, but their line widths overlap with those of the brighter HIIRs. 

747: %

748: Not shown are common Galactic HIIRs, which exhibit narrower H recombination line FWHMs of $20-40$ km s$^{-1}$ and  $Q[H^+] = 4 \times 10^{47} -

749: 10^{51}$ s$^{-1}$ \citep{churchwell78}.  

750: %

751:  A loose correlation between FWHM and $Q[H^+]$ is evident, and its slope for bELCs is consistent with that reported for GHIIR populations of nearby spirals \citep[i.e. NGC~4449 and NGC~4321,][]{arsenault90,rozas98,fuentesmasip00a}.

752: 

753: We also compare measurements of size (diameter $d$) and $n_e$ for several types of HIIRs in Figure~\ref{fig:lumsig}.  The variations in $d$ and $n_e$ within the (U)CHII, UDHII, and bELC populations extend along tracks of constant $Q[H^+]$ for a constant-density Str\"omgren sphere: $n_e \propto d^{-1.5}$ (solid lines in Figure~\ref{fig:lumsig}), while GHIIRs cover a broader range in $Q[H^+]$ values.  While GHIIRs are similar to bELCs in terms of luminosity or inferred stellar mass, they are  larger ($d \gs 100 $ pc) and much less dense \citep[e.g. ELC B1 has \protect{$n_e = 10^{3.5}- 10^4$ cm$^{-3}$ from [Fe~{\sc iii}]} ratios,][and see below]{gilbert00}.  GHIIRs have rms densities of a few to a few $100$ cm$^{-3}$ (derived from emission measure profiles), but forbidden-line diagnostics, e.g. [S~{\sc ii}] and [O~{\sc ii}], yield densities that are ten times greater \citep[e.g.,][]{kennicutt84}.  This range of densities is commonly  interpreted as evidence for clumpiness of the HIIRs, and it suggests filling factors of dense gas of $0.01-0.1$ \citep{kennicutt84}.  Such a variable-density medium in GHIIRs is certainly plausible: inhomogeneities are common in resolved HIIRs like 30 Dor \citep[e.g.]{chu94}, and density variations have been measured in both Galactic and extragalactic HIIRs  \citep{copetti00,castaneda92}.

754: 

755: (U)CHIIs are included in Figure~\ref{fig:lumsig} because they represent extremes in all of the plotted parameters.  Although their low $Q[H^+]$ values, small implied stellar masses, and sub-parsec sizes clearly differentiate Galactic (U)CHIIs from GHIIRs and bELCs, their extremely high densities are similar to that of ELC B1:  compact (diameter $d < 0.5$ pc) and ultra-compact ($d < 0.05$ pc) HIIRs have $n_e$ ranging over $\sim 10^3 - 10^5$ cm$^{-3}$ \citep{wood89}.

756: Common Galactic HIIRs (not plotted) have comparable luminosities to those of (U)CHIIs, but they are larger ($0.5-15$ pc) and have slightly lower densities ($10^2 - 10^4$ cm$^{-3}$) than their compact cousins.

757: Moreover, UCHIIs have dynamical ages indicating extreme youth \citep[5000 years,][]{wood89}, although they could live much longer ($\sim 10^5$ years) \citep{garay99}, so the UCHIIs may be the precursors of common HIIRs.  The bELCs appear to be $3-7$ Myr old (\S~\ref{sec:agemass}) and therefore much more evolved than UCHIIs, yet they are clearly more compact and perhaps at an earlier stage of evolution than the diffuse GHIIRs.  

758: 

759: We estimate a volume-averaged $n_e$ for all ELCs by assuming ionization and recombination balance for a $10^4$ K HIIR with Q[$H^+$] (extinction-corrected, where possible) from Table~\ref{tab:agemass} along with their \brg\ sizes (i.e. the FWHM measured in our 2.17 \um\ narrow-band image, see Figure~\ref{fig:sizes}).

760:  The resulting values range from 30 to 600 cm$^{-3}$ with an average of 180

761: cm$^{-3}$, and they represent lower limits on the electron densities

762: of ELCs.  For the four bELCs that have thermal radio counterparts, \citet{neff00} compute similar rms $n_e$ values of $30-360$ cm$^{-3}$ by assuming measured diameters of up to $70-120$ pc (which in some cases gives an upper limit on size and lower limit on $n_e$ because the sources are not resolved by their $\sim 80$ pc beam). 

763: %

764: \citet{neff00} also consider the possibility that their thermal radio sources have $8-30$ pc diameters that are nearer to those observed for SSCs \citep[$\sim 8$ pc,][]{whitmore99}, and thereby obtain higher estimates of $n_e = 1 - 5 \times 10^3$ cm$^{-3}$.  If the $K$-band emission from ELCs is concentrated in compact, unresolved cores that

765: are the size of the optical SSCs, then the density estimates would increase correspondingly, to $700-8500$ cm$^{-3}$ with a mean of 2250 cm$^{-3}$; the maximum value is that of ELC B1, and it is consistent with the range inferred from [Fe~{\sc iii}] ratios by \citet{gilbert00}. 

766: Thus the radio-inferred densities for a subset of ELCs are comparable with the IR-inferred ones.  For consistency, the $n_e$ values plotted in Figure~\ref{fig:lumsig} are the lowest (rms $n_e$) estimates for GHIIRs, ELCs, and UDHIIs, which we discuss next.

767: 

768: We estimate densities for the UDHIIs of He

769: $2-10$ in the same way as for the ELCs, using the $Q[H^+]$ values from 0.7 cm data of

770: \citet{johnson03} and the 10\um\ sizes of \citet{vacca02}, where we

771: combine their objects 1 and 2 since they are blended in the observations of the latter work.

772: Thus we assume that the mid-IR and near-IR sizes are comparable, and that the stars and warm dust share the same volume.

773: We find densities for the two compact UDHIIs (4 and 5, with radii

774: of 4 and 5 pc) of 2200 and 1280 cm$^{-3}$.  For the two larger UDHII

775: regions (1+2 and 3, with radii of 16 and 17 pc), we find $n_e \approx

776: 390 $ and 180 cm$^{-3}$.  The lowest estimates of average ELC densities are thus similar to those

777: of the more extended UDHIIs in He $2-10$, and the SSC-size estimates are similar to those of the compact UDHIIs.  

778: 

779: However, \citet{vacca02} do not assume that the stars and warm dust in UDHIIs share the same volume; they estimate $n_e$ for the He $2-10$ UDHIIs by fitting the radio-NIR SEDs. \citet{vacca02} model a UDHII

780: as a central HIIR that envelopes a young SSC and extends out to the

781: inner radius of a surrounding annular dust cocoon that extends to an

782: outer radius.  Their best-fit inner radii are similar to the sizes of

783: optical SSCs, $2-5$ pc, and they are consistent with fits to more recent radio data of \citet{johnson03}.  This leads to average values for $n_e$ of

784: $1400-5720$ cm$^{-3}$, which agree well with the radio estimates of

785: the average $n_e$ from \citet{kobulnicky99} and \citet{johnson03}.

786: For the compact radio counterpart of the NGC~5253 HIIR,

787: \citet{turner00} report an even larger estimate of the rms $n_e$ of $

788: 4 \times 10^4$ cm$^{-3}$, which is in the range of UDHII $n_e$.

789: Adopting the higher estimates of $n_e$ for UDHIIs or bELCs\footnote{We discuss further evidence for high-electron-density material in bELCs in

790: Paper II, where we derive radial density distributions for the bELC

791: HIIRs by fitting the \brg\ profiles with a kinematic model for a

792: cluster outflow; this yields even higher values for the

793: emissivity-weighted average $n_e$: $10^3-10^5 $ cm$^{-3}$.}

794:  would move them toward to the (U)CHIIs in the $n_e-d$ panel of Figure~\ref{fig:lumsig}, while adopting the higher estimates for GHIIRs would move them up; the figure would still display a strong anticorrelation between $n_e$ and diameter in which bELCs and UDHIIs occupy a central region that is distinct from that of the other massive systems plotted, the GHIIRs.

795: 

796: % add a caveat about sample diffs/selection effects

797: Although the populations of HIIRs that are discussed above exist in a variety of environments and were all observed in different ways and subject to different selection effects from ELCs, their comparison shows how ELCs fit into the broader context of star formation.

798: %

799: Their nearest relatives are the UDHIIs, whose Lyman

800: continuum luminosities and electron densities are similar, but UDHIIs are so heavily obscured that they are rarely visible in the optical range and are thought to be younger than ELCs.  UDHIIs are classified based on their radio properties, and ELCs based on their recombination lines, but we have already conflated the two groups by assuming the \brg\ line fluxes of \citet{henry07} emerge from the radio/MIR sources of \citet{kobulnicky99} and \citet{vacca02}.  That is because all evidence suggests that ELCs and UDHIIs represent the same objects, perhaps at slightly different evolutionary stages, so that massive clusters form within gas- and dust-shrouded UDHIIs, which become ELCs as their massive hot stars drive winds (that are visible in recombination line emission) and expel gas to reveal young SSCs with low extinctions at visible wavelengths.  If these massive clusters form hierarchically from clusters of subclusters \citep[e.g.][]{testi00,bonnell03}, then the ELCs may be comprised of several evolving UDHIIs that our data cannot resolve.

801: %

802: We conclude that bELCs and UDHIIs have some properties in common with

803: other types of HIIRs, but taken together, their broad

804: lines, high densities, intermediate sizes, and large luminosities constitute a new class

805: of massive HIIRs that is distinct from GHIIRs.  

806: % "objects" changed to "massive hiirs"

807: 

808: % added par:

809: While our small sample spans two orders of magnitude in $Q[H^+]$ and inferred cluster mass, it may represent only the bright end of the full young cluster population in the Antennae.  If the mass function of young clusters in the Antennae starburst reaches down to single-star scales, would those objects resemble Galactic UCHIIs?  The analogy between UCHIIs and UDHIIs may extend beyond extreme densities, compactness, and extinctions:  the massive stars within UCHIIs are thought to spend $10-20$ \%  of their main-sequence lifetimes in an embedded phase \citep{wood89}, and those in UDHIIs may spend even longer fractions of their lives hidden from optical view, e.g. 15 \% in He~$2-10$ \citep[][]{kobulnicky99} to 40 \% in NGC~5253 \citep[][]{martinhernandez05}, until their cluster winds remove much of the obscuring gas.  Such an extended embedded phase for massive stars in starbursts is proposed  to explain discrepancies between measured mid-IR line ratios and stellar population synthesis models with normal upper mass limits by \citet{rigby04}, who point out that the higher ISM pressures in starbursts should lengthen their embedded compact HIIR phase relative to that in the Galaxy.  The implications of a long embedded phase in which IR nebular lines are suppressed are serious for starburst population synthesis modeling \citep{rigby04}.

810: % and the interpretation of the correlation between UV slope and IR-to-UV flux ratio in starbursts \citep{vacca02}.

811: 

812: 

813: \section{Evolution of ELCs}

814: \label{sec:evolution}

815: 

816: % mean sizes shift a bit

817: In an evolutionary sequence determined by cluster age and embeddedness, massive stellar clusters form in heavily obscured UDHIIs, evacuate their surroundings as wind-blowing ELCs, and, if they survive the mass-loss phase, emerge as SSCs, which may ultimately become GCs if they survive further destructive cluster dynamical processes and tidal interactions with their host galaxy.  This scenario requires some explanation of cluster size evolution, in particular how the large $K$-band radii of ELCs ($10-50$ pc, Table~\ref{tab:phottable} and Figure~\ref{fig:sizes}) can be reconciled with the typical half-light radii of optically revealed SSCs \citep[$\sim 4$ pc in the Antennae,][]{whitmore99}.  We assume that ELCs are young SSCs embedded in HIIRs because their inferred stellar masses are similar to those of SSCs, which can be factors of a few to ten times more compact than ELCs.  Figure~\ref{fig:sizes} shows a wide scatter in the 2.25 \um\ stellar continuum sizes of 10 ELCs at ages of $6-7$ Myr: their mean ($\pm$ standard deviation) half-light radius is $17 \pm 8$ pc.  Younger ELCs (5 have ages of $3-6$ Myr) have a larger mean half-light radius of $35 \pm 11$ pc at 2.25 \um.   Additional evidence for a decrease in Antennae cluster size with age, as well as a large scatter in size, is presented by \citet{mengel05}, who report average ($\pm$ standard deviation) WFPC2 $I$-band effective radii for 18 isolated clusters with ages below 4 Myr and in the range $8-11$ Myr of $16 \pm 15$ pc and $6.5 \pm 5.3$ pc, respectively.  

818: 

819: This tendency of older clusters to be more compact could be explained by cluster contraction with age or by the preferential survival of more concentrated clusters at any given age.  Clusters that form hierarchically would contract as subclusters merge to form a single cluster.  Simulations by \citet{bonnell03} show that star clusters may form hierarchically, such that fragmentation in a molecular cloud produces sheets and filaments which form stars before the rest of the cloud material, and these stars form high-density subclusters that merge after several free-fall times to create a centrally condensed cluster.  Observations of the Serpens molecular cloud core provided the first evidence for hierarchical cluster formation: \citet{testi00} find that star formation is not homogeneous in the core, but rather is concentrated in dense subclusters that are well separated spatially and kinematically.  If ELCs are young enough to be undergoing the final merger of their initial subclusters, this might explain their size trend in Figure~\ref{fig:sizes}.  

820: 

821: However, even if hierarchical formation is still underway in ELCs and causing contraction, it must compete with the countervailing influence of mass loss due to stellar winds that lead to cluster outflows.  Cluster expansion is caused by the loss of gas via stellar evolution and stars via relaxation, tidal stripping, and dynamical friction.  Mass loss and consequent expansion can unbind a cluster and cause it to dissolve below detection limits into the background stellar population \citep[see][for an analysis of these effects across the cluster mass spectrum]{fall01}; hence compact clusters live longer than less compact ones, and they increasingly dominate the cluster population as they age.  Two recent studies of the age distribution of stellar clusters in the Antennae argue that the sharp decrease in the number of clusters as a function of age is strong evidence for rapid dissolution of the majority of young clusters on timescales of order 10 Myr \citep{mengel05,fall05}.  Thus we favor expansion and rapid dissolution over contraction due to hierarchical formation in explaining the smaller sizes of older ELCs.

822: 

823: Large ELCs may therefore be puffed-up young SSCs that have expanded due to prodigious mass loss, which limits their chances for survival if they are not compact enough, if their IMF is top-heavy or truncated at low masses, or if their star-formation efficiency (SFE) is not high enough.  \citet{boily03b} simulate cluster evolution in the presence of rapid (i.e. instantaneous) mass loss; after equilibrium is re-established, they find expansion factors for the half-mass radii of up to a factor of ten or more for bound clusters having Plummer and King distribution functions with varied concentrations and assumed SFEs.  This is a plausible explanation for the sizes of ELCs if they were born as more compact clusters within UDHIIs; however, the mass loss in ELCs proceeds over the course of at least several Myr, so they would not suffer such extreme expansion as the clusters simulated by \citet{boily03b}. 

824: 

825: The expansion of young clusters due to dynamical evolution is observed in nearby resolved clusters: cluster core radii in the LMC and SMC increase with age, at varying rates, up to ages of at least 1 Gyr \citep{elson89,mackey03b,mackey03a}.  These young clusters ($10^6-10^9$ yrs old) have power-law luminosity profiles that often show bumps or dips that suggest the presence of substructures that have not been smoothed out by relaxation \citep{elson87,mackey03a,mackey03b}.  These profiles do not display the tidal truncation of a King profile \citep{king66}, which best describes the luminosity profiles of much more evolved GCs.  There is evidence that the more massive SSCs in nearby galaxies also have power-law light profiles, some extending beyond 100 pc in radius, some with clumpy substructure \citep[see references in][]{schweizer04}; evidence for the onset of tidal truncation in older SSCs (beyond a few 100 Myr) and intermediate-age GCs has also been found \citep{whitmore99, schweizer04}.  \citet{larsen04c} finds that extended envelopes relative to King profiles are common among the youngest clusters in a sample of spiral galaxies observed with HST WFPC2, but his sample does not show the trend of increasing cluster radii with age that is seen in the LMC and SMC. 

826: 

827: The interpretation of large ELCs as puffed-up SSCs that are caught in the act of dissolving assumes that their $K$-band light at large radii is dominated by stellar continuum emission.  Alternative explanations for their sizes might, however, involve dust scattering and emission from a shell at larger radii surrounding an SSC-sized cluster, 

828: or differential extinction that preferentially obscures the cluster core relative to its outer radii, making the observed FWHM larger than the intrinsic one.  A final possibility is that SSCs may form hierarchically from clusters of smaller subclusters that merge rapidly to make a single massive cluster \citep[within a couple of free-fall times for a low-mass cluster, ][]{bonnell03};  in this picture, prior to the final merger into a single cluster, the average stellar density  is smaller than in individual subclusters, and the overall structure is probably not spherically symmetric.  Observations at higher angular resolution at infrared wavelengths, either space-based or adaptive optics-assisted, will be able to test these scenarios.

829: 

830: \section{Star Formation in Massive Clusters}

831: \label{sec:sfr}

832: 

833: Comparing the star-formation rate (SFR) in ELCs with that of the entire system is difficult

834: because the SFRs of galaxies are determined from observed quantities

835: such as \ha\ and far-infrared (FIR) luminosities averaged over large

836: spatial scales, and they consider continuous star formation

837: % and a Salpeter IMF down to 0.1 \Msun\ 

838: \citep[e.g.][]{kennicutt98}.  Assuming the calibrations of

839: \citet{kennicutt98}, adopting an IRAS luminosity of $10^{11}$ \Lsun\ 

840: for the Antennae leads to an SFR$_{\rm FIR} \approx 17$ \Msun\ 

841: yr$^{-1}$; \citet{zhang01} use Kennicutt's \ha\ calibration to

842: estimate SFR$_{\ha} \approx 20$ \Msun~yr$^{-1}$.  If we apply this

843: \ha\ calibration to the sample of ELCs by assuming case B

844: recombination, we derive a total SFR$_{\ha}$ of 3.2 \Msun\ yr$^{-1}$.

845: Thus the fraction of the recent SFR of the Antennae

846: that is provided by ELCs is at least 15\%, and half of this

847: contribution is provided by ELC B1 alone.  This is a lower limit because we have neglected extinction in two-thirds of the ELCs, our sample probably does not include all ELCs, and there could be a significant escape fraction for ionizing photons from the clusters.

848: %

849: From a photometric study of the current star cluster population in the Antennae, \citet{mengel05} estimate a total SFR in clusters over the past 25 Myr of at least 20 \Msun\ yr$^{-1}$ (in agreement with the \ha\ estimate for all star formation), but they argue that the age distribution of SSCs implies rapid dissolution of most clusters, and that accounting for this could raise the SFR to at least 50 \Msun\ yr$^{-1}$.

850: 

851: An estimate of the fraction of stars that are formed in clusters rather than in the field may be derived from imaging observations that compare amounts of discrete and diffuse emission from

852: star-formation tracers: \citet{fischer96} reported that about 70\% of

853: the \brg\ emission in a 30\arcsec\ aperture including the nucleus of

854: NGC~4039 and the southern overlap region is diffuse, while the

855: remaining 30\% is associated with \ha\ and radio knots.  Similarly,

856: \citet{hummel86} report that the total fraction of radio continuum

857: emission from discrete knots (at 6\arcsec\ resolution) is about 35\%,

858: with errors of $20-30$\%.  Our lower limit to the fraction of the

859: star-formation rate in ELCs is about half of what one might expect

860: from these imaging studies.  These results are consistent with the range deduced from UV observations of samples of starbursts: about 15--50\% of the UV light

861: is typically associated with stellar clusters

862: \citep[e.g.][]{meurer95b,maoz96}.  Radio continuum imaging of the

863: Wolf-Rayet starburst He 2-10 reveals that 15\% of the O stars in the

864: galaxy reside in UDHIIs \citep{kobulnicky99}, which are probably the

865: progenitors of ELCs and SSCs. Since our measurement is a lower limit

866: on the fraction of star formation in Antennae clusters, it appears

867: that ELCs constitute the majority of the {\it clustered} star-formation rate

868: in the Antennae, and at least 15\% of the \ha-derived total SFR.

869: 

870: The actual proportion of clustered star formation in the Antennae may be 

871: nearer to 100\% if many of the young SSCs are quickly dissolved or disrupted \citep[e.g.][]{mengel05,fall05}.  \citet{whitmore99} find a power-law form for the luminosity function for Antennae SSCs that closely resembles that of other SSCs in other starbursts \citep[see e.g.][]{meurer95a,larsen02} as well as that of OB associations and GHIIRs \citep[e.g.][]{kennicutt89,mckee97}.  Many of the low-mass SSCs must dissolve into the field population in order for a power-law SSC population to evolve into a log-normal globular cluster population; simulations indicate that this is plausible \citep{fall01}, and observations have begun to reveal the process \citep[e.g.][]{degrijs03,lamers05}.

872: Indeed, \citet{mengel05} and \citet{fall05} argue that rapid dissolution into the field population is the normal fate for SSCs in the Antennae; those that survive mass loss in the ELC phase are subject to further destructive dynamical and evolutionary processes that may prevent them from becoming GCs.  

873: 

874: \section{Conclusions}

875: \label{sec:conclusions}

876: High-resolution spectroscopy of the youngest SSCs in the Antennae

877: reveals broad, spatially extended, non-Gaussian, supersonic \brg\ 

878: emission ($60-110$ km s$^{-1}$ FWHM) in the ELCs.  Typical GHIIRs also have supersonic linewidths and

879: luminosities at the low end of the ELC ranges, but they are much

880: larger and less dense than ELCs, and are generally excited by

881: lower-mass clusters and OB associations.  Thus ELCs appear to

882: represent a different class of dense extragalactic HIIR that is powered by young SSCs and may be associated with high-pressure environments.  Their closest relatives

883: are the UDHIIs, which are younger (age $\ls 1$ Myr), more heavily

884: embedded radio and mid-IR nebulae with high electron densities. UDHII Lyman continuum luminosities and infrared sizes are similar to those of ELCs, so UDHIIs

885: require similar stellar masses for their embedded ionizing clusters as ELCs,

886: $10^5-10^7$ \Msun.  Thus UDHIIs are the probable precursors of ELC.  

887: 

888: The youngest ELCs (ages $3-6$ Myr) are concentrated in the southern

889: overlap region of the Antennae, while those in the northern overlap

890: region are slightly older ($6-7$ Myr).  The rms radial velocity

891: dispersion of the youngest ELCs is consistent with

892: that measured for molecular gas via CO in the same region, so the

893: young clusters and dense gas share similar kinematics.  ELCs in our incomplete sample comprise about 15\% of the current SFR in the Antennae (half of that is concentrated in the brightest cluster) and they make up a significant fraction of the expected clustered star formation in the system.  

894: 

895: Because bELC \brg\ linewidths typically exceed the most conservative estimates of the cluster escape velocities, much of the H~{\sc ii} gas is not bound to the

896: clusters--it is flowing out in a cluster wind.  The greater extent of \brg\ emission relative to continuum emission is consistent with escape of ionized gas in winds.

897: Although this mass loss takes place over the course of several Myr rather than instantaneously, it may be dramatic enough to prevent some of the ELCs from surviving as bound SSCs; their large $K$-band sizes and the decrease in mean ELC size with age may be the signature of rapid cluster expansion and dissolution, which leads to the preferential survival of the most compact clusters in each age group.  In Paper II we present a kinematic model for cluster outflows to infer their mass-loss rates and energetics, and discuss ELC feedback in the context of similar systems such as wind-blown superbubbles and galactic

898: superwinds.

899: 

900: 

901: %%%%%%%%%%%%%%%%%%%

902: % Acknowledgments %

903: %%%%%%%%%%%%%%%%%%%

904: 

905: \acknowledgments

906: 

907: We thank M. Lehnert and W. Vacca for discussions and comments, and we thank the 

908: referee for comments that improved the paper. 

909: We also  thank the staffs of the European Southern Observatory and the

910: Keck Observatory, and

911: observing assistant Ron Quick in particular.  The authors wish to

912: recognize and acknowledge the very significant cultural role and

913: reverence that the summit of Mauna Kea has always had within the

914: indigenous Hawaiian community.  We are most fortunate to have the

915: opportunity to conduct observations from this mountain.  

916: 

917: This work was supported in part by National Science Foundation (NSF)

918: grant AST-0205999, a NASA GSRP fellowship, and the NSF Science and

919: Technology Center for Adaptive Optics, managed by the University of

920: California at Santa Cruz under cooperative agreement No. AST-9876783.

921: Part of this work was performed under the auspices of the U.S. Department of Energy,

922: National Nuclear Security Administration by the University of California,

923: Lawrence Livermore National Laboratory under contract No. W-7405-Eng-48.

924: 

925: %%%%%%%%%%%%%%

926: % References %

927: %%%%%%%%%%%%%%

928: 

929: %\bibliographystyle{apj} 

930: %\bibliography{ssc}

931: 

932: 

933: \begin{thebibliography}{110}

934: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

935: 

936: \bibitem[{{Abbott} \& {Conti}(1987)}]{abbott87}

937: {Abbott}, D.~C. \& {Conti}, P.~S. 1987, \araa, 25, 113

938: 

939: \bibitem[{{Alonso-Herrero} {et~al.}(2002){Alonso-Herrero}, {Rieke}, {Rieke}, \&

940:   {Scoville}}]{alonso02}

941: {Alonso-Herrero}, A., {Rieke}, G.~H., {Rieke}, M.~J., \& {Scoville}, N.~Z.

942:   2002, \aj, 124, 166

943: 

944: \bibitem[{{Alonso-Herrero} {et~al.}(2004){Alonso-Herrero}, {Takagi}, {Baker},

945:   {Rieke}, {Rieke}, {Imanishi}, \& {Scoville}}]{alonsoherrero04}

946: {Alonso-Herrero}, A., {Takagi}, T., {Baker}, A.~J., {Rieke}, G.~H., {Rieke},

947:   M.~J., {Imanishi}, M., \& {Scoville}, N.~Z. 2004, \apj, 612, 222

948: 

949: \bibitem[{{Amram} {et~al.}(1992){Amram}, {Marcelin}, {Boulesteix}, \& {Le

950:   Coarer}}]{amram92}

951: {Amram}, P., {Marcelin}, M., {Boulesteix}, J., \& {Le Coarer}, E. 1992, \aap,

952:   266, 106

953: 

954: \bibitem[{{Arsenault} \& {Roy}(1988)}]{arsenault88}

955: {Arsenault}, R. \& {Roy}, J.-R. 1988, \aap, 201, 199

956: 

957: \bibitem[{{Arsenault} {et~al.}(1990){Arsenault}, {Roy}, \&

958:   {Boulesteix}}]{arsenault90}

959: {Arsenault}, R., {Roy}, J.-R., \& {Boulesteix}, J. 1990, \aap, 234, 23

960: 

961: \bibitem[{{Ashman} \& {Zepf}(1992)}]{ashman92}

962: {Ashman}, K.~M. \& {Zepf}, S.~E. 1992, \apj, 384, 50

963: 

964: \bibitem[{{B{\" o}ker} {et~al.}(1999){B{\" o}ker}, {van der Marel}, \&

965:   {Vacca}}]{boeker99}

966: {B{\" o}ker}, T., {van der Marel}, R.~P., \& {Vacca}, W.~D. 1999, \aj, 118, 831

967: 

968: \bibitem[{{Beck} {et~al.}(2001){Beck}, {Turner}, \& {Gorjian}}]{beck01}

969: {Beck}, S.~C., {Turner}, J.~L., \& {Gorjian}, V. 2001, \aj, 122, 1365

970: 

971: \bibitem[{{Beck} {et~al.}(2000){Beck}, {Turner}, \& {Kovo}}]{beck00}

972: {Beck}, S.~C., {Turner}, J.~L., \& {Kovo}, O. 2000, \aj, 120, 244

973: 

974: \bibitem[{{Beckman} {et~al.}(2000){Beckman}, {Rozas}, {Zurita}, {Watson}, \&

975:   {Knapen}}]{beckman00}

976: {Beckman}, J.~E., {Rozas}, M., {Zurita}, A., {Watson}, R.~A., \& {Knapen},

977:   J.~H. 2000, \aj, 119, 2728

978: 

979: %\bibitem[{{Binney} \& {Merrifield}(1998)}]{binney98} {Binney}, J. \& {Merrifield}, M. 1998, Galactic Astronomy (Princeton: Princeton University Press)

980: 

981: \bibitem[{{Boily} \& {Kroupa}(2003)}]{boily03b}

982: {Boily}, C.~M. \& {Kroupa}, P. 2003, \mnras, 338, 673

983: 

984: \bibitem[{{Boily} {et~al.}(2005){Boily}, {Lan{\c c}on}, {Deiters}, \&

985:   {Heggie}}]{boily05}

986: {Boily}, C.~M., {Lan{\c c}on}, A., {Deiters}, S., \& {Heggie}, D.~C. 2005,

987:   \apjl, 620, L27

988: 

989: \bibitem[{{Bonnell} {et~al.}(2003){Bonnell}, {Bate}, \& {Vine}}]{bonnell03}

990: {Bonnell}, I.~A., {Bate}, M.~R., \& {Vine}, S.~G. 2003, \mnras, 343, 413

991: 

992: \bibitem[{{Bottema}(1993)}]{bottema93}

993: {Bottema}, R. 1993, \aap, 275, 16

994: 

995: \bibitem[{{Cabanac} {et~al.}(2005){Cabanac}, {Vanzi}, \& {Sauvage}}]{cabanac05}

996: {Cabanac}, R.~A., {Vanzi}, L., \& {Sauvage}, M. 2005, \apj, 631, 252

997: 

998: %\bibitem[{{Calzetti}(1997)}]{calzetti97} {Calzetti}, D. 1997, \aj, 113, 162

999: 

1000: \bibitem[{{Castaneda} {et~al.}(1992){Castaneda}, {Vilchez}, \&

1001:   {Copetti}}]{castaneda92}

1002: {Castaneda}, H.~O., {Vilchez}, J.~M., \& {Copetti}, M.~V.~F. 1992, \aap, 260,

1003:   370

1004: 

1005: \bibitem[{{Castellanos} {et~al.}(2002){Castellanos}, {D{\' i}az}, \&

1006:   {Tenorio-Tagle}}]{castellanos02}

1007: {Castellanos}, M., {D{\' i}az}, {\' A}.~I., \& {Tenorio-Tagle}, G. 2002, \apjl,

1008:   565, L79

1009: 

1010: \bibitem[{{Chu} \& {Kennicutt}(1994)}]{chu94}

1011: {Chu}, Y. \& {Kennicutt}, R.~C. 1994, \apj, 425, 720

1012: 

1013: \bibitem[{{Churchwell} {et~al.}(1978){Churchwell}, {Smith}, {Mathis}, {Mezger},

1014:   \& {Huchtmeier}}]{churchwell78}

1015: {Churchwell}, E., {Smith}, L.~F., {Mathis}, J., {Mezger}, P.~G., \&

1016:   {Huchtmeier}, W. 1978, \aap, 70, 719

1017: 

1018: \bibitem[{{Copetti} {et~al.}(2000){Copetti}, {Mallmann}, {Schmidt}, \&

1019:   {Casta{\~ n}eda}}]{copetti00}

1020: {Copetti}, M.~V.~F., {Mallmann}, J.~A.~H., {Schmidt}, A.~A., \& {Casta{\~

1021:   n}eda}, H.~O. 2000, \aap, 357, 621

1022: 

1023: \bibitem[{{de Grijs} {et~al.}(2003){de Grijs}, {Bastian}, \&

1024:   {Lamers}}]{degrijs03}

1025: {de Grijs}, R., {Bastian}, N., \& {Lamers}, H.~J.~G.~L.~M. 2003, \apjl, 583,

1026:   L17

1027: 

1028: \bibitem[{{Dove} {et~al.}(2000){Dove}, {Shull}, \& {Ferrara}}]{dove00}

1029: {Dove}, J.~B., {Shull}, J.~M., \& {Ferrara}, A. 2000, \apj, 531, 846

1030: 

1031: \bibitem[{{Elson} {et~al.}(1987){Elson}, {Fall}, \& {Freeman}}]{elson87}

1032: {Elson}, R.~A.~W., {Fall}, S.~M., \& {Freeman}, K.~C. 1987, \apj, 323, 54

1033: 

1034: \bibitem[{{Elson} {et~al.}(1989){Elson}, {Freeman}, \& {Lauer}}]{elson89}

1035: {Elson}, R.~A.~W., {Freeman}, K.~C., \& {Lauer}, T.~R. 1989, \apjl, 347, L69

1036: 

1037: \bibitem[{{Fall} {et~al.}(2005){Fall}, {Chandar}, \& {Whitmore}}]{fall05}

1038: {Fall}, S.~M., {Chandar}, R., \& {Whitmore}, B.~C. 2005, \apjl, 631, L133

1039: 

1040: \bibitem[{{Fall} \& {Zhang}(2001)}]{fall01}

1041: {Fall}, S.~M. \& {Zhang}, Q. 2001, \apj, 561, 751

1042: 

1043: \bibitem[{{Figer} {et~al.}(2000){Figer}, {Becklin}, {McLean}, {Gilbert},

1044:   {Graham}, {Larkin}, {Levenson}, {Teplitz}, {Wilcox}, \& {Morris}}]{figer00}

1045: {Figer}, D.~F., {Becklin}, E.~E., {McLean}, I.~S., {Gilbert}, A.~M., {Graham},

1046:   J.~R., {Larkin}, J.~E., {Levenson}, N.~A., {Teplitz}, H.~I., {Wilcox}, M.~K.,

1047:   \& {Morris}, M. 2000, \apjl, 533, L49

1048: 

1049: \bibitem[{{Figer} {et~al.}(1997){Figer}, {McLean}, \& {Najarro}}]{figer97}

1050: {Figer}, D.~F., {McLean}, I.~S., \& {Najarro}, F. 1997, \apj, 486, 420

1051: 

1052: \bibitem[{{Fischer} {et~al.}(1996){Fischer}, {Shier}, {Luhman}, {Satyapal},

1053:   {Smith}, {Stacey}, {Unger}, {Greenhouse}, {Spinoglio}, {Malkan}, {Lord},

1054:   {Miles}, {Shure}, {Clegg}, {Ade}, {Armand}, {Burgdorf}, {Church}, {Davis},

1055:   {di Giorgio}, {Ewart}, {Furniss}, {Glencross}, {Gry}, {Lim}, {Molinari},

1056:   {Nguyen-Q-Rieu}, {Price}, {Sidher}, {Smith}, {Swinyard}, {Texier}, {Trams},

1057:   \& {Wolfire}}]{fischer96}

1058: {Fischer}, J., {Shier}, L.~M., {Luhman}, M.~L., {Satyapal}, S., {Smith}, H.~A.,

1059:   {Stacey}, G.~J., {Unger}, S.~J., {Greenhouse}, M.~A., {Spinoglio}, L.,

1060:   {Malkan}, M.~A., {Lord}, S.~D., {Miles}, J.~W., {Shure}, M.~A., {Clegg},

1061:   P.~E., {Ade}, P.~A.~R., {Armand}, C., {Burgdorf}, M., {Church}, S.~E.,

1062:   {Davis}, G.~R., {di Giorgio}, A., {Ewart}, D., {Furniss}, I., {Glencross},

1063:   W.~M., {Gry}, C., {Lim}, T., {Molinari}, S., {Nguyen-Q-Rieu}, {Price}, M.~C.,

1064:   {Sidher}, S.~D., {Smith}, A., {Swinyard}, B.~M., {Texier}, D., {Trams},

1065:   N.~R., \& {Wolfire}, M.~G. 1996, \aap, 315, L97

1066: 

1067: \bibitem[{{Fritze-v.~Alvensleben} \& {Gerhard}(1994)}]{fritzevalvensleben94}

1068: {Fritze-v.~Alvensleben}, U. \& {Gerhard}, O.~E. 1994, \aap, 285, 775

1069: 

1070: \bibitem[{{Fuentes-Masip} {et~al.}(2000){Fuentes-Masip}, {Mu{\~ n}oz-Tu{\~

1071:   n}{\' o}n}, {Casta{\~ n}eda}, \& {Tenorio-Tagle}}]{fuentesmasip00a}

1072: {Fuentes-Masip}, O., {Mu{\~ n}oz-Tu{\~ n}{\' o}n}, C., {Casta{\~ n}eda}, H.~O.,

1073:   \& {Tenorio-Tagle}, G. 2000, \aj, 120, 752

1074: 

1075: \bibitem[{{Garay} \& {Lizano}(1999)}]{garay99}

1076: {Garay}, G. \& {Lizano}, S. 1999, \pasp, 111, 1049

1077: 

1078: %\bibitem[{{Gerssen} {et~al.}(1997){Gerssen}, {Kuijken}, \& {Merrifield}}]{gerssen97} {Gerssen}, J., {Kuijken}, K., \& {Merrifield}, M.~R. 1997, \mnras, 288, 618

1079: 

1080: \bibitem[{{Gilbert}(2002)}]{gilbert02}

1081: {Gilbert}, A.~M. 2002, Ph.D.~Thesis, University of California, Berkeley

1082: 

1083: \bibitem[{{Gilbert} \& {Graham}(2001)}]{gilbert01b}

1084: {Gilbert}, A.~M. \& {Graham}, J.~R. 2001, in IAU Symp. Ser. 207: Extragalactic

1085:   Star Clusters (Astronomical Society of the Pacific), 471

1086: 

1087: \bibitem[{{Gilbert} \& {Graham}(2007)}]{gilbert07b}

1088: {Gilbert}, A.~M. \& {Graham}, J.~R. 2007, \apj\ submitted

1089: 

1090: \bibitem[{{Gilbert} {et~al.}(2000){Gilbert}, {Graham}, {McLean}, {Becklin},

1091:   {Figer}, {Larkin}, {Levenson}, {Teplitz}, \& {Wilcox}}]{gilbert00}

1092: {Gilbert}, A.~M., {Graham}, J.~R., {McLean}, I.~S., {Becklin}, E.~E., {Figer},

1093:   D.~F., {Larkin}, J.~E., {Levenson}, N.~A., {Teplitz}, H.~I., \& {Wilcox},

1094:   M.~K. 2000, \apjl, 533, L57

1095: 

1096: \bibitem[{{Grebel} \& {Chu}(2000)}]{grebel00}

1097: {Grebel}, E.~K. \& {Chu}, Y. 2000, \aj, 119, 787

1098: 

1099: \bibitem[{{Hanson} {et~al.}(1996){Hanson}, {Conti}, \& {Rieke}}]{hanson96}

1100: {Hanson}, M.~M., {Conti}, P.~S., \& {Rieke}, M.~J. 1996, \apjs, 107, 281

1101: 

1102: \bibitem[{{Harris}(1991)}]{harris91}

1103: {Harris}, W.~E. 1991, \araa, 29, 543

1104: 

1105: \bibitem[{{Heckman} {et~al.}(1995){Heckman}, {Dahlem}, {Lehnert}, {Fabbiano},

1106:   {Gilmore}, \& {Waller}}]{heckman95}

1107: {Heckman}, T.~M., {Dahlem}, M., {Lehnert}, M.~D., {Fabbiano}, G., {Gilmore},

1108:   D., \& {Waller}, W.~H. 1995, \apj, 448, 98

1109:   

1110: \bibitem[{Henry} {et~al.}(2007){Henry}, {Turner}, {Beck}, {Crosthwaite}, \& {Meier}]{henry07}

1111: {Henry}, A.~L., {Turner}, J.~L., {Beck}, S.~C., {Crosthwaite}, L.~P., \& {Meier}, D.~S. 2007, \aj, 133, 757

1112: 

1113: \bibitem[{{Ho} \& {Filippenko}(1996)}]{ho96a}

1114: {Ho}, L.~C. \& {Filippenko}, A.~V. 1996, \apjl, 466, L83

1115: 

1116: \bibitem[{{Hummel} \& {Van Der Hulst}(1986)}]{hummel86}

1117: {Hummel}, E. \& {Van Der Hulst}, J.~M. 1986, \aap, 155, 151

1118: 

1119: \bibitem[{{Hummer} \& {Storey}(1987)}]{hummer87}

1120: {Hummer}, D.~G. \& {Storey}, P.~J. 1987, \mnras, 224, 801

1121: 

1122: \bibitem[{{Johnson} \& {Kobulnicky} (2003)}]{johnson03}

1123: {Johnson}, K.~E.  \& {Kobulnicky}, H.~A. 2003, \apj, 597, 923

1124: 

1125: \bibitem[{{Johnson} {et~al.}(2000){Johnson}, {Leitherer}, {Vacca}, \&

1126:   {Conti}}]{johnson00}

1127: {Johnson}, K.~E., {Leitherer}, C., {Vacca}, W.~D., \& {Conti}, P.~S. 2000, \aj,

1128:   120, 1273

1129: 

1130: \bibitem[{{Kennicutt} \& {Chu}(1988)}]{kennicutt88}

1131: {Kennicutt}, R.~C. \& {Chu}, Y.-H. 1988, \aj, 95, 720

1132: 

1133: \bibitem[{{Kennicutt}(1984)}]{kennicutt84}

1134: {Kennicutt}, R.~C. 1984, \apj, 287, 116

1135: 

1136: \bibitem[{{Kennicutt}(1998)}]{kennicutt98}

1137: ---. 1998, \apj, 498, 541

1138: 

1139: \bibitem[{{Kennicutt} {et~al.}(1989){Kennicutt}, {Edgar}, \&

1140:   {Hodge}}]{kennicutt89}

1141: {Kennicutt}, R.~C., {Edgar}, B.~K., \& {Hodge}, P.~W. 1989, \apj, 337, 761

1142: 

1143: \bibitem[{{King}(1966)}]{king66}

1144: {King}, I.~R. 1966, \aj, 71, 64

1145: 

1146: \bibitem[{{Kobulnicky} \& {Johnson}(1999)}]{kobulnicky99}

1147: {Kobulnicky}, H.~A. \& {Johnson}, K.~E. 1999, \apj, 527, 154

1148: 

1149: \bibitem[{{Kroupa}(2001)}]{kroupa01}

1150: {Kroupa}, P. 2001, \mnras, 322, 231

1151: 

1152: \bibitem[{{Lamers} {et~al.}(2005){Lamers}, {Gieles}, \& {Portegies

1153:   Zwart}}]{lamers05}

1154: {Lamers}, H., {Gieles}, M., \& {Portegies Zwart}, S. 2005, \aap, 429, 173

1155: 

1156: \bibitem[{{Larsen}(2002)}]{larsen02}

1157: {Larsen}, S.~S. 2002, \aj, 124, 1393

1158: 

1159: \bibitem[{{Larsen}(2004)}]{larsen04c}

1160: {Larsen}, S.~S. 2004, \aap, 416, 537

1161: 

1162: \bibitem[{{Larsen} {et~al.}(2001){Larsen}, {Brodie}, {Elmegreen}, {Efremov},

1163:   {Hodge}, \& {Richtler}}]{larsen01}

1164: {Larsen}, S.~S., {Brodie}, J.~P., {Elmegreen}, B.~G., {Efremov}, Y.~N.,

1165:   {Hodge}, P.~W., \& {Richtler}, T. 2001, \apj, 556, 801

1166: 

1167: \bibitem[{{Leitherer} {et~al.}(1999){Leitherer}, {Schaerer}, {Goldader},

1168:   {Delgado}, {Robert}, {Kune}, {de Mello}, {Devost}, \&

1169:   {Heckman}}]{leitherer99}

1170: {Leitherer}, C., {Schaerer}, D., {Goldader}, J.~D., {Delgado}, R.~M.~G.,

1171:   {Robert}, C., {Kune}, D.~F., {de Mello}, D.~F., {Devost}, D., \& {Heckman},

1172:   T.~M. 1999, \apjs, 123, 3

1173: 

1174: \bibitem[{{Leitherer} {et~al.}(1996){Leitherer}, {Vacca}, {Conti},

1175:   {Filippenko}, {Robert}, \& {Sargent}}]{leitherer96a}

1176: {Leitherer}, C., {Vacca}, W.~D., {Conti}, P.~S., {Filippenko}, A.~V., {Robert},

1177:   C., \& {Sargent}, W.~L.~W. 1996, \apj, 465, 717

1178: 

1179: \bibitem[{{Mackey} \& {Gilmore}(2003{\natexlab{a}})}]{mackey03b}

1180: {Mackey}, A.~D. \& {Gilmore}, G.~F. 2003{\natexlab{a}}, \mnras, 338, 120

1181: 

1182: \bibitem[{{Mackey} \& {Gilmore}(2003{\natexlab{b}})}]{mackey03a}

1183: ---. 2003{\natexlab{b}}, \mnras, 338, 85

1184: 

1185: \bibitem[{{Maoz} {et~al.}(1996){Maoz}, {Barth}, {Sternberg}, {Filippenko},

1186:   {Ho}, {Macchetto}, {Rix}, \& {Schneider}}]{maoz96}

1187: {Maoz}, D., {Barth}, A.~J., {Sternberg}, A., {Filippenko}, A.~V., {Ho}, L.~C.,

1188:   {Macchetto}, F.~D., {Rix}, H.-W., \& {Schneider}, D.~P. 1996, \aj, 111, 2248

1189: 

1190: \bibitem[{{Mart\'in-Hern\'andez} {et~al.}(2005){Mart\'in-Hern\'andez}, {Schaerer}, \& {Sauvage}}]{martinhernandez05}

1191: {Mart\'in-Hern\'andez}, N.~L., {Schaerer}, D., \& {Sauvage}, M.  2005, \aap, 429, 449

1192: 

1193: \bibitem[{{Massey} \& {Hunter}(1998{\natexlab{a}})}]{massey98a}

1194: {Massey}, P. \& {Hunter}, D.~A. 1998{\natexlab{a}}, \apj, 493, 180

1195: 

1196: \bibitem[{{Massey} \& {Hunter}(1998{\natexlab{b}})}]{massey98}

1197: ---. 1998{\natexlab{b}}, \apj, 493, 180

1198: 

1199: \bibitem[{{McCrady} {et~al.}(2003){McCrady}, {Gilbert}, \&

1200:   {Graham}}]{mccrady03}

1201: {McCrady}, N., {Gilbert}, A.~M., \& {Graham}, J.~R. 2003, \apj, 596, 240

1202: 

1203: \bibitem[{{McCrady} {et~al.}(2005){McCrady}, {Graham}, \& {Vacca}}]{mccrady05}

1204: {McCrady}, N., {Graham}, J.~R., \& {Vacca}, W.~D. 2005, \apj, 621, 278

1205: 

1206: \bibitem[{{McKee} \& {Williams}(1997)}]{mckee97}

1207: {McKee}, C.~F. \& {Williams}, J.~P. 1997, \apj, 476, 144

1208: 

1209: \bibitem[{{McLean} {et~al.}(1998){McLean}, {Becklin}, {Bendiksen}, {Brims},

1210:   {Canfield}, {Figer}, {Graham}, {Hare}, {Lacayanga}, {Larkin}, {Larson},

1211:   {Levenson}, {Magnone}, {Teplitz}, \& {Wong}}]{mclean98}

1212: {McLean}, I.~S., {Becklin}, E.~E., {Bendiksen}, O., {Brims}, G., {Canfield},

1213:   J., {Figer}, D.~F., {Graham}, J.~R., {Hare}, J., {Lacayanga}, F., {Larkin},

1214:   J.~E., {Larson}, S.~B., {Levenson}, N., {Magnone}, N., {Teplitz}, H., \&

1215:   {Wong}, W. 1998, in Proc. SPIE: Infrared Astronomical Instrumentation, ed.

1216:   A.~M. Fowler, Vol. 3354, 566--578

1217: 

1218: \bibitem[{{McLean} {et~al.}(2000){McLean}, {Graham}, {Becklin}, {Figer},

1219:   {Larkin}, {Levenson}, \& {Teplitz}}]{mclean00}

1220: {McLean}, I.~S., {Graham}, J.~R., {Becklin}, E.~E., {Figer}, D.~F., {Larkin},

1221:   J.~E., {Levenson}, N.~A., \& {Teplitz}, H.~I. 2000, in Proc. SPIE: Optical

1222:   and IR Telescope Instrumentation and Detectors, ed. M.~{Iye} \& A.~F.

1223:   {Moorwood}, Vol. 4008, 1048--1055

1224: 

1225: \bibitem[{{Mengel} {et~al.}(2002){Mengel}, {Lehnert}, {Thatte}, \&

1226:   {Genzel}}]{mengel02}

1227: {Mengel}, S., {Lehnert}, M.~D., {Thatte}, N., \& {Genzel}, R. 2002, \aap, 383,

1228:   137

1229: 

1230: \bibitem[{{Mengel} {et~al.}(2005){Mengel}, {Lehnert}, {Thatte}, \&

1231:   {Genzel}}]{mengel05}

1232: ---. 2005, \aap, 443, 31

1233: 

1234: \bibitem[{{Mengel} {et~al.}(2001){Mengel}, {Lehnert}, {Thatte},

1235:   {Tacconi-Garman}, \& {Genzel}}]{mengel01a}

1236: {Mengel}, S., {Lehnert}, M.~D., {Thatte}, N., {Tacconi-Garman}, L.~E., \&

1237:   {Genzel}, R. 2001, \apj, 550, 280

1238: 

1239: \bibitem[{{Meurer}(1995)}]{meurer95a}

1240: {Meurer}, G.~R. 1995, \nat, 375, 742

1241: 

1242: \bibitem[{{Meurer} {et~al.}(1995){Meurer}, {Heckman}, {Leitherer}, {Kinney},

1243:   {Robert}, \& {Garnett}}]{meurer95b}

1244: {Meurer}, G.~R., {Heckman}, T.~M., {Leitherer}, C., {Kinney}, A., {Robert}, C.,

1245:   \& {Garnett}, D.~R. 1995, \aj, 110, 2665

1246: 

1247: \bibitem[{{Mirabel} {et~al.}(1998){Mirabel}, {Vigroux}, {Charmandaris},

1248:   {Sauvage}, {Gallais}, {Tran}, {Cesarsky}, {Madden}, \& {Duc}}]{mirabel98}

1249: {Mirabel}, I.~F., {Vigroux}, L., {Charmandaris}, V., {Sauvage}, M., {Gallais},

1250:   P., {Tran}, D., {Cesarsky}, C., {Madden}, S.~C., \& {Duc}, P.-A. 1998, \aap,

1251:   333, L1

1252: 

1253: \bibitem[{{Moorwood} {et~al.}(1998){Moorwood}, {Cuby}, {Biereichel}, {Brynnel},

1254:   {Delabre}, {Devillard}, {van Dijsseldonk}, {Finger}, {Gemperlein},

1255:   {Gilmozzi}, {Herlin}, {Huster}, {Knudstrup}, {Lidman}, {Lizon}, {Mehrgan},

1256:   {Meyer}, {Nicolini}, {Petr}, {Spyromilio}, \& {Stegmeier}}]{moorwood98}

1257: {Moorwood}, A., {Cuby}, J.-G., {Biereichel}, P., {Brynnel}, J., {Delabre}, B.,

1258:   {Devillard}, N., {van Dijsseldonk}, A., {Finger}, G., {Gemperlein}, H.,

1259:   {Gilmozzi}, R., {Herlin}, T., {Huster}, G., {Knudstrup}, J., {Lidman}, C.,

1260:   {Lizon}, J.-L., {Mehrgan}, H., {Meyer}, M., {Nicolini}, G., {Petr}, M.,

1261:   {Spyromilio}, J., \& {Stegmeier}, J. 1998, The Messenger, 94, 7

1262: 

1263: \bibitem[{{Neff} \& {Ulvestad}(2000)}]{neff00}

1264: {Neff}, S.~G. \& {Ulvestad}, J.~S. 2000, \aj, 120, 670

1265: 

1266: \bibitem[{{Nikola} {et~al.}(1998){Nikola}, {Genzel}, {Herrmann}, {Madden},

1267:   {Poglitsch}, {Geis}, {Townes}, \& {Stacey}}]{nikola98}

1268: {Nikola}, T., {Genzel}, R., {Herrmann}, F., {Madden}, S.~C., {Poglitsch}, A.,

1269:   {Geis}, N., {Townes}, C.~H., \& {Stacey}, G.~J. 1998, \apj, 504, 749

1270: 

1271: \bibitem[{{O'Connell} {et~al.}(1994){O'Connell}, {Gallagher}, \&

1272:   {Hunter}}]{oconnell94}

1273: {O'Connell}, R.~W., {Gallagher}, J.~S., \& {Hunter}, D.~A. 1994, \apj, 433, 65

1274: 

1275: \bibitem[{{O'Connell} {et~al.}(1995){O'Connell}, {Gallagher}, {Hunter}, \&

1276:   {Colley}}]{oconnell95}

1277: {O'Connell}, R.~W., {Gallagher}, J.~S., {Hunter}, D.~A., \& {Colley}, W.~N.

1278:   1995, \apjl, 446, L1

1279: 

1280: \bibitem[{{Osterbrock}(1974)}]{osterbrock74}

1281: {Osterbrock}, D.~E. 1974, {Astrophysics of gaseous nebulae} (San Francisco:

1282:   W.~H.~Freeman and Co.)

1283: 

1284: \bibitem[{{Persson} {et~al.}(1998){Persson}, {Murphy}, {Krzeminski}, {Roth}, \&

1285:   {Rieke}}]{persson98}

1286: {Persson}, S.~E., {Murphy}, D.~C., {Krzeminski}, W., {Roth}, M., \& {Rieke},

1287:   M.~J. 1998, \aj, 116, 2475

1288: 

1289: \bibitem[{{Rigby} \& {Rieke}(2004)}]{rigby04}

1290: {Rigby}, J.~R. \& {Rieke}, G.~H. 2004, \apj, 606, 237

1291: 

1292: \bibitem[{{Rieke} \& {Lebofsky}(1985)}]{rieke85}

1293: {Rieke}, G.~H. \& {Lebofsky}, M.~J. 1985, \apj, 288, 618

1294: 

1295: \bibitem[{{Rousselot} {et~al.}(2000){Rousselot}, {Lidman}, {Cuby}, {Moreels},

1296:   \& {Monnet}}]{rousselot00}

1297: {Rousselot}, P., {Lidman}, C., {Cuby}, J.-G., {Moreels}, G., \& {Monnet}, G.

1298:   2000, \aap, 354, 1134

1299: 

1300: \bibitem[{{Rozas} {et~al.}(1998){Rozas}, {Sabalisck}, {Beckman}, \&

1301:   {Knapen}}]{rozas98}

1302: {Rozas}, M., {Sabalisck}, N., {Beckman}, J.~E., \& {Knapen}, J.~H. 1998, \aap,

1303:   338, 15

1304: 

1305: \bibitem[{{Rubin} {et~al.}(1970){Rubin}, {Ford}, \& {D'Odorico}}]{rubin70}

1306: {Rubin}, V.~C., {Ford}, W.~K., \& {D'Odorico}, S. 1970, \apj, 160, 801

1307: 

1308: \bibitem[{{Salpeter}(1955)}]{salpeter55}

1309: {Salpeter}, E.~E. 1955, \apj, 121, 161

1310: 

1311: \bibitem[{{Saviane} {et~al.}(2004){Saviane}, {Hibbard}, \& {Rich}}]{saviane04}

1312: {Saviane}, I., {Hibbard}, J.~E., \& {Rich}, R.~M. 2004, \aj, 127, 660

1313: 

1314: \bibitem[{{Schweizer}(1987)}]{schweizer87}

1315: {Schweizer}, F. 1987, in Nearly Normal Galaxies. From the Planck Time to the

1316:   Present, 18--25

1317: 

1318: \bibitem[{{Schweizer}(2004)}]{schweizer04}

1319: {Schweizer}, F. 2004, in ASP Conf. Ser. 322: The Formation and Evolution of

1320:   Massive Young Star Clusters, 111

1321: 

1322: \bibitem[{{Schweizer} \& {Seitzer}(1998)}]{schweizer98}

1323: {Schweizer}, F.~. \& {Seitzer}, P. 1998, \aj, 116, 2206

1324: 

1325: \bibitem[{{Schweizer} {et~al.}(1996){Schweizer}, {Miller}, {Whitmore}, \&

1326:   {Fall}}]{schweizer96}

1327: {Schweizer}, F., {Miller}, B.~W., {Whitmore}, B.~C., \& {Fall}, S.~M. 1996,

1328:   \aj, 112, 1839

1329: 

1330: \bibitem[{{Shopbell} \& {Bland-Hawthorn}(1998)}]{shopbell98}

1331: {Shopbell}, P.~L. \& {Bland-Hawthorn}, J. 1998, \apj, 493, 129

1332: 

1333: \bibitem[{{Smith} \& {Gallagher}(2001)}]{smith01}

1334: {Smith}, L.~J. \& {Gallagher}, J.~S. 2001, \mnras, 326, 1027

1335: 

1336: \bibitem[{{Smith} \& {Weedman}(1970)}]{smith70}

1337: {Smith}, M.~G. \& {Weedman}, D.~W. 1970, \apj, 161, 33

1338: 

1339: \bibitem[{{Spitzer}(1987)}]{spitzer87}

1340: {Spitzer}, L. 1987, Dynamical Evolution of Globular Clusters (Princeton:

1341:   Princeton Univ. Press, 1987)

1342: 

1343: \bibitem[{{Stanford} {et~al.}(1990){Stanford}, {Sargent}, {Sanders}, \&

1344:   {Scoville}}]{stanford90}

1345: {Stanford}, S.~A., {Sargent}, A.~I., {Sanders}, D.~B., \& {Scoville}, N.~Z.

1346:   1990, \apj, 349, 492

1347: 

1348: \bibitem[{{Sternberg}(1998)}]{sternberg98}

1349: {Sternberg}, A. 1998, \apj, 506, 721

1350: 

1351: \bibitem[{{Stetson}(1987)}]{stetson87}

1352: {Stetson}, P.~B. 1987, \pasp, 99, 191

1353: 

1354: %\bibitem[{{Strickland} \& {Stevens}(1999)}]{strickland99} {Strickland}, D.~K. \& {Stevens}, I.~R. 1999, \mnras, 306, 43

1355: 

1356: \bibitem[{{Testi} {et~al.}(2000){Testi}, {Sargent}, {Olmi}, \&

1357:   {Onello}}]{testi00}

1358: {Testi}, L., {Sargent}, A.~I., {Olmi}, L., \& {Onello}, J.~S. 2000, \apjl, 540,

1359:   L53

1360: 

1361: \bibitem[{{Turner} {et~al.}(2003){Turner}, {Beck}, {Crosthwaite}, {Larkin},

1362:   {McLean}, \& {Meier}}]{turner03}

1363: {Turner}, J.~L., {Beck}, S.~C., {Crosthwaite}, L.~P., {Larkin}, J.~E.,

1364:   {McLean}, I.~S., \& {Meier}, D.~S. 2003, \nat, 423, 621

1365: 

1366: \bibitem[{{Turner} {et~al.}(2000){Turner}, {Beck}, \& {Ho}}]{turner00}

1367: {Turner}, J.~L., {Beck}, S.~C., \& {Ho}, P.~T.~P. 2000, \apjl, 532, L109

1368: 

1369: \bibitem[{{Vacca} {et~al.}(2002){Vacca}, {Johnson}, \& {Conti}}]{vacca02}

1370: {Vacca}, W.~D., {Johnson}, K.~E., \& {Conti}, P.~S. 2002, \aj, 123, 772

1371: 

1372: \bibitem[{{Walborn}(1991)}]{walborn91}

1373: {Walborn}, N.~R. 1991, in IAU Symp. 148: The Magellanic Clouds, ed. R.~{Haynes}

1374:   \& D.~{Milne}, Vol. 148 (Dordrecht: Kluwer Academic Publishers), 145

1375: 

1376: \bibitem[{{Whitmore} {et~al.}(2005)}]{whitmore05}

1377: {Whitmore}, B.~C., {Gilmore}, D., {Leitherer}, C., {Fall}, S.~M., {Chandar}, R., 

1378: {Blair}, W.~P., {Schweizer}, F., {Zhang}, Q., \& {Miller}, B.~W. 2005, \aj, 130, 2104

1379: 

1380: \bibitem[{{Whitmore} \& {Schweizer}(1995)}]{whitmore95}

1381: {Whitmore}, B.~C. \& {Schweizer}, F. 1995, \aj, 109, 960

1382: 

1383: \bibitem[{{Whitmore} \& {Zhang}(2002)}]{whitmore02}

1384: {Whitmore}, B.~C. \& {Zhang}, Q. 2002, \aj, 124, 1418

1385: 

1386: \bibitem[{{Whitmore} {et~al.}(1999){Whitmore}, {Zhang}, {Leitherer}, {Fall},

1387:   {Schweizer}, \& {Miller}}]{whitmore99}

1388: {Whitmore}, B.~C., {Zhang}, Q., {Leitherer}, C., {Fall}, S.~M., {Schweizer},

1389:   F., \& {Miller}, B.~W. 1999, \aj, 118, 1551

1390: 

1391: \bibitem[{{Williams} \& {McKee}(1997)}]{williams97}

1392: {Williams}, J.~P. \& {McKee}, C.~F. 1997, \apj, 476, 166

1393: 

1394: \bibitem[{{Wilson} {et~al.}(2000){Wilson}, {Scoville}, {Madden}, \&

1395:   {Charmandaris}}]{wilson00}

1396: {Wilson}, C.~D., {Scoville}, N., {Madden}, S.~C., \& {Charmandaris}, V. 2000,

1397:   \apj, 542, 120

1398: 

1399: \bibitem[{{Wood} \& {Churchwell}(1989)}]{wood89}

1400: {Wood}, D.~O.~S. \& {Churchwell}, E. 1989, \apjs, 69, 831

1401: 

1402: \bibitem[{{Zhang} {et~al.}(2001){Zhang}, {Fall}, \& {Whitmore}}]{zhang01}

1403: {Zhang}, Q., {Fall}, S.~M., \& {Whitmore}, B.~C. 2001, \apj, 561, 727

1404: 

1405: \end{thebibliography}

1406: 

1407: % Fig 1 revised new names, inset size correction, mag x5 not x8

1408: \begin{figure}[ph]

1409: \begin{center}

1410: \plotone{f1.eps}

1411: \caption[ISAAC 2.17 \um\ image of the Antennae] {Image of the Antennae using

1412:   ISAAC 2.17 \um\ narrow-band filter, which includes \brg.  Units are

1413:   arcseconds of R.A. and Dec.; north is up and east is to the left.  Diamonds mark positions of NIRSPEC targets (except for the circled sources 10 and 16 where they are omitted to avoid crowding).  Circles mark positions of the brightest young clusters in HST F555W (Whitmore et al.

1414:   1999).  SSC B is shown in an inset, magnified by a factor of five, in order to reveal IR structure and separate HST sources.  At the origin is Star 4 of \citet{whitmore99}. 

1415: \label{fig:ant_hst}}

1416: \end{center}

1417: \end{figure}

1418: 

1419: % add a new Fig 2 in revised version, name change

1420: \begin{figure}[h] 

1421: \begin{center}

1422: %\epsscale{0.85} 

1423: \plotone{f2.eps}

1424: \caption[Comparison of ELC images from ISAAC and HST] {Comparison of ELCs observed from the ground and by HST in several wavebands.  Bottom row shows the ISAAC 2.17 \um\ image (\protect{0\farcs4} seeing); subsequent rows show NIC3 F222M image (\protect{0\farcs2/}pixel), NIC2 F187N and F160W images (\protect{0\farcs075}/pixel), and ACS F814W image (\protect{0\farcs05}/pixel).  Horizontal bar is \protect{1\arcsec}\ long. 

1425: \label{fig:thumbs} }

1426: \end{center}

1427: \end{figure}

1428: 

1429: 

1430: % Fig 2-> 3 revised names, order

1431: \begin{figure}[ph] 

1432: \begin{center}

1433: \epsscale{0.85} 

1434: % was .65

1435: \vbox{

1436: \plottwo{f3a.eps}{f3b.eps}

1437: % was 1.0

1438: \epsscale{0.85}

1439: \plottwo{f3c.eps}{f3d.eps}

1440: }

1441: \caption[Br$\gamma$ echelle spectra of Antennae targets] {Br$\gamma$ spectra

1442:   of young Antennae SSCs and nuclear regions, featuring broad,

1443:   non-Gaussian wings.  The spectral resolution is about 12 \kms\ FWHM (4

1444:   pixels, shown as a bar in upper right corner of all panels), so the lines are well resolved and their widths are not

1445:   simply due to thermal broadening.  He~{\sc i} emission at 2.1647

1446:   $\mu$m ($\Delta v = -191$ \kms from \brg) is also visible in the

1447:   brightest sources (top left panel). Note the different velocity

1448:   scale in top left panel, and varied flux scales throughout.}

1449: \label{fig:brg}

1450: \end{center}

1451: \end{figure}

1452: 

1453: 

1454: % Fig 3->4 revised names

1455: \begin{figure}[th] 

1456: \begin{center}

1457: \epsscale{0.75}

1458: \plotone{f4.eps}

1459: \caption[Br$\gamma$ position-velocity diagrams for Antennae ELCs] {Br$\gamma$

1460:   position-velocity diagrams for ELCs B1, B, C, and D, overlaid with

1461:   arbitrarily scaled integrated spatial profiles and extracted spectra.  

1462:   The \brg\ emission is 

1463:   extended with respect to the continuum, and its spatially resolved

1464:   velocity gradients are suggestive of nonspherical flows.

1465: \label{fig:elcpv}}

1466: \end{center}

1467: \end{figure}

1468: 

1469: % Fig. 5 added in revised version

1470: \begin{figure}[th] 

1471: \begin{center}

1472: \epsscale{0.75}

1473: \plotone{f5.eps}

1474: \caption[Antennae ELC light profiles]{Normalized, azimuthally averaged radial profiles of ISAAC 2.17 and 2.25 \um\ image counts for two ELCs and Star 4 demonstrate that the ELCs are resolved relative to the observed PSF.  Error bars indicate $1\sigma$ Poisson errors.

1475: \label{fig:radprofs}}

1476: \end{center}

1477: \end{figure}

1478: 

1479: % was Fig 4, becomes 6 

1480: % lorentz_sizes.eps -> f6.eps

1481: % figure/numbers updated with plate scale corrections.  new fits, errors, names. C1 -> d1.

1482: % 7 -> 9 pc.  25-40 pc -> 30-50 pc.

1483: \begin{figure}[th] 

1484: \begin{center}

1485: %\epsscale{0.75}

1486: \plotone{f6.eps}

1487: \caption[Sizes of Antennae ELC as function of age] {Above: Deconvolved sizes of ELCs measured from 2.25 \um\ ISAAC narrow-band image, plotted as a function of ELC age.   At ages of $6-7$ Myr, the population of ELCs has a large size spread (down to half-light radii of 9 pc for ELC F), but younger ELCs are larger, with radii of $30-50$ pc.  Below:  The ratio of 2.17 \um\ (stellar continuum plus \brg\ emission) to 2.25 \um\ (stellar continuum) half-light radii for bELCs increases slightly with age.  Most bELCs (which exclude ELCs F and D1) have larger sizes in the \brg\ filter than in the continuum one.  Sizes are geometric mean FWHMs determined from 2D Lorentzian fits with Star 4 geometric mean radius subtracted in quadrature, and error bars are computed from fit errors. \label{fig:sizes}}

1488: \end{center}

1489: \end{figure}

1490: 

1491: 

1492: % Fig 5a,b--becomes 7a,b, revised sizes--add note that width has instr/thermal components removed, also specify that bELCs not all ELCs are plotted.

1493: \begin{figure}[th]

1494: \begin{center}

1495: \epsscale{1.15}

1496: \plottwo{f7a.eps}{f7b.eps}

1497: \caption[Relation between bELC Lyman continuum rate and linewidth, and electron density and diameter] {Above: bELC Lyman

1498:   continuum rate versus line width, with instrumental response and thermal component removed, compared with other samples of HIIRs:

1499:   nearby massive cluster 30 Dor \protect{\citep[luminosity and line width

1500:   from][respectively]{walborn91,chu94}, UDHIIs/bELCs in NGC 5253

1501:   \citep{turner03} and He $2-10$ \protect{\citep{kobulnicky99,johnson03,vacca02,henry07}}, Galactic (U)CHIIs

1502:   \citep[][]{garay99}, and first-ranked GHIIRs in several galaxies

1503:   \citep[][]{arsenault88}.   Below: Mean electron density versus diameter for bELCs compared with the other HIIR populations.  Solid lines indicate the Str\"omgren sphere relation for an HIIR with constant density: $n_{e} \propto {d}^{-1.5}$, for $Q[H^+] = 10^{49}, 10^{51}$, and $10^{53}$ s$^{-1}$}.  Sizes are measured from radio data for all but the bELCs and UDHIIs, whose sizes are from near- and mid-IR images, respectively; adopting the pc-scale radio sizes for UDHIIs or the typical optical SSC sizes for bELCs moves them closer to the (U)CHIIs.  All densities plotted are average values; the GHIIRs would move up by one or two orders of magnitude if we adopted the higher densities that are derived from optical forbidden lines and are weighted toward dense, presumably low-filling-factor gas.}

1504: \label{fig:lumsig}

1505: \end{center}

1506: \end{figure}

1507:   

1508: \clearpage

1509: 

1510: % TAb 1--modified

1511: \input{tab1}

1512: 

1513: \clearpage

1514: 

1515: % Table 2

1516: \input{tab2}

1517: 

1518: \clearpage

1519: 

1520: % Tab 3--modified

1521: \input{tab3}

1522: 

1523: \clearpage

1524: 

1525: % Tab 4--modified

1526: \input{tab4}

1527: 

1528: \end{document}

1529: