1: \documentclass[12pt,preprint]{aastex}
2:
3: \begin{document}
4:
5: \title{A First Estimate Of The X-Ray Binary Frequency As A Function Of
6: Star Cluster Mass In A Single Galactic System}
7:
8: \author{D.~M. Clark \altaffilmark{1,2}, S.~S. Eikenberry \altaffilmark{2},
9: B.~R. Brandl \altaffilmark{3}, J.~C. Wilson \altaffilmark{6},
10: J.~C. Carson \altaffilmark{5}, C.~P. Henderson \altaffilmark{4},
11: T. ~L. Hayward \altaffilmark{7}, D.~J. Barry \altaffilmark{4},
12: A.~F. Ptak \altaffilmark{8}, E.~J.~M. Colbert \altaffilmark{8}}
13:
14: \altaffiltext{1} {Instituto de Astronom\'{i}a, Universidad Nacional
15: Aut\'{o}noma de M\'{e}xico, Apdo Postal 877, Ensenada, Baja
16: California, M\'{e}xico; dmclark@astrosen.unam.mx}
17: \altaffiltext{2}{Department of Astronomy, University of Florida,
18: Gainesville, FL 32611; dmclark@astro.ufl.edu}
19: \altaffiltext{3}{Leiden University, P.O. Box 9513, 2300 RA Leiden,
20: Netherlands.}
21: \altaffiltext{4}{Astronomy Department, Cornell University, Ithaca, NY
22: 14853.}
23: \altaffiltext{5}{Max Planck Institute for Astronomy, K\"{o}nigstuhl 17, D-69117 Heidelberg, Germany}
24: \altaffiltext{6}{Department of Astronomy, P.O Box 400325, University
25: of Virginia, Charlottesville, VA 22904.}
26: \altaffiltext{7}{Gemini Observatory, AURA/Casilla 603, La Serena,
27: Chile}
28: \altaffiltext{8}{Department of Physics and Astronomy, Johns
29: Hopkins University, 3400 North Charles St., Baltimore, MD 21218.}
30:
31: \begin{abstract}
32:
33: We use the previously-identified 15 infrared star-cluster counterparts
34: to X-ray point sources in the interacting galaxies NGC 4038/4039 (the
35: Antennae) to study the relationship between total cluster mass and
36: X-ray binary number. This significant population of X-Ray/IR
37: associations allows us to perform, for the first time, a statistical
38: study of X-ray point sources and their environments. We define a
39: quantity, $\eta$, relating the fraction of X-ray sources per unit mass
40: as a function of cluster mass in the Antennae. We compute cluster
41: mass by fitting spectral evolutionary models to $K_s$ luminosity.
42: Considering that this method depends on cluster age, we use four
43: different age distributions to explore the effects of cluster age on
44: the value of $\eta$ and find it varies by less than a factor of four.
45: We find a mean value of $\eta$ for these different distributions of
46: $\eta$ = 1.7$\times$10$^{-8}$ $M_{\sun}^{-1}$ with $\sigma_{\eta}$ =
47: 1.2$\times$10$^{-8}$ $M_{\sun}^{-1}$. Performing a $\chi^2$ test, we
48: demonstrate $\eta$ could exhibit a positive slope, but that it depends
49: on the assumed distribution in cluster ages. While the estimated
50: uncertainties in $\eta$ are factors of a few, we believe this is the
51: first estimate made of this quantity to ``order of magnitude''
52: accuracy. We also compare our findings to theoretical models of open
53: and globular cluster evolution, incorporating the X-ray binary
54: fraction per cluster.
55:
56: \end{abstract}
57:
58: \keywords{galaxies: starburst -- galaxies: star clusters -- X-rays:
59: binaries}
60:
61: \section{Introduction}
62:
63: The Antennae are a pair of colliding galaxies with an unusually large
64: number of X-ray point sources. High resolution X-ray observations
65: taken with {\it Chandra} revealed 49 new individual sources
66: \citep{fab01}, where previous observations only indicated extended
67: filamentary structure \citep{fab97}. These X-ray sources range in
68: luminosity from $10^{38}-10^{40}$ ergs s$^{-1}$. Most of these are
69: thought to be X-ray binaries (XRBs) with a black hole compact
70: companion \citep{fab01}.
71:
72: In addition to numerous X-ray sources, the Antennae contain many
73: bright, massive young clusters that are evident in both optical, {\it
74: HST} \citep{whi99} and infrared (IR) \citep[henceforth Paper I]{bra05}
75: images. This makes this pair of interacting galaxies an ideal target
76: for studying the environments of XRBs. In \citet[henceforth Paper
77: II]{cla07} we performed an extensive study of the XRB environments
78: using {\it Chandra} X-ray images and $J$ and $K_s$ IR images
79: (Paper I). Our present paper will expand on our previous study by
80: exploring the relationship between XRBs and cluster mass in the
81: Antennae.
82:
83: Recent theoretical models of young, massive cluster evolution provide
84: a framework for comparison to our observational study. Two in
85: particular, \citep{osk05,sep05}, incorporate the fraction of XRBs per
86: cluster in their models. \citet{osk05} use a population synthesis
87: code to study the evolution of X-ray emission in young, massive
88: clusters. \citet{sep05} investigate the role of supernova kicks in
89: XRB expulsion from the parent cluster using the population synthesis
90: code, StarTrack. They also incorporate the number of XRBs for a range
91: in cluster mass. We will compare our measurements of the fraction of
92: XRBs per cluster in the Antennae to those predicted by these models.
93:
94: We organize our paper as follows: In \S2 we give a brief summary of
95: our previous work on the Antennae and then define a quantity, $\eta$,
96: relating the XRB fraction to cluster mass in the Antennae. In our
97: analysis, we estimate cluster mass using $K_s$ luminosity, which
98: depends non-trivially on the assumed cluster age. We explore cluster
99: age/luminosity relations and their impact on our mass estimates in
100: \S3. We compare $\eta$ to the measured value predicted by theoretical
101: cluster evolutionary models and present conclusions in \S4.
102:
103: \section{Observations and Data Analysis}
104:
105: \subsection{Infrared Images}
106:
107: This paper is based on infrared (IR) $J$ (1.25 $\mu$m) and $K_s$ (2.15
108: $\mu$m) images of the Antennae galaxies. We initially presented these
109: data and discussed the details of their reduction in Paper I. In
110: summary, 20-minute total exposures in each filter were acquired using
111: the Wide-field InfraRed Camera (WIRC -- see \citet{wil03} for details)
112: on the Palomar 5-m telescope during the night of March 22nd, 2002. In
113: Paper II, we made a frame-tie between the IR and X-ray images using IR
114: counterparts to circumvent the poor absolute astrometric accuracy of
115: {\it Chandra} ($\sim1\farcs5$). We matched seven IR sources from the
116: WIRC images with {\it Chandra} X-ray point sources. Using a least
117: squares fit of a linear matching function we tied {\it Chandra} right
118: ascension and declination to WIRC {\it x, y} pixel positions. The rms
119: positional uncertainty is $\sim0\farcs5$. With a strong astrometric
120: frame-tie in place we were able to accurately identify IR counterparts
121: to X-ray sources. We found 19 IR counterparts within $1\farcs5$ of an
122: X-ray source, 13 of which were within 1$\farcs$0 of an X-ray source.
123: After estimating the IR source density, we predict only two of
124: the ``strong'' matches (separations $<1\farcs0$) and three of the
125: ``possible'' matches (separations between 1$\farcs$0 -- 1$\farcs$5)
126: are due to chance superpositions of unrelated objects.
127:
128: In Paper II, we pointed out two important implications for these
129: results. First, that there is clearly a significant excess of IR
130: counterparts within $1\farcs0$ of the X-ray sources -- 13, where we
131: expect only two in the null hypothesis of no physical counterparts.
132: Even including the ``possible'' counterparts out to $1\farcs5$, we
133: have a total of 19 counterparts, where we expect only five are chance
134: superpositions. Secondly, this implies that for any given ``strong''
135: IR counterpart, we have a probability of $\sim$ 85\% ($11/13$ with a
136: $1\sigma$ uncertainty of 0.3\footnote{Found using confidence levels
137: for small number statistics listed in Tables 1 and 2 of
138: \citet{geh86}.}) that the association with an X-ray source is real.
139: Even for the ``possible'' counterparts, the probability of true
140: association is $\sim$50\%. Therefore, regardless of the physical
141: separations between the X-ray sources and their IR counterpart, we are
142: confident that the majority of these associations are real.
143:
144: We note that of the 19 X-ray sources with counterparts, two are the
145: nuclei \citep{zez02a}, one is a background quasar \citep{cla05}, and
146: two share the same IR counterpart. Therefore, in this paper we will
147: only consider the 15 IR counterparts (of the original 19) that are
148: star clusters in the Antennae.
149:
150:
151: \subsection{Photometry}
152:
153: We performed aperture photometry in both the $J$ and $K_s$ bands on
154: all 15 IR cluster counterparts plus an additional 204 clusters
155: identified by eye in these IR images of the Antennae (see also Paper I
156: and the tables there-in). We defined our aperture as $\sim3\sigma$ of
157: the Gaussian PSF, where $J$ had a full width at half maximum (FWHM) of
158: $1\farcs2$ and $K_s$ had a FWHM of $0\farcs9$. We measured a mean and
159: median sky background flux in two separate annuli between $\sim$6 --
160: 10$\sigma$ of the PSF. To account for the exceptionally crowded field
161: of the Antennae, we employed the use of background arcs instead of
162: annuli. Multiplying these four measurements by the area of the central
163: aperture and then subtracting these from the flux in the central
164: aperture yielded four separate source flux measurements. We defined
165: error in sky background, $\sigma_{sky}$, as the standard deviation of
166: the four measured source fluxes. We also considered Poisson noise,
167: $\sigma_{adu}$, defined as the total source flux divided by the square
168: root of the gain for the WIRC instrument. The gain for WIRC during
169: the observations was $2e^{-} DN^{-1}$ \citep{wil03}\footnote{At the
170: time of the Antennae observations, WIRC was equipped with a Hawaii-1
171: 1K$\times$1K detector and this is the gain for it.}. Adding
172: $\sigma_{sky}$ and $\sigma_{adu}$ in quadrature, we computed the total
173: error in flux, $\sigma_{flux}$. We converted fluxes to magnitudes
174: using a bright, 2MASS star in the field and defined the error in
175: magnitude, $\sigma_m$, as $\sigma_{flux}$ divided by the mean flux.
176: Typical errors in magnitude were $\sim$0.06 mag in both bands, with no
177: error above 1.0 mag.
178:
179: To estimate cluster masses we needed to compute $K_s$ luminosity
180: ($M_{K_s}$). We computed $M_{K_s}$ using reddening derived from $(J -
181: K_s)$ colors (Paper II). Assuming all clusters are dominated by O
182: and B stars, their intrinsic $(J-K_s)$ colors are $\sim$0.2 mag.
183: Approximating this value as 0 mag, this allowed us to estimate
184: $A_{K_s}$ as $\simeq$ $(J-K_s)_{obs}$/1.33 using the extinction law
185: defined in \citet{car89}.
186:
187: \subsection{XRB-to-Cluster Mass Fraction}
188:
189: We assume for now that cluster mass is proportional to $K_s$
190: luminosity -- i.e. that the stellar composition of all clusters is the
191: same. We defined a luminosity cutoff for statistical purposes as
192: $M_{K_s}$ = $-13.2$ mag (see Paper II for details). We binned the
193: data by 0.2 mag in $M_{K_s}$ and then calculated an average flux per
194: bin. Computing the fraction of the total number of clusters per
195: average flux of each bin, we took this as the probability of finding a
196: cluster with a specific mass. In Figure 1 we compare this probability
197: for both clusters with X-ray sources and all clusters in the Antennae.
198: This shows that XRBs are more common in more massive clusters.
199:
200: This result is not surprising -- as star cluster mass increases, so
201: does the number of massive stars in it. Through stellar evolution, a
202: certain fraction of these stars will die in supernova explosions,
203: leaving behind neutron star or black hole remnants. In turn, a
204: fraction of these stellar remnants will retain/acquire a mass-donating
205: companion star, becoming a detectable XRB. Thus, through sheer
206: numbers of stars in more massive clusters, we expect a greater
207: likelihood of finding XRBs in them. This leads us to two interesting
208: questions: 1) quantitatively, what cluster mass will more likely
209: produce an XRB and 2) is there some intrinsic property of massive
210: cluster physics that favors the production of XRBs, beyond simple
211: scaling with mass?
212:
213: We believe that our large sample of IR-to-X-ray associations provides
214: the first dataset sufficient to estimate the answers to these
215: questions for the Antennae galaxies. In our approach to answer these
216: questions we explore the relationship between the number of X-ray
217: detections per unit mass as a function of cluster mass in the
218: Antennae. We can formalize this expression in the following equation:
219:
220: \begin{equation}
221: N_X(M_c) = N_{Cl}(M_c)\cdot\eta(M_c)\cdot M_c
222: \end{equation}
223:
224: Here, $N_X(M_c)$ is the number of detected X-ray sources with an IR
225: cluster counterpart, $N_{Cl}(M_c)$ is the number of detected clusters,
226: and $\eta(M_c)$ is the fraction of X-ray sources per unit mass, all as
227: a function of cluster mass, $M_c$.
228:
229: If $\eta(M_c)$ increases or decreases over a range in $M_c$, this
230: means there could be something peculiar about massive cluster physics
231: to favor or suppress XRB formation. In contrast, a constant
232: $\eta(M_c)$ across all $M_c$ would indicate that more massive clusters
233: are more likely to have an XRB simply because they have more stars.
234:
235: While $\eta(M_c)$ is a powerful tool in studying the number of XRBs
236: per cluster, it requires that we know the mass of each star
237: cluster. However, extrapolating the masses of the Antennae clusters
238: from our photometric data required models that called for estimates of
239: ages and metallicities. While we successfully constrained these
240: inputs and determined cluster masses (see below and \S3), we first
241: sought to compute $\eta(M_c)$ in terms of a purely observable quantity
242: -- flux. Calculating $\eta(M_c)$ as a function of $K_s$-band flux,
243: $\eta(F_{K_s})$, allowed us to investigate non-model dependent trends
244: in $\eta(M_c)$.
245:
246: We calculated $\eta(F_{K_s})$ for clusters with a $K_s$-band
247: luminosity brighter than the -13.2 mag cutoff using bin sizes of
248: $F_{K_s} = 4\times10^6$ in counts (DN) (Figure 2). This bin size was
249: small enough to show a trend in $\eta(F_{K_s})$, but large enough to
250: contain at least two clusters with X-ray sources, allowing us to
251: assign error bars to each value of $\eta(F_{K_s})$. The errors
252: plotted on the graph are the measurement uncertainty in the mean value
253: of the four $\eta(F_{K_s})$ added in quadrature with the Poisson
254: uncertainty of the mean $\eta(F_{K_s})$ in each bin. Due to the small
255: sample size per bin, we computed these uncertainties using the small
256: number statistics formulae described in \citet{kee62}. Figure 2 shows
257: that $\eta(F_{K_s})$ is roughly consistent with a constant value of
258: $5.4\times10^{-8}$ $F_{K_s}^{-1}$ with an uncertainty of
259: $\sigma_{\overline{\eta}}= 1.8\times10^{-8}$ $F_{K_s}^{-1}$.
260:
261: In essence, $\eta(F_{K_s})$ is comparing two different mass
262: distributions, $N_{Cl}(F_{K_s})$, the mass distribution for all
263: clusters in the Antennae and $N_X(F_{K_s})/F_{K_s}$, the mass
264: distribution for clusters with X-ray sources, normalized by flux. If
265: there is a constant number of X-ray sources per unit cluster mass as
266: suggested by Figure 2, then these two mass distributions should be the
267: same. We can further corroborate this result by comparing
268: $N_X(F_{K_s})/F_{K_s}$ and $N_{Cl}(F_{K_s})$ using a two-sided
269: Kolmogorov-Smirnov (K-S) test. The K-S test yielded a D-statistic of
270: 0.75 and a probability of 0.107 that they are related. Considering
271: the separate cluster mass populations as two probability
272: distributions, each can be expressed as a cumulative distribution.
273: The D-statistic is then the absolute value of the maximum difference
274: between each cumulative distribution. This test quantitatively
275: demonstrates that there is nothing peculiar about massive clusters in
276: the Antennae with associated XRBs. We also computed the Pearson {\it
277: r} linear correlation coefficient between $N_X(F_{K_s})/F_{K_s}$ and
278: $N_{Cl}(F_{K_s})$, finding a value of 0.99. Since a value of 1 means
279: a perfect linear fit, this value of {\it r} further substantiates the
280: observed relationship in $\eta(F_{K_s})$.
281:
282: We then converted $\eta(F_{K_s})$ into the more conventional units of
283: solar mass. Here we assume all clusters are coeval. Selecting
284: $M_{K_s}$ listed in the Bruzual-Charlot (BC) cluster evolutionary
285: models \citep{bru03} for a 20 Myr, $1M_{\sun}$ cluster as a typical
286: value in the Antennae \citep{whi99}, we converted the model $M_{K_s}$
287: to $F_{K_s}$ using the standard relationship between luminosity and
288: flux. Multiplying $\eta(F_{K_s})$ by $F_{K_s}$ we converted
289: $\eta(F_{K_s})$ to solar masses: assuming a cluster metallicity of z =
290: 0.02, $\eta = 5.8\times10^{-8} M_{\sun}^{-1}$ with an uncertainty of
291: $\sigma_{\overline{\eta}} = 1.9\times10^{-8}M_{\sun}^{-1}$, while
292: assuming a metallicity of z = 0.05, $\eta =
293: 8.9\times10^{-8}M_{\sun}^{-1}$ with an uncertainty of
294: $\sigma_{\overline{\eta}} = 3.0\times10^{-8}M_{\sun}^{-1}$.
295:
296: While we assume all clusters are $\sim$20 Myr old, we note that the
297: actual range in ages should be $\sim$1--100 Myr \citep{whi99}. The BC
298: models indicate that clusters in this age range could vary by a factor
299: of as much as 100 in mass for a given $K_s$ luminosity. Since $\eta$
300: is a function of mass, incorrectly assigning cluster ages has the
301: potential to significantly impact $\eta$. In the next section, we
302: explore how differences in cluster age can affect the value of $\eta$.
303:
304: \section{Effects of Age and Slope in $\eta$}
305:
306: We investigated the effect differences in cluster age has on $\eta$ by
307: assuming three individual age distributions for the Antennae clusters:
308: an instant burst in which all clusters are the same age, a uniform
309: distribution, and a distribution of the form, $dN/d\tau \propto
310: \tau^{-1}$ \citep{fal05}. In each case, we assumed all clusters have
311: solar metallicity (z = 0.02) \citep{whi99}.
312:
313: In addition, we address the issue of whether the functional form of
314: $\eta(M_c)$ can be fit by a slope or if, indeed, $\eta(M_c)$ is
315: consistent with a single value, by performing a $\chi^2$ test. We
316: computed the $\Sigma\chi^2$ between the $\eta$ values for each bin and
317: the mean value of $\eta$, and computed the $\Sigma\chi^2$ between the
318: $\eta$ values for each bin and a fitted line to these values. The
319: difference between the two value of $\Sigma\chi^2$,
320: $\Delta\Sigma\chi^2$, indicated the significance of the fitted
321: slope. If $\Delta\Sigma\chi^2$ was less than one, with in one
322: $\Sigma\chi^2$ deviation, then this indicated that the fitted slope
323: was insignificant. While a value of $\Delta\Sigma\chi^2$ between one
324: and two, with in two $\Sigma\chi^2$ deviations, indicated only a weak
325: slope. Any value of $\Delta\Sigma\chi^2$ greater than two meant a
326: significant, nonzero slope in $\eta$. For our initial case, where we
327: estimated cluster mass using $F_{K_s}$ we found $\Delta\Sigma\chi^2 =
328: 1.9$, suggesting a weak, but non-negligible slope, in $\eta$.
329:
330: Considering the case of an instant burst in star formation, we
331: assigned the same age to describe all clusters and picked several such
332: values in the range 1 -- 100 Myr. Applying the BC models, we
333: converted $\eta(F_{K_s})$ to units of solar mass for a range in ages.
334: For each distribution in $\eta$ we computed a mean value (Figure 3).
335: The mean for these values is $\eta_{instant}$ = 3.3$\times$10$^{-8}$
336: $M_{\sun}^{-1}$ with a standard deviation of $\sigma_{instant}$ =
337: 2.9$\times$10$^{-8}$ $M_{\sun}^{-1}$.
338:
339: Performing our $\chi^2$ test for each distribution in $\eta$ (see
340: Figure 3), we found a mean in $\Delta\Sigma\chi^2$ of 0.86, with a
341: standard deviation, $\sigma_{\Delta\Sigma\chi^2}$ = 1.7. This seems
342: to indicated conflicting results. For some cases in assumed cluster
343: age there is no significant slope in $\eta$, while other cases show a
344: pronounced slope. This becomes more evident by examining Figure 3,
345: which shows a range in $\Delta\Sigma\chi^2$ of $\sim$0--6.5.
346: Incidentally, for $\eta_{20}$, $\Delta\Sigma\chi^2$ = 2.2, while for
347: $\eta_{100}$, $\Delta\Sigma\chi^2$ = 4.8$\times$10$^{-3}$.
348:
349: Next, we assumed a uniform age distribution for our Antennae cluster
350: sample. Picking ages at random from a uniform distribution between 1
351: -- 100 Myr, we assigned an age to each cluster in our sample.
352: Applying the BC models, we computed each cluster's mass based on the
353: assigned age, produced a mass distribution, and then calculated a mean
354: $\eta$. Performing a Monte Carlo (MC) simulation, we recreated
355: cluster mass distributions 10,000 times, producing a large sample of
356: $\eta$'s with a mean of $\overline{\eta}_{uniform}$ = 5.4$\times$10$^{-9}$
357: $M_{\sun}^{-1}$ and $\sigma_{\overline{\eta}_{uniform}}$ = 6.3$\times$10$^{-10}$
358: $M_{\sun}^{-1}$ (Figure 4).
359:
360: We then administered our $\chi^2$ test for each realized distribution
361: of $\eta$, finding a mean $\overline{\Delta\Sigma}\chi^2$ = 5.9 and
362: $\sigma_{\overline{\Delta\Sigma}\chi^2}$ = 5.2. In some cases, $\eta$
363: didn't exhibit a positive slope. Therefore, we investigated whether
364: the slope in $\eta$ tended to be mostly positive or mostly negative by
365: multiplying each $\overline{\Delta\Sigma}\chi^2$ statistic by the sign
366: of the slope. In doing so, we found a mean
367: $\overline{\Delta\Sigma}\chi^2$ = 5.8 and
368: $\sigma_{\overline{\Delta\Sigma}\chi^2}$ = 5.3. Thus almost all
369: distributions in $\eta$ for the uniform case had a positive slope.
370:
371: A more realistic approach is assuming the cluster ages are defined by
372: a power law (PL): $dN/d\tau \propto \tau^{-1}$ \citep{fal05}. These
373: authors derived their relationship using {\it HST} {\it
374: UBVI}H$_{\alpha}$ observations of $\sim$11,000 clusters. Fitting BC
375: models to photometry of each cluster, they generated an age
376: distribution. In our analysis we picked ages at random according to
377: this distribution. Mirroring the procedure used for the uniform
378: distribution case above, we created a sample of 10,000 $\eta$ values.
379: We found a mean for this sample of $\overline{\eta}_{PL}$ =
380: 8.7$\times$10$^{-9}$ $M_{\sun}^{-1}$ and a
381: $\sigma_{\overline{\eta}_{PL}}$ = 1.2$\times$10$^{-9}$
382: $M_{\sun}^{-1}$.
383:
384: Performing our $\chi^2$ test as we did with the uniform age
385: distribution case, we found a mean $\overline{\Delta\Sigma}\chi^2$ =
386: 5.1 and $\sigma_{\overline{\Delta\Sigma}\chi^2}$ = 6.0. Accounting
387: for variations in the sign of the slope in $\eta$, we found
388: $\overline{\Delta\Sigma}\chi^2$ = 4.3 and
389: $\sigma_{\overline{\Delta\Sigma}\chi^2}$ = 6.6. Therefore, the trend
390: in the slope of $\eta$ remains positive, but not as significantly as
391: the uniform age distribution case.
392:
393: In a final scenario, we used the cluster ages listed in the
394: electronic table available through \citet{men05} to fit ages to 144
395: clusters in our sample, including all 15 clusters associated with
396: X-ray sources. \citet{men05} derived ages by using three age
397: indicators -- $UBVI$ and $K_s$ broadband photometry to break the
398: age/reddening degeneracy, H$\alpha$ and Br$\gamma$ emission to
399: identify clusters less than 7 Myr, and CO band-head absorption from
400: narrow-band images for clusters $\sim$10 Myr. They then fit these
401: data to theoretical spectra for ages $<500$ Myr using a $\chi^2$
402: minimization technique \citep[for details, see][]{men05}.
403:
404: Following the method discussed in \S2.3, we used the BC models to
405: convert $M_{K_s}$ to mass for those clusters with \citet{men05} age
406: estimates. Using cluster bins of 1.0$\times10^7M_{\sun}$ in mass, we
407: computed four values for $\eta$ (see Figure 5). The errors plotted on
408: the graph are uncertainties in the mean value of $\eta$ added in
409: quadrature with the Poisson uncertainty in each bin. Again, we used
410: the small number statistic formulae in \citet{kee62} to compute these
411: errors. We found a mean in $\eta$ of 2.2$\times10^{-8}M_{\sun}^{-1}$
412: and $\sigma_{\overline{\eta}} = 1.2\times10^{-8}M_{\sun}^{-1}$.
413: Applying our $\chi^2$ test, we found $\Delta\Sigma\chi^2$ =
414: 1.0$\times$10$^{-2}$, implying no significant slope in $\eta$.
415:
416: We summarize our age analysis in Figure 6. Comparing the
417: distributions in $\eta$ for the fitted ages, instant burst, uniform
418: and power law age distributions, all values for $\eta$ are within a
419: factor of four. Assuming an instant burst of 20 Myr, $\eta$ differs
420: by a factor of ten (see Figure 6). Excluding specific ages for instant
421: bursts of cluster formation, there is little variation in the value of
422: $\eta$.
423:
424: Comparing our $\chi^2$ test for each of the four different age
425: assumptions indicates inconsistent results (see Table 1). In the
426: instant burst, uniform and power law cases, the slope in $\eta$ varies
427: from insignificant to distinctly positive, while $\eta$ has no
428: significant slope when the ages derived by \citet{men05} are fit to
429: our clusters. Therefore, we can not ignore that $\eta$ might have a
430: non-zero slope and we will discuss the implications of this in the
431: following section.
432:
433: \section{Summary and Conclusions}
434:
435: Through the quantity $\eta$, our investigation revealed conflicting
436: results with respect to the relationship between observed number of
437: XRBs and cluster mass. We performed a $\chi^2$ test on the slope in
438: $\eta$ for a variety of assumed star cluster age distributions and
439: found some cases where the slope was insignificant, while other cases
440: showed a distinctly positive slope. No slope indicates the observed
441: number of XRBs per unit mass is independent of cluster mass, while a
442: positive slope in $\eta$ suggests more XRBs per unit mass are produced
443: in more massive clusters. In the following discussion, we will
444: consider both a constant value in $\eta$ as well as a slope in $\eta$,
445: and the implications of each result.
446:
447: Initially, we estimated cluster mass by fitting BC spectrophotometric
448: models to cluster $M_{K_s}$, assuming all clusters are $\sim$20 Myr.
449: Recognizing that this method depends on cluster age, we explored how
450: different assumptions of a cluster age distribution for the Antennae
451: affect $\eta$ and showed that $\eta$ varies by a factor of roughly
452: four; although including individual ages for the instant bursts case
453: will increase the variations in $\eta$ to a factor of 10.
454:
455: We now proceed by comparing the mean value of $\eta$ for the four
456: different assumed age distributions, $\eta$ = 1.7$\times$10$^{-8}$
457: $M_{\sun}^{-1}$, to that predicted by models of young, massive
458: clusters. We will compare $\eta$ to theoretical models discussed in
459: \citet{osk05} and \citet{sep05}.
460:
461: In a recent study presented in \citet{osk05}, the author modeled X-ray
462: emission from young, massive star clusters, assuming a closed system
463: with constant mass, no dynamics and all stars are coeval, with cluster
464: metallicities of either z = 0.02 or z = 0.008. These models predict
465: $\sim$2-5\% of all OB stars in a cluster should produce high mass
466: X-ray binaries (HMXBs). In the models in \citet{osk05}, all clusters
467: are assumed to have masses of M$_{cl}$ = 10$^6$ $M_{\sun}$ with
468: stellar masses ranging from 1 -- 100 $M_{\sun}$. Considering the
469: Salpeter initial mass function (IMF) of the form $\xi(M)$ =
470: $M_0M^{-2.35}$ and defining stars with masses $>8$ $M_{\sun}$ as ``OB
471: stars'', for our purposes here, we estimated 6\% of all stars in the
472: model clusters are OB stars. Therefore, 1 -- 3$\times$10$^{-3}$ of
473: all stars in a cluster with an initial mass of 3$\times$10$^6$
474: $M_{\sun}$ (set by the Salpeter IMF) should produce an XRB. Since the
475: Salpeter IMF implies there are 7$\times$10$^5$ stars in a cluster,
476: then these stars should produce 7 -- 22$\times$10$^2$ XRBs -- orders
477: of magnitude greater than the $\sim$49 observed in the Antennae.
478: Expressing $\eta$ as a fraction of XRBs-to-cluster mass, the models in
479: \citet{osk05} suggest $\eta$ ranges from 3--7$\times$10$^{-4}$
480: $M_{\sun}^{-1}$. These values are greater by at least a factor of
481: 1000 from our estimates for $\eta$. Clearly, this predicts a much
482: larger number of compact object binaries than what we observed in the
483: Antennae. \citet{osk05} note that they were unable to detect HMXBs in
484: three massive ($\sim10^4$ $M_{\sun}$) clusters which they predict
485: should contain between 1-3 HMXBs. If our measured value for $\eta$
486: accurately describes the number of HMXBs formed, then it is not
487: surprising that \citet{osk05} fail to find any. As pointed out by
488: these authors, future modeling of HMXB formation is needed to
489: understand the discrepancy between the predictions and observations of
490: XRB populations in starburst galaxies.
491:
492: In another study, \citet{sep05} use the binary evolution and
493: population synthesis code, StarTrack \citep{bel02}, to investigate the
494: rate of XRB formation and ejection from young, massive clusters. This
495: program tracks stellar parameters such as radius, luminosity, mass and
496: core mass. The simulations are stopped at the formation of a compact
497: object. The models include mass transfer in binaries and include
498: transient XRBs. \citet{sep05} consider cluster masses ranging from
499: $5\times10^4$ $M_{\sun}$ to $5\times10^6$ $M_{\sun}$ and cluster ages
500: from 1 to $\sim$20 Myr and compute the average number of XRBs within
501: 1--1000 pc of the cluster center.
502:
503: Considering the typical cluster age in the Antennae is 20 Myr,
504: \citet{sep05} predict a $5\times10^4$ $M_{\sun}$ cluster should
505: contain 0.13 XRBs, while 15 XRBs should reside in a $5\times10^6$
506: $M_{\sun}$ cluster. Here we assume that an XRB is associated with a
507: cluster if it is within 100 pc. This separation is equivalent to
508: $1\farcs0$ at the distance of the Antennae (for $H_{0}$=75 km s$^{-1}$
509: Mpc$^{-1}$), which is our criteria for an XRB-cluster association
510: (Paper II). These model predictions for XRB detections assume a
511: limiting X-ray luminosity of $L_X$ = 5$\times$10$^{35}$ ergs s$^{-1}$,
512: but the observed limiting luminosity in the Antennae is
513: 2$\times$10$^{37}$ ergs s$^{-1}$. Using the X-ray luminosity function
514: (XLF) for the Antennae defined in \citet{zez02c}, we scaled the XRB
515: results of \citet{sep05} to estimate what these models would predict
516: for the observed number of XRBs in the Antennae clusters. Using a XLF
517: power law slope of $\alpha$ = -0.45 \citep{zez02c}, the models predict
518: 0.02 XRBs are observed in a $5\times10^4$ $M_{\sun}$ cluster, while
519: 2.7 XRBs should be seen in a $5\times10^6$ $M_{\sun}$ cluster, at the
520: luminosity limits of the X-ray observations. Expressing these model
521: results as a fraction of XRBs-to-cluster mass, we can directly compare
522: them to our measured value for $\eta$ in the Antennae. Doing so,
523: \citet{sep05} predict $\eta$ ranges from $4-5\times10^{-7}$
524: $M_{\sun}^{-1}$, with in a factor of five from our predictions for
525: $\eta$. As mentioned by \citet{sep05}, several caveats exist for
526: their models including: 1) assumed binary fraction of unity which
527: could lead to over estimates of the mean number of XRBs per cluster,
528: 2) the stellar, power-law IMF can affect the XRB fraction per cluster,
529: and 3) changes in the half-mass radius can strongly influence the
530: median XRB distance from the cluster. These factors could potentially
531: explain the discrepancies between their models and our observations.
532:
533: Since some forms of $\eta$ exhibit a distinctly positive slope, these
534: models may not always apply to $\eta$. More importantly, a positive
535: slope has implications for star formation scenarios in clusters. As
536: mentioned above, a positive slope implies more XRBs per unit mass are
537: produced in more massive clusters. If an abnormally large number of
538: XRBs exist in a star cluster, then the progenitors of their compact
539: objects should also posses an over abundance. Since the progenitors
540: are massive stars, this implies star formation in massive clusters
541: favors stars at the heavier end in mass. This is not unusual. Work
542: by \citet{sto05} suggest some of the largest clusters in the Milky Way
543: could have a top heavy mass function.
544:
545: In this paper we introduced the quantity, $\eta$, relating the
546: fraction of X-ray sources per unit mass as a function of cluster mass.
547: Applying this function to the Antennae, we revealed several important
548: environmental implications for the X-ray sources in the Antennae.
549: Specifically, $\eta$ predicts a far different relationship between XRB
550: formation and cluster mass than that predicted by \citet{osk05} and is
551: broadly consistent with that predicted by \citet{sep05}. Clearly,
552: future cluster modeling with particular emphasis on the relationship
553: between the number of XRBs in a galaxy and the galactic cluster
554: environment is essential to explain our current observations.
555: Furthermore, a $\chi^2$ test demonstrated the functional form of
556: $\eta$ did not always remain consistent with a single value, but for
557: some assumptions for a cluster age distributions, the slope had a
558: significantly positive value. While this could imply a top heavy mass
559: function in massive clusters, our statistics are small. We plan to
560: enlarge our statistical base by extending our observational study to
561: additional starburst galaxies. We can then address whether the
562: properties of $\eta$ depend on an individual galaxy or are consistent
563: across all galactic environments.
564:
565: \acknowledgments
566:
567: The authors thank the staff of Palomar Observatory for their excellent
568: assistance in commissioning WIRC and obtaining these data. WIRC was
569: made possible by support from the NSF (NSF-AST0328522), the Norris
570: Foundation, and Cornell University. S.S.E. and D.M.C. are supported
571: in part by an NSF CAREER award (NSF-9983830). We also thank
572: J.R. Houck for his support of the WIRC instrument project. The authors
573: are also grateful for many long and insightful discussions with
574: M.L. Edwards.
575:
576:
577: \vfill \eject
578:
579: \begin{thebibliography}{}
580:
581: \bibitem[Anders et al.(2004)]{and04} Anders, P., de Grijs, R.,
582: Fritze-v. Alvensleben, U., \& Bissantz, N. 2004, \mnras, 347, 17
583:
584: \bibitem[Belczynski et al.(2002)]{bel02} Belczynski, K., Kalogera, V., \&
585: Bulik, T. 2002, \apj, 572, 407
586:
587: \bibitem[Brandl et al.(2005)]{bra05} Brandl, B.R., et al. 2005, \apj, 635, 280
588:
589: \bibitem[Bruzual \& Charlot(2003)]{bru03} Bruzual, G. \& Charlot, S. 2003,
590: \mnras, 344, 1000
591:
592: \bibitem[Cardelli, Clayton, \& Mathis(1989)]{car89} Cardelli, J.A., Clayton,
593: G.C., \& Mathis, J.S. 1989, \apj, 345, 245
594:
595: \bibitem[Clark et al.(2005)]{cla05} Clark, D.M., et al. 2005, \apj,
596: 631, L109
597:
598: \bibitem[Clark et al.(2007)]{cla07} Clark, D.~M., et al. 2007, \apj,
599: 658, 319
600:
601: \bibitem[Fabbiano(1995)]{fab95} Fabbiano, G. 1995, in X-Ray
602: Binaries, ed. W.H.G. Lewin, J. van Paradijs, \& E.P.J. van den
603: Heuvel (Cambridge: Cambridge Univ. Press), 390
604:
605: \bibitem[Fabbiano et al.(1997)]{fab97} Fabbiano, G., Schweizer, F., \&
606: Mackie, G. 1997, \apj, 478, 542
607:
608: \bibitem[Fabbiano, Zezas, \& Murray(2001)]{fab01} Fabbiano, G., Zezas,
609: A., \& Murray, S.S. 2001, \apj, 554, 1035
610:
611: \bibitem[Fall(2006)]{fal06} Fall, S.M. 2006, \apj, in press
612:
613: \bibitem[Fall, Chandar, \& Whitmore(2005)]{fal05} Fall, S.M., Chandar,
614: R., \& Whitmore, B.C. 2005, 631, L133
615:
616: \bibitem[Gehrels(1986)]{geh86} Gehrels, N. 1986, \apj, 303, 336
617:
618: \bibitem[Harris(1996)]{har96} Harris, W.E. 1996, \aj, 112, 1487
619:
620: \bibitem[Keeping(1962)]{kee62} Keeping, E.S. 1962, Introduction To
621: Statistical Inference, (Princeton; van Nostrano), p.202
622:
623: \bibitem[Mengel(2005)]{men05} Mengel, S., Lehnert, M.D., Thatte, N.,
624: \& Genzel, R. 2005, \aa, 443, 41
625:
626: \bibitem[Oskinova(2005)]{osk05} Oskinova, L.M. 2005, \mnras, 361, 679
627:
628: \bibitem[Pooley et al.(2003)]{poo03} Pooley, D., et al. 2003, \apjl,
629: 591, L131
630:
631: \bibitem[Portegies Zwart(2004)]{por04} Portegies Zwart, S.F., Hut,
632: P., McMillan, S.L.W., \& Makino, J. 2004, \mnras, 351, 473
633:
634: \bibitem[Sepinsky, Kalogera, \& Belczynski(2005)]{sep05} Sepinsky, J.,
635: Kalogera, V., \& Belczynski, K. 2005, \apj, 621, L37
636:
637: \bibitem[Stolte et al.(2005)]{sto05} Stolte, A., Brandner, W., Grebel,
638: E.K., Lenzen, R., \& Lagrange, A. 2005, \apj, 628, L113.
639:
640: \bibitem[Wilson et al.(2003)]{wil03} Wilson, J.C., et al. 2003,
641: Proc. SPIE, 4841, 451
642:
643: \bibitem[Zezas et al.(2002a)]{zez02a} Zezas, A., Fabbiano, G., Rots, A.H., \&
644: Murray, S.S. 2002, \apj, 142, 239
645:
646: \bibitem[Zezas et al.(2002b)]{zez02b} Zezas, A., Fabbiano, G., Rots, A.H., \&
647: Murray, S.S. 2002, \apj, 577, 710
648:
649: \bibitem[Zezas \& Fabbiano (2002)]{zez02c} Zezas, A., \& Fabbiano, G. 2002,
650: \apj, 577, 726
651:
652: \bibitem[Whitmore et al.(1999)]{whi99} Whitmore, B.C., Zhang, Q.,
653: Leitherer, C., \& Fall, S.M. 1999, \aj, 118, 1551
654:
655: \end{thebibliography}
656:
657: \clearpage
658:
659: \begin{deluxetable}{lccc}
660: \tablecaption{$\Delta\Sigma\chi^2$ Statistics}
661: \tablewidth{0pt}
662: \startdata
663: \tableline
664: \tableline
665: Age Test & $\Delta\Sigma\chi^2$ & $\sigma_{\Delta\Sigma\chi^2}$ & Median $\Delta\Sigma\chi^2$ \\
666: \tableline
667: flux & 1.9 & --- & --- \\
668: 20 Myr & 2.2 & --- & --- \\
669: 100 Myr & 4.8$\times$10$^{-3}$ & --- & --- \\
670: Mengle & 1.0$\times$10$^{-2}$ & --- & --- \\
671: Instant & 0.86 & 1.7 & 0.10 \\
672: Uniform & 5.9 & 5.2 & 4.7 \\
673: Uniform\tablenotemark{1} & 5.8 & 5.3 & 4.7 \\
674: Power Law & 5.1 & 6.0 & 2.8 \\
675: Power Law\tablenotemark{1} & 4.3 & 6.6 & 2.1 \\
676: \enddata
677: \tablecomments{For instant, uniform and power law cases,
678: $\Delta\Sigma\chi^2$ statistics are mean values.}
679: \tablenotetext{1}{Multiplied $\Delta\Sigma\chi^2$ by the sign of the
680: slope for the fitted line.}
681: \end{deluxetable}
682:
683: \clearpage
684:
685: \begin{figure}
686: \figurenum{1}
687: \plotone{f1.eps}
688: \caption{Here we plot $M_{K_s}$ versus the number of clusters in each
689: bin divided by mean flux in that bin. Each bin is 0.2 mag. Arguing
690: that mass is proportional to flux, this graph shows the probability
691: of finding a cluster with a given mass. The dashed line signifies
692: the magnitude cutoff, $M_{K_s} = -13.3$ mag.
693: \label{Fig.1}}
694: \end{figure}
695:
696: \clearpage
697:
698: \begin{figure}
699: \figurenum{2}
700: \plotone{f2.eps}
701: \caption{This figure displays $\eta(F_{K_s})$ plotted versus
702: $M_{K_s}$. The bins are $F_{K_s}$ = $4\times10^6$ DN$^{-1}$ in
703: size. Error bars are the uncertainties in the mean value of $\eta$
704: added in quadrature with the Poisson uncertainty in each bin. The
705: dotted line is the mean of the four $\eta(F_{K_s})$ values.
706: \label{Fig.2}}
707: \end{figure}
708:
709: \clearpage
710:
711: \begin{figure}
712: \figurenum{3}
713: \plotone{f3.eps}
714: \caption{Assuming an instant burst of star formation in the Antennae,
715: we plot the mean value of $\eta$ for a range in ages between 1 - 100
716: Myr. Notice the factor of $\sim$10 range in $\eta$ as well as the
717: degeneracy in $\eta$ in this age range. Error bars are uncertainties
718: in $\eta$.
719: \label{Fig.3}}
720: \end{figure}
721:
722: \clearpage
723:
724: \begin{figure}
725: \figurenum{4}
726: \plotone{f4.eps}
727: \caption{Comparison between uniform and PL Monte Carlo simulations of
728: $\eta$. The peaks of each distribution vary by a factor of $\sim$2
729: in $\eta$, indicating $\eta$ does not significantly change when we
730: assume different age distributions for the Antennae. We also plot
731: the values of $\eta$ for instantaneous bursts at three different
732: ages.
733: \label{Fig.4}}
734: \end{figure}
735:
736: \clearpage
737:
738: \begin{figure}
739: \figurenum{5}
740: \plotone{f5.eps}
741: \caption{Here $\eta$ is plotted versus cluster mass in units
742: of $M_{\sun}$. In this case, we computed cluster mass using ages
743: provided by \citet{men05} (see \S3). The bins are $M_{\sun}$ =
744: 1$\times10^7$ $M_{\sun}$ in size. Error bars are the uncertainties
745: in the mean value of $\eta$ added in quadrature with the Poisson
746: uncertainty in each bin. The dotted line is the mean value of the
747: four $\eta(M_{\sun})$.
748: \label{Fig.5}}
749: \end{figure}
750:
751: \clearpage
752:
753: \begin{figure}
754: \figurenum{6}
755: \plotone{f6.eps}
756: \caption{Here we summarize how age affects $\eta$, assuming four
757: different age distributions for the Antennae clusters: instant burst
758: (a), uniform (b), power law (c) and derived ages by \citet{men05} (d).
759: Each value is the mean $\eta$ and includes 1-$\sigma$ error bars. See
760: text for details. Also included is $\eta$ for an instant starburst of
761: 20 Myr (e) and 100 Myr (f).
762: \label{Fig.6}}
763: \end{figure}
764:
765:
766: \end{document}
767:
768:
769: