The very slow rotation of the magnetic O9.7 V star HD 54879

ABSTRACT

The first FOcal Reducer low dispersion Spectrograph (FORS 2) spectropolarimetric observation of the longitudinal magnetic field of HD 54879 of the order of −600 G with a lower limit of the dipole strength of ∼2 kG dates back to 2014. Since then, observations showed a gradual decrease of the absolute value of the mean longitudinal magnetic field. In the course of the most recent monitoring of HD 54879 using FORS 2 spectropolarimetric observations from 2017 October to 2018 February, a longitudinal magnetic field strength change from about −300 G down to about −90 G was reported. A sudden increase of the absolute value of the mean longitudinal magnetic field and an accompanying spectral variability was detected on 2018 February 17. New FORS 2 spectropolarimetric data obtained from 2018 December to 2019 February confirm the very slow magnetic field variability, with the field decreasing from about 150 G to −100 G over 2 months. Such a slow magnetic field variability, related to the extremely slow rotation of HD 54879, is also confirmed using high-resolution High Accuracy Radial velocity Planet Searcher in polarimetric mode and Echelle SpectroPolarimetric Device for the Observation of Stars spectropolarimetry. The re-analysis of the FORS 2 polarimetric spectra from 2018 February indicates that the previously reported field increase and the change of the spectral appearance was caused by improper spectra extraction and wavelength calibration using observations obtained at an insufficient signal-to-noise ratio. The presented properties of HD 54879 are discussed in the context of the Of?p spectral classification.

1 INTRODUCTION

Only few massive O-type stars possess strong large-scale organized magnetic fields. One of them, the O9.7 V star HD 54879 was detected as magnetic by Castro et al. (2015) using the FOcal Reducer low dispersion Spectrograph (FORS 2; Appenzeller et al. 1998) and the High Accuracy Radial velocity Planet Searcher in polarimetric mode (HARPSpol; Snik et al. 2008). The authors reported the presence of a −600 G longitudinal magnetic field with a lower limit of the dipole strength of ∼2 kG. Numerical MHD simulations describing the interaction of the magnetic field of this star with its wind were recently discussed by Hubrig et al. (2019). In the same work, Hubrig et al. (2019) presented 26 new FORS 2 spectropolarimetric observations of HD 54879 obtained from 2017 October 4 to 2018 February 21 showing a change of the mean longitudinal magnetic field from about −300 G down to about −90 G. The authors also reported on a sudden, short-term increase of the absolute value of the mean longitudinal magnetic field on the night of 2018 February 17, measuring a longitudinal magnetic field of −833 G. The inspection of the FORS 2 spectrum acquired during that night also indicated a change in spectral appearance and a decrease of the radial velocity by several 10 km s⁻¹.

Since such an unusual behaviour was not observed for other massive magnetic stars, we decided to monitor the magnetic field variability again and obtained data between 2018 December and 2019 February. In the following, we report on our results obtained from these most recent FORS 2 spectropolarimetric observations, indicating a very slow change of the magnetic field related to slow stellar rotation. High-resolution spectropolarimetric observations confirm the results of the magnetic field measurements made using low-resolution FORS 2 spectropolarimetric observations. Our new results are also discussed in the context of the previous detection of a short-term field increase and the change of spectral appearance.

2 OBSERVATIONS AND MAGNETIC FIELD ANALYSIS

In total, 24 new spectropolarimetric observations of HD 54879 with FORS 2 were obtained between 2018 December 16 and 2019 February 15 in the framework of our programme 0102.D-0285 executed in service mode at the european southern observatory's (ESO) very large telescope (VLT). Among them, five observations were obtained outside the weather specifications and appear underexposed with signal-to-noise ratios (S/N) less than 1300, whereas one observation yielded saturated spectra. In the following, we will discuss the 18 unsaturated high S/N spectra, while we discuss the five lower S/N spectra in more detail in Section   5.

All spectra were observed with the GRISM 600B and the narrowest available slit width of 0|${^{\prime\prime}_{.}}$|4 to obtain a spectral resolving power of R ≈ 2000. The observed spectral range from 3250 to 6215 Å includes all Balmer lines, apart from H α, and numerous helium lines. Further, in our observations, we used a non-standard readout mode with low gain (200kHz,1×1,low), which provides a broader dynamic range, hence allowing us to reach a higher S/N in the individual spectra. The extraction of the parallel and perpendicular beams on the FORS 2 raw data was carried out using a pipeline written in the MIDAS environment by T. Szeifert, (in the following called the MIDAS pipeline). More details on the observational and reduction methods can be found in Hubrig et al. (2019) and references therein.

The results of the measurements are listed in Table 1, where the first two columns give the modified Julian dates (MJDs) for the middle of the exposure and the S/N values of the spectra. The results of our magnetic field measurements, those for the entire spectrum and those using only the hydrogen lines, are presented in Columns 3 and 4, followed by the measurements using all lines in the null spectra. Null spectra N are calculated as pairwise differences from all available Stokes V profiles so that the real polarization signal should cancel out.

The distribution of the mean longitudinal magnetic field values as a function of MJD is presented in Fig. 1. In the top panel of this figure, we show all FORS 2 longitudinal magnetic field measurements acquired between 2014 February 8 and 2019 February 20. The most recent measurements obtained from December 2018 to February 2019 are presented in the bottom panel. While the few observations from 2014 and 2015 indicated a mean longitudinal magnetic field of the order of −500 to −900 G, we observe in the last years a significantly weaker longitudinal magnetic field with values between −300 and +150 G. The strongest longitudinal magnetic field of positive polarity of 150 G was measured on the night 2018 December 16. After this date, the longitudinal magnetic field is gradually decreasing, reaching a value of about −100 G on 2019 February 15. These measurements together with those presented by Hubrig et al. (2019) indicate that the rotation period of HD 54879 is very long, at least several years, and that from 2017 October to 2019 February we are observing the star at rotational phases with the best visibility of the magnetic equator. A small scatter in the measurements of the magnetic field is frequently observed in massive O-type stars and is most likely due to contamination by the immediate stellar environment with a denser cooling disc, confined to the magnetic equatorial plane (e.g. Hubrig et al. 2015; Shultz & Wade 2017). However, the scatter we see in the measurements presented in Fig. 1 is within the uncertainties represented by the error bars.

To check the consistency of the low-resolution FORS 2 magnetic field measurements with measurements using high-resolution spectropolarimeters, we overplotted in Fig. 1 these FORS 2 measurements with high-resolution longitudinal magnetic field measurements obtained from archival data. We discuss these observations in the following section.

3 HIGH-RESOLUTION SPECTROPOLARIMETRY OF HD 54879

As already reported by Järvinen et al. (2017), three spectropolarimetric observations of HD 54879 were obtained with the HARPS polarimeter attached to ESO’s 3.6 m telescope (La Silla, Chile) at a spectral resolution of about 115 000 on 2014 April 22, and on 2015 March 11 and March 14. The published values of the mean longitudinal magnetic field showed a change from −578 G in 2014 to −487 G in 2015. Recently, nine Echelle SpectroPolarimetric Device for the Observation of Stars (ESPaDOnS; Donati et al. 2006) spectra with a spectral resolution of 65 000 obtained between 2014 November and 2018 January became publicly available in the CFHT archive. Among them, two observations were obtained on the same night on 2014 November 9. To measure the longitudinal magnetic field, we employed the least-squares deconvolution (LSD) technique, allowing us to achieve a much higher S/N in the polarimetric spectra. The LSD technique combines line profiles (assumed to be identical) centred on the position of the individual lines and scaled according to the line strength and sensitivity to a magnetic field. The details of this technique and of the calculation of the Stokes I and Stokes V parameters can be found in the work of Donati et al. (1997). The line masks employed in the measurements of the longitudinal magnetic fields, one with 127 lines including the He and the metal lines and the second one with 113 metal lines, were constructed using the Vienna Atomic Line Database (Kupka et al. 2011) and the stellar parameters T_eff = 30.5 kK and |$\log \, g=4.0$| reported by Shenar et al. (2017). The calculated LSD Stokes I and Stokes V profiles for each observing epoch are presented in Fig. 2. The measurements using both line masks are listed in Table 2 along with the MJDs for the middle of the exposure, the exposure times, and the S/N values of the spectra. In all cases, the false alarm probability was less than 10⁻¹⁰. Notably, the distribution of data points obtained for the high-resolution spectropolarimetric observations in Fig. 1 fits the low-resolution FORS 2 measurements very well, indicating that the measurements using low-resolution spectropolarimetry are fully consistent with those made with other instruments.

4 ROTATION PERIOD AND SHORT-TERM VARIABILITY

The magnetic field measurements presented in Fig. 1 at MJDs between 56696 and 58529 suggest that the rotation period of HD 54879 is very long. Assuming that the negative field extremum reaches a value of −500 to −900 G, measured in 2014 February, and not having reached yet this value by 2019 February, it is very likely that the rotation cycle is longer than 5 yr.

Additional evidence for a very long rotation period follows from the consideration of the variability of the H α line profiles. Previous studies of magnetic massive stars revealed a correlation between the absolute value of the mean longitudinal magnetic field and the strength of the H α emission in the sense that the strongest H α emission appears at phases of maximum absolute value of the mean longitudinal magnetic field (Sundqvist et al. 2012). Also, the light curves in the visible and the X-ray emission strengths usually display a positive correlation with the absolute value of the mean longitudinal magnetic field. In Fig. 3(a), we present profiles of H α emission observed in the high-resolution ESPaDOnS spectra. The shape of the H α emission observed in these spectra obtained from 2014 November to 2018 January is extremely variable, changing between a triple-peak and a double-peak emission line profile. Such a strongly variable shape of the H α emission reflects a composite structure of the surrounding circumstellar material. We see that the lowest emission contributions with the deepest V-shaped central depression have appeared in 2017 January. This is in agreement with the H α equivalent width (EW) measurements presented in table A1 in the work of Shenar et al. (2017), who detect the lowest emission EW in 2017 February. The observations of the H α profile variability presented by these authors in their fig. 13 go back to 2009. However, none of the displayed H α profiles shows such a low emission level as observed in 2017 January in the ESPaDOnS spectra. Since other massive magnetic stars show a variability of the H α emission strength with the stellar rotation period, the fact that the lowest H α emission profile has been observed only once since 2009 may suggest that the rotation period of HD 54879 is longer that 9 yr.

In Figs 3(b) and (c), we show the variablity of H α on different, much shorter time-scales. The two profiles overplotted in Fig. 3(b) were observed in ESPaDOnS spectra on a time-scale of 17 d, whereas those presented in Fig. 3 (c) have a time lapse of only 88 min. A small variability in the line cores of C iii 5696 and Si iii 5697 on a time-scale of a few minutes can be seen in Fig. 3(d), which is, however, on the same scale as the noise in the continuum and will need higher S/N data for verification. The observed short-term spectral variability is not expected to be caused by changes in the star’s stellar parameters, but might be related to the wind or immediate environment of the star, including a denser cooling disc confined to the magnetic equatorial plane (e.g. Martins et al. 2012). In the absence of sufficient centrifugal support due to the slow rotation, material accumulated in the disc and located below the corotation radius falls back on to the stellar surface.

It is of interest that the typical Of?p star HD 108 with a rotation period of several decades shows He i and Balmer line variability on time-scales of days, similar to the observed short-term variability of HD 54879. A short-term variability on the time-scale of hours was reported for another Of?p star, CPD −28° 2561, by Hubrig et al. (2015). The variability domain of minutes or tens of minutes in magnetic O-type stars remains, however, until now unexplored due to the lack of suitable spectroscopic and photometric time-series.

Apart from H α, also the He ii 4686 line is a very sensitive stellar wind indicator and can be used to study the variability in high-resolution HARPSpol and ESPaDOnS spectra. In Fig. 4, we present our measurements of EWs and radial velocities of this line in spectra obtained from 2014 October to 2018 January. The EWs measured on the HARPSpol spectra appear larger than those measured in the ESPaDOnS spectra, probably due to the much higher spectral resolution of HARPSpol. In spite of the large dispersion of measurements in the spectra obtained in the last years, it appears quite possible that the distribution of the data points for the ESPaDOnS spectra indicates a small decrease of the strength of the He ii 4686 line in 2018. The results of the radial velocity measurements presented in this figure on the right-hand side show a slightly decreasing trend, which, however, is of the order of the measurement accuracies.

The first radial velocity measurement reported in the literature, v_rad = 15.6 ± 1.4 km s⁻¹ by Neubauer (1943), was followed by the work of Boyajian et al. (2007), who reported v_rad = 35.4 ± 1.4 km s⁻¹. Castro et al. (2015) compared in their table 4 the radial velocity measurement of 29.5 ± 1.0 km s⁻¹ from one HARPS spectrum with other measurements in the literature and concluded that HD 54879 could be a member of a long-period binary system. Our own measurements using the ESPaDOnS Stokes I spectra and employing the LSD technique indicate that the radial velocity slightly decreased to v_rad = 27.0 ± 0.1 km s⁻¹ measured on the most recent ESPaDOnS spectra obtained in 2018 January. In Fig. 5, we present the compiled literature measurements complemented by the three HARPS and nine most recent ESPaDOnS observations. A very slow but gradual decrease in radial velocity can probably be considered as real in view of the high measurement accuracy reached in high-resolution spectropolarimetric observations using the LSD technique and indicates that HD 54879 could be a member of a binary system with a very long orbital period. Obviously, future monitoring of HD 54879 with high-resolution spectroscopy is necessary to confirm the observed decrease in EWs and radial velocities.

We also tested whether it is possible to use the low resolution of FORS 2 observations to study the variability of the He ii 4686 line over the time interval from 2017 October 4 to 2019 February 15. The radial velocity changes for the He ii 4686 line in FORS 2 spectra acquired between 2017 October 4 and 2018 February 21 are presented in Fig. 6 on the left-hand side and those measured on the spectra obtained between 2018 December 16 and 2019 February 15 on the right-hand side of the same figure. It is obvious that the scatter of the data points presented in both panels is too big to allow us to make any conclusion on the variability of the radial velocities. This result is in line with the previous inconclusive search of periodicity by Hubrig et al. (2019) using FORS 2 radial velocities.

On the left-hand panel of Fig. 7, we present the EWs of the He ii 4686 line measured in the FORS 2 spectra obtained from 2017 October 4 to 2018 February 21 and those from 2018 December 16 to 2019 February 15 on the right-hand side. No significant changes in line intensities are detected during both observing runs.

5 STUDYING FORS 2 SPECTRA WITH INSUFFICIENT SIGNAL-TO-NOISE RATIO

Rather unexpectedly, our analysis of the underexposed FORS 2 spectra from the observations between 2018 December and 2019 February showed that they are similar to the spectra obtained in the previously reported observations acquired with an S/N of about 1130 on 2018 February 17 (Hubrig et al. 2019), for which an increase of the absolute value of the mean longitudinal magnetic field, a change in spectral appearance, and a decrease of the radial velocity by several 10 km s⁻¹ were reported. While such a behaviour was observed only once in the FORS 2 observations acquired from 2017 October to 2018 February, it was detected in all five recent observations with an S/N below 1300. As an example, we present in Fig. 8 two FORS 2 observations recorded with an S/N of about 830 on the same night on 2019 January 1 and separated by just 25 min. The first observation at MJD 58485.1143 resulted in a 〈B_z〉 measurement compatible with a non-detection, while the second observation at MJD 58485.1234 gave 〈B_z〉 = −1300 ± 220 G, using all lines. This spurious field increase at MJD 58485.1234 is accompanied by a spurious change of spectral appearance, including the increase of all absorption hydrogen and He i lines and the decrease of higher ionization lines like He ii, C iii, and Si iv, and by a radial velocity shift of over 100 km s⁻¹.

In Fig. 9, we show the overplotted Stokes I spectra for He i 4922 line profiles corresponding to subexposures in observations of varying quality recorded on two different nights, the observation obtained with an S/N = 2340 at MJD 58480.0776 and that with an S/N = 1120 at MJD 58487.2366. According to the atmospheric parameters presented by Castro et al. (2015), HD 54879 has already evolved from the ZAMS and is passing through the β Cephei instability strip. However, we are convinced that such short-term spectral variability is not real, as we never detected it in previous observations of this and other targets and it solely appears in data with insufficient S/N.

Further, we investigated if an alternative spectrum extraction could result in more stable wavelengths also for the spectra with insufficient S/N. For this, we employed the ESO FORS pipeline based on the Reflex toolkit and tailored to the polarimetric mode of FORS 2 (PMOS). However, as can be seen in Fig. 10, the ESO FORS pipeline has issues with wavelength stability even for higher S/N data. While most of the higher S/N spectra gave similar results when determining the longitudinal magnetic field, when compared to the MIDAS pipeline results, we concluded that the ESO FORS PMOS pipeline in its current form is not delivering proper results.

The observations with lower S/N at MJD 58487.2366 show a strong shift in the Stokes I profile recorded in one subexposure. For the same night, we also present in Fig. 11 ordinary and extraordinary circularly polarized line profiles of the He i 4922 line. A clear shift in the spectra for one subexposure indicates that the presence of a strong longitudinal magnetic field is spurious.

The inspection of all FORS 2 extracted spectra indicates no impact of high airmass or variable seeing during the observations. The short-term spectral and magnetic variability is detected only in observations with an S/N below 1300. We conclude that the spectral extraction with the MIDAS pipeline is not working properly for underexposed spectra and is producing spectral artefacts. While different scenarios were previously discussed by Hubrig et al. (2019) in an attempt to interpret the observation on 2018 February 17, it appears now that this observation was affected by an imperfect spectral extraction.

6 DISCUSSION

The new FORS 2 spectropolarimetric data obtained from 2018 December to 2019 February confirm the very slow magnetic field strength variability in HD 54879. While the few observations from 2014 and 2015 indicated a mean longitudinal magnetic field value of the order of −500 to −900 G, we observe in the last years a significantly weaker magnetic field with a mean longitudinal magnetic field value between −300 and +150 G. The strongest longitudinal magnetic field of positive polarity of 150 G was measured on the night of 2018 December 16. After this date, the longitudinal magnetic field is gradually decreasing, reaching a value of about −100 G on 2019 February 15. This slow magnetic field variability, related to the extremely slow rotation of HD 54879, is also confirmed using high-resolution HARPS and ESPaDOnS spectropolarimetry. Assuming that the negative field extremum reaches a value of −500 to −900 G, measured in 2014 February, the rotation cycle is expected to be longer than 5 yr. Additional evidence for a very long rotation period, longer than 9 yr, follows from the consideration of the variability of the H α line profiles. However, since very long rotation periods are best determined from magnetic field variability, future monitoring of HD 54879 should include both, the follow-up of the changes of the H α line profile and of the measurements of the longitudinal magnetic field.

The analysis of the new FORS 2 polarimetric spectra indicates that the previous detection of a significant field increase and a change of the spectral appearance is due to improper spectra extraction and wavelength calibration, using observations obtained at an insufficient S/N ratio.

Among the previously detected magnetic O-type stars, five are classified as Of?p stars. The primary characteristic for the Of?p stars, according to the definiton by Walborn (1972), is a variable and comparable emission strength of the C iii blend (C iii λλ4647-4650-4652) with respect to the neighbouring variable emission N iii blend (N iii λλ4634-4640-4642). The observed C iii blend in these stars is strongly variable, exhibiting transitions from absorption line profiles to emission line profiles at certain rotation phases. The presence of variable emission in the C iii blend is indicative of circumstellar structure around the Of?p stars, related to their magnetospheres. However, the emission in the C iii blend disappears entirely in late O-type stars (Walborn et al. 2010) and is thus missing in HD 54879, meaning that a selective emission effect cannot be observed in HD 54879. Obviously, the Of?p classification is very narrow as it is limited to stars with spectral types in the range O4f?p–O8f?p, with temperatures between 34.5 and 41 kK. HD 54879 is significantly cooler with T_eff = 30.5 kK and this is the main reason why no association with the Of?p class has been done so far.

On the other hand, many properties of HD 54879 are similar to those of Of?p stars. The H α emission line in HD 54879 presented in Fig. 3 is highly variable. In analogue to the C iii blend, the presence of the H α emission in Of?p stars is related to their magnetospheres and its variability is expected to trace dense environments in Of?p stars. This line is frequently used to determine their rotation periods.

A remarkable resemblance of the ultraviolet (UV) spectra of HD 54879 and the Of?p star NGC 1624-2 with a dipole strength of ∼20 kG estimated by Wade et al. (2012) was recently discussed by David-Uraz et al. (2019). The authors report that despite of the later spectral type of HD 54879, its UV spectrum is surprisingly similar to the UV spectrum of NGC 1624-2 obtained at a rotational phase of nearly magnetic equator-on view. Both stars exhibit the lowest vsin i and macroturbulent velocity v_mac values known among the magnetic O-type stars (Shenar et al. 2017) and do not show nitrogen excess (Castro et al. 2015). Alike the spectral appearance of the Of?p star NGC 1624-2, the high-resolution HARPS spectra display weak emission lines belonging to various metal lines, indicating that the line formation in the atmosphere of HD 54879 can be similarly complex.

As we discussed in Section 4, the rotation period of HD 54879 is probably longer than 9 yr. Also Of?p stars are known as a class of slowly rotating magnetic massive stars with the longest rotation period of about 50–60 yr suggested for the Of?p star HD 108 (Nazé, Vreux & Rauw 2001). Furthermore, recent analyses of the XMM–Newton spectra of HD 54879 by Shenar et al. (2017) and Hubrig et al. (2019) indicate overluminosity by at least one order compared to other O-type stars with similar spectral types. Such an excess of X-ray luminosity is typical for all Of?p stars, for which X-ray spectra are usually well described by multitemperature thermal plasma models. All of these properties suggest that HD 54879 is an analogue to the Of?p stars and only misses the Of?p classification criteria because of its lower temperature.

In respect to the possible binary nature of HD 54879, it was suggested in recent years that three of five Of?p stars are members of binary systems. Long-term radial velocity changes indicating binarity were reported for the Of?p star HD 191612 (Howarth et al. 2007) who suggested P_orb = 1542 d. According to Wade et al. (2019), the Of?p star HD 148937 is likely a high-mass, double-lined spectroscopic binary. Also for the Of?p star HD 108 Nazé, Walborn & Martins (2008) reported that a very long-term binary cannot be excluded. Similarly, the compilation of radial velocity measurements over tens of years indicates that HD 54879 could be a member of a long-period binary system.

The knowledge of the frequency of membership of upper main-sequence stars with radiative envelopes in wide binary systems is very important, as it can be related to the origin of their magnetic fields. It was suggested that magnetic fields may be generated by strong binary interaction, i.e. in stellar mergers, during a mass transfer, or in the course of a common envelope evolutionary phase (Tout et al. 2008). The resulting strong differential rotation (Petrovic et al. 2005) is then considered as a key ingredient for the magnetic field generation. Requiring mergers to produce magnetic stars implies that there should be almost no magnetic star in a close binary. Indeed, magnetic components in close binaries are very rare: Only three close binaries with a magnetic Ap component, HD 98088 (P_orb = 5.9 d; Babcock 1958), HD 25267 (P_orb = 5 d; Borra & Landstreet 1980), and HD 161701 (P_orb = 12.5 d; Hubrig et al. 2014), are currently known, and only two late-B type binaries with magnetic components and orbital periods below 20 d were recently detected, HD 5550 (Alecian et al. 2016) and BD−19°5044 (Landstreet et al. 2017). The situation among early-B type stars is even more extreme, as only one early-B type short-period magnetic binary, HD 136504 (Shultz et al. 2015), is currently known. As wide binaries are widespread among Ap and late Bp stars (Mathys 2017), such wide binaries could have been born as hierarchical triple stars, where the inner binary merged.

Observations in star-forming regions indicate that almost all stars form in clusters (Lada & Lada 2003), and that the number of multiple systems within these clusters is remarkably high. Binary population synthesis simulations predict that the rate of main-sequence mergers increases with mass (e.g. de Mink et al. 2014; Schneider et al. 2015). de Mink et al. (2014) found a merger fraction of 8 per cent in a population of B-type stars and 12 per cent in O-type stars. Also, recent observations of Ap and late Bp stars support a scenario where mergers produce magnetic stars (Mathys 2017). Clearly, future monitoring of radial velocities of massive magnetic stars is important to be able to conclude on the role of binarity for the magnetic field generation.

ACKNOWLEDGEMENTS

Based on observations made with ESO Telescopes at the La Silla Paranal Observatory under programme IDs 191.D-0255, 0100.D-0110, and 0102.D-0285. We thank the referee G. Mathys for his constructive comments that helped to improve the paper.

REFERENCES