astro-ph0310200/ms.tex
1: \documentclass[]{emulateapj}
2: \usepackage{psfig}
3: \submitted{Accepted to The Astrophysical Journal, Letters (December 16, 2003)}
4: 
5: \advance \voffset by 0.00cm\relax
6: \def\msun{{\rm ~M}_{\odot}}
7: \def\rsun{{\rm ~R}_{\odot}}
8: \def\mdot{\dot M}
9: \def\mpy{{\rm ~M}_{\odot} {\rm ~yr}^{-1}}
10: 
11: \shorttitle{X-RAY BINARY POPULATIONS: NGC1569}
12: \shortauthors{BELCZYNSKI, KALOGERA, ZEZAS, \& FABBIANO }
13: 
14: %\received{2003 October 7}
15: \begin{document}
16: 
17: \title{X-RAY BINARY POPULATIONS: THE LUMINOSITY FUNCTION OF NGC1569}
18: 
19: \author{K.\ Belczynski\altaffilmark{1,2}, V.\ Kalogera\altaffilmark{1}, 
20:         A.\ Zezas \altaffilmark{3}, and G.\ Fabbiano\altaffilmark{3}}
21: 
22: \affil{ $^{1}$ Northwestern University, Physics \& Astronomy,
23:        2145 Sheridan Rd., Evanston, IL 60208\\ 
24:        $^{2}$ Lindheimer Postdoctoral Fellow\\
25:        $^{3}$ Harvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, 
26:        MA 02138\\
27:        belczynski, vicky@northwestern.edu; azezas, pepi@head.cfa.harvard.edu}
28: 
29:  \begin{abstract} 
30:  Using the population synthesis code {\em StarTrack} we construct the first
31:  synthetic X-ray binary populations for direct comparison with the X-ray 
32:  luminosity function (XLF) of NGC~1569 observed with {\em Chandra}. 
33:  Our main goal is to examine whether it is possible to reproduce the XLF 
34:  shape with our models, given the current knowledge for the star-formation 
35:  history of this starburst galaxy. We thus produce hybrid models meant 
36:  to represent the two stellar populations: 
37:  one old, metal-poor with continuous star-formation for $\sim$ 1.5\,Gyr and 
38:  another recent and metal-rich population. To examine the validity of the models 
39:  we compare XLFs calculated for varying ages of the populations and varying 
40:  relative weights for the star-formation rates in the two populations. 
41:  We find that, for typical binary evolution parameters, it is 
42:  indeed possible to quite closely match the observed XLF shape. 
43:  The robust match is achieved for an age of
44:  the young population and a ratio of star formation rates in the two
45:  populations that are within factors of 1.5 and 2, respectively, of those 
46:  inferred from HST observations of NGC~1569. 
47:  In view of this encouraging first step, we discuss the implications of our X-ray
48:  binary models and their potential as tools to study binary populations in 
49:  galaxies.
50: 
51:  \end{abstract}
52: 
53: \keywords{binaries: close -- stars: evolution -- galaxies: individual (NGC~1569) -- 
54:           X-rays: binaries}
55: 
56: \section{INTRODUCTION}
57: \label{sec:intro}
58: 
59: X-ray binaries (XRB) in nearby galaxies  have been known since 
60: the {\em Einstein} era, but detections of significant samples were limited to 
61: just a few sources
62: (e.g., Magellanic Clouds or M31) and interpretation of the source properties was 
63: often hampered by confusion problems. {\em Chandra} observations 
64: have revolutionized XRB studies with the discovery of 
65: large numbers of point X-ray sources in galaxies even beyond the Local 
66: Group (e.g., Fabbiano \& White 2003). Short-term variability of many sources  
67: excludes the possibility of source confusion and 
68: strongly points toward accretion as the origin of the X-rays. 
69: The samples in most cases are large enough that we are now able to examine XRB 
70: {\em populations} in a wide range of galactic environments with different star 
71: formation histories. The samples are characterized 
72: by cumulative X-ray luminosity functions (XLFs) fitted by single or broken 
73: power-laws (e.g., Grimm et al.\  2003; Zezas \& Fabbiano 2002). 
74: 
75: It has been noted that the XLF slopes follow a rather systematic behavior with  
76: population age and possibly star formation rate (SFR) (Grimm et al. 2003; 
77: Sarazin et al. 2003). 
78: Such  behavior could constrain the SF history 
79: and properties of XRB populations in nearby galaxies. However, the 
80: development of  reliable diagnostics requires a sound physical understanding of the 
81: observed correlations. With the exception of a recent study (Sipior et al. 2003, 
82: which does not focus on a specific set of observations),
83: attempts to gain physical insight have been based so far on
84: analytical models that {\em assume} the existence of power-law XLFs, and that X-ray 
85: lifetime is  inversely proportional to 
86: X-ray luminosity and XRB properties do not depend on the SF history of the host 
87: galaxy (Wu 2001; Kilgard et al. 2002). 
88: 
89: In order to develop a set of useful diagnostics, it is 
90: important to develop theoretical models for XRB formation and evolution that depend 
91: on the star-formation history of the host galaxies, and allow us to identify the 
92: main physical elements that determine the XLF slopes. Given that this study focuses on 
93: observed {\em populations}, it is clear that the employment of population synthesis 
94: models is necessary. 
95: First pre-Chandra XRB models for starbursts were developed and compared to ASCA 
96: observations of the total X-ray luminosity of WR
97: galaxy He2-10 by Van Bever \& Vanbeveren 2000.
98: However, only now we may compare the specific theoretical models with the
99: observed point source populations.  
100: A collaborative effort has led to the development of a detailed    
101: population synthesis code {\em StarTrack} (see \S\,\ref{sec:models}) 
102: that specifically focus on careful, self-consistent XRB calculations.
103: Ultimately our goal is to construct a coherent picture of XRB 
104: formation and evolution based on 
105: theoretical models that have been first calibrated against observations of well-studied 
106: galaxies, and thus can be used in the interpretation of other XLF observations.
107: 
108: In this {\em Letter} we compare our model XLFs to {\em Chandra} observations of 
109: NGC~1569, a blue dwarf Irregular at a distance of 2.2Mpc (Israel 1988), characterized 
110: as a (post-)starburst galaxy. It has been selected as a good test case because its star 
111: formation history is {\em relatively} well constrained by HST optical and infrared 
112: observations (Aloisi et al. 2001; Greggio et al.1998; Vallenari \& Bomans 1996) 
113: {\em and} has a long ($\sim 80$\,ks) {\em Chandra} exposure providing a detection limit of 
114: $\simeq 10^{36}$\,erg\,s$^{-1}$ (Martin, Kobulnicky \& Heckman 2002; hereafter M02). 
115: We are mainly concerned with two questions: (i) is it at all possible to theoretically 
116: reproduce the observed XLF? and  (ii) do our models agree with the current constraints 
117: on the star-formation history of NGC~1569 derived by observations in other wavelenghts? 
118: 
119: 
120: \section{THEORETICAL MODELS}
121:  \label{sec:models}
122: 
123: The {\em StarTrack} code was originally developed for the modeling of binaries with 
124: two compacts (BKB), but has recently undergone major revisions 
125: (Belczynski et al. 2003, in preparation) 
126: intended to treat in detail the formation and evolution of XRBs, for any choice of 
127: SF history and metallicity. The main revisions include a 
128: detailed treatment of tidal synchronization and circularization (Hut 1981), 
129: individual treatment of various mass-transfer (MT) episodes, 
130: full numerical orbit evolution with angular momentum losses due to magnetic breaking, 
131: gravitational radiation, mass transfer/loss, and tides. 
132: 
133: We have calibrated the tidal implementation using observations of binary 
134: eccentricities and periods in stellar clusters (Mathieu et al.\ 1992) and of orbital 
135: decay in high-mass XRBs  (Levine et al.\ 2000). 
136: Our MT implementation involves the detailed 
137: calculation of Roche-lobe overflow MT rates based on radius-mass exponents 
138:  both for the donor stars and their Roche lobes. We have compared our results 
139: for a set of MT sequences to both published MT calculations (e.g., Beer \& 
140: Podsiadlowski 2002) and results 
141: obtained within our group using  an updated stellar evolution code (Ivanova et al. 2003) 
142: with very satisfactory agreement, much better than typical MT implementations. 
143: The modeled X-ray phases are 
144: identified as (i) stable, driven by nuclear evolution of the donor, magnetic braking 
145: or gravitational radiation 
146: (ii) thermally unstable with possibly anisotropic emission (King et al. 2001); 
147: (iii) Eddington-limited MT (cf.\ Kalogera et al.\ 2003, in preparation); and 
148: (iv) persistent or transient (critical MT rate from Dubus et al. 1999). We also 
149: account for wind 
150: accretion onto compact objects following Hurley, Tout \& Pols (2002). 
151: The updated code has been already tested and used for a study of 
152: Galactic ultracompact binaries (Belczynski \& Taam 2003).
153: 
154: In this paper we examine whether XRB models with rather standard binary-evolution 
155: parameters (i.e., not specifically selected for this study), but with star-formation 
156: history and metallicity consistent with what is known about NGC~1569 can produce an XLF 
157: shape 
158: in agreement with observations. We choose parameters from the reference model  in BKB, 
159: with just a few differences: the maximum neutron star (NS) mass is set equal to $2 \msun$, 
160: the most 
161: recently inferred natal NS kick distribution is incorporated (Arzoumanian, Chernoff \& 
162: Cordes 2002), the primary 
163: masses are selected in the range $4-100 \msun$ (with an initial-mass-function slope of 
164: -2.7; see Kroupa, Tout \& Gilmore 1993), and the secondary masses in the range 
165: $0.08-100 \msun$ with a flat mass ratio (secondary divided by primary mass) distribution. 
166: 
167: We construct our XRB models with focus on the estimated metallicities and
168: SFR properties of stellar populations in NGC~1569. A coherent picture has been 
169: developed (Aloisi et al. 2001; Greggio et al. 1998; Vallenari \& Bomans 1996) for two 
170: populations: (i) one 
171: metal-rich~($\sim 0.25$\,Z$_\odot$, although it could be comparable to 1\,Z$_\odot$; M02), 
172: young population formed in a global burst of star formation that lasted about 100 Myr and 
173: seems to have stopped 5-10Myr ago, and (ii) one metal-poor ($Z=0.004-0.0004$), old 
174: population formed by less vigorous (possibly by a factor of $10-20$ in SFR), continuous 
175: star formation for the past 1-1.5 Gyr. For consistency with these estimates we adopt 
176: $Z=0.005$ and $Z=0.0022$, for the young and the old populations, respectively. 
177: We examine the consistency of our models by allowing the relative SFR
178: weight and the population age and metallicity  to vary. 
179: 
180: 
181: \section{THE X-RAY LUMINOSITY FUNCTION} 
182:  \label{sec:results} 
183:  
184: We analyzed the archival {\em Chandra} data using the Ciao 
185: v3.0 data analysis suite. After screening for high background intervals we
186: searched for sources in the full 0.3-7.0~keV band with the
187: {\textit{wavedetect}} source detection algorithm. Then, for each source
188: within the D25 ellipse of the galaxy ($\rm{D_{max}=3.6'}$, $\rm{D_{min}=1.8'}$), 
189: we estimated the net 
190: number of counts using an aperture including all the
191: emission from a source. The background was determined locally from a
192: source-free annulus around each source. We detected 14 sources 
193: with a significance greater than 3$\sigma$ above the local background, in excellent 
194: agreement with M02, but we exclude one that has been identified as a supernova 
195: remnant by M02. Based on the $\textit{LogN-LogS}$ relation from the $\textit{Champ}$
196: survey (Kim et al.\ 2003) we estimate that at most two of our sources
197: within the D25 area are associated with background or foreground objects.
198: 
199: The relatively high number of counts of the faintest sources 
200: ($\sim15-20$ counts) compared to their typical background ($\sim5-10$ counts) and the 
201: lack of a turnover in the observed cumulative XLF (see Figure 2)  
202: indicate that incompleteness is not significant in the low luminosity end of the XLF. 
203: The higher level of diffuse emission in the central region of the galaxy (M02) may result 
204: to a higher detection threshold, but, given the very small number of sources in this region, 
205: we do not attempt to correct for this effect. In our XLF, apart from the standard 
206: Gehrels errors (Gehrels 1986), we include errors associated with the uncertainties in 
207: the observed count rate of each source, following Zezas \& Fabbiano (2002). The 
208: luminosity of the sources is in the 0.1-10.0~keV band assuming a power-law spectrum 
209: ($\Gamma=1.7$) with Galactic column density ($2.3\times10^{20}~\rm{cm^{-3}}$) and is 
210: corrected for absorption.
211:    
212: In our theoretical models X-ray luminosities ($L_X$) are calculated based on the 
213: {\em accretion} rate $\dot{M}$ with appropriate efficiencies for 
214: NS and black holes (BH). At present no corrections specific to the {\em Chandra} 
215: energy band are taken into account, due to the lack of a 
216: reliable spectral model across a wide range of frequencies. We further apply the Eddington 
217: limit and calculate the effects of associated non-conservative mass transfer with the lost 
218: matter carrying away the specific angular momentum of the accretor.   
219: 
220: To determine the contribution of transient systems to the XLF we need information on their 
221: X-ray duty cycle (DC), which however cannot be provided reliably by the disk instability 
222: theory.  Empirically it is thought that $DC \lesssim 1\%$ (e.g., Taam, King \& Ritter 
223: 2000). In {\em StarTrack} we adopt a DC value (in this study DC=1\%) and, for each 
224: system we randomly sample the probability that the source is
225: in outburst and then assign an X-ray luminosity equal to the Eddington limit,
226: whereas for the systems in quiescence we assume that their X-ray luminosities 
227: are too low to be detectable.
228: Donors more massive than the accretors can drive  mass transfer on their thermal 
229: timescale (e.g., Kalogera \& Webbink1996) and anisotropic emission is 
230: possible (King et al. 2001). We identify such systems based on the calculated radius-mass 
231: exponents. At present we do not account for the anisotropic-emission possibility, but 
232: do account for the possibility of super-Eddington transfer as described above. 
233: Both transient and thermal-timescale XRBs are Roche-lobe overflow (RLOF) systems,
234: and typically dominate old ($\gtrsim 100$\,Myr) populations (see also Sipior 
235: et al.\ 2003).  
236: 
237: 
238: \begin{figure}
239: \begin{center}
240: \psfig{figure=f1.eps,width=2.85in}
241: \caption{Non-monotonic behavior with time of theoretical normalized XLFs for 
242: two stellar populations: one old at 1.5 Gyr  and one young at age 10, 70, and 
243: 170\,Myrs (continuous SFR through 1.5\,Gyr, and 10, 70, and 100\,Myrs, 
244: respectively). the average SFR in the old population is assumed to be 20 times 
245: smaller than that in the young population.}
246: \vspace*{-1.3cm}
247: \label{fig:mod00}
248: \end{center}
249: \end{figure}
250: 
251: \begin{figure}
252: \begin{center}
253: \psfig{figure=f2.eps,width=2.85in}
254: \caption{The observed NGC1569 XLF with error bars (solid lines) plotted against 
255: predicted XLFs (dashed curves) shown with $1\,\sigma$\ sampling errors
256: (shaded areas).  All curves are normalized to the total number (13) of the 
257: detected sources. Each panel corresponds to different choices for the age 
258: of the young and old populations and the SFR ratio of the young relative to 
259: the old, respectively.  
260: {\em Top panel:} 110\,Myr, 1.5\,Gyr, and 20;     
261: {\em Middle panel:} 70\,Myr, 1.5\,Gyr, and 20;  
262: {\em Bottom panel:} 70\,Myr, 1.3\,Gyr, and 40.}
263: \vspace*{-1.3cm}
264: \label{fig:mod01}
265: \end{center}
266: \end{figure}
267: 
268: 
269: Based on the current understanding of the SF history of NGC1569 
270: (\S\,\ref{sec:models}), we calculate combined XLFs for ages 1--1.5\,Gyr and 
271: 70--110\,Myr, for the old and young populations, respectively. 
272: The choice of the relative weight of the two populations also affects the XLF 
273: shape and we choose SFR weight factors in the range 10--50. Population 
274: models tend to have rather uncertain absolute normalizations and the absolute
275: SFRs in NGC~1569 are not well constrained either (Vallenari \& Bomans1996). 
276: Although, the total number 
277: of sources in a galaxy (corrected carefully for selection effects, partial galaxy 
278: coverage, 
279: etc) can provide additional model constraints, in this first step of our studies we 
280: restrict ourselves to calibrations and comparisons to observations based on the XLF 
281: {\em shape}. Therefore we used observed and model XLFs normalized to the total number 
282: of objects, and we take into account the small-number bias by randomly selecting 
283: from the simulated source populations samples of 13 sources at a time. 
284: The sampling of the
285: underlaying model parent population is then repeated number of times ($\sim 10^4$) to 
286: yield the predicted XLFs and their associated errors.
287: 
288: 
289: Figure 1 indicates the behavior of XLF as a function of age of young XRB population 
290: of NGC1569, while the age of the old population is fixed at 1.5\,Gyr.
291: Unlike earlier suggestions (Wu 2001; Kilgard et al. 2002), we find that the 
292: dependence of the 
293: XLF slope on age is {\em non-monotonic} and this can be actually understood physically. 
294: At early times ($\sim 10$\,Myr) the XLFs are by far dominated by MT-fed
295: systems of the older population. The recent star formation episode barely begun
296: producing wind-fed systems with the most massive compact objects (formed earlier)
297: that accrete from strong winds from the next most massive donors. 
298: The combination of young and bright wind-fed sources with the MT-fed systems of
299: older population tends to make XLFs rather flat.
300: However, relatively quickly ($\sim 70$\,Myr) the very massive stars of young population 
301: end their lives 
302: and are removed from the XRB sample. They are replaced by less massive, and fainter 
303: wind-fed systems and the XLF becomes steeper. 
304: As time passes ($\sim 170$\,Myr) RLOF systems of younger population, that tend to be 
305: brighter than wind-fed sources of similar binary component masses, start becoming 
306: important in the total population and the XLF slope once again becomes flatter.
307: 
308: In Figure 2 we present XLFs calculated for NGC1569. 
309: Model A with parameters taken at face 
310: value from the HST observations: young population at 110\,Myrs 
311: with SFR of 20 times larger than that of the old population at 1.5\,Gyr.
312: The calculated function tends to be flatter and extends to higher $L_X$ 
313: values compared to the observed XLF. 
314: At the age of 110\,Myr since the onset of the burst (10\,Myr after its end), 
315: the young population includes a significant number of a bright RLOF sources 
316: flattening the XLF. 
317: Model B corresponds to the young population at an earlier age of 70\,Myrs 
318: (equal to the burst duration)
319: when it is dominated by high mass XRBs, and we already see that 
320: the XLF becomes steeper, but not quite enough to match the observed one. 
321: There are still quite a few RLOF systems formed in the old star formation 
322: episode. However, the duration of that episode as well as its end time  
323: are not very precisely established.
324: Model C corresponds to the young population at 
325: 70\,Myrs (equal to the burst duration), but the old one at 1.3\,Gyr with continuous 
326: star formation for 1\,Gyr. 
327: The predicted XLF quite closely match the observed one within the relevant errors. 
328: We note the tendency for higher luminosities in the models, but we note that we 
329: adopt bolometric luminosities, which are higher than those in the {\em Chandra} 
330: band.
331: In Table 1 we present the content of the X-ray binary population
332: for our best model (C). The population is dominated 
333: by young high-mass XRBs, but with significant contribution of 
334: old RLOF systems. In the young population, where only the most massive 
335: stars have ended evolution, many accretors are BH, while in the 
336: older population sources with NS dominate. 
337: 
338:                           
339: \begin{deluxetable}{crr}
340: \tablewidth{500pt}
341: \tablecaption{Model C XRB Sub-Populations\tablenotemark{a}}
342: \tablehead{ Type                        & Old Pop.     & Young Pop. }
343: \startdata
344: RLOF systems                            & 24.4\%       & 3.7\%     \\
345: NS/BH accretor                          &  1.5         & 4.0       \\
346: transient\tablenotemark{b}/persistent   &  0.1         & no trans. \\
347: thermal/rest                            &  0.38        & 0.25      \\
348:                                         &              &           \\
349: WIND-fed systems                        &  1.5\%       & 70.4\%    \\
350: NS/BH accretor                          &  1.0         & 0.53      
351: 
352: \enddata
353: \label{numbers01}
354: \tablenotetext{a}{Corresponds to all active ($L_x>1.22 \times 10^{36} erg\ s^{-1}$) 
355: sources.} 
356: \tablenotetext{b}{Only transients at outburst stage are listed here assuming DC=1\%.}
357: \end{deluxetable}
358: 
359: 
360: \section{DISCUSSION} 
361: \label{sec:discussion} 
362:  
363: We present our first results from XRB population models developed for comparison 
364: with current and future {\em Chandra} observations of nearby galaxies. 
365: We choose NGC~1569 as our first test case and we find good agreement
366: between our models and the observations. This agreement is even more 
367: remarkable in view of the fact that we did not
368: attempt to fine tune any of the model parameters related to X-ray binary 
369: evolution. 
370: However, we have explored other models with varying metallicities 
371: and IMF slopes, and found that both our quantitative and qualitative conclusions 
372: remain robust  and the other models do not
373: offer a better match to the observed XLF shape. This is true even when we account for 
374: the fraction of systems that can escape NGC~1569, due to systemic
375: velocities acquired at supernova explosions.
376: 
377: Examination of various models with properties consistent with NGC~1569 
378: constraints, lead us to conclusion that an age of 70\,Myr for the young
379: and 1.3 Gyr for the old population and a SFR relative weight of 40 are 
380: favored.
381: In order to get agreement between the model and the observed 
382: XLF, we require a recent burst that is younger than inferred 
383: from the optical/infrared data.
384: This slight discrepancy could be due to the fact that at this point we do not 
385: consider different black-hole binary spectral states and anisotropic emission 
386: from pulsar binaries. On the other hand the parameters of the older population
387: (which is dominated by old non-magnetized neutron star binaries) are very 
388: consistent with the latest picture from the HST data (Angeretti et al.\ 2003, 
389: private communication).
390: 
391: We consider these encouraging results only a small, first step in our exploration of XRB 
392: models and their comparison to observations. As we gain experience with the 
393: study of specific galaxies, we expect to develop a reliable calibration system that will 
394: then allow us to extract information about origin of XRBs in other galaxies. A natural 
395: extension of this study will include two elements: the exploration of constraints on the 
396: absolute normalization of the XLF in addition to its shape, and the comparison models 
397: with a sample of starburst galaxies that form a time sequence with ages in a wide range 
398: of values to address the theoretical basis for correlations suggested by Grimm et al. 
399: (2003), for example. Detailed examination of degeneracies in the derived constraints is 
400: also important.  
401: Moreover, the modeling of supernovae remnants may prove to
402: be necessary, since they may {\em i)} contribute significantly to and {\em
403: ii)} be hard
404: to remove from observed point source samples.
405: 
406:   
407: 
408: \acknowledgments
409: We thank the referee J.~Irwin, and also A.~King and 
410: T.~Maccarone for useful comments, and the Aspen Center for Physics and the 
411: NU Visitors' fund (to AZ) for support. 
412: We also thank M.~Tosi and L.~Angeretti for discussing their results prior to 
413: publication.
414: This work is partially supported by a Packard Fellowship, and 
415: a {\em Chandra} theory grant to VK, NASA LTSA grant NAG5-13056 to AZ and VK
416: and NASA grant NAS8-39073 to GF.  
417: 
418: 
419: \begin{references}
420: 
421: \reference{} Aloisi et al.\ 2001, \apj , 121,1425
422: \reference{} Arzoumanian, Z., Chernoff, D.\ F., \& Cordes, 
423:              J.\ M.\ 2002, \apj, 568, 289
424: \reference{} Beer, M.E.\ \& Podsiadlowski, P.\ 2002, \mnras , 331, 351
425: \reference{} Belczynski, K., Kalogera, V., \& Bulik, T.\ 2002,
426:              \apj, 572, 407 (BKB)
427: \reference{} Belczynski, K., \& Taam, R.E.\ 2003, \apj, submitted
428: \reference{} Dubus, G.\ et al.\ 1999, \mnras , 303, 139
429: \reference{} Fabbiano, G.\ \& White, N.\ 2003, review 
430: \reference{} Gehrels, N.\ 1986, \apj, 303, 336 
431: \reference{} Greggio, L., et al.\ 1998, \apj , 504, 725
432: \reference{} Grimm, H.-J, Gilfanov, M., \& Sunyaev, R.\ 2003, \mnras , 339, 793
433: \reference{} Hurley, J. R., Tout, C. A., \& Pols, O. R.\ 2002, \mnras, 329, 897
434: \reference{} Hut, P.\ 1981, \aap, 99, 126 
435: \reference{} Israel, F. P.\ 1988, \aap, 194, 24
436: \reference{} Ivanova, N., Belczynski, K., Kalogera, V., Rasio, F., \& 
437:              Taam, R. E.\ 2003, \apj, 592, 475
438: \reference{} Kalogera, V.\ \& Webbink, R.F.\ 1996, \apj , 458, 301
439: \reference{} Kilgard, R.E., et al.\ 2002, \apj , 573, 138
440: \reference{} Kim et al.\ 2003, \apj, in press (astro-ph/0308493)
441: \reference{} King, A.R.\ et al.\ 2001, \apj , 552, L109
442: \reference{} Kroupa, P., Tout, C.\ A., \& Gilmore, G.\ 1993,
443:              \mnras, 262, 545
444: \reference{} Levine, A., Rappaport, S.A., \& Zojcheski, G.\ 2000, 541, L194
445: \reference{} Martin, C.~L., Kobulnicky, H.~A., \& Heckman, T.~M.\ 2002, \apj, 
446:              574, 663 (M02) 
447: \reference{} Mathieu, R. D., et al.\ 1992, "Binaries as Tracers of 
448:              Stellar Formation", ed. Duquennoy, A. \& Mayor, M., Cambridge 
449:              University Press, p.278 
450: \reference{} Sarazin, C.L., et al.\  2003, \apj , 595, in press 
451: \reference{} Sipior, M., Eracleous, M., \& Sigurdsson, S.\ 2003, \apj , submitted 
452:              (astro-ph/0308077)
453: \reference{} Taam, R.E., King, A.R., \& Ritter H.\ 2000, \apj, 541, 329
454: \reference{} Vallenari, A.\ \& Bomans, D.J.\ 1996, \aap , 313, 713
455: \reference{} Van Bever, J., \& Vanbeveren, D.\ 2000, \aap, 358, 462
456: \reference{} Wu, K.\ 2001, Publications of the Astronomical Society of Australia, 18, 443
457: \reference{} Zezas, A.\ \& Fabbiano, G.\ 2002, \apj , 577, 726
458: 
459: \end{references} 
460: 
461: 
462: 
463: 
464: \end{document}
465: 
466: 
467: