astro-ph0204495/ms.tex
1: %\documentclass[12pt,preprint]{aastex}
2: \documentstyle[emulateapj,apjfonts,epsfig]{article}
3: 
4: \newcommand{\ergsec}{erg sec$^{-1}$}
5: 
6: \begin{document}
7: 
8: \title{Transient X-ray Binaries in Elliptical Galaxies}
9: 
10: \author{Anthony L. Piro}
11: \affil{Department of Physics, Broida Hall, University of California
12: 	\\ Santa Barbara, CA 93106; piro@physics.ucsb.edu}
13: %\email{piro@physics.ucsb.edu}
14: 
15: \and
16: 
17: \author{Lars Bildsten}
18: \affil{Institute for Theoretical Physics and Department of Physics,
19: Kohn Hall, University of California
20: 	\\ Santa Barbara, CA 93106; bildsten@itp.ucsb.edu}
21: %\email{bildsten@itp.ucsb.edu}
22: 
23: \begin{abstract}
24:  
25: {\it Chandra} observations of elliptical galaxies have revealed large
26: numbers of Low Mass X-ray Binaries (LMXBs) accreting at rates $>
27: 10^{-9} M_\odot$ yr$^{-1}$. One scenario which generates this
28: $\dot{M}$ from an old stellar population is nuclear driven mass
29: transfer onto a neutron star or black hole from a Roche lobe filling
30: red giant. However, in our Galaxy, most of these systems accrete
31: sporadically as transients due to a thermal instability in the
32: accretion disk. Using the common criterion for disk instability
33: (including irradiation), we find that this mode of mass transfer
34: leads to transient accretion for at least $75\%$ of the binary's
35: life. Repeated {\it Chandra} observations of elliptical galaxies
36: should reveal this population. The recurrence times might be very long
37: \mbox{($\sim$ 100-10,000 years} depending on the orbital period at the
38: onset of mass transfer), and outbursts might last for 1-100
39: years. Mass transfering binaries can also be formed in old populations
40: via interactions in dense stellar environments, such as globular
41: clusters (GCs). These tend to have shorter orbital periods and are
42: more likely stable accretors, making them apparently a large fraction
43: of the elliptical's LMXB population.
44: 
45: \end{abstract}
46: 
47: \keywords{accretion, accretion disks:
48: 	binaries: close --- galaxies: elliptical and lenticular ---
49: 	pulsars: individual: GRO J1744--28 ---
50: 	stars: neutron --- X-rays: binaries}
51: 
52: \section{Introduction}
53: 
54:   {\it Chandra } observations of nearby elliptical and S0 galaxies
55: (e.g. Sarazin, Irwin \& Bregman 2000; Blanton, Sarazin \& Irwin 2000;
56: Sarazin, Irwin, \& Bregman 2001;
57: Finoguenov \& Jones 2001; Angelini et al. 2001; Kraft et
58: al. 2001) often find $\approx $ 20 X-ray sources with luminosities
59: greater than $10^{38}$ \ergsec and even more with luminosities greater
60: than $10^{37}$ \ergsec. For a neutron star (NS) accretor, this
61: corresponds to accretion rates as large as $10^{-8} M_\odot$
62: yr$^{-1}$, implying a short lifetime compared to the age of the
63: stellar population. One way to generate such rapid accretion rates in
64: an old stellar population (neglecting formation of tight binaries in
65: dense environments, such as globular clusters) is to have a Roche
66: lobe-filling binary driven by nuclear expansion of a low-mass star on
67: the red giant branch (orbital periods are days or longer). Ritter
68: (1999) has shown an analytic solution for such a Low Mass X-Ray Binary
69: (LMXB), which is further simplified in ellipticals where the initial
70: companion masses should all be $\approx 0.9M_\odot$ for an old
71: population. 
72: 
73:   However, such wide binaries are subject to thermal instabilities in
74: the accretion disk (King, Kolb \& Buderi 1996; see Dubus, Hameury \&
75: Lasota 2001 for an up-to-date discussion) that cause transient
76: accretion.  Using the current instability criterion, we follow the
77: evolution of such an LMXB and find that over 75\% of its life is spent
78: as a transient. We thus expect that a large fraction of the field
79: LMXBs in elliptical galaxies are long-term transients. Our ability to
80: detect them with repeated {\it Chandra} observations depends on their
81: outburst durations, which can be as long as years.
82: 
83:  We overview the binary evolution scenario in \S 2 and summarize the
84: challenge of finding the bright, blue optical counterparts to these
85: LMXBs in elliptical galaxies. In \S 3, we apply the current
86: understanding of accretion disk instabilities in X-ray sources and
87: estimate the outburst recurrence times. In \S 4, we compare our work
88: to galactic systems, noting that the recently discovered galactic
89: transients with outburst durations in excess of ten years
90: (e.g. Wijnands 2002) might well be wide binaries. We close in \S 5
91: with our conclusions and a discussion of alternative binary scenarios
92: that lead to more stable mass transfer, especially relevant for those
93: LMXBs found in globular clusters (Sarazin, Irwin \& Bregman 2001;
94: Angelini, Loewenstein \& Mushotzky 2001).
95: 
96: \section{Binary Evolution with Red Giant Donors}
97: 
98:   We only consider binaries with initial orbital periods $\ge$1
99: day. If the binary is tighter ($P_{\rm orb} < 1$ day), rapid
100: accretion may be initiated due to orbital angular momentum loss from
101: magnetic braking, resulting in mass transfer while a low-mass star is
102: still on the main sequence (see Tauris \& Savonije 1999). Whenever
103: there are multiple parameters to choose, we always take the value that
104: maximizes the stability of the accretion disk. This assures that
105: our calculated fraction of the evolutionary lifetime spent as a
106: transient is a firm lower limit. 
107: 
108:   We use Ritter's (1999) analytical solution for the LMXB evolution,
109: which assumes that the donor star is Roche lobe filling on the red
110: giant branch and expanding due to an increasing luminosity. The mass
111: transfer is slow enough that the donor star always remains close to
112: thermal equilibrium; thus $R_2 = R_{2,R}$, where $R_2$ is the radius
113: of the donor star and $R_{2,R}$ is the radius of its Roche lobe. For
114: simplicity, we assume that mass and angular momentum are conserved,
115: which implies that at the end of the LMXB's evolution the NS (of mass
116: $M_1$) will have a mass of 1.7-2.1$M_\odot$. We assume that the
117: initial donor mass is $0.9 M_\odot$ with a population I metallicity,
118: consistent with recent determinations of the age and metallicity of
119: the predominant stellar populations in these elliptical galaxies
120: (Trager et. al. 2000). 
121: 
122:   Once the evolution is solved, all parameters of the binary can be
123: found in terms of the orbital period at the onset of mass transfer,
124: $P_i$, which is measured in days, and the donor's mass, $M_2$, which
125: is measured in solar masses. The mass transfer rate within the binary
126: is
127: \begin{eqnarray}
128:         - \dot{M_2}=6.4 \times 10 ^{-10}
129:                 M_\odot \textrm{ yr}^{-1}
130:                 \mbox{ }P_i ^{ 14 / 15 }
131:                 \left( \frac{M_2}{0.9} \right) ^{ - 7 / 3 }
132:                 \nonumber \\
133:                 \times \left( \frac{2.3 - M_2}{1.4} \right) ^{ - 14 / 5 }
134:                 \left( \frac{5}{3} \frac{1}{M_2} - \frac{2}{2.3-M_2}
135:                          \right) ^{-1},
136: 	\label{eqn_mdot}
137: \end{eqnarray}
138: where $1.4M_\odot$ is the initial NS mass and $2.3M_\odot$ is the
139: total mass of the binary. Since we will show in \S 3 that the disk is
140: likely unstable, we will not concern ourselves here with the implied
141: super Eddington accretion rates at large $P_i$ 
142: should this mass transfer rate proceed
143: onto the NS. Using equation ($\ref{eqn_mdot}$) the evolution of the
144: binary system is shown for varying $P_i$ by the
145: solid lines in Figure 1. The long dashed line is the boundary where
146: the hydrogen envelope of the donor has vanished, halting the
147: evolution at the ending orbital period noted.  The evolving orbital period is
148: \begin{eqnarray}
149:          P_{\rm orb} = P_i
150:                 \left( \frac{M_2}{0.9} \right) ^{-3}
151:                 \left( \frac{2.3 - M_2}{1.4} \right) ^{-3}, 
152: \end{eqnarray}
153: shown in Figure 2 by the solid lines. 
154: 
155:  For such an LMXB in an elliptical galaxy, we expect: (1) the orbit
156: must be wide in order to reach such high $\dot M$'s, and (2) the
157: reprocessing of the X-rays by the large accretion disk should make it
158: a bright, very blue object. At the Eddington luminosity, an X-ray
159: binary with $P_{\rm orb}=10$ days (much like Cygnus X-2) has $M_V=-2$
160: (van Paradijs \& McClintock 1994) and is blue ($B-V\approx 0,
161: U-B\approx -1$; van Paradijs \& McClintock 1995).  For NGC 4697 (at a
162: distance modulus of $\approx 30.35$; Tonry et al. 2001), this
163: corresponds to an apparent V magnitude of 28.4.
164:  
165:   Assuming that the LMXB is not associated with a globular cluster, the
166: brightest star which could confuse a search for the optical
167: counterpart would be a red giant branch star just about to reach
168: helium core ignition (a TRGB star). For metallicities $[{\rm
169: Fe/H}]<-1$, $M_V(TRGB)\approx -2.5$, whereas for $[{\rm Fe/H}]\approx
170: -0.7$, $M_V(TRGB)\approx -1.5$ (Saviane et al. 2000). For a slightly
171: metal-rich star (such as is likely the case for these ellipticals;
172: Trager et al. 2000), the peak brightness is even fainter,
173: $M_V(TRGB)\approx -1.0$ (Garnavich et al. 1994). Hence, the optical
174: counterpart to a wide LMXB will be brighter at $V$ (and especially so
175: in $B$ or $U$) than any red giant branch star in an elliptical galaxy.
176: 
177:  The surface brightness of an elliptical galaxy would make such a
178: search difficult in the central regions. However, beyond $\approx$ 2
179: arcminutes, the surface brightness at $V$ (in NGC 4697, for example) is
180: below the sky level (for HST) of 22-23 ${\rm Mag}/{\rm arc-sec^2}$
181: (Goudfrooij et al. 1984). In that limit, and using the measured {\it
182: Chandra} position of sources at such locations, such a wide (10 day)
183: binary would be detectable at B or V in a $5\times 10^4\ {\rm s}$
184: observation with the {\it Advanced Camera for Surveys} on HST.  Of
185: course, such optical counterparts are easy to find in the nearby bulge
186: of M31, where $V\approx 22$.
187: 
188: \section{Disk Instability, Transient Fraction, and Recurrence Times}
189: 
190:   If the disk is thermally stable and transfers matter to the NS at the
191: supplied rate, there is nearly a one-to-one relation between $P_i$ 
192: and the X-ray luminosity (see Figure 1) that would allow the 
193: orbital period distribution to be inferred from the luminosity
194: distribution (Webbink et al. 1983). However, this argument fails
195: because these wide binaries are almost always transient accretors. 
196: 
197:   The usual condition used for the disk instability is that some portion
198: of the disk is below the hydrogen ionization temperature ($\approx
199: 6500$ K). Since the effective temperature of the disk decreases with
200: increasing radius, it is usually adequate to check that the outer
201: radius of the disk is below the critical temperature. In X-ray
202: binaries, however, this temperature is fixed by irradiation 
203: from the central accreting source (van Paradijs 1996). The
204: surface temperature is then estimated as (Dubus et al. 2001) 
205: \begin{eqnarray}
206:         T_{\rm irr}^4 \approx
207: 		\frac{C \eta \dot{M_1} c^2 }{ 4 \pi \sigma R^2 }, 
208: \end{eqnarray}
209: where $R$ is the disk radius, $\sigma$ is the Stefan-Boltzmann
210: constant, $\eta=0.1$ is the conversion efficiency for accretion onto
211: the NS, and $C$ measures the fraction of the X-ray luminosity that
212: reaches the disk (and contains information on the irradiation
213: geometry). We take $C=5.7\times 10^{-3}$ (slightly larger than
214: the preferred value, $C=5\times 10^{-3}$ of Dubus et al. 2001) so as
215: to maximize the persistent fraction. The disk is assumed to extend out
216: to 70\% of the Roche lobe around the NS (King et al. 1997).
217: 
218:   Setting $T_{\rm irr}=6500 $ K allows us to solve for the critical time,
219: accretion rate, and orbital period at which the instability sets in as
220: a function of the donor's mass.  The critical mass transfer rate is then
221: \begin{eqnarray}
222:         - \dot{M}_{\rm 2,crit}  = 3.3 \times 10 ^{-9}
223: 		M_\odot \textrm{ yr}^{-1}
224: 		M_2 ^{ 14 / 9 }
225:                 % \nonumber \\
226:                 \left( 2.3 - M_2 \right) ^{ - 14 / 9 }
227: 		\nonumber \\
228:                 \times \left( \frac{5}{3} \frac{1}{M_2} - \frac{2}{2.3-M_2}
229:                          \right) ^{-10/3}, 
230: \end{eqnarray}
231: which is used for the instability boundary shown in Figures 1 and 2 as
232: the dotted line. The critical orbital period is 
233: \begin{eqnarray}
234:          P_{\rm orb,crit} = 5.5 \textrm{ days }
235:                 M_2^{ 7 / 6 }
236:                 \left( 2.3 - M_2 \right) ^{ - 5 / 3}
237: 		\nonumber \\
238: 		\times \left( \frac{5}{3} \frac{1}{M_2}
239: 			- \frac{2}{2.3-M_2} \right) ^{ -5 / 2 },
240: \end{eqnarray}
241: and this is used for the instability boundary in Figure 2.  The
242: fraction of time spent transient is then found (numerically to better
243: than a percent) as a
244: function of the initial orbital period,
245: \begin{eqnarray}
246:  	{\rm Transient \ Fraction} = -0.0500 \chi ^2 + 0.2385 \chi + 0.7655,
247: \end{eqnarray}
248: where $\chi = \log_{10} \left( P_i / {\rm d} \right)$.  For $P_i$ 
249: in excess of a day, the binary is transient at least $75\%$ of
250: the time, and this does not include binaries containing black holes
251: which are transient almost all the time (e.g. King et al. 1996). 
252: 
253:   We estimate a maximum recurrence time for outbursts by assuming the
254: outside-in model. We assume that the binary is quiescent until a
255: critical surface density in the outer accretion disk (at radius $R$)
256: has been reached (Dubus, Hameury,\&  Lasota 2001),
257: \begin{eqnarray}
258:         \Sigma_{\rm max} = 644 \textrm{ g cm}^{-2}
259:                 ( M_1/M_\odot )^{ - 0.37 }
260:                 ( R/R_\odot )^{ 1.11 },
261: \end{eqnarray}
262: where we assume that the viscosity parameter $\alpha=0.1$ and that
263: there is no irradiation of the disk in quiescence. This expression 
264: gives a maximum recurrence time
265: \begin{eqnarray}
266:         t_{\rm rec,max} =  15 \textrm{ yrs }
267:                 P_i ^{1.14}
268:                 M_2 ^{-3.88}
269:                 \left( 2.3 - M_2 \right) ^{-2.75}
270: 		\nonumber \\
271: 		\times
272:                 \left( \frac{5}{3} \frac{1}{M_2}
273:                         - \frac{2}{2.3-M_2} \right), 
274: \end{eqnarray}
275: that should be viewed with some caution given the uncertainties in the
276: accretion disk physics. 
277: This recurrence time is $\sim$ 100-10,000 years (depending on the initial
278: orbital period; see Figure 2) and is consistent with current estimates
279: from observations of galactic ms radio pulsar/He white dwarf binaries
280: (Ritter \& King 2002). Our estimate is maximal because it is possible
281: (and maybe even likely) that the LMXB may outburst before the outer disk
282: reaches the critical mass, such as an outburst initiated in the inner disk.  
283: 
284: \section{Comparison with Galactic Sources} 
285: 
286:   One difficulty with comparing these evolutionary models with
287: observations is the lack of long orbital period LMXBs in our galaxy.
288: Many persistent sources have low enough orbital periods ($\sim1-20$
289: hours) that they are probably not accurately described by the
290: evolutionary model used above. However, a lack of persistent sources
291: at $P_{\rm orb} > 1 $ day is in agreement with our calculations and
292: generally expected from previous work (for example King, Frank, Kolb \&
293: Ritter 1997).  Figure 1 shows that such NS binaries spend most of
294: their lifetime as transient rather than persistent sources ($>75\%$ of
295: the time).  Though all these relations are consistent, it is still
296: hard to claim that the instability criterion has been fully
297: scrutinized by the galactic population.
298: 
299:   Indeed, the outburst durations of these systems might be so long that
300: they appear to us as ``persistent'' sources (with accretion rates most
301: likely in excess of the mass transfer rate expected from binary
302: evolution) until they suddenly disappear. Indeed, one such source (and
303: maybe others; Wijnands 2002) has now been clearly identified:
304: KS~1731-260 (Wijnands et al. 2001; Rutledge et al. 2002). It was
305: accreting at $\dot M\approx 3\times 10^{-9}M_\odot {\rm yr^{-1}}$ for
306: at least a decade prior to its sudden decline in luminosity by a
307: factor of $>10^4$ (Rutledge et al. 2002). Analysis of the immediate
308: quiescent emission in terms of thermal emission from the NS surface
309: yielded constraints on the recurrence time from 300-1,000 years
310: (Wijnands et al. 2001; Rutledge et al. 2002). Until this system's
311: orbital period is found (current knowledge of the infrared counterpart
312: points to a possibly evolved companion; Orosz, Bailyn \& Whitman 2001)
313: we will not be able to say whether it is an LMXB of the type we are
314: discussing. However, its existence does point to a likelihood of
315: finding such similar systems in elliptical galaxies.
316: 
317:   One galactic binary that does have a long orbital period is GRO
318: J1744-28 (11.8 days; Finger et al. 1996).  This system is transient
319: and was shown by Bildsten and Brown (1996) to have a companion with
320: mass $M_2 \approx 0.334 M_\odot$ and core mass \mbox{$M_c \approx 0.22
321: M_\odot$} (if the companion is a population I metallicity star). When
322: this binary's evolutionary path is predicted and plotted, it is found
323: to be transient (as it is) with a maximum outburst recurrence time of
324: \mbox{$\sim$1,000 yrs} (see dot-dashed line in Figure 2), consistent
325: with a single outburst.
326: 
327: \section{Conclusions and Discussion}
328: 
329:   We have shown that the expected mass transfer scenario for low mass
330: X-ray binaries in the field of an elliptical galaxy (where the donors
331: must be $<M_\odot$) that yield $\dot M\approx 10^{-8} M_\odot
332: {\rm yr^{-1}}$ (as observed by {\it Chandra}) have thermally unstable
333: disks. Such binaries are expected to be transient accretors with
334: maximum recurrence times of 100-10,000 years for at least $75\%$ of
335: their lifetimes. When active, it appears likely that they would easily
336: reach the NS's Eddington limit.
337: 
338:   Our inability to calculate the duration of the bright outbursting
339: phase makes it difficult to predict how many such transients will be
340: identified in repeated {\it Chandra} visits. If the duty cycle is 
341: bright for ten years out of a thousand (like implied for KS 1731-260;
342: Rutledge et al. 2002), and there are 100 such binaries, then a year
343: later, roughly 10 would have faded into quiescence to be replaced by
344: new transients. This is not that far off from what was seen over a
345: five month interval in Cen A (Kraft et al. 2001), where four sources
346: (out of 120 exposed in both epochs) were either non-detectable
347: ($L_x<(1-2)\times 10^{36} \ {\rm erg \ s^{-1}}$) or appeared bright
348: ($L_x>4\times 10^{37} {\rm erg \ s^{-1}}$) in either epoch. There were a
349: larger number ($\approx 35$) which showed significant variability
350: (Kraft et al. 2001).
351: 
352:   Such comparisons are also complicated by the discovery that many of
353: the X-ray sources in ellipticals are coincident with GCs (Sarazin et
354: al. 2001; Angelini et al. 2001; White, Sarazin \& Kulkarni 2002). 
355: Other possible accretion scenarios then include accretion from
356: lower-mass main sequence stars that are formed in this dense stellar
357: environment (see Rasio, Pfahl \& Rappaport 2000 for a recent
358: discussion). These tight systems would most likely be persistent
359: accretors, thus apparently reducing the fraction of transients amongst
360: the detected sources. This might just be an impression however, as if
361: the duty cycle of the field transients is really 1\%, then the
362: underlying population is indeed larger than that in GCs. Namely, it is
363: possible that the GC population is only {\it apparently } a large
364: fraction because LMXBs in GCs are persistently bright. The census of
365: the much more numerous population of field transients can only be
366: taken over a few thousand years.
367: 
368:   We thank G. Dubus, J. Irwin, H. Ritter, and C. Sarazin for helpful
369: discussions and the anonymous referee for a very informative critique of
370: our work. This research was supported by NASA via grant NAG 5-8658
371: and by the NSF under Grants PHY99-07949 and AST01-96422. L. B. is a
372: Cottrell Scholar of the Research Corporation.
373: 
374: \begin{references}
375: 
376: \noindent
377: 	Angelini, L., Loewenstein, M. \& Mushotzky, R. F. 2001, \apjl,
378: 	557, L35
379:    
380: \noindent
381: 	Bildsten, L. \& Brown, E. F. 1996, \apj, 477, 897
382: 
383: \noindent
384: 	Blanton, E. L., Sarazin, C. L. \& Irwin, J., A., 2000, \apj, 552, 106
385: 
386: %\noindent
387: %	de Jong, J. A., van Paradijs, J. \& Augusteijn, T.
388: %	1996, \aap, 314, 484
389: 
390: \noindent
391: 	Dubus, G., Hameury, J.-M. \& Lasota, J.-P. 2001, \aap, 371, 251
392: 
393: %\noindent
394: %Dubus, G., Lasota, J.-P., Hameury, J.-M. \& Charles, P. 1999, \mnras,
395: %	303, 139
396: 
397: %\noindent
398: %	Faber, S. M., Wegner, G., Burstein, D., Davies, R. L., Dressler, A.,
399: %	Lynden-Bell, D. \& Terlevich, R. J. 1989, \apjs, 69, 763
400: 
401: \noindent
402: 	Finger, M. H., Koh, D. T., Nelson, R. W., Prince, T. A.,
403: 	Vaughan, B. A. \&  Wilson, R. B. 1996, \nat, 381, 291
404: 
405: \noindent
406: 	Finoguenov, A. \& Jones, C. 2001, \apj, 547, L107
407: 
408: \noindent
409: 	Garnavich, P. M., Vandenberg, D. A., Zurek, D. R., Hesser
410: 	\& J. E. 1994, AJ, 107, 1097
411: 
412: \noindent
413: 	Goudfrooij, P., Hansen, L., Jorgenson, H.E., Norgaard-Nielsen, H.U.,
414: 	de Jong, T. \& van den Hoek, L. B. 1994, A\&AS, 104, 179
415: 
416: %\noindent
417: %Frank, J., King, A. R. \& Raine, D. J. 1992, Accretion Power in
418: %	Astrophysics (Cambridge: Cambridge Univ. Press)
419: 
420: \noindent
421: 	King, A.R., Frank, J., Kolb, U. \& Ritter, H. 1997, \apj, 484, 844
422: 
423: \noindent
424: 	King, A. R., Kolb, U. \& Burderi, L. 1996, \apj, 464, 127
425: 
426: \noindent
427: 	Kraft, R. P., Kregenow, J. M., Forman, W. R., Jones, C. \&
428: 	Murray, S. S. 2001, \apj, 560, 675 
429: 
430: \noindent
431: 	Orosz, J. A., Bailyn, C. D. \& Whitman, K. 2001, Atel, No. 75 
432: 
433: \noindent
434: 	Rasio, F. A., Pfahl, E. D. \& Rappaport, S. 2000, \apj, 532, L47
435: 
436: %\noindent
437: %	Refsdal, S. \& Weigert, A. 1970, \aap, 6, 426
438: 
439: \noindent
440: 	Ritter, H. 1999, \mnras, 309, 360
441: 
442: \noindent
443: 	Ritter, H. \& King, A. R. 2002, The Physics of Cataclysmic Variables
444: 	and Related Objects, ASP Conference Series, in press
445: 
446: \noindent
447: 	Rutledge, R. E., Bildsten, L., Brown, E. F., Pavlov, G. G.,
448: 	Zavlin, V. E. \& Ushomirsky, G. 2002, \apj, submitted
449: 	(astro-ph/0108125) 
450: 
451: \noindent
452: 	Sarazin, C. L., Irwin, J. A. \& Bregman, J. N. 2000, \apj, 554, L101
453: 
454: \noindent
455: 	Sarazin, C. L., Irwin, J. A. \& Bregman, J. N. 2001, \apj, 556, 533
456: 
457: \noindent
458: 	Saviane, I., Rosenberg, A., Piotto, G. \& Aparicio, A. 2000, A\&A,
459: 	355, 966
460: 
461: \noindent
462: 	Tauris, T. M. \& Savonije, G. J. 1999, \aap, 350, 928
463: 
464: \noindent
465: 	Tonry, J. L., et al. 2001, ApJ, 546, 681
466: 
467: \noindent
468: 	Trager, S. C., Faber, S. M., Worthey, G. \& Gonzalez, J. J. 2000,
469: 	AJ, 120, 165
470: 
471: \noindent
472: 	van Paradijs, J. 1996, \apj, 464, 139
473: 
474: \noindent
475: 	van Paradijs, J. \& McClintock, J. E. 1994, \aap, 290, 133
476: 
477: \noindent
478: 	van Paradijs, J. \& McClintock, J. E. 1995, in 
479: 	``X-Ray Binaries'', ed. W.H.G. Lewin, J. van Paradijs \& E. P. J. van
480: 	den Heuvel (Cambridge: Cambridge), p. 58
481: 
482: \noindent
483: 	Webbink, R.F., Rappaport, S. \& Savonije, G. J. 1983, \apj, 270, 678
484: 
485: \noindent
486: White, R. E., Sarazin, C. L. \& Kulkarni, S. R. 2002, to appear in
487: \apj 
488: 
489: \noindent
490:        Wijnands, R., Miller, J. M., Markwardt, C., Lewin, W. H. G., \&
491:        van der Klis, M. 2001, \apj, 560, L159 
492: 
493: \noindent
494: 	Wijnands, R. 2002, in  "The X-ray
495: 	Universe at Sharp Focus", eds. S. Vrtilek, E. Schlegel and
496: 	L. Kuhi, ASP Conference Series, in press (astro-ph/0107600) 
497: 
498: \end{references}
499: 
500: \begin{figure}
501: \plotone{f1.eps} 
502: \figcaption{Mass transfer rate as a function of
503: time for LMXBs with red giant donors. The solid lines are the
504: evolutionary path of the binary given an initial orbital period when
505: Roche lobe overfilling begins (evolution is from left to right).
506: The dashed line is the boundary where the Hydrogen envelope of the
507: donor has vanished, thus halting the evolution. 
508: The dotted line is the boundary at which the
509: accretion disk instability begins.}
510: % \label{fig1}
511: \end{figure}
512: 
513: \begin{figure}
514: \plotone{f2.eps}
515: \figcaption{Estimated 
516: maximum recurrence time as a function of the orbital period.
517: The curves are the same as in Figure 1. The additional dot-dashed line is
518: the predicted evolutionary path of GRO J1744-28. The filled circle
519: denotes this binary's current orbital period of 11.8 days.}
520: % \label{fig2}
521: \end{figure} 
522: 
523: \end{document}
524: 
525: 
526: 
527: 
528: