Mon.Not.R.Astron.Soc.000,000–000(0000) Printed22January2013 (MNLATEXstylefilev2.2) IGR J19308+0530: Roche lobe overflow on to a compact object from a donor 1.8 times as massive (cid:63) E.M. Ratti1†, T.F.J. van Grunsven1,2, M.A.P. Torres1,3, P.G. Jonker1,2,3, 3 J. C. A. Miller-Jones4, J. W.T. Hessels5,7, H. Van Winckel6, M. van der Sluys2,8, 1 0 G. Nelemans2,6 2 1SRON, Netherlands Institute for Space Research, Sorbonnelaan 2, 3584 CA, Utrecht, The Netherlands n 2Department of Astrophysics/IMAPP, Radboud University Nijmegen, Heyendaalseweg 135,6525 AJ, Nijmegen, The Netherlands a 3Harvard–Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, U.S.A. J 4International Centre for Radio Astronomy Research - Curtin University, GPO Box U1987, Perth, WA 6845, Australia 5ASTRON, the Netherlands Institute for Radio Astronomy, Postbus 2, 7990 AA, Dwingeloo, The Netherlands 1 6Instituut voor Sterrenkunde, K.U.Leuven, Celestijnenlaan 200D, B-3001 Leuven, Belgium 2 7Astronomical Institute ”Anton Pannekoek,” University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands 8Nikhef National Institute for Subatomic Physics, Science Park 105, 1098 XG Amsterdam, The Netherlands ] E H 22January2013 . h p - ABSTRACT o Wepresentphase-resolvedspectroscopy and photometryof theoptical counterpart to r t the X–ray binary IGR J19308+0530. Ellipsoidal modulations in the lightcurve show s a thattheF-typecompanionstarinthesystemisRoche-lobefilling.Theopticalspectra [ are dominated by absorption features from the donor star, with ∼10-20% disc contri- butiontotheopticalcontinuum.Wemeasureanorbitalperiodof14.662±0.001hours, 1 aradialvelocitysemi-amplitudeforthecompanionstarofK =91.4±1.4kms−1anda v 2 rotationalbroadeningofvsini=108.9±0.6kms−1.FromK andvsini,giventhatthe 6 2 donor star is filling its Roche lobe, we derive a mass ratio of q =M /M = 1.78±0.04, 9 2 1 8 which is typically considered to be too large for stable Roche-lobe overflow. Our ob- 4 servations support an inclination of ∼50 degrees. The accretor in IGR J19308+0530 . is most likely a white dwarf, although a neutron star cannot entirely be excluded. 1 0 Key words: stars: individual (IGR J19308+0530) — accretion: accretion discs — 3 stars: binaries — X-rays: binaries 1 : v i X r 1 INTRODUCTION jority of accreting WDs in binaries, in fact, belongs to the a class of cataclysmic variables (CVs), which have late type Intermediate-mass X–ray binaries (IMXBs) are binary sys- secondaries (M (cid:46) M ) and P (cid:46) 6 hours (Knigge et al. temswhereacompactobject-blackhole,neutronstar(NS) 2 1 2011). NSs or BHs, instead, are typically observed in X– or white dwarf (WD) - is accreting matter from a compan- ray binaries (XRBs) hosting either a massive O-B donor ionstarofspectraltypeAorF.IMXBsarerarelyobserved star (high-mass XRBs, HMXBs) driving accretion via stel- (see, e.g., the catalogue from Liu et al. 2007 and the 2012 lar wind, or a late M or K dwarf secondary star (low-mass versionoftheRKcatalogueRitter&Kolb20031).Thema- XRBs,LMXBs)accretingviaRochelobeoverflow.Therea- son for the observed rarity of IMXBs especially among NS andWDsystemsisthat,whenthecompanionismoremas- (cid:63) Observations made with the HERMES spectrograph at the sivethantheaccretorbutnotmassiveenoughtohavestrong MercatorTelescope,operatedonLaPalmabytheFlemishCom- winds,windaccretionproceedsataverylowrateandRoche- munity, at the Spanish Observatorio del Roque de los Mucha- lobe accretion is thought to be unstable. For NSs and WDs chosoftheInstitutodeAstrofsicadeCanarias.HERMESissup- in IMXBs mass flows from the more massive to the lighter ported by the Fund for Scientific Research of Flanders, the Re- starandangularmomentumconservationshrinkstheorbit, searchCouncilofK.U.LeuvenandtheFondsNationalRecherches leadingtoenhancedmasstransfer.ThebrightX-raybinary Scientific (FNRS), Belgium, the Royal Observatory of Belgium, phaseisthereforeintenseandshort-lived,causinganobser- the Observatoire de Gen´eve, Switzerland and the Thringer Lan- dessternwarteTautenburg,Germany. vational bias towards LMXBs, CVs and HMXBs (Tauris & † email:[email protected] vandenHeuvel2006).Nevertheless,IMXBscouldbealarge 1 http://physics.open.ac.uk/RKcat/ (cid:13)c 0000RAS 2 Ratti et al. fractionoftheXRBpopulationandhaveanimportantrole inunderstandingtheirevolution(Podsiadlowskietal.2001). CygX–2andHerX–1arethoughtohavestartedasIMXBs, even though the measured mass ratio is currently <1. IGR J19308+0530 was discovered by INTEGRAL (Bird et al. 2006) and observed by Swift (Rodriguez et al. 2008). An association with the star TYC486-295-1, classified as an F8 star in the survey by McCuskey (1949), was made using the Swift position (Rodriguez et al. 2008). This was confirmed using an accurate Chandra position of the X- ray source (Ratti et al. 2010). Considering typical param- eters of an F8 star, Rodriguez et al. (2008) suggested IGR J19308+0530 to be a L/IMXB in quiescence or a CV at a distance of (cid:46)1 kpc. Here, we present phase-resolved optical spectroscopy and photometry of IGR J19308+0530, in order to measure the orbitalperiodP,theradialvelocitysemi-amplitudeK and Figure1. OptimalsubtractionofthespectrumoftheF4Vstellar 2 the projected rotational velocity vsini of the companion templateHD185395fromthespectrumofIGRJ19308+0530.The star, and the system inclination i. In a Roche lobe filling imageshowsthepartofspectralregionS2usedtomeasurevsini. The interstellar feature at ∼6280 ˚A was masked. The template system,K andvsiniallowustoinfertheratioq=M /M 2 2 1 andtargetspectrumareoffsetby0.4and0.2iny. between the mass of the secondary and the primary star in the system (Wade & Horne 1988, see also Gray 1992) and, knowing P and i, to solve the system mass function. not taken due to non-photometric weather. After debias- ing and flat-fielding the images using dome flat-field obser- vations (with standard routines in IRAf) the instrumental 2 OBSERVATIONS AND DATA REDUCTION magnitudesofIGRJ19308+0530andfourcomparisonstars were computed by means of aperture photometry. Differen- In total twenty-two high-resolution spectra of tial lightcurves were then obtained for IGR J19308+0530 IGR J19308+0530 were collected. Observations were with respect to the comparison star TYC 486-968-1. We made on two nights in 2010 Mar., one night in 2010 alsoextractedlightcurvesforTYC486-968-1usingtheother Apr.and9nightsin2010Jun.usingthefiberspectrograph comparison stars. No significant variability was detected High Efficiency and Resolution Mercator Echelle Spec- in the i(cid:48)-band (r.m.s.(cid:46) 0.006 mag), whereas for the other trograph (HERMES), mounted at the Mercator telescope bands a larger scatter was observed due to weather condi- in La Palma (Raskin et al. 2011). The typical exposure tions (clouds and Calima). The maximum departure from time was 1200 seconds. The spectra have a dispersion of the mean value was of 0.1 mag. Given that the ellipsoidal 0.027 ˚A/pixel at 5000 ˚A and cover the wavelength range modulation in IGR J19308+0530 is small in amplitude we 3770-7230 ˚A. The fiber aperture is 2(cid:48).(cid:48)5 on the sky, but decided to model only the i(cid:48)-band lightcurves. the presence of a slicer mimics a narrow slit providing a resolution of ∼85000 irrespective of the seeing. The templatestarHD185395,ofspectraltypeF4V(laterfound to be the closest match to the spectral type of the target, 3 ANALYSIS AND RESULTS see below) was observed with the same settings for 360 3.1 Spectroscopy seconds on 2009 Aug. 5. The extraction of the spectra was performedthroughthededicatedautomateddatareduction The spectra are dominated by absorption features from the pipeline HermesDRS. For each spectrum we selected two secondary star (??). Although no emission line is directly regions for the analysis, one covering the Hγ and Hβ lines visible, residual emission in the Balmer lines appears when (4280-5250 ˚A) and one around the Hα line (5950-6700 ˚A), subtracting the spectra one from another, after correcting which we will refer to as S1 and S2, respectively. These for the orbital shift of the lines. The emission component regions were selected as they are rich in stellar lines, with is variable in intensity and wavelength and slightly shifted little contamination from interstellar features. A good with respect to the absorption line. fit of the continuum was achieved in each region with a Using the package molly developed by T.Marsh, we polynomial function of order 9. We normalized the spectra measured the orbital velocity of the companion star in dividing by the polynomial fit. IGR J19308+0530 by cross-correlating the spectra of the We also performed time-resolved photometry of targetwiththatofthetemplatestar.Astheabsorptionlines IGR J19308+0530, with the 80-cm IAC80 telescope at arerotationallybroadenedinthetargetspectrum,abroad- the Observatorio del Teide in Tenerife equipped with the ening of 100 kms−1was applied to the template (see below CAMELOT CCD imager. The observations were obtained on rotational broadening), improving the cross-correlation. duringpartofthreenightsbetween2010Jul.30andAug.02, The Balmer lines and interstellar features were masked. by cycling through the Sloan g(cid:48), r(cid:48), i(cid:48) and z(cid:48)-band filters. We performed a fit of the velocities versus time with a Three consecutive exposures were obtained in each filter, sine function, with K , the systemic radial velocity γ, T , 2 0 with integration times ranging from 6 to 24 sec depend- and P as free parameters. T was constrained to be near 0 ing on the filter and seeing conditions. Standard stars were the middle of the time span over which the observations (cid:13)c 0000RAS,MNRAS000,000–000 The mass ratio in IGR J19308+0530 3 spectraaroundphase0.5,suggestinglittleirradiationonthe inner face of the companion star. To measure vsini, we compared the spectrum of the template star with the Doppler-corrected average of the IGR J19308+0530 spectra between phase 0.9 and 0.1. The observed full-width at half-maximum (FWHM) of the absorption lines in the target spectra is determined by the intrinsic line width (expected to be dominated by vsini),broadenedbytheinstrumentalresolutionprofileand smearedbythemotionofthecompanionstarduringthein- tegration time of one observation. In order to account for thesmearing,wemadeasmanycopiesofthetemplatespec- Figure 2. Radial velocity curve from the absorption lines in trum as the number of target spectra we used for the av- the IGR J19308+0530 spectra (region S1) and its best fitting erage and we artificially smeared each copy of the template sinusoid(solidline).Forcomparisonweplotcircularorbitswith by 2πTKcos(2πφ)/P, where T is the duration of one ex- K2 =112kms−1(dotted line) and K2 =140kms−1(dashed line), posure on IGR J19308+0530 and φ the phase of one of the providingq=1.4andq=1respectivelyfortheobservedvsini. IGR J19308+0530 spectra we averaged. After that, we av- eraged the smeared template and broadened the resulting spectrum with different values of vsini. For each vsini, we were taken, and such that phase 0 is at the inferior con- performedanoptimalsubtractionofthebroadenedtemplate junction of the companion star. The best-fitting sinusoid fromtheaveragedspectrumofIGRJ19308+0530:againthe provided a χ2 of 57.6 for the region S1, and 41.76 for S2 broadeningwhichgivestheminimumχ2 providesameasure (19 d.o.f.). In both cases, the uncertainties on the param- of the actual vsini (Figure 1). A value of 0.5 was assumed eters were estimated assuming that the sinusoidal model forthelimbdarkening.Wemaskedinterstellarfeaturesand was correct, and we scaled the errors on the velocities to the lines from the Balmer series. reach a reduced χ2 of ∼1. The values of P, T and γ mea- In order to estimate the uncertainty on vsini, we in- 0 sured in S1 and S2 are consistent at the 1σ level, K is cludedthisprocedureinaMonteCarlosimulation,following 2 consistent at the 2 σ level. The error-weighted average of Steeghs & Jonker (2007). We copied each target spectrum the parameters gives P = 0.61092±0.00003 days, T = 500times,usingabootstraptechniquewheretheinputspec- 0 2455330.8169±0.0023 HJD/UTC, K = 91.4±1.4 kms−1 trum is resampled by randomly selecting data points from 2 and γ = −18.5±0.9 kms−1. Figure 2 shows the radial ve- it. The bootstrapping maintains the total number of data locity curve (rvc) folded on the above period. The value points in the spectrum. For each bootstrap copy, one value of γ is in the reference frame of the template star used of vsini is measured as described above. The distribution for the cross-correlation, whose systemic radial velocity is ofvsiniobtainedfromthe500copiesiswelldescribedbya −28±0.9 kms−1(Wilson 1953). The systemic radial veloc- Gaussian,whosemeanandr.m.s.providesthebest-fitvsini ity of IGR J19308+0530 is therefore −46.5±1.2 kms−1. andits1σerror.Astemplateandtargetspectraareacquired We obtained a set of high signal-to-noise template with the same instrument, the instrumental resolution pro- spectra from the UVES Paranal Observatory Project file is not affecting our measurement. (UVESPOP,Bagnuloetal.2003)ofA,FandearlyGstars The weighted average of the results from S1 and S2, con- of luminosity classV, III and IV. We subtractedeachtem- sistent at the 1σ level, is vsini=108.9±0.6kms−1. With plate spectrum from the Doppler-corrected average of the vsini and K2, we calculated the system mass ratio q = IGR J19308+0530 spectra between orbital phase 0.9 and M /M from the relation vsini = (1+q) 0.49q2/3 2 1 K2 0.6q2/3+ln(1+q1/3) 0.1. The reason for choosing this range in phase is that the (Horne et al. 1986), obtaining q=1.78±0.04. oblate shape of the Roche-lobe filling companion star and the possible presence of irradiation from the compact ob- ject could cause asymmetries in the line profiles, which are 3.2 Ellipsoidal modulation and system inclination minimised close to phase 0. We performed a χ2 test on the residuals of the subtraction: the template resulting in the Figure3showsthei(cid:48)-bandlightcurveforIGRJ19308+0530 minimumχ2 providesourbestestimateforthesourcespec- obtained by phase folding the CAMELOT data on the traltype.Inparticular,weadoptedtheoptimalsubtraction ephemeris determined in Section 3.1. The lightcurve dis- procedure implemented in molly, where the templates are plays the typical signature of ellipsoidal variation, with multiplied by a factor 0<f< 1 before the subtraction, rep- two unequal minima, but in addition asymmetric maxima resenting the fractional contribution of light from the sec- (O’Connell 1951 and Wilsey & Beaky 2009). ondarystar(1minusthediscveiling).Thefactorf isfound Wemodeledthei(cid:48)-bandlightcurveusingtheXRbinarypro- by minimizing the difference between the residuals and a gramwrittenbyE.L.Robinson.Areasonablefittothedata smoothedversionofitself.Beforedoingthesubtraction,the (reducedχ2 ∼7,403d.o.f.)isobtainedwithamodelassum- UVESPOP spectra were shifted to the rest frame of the ing an F4V secondary star and including 30% disc contri- average target spectrum, and degraded to match the sam- bution to the total light plus a disc hot-spot at phase 0.75. pling and line broadening of the latter. The disc veiling in the model is larger than observed in the The procedure favors an F4-F6V companion star, with a spectra, but variability is possible as the photometry was disc veiling of 10−20%. The same spectral type and veil- performed one month after the last spectrum was acquired. ing are obtained when considering an average of the target Similar models with different assumptions about the disc (cid:13)c 0000RAS,MNRAS000,000–000 4 Ratti et al. 4 DISCUSSION AND CONCLUSION We performed a dynamical study of the system IGR J19308+0530 through optical spectroscopy and photometry. The optical spectra are dominated by the companion star, with no evidence of irradiation, no emission features visible from the accretion flow besides a partial filling in of the Balmer lines and 10 − 20% disc contribution to the continuum. The secondary star is most likely of spectral type F4-6V. Ellipsoidal mod- ulations are detected on the ∼14.6 hour orbit. From phase-resolved spectroscopy we measure an extreme value for the binary mass ratio of q =1.78±0.04. The lightcurve modeling provides a reasonable fit to the data with a disc+hot-spot model at i∼52◦. Solving the mass function f(M ) : M sin3i = PK23 =0.03M with this inclination, 2 1(1+q)2 2πG (cid:12) Figure 3. i(cid:48)-band CAMELOT lightcurve, folded using the weobtainthefollowingindicativemasses:M1 ∼0.8M(cid:12) and ephemerisfromtheradialvelocitydata.Thecoloursindicatethe M2 ∼1.4M(cid:12).ThemassesareconsistentwithaWDaccretor threeobservingnights(Jul.30inred,Jul.31ingreenandAug.1 and an F4V donor (typical mass of ∼1.37M(cid:12), Mamajek’s inblue).Thedrawnlineshowsafitwithellipsoidalmodulations list 20112). If the inclination is lower, which we cannot plusadiscwithabrightdiscspot. exclude based on these data, the masses will increase, allowing a scenario with a NS primary if i(cid:46)45◦. However, in this case the companion star would be over-massive for the spectral type, which is unusual for XRB and CVs. Assuming an F4V mass donor with a radius equal properties(discradius,heightandtemperatureprofile)also to that of the Roche lobe for our best estimated masses, givereasonablefits.Aswehavenoindicationstosingleout the magnitude of the companion is M = 2.7 in the vi- onepreferredsetofparameters,wedonotprovideaformal V sual band (half a magnitude brighter than for a typical uncertaintyoni,butanindicativevalueof∼52◦asaguide. F4V star, Mamajek 2011). Comparing with the apparent magnitude of IGR J19308+0530 (m = 10.95±0.13, con- V verted from m in the Tycho catalogue), we estimate a VT distance range of 300−450pc, for an extinction between N =(0−2.6)×1021cm−2(Dickey&Lockman1990,where H 3.3 Upper limits in the radio waveband N is converted into A following Gu¨ver & O¨zel 2009). H V Combining the systemic radial velocity with the source We searched for a radio counterpart to IGR J19308+0530 proper motion reported in the UCAC3 catalogue (?), with a Karl G. Jansky Very Large Array observation at 4.6 we computed the Galactic space velocity components and7.9GHz(120MHzbandwidth)(proposalID10B-238), of IGR J19308+0530 using the method of Johnson & taken on 2010 Aug. 19 with the array in its most compact Soderblom (1987). Assuming that the Local Standard D-configuration. The on-source time was 33 minutes and of Rest (LSR) participates in the Galactic rotation at thedatawerereducedaccordingtothestandardprocedures 254kms−1(Reid et al. 2009) and a distance of 375±75pc, within the Common Astronomy Software Application (Mc- the derived peculiar velocity is 45.3±2.9kms−1. This rules Mullin et al. 2007) software package, using the calibrator out a large asymmetric kick from a supernova. 3C 286 to set the amplitude scale and J1922+1530 to cali- The X–ray luminosity of IGR J19308+0530 measured brate the amplitude and phase gains for the target source. by Chandra in 2007 (Ratti et al. 2010) is 5×1029 −4× IGR J19308+0530 was not detected at either frequency. A 1030ergs−1, for distance in the above range of 300−450pc 3-sigma upper limit to the source flux of 54 µJy/beam was and the corresponding N . In the same way, UV Swift ob- derived H servationsprovide4×1030−1×1031ergs−1˚A−1 at2500˚A. Tosearchforradiopulsationsfromapotentialpulsarin A scenario where IGR J19308+0530 is not in full contact this system, we used the Westerbork Synthesis Radio Tele- and the X–rays are from coronal activity of the compan- scope and the PuMa2 backend (Karuppusamy et al. 2008). ion seems unlikely. First, coronal activity usually does not We observed for 1 hr from both 310−380 MHz (Obs. ID produce prolonged high energy emission, while the source 11205182, 29 Aug. 2012) and from 1300−1460 MHz (Obs. was discovered from a stack of INTEGRAL observations. ID11205210,31Aug.2012).Thepredicteddispersionmea- Second, coronal activity could explain the observed X–ray sure(DM)fora300pcdistancealongthisline-of-sight(see Discussion)isonly∼4pccm−3 (Cordes&Lazio2002).For luminosity only for the earliest spectral type allowed by the observations combined with very little N , as for F eachdataset,weusedthePRESTOsoftwaresuite(Ransom H 2001)tosearchasetoftrialDMsupto30pccm−3.Noob- stars the ratio between the X–rays and bolometric flux is log(L /L ) (cid:46) −4.6 for vsini ∼ 100 (Walter 1983). Fi- vious radio pulsar signal was detected after an acceleration X bol search. From these observations, we can place conservative flux density limits of S < 0.9 mJy and S < 0.4 mJy 350 1400 foranypulsarpresentinthesystem,assumingitisbeamed 2 http://www.pas.rochester.edu/∼emamajek/ towards us. EEM dwarf UBVIJHK colors Teff.dat (cid:13)c 0000RAS,MNRAS000,000–000 The mass ratio in IGR J19308+0530 5 nally, the spectra indicate disc contribution to the contin- ence Foundation operated under cooperative agreement by uum,andtheemissiondetectedintheBalmerlinesshowsa Associated Universities, Inc. slight, variable velocity offset with respect to the radial ve- locity curve of the companion star which is not expected in case they originate from coronal activity. The emission line REFERENCES componentalsoseemsweakforanhighlyactivestar.Ahot WD of ∼60000K alone could account for the UV emission, Bagnulo, S., Jehin, E., Ledoux, C., Cabanac, R., Melo, but not for LX unless thermonuclear burning is happening C.,Gilmozzi,R.,ESOParanalScienceOperationsTeam, onitssurface.Thefewintermediate-massCVswithlongor- 2003, The Messenger, 114, 10 bitalperiodthatareknowndoappearassupersoftsources Bird, A. J., et al., 2006, ApJ, 636, 765 (SSSs),showingsoftX–rayspectrapossiblyduetostablehy- Cordes, J. M., Lazio, T. J. W., 2002, ArXiv Astrophysics drogen burning on the WD (e.g. Kahabka 2006). However, e-prints withaluminosityof1036−1038ergs−1,SSSsaremuchmore Cranmer, S. R., Saar, S. H., 2011, ApJ, 741, 54 luminousthanwhatweobservefromIGRJ19308+0530.As Dickey, J. M., Lockman, F. J., 1990, ARA&A, 28, 215 thewindmasslossexpectedforthesecondaryspectraltype Gray, D. F., 1992, The observation and analysis of stellar is low (∼ 10−14M(cid:12) yr−1, Cranmer & Saar 2011) wind ac- photospheres., Camb. Astrophys. Ser., Vol. 20, C.U.P cretion can not account for the observed X–ray luminosity. Gu¨ver, T., O¨zel, F., 2009, MNRAS, 400, 2050 We conclude that IGR J19308+0530 is most likely in con- Horne, K., Wade, R. A., Szkody, P., 1986, MNRAS, 219, tact,consistentwiththeellipsoidalmodulationsobservedin 791 the lightcurve. This makes the source particularly interest- Johnson, D. R. H., Soderblom, D. R., 1987, AJ, 93, 864 ing among the scarcely populated class of IMXBs since it Kahabka, P., 2002, ArXiv Astrophysics e-prints shows Roche lobe overflow at a low accretion rate m˙. This Kahabka, P., 2006, Advances in Space Research, 38, 2836 is unusual, as Roche lobe accretion with an intermediate Karuppusamy, R., Stappers, B., van Straten, W., 2008, mass companion and such a large mass ratio is typically PASP, 120, 191 considered unstable, with an intense and short-lived accre- Knigge, C., Baraffe, I., Patterson, J., 2011, ApJS, 194, 28 tionphase(e.g.,Tauris&vandenHeuvel2006).Donorstars Liu,Q.Z.,vanParadijs,J.,vandenHeuvel,E.P.J.,2007, ofmassaround1.4M(cid:12) havethinornon-existentconvective A&A, 469, 807 envelopes (Verbunt & Zwaan 1981) which implies that the McCuskey, S. W., 1949, ApJ, 109, 426 instabilitydoesnotproceedonadynamicaltimescale.While McMullin, J. P., Waters, B., Schiebel, D., Young, W., Go- a convective envelope star would react to mass loss by ex- lap, K., 2007, in R. A. Shaw, F. Hill, & D. J. Bell, ed., panding,astarwithamostlyradiativeenvelopewillshrink, Astronomical Data Analysis Software and Systems XVI, slowing down the mass transfer. However, with a mass ra- vol.376ofAstronomicalSocietyofthePacificConference tio of 1.8 the radius of the Roche lobe will reduce faster Series, p. 127 in response to mass transfer than the radius of the com- O’Connell, D. J. K., 1951, Publications of the Riverview panion, triggering a thermal instability (Tauris & van den College Observatory, 2, 85 Heuvel 2006). Using the binary stellar-evolution code orig- Podsiadlowski, P., Rappaport, S., Pfahl, E., 2001, The inally developed by Eggleton (Yakut & Eggleton 2005 and influence of binaries on stellar population studies, Dor- referencestherein),wemodeledevolutionarytracksthatal- drecht: Kluwer Academic Publishers, 2001, xix, 582 low for periods of stable Roche-lobe accretion onto a WD p. Astrophysics and space science library (ASSL), with masses and orbital periods consistent with our find- Vol. 264. ISBN 0792371046, p.355, 264, 355 ings, although at higher m˙ values than implied by the LX Podsiadlowski,P.,Han,Z.,Rappaport,S.,2003,MNRAS, (see also the appendix in Podsiadlowski et al. 2003). The 340, 1214 low LX can be explained if we are observing a short lived Ransom, S. M., 2001, New search techniques for binary phase of low m˙, or if the mass transfer is non-conservative pulsars, Ph.D. thesis, Harvard University (a fully non-conservative scenario requires a mass loss rate Raskin, G., et al., 2011, A&A, 526, A69 fromthesystemof10−10−10−9M(cid:12)/yr).Thereiscurrently Ratti, E. M., Bassa, C. G., Torres, M. A. P., Kuiper, L., no direct evidence in favour of or against the presence of Miller-Jones,J.C.A.,Jonker,P.G.,2010,MNRAS,408, mass loss from the system. 1866 From an observational point of view, an overestimate Reid,M.J.,Menten,K.M.,Brunthaler,A.,Zheng,X.W., of q could be due to uncertainties on the limb darkening. Moscadelli, L., Xu, Y., 2009, ApJ, 693, 397 However, even assuming a limb darkening of 0 (instead of Ritter, H., Kolb, U., 2003, A&A, 404, 301 0.5) reduces vsini by only a few kms−1, still providing a Rodriguez, J., Tomsick, J. A., Chaty, S., 2008, A&A, 482, high q of ∼1.6. 731 Steeghs, D., Jonker, P. G., 2007, ApJ, 669, L85 Tauris, T. M., van den Heuvel, E. P. J., 2006, Forma- tion and evolution of compact stellar X-ray sources, ACKNOWLEDGMENTS C.U.P., p. 623 PGJ andGN acknowledge support from a VIDIgrant from Verbunt, F., Zwaan, C., 1981, A&A, 100, L7 the Netherlands Organisation for Scientific Research. We Wade, R. A., Horne, K., 1988, ApJ, 324, 411 thankT.MarshforMolly,E.L.RobinsonforhisXRbinary Walter, F. M., 1983, ApJ, 274, 794 codeandF.Verbuntforusefuldiscussion.TheNationalRa- Wilsey,N.J.,Beaky,M.M.,2009,SocietyforAstronomical dioAstronomyObservatoryisafacilityoftheNationalSci- Sciences Annual Symposium, 28, 107 (cid:13)c 0000RAS,MNRAS000,000–000 6 Ratti et al. Wilson, R. E., 1953, Carnegie Institute Washington D.C. Publication, 0 Yakut, K., Eggleton, P. P., 2005, ApJ, 629, 1055 (cid:13)c 0000RAS,MNRAS000,000–000