astro-ph0510430/ms.tex
1: \documentclass[12pt,preprint]{aastex}
2: 
3: \newcommand{\myemail}{andrea@astro.soton.ac.uk}
4: 
5: \shorttitle{An ultracompact X-ray binary in M\,15}
6: \shortauthors{Dieball et al.}
7: 
8: \begin{document}
9: 
10: \title{An ultracompact X-ray binary in the globular cluster M\,15
11:   (NGC\,7078)$^1$}
12: \footnotetext[1]{Based on observations with the NASA/ESA Hubble Space
13:   Telescope, obtained at the Space Telescope Science Institute, which
14:   is operated by the Association of Universities for Research in
15:   Astronomy, Inc. under NASA contract No. NAS5-26555.} 
16: 
17: \author{A.~Dieball$^2$, C.~Knigge$^2$, D.~R.~Zurek$^3$,
18:   M.~M.~Shara$^3$, K.~S.~Long$^4$, P.~A.~Charles$^{5,2}$,
19:   D.~C.~Hannikainen$^6$ and L.~van~Zyl$^{7,8}$}
20: \footnotetext[2]{School of Physics and Astronomy, University of Southampton,
21:   SO17 1BJ, UK}
22: \footnotetext[3]{Department of Astrophysics, American Museum of
23:   Natural History, New York, NY 10024}
24: \footnotetext[4]{Space Telescope Science Institute, Baltimore, MD 21218}
25: \footnotetext[5]{South African Astronomical Observatory, PO Box 9,
26:   Observatory, 7935, South Africa} 
27: \footnotetext[6]{University of Helsinki, P.O. Box 14, SF-00014
28:   Helsinki, Finland}  
29: \footnotetext[7]{Astrophysics Group, School of Physics, Keele University,
30:   Staffordshire ST5 5BG, UK}  
31: \footnotetext[8]{Department of Astrophysics, Oxford University, Oxford
32:   OX1 3RH, UK} 
33: \begin{abstract}
34: 
35: We have used the Advanced Camera for Surveys on board the Hubble Space
36: Telescope to image the core of the globular cluster M\,15 in the
37: far-ultraviolet (FUV) waveband. Based on these observations, we identify
38: the FUV counterpart of the recently discovered low-mass X-ray binary
39: M\,15 X-2. Our time-resolved FUV photometry shows a modulation with
40: $0.062 \pm 0.004$ mag semi-amplitude and we clearly detect a period of
41: $22.5806 \pm 0.0002$ min. We have carried out extensive Monte Carlo
42: simulations which show that the signal is consistent with being
43: coherent over the entire observational time range of more than 3000
44: cycles. This strongly suggests that it represents the orbital period
45: of the binary system. M\,15 X-2 is FUV bright ($\rm{FUV} \simeq 17$ mag)
46: and is characterized by an extremely blue spectral energy distribution
47: ($F_{\lambda} \propto \lambda^{-2.0}$). We also find evidence for an
48: excess of flux between 1500~\AA\ and 1600~\AA\ and probably between
49: 1600~\AA\ and 2000~\AA\, which might be due to C\,IV 1550~\AA\  and
50: He\,II 1640~\AA\  emission lines. We also show that M15 X-2's X-ray
51: luminosity can be powered by accretion at the rate expected for
52: gravitational-wave-driven mass transfer at this binary period. The
53: observed FUV emission appears to be dominated by an irradiated
54: accretion disk around the neutron star primary, and the variability
55: can be explained by irradiation of the low-mass white dwarf donor if
56: the inclination of the system is $\approx 34^\circ$. We conclude that
57: all observational characteristics of M15 X-2 are consistent with it
58: being an ultracompact X-ray binary, only the third confirmed such
59: object in a globular cluster.   
60: 
61: \end{abstract}
62: 
63: \keywords{globular clusters: individual(\objectname{M\,15}) -- stars:
64:   close binaries -- stars: individual {(M\,15\,X-2)} -- ultraviolet: stars}
65: 
66: \section{Introduction}
67: 
68: Ultracompact X-ray binaries (UCXBs) are very tight, interacting
69: binaries ($\approx 10^{10}$ cm orbit) with periods $\le 80$
70: min. They contain neutron stars (NSs) or black holes accreting from a
71: low-mass ($\le 0.1 {\rm M}_{\odot}$) degenerate companion. UCXBs are a
72: sub-class of the low-mass X-ray binaries (LMXBs). It has been known
73: for a long time that bright LMXBs are found to be overabundant in
74: globular clusters (GCs) compared to the Galactic field (Katz 1975,
75: Clark 1975). This leads to the suggestion that LMXBs, and 
76: consequently UCXBs, are preferentially formed in the dense environment
77: of the GC cores through various stellar interactions, such as
78: direct collisions between a NS and a red giant (Verbunt 1987, Davies
79: et al.~1992, Ivanova et al.~2005), tidal capture of a main sequence
80: star by a NS (Bailyn \& Grindlay 1987), or exchange interactions
81: between NSs and primordial binaries (Rasio et al.~2000). 
82: It has been suggested that most of the 13 bright LMXBs in GCs might be
83: UCXBs (Bildsten \& Deloye 2004, Ivanova et al.\ 2005). However, only
84: two are confirmed in GCs to date. These are 4U\,1820-30 in NGC\,6624 
85: ($P_{\rm{orb}} = 11.4$ min, Stella et al.~1987)
86: and 4U\,1850-087 in NGC\,6712 ($P_{\rm{orb}} = 20.6$ min, Homer et
87: al.~1996). Two other GC sources have been suggested as particularly
88: strong UCXB candidates (Homer 2003).
89: One of these candidates is CXO\,J212958.1+121002
90: in M\,15, also known as M\,15 X-2 (White \& Angelini 2001). 
91: 
92: M\,15 is the only galactic GC known to harbour {\it two} bright
93: LMXBs. A single source was detected in early X-ray studies,
94: 4U\,2127+119, and that was identified  with the optical counterpart
95: AC\,211 (Auri{\`e}re et al. 1984, Charles et al. 1986). 
96: {\it Chandra} observations later on resolved 4U\,2127+119 into {\it
97:   two} X-ray sources (White \& Angelini 2001). One of these was the
98: previously known LMXB AC\,211. The second source, CXO\,J212958.1+121002 
99: or M\,15 X-2, is actually 2.5 times brighter than AC\,211 in X-rays. 
100: Based on {\it Hubble Space Telescope (HST)} data from Guhathakurta et
101: al.\ (1996), White \& Angelini (2001) identified a blue $U = 18.6$ mag
102: star as the optical counterpart to the second source (star 590 in
103: De\,Marchi \& Paresce 1994). However, the orbital period of
104: M\,15 X-2 has so far not been determined. 
105: 
106: Here, we present FUV data of M\,15 X-2 taken with the {\it HST} that allow
107: us to classify the source as an UCXB. In 
108: Sect.~\ref{data} we describe the data and their reduction. We present
109: the analysis of the photometry and determine the period of M\,15 X-2
110: in Sect.~\ref{analysis}. We summarize our results and conclusions in
111: Sec.~\ref{summary}. 
112: 
113: \section{Observations and data reduction}
114: \label{data}
115: 
116: M\,15 was observed with the Advanced Camera for Surveys (ACS) on board
117: the {\it HST} in  September and October 2003, and October to December
118: 2004. Images were taken using the far-ultraviolet (FUV) filters F140LP,
119: F150LP, and F165LP in the Solar Blind Channel (SBC), and the near-UV (NUV)
120: F220W filter in the High Resolution Channel (HRC). 
121: 
122: For our variability study (see Sect.~\ref{analysis}) only relative
123: magnitudes are needed. These were derived directly from the
124: individual flatfielded images. Aperture photometry was carried out
125: using {\tt daophot} (Stetson 1991) running under {\tt IRAF}
126: \footnote{{\tt IRAF} (Image 
127:   Reduction and Analysis Facility) is distributed by the National
128:   Astronomy and Optical Observatories, which are operated by AURA,
129:   Inc., under cooperative agreement with the National Science
130:   Foundation.} 
131: using an aperture radius of 4 pixels and a sky annulus of 50 to 60
132: pixels. For this purpose, we only used the 90 exposures taken in
133: SBC/F140LP, since these data provide the longest  time coverage
134: (from October 14 to December 5, 2004). Eighty of these images had
135: exposure times of 300 sec, four were exposed for 140 sec and six for
136: 40 sec. One orbit of observation yielded 9 data points.   
137: 
138: In order to determine the time-averaged spectral energy distribution
139: (SED) of our counterpart, we also carried out {\it absolute}
140: photometry on combined and geometrically corrected images for each
141: filter. These master images were created using {\tt multidrizzle}
142: running under {\tt PyRAF}. For our aperture photometry, we used an
143: aperture radius of 5 pixels for all SBC FUV data, a smaller radius of
144: 4 pixels in the HRC NUV data, and a sky annulus of 5 to 7 pixels. The
145: smaller aperture in the NUV was chosen to avoid the effects of severe
146: crowding. In the FUV, aperture corrections were determined via curves
147: of growth constructed from isolated stars in our master images. For
148: the NUV data, we used the encircled energy fractions published by
149: Sirianni et al. (2005). 
150: In order to derive a reliable SED, we ideally want mean fluxes in
151: non-overlapping wavelength windows. However, the ACS/SBC/F140LP
152: bandpass fully includes F150LP, and F150LP in turn includes F165LP. We
153: therefore created two artificial narrow-band filters that are defined
154: as the differences between the actual filters, i.e.\ we
155: define  F140N = F150LP - F140LP and F150N = F150LP - F165LP. Thus the
156: count rate of a source in F140N, for example, is simply obtained by
157: subtracting its count rate in F150LP from that in
158: F140LP. Fig.~\ref{spect} (top panel) shows the resulting
159: throughput curves for the artificial filters. As can be seen, they  barely
160: overlap and are thus ideal for characterizing the UV SED of our source.   
161: In order to convert count rates into STMAGs, we used the {\tt synphot}
162: package running under {\tt IRAF}. Full details on our analysis
163: procedure will be provided in a forthcoming publication that will
164: present our entire FUV/NUV data set for M\,15. 
165: 
166: \section{Analysis and Discussion}
167: \label{analysis}
168: 
169: \subsection{Source Identification}
170: 
171: We identified the FUV counterparts to AC\,211 and to M\,15 X-2 on the
172: F140LP images, using the Guhathakurta et al.~(1996) {\it HST} images and the
173: positions provided by White \& Angelini (2001) as a reference
174: guide. Fig.~\ref{XB2} shows a close-up of the FUV (SBC/F140LP) and NUV
175: (HRC/F220W) master images, centred on AC\,211 and M\,15 X-2. The
176: {\it Chandra} positions for these sources are also indicated, after shifting
177: them by $\approx 1\farcs2$~south in order to optimally align the X-ray
178: and FUV positions of AC\,211. The centre of the FUV counterpart to
179: M\,15 X-2 is slightly offset ($0\farcs165$ south) of the {\it
180:   Chandra} position, but still well within the internal $0\farcs25$
181: {\it Chandra} error radius for this source (White \& Angelini
182: 2001). Our offset is consistent with the $0\farcs13$ offset that White
183: \& Angelini (2001) found between their {\it Chandra} positions and the
184: optical counterpart from De\,Marchi \& Paresce (1994). Both AC\,211
185: and M\,15 X-2 are clearly detected as strong FUV and NUV sources in
186: our ACS images.    
187: 
188: \subsection{Time-Series Analysis}
189: 
190: Fig.~\ref{lightcurve} shows the mean-subtracted light curve for all
191: observing epochs. Low-amplitude variations with a peak-to-peak
192: amplitude $> 0.1$ mag can be seen. We searched for a periodic signal
193: by carrying out $\chi^{2}$ fits for a grid of trial frequencies. 
194: The resulting periodogram is shown in
195: Fig.~\ref{power}. The best fit yields a $\chi_{\nu}^{2}=1.31$ and
196: suggests a period of $22.5806$ 
197: min. In order to test the coherence of this period, we carried out
198: Monte Carlo simulations. Briefly, we created 10000 fake data sets with
199: the same time sampling, periodic signal and noise
200: characteristics as our real data, but with a random phase offset
201: assigned to each of the six observing epochs. Thus the phase was fully
202: coherent within each epoch, but fully randomized between them. 
203: We then again carried out a sequence of $\chi^{2}$\ fits for each data
204: set. Next, we fixed the period at the best global value and fitted
205: each epoch separately with both phase and amplitude as free
206: parameters. We then defined a {\it phase coherence index} (PCI, see
207: e.g.\ Haswell et al.\ 1997), which in our case is simply the
208: $\chi^{2}$ of the phase estimates for the individual epochs, with the
209: phase of the global fit as the reference value. {\it We find that only
210:   0.9\% of the fake data with randomized phases have a PCI as good as
211:   the real data.} We can therefore reject the null hypothesis that the
212: periodic signal loses coherence completely over time-scales comparable
213: to our inter-epoch spacing. The latter is $\approx 11$ days,
214: corresponding to $\approx 700$ cycles. We can view this as a
215: constraint on the period derivative, i.e. $\dot{P}$ must be small
216: enough so that less than one cycle is lost over $N \simeq 700$ cycles,
217: i.e.\ $\dot{P} \lesssim N^{-1}$. The quality factor of the 22.58~min
218: signal must therefore be $Q = \dot{P}^{-1} \gtrsim 700$. By contrast,
219: mHz QPOs tend to have $Q \approx 10$ (e.g. Chakrabarty et al.\ 2001,
220: Boroson et al.\ 2000). As a further check, we carried out Monte Carlo
221: simulations in which the input sinusoids were coherent, i.e. the phase
222: was fixed at the same value for each epoch. This produced a PCI
223: distribution that was consistent with the PCI of the real data
224: set. Thus the 22.58~min signal is consistent with being fully coherent
225: over the entire $\simeq 3300$\ cycles spanned by our observations. All
226: of these tests support the orbital nature of this signal.   
227: 
228: We also used our Monte Carlo simulations to estimate the statistical
229: error on the parameters of the observed signal. For this purpose, we
230: again used coherent input signals and conservatively used error bars
231: scaled so as to yield $\chi_{\nu}^{2} =1$ for the fit to the real
232: data. We then used the standard deviation of the periods and amplitudes found
233: for the fake data sets to estimate the errors on the measured
234: parameters. The final results were $P = 22.5806 \pm 0.0002$ min for
235: the orbital period and $a = 0.062 \pm 0.004$ mag for the
236: semi-amplitude. 
237: This yields an ephemeris for the time of maximum light
238: \begin{equation}
239: T_{max}\rm{(BJD)} = 2453308.88743(16) + 0.01568096(13) \times E,
240: \end{equation}
241: where the numbers in brackets give the errors on the last two digits. 
242:   
243: A sine wave with M\,15\,X-2's period, amplitude and fixed phase is
244: overplotted on the lightcurves in Fig.~\ref{lightcurve}. 
245: This visually confirms that there is no sign of a loss of coherence
246: over the entire observational time span of 3312 cycles. We conclude
247: that the periodic signal is almost certainly an orbital modulation. 
248: \footnote{We note that the average time resolution of our data is
249:   $\approx 5.5$ min which corresponds to a Nyquist frequency of
250:   $\nu_{Ny} \approx 130 \rm{d}^-1$. A period of $\approx 7.4$ min
251:   above $\nu_{Ny}$ could be reflected to yield our observed signal of
252:   22.6 min. However, such short orbital periods are extremely unlikely
253:   (see e.g. Deloye \& Bildsten 2003, Homer 2003).}
254: 
255: \subsection{Continuum Spectral Energy Distribution}
256: 
257: The SED of M\,15 X-2 is shown in Fig.~\ref{spect} (top panel). Each
258: point is plotted at the average wavelength of the corresponding
259: filter. As expected for compact, interacting binaries, the SED of M\,15 X-2
260: rises towards the blue. However, it is also worth noting that there
261: seems to be an excess of flux in F150N, i.e.\ around 1550~\AA. This
262: excess flux can be caused by additional C\,IV $\lambda = 1550$~\AA\
263: and/or He\,II $\lambda = 1640$~\AA\ line emission. We therefore fit a
264: power law to the two bracketing data points only. The best fit is found for a
265: power-law index $-2.0 \pm 0.2$. This power-law spectrum is
266: overplotted in Fig.~\ref{spect}. The depression at $\approx 2200$ \AA\
267: is due to a well known reddening feature there.  
268: 
269: \subsection{Evidence for Line Emission}
270: \label{ev_line}
271: 
272: The excess flux in F150N seems to indicate line emission due to C\,IV
273: at~$\lambda = 1550$~\AA\ and/or He\,II $\lambda = 1640$~\AA\ (the
274: latter would also contribute to F165LP). However, such a peak might
275: also be caused simply by a turnover in an otherwise smooth 
276: continuum. We have therefore carried out synthetic photometry 
277: for blackbodies (BBs) with temperatures
278: $100000 \ge T_{eff} \ge 10000$. Fig.~\ref{spect} (bottom panel) shows
279: the resulting BB sequence in the F140N-F150N vs F150N-F165LP
280: colour-colour diagram (CC diagram). Note that we reddened all
281: synthetic photometry by M\,15's $E_{(B-V)}=0.1$ mag (Harris 1996). As
282: expected, the BBs are located on a sequence going from blue (for hot
283: sources) to red colours (for cool sources). The cross on the sequence marks
284: $T_{eff} = 20200$ K, which is the temperature of a BB peaking at
285: $\lambda = 1550$~\AA, in the centre of the F150N filter.
286: The observed location of M\,15 X-2 in the CC diagram is
287: also marked and is clearly distinct from the blackbody sequence. The
288: reason is simply that the F150N filter is much narrower than the peaks
289: of the BB distributions. Thus, the latter cannot cause a strong excess
290: in this filter alone. We caution that true stellar spectra {\it can}
291: have turnovers more sharp than suggested from a comparison with BB
292: spectra, as the example of AC\,211 shows (Downes et al.\ 1996).   
293: 
294: The location of a power-law $F_{\lambda} \propto \lambda^{-2.0}$
295: spectrum is also marked in the CC diagram.  
296: In order to check how strong a line might be needed to account for the
297: observed flux excess, we have also carried out synthetic photometry of
298: power law spectra (with index -2.0) {\it and} an emission line at C\,IV 
299: $\lambda = 1550$~\AA\ with equivalent widths (EW) of 10~\AA\ to
300: 50~\AA. As can be seen, M\,15 X-2 is located close to the power-law +
301: C\,IV 1550~\AA\ sequence, but slightly above, which suggests additional He\,II
302: emission. We then carried out synthetic photometry of power law
303: spectra and C\,IV $\lambda = 1550$~\AA\ ($\rm{EW} \simeq 30$ \AA) {\it
304:   and} He\,II $\lambda = 1640$~\AA\ emission line with EWs of 10~\AA\
305: to 60~\AA. We conclude that the SED of M\,15 X-2 can be described by
306: a power-law $F_{\lambda} \propto \lambda^{-2.0}$ with an additional
307: C\,IV 1550~\AA\ and He\,II 1640~\AA\ emission line with $\rm{EW}
308: \simeq 30$~\AA\ each. However, given that there are three free
309: parameters in this model, a perfect match to just three colours is of
310: course guaranteed. Spectroscopy will be needed to confirm the spectral
311: shape and the existence of line emission.  
312: 
313: 
314: \section{Discussion}
315: \label{summary}
316: 
317: Knowledge of the orbital period of M\,15 X-2 allows us to derive a
318: more defined picture of this ultracompact system. Eggleton (1983)
319: showed that for small mass ratios $0.01 \le q = M_{2}/M_{1} \le 1$ the
320: mean density $\rho$ of the Roche lobe-filling companion becomes a
321: function of $P_{orb}$ mainly, i.e.\ $P_{orb} \times \rho^{1/2} \simeq
322: 0.438$. For an orbital period of 22.6 min, this gives $\rho = 786~
323: \rm{g}~ \rm{cm}^{-3}$ which is consistent with the mean density of a
324: low-mass white dwarf (WD). We can then use the mass-radius relationships
325: published by Deloye \& Bildsten (2003, their Fig.~4) to constrain the
326: minimum mass of the donor star to $0.02~\rm{M}_\odot \le M_{2,min}
327: \le 0.03~\rm{M}_\odot$ and its minimum radius to $0.03~\rm{R}_\odot
328: \le r_{2,min} \le 0.04~\rm{R}_\odot$, depending on composition. These
329: lower limits correspond to low-temperature donors. Using Kepler's 3rd
330: law and assuming a NS-dominated system mass near $1.4~\rm{M}_\odot$,
331: we estimate the binary separation to $\approx 2.1 \times 10^{10}$ cm.     
332: 
333: A blackbody of $T_{eff} \approx 32000$ K has a blue spectral slope
334: most similar to the one we fitted to M\,15 X-2 (see
335: Fig.~\ref{spect}). Placed at M\,15's distance, such a blackbody would
336: have a radius of $r_{bb} \approx 6.5 \times 10^{9}$ cm or $0.1\
337: \rm{R}_\odot$ if it is to have the same flux that we measured for
338: M\,15 X-2. This is larger than expected for the radius $r_{2}$ of the
339: degenerate companion, but comparable to the circularization radius of 
340: $r_{circ} \approx 0.2\ \rm{R}_\odot$ of the accretion disk.
341: We therefore conclude that the FUV light is coming from the accretion
342: disk rather than from the WD donor. 
343: This is consistent with the UCXB model of Arons \& King (1993)
344: in which the orbital modulation is then caused by the irradiation
345: of the WD donor. Using their Eq.~15 we then estimate the
346: inclination angle of the system $i \approx 34^{\circ}$. This face-on
347: inclination is consistent with the absence of eclipses. 
348: We note that no modulation can be seen in X-rays (Hannikainen et
349: al. 2005).
350:   
351: M\,15 X-2's X-ray luminosity was found to be $L_{X} \approx 1.4 \times
352: 10^{36}\ \rm{erg}\ \rm{s}^{-1}$ (White \& Angelini 2001, Hannikainen
353: et al.\ 2005). Assuming a 10 km radius and a mass of $1.4\
354: \rm{M}_\odot$ for the NS, this requires $\dot{M} > L_X R_\star / 
355: G M_\star \approx 10^{-10}\ \rm{M}_\odot\ \rm{yr}^{-1}$. This can be
356: compared to the accretion rate expected from conservative mass
357: transfer driven by angular momentum loss via gravitational radiation
358: in an UCXB  
359: \begin{equation}
360: \dot{M_{gr}} = 1.27 \times 10^{-8} \times q^{2} \times M_\star^{8/3}
361:   \times P_{orb}^{-8/3}(\rm{h}) \times (1+q)^{-1/3} \times (5/6 + n/2
362:   -q)^{-1}. \label{mdot}
363: \end{equation}
364: Taking our minimum donor mass and corresponding mass-radius index
365: $n \simeq -0.1$ (e.g. Deloye \& Bildsten 2003), we 
366: derive a lower limit of $\dot{M_{gr}} \approx 4 \times 10^{-10}\ \rm{M}_\odot
367: yr^{-1}$. This suggests that the observed X-ray emission can be
368: powered by gravitational radiation-driven mass transfer. We conclude
369: that M\,15 X-2 can be classified as an UCXB, only the third confirmed
370: such system in a GC. Our results are consistent with the idea that
371: indeed many GC LMXBs are UCXBs.
372: 
373: \acknowledgments
374: 
375: We thank Tom Maccarone, Tom Marsh, Geoff Daniell and Chris Deloye for
376: valuable discussions. This work was supported by NASA through grant
377: GO-9792 from the Space Telescope Science Institute, which is operated
378: by AURA, Inc., under NASA contract NAS5-26555. 
379: 
380: \begin{thebibliography}{}
381: \bibitem[]{401} Arons, J. \& King, I. R. 1993, ApJ, 413, L121	
382: \bibitem[]{402} Auri{\`e}re, M., Le F{\`e}vre, O. \& Terzan, A. 1984,
383:   A\&A, 138, 415 
384: \bibitem[]{403} Bailyn, C. D. \& Grindlay, J. E. 1987, ApJ, 316, L25
385: \bibitem[]{404} Bildsten, L. \& Deloye, C. J. 2004, ApJ, 607, L119
386: \bibitem[]{405} Boroson, B. et al. 2000, ApJ, 545, 399
387: \bibitem[]{407} Chakrabarty, D., Homer, L., Charles, P. A. \&
388:   O'Donoghue, D. 2001, ApJ, 562, 985 	
389: \bibitem[]{} Charles, P. A., Jones, D. C. \& Naylor, T. 1986, Nature
390:   323, 417 
391: \bibitem[]{408} Clark, G. W. 1975, ApJ, 199, L143
392: \bibitem[]{409} Davies, M. B., Benz, W. \& Hills, J. G. 1992, ApJ, 401, 246 
393: \bibitem[]{410} Deloye, C. J. \& Bildsten, L. 2003, ApJ, 598, 1217
394: \bibitem[]{411} De\,Marchi, G. \& Paresce, F. 1994, ApJ, 422, 597
395: \bibitem[]{412} Downes, R. A., Anderson, S. F. \& Margon, B. 1996,
396:   PASP, 108, 688 
397: \bibitem[]{414} Eggleton, P. P. 1983, ApJ, 268, 368  
398: \bibitem[]{415} Guhathakurta, P., Yanni, B., Schneider, D. P. \&
399:   Bahcall, J. N. 1996, AJ, 111, 267 
400: \bibitem[]{} Hannikainen, D. C., Charles, P. A., van Zyl, L. et
401:   al. 2005, MNRAS, 357, 325
402: \bibitem[]{416} Harris, W.E. 1996, AJ, 112, 1487
403: \bibitem[]{417} Haswell, C. A., Patterson, J., Thorstensen, J. R.,
404:   Hellier, C. \& Skillman, D. R. 1997, ApJ, 467, 847 
405: \bibitem[]{418} Homer, L., Charles, P. A., Naylor, T. et al. 1996,
406:   MNRAS, 282, L37   
407: \bibitem[]{419} Homer, L. 2003, KITP Workshop: ``The Physics of
408:   Ultracompact Stellar Binaries'' (Feb 1-2, 2003), Coordinators:
409:   L. Bildsten, D. Chakrabarty, G. Nelemans,
410:   http://online.itp.ucsb.edu/online/ultra\_c03/homer\/  
411: \bibitem[]{421} Ivanova, N., Rasio, F. A., Lombardi, J. C., Dooley,
412:   K. L. \& Proulx, Z. F. 2005, ApJ, 621, L109 
413: \bibitem[]{422} Katz, J. I. 1975, Nature, 253, 698
414: \bibitem[]{423} Rasio, F. A., Pfahl, E. D. \& Rappaport S. 2000, ApJ, 532, L47
415: \bibitem[]{424} Sirianni, M. et al.\ 2005, PASP, 117, 1049
416: \bibitem[]{427} Stella, L., Priedhorsky, W., \& White, N. E. 1987a,
417:   ApJ, 312, L17  
418: \bibitem[]{428} Stetson, P. B.\ 1991, in 3rd ESO/ST-ECF Garching -
419:   Data Analysis Workshop, eds. Grosb{\o}l P. J., Warmels R. H., p. 187  
420: \bibitem[]{429} Verbunt, F. 1987, ApJ, 312, L23
421: \bibitem[]{430} White, N. E. \& Angelini, L. 2001, ApJ, 561, L101
422: \end{thebibliography}
423: 
424: \begin{figure}
425: \epsscale{1.09}
426: \includegraphics[scale=0.565]{f1a.eps}
427: \includegraphics[scale=0.5]{f1b.eps}
428: \caption{Left: FUV image (taken in SBC/F140LP) of M\,15 X-2. North
429:   is up and east to the left. The field-of-view of this close-up is
430:   $\approx 8\farcs9 \times 8\farcs9$. The {\it Chandra} positions for
431:   AC\,211 and M\,15 X-2 are indicated. 
432:   Right: The same but for the NUV. Note
433:   the severe crowding in this image. Both the LMXB AC\,211 and the
434:   UCXB M\,15 X-2 stand out as bright objects in both
435:   images. \label{XB2}} 
436: \end{figure}
437: 
438: \begin{figure}
439: \epsscale{0.98}
440: \includegraphics[angle=270,scale=0.63]{f2.eps}
441: \caption{Lightcurves of M\,15\,X-2 (mean subtracted photometric data) with
442:   overplotted sine wave of 22.5806 min period. All six observing
443:   epochs are plotted on the same scale, note that the top panel
444:   displays the first observing epoch which lasted for 5 orbits, the
445:   bottom panel displays the following 5 epochs which lasted for one
446:   orbit each. \label{lightcurve}}
447: \end{figure}
448: 
449: \begin{figure}
450: \includegraphics[angle=270,scale=0.45]{f3_small.ps}
451: \caption{$\chi^{2}$ vs. frequency periodogram of M\,15\,X-2, the
452:   strong peak corresponds to the period of 22.5806 min. The inset
453:   figure shows a zoom on the peak frequencies at 63.77 d$^{-1}$. See
454:   the text for details. \label{power}}      
455: \end{figure}
456: 
457: \begin{figure}
458: \plotone{f4.eps}
459: \caption{Top panel: Spectral energy distribution of M\,15\,X-2. The
460:   dashed line corresponds to a power-law $F_{\lambda} \propto
461:   \lambda^{-2.0}$. Filter transmission curves 
462:   are overplotted as dotted lines. Bottom panel: Colour-Colour diagram for 
463:   BBs of different $T_{eff}$ (dots). The cross on the BB sequence
464:   marks $T_{eff} = 20200$ K, which is the temperature of a BB peaking
465:   at $\lambda = 1550$ \AA. The filled diamond denotes a power-law with
466:   index -2.0. The stars denote power law spectra (with index -2.0)
467:   with additional line emission with equivalent widths (EW) of 10\AA\
468:   to 60 \AA. M\,15\,X-2 is also marked. 
469:   See the text for details. \label{spect}} 
470: \end{figure}
471: 
472: \end{document}
473: 
474: