1: % File draft date:  2 Jan 2002

2: % \documentstyle[12pt,../aasms]{article}

3: %\documentstyle[12pt,emulateapj5]{article}

4: \documentclass[12pt,preprint,flushrt]{aastex}

5: %\doublespace

6: 

7: \begin{document}

8: \newcommand{\etal}{{\it et al.}}

9: \include{idl}

10: 

11: \title{Phase Coherent Timing of RX J0806.3+1527 with ROSAT and CHANDRA}

12: \author{Tod E. Strohmayer}

13: \affil{Laboratory for High Energy Astrophysics, NASA's Goddard Space Flight 

14: Center, Greenbelt, MD 20771; stroh@clarence.gsfc.nasa.gov}

15: 

16: \begin{abstract}

17: 

18: RX J0806.3+1527 is an ultra-compact, double degenerate binary with the

19: shortest known orbital period (321.5 s).  Hakala et al. (2003) have

20: recently reported new optical measurements of the orbital frequency of

21: the source which indicate that the frequency has increased over the

22: $\approx 9$ years since the earliest ROSAT observations.  They find

23: two candidate solutions for the long term change in the frequency;

24: $\dot\nu \approx 3$ or $6\times 10^{-16}$ Hz s$^{-1}$. Here we present

25: the results of a phase coherent timing study of the archival ROSAT and

26: Chandra data for RX J0806.3+1527 in the light of these new

27: constraints.  We find that the ROSAT -- Chandra timing data are

28: consistent with both of the solutions reported by Hakala et al., but

29: that the higher $\dot\nu = 6.1 \times 10^{-16}$ Hz s$^{-1}$ solution

30: is favored at the $\approx 97\%$ level. This large a $\dot\nu$ can be

31: accomodated by an $\approx 1 M_{\odot}$ detached double degenerate

32: system powered in the X-ray by electrical energy (Wu et al. 2002).

33: With such a large $\dot\nu$ the system provides a unique opportunity

34: to explore the interaction of gravitational radiation and

35: electromagnetic torques on the evolution of an ultracompact binary.

36: 

37: \end{abstract}

38: 

39: \keywords{Binaries: close - Stars: individual (RX J0806.3+1527, RX

40: J1914.4+2456) - Stars: white dwarfs -- X-rays: binaries -- Gravitational Waves}

41: 

42: \section{Introduction}

43: 

44: Double degenerate systems containing a pair of interacting white

45: dwarfs are the most compact binaries known. Evidence has been mounting

46: recently that the two systems with the shortest inferred orbital

47: periods, RX J1914.4+2456 (569 s, hereafter J1914) and RX J0806.3+1527

48: (321 s, hereafter J0806), may represent a new class of double

49: degenerates whose X-ray flux is powered by a unipolar induction

50: process rather than accretion (see Wu et al. 2002; Strohmayer 2002;

51: and Hakala et al. 2003). These objects have much in common (see

52: Cropper et al. 2003 for a recent review). Their X-ray lightcurves show

53: 100\% modulation with a sharp rise and more gradual decay, suggestive

54: of a small X-ray emitting region on the primary (Cropper et

55: al. 1998). In both systems the optical maximum lags the X-ray maximum

56: by about 0.6 of a cycle, consistent with the optical variations

57: resulting from X-ray heating of the secondary (Ramsay et al.  2000;

58: Israel et al. 2003).

59: 

60: Of the models that have been proposed for J1914 and J0806, the

61: Intermediate Polar (IP) scenario is looking increasingly unlikely (see

62: Cropper et al. 2003 for a discussion). All the remaining models are

63: variants of a double degenerate scenario. In the Polar model the

64: primary is magnetic, the binary is synchronized and the X-ray flux is

65: accretion driven (Cropper et al. 1998). Some difficulties with this

66: model are the lack of circular polarization in the optical (although

67: magnetic fields weak enough to preclude detection of polarization may

68: still be able to enforce synchronization), and the lack of any hard

69: ($> 2$ keV) spectral component, as is usually seen in Polars. Another

70: variant is the direct accretor model (Marsh \& Steeghs 2002; Ramsay et

71: al. 2002). In this model both components are non-magnetic, and the

72: accretion stream impacts directly onto the surface of the primary

73: (i.e. without the formation of an accretion disk). Shortcomings of

74: this model include the stable X-ray to optical phasing of the

75: lightcurves, and the need for a relatively low mass ($< 0.5

76: M_{\odot}$) primary.

77: 

78: An important constraint on the models can be obtained by measuring the

79: orbital evolution in these systems.  If the donors are degenerate,

80: then stable accretion will lead to a widening of the orbit and a

81: decrease in the orbital frequency, $\dot\nu$. From a timing study of

82: archival ROSAT and ASCA data Strohmayer (2002) found evidence that the

83: orbital frequency of J1914 is increasing at a rate consistent with

84: loss of orbital angular momentum to gravitational radiation. Recently,

85: Hakala et al. (2003) have used archival ROSAT and new optical timing

86: measurements of J0806 to constrain the orbital evolution. They find

87: that, similarly to J1914, the orbital frequency of J0806 is also

88: increasing, but at a rate that is factor of about 30 or more higher

89: than for J1914.  Considering the more compact orbit in J0806, and the

90: strong $\nu^{11/3}$ orbital frequency dependence of gravitational

91: radiation, a substantially higher $\dot\nu$ is not unexpected.

92: 

93: Hakala et al. (2003) used period measurements from ROSAT in 1994 -

94: 1995 as well as more recent optical data from the ESO VLT and the

95: Nordic Optical Telescope (NOT) on La Palma to explore the orbital

96: evolution of J0806. They derived a frequency history at three epochs

97: and deduced a frequency increase.  Here we report the results of a

98: coherent, phase timing study of J0806 using the 1994 - 1995 ROSAT data

99: combined with the public Chandra observations from November,

100: 2001. Hakala et al. (2003) did not use the Chandra data in their study

101: because it is too short in duration for a precision frequency

102: measurement. By themselves the ROSAT and Chandra data are insufficient

103: to unambiguously constrain the frequency evolution because of their

104: sparseness. However, in the light of the additional constraints on

105: $\dot\nu$ provided by the Hakala et al. study, we can use phase timing

106: analysis of the ROSAT and Chandra data to test for consistency with

107: and to see if the data favor one of the possible solutions found by

108: Hakala et al. (2003).  We find that the ROSAT and Chandra phase timing

109: data are consistent with the two possible $\dot\nu$ solutions found by

110: Hakala et al. (2003), and that the solution with the larger $\dot\nu

111: \approx 6 \times 10^{-16}$ Hz s$^{-1}$ provides a better fit to the

112: phase timing data and is preferred at about the $97\%$ confidence

113: level. We also briefly discuss the implications of this finding for

114: the models for J0806 and its cousin source J1914.

115: 

116: \section{Data Extraction and Analysis}

117: 

118: The ROSAT HRI observed J0806 for a total of $\approx$ 16 ksec in the

119: interval from October, 1994 through April, 1995. Burwitz \& Reinsch

120: (2001) reported on the pulse timing from these observations.  Chandra

121: observed J0806 for 20 ksec in November, 2001. Spectroscopy from this

122: observation is reported by Israel et al. (2003).  All these data are

123: now in the public domain.  We used standard HEASARC FTOOLS data

124: analysis packages (i.e., XSELECT) to extract and analyse the data.  We

125: began by producing images and extracting source events. We next

126: applied barycentric corrections to the event times for each

127: observation. We used the standard mission FTOOLS in conjunction with

128: the ROSAT and Chandra orbital and JPL (DE200) solar system ephemerides

129: (Standish et al. 1992). For the source position we used the

130: coordinates, ($\alpha = 08^h 06^m 23.2^s$, $\delta = 15^{\circ} 27'

131: 30.2''$ :J2000), obtained by Israel et al. (2002) from their study of

132: the optical-infrared counterpart to J0806. Our analysis resulted in a

133: total of 725 and 6076 photons from the ROSAT and Chandra observations,

134: respectively.

135: 

136: \subsection{Coherent Timing Methods}

137: 

138: We performed our coherent timing studies using the $Z_n^2$ statistic

139: (Buccheri 1983; see also Strohmayer 2002 and Strohmayer \& Markwardt

140: 2002 for examples of the use of this statistic in a similar

141: context). We use a two parameter phase model, with $\phi (t) = \nu_0

142: (t-t_0) + \frac{1}{2} \dot\nu (t - t_0)^2$. Since the methods are

143: described elsewhere we do not repeat the details here. From their

144: study of the ROSAT data, Burwitz \& Reinsch (2001) found two candidate

145: periods, 321.5393 or 321.5465 s, with an uncertainty of 0.4 ms. We

146: began by first analyzing just the ROSAT events and confirm these

147: results. In terms of frequencies they are 3.1100392 and 3.10997285

148: mHz, respectively, with an uncertainty of about 4 nHz.

149: 

150: From their analysis of the orbital frequency at three different epochs

151: spanning about 9.2 years Hakala et al. (2003) found two possible

152: solutions for $\dot\nu$.  Two solutions were possible because of the

153: ambiguity in the period obtained from the ROSAT observations.  We used

154: the ROSAT and Chandra data to perform a grid search in the $\nu -

155: \dot\nu$ plane in the vicinity of these two possible solutions in

156: order to determine whether the combined ROSAT -- Chandra data are

157: consistent with them and to see if perhaps one of them is favored.  We

158: calculated a $\chi^2$ statistic by comparing the model predicted

159: phases with the observed phases (see Strohmayer 2002 for details). We

160: had a total of 31 phase measurements, 10 from the ROSAT epoch and 21

161: from the Chandra data, giving us fits with 29 degrees of freedom. The

162: results of our grid search are summarized in Figure 1. Panel (a) shows

163: the $1\sigma$ confidence contours in the vicinity of $\dot\nu = 3.1

164: \times 10^{-16}$ Hz s$^{-1}$, while panel (b) shows the same

165: quantities in the vicinity of the larger $\dot\nu$ solution, $\dot\nu

166: = 6.1 \times 10^{-16}$ Hz s$^{-1}$. In each figure the best solution

167: and errors from the Hakala et al. study are marked by the solid and

168: dashed horizontal lines, respectively. We find that the ROSAT --

169: Chandra timing data have solutions which are consistent with both of

170: the Hakala et al. (2003) solutions. However, we also find that the

171: $\dot\nu = 6.11 \times 10^{-16}$ Hz s$^{-1}$ gives a better fit. This

172: solution has a minimum $\chi^2 = 33.1$, while the other $\dot\nu$

173: solution has a minimum $\chi^2 = 40.8$. The $\Delta\chi^2 = 7.7$

174: between the two solutions favors the higher $\dot\nu$ solution at

175: about the $97\%$ level. Although this level of significance does not

176: provide concrete proof, it is suggestive that the higher $\dot\nu$

177: solution is the correct one.

178: 

179: Figures 2a and 2b compare the phase residuals computed from the models

180: with $\dot\nu = 3.1 \times 10^{-16}$ and $6.1 \times 10^{-16}$ Hz

181: s$^{-1}$, respectively.  Interestingly, much of the improvement in the

182: fit results from the improvement of the residuals to the ROSAT

183: data. Although this may at first seem surprising, these data span an

184: interval of about 7 months. A $\dot\nu$ of $6 \times 10^{-16}$ Hz

185: s$^{-1}$ produces a phase advance on the order of $\approx 0.2$ cycles

186: over such a time span, so that the ROSAT data are in principle

187: sensitive to a $\dot\nu$ of this size. We note that without the {\it a

188: priori} constraints on $\dot\nu$ provided by the Hakala et

189: al. results, the ROSAT -- Chandra data can be fit by a number of

190: different combinations of $\nu$ and $\dot\nu$. Some of these ``alias''

191: solutions can be seen in Figure 1a and b. 

192: 

193: \subsection{Discussion and Implications}

194: 

195: The observations by Hakala et al. (2003) of the orbital frequency

196: history of J0806 provide strong evidence that the orbit is decaying at

197: a rate as predicted by the ultra-compact binary model in which the

198: loss of orbital angular momentum is dominated by gravitational

199: radiation and there is no mass exchange between the two stars.  We

200: have shown that phase coherent timing of the ROSAT and Chandra data

201: are consistent with the rate of orbital frequency increase deduced by

202: Hakala et al (2003), and that the ROSAT - Chandra data provide

203: suggestive evidence that the higher $\dot\nu$ value is the correct

204: one.

205: 

206: For a detached binary with a circular orbit the rate of change of the

207: orbital frequency due to gravitational radiation is (see for example,

208: Evans, Iben \& Smarr 1987; Taylor \& Weisberg 1989)

209: \begin{equation} 

210: \dot\nu_{gr} = 1.64 \times 10^{-17} \ \left ( \frac{\nu}{10^{-3}\; 

211: {\rm Hz}} \right )^5 \; \left ( \frac{\mu}{M_{\odot}} \right ) \; 

212: \left ( \frac{a}{10^{10} \; {\rm cm}} \right )^2 \;\; {\rm Hz} 

213: \ {\rm s}^{-1} \; ,

214: \end{equation}

215: where, $\nu$, $\mu$, and $a$ are the orbital frequency, reduced mass

216: and orbital separation of the components, respectively. If the orbital

217: decay results only from gravitational radiation losses and there is no

218: mass transfer, then the constraint on $\dot\nu$ implies a constraint

219: on the so called ``chirp mass,''

220: \begin{equation} 

221: \left ( \frac{M_{ch}}{M_{\odot}} \right )^{5/3}  = 

222: \left ( \frac{\mu}{M_{\odot}} \right ) \ \left ( 

223: \frac{m_1 + m_2}{M_{\odot}} \right )^{2/3}  = 2.7 \times 10^{16} \left ( 

224: \frac{\nu}{10^{-3} \ {\rm Hz}} \right )^{-11/3} \ \dot\nu \; .

225: \end{equation}

226: This constraint follows directly from equation (1) and the use of

227: Kepler's law to substitute for the orbital separation, $a$.  This

228: expression also assumes no spin - orbit coupling. If the stars remain

229: synchronized with the orbit, then the expression is modified to

230: account for the angular momentum required to spin up the components as

231: the orbit shrinks.  We can use these expressions to constrain the

232: component masses of J0806 under the assumption that the orbital

233: dissipation is dominated by gravitational radiation. Figure 3 shows

234: the $1\sigma$ contours in the mass plane for the two possible

235: $\dot\nu$ solutions. Contours calculated assuming no spin - orbit

236: coupling (solid curves), and full synchronization (dashed curves) are

237: also shown.  The dotted lines mark the region where the mass ratio, $q

238: = m_{sec} / m_{prim}$ is between 0.5 and 1. Binary evolution

239: calculations for ultra-compact, double degenerate systems suggest that

240: they preferrentially form with mass ratios $\sim 1$ (see Nelemans et

241: al. 2001, Napiwotzki et al. 2002). The inferred masses are consistent

242: with double degenerate system with a total system mass $\sim 1

243: M_{\odot}$ (see also Hakala et al. 2003).

244: 

245: The inferred rate of orbital evolution in J0806 is so large that it

246: opens up a unique opportunity for phase timing observations to track

247: its evolution in detail.  For example, at a rate of $6 \times

248: 10^{-16}$ Hz s$^{-1}$ the phase will advance by $\approx 0.3$ cycles

249: in only a year. A glance at Figure 2 confirms that such changes can

250: be easily measured with Chandra in less than a year.  

251: 

252: The inferred spin-up of the system is difficult to understand in the

253: context of an accreting double-degenerate system.  Stable accretion

254: requires that the orbital separation increases, contrary to the

255: observations. (see for example, Rappaport, Joss \& Webbink 1982;

256: Nelemans et al. 2001).  As noted by Strohmayer (2002) and Hakala et

257: al. (2003), the observation of spin-up in J1914 and J0806 can be

258: accomodated in the unipolar induction, or electric star model of Wu et

259: al. (2002). Indeed, for the large $\dot\nu$ inferred from J0806, it is

260: possible that a majority of the orbital dissipation may arise from the

261: electromagnetic torques.  It is likely that instabilities in this

262: process as, for example, caused by induced magnetic fields, will

263: introduce variability both in the X-ray flux and the electromagnetic

264: torque (Cropper et al. 2003). With precise, phase coherent timing, it

265: is possible that such variations could be detected, and the electric

266: star model explored in detail.  As noted by Hakala et al. (2003), this

267: would also be possible with independent optically derived constraints

268: on the component masses, and would also be facilitated by future

269: gravitational wave measurements (with, for example LISA).

270: 

271: \centerline{\bf References}

272: 

273: \noindent{} Buccheri, R. et al. 1983, A\&A, 128, 245

274: 

275: \noindent{} Burwitz, V. \& Reinsch, K. 2001, X-ray Astronomy: stellar

276: endpoints, AGN, and the diffuse X-ray background, Bologna, Italy, eds.

277: White, N. E., Malaguti, G., and Palumbo, G., AIP conference

278: proceedings, 599, 522.

279: 

280: \noindent{} Cropper, M., Ramsay, G., Wu, K. \& Hakala, P. 2003, in

281: Proc. Third Workshop on Magnetic CVs, Cape Town, (astro-ph/0302240).

282: 

283: \noindent{} Cropper, M. et al. 1998, MNRAS, 293, L57.

284: 

285: \noindent{} Evans, C. R., Iben, I. \& Smarr, L. 1987, ApJ, 323, 129.

286: 

287: \noindent{} Hakala, P. et al. 2003, MNRAS, in press (astro-ph/0305283).

288: 

289: \noindent{} Iben, I. \& Tutukov, A. V. 1991, ApJ, 370, 615.

290: 

291: \noindent{} Israel, G. L. et al. 2003, in Proc. Third

292: Workshop on Magnetic CVs, Cape Town, (astro-ph/0303124).

293: 

294: \noindent{} Israel, G. L. et al. 2002, A\&A, 386, L131.

295: 

296: \noindent{} Israel, G. L. et al. 1999, A\&A, 349, L1.

297: 

298: \noindent{} Marsh, T. R. \& Steeghs, D. 2002, MNRAS, 331, L7.

299: 

300: \noindent{} Motch, C. et al. 1996, A\&A, 307, 459.

301: 

302: \noindent{} Napiwotzki, R. et al. 2002, A\&A, 386, 957.

303: 

304: \noindent{} Nelemans, G., Portegies Zwart, S. F., Verbunt, F. \& Yungelson, 

305: L. R. 2001a, A\&A, 368, 939.

306: 

307: \noindent{} Ramsay, G., Cropper, M., Wu, K., Mason, K.~O., \& Hakala, P.\ 

308: 2000, MNRAS, 311, 75.

309: 

310: \noindent{} Ramsay, G., et al. 2002, MNRAS, 333, 575.

311: 

312: % \noindent{} Ramsay, G. Hakala, P. \& Cropper, M. 2002, MNRAS, 332, L7.

313: 

314: \noindent{} Rappaport, S., Joss, P. C. \& Webbink, R. F. 1982, ApJ, 254, 616.

315: 

316: \noindent{} Standish, E. M., Newhall, X. X., Williams, J. G. \& Yeomans, D. K

317:             1992, in Explanatory Supplememt to the Astronomical Almanac, ed. 

318:             P. K. Seidelmann (Mill Valley: University Science), 239.

319: 

320: \noindent{} Strohmayer, T. E. 2002, ApJ, 581, 577.

321: 

322: \noindent{} Strohmayer, T. E., \& Markwardt, C. B., 2002, ApJ, 577, 337.

323: 

324: \noindent{} Strohmayer, T. E., \& Markwardt, C. B., 1999, ApJ, 516, L81.

325: 

326: \noindent{} Taylor, J. H. \& Weisberg, J. M. 1989, ApJ, 345, 434.

327: 

328: \noindent{} Wu, K., Cropper, M., Ramsay, G. \& Sekiguchi, K. 2002, MNRAS, 

329: 331, 221.

330: 

331: \pagebreak

332: \centerline{\bf Figure Captions}

333: 

334: \figcaption[f1a.ps, f1b.ps]{Results of our ROSAT -- Chandra timing

335: study for J0806. (a) The $1\sigma$ confidence contours in the $\nu$,

336: $\dot\nu$ plane are shown in the vicinity of the lower $\dot\nu

337: \approx 3\times 10^{-16}$ Hz s$^{-1}$ solution of Hakala et

338: al. (2003). The best $\chi^2$ for this solution was $\chi^2_{min} =

339: 40.8$. The Hakala et al. solution and $1\sigma$ errors are marked by

340: the horizontal solid and dashed lines, respectively. The nearby,

341: ``alias'' solutions result from the sparseness of the ROSAT and

342: Chandra observations. (b) Same as for Figure 1a, but in the vicinity

343: of the $\dot\nu = 6\times 10^{-16}$ Hz s$^{-1}$ solution of Hakala et

344: al. (2003). This solution has a better fit with $\chi^2_{min} = 33.1$.

345: \label{f1}}

346: 

347: \figcaption[f2a.ps, f2b.ps]{Phase residuals from our ROSAT -- Chandra

348: timing study for J0806.  (a) The phase residuals from the best fit

349: consistent with the $\dot\nu \approx 3\times 10^{-16}$ Hz s$^{-1}$

350: solution of Hakala et al. (2003) are shown.  The individual phase

351: measurements are on the x-axis, and the vertical dotted line denotes

352: the ROSAT and Chandra epochs.  The long time gaps are essentially

353: removed in this representation. (b) Same as Figure 2a, but for the

354: $\dot\nu \approx 6\times 10^{-16}$ Hz s$^{-1}$ solution of Hakala et

355: al. (2003). This solution has the better overall $\chi^2$. Most of the

356: improvement comes from the better fit to the ROSAT phases.

357: \label{f2}}

358: 

359: \figcaption[f3.ps]{Constraints on the component masses of J0806

360: assuming that the orbital decay results only from gravitational

361: radiation losses. The $1\sigma$ constraints for the $6 \times

362: 10^{-16}$ Hz s$^{-1}$ (upper) and $3 \times 10^{-16}$ Hz s$^{-1}$

363: (lower) are shown.  In each case the solid and dashed curves are

364: derived assuming no spin - orbit coupling, and full synchronization,

365: respectively. The region between the dotted lines marks the phase

366: space for which $0.5 < q = m_{sec}/m_{prim} < 1.0$.

367: \label{f3}}

368: 

369: \pagebreak

370: 

371: \begin{figure}

372: \begin{center}

373:  \includegraphics[width=6in, height=6in]{f1a.ps}

374: \end{center}

375: Figure 1a: Results of our ROSAT -- Chandra timing study for J0806. (a)

376: The $1\sigma$ confidence contours in the $\nu$, $\dot\nu$ plane are

377: shown in the vicinity of the lower $\dot\nu \approx 3\times 10^{-16}$

378: Hz s$^{-1}$ solution of Hakala et al. (2003). The best $\chi^2$ for

379: this solution was $\chi^2_{min} = 40.8$. The Hakala et al. solution

380: and $1\sigma$ errors are marked by the horizontal solid and dashed

381: lines, respectively. The nearby, ``alias'' solutions result from the

382: sparseness of the ROSAT and Chandra observations. (b) Same as for

383: Figure 1a, but in the vicinity of the $\dot\nu = 6\times 10^{-16}$ Hz

384: s$^{-1}$ solution of Hakala et al. (2003). This solution has a better

385: fit with $\chi^2_{min} = 33.1$.

386: \end{figure}

387: \clearpage

388: 

389: \begin{figure}

390: \begin{center}

391: \includegraphics[width=6in, height=6in]{f1b.ps}

392: \end{center}

393: Figure 1b:

394: \end{figure}

395: \clearpage

396: 

397: \begin{figure}

398: \begin{center}

399:  \includegraphics[width=6in, height=6in]{f2a.ps}

400: \end{center}

401: 

402: Figure 2a: Phase residuals from our ROSAT -- Chandra timing study for

403: J0806.  (a) The phase residuals from the best fit consistent with the

404: $\dot\nu \approx 3\times 10^{-16}$ Hz s$^{-1}$ solution of Hakala et

405: al. (2003) are shown.  The individual phase measurements are on the

406: x-axis, and the vertical dotted line denotes the ROSAT and Chandra

407: epochs.  The long time gaps are essentially removed in this

408: representation. (b) Same as Figure 2a, but for the $\dot\nu \approx

409: 6\times 10^{-16}$ Hz s$^{-1}$ solution of Hakala et al. (2003). This

410: solution has the better overall $\chi^2$. Most of the improvement

411: comes from the better fit to the ROSAT phases.

412: \end{figure}

413: \clearpage

414: 

415: \begin{figure}

416: \begin{center}

417:  \includegraphics[width=6in, height=6in]{f2b.ps}

418: \end{center}

419: Figure 2b: 

420: \end{figure}

421: \clearpage

422: 

423: \begin{figure}

424: \begin{center}

425:  \includegraphics[width=6in, height=6in]{f3.ps}

426: \end{center}

427: Figure 3: Constraints on the component masses of J0806 assuming that

428: the orbital decay results only from gravitational radiation

429: losses. The $1\sigma$ constraints for the $6 \times 10^{-16}$ Hz

430: s$^{-1}$ (upper) and $3 \times 10^{-16}$ Hz s$^{-1}$ (lower) solutions

431: are shown.  In each case the solid and dashed curves are derived

432: assuming no spin - orbit coupling, and full synchronization,

433: respectively. The region between the dotted lines marks the phase

434: space for which $0.5 < q = m_{sec}/m_{prim} < 1.0$.

435: \end{figure}

436: 

437: \clearpage

438: 

439: 

440: %%%%%%%%%%%%%%%%%%%%%%%%%%%

441: %%%%% End of document %%%%%

442: %%%%%%%%%%%%%%%%%%%%%%%%%%%

443: 

444: \end{document}

445: