1: 

2: 

3: \documentclass[

4:   ,draft            % use draft while you are working on the paper

5:   ,mathptm                 % add further options here if necessary

6:   ]

7:   {aipproc}

8: 

9: \layoutstyle{8x11double}

10: 

11: 

12: \def\lesssim{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}

13: \def\gtrsim{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}

14: 

15: \newcommand{\apj}{ApJ}

16: \newcommand{\apjs}{ApJS}

17: \newcommand{\apjl}{ApJL}

18: \newcommand{\aj}{AJ}

19: \def\mnras   {{\it MNRAS}}

20: \def\pasj    {{\it PASJ}}

21: \newcommand{\fpk}{$F_{\rm peak}$}

22: \newcommand{\fpkre}{$F_{\rm peak,RE}$}

23: \newcommand{\fper}{$F_{\rm per}$}

24: \newcommand{\eps}{{\rm ergs\,s^{-1}}}

25: \newcommand{\epcs}{{\rm ergs\,cm^{-2}\,s^{-1}}}

26: \newcommand{\epc}{{\rm ergs\,cm^{-2}}}

27: \newcommand{\cts}{{\rm count\,s^{-1}}}

28: \newcommand{\leddh}{$L_{\rm Edd,H}$}

29: \newcommand{\leddhe}{$L_{\rm Edd,He}$}

30: \newcommand{\pasm}{P_{\rm ASM}}

31: \newcommand{\pflux}{P_{\rm peak}}

32: \newcommand{\ans}{{\it ANS}}	% Astronomical Netherlands Satellite

33: \newcommand{\vfb}{{\it Vela 5-B}}

34: \newcommand{\xte}{{\it RXTE}}	% Rossi X-ray Timing Explorer

35: \newcommand{\sax}{{\it BeppoSAX}}

36: \newcommand{\ros}{{\it ROSAT}}

37: \newcommand{\cgro}{{\it CGRO}}

38: \newcommand{\uhu}{{\it Uhuru}}

39: \newcommand{\exo}{{\it EXOSAT}}

40: \newcommand{\ein}{{\it Einstein}}

41: \newcommand{\asca}{{\it ASCA}}

42: \newcommand{\osos}{{\it OSO-7}}

43: \newcommand{\osoe}{{\it OSO-8}}

44: \newcommand{\sas}{{\it SAS-3}}	% Small Astronomy Satellite (MIT)

45: \newcommand{\arv}{{\it Ariel-5}}

46: \newcommand{\cop}{{\it Copernicus}}

47: \newcommand{\hak}{{\it Hakucho}}

48: \newcommand{\gin}{{\it Ginga}}

49: \newcommand{\chandra}{{\it Chandra}}

50: \newcommand{\xmm}{{\it XMM-Newton}}

51: \newcommand{\hst}{{\it HST}}

52: 

53: \begin{document}

54: 

55: \title{Thermonuclear burst physics with RXTE}

56: 

57: 

58: \author{Duncan K. Galloway}{

59:   address={Massachusetts Institute of Technology}

60: }

61: 

62: \author{Deepto Chakrabarty}{

63:   address={Massachusetts Institute of Technology}

64: }

65: 

66: \author{Andrew Cumming}{

67:   address={University of California, Santa Cruz}

68: }

69: 

70: \author{Erik Kuulkers}{

71:   address={ESTEC}

72: }

73: 

74: \author{Lars Bildsten}{

75:   address={University of California, Santa Barbara}

76: }

77: 

78: \author{Richard Rothschild}{

79:   address={University of California, San Diego}

80: }

81: 

82: 

83: \begin{abstract}

84: Recently we have made measurements of thermonuclear burst energetics and

85: recurrence times which are unprecedented in their precision, largely

86: thanks to the sensitivity of the {\it Rossi X-ray Timing Explorer}\/ ({\it

87: RXTE}). In the "Clocked Burster", GS~1826$-$24, hydrogen burns during the

88: burst via the rapid-proton (rp) process, which has received particular

89: attention in recent years through theoretical and modelling studies. The

90: burst energies and the measured variation of alpha (the ratio of

91: persistent to burst flux) with accretion rate strongly suggests solar

92: metallicity in the neutron star atmosphere, although this is not

93: consistent with the corresponding variation of the recurrence time.

94: Possible explanations include extra heating between the bursts, or a

95: change in the fraction of the neutron star over which accretion takes

96: place. I also present results from 4U~1746$-$37, which exhibits regular

97: burst trains which are interrupted by ``out of phase'' bursts.

98: \end{abstract}

99: 

100: \maketitle

101: 

102: 

103: \section{Introduction}

104: 

105: Unstable thermonuclear ignition of accreted fuel on neutron stars (NSs) in

106: low-mass X-ray binaries (LMXBs) is triggered once a critical column

107: density is reached in the fuel layer (e.g. \cite[]{bil00}).  Regular

108: bursting is surprisingly uncommon, and only one source is known to

109: consistently burst regularly (GS~1826$-$24; \cite[]{gal03d}).  The

110: conditions required for regular bursting likely include steady accretion

111: and uniform spreading of the accreted fuel over the NS surface, as well as

112: complete fuel consumption.  If the accretion rate is sufficiently steady

113: the critical density for ignition will be reached after a fixed time. If,

114: additionally, all the accreted fuel is burnt during each burst, then each

115: burst will be ignited after the same interval, leading to regular

116: bursting. Clearly, relaxation of any one of these conditions will lead to

117: variations in the burst interval, and deviations away from regular

118: bursting.

119: 

120: Here we present recent results obtained via measurement of thermonuclear

121: burst properties in GS~1826$-$24 and the globular cluster source

122: 4U~1746$-$37 with \xte. In GS~1826$-$24 we found regular bursting at a

123: range of accretion rates, which allows us to constrain the composition of

124: the burning fuel. In 4U~1746$-$37 we found trains of regular bursts

125: interrupted by bursts which were ``out of phase''. We discuss possible

126: mechanisms for this phenomenon, as well as future observational tests to

127: distinguish between them.

128: 

129: \section{Observations and analysis}

130: 

131: We obtained public \xte\/ data from the HEASARC archive at

132: \url{http://heasarc.gsfc.nasa.gov} and searched for bursts in 1-s binned

133: Standard-1 data. We found a total of 24 bursts from GS 1826$-$24, and 28

134: bursts from 4U~1746$-$37. For each burst we extracted full-range

135: (2--60~keV) time-resolved spectra every 0.25~s from high-time resolution

136: PCA data modes (GoodXenon or Generic Event, where available), and fitted

137: an absorbed blackbody to each spectrum after subtracting the persistent

138: emission. We estimated the observed bolometric flux from the blackbody fit

139: parameters, and accumulated the measured fluxes to derive the burst

140: fluence. We also extracted non-burst spectra from both PCA and HEXTE

141: (15--200~keV) and fitted to absorbed Comptonisation models in order to

142: estimate the persistent flux level, and hence the accretion rate

143: $\dot{M}$. Full details of the analysis procedures may be found in

144: Galloway et al. 2004, in preparation.

145: 

146: We also made simulations to predict ignition conditions for thermonuclear

147: bursts following \cite[]{cb00}.  We calculate the temperature profile of

148: the accumulating layer of hydrogen and helium, and adjust its thickness

149: until a thermal runaway occurs at the base. The temperature is mostly set

150: by hydrogen burning via the hot CNO cycle, and therefore the CNO mass

151: fraction $Z$, which we refer to as the metallicity. We assume a constant

152: level of flux from the crust, at $Q_{\rm crust}=0.1$ MeV per nucleon

153: \cite[]{brown00}.

154: 

155: \section{Results: GS 1826$-$24}

156: 

157: We observed bursts from GS~1826-24 over a factor of 1.7 range in

158: persistent intensity.  The bursts from a given epoch were consistent with

159: a single recurrence time, which varied with accretion rate as $\Delta

160: t\propto\dot{M}^{-1.05}$ (assuming that $\dot{M}$ is proportional to the

161: persistent intensity $F_{X}$; Fig. \ref{vstheory}).  All the bursts had

162: similar lightcurves, and exhibited long tails likely powered by rp-process

163: hydrogen burning.  The burst fluence increased by $\approx5$\% over the

164: observed range of $F_X$, and the ratio of persistent to burst fluence

165: $\alpha$ decreased by $\approx 10$\%. The mean value of

166: $\alpha=41.7\pm1.6$ is in the range expected for mixed H/He burning during

167: the bursts.

168: 

169: \begin{figure}

170:   \includegraphics[height=.5\textheight]{f5.eps}

171:   \caption{ Variation of the burst recurrence time ({\it upper

172: panel}) and the burst fluence ({\it lower panel}) as a function of the

173: estimated bolometric persistent flux in GS~1826$-$24, from \xte\/ measurements

174: between 1997--2002.

175: Error bars indicate the $1\sigma$ uncertainties. 

176: The curves show theoretical calculations for

177: a range of metallicities: $Z=0.02, 0.01, 0.003,$ and $0.001$. The solid

178: angle $(R/d)$ and gravitational energy have been chosen in each case to

179: match the observed fluence and recurrence time at $F_{\rm p}=2.25\times

180: 10^{-9}\ {\rm erg\,cm^{-2}\,s^{-1}}$. For $Z=0.02, 0.01, 0.003,$ and

181: $0.001$, this gives $R/d=13, 10, 8, 6\ {\rm km}\ @\ 10\ {\rm kpc}$, and

182: $Q_{\rm grav}=175, 196, 211, 215$ MeV per nucleon, where $Q_{\rm

183: grav}=GM/R$ is the gravitational energy from accretion.

184: Reproduced from \cite{gal03d}.

185:   \label{vstheory} }

186: \end{figure}

187: 

188: The decrease in $\alpha$ with $\dot M$ implies that stable

189: burning of hydrogen takes place between the bursts, and suggests solar

190: metallicity ($Z\approx 0.02$) in the accreted layer. Solar metallicity

191: models also give good agreement with the observed burst energies, while

192: low-metallicity models do not.

193: However, the relatively steep variation in $\Delta t$ with $\dot M$

194: suggests little variation in the fuel composition at ignition, which in

195: turn implies low CNO metallicity. Otherwise, hydrogen burning between the

196: bursts would lead to an increased H-fraction at ignition as $\Delta t$

197: became shorter, leading to (relatively) delayed ignition and increased burst

198: fluence.

199: 

200: There are several possible ways to reconcile the solar metallicity

201: models with the fluence and $\Delta t$ measurements. Firstly, studies of

202: timing and spectra of LMXBs indicate that $L_X$ is not always a good

203: indicator of $\dot M$ (e.g.~\cite[]{vdk90}), while we have assumed

204: $L_X\propto\dot M$ here. Secondly, extra heating of the accumulating layer would

205: act to reduce the critical mass and bring the observations and theory

206: into agreement. One possibility is that residual heat from the ashes

207: of previous bursts heats the layer \cite[]{taam93,woos03},

208: although time-dependent simulations are required to test this. Thirdly,

209: if the

210: fraction of the NS surface covered by fuel changes with

211: $\dot M$, the changing local accretion rate per unit area could also

212: reconcile the models and observations. We found that the blackbody

213: radius $R_{\rm bb}$ in the tail of the bursts decreased by $\approx $20\%

214: between the observed epochs. If this indicates a change in covering

215: fraction, it would almost be enough to explain the discrepancy. However,

216: the covering fraction decrease with $\dot M$ is opposite to the increase

217: suggested by \cite[]{bil00} to explain trends in burst properties.

218: 

219: \section{4U 1746$-$37}

220: 

221: A well-known dipper and burst source with an orbital period of 5.7~hr

222: \cite[]{sansom93,homer02}, 4U~1746$-$37 is located in the globular cluster

223: NGC~6441.  The peak fluxes of the 28 bursts in public \xte\/ observations

224: were bimodally distributed, with 15 bursts reaching fluxes between

225: (1.0--$3.5)\times10^{-9}\ \epcs$ (here we refer to these as ``faint''

226: bursts), and the remaining 13 peaking at between (4.3--$7.1)\times10^{-9}\

227: \epcs$ (``bright'' bursts).

228: 

229: \begin{figure}

230:   \includegraphics[height=.28\textheight]{example_lc.eps}

231:   \caption{

232: 2--60~keV intensity measured over the field of 4U~1746$-$37 on 1998

233: November 7--8. The regular dips in the lightcurve are due to occultations

234: of the source by the Earth due to the $\approx90$~min satellite orbit.

235: The triangles show the predicted times of bursts according to the

236: ephemeris determined from bursts \#12--15, 17 and 18. The crosses show

237: where regular bursts may have occurred, but could not be observed due to

238: data gaps.  

239:  \label{example} }

240: \end{figure}

241: 

242: The bursts appeared to occur regularly on two occasions.  On 1998 November

243: 7th we observed 5 faint bursts (as well as the tail of a 6th burst), each

244: separated by $\approx1.2$~hr, and with rather uniform properties (fluence,

245: peak flux, timescale etc.).  However, we also observed a bright burst

246: which was not consistent with a $\approx1.2$~hr recurrence time. Burst

247: \#16, which exhibited photospheric radius-expansion (PRE) and reached a

248: peak of $5.1\times10^{-9}\ \epcs$, occurred 1.84~hr after the previously

249: observed burst, which was inferred from the observation of a burst tail at

250: the end of the occultation (\#15; Fig.  \ref{example}).  Neglecting burst

251: \#16, the remaining bursts were consistent with a rather steady recurrence

252: time of $1.23\pm0.01$~hr, with an rms error between the observed and

253: predicted burst times of just 0.034~hr.  The actual recurrence time for

254: burst \#16 could have been as short as 35~min if a faint burst had

255: occurred 1.23~hr after burst \#15, although the source was occulted at the

256: time.

257: 

258: On 1996 October 25--27, when the source was substantially fainter, we also

259: observed two successive bursts on three instances, again with rather

260: uniform properties.  However, the majority of the bursts this time were

261: bright, reaching peak fluxes $\gtrsim 4.5\times10^{-9}\ \epcs$; and

262: secondly, the recurrence time was 3--3.3~hr. Again, we observed a regular

263: train of bursts with rather homogeneous properties, interrupted

264: this time by a faint burst

265: at a time which was not consistent with the approximately periodic

266: recurrence of the other bursts.

267: 

268: These deviations from patterns of otherwise regular bursting are puzzling

269: for two reasons. Firstly, the regular bursting indicates that the critical

270: column depth for burst ignition occurs on a regular basis; for the

271: out-of-phase bursts, what causes the fuel layer to ignite prior to

272: achieving the critical density?  Secondly, assuming constant $\dot{M}$ and

273: complete consumption of the available fuel in the regular bursts, the

274: fluence of the out-of-phase bursts appear to be inconsistent with the

275: amount of fuel accreted since the previous burst.

276: 

277: We suggest three possible explanations for the

278: observed burst properties:

279: 

280: \noindent {\bf 1. The bright bursts originate from 4U~1746$-$37, but the

281: faint bursts originate from a second bursting source not positionally

282: coincident with 4U~1746$-$37}.

283: Two bursting sources would naturally explain the bimodal distribution

284: burst properties, as well as the two distinct patterns of regular

285: bursting.  If the second source was located significantly off the

286: satellite aimpoint, the observed flux would be reduced due to the decrease

287: in collimator efficiency with increasing off-axis angle, thus explaining

288: the relative weakness of the fainter bursts.

289: However, neither a deep sky map produced from ASM scans of the region

290: around 1998 November 7--8

291: nor \chandra\/ observations detected a second source in the

292: field \cite[]{homer02}, although this could be because the second source

293: was quiescent at the time.

294: We also attempted to constrain the origin of the bursts by exploiting

295: the small differences in pointing between the different proportional

296: counter units (PCUs) comprising the \xte\/ PCA. We combined the count

297: rates from individual PCUs from 7 faint bursts observed with similar

298: spacecraft attitude, in order to increase the signal-to-noise ratio. We

299: found that the most probable region for the origin of the faint bursts

300: runs accross the field of view approximately NW to SE, centered on the

301: position of 4U~1746$-$37.

302: Thus, we found no evidence that the faint bursts originate from a different

303: position within the field of NGC~6441.

304: 

305: \noindent {\bf 2. The bright bursts originate from 4U~1746$-$37, and the

306: faint bursts originate from a second bursting source within NGC~6441}.

307: Given the low probability of positional coincidence for two unrelated

308: bursting LMXBs within $\approx 1^\circ$ of each other, combined with the

309: concentration of NSs in globular clusters, the most likely site

310: for a second burst source in the field of 4U~1746$-$37\ is NGC~6441. One

311: final piece of evidence comes from measurements of the asymptotic

312: blackbody radius for the bursts.  If the second bursting source was not

313: related to NGC~6441, then we would expect the distance to the source to be

314: different.  For two NSs emitting X-ray bursts at different

315: distances, we then expect that the blackbody radii should also be

316: significantly different. Instead, we find that the blackbody radii are

317: essentially identical for both classes of bursts.

318: This suggests the second source has approximately the same distance as

319: 4U~1746$-$37, and hence is most likely within NGC~6441.

320: 

321: \noindent{\bf 3. The bursts all originate from a single bursting source in

322: NGC~6441, 4U~1746$-$37.}

323: This hypothesis is somewhat difficult to accept, as discussed above,

324: because of the bimodal distribution of properties of the non-PRE bursts

325: from the area, the apparently bimodal distribution of burst recurrence

326: times, and the interruption of regular trains of one type of burst by

327: bursts of a second type.

328: 

329: One source could produce both types of bursts if it had undergone a

330: transition from the regime of unstable hydrogen ignition to that of

331: unstable He ignition.

332: In 1996 October for the steady bright

333: bursts from 4U~1746$-$37 we estimate $\alpha\approx50$, while in 1998

334: November the $\alpha$-value for the steady faint bursts was much higher, at

335: $\approx220$.  The former value is typical for bursts which burn mixed

336: H/He \cite[]{gal03d}, while the latter is more typical for pure He bursts.

337: Thus, the bright steady bursts may arise from mixed H/He burning triggered

338: by unstable H ignition at very low accretion rates, while the intermittent

339: faint bursts burn the resulting He ashes; while in 1998 November, when

340: $\dot{m}$ was higher, the regular faint bursts are triggered instead by

341: unstable He burning.  We estimate the accreted column at the time of

342: ignition for the regular bright bursts assuming cosmic abundances (giving

343: energy generation of $Q_{\rm nuc}=4.4$~MeV per nucleon) to be

344: $y\approx2.5\times10^{7}\ {\rm g\,cm^{-2}}$. This is consistent with a H

345: flash but is at the lower limit of the He ignition curve (e.g.

346: \cite[]{cb00}).  For the faint bursts, assuming they burn pure helium so

347: that $Q_{\rm nuc}=1.6$~MeV per nucleon, the column is

348: $y\approx5.2\times10^7\ {\rm g\,cm^{-2}}$ which is consistent with a He

349: flash.

350: 

351: Additional observations at different accretion rates are a crucial

352: test of such a mechanism for giving rise to the bursts observed so far.

353: If the properties of the two classes of bursts can be shown to vary

354: independently, this would strengthen the case for two separate sources.

355: High spatial resolution observations with {\it Chandra}\/ or {\it

356: XMM-Newton}\/ may then be used to test for two distinct origins for the

357: two classes of bursts.

358: 

359: 

360: A final point which presents additional difficulties for

361: understanding the properties of this source is the overall frequency of

362: the bursts.  The broadband persistent flux from the source on 1996 October

363: and 1998 November was $(0.268\pm0.012)$ and $(1.93\pm0.07)\times10^{-9}\

364: \epcs$, respectively. For a source distance of 11~kpc, these fluxes

365: correspond to a spherically-averaged accretion rate of 0.025 and $0.18\,

366: \dot{m}_{\rm Edd}$ (where we assume $\dot{m}_{\rm Edd}=8.8\times10^4\

367: {\rm g\,cm^{-2}\,s^{-1}}$). For a NS accreting solar metallicity

368: material with a hydrogen fraction $X=0.7$ at these rates the model

369: predicts recurrence times of 44 and 4.1~hr, respectively.  That the

370: observed recurrence times (3.1 and 1.22~hr) were so short in comparison

371: may be an indication that the accretion does not completely cover the NS

372: surface, so that the local accretion rate is much higher.

373: 

374: \section{DISCUSSION}

375: 

376: The precise measurements of burst properties possible with \xte\/ allow

377: tests of burst theory to unprecedented levels of precision. For both

378: sources discussed here, the bursting behaviour may be significantly

379: altered because accretion is not taking place over the entire NS

380: surface. Measurements of blackbody radii from burst spectra offer at best a

381: qualitative way to measure the area of accretion, unless the deviations of

382: the spectra from pure blackbodies can be accounted for and the effective

383: temperature measured accurately. This can likely only be achieved with

384: measurements by dedicated spectroscopic instruments like {\it Chandra}\/

385: or {\it XMM-Newton}.  Such studies, when combined with the extensive

386: archival observations of bursters accumulated by \xte\/ over its lifetime

387: are thus an excellent way to improve our understanding of burst physics.

388: Future observations in previously unexplored ranges of accretion rate will

389: also further constrain the burst physics.

390: 

391: 

392: 

393: 

394: \bibliographystyle{aipproc}   % if natbib is available

395: 

396: \begin{thebibliography}{9}

397: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

398: \providecommand{\enquote}[1]{``#1''}

399: \expandafter\ifx\csname url\endcsname\relax

400:   \def\url#1{\texttt{#1}}\fi

401: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi

402: 

403: \bibitem[{Bildsten}(2000)]{bil00}

404: {Bildsten}, L., \enquote{Theory and Observations of Type I X-Ray Bursts from

405:   Neutron Stars,} in \emph{Cosmic Explosions, the 10th Annual October

406:   Astrophysics Conference, Maryland, October 11--13 1999, AIP Conf. 522},

407:   edited by S.~{Holt} and W.~{Zhang}, AIP, Woodbury NY, 2000, pp.  359--369.

408: 

409: \bibitem[{Galloway} et~al.(2004)]{gal03d}

410: {Galloway}, D.~K., {Cumming}, A., {Kuulkers}, E., {Bildsten}, L.,

411:   {Chakrabarty}, D., and {Rothschild}, R.~E., \emph{\apj}, \textbf{601},

412:   466--473 (2004).

413: 

414: \bibitem[{Cumming} and {Bildsten}(2000)]{cb00}

415: {Cumming}, A., and {Bildsten}, L., \emph{\apj}, \textbf{544}, 453--474 (2000).

416: 

417: \bibitem[{Brown}(2000)]{brown00}

418: {Brown}, E.~F., \emph{\apj}, \textbf{531}, 988--1002 (2000).

419: 

420: \bibitem[{van der Klis} et~al.(1990)]{vdk90}

421: {van der Klis}, M., {Hasinger}, G., {Damen}, E., {Penninx}, W., {van Paradijs},

422:   J., and {Lewin}, W.~H.~G., \emph{\apjl}, \textbf{360}, L19--L22 (1990).

423: 

424: \bibitem[{Taam} et~al.(1993)]{taam93}

425: {Taam}, R.~E., {Woosley}, S.~E., {Weaver}, T.~A., and {Lamb}, D.~Q.,

426:   \emph{\apj}, \textbf{413}, 324--332 (1993).

427: 

428: \bibitem[{Woosley} et~al.(2004)]{woos03}

429: {Woosley}, S.~E., {Heger}, A., {Cumming}, A., {Hoffman}, R.~D., {Pruet}, J.,

430:   {Rauscher}, T., {Fisker}, J.~L., {Schatz}, H., {Brown}, B.~A., and

431:   {Wiescher}, M., \emph{\apjs}, \textbf{151}, 75--102 (2004).

432: 

433: \bibitem[{Sansom} et~al.(1993)]{sansom93}

434: {Sansom}, A.~E., {Dotani}, T., {Asai}, K., and {Lehto}, H.~J., \emph{\mnras},

435:   \textbf{262}, 429--434 (1993).

436: 

437: \bibitem[{Homer} et~al.(2002)]{homer02}

438: {Homer}, L., {Anderson}, S.~F., {Margon}, B., {Downes}, R.~A., and {Deutsch},

439:   E.~W., \emph{\aj}, \textbf{123}, 3255--3262 (2002).

440: 

441: \end{thebibliography}

442: 

443: \end{document}

444: 

445: \endinput

446: