MNRAS000,1–11(2017) Preprint20January2017 CompiledusingMNRASLATEXstylefilev3.0 How unique is Plaskett’s star? A search for organized magnetic fields in short period, interacting or post-interaction massive binary systems⋆ Ya¨el Naz´e1†, Coralie Neiner2, Jason Grunhut3,4, Stefano Bagnulo5, Evelyne Alecian6, 7 Gregor Rauw1, Gregg A. Wade7, and the BinaMIcS collaboration 1 1 Groupe d’Astrophysique desHautes Energies- STAR, Institut d’Astrophysique et de G´eophysique - B5c, Universit´ede Li`ege,19c 0 All´ee du 6 Aouˆt, B-4000 Sart Tilman, Belgium 2 2 LESIA, Observatoirede Paris, PSL Research University,CNRS, Sorbonne Universit´es,UPMCUniv. Paris 06, Univ. Paris Diderot, n Sorbonne Paris Cit´e, 5place Jules Janssen, 92195 Meudon, France a 3 European Southern Observatory (ESO), Karl Schwarzschild Strasse, 85 748, Garching bei Mu¨nchen, Germany J 4 Dunlap Institutefor Astronomy and Astrophysics, Universityof Toronto, Rm101, 50 St. George Street, Toronto, ON, M5S3H4, Canada 5 Armagh Observatory and Planetarium, College Hill, Armagh, BT61 9DG, UK 9 6 Universit´eGrenoble Alpes, IPAG, CNRS, 38000 Grenoble, France 1 7 Dept. of Physics, Royal Military College of Canada, P.O. Box 17000, StationForces, Kingston,ON, Canada, K7K 4B4 ] R S 20January2017 . h p ABSTRACT - o Amongst O-type stars with detected magnetic fields, the fast rotator in the close r binarycalledPlaskett’sstarshowsavarietyofunusualproperties.Sincestrongbinary t s interactions are believed to have occurred in this system, one may wonder about a their potential role in generating magnetic fields. Stokes V spectra collected with the [ low-resolutionFORS2andhigh-resolutionESPaDOnSandNarvalspectropolarimeters 1 were therefore used to search for magnetic fields in 15 interacting or post-interaction v massive binaries. No magnetic field was detected in any of them, with 0G always 0 being within 2σ of the derived values. For 17 out of 25 stars in the systems observed 7 at high-resolution, the 90% upper limit on the individual dipolar fields is below the 3 dipolar field strength of Plaskett’s secondary; a similar result is found for five out of 5 six systems observed at low resolution. If our sample is considered to form a group 0 of stars sharing similar magnetic properties, a global statistical analysis results in a . 1 stringentupper limit of∼200G onthe dipolar field strength.Moreover,the magnetic 0 incidencerateinthefullsampleofinteractingorpost-interactionsystems(ourtargets 7 + Plaskett’s star) is compatible with that measuredfrom large surveys,showing that 1 they are not significantly different from the general O-star population. These results : v suggest that binary interactions play no systematic role in the magnetism of such i massive systems. X Key words: stars: early-type – stars: magnetic field – stars: individual: Plas- r a kett’sstar,HD1337,HD25638,HD35652,HD35921,HD57060,HD100213,HD106871, HD115071, HD149404, HD152248, HD190967, HD209481, HD228854, LSS3074, XZ Cep 1 INTRODUCTION were strongly suspected to exist based on several indirect pieces of evidence, such as the presence of strong magnetic Prior to the early 2000s, no direct detection of magnetic fieldsin the remnantsof massive stars (pulsars and magne- fieldsinO-typestarshadbeenreported.However,suchfields tars),ofsynchrotronradioemissioninsomemassivebinaries (for a review on non-thermal emitters, see Benaglia 2010), orofparticularphenomenasuchasperiodicmodulationsof ⋆ BasedonobservationscollectedattheEuropeanOrganisation lineprofiles(ase.g.inθ1OriC,Stahlet al.1996).Directev- for Astronomical Research in the Southern Hemisphere under idenceofmagnetismwasfinallyacquiredinthebeginningof ESOprogramme095.D-0075. the21st century.θ1OriCwas thefirstO-typestarfoundto † FNRSResearchAssociate,E-mail:[email protected] (cid:13)c 2017TheAuthors 2 Y. Naz´e et al. bemagnetic(Donatiet al.2002),HD191612, anOf?p star, rectly responsible for the generation of a global magnetic was the second (Donati et al. 2006), and many more have field.This could occur through the triggering of differential followed since then. Currently, about a dozen O-stars and rotation (e.g. Spruit 2002; Braithwaite 2006). Such a sce- several tens of early B-stars are known to be magnetic (see nario was proposed by several authors, in particular to ex- e.g. a summary by Petit et al. 2013), leading to an inferred plainmagnetisminmassivestarsasaresultofstellarmerg- incidencerateofsuchfieldsofabout6–8%amongstmassive ers(Ferrario et al.2009;Langer2014).Magnetism acquired stars (Fossati et al. 2015b; Grunhutet al. 2016). after a merging event would naturally concern (currently) Some of the first detections were made in binaries: single magnetic objects or magnetic stars in long-period for example, θ1OriC and HD191612 are known to have systems, but this idea cannot be applied directly to Plas- companions. However, the orbital periods of these sys- kett’sstar as it is a short-period system. However,“almost- tems are long, implying that the binarity has little in- merging”systems,afteracommon-envelopephaseoramass- fluence on the stellar properties. The case of Plaskett’s transferepisode,shouldalsoexperiencestrongshear,though star is very different: this “star” actually is a binary sys- discussionsaroseonwhethersuchprocesseswouldalsocon- tem harbouring two late-type O-stars in a rather tight stitute a viable road towards magnetism. Indeed, some Be orbit with a period of 14.4d (as first demonstrated by starsinbinarysystemsmaybetheproductsofapastmass- Plaskett 1922). It displays several peculiarities. The sec- transfer (van Bever& Vanbeveren 1997; McSwain & Gies ondary star appears slightly more massive than the pri- 2005; de Mink et al. 2013, and references therein) but no mary androtates morethanfour timesfaster thanits com- large-scalemagneticfieldwasdetectedforthoseobjects(see panion(Bagnuolo et al.1992;Linderet al.2008).Thestars Wadeet al. 2016b andSect.3.3 ofRiviniuset al.2013,and wereclassifiedO7.5I+O6IbyBagnuolo et al.(1992)andO8 references therein). Nevertheless, Plaskett’s star clearly ap- III/I+O7.5 III by Linderet al. (2008) - but the latter au- pearsas apotentialcandidatefor thisscenario (e.g. Langer thorsnotedthatthesecondarymay actually beaO7Vstar 2014;Schneideret al.2016).Iftrue,itwouldhaveimportant that shows an apparent O7.5III spectrum because of its implicationsforunderstandingthegeneralpathwaysleading rapidrotation.Amismatchthenexistsbetweenthespectro- tothe productionof magnetic fieldsin massive stars. scopic masses and the dynamical ones, the latter being too In this study we set out to directly assess the valid- large (Linderet al. 2008; Grunhutet al. 2013). Analyzing ity of this scenario, which has never before been done, theopticalspectrawithatmospheremodelsfurtherrevealed by verifying the magnetic status of “twins” of Plaskett’s that both stars are enriched in helium, that the secondary star, defined as short-period massive binaries that inter- surface isdepleted in nitrogen,and thatthenitrogen abun- act or have interacted in the recent past. If binary in- danceoftheprimarystar isabout 16 timesthatof theSun teractions, in particular mass transfer events, constitute while its carbon abundance is depleted by a factor of five a dominant process for the generation of magnetic fields (Linderet al. 2008, the N enrichment was already detected of massive stars, then most (if not all) of these sys- in X-ray data by Linderet al. 2006). In order to explain tems would possess a magnetic component. In that case, the unusual properties of the components, Bagnuolo et al. we would thus have identified a specific group of mag- (1992)andLinderet al.(2008)proposedthatthesystemisa netic O-stars, the second such category after the group postmasstransferbinaryinwhichthe(currently)rapidlyro- of Of?p stars which were all found to be magnetic since tatingsecondarystarreceivedmassandangularmomentum the pioneering detection of 2006 (HD191612, Donati et al. from its companion. In parallel, the Doppler maps derived 2006; Wadeet al. 2011; HD148937, Hubriget al. 2008; from atomographicanalysisoftheHαandHeiiλ4686 lines Wadeet al. 2012a; HD108, Martins et al. 2010; NGC1624- revealedanannularemissionregionwhichwasinterpretedin 2, Wade et al. 2012b; CPD−28◦2561, Hubriget al. 2012; terms of a flattening of the secondary’s wind (Linderet al. Wadeet al. 2015). On the contrary, if mass transfer events 2008).Thiswassubsequentlyunderstoodasaregionofmag- are not a source of strong, large-scale magnetic fields, then netically confined winds when a strong magnetic field was the magnetic incidence in these systems would not dif- detectedonthisstar(Bd>2.85kG, Grunhutet al.2013).It fer significantly from that found for large stellar samples isworthnotingthatPlaskett’ssecondaryistheonlyknown (Fossati et al. 2015b; Grunhutet al. 2016). In this paper, rapidly-rotating magnetic O-star, hence the only O-star to we report the results of such a campaign. Sections 2 and 3 appear in the centrifugally-supported magnetosphere zone present the targets and theobservations used in this study, of theconfinement-rotation diagram (Petit et al. 2013). respectively, while Sect. 4 provides the results and Sect. 5 The peculiar properties and evolutionary history of reports ourconclusions. this system lead to speculations regarding the origin of the magnetic field of Plaskett’s secondary, with two main possibilities. On one hand, Plaskett’s secondary is a mas- 2 THE SAMPLE sive star, and global (i.e. not localized) magnetic fields in massive stars are generally thought to be fossil (e.g. Moss The first step of the project was to select interacting or 2001, for a review see Neiner et al. 2015b and references post-interaction massive binaries. The key features to iden- therein). This could be the case of Plaskett’s secondary tify such systems are the sizes of the stars (e.g. filling their and, in this case, it would simply be amongst the few per- Roche lobes or close to doing so), but also the detection of cent of magnetic objects in themassive star population. Its non-solar abundances and/or of rotational asynchronicity, peculiarities - linked to binary interactions (stellar tides, as well as a mismatch between predictions from evolution- see Palate & Rauw 2014, and previousmass transfer event, ary models of isolated stars and the observed stellar prop- Linderet al. 2008) - would then be a coincidence. On the erties (age, masses, abundances,and luminosities). In addi- otherhand,binarity could haveplayeda keyrole, beingdi- tion, these binaries need to be bright enough to be observ- MNRAS000,1–11(2017) How unique is Plaskett’s star? 3 ablewiththecurrentgenerationofspectropolarimeters,but secondarywind,similar tothatfoundinPlaskett’sstar.Fi- theyalso oughttobewell known,i.e.tohavereliable phys- nally, a wind-wind collision is also present in the system ical parameters, so that there is no ambiguity on their evo- (Linder 2008). lutionary status. Gies (2003) provided lists of well-studied • HD100213=TUMus (O7.5V+O9.5V) is a binary with double-lined spectroscopic (SB2) binary systems located in e=0 and P=1.4d (Linderet al. 2007; Pennyet al. 2008). the Galaxy and the Magellanic Clouds, among which he Mutual irradiation of the stars in this very short-period separately listed unevolved cases (his Table 1) and evolved system leads to the observation of line strength variations cases (his Table 2), the latter ones being subdivided be- along the orbit (Linderet al. 2007; Palate et al. 2013). The tween systems before contact, in contact, and after mass eclipses of the system indicate that the stars are in con- transfer.Penny et al.(2008)alsolistedbinariesinwhichone tact while the masses derived from the orbital solution are starfillsitsRochelobe,complementingthelist.Eliminating lowerthanexpectedfromevolutionarymodels,suggestinga thecomplexhigh-multiplicitysystemsδOriandCygOB25, pastepisodeofmass-transfer(Linderet al.2007;Qian et al. which could blur the picture, while adding HD149404 and 2007; Pennyet al. 2008,and references therein). LSS3074,twootherpost-RocheLobeOverflow(RLOF)sys- • HD106871=AB Cru (O8V+B0.5:) appears as a semi- tems not listed in these references but whose interactions detached, eclipsing binary with a circular orbit and a pe- wererecentlyanalyzed(seedetailsbelow),weendedupwith riod of3.4d(Lorenz et al.1994).A consistent fittothede- 15 relatively bright targets: rived properties of the stars (effective temperatures, radii, masses, luminosities) cannot be found with evolutionary • HD1337=AOCas(O9.7III+O9.5V)consistsoftwostars tracks considering the absence of interaction while anoma- in a circular orbit of 3.5d period (Stickland 1997; Linder lous hydrogen and helium abundances are detected, both 2008). The analysis of the eclipses yields a semi-detached factssuggestingthesystemtohaveundergonemass-transfer configuration, with a ‘hot spot’ on the secondary possibly (Lorenzet al. 1994). correspondingtotheimpactofanaccretingstream(Linder • HD115071(O9.5V+B0.2III)consistsofasemi-detached 2008). Furthermore, stellar masses and radii appear too binary with a circular orbit and a period of 2.7d small compared to typical values for massive stars of the (Pennyet al. 2002). The ellipsoidal variations of its same spectral types, reinforcing the mass-transfer scenario lightcurve indicate that the secondary star fills its Roche (Linder 2008). lobe. The derived masses are smaller than expected, sug- • HD25638=SZCam(O9.5V-9IV+B0-0.5V)comprisesan gestingmass-transfertohaveoccurredinaveryrecentpast eclipsing,circularSB2systemwithaperiodof2.7d(Gorda (Pennyet al. 2002, and references therein). 2015, and references therein). The derived radius of the • HD149404 (O7.5If+ON9.7I) is a non-eclipsing, de- primary is too large for a main sequence star while the tached binary with a circular orbit and a period of 9.8d mass of the secondary is too small for its spectral type (Rauwet al. 2001; Thaller et al. 2001). Emission lines such (Lorenzet al.1998).Furthermore,thederivedmasses,lumi- asHαandHeii4686displayphase-lockedvariationsindicat- nosities,andtemperaturescouldnotbefittedbytheoretical ingthepresenceofawind-windcollision (Rauwet al.2001; evolutionarytracks,rulingoutan‘interaction-freeevolution’ Thaller et al. 2001). In addition, the secondary star is of (Harries et al.1998;Lorenzet al.1998;Tamajo et al.2012). the ON type, i.e. it presents anomalous abundances which • HD35652=IUAur (O9.5V+B0.5IV-V) comprises an hint at a past interaction (Rauwet al. 2001; Thaller et al. eclipsing, semi-detached binary with a circular orbit and a 2001).Arecent,moredetailedstudyfurtherrevealsanasyn- period of P=1.8d (Harries et al. 1998; O¨zdemir et al. 2003, chronous rotation of the two stars, a significant nitrogen andreferencestherein).Thederivedmassesandgravitiesof overabundanceofthesecondarystar,andpeculiarpositions the stars cannot be explained by the separate evolution of of the stars in the HR diagram (Raucqet al. 2016). All of massivestarsbornatthesametime,suggestingapastmass- these properties clearly disagree with expectations of sin- transfer event (Harries et al. 1998, see in particular their glestarevolutionarymodels,yieldingfurthersupporttothe Fig. 9). past mass-exchange scenario (Raucq et al. 2016). • HD35921=LYAur (O9III+O9.5III) is a circular binary • HD152248 (O7.5III(f)+O7III(f)) is a slightly eccen- with a period of 4.0d (Stickland et al. 1994; Zhao et al. tric and eclipsing binary with a period of 5.8d (Sanaet al. 2014, and references therein). The analysis of the eclipses 2001; Mayeret al. 2008). Phase-locked variations in Hα suggests a semi-detached or slightly overcontact configu- and Heii4686 lines as well as in the X-ray emission arise ration (Li & Leung 1985; Drechsel et al. 1989; Mayer et al. from a wind-wind collision located between the two stars 2013) and the derived masses appear slightly too low (Sanaet al.2001,2004).Thederivedmassesarelowerthan (Sticklandet al.1994;Mayer et al.2013),favouringanevo- expectedfromevolutionarytracksandbothcomponentsap- lutionary scenario where mass-transfer has occurred. pear close to filling their Roche lobes, suggesting a past • HD57060=29CMa (O8.5I+O9.7V) is a slightly eccen- mass-transferepisode(Sanaet al.2001).Notethattidalef- tric binary with a period of 4.4d (Bagnuolo et al. 1994; fects are reinforced by a small, non-zero,eccentricity. Stickland1997;Linder 2008).Thephotometriceclipsessug- • HD190967 (O9.5V+B1I-II) consists of an eclipsing bi- gestsomeinteractionbetweenthestars(fromsemi-detached nary in a semi-detached configuration and with a period to overcontact configuration, Bagnuolo et al. 1994; Linder of 6.5d and e=0 (Harries et al. 1997; Kumsiashvili et al. 2008;Antokhinaet al.2011)andtheanomalousluminosities 2005). In a mass-gravity diagram, the derived values are at also seem to indicate an on-going mass-transfer. Further- oddswithpredictionsofevolutionarymodelsforstarsofthe more, atomographic analysis of theHαandHeii4686 lines same age, indicating a past episode of rapid mass-transfer (Linder 2008) and discrepancies in themass ratio determi- (Harries et al.1997;Kumsiashvili et al.2005).Notethat,to nation(Antokhinaet al.2011) suggest aflattened,disk-like best fit the lightcurve, either bright spots or an accretion MNRAS000,1–11(2017) 4 Y. Naz´e et al. Table 1. Summaryofthepropertiesofourtargets andofPlaskett’sstar ID V Sp.Types P(d) e C/SD? AA? Mass vsin(i)(kms−1) gainer prim-sec-tert Plaskett’sstar 6.06 O8III/I+O7.5III 14.4 0 N Y sec. 77-370∗ HD1337 6.14 O9.7III+O9.5V 3.5 0 Y sec. 118-82∗ HD25638 6.93 O9.5V-9IV+B0-0.5+B0III-III 2.7 0 N prim. 137-69-40∗ HD35652 8.39 O9.5V+B0.5IV-V+B 1.8 0 Y prim. 172-159-71∗ HD35921 6.85 O9III+O9.5III+B 4.0 0 Y prim.? 202-122-27∗ HD57060 4.95 O8.5I+O9.7V 4.4 0.1 Y sec. 107-193∗ HD100213 8.41 O7.5V+O9.5V 1.4 0 Y 222-191 HD106871 8.48 O8V+B0.5: 3.4 0 Y Y prim. HD115071 7.97 O9.5V+B0.2III 2.7 0 Y prim. 101-132 HD149404 5.47 O7.5If+ON9.7I 9.8 0 N Y prim. 56-80∗ HD152248 6.05 O7.5III(f)+O7III(f) 5.8 0.1 N 139-128∗ HD190967 8.16 O9.5V+B1I-II 6.5 0 Y prim. 230-115∗ HD209481 5.55 O9III+ON9.7V 3.1 0 N Y prim. 127-93∗ HD228854 8.65 O7.3V+O7.7V 1.9 0 Y 223-203∗ LSS3074 11.7 O4If+O7.5 2.2 0 N Y sec. 110-127 XZCep 8.51 O9.5V+B1III 5.1 0 Y prim. 104-172∗ Evidenceforcurrentorpastinteraction(forreferences,seetext):“C/SD”=inacontact orsemi-detachedconfiguration,“AA”= abundances anomaliesdetected -notethattherearealsoother anomaliesdetected (e.g.toolowmasses,asynchronous rotation,...),see textfordetails.Theprojectedrotational velocitiesvsin(i)aregivenforprimaryandthensecondarythentertiary,theycomefromour analysisofhigh-resolutionspectrawhenanasteriskisaddedandfromliteratureotherwise(Howarthetal.1997forHD100213, otherwiseseededicated itemsintextforliteraturereferences). disk are needed (Kumsiashviliet al. 2005; Djuraˇsevi´c et al. ondary star fills its Roche lobe while the derived stellar 2009). properties are at odds with predictions from evolutionary • HD209481=LZCep (O9III+ON9.7V) is a binary with models(incompatibleages forthederivedmasses andgrav- a circular orbit and a period of 3.1d. The ellipsoidal varia- ities, Harries et al. 1997), suggesting a previous episode of tionsofitsopticallightcurveindicatethatbothstarsalmost mass-transfer. filltheirRochelobes(Mahy et al. 2011;Palate et al. 2013). Severallinesofevidencepoint towardsapastmass-transfer Evidence for past or current interactions (abundance episode (Harries et al. 1998; Mahy et al. 2011): the photo- anomalies, Roche lobe filling,...) in the chosen systems has sphereofthesecondarystarappearsstronglyenrichedinhe- just been mentioned above (see Table 1 for a summary), liumandnitrogenbutdepletedincarbonandoxygen,there but the presence of such interactions does not necessarily is a slight asynchronicity between therotation periods, and implythattheywereactuallyefficientinmodifyingthestel- theevolutionarymassesarehigherthanthemassesinferred lar structure. In particular, Plaskett’s secondary displays from theorbital solution. a high rotation rate (∼300kms−1, Bagnuolo et al. 1992; • HD228854 (O7.3V+O7.7V) is a system with a circu- Linderet al.2008;Grunhutet al.2013),demonstratingthat lar orbit and a period of 1.9d (Ya¸sarsoy & Yakut 2013, momentum exchange was very efficient for that object. For and references therein). The analysis of the eclipses in- the stars in our sample, projected rotational velocities are dicates filling of the Roche lobe(s) (either overcontact, also rather high and generally amount to 100–200kms−1, Deˇgirmenci et al. 1999; Qian et al. 2007, or contact/semi- with components in four systems (HD35921, HD100213, detached, Mayer et al. 2002; Ya¸sarsoy & Yakut 2013) and HD190967, and HD228854) reaching even higher velocities the derived physical properties (in particular masses and (vsin(i)>200kms−1 see Table 1). This confirms that our gravities) disagree with theoretical expectations for (sin- sample is well suited to find out whether magnetic fields gle) massive star evolution (Harries et al. 1997), pointing are triggered in massive binary interactions. towards a post mass-transfer situation. • LSS3074 (O4If+O7.5f) is a circular binary with a pe- riod of 2.2d (Niemela et al. 1992; Haefner et al. 1994, and Raucq et al. 2017 submitted). Hα and Heii4686 lines present line profile modulations that are reminiscent of those observed in HD149404, suggesting wind-wind colli- sion to be present (Raucq et al., 2017, submitted). Pho- 3 OBSERVATIONS tometric, spectroscopic, and polarization measurements led to the derivation of the physical parameters for the system For the Northern targets, high-resolution spectropolarime- (Niemela et al.1992,andRaucqetal.2017submitted).The try was acquired with ESPaDOnS at CFHT (Donati 2003) anomalousabundancesaswellasthemismatchbetweenevo- andNarvalatTBL(Auri`ere2003).FortheSoutherntargets, lutionary tracks and observed masses and luminosities sug- low-resolutionspectropolarimetrywasacquiredwithFORS2 gest a mass-transfer in the system (Raucq et al. 2017 sub- at ESO (Appenzeller& Rupprecht 1992; Szeifert et al. mitted). 1998). Two targets, HD152248 and HD149404, were ob- • XZCep(O9.5V+B1III)isanon-eccentric,eclipsingsys- served at both low and high resolutions, as they are visible tem with a period of 5.1d (Harries et al. 1997). The sec- from Chile as well as from Hawaii. MNRAS000,1–11(2017) How unique is Plaskett’s star? 5 3.1 Low-resolution FORS2 spectropolarimetry Low-resolution spectropolarimetric data of six targets were obtained with the Very Large Telescope equipped with FORS2inSpring2015(ESO095.D-0075,PINaz´e,seeTable 2).ThesedataweretakeninservicemodewiththeredCCD (a mosaic composed of two 2k×4k MIT chips), no binning, aslit of1”andthe1200B grating(R∼1400).Theobserving sequence consisted of 8 subexposures with retarder wave- plate positions of +45◦, −45◦, −45◦, +45◦, +45◦, −45◦, −45◦, +45◦.WereducedthesespectropolarimetricdatawithIRAF as explained by Naz´eet al. (2012): the aperture extraction radius was fixed to 20 pixels, the nearby sky background was subtracted, and wavelength calibration was performed from 3675 to 5128A˚ (with pixels of 0.25A˚) considering arc lampdatatakenatonlyoneretarderwaveplateposition(in our case, −45◦). There was no indication of variation of the StokesIspectrabetweensubexposures.Wethenconstructed the normalized Stokes V/I profile, as well as a diagnostic “null” profile N/I (see Donatiet al. 1997; Bagnulo et al. 2009,fordetailsontheprocedureandFig.1foranillustra- tion in our case). Finally, the associated longitudinal mag- netic field was estimated by minimizing χ2=P (yi−Bzxi−a)2 i σ2 i withyi eitherV/Iorthenullprofileatthewavelengthλiand xi=−geff 4.67×10−13 λ2i 1/Ii (dI/dλ)i (Bagnulo et al. 2002). Thiswasdoneforxiintheintervalbetween−10−6and+10−6, Figure 1. Top Left: The spectrum of HD100213 taken on the second night of observation with FORS2, with the selected win- afterdiscardingedgesanddeviantpointsandafterselecting dowsshownbyagreenthickline.Top right:AssociatedV/I (top spectral windows centered on stellar lines (see Naz´e et al. blue)andN/I (bottomred)asafunctionofwavelength.Bottom: 2012, for furtherdiscussion). These windows comprise both V/I (left) and N/I (right) as a function of xi, with their best- primaryandsecondary absorption lines,butavoidemission fit straight lines shown in cyan. Rejected values, not considered lines (as some of them arise in wind-wind collision regions forthe fitting, appear inred. Note the null slopes ofthe best-fit and are thus not representative of the stellar photosphere). straightlines,thatindicatetheabsenceofasignificantmagnetic The values reported in Table 2 were obtained after rectify- field. ing the Stokes profiles, but similar values are found if no rectification is applied. an illustration). We first constructed a line mask from the VALD2 database (Piskunov et al. 1995; Kupkaet al. 1999) usingtheappropriatetemperatureandlog(g)valuesforeach 3.2 High-resolution ESPaDOnS and Narval star, assuming solar abundances and considering only lines spectropolarimetry deeperthan1%ofthecontinuum.Wethenexcludedallhy- Mosthigh-resolutionspectropolarimetricdataofourtargets drogen lines and lines blended with hydrogen. This yielded were collected in the context of the MiMeS (Wadeet al. masks with between 430 and 1220 lines, depending on the 2016a) and the BinaMIcS (Alecian et al. 2015) large pro- star, over the 3700–9800A˚ interval. The line depths were grammes. They were acquired with ESPaDOnS, an echelle thenadjustedtoprovidethebest fittotheobserved Stokes spectropolarimeter installed at the Canada-France-Hawaii I spectrum of each target. Note that regularisation (see Telescope. In addition, archival data from dedicated pro- Kochukhovet al. 2010) and clipping of the deviant points grammes(PIsNeiner,Bouret)wereobtainedforHD209481 werebothappliedtoimprovethefinalsignal-to-noiseofthe in 2006 and 2009 using Narval, a similar spectropolarime- LSDprofiles. terinstalledatPicduMidi(France).Biases, flat-fields,and Since our targets are binaries, the second step was to ThAr calibrations were obtained at the beginning and at derivetheindividualpropertiesofeachcomponent(seeFig. the end of each night. The data reduction was performed 2and Grunhutet al. 2016 for details on theprocedure).To using Libre-Esprit, the dedicated reduction software based this aim, multiple absorption components with profiles de- on Esprit (Donati et al. 1997). There was no indication of rived from theconvolution of a rotationally broadened pro- variationoftheStokesIspectrabetweensubexposures.Nor- file with a radial-tangential macroturbulence profile were malizationoftheindividualorderswasthenperformedusing fitted to the Stokes I profiles. Since this disentangling pro- low-degree polynomial fits. cess often leads to degenerate solutions, relevant informa- After the Stokes spectra were calculated, we followed tion (e.g. vsin(i)) from the literature was used to constrain the same procedure as Grunhutet al. (2016). First, the thefits,wheneverneeded.ThederivedRVswerecheckedfor Least-Squares Deconvolution technique(LSD,Donati et al. compatibility with theknownorbital solutions of thesesys- 1997; Kochukhovet al. 2010) was applied to all polarimet- tems, i.e. we verified that each observed pair of RVs agreed ric spectra to increase the signal-to-noise, thereby helping with the solution at some phase during the orbit (accurate to detect weak magnetic Zeeman signatures (see Fig. 2 for orbital phases at current epoch cannot be computeddueto MNRAS000,1–11(2017) 6 Y. Naz´e et al. HD1337 HD25638 HD35652 50*N/I & V/Icc111...000012 0*N/I & V/Icc1111....000000110505 15*N/I & V/Icc1111....000000110505 5 I/Ic0.99 I/Ic0.995 I/Ic0.995 0.990 0.990 -400 -200 0 200 400 600 -400 -200 0 200 400 -600-400-200 0 200 400 600 velocity (km/s) velocity (km/s) velocity (km/s) HD35921 HD57060 HD149404 V/Ic V/Ic1.02 V/Ic & 1.010 & & 1.02 N/I c1.005 N/I c1.01 N/I c1.01 40*1.000 120*1.00 100*1.00 I/Ic0.995 I/Ic0.99 I/Ic0.99 -800-600-400-200 0 200 400 600 -400 -200 0 200 400 -400 -200 0 200 400 velocity (km/s) velocity (km/s) velocity (km/s) HD152248 2014-04-14 HD190967 HD209481 2011-06-22 30*N/I & V/Icc111...000012 70*N/I & V/Icc1111....00000123 100*N/I & V/Icc111...000012 I/Ic0.99 I/Ic0.99 I/Ic0.99 0.98 -400 -200 0 200 400 -400 -200 0 200 400 -400 -200 0 200 400 velocity (km/s) velocity (km/s) velocity (km/s) HD228854 XZ Cep V/Ic1.010 V/Ic1.02 & & N/I c1.005 N/I c1.01 5* 0* 11.000 5 1.00 I/Ic0.995 I/Ic0.99 -600 -400 -200 0 200 400 600 -400 -200 0 200 400 velocity (km/s) velocity (km/s) Figure 2. LSD profiles derived for the BinaMIcS sample (coming from Grunhutetal. 2016 for the three MiMeS targets - HD1337, HD35921,HD209481).Eachpanelprovides,fromtoptobottom,thesystemnameandtheStokesV,N,andIprofiles,respectively.The individualcomponents ofthe I profiles,derivedfromthedisentanglingproceduredescribedinSect.3.2,areaddedwithcolours(yellow, blue,orgreensolidlines)alongwiththeirsum(dashedredlines). Table 2.Resultsfromthelow-resolutionspectropolarimetryobtainedwithFORS2. ID L1/L2 Exp.(s) SNR Date HJD-2450000. Bz (G) Nz (G) HD100213 ∼2 8×70 2470 2015-04-03 7115.687 34±48 2±47 8×70 2365 2015-04-05 7117.732 24±50 37±49 HD106871 ∼2.5 8×80 2210 2015-04-03 7115.733 −31±45 3±44 8×80 2380 2015-04-05 7117.754 32±39 60±38 HD115071 ∼0.9 8×60 2350 2015-04-05 7117.837 −22±34 −44±33 8×60 2520 2015-05-01 7143.618 −42±32 16±33 HD149404 ∼0.7 8×7 1650 2015-05-16 7158.907 214±119 −234±117 8×7 2130 2015-05-17 7159.585 −64±80 3±69 HD152248 0.8–1. 8×10 2100 2015-05-02 7144.620 −25±69 36±63 8×10 1845 2015-05-17 7159.570 29±59 90±59 LSS3074 1.–2.5 8×536 1400 2015-04-03 7115.809 −591±619 200±427 8×536 1450 2015-04-05 7117.796 −452±391 187±362 Fromlefttoright,thecolumnsshowthebinaries’names,thelightratios(seeSect. 2forreferences),theexposurelengths,thepeak signal-to-noiseratiointhe4000–5000A˚range,theobservingtimes(bothinYYYY-MM-DDformatandinheliocentricJuliandates), andfinallythederivedmagneticfieldvaluesandtheirassociatederrorsforboththesystem’sStokes V profileandfromthenull diagnosticprofile.These Bz and Nz valueswerefoundusingrectification andwithinselectedspectralwindows. MNRAS000,1–11(2017) How unique is Plaskett’s star? 7 the large accumulated phase uncertainties). Disentangled I Table 3 provides the field values derived from each high- profileswerethenobtainedforeachsystemcomponent,and resolution spectropolarimetric observation for individual used to estimate individual V and N contributions. Note components and for each system as well as upper limits on that,iflineprofilesoverlapandonecomponentismagnetic, non-detectedmagneticfieldspercomponentfromindividual this procedure does not account for the possible magnetic andcombinedobservations.Itisclearfromthesetables(see contaminationcausedbytheStokesV signalofthecompan- also Fig. 3) that (1) no problem due to spurious signals af- ion. fectsthedata(thenulldiagnosticsbeingalwayscompatible Finally, the magnetic properties were derived from the withzerowithin2σatmost-andgenerally wellwithin1σ) StokesV profiles.Wecalculatedineachcasetheprobability and (2) no significant magnetic field is detected for any of that deviations of the observed V/Ic signal from the nor- thesystems ortheir components. malized value of 0 occur by chance (i.e. because of noise One caveat must however be mentioned: when using only - see Donati et al. 1992, but note that this reference low-resolution spectropolarimeters likeFORS2,even strong reports the complement probability). This was done using fieldsmaybecomeunobservable,dependingonthemagnetic an optimal velocity grid (with a minimum resolution of geometry of the star and the observing phase. Indeed, as 1.8kms−1) representing the best compromise between in- thesinusoidal Bz curveofmagneticobliquerotatorsreaches creasing the signal-to-noise and not smearing the Zeeman or nears a value of 0G at cross-over phases (i.e. when the signature. Next, the longitudinal field values Bz were com- dipole is seen equator-on), the field would become unde- puted from the first-order moment of the unbinned Stokes tectable in low resolution observations. However, this oc- V profiles(Mathys1989;Donatiet al.1997),witherrorsde- curs during a small fraction of the rotational period only, rivedbypropagatingtheknownuncertaintiesforeachpixel. an interval depending on the error bars. For example, we Similarmeasurementsweremadeforthediagnosticnullpro- mayconsiderPlaskett’s Bzvalues(Grunhutet al.2013)and files N. Note that the field values for the MiMeS targets assume a sinusoidal Bz varying from −810G to 680G. Non- (HD1337,HD35921,andHD209481)werealreadyreported detectionsimply Bz∼0,i.e.|Bz|<NσwithN generallytaken byGrunhutet al. (2016). tobe3(thoughavalueof5appearsbettersuitedforFORS, Since the significance level (or false-alarm probabil- see Bagnulo et al. 2002). Since the typical error on the ten ity) was larger than 10−3 in each case, corresponding to FORSmeasurementsamountsto∼55G,excludingthemore clear non-detections, upper limits on the dipolar field val- uncertain case of LSS3074, non-detections would then oc- ueswerederivedusingMonte-Carlosimulationsasdescribed cur during 15% of the orbit if we consider 3σ, or 25% for byNeiner et al. (2015a) and Blaz`ere et al. (2015). Forboth a 5σ threshold. Missing a detection is less probable when theprimaryandsecondary,wesimulated1000obliquedipole severalobservationsofthesametarget areacquired,aswas modelsforvariousvaluesofthepolarmagneticfieldstrength doneforallsystemsobservedatlowresolution.Indeed,itis withrandominclinationanglesi,obliquityanglesβ,andro- unlikely to have two random exposures sampling cross-over tational phases. White Gaussian noise with a null average phases for one target. Moreover, if it is already unlikely for and a variance corresponding to the signal-to-noise of each onetarget,itisevenmoresoforseveralindependentobjects. observed profile was added. Using the best-fit disentangled Considering a 15% chance to miss detecting a Plaskett-like LSD I profiles, we calculated local Stokes V profiles assum- field in one observation, the probability to miss field detec- ing the weak-field case, which were then integrated over tionsinalltenobservationsis(0.15)10∼10−8 (or10−6 if25% the visible hemisphere of the star. The derived synthetic is considered). Of course, there is a high chance (∼80%) to Stokes V profiles were normalized to the intensity of the miss one detection overten observations, but since we have continuum, and the Neyman-Pearson likelihood ratio test two observations of each target and two targets in common was then applied to estimate the probability of detecting a with the high-resolution sample, then thisrisk may be mit- dipolar oblique magnetic field (with a threshold of 10−3 for igated.Onecould object that some oftheobservation pairs the false alarm probability). A 90% detection rate was re- are separated by only one or two days, so that there would quiredtoconsiderthatthesimulatedfieldwouldstatistically benolarge phasechangeoverthatintervaliftherotational be detected in the data. This translates into upper limits period is long. However, it must be remembered that, con- for the possible undetected dipolar field strength for each trary to the“usual”magnetic massive stars, our targets are star and each observation. For HD152248 and HD209481, relatively fast rotators in short-period systems so that Bz multipleobservationswereavailableandsincenoneofthem should change rapidly, with significant variations expected resulted in a magnetic detection, single-observation statis- on thetimescale of one day, as observed for Plaskett’s star. tics were combined to extract a stricter upper limit on the Moreover,evenifwedoconsiderthatthetwoobservationsof non-detected field (see Sect. 4.2 of Neineret al. 2015a for atargetweretakenatthesamerotationalphase,theproba- details). bilitytomissafieldinallfivecasesisstilllow:(0.15)5∼10−4 considering a 15% chance of missing the field detection, or 10−3 if a 25% chance is used instead. Therefore, statisti- 4 RESULTS AND DISCUSSION cally, we could have missed detecting the magnetic field of Table2reportstheresultinglongitudinalmagneticfieldval- ues1 for the low-resolution spectropolarimetric data, while However, deriving individual limits from the system’s values is notstraightforward,sincelongitudinalfieldsarederivedfromthe analysis of sets of lines, whose strengths for primary and sec- 1 They are weighted averages of the contributions of all compo- ondary do not necessarily reflect the light ratios between com- nents in each system, i.e. each value is valid for the entire sys- ponents and furthermore vary with ion under consideration. We tem. The light ratios L1/L2 of the systems are known (Table 2). thereforerefrainfromderivingindividuallimits. MNRAS000,1–11(2017) 8 Y. Naz´e et al. Table 3.Resultsfromthehigh-resolutionESPaDOnSandNarvalspectropolarimetry. ID Exp. Date HJD Obj Vr Bz Nz σ up.lim. /SNR -2.45e6 (kms−1) (G) (G) (G) (kG) HD1337 4×720 2009-10-09 5113.845 prim 178 46 −37 39 1.0 1280 sec −160 −46 −134 74 1.9 system 35 40 60 HD25638 8×840 2013-11-20 6616.786 prim 193 −62 −7 69 1.7 1557 sec −272 −112 −154 241 6.8 tert 17 −12 −62 55 2.0 system −55 85 101 HD35652 8×840 2016-02-27 7445.812 prim −198 −122 −42 267 7.1 942 sec 315 216 120 205 4.8 tert 124 189 −368 329 8.7 system 37 −161 249 HD35921 4×940 2011-11-13 5878.872 prim 130 −102 −44 75 1.7 1215 sec −287 76 −48 104 2.8 tert −18 16 −45 74 2.0 system 74 −15 98 HD57060 12×540 2013-03-03 6354.835 prim 160 17 −5 15 0.6 2942 sec −141 173 286 196 3.3 system 92 102 47 HD149404 16×390 2014-06-08 6816.897 prim −108 14 −10 25 0.7 3038 sec 7 34 4 16 0.3 system 35 9 16 HD152248 4×540 2014-04-13 6761.045 prim −135 −175 −72 155 4.1,comb: 1.5 794 sec 111 416 −148 206 5.4,comb: 1.9 system −9 −39 148 HD152248 12×540 2014-04-14 6762.060 prim 129 112 88 89 2.2 1358 sec −192 −153 −135 104 2.5 system −46 7 97 HD190967 8×840 2016-05-18 7527.097 prim −32 −15 −15 87 2.1 989 sec 80 −3 21 32 0.9 system −14 −13 46 HD209481 4×600 2006-12-14 4084.274 prim 74 14 60 74 1.7,comb: 0.4 867 sec −226 −36 51 120 3.1,comb: 0.6 system −34 107 91 HD209481 4×675 2009-07-20 5033.483 prim 27 −73 124 97 2.3 667 sec −100 160 −168 134 3.4 system −46 −10 65 HD209481 4×825 2009-07-24 5037.485 prim −98 85 −32 48 1.1 1191 sec 212 213 24 83 2.1 system 99 −87 63 HD209481 4×675 2009-07-25 5038.492 prim 26 59 2 66 1.5 916 sec −95 −51 60 107 2.7 system −34 33 45 HD209481 4×675 2009-07-26 5039.492 prim 46 192 67 62 1.4 804 sec −154 −225 9 90 2.3 system 16 0 54 HD209481 4×675 2009-07-27 5040.490 prim −97 142 −112 180 4.2 315 sec 212 −459 −307 322 8.3 system −12 −80 238 HD209481 4×675 2009-07-28 5041.480 prim 10 26 −9 54 1.2 1058 sec −66 5 1 76 1.9 system −19 −2 29 HD209481 4×675 2009-07-29 5042.496 prim 54 23 −62 48 1.1 1199 sec −174 −21 128 70 1.8 system 26 −19 46 HD209481 4×675 2009-07-30 5043.476 prim −96 85 −17 60 1.4 1037 sec 201 −62 −49 100 2.5 system 76 25 69 HD209481 4×675 2009-07-31 5044.496 prim −4 2 −39 57 1.3 1105 sec −40 40 −17 86 2.2 system −19 −32 30 MNRAS000,1–11(2017) How unique is Plaskett’s star? 9 Table 3.Continued. ID Exp. Date HJD Obj Vr Bz Nz σ up.lim. /SNR -2.45e6 (kms−1) (G) (G) (G) (kG) HD209481 4×675 2009-08-03 5047.461 prim −24 39 −69 69 1.6 968 sec 6 71 5 89 2.2 system 46 −2 32 HD209481 12×400 2011-06-22 5735.027 prim −60 3 2 22 0.5 2417 sec 108 −3 −6 39 1.0 system 1 −16 17 HD228854 8×840 2016-05-17 7526.067 prim −206 −142 −146 188 5.7 975 sec 275 −122 142 256 7.6 system 27 208 269 XZCep 8×840 2013-08-19 6523.930 prim −265 −30 5 45 1.1 609 sec 103 621 −473 488 11.1 system 37 −180 184 Fromlefttoright,thecolumnsprovidethebinaries’names,theexposurelengthsandpeaksignal-to-noiseratiosinthe5000–6500A˚ range,theobservingtimes(bothinYYYY-MM-DDformatandinheliocentricJuliandates),theconsideredcomponent,itsheliocentric velocity, themagneticfieldvaluesandtheirassociatederrors(σ)derivedfromthesystem’sStokes V profile(Bz)andfromthenull diagnosticsprofiles(Nz),andfinallytheupperlimitsonthedipolarfieldstrengthwith90%chanceofbeingdetected. Whenseveral observationsofthesametargetareavailable(asforHD152248 andHD209481), theupperlimitontheundetected fieldfromthe combinedstatisticsisquotedinitalicsinthelastcolumnofthefirstobservationlineaftera“comb”prefix. onestar in theFORS2sample, butnot more. Notethat for byfollowingtheBayesianapproachofKolenberg& Bagnulo high-resolutiondata,thisproblemdoesnotexistasasingle (2009, see in particular their Eq. (7), later generalised by randomobservationhasproventobeefficienttodetectmag- Asensio Ramos et al. 2015). It assumes that all stars of our netism(whenitexists)inamassivestarduringtheMiMeS, sample are similar, i.e. they have a dipolar field with iden- BinaMIcS,andothermagneticsurveys(seealsoamorethe- tical field strength, but their magnetic axes are randomly oreticaldiscussionbyPetit & Wade2012).Indeed,exceptin oriented with respect to the line-of-sight, and it combines veryrarecases2,evenifthederivedlongitudinalfieldiszero, the observational values into a constraint on the common the high-resolution Stokes V profile of a magnetic object is fieldstrength.WenotethatEq.(7)ofKolenberg& Bagnulo not flat and the field can be detected whatever the phase. (2009)isvalidonlyforindependentmeasurementsondiffer- Missingthedetectionofastrongfieldinthehigh-resolution entstars,thereforeforthoseobjectsthathavebeenobserved sample thusis unlikely. more than once we have considered only the field measure- It is important to note that the error bars on Bz are ment with the smallest error bar. Note also that Eq. (7) small (σ smaller than 100G), except for a few cases - for hasbeenappliedtwice(seeFig.3):firsttomeasurementsof FORS data, only LSS3074 has large error bars (σ>100G) individual stars (results from high-resolution data for stars becauseitisafaintobject(B∼13magcomparedtoB=5−9 labelled “primary”, “secondary”, or “tertiary” in Table 3), fortheothers);forESPaDOnS/Narvaldata,largeerrorsare second tofield estimates for thesystems(combiningresults only found for all components of HD35652 and HD228854, fromTable2forlow-resolutionobservationsandthoselisted as well as the secondary stars of HD25638, HD57060, and under the“system”label in Table 3 for high-resolution ob- XZCep.Henceoursurveyisgenerally verysensitive.Inthe servations).Wethenderivethatthereis a90% chancethat high-resolutiondata,thisisalsoreflectedbythedetermina- the dipolar field strength at the pole is Bd≤140−220G, a tionoftheupperlimitsonthedipolarfieldswhichcouldhave very stringent limit well below Plaskett’s secondary field. remainedhiddeninthenoiseforeachtarget.Thelimitsare Hence our sample and Plaskett’s star, while having all un- wellbelowthefieldstrengthofPlaskett’sstar(Bd>2.85kG, derwent binary interactions, do not seem to form a specific Grunhutet al. 2013) for the vast majority of the cases (17 and coherent group of magnetic stars. out of 25 limits on individualcomponents). The objectiveof this paperis to test theplausibility of To shed more light onto our results, we can compare binary interactions as efficient processes at generating sta- themtothecurrentknowledgeoflarge-scalemagneticfields ble magnetic fields in massive systems. If this were true, in massive stars. Such fields are generally considered to we would expect all our targets to share similar properties, be (mostly) dipolar. In addition, except for ζOriA (a su- different from the usual massive star population. In such pergiant O-star with an extremely low field, Blaz`ere et al. a case, we can further statistically characterize our sample 2015) and βCMa (Fossati et al. 2015a), the detected fields arestrong,withBdlargerthan0.5kGforOandB0.5–1stars (e.g.Petit et al.2013).ThisisnotablythecaseofPlaskett’s secondary. The detection rate of such fields, found in large 2 Whenbothcomponentsaremagnetic,whentheobservationhas samplescomposedofsinglemassivestarsand/orlong-period been obtained at conjunction, and when the signatures of both massive binaries, amounts to 6–8% (Fossati et al. 2015b; components are of opposite shape and cancel each other either Grunhutet al. 2016). How does that compare to the case partiallyortotally,themagneticsignalcouldbedilutedandnot of massive interacting or post-interaction binaries? Overall, bedetectedifpresent.Notehoweverthatnotonlyisthiscasevery unlikelybutnone ofour targets was observedat conjunction, as wehaveonemagneticdetection(Plaskett’sstar)in16inves- demonstratedbytheverydifferentvelocitiesinTable3(seealso tigatedmultiplesystems(oursample+Plaskett’sstar,com- Fig.2foravisualinspectionofthecomponents’ separations). prising 35 individual stars in total). If we exclude the tar- MNRAS000,1–11(2017) 10 Y. Naz´e et al. sampleof15interactingandpost-interactionsystems.Ded- icated spectropolarimetry did not lead to any magnetic de- tectioninthesesystems.Infact,forthevastmajorityofour targets,thelongitudinalfieldsare<300Gandtheindividual upperlimitsonthedipolarfieldsarebelowthedipolarfield strengthofPlaskett’ssecondaryfor17outof25binarycom- ponents. Considering our targets as an homogeneous group ofsimilarobjects,aglobalstatisticalanalysisofthederived field values leads to a 90% upper limit on the (common) dipolar field strength of only 140–220G, underlining once again the dissimilarity between our targets and Plaskett’s star. Moreover, the low rate of magnetic detection in these interacting and post-interaction systems is compatible with the rates found from general surveys of O-stars. Together with the lack of magnetic detection in Be stars, some of whicharealsolikelyproductsofmass-transfer,thissuggests that binary interactions do not systematically trigger sta- ble,strongmagneticfieldsinsuchsystems,andthatafossil origin is still the best scenario for explaining the magnetic fieldsof massive stars. ACKNOWLEDGEMENTS YN and GR acknowledge support from the Fonds Na- Figure 3. Cumulative (top) or simple (bottom) distribution of tional de la Recherche Scientifique (Belgium), the Com- the Bz values normalizedtotheir errorbars forour sample(left, munaut´e Franc¸aise de Belgique, the PRODEX XMM con- forBzvaluesofindividualstars-labelled“primary”,“secondary”, tract(Belspo), and an ARC grant for concerted research or “tertiary” in Table 3; right, for Bz values estimated for the actions financed by the French community of Belgium wholesystem-fromTable2forFORSmeasurements andunder (Wallonia-BrusselsFederation).GAWacknowledgesDiscov- “system”label in Table 3 for high-resolution observations). The eryGrantsupportfromtheNaturalScienceandEngineering dotted red lines are for the Nz values, while the dashed vertical Research Council (NSERC) of Canada. linescorrespondtoa3σdetection level.Notehow themeasured valuesareallwithin2σfromzero. REFERENCES getswithlessstringentconstraints,thenthehigh-resolution datayieldonemagneticdetection(Plaskett’ssecondary)on Alecian,E., Neiner,C.,Wade, G. A.,et al.2015, New Windows 19stars.Ifwekeeponlyindividualmeasurementswithfield onMassiveStars,307,330 Antokhina,E.A.,SrinivasaRao,M.,&Parthasarathy, M.2011, limits lower than Plaskett’s field (2.85kG), one on 17 stars NewAstron.,16,177 for a limit of 2kG, and one on 9 stars for a limit of 1kG. Appenzeller,I.,&Rupprecht,G.1992,TheMessenger,67,18 Focusing only on the mass gainers (Plaskett’s secondary + Asensio Ramos, A., Mart´ınez Gonz´alez, M. J., & Manso Sainz, those of our sample - when known, see Table 1), there is R.2015,A&A,577,A125 one magnetic object out of 8 stars measured at high res- Auri`ere,M.2003,EASPublicationsSeries,9,105 olution with small error bars. In summary, whatever the Bagnulo, S., Szeifert, T., Wade, G. A., Landstreet, J. D., & case under consideration, these incidence levels appear en- Mathys,G.2002,A&A,389,191 tirely compatible with that derived from general samples, Bagnulo, S., Landolfi, M., Landstreet, J. D., et al. 2009, PASP, considering the small number statistics. The incidence rate 121,993 intheinteracting/post-interactionbinarysampleisthusnot Bagnuolo, W. G., Jr., Gies, D. R., & Wiggs, M. S. 1992, ApJ, 385,708 much larger than for the general O-star population. There- Bagnuolo, W. G., Jr.,Gies, D. R., Hahula, M. E., Wiemker, R., fore, there is no need to include additional processes linked &Wiggs,M.S.1994,ApJ,423,446 tobinaryinteractionstoexplainthepresenceofamagnetic Benaglia,P.2010,HighEnergyPhenomenainMassiveStars,422, field in Plaskett’s secondary. 111 Blaz`ere,A.,Neiner,C.,Tkachenko,A.,Bouret,J.-C.,&Rivinius, T.2015,A&A,582,A110 Braithwaite,J.2006,A&A,449,451 5 CONCLUSION Carrier,F.,North,P.,Udry,S.,&Babel,J.2002,A&A,394,151 Deˇgirmenci, O¨. L., Sezer, C., Demircan, O., et al. 1999, A&AS, Amongst magnetic O-stars, one object is particularly re- 134,327 markable: the secondary of Plaskett’s star, a binary system deMink,S.E.,Langer,N.,Izzard,R.G.,Sana,H.,&deKoter, which hasvery likely undergonemass-transfer in therecent A.2013,ApJ,764,166 past.Thissituation ledtotheintriguingpossibilitythatbi- Djuraˇsevi´c, G., Vince, I., Khruzina, T. S., & Rovithis-Livaniou, naryinteractionscouldbeasourceofmagnetisminmassive E.2009,MNRAS,396,1553 stars.Totestthevalidityofthisscenario,wehaveobserveda Donati,J.-F.2003,SolarPolarization,307,41 MNRAS000,1–11(2017)