1: \documentclass[12pt,preprint]{aastex}
2:
3: \newcommand{\myemail}{andrea@astro.soton.ac.uk}
4:
5: \shorttitle{Unveiling the core of the Globular Cluster M\,15 in the
6: Ultraviolet}
7: \shortauthors{Dieball et al.}
8:
9: \begin{document}
10:
11: \title{Unveiling the core of the Globular Cluster M\,15 in the
12: Ultraviolet\footnote{Based on observations with the NASA/ESA
13: {\it Hubble Space Telescope}, obtained at the Space Telescope Science
14: Institute, which is operated by the Association of Universities
15: for Research in Astronomy, Inc., under NASA contract NAS5-26555.}}
16:
17: \author{A. Dieball, C. Knigge}
18: \affil{Department of Physics and Astronomy, University Southampton,
19: SO17 1BJ, UK}
20:
21: \author{D. R. Zurek, M. M. Shara}
22: \affil{Department of Astrophysics, American Museum of Natural History,
23: New York, NY 10024}
24:
25: \author{K. S. Long}
26: \affil{Space Telescope Science Institute, Baltimore, MD 21218}
27:
28: \author{P. A. Charles}
29: \affil{South African Astronomical Observatory, PO Box 9, Observatory,
30: 7935, South Africa}
31:
32: \and
33:
34: \author{D. Hannikainen}
35: \affil{University of Helsinki, P.O. Box 14, SF-00014 Helsinki, Finland}
36:
37: \begin{abstract}
38:
39: We have obtained deep FUV and NUV images of the inner region of the
40: dense globular cluster M\,15 with the HST ACS.
41: The FUV--NUV colour-magnitude diagram shows a well-defined
42: track of horizontal branch stars, as well as a trail of blue
43: stragglers and white dwarfs. The main-sequence turnoff is clearly
44: visible at FUV $\simeq 23.5$ mag and FUV--NUV$ \simeq 3$ mag, and the
45: main-sequence stars form a prominent track that extends at least
46: 2 mag below the main-sequence turnoff. As such, this is the
47: deepest FUV--NUV colour-magnitude diagram of a globular cluster
48: presented so far. Cataclysmic variable and blue straggler candidates
49: are the most centrally concentrated stellar populations, which might
50: either be an effect of mass segregation or reflect the preferred
51: birthplace in the dense cluster core of such dynamically-formed
52: objects.
53: We find 41 FUV sources that exhibit significant variability. We
54: classify the variables based on an analysis of their UV colours and
55: variability properties. We find four previously known RR Lyrae and 13
56: further RR Lyrae candidates, one known Cepheid and six further
57: candidates, six cataclysmic variable candidates, one known and one
58: probable SX Phoenicis star, and the well known low-mass X-ray binary
59: AC\,211. Our analysis represents the first detection of SX Phoenicis
60: pulsations in the FUV. We find that Cepheids, RR Lyrae stars and SX
61: Phoenicis exhibit massive variability amplitudes in this wave band
62: (several magnitudes).
63:
64: \end{abstract}
65:
66: \keywords{globular clusters: individual(\objectname{M\,15}) -- stars:
67: close binaries -- stars: variables -- ultraviolet: stars}
68:
69: \section{Introduction}
70:
71: Globular clusters (GCs) in the Milky Way are old, gravitationally bound stellar
72: systems. Stellar densities in their cores can be extremely high,
73: reaching up to $10^{6}\ {\rm stars}/{\rm pc}^{3}$, hence making them
74: ideal crash-test laboratories for studying the dynamics in such dense
75: environments. Close encounters and even direct collisions with
76: resulting mergers between the cluster stars are inevitable, leading to
77: a variety of dynamically formed, exotic stellar populations like blue
78: stragglers (BSs), cataclysmic variables (CVs), low-mass X-ray binaries
79: (LMXBs), and probably binary horizontal branch (HB) stars and other
80: close binary (CB) systems. CBs are important for our understanding of
81: GC evolution, since the binding energy of a few CBs can rival that of
82: a GC. Thus, by transferring their orbital energy to
83: passing single stars, CBs can significantly affect the dynamical
84: evolution of the cluster (e.g.\ Elson et al.\ 1987, Hut et al.\ 1992,
85: and references therein). This depends critically on the number of CBs;
86: if there are only a few CBs, long-term interactions dominate the cluster
87: evolution, but the presence of many CBs can lead to violent
88: interactions that heat the cluster and cause its expansion and
89: evaporation on significantly shorter time-scales.
90: Therefore, knowledge of the CB populations in the cores of clusters
91: is crucial to understanding the evolution of clusters.
92:
93: However, despite their impact on cluster evolution and their
94: importance for our understanding of dynamical binary formation and
95: binary evolution, there have been only a few detections of binaries in
96: GCs during the past decades. The study of binaries has been
97: extremely difficult due to the spatial resolutions and detection
98: limits of the available telescopes. Only with the arrival of {\it Chandra}
99: and the {\it Hubble Space Telescope (HST)},
100: with their improved imaging capabilities, has it been possible to finally
101: detect large numbers of CVs and other binary systems (e.g.\ Grindlay
102: et al.\ 2001; Edmonds et al.\ 2003a, 2003b; Knigge et al.\ 2002, 2003;
103: Dieball et al.\ 2005a and references therein).
104:
105: CBs and other dynamically formed systems share an observational
106: characteristic: they show a spectral energy distribution that is bluer
107: than that of other cluster members and emit radiation in the
108: far-ultraviolet (FUV). ``Ordinary'' stars, like main-sequence (MS)
109: stars and red giants (RGs), are too cool to show up at these
110: wavelengths. As a consequence, crowding is generally not a severe
111: problem in FUV imaging, very much unlike optical imaging, which is
112: often nearly impossible in dense cluster cores. Thus, a deep,
113: high-resolution FUV-survey is the ideal tool to detect and study
114: these stellar species.
115:
116: So far, deep FUV studies have been carried out in detail only for two
117: clusters: 47\,Tuc \citep{knigge1,knigge2} and NGC\,2808 (Brown et
118: al. 2001; Dieball et al. 2005). In 47\,Tuc, Knigge et al.\ (2002)
119: found FUV counterparts to all of the {\it Chandra} CV candidates known at
120: that time in the core of this cluster (Grindlay et
121: al. 2001)\footnote{Note that recently \citet{heinke2005} published a
122: catalog of X-ray sources of 47\,Tuc comprising 300 objects within
123: one half-mass radius. Work is in progress to identify counterparts
124: to all the sources located within the FUV field of view (C. Knigge
125: 2006, private communication).} and
126: confirmed their CV nature. In addition, several additional CV
127: candidates were suggested, and a clean, well-populated FUV BS
128: sequence and numerous young white dwarfs (WDs) on the upper end of the
129: cooling sequence were found. In NGC\,2808, Brown et al.\ (2001)
130: used FUV and NUV imaging to uncover a population of sub-luminous
131: hot HB stars that probably underwent a late helium-core flash on the
132: WD cooling curve. Our own reanalysis of this data set pushed
133: the detection limits deeper and revealed numerous BSs, CV candidates,
134: and hot young WDs in this cluster \citep{dieball1}.
135:
136: M\,15 is one of the oldest (13.2 Gyr; McNamara et al.~2004)
137: low-metallicity ($\rm{[Fe/H]} = -2.26$; Harris 1996) GCs in our
138: Galaxy, at a distance of 10.3 kpc (van den Bosch et al.~2006). It has
139: received considerable attention in the literature for numerous
140: reasons. It is probably a core-collapsed cluster, with a very small
141: and very dense core ($\rm{R}_{c} \approx 0\farcm07$,
142: $\rho_{c} \approx 7 \times 10^{6} M_\odot {\rm pc}^{-3}$;
143: van den Bosch et al.~2006).
144: Because of its unusual compact core, high central velocity
145: dispersion, and luminosity profile, it has even been suggested that
146: M\,15 hosts an intermediate-mass black hole (IMBH) of up to a few
147: thousand solar masses (e.g.\ Peterson et al.~1989; Guhathakurta et
148: al.~1996; Gebhardt et al.~1997; Gerssen et al.~2002). However,
149: Baumgardt et al.\ (2003) presented $N$-body models that are capable of
150: explaining the observed velocity dispersion and luminosity profile
151: without the need for a central IMBH, but with a concentration of
152: neutron stars (NSs) and massive WDs in the centre due
153: to mass segregation. McNamara et al.\ (2003) also found that there is
154: little evidence for an IMBH, their proper-motion dispersion profile of
155: the inner region of M\,15 being in good agreement with the predictions
156: from Baumgardt et al.\ (2003). Most recently, van\,den\,Bosch et
157: al.\ (2006) suggested a mass of 3400 $M_\odot$ in the central 0.05 pc,
158: which could either be an IMBH, or a large number of compact objects,
159: or a combination of both.
160:
161: M\,15 is also the only Galactic GC known to harbour {\it two} bright
162: LMXBs. The first one, 4U\,2127+119, was identified by Auri{\`e}re et
163: al. (1984) with the optical counterpart AC\,211. It has an orbital
164: period of 17.1 hr (Ilovaisky et al.~1993) and is one of the optically
165: brightest LMXBs known. Its high $L_{\rm{opt}}/L_{\rm{X}}$ ratio
166: suggests that the system is an accretion disk corona source. White \&
167: Angelini (2001) used {\it Chandra} observations to
168: resolve 4U\,2127+119 into {\it two} X-ray sources, separated by just
169: $2\farcs7$ (see also Charles et al.\ 2002; Hannikainen et al.\
170: 2005). One of these was the previously known LMXB AC\,211. The
171: second source, CXO\,J212958.1+121002 or M\,15 X-2, is actually 2.5
172: times brighter than AC\,211 in X-rays and is the likely source of the
173: X-ray bursts observed in 1988 and 2000 (Dotani et al.~1990; Smale
174: 2001). Its optical counterpart is a blue $U \approx 19$ mag
175: star (star 590 in De\,Marchi \& Paresce 1994). Dieball et al.\ (2005b)
176: found that M\,15 X-2 is an ultra-compact X-ray binary (UCXB) with an
177: orbital period of 22.6 minutes, only the third confirmed such object in a
178: GC.
179:
180: M\,15 is one of the three original Oosterhoff type II clusters
181: identified in \citet{oosterhoff}. More than 200 variable stars are
182: known in this cluster; the majority of them are RR Lyrae stars (e.g.\
183: Silbermann \& Smith 1995; Clement et al.\ 2001; Arellano et al.\
184: 2006). \citet{clement01} list 204 entries in their catalog
185: of variable stars in M\,15, out of which 149 are RR Lyrae stars (at
186: least 55 are of type RRab, 52 of type RRc, and 23 of type RRd).
187: \citet{tuairisg} classified 19 RRc (10 new ones), 12 RRab (six
188: previously unknown) and two RRd (one previously unknown). However, the
189: classification of the RR Lyrae subtype is not always consistent with
190: \citet{clement01}.
191:
192: The cluster also hosts other interesting stellar populations, among
193: them a large number of BSs (e.g.\ Yanni et al.~1994). At least
194: three SX Phoenicis stars (Jeon et al.\ 2001; Kravtsov \& Zheleznyak
195: 2003), eight millisecond pulsars (Phinney 1996) and three dwarf nova (DN)
196: candidates are known (Shara et al.~2004; Hannikainen et al.~2005). De
197: Marchi \& Paresce (1994, 1996) also found extreme blue stars in their
198: broadband $UV$ colour-magnitude diagram (CMD), which were explained by
199: D'Antona et al.\ (1995) as the merger product of two low-mass helium
200: white dwarfs. Most of the above mentioned exotica are indicators of
201: strong dynamical interaction. M\,15's optical CMD (Van der Marel et
202: al.~2002) shows a prominent blue horizontal branch (BHB) and extreme
203: blue horizontal branch (EHB) stars with a few gaps along the HB (see
204: below our Fig.~\ref{cmd_opt}, right panel).
205:
206: Here we present FUV and NUV data of M\,15 taken with the Advanced
207: Camera for Surveys (ACS) on board the {\it HST}. In Sect.~\ref{data}
208: we describe the data and their reduction. We present the analysis of
209: the photometry and the FUV versus NUV CMD in Sect.~\ref{cmd}. Our
210: search for variability among our catalog stars and a study of the
211: variable sources are given in Sect.~\ref{variability}. Our results
212: are summarized in Sect.~\ref{summary}.
213:
214: \section{Observations and data reduction}
215: \label{data}
216:
217: M15 was observed with the ACS onboard {\it HST} for a total of 10
218: orbits in 2003 October and 2004 October to December. The observations
219: were spaced out through this period in order to improve our chances of
220: measuring time-variable phenomena. On the first visit in 2003 October,
221: eight images with individual exposures of 290 s were obtained using
222: the NUV F220W filter in the High Resolution Channel (HRC). The
223: remaining visits were used to obtain a total of 90 images using
224: the FUV F140LP filter in the Solar Blind Channel (SBC). Each of the
225: SBC visit consisted of eight 300 s exposures, followed by either a
226: 40 s or a 140 s exposure at the end of the visit. The same pointing
227: and no dither pattern was used for all of the observations. An
228: observing log can be found in Table~\ref{obslog}. The pixel sizes are
229: $0\farcs034 \times 0\farcs030$ for the SBC, resulting in a nominal
230: field of view of $35\arcsec \times 31\arcsec$, and $0\farcs028 \times
231: 0\farcs025$ pixels for the HRC, covering a nominal $29\arcsec \times
232: 26\arcsec$ field of view.
233:
234: \begin{table}
235: \begin{center}
236: \caption{\label{obslog} Observation log. Col. (1): Visit. Col. (2):
237: Data set. Col. (3): Start date of the observations. Col. (4): Total exposure
238: time for that data set. Cols. (5) and (6): Detector and Filter,
239: respectively. See text for details.}
240: \begin{tabular}{llcrcl}
241: \tableline
242: visit & dataset & start date & total exp.time & detector (ACS) & filter\\
243: \tableline\tableline
244: 1 & J8SI08010 & 2003-10-27 02:03:00 & 2320 & HRC & F220W\\\tableline
245: 2 & J8SI01011 & 2004-10-14 05:44:00 & 2400 & SBC & F140LP\\
246: 2 & J8SI01RKQ & 2004-10-14 06:29:00 & 40 & SBC & F140LP\\\tableline
247: 3 & J8SI01021 & 2004-10-14 07:15:00 & 2400 & SBC & F140LP\\
248: 3 & J8SI01S8Q & 2004-10-14 08:00:00 & 140 & SBC & F140LP\\\tableline
249: 4 & J8SI01031 & 2004-10-14 08:51:00 & 2400 & SBC & F140LP\\
250: 4 & J8SI01SIQ & 2004-10-14 09:36:00 & 140 & SBC & F140LP\\\tableline
251: 5 & J8SI01041 & 2004-10-14 10:27:00 & 2400 & SBC & F140LP\\
252: 5 & J8SI01T4Q & 2004-10-14 11:12:00 & 140 & SBC & F140LP\\\tableline
253: 6 & J8SI01051 & 2004-10-14 12:03:00 & 2400 & SBC & F140LP\\
254: 6 & J8SI01TEQ & 2004-10-14 12:48:00 & 140 & SBC & F140LP\\\tableline
255: 7 & J8SI02011 & 2004-10-24 10:27:00 & 2400 & SBC & F140LP\\
256: 7 & J8SI02LTQ & 2004-10-24 11:13:00 & 40 & SBC & F140LP\\\tableline
257: 8 & J8SI03011 & 2004-11-06 19:55:00 & 2400 & SBC & F140LP\\
258: 8 & J8SI03GCQ & 2004-11-06 20:40:00 & 40 & SBC & F140LP\\\tableline
259: 9 & J8SI04011 & 2004-11-17 11:49:00 & 2400 & SBC & F140LP\\
260: 9 & J8SI04JDQ & 2004-11-17 12:34:00 & 40 & SBC & F140LP\\\tableline
261: 10 & J8SI05011 & 2004-11-26 16:30:00 & 2400 & SBC & F140LP\\
262: 10 & J8SI05YYQ & 2004-11-26 17:16:00 & 40 & SBC & F140LP\\\tableline
263: 11 & J8SI06011 & 2004-12-05 03:36:00 & 2400 & SBC & F140LP\\
264: 11 & J8SI06BSQ & 2004-12-05 04:22:00 & 40 & SBC & F140LP\\
265: \tableline
266: \end{tabular}
267: \end{center}
268: \end{table}
269:
270: As the ACS focal surfaces are inclined towards the principal rays, the
271: resulting exposures are geometrically distorted. This distortion consists
272: of two effects, namely, a tilted elongation of the ACS apertures,
273: causing the pixel scale to be smaller along the radial direction than
274: along the tangential direction of the Optical Telescope Assembly field
275: of view, and a variation of pixel area across the detector. In order
276: to correct these effects, each pixel of the distorted original image
277: has to be mapped onto pixels in a rectified, undistorted output image,
278: taking into account shifts and rotations between images and the -
279: known - optical distortion of the camera. For a detailed discussion on
280: the distortion effects of ACS imaging and the use of {\tt multidrizzle} we
281: refer the reader to the {\it HST} ACS data handbook and references therein.
282:
283: We created master images for each filter using {\tt multidrizzle}
284: running under {\tt PyRAF}. As a first step, we geometrically corrected
285: the single flat-fielded images. To account for small shifts between the
286: desired and actual pointing positions, we used the
287: {\tt IRAF}\footnote{{\tt IRAF} (Image
288: Reduction and Analysis Facility) is distributed by the National
289: Astronomy and Optical Observatory, which is operated by AURA,
290: Inc., under cooperative agreement with the National Science
291: Foundation.}
292: tasks {\tt crossdriz} and {\tt shiftfind} on the single drizzled
293: (i.e. geometrically corrected) images. The shifts are usually less than a
294: pixel but can be as high as 1.8 pixels for one orbit in
295: SBC/F140LP. For the HRC/F220W images we found that it is sufficient to
296: use the pipeline combined and drizzled image that is created for each
297: orbit. Note that the combined and geometrically corrected ACS/SBC and
298: HRC images produced by {\tt multidrizzle} have a pixel scale of
299: $0.025\arcsec / {\rm pixel}$ and are normalized to 1 s
300: exposure time.
301: The total exposure times are 24800 s for the FUV and
302: 2320 s for the NUV master image and have to be considered when
303: calculating the photometric errors. The combined and geometrically
304: corrected FUV (SBC/F140LP) and NUV
305: (HRC/F220W) master images are shown in Fig.~\ref{image1} and
306: \ref{image2}, respectively. As expected, the FUV image is considerably less
307: crowded than the NUV image. Both images show significant
308: concentrations of sources towards the cluster core.
309:
310: \begin{figure}
311: \includegraphics[width=16cm]{f1_small.ps}
312: \caption{\label{image1} Combined and geometrically corrected
313: master image of all FUV SBC/F140LP exposures taken from the
314: central region of M\,15. North is up and east to the left. The pixel
315: scale is $0\farcs025$ pixel$^{-1}$, and the field of view is $35\arcsec
316: \times 31\arcsec$. Variables are marked with
317: circles and their number given in
318: Tables~\ref{periods1} and \ref{periods2}. The colours denote: red -
319: RR Lyrae and candidates, blue - CV candidates, green - SX Phoenicis
320: and candidates, cyan - Cepheid and candidates, magenta squares -
321: LMXBs, black - other variables. The locations of two further X-ray
322: sources, HCV\,2005-A and HCV\,2005-B, which are situated in our field of
323: view and are suspected DNe, are indicated with arrows. Note
324: that their FUV counterparts are too faint to show up in our FUV
325: master image, but the sources could be identified in the NUV. See
326: Sect.~\ref{dn} for details.}
327: \end{figure}
328:
329: \begin{figure}
330: \includegraphics[width=16cm]{f2_small.ps}
331: \caption{\label{image2} Master image of all NUV exposures
332: of the same region of M\,15. The orientation is the same as in
333: Fig.~\ref{image1}.
334: %Note that the NUV master image is displayed with the same scale as
335: %the FUV master image in Fig.~\ref{image1}, but actually the HRC field
336: %of view is just $29\arcsec \times 25\arcsec$ and thus smaller than
337: %the SBC one.
338: Both Fig.~\ref{image1} and \ref{image2} are
339: displayed on a logarithmic intensity scale in order to bring out the
340: fainter sources.}
341: \end{figure}
342:
343:
344: \subsection{Source detection}
345: \label{detection}
346:
347: As one of our main goals of this study is to construct a deep FUV--NUV
348: CMD, we took special care to identify as many of the weak sources as
349: possible. We first used {\tt daofind} (Stetson 1991) running under
350: {\tt IRAF} to create an initial list of source detections. After some
351: experimentation, we used an FWHM of 5 pixels and a rather high
352: threshold to maximize the number of faint sources detected and
353: minimize the number of multiple detections in and around bright
354: sources.
355: We caution that the FWHM used in {\tt daofind} does not reflect the ``true''
356: FWHM of the SBC point-spread-function (PSF), but is rather a
357: compromise to minimize the number
358: of false detections at the image rims and in the defect or shaded areas
359: of the MAMA detectors (due to the broken anode of the ACS MAMA; see
360: the ACS handbook for details). We overplotted the source coordinates
361: on the FUV master image and inspected them by eye. Overall, the
362: process worked well. However, we found that the automatic process left
363: some faint sources missing and others (bright sources) that were
364: multiply detected. We inserted missing faint sources by hand and
365: deleted false detections at the rims of the image, in the defect area,
366: and multiple detections around bright stars. In the end, our resulting
367: catalog of FUV sources contained 2731 entries.
368:
369: Source detection in the NUV master image worked well with standard
370: values for the threshold (Stetson 1991), and 9164 sources were
371: detected.
372: In order to match the FUV and NUV catalogs, we first transformed
373: the FUV coordinates onto the NUV frame using the {\tt geomap} and
374: {\tt geoxytran} tasks running under {\tt IRAF}. We used 30 bright stars
375: that are common in both the FUV and NUV image as a reference list
376: required by {\tt geomap} to determine the geometrical transformation
377: between the two frames.
378: Small shifts between the transformed FUV and the NUV coordinate of
379: sources common in both frames are to be expected. These shifts depend
380: on how well the applied coordinate transformation actually worked, and
381: also on the width of the stars' PSF in the images. We thus checked for
382: common pairs within a tolerance of 1.5 pixels, which resulted in 1686
383: matches. Increasing the tolerance to 2.5 pixels added another 193
384: matches. These additional matches are uniformly distributed across the
385: NUV master image with no preferred location, suggesting that our
386: coordinate transformation worked very well and no distortions are
387: visible. A total of 2137 FUV sources are located within the NUV field of
388: view. Out of these, 258 objects were not matched with an NUV
389: counterpart. Inspecting these unmatched sources by eye, we added
390: further 34 NUV sources to our initial NUV source
391: list. Photometry on these 34 additional NUV sources was then
392: performed without a Gaussian recentering as these additional sources
393: are very faint and a recentering algorithm will actually centre on the
394: next brighter star, see Sect.~\ref{photometry} and
395: \ref{variability}. Counterparts to the remaining 224 unmatched FUV
396: sources are too faint to show up in the NUV master image.
397:
398: We expect that a few false matches are among our FUV--NUV
399: pairs. The number of false matches depends on the number of objects in
400: the two source lists, the number of found pairs, and the matching
401: tolerance (here we chose 2.5 pixels). Following Knigge et al. (2002), we
402: estimate that less than $\approx 37$ pairs (1.9\% of all found matches) are
403: chance coincidences due to the source density in the NUV field of view.
404: Most of the remaining unmatched FUV sources are located
405: outside or at the rim of the NUV frame.
406:
407: We chose 2.5 pixels as our maximum tolerance radius as a compromise in
408: order to get all common pairs, but to keep the number of false matches at
409: an acceptable level. As can be seen from Table~\ref{catalogue}, the
410: actual matching radius is in most cases much smaller, with a mean
411: matching radius of just 0.5 pixels.
412:
413: A catalog listing all our FUV--NUV matches is available in the
414: electronic edition of the Journal. For reference, Table~\ref{catalogue} lists
415: only 10 entries. We also include sources without NUV
416: counterparts in our catalog because these are likely to
417: include additional WD and CV candidates.
418:
419:
420: \begin{deluxetable}{ccccccccccccccc}
421: \tabletypesize{\scriptsize}
422: \tablewidth{20cm}
423: \setlength{\tabcolsep}{0.1cm}
424: \rotate
425: \tablecolumns{15}
426: \tablecaption{Catalog of all sources in our FUV
427: field of view. The first nine columns are self explanatory. The
428: tenth column denotes the matching radius between the NUV and
429: transformed FUV Cartesian coordinates. The radius from the cluster
430: centre (at x$\approx$900 and y$\approx$850 in our FUV master
431: image, determined from star number counts in both master images) is
432: given in column eleven. The optical $B, V, U$ magnitudes and the
433: corresponding ID from the \citet{vandermarel} optical catalog are
434: given in the last four columns. An asterisk means that there is no
435: entry because there is no match with an NUV source and/or
436: with an optical source. Only entries 1 -- 5 and 1000 -- 1005 are
437: given. Table \ref{catalogue} is published in its entirety in the
438: electronic edition of the {\it Astrophysical Journal}. A portion is
439: shown here for guidance regarding its form and content. \label{catalogue}}
440: \tablehead{
441: \colhead{id} & \colhead{alpha} & \colhead{delta} & \colhead{x$_{FUV}$} & \colhead{y$_{FUV}$} & \colhead{$FUV$} & \colhead{$\Delta FUV$} & \colhead{$FUV-NUV$} & \colhead{$\Delta(FUV-NUV)$} & \colhead{tolerance} & \colhead{radius} & \colhead{$V$} & \colhead{$B$} & \colhead{$U$} & \colhead{$\rm{id}_{VDM}$}\\
442: \colhead{} & \colhead{[$^{\rm h} ~^{\rm m} ~^{\rm s}$]} & \colhead{[$^{\circ} ~\arcmin ~\arcsec$]} &\colhead{[pixel]} &\colhead{[pixel]} &\colhead{[mag]} & \colhead{[mag]} & \colhead{[mag]} & \colhead{[mag]} & \colhead{[pixel]} & \colhead{[$\arcsec$]} & \colhead{[mag]} & \colhead{[mag]} & \colhead{[mag]} &\colhead{}}
443: \startdata
444: 1 & 21 29 59.519 & 12 9 51.352 & 164.820 & 516.783 & 15.892 & 0.003 & * & * & * & 20.179 & * & * & * & *\\
445: 2 & 21 29 59.514 & 12 9 53.303 & 167.642 & 594.829 & 24.823 & 0.116 & * & * & * & 19.388 & * & * & * & *\\
446: 3 & 21 29 59.507 & 12 9 53.013 & 171.709 & 583.227 & 25.584 & 0.178 & * & * & * & 19.390 & 20.125 & 20.674 & 20.576 & 25120\\
447: 4 & 21 29 59.503 & 12 9 53.360 & 174.173 & 597.098 & 23.667 & 0.058 & * & * & * & 19.216 & * & * & * & *\\
448: 5 & 21 29 59.502 & 12 9 54.293 & 174.907 & 634.400 & 24.352 & 0.084 & * & * & * & 18.912 & 18.659 & 19.282 & 19.174 & 24915\\
449: 1000 & 21 29 58.603 & 12 9 46.293 & 702.035 & 314.413 & 23.500 & 0.049 & * & * & * & 14.275 & * & * & * & *\\
450: 1001 & 21 29 58.602 & 12 9 45.692 & 702.384 & 290.392 & 23.018 & 0.039 & * & * & * & 14.837 & 18.232 & 18.775 & 18.636 & 8759\\
451: 1002 & 21 29 58.602 & 12 9 56.222 & 702.513 & 711.564 & 23.993 & 0.072 & 3.154 & 0.080 & 0.690 & 6.029 & 19.673 & 20.072 & 20.012 & 5498\\
452: 1003 & 21 29 58.602 & 12 10 12.702 & 702.642 & 1370.79 & 17.520 & 0.005 & -0.204 & 0.007 & 0.072 & 13.923 & * & * & * & *\\
453: 1004 & 21 29 58.601 & 12 9 52.474 & 702.917 & 561.646 & 23.815 & 0.067 & * & * & * & 8.732 & * & * & * & *\\
454: 1005 & 21 29 58.601 & 12 9 56.542 & 703.298 & 724.359 & 23.182 & 0.051 & 2.973 & 0.058 & 0.350 & 5.835 & 18.920 & 19.424 & 19.271 & 5396\\
455: \enddata
456: \end{deluxetable}
457:
458: \subsection{Aperture photometry}
459: \label{photometry}
460:
461: Photometry was carried out on the combined and geometrically corrected
462: images for each filter using {\tt daophot} (Stetson 1991) running
463: under {\tt IRAF}. We used apertures of 3 pixels for both the
464: FUV and the NUV data and allowed also for a Gaussian recentering
465: of the input coordinates derived in {\tt daofind}. In the FUV,
466: aperture corrections were
467: determined via curves of growth constructed from isolated stars in our
468: master images. For the NUV data, we used the encircled energy
469: fractions published by Sirianni et al.~(2005). Because of close
470: neighbour stars in the centre of M\,15 (especially important in the
471: HRC/F220W images; see Fig.~\ref{image2}), we chose a small sky annulus
472: of 5 -- 7 pixels to minimize the number of stars within the sky annulus.
473: However, our small sky annulus also contains flux from the target
474: star. Thus, too much flux will be subtracted during sky
475: subtraction. For a few isolated stars, we were able to choose a larger
476: sky annulus of 50 -- 60 pixels. Assuming that we get the ``true'' flux
477: for the source we want to measure if we use the larger sky annulus,
478: we can then apply a flux correction for the small sky annulus to all
479: sources. These flux corrections depend on the PSF of the star, which is
480: different for the SBC/F140LP and HRC/F220W. For the sake of
481: simplicity, we assume the corrections to be the same for all stars in the
482: master image in each filter. Table \ref{corr} lists all encircled
483: energy fractions and sky corrections for apertures of 1 -- 10 pixels
484: for both the SBC/F140LP and HRC/F220W filters.
485:
486: In order to convert count rates into fluxes and STMAGs, we used
487: conversion factors of
488: \nonumber
489: \begin{eqnarray}
490: \nonumber
491: \rm{PHOTFLAM}_{SBC/F140LP} = 2.7182\times10^{-17} \rm{erg\,cm}^{2} \AA^{-1} \rm{counts}^{-1}\\
492: \nonumber
493: \rm{PHOTFLAM}_{HRC/F220W} = 7.9999\times10^{-18} \rm{erg\,cm}^{2} \AA^{-1} \rm{counts}^{-1}
494: \end{eqnarray}
495: Fluxes are then converted into STMAGs using
496: \begin{eqnarray}
497: \nonumber
498: \rm{STMAG} = -2.5 \times log_{10} (flux/(ee \times skycorr)) + 21.1,
499: \end{eqnarray}
500: where $\rm{flux} = \rm{count rate} \times \rm{PHOTFLAM}$ (i.e.\ the
501: measured flux inside our chosen aperture), ee is the encircled energy
502: fraction and skycorr is the flux correction for the small sky annulus
503: listed in Table~\ref{corr}. The conversion factors PHOTFLAM are given
504: in the image headers; see also the ACS data handbook.
505:
506: \begin{table}
507: \begin{center}
508: \caption{\label{corr} Encircled energy fractions and sky corrections
509: for apertures of 1 to 10 pixels. The encircled energy fractions for
510: the SBC/F140LP data were determined using curves of growth for
511: isolated stars in the SBC/F140LP master image; the encircled energy
512: fractions for the HRC/F220W were taken from Sirianni et al.~(2005).}
513: \begin{tabular}{cccccc}
514: \tableline
515: \multicolumn{2}{c}{aperture} & \multicolumn{2}{c}{encircled energy
516: fraction}
517: & \multicolumn{2}{c}{sky correction} \\
518: {[pixels]} & [$\arcsec$] & HRC/F220W & SBC/F140LP & HRC/F220W & SBC/F140LP \\
519: \tableline
520: 1 & 0.025 & 0.235 & 0.10 & 1.014 & 1.02 \\
521: 2 & 0.050 & 0.506 & 0.29 & 1.014 & 1.03 \\
522: 3 & 0.075 & 0.623 & 0.45 & 1.025 & 1.04 \\
523: 4 & 0.100 & 0.686 & 0.54 & 1.039 & 1.06 \\
524: 5 & 0.125 & 0.733 & 0.60 & 1.057 & 1.09 \\
525: 6 & 0.150 & 0.758 & 0.64 & 1.081 & 1.13 \\
526: 7 & 0.175 & 0.771 & 0.66 & 1.110 & 1.17 \\
527: 8 & 0.200 & 0.782 & 0.68 & 1.145 & 1.23 \\
528: 9 & 0.225 & 0.791 & 0.70 & 1.186 & 1.29 \\
529: 10 & 0.250 & 0.801 & 0.72 & 1.235 & 1.37 \\
530: \tableline
531: \end{tabular}
532: \end{center}
533: \end{table}
534:
535:
536: \section{The FUV--NUV CMD}
537: \label{cmd}
538:
539: \begin{figure}[h]
540: \centerline{
541: \includegraphics[width=16cm]{f3_small.ps}
542: }
543: \caption{\label{cmdps} FUV--NUV CMD of the core region of M\,15. For
544: orientation purposes, we include a theoretical WD cooling sequence,
545: a ZAMS, and a ZAHB (see
546: text for details). The circles and the corresponding numbers denote
547: variable FUV sources, as discussed in Sect.~\ref{variability}. The
548: location of the UCXB M\,15 X-2 is also indicated with a magenta
549: square. The brightest FUV star detected, V21, corresponds to the
550: luminous LMXB AC\,211, see Sect.~\ref{lmxb}.}
551: \end{figure}
552:
553: The FUV--NUV CMD of the core region of M\,15 is shown in
554: Fig.~\ref{cmdps}. Several stellar populations show up, such as the
555: bright HB stars, BSs, and WD candidates. A number of sources are located
556: between the WD cooling sequence and the MS. These sources include both
557: detached and interacting WD-MS binaries, and we refer to these sources
558: as ``gap objects'', see Sect.~\ref{cv}. The region of data
559: points in the lower right (i.e.\ faint and red) part of the CMD
560: denotes the MS stars and RGs. The turnoff is at
561: FUV $\approx 23.5$ mag and FUV--NUV $\approx 3$ mag at the bright
562: end of this data region. As can be seen,
563: our CMD reaches at least 2 mag below the FUV MS
564: turnoff. A trail of stars extends from the MS turnoff
565: towards red and fainter magnitudes. Sources that match with optical
566: RGs are all located on this trail. Error bars for each data
567: point are also plotted.
568:
569: For orientation purposes, we have also calculated and plotted a set of
570: theoretical tracks representing the zero-age main-sequence (ZAMS), the
571: zero-age horizontal branch (ZAHB), and the WD cooling sequence. The
572: sequences were constructed using theoretical ZAHB models from
573: \citet{dorman}, the fitting formulae of \citet{tout},
574: and the \citet{wood} grid of theoretical WD cooling curves for our WD
575: cooling sequence. We then used {\sc synphot} within {\sc STSDAS} to
576: calculate the FUV and NUV magnitudes of stars on the corresponding
577: sequence. This was achieved by interpolating on the Kurucz grid of
578: model stellar atmospheres and folding the resulting synthetic spectra
579: through the response of the appropriate filter+detector
580: combinations. For the WD cooling sequence, we used a grid of synthetic
581: DA WD spectra kindly provided by Boris G\"{a}nsicke (see G\"{a}nsicke
582: et al.\ 1995). The procedure is described in more
583: detail in \citet{dieball1}. For all our synthetic tracks we adopted a
584: distance of 10.3 kpc, a reddening of $E_{B-V}=0.1$, and a
585: cluster metallicity of $[Fe/H] = -2.26$ \citep{harris}.
586:
587: In the following, we describe the stellar populations within our CMD
588: in more detail.
589:
590: \subsection{HB stars}
591: \label{hb}
592:
593: M\,15 is one of the most metal-poor GCs in our Galaxy. Like other
594: metal-poor GCs, its optical CMD reveals a blue HB and a prominent blue
595: tail extending to fainter magnitudes (Van der Marel et al.~2002; see
596: also Fig.~\ref{cmd_opt}). This tail is the so-called EHB, which is
597: separated from the horizontal part by
598: a gap. \citet{brown} suggested that the subluminous EHB stars in
599: NGC\,2808 (as well as the EHB gap in its optical CMD) can be explained
600: by a late helium-core flash that these stars undergo while they
601: descend the WD sequence. In NGC\,2808's FUV--NUV CMD, the HB stars
602: cluster in two distinct data clumps around FUV--NUV $\approx -0.7$ mag
603: (BHB) and FUV--NUV $\approx -1.2$ mag (EHB).
604:
605: Our FUV--NUV CMD of M\,15 reveals a strong population of 133 HB
606: stars along a well-defined HB sequence and within a bright data region
607: at FUV $\approx 15.5$ mag and FUV--NUV $\approx -0.6$ mag. This is
608: different from the FUV--NUV CMD of NGC\,2808 \citep{brown, dieball1},
609: which shows two distinct clumps of HB stars, corresponding to the
610: optical EHB and BHB. However, our HB stars within the bright data clump
611: match with optical EHB stars, whereas the sources along the ZAHB match
612: with BHB stars, see Sect.~\ref{optical} and Fig.~\ref{cmd_opt}
613: below. Thus, we possibly see a similar population of EHB stars here as
614: in NGC\,2808.
615:
616: \subsection{Blue Stragglers}
617: \label{bs}
618:
619: BSs are thought to be the end product of a collision or coalescence of
620: two or more MS stars; i.e., they are dynamically formed objects and as
621: such are expected to be preferentially found in the dense cluster
622: cores. As they are more massive than ordinary cluster MS stars, we
623: expect them to be slightly evolved. In optical CMDs, they are
624: therefore located above and slightly to the red of the MS. In our
625: FUV--NUV CMD, see Fig.~\ref{cmdps}, we see a trail of stars starting
626: from the clump of MS stars and RGs in the lower right corner
627: and reaching along the ZAMS up towards the clump of EHB stars at FUV
628: $\approx 15.5$ mag. This trail of stars is below the ZAHB and less clearly
629: defined than the sequence formed by the HB stars. The location of
630: these sources roughly agrees with the expected location of BSs, i.e.\
631: along the ZAMS but brighter than the MS turnoff. We find 69 BS
632: candidates but caution that a small number of these may not be BSs
633: since the various stellar populations in our CMD partly overlap. The
634: discrimination between the gap objects and BS candidates is particularly
635: difficult and would require further observations.
636:
637: \subsection{White Dwarfs}
638: \label{wd}
639:
640: A population of about 28 sources can be seen in Fig.~\ref{cmdps} that
641: lie along or close to the theoretical WD cooling curve and are
642: therefore probably hot, young WDs. Twenty-five of our WD candidates are
643: brighter than 22.5 mag.
644:
645: Following \citet{knigge1}, we can estimate the number of expected WDs
646: by scaling from the number of HB stars in the same field of view. We
647: count $\approx 133$\ BHB and EHB stars in our CMD. Given that the
648: lifetime of stars on the HB is approximately $\tau_{HB} \simeq
649: 10^{8}$~yr (e.g. Dorman 1992), we can predict the number of WDs above
650: a given temperature on the cooling curve from the relation
651:
652: $\frac{N_{WD}}{N_{HB}} \sim \frac{\tau_{WD}}{\tau_{HB}}$
653:
654: where $\tau_{WD}$ is the WD cooling age at that temperature (e.g. Knigge
655: et al.~2002). WDs at FUV $\sim 22.5$ mag on our WD sequence have
656: a temperature of $T_{eff} = 26,000$ K and a corresponding cooling age
657: of $\tau_{WD} \simeq 3.99 \times 10^{7}$~yr. We therefore predict
658: a population of approximately 53 WDs brighter than FUV = 22.5, which
659: strongly suggests that most if not all of our candidates are indeed WDs.
660: Note that more and possibly fainter WDs are present in our FUV field
661: of view but are not detected in the NUV and thus do not appear in our
662: FUV--NUV CMD.
663:
664: We caution that this is a simplistic approach and we do not take
665: the effects of mass segregation into account. Richer et al.\ (1997)
666: suggested that if the WD cooling times are an appreciable fraction of
667: the cluster age (or rather the relaxation time), the slope of the
668: cluster initial mass function and the relation between stellar mass
669: and MS lifetime become important. This can increase the expected
670: number of faint WDs up to a factor of 3.
671:
672: \subsection{CV candidates}
673: \label{cv}
674:
675: As can be seen from our Fig.~\ref{cmd}, there are a number of sources
676: that are located between the WD cooling sequence and the ZAMS. This is
677: a region in the CMD where we expect to find CVs, but also detached
678: WD-MS binaries. As we cannot discriminate between true CVs and
679: detached WD-MS binaries based on our FUV--NUV CMD alone, we refer to
680: these sources as ``gap objects'', thus denoting sources that are
681: located in the gap region between the WD cooling sequence and the
682: MS. We find $\approx 57$ sources in this gap zone but caution again
683: that this figure is not to be taken as a strict number as the
684: discrimination especially between these gap objects and other
685: populations like WDs, MS stars and BSs is somewhat difficult. In the
686: following, we compare this number with theoretical predictions
687: for CVs. For this purpose, we scale the predicted number of
688: dynamically-formed CVs in 47\,Tuc to M\,15.
689:
690: For 47\,Tuc, \citet{distefano} predicted a population of $\sim 190$
691: active CVs formed via tidal capture. \citet{ivanova} investigated all
692: sorts of dynamical and primordial formation channels and predict a few
693: hundred CVs that should be present in the cluster core in 47\,Tuc. However,
694: \citet{shara06} found that dynamically produced CVs have shorter lives
695: than their field counterparts, which decreases the expected number of
696: CVs by a factor of 3. They based their study on $N$-body simulations
697: starting with 100,000 objects and a binary fraction of 5\% and found
698: that exchange interactions are more likely to make CVs with higher
699: mass MS donors (donor mass $> 0.7 M_{\odot}$), which are more likely
700: to be short-lived than their field counterparts.
701:
702: We follow the simplified estimate for the capture rate in GC cores
703: \citep{heinke},
704:
705: $\Gamma \propto \rho_{c}^{1.5} \cdot r_{c}^2 $,
706:
707: where $\rho_{c}$ is the central luminosity density and $r_{c}$ is the
708: core radius. Taking $\rho_{c}$, $r_{c}$, and the distance to the clusters
709: from \citet{harris}, we find that the expected number of dynamically-formed
710: CVs in M\,15 is $\approx 200$, comparable to what is expected for
711: 47\,Tuc. \citet{distefano} found that approximately half of the
712: captures would take place within 1 core radius. Given our detection
713: limits, we cannot hope to detect the very oldest and faintest CVs.
714: The coolest WDs that we detect have $T_{eff} = 26,000$ K (see
715: Sect.~\ref{wd}). This suggests that we can only expect to find
716: relatively bright, long-period CVs above the period gap (see Townsley
717: \& Bildsten, 2003, their Figs.~1 and 2). About 20 of these
718: long-period CVs should exist in 47\,Tuc (Di\,Stefano \& Rappaport,
719: 1994, their Fig.~3 and Table~5).
720: Scaling this number, we expect 21 such systems in M\,15. As our image
721: covers more than $5 \times r_{core}$, we expect to find most, but
722: perhaps not all, of these sources. As a rough estimate, we suggest that
723: 10 -- 21 of the long-period, active CVs should be within our field of view.
724: If dynamical effects shorten the lifetime of CVs by a factor of about
725: 3, as suggested by \citet{shara06}, the expected number of CVs in our
726: data would drop accordingly. We caution that these are extremely rough
727: estimates, but they suggest that some, but not all, of our gap objects
728: are likely to be CVs.
729:
730: \subsection{Optical counterparts}
731: \label{optical}
732:
733: An optical catalog of 31,983 sources in the central region of M\,15
734: was presented by \citet{vandermarel}. These data were obtained in
735: 1994 April in the context of {\it HST} program GO-5324, using the
736: WFPC2/F336W, F439W, and F555W camera/filter combination, with the PC
737: centred on the cluster centre. The chosen WFPC2 filters correspond
738: roughly to the Johnson $U$, $B$, and $V$ bands. First results from these
739: observations were reported by \citet{guhathakurta}. \citet{vandermarel}
740: applied an improved astrometric and photometric calibration to the
741: original star list, and we used this catalog to find optical
742: counterparts for our FUV sources.
743:
744: We first transformed the $\alpha, \delta$ coordinates in the optical
745: catalog to the Cartesian ($x$, $y$) FUV coordinate system, using the
746: {\tt IRAF} task {\tt wcsctran}. Note that since the optical $\alpha,
747: \delta$ coordinates do not correspond exactly to the FUV ones, the
748: Cartesian coordinates also do not match exactly and require a further
749: transformation. Thus, we selected 80 HB stars that are common to
750: both catalogs and could be easily identified. These 80 sources
751: served as our reference list for a coordinate transformation. Using
752: the {\tt IRAF} tasks {\tt geomap} and {\tt geoxytran}, we transformed the
753: physical FUV coordinate system to the optical ($x$, $y$)
754: one.\footnote{We found a rather large shift between our FUV and the
755: optical coordinates given in the electronic version of the optical
756: catalog available via CDS, Strasbourg. For example, the
757: coordinates for the well-known LMXB AC\,211 (optical id 5435) are
758: $21^{h} 29^{m} 57.91^{s} +12^{\circ} 10\arcmin 06\farcs8$ in the
759: VizieR version of the optical catalog. In our catalog presented in
760: Table~\ref{catalogue}, AC\,211 (FUV ID 1580) is at $21^{h} 29^{m}
761: 58.238^{s} +12^{\circ} 10\arcmin 1\farcs65$, which is much closer to
762: previously published figures \citep{wa, kulkarni}. However, the
763: original electronic table in \citet{vandermarel} listed only $\Delta
764: \alpha$ and $\Delta \delta$ with respect to cluster centre and no
765: absolute $\alpha, \delta$. Applying $\Delta \alpha$ and $\Delta
766: \delta$ to the cluster centre coordinates given in
767: \citet{vandermarel} results in $21^{h} 29^{m} 57.965^{s}
768: +12^{\circ} 10\arcmin 02\farcs85$ for AC\,211, closer to previous
769: published values and different from the coordinates computed by
770: VizieR.}
771: Allowing for a matching radius of 1.5 pixel, we then found 1632 sources
772: that are common to both the optical and our FUV catalog.
773:
774: We checked the location of the common sources in both the FUV--NUV
775: and the optical CMD and found that indeed all stellar populations are
776: at their expected location in the optical CMD, i.e.\ our matched HB
777: stars are on the optical blue or extended blue horizontal branch (see
778: Sect.~\ref{hb}), and the matched sources within our BS region
779: are in the BS region in the optical CMD, and so on. This is
780: illustrated in Fig.~\ref{cmd_opt}, where we indicate the different
781: stellar populations found in our FUV--NUV CMD with different colours
782: and overplotted them on the optical CMD. A few BSs are within the
783: optical MS region, which indicates that these BSs are
784: formed from the coalescence of lower mass MS stars and thus are
785: of a lower
786: mass than the BS above the optical MS turnoff. All matched sources
787: that are located in the MS/RG clump in our FUV--NUV CMD correspond to
788: optical MS stars or RGs. We did not find any
789: counterparts to our WD candidates, which is to be expected as WDs
790: are extremely faint in the optical. We found 12 matches among
791: our gap objects. Ten of these are located along the optical
792: MS. These systems might be CVs or detached WD-MS binaries that are
793: dominated by the MS star in the optical, whereas the systems are
794: dominated by the accretion disk or a hot WD in the FUV. Note that
795: the gap object with the bluest optical counterpart is actually the
796: UCXB M\,15 X-2.
797:
798: \begin{figure}
799: \centerline{
800: \includegraphics[width=16cm]{f4_small.ps}
801: }
802: \caption{\label{cmd_opt} Comparison between the various stellar
803: populations found in our FUV--NUV CMD (left) and the optical CMD
804: (right; based on the \citet{vandermarel} catalog). The stellar
805: populations are at their expected location in the optical CMD;
806: i.e., our BS candidates are mainly above the MS turnoff in the
807: optical CMD, etc. (see text for details). Gap objects (which
808: include CV candidates) are plotted in green, BS candidates in blue,
809: MS stars in yellow, RGs in red, EHB stars in magenta and BHB stars
810: in cyan. Note that we did not find any optical counterparts to our
811: WD candidates. The counterpart to our variable candidate V6 is
812: indicated in the optical CMD (right).}
813: \end{figure}
814:
815: \subsection{Radial distributions of the stellar populations}
816: \label{cumulativetxt}
817:
818: \begin{figure}[ht!]
819: \centerline{
820: \includegraphics[width=15cm]{f5.eps}
821: }
822: \caption{\label{cumulative1} Cumulative radial distributions of the
823: various stellar populations that show up in our FUV--NUV
824: CMD. Top: Only sources with NUV $\le 21$ mag are
825: considered. We only compare BS candidates (dotted line), gap
826: objects (dashed line), and HB stars (solid line), as
827: only two WD candidates are present that are brighter than 21 mag in the
828: NUV. Bottom: Radial distribution of BHB stars (dotted
829: line) and EHB stars (dashed line). EHB stars seem to be
830: slightly more concentrated in the core. See text for details.}
831: \end{figure}
832:
833: \begin{figure}[ht!]
834: \centerline{
835: \includegraphics[width=15cm]{f6.eps}
836: }
837: \caption{\label{cumulative2} Cumulative radial distributions of EHB
838: and BHB stars in the inner $5\arcsec$ (top) and outside the
839: core region (bottom).}
840: \end{figure}
841:
842: Fig.~\ref{cumulative1} shows the cumulative radial distribution of the
843: stellar populations that show up in our CMD and that are discussed
844: above. As our CMD is limited by our NUV data, we only consider data
845: that are brighter than NUV = 21 mag. We note that the radial
846: distribution of MS stars is not shown because the difference between
847: the MS stars and all other populations might be an effect of
848: incompleteness in the cluster core in the FUV image. A few bright
849: stars are concentrated in the cluster centre, and the broad PSF wings
850: of the bright sources hamper the detection of faint objects,
851: especially MS stars and WDs. As only two WD candidates with NUV < 21
852: mag are present in our limited data set, we show only the radial
853: distributions of BS candidates, gap objects, and HB stars, see the top
854: panel in Fig.~\ref{cumulative1}.
855: All of the gap sources are CV candidates to start with, but we caution
856: again that this sample might also contain detached WD-MS binary
857: candidates. The gap objects seem to be the most centrally concentrated
858: stellar population, followed by BS candidates and then HB stars.
859:
860: We applied a Kolmogorov-Smirnov (K-S) test on the restricted
861: data set in order to determine the statistical significance of the
862: differences between the various stellar populations. However, we
863: caution that HB stars, gap sources, and BS candidates cannot be clearly
864: distinguished based on the CMD alone, and we might have sources
865: in one sample that actually belong to another. Our selected
866: data set contains 133 HB stars, 69 BS candidates, and 16 gap sources.
867:
868: The K-S test returns the probability that the maximum difference between
869: the two distributions being compared should be as large as observed,
870: under the null hypothesis that the distributions are drawn from the
871: same parent populations. In this sense, a high K-S probability agrees
872: with both distributions to be likely from the same parent
873: distribution. The smaller the returned K-S probability, the more
874: unlikely it is that they originate from the same parent distribution,
875: i.e.\ the more different are the two compared distributions.
876: BS candidates and gap sources differ from the HB population
877: with K-S test probabilities of 14\% (BSs vs. HBs) and 2\% (gap sources
878: vs. HBs). Gap objects and BS candidates are different with K-S
879: probability of 8\%.
880:
881: As shown in Fig.~\ref{cumulative1}, gap sources and BS candidates are
882: the most centrally concentrated populations.
883: There are two possible physical explanations for this: (1) Since CVs
884: and BSs are expected to be more massive than MS and HB stars, the
885: differences in radial distribution could be an indication of mass
886: segregation; i.e., the heavier objects sink towards the cluster core
887: and end up more centrally concentrated than the lighter MS and HB
888: stars (and WDs). (2) Alternatively, since CVs and BSs in GCs can form
889: through two- or three-body interactions that take
890: place preferentially in the denser cluster core, they may be
891: overabundant there because this is their birthplace.
892:
893: In any case, the enhanced central concentration of the gap sources
894: provides additional evidence that most of these sources are indeed CVs
895: or non-interacting WD/MS binaries (as opposed to chance
896: superpositions, foreground stars, etc.), i.e.\ they either segregated
897: towards the cluster centre or formed in the dense core region.
898:
899: The bottom panel of Fig.~\ref{cumulative1} shows the radial
900: distribution of only HB stars. We compare EHB and BHB stars, and it
901: seems that EHB stars are slightly more concentrated in the inner $5
902: \arcsec$, i.e.\ the cluster core.
903: As a next step, we compare the EHB and BHB stars only in the cluster
904: core (Fig.~\ref{cumulative2}, top panel) and outside the core region
905: (Fig.~\ref{cumulative2}, bottom panel). We found that the two HB
906: populations in the inner $5\arcsec$ are different with K-S probabilities of
907: 7.5\%, whereas the EHB and BHB populations in the outskirts (i.e.\
908: outside the cluster core) of M\,15 appear not to be different, with a K-S
909: test probability of 93.2\%. This might indicate that EHB stars are
910: indeed more centrally concentrated than BHB stars. One possible
911: explanation could be that EHB stars have a dynamical origin
912: \citep{fusipecci} and could have formed from mergers of WD binary
913: systems \citep{iben}.
914:
915: \section{Variable candidates}
916: \label{variability}
917:
918: \subsection{Detection of Variability}
919: \label{vardetect}
920:
921: For our variability study only relative magnitudes are needed. For
922: this purpose, we performed photometry on the individual drizzled,
923: i.e. geometrically corrected, images, allowing also for a Gaussian
924: recentering of the input coordinates. We only used exposures
925: taken in SBC/F140LP, since this data set provides the longest time
926: coverage (from 2004 October 14 to December 5). In total 90 exposures
927: were used for this analysis. Eighty of these had exposure times of 300
928: s, six were exposed for 40 s and four had exposure times of 140
929: s (see Table~\ref{obslog}).
930:
931: For each star in our input catalog of FUV sources, its mean magnitude
932: and the corresponding standard derivation were calculated from all
933: single magnitudes. Fig.~\ref{varsigma} shows a plot of all mean
934: magnitudes versus their corresponding $\sigma_{mean}$ for all stars.
935: Based on this plot, we selected 53 sources that show a
936: $\sigma_{mean}$ exceedingly higher than their companions with similar
937: brightness. These initial variable candidates are marked with plus signs
938: in Fig.~\ref{varsigma}. As a next step, we checked each of our initial
939: variable candidates by eye on all 90 input images. We found that in
940: a few cases, the recentering algorithm
941: actually centred on the nearest neighbour instead of the source
942: in question. This happened especially for our variable candidate
943: 32, which for a few orbits became extremely faint. In these cases,
944: we forced the photometry on the initial input coordinates without
945: recentering. We caution that for accurate photometry a recentering is
946: usually better, as small shifts between the input images might occur
947: (see Sect.~\ref{detection}).
948: However, in the case mentioned above, the recentering fails to work
949: and photometry forced on the initial input coordinates is then the
950: only option available, but it should be considered with some care.
951:
952: A few initial variable candidates were located in the halos of close
953: bright stars. These stars either are only detectable in the combined
954: master image, but are too faint to show up in the single input images
955: (which can have exposure times as low as 40 s), or they might just
956: be false detections in the halos of bright stars. For our further
957: analysis, these questionable candidates are not considered. Most of
958: these doubtful objects have no NUV counterpart (except for source
959: 2388), which again indicates that they are either very blue and faint
960: sources or indeed false detections. Two variable candidates
961: turned out not to be true sources at all but bad pixels and were
962: consequently deleted from our list of FUV sources in
963: Table~\ref{catalogue}. We checked that there are no further known bad
964: pixels that coincided with suspected sources. The remaining 41
965: variable candidates are marked with diamonds
966: in Fig.~\ref{varsigma} and are discussed in detail in the
967: following. Note that six of our good variable candidates (V1, V2,
968: V3, V4, V6, and V35) are located outside the NUV field of view. V15
969: and V40 are inside the common field of view but do not have an NUV
970: counterpart within our chosen tolerance radius of 2.5
971: pixels. Consequently, these eight variable candidates do not appear in the
972: CMD in Fig.~\ref{cmdps}. Our notation ``Vxx'' refers to the variable
973: candidates that we discuss in the present paper. We caution that
974: \citet{clement01} use the same nomenclature but for different stars.
975:
976: Note that the UCXB M\,15 X-2 does not show a noticeable
977: $\sigma_{mean}$. This source has an orbital period of $22.5806 \pm
978: 0.0002$ minutes, and the signal is clearly significant, see
979: \citet{dieball2}. However, the semi-amplitude of its modulation is
980: only $0.062 \pm 0.004$ mag, which is below our selection criterion in
981: the context of this paper. All our variable candidates have
982: $\sigma_{mean} > 0.2$ mag, and M\,15 X-2 would not have been detected
983: as a variable based on Fig.~\ref{varsigma}, in contrast to the LMXB
984: AC\,211, which corresponds to our V21.
985: This shows that we detect stars that show strong variability, but we
986: do not claim to be complete and detect ``all'' variable stars in the
987: core of M\,15. For this, a more thorough variability analysis would be
988: needed, which is beyond the scope of the present paper.
989: We marked the position of the UCXB with a star and indicate it with an
990: arrow in Fig.~\ref{varsigma}.
991:
992: In order to check for systematic trends in our data, we analyzed the
993: light curves of the 10 brightest, non-variable stars in our
994: sample. A slight trend of somewhat fainter magnitudes towards the end
995: of each orbit is visible in all of them. The resulting periodograms
996: all peak at a period that is consistent with the {\it HST's} orbital
997: period.
998:
999: \begin{figure}
1000: \centerline{
1001: \includegraphics[width=15cm]{f7_small.ps}
1002: }
1003: \caption{\label{varsigma} Mean FUV magnitudes vs.\
1004: $\sigma_{mean}$ derived from the individual photometry of each
1005: star. The 53 initial variable candidates are marked with plus signs.
1006: The 41 variable candidates that remain after closer inspection on
1007: the individual input images are additionally marked with
1008: diamonds. The position of M\,15 X-2 is marked with a star and
1009: indicated with an arrow. The gap objects/CV candidates
1010: that we selected based on our FUV--NUV CMD are marked with red filled
1011: circles. The two additional objects (V15 and V40) that do not have an
1012: NUV counterpart but are likely CVs because of their variability are
1013: marked with green filled circles. See text for details.}
1014: \end{figure}
1015:
1016: Fig.~\ref{lcurve1} to Fig.~\ref{lcurve3} show the mean-subtracted
1017: light curves over all observing epochs for all remaining good variable
1018: candidates. We computed $\chi^{2}$ periodograms for all data
1019: sets in order to determine periods for all variable candidates. However,
1020: even if present, a periodicity can only be determined if the
1021: observations sample that period sufficiently well. This is the case
1022: for many but not all variables. A sine wave representing the
1023: corresponding period is then overplotted on their light curve. The
1024: folded light curves are shown in Fig.~\ref{fold}. Tables~\ref{periods1}
1025: and \ref{periods2} give an overview over all variable candidates,
1026: their periods, and their classification based on the following discussions.
1027:
1028: \begin{figure}
1029: \centerline{
1030: \includegraphics[width=17cm]{f8_small.ps}
1031: }
1032: \caption{\label{lcurve1} Light curves for our variable sources V1
1033: -- V15. The mean subtracted instrumental magnitudes are plotted
1034: vs.\ time.}
1035: \end{figure}
1036:
1037: \begin{figure}
1038: \centerline{
1039: \includegraphics[width=17cm]{f9_small.ps}
1040: }
1041: \caption{\label{lcurve2} Same as Fig.~\ref{lcurve1}, but for the
1042: variable sources V16 -- V30.}
1043: \end{figure}
1044:
1045: \begin{figure}
1046: \centerline{
1047: \includegraphics[width=17cm]{f10_small.ps}
1048: }
1049: \caption{\label{lcurve3} Same as Fig.~\ref{lcurve1}, but for the
1050: variable sources V31 -- V41.}
1051: \end{figure}
1052:
1053: \begin{figure}
1054: \centerline{
1055: \includegraphics[width=16cm]{f11_small.ps}
1056: }
1057: \caption{\label{fold} Folded light curves for the variables
1058: for which we derived periods. See also Table~\ref{periods1} and
1059: \ref{periods2}.}
1060: \end{figure}
1061:
1062: \begin{figure}
1063: \centerline{
1064: \includegraphics[width=16cm]{f12_small.ps}
1065: }
1066: \caption{\label{foldrr} Folded light curves for the RR Lyrae V8, V9, and
1067: V14. The left panels shows the light curves folded with the periods
1068: derived in this paper, while the right panels shows the light curves folded
1069: with a period derived by \citet{tuairisg}. See text for details.}
1070: \end{figure}
1071:
1072: \begin{table}
1073: \begin{center}
1074: \caption{\label{periods1} Known RR Lyrae stars and RR Lyrae
1075: candidates. Reference for the known RR Lyrae is \citet{tuairisg}.}
1076: \begin{small}
1077: \begin{tabular}{lccll}
1078: \\
1079: \tableline
1080: id & $<FUV>$ & $<FUV-NUV>$ & period & other name\\
1081: & [mag] & [mag] & [d] & \\
1082: \hline
1083: V2 & 18.845 & -- & $0.3674$ & BSR\,1856, V144?\\
1084: V3 & 18.795 & -- & $0.2768$ & BSR\,1856, V144?\\
1085: V4 & 18.619 & -- & $0.2995$ & BSR\,1856, V144 (most likely counterpart)\\
1086: V8 & 18.999 & 1.300 & $0.3644$ & V165?\\
1087: V9 & 19.485 & 1.917 & $0.3492$ & V128\\
1088: V12 & 20.790 & 2.700 & $0.5350$ & \\
1089: V14 & 18.526 & 1.290 & $0.2997$ & V137?\\
1090: V16 & 18.722 & 1.643 & $0.3083$ & \\
1091: V17 & 18.485 & 1.452 & $0.2811$ & \\
1092: V19 & 20.485 & 3.483 & $0.5408$ & \\
1093: V23 & 18.801 & 1.768 & $0.6161$ & \\
1094: V27 & 18.849 & 1.314 & $0.3531$ & V160 \\
1095: V28 & 19.189 & 1.845 & $0.3615$ & V135 \\
1096: V29 & 19.512 & 2.379 & $0.2483$ & \\
1097: V32 & 18.525 & 0.493 & $0.2110$ & \\
1098: V34 & 18.972 & 1.854 & $0.3973$ & \\
1099: V35 & 19.398 & -- & $0.4127$ & \\
1100: \tableline
1101: \end{tabular}
1102: \end{small}
1103: \end{center}
1104: \end{table}
1105:
1106: \begin{table}
1107: \begin{center}
1108: \caption{\label{periods2} Classification of all other variables.
1109: Periods are in days unless otherwise stated. References
1110: for the known variables are: (a) \citet{tuairisg}, (b) Ilovaisky et
1111: al.~(1993), (c) \citet{kravtsov}, (d) \citet{dieball2}.}
1112: \begin{small}
1113: \begin{tabular}{lcclll}
1114: \\
1115: \tableline
1116: id & $<FUV>$ & $<FUV-NUV>$ & period & type & other name\\
1117: & [mag] & [mag] & [d] & & \\
1118: \hline
1119: V1 & 20.967 & -- & $2.6$ & CEP candidate & \\
1120: V5 & 20.573 & 2.301 & -- & ? & \\
1121: V6 & 19.985 & -- & -- & ? & \\
1122: V7 & 21.604 & -1.238 & -- & CV candidate & \\
1123: V10 & 19.157 & 2.389 & $1.6$ & CEP & V16 (a)\\
1124: V11 & 22.277 & -0.313 & -- & CV candidate & \\
1125: V13 & 20.739 & 2.656 & $3.4$ & CEP candidate & \\
1126: V15 & 21.364 & -- & -- & CV candidate & \\
1127: V18 & 20.347 & 3.530 & $1.4$ & CEP candidate & \\
1128: V20 & 18.043 & 0.641 & $1.5$ & ? (BHB zone) & \\
1129: V21 & 13.881 & -0.598 & $17.1$ [hrs] (b) & LMXB & AC211 (b)\\
1130: V22 & 19.599 & 0.968 & $1.06$ [hrs] & SXPHE candidate & \\
1131: V24 & 21.030 & 3.209 & -- & ? (BHB zone) & \\
1132: V25 & 21.153 & 3.613 & -- & ? (BHB zone) & \\
1133: V26 & 21.492 & 1.919 & -- & ? (BS zone) & \\
1134: V30 & 19.635 & 1.938 & $3.8$ & CEP candidate & \\
1135: V31 & 20.328 & 2.218 & $3.8$ & CEP candidate & \\
1136: V33 & 19.954 & 0.670 & $1.32$ [hrs] & SXPHE & VGG 10041 (c)\\
1137: V36 & 20.285 & 3.137 & $4.5$ & CEP candidate & \\
1138: V37 & 20.651 & 1.342 & -- & ? (BS zone) & \\
1139: V38 & 20.300 & 1.689 & -- & ? (BS zone) & \\
1140: V39 & 19.962 & 0.333 & -- & CV candidate & \\
1141: V40 & 21.692 & -- & -- & CV candidate & \\
1142: V41 & 17.676 & -0.469 & -- & CV candidate & \\
1143: -- & 16.964 & -1.062 & $22.6$ [min] (d) & UCXB & M\,15 X-2 (d)\\
1144: \tableline
1145: \end{tabular}
1146: \end{small}
1147: \end{center}
1148: \end{table}
1149:
1150: \subsection{RR Lyrae stars with possible counterparts: V2, V3, V4, V8,
1151: V9, V14, V27, V28}
1152: \label{rrlyrae}
1153:
1154: RR Lyrae stars are radially pulsating giant stars of spectral class A
1155: -- F that change their brightness with periods of 0.2 to 1.2 days and
1156: amplitudes from 0.5 -- 2 mag in $V$ \citep{GCVS}\footnote{The GCVS is
1157: also available at: http://www.sai.msu.su/groups/cluster/gcvs/gcvs/}.
1158: RR Lyrae are divided into (at least) two subgroups:
1159: RRab show asymmetric light curves and pulsate in the fundamental
1160: tone. They show periods from 0.3 to 1.2 days and amplitudes from 0.5
1161: to 2 mag in $V$. RRc show first-overtone pulsations, their light curves
1162: are roughly sinusoidal with periods of 0.2 -- 0.5 days and their
1163: amplitudes are not greater than 0.8 mag in $V$.
1164:
1165: The light curve of our variables V2, V3 and V4 show long-term
1166: variability, with periods of 0.3674, 0.2768 days, and
1167: 0.2995 days, respectively. Periods and amplitudes agree
1168: with these variables being RR Lyrae stars. All of these three sources
1169: match within $2\arcsec$ with the known RR Lyrae V144 in
1170: \citet{clement01}, or BSR\,1856 in \citet{tuairisg}, and constitute
1171: possible counterparts to this object. V2's coordinates match within
1172: $0\farcs5$ with the known RR Lyrae and are thus the closest positional
1173: match, whereas our V3 and V4 both match within $1\farcs8$ with BSR
1174: 1856. \citet{tuairisg} derived a period of 0.2990 days for BSR\,1856,
1175: which agrees very well with the period of our V4. It is likely that
1176: V2 and V3 maybe are newly discovered RR Lyrae whereas V4 is the true
1177: counterpart to BSR 1857.
1178:
1179: The coordinates of V8 match within $0\farcs9$ with V165 in
1180: \citet{tuairisg}, for which they derived a period of 0.4339 days. We
1181: derived a much smaller period of 0.3644 days, which fits our
1182: observed light curve better than a period of 0.4339 days. As a further
1183: test, we folded the light curve with our period
1184: and a period of 0.4339 days, see Fig.~\ref{foldrr}. As can be seen, the
1185: resulting plot with the longer period does not result in a smooth
1186: curve. Although the coordinates agree well enough, the different
1187: periods suggest that our V8 is probably not the same source as V165 in
1188: \citet{tuairisg}.
1189:
1190: V9 matches within $1\farcs1$ with V128 in \citet{clement01}. However,
1191: our period is 0.3492 days and thus somewhat shorter than the one
1192: \citet{tuairisg} derived for this RR Lyrae, namely, 0.4034 days.
1193: Folding the observed light curve with our period and folding with the longer
1194: period of 0.4034 days (see Fig.~\ref{foldrr}) both give reasonably
1195: smooth curves. Thus, we cannot exclude that the period derived by
1196: \citet{tuairisg} might be the correct one for this RR Lyrae.
1197:
1198: V14 and the previously known RR Lyrae V137 in Clement et al.\ (2001)
1199: match within $1\farcs8$, but the periods do not agree with each
1200: other. We derived a period of 0.2997 days, which is smaller than the
1201: 0.3520 day period \citet{tuairisg} derived. The fit to the observed
1202: light curve is better with our shorter period than with a 0.3520 day
1203: period. Also, the folded light curves suggest that 0.3520 days is not
1204: the correct period for this variable star.
1205:
1206: V27 shows a long-term variability with a period of 0.3531 days, which
1207: agrees very well with the known RR Lyrae V160 in Clement et al.\
1208: (2001), for which \citet{tuairisg} derived a period of 0.3529 days. The
1209: coordinates of V27 and Tuairisg's V160 match within
1210: $2\arcsec$. NUV information is available for V27, and its location
1211: in our FUV--NUV CMD is above the ZAMS and close to the ZAHB. This
1212: agrees well with the expected location of RR Lyrae stars.
1213:
1214: V28 is within $2\arcsec$ from the known RR Lyrae V135 (Clement et al.,
1215: 2001). We derived a period of 0.3615 days for this variable star,
1216: which agrees very well with the 0.3619 day period obtained by
1217: \citet{tuairisg}.
1218:
1219: \subsection{Newly detected RR Lyrae: V12, V16, V17, V19, V23, V29,
1220: V32, V34, V35}
1221: \label{rrlyraecand}
1222:
1223: The coordinates of our RR Lyrae candidates V12, V16, V17, V19, V23, V29,
1224: V32, V34 and V35 do not match those of any known RR Lyrae stars within
1225: $2\arcsec$. All of these sources have FUV$_{STMAG} =$ 18.5 -- 21.3
1226: mag and brightness variations of about 2 -- 4 mag on a timescale of 0.2
1227: -- 0.6 days (see Table~\ref{periods1}). Except for V35, NUV
1228: information is available for all of them, and V12 and V32 are located
1229: between the BS and ZAHB sequence in our CMD in Fig.~\ref{cmdps},
1230: whereas V16, V17, V23, V29, and V34 lie more or less exactly on the
1231: ZAHB track. V19 is somewhat fainter and redder than most of the other
1232: RR Lyrae candidates and is located above the faint end of the
1233: ZAHB. The positions of all these variable sources agree with the
1234: expected location of RR Lyrae stars in the FUV--NUV CMD.
1235:
1236: All of the above-mentioned variables can be called RR Lyrae with good
1237: confidence, as their periods, amplitudes, and location in
1238: the CMD agree with what we expect for this type of variable. However,
1239: V12 shows a larger scatter in its light curve than the other RR Lyrae,
1240: and its location in the CMD is towards fainter FUV magnitudes. We
1241: call this variable a suspected RR Lyrae.
1242:
1243: The reason why these RR Lyrae were not found in previous studies could
1244: be that optical surveys suffer from severe crowding that increases
1245: towards the cluster core. Also, the FUV amplitudes for RR Lyrae are
1246: more significant than the optical ones, making them easy to
1247: detect. However, we point out that our variability census is not
1248: complete as we only consider bright variables with large FUV
1249: amplitudes. Variables that either are FUV fainter than 22 mag
1250: or have smaller amplitudes, see Fig.~\ref{varsigma}, are not further
1251: discussed in the present paper.
1252:
1253: \subsection{Previously Suspected Dwarf Novae: HCV2005-A (CV\,1) and HCV2005-B}
1254: \label{dn}
1255:
1256: DNe are CVs that show multiple outbursts ranging in
1257: (optical) brightness from 2 to 5 mag. The outburst intervals
1258: are quasi-periodic but can range from days to decades. The duration of
1259: an outburst is typically days to weeks. These outbursts are thought to
1260: be due to the release of gravitational energy, which results from an
1261: instability in the accretion disk.
1262:
1263: In addition to the two LMXBs AC\,211 and M\,15 X-2,
1264: \citet{hannikainen} found four faint X-ray sources within the inner
1265: $50 \arcsec$ of M\,15. Two of these sources are within our field of
1266: view, namely, HCV\,2005-A and HCV\,2005-B (see Table~\ref{dntab}).
1267:
1268: \citet{hannikainen} suggested that HCV\,2005-A is a probable DN. This
1269: source is the same as the DN CV\,1 in \citet{shara}. This object does not
1270: appear in our initial list of FUV sources and was not noticed when
1271: we inspected the FUV master image by eye. The reason for this is that
1272: it is very faint and so close to the nearest bright star that the
1273: two blend into each other and could not be discriminated. As
1274: \citet{shara} pointed out, the DN is in quiescence during their FUV
1275: observations and too faint to be detected. \citet{shara} used the same
1276: data set as we do but could only use a part of the FUV epochs as the
1277: remaining observations were not yet available at that time. In fact,
1278: CV\,1 is in outburst in our fourth observing epoch, and so we were able
1279: to identify this DN (see Fig.~\ref{dn_cv1}). This source appears to be
1280: undetectable in the WFPC2 1994 April data, on which the \citet{vandermarel}
1281: catalog is based, whereas the WFPC2 data taken in 1994 October show
1282: a bright source (see also Hannikainen et al.\ 2005).
1283:
1284: The second X-ray source HCV\,2005-B is a faint NUV = 21.36 mag object
1285: that appears only in our NUV images and is too faint to be detected
1286: in our FUV data in any epoch (Fig.~\ref{dn_hcv}).
1287: \citet{hannikainen} suggested that this source is most likely a DN
1288: but might also be a quiescent soft X-ray binary. Unfortunately, we
1289: cannot shed more light on the true nature of this object.
1290:
1291: \begin{figure}
1292: \centerline{
1293: \includegraphics[width=5cm]{f13a.eps}
1294: \includegraphics[width=5cm]{f13b.eps}
1295: \includegraphics[width=5cm]{f13c.eps}
1296: }
1297: \caption{\label{dn_cv1} Close-up of the DN CV\,1 or HCV\,2005-A that
1298: was in outburst only in our fourth FUV observing epoch (left) but
1299: was in quiescence during all other FUV observing periods
1300: (middle). The right panel shows a close-up in the NUV. Note
1301: that we used the same orientation for these finding
1302: charts as \citet{hannikainen} and \citet{charles}; i.e., north is to
1303: the upper left corner and east to the lower left. CV\,1 is marked with
1304: a circle and is located at $\alpha = 21^{h} 29^{m} 58.275^{s}$,
1305: $\delta = 12^{\circ} 10\arcmin 00\farcs1$. The field of view of these
1306: close-ups is $\approx 1\arcsec \times 0\farcs7$.}
1307: \end{figure}
1308:
1309: \begin{figure}
1310: \centerline{
1311: \includegraphics[height=5cm]{f14a_small.ps}
1312: \includegraphics[height=5cm]{f14b.eps}
1313: }
1314: \caption{\label{dn_hcv} Same as Fig.~\ref{dn_cv1}, but for the DN
1315: HCV\,2005-B. North is to the top left and east to the bottom
1316: left. HCV\,2005-B appears as a faint NUV source (right), but is too
1317: faint to be detected in our FUV data (left).The field of view of
1318: these close-ups is $\approx 4\farcs5 \times 3\arcsec$. The location
1319: of HCV\,2005-B in our image is at $\alpha = 21^{h} 29^{m} 58.226^{s}$,
1320: $\delta = 12^{\circ} 10\arcmin 11\farcs4$.}
1321: \end{figure}
1322:
1323: \begin{table}
1324: \begin{center}
1325: \caption{\label{dntab} Cartesian coordinates in the NUV master image
1326: and $NUV$ magnitudes for the two DNe HCV\,2005-A (or CV\,1) and HCV\,2005-B.}
1327: \begin{tabular}{ccccc}
1328: name & x & y & NUV & $\Delta$ NUV \\
1329: \hline
1330: HCV\,2005-A & 744.462 & 722.678 & 18.913 & 0.031 \\
1331: HCV\,2005-B & 771.660 & 1180.19 & 21.361 & 0.051 \\
1332: \hline
1333: \end{tabular}
1334: \end{center}
1335: \end{table}
1336:
1337: \subsection{Other Cataclysmic Variables: V7, V11, V15, V39, V40, V41}
1338: \label{cvcand}
1339:
1340: The light curve of our variable V7 shows brightness
1341: variations of a few tenths of a magnitude on short timescales of
1342: minutes. CVs can show sizeable variability on a wide range of
1343: timescales, and the variation we observe here might be the so-called
1344: flickering, which is a random, aperiodic variability seen on
1345: timescales of seconds to several minutes and typical for CVs
1346: (e.g. Bruch 1992).
1347: The location of this variable source in the FUV--NUV CMD is close to
1348: the WD cooling sequence, which makes this source an excellent CV
1349: candidate. V11, V39, and V41 also are very good CV candidates as their
1350: light curves show flickering and they are all located in the gap region
1351: in our CMD, see Fig.~\ref{cmdps}.
1352:
1353: V15 and V40 show similar small time-scale variations. As no NUV
1354: information is available for those two sources, we call these
1355: variables suspected CVs.
1356:
1357: \subsection{Cepheids: V10}
1358: \label{cepheids}
1359:
1360: Cepheids are radially pulsating, bright variables with periods of
1361: $\approx 1 - 135$ days and variability amplitudes from several
1362: hundredths to 2 mag in $V$. The amplitude in $B$ is larger
1363: and increases towards the FUV.
1364: Traditionally, both Delta Cep and W Vir stars are called Cepheids as
1365: it is often impossible to discriminate between them on the basis of
1366: the light curves for periods in the range 3 -- 10 days. However, these
1367: are distinct groups of entirely different objects in different
1368: evolutionary stages. Delta Cep stars, the classical Cepheids,
1369: are relatively young objects belonging to the young disk
1370: population. They can be found in open clusters and have just evolved
1371: into the instability strip of the Hertzsprung-Russell diagram. W
1372: Virginis variables, on the other hand, belong to the old disk
1373: population and can be found in GCs and at high Galactic
1374: latitudes. They show periods of approximately 0.8 -- 35 days and
1375: amplitudes from 0.3 to 1.2 mag in $V$. The period-luminosity
1376: relations are different for Delta Cep and W Virginis variables. For
1377: more details, see \citet{GCVS}.
1378:
1379: The coordinates of V10 closely match (within $0\farcs2$) with the
1380: Cepheid FP V16, and its period of 1.62 days also agrees with this
1381: variable source being a Cepheid star. \citet{tuairisg} found a shorter
1382: period of only 1.2411 days for this object, but we note that their
1383: estimate is close to one of our aliases in our periodogram. The
1384: location of V10 in our CMD is above the ZAHB sequence.
1385:
1386: \subsection{Other possible Cepheid candidates: V1, V13, V18, V30, V31, V36}
1387: \label{cephcand}
1388:
1389: The most striking feature in the light curve of our variable
1390: V1 is that it brightens by $\approx 3$ mag in the fifth observing
1391: epoch. This might indicate an outburst, as in DNe. On the other hand,
1392: V1 might be a variable with a long-term period, possibly 2.6 days,
1393: which might suggest that this source is a Cepheid. We
1394: do not have additional NUV information, as this star is located
1395: outside the NUV field of view. However, the counterpart to V1 in
1396: \citet{vandermarel} is a bright source located on the BHB in the
1397: optical CMD (Fig.~\ref{cmd_opt}). This suggests that this
1398: variable is a Cepheid.
1399:
1400: V13 is $\approx 1 - 2$ mag brighter in the fifth and sixth
1401: observing epochs compared to the remaining epochs. This again might indicate
1402: an outburst as in DNe, or it might indicate a long-term variability of
1403: possibly 3.4 days, which is consistent with a Cepheid periodicity. Its
1404: location in our FUV--NUV CMD is above the ZAMS and close to the
1405: ZAHB. As CV candidates are expected in the region between the ZAMS and
1406: the WD cooling sequence, we suggest that V13 is a Cepheid candidate
1407: rather than a CV candidate.
1408: The same is true for our variable sources V18, V30, V31, and
1409: V36. All sources show an increase in brightness of about 2 mag in at
1410: least one observing epoch, which might be indicative of a DN
1411: outburst. However, periods of 1.4 days for V18, 4.5 days for V36, and
1412: 3.8 days for V30 and V31 fit to the corresponding light curves which
1413: might suggest that these stars are Cepheids. V18 and V36 are located above
1414: the ZAHB in our FUV--NUV CMD, see Fig.~\ref{cmdps}, whereas V30 and
1415: V31 are located above the ZAMS and below the ZAHB, but close to the
1416: location of the previously known Cepheid (our V10). The location of
1417: all these variables agrees with the expected location of Cepheids in
1418: our CMD.
1419:
1420: We caution that our variable classification is a tentative one, based
1421: on the derived periods and the photometric quantities of the variables.
1422: This is especially difficult for our Cepheid candidates, as
1423: the long periods are not evenly or completely covered by our
1424: data set. This is reflected in Fig.~\ref{fold}, which shows the folded
1425: light curves for all our variables. As can be seen, there are
1426: quite large gaps in the folded light curves for our Cepheid candidates,
1427: rendering their classification somewhat uncertain. However, except for
1428: V31, all our Cepheid candidates have optical counterparts, and these
1429: are all located along the BHB in the optical CMD, the expected
1430: location for Cepheid variables.
1431:
1432: \subsection{Low-mass X-ray binaries: V21/AC\,211 and M\,15 X-2}
1433: \label{lmxb}
1434:
1435: V21 is the known LMXB 4U\,2129+12 or AC\,211. It is not only the
1436: brightest variable but also the brightest FUV source in our
1437: catalog. This source has a period of 17.1 hr (Ilovaisky et
1438: al.~1993) and is one of the optically brightest LMXBs known. The
1439: light curve of V21/AC\,211 is presented in Fig.~\ref{lcurve2} and seems
1440: to indicate an eclipse during the sixth observing
1441: epoch. Unfortunately, our own data coverage is not good enough to
1442: derive a reliable period for this source.
1443:
1444: As mentioned earlier, the second LMXB in this cluster, M\,15 X-2, does
1445: not show a large $\sigma_{mean}$ in Fig.~\ref{varsigma}. We refer the
1446: reader to \citet{dieball2} for a thorough discussion of this UCXB. For
1447: the sake of completeness, we include M\,15 X-2 in Table~\ref{periods2}
1448: and mark this source in our CMD in Fig.~\ref{cmdps}.
1449:
1450: \subsection{SX Phoenicis: V33}
1451: \label{sxphoen}
1452:
1453: SX Phoenicis stars are pulsating BSs with short periods of less
1454: than 0.1 days \citep{jeon}. They are found in the old disk population
1455: and in GCs, and show variability amplitudes of up to 0.7
1456: mag in $V$.
1457:
1458: V33 shows large brightness variation of $\approx 3$ mag with a
1459: period of 1.32 hr. A sine wave with a corresponding period fits
1460: the observed light curve well, see Fig.~\ref{lcurve3}. The
1461: coordinates for V33 agree within $0\farcs8$ with the known SX
1462: Phoenicis star VGG 10041, which is ZK\,62 in \citet{kravtsov}, who derived a
1463: period of 1.248 hr for this variable star. The FUV amplitude
1464: is much larger than the optical one. The same effect is known for RR
1465: Lyrae stars and Cepheids. Large FUV amplitudes are to be expected
1466: since the amplitudes of the pulsations increase towards the
1467: FUV \citep{downes, wheatley}.
1468:
1469: \subsection{Another SX Phoenicis candidate: V22}
1470: \label{sxphoencand}
1471:
1472: Our variable V22 shows a 2 mag brightness variation with a
1473: period of 1.06 hr. Its location in our CMD, see Fig.~\ref{cmdps}, is
1474: slightly above the ZAMS and BS track. Its characteristics suggest that
1475: it might be an SX Phoenicis star.
1476:
1477: \subsection{Other suspected variables: V5, V6, V20, V24, V25, V26, V37, V38}
1478: \label{interact}
1479:
1480: The light curve of V5 shows a small-amplitude scatter, but the most
1481: striking feature is the rise in brightness at the end of the first
1482: observing epoch, which might correspond to an egress from an
1483: eclipse, or the beginning of an outburst as in DNe. The source is faint
1484: in the second, third, fourth, and sixth observing epochs, but brighter
1485: in the fifth observing epoch, which fits the characteristics of a
1486: DN. The source is located between the BS and ZAHB sequence in our
1487: FUV--NUV CMD, see Fig.~\ref{cmdps}. However, its optical counterpart
1488: is located on the BHB in the optical CMD. If this source is indeed a
1489: DN in M15, its optical counterpart is unusually bright. Further
1490: observations are required to shed more light on this variable source.
1491:
1492: The light curve of V6 shows a considerable drop in brightness in the
1493: second, third, and fifth observing epochs. The characteristics of the
1494: light curve match that of a DN. In this case, V6 seems to be in
1495: outburst during the first, fourth, and sixth observing
1496: epochs. This would imply an unusually high duty cycle of 70\% of the
1497: time coverage. Unfortunately, this star is located outside the NUV
1498: field of view.
1499: However, we do have an optical counterpart to this source, which
1500: is a rather unexpected one: it is one of the brightest optical
1501: sources, and its location in the optical CMD is blueward of the tip
1502: of the RG branch at $V=12.953$ and $B-V = 0.737$ mag (see
1503: Fig.~\ref{cmd_opt}). Its optical magnitude is far too bright to agree
1504: with a DN at the distance of M\,15. A possible explanation is that the
1505: optical counterpart is a ``mismatch'' (although the FUV and optical
1506: coordinates agree within $0\farcs0165$),
1507: and the ``true'' optical source might be an undetected, faint source
1508: in nearly the same line of sight as the bright one. Another, and
1509: perhaps more likely, explanation is that this source is a variable
1510: field star that happens to be in our line of sight. If it is a field
1511: DN, it would be much closer to us than M\,15, i.e.\ at least 400 pc
1512: assuming an absolute $M_V = 5$ mag for a DN in outburst.
1513:
1514: Our variable V20 shows flickering and in addition reveals some
1515: long-term variability. It is brightest during our first, third, fourth,
1516: and fifth observing epochs. This might indicate outburst activity,
1517: as in DNe. It is also possible that V20 is a CV of VY Scl type, which
1518: show sudden drops in brightness. If this is true, V20 might be in a
1519: low state in the second and sixth observing epochs. On the other hand,
1520: its light curve could also be explained with a Cepheid-type long-term
1521: periodicity of 1.5 days. This suspected variable is a bright source
1522: and is located just below the ZAHB in our CMD. Based on its light curve
1523: and location in the CMD alone, we cannot decide whether this source
1524: might be a bright CV or a bright Cepheid, and further observations are
1525: required to decide on the nature of this variable.
1526:
1527: V24 is located above the clump of MS stars and RGs and on the
1528: lower (fainter) part of the ZAHB in our FUV--NUV CMD. From its location
1529: in the FUV--NUV CMD, it might well be an FUV fainter but redder BS
1530: or a BHB star. However, the optical counterpart to this source is
1531: located on the BHB in the optical CMD. This suggests that this source
1532: might be a binary or pulsating HB star.
1533:
1534: The light curve of V25 shows some small-scale brightness variation on
1535: short time-scales, possibly flickering. However, in addition to the
1536: flickering, a long-term variability seems to be superposed, which can
1537: be seen in the first observing epoch where the source is $\approx 2$
1538: mag fainter towards its end. The location of V25 in our CMD is on
1539: the lower, fainter part of the ZAHB. This source is difficult to
1540: classify, and it might be a binary or pulsating BS or HB star.
1541:
1542: V26, V37, and V38 show flickering with a few tenths of a magnitude
1543: variation on short time-scales. Their location in the FUV--NUV CMD is
1544: close to the ZAMS and they might be either bright CVs or binary or
1545: pulsating BSs. Only V37 has an optical counterpart, but its optical
1546: magnitudes agree again with a BS or a bright CV.
1547:
1548: \subsection{Variability Census and its Implications}
1549:
1550: We found 41 sources that show strong signs of variability. Based
1551: on the appearance of their light curve (Figs.~\ref{lcurve1} to
1552: \ref{lcurve3}), their period, and their location in the FUV--NUV and
1553: in the optical CMDs (Figs.~\ref{cmd} and \ref{cmd_opt}), we suggest
1554: classifications for our variable sources.
1555: We compared the coordinates and periods, if available, of our variable
1556: sources with known variables in the core region of M\,15. In total,
1557: we found 17 RR Lyrae stars (4 of which are known), 6 CV
1558: candidates, 1 known SX Phoenicis star, 1 SX Phoenicis
1559: candidate, 1 previously known Cepheid and 6 further candidates,
1560: and the known LMXB AC\,211. The 8 remaining suspected variables
1561: cannot be classified clearly, but most of them might be CVs or
1562: pulsating/binary BSs.
1563:
1564: Note that only six out of our gap objects show strong signs of
1565: variability. However, the remaining gap objects might also be
1566: variable, but they are either fainter or have smaller amplitudes and
1567: thus were not detected in our variability study, see
1568: Fig.~\ref{varsigma}. (We note again that, for
1569: example, the UCXB M\,15 X-2 was not detected as a strongly variable
1570: source in this study, as it shows a modulation with only 0.062 mag
1571: semi-amplitude; see Dieball et al.\ 2005b.) We marked the gap objects
1572: that we selected based
1573: on our FUV--NUV CMD as red filled circles in Fig.~\ref{varsigma}. The
1574: green filled circles denote the two sources that do not have an NUV
1575: counterpart, and thus do not appear in our FUV--NUV CMD, but are
1576: likely CVs because of
1577: their variability. There are $\approx 60$ gap objects/CV candidates.
1578: Typical short-timescale variability for a CV is flickering with
1579: amplitudes of a few tenths of a magnitude. As can be seen from
1580: Fig.~\ref{varsigma}, there are 25 gap sources that show a
1581: $\sigma_{mean} > 0.5$ mag, but most of these are located in the bulge
1582: of faint sources at FUV$_{mean} > 22$ mag and thus were not considered
1583: in the above study. We consider these sources good CV candidates. This
1584: number agrees with our previous estimate, see Sect.~\ref{cv}.
1585:
1586:
1587: \section{Summary}
1588: \label{summary}
1589:
1590: We have analyzed deep FUV {\it HST} ACS observations of the core of
1591: the metal-poor GC M\,15. Based on our FUV and NUV
1592: data, we constructed a CMD in which various stellar populations are
1593: present. Our CMD is deep enough that MS stars and RGs show
1594: up. The MS turnoff is visible at FUV $\simeq 23.5$ mag and FUV--NUV
1595: $\simeq 3$ mag, and the MS stars and RGs form a prominent data
1596: clump that extends at least 2 mag below the MS turnoff.
1597: As such, this is the deepest FUV CMD presented so far. A line of
1598: $\approx 70$ BS candidates and a well-defined sequence of $\approx
1599: 130$ HB stars are present in our CMD. BHB stars are located along the
1600: well-defined HB sequence, whereas EHB stars are clustered within a
1601: bright data clump. Approximately 30 WD candidates are found close to
1602: the WD cooling sequence, which is well within the expected number of
1603: WDs in our field of view down to FUV $\simeq 22.5$ mag. The region
1604: between the WD cooling sequence and the ZAMS and BS track is occupied
1605: by $\approx 60$ objects. Theoretical considerations suggest that at
1606: least some of these are CVs. We caution that these numbers should not
1607: be taken as exact, since the discrimination between the various zones
1608: in our CMD is difficult. The cumulative radial distributions of the stellar
1609: populations in our FUV--NUV CMD suggest that gap sources and BS
1610: candidates are the most centrally concentrated populations. This
1611: might be the expected effect of mass segregation; i.e., more massive
1612: stars and binaries like CVs and BSs sink towards the cluster
1613: core. Alternatively, it might reflect the preferred birthplace of such
1614: objects, as CVs and BSs are thought to be dynamically formed
1615: and as such are preferentially found in the dense cluster core. We
1616: found that also EHB stars are more concentrated in the core region than BHB
1617: stars. This might indicate that at least some EHB stars also have a
1618: dynamical origin.
1619:
1620: We searched for variability among our FUV catalog stars and found
1621: 41 variable sources. They are distributed over the entire CMD,
1622: but most of them are located around either the ZAHB or the
1623: ZAMS. Periods could be derived for 27 variables. Based on their
1624: variability properties and their location in the FUV--NUV
1625: CMD, we suggest classifications for 33 variables. Seven
1626: variable sources could be identified with previously known variables.
1627:
1628: In total, we found four known RR Lyrae stars, 13 additional RR Lyrae
1629: candidates, six CV candidates, two SX Phoenicis stars (of which one
1630: is known), seven Cepheids (one previously known), and the known LMXB
1631: AC\,211. The eight remaining suspected variables
1632: cannot be classified clearly, but most of them might be CVs or
1633: pulsating/binary BSs. We found that the brightness variations for RR
1634: Lyrae stars, Cepheids, and SX Phoenicis stars are a few magnitudes
1635: in the FUV and thus much larger than in the optical. This is to be
1636: expected since the amplitudes of the pulsations increase towards the
1637: FUV \citep{downes, wheatley}. However, this is the first time such
1638: large-scale FUV amplitudes have been observed for SX Phoenicis stars.
1639: Note that out of our 60 gap objects/CV candidates, only 6 show strong
1640: signs of variability. However, 25 of the gap sources show
1641: $\sigma_{mean} \ge 0.5$ mag. We consider at least these to be good CV
1642: candidates. This number agrees with theoretical predictions.
1643:
1644: All previously known X-ray sources and DNe that are located in our
1645: field of view are detected in our UV study. The LMXB AC\,211 is the
1646: brightest FUV source in our sample, and also the UCXB M\,15 X-2 is
1647: among the brightest sources. HCV\,2005-A and HCV\,2005-B
1648: \citep{hannikainen} are both detected in our NUV data set, although
1649: HCV\,2005-B appears only as a faint NUV source and is not detected in
1650: the FUV. HCV\,2005-A or CV\,1 \citep{shara} is in outburst in our
1651: fourth FUV observing epoch, but otherwise it is too faint to be detected
1652: in the FUV.
1653:
1654: Overall, the results of our study confirm that FUV observations are
1655: particularly well suited in studying hot, and especially dynamically-formed,
1656: stellar populations like CVs, BSs, and X-ray binaries in the cores of
1657: GCs. They are also a powerful tool in detecting new variable stars in
1658: the dense core regions of GCs where optical studies are hampered by
1659: the immense crowding. In addition, the large FUV amplitudes of
1660: especially RR Lyrae stars, Cepheids, and SX Phoenicis stars make them easily
1661: detectable as variables in this wave band.
1662:
1663: All clusters that have been studied so far in the FUV, namely, 47\,Tuc,
1664: NGC\,2808, and M\,15, revealed large numbers of gap objects that were
1665: more or less consistent with theoretical predictions for dynamically
1666: formed CVs. It is interesting to directly compare NGC\,2808 and M\,15,
1667: as both clusters now have been studied in both the FUV and NUV,
1668: with a full coverage of the cluster cores. Table~\ref{gc} lists the
1669: cluster parameters taken from \citet{harris} and the number of gap
1670: objects and BS and WD candidates found in the clusters. As can be
1671: seen, slightly more gap sources were detected in NGC\,2808 than in
1672: M\,15, whereas more BS candidates are present in M\,15 than in
1673: NGC\,2808. Evidence for a concentration of the more massive CVs and
1674: BSs is more evident in NGC\,2808 (see Dieball et al.~2005a, their Fig.~4),
1675: which is somewhat contrary to the longer relaxation time for the
1676: cluster's core (see Table~\ref{gc}). However, we caution that these
1677: are still small number statistics, so their difference is not
1678: particularly significant. We are currently working on data of more
1679: clusters. Our current sample with just two GCs is too small to draw
1680: any conclusions about the impact of a difference in the binary
1681: populations on the cluster evolution. However, the large number of
1682: exotica and massive binaries found in the cores of these clusters
1683: agrees with theoretical expectations for evolved clusters.
1684:
1685: \begin{table}[hb!]
1686: \begin{center}
1687: \caption{\label{gc} Number of gap sources and WD and BS candidates detected in
1688: GC. The cluster metallicity and the logarithmic core relaxation time
1689: log(tc) and halfmass relaxation time log(th) are taken from \citet{harris}.}
1690: \begin{tabular}{lcccccc}
1691: \tableline
1692: name & gap & WDs & BS & [Fe/H] & lg(tc) & lg(th) \\
1693: \tableline
1694: NGC\,2808 & 60 & 40 & 61 & -1.15 & 8.30 & 9.13 \\
1695: M\,15 & 57 & 28 & 69 & -2.26 & 7.02 & 9.35 \\
1696: \tableline
1697: \end{tabular}
1698: \end{center}
1699: \end{table}
1700:
1701:
1702: \acknowledgments
1703:
1704: We are grateful to an anonymous referee for a valuable discussion that
1705: helped us to improve this paper. DH acknowledges the Academy of Finland.
1706: This work was supported by NASA through grant GO-9792 from the Space
1707: Telescope Science Institute, which is operated by AURA, Inc., under NASA
1708: contract NAS5-26555.
1709:
1710: \begin{thebibliography}{}
1711: \bibitem[Arellano et al.~(2006)]{arellano} Arellano, F. A., García,
1712: L. G. \& Rosenzweig, P. 2006, RMxAA, 42, 75
1713: \bibitem[Auri{\`e}re et al.~(1984)]{auriere} Auri{\`e}re, M., Le F{\`e}vre,
1714: O. \& Terzan, A. 1984, A\&A, 138, 415
1715: \bibitem[Baumgardt et al.~(2003)]{baumgardt} Baumgardt, H., Hut, P.,
1716: Makino, J., McMillan, S. \& Portegies Zwart, S. 2003, ApJ, 582, L21
1717: \bibitem[Brown et al.~(2001)]{brown} Brown, T. M., Sweigart, A. V.,
1718: Wayne B. L.\ \& Hubeny, I.\ 2001, ApJ, 562, 368
1719: \bibitem[Bruch (1992)]{bruch} Bruch, A.\ 1992, A\&A, 266, 237
1720: \bibitem[Charles et al.\ (2002)]{charles} Charles, P. A., Clarkson,
1721: W. I. \& van Zyl, L. 2002, New Astronomy, 7, 21
1722: \bibitem[Clement et al. (2001)]{clement01} Clement, C., Muzzin, A., Dufton,
1723: Q. et al. 2001, AJ, 122, 2587
1724: \bibitem[D'Antona et al.~(1995)]{dantona} D'Antona, F., Vietri,
1725: M. \& Pesce, E. 1995, MNRAS, 272, 730
1726: \bibitem[De Marchi \& Paresce (1994)]{} De Marchi, G. \& Paresce,
1727: F. 1994, ApJ, 422, 597
1728: \bibitem[De Marchi \& Paresce (1996)]{demarchiparesce} De Marchi,
1729: G. \& Paresce, F. 1996, ApJ, 467, 658
1730: \bibitem[Dieball et al.~(2005a)]{dieball1} Dieball, A., Knigge, C., Zurek, D.,
1731: Shara, M.M. \& Knox, L. 2005a, ApJ, 625, 156
1732: \bibitem[Dieball et al.~(2005b)]{dieball2} Dieball, A., Knigge, C., Zurek, D.,
1733: Shara, M.M. \& Knox, L. 2005b, ApJL, 634, L105
1734: \bibitem[Di\,Stefano \& Rappaport (1994)]{distefano} Di\,Stefano, R.\
1735: \& Rappaport, S.\ 1994, ApJ, 423, 274
1736: \bibitem[Dorman (1992)]{dorman} Dorman, B.\ 1992, ApJS, 81, 221
1737: \bibitem[Dotani et al.~(1990)]{dotani} Dotani, T., Inoue, H., Murakami, T.,
1738: Nagase, F. \& Tanaka, Y. 1990, Nature, 347, 534
1739: \bibitem[Downes et al.~(2004)]{downes} Downes, R. A., Margon, B.,
1740: Homer, L.\ \& Anderson, S. F. 2004, AJ, 128, 2288
1741: \bibitem[Edmonds et al.~(2003a)]{}Edmonds, P. D., Gilliland, R. L.,
1742: Heinke, C. O. \& Grindlay, J. E. 2003a, ApJ, 596, 1177
1743: \bibitem[Edmonds et al.~(2003b)]{} Edmonds, P. D., Gilliland, R. L.,
1744: Heinke, C. O. \& Grindlay, J. E. 2003b, ApJ, 596, 1197
1745: \bibitem[Elson et al.~(1987)]{elson} Elson R, Hut P.\ \& Inagaki S.\
1746: 1987, \araa, 25, 565
1747: \bibitem[Fusi Pecci et al.~(1993)]{fusipecci} Fusi Pecci, F., Ferraro,
1748: F. R., Bellazzini, M. et al. 1993, AJ, 105, 1145
1749: \bibitem[G\"{a}nsicke et al.~(1995)] G\"{a}nsicke, B. T., Beuermann, K. \&
1750: de Martino, D. 1995, A\&A, 303, 127
1751: \bibitem[Gebhardt et al.~(1997)]{gebhardt} Gebhardt, K., Pryor, C.,
1752: Williams, T. B., Hesser, J. E. \& Stetson, P. B. 1997, AJ, 113, 1026
1753: \bibitem[Gerssen et al.~(2002)]{gerssen} Gerssen, J., Van der Marel,
1754: R. P., Gebhardt, K. et al.\ 2002, AJ, 124, 3270
1755: \bibitem[Grindlay et al.~(2001)]{} Grindlay, J. E., Heinke, C.,
1756: Edmonds, P. D. \& Murray, S. S. 2001, Science, 292, 2290
1757: \bibitem[Guhathakurta et al.~(1996)]{guhathakurta} Guhathakurta, P.,
1758: Yanny, B., Schneider, D. P. \& Bahcall, J. N. 1996, AJ, 111, 267
1759: \bibitem[Hannikainen et al.~(2005)]{hannikainen} Hannikainen, D. C.,
1760: Charles, P. A., van Zyl, L. et al. 2005, MNRAS, 357, 325
1761: \bibitem[Harris (1996)]{harris} Harris, W.E.\ 1996, Astronomical
1762: Journal, 112, 1487
1763: \bibitem[Heinke et al.~(2005)]{heinke2005} Heinke, C. O., Grindlay,
1764: J. E., Edmonds, P. D. et al. 2005, ApJ, 625, 796
1765: \bibitem[Heinke et al.~(2003)]{heinke} Heinke, C. O., Grindlay, J. E.,
1766: Lugger, P. M.\ et al. 2003, ApJ, 598, 501
1767: \bibitem[Hut et al.~(1992)]{hut} Hut, P., McMillan, S., Goodman, J.,
1768: Mateo, M., Phinney, E. S.\ et al.\ 1992, PASP, 104, 981
1769: \bibitem[Iben (1990)]{iben} Iben, I. J. 1990, ApJ, 353, 215
1770: \bibitem[Ilovaisky et al.~(1993)]{ilovaisky} Ilovaisky, S. A., Auri{\`e}re, M.,
1771: Kock-Miramond, L. et al. 1993, A\&A, 270, 139
1772: \bibitem[Ivanova et al.~(2006)]{ivanova} Ivanova, N., Heinke, C. O.,
1773: Rasio, F. A. et al.~2006, MNRAS, 372, 1043
1774: \bibitem[Jeon et al.~(2001)]{jeon} Jeon, Y.-B., Kim, S.-L., Lee, H. \& Lee,
1775: M. G. 2001, AJ, 121, 2769
1776: \bibitem[Kholopov et al.~(1988)]{GCVS} Kholopov, P.N., Samus, N.N.,
1777: Frolov, M.S., Goranskij, V.P., Gorynya, N.A. et al. 1988, General
1778: Catalog of Variable Stars (4th ed.; Moscow: Nauka)
1779: \bibitem[Knigge et al.~(2002)]{knigge1} Knigge, C., Zurek, D. R., Shara,
1780: M. M.\ \& Long, K. S.\ 2002, ApJ, 579, 752
1781: \bibitem[Knigge et al.~(2003)]{knigge2} Knigge, C., Zurek, D. R., Shara,
1782: M. M., Long, K. S.\ \& Gilliland, R. L.\ 2003, ApJ, 599, 1320
1783: \bibitem[Kravtsov \& Zheleznyak (2003)]{kravtsov} Kravtsov, V. V. \&
1784: Zheleznyak, A. P. 2003, Information Bulletin on Variable Stars, 5452, 1
1785: \bibitem[Kulkarni et al.~(1990)]{kulkarni} Kulkarni, S. R., Goss,
1786: W. M., Wolszczan, A.\ \& Middleditch, J. 1990, ApJ, 363, L5
1787: \bibitem[McNamara et al.~(2003)]{} McNamara, B. J., Harrison, T. E. \&
1788: Anderson, J. 2003, ApJ, 595, 187
1789: \bibitem[McNamara et al.~(2004)]{mcnamara} McNamara, B. J., Harrison,
1790: T. E. \& Baumgardt, H. 2004, ApJ, 602, 264
1791: \bibitem[Oosterhoff (1939)]{oosterhoff} Oosterhoff, P. T. 1939, The
1792: Observatory, 62, 104
1793: \bibitem[Peterson et al.~(1989)]{peterson} Peterson, R. C., Seitzer,
1794: P. \& Cudworth, K. M. 1989, ApJ, 347, 251
1795: \bibitem[Phinney (1996)]{phinney} Phinney, F. S. 1996, The origins,
1796: evolution, and destinies of binary stars in clusters, Astronomical
1797: Society of the Pacific Conference Series, Volume 90, p. 163
1798: \bibitem[Richer et al.\ (1997)]{richer97} Richer, H. B., Fahlman,
1799: G. G., Ibata, R. A. et al. 1997, ApJ, 484, 741
1800: \bibitem[Shara et al.~(2004)]{shara} Shara, M. M., Hinkley, S., Zurek,
1801: D. R., Knigge, C. \& Bond, H. E. 2004, AJ, 128, 2847
1802: \bibitem[Shara \& Hurley (2006)]{shara06} Shara, M. M. \& Hurley,
1803: J. R. 2006, ApJ, 646, 464
1804: \bibitem[Silbermann \& Smith (1995)]{silbermann} Silbermann, N. A. \&
1805: Smith, H. A. 1995, AJ, 110, 704
1806: \bibitem[Sirianni et al.~(2005)]{sirianni} Sirianni, M., Jee, M.J.,
1807: Benítez, N. et al.\ 2005, PASP, 117, 1049
1808: \bibitem[Smale (2001)]{smale} Smale, A. P. 2001, ApJ, 562, 957
1809: \bibitem[Stetson (1991)]{stetson} Stetson, P. B.\ 1991, in 3rd ESO/ST-ECF
1810: Garching - Data Analysis Workshop, eds. Grosb{\o}l P. J., Warmels
1811: R. H., p. 187
1812: \bibitem[Tout et al.~(1996)]{tout} Tout, C. A., Pols, O. R.,
1813: Eggleton, P. P. K., de Martino, D.\ \& Han, Z. 1996, MNRAS, 281, 257
1814: \bibitem[Townsley \& Bildsten (2003)]{tb} Townsley, D. M. \& Bildsten,
1815: L.\ 2003, ApJ, 596, L227
1816: \bibitem[Tuairisg et al.~(2003)]{tuairisg} Tuairisg, S. O., Butler, R. F.,
1817: Shearer, A. et al.\ 2003, MNRAS, 345, 960
1818: \bibitem[van den Bosch et al.~(2006)]{vandenbosch} van den Bosch, R.,
1819: de Zeeuw, T., Gebhardt, K., Noyola, E.\ \& van de Ven, G.\ 2006,
1820: ApJ, 641, 852
1821: \bibitem[Van der Marel et al.~(2002)]{vandermarel} Van der Marel,
1822: R. P., Gerssen, J., Guhathakurta, P., Peterson, R. C. \& Gebhardt,
1823: K. 2002, AJ, 124, 3255
1824: \bibitem[Wheatley et al.~(2004)]{wheatley} Wheatley, J. M., Welsh,
1825: B. Y., Siegmund, O. H. W. et al.\ 2004, ApJ, 619, L123
1826: \bibitem[White \& Angelini (2001)]{wa} White, N. E. \& Angelini,
1827: L. 2001, ApJ, 561, L101
1828: \bibitem[Wood (1995)]{wood} Wood, M. A. 1995, Lecture Notes in Physics, 443, 41
1829: \bibitem[Yanny et al.~(1994)]{yanny} Yanny, B., Guhathakurta, P., Bahcall,
1830: J. N. \& Schneider, D. P. 1994, AJ, 107, 1745
1831: \end{thebibliography}
1832:
1833: \end{document}
1834:
1835: