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

2: \documentclass{emulateapj}

3: %\usepackage{amssymb}

4: %\usepackage{psfig}

5: \usepackage{graphicx}

6: %\usepackage{color,calc}

7: 

8: %\makeatletter

9: 

10: \providecommand{\tabularnewline}{\\}

11: 

12: \shorttitle{Clearing the Dust from Globular Clusters}

13: \shortauthors{Umbreit, Chatterjee \& Rasio}

14: 

15: %\usepackage{babel}

16: %\makeatother

17: \begin{document}

18: 

19: \title{Clearing the Dust from Globular Clusters}

20: 

21: 

22: \author{Stefan Umbreit,\altaffilmark{1} Sourav Chatterjee, 

23: \altaffilmark{1} and

24: Frederic A.\ Rasio\altaffilmark{1}}

25: 

26: \affil{$^{1}$ Department of Physics and Astronomy, Northwestern  

27: University,

28: 2131 Tech Drive, Evanston, IL, 60208}

29: \email{s-umbreit@northwestern.edu}

30: 

31: 

32: 

33: 

34: \begin{abstract}

35: Recent \emph{Spitzer} observations of the globular cluster M15

36: detected dust associated with its intracluster medium.

37: Surprisingly, these observations imply that the dust must be very  

38: short-lived

39: compared to the time since the last Galactic plane crossing of the

40: cluster. Here we propose

41: a simple mechanism to explain this short lifetime. We argue that the  

42: kinetic

43: energy of the material

44: ejected during a stellar collision may be sufficient to remove the

45: gas and dust entirely from a cluster,

46: or to remove the gas as a wind, in addition to partially destroying the dust.

47: By calculating the

48: rate of stellar collisions using an $N$-body model for the cluster,

49: we find remarkable agreement between the average time between collisions

50: and the inferred dust lifetime in this cluster, suggesting a possible  

51: close

52: relation between the two phenomena. Furthermore, we also obtain

53: the birthrate of blue stragglers formed through collisions in M15. By  

54: comparing

55: with the observed number of blue stragglers, we derive an upper limit

56: for their average lifetime which turns out to be consistent with  

57: recent model

58: calculations, thereby lending further support to our model.

59: \end{abstract}

60: 

61: \keywords{stellar dynamics -- shock waves -- blue stragglers -- stars: mass loss-- globular clusters: general -- globular clusters: individual (M15)}

62: 

63: 

64: \section{Introduction}

65: 

66: \nocite{*}In recent years, infrared space telescopes like \emph{IRAS,

67: ISO} and especially \emph{Spitzer} made it possible to investigate

68: not only the stellar content of globular clusters (GCs) but also their

69: intracluster medium (ICM), consisting of gas and dust lost by giants.  

70: Evans

71: et al.\ (2003) were the first to clearly detect thermal dust emission

72: from the core of the GC M15 (NGC~7078) with {\em ISO}. They found a dust

73: mass an order of magnitude lower than one would expect based

74: on the number, mass-loss rates and lifetimes of horizontal branch

75: (HB) stars, even at the extremely low metallicity of M15. Subsequent,

76: more sensitive measurements with {\em Spitzer}

77: resulted in a dust mass of $9\times10^{-4} 

78: \,{\rm M}_{\odot}$ indicating an accumulation time-span not much longer than 

79: $10^{6}\,$yr (Boyer et al.\ 2006). While such a short

80: time is commonly explained by ram-pressure stripping during Galactic

81: plane passages, the time since the last passage for M15 is about

82: $4\times10^{7}\,$yr, an order of magnitude larger. Therefore,

83: additional processes have been suggested to explain short ICM

84: lifetimes, including blow-out of nova explosions, fast winds from

85: stars, relativistic winds from millisecond pulsars

86: (Spergel 1991), and radiative ejection by the strong radiation

87: field in M15 (Smith 1999).

88: 

89: Another possibility that has not

90: yet been investigated is related to the outcome of stellar collisions.

91:  From hydrodynamic simulations it is known that colliding main-sequence

92: stars (MSS) lose typically $\sim1\%$ of their mass 

93: (see, e.g., Lombardi et al.\ 1996).

94: As this material is ejected with speeds typically a few

95: times the escape speed from the stellar surface, it will transfer a

96: significant amount of energy to the ICM and possibly reduce the  

97: amount of observable dust considerably. For instance, if the gas, released

98: through a collision between two $0.5\,{\rm M}_{\odot}$ MSSs,

99: has a mass of $0.02\,{\rm M}_{\odot}$ and leaves with twice the

100: escape velocity ($\approx1200\,{\rm km\, s}^{-1}$), it possesses

101: enough kinetic energy to accelerate more than $15\,{\rm M}_{\odot}$,

102: the estimated total mass of ICM gas in M15 (Boyer et al.\ 2006), up to a

103: speed of $40\,{\rm km\, s}^{-1}$, the escape speed from the center of  

104: M15.

105: 

106: The inferred dust lifetime in the ICM would then require that

107: such a stellar collision happens every $\sim 1\,$Myr. Given that M15

108: has a central density of $n\gtrsim10^{6}\,{\rm M}_{\odot}\,{\rm pc}^{-3}$

109: and a velocity dispersion of $\sigma\approx10\,{\rm km}\,{\rm s}^{-1}$ 

110: (see, e.g., McNamara et al.\ 2004), the collision time $T_{\rm coll}=

111: 7\times10^{11}\,{\rm yr}\,\left(\frac{10^{5}{\rm pc}^{-3}}{\nu}\right) 

112: \,\left(\frac{\sigma}{100\,{\rm km}\,{\rm s}^{-1}}\right)$

113: for a $1\,{\rm M}_{\odot}$ MSS is $\sim 10^{10}\,{\rm yr}$ (Binney \&  

114: Tremaine 1987).

115: This results in at least one collision every $<5\,{\rm Myr}$ for the  

116: more

117: than $2000$ stars in the core (e.g., Dull et al.\ 1997). From this

118: rough estimate one can already see that this mechanism

119: is indeed promising.

120: 

121: However, there are several factors in M15 that complicate

122: a better estimate of the collision time. Firstly, in GCs as old

123: as M15, all MSSs have masses below $0.8\,{\rm M}_{\odot}$, while initially

124: more massive stars have produced remnants like white dwarfs

125: (WDs) and neutron stars (NSs), which will have different interaction  

126: rates.

127: Secondly, because of mass segregation the lower-mass MSSs

128: may concentrate in lower density regions, while more massive

129: WDs and NSs concentrate near the center. Finally,

130: it is not {\em a priori} clear that enough energy from the

131: gas ejected in a stellar collision can be effectively transferred to  

132: the ICM

133: since losses through radiative shocks may be important.

134: 

135: We address the issue of mass-segregation considering a specific

136: model of M15 given in McNamara et al.\ (2004), which is based on an

137: $N$-body model of Baumgardt et al.\ (2003). This

138: allows us to estimate collision times involving different stellar 

139: populations in the cluster. Furthermore, we obtain a minimum formation 

140: rate for blue stragglers (BSs) through MS-MS mergers. Using 

141: the number of observed BSs in the core of M15 we estimate an average 

142: BS lifetime and compare it with evolutionary models, as an extra 

143: check on our basic model (\S\ref {sec:Main-Sequence-Collision-Time}).

144: In \S\ref{gas_ICM} we discuss how the ejected

145: material from a stellar collision interacts with the ICM and how it can

146: clear signatures of dust emission. We conclude in

147: \S\ref{conclusion}.

148: 

149: 

150: \section{Collision Times}

151: \label{sec:Main-Sequence-Collision-Time}

152: 

153: We define a collision between two stars with radii $R_{1}$ and $R_{2}$

154: to occur whenever their distance $d\leq R_{1}+R_{2}$. The

155: average local collision time $T_{\rm coll}^{(1,2)}(r)$ for one star of

156: type {}``1'' to collide with a star of type {}``2'' can be written

157: as (compare to Binney \& Tremaine 1987 their eq. 8-122)

158: \begin{eqnarray}

159: \frac{1}{T_{\rm coll}^{(1,2)}(r)} & = & 4\sqrt{\pi}n_{2}(r)\sigma(r)\label{eq:local-coll-time}\\

160:   &  & \times\left(\left(R_{1}+R_{2}\right)^{2}+

161:   \frac{G\left(M_{1}+M_{2}\right)\left(R_{1}+R_{2} 

162: \right)}{2\sigma(r)^{2}}\right)\,.\nonumber 

163: \end{eqnarray}

164: where $n_{1,2}$ is the number density of field stars {}``2'' and stars

165: of type {}``1'', $M_{1,2}$

166: their masses, and $\sigma$ the one-dimensional

167: velocity dispersion (assuming a Maxwellian velocity distribution for

168: both species with 

169: $\sigma(r)=\sqrt{(\sigma_1(r)^2+\sigma_2(r)^2)/2}$).

170: To get the total number of collisions per unit time in the

171: cluster between these two species, eq.~\ref{eq:local-coll-time}

172: is integrated over the whole cluster,

173: \begin{equation}

174: \frac{dN_{coll,tot}}{dt}=4\pi\int_{0}^{\infty}dr\, r^{2}\, n_{1}(r)\, 

175: \frac{1}{T_{coll}(r)}\label{eq:totat-coll-rate}

176: \end{equation}

177: where $r$ is the radial position in the cluster.

178: 

179: In order to account for

180: a continuous stellar mass spectrum and the mass-radius relationship of

181: stars, we take local averages $R_{1,2}(r)$ and $M_{1,2}(r)$

182: at position $r$. 

183: We find that in the N-body model for M15 these profiles, as well as 

184: $n_{1,2}(r)$, can be well represented

185: by power-laws over a sufficiently large range of 

186: $0.025\,{\rm pc}<r<2\,{\rm pc}$. 

187: Inside of $0.025\, {\rm pc}$

188: there are almost no MSSs, while outside of $\approx1\,{\rm pc}$ the

189: contribution of the integrand in eq.~\ref{eq:totat-coll-rate} becomes

190: rapidly negligible. We do not consider collisions between MSSs

191: and NSs as the NS retention fraction in GCs is expected to be low  

192: (Pfahl \& Rappaport 2001; Dull et al. 1997). We also do not consider 

193: collisions between giants and MSSs

194: as the escape velocity at the surface of a giant, and even more so the

195: expected energy of the ejected material, is an order of magnitude

196: lower than for MSSs. For simplicity we choose a constant $\sigma=11\, 

197: {\rm km\, s}^{-1}$, as $\sigma$ does not vary much within 

198: $1\, {\rm pc}$ ($\pm 1\, {\rm km\, s}^{-1}$; compare Dull et al. 1997) 

199: and also agrees with the value obtained by McNamara \& Baumgardt (2004) for M15 

200: for a similar region $(r<0.3'\approx 0.8\rm{ pc})$.

201: The slight variations within $1\, {\rm pc}$ are 

202: accounted for in the error estimates of the collision rates.  

203: With all quantities given as power-laws, we solve 

204: eq.~\ref{eq:totat-coll-rate}

205: analytically. Our calculations are based on the results of

206: $N$-body simulations by Baumgardt et al.\ (2003), which were

207: scaled to fit the velocity dispersion profile of M15 in McNamara et

208: al.\ (2004). Their model consisted of initially $130,072$ stars with

209: a realistic mass spectrum, and included a treatment of stellar  

210: evolution and

211: the Galactic tidal field. They also take into account velocity kicks  

212: imparted to

213: NSs at birth and consider two extreme cases, one where all

214: NSs are retained, and one where all NSs are removed from the cluster.

215: Here we only consider the latter case, since, as mentioned before, 

216: the actual NS retention fraction is expected to be very low.

217: 

218: 

219: 

220: In Table~\ref{tab:Parameters-fits} 

221: \begin{table}

222: \begin{center}

223: \caption{\label{tab:Parameters-fits}Parameters of the power-law fits for

224: the mass ($M$), density ($n$), and stellar radius ($R$) for

225: WDs and MSSs. The power-laws are of

226: the form $a=a_{0}\,(r/r_{0})^{b}$ with $r_{0}=1\,{\rm pc}$.}

227: \begin{tabular}{|c|c|c|}

228: \hline

229: &

230: $a_{0}$&

231: $b$\tabularnewline

232: \hline

233: \hline

234: $n_{MS}$&

235: $(3.9\pm0.2)\times10^{3}{\rm pc}^{-3}$&

236: $-1.620\pm0.006$\tabularnewline

237: \hline

238: $n_{WD}$&

239: $(3.3\pm0.1)\times10^{3}{\rm pc}^{-3}$&

240: $-2.280\pm0.009$\tabularnewline

241: \hline

242: $R_{MS}$&

243: $(0.51\pm0.05)\,{\rm R}_{\odot}$&

244: $-0.143\pm0.008$\tabularnewline

245: \hline

246: $R^2_{MS}$&

247: $(0.32\pm0.05)\,{\rm R}_{\odot}$&

248: $-0.26\pm0.01$\tabularnewline

249: \hline

250: $M_{MS}$&

251: $(0.49\pm0.02)\,{\rm M}_{\odot}$&

252: $-0.105\pm0.002$\tabularnewline

253: \hline

254: $M_{WD}$&

255: $(0.74\pm0.02)\,{\rm M}_{\odot}$&

256: $-0.117\pm0.002$\tabularnewline

257: \hline

258: \end{tabular}

259: \end{center}

260: \end{table}

261: the fit parameters for $n$, $R$, $R^2$

262: and $M$ for WDs and MSSs in M15 are shown. As can

263: be seen, through mass segregation, MSSs

264: are more abundant outside $1\,{\rm pc}$ while the more massive WDs

265: dominate the central $0.75\,{\rm pc}$ where collisions are most likely

266: to happen. As a consequence, we should expect more collisions between

267: WDs and MSSs than between MSSs. From eq.~\ref{eq:totat-coll-rate}, we obtain

268: $dN_{coll,MS-WD}/dt=(2\pm1)\times10^{-7}\,{\rm yr}^{-1}$

269: and $dN_{coll,MS-MS}/dt=(4\pm1)\times10^{-8}\,{\rm yr}^{-1}$, where

270: $N_{coll,X-Y}$ is the number of collisions between $X$ and $Y$. It

271: follows that, given the current state of M15, we expect one

272: collision every $(5\pm2)\times10^{6}\,{\rm yr}$ between a WD and

273: a MSS and every $(3\pm1)\times10^{7}\,{\rm yr}$ a collision between

274: two MSSs, resulting in one collision every $(4\pm2)\times10^{6} 

275: \,{\rm yr}$ that releases high-velocity gas into the ICM.

276: 

277: It is rather remarkable how closely this timescale

278: coincides with the estimated ICM dust lifetime,

279: suggesting a direct connection between dust clearing and

280: stellar collisions.

281: 

282: %Even considering only collisions between MSSs,

283: %the resulting collision timescale is much smaller, at least by

284: %a factor of $5$, than the time since the last Galactic-plane passage.

285: 

286: Using a

287: similar approach, we estimate the BS formation rate, defining BSs

288: as two merged MSSs with a combined mass $>1\,{\rm M}_{\odot}$,  

289: significantly

290: larger than the MS turn-off mass ($0.8\,{\rm M}_{\odot}$).  

291: After binning

292: all MSSs into $0.1\,{\rm M}_{\odot}$ bins, we determine the collision

293: rate between all mass bins with combined mass $>1\,{\rm M}_{\odot}$. This also

294: allows us to determine the expected mass spectrum for BSs for comparison

295: with future observations.

296: We obtain a total collision rate of $(6.4\pm0.7)\times10^{-3}\,{\rm  

297: Myr}^{-1}$

298: which, together with the $6-7$ BSs observed in M15, implies

299: an average BS lifetime of about $1\,{\rm Gyr}$. This is also  

300: consistent with recent BS

301: evolution models (e.g., Leigh et al.\ 2007; Sills et al.\ 2001). 

302: However, we note that this is certainly an upper limit, given that

303: the presence of binaries would increase the formation rate through 

304: binary mergers and resonant interactions 

305: (Fregeau et al.\ 2004; Mapelli et al.\ 2004).  

306: For example, for a Plummer sphere with a Kroupa mass function ranging

307: from $0.3$ to $3.0\,\rm{M_\odot}$ and a binary fraction of $30\%$, the

308: rate of collisions mediated by binary-single and binary-binary interactions 

309: can be $\sim 1$ order of magnitude larger

310: than from single-single interactions (Chatterjee et al.\ 2008). In Figure~\ref{fig:bs_mass} 

311: \begin{figure}

312: \begin{center}

313: \plotone{f1.eps}

314: \caption{Estimated mass distribution of BS candidates created via

315: collisions between two MSSs with combined mass $>0.8 \rm{M}_{\odot}$.

316: $n_{i}$, $b$, and $N$ are the number of BSs in the $i$th bin, the bin

317: size, and the total number of BSs, respectively.}

318: \label{fig:bs_mass}   

319: \end{center}

320: \end{figure}

321: the mass distribution for BSs is shown. As expected 

322: it decreases for higher mass BSs, since the number 

323: density for the lower mass MSSs is much higher compared to that for the close 

324: to turn-off MSSs.

325: 

326: \section{Interaction of Collision Ejecta with the ICM}

327: \label{gas_ICM}

328: 

329: Interactions between very fast moving gas and a low-density interstellar

330: ambient medium have been extensively investigated in the context of

331: supernova remnants. Their evolution progresses in several distinct  

332: stages

333: (Chevalier 1977): the {}``ejecta dominated'' (ED) stage, 

334: characterized by a freely expanding blast-shock wave until the mass

335: of the swept up material is comparable to the mass of the ejecta,

336: the {}``Sedov-Taylor'' (ST) stage, where the blast-wave expands

337: adiabatically, and the {}``pressure-driven snowplow'' (PDS) stage,

338: where a thin shell forms which {}``snowplows'' through the ambient

339: medium, driven by the pressure of the hot interior in addition to

340: its own momentum (Cox 1972; Cioffi et al.\ 1988). 

341: Cioffi et al.\  (1988) obtained a simple offset power-law solution that

342: describes the kinematics of the blast wave in the PDS stage,  

343: which can be written as $v_{s}=v_{PDS}\left(R_{s}/R_{PDS}\right)^{-7/3}$, where

344: $v_{s}$ and $R_{s}$ are the velocity and radial position of the

345: shock front, respectively, relative to the site of the collision, and

346: $v_{PDS}$ and $R_{PDS}$ are the velocity and radial position of

347: the shock front at the transition from the ST stage to the PDS stage.  

348: The transition values are given by 

349: $v_{PDS}=413n_{0}^{1/7}\zeta^{3/14}_{m}E_{51}^{1/14}\,{\rm km\, s}^{-1}$

350: and $R_{PDS}=14.0\, E_{51}^{2/7}\zeta^{-1/7}_{m}n_{0}^{-3/7}

351: \,{\rm pc}$, where $E_{51}$

352: is the kinetic energy of the ejected material in units of $10^{51}\, 

353: {\rm ergs}$, $n_{0}$ the density of the ambient medium in 

354: ${\rm cm}^{-3}$, and $\zeta_m$ is the heavy element abundance relative to

355: solar abundances.

356: An interesting property of the flow of the post-shock gas is that in the 

357: PDS stage its mean velocity has the same value as the shock speed 

358: (Cioffi et al.\ 1988).

359: For our problem this means that, provided $v_{s}$ is larger than the 

360: cluster escape speed after the 

361: shock swept up all of the ICM, the shocked gas can entirely escape the 

362: cluster.

363: 

364: It is now interesting to see if such an ICM removal mechanism might be

365: applicable to M15. For this estimate we assume an ICM 

366: mass of $15\,{\rm M}_{\odot}$, which is expected given the amount of dust

367: detected by Boyer et al.\  (2006) and the low metallicity of M15

368: ($\zeta_m=10^{-2.4}$).

369: In order to estimate $E_{51}$ for collisions between MSS 

370: we can use the results of Lombardi et al.\ (2002) (their Table 3),

371: obtaining $E_{51}=1-11\times10^{-4}$. Unfortunately, there are no similar

372: results for MS--WD collisions. However, given that the mass loss

373: is presumably between $15\%$ to $50\%$ (Ruffert \& Mueller

374: 1991; Ruffert 1992) and that the velocity of the ejecta should be of the

375: order of the escape speed of the MSS ($\approx500\, {\rm km\, s}^{-1}$)

376: we obtain a similar range for $E_{51}$.  Note that,

377: although the ejecta are ejected nearly isotropically 

378: (Lombardi et al.\ 1996), the ICM is likely to have a more irregular 

379: structure, as the patchy dust emission in M15 indicates. Thus, only a 

380: fraction of the released energy might actually be transferred to the ICM. 

381: As the 3D structure of the ICM is unknown, we limit our analysis to a 

382: fiducial ejecta energy of $E_{51}= 3\times10^{-4}$ but also determine the 

383: minimum $E_{51}$ value to remove all of the ICM gas from the cluster, noting 

384: that these values are to be understood as {}``effective'' energies ramming 

385: into the ICM gas.

386: Similarly, estimates for the ICM density are also rather uncertain. 

387: If we assume that all the gas is contained within a radius of 

388: $1-2\, \rm{pc}$ (the approximate radial position of the dust emission), 

389: the density of a $15\, \rm{M}_\odot$ gas cloud would be between 

390: $20-150\, \rm{cm}^{-3}$. 

391: For simplicity, we also assume that the shock, once it leaves the ST stage,  

392: remains radiative, which might not necessarily be the case for the  

393: low temperatures ($T< 10^5\, \rm{K}$) and shock speeds ($<50\, {\rm km s}^{-1}$)

394: when the shock has swept up most of the ICM, as cooling is less efficient in 

395: this regime (Sutherland \& Dopita 1993).

396: 

397: Using $E_{51}=3\times10^{-4}$ and $n=20-150, \rm {cm}^{-3}$ we obtain 

398: $v_{s}\approx13-15\,{\rm km\, s}^{-1}$ after the shock has

399: swept up $15\, {\rm M}_{\odot}$. Since this velocity is lower than

400: the escape speed of $\approx40\, {\rm km\, s}^{-1}$ for M15, it 

401: follows that the energy of the ejected material from one MSS collision may 

402: not be sufficient to remove the ICM completely. In

403: fact, only for $E_{51}\gtrsim8\times10^{-4}$, which is at the upper end of

404: the estimated ejecta energy interval, the shocked 

405: ICM gas attains a velocity sufficiently large to leave the cluster.

406: 

407: On the other hand, the shock

408: also heats up the gas to high temperatures,  

409: ($\approx10-15\times10^{4}\,{\rm K}$ at those shock speeds). 

410: Gas at such temperatures expands beyond

411: a critical radius and flows out of the cluster as a wind, reducing

412: the amounts of ionized gas down to less than $1\,\rm{M}_{\odot}$ 

413: (Knapp et al.\ 1996).

414: Assuming that the dust follows the gas, the dust would therefore

415: leave the cluster on a timescale as short as $\sim10^{5}\,{\rm yr}$.

416: In fact, the wind should be even stronger in our scenario than for the 

417: static model considered in Knapp et al.\ (1996), since here the gas itself

418: has a considerable outwards speed through the shock.

419: 

420: So far, we assumed the existence of rather large amounts of gas,  

421: based on the amount of observed dust and M15's extremely low metallicity.

422: However, searches for gas in M15 have had very limited success, and 

423: Smith et al.\ (1995) estimate an upper limit for the total ICM mass of about 

424: $3\, {\rm M}_{\odot}$. The reason for the

425: much larger observed dust-to-gas ratio is not well understood

426: (see, e.g., van Loon et al.\ 2006). As the dust-to-gas ratio of the

427: material lost in the winds of red giants should scale with the  

428: metallicity of the stars (van Loon et al.\ 2005), thus resulting in more 

429: than $10\,{\rm M}_\odot$ of gas, it appears that additional processes may be 

430: at work that remove the gas more easily than the dust. Nevertheless, it is

431: also possible that most of the ICM is in molecular form as the

432: CO-to-H$_2$ conversion factor is not known for such low metallicities

433: and extreme radiative environments (van Loon et al.\ 2006). Repeating the 

434: previous calculation for an ICM mass 

435: of $M=3\, {\rm M}_{\odot}$ and $n=3-30\, \rm{cm}^{-3}$ accordingly, we obtain

436: $v_{s} = 52-63\,{\rm km\, s}^{-1}$ for $E_{51}= 3\times10^{-4}$, while

437: for $E_{51}= 2\times10^{-4}$ $v_{s}=39-47\, {\rm km\, s}^{-1}$

438: which is close to and larger than the cluster escape speed of 

439: $40 \rm{km\, s}^{-1}$. In this case, the ejecta of one 

440: MSS collision would likely be able to accelerate this gas 

441: out of the cluster.

442: 

443: As a final caveat, we note that it is not very clear whether

444: the dust will follow the rather low-density ionized gas. This

445: strongly depends, among other quantities, on the dust grain properties

446: and their electric potential relative to the ionized gas in a rather

447: complicated way (Draine \& Salpeter 1979). For instance, Nozawa et

448: al.\ (2006) and similarly Slavin et al.\ (2004) found in their  

449: simulations of shocks driven into dusty interstellar medium that, 

450: while small grains with sizes of $\approx0.01\,{\rm \mu m}\,$ get 

451: destroyed by sputtering and grain-grain collisions, only grains with 

452: sizes $\approx0.1\,{\rm \mu m}$ are actually dragged along with the gas, 

453: while grains with sizes $\geq1\, {\rm \mu m}$

454: remain almost unaffected and do not follow the shock wave. On the

455: other hand, if the size distribution of the dust grains is similar

456: to the one for the local interstellar medium, we see (e.g., Mathis 1996)

457: that most of the dust mass is in grains with sizes of $\approx0.1\, 

458: \rm{\mu m}$. This means that, even if not all dust particles follow the gas

459: flow, we can nevertheless expect that most of the dust mass will

460: remain sufficiently well coupled to the gas and consequently be removed.

461: 

462: \section{Conclusions}

463: \label{conclusion}

464: 

465: In this Letter we proposed a new mechanism to explain the  

466: relatively

467: short lifetimes of the ICM dust in a dense GC, developing our  

468: arguments in detail

469: for the case of M15. By calculating the

470: rate of stellar collisions using the detailed model for M15 by  

471: McNamara et al.\ (2004),

472: we find a remarkable, close agreement between the average time between  

473: collisions

474: and the inferred dust lifetime of $\simeq10^{6}\,{\rm yr}$ (Boyer et  

475: al.\ 2006) in this

476: cluster, pointing to a direct link between the two phenomena. We  

477: argue that

478: the kinetic energy of the material ejected during a stellar

479: collision may be sufficient to remove the dust from the cluster, 

480: depending on the assumed ICM mass, either directly by  

481: accelerating

482: dust and gas to velocities larger than the cluster escape speed, or  

483: indirectly,

484: by accelerating and heating the gas, which then expands and leaves

485: the cluster as a wind, carrying the dust along with it. Although there

486: are some uncertainties as to how well the dust will couple

487: to the gas, especially at the low shock speeds expected for this

488: problem, there are some indications from simulations of supernova

489: remnants that might support sufficient coupling (e.g., Nozawa et al.\  

490: 2006).

491: In addition, at least some dust grains can also

492: be efficiently destroyed by grain-grain collisions or sputtering (Slavin

493: et al.\ 2004; Nozawa et al.\ 2006), which further helps to reduce the

494: amount of observable dust in the cluster. On the other hand,

495: the results of these studies may not be directly applicable to our

496: scenario because, e.g., the intense UV field

497: present in a cluster like M15 could strongly affect the electric  

498: potential of the grains and, therefore, their coupling to the ionized gas

499: (Draine \& Salpeter 1979).

500: 

501: With a detailed model for M15 we were also able to calculate the

502: formation rate and mass distribution of BSs through MS-MS collisions. 

503: By comparing with the observed number of BSs

504: in the cluster, we derive an upper limit to their average lifetime

505: of $\simeq1\,{\rm Gyr}$, consistent with current stellar structure  

506: and evolution

507: models for BSs (e.g., Leigh et al.\ 2007).

508: 

509: We conclude that the interaction of ejected gas from

510: stellar collisions with the ICM will strongly influence the  

511: observable signal of the dust

512: in the ICM, and, given the remarkable agreement between dust  

513: lifetimes and collision times

514: in M15, this represents a promising mechanism to explain the very  

515: short dust and ICM lifetimes

516: in GCs.

517: 

518: 

519: 

520: 

521: \acknowledgements{We thank Holger Baumgardt for providing us with

522: snapshots of his M15 models and for helpful comments.

523: We also thank James Lombardi for helpful discussions regarding

524: stellar collisions, Bruce Draine for valuable advice during a

525: recent visit to Northwestern, and the referee, Jacco van Loon, who

526: helped us improving the clarity of the paper.

527: This work was supported by NASA Grants NNG04G176G and NNX08AG66G.}

528: 

529: %\bibliographystyle{/home/stefan/.TeX/bibtex/apj}

530: %\bibliography{DusBS}

531: 

532: \begin{thebibliography}{24}

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

534: 

535: \bibitem[{{Baumgardt} \& {Makino}(2003)}]{2003MNRAS.340..227B}

536: {Baumgardt}, H. \& {Makino}, J. 2003, \mnras, 340, 227

537: 

538: \bibitem[{{Binney} \& {Tremaine}(1987)}]{1987gady.book.....B}

539: {Binney}, J. \& {Tremaine}, S. 1987, {Galactic dynamics} (Princeton, NJ,

540:   Princeton University Press, 1987, 747 p.)

541: 

542: \bibitem[{{Boyer} {et~al.}(2006)}]{2006AJ....132.1415B}

543: {Boyer}, M.~L., {Woodward}, C.~E., {van Loon}, J.~T., {Gordon}, K.~D., {Evans},

544:   A., {Gehrz}, R.~D., {Helton}, L.~A., \& {Polomski}, E.~F. 2006, \aj, 132,

545:   1415

546: 

547: \bibitem[{{Chatterjee} {et~al.}(2008)}]{2008ApJ.........}

548:   {Chatterjee}, S., {Fregeau}, J.~M., \& {Rasio}, F.~A. 2008, ApJ,

549:   subm.

550: 

551: \bibitem[{{Chevalier}(1977)}]{1977ARA&A..15..175C}

552: {Chevalier}, R.~A. 1977, \araa, 15, 175

553: 

554: \bibitem[{{Cioffi} {et~al.}(1988){Cioffi}, {McKee}, \&

555:   {Bertschinger}}]{1988ApJ...334..252C}

556: {Cioffi}, D.~F., {McKee}, C.~F., \& {Bertschinger}, E. 1988, \apj, 334, 252

557: 

558: \bibitem[{{Cox}(1972)}]{1972ApJ...178..159C}

559: {Cox}, D.~P. 1972, \apj, 178, 159

560: 

561: \bibitem[{{Draine} \& {Salpeter}(1979)}]{1979ApJ...231...77D}

562: {Draine}, B.~T. \& {Salpeter}, E.~E. 1979, \apj, 231, 77

563: 

564: \bibitem[{{Evans} {et~al.}(2003)}]{2003A&A...408L...9E}

565: {Evans}, A., {Stickel}, M., {van Loon}, J.~T., {Eyres}, S.~P.~S., {Hopwood},

566:   M.~E.~L., \& {Penny}, A.~J. 2003, \aap, 408, L9

567: 

568: \bibitem[{{Knapp} {et~al.}(1996)}]{1996ApJ...462..231K}

569: {Knapp}, G.~R., {Gunn}, J.~E., {Bowers}, P.~F., \& {Vasquez Poritz}, J.~F.

570:   1996, \apj, 462, 231

571: 

572: \bibitem[{{Leigh} {et~al.}(2007){Leigh}, {Sills}, \&

573:   {Knigge}}]{2007ApJ...661..210L}

574: {Leigh}, N., {Sills}, A., \& {Knigge}, C. 2007, \apj, 661, 210

575: 

576: \bibitem[{{Lombardi} {et~al.}(1996){Lombardi}, {Rasio}, \&

577:   {Shapiro}}]{1996ApJ...468..797L}

578: {Lombardi}, Jr., J.~C., {Rasio}, F.~A., \& {Shapiro}, S.~L. 1996, \apj, 468,

579:   797

580: 

581: \bibitem[Lombardi et al.(2002)]{2002ApJ...568..939L} Lombardi, J.~C., Jr., 

582: Warren, J.~S., Rasio, F.~A., Sills, A., 

583: \& Warren, A.~R.\ 2002, \apj, 568, 939 

584: 

585: \bibitem[{{Mapelli} {et~al.}(2004)}]{2004ApJ...605L..29M}

586: {Mapelli}, M., {Sigurdsson}, S., {Colpi}, M., {Ferraro}, F.~R., {Possenti}, A.,

587:   {Rood}, R.~T., {Sills}, A., \& {Beccari}, G. 2004, \apjl, 605, L29

588: 

589: \bibitem[{{Mathis}(1996)}]{1996ApJ...472..643M}

590: {Mathis}, J.~S. 1996, \apj, 472, 643

591: 

592: \bibitem[{{McNamara} {et~al.}(2004){McNamara}, {Harrison}, \&

593:   {Baumgardt}}]{2004ApJ...602..264M}

594: {McNamara}, B.~J., {Harrison}, T.~E., \& {Baumgardt}, H. 2004, \apj, 602, 264

595: 

596: \bibitem[{{Nozawa} {et~al.}(2006){Nozawa}, {Kozasa}, \&

597:   {Habe}}]{2006ApJ...648..435N}

598: {Nozawa}, T., {Kozasa}, T., \& {Habe}, A. 2006, \apj, 648, 435

599: 

600: \bibitem[{{Ruffert}(1992)}]{1992A&A...265...82R}

601: {Ruffert}, M. 1992, \aap, 265, 82

602: 

603: \bibitem[{{Sills} {et~al.}(2001)}]{2001ApJ...548..323S}

604: {Sills}, A., {Faber}, J.~A., {Lombardi}, Jr., J.~C., {Rasio}, F.~A., \&

605:   {Warren}, A.~R. 2001, \apj, 548, 323

606: 

607: \bibitem[{{Slavin} {et~al.}(2004){Slavin}, {Jones}, \&

608:   {Tielens}}]{2004ApJ...614..796S}

609: {Slavin}, J.~D., {Jones}, A.~P., \& {Tielens}, A.~G.~G.~M. 2004, \apj, 614, 796

610: 

611: \bibitem[{{Smith}(1999)}]{1999PASP..111..980S}

612: {Smith}, G.~H. 1999, \pasp, 111, 980

613: 

614: \bibitem[{{Smith} {et~al.}(1995){Smith}, {Woodsworth}, \&

615:   {Hesser}}]{1995MNRAS.273..632S}

616: {Smith}, G.~H., {Woodsworth}, A.~W., \& {Hesser}, J.~E. 1995, \mnras, 273, 632

617: 

618: \bibitem[{{Spergel}(1991)}]{1991Natur.352..221S}

619: {Spergel}, D.~N. 1991, \nat, 352, 221

620: 

621: \bibitem[{{Sutherland} \& {Dopita}(1993)}]{1993ApJS...88..253S}

622: {Sutherland}, R.~S. \& {Dopita}, M.~A. 1993, \apjs, 88, 253

623: 

624: \bibitem[{{van Loon} {et~al.}(2005)}]{2005A&A...438..273V}

625: {van Loon}, J.~T., {Cioni}, M.-R.~L., {Zijlstra}, A.~A., \& {Loup}, C. 2005,

626:   \aap, 438, 273

627: 

628: \bibitem[{{van Loon} {et~al.}(2006)}]{2006MNRAS.365.1277V}

629: {van Loon}, J.~T., {Stanimirovi{\'c}}, S., {Evans}, A., \& {Muller}, E. 2006,

630:   \mnras, 365, 1277

631: 

632: \end{thebibliography}

633: 

634: 

635: 

636: \end{document}

637: