Are magnetic fields universal in O-type multiple systems?

ABSTRACT

Although significant progress has been achieved in recent surveys of the magnetism in massive stars, the origin of the detected magnetic fields remains to be the least understood topic in their studies. We present an analysis of 61 high-resolution spectropolarimetric observations of 36 systems with O-type primaries, among them 10 known particle-accelerating colliding-wind binaries exhibiting synchrotron radio emission. Our sample consists of multiple systems with components at different evolutionary stages with wide and tight orbits and different types of interactions. For the treatment of the complex composite spectra of the multiple systems, we used a special procedure involving different line masks populated for each element separately. Out of the 36 systems, 22 exhibit in their least-squares deconvolution Stokes V profiles definitely detected Zeeman features, among them 7 systems with colliding winds. For 14 systems, the detected Zeeman features are most likely associated with O-type components, whereas for 3 systems we suggest an association with an early B-type component. For the remaining five systems, the source of the field is unclear. Marginal evidence for the detection of a Zeeman feature is reported for 11 systems and non-detection for 3 systems. The large number of systems with definitely detected Zeeman features presents a mystery, but probably indicates that multiplicity plays a definite role in the generation of magnetic fields in massive stars. The newly found magnetic systems are supreme candidates for spectropolarimetric monitoring over their orbital and rotation periods to obtain trustworthy statistics on the magnetic field geometry and the distribution of field strength.

1 INTRODUCTION

Most recent observational surveys indicate that ∼7 per cent of O-type stars host strong, large-scale organized fields (Grunhut et al. 2017; Schöller et al. 2017). The presence of a magnetic field in these stars can significantly modify their mass-loss and alter their angular momentum, subsequently affecting their evolution (e.g. ud-Doula, Owocki & Townsend 2008; Takahashi & Langer 2021). On the other hand, the principal question on the origin of the detected magnetic fields still remains unanswered, presenting currently the least understood topic in studies of massive stars. It has been argued that magnetic fields could be fossil, dynamo generated, or may be generated by strong binary interaction, i.e. in stellar mergers, or during a mass transfer or common envelope evolutionary phase (e.g. Tout et al. 2008; Ferrario et al. 2009; Schneider et al. 2016). The currently most popular scenarios consider strong binary interaction and merging events. Mass transfer or stellar merging may rejuvenate the mass gaining star, while the induced differential rotation is thought to be the key ingredient to generate a magnetic field (e.g. Wickramasinghe, Tout & Ferrario 2014).

In the context of this scenario, O stars are of special interest as the fraction of close binaries for these objects is so high that many of them interact with a companion during their main-sequence evolution through mass transfer or merging (e.g. Sana et al. 2012). As an example, by combining spectroscopy, interferometry, and photometry for the triple system HD 150136, Mahy et al. (2018) showed that the masses estimated through their analysis (42.8, 29.5, and 15.5 M_⊙) are much smaller than those expected from evolutionary models (55.4, 30.6, and 22.4 M_⊙). These results imply that evolutionary tracks computed for single stars cannot be used in studies of a majority of the O-type stars that interact with the other components in binary and multiple systems and that this interaction dominates the evolution of massive stars (e.g. Holgado et al. 2020). No search for the presence of a magnetic field in HD 150136 using high-resolution spectropolarimetry has been carried out yet. With respect to already known magnetic massive stars, Frost et al. (2021) combined 10 yr of spectroscopic and interferometric data for the magnetic Of?p binary HD 148937 with an orbital period of 29 yr and reported that the magnetic primary, although more massive, appears younger, suggesting that a merge or mass transfer took place in this system.

Offner et al. (2022) reported that the formation of multiple star systems with separations of about 0.1 pc takes place during the earliest phases of star formation and that the majority of such systems form and evolve to their final configuration during the time period spanned by the collapse of dense cores through the end of mass accretion. To study the impact of magnetic fields in such a process, Wurster, Bate & Price (2019) used non-ideal MHD simulations including ambipolar diffusion, Ohmic dissipation, and the Hall effect. The authors concluded that non-ideal magnetic processes do not inhibit the formation of multiple systems and that such systems form independent of the initial field strength.

Surprisingly, despite of the fact that multiplicity is widespread among massive stars, previous spectropolarimetric surveys of massive stars indicated that the incidence rate of magnetic component(s) in close binaries with P_orb < 20 d is very low, less than 2 per cent (e.g. Alecian et al. 2015). Of the dozen O stars currently confirmed to possess magnetic fields, about half are known to belong to wide binary systems where components can be treated in the spectropolarimetric analysis as single stars where the structure and evolution depend only on intrinsic stellar properties. However, for stars in dense clusters, or in binary and multiple systems, the evolution can be influenced by interaction with neighbouring stars. As of today, only one short-period binary system, Plaskett’s star, is known to contain a hot, massive, magnetic star (Grunhut et al. 2013). Obviously, these results are difficult to comprehend, pointing out the need for a re-analysis of available spectropolarimetric observations using a special procedure for the treatment of complex composite spectra of double and multiple systems.

In particular, the employment of high-resolution high signal-to-noise (S/N) observations is necessary to properly interpret the features appearing in the Stokes V spectra. Advantageously, the ESO archive contains a number of high-resolution (⁠|$R\approx 110\, 000$|⁠) HARPSpol (High Accuracy Radial velocity Planet Searcher polarimeter; Snik et al. 2008) observations acquired for O-type stars belonging to binary and multiple systems. Almost all of them were observed within the framework of the ESO Large programme ‘The B fields in OB stars’ (BOB). Spectropolarimetric observations of a number of multiple systems with O-type primaries were acquired with ESPaDOnS (the Echelle SpectroPolarimetric Device for Observations of Stars) with |$R\approx 65\, 000$| within the framework of the CFHT (the Canada-France-Hawaii Telescope) Large program ‘Magnetism in Massive Stars’ (MiMeS), available in the CFHT science archive. For our re-analysis of spectropolarimetric data, we initially concentrated on HAPRSpol data, which have a higher spectral resolution than ESPaDOnS observations. However, as in the process of this analysis a number of multiple systems indicated the presence of Zeeman features in their Stokes V spectra, we decided to also analyse ESPaDOnS observations of already known multiple systems with O-type primaries. In total, we were able to gather reduced spectropolarimetric data for 36 multiple systems with O-type primaries. Half of the targets were observed only on single epochs, 13 targets were observed twice, and 5 targets were observed three times.

Importantly, our target list includes 10 particle-accelerating colliding-wind binaries (PACWBs) exhibiting synchrotron radio emission. The catalogue of PACWBs based on radio observations was presented by De Becker & Raucq (2013). It included exclusively massive non-degenerate stars in binary or higher multiplicity systems with wind–wind interaction regions. The radiation mechanism of such systems requires two main ingredients: a population of relativistic electrons and a magnetic field, which can also play an important role in the process of the electron acceleration. Although no magnetic field detection was previously achieved in the few systems studied by Neiner et al. (2015), we believe that a magnetic survey of a more representative sample of PACWBs is necessary to investigate the role of magnetic fields in the generation of their synchrotron radio emission. Further, in addition to PACWBs, our sample contains two X-ray colliding wind binaries (X-rayCMBs), one binary suggested to present colliding winds, and one high-mass X-ray binary (HMXB). Apart from colliding wind binaries, we included in our sample a number of ordinary main-sequence systems with non-interacting components, systems called in the literature ‘twins’ of Plaskett’s Star that are defined as short-period massive binaries that interact or have interacted in the recent past, two systems with blue supergiant components, and two systems with Wolf–Rayet (WR) stars.

The structure of this paper is as follows. In Section 2, we give details on the available archival spectropolarimetric observations and describe the methodology of our analysis. The results of the magnetic field measurements for each target are presented in Section 3. In Section 4, we discuss the implication of the reported field detections on future studies of massive stars.

2 OBSERVATIONS AND ANALYSIS

The list of targets, their visual magnitudes, spectral classifications, and multiplicity information with related literature sources are presented in Table 1. Further, we added notes concerning the nature of the targets.

HARPSpol and ESPaDOnS spectropolarimetric observations usually consist of four subexposures recorded at different positions of the quarter-wave retarder plate. The HARPSpol spectra cover the spectral range 3780–6910 Å, with a small gap between 5259 and 5337 Å, whereas the ESPaDOnS spectra extend from 3700 to 10480 Å. The normalization of the spectra to the continuum level was described in detail by Hubrig et al. (2013a). The assessment of the longitudinal magnetic field measurements is presented in our previous papers (e.g. Hubrig et al. 2018; Järvinen et al. 2020). Similar to our previous studies, to increase the signal-to-noise ratio (S/N) by a multiline approach, we employed the least-squares deconvolution (LSD) technique. The details of this technique, as well as how the LSD Stokes I, Stokes V, and diagnostic null spectra are calculated, were presented by Donati et al. (1997).

As already mentioned above, a careful analysis using high-quality spectropolarimetric observations and the utilisation of a special procedure are crucial to detect and properly interpret the Zeeman features observed in the polarimetric spectra of multiple systems. In such systems, the shapes of blended spectral lines look different depending on the visibility of each component at different orbital phases. Additionally, the amplitudes of the Zeeman features are much lower in multiple systems in comparison to the size of these features in single stars. Thus, for the LSD analysis of multiple systems, it is preferable not to populate the line mask with weak spectral lines. In Fig. 1, we show the effect of the different strengths in the lines used in the magnetic field measurements on the amplitude of the corresponding Zeeman features and therefore on the measurement accuracy in the magnetic star HD 54879. The amplitude of the Zeeman feature corresponding to the weaker O ii line at 4602 Å is significantly smaller than for the stronger O ii line at 4596 Å having both a Landé factor of 0.9. When extracting the magnetic field strength of the two lines individually, the accuracy of the mean longitudinal magnetic field measurement <B_z> using the weaker line has decreased by more than a factor of 2. Moreover, in contrast to wide binary systems where the magnetic components can be treated in the spectropolarimetric analysis as single stars, Zeeman features are usually blended in composite spectra of tight multiple systems. The presence of severe distortions in the shapes of the Zeeman features hinders magnetic field detections in multiple systems and only dedicated spectropolarimetric monitoring over the orbital and rotation periods can permit to make solid conclusions on the magnetic field strength and geometry.

The effect of the blending and severe distortion of Zeeman features in multiple systems was demonstrated in the work of González et al. (2017), who studied the massive magnetic triple system HD 164492C in the Trifid nebula. This system consists of a close binary (component Ca) composed of B-type stars and a more distant rapidly rotating early-B He-rich magnetic component Cb. As all three components in the HARPSpol spectra remain blended at all observing epochs, Zeeman features in the Stokes V spectra exhibit complex configurations and atypical shapes compared to the characteristic Zeeman S-shape features usually detected in single magnetic massive stars. In Fig. 2, we present our LSD analysis of two HARPSpol observations of this system at two different epochs using a line mask containing He i and Si iii lines. The clearly visible features in the Stokes V spectra indicate the magnetic nature of at least one of the components, even if their shapes significantly deviate from the typical Zeeman Stokes V profiles observed in single magnetic stars. Due to severe blending, the presence of a magnetic field can remain undetected in some targets when only single or very few observations are available or when spectropolarimetric observations are obtained with low S/N. We note that a considerable departure of the Stokes V profiles from an S-shape can also be due to the presence of magnetic fields of mixed polarities over the visible hemisphere (see fig. 7 in Järvinen et al. 2020), the rotational Doppler effect in fast rotating stars (e.g. Mathys 1995), or the presence of gradients in velocity and magnetic field as a function of optical depth (e.g. Järvinen et al. 2021, and references therein).

Based on the complications existing in the analysis of the presence of magnetic fields in multiple systems discussed above, it appears reasonable to search in the Stokes V spectra of O-type systems for both Zeeman features with typical and atypical shapes, similar to the study of the triple system HD 164492C. To evaluate, whether the detected features are spurious or definite detections, we followed the generally adopted procedure to use the false alarm probability (FAP) based on the reduced χ² test statistics (Donati, Semel & Rees 1992): the presence of the Zeeman feature is considered as a definite detection (DD) if FAP ≤ 10⁻⁵, as a marginal detection (MD) if 10⁻⁵ < FAP ≤ 10⁻³, and as a non-detection (ND) if FAP> 10⁻³.

In the LSD technique, the line masks corresponding to the individual spectral types of the studied targets are generally constructed using the Vienna Atomic Line Database (VALD3; Kupka, Dubernet & the VAMDC Collaboration 2011). As our targets are binaries or higher order multiple systems exhibiting in their spectra very different spectral signatures depending on the spectral classification of the individual components, special care has to be taken to populate the line masks for each system. Only for a few systems in our sample the Doppler separation between the components is large enough to allow us to search for the presence of a magnetic field in each component using individual line masks. But before such masks for each component can be created, the line spectra of each component in each studied system has to be carefully surveyed to check for the presence and strength of spectral lines belonging to different elements.

The appearance of the spectra of the components usually depends not only on the orbital phase and the rotation rate but also on the evolutionary state: some of the stars are nitrogen enriched, exhibiting strong nitrogen lines clearly identified in their spectra. In some other systems, the components are strongly enriched in helium but depleted in carbon and oxygen. Overabundances and underabundances of different elements can also be caused by mass-transfer episodes expected in interacting and post-interaction systems. Multiple systems with short orbital periods undergo tidal interaction or mass transfer, and can suffer from X-ray irradiation. Obviously, combining lines of all elements together in the LSD line masks, without checking their presence and strength in the spectra, may lead to the dilution of the magnetic signal, or even to its partial cancellation.

As mentioned above, also the shape of blended spectral lines in composite spectra can be different depending on the visibility of each component. Therefore, for observations obtained at different epochs, it is sometimes useful to carry out the LSD analysis using a different line mask tailored specifically for the best visible component. In our LSD magnetic field analysis, to take into account the different spectral types of the components, we usually constructed separate masks for each individual element in different ionization states, namely He i and He ii masks, and masks with lines belonging to different ionization states of CNO elements and Si. As different spectral lines in massive stars have different formation depths, the use of masks constructed for individual elements to search for the presence of a magnetic field appears especially promising, in particular in view of the recent work by Järvinen et al. (2022), who reported on different field strengths measured using spectral lines belonging to different elements in the magnetic O-type star HD 54879: the measurements using helium lines that are formed higher in the stellar atmosphere compared to the formation depths of metal lines showed significantly lower field strengths than those obtained from the measurements using silicon and oxygen lines. Based on these results, it is clear that an LSD analysis using masks containing individual elements is extremely useful to select masks substantially contributing to the magnetic signal.

There are other factors playing an important role in the magnetic field measurements of massive O stars, such as the much lower number of spectral lines in comparison to less massive stars, causing a relatively high noise level in the LSD profiles. In systems with severe blending between the line profiles of the system components, it is sometimes possible to select unaffected or to a lesser extent blended lines if the spectral types of the components are different. Yet, already a small degree of contamination can have an impact on the magnetic field measurements. This is also true if some lines are contaminated by emission. Sometimes it is also not clear whether the observed features are of stellar origin, or whether they are caused by a nebular contamination around the targets. Massive O stars can be responsible for the ionization of gas in a surrounding nebula, but also for the creation of interstellar bubbles.

The results of our analysis of 36 binary and multiple systems using 61 HARPSpol and ESPaDOnS observations are presented in Table 2, where we indicate the spectropolarimeter used for the observations, the MJD values at the middle of the exposures, S/N values measured in the Stokes I spectra in the spectral region around 5000 Å, the applied line mask, and the FAP values with the detection flags DD, MD, or ND. For the few systems with components appearing clearly separated in the recorded spectra, we also present in the last column of this table the results of our measurements of the mean longitudinal magnetic field. Notably, in contrast to chemical abundance studies, spectrum disentangling is usually not needed to detect the presence of a magnetic field as Zeeman features should clearly stand out in high S/N LSD Stokes V spectra, if one of the binary or multiple system components possesses a magnetic field. However, if several spectropolarimetric observations phased over the orbital and rotation period for each target are available, an iterative disentangling process (e.g. González et al. 2017) should be applied to properly estimate the field strength and deduce the most likely field configuration.

In the following, we give a brief description of the targets and discuss the results of the magnetic field measurements for each target individually. The studied systems are divided into five different groups: the first group is composed of PACWBs (Section 3.1), the second of X-ray binaries with colliding winds (Section 3.2), the third group includes all other systems with definite detections of magnetic fields (Section 3.3), the fourth group contains all other systems with marginal field detections (Section 3.4), and the fifth group contains the remainder, all other systems with non-detections (Section 3.5). Mosaics with the plots presenting the LSD Stokes I, V, and diagnostic null N spectra calculated for the available HARPSpol and ESPaDOnS observations for the first and second groups are presented in Figs 3 and 4, respectively. The LSD spectra for the third group are presented in Fig. 5 and in Figs 7 and 8 for the fourth and fifth groups.

3 RESULTS OF OUR LSD ANALYSIS FOR INDIVIDUAL TARGETS

3.1 Particle-accelerating colliding-wind binaries

HD 36486 (δ Ori):According to Harvin et al. (2002), this system consists of an eclipsing binary with an orbital period of P_orb = 5.732 d and a distant early B-type tertiary at an angular separation of about 0.3 arcsec relative to the binary with an orbital period of the order of several thousand days. The authors suggested that the close binary may have suffered extensive mass-loss through binary interaction. The system was classified as a PACWB exhibiting synchrotron radio emission (De Becker & Raucq 2013). Neiner et al. (2015) analysed two Narval spectropolarimetric observations of HD 36486, but reported no magnetic field detection.

This system was observed three time with HARPSpol in 2013 December, where the first two observations were acquired on two consecutive nights. The LSD analysis of these spectropolarimetric observations was carried out using a mask containing only He i lines. As presented in Fig. 3 and Table 2, we detect a broad Zeeman feature in the second observation, achieving a marginal detection <B_z> = 358 ± 17 G with FAP = 4 × 10⁻⁴. Taking into account fig. 2 in Harvin et al. (2002), this feature most probably corresponds to the O9.5 II primary component.

HD 37468 (σ Ori): According to Maíz-Apellániz et al. (2019, and references therein), this system is a SB2 system with a visual B0.2 V(n) companion. The distant component is probably also an SB2 system (Maíz-Apellániz et al. 2021) in a nearly circular orbit around the close AaAb system with a period of 159.90 yr and a current separation of 0.26 arcsec. The AaAb has an orbital period of 143.198 d (Maíz Apellániz et al. 2018, and references therein). The system was mentioned as a PACWB exhibiting synchrotron radio emission by De Becker & Raucq (2013). Neiner et al. (2015) reported the absence of a magnetic field in this system using one ESPaDOnS observation acquired in 2008.

We confirm the field non-detection from the same archival observations, but achieve marginal magnetic field detections in more recent HARPSpol and ESPaDOnS observations. As presented in Fig. 3 and Table 2, the LSD analysis of the HARPSpol observation acquired in 2013 shows the presence of a Zeeman feature with FAP = 8 × 10⁻⁵ in the Stokes V profile calculated with a mask containing He i, He ii, Si iii, and Si iv lines. For the ESPaDOnS observation acquired in 2014, we detect a feature with FAP = 9 × 10⁻⁴ in the Stokes V spectrum using a mask with He i and He ii lines. Since the observed composite spectra show insufficient velocity separation of the system components, it is not clear to which component the detected Stokes V features belong, although their shifted positions relative to the centre of the corresponding Stokes I profiles seem to indicate the magnetic nature of the secondary B-type component.

HD 47839 (15 Mon): According to Maíz-Apellániz et al. (2019, and references therein), this system is a long-period SB1 binary in the cluster NGC 2264. From the analysis of TESS data, Burssens et al. (2020) found in HD 47839 the presence of stochastic low-frequency (SLF) photometric and β Cep variability, where the β Cep pulsations most probably belong to the fast-rotating early-type B component. Kaper et al. (1996) and Walborn (2006) reported that this system exhibits distinct peculiarities in its spectrum, which could be typical for stars possessing magnetic fields. HD 47839 was classified as a PACWB exhibiting synchrotron radio emission (De Becker & Raucq 2013). Neiner et al. (2015) used several Narval and one ESPaDOnS observations to search for the presence of a magnetic field and reported that no field was detected.

We downloaded from the CFHT archive the same ESPaDOnS observation from 2012 and detect, using a mask containing He i, He ii, and C iv lines, a clear Zeeman feature with FAP < 10⁻¹⁰ corresponding to the B1.5/2V component. As presented in Table 2 and in Fig. 3, we achieve a definite detection of the mean longitudinal magnetic field <B_z> = 282 ± 58 G. The previous measurements of the longitudinal magnetic field in HD 47839 by Hubrig et al. (2013b) using the full FORS 1 spectrum coverage showed a magnetic field of positive polarity at a 2.6σ significance level, <B_z> = 134 ± 52 G, whereas a magnetic field of the order of −160 G at a significance level of 4.4σ was reported using observations with the SOFIN spectropolarimeter attached to the 2.6 m Nordic Optical Telescope.

HD 93129A: According to Benaglia & Koribalski (2004), this visual multiple system stands out as the earliest, most luminous, and most massive star known in our Galaxy. Only 2.6 arcsec from HD 93129A, Walborn (1973) discovered a second star, HD 93129B, of spectral type O3.5 V((f)). The most recent study by Sana et al. (2014) showed that the separation between the Aa and Ab components is about 30 mas. This system dominates the core of Trumpler 14 and, in total, six companions within 5 arcsec from the central star were reported in the literature (e.g. Sana et al. 2014). HD 93129 was identified as a PACWB by De Becker & Raucq (2013).

The LSD analysis of the single HARPSpol observation of this system obtained in 2016 was carried out using a mask containing He ii lines. As presented in Table 2 and in Fig. 3, we achieve a definite detection with FAP < 10⁻¹⁰. Given the location of the detected Zeeman feature in the Stokes V spectrum relative to the broad Stokes I line profile presenting a blend composed of the Aa and Ab components, it is possible that the secondary component with spectral type O3.5V((f)) possesses a magnetic field.

HD 151804 (HR 6245): This system was classified as a PACWB exhibiting synchrotron radio emission by De Becker & Raucq (2013). Kervella et al. (2022) show a significant detection of a companion using Gaia EDR3 data (S/N = 5.2) but not using Gaia DR2 data (S/N = 0.82). This star falls into the 9 per cent of their sample that has an improved S/N after the release of EDR3. According to Burssens et al. (2020), this system shows SLF photometrical variability.

HD 151804 was previously observed with FORS 1 in spectropolarimetric mode by Hubrig et al. (2008). However, no field detection at a significance level of 3σ was achieved in this study. Also, no magnetic field was detected in the study of Neiner et al. (2015), who analysed one HARPSpol observation acquired in 2011. As presented in Table 2 and in Fig. 3, analysing the same observation, we achieved a marginal detection with FAP = 6 × 10⁻⁴ using a mask containing He i and He ii lines. Given the location of the detected feature in the Stokes V spectrum relative to the broad Stokes I line profile, it is not clear whether this feature is due to the presence of a background or foreground object or an unresolved second body in the system.

HD 152248 (V1007 Sco): According to Sana et al. (2001), this PACWB eccentric eclipsing SB2 binary has an orbital period close to 6 d (Hill et al. 1974). It belongs to the young rich open cluster NGC 6231. Both components are close to filling their Roche lobe at periastron passage (Sana et al. 2001). However, the authors do not find any evidence that mass transfer is currently taking place in this system. On the other hand, it cannot be excluded that such a process has already been taking place in the past.

As presented in Table 2 and in Fig. 3, using ESPaDOnS observations of HD 152248 obtained in 2013 and 2014, we detect distinct features in the LSD Stokes V spectra with FAP < 10⁻¹⁰ using a mask with He i, He ii, and Si iv lines for the first observation and FAP = 4 × 10⁻⁷ using a mask containing He i, He ii, and O iii lines for the second observation. Given the change of the position of the detected features in velocity space from about −230 to −50 km s⁻¹ just over 1 d, and the fact that these features do not correspond to the position of the LSD Stokes I profiles of any of the O-type components, it is possible that a yet undetected third component is present in this system.

HD 152408 (HR 6272, WR79a): This target is a member of NGC 6231 and was classified as O8Iape by Sota et al. (2014). It is listed in the Seventh Catalogue of Galactic WR stars presented by van der Hucht (2001). Skinner et al. (2010, and references therein) reported on the X-ray detection in HD 152408 with XMM–Newton, emphasizing that this detection is of special importance as it is achieved in the object belonging to the latest WN subtype in the WN 7–9 range. The authors also reported on the existence of a fainter near-IR source (2MASSJ165458.09-410903.6) at an offset of 4.8 arcsec from HD 152408 (see also Mason et al. 1998). According to De Becker & Raucq (2013), this target is a PACWB. No magnetic field was detected by Neiner et al. (2015) using three consecutive ESPaDOnS observations obtained during the same night on 2012 July 6.

Hubrig et al. (2008) reported on the detection of a mean longitudinal magnetic field of the order of 90 G at a significance level of 3σ using low-resolution FORS 1 spectropolarimetric observations. For our LSD analysis, we used the same three consecutive ESPaDOnS observations with a time lapse of 0.56 h over 1.7 h during the same night as Neiner et al. (2015) and a mask containing He i and He ii lines for the first observation and a mask containing He ii and C iv lines for the second and the third observations. As presented in Table 2 and in Fig. 3, we detect in the LSD Stokes V spectra small Zeeman features with FAP = 5 × 10⁻⁴ for the first observations and FAP < 10⁻¹⁰ for both the second and third observations, indicating the magnetic nature of this target. Given the slightly decentred position of the detected features relative to the observed LSD Stokes I profile, it is quite possible that the detected Zeeman features correspond to an unresolved second body.

HD 164794 (9 Sgr): According to Fabry et al. (2021), this target is an astrometric SB2 system in the Lagoon Nebula with a 53 M_⊙ primary and a 39 M_⊙ secondary with an orbital period of 3261 d. The system was classified as a PACWB exhibiting synchrotron radio emission by De Becker & Raucq (2013). Rauw et al. (2016) reported that a maximum in the X-ray emission during the periastron passage is coming from shocked gas in the interaction zone of the stellar winds. Neiner et al. (2015) have not detected the presence of a magnetic field in this system using several ESPaDOnS observations and one HARPSpol observation.

The presence of a weak magnetic field of about 200 G was reported by Hubrig et al. (2008) using low-resolution spectropolarimetric observations with FORS 1 at the VLT. Also subsequent high-resolution SOFIN and HARPSpol observations analysed using the moment technique indicated the presence of a magnetic field of a similar order (Hubrig et al. 2013b). As presented in Table 2 and in Fig. 3, our LSD analysis of the two high-resolution HARPSpol observations acquired in 2015 and 2016 and one ESPaDOnS observation from 2005 shows for all these observations definite detections of clear Zeeman features of positive and negative field polarities with FAP < 10⁻¹⁰. For the first HARPSpol observation and the ESPaDOnS observation, we used a mask containing He i, He ii, N iv, and O iii lines, whereas for the second HARPSpol observation we used a mask containing He ii and N iv lines. As the components in HD 164794 appear overlapped in all LSD Stokes I spectra and their K1 and K2 amplitudes are rather low, 25 km s⁻¹ for the primary and 39 km s⁻¹ for the secondary (Rauw et al. 2016), it is difficult to decide about the origin of these features shifted to the red relative to the centers of the Stokes I profiles. We believe that the origin of these features is not related to a magnetic background or foreground object: as is shown in Fig. 3, their position in velocity space is slightly changing over the years in the range of 150−180 km s⁻¹ and thus we cannot exclude the presence of an additional component in this system.

HD 167971 (MY Ser): According to Ibanoglu et al. (2013), this SB3 system is one of the rare massive O-type triple systems where the secondary and the tertiary components compose an eclipsing binary. It is the brightest member in the NGC 6604 cluster embedded in the Ser OB2 association. The brighter component of this triple system is an overcontact eclipsing binary with an orbital period of 3.32 d, just filling the entire outer contact surface with a fill-out factor of 0.99. The outer pair has a period of about 22 yr. HD 167971 was classified as a PACWB exhibiting synchrotron radio emission by De Becker & Raucq (2013). Neiner et al. (2015) reported a magnetic field non-detection using one ESPaDOnS spectrum acquired in 2013.

For our LSD analysis, we used two archival ESPaDOnS spectra, one from 2013 and another one from 2015. As presented in Table 2 and in Fig. 3, all three components are well visible in both ESPaDOnS observations. For the first observation, we detect the presence of a Zeeman signature with FAP < 10⁻¹⁰ only in the weakest red component and measure a very strong magnetic field <B_z> = 1324 ± 582 G. Given the rather large uncertainty affecting this measurement, it is necessary to confirm the presence of such a strong magnetic field with future observations obtained at higher S/N. Zeeman features for all three components are detected in the second ESPaDOnS observation where a marginal detection with FAP = 9 × 10⁻⁴ corresponding to <B_z> = − 977 ± 437 G is achieved for the blue component, and a definite detection with FAP = 6 × 10⁻⁶ corresponding to <B_z> = − 57 ± 33 G is achieved for the central component showing the deepest intensity profile in the Stokes I spectrum. No detection is achieved for the third component. We also detect a structure in the diagnostic null N spectrum calculated for the second observation. The origin of this structure corresponding to the central component is probably related to the presence of pulsations. For both observations, we used a line mask containing He ii, C iv, and N iii lines.

HD 190918 (WR 133): This SB2 system, also known as WR 133, was classified as a PACWB exhibiting synchrotron radio emission by De Becker & Raucq (2013). According to Richardson et al. (2021), the orbital period is 112.8 d with a moderate eccentricity of 0.36, and a separation of 0.79 mas on the sky. Since the derived masses are low compared to the spectral types of the components, the authors suggest that WR 133 should have been formed through binary interactions.

This system was mentioned as a non-magnetic PACWB by Neiner et al. (2015), who used one ESPaDOnS observation acquired in 2010. We used for our LSD analysis the same ESPaDOnS observation from 2010 and another one from 2008 using a mask with He i and He ii lines. As presented in Table 2 and in Fig. 3, there is evidence for the presence of a definite Zeeman feature in the observation acquired in 2008 with FAP < 10⁻¹⁰ corresponding to <B_z> = − 355 ± 44 G. A marginal detection with FAP = 3 × 10⁻⁴ corresponding to <B_z> = 56 ± 8 G is achieved for the observation acquired in 2010. Since the components of this system are overlapped in the LSD Stokes I spectra, it is not clear to which component the detected features belong. Notably, studies of systems with WR components are of special interest: according to Dsilva et al. (2022), the multiplicity properties of WR stars, which are the immediate progenitors of black holes, directly affect the properties of black-hole binaries (e.g. through black-hole kicks in a core-collapse scenario), and hence the predictions for gravitational-wave progenitors.

3.2 X-ray binaries with colliding winds

HD 1337 (AO Cas): According to Linder (2008), this O9.2II + O8V((f)) system is probably an interacting eclipsing binary in a semidetached configuration with an orbital period of 3.5 d, where the secondary is probably impacted by an accretion stream. Portegies Zwart et al. (2002) listed this target among the systems with X-ray colliding winds, where X-rays are produced in the collisions between the winds. Using Hipparcos photometry, Lefévre et al. (2009) reported that HD 1337 is variable with a mean amplitude of 0.198 mag.

The LSD analysis of the single ESPaDOnS observation of HD 1337 obtained in 2009 was carried out using a mask containing He i, He ii, and O iii lines. As presented in Table 2 and in Fig. 4, we detect a clear Zeeman feature with FAP < 10⁻¹⁰ in the LSD Stokes V spectrum achieving a definite detection of the mean longitudinal magnetic field with a strength <B_z> = − 274 ± 52 G in the more luminous O9.2 II component.

HD 48099: This SB2 system with a 3.1 d orbital period belongs to the Mon OB2 association (Garmany et al. 1980). The detection of a strongly clumped wind and a nitrogen abundance of about eight times the solar value in the primary probably indicate that the system went through an evolutionary phase of mass exchange or mass loss through Roche lobe overflow (Mahy et al. 2010). Berdyugin et al. (2016) investigated the structure of the binary system by measuring linear polarization that arises due to a light scattering process. Their model fit suggested that light is scattered on a cloud produced by the colliding stellar winds. The available X-ray data provided additional evidence for the existence of colliding stellar winds in this system. Burssens et al. (2020) reported that one-sector TESS observations of HD 48099 show SLF variability.

Two components are clearly visible in the LSD Stokes I profile presented in Fig. 4. The secondary component shows a narrow-lined profile, while the primary is rotating fast with a |$v\, \sin \, i$| of about 91 km s⁻¹ (Mahy et al. 2010). As reported in Table 2, we detect a Zeeman feature with FAP = 7 × 10⁻⁶ in the LSD Stokes V profile calculated for the single available ESPaDOnS observation obtained in 2012, using a line mask containing He i, He ii, C iv, and Si iv lines. The position of the Zeeman feature relative to the composite Stokes I spectrum indicates that the magnetic field is detected in the broad-lined O5V primary component.

HD 149404 (V918 Sco): According to Rauw et al. (2001), the secondary in this system with an orbital period of 9.81 d seems to be the most evolved component and its current evolutionary status could best be explained if the system has undergone a Roche lobe overflow episode during the past. Thaller (1998) suggested that the observed double-peaked structure of the H α emission is due to a colliding wind interaction. Also, Rauw et al. (2001) favour a model where the H α emissions arise in the arms of a colliding wind shock region.

We used for the LSD analysis of one HARPSpol observation obtained in 2016 a mask containing He i, He ii, C iv, Si iii, and Si iv lines, whereas for the ESPaDOnS observation acquired in 2014 the mask was populated using He i, He ii, Si iii, and Si iv lines. As presented in Table 2 and in Fig. 4, we achieve for both observations marginal detections with FAP = 5 × 10⁻⁴ and FAP = 7 × 10⁻⁴, respectively.

HD 153919 (4U 1700-37): According to Falanga et al. (2015), this runaway HMXB is the potentially most massive primary in the known sample of HMXBs in the Galaxy. Previous studies suggested that the companion in the HD 153919 system with P_orb = 3.41 d is either a massive neutron star or a low-mass black hole (Clark et al. 2002; Abbott et al. 2020). Ankay et al. (2001) proposed that HD 153919 originates from the OB association Sco OB1 of which NGC 6231 is the suggested core. The authors also reported that the progenitor of this system went into supernova within 6 Myr. According to van der Meij et al. (2021), HD 153919 might be a prototype in the Milky Way for the progenitor of gravitational wave events such as GW190412. Also a presence of X-ray flares was reported in different studies (e.g. Boroson et al. 2003). Based on a model of an entirely wind-accreting system, Brinkmann (1981) suggested that the flares were associated with accretion from the magnetotail of a neutron star.

Using one low-resolution FORS 2 spectropolarimetric observation of HD 153919 obtained in 2010, Hubrig et al. (2011) reported the presence of a mean longitudinal magnetic field <B_z> = 213 ± 68 G. In contrast, no field was detected in the single FORS 2 observation acquired in 2013 (Hubrig et al. 2013b). As presented in Table 2 and in Fig. 4, no obvious Zeeman features were detected in the HARPSpol observation obtained in 2016 and in the ESPaDOnS observation from 2011, using a line mask containing He i, He ii, C iv, and O iii lines. It is possible that the tiny features visible in both observations close to 250 km s⁻¹ in velocity space are due to a background or a foreground object.

3.3 Other targets with definitely detected magnetic fields

HD 17505: In contrast to Maíz-Apellániz et al. (2004), who reported for this system the spectral classification O6.5III((f))n + O8V, in a more recent work Sota et al. (2014) characterized it as at least a triple system. HD 17505 is located in the centre of the cluster IC 1848 within the Cassiopeia OB6 association. According to Raucq et al. (2018), the components of the inner binary with an orbital period of 8.57 d are both well inside their Roche lobe suggesting that this system has not yet experienced binary interaction, while the outer orbit has a period of less than 61 yr.

The LSD analysis of the single ESPaDOnS observation of HD 17505 obtained in 2009 was carried out using a mask containing He i, He ii, and O iii lines. As presented in Table 2 and Fig. 5, the detected Zeeman feature in the Stokes V spectrum is highly significant with FAP < 10⁻¹⁰. Given the fact that the LSD Stokes I profile consists of several components that are heavily blended, it is unclear to which component the detected Zeeman feature belongs.

HD 35921 (LY  Aur): According to Maíz-Apellániz et al. (2019, and references therein), this system is composed of an SB2 eclipsing system with a circular orbit with a period of about 4 d and a distant B0.2 IV companion located 0.6 arcsec away and which is a SB1 system with an orbital period of 20.46 d. Using Hipparcos photometry, Lefévre et al. (2009) reported that HD 35921 is variable with a mean amplitude of 0.722 mag over the orbital period of the SB2 eclipsing system. Mayer et al. (2013) reported that the eclipsing SB2 binary belongs to the rare class of early-type contact systems in a phase of non-conservative mass exchange.

The LSD analysis of the single ESPaDOnS observation of HD 35921 obtained in 2011 was carried out using a mask containing He i, He ii, O ii, and O iii lines. As presented in Table 2 and Fig. 5, the detected Zeeman feature in the Stokes V spectrum is highly significant with FAP < 10⁻¹⁰. According to fig. 3 in Mayer et al. (2013) displaying composite Stokes I profiles in the spectra of this system, the measured longitudinal magnetic field <B_z> = 538 ± 172 G corresponds to the secondary component.

HD 36879: This target is a runaway candidate from the analysis of Hipparcos proper motions (Mdzinarishvili & Chargeishvili 2005). Walborn & Panek (1984) reported that HD 36879 shows unexplained spectral variability and peculiarities, among them peculiar narrow, variable Si iv emission lines at λ1394 and λ1403.

A magnetic field of HD 36879 at a 3.5σ significance level was for the first time detected in 2007 using low-resolution FORS 1 spectropolarimetric observations (Hubrig et al. 2008). Its presence was later confirmed with the moment technique in high-resolution spectropolarimetric observations acquired with SOFIN (Hubrig et al. 2013b). Our analysis of one HARPSpol observation acquired in 2015 and one ESPaDOnS observation from 2009 using the LSD technique with a line mask containing He i, He ii, C iv, N iv, O iii, and Si iv lines shows the presence of a clear Zeeman feature in the LSD Stokes V spectrum calculated for the HARPSpol observation. As presented in Fig. 5 and Table 2, the Zeeman feature with FAP = 6 × 10⁻¹⁰ detected in this observation corresponds to <B_z> = 301 ± 42 G. Only a small feature with FAP = 4 × 10⁻² was detected in the LSD Stokes V spectrum calculated for the ESPaDOnS observation using the same line mask.

Interestingly, the HARPSpol observation indicates that the shapes of the line profiles for different elements are variable over a time interval of 20.5 min, the time difference between the beginning of subexposures. The shapes of the line profiles of He ii and O iii lines show a different character of variability compared to the variability of He i and Si iv lines. Notably, spectrum variability based on two different observing epochs separated by hundreds of days was already reported in the past by Hubrig et al. (2013b).

HD 57061 (τCMa): According to Stickland, Lloyd & Sweet (1998), this target is the brightest in NGC 2362 and produces a bow shock in nearby gas. NGC 2362 is a rich young open cluster where HD 57061 is the only target evolved away from the zero-age main sequence. Using spatially resolved spectroscopy, Maíz-Apellániz & Barbá (2020) reported that HD 57061 is a high-order multiple system. The components Aa and Ab are separated by 0.151 arcsec and have O-type spectral classifications with the brighter Aa component classified as an O9.2 of an uncertain luminosity classification and the weaker Ab component classified as O9 III. The Aa component itself is composed of an inner spectroscopic binary with a 154.92 d period. In addition, one of the components shows eclipses with a much shorter period of 1.282 d (van Leeuwen & van Genderen 1997). Additional visual companions B, C, D, and E are listed in different literature sources.

Two components are clearly visible in the LSD Stokes I profile calculated using the HARPSpol observation obtained in 2015. We detect a clear Zeeman feature in the LSD Stokes V profile of the secondary calculated using a mask containing He i, C iii, and O iii lines. As presented in Fig. 5 and in Table 2, the detected Zeeman feature is highly significant with FAP = 3 × 10⁻¹⁰ and corresponds to a mean longitudinal magnetic field <B_z> = −285 ± 144 G.

HD 92206C: According to Campillay et al. (2007), this system with an orbital period of 2.02 d is the second brightest member of the open cluster NGC 3324 after HD 92206 AB. Walborn (1982) reported that the spectrum of HD 92206C displays very strong, broad hydrogen lines, possibly similar to those in the Orion Trapezium cluster, in particular θ¹ Ori C, and that the appearance of these lines is indicative of extreme youth. Based on the BESO (Bochum Echelle Spectrograph for Observatorio Cerro Armazones) spectra, Mayer et al. (2017) reported the clear presence of a third companion visible in the He i profiles at an estimated separation from the SB2 system by less than 1 arcsec. Using low-resolution FORS 2 spectra, Hubrig et al. (2013b) reported the detection of a mean longitudinal magnetic field <B_z> = 204 ± 46 G using metal, helium, and hydrogen lines.

LSD Stokes I, V, and diagnostic null N spectra have been calculated for the two available HARPSpol observations of HD 92206C obtained in 2013 on two consecutive nights. As shown in Fig. 5 and in Table 2, using a mask containing He i lines, a clear Zeeman feature in the LSD Stokes V spectrum with FAP < 10⁻¹⁰ corresponding to the third component is detected in the observations obtained on the first night. No feature in the LSD Stokes V spectrum using a line mask with He i and He ii lines is detected on the second night, where the third component is not visible in the LSD Stokes I spectrum.

HD 101205 (V871 Cen): According to Sana et al. (2014, and references therein), this target is a complex multiple system with three visual pairs previously resolved at separations of 0.36 arcsec (A,B), 1.7 arcsec (AB,C), and 9.6 arcsec (AB,D). Further, the authors report that the A,B pair contains an eclipsing binary with an orbital period of about 2 d and a spectroscopic binary with a period of 2.8 d. However, it is impossible to decide which component of the A,B pair is the eclipsing one and which is the spectroscopic one.

Our LSD analysis of the single available HARPSpol observation of HD 101205 obtained in 2013 with a mask containing He i, He ii, and O iii lines reveals a Zeeman feature with FAP = 9 × 10⁻⁶. As shown in Fig. 5, all components of the A,B pair are blended with each other. Therefore, it is not clear, to which component the detected Zeeman feature should be assigned.

HD 152218A (V1294 Sco): According to Rosu et al. (2022), this system is an eclipsing eccentric binary system with an orbital period of 5.6 d and rotation periods for the primary and the secondary of 2.69 and 2.56 d, respectively. It belongs to the rich open cluster NGC 6231 located at the core of the Sco OB1 association. From the analysis of TESS data, Kolaczek-Szymanski et al. (2021) reported that HD 152218A exhibits a characteristic shape of brightness changes close to the periastron passage. During this passage the variable tidal potential can drive tidally excited oscillations (TEOs), which are usually gravity modes. HD 152218A also shows low-amplitude intrinsic variability in the low-frequency range, and can be regarded as a SPB pulsator.

As presented in Fig. 5, in the first HARPSpol observation obtained in 2013 both components of this SB2 system are blended with each other, whereas in the second HARPSpol observation obtained in 2016 they are clearly separated. Only for the second observation we detect a clear Zeeman feature with FAP < 10⁻¹⁰ corresponding to a longitudinal magnetic field <B_z> = − 307 ± 94 G in the secondary component. As shown in Table 2, we used for the LSD analysis of the first observation a mask containing He i, He ii, C iv, Si iii, and Si iv lines and for the second observation a mask containing He i, He ii, C iii, C iv, and Si iv lines.

HD 152236 (ζ⁰¹ Sco): In the recent study by Mahy et al. (2022, and references therein), this system belonging to the open cluster NGC 6231 is considered as a candidate LBV with a companion at a distance of about 11.5 mas and a Δm of 6.3 mag. LBVs are suggested to be a brief evolutionary phase describing massive stars in transition to the WR phase. Using linear polarization observations, Clark et al. (2012) detected position-angle rotation through the H α emission line, which can be associated either with large-scale, axisymmetric structures, or irregular wind clumps.

As shown in Fig. 5 and in Table 2, we used for the first and second observations obtained with HARPSpol in 2014 a mask containing He i lines, whereas for the third observation obtained in 2015 we used a mask containing He i, Si iii, and Si iv lines. Distinct features in the LSD Stokes V spectra with FAP < 10⁻¹⁰ and FAP = 2 × 10⁻⁷ are detected for the second and third observations, respectively. Given the position of the definitely detected features in the Stokes V spectrum, it is possible that the very weak secondary component possesses a magnetic field.

HD 152246: Nasseri et al. (2014) reported that HD 152246 belonging to the Sco OB1 association is at least a high-mass hierarchical triple system consisting of an inner pair Ba,Bb with an orbital period of 6 d and with approximately 17 and 6 M_⊙ component mass, respectively, and an outer component A with a mass of approximately 20 M_⊙ (see also Mayer et al. 2017). The A,B pair in HD 152246 with an orbital period of about 470 d has an eccentricity significantly higher than expected from the period-eccentricity diagram of other O-type systems (Sana et al. 2012). Nasseri et al. (2014) suggested that HD 152246 was probably born as a hierarchical quadruple system formed by two close binaries and that the Aa,Ab initial binary may have been driven into coalescence because of stellar evolution.

One HARPSpol observation of HD 152246 was obtained in 2013 and a second one in 2016. As shown in Fig. 5 and in Table 2, our LSD analysis using for the first observation a mask containing He i, He ii, C iii, and C iv, and a mask with He i, He ii, C iv, Si iii, and Si iv lines for the second observation reveals the presence of distinct features in the LSD Stokes V spectra with FAP = 6 × 10⁻⁷ for the first observation and FAP < 10⁻¹⁰ for the second observation. Given the location of the detected features in the LSD Stokes V spectra relative to the LSD Stokes I profiles, it is possible that they belong to the faster rotating outer component A.

HD 155775 (V1012 Sco): Lefévre et al. (2009) reported that HD 155775 belonging to the Sco OB1 association is an eclipsing binary with a period of 1.515 d and an amplitude of 0.061 mag. Our LSD analysis of the single available HARPSpol observation of HD 155775 acquired in 2013 shows that the primary O-type component and the lower luminosity secondary component are clearly separated in the LSD Stokes I spectrum. As presented in Fig. 5 and in Table 2, a definite Zeeman feature with FAP < 10⁻¹⁰ corresponding to a mean longitudinal magnetic field <B_z> = − 9 ± 37 G is detected in the primary component using a mask containing He i, N iii, and Si iii lines.

HD 166546: Our LSD analysis of the single available HARPSpol observation obtained in 2013 using a mask with exclusively Si iii lines and presented in Fig. 6 confirms that this target is a binary system and could even be a triple system. As presented in Fig. 5 and in Table 2, the LSD analysis using a mask containing He i and Si iii lines reveals a definite Zeeman feature with FAP < 10⁻¹⁰, probably corresponding to the third component in this system. It can however not be excluded that the detected feature is related to a background or foreground object.

HD 209481 (14 Cep): Using Hipparcos photometry, Lefévre et al. (2009) reported that this semidetached Algol eclipsing system with an orbital period of 3.1 d is variable with a mean amplitude of 0.099 mag. Mahy et al. (2011, and references therein) showed that the Roche lobe filling component, either the primary or the secondary, exhibits a strong helium and nitrogen mass fraction enhancement. The system is most likely in a configuration where the mass transfer happened while both components were still in the hydrogen-burning core phase (e.g. Howarth et al. 1991). Thus, this rare system composed of two evolved early-type stars is an important target for a study of binary interaction and its impact on massive star evolution.

One ESPaDOnS spectrum of HD 209481 was acquired in 2011 and another one in 2015. As shown in Fig. 5 and in Table 2, our LSD analysis using a mask containing He i lines for the first observation in 2011 reveals the presence of two Zeeman features in the LSD Stokes V spectrum, one with FAP < 10⁻¹⁰ located close to the blue wing of the LSD Stokes I profile and another one with FAP = 3 × 10⁻⁴ corresponding to the center of the LSD Stokes I profile. For the second observation, using a mask containing He i and He ii lines, a Zeeman feature is detected with FAP < 10⁻¹⁰, corresponding to the red wing of the LSD Stokes I profile. Given the location of the detected features relative to the observed composite LSD Stokes I profiles, it is not clear whether they may belong to the additional component(s) in this system.

HD 305524: This target belonging to the open cluster Collinder 228 is not known to be a binary or a multiple system. According to Levato et al. (1990), HD 305524 shows radial velocity variability, probably caused by the presence of a companion. Nazé et al. (2012) used two low-resolution FORS 2 spectropolarimetric observations of HD 305524 to search for the presence of a magnetic field, but no field at a significance level of 3σ was detected. One available HARPSpol observation of HD 305524 obtained in 2016 was downloaded from the ESO archive. We note that the shape of the Stokes I profile calculated in our LSD analysis using a mask containing He i and He ii lines indicates that this target is probably a SB2 system. As presented in Fig. 5 and in Table 2, we detect a definite Zeeman feature with FAP < 10⁻¹⁰, probably corresponding to the secondary less luminous component in this system.

3.4 Other targets with marginal magnetic field detections

HD 36512 (υ Ori): According to Burssens et al. (2020), TESS observations of this SB1 system suggest the presence of SLF variability, but of much lower amplitude than typical for stars with similar spectral type. The authors also report on the dominant frequency of 0.306 d⁻¹. Smith (1981) announced the presence of non-radial pulsations. Although HD 36512 is mentioned as a SB1 system in Burssens et al. (2020), our LSD analysis using a mask containing He i lines shows that HD 36512 is a SB2 system.

The results of our LSD analysis of available HARPSpol observations, where two observations were acquired in 2013 and one in 2015, are presented in Fig. 7 and Table 2. We achieve a marginal detection with FAP = 3 × 10⁻⁴ corresponding to <B_z> = 123 ± 10 G for the first observation using a mask containing He i, He ii, C iii, O ii, O iii, Si iii, and Si iv lines. A second marginal detection is achieved for the third observation, with FAP = 3 × 10⁻⁴ corresponding to <B_z> = − 19 ± 2 G using a mask containing He i and He ii lines. No detection was achieved for the second HARPSpol observation.

HD 37366: According to Maíz-Apellániz et al. (2019, and references therein), this SB2 system has a dim companion 0.6 arcsec away. The components of HD 37366 are in an eccentric orbit (P_orb = 31.9 d) and show different line broadening with projected rotational velocities of 30 and 100 km s⁻¹ (Boyajian et al. 2007). The secondary is a broad-lined early B-type main-sequence star. Our LSD analysis of the single ESPaDOnS observation of HD 37366 obtained in 2008 using a mask containing He i, He ii, C iii, C iv, Si iii, and Si iv lines reveals the presence of a Zeeman feature with FAP = 2 × 10⁻⁴ corresponding to a marginal detection of a magnetic field in the sharp-lined O9.5IV primary component (see Fig. 7 and Table 2).

HD 75759 (HR 3525): According to Thackeray (1966), this SB2 system which is the central exciting source in the large H i region Gum 17, has a long orbital period of 33.3 d and a large eccentricity. Burssens et al. (2020) mention for HD 75759 the presence of SLF variability.

As presented in Fig. 7 and Table 2, our LSD analysis of the HARPSpol spectrum obtained in 2013 using a mask containing He i, C iii, N iii, Si iii, and Si iv lines indicates a marginal detection in the secondary component with FAP = 2 × 10⁻⁴ corresponding to <B_z> = − 92 ± 9 G. Also the LSD analysis of the second available HARPSpol spectrum obtained in 2015 using a mask containing He ii and C iv lines reveals a distinct feature in the Stokes V spectrum with FAP = 3 × 10⁻⁴, but this time in the primary component. For this observing epoch, we measure <B_z> = − 113 ± 10 G.

Using lower-resolution spectropolarimetric observations with the FOcal Reducer low dispersion Spectrograph in spectropolarimetric mode (FORS 2; Appenzeller et al. 1998), Schöller et al. (2017) reported for this system a magnetic field non-detection with <B_z> = − 103 ± 58 G employing for the measurements metal, He, and hydrogen lines. This non-detection can probably be explained by the too low resolution of FORS 2, only of the order of about 2000, which does not permit to study the presence of a magnetic field in each component separately.

HD 147165 (σ Sco):According to Pigulski (1992, and references therein), this target is a quadruple system, where the primary is a high-amplitude β Cep-type star in a SB2 system with an orbital period of 33 d. The tertiary component is only about 0.4 arcsec away, whereas the fourth component is located at a much larger distance of about 20 arcsec. Further, the authors reported that the primary in the SB2 system is in the shell-hydrogen burning phase presenting the light-time effect contributing to the observed changes of the pulsation period. Bodensteiner et al. (2018) visually inspected Wide-field Infrared Survey Explorer22 |$\mu$|m observations and detected unaligned offset bow shocks around HD 147165, probably indicating binary interaction. A spectral classification O9.5 + B7 was reported by Beavers & Cook (1980). A later study of this system by Maíz-Apellániz et al. (2021) suggests that the Aa component is a single-lined B1 III object and the secondary should be a B dwarf with a spectral type around B1.

One HARPSpol observation has been acquired in 2014 and a second one in 2015. For the LSD analysis, we used for the first observation a mask containing He i lines, whereas for the second observation the mask was populated using He i, He ii, C ii, N ii, Si iii, and Si iv lines. As presented in Fig. 7 and Table 2, a small Zeeman feature for the first observation with FAP = 1 × 10⁻⁵, probably corresponding to the primary component, is visible in the Stokes V spectrum indicating a marginal detection. No magnetic field is detected for the second observation. The calculated diagnostic null N spectra exhibit strong distinct features shifted to the blue in the first observation and to the red in the second observation. We assume that they are related to the high-amplitude pulsations in the β Cep primary.

HD 152590 (V1297 Sco): Although Pourbaix et al. (2004) and Burssens et al. (2020) list this object as an eclipsing SB1 system with a period of 4.5 d, HD 152590 is clearly a SB2 system with a weak component easily recognizable in the HARPSpol spectra using a LSD analysis with a mask containing He i lines.

As presented in Fig. 7 and Table 2, using for our LSD analysis of the two available HARPSpol observations acquired in 2013 and 2015 a mask containing He i, He ii, O iii, and Si iv lines, we achieve a marginal detection with FAP = 1 × 10⁻⁴ in the observation acquired in 2013.

HD 167659: According to Sana et al. (2014), this target in the young open cluster NGC 6231 is a binary system containing a faint companion with a Δm of 2.54 mag. The presence of a companion is also detectable in the blue wing of the LSD Stokes I profile in the single available HARPSpol observation of HD 167659 acquired in 2013. As shown in Fig. 7 and Table 2, using for our LSD analysis a mask containing He ii lines, we detect a small Zeeman feature with FAP = 1 × 10⁻⁵ corresponding to the red wing of the LSD Stokes I profile. It is not clear whether this feature belongs to an additional component in this system or to a background or foreground magnetic star.

HD 204827: Kervella et al. (2022) show a significant detection of a companion using Gaia EDR3 data (S/N = 3.2), but not using Gaia DR2 data (S/N = 2.47). This star falls into the 9 per cent of their sample that has an improved S/N after the release of EDR3. Peter et al. (2012) studied massive binaries in the Cepheus OB2/3 region and reported that HD 204827 has a companion at a separation of 0.09 arcsec.

The presence of both components is clearly visible in the single available ESPaDOnS observation obtained in 2011. As shown is Fig. 7 and Table 2, our LSD analysis using a mask containing He i, He ii, and O iii lines reveals a marginal detection of a feature with FAP = 4 × 10⁻⁴, probably corresponding to the component exhibiting sharp lines in the LSD Stokes I spectrum.

3.5 Other targets with no detections of magnetic fields

HD 93161A: HD 93161 is a visual binary with a separation of 2 arcsec in the open cluster Tr 16. According to Nazé et al. (2005), the primary HD 93161A is a non-interacting SB2 system with an orbital period of 8.6 d. This system must be older and more evolved than main-sequence stars in the core of the Trumpler 16 cluster (Rauw et al. 2009).

As presented in Fig. 8 and in Table 2, our LSD analysis of the single HARPSpol observation of HD 93161A obtained in 2016 using a mask containing He i, He ii, C iv, and O iii lines shows a non-detection, although we cannot exclude the possible presence of a very weak broad feature in the LSD Stokes V spectrum. Future high S/N observations of this system at different epochs are required to decide on the presence of a magnetic field.

HD 191201: This target is a visual binary with an SB2 component. The SB2 detached system with a previously known late O-type primary and an early B-type secondary on an orbital period of 8.3 d is a member of the open cluster NGC 6871 (e.g. Burkholder et al. 1997). According to Maíz-Apellániz et al. (2019), the B component at a distance of 1 arcsec and a Δm of 1.8 mag is an O star with spectral type O9.7 III.

As presented in Fig. 8 and in Table 2, using a mask with He i, Si iii, and Si iv lines in our LSD analysis of the ESPaDOnS observation of HD 191201 acquired in 2011, a weak feature corresponding to a third component is clearly visible. However, the FAP value calculated for this feature, FAP > 4 × 10⁻³, indicates a non-detection. As there is also a feature in the diagnostic null N spectrum, it is possible that the third weak component is pulsating. Additional observations are necessary to better characterize the third component and to search for the presence of a magnetic field.

4 DISCUSSION

We present for the first time a search for magnetic fields in a sample of binary and multiple systems with O-type primaries, utilizing a special procedure not previously applied in spectropolarimetric studies of massive multiple systems. The difficulty in detecting magnetic fields in such systems is related to a number of obstacles in the treatment of their polarimetric spectra, among them a low number of useful spectral lines and severe blending between the spectral lines of the individual components. In the presented study, only eleven systems show in their spectra sufficient component separations to allow us to study the presence of a magnetic field in the individual components. Moreover, since nearly all massive systems belong to open clusters that usually form the densest parts in OB associations, the polarimetric spectra can be contaminated by background or foreground stars and this contamination can lead to a misinterpretation of the detected Zeeman features in the Stokes V spectra.

Our spectropolarimetric survey of 36 systems reveals that 22 systems exhibit in their LSD Stokes V profiles definitely detected Zeeman features with FAP ≤ 10⁻⁵. Among them, for 14 systems the detected features are most likely associated with an O-type component, whereas for 3 systems we suggest an association with a B-type component. For five systems with definite Zeeman feature detections, the quadruple system HD 101205, the PACWB systems HD 152248 and HD 164794, the presumably triple system HD 166546, and the Algol system HD 209481, it is not clear whether the origin of the Zeeman features is related to one of the already known companions, or to additional companions, background, or foreground stars. Since all five stars belong to open clusters or associations, the presence of background or foreground stars cannot be excluded.

For eight systems, HD 36512, HD 37366, HD 75759, HD 147165, HD 204827, the PACWBs HD 36486 and HD 151804, and the CWB system HD 149404, only marginal evidence for the detection of a Zeeman feature in the O-type components was found. Marginal field detections have been achieved for two systems with early B-type components, HD 152590 and the PACWB HD 37468. The correspondence to the known system components is not clear for the system HD 167659 with a marginal detection, belonging to the young open cluster NGC 6231. Detections were not achieved in three systems, in the visual binary HD 93161A, the HMXB system HD 153919, and in the triple system HD 191201. A schematic representation of the results of our LSD analysis is shown in the pie chart of Fig. 9.

Definite detections of Zeeman features were achieved in two systems classified in the literature as WR stars, HD 152408 (=WR 79a) and HD 190918 ( =WR 133) with <B_z> = − 355 ± 44 G. Reports on the detection of magnetic fields in WR stars, which are descendants of massive O stars, are still very scarce. Due to the strong line broadening, predominantly low-resolution spectropolarimetric data obtained with FORS 2 were used in the past (e.g. Hubrig et al. 2016; Hubrig et al. 2020). We also report on definite detections of a Zeeman feature in two observations of the system containing the candidate LBV HD 152236. These two features can probably be associated with a very weak secondary component in this system.

Interestingly, out of the 36 systems, 6 binary and multiple systems, HD 75759, HD 92206C, HD 152218A, HD 152246, HD 152590, and HD 164794, were previously observed using low-resolution (R ≈ 2000) spectropolarimetry with FORS 2, but, in contrast to our study, no detection was achieved for any of these systems (e.g. Schöller et al. 2017). These results imply that exclusively high-resolution spectropolarimetric observations should be used to study magnetic fields in massive binary and multiple systems.

In this work, we present the first observational evidence that PACWBs host magnetic stars that are possibly responsible for the generation of their synchrotron radio emission. The LSD analysis of these systems presented in Fig. 3 and Table 2 shows that among the ten PACWBs, definite detections of Zeeman features are achieved in seven systems. Marginal detections were achieved for three systems, HD 36486, HD 37468, and HD 151804. For the systems HD 36486, HD 167971, and HD 190918, it was possible to estimate their mean longitudinal magnetic field strengths, which are in the range from a few hundred Gauss up to the highest value of 1.3 kG measured in one of the three components of the blue supergiant system HD 167971. This triple system is of extraordinary interest for future studies of magnetism in O stars because for all three components either definite or marginal detections are achieved. Furthermore, this system with two colliding-wind regions (De Becker 2018) contains interacting components: the third component is orbiting the short period (P_orb = 3.3 d) eclipsing overcontact pair with both components filling their corresponding Roche lobes. Obviously, the confirmation of the presence of such a strong field of 1.3 kG in this system using additional polarimetric observations obtained at higher S/N is highly desirable. Definite detections of Zeeman features have also been achieved in the two interacting X-rayCWB systems HD 1337 and HD 48099 having short orbital periods of 3.5 d and 3.1 d, respectively.

Especially studies of magnetic fields in systems with periods short enough to allow for interaction between the components are important because they can permit to get a deeper insight into the role of the magnetic field not only in binary and multiple system evolution, but also in the context of supernova progenitors. According to Sana et al. (2012), massive stars in close binaries with primaries of mass |$M\gtrsim 15\, {\rm M}_{\odot }$| will interact with a stellar companion before they explode as core-collapse supernovae with neutron stars or black holes as end products. A significant fraction of the targets in our sample have rather short orbital periods, below 10 d. This is probably due to the fact that systems with massive components usually belong to open clusters: according to van Leeuwen & van Genderen (1997), massive binary and multiple systems play an important role for the dynamical evolution of star clusters because they can serve as an energy exchange mechanism, leading to the creation of very tight binaries and escaping cluster members, but also to temporary and usually very unstable three to four star configurations. Nine targets in our sample, HD 1337, HD 35921, HD 36486, HD 48099, HD 149404, HD 152246, HD 167971, HD 190916, and HD 209481 are mentioned in the literature as interacting systems where the components, depending on their evolutionary state, undergo or in the past have already undergone episodes of conservative or non-conservative mass transfer or even coalescence.

The role of magnetic fields in binary interactions as well as magnetic field geometries in components of binary systems and multiple systems are currently almost unexplored: among O-type systems, only Plaskett’s star with P_orb = 14.4 d is known to harbour a broad-line secondary with a longitudinal magnetic field of a semi-amplitude of about 500 G (Grunhut et al. 2013). However, significant difficulties in the interpretation of the observed Stokes V profiles were recently reported by Grunhut et al. (2022). As for systems with B-type stars, Hubrig et al. (2014a) reported on a study of the field geometry of a magnetic Ap secondary in the close SB2 system HD 161701 with a B-type HgMn primary: the obtained measurements indicated that the field on the surface of the Ap star permanently facing the primary component is positive and that it is negative on the far side. The authors suggested that the alignment of the magnetic axis with the orbital radius vector may indicate that the generation of the magnetic field was a dynamic process during tidal synchronization. Obviously, binary systems and multiple systems with more than one component possessing a magnetic field are of considerable interest for observations of the configuration and strength of their fields. In this study, we identified only two such systems, the SB2 system HD 75759 with P_orb = 33.3 d for which marginal detections were achieved in both components, and the triple system HD 167971 with P_orb = 3.3 d for the inner eclipsing binary and a P_orb of about 22 yr for the outer system. As already mentioned above, for this triple system we achieved definite and marginal detections in all three spectroscopic components.

Notably, previously reported spectropolarimetric observations of magnetic Bp and O stars in double and multiple systems seem to indicate that for the majority of the systems only one of the components is magnetic. No magnetic field was detected in the primary component of Plaskett’s star (Grunhut et al. 2022). Hubrig et al. (2014b) presented low-resolution FORS 2 spectropolarimetric observations of three components in the multiple system ADS 10991 in the Trifid nebula with an age of just a few 0.1 Myr: the O7.5 star HD 164492A, the B1 triple system HD 164492C (González et al. 2017), and the Herbig Be star HD 164492D. No magnetic field at a significance level of 3σ was detected in HD 164492A and HD 164492D, whereas a longitudinal magnetic field of the order of 600 G is present in the primary of the system HD 164492C. Among the SB2 systems with early B-type primaries studied by Shultz et al. (2018), most systems contain only one magnetic component.

Our survey of high-resolution ESO HARPSpol and CFHT ESPaDOnS archival spectropolarimetric observations of O-type binary and multiple systems revealed a sizable sample of such systems with potential magnetic components, pointing out that the incidence rate of magnetic fields in massive binaries has probably been greatly underestimated. This is a puzzling and rather unexpected result, indicating that multiplicity probably plays a role in the generation of magnetic fields in massive stars. Assuming that interactions between the components in massive binary and multiple systems constitutes a dominant process for the generation of magnetic fields, we should indeed expect that a majority of these systems will harbour a magnetic component.

Evidently, to answer the principal question of the possible origin of magnetic fields in massive stars, further work based on dedicated spectropolarimetric monitoring of the studied systems is needed. As most of the targets have been observed only once or twice, our LSD analysis solely targeted the detection of their magnetic fields using the presence of conspicuous features in their LSD Stokes V spectra. Clearly, to establish the magnetic nature of the surveyed targets with more confidence, additional spectropolarimetric observations are needed to investigate the temporal variability of the detected Zeeman features. Furthermore, to better understand the physical processes playing a role in massive stars, it is also important to investigate the three-dimensional structure of these systems based on monitoring of the strength of the magnetic field, the intensity variability of different spectral lines, and of the radial velocity shifts. Spectropolarimetric monitoring will allow the disentangling of the individual stellar spectra and will provide crucial information on the magnetic field geometries, rotation periods, and the geometry of the orbits as well as on the peculiar properties and the evolutionary history of the components from the study of their radii, abundances, and rotation rates. We also plan to use TESS and Kepler data to look more closely at the O-type systems studied in this paper in order to further constrain their properties, as well as search for low-frequency power excess associated with massive stars (e.g. Bowman et al. 2019).

ACKNOWLEDGEMENTS

We thank the referee, Gautier Mathys, for insightful comments. Our work is based on observations made with ESO telescopes at the La Silla and Paranal observatories under programmes 187.D-0917(A) and 191.D-0255 and observations collected at the Canada-France-Hawaii Telescope (CFHT), which is operated by the National Research Council Canada, the Institut National des Sciences de l’Univers of the Centre National de la Recherche Scientifique of France, and the University of Hawaii.

DATA AVAILABILITY

The HARPS and ESPaDOnS spectropolarimetric observations underlying this article can be obtained from the ESO and CFHT archives, which can be found under https://www.eso.org/UserPortal/ and https://www.cadc-ccda.hia-iha.nrc-cnrc.gc.ca/en/cfht/, respectively.

References