© The Authors 2025

1. Introduction

The classical B emission line stars, or Be stars, are typically understood as rapidly rotating B-type stars in which an additional mechanism acts to eject photospheric material, forming a gaseous, Keplerian, slowly outflowing, viscous decretion disk (VDD). Knowledge about the disk and its physics has significantly increased over recent decades (as summarized by Rivinius et al. 2013). This additional mechanism is largely suspected to be related to pulsation. The origin of the rapid rotation, on the other hand, has been suggested to lie either in the interior evolution of a single star, via angular momentum transport from the contracting core to the surface (see, e.g., Granada et al. 2013), or in binary evolution, where the rotation is the product of mass transfer between a more rapidly evolving component and a mass gainer that is observed as a Be star. The former case could be proven, for instance, by identifying main sequence (MS) companions to Be stars in orbits too close to accommodate a previous mass transfer (Bodensteiner et al. 2020b find no such systems among early-type Be stars, but see also Kervella et al. 2022, who suggest Achernar as one case, albeit of later subtype than the cutoff of Bodensteiner et al. 2020b), or through statistical means. If the origin were angular momentum transport by mass transfer, we would typically either see properties such as anomalous space velocities, for example if the former donor exploded and the system were disrupted (Berger & Gies 2001), or the remnant of the previous mass donor. Such a remnant, unless it were to produce strong X-rays through accretion onto a compact object, would nevertheless be difficult to detect, as it would contribute only a low percentage of the total flux (see Götberg et al. 2018; Jones et al. 2022; Klement et al. 2022b; Wang et al. 2023, for descriptions of spectral modeling and reports on the most recent search efforts).

However, there is one short period of time in which such a remnant could be easily seen next to the Be star, and possibly even outshine it, that is, when the mass transfer is fully complete, and the Be star is already a classical Be star in the above sense (i.e., when it is producing its own decretion disk through mass ejection), but the former donor has not yet contracted to its later equilibrium configuration of a hot subdwarf and could be dubbed a pre-subdwarf. In this case, the spectral appearance might be dominated by the photospheric spectrum of the former donor, and show some quite unique features (Schürmann et al. 2022; Irrgang et al. 2022). Such subdwarfs would later go on to become white dwarfs, but in the context of this work only the immediate, post-overflow subdwarf phase is of concern. Be+WD systems relevant to the binarity connected evolution and formation of Be stars could be formed via another channel (case BB mass transfer, as described by Gies et al. 2023) and are only discussed in Sect. 4.

The current discussion on Be star binary systems that may have recently completed mass transfer was triggered observationally, when Liu et al. (2019) initially suggested that LB-1, also known as ALS 8775, might be a very massive stellar black hole with a Be star companion. This was quickly refuted, and of the many alternative hypotheses, that of a classical Be star with a stripped and bloated post-Roche-lobe-overflow companion that has not yet contracted to its equilibrium radius and temperature has become the generally accepted one (Shenar et al. 2020).

In the wake of the discovery of LB-1, Rivinius et al. (2020) proposed that HR 6819, as a Be star with a quiescent stellar-mass black-hole companion, may be one of the alternative scenarios also applicable to LB-1. However, a little later Bodensteiner et al. (2020a) and El-Badry & Quataert (2021) suggested that HR 6819 could also be a pre-subdwarf companion. Using observations with the Very Large Telescope Interferometer (VLTI), this interpretation was proven correct by Frost et al. (2022), who were able to determine an initial astrometric orbit of two similarly bright objects from the two first observations of an ongoing study.

The field is evolving rapidly, and more objects are frequently proposed. For example, Ramachandran et al. (2023) and Villaseñor et al. (2023) suggested systems, but both are in the Magellanic Clouds, while the current study focuses on the local environment of bright stars, making it much easier for follow-up studies to look at the reported objects in detail.

Here, we identify and describe five other systems that bear the characteristics of a stripped donor and a spun-up companion star. In the following, Sect. 2 introduces the candidate selection criteria and the observations on which this work is based –both from archives and our own–, while Sect. 3 discusses the systems consisting of a Be star and a confirmed or candidate post-overflow companion. In addition to LB-1 and HR 6819, introduced above, we present several more systems, ranging from systems with an observationally confirmed nature, to candidates identified with moderate confidence. El-Badry et al. (2022) propose that HD 15124 is an immediate progenitor of these systems, but as shown in Sect. 4, many such systems may be known already, as the system parameters of HD 15124 do not seem to be particularly exceptional. Section 4 further considers statistics, the relation to the general paths of binary evolution, and the phases of stellar evolution and binary interaction in which the individual systems might be, albeit without attempting an exhaustive quantitative analysis, which is deferred to later works. The Appendices show spectroscopic data (Appendix A), interferometric data (Appendix B), and list additional Be stars with composite spectra and thus potential binaries that were found to be of interest, but do not quite fit into the main body (Appendix C).

2. Observational data

2.1. Candidate identification

Chochol & Mayer (2002) presented a compilation of six B-type binaries with invisible but supposedly massive companions according to their high mass function. Considering the properties of LB-1 and HR 6819, these are of particular interest, and indeed five of the companions appear to be good candidates for stripped and bloated stars. Additional potential systems were identified by visually inspecting archival data, as detailed below. Again following the leads provided by LB-1 and HR 6819, but also by examining other systems presented here, candidates were identified by looking for Be stars with drastically narrower absorption lines than would be expected from the emission-line shape (Rivinius et al. 2013). Further criteria were that they either show radial velocity (RV) variations in these absorption lines, but not in the emission, or a potentially abnormal strength pattern in the spectral lines of H, He, and CNO. The spectra of these candidates can be seen in Appendix A.

Considering the inhomogeneity of the sources from which the candidates were drawn, there is no statistically meaningful parent sample. Such a sample can nevertheless be constructed a posteriori, as shown in Sect. 4.

2.2. Interferometry

Several of the program stars were observed using optical or near-infrared (NIR) interferometry in order to detect the companions (see below for a more detailed description of the facilities used). In this work, we present new interferometric observations for eight of the program stars: VLTI data for LB-1, HR 2309, HR 3195, HD 44637, and V1371 Tau, and CHARA Array data for V742 Cas, V447 Sct, and V1362 Cyg. Close, low-contrast companions were successfully detected for four stars –V742 Cas, HR 2309, HD 44637, and V1371 Tau– and the astrometric orbit of V742 Cas was subsequently mapped with additional measurements.

The log of the interferometric data can be found in Table B.1. The calibrator stars were selected using the Searchcal software developed and maintained by JMMC¹ (Chelli et al. 2016). The calibrator angular diameters were adopted from the JMMC catalog of stellar diameters (Bourgés et al. 2014, 2017) and are listed in Table B.2.

For interferometric data with low spectral resolution (LB-1, V742 Cas, V447 Sct, and V1362 Cyg), the CANDID code was used to obtain (1) the relative positions and flux ratios of the companion in the case of successful detections, and (2) minimum magnitude differences for the unseen companions in the case of non-detections. For details about the CANDID code, we refer the reader to the dedicated work presenting the code (Gallenne et al. 2015) as well as more recent works that made use of it (e.g., Klement et al. 2022b). To facilitate easier manipulation of interferometric data with high spectral resolution (HR 2309, HR 3195, HD 44637, and V1371 Tau), the more versatile code for interferometric modeling, PMOIRED (Mérand 2022), was employed to achieve the same goals. Furthermore, high-spectral-resolution K-band data enable the Brγ line to be resolved, which is useful for studying the kinematics of the circumstellar environment and for identifying whether the emission-line star is the brighter or fainter component in a low-contrast binary. Figures in Appendix B show examples of the data analysis.

2.2.1. Very Large Telescope Interferometer

The Very Large Telescope Interferometer (VLTI) is a facility located on Cerro Paranal in Chile that can be fed by either the four 8.2m unit telescopes (UTs) or an array consisting of four movable 1.8m auxiliary telescopes (ATs). The angular resolution, which is defined as the angular size of an object where visibility goes through the first null, will depend on the projected length of the baseline configuration used: for the VLTI with UTs, the angular resolution has a maximum value of 4 mas, and for the VLTI with the ATs put in large configuration it is 2 mas at most. The instrument used here, GRAVITY, takes four telescope beams and combines them in the K-band, with a choice of spectral resolutions of R ∼ 22 500 and 4000 (GRAVITY Collaboration 2017). Observations were reduced with the GRAVITY pipeline workflow provided under ESO Reflex (Freudling et al. 2013). The VLTI was used with the UTs for LB-1 in low spectral resolution. The ATs were located in the astrometric baseline configurations for HR 2309, HD 44367, and HR 3195, while for V1371 Tau they were in the medium baseline configuration, all observed in high spectral resolution.

2.2.2. CHARA Array

The Center for High Angular Resolution Astronomy (CHARA) Array (ten Brummelaar et al. 2005; Schaefer et al. 2020) is an optical/NIR interferometer located on Mt. Wilson, California, USA, and consists of six 1m telescopes in a Y-shaped configuration. The maximum baseline of B_max = 330 m yields an angular resolution of λ/(2B_max)∼0.5 milliarcsec (mas) in the NIR H band and ∼0.7 mas in the K band. The six-telescope beam combiners MIRC-X (Anugu et al. 2020) and MYSTIC (Setterholm et al. 2023), operating simultaneously in the NIR H and K bands, respectively, are capable of detecting binary companions with snapshot observations (on-source integration time of 10–20 minutes) down to a contrast of ∼0.5%. The default spectral resolution of R ∼ 50 corresponds to an interferometric field of view (FoV) of ∼50 and ∼65 mas for the two instruments, respectively. The data for V742 Cas, V447 Sct, and V1362 Cyg taken with the default settings were reduced and calibrated using dedicated software² (pipeline version 1.3.5, Anugu et al. 2020) and will be made publicly available in the Optical Interferometry Database³ (Haubois et al. 2014) and the CHARA Data Archive⁴.

2.3. Spectroscopy

2.3.1. UVES

The cross-dispersed UV-Visual Echelle Spectrograph (UVES) is mounted at UT2-Kueyen of the VLT at Cerro Paranal (Dekker et al. 2000). It was used to obtain high-resolution spectra of LB-1, HR 2309, HR 3195, HD 44637, and V1371 Tau. The slit width was set to 1″, providing a resolving power of R ∼ 40 000. LB-1 was observed in the 437/760 cross-disperser setting, giving simultaneous spectral coverage in the ranges of about 380-495 nm and 570-950 nm. HR 2309, HR 3195, HD 44637, and V1371 Tau were observed in the 390/580 setting, providing coverage of the 330–450 nm and 480–680 nm spectral regions. The data were reduced with the ESO Reflex pipeline workflow for UVES (Freudling et al. 2013). In all cases, including LB-1 as the faintest target, and data from the two smaller telescopes below, the signal-to-noise ratio was well above 100 over the entire spectrum.

2.3.2. FEROS

The HR 6819 spectra were taken with the Fibre Extended Range Optical Spectrograph (FEROS, Kaufer et al. 1999) and were reduced with the standard FEROS pipeline⁵. The instrument provides a coverage of about 370–920 nm with a resolving power of R ∼ 48 000. These are the same spectra as already used by Rivinius et al. (2020) and introduced there in detail. In addition, this study uses two archival spectra of HR 4930⁶ that were reduced the same way.

2.3.3. ARCES

The Astrophysical Research Consortium Echelle Spectrograph (ARCES; Wang et al. 2003), mounted at the 3.5 m ARC-telescope of the Apache Point Observatory, is a high-resolution, cross-dispersed visible light spectrograph with a coverage of 320–1000 nm and a resolving power of R ∼ 31 500. ARCES observed 141 Tau, V742 Cas, V1362 Cyg, V2174 Cyg, V447 Sct, HR 2309, HR 8107, and HD 44637 for this study. The data were reduced using procedures from the Image Reduction and Analysis Facility (IRAF⁷) including 1D extraction, bias subtraction, removal of scattered light and cosmic rays, division by flat field exposures, wavelength calibration via ThAr lamp exposures, continuum normalization, and merger of the 107 orders.

2.4. Other

As mentioned above, candidates were also partly identified by visual inspection of archival data, in particular the spectra in the BeSS database (Neiner et al. 2011; Neiner 2018). This is a large collection of Be star spectra taken by amateur astronomers worldwide over the past two decades. Spectrographs range from long-slit low-resolution instruments over relatively high-resolution Hα spectrographs to fiber-coupled echelle instruments spanning most of the visual domain with a resolving power of typically R ∼ 8000–11 000. We used the spectra of the highest available quality for the identified candidates in this study; their technical details can be obtained from the BeSS database⁸. Upon request, additional spectra were taken by BeSS observers to improve or clarify period values. Further, a few single spectra were obtained from other public archives or published resources, which are referenced in the text where applicable.

All parallaxes mentioned in this work have been taken from Gaia DR3 (Gaia Collaboration 2023). However, because the confirmed targets are all astrometric binaries as well, these parallaxes cannot be translated into distances directly, and indeed all stars mentioned below show signs of acceleration anomalies in their astrometric data (Kervella et al. 2019; Brandt 2021). However, none of the targets have an entry in the binary tables of Gaia DR3 (Gaia Collaboration 2022). For some candidates, we used photometric data obtained by TESS (Transiting Exoplanet Survey Satellite, Ricker et al. 2015)⁹.

3. Individual systems

In this section, the individual systems are discussed, first in order of discovery (LB-1 and HR 6819), and then in order of confidence level, where the systems with interferometric confirmation come first (V742 Cas, HD 44637), followed by the systems with a well-determined spectroscopic orbit and sufficient data for an independent analysis, where a distant orbit could be excluded by interferometry (V447 Sct, V1362 Cyg), and then systems with only some of our own data (V2174 Cyg), and finally systems that will need further investigation before any conclusion can be drawn (HR 8107, HR 4930, V505 Mon, V658 Car) or rejected candidates (141 Tau, HR 2309, HR 3195, V1371 Tau). Hα profiles and the blue and red spectral regions for most of those objects are shown in the Appendix as Figs. A.1, A.2, and A.3.

The orbits were determined using orbfit-lib¹⁰. In all cases, a free orbital solution resulted in an eccentricity of zero within the uncertainties, and so the solution was fixed to a circular orbit.

3.1. LB-1 (LS V +22 25 and ALS 8775)

This is the first system that was suggested to consist of a Be star with a stripped and bloated low-mass (i.e., immediately post-overflow) companion; this latter dominates the visual absorption, while the Be star is mostly seen by virtue of its line emission, with only a marginal contribution due to the absorption (Shenar et al. 2020). Abdul-Masih et al. (2020) concluded that the stripped companion is almost non-rotating, with v sin i = 7.5 km s⁻¹ and a macroturbulence of ξ = 4 km s⁻¹. A more detailed chemical analysis by Irrgang et al. (2020) revises the rotation to v sin i = 8.7 ± 0.2 km s⁻¹ with macroturbulence ξ ≤ 0.1 km s⁻¹.

The spectrum of LB-1 is shown in Appendix A. It is relatively cool, in the late-B-type regime (about 13 kK judging from the Si II/III balance, and 12 720 K according to Irrgang et al. 2020), and shows an obvious CNO-process-affected abundance pattern, or as pointed out by Irrgang et al. (2020), He and N are strongly enriched, while C and O are under-abundant. These authors also report a systematic under-abundance of Mg, Al, Si, S, Ar, and Fe, although their analysis was undertaken with the assumption that the spectrum originates from a single star. The dilution of these lines due to the continuum contribution of the Be star might explain the systematic weakness of these lines, even for solar abundances. The composite nature also explains the shape of the helium lines, for which the authors note that the observed line wings are excessively strong while the line cores appear excessively weak for a normal stellar atmosphere. The spectrum of the Be star is most clearly seen the in He I lines, suggesting a much earlier B subtype.

The first of the two interferometric observations taken one year apart with VLTI/GRAVITY had to be discarded due to a bad calibrator, which displayed a strong binary-like signal. The second one does not show any indication of a third, distant companion, as the closure phases are consistent with zero. The bandwidth-smearing FoV of the measurements is ∼84 mas, and searching for a companion up to a separation of 120 mas with CANDID led to a null result. The mean of the minimum magnitude difference ΔK between the primary Be star and a possible companion at separations of < 120 mas is 4.69 ± 0.25 mag. This rules out the black hole scenario within a triple system, as has been suggested (and excluded; see below) for HR 6819. The angular resolution of the VLTI with UTs of the order of 2 mas, together with the distance to LB-1 above 1 kpc according to the Gaia DR3 parallax, do not enable a binary system with the given orbital period to be resolved.

3.2. HR 6819 (HD 167128 and QV Tel)

This is the second system for which a Be+pre-subdwarf nature was suggested, by Bodensteiner et al. (2020a), and then proven by Frost et al. (2022). The former qualitatively find T_eff = 16 ± 1 kK for the narrow-lined component and the indication of an abundance pattern that is in line with expectations for CNO-processed material, but find no significant helium overabundance. This suggests that the core has not been stripped very deep into the hydrogen-burning shell. However, it should be noted that Bodensteiner et al. explicitly only mention nitrogen and oxygen, and not carbon. The latter should be depleted, but seems quite strong in the spectra (see Fig. 1). It is also noteworthy that HR 6819 has a rather strong Ne I λ6402 line compared to Si II λ6437, yet also obvious C II lines, which may indicate a more progressed stage of nucleosynthesis than CNO burning; although this needs to be modeled in detail.

Fig. 1. Spectral regions of several candidates with high-quality spectra showing lines of interest with respect to the chemical nucleosynthesis. From bottom to top, spectra are sorted in approximate order of increasing effective temperature. The presence or absence of the C II λ4267 in the various spectra and the strength of nitrogen lines indicates non-solar chemistry. The RV-variable spectra have been shifted to line up their spectral features. With the exception of LB-1 (UVES) and HR 6819 (FEROS), all spectra were obtained with ARCES. See Figs. A.2 and A.3 for more comprehensive plots.

Bodensteiner et al. (2020a) give a low rotational velocity of v sin i = 15 km s⁻¹, albeit with a very high macroturbulence of ξ = 35 km s⁻¹, but also note that they exclude v sin i > 25 km s⁻¹ in any case, which we consider a safe upper limit. Further spectral properties for the pre-subdwarf are discussed by Rivinius et al. (2020); in particular their Figs. 2 and C.2 are of interest for comparison with similar figures of the candidate objects shown below. For the Be star, Bodensteiner et al. (2020a) estimate an effective temperature of 20 kK, again an earlier spectral subtype than for the pre-subdwarf.

3.3. V742 Cas (HD 698)

This system is listed by Chochol & Mayer (2002) as having an invisible companion, with reference to Sahade et al. (1992), although it has been known to be a single-lined binary since 1932 (Pearce). Sahade et al. (1992) derived the orbit from UV spectroscopy taken with the International Ultraviolet Explorer (IUE) satellite and discuss wind properties based on the typical wind lines expected in an OB-type star. Spectroscopic observations taken from the BeSS database confirm the orbit with a period of 55.9 d and a RV amplitude K of about 80 km s⁻¹ in a circular orbit. The system was then observed with ARCES and CHARA, which again confirmed the spectroscopic orbit, and revealed it to be a low-contrast binary, that is, the two components are of similar brightness.

A combined orbital solution was then obtained using BeSS, ARCES, and CHARA observations. For the RVs, the He I λ6678 line was used (see Appendix A). This is present in 12 amateur spectra of the Hα region, 7 echelle spectra taken by amateur observers (all those are available from the BeSS database), and 12 ARCES spectra (see Table A.3).

The ARCES spectra show relatively narrow but well-resolved deep absorption lines, with a typical full width at half maximum (FWHM) of ∼35 km s⁻¹ for lines of Fe II or Si II. This can be compared to Bodensteiner et al.’s analysis of HR 6819, which has a FWHM of about 60 km s⁻¹. A preliminary spectroscopic analysis, which will be published elsewhere in full once complete (Przybilla, priv. comm.), suggests a v sin i of 11 km s⁻¹ with a macroturbulence ξ of 19 km s⁻¹.

The spectrum does not show solar abundances, but is He-rich, with enhanced nitrogen and depleted carbon (Przybilla, private communication). As a quantitative analysis is deferred to future work, it is difficult to estimate the photospheric temperature from the habitually used indicators. It should be noted that lines of both Fe III and Si III are present, but they are much weaker than those of Fe II and Si II, and that, apart from the broader lines (see Fig. 2 and Appendix A), the spectrum is generally similar to that of LB-1 (which has ∼13 kK), including the chemical peculiarities. The Be-typical emission lines are stable in RV, although their general weakness other than in Hα makes it difficult to give an upper limit to any RV variability that would mirror that of the narrow-line component. Attempts to measure the Hα RV variations through spectral disentangling were unsuccessful, and no traces of the Be star absorption spectrum were found.

Fig. 2. ARCES spectra of V742 Cas folded with the orbital period for selected lines as labeled. The absorption components in the O I, SiII, and He I lines (seen as dark trails) originate from the photosphere of the stripped star. With the exception of the RV variable peak in Hα, the emission originates from the disk regions close to the Be star.

The interferometric results are summarized in Table B.3. From the four measurements with separations of > 0.5 mas, the fainter component contributes 31.1 ± 0.3% of the total flux in the H band on average and 35.5 ± 0.4% of the total flux in the K band. However, with the data at hand it is not possible to distinguish whether it is the Be star or the narrow-lined component that is the brightest, as the Be star cannot be identified with the low spectral resolution of MIRC-X or MYSTIC. The fact that the fainter star seems to be brighter in K than in H –that is, it is redder– would suggest that the Be star is the fainter component, because the disk contribution to the Be-star light reddens the spectral energy distribution (SED).

The combined solution, presented in Fig. 3, that is, taking both spectroscopy and astrometry into account, agrees well with the IUE-based solution by Sahade et al. (1992, see Table 1). Nevertheless, this solution translates to unphysical parameters when the Gaia DR3 distance of 708 pc is used, such as a negative mass for one component. However, this is not necessarily a reliable distance, because even though it has a renormalized unit weight error (RUWE) of only 1.46, it is listed by Kervella et al. (2019) and Brandt (2021) as a star with a proper motion anomaly. With a semi-major axis of 0.663 mas, this is the smallest visually confirmed orbit to date, with the previous record-holder being WR 133 (Richardson et al. 2021). Finally, as a binary, it should be removed from the interferometric calibrator catalog of Cruzalèbes et al. (2019).

Fig. 3. Circular orbit solution for V742 Cas (Table 1). Left: RV curve of the narrow-lined component (dashed line) with the RVs measured in the He I λ6678 line (black circles). The bottom panel shows the residuals in units of σ. Right: Relative astrometric orbit derived from interferometry. The orbital motion is clockwise. The black plus symbol corresponds to the location of the primary Be star, while the best-fit companion orbit is plotted as a black dashed line. Error ellipses correspond to the interferometric measurements with 5σ uncertainties (MIRC-X in blue and MYSTIC in red). The associated calculated positions are plotted as black crosses. Finally, the line of nodes is plotted as a gray solid line, the green cross shows the ascending node, and the smaller yellow cross shows the superior conjunction of the Be star that is used as T0 for the phased plots. The two MYSTIC points with larger error ellipses correspond to angular separations below the nominal resolution of CHARA in the K-band.

A more detailed analysis of the orbital parameters, and spectral disentangling of the components, the distance, the component masses, and the physical properties –including the abundances–, is deferred to a later study. A preliminary analysis along those lines carried out by Przybilla (private communication) confirms that, despite the light ratio of approximately unity, one component is of a normal mass for an intermediate B star and has largely solar composition, while the other has less than one-fifth of that mass and shows CNO-cycle-modified chemistry.

3.4. HD 44637

HD 44637 was identified in the BeSS database as an object with narrow and RV-variable spectral absorption lines, while the emission did not show the same behavior. This system was observed a few times with ARCES and UVES, and in a coordinated campaign through BeSS. The absorption line RV is variable, but the period proved to be too long to obtain a solid value before the end of the observability period in 2022/23; it is certainly longer than half a year.

Two GRAVITY/VLTI observations show an unambiguous detection of two unresolved sources separated by about 1.5 mas, with a contrast ratio of about 40% to 60%, where the Be star could be identified as the brighter source by virtue of the phase signature across the Brγ line. The two observations were taken about two months apart, and the position angle between the two components changed by about 90 degrees. This is in agreement with RV measurements obtained from BeSS spectra that show a slow redward acceleration by about 25 km s⁻¹ over the same time span. This also indicates a long period of at least half a year or more. As shown below in a comparison of all systems, this is unusually long but is not unique. Hα shows the strongest emission of the sample, at an emission height in units of the continuum of E/C = 8 in the combined spectrum, and so for the Be star alone it must be well above 10 (Fig. A.1), and Brγ still E/C = 3 (for the Be star alone, see Fig. B.1). Such a well-developed disk is another indication of a long period, as the radius of the tidal truncation of the disk must be accordingly large.

The stellar spectrum is most comparable to that of HR 6819, as, judging from the Si II/Si III ratio, it is only slightly cooler, and has a similar rotational velocity v sin i. It also shows nitrogen lines of enhanced strength and even less carbon and oxygen than HR 6819, but even though it is cooler, the helium lines are stronger.

Although a full determination of the orbital elements is ongoing due to the long period, and results must be deferred to a later study, this is not needed to count HD 44637 as a binary of the same nature as HR 6819, and with the same confidence, as the interferometric and spectroscopic variability is fully clear in that respect.

3.5. V447 Sct (HD 173219)

This system is also listed by Chochol & Mayer (2002) as a B-type star with an invisible companion. While it is sometimes classified as a B0 supergiant, this is probably not the case. Its parallax is 0.5 mas. For comparison, the actual B1 supergiant ζ¹ Sco has the same parallax value, but is 3.5 magnitudes brighter in V. Unless there is highly unusual extinction towards V447 Sct, its magnitude is more in line with a core-hydrogen-burning early B star, and even more so if it is a low-contrast binary, that is, with similarly bright components. This is also reflected by some earlier classifications as B1:V:npe (Morgan et al. 1955; Hiltner 1956). The spectrum was analyzed in more detail by Frémat et al. (2006), who confirm a more classical Be-like spectral type of B0.5 IV, and also give v sin i = 61 km s⁻¹ for the absorption lines, which nevertheless do not belong to the Be star in the system.

Hutchings & Redman (1973) identify the system as very peculiar, and –as with other systems mentioned here– as having a very massive, invisible secondary, which they speculate could be a collapsed object, like a black hole. The spectroscopic orbit obtained by these authors also puts it squarely in the same regime as the other candidates, with P = 58.41 d, K = 87.4 km s⁻¹, and a circular orbit (see Figs. 4 and 5). The full width at half maximum (FWHM) of the absorption lines is quite high, in agreement with the v sin i value of Frémat et al. (2006), but the profiles are Gaussian shaped, similar to those of HR 6819, and so, as well as rotation, macroturbulence might play a significant role in their broadening.

Fig. 4. Same as the left panel of Fig. 3 but for a circular orbit solution for V447 Sct (Table 2), and RVs measured in the Si III λ4553 line.

Fig. 5. ARCES spectra of V447 Sct folded with the orbital period for selected lines, as labeled. Similar to V742 Cas shown in Fig. 2, except for the single peaked emission in Si II, which is something also seen in HR 6819 (Rivinius et al. 2020).

The Hα emission is morphologically unique. It certainly does not have a P Cygni wind-type profile but neither is it akin to that of a classical Be star. It has a broad base that does not vary in RV with any obviously seen amplitude in the given spectra. In addition, there is a single peak that moves in RV with the binary period (see Fig. A.1). This is similar to what is seen in V742 Cas, but here the RV amplitude of the Hα emission peak is only half that of the photospheric absorption lines, indicating either some active accretion stream still present in the system, or a wind–wind interaction. However, in lines other than Hα with single-peaked emission, the amplitude is comparable to that of the absorption lines. These include He I λ5876, He I λ7061, He I λ4471 (where it mimics a double absorption line), and Si II λ6347 (Fig. 5).

The absorption spectrum is the earliest among the candidates, showing strong Si III lines and a He I λ4471/Mg II λ4481 balance indicative of an early-type Be star. The spectrum has very obvious strong nitrogen and helium lines, but nearly absent carbon and oxygen lines. This strongly suggests CNO-processed material.

V447 Sct was observed with CHARA, but this did not result in the detection of a companion. This is not overly unexpected, because V447 Sct is fainter and farther away than V742 Cas. Two observations with CHARA were taken 14 days apart close to phases ∼0.32 and ∼0.55 according to the ephemeris given in Table 2, at least partly corresponding to epochs close to RV maxima. No companion was revealed, with a minimum magnitude difference of ΔH_min = 3.70 ± 0.16 mag and ΔK_min = 3.35 ± 0.19 mag at the first epoch, and ΔH_min = 2.75 ± 0.22 mag and ΔK_min = 4.11 ± 0.17 mag at the second epoch for angular separations of between 0.5 and 25 mas.

3.6. V1362 Cyg (HD 190467)

This is another system listed by Chochol & Mayer (2002) as a binary with a high mass function and an invisible companion. The source given there (Hill et al. 1976) only refers to an unpublished orbit with ∼57 d, but states that it has been thoroughly verified to rule out other periods. There are 24 spectra in BeSS –taken over more than 10 years– suitable for RV measurement of He I λ6678, as well as a number of recent ARCES spectra; see Table A.2. From these, the orbit was determined independently, confirming a circular orbit with a period of 56.81 d and an RV amplitude of K = 85 kms for the absorption lines (Fig. 6). This translates to an orbital radius of about 0.5 au/sin i. Considering the parallax of less than 1 mas, the interferometric non-detection is unsurprising: the CHARA/MIRC-X measurement in the H band revealed no low-contrast companion and a mean ΔH_min of 3.19 ± 0.12 mag for angular separations of between 0.5 and 25 mas. However, it should be kept in mind that the observing date was close to phase zero according to Table 3, and so this is not a fully conclusive non-detection, depending on inclination.

Fig. 6. Same as Fig. 4, but for a circular orbit solution for V1362 Cyg (Table 3), and RVs measured in the He I λ6678 line.

The classification as an eclipsing binary present in certain catalogs (Popova & Kraicheva 1984; Morris 1985; Malkov et al. 2006; Avvakumova et al. 2013), with a period of 7 d, can be traced back to Percy (1970). In the one available TESS sector, a variability timescale of about 3.5 d is present, but with semi-regular characteristics, which are unlike any signature of binarity. From a modern perspective, judging by the original plots and the TESS data, the variations are similar to the typical Be-type photometric variability that arises in the inner region of the disk as material is ejected from the star (Labadie-Bartz et al. 2022). Therefore, V1362 Cyg should be removed from the respective catalogs. Likewise, being a binary, it should be deleted from the catalog of potential interferometric calibrators (Cruzalèbes et al. 2019).

Percy (1984) mentions a v sin i of 75 km s⁻¹, but also notes a “variable sharpness” of the lines. The typically measured FWHM in both the BeSS and ARCES spectra is about 55 km s⁻¹, which is slightly narrower than the FWHM of HR 6819 of 60 km s⁻¹. For Table 4, which summarizes the properties of the identified systems, we therefore assume that it has a similar v sin i, which is less than 25 km s⁻¹. The spectrum is not obviously non-solar in its abundances, since C II is relatively strong, while Ne I is present but is not clearly over-abundant. However, N II might have stronger lines than one would expect of a star with this temperature.

The Be-typical emission lines of V1362 Cyg appear stationary. However, there are two exceptions, as seen in Fig. 7. The first is a single, weak emission peak at a wavelength of 7513 Å, which traces the absorption line RV curve. This is similar to what is seen in HR 6819 (Rivinius et al. 2020, their Fig. C.2). As opposed to HR 6819, however, a single peak component is also apparent in the Hα line, again following the absorption RV.

Fig. 7. As Fig. 5, BeSS echelle and ARCES spectra of V1362 Cyg folded with the orbital period for selected lines, as labeled. The infrared region at 7500 Å is only covered by ARCES.

3.7. V2174 Cyg (HD 235679)

This is also one of the binaries noted by Chochol & Mayer (2002) as having an invisible massive companion. Spectra secured through BeSS and with ARCES corroborate the literature description as nitrogen rich. The identification of this star as a system with a post-mass-overflow component is strongly suggested already by the literature. It was first noted by Abt et al. (1970) as a binary, though the period provided by these authors of 111 d was later rejected (see below). Walborn (1971) then found it to be a nitrogen-rich B-type supergiant with strongly depleted carbon and oxygen, and suggested that it might be a new type of helium star given their finding of an abnormally large helium-to-hydrogen ratio. Bolton & Rogers (1978) confirmed the hydrogen deficiency and also noted strong emission in the hydrogen lines and weaker emission in the He I lines. These authors derived a period of 225.16 d, an RV amplitude of K = 63.6 kms, and a low eccentricity compatible with zero within e = 0.1 ± 0.1, implying a very high mass function of 5.9.

Finally, Bolton & Hurkens (2001) investigated the system in some detail and revised the period to 225.33 d. These authors could not find any trace of a potentially Roche-lobe-filling companion, and consequently considered the system to consist of a Be-type supergiant and an invisible companion. However, they also state that the Be star, which is supposed to be a supergiant from previous works, is subluminous considering its spectral type. Bolton & Hurkens (2001) largely confirm the earlier determined orbital parameters, with the exception of elements computed from hydrogen-line measurements, which they characterize as “very different from those derived from the other line groups”. The authors even say that the emission lines, which are strong with a height above the continuum of E/C ∼ 3 to 4, are approximately stationary. They consider a black hole as the companion, but reject it in favor of a model where the emission stems from a wind–wind collision. They also state that the spectroscopically invisible component would then either have to be a rapid rotator, or at least two magnitudes fainter, and even suggest a post-overflow companion. They do not, however, consider the possibility that it is this post-overflow companion that produces the absorption spectrum they observed.

Three echelle spectra were taken with ARCES and an additional one was made available in the BeSS database. These spectra show a largely identical spectral pattern to that seen in the spectra of V447 Sct, namely a strongly enhanced nitrogen abundance and helium enrichment, while carbon and oxygen are depleted. The Si II/Si III ionization balance points to a slightly lower temperature than that of V447 Sct, but the red Si II lines at 6347/71 Å are probably filled in by emission. The line width is similar to that of V447 Sct.

The findings for LB-1 and HR 6819, the description of the other objects above, and the available spectra leave little doubt that V2174 Cyg is perfectly described by a model consisting of a classical, recently spun-up Be star, which is responsible for the line emission and the broad hydrogen line wings, and a low-gravity, stripped and bloated core of a post-overflow object, which contributes the narrower photospheric absorption lines and exhibits strong signs of a processed chemistry. At V = 9 mag and a distance of above 1 kpc (Bolton & Hurkens 2001), the system is very much at the limit of the current interferometric capabilities for astrometric confirmation, although with knowledge of the orbit, it might just be possible to resolve the system at quadrature. For the orbital elements in Table 4, we adopt those derived from N II by Bolton & Hurkens, as they have the smallest uncertainties.

3.8. Other potential candidates

3.8.1. HR 8107 (HD 201836)

HR 8107 was discovered to be a Be star by Hirata et al. (1986), and so it was not part of the often meticulous observing campaigns targeting bright Be stars from the first half of the 20th century. This system shows a moderately broad-lined B5-type spectrum (v sin i ≈ 160 km s⁻¹) superimposed with a much narrower and perhaps slightly later B-type spectrum (v sin i ≈ 15 km s⁻¹). There are six ARCES spectra, three BeSS echelle spectra, and a number of BeSS Hα spectra, some of which also cover He I λ6678. The narrow-lined component is clearly RV-variable, but the amplitude is not more than K ∼ 20 km s⁻¹. The Be emission-line shape of Hα suggests a low inclination, but not full polarity. Hα also shows a tidally disturbed appearance in some spectra, where the infrared Ca II triplet can also be seen in emission; both characteristics suggest a Be star with a nearby binary companion (Shokry et al. 2018), and not a hierarchical system of more stars in which the Be star would be an isolated companion. In the above sample, for instance, Ca II emission is obviously present in five out of seven systems (LB-1, V742 Cas, HD 44637, V447 Sct, and V2174 Cyg), which is far more than the expected one-fifth from the general statistics regarding Be stars (Shokry et al. 2018).

In TESS data, the object appears as an eclipsing binary with a period of 12.14569 d (IJspeert et al. 2021). However, the spectroscopic data cannot be folded into a coherent curve with this period. The eclipses point to an eccentric binary of two stars of quite different radii and intrinsic brightness, but even the primary eclipse is just about 1.5% deep. Together with the U-shape of the dip, this suggests that neither the Be star, its disk, nor the narrow-lined component are likely to be involved in the eclipse. Further observations are required to decipher whether HR 8107 is a hierarchical system or –considering the size of the TESS point spread function of about 21 arcseconds– the eclipsing signal comes from an unrelated background object. Interferometry in particular might also help to explore how the photometric signal is related to the spectroscopic properties.

However, disregarding the eclipsing light curve, HR 8107 is a relatively strong potential candidate for a more pole-on case of a post-interaction system. In order to develop a more complete characterization of this system, the period and amplitude need to be established, and ideally also the astrometric orbit, which should be easily accessible at a Gaia parallax of 1.4 mas.

3.8.2. HR 4930 (HD 113120 and LS Mus)

The ESO archive (see Sect. 2.3.2) contains two FEROS spectra for HR 4930, which were obtained in direct succession at MJD = 53128.1. The spectra show a clear composite spectrum, with both objects being B-type stars, one broad- and one narrow-lined (see Fig. 8, uppermost spectrum). Judging by the He I λ4471/Mg II λ4481 ratio, as well as the presence of other species, the broad-lined component is a B star of around spectral type B2, while the narrow-lined star is of type B8 or later. It seems unlikely that the 45 kK template used by Wang et al. (2018, 2021) would produce a signal for any B8-type object, in particular considering the UV-weakness of such a spectral type versus the suggested 4% flux contribution.

Fig. 8. He I λ4471/Mg II λ4481 region of HR 4930, clearly showing the composite nature of the spectrum. In the FEROS spectrum (the upper one here), the secondary component of the He I line is also seen. The difference in strength of the secondary features in the spectra of Chauville et al. (2001) with respect to the FEROS spectrum is present in the original data and remains unexplained here.

Narrow components are also clearly present in Si II, C II, and Fe II lines. Their typical width is ∼25 km s⁻¹, which fits the upper limit by Wang et al. (2021). The depression of the narrow Mg II feature below the ambient spectrum is about 4%, while that of the Si II λ4128/32 doublet is about 3%. If these components were to have a depth of 100% in the spectrum of the narrow-lined star, these percentages would be lower limits on the fractional flux contribution of that star. Chauville et al. (2001) obtained two spectra of the He I λ4478/Mg II λ4481 region in HR 4930, which are of a lower quality than the FEROS data, but sufficient to identify the composite nature of Mg II λ4481 (see Fig. 8, lower two spectra). These observations were obtained at MJD = 47929.2 and 48641.3, respectively, and show a clear difference between the positions of the secondary Mg II λ4481 component, implying a radial-velocity amplitude in Mg II λ4481 of K > 40 km s⁻¹. Unfortunately, the period by Wang et al. (2023) is not sufficiently precise to phase these measurements –which were taken about two decades and longer ago– with their SB2 solution.

There is a component at a distance of 0.6″ noted by Wang et al. (2021) that is about three magnitudes fainter, although these authors exclude that this could be the source of the subdwarf spectrum based on the spatial resolution of HST/STIS.

This leaves one option for HR 4930: it could be a highly complex and hierarchical system, which would offer the only possibility to reconcile the UV observations by Wang et al. (2021) with the observations at optical wavelengths described above. In such a system, components Aa and Ab would be the Be star and a subdwarf, respectively, and Ba is the B8 or later-type star with an unseen companion Bb. In order to test this hypothesis, the orbit of the visual component B has to be determined –either spectroscopically or interferometrically– and compared to the observations of the UV component by Wang et al. (2023). If the orbit is different, it is a hierarchical quadruple system; if the orbit is the same, the companion cannot be a subdwarf of 45 kK, as such an object would have none of the singly ionized metal lines seen at optical wavelengths. Whether, in that case, the narrow-lined companion is a normal late-B MS star or a mass-transfer remnant will depend on the flux ratio, which needs to be measured interferometrically.

3.8.3. V505 Mon (HD 48914)

Two of the stars mentioned by Chochol & Mayer (2002) are not in the above list of candidates. FY Vel is described as a β Lyr system, with P = 33 d and K = 130 to 140 km s⁻¹ (Thackeray 1971). It probably does not fit here, but should be understood in the context of more similar systems, such as β Lyr and V453 Sco. The other one, V505 Mon, is probably more interesting in this context, and may be another candidate. However, from the literature alone this system is much more difficult to assess than V2174 Cyg, and there are no spectra of V505 Mon available to us. Mayer et al. (2001) identify a gaseous disk in the system and mention the clear similarity of this system to V742 Cas, V447 Sct, and V1362 Cyg: it has an orbital period of 53.8 d and an RV semi-amplitude of K = 92.1 km s⁻¹ for the absorption lines, which, at v sin i = 45 ± 5 km s⁻¹, are relatively narrow. In addition, Chochol & Mayer (2002) mention a low surface gravity and strong helium lines, also fitting the picture. The problem is that the system is also an eclipsing binary, which introduces phase-dependent phenomena in the absorption spectrum. Without a more detailed investigation of these variations, it is difficult to judge whether or not the disk in the system is already that of a classical Be star, that is, whether it is a slowly outflowing decretion disk, as in classical Be stars (Rivinius et al. 2013), is in the final stages of accretion, or indeed needs to be understood in an entirely different framework.

3.8.4. V658 Car (HD 92406)

V658 Car was reported as an eclipsing “post-Algol” system by Hauck (2018), with one component being an A0 shell star and the other a contracting white-dwarf precursor with a mass of 0.28 M_⊙. The system is reported to have P = 32.1854 d, K₁ = 10 km s⁻¹, and K₂ = 80 km s⁻¹, with the higher-mass component being cooler and less bright. The pre-subdwarf is estimated to have T_eff = 13 kK, which fits well into the parameter space spanned by the other candidates. However, most of the above parameters were derived from photometric data and 19 Hα-only spectra from BeSS, which nicely span two orbital cycles. These spectra seem to be similarly complicated to those of V505 Mon, and so a final judgment on the nature of the system should only be made with additional, more modern, spectroscopic extended-range and possibly also interferometric observations. Unfortunately, the Hα spectra in BeSS cannot be combined to form an RV curve amenable to meaningful quantitative analysis beyond the work by Hauck (2018). TESS observations obtained from the archive (see Sect. 2.4) also cover the eclipses and confirm the period.

3.9. Rejected candidates

3.9.1. V1371 Tau (HD 36665)

Bodensteiner et al. (2020b) present V1371 Tau as a Be star, for which no indication of a close companion has been reported. Inspection of the BeSS spectra, which prior to our campaign consisted of one echelle spectrum and four Hα spectra, revealed possible RV variability in Hα and He I λ6678, but the star does not exhibit signatures of the narrow absorption lines found in all other candidates. Moreover, spectroscopically, V1371 Tau resembles an unremarkable B1 III star with moderate (but not slow) rotation that sometimes exhibits weak Hα emission (Fig. 9, bottom). The star was consecutively observed by TESS in sectors 43–45. It is a very obvious eclipsing binary, as seen in the top of Fig. 9. Over the three months of observation, there are four almost identical, deep and narrow eclipses, signaling a binary of two photometrically similar stars in an eccentric orbit. The period is 33.619 d with an eccentricity of 0.26. This mismatch of showing only one set of spectral lines in a system with (at least) two objects of similar size and luminosity, as is indicated by the eclipses, make it a candidate worthy of investigation.

Fig. 9. Photometric and spectroscopic observations of V1371 Tau. Top: TESS photometry of V1371 Tau normalized and folded with the orbital period. Each phase was observed at least twice. Bottom: Hα profiles available from BeSS.

One VLTI/GRAVITY observation was taken just four days after a conjunction according to the photometric orbit. The system was resolved to be a low-contrast system with two components, at a distance of about 5 mas. Considering the Gaia DR3 parallax of 0.7 mas, there is no plausible orbit that would have taken the components to such a separation in just about 100 hours. V1371 Tau is therefore a hierarchical system of at least three stars, similar to ν Gem (Klement et al. 2021) or HR 2309 (below), and will be analyzed in greater detail in a future study.

3.9.2. HR 2309 (HD 44996)

By visual inspection of spectra in the BeSS database, this star was identified as RV-variable with narrow absorption lines yet stationary emission. It was subsequently observed with ARCES and UVES and in a coordinated campaign through BeSS, which confirmed the RV variability of the absorptions and the quasi-stationary Be-type emission. The absorption spectrum is very narrow, similar to that of LB-1, and is not fully resolved by ARCES. Indeed, Slettebak (1982) exclaims “Here is that rare thing: a sharp-lined Be star!” (but inexplicably then lists v sin i = 50 km s⁻¹ in his Table 2). HR 2309 was thus initially considered a high-confidence candidate for the present study.

GRAVITY/VLTI observations show an unambiguous detection of two unresolved sources at a separation of 34.4 mas, with a contrast ratio of 44%–56%. The fainter companion is the Be star, as is evident from the interferometric phase signature across the Brγ emission, and is hence dubbed component B. At a parallax of 2.5 mas, the rather rapid RV changes by up to 50 km s⁻¹ per day, but cannot be due to the interferometrically resolved orbit, meaning that the system must be hierarchical. With newly obtained data, the orbit of the narrow-lined component could be constrained to a period of about 5.6 d with K_Aa ∼ 60 km s⁻¹. Finally, in Mg II λ4481, a third component Ab could be detected spectroscopically, making the system SB3, with two inner narrow-lined components of a mass ratio of about 1:2 and the outer Be-star. Further parameters and an analysis of the system will be presented in a dedicated future work on Be stars in hierarchical systems. However, the striking spectral similarity to a stripped core should cause us to pursue this system with caution, noting that this is quite similar to the architecture originally proposed for HR 6819 (Rivinius et al. 2020).

3.9.3. 141 Tau (HR 2116 and HD 40724)

141 Tau was discovered as a Be star recently by Chojnowski et al. (2015) and is not listed in the BeSS catalog. There are nine ARCES spectra showing a clear composite nature of a B8e spectrum with v sin i ≈ 120 km s⁻¹ and an early A-type spectrum with v sin i ≈ 3 km s⁻¹. The analyses of these spectra suggest an orbital period of 11.44 d, with K = 36.5 km s⁻¹ and e = 0 for the narrow-lined component, but no detectable motion of the broad-lined component. This would, however, be a very close orbit to form a stable Be disk within that radius, and upon close inspection a third spectral component can be marginally detected in Mg II λ4481 at quadrature. This makes the system similar to HR 2309, as discussed above, and not of the nature searched for in the present work. Nevertheless, interferometric confirmation of a hierarchical nature would be desirable.

3.9.4. HR 3195 (HD 67888 and PQ Pup)

In BeSS, there are only one Hα and one echelle spectrum available for this object. Andersen & Nordstrom (1983) note it as potentially RV variable. Like many other systems studied in this latter work, Kervella et al. (2019) and Brandt (2021) consider it to be a star with a proper motion anomaly.

HR 3195 has unresolved narrow absorption lines in the BeSS echelle spectrum, but differs from the other systems by its Hα emission profile. The narrow single peak flanked by shoulder inflections (a so-called wine-bottle profile; Hummel & Dachs 1992; Hanuschik et al. 1996) suggests that HR 3195 could be a genuinely single classical Be star viewed pole-on. However, the Hα line profile at times shows an additional absorption, making it triple peaked (Fig. A.1), which we usually find to be a good indicator of binarity. On these grounds, HR 3195 did qualify as a candidate. One interferometric observation did not, however, show a low-contrast companion, but rather no binarity signature was found. As the arguments for a face-on orbit remain valid, this means that either the star is single, or if there is a companion, it must be very faint, and in either case the observed narrow-line absorption spectrum belongs to the Be star.

4. Discussion

4.1. Statistics of Be stars as binary products

The BeSS database holds spectra of hundreds of Be stars, many of them observed regularly. Most often these spectra include Hα, and sometimes, when echelle instruments were used, also Hβ. However, it is not a homogeneous database in terms of sky and magnitude coverage. For the purpose of statistics, a subsample of BeSS objects has to be defined that is reasonably complete and from which candidates could be identified with certainty, namely those stars with observations of sufficient quality and quantity that the criteria laid out in Sect. 2 could be evaluated without ambiguity. As most observers are located in the northern hemisphere, and have limited telescope power, this subsample was chosen as objects listed as classical Be stars in the BeSS database, north of −30°, and brighter than or equal to V = 8.0 mag. Stars north of that declination and brighter than the cut-off typically have more than 100 spectra in the database, while those south of that declination have fewer. It should be kept in mind, however, that for echelle data, BeSS counts each order as one spectrum. This gave 328 objects, the vast majority of which have a sufficient number of observations. Applying the same subdivision as Shokry et al. (2018, B0 to B2, B3 to B6, and B7 to A1, but see Sect. 4.1 of Shokry et al. 2018 for caveats and details), these can be grouped into 136 early-, 96 mid-, and 96 late-type Be stars.

On close inspection, not all of these qualify as classical Be stars (as shown e.g., in Appendix C). In turn there are also newly identified Be stars that are not in the BeSS catalog of classical Be stars, while other candidates, such as V2174 Cyg or V505 Mon, are also not in the BeSS database. A detailed discussion of the statistics of Be stars as found in the BeSS database will be published elsewhere, but for now it can be noted that the non-Be stars and the occasional star without sufficient data make up between 5 and 10% of the sample, and so we estimate that the sample population from which candidates are identified contains about 300 Be stars.

Based on the V-band magnitudes given in BeSS, this sample then includes three of the confirmed candidates: V447 Sct, V742 Cas, and HD 44637. The latter is probably fainter, but still has a listed magnitude of V = 8 in the BeSS database, and is thus included. The other confirmed candidates are either too faint (LB-1, V1362 Cyg) or too southerly (HR 6819). Of the remaining unconfirmed candidates, HR 8107 and V505 Mon would be included in the sample, but the latter is not listed in BeSS.

In a typical Be star, almost all light in the visual domain comes from one B-type star. In the binary systems discussed here, this is not true, which would introduce an over-detection bias for a magnitude-limited sample. Assuming the components to be of equal brightness, the found systems will be twice as bright, or detected out to a distance of 1.4 times that of the non-binaries (or probably somewhat less if considering extinction). As Be stars are strongly concentrated on the Galactic plane, that is, their distribution is largely two-dimensional, this translates to a correction factor of about two, which is applied in the following discussion.

This means that, in a magnitude-limited sample of Be stars, there are about 0.5–1% of binary systems in late or immediate post-mass-transfer phases. If, instead, the spectral subtype is also taken into account, this incidence more than doubles for early subtypes, and goes to zero for mid and late ones. As a side note, there is a similar number of hierarchical triples, where the Be star is the outer object but is not the dominant light source (such as ν Gem, Klement et al. 2021, or HR 2309 and 141 Tau mentioned above). This number can now be compared to other Be-type binaries in the BeSS sample: There are 16 or 17 known or strongly suspected Be+sdOB binaries, which constitute 5% of the sample (φ Per, κ Dra, o Pup, 7 Vul, 8 Lac A, 28 Cyg, 59 Cyg, 60 Cyg, HR 2142, HR 2249, HR 2855, HR 2921, HR 7807, QY Gem, and V1150 Tau; see Wang et al. 2018, 2021; Klement et al. 2022a,b and references therein, as well as HR 1772, Klement et al., in prep., and possibly V1294 Aql Harmanec et al. 2022). Further, there are six γ Cas-like X-ray stars in the BeSS list, which would be an incidence of 2% (γ Cas, π Aqr, HR 2284, HR 2370, V782 Cas, and V558 Lyr, see Nazé et al. 2022). The stars with such γ Cas-like X-ray properties form a class that, among other hypotheses, has been proposed to consist of Be+WD binaries (see, e.g., Gies et al. 2023, for the most recent work). Most of these other hypotheses also consider evolved companions, and a search for radial-velocity variations resulted in a considerably higher number of positive detections and candidates than for other Be stars (Nazé et al. 2022), meaning that a binary connection of the γ Cas phenomenon seems likely, even if a single-star magnetic hypothesis cannot be firmly ruled out yet.

The detection of Be+pre-subdwarf systems in the sample as defined above in this section is likely reasonably complete. The very sharp lines and high radial-velocity amplitudes are easily distinguished once these properties have been identified as hallmarks of such systems. Fully face-on orbits therefore pose a challenge, but as shown by the example of HR 8107, even those might be detected. Indeed, at a typical value of K > 50 km s⁻¹, a system would have to be very close to a face-on orientation to go unnoticed and considering the size of the sample the likelyhood can be neglected.

The γ Cas-like X-ray stars are probably also fully complete, as they are easily discovered through X-ray properties at the distances in question. While it is conceivable that there is a currently inactive –that is, diskless– star of this class within the sample, in practice this is probably not the case: all such stars are of early type, and tend to be more active on short timescales, meaning that an extended inactivity period preventing their detection is unlikely.

However, the list of Be+sdOB systems in the sample is certainly incomplete. They are difficult to detect even with interferometry, and many of the ∼300 stars are yet to be observed with this technique. Therefore, in summary, if one accepts the proposition of γ Cas-like X-ray stars as binary products by Gies et al. (2023), this puts the number of Be stars that are binary products and remain bound at least 8% for all Be stars.

It is particularly noteworthy that none of these systems seem to have a truly late-type Be component; they are all early, or at most mid-type Be stars up to about B3, although the case of 7 Vul as B4/5+sdO (Harmanec et al. 2020) and κ Dra as B6IIIe+sdB (Klement et al. 2022a) should be kept in mind. While a precise temperature determination is not yet available for most of the Be stars with pre-subdwarf companions, in all the composite spectra, the strongest lines from the Be component seem to be of hydrogen and helium, suggesting early subtypes as well. In contrast, not only are late-type Be stars well represented in the BeSS sample, but some of them are also well known to be binaries, such as the B8 star Pleione, for instance (Nemravová et al. 2010).

Interestingly, Pleione has a highly eccentric orbit of e = 0.7, whereas all of the above objects, where known, have circular or near-circular orbits: the two most eccentric Be+sdO binaries, 59 Cyg and 60 Cyg, have e = 0.14 and e = 0.2, respectively (Peters et al. 2013; Klement et al. 2022b). With that in mind, the bound binary products make up almost 20% of early-type Be stars, and nearly none of the mid- to late-type ones.

Highly eccentric Be binaries, on the other hand, such as δ Sco with e = 0.94 (Tycner et al. 2011), exist throughout the Be star spectral range. For binary interaction products, circularization is expected as part of the mass transfer, although this might not be completely justified (see Sepinsky et al. 2007, 2009, for a discussion). However, such high eccentricities indicate that not all early-type, and probably few, if any, late-type Be stars, are mass-transfer products, as these would have pre-interaction orbits. This has even been proven for the mid-type Be star binary α Eri, which has e = 0.73, where the MS companion spectrum is detected and the orbit is too close to fit a stable hierarchical system within that size that could have undergone mass transfer (Kervella et al. 2022).

4.2. Potential progenitor systems

El-Badry et al. (2022) report HD 15124 as a semidetached system of early/mid-B+F type stars with about 5 and 1 M_⊙, respectively, and a period of 5.47 d. They model the system as on the path toward an LB-1/HR6819-like configuration, although the total mass of the system seems to be slightly too low to fit this scenario. Such systems are indeed relatively common, and are often initially classified as normal Be stars, simply on account of the presence and morphology of their Balmer emission. Examples, sorted by increasing orbital period, include CX Dra (Simon 1996, B2.5Ve+F5III, P = 6.69603 d, K_Be = 34 km s⁻¹, K_F = 156 km s⁻¹) FF Cam (Garrel et al. 2013, Be+K, P = 7.788 d, K_Be ≪ K_K= 85 km s⁻¹), 14 Lac (Hill et al. 1997; Linnell et al. 2006, B3e+F9IV, P = 10.0854 d, K_Be = 25 km s⁻¹, K_F = 158 km s⁻¹), HL Lib (Dempsey et al. 1990, B9IVe+FIII, P = 24.615 d, K_Be ≪ K_F = 84.1 km s⁻¹), and HD 81357 (Koubský et al. 2019, B8e+K?, P = 33.77458 d, K_Be ∼ 10 km s⁻¹, K_K = 82 km s⁻¹), as well as the class of double-periodic variables (DPVs; see Mennickent et al. 2016). Not all of those systems are massive enough to form an early type Be star, especially the last two in this list. However, nothing in the statistics excludes that some late-type Be stars may be spun up through binary interaction –it is merely the balance of the various spin-up processes that seems to change drastically from early- to late-type Be stars.

However, which of those will eventually become classical Be stars can only be determined with detailed modeling to predict the final orbit and masses of the post-interaction system (see Götberg et al. 2018, for an example). Also, although most of the above systems would make the cut considered in the previous subsection as being northern and bright, this is probably not a statistically very meaningful comparison, as the extended donor companion, and not the B star component, dominates the light from the system. In the case of very large rates of mass loss and gain, such a system might also be identified as of β Lyr type, where the companion is hidden in an obscuring torus, and so the basic components of the binary are an evolved, hot donor star and an invisible (but not black hole) companion, as suggested by Chochol & Mayer (2002) for FY Vel (also see Sect. 3.8.3).

4.3. Accretion flows

Four systems, LB-1, V742 Cas, V447 Sct, and V1362 Cyg, show RV-variable Hα emission components that trace the RV curve of the donor in phase, but not necessarily with the same amplitude. V447 Sct also shows an example of this in some He I lines. This is not typical for a Be star disk in a binary. While tidal distortions may lead to density spirals in the disk and some phase-locked variability, this is typically observed as a slight variation of the violet-to-red height ratio of the emission peaks, and not as a clearly traceable, strong, additional single emission peak going back and forth (Panoglou et al. 2018). The fact that such a peak is also not expected in a fully detached binary suggests that either there is residual mass transfer ongoing in those systems, or that the situation has reversed and the former donor is now re-accreting material from the Be disk.

4.4. Rotation

4.4.1. Pre-subdwarf rotation

A striking similarity between all non-rejected candidates is the slow rotation of the mass donor, that is, the pre-subdwarf component. It should be noted, however, that this is not necessarily representative, as the presence of absorption lines with anomalously low width was used as one criterion for the identification of possible candidate systems. Schürmann et al. (2022) make the case that the slow rotation of the former donor can only be explained by efficient magnetic angular momentum transport in its stellar interior during the post-MS expansion. If this were a common property of early-type stars, the ubiquity of slow rotation would be expected.

The slow rotation can also be misleading, however, as is seen in HR 2309 above. Together with LB-1, this object has the smallest v sin i, but was shown to likely be a normal MS star in a hierarchical system.

4.4.2. Be star rotation

Other than the donor, the mass gainer, that is, the future Be star, would spin up during mass transfer. Be stars are widely agreed to be rapid rotators (Rivinius et al. 2013; Zorec et al. 2016). The interior stellar evolution during core-hydrogen burning, as the core contracts, transports angular momentum outwards, resulting in an evolutionary spin up in terms of the critical fraction w = v_eq./v_crit.. The viscous decretion disk is the most efficient way for the star to shed that angular momentum, and so a disk will form as soon as mechanisms for mass and angular momentum transfer into the disk become available to the star (Ghoreyshi et al. 2021; Rímulo et al. 2018). Not all immediate post-interaction systems would be of the type discussed in this work, as mergers are also a possibility.

This also means that, once a B star has entered the Be regime, it is likely to remain a Be star until the end of its MS life. This suggests a possible solution for the difference between early and late-type Be stars discussed above: the longer MS lifetime of late-type Be stars may make the single-star evolutionary spin-up route more viable for them, as they may have had enough time to spin up by single-star evolution. This does not mean there are no binary products in late-type Be stars at all, as demonstrated by cases such as KOI-81 and Regulus, which are both rapidly rotating B8 stars with a white dwarf companion; however, these are not observed as Be stars (Matson et al. 2015; Gies et al. 2020).

The earlier stars, due to their shorter life time on the MS, and possibly also due to their higher binarity fraction, would more typically have required the binary evolution path to spin up sufficiently.

4.5. Surface abundances

All donors seem to have surface hydrogen, as can be seen when the hydrogen lines are phased with the orbital period. LB-1, V742 Cas, V2174 Cyg, and V447 Sct show obvious signs of CNO-processing, namely depleted carbon and enhanced nitrogen. V447 Sct has the most obvious nitrogen enhancement and carbon/oxygen depletion, but no further sign of any enhancement. For HR 6819, HD 44637, and V1362 Cyg, the picture is less clear. These systems do show a CNO-typical pattern in helium, nitrogen, and oxygen, but not the expected strong carbon depletion.

5. Conclusions

In summary, there are now six systems that have been confirmed as Be binaries with out-of-equilibrium post-overflow companions. Five are spectroscopic binaries in circular orbits (one orbit remaining undetermined until further observations in the coming season), where narrow-lined B-type features dominate the absorption spectrum with high RV, whereas the Be-star emission lines remain largely stationary, pointing to either a binary with an extreme mass ratio –and considering the other observational properties, an out-of-equilibrium post-overflow companion– or a hierarchical system, where the Be star would be an outer component. It is noteworthy that the Be stars in the vast majority of systems that are found as binary products are of early B subtype, suggesting differences in the formation channels of early- and late-type Be stars.

We positively identified three of those six systems –HR 6819, V742 Cyg, and HD 44637– using interferometry as close, low-contrast binaries, without a wider companion. For another three, LB-1, V447 Sct, and V1362 Cyg, only the absence of a wider companion was shown interferometrically, but this is sufficient to include them as confirmed cases.

We also identified three of the candidates –141 Tau, HR 2309, and V1371 Tau– using interferometry and/or spectroscopy as hierarchical systems and these will be analyzed elsewhere. Of the remaining candidates, V2174 Cyg has all the signs of a post-overflow binary, but without interferometric confirmation, it may still be of a similar kind to HR 2309. For HR 8107, HR 4930, V505 Mon, and V686 Car, further studies are needed, but these remain promising candidates. HR 3195 could be rejected as being a single star, or at least as not having any detectable companion.

This group is sufficient to allow some preliminary insights into the statistical properties, first as a subset of Be stars, of which they form about 0.5–1% in a magnitude-limited sample, but also objects in their own right: all systems have zero eccentricity, all show some sign of processed material in their spectra, and all of the former donors (except for LB-1) have rather slow –but not quite zero– rotation, while the Be stars have naturally rapid rotation.

Further studies are required to explore the chemical abundances of the donor stars in more detail, and to better constrain orbital parameters. In particular, we will attempt to obtain a combined astrometric and RV (SB2) solution for the three interferometrically resolved systems through spectral disentanglement.

Data availability

Tables A.1 to A.4 and B.1 to B.3 are available at the CDS via anonymous ftp to cdsarc.cds.unistra.fr (130.79.128.5) or via https://cdsarc.cds.unistra.fr/viz-bin/cat/J/A+A/694/A172. Appendices are available at Zenodo.

Acknowledgments

Part of this work is based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere under ESO programmes 073.D-0274(A), (HR6819 FEROS) 073.C-0337(A), (LS MUS FEROS) 2104.D-5024(A), (LB-1 UVES) 0106.D-0994(A), (LB-1 GRAVITY) 0110.D-0400(A), (HD44996 & HD44673 & HD36665 UVES) 0110.D-4381(A), (HD44996 GRAVITY) 0110.D-4381(B), (HD44637 & HD36665 GRAVITY) 2110.D-5034(A), and 2110.D-5034(B) (HR3195 UVES & GRAVITY) available from the ESO archive at https://archive.eso.org. Partly based on observations obtained with the Apache Point Observatory 3.5-meter telescope, which is owned and operated by the Astrophysical Research Consortium. This work is based upon observations obtained with the Georgia State University Center for High Angular Resolution Astronomy Array at Mount Wilson Observatory. The CHARA Array is supported by the National Science Foundation under Grant No. AST-1636624 and AST-2034336. Institutional support has been provided from the GSU College of Arts and Sciences and the GSU Office of the Vice President for Research and Economic Development. This work has made use of the BeSS database, operated at LESIA, Observatoire de Meudon, France. BeSS observers: P. Berardi, E. Bertrand, C. Buil, S. Charbonnel, A. de Bruin, V. Desnoux, J. Foster, O. Garde, J. Guarro Fló, B. Heathcote, F. Houpert, R. Leadbeater, T. Lemoult, T. Lester, M. Pujol, O. Thizy. This research has made use of NASA’s Astrophysics Data System Bibliographic Services. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France. This research has made use of the Jean-Marie Mariotti Center Aspro and SearchCal services. This paper includes data collected with the TESS mission, obtained from the MAST data archive at the Space Telescope Science Institute (STScI). Funding for the TESS mission is provided by the NASA Explorer Program. STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5–26555. This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement. SK acknowledges funding for MIRC-X received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Starting Grant No. 639889 and Consolidated Grant No. 101003096). JDM acknowledges funding for the development of MIRC-X (NASA-XRP NNX16AD43G, NSF-AST 1909165) and MYSTIC (NSF-ATI 1506540, NSF-AST 1909165).

References

All Tables

All Figures

Fig. 1. Spectral regions of several candidates with high-quality spectra showing lines of interest with respect to the chemical nucleosynthesis. From bottom to top, spectra are sorted in approximate order of increasing effective temperature. The presence or absence of the C II λ4267 in the various spectra and the strength of nitrogen lines indicates non-solar chemistry. The RV-variable spectra have been shifted to line up their spectral features. With the exception of LB-1 (UVES) and HR 6819 (FEROS), all spectra were obtained with ARCES. See Figs. A.2 and A.3 for more comprehensive plots.

In the text

	Fig. 2. ARCES spectra of V742 Cas folded with the orbital period for selected lines as labeled. The absorption components in the O I, SiII, and He I lines (seen as dark trails) originate from the photosphere of the stripped star. With the exception of the RV variable peak in Hα, the emission originates from the disk regions close to the Be star.
In the text

Fig. 3. Circular orbit solution for V742 Cas (Table 1). Left: RV curve of the narrow-lined component (dashed line) with the RVs measured in the He I λ6678 line (black circles). The bottom panel shows the residuals in units of σ. Right: Relative astrometric orbit derived from interferometry. The orbital motion is clockwise. The black plus symbol corresponds to the location of the primary Be star, while the best-fit companion orbit is plotted as a black dashed line. Error ellipses correspond to the interferometric measurements with 5σ uncertainties (MIRC-X in blue and MYSTIC in red). The associated calculated positions are plotted as black crosses. Finally, the line of nodes is plotted as a gray solid line, the green cross shows the ascending node, and the smaller yellow cross shows the superior conjunction of the Be star that is used as T0 for the phased plots. The two MYSTIC points with larger error ellipses correspond to angular separations below the nominal resolution of CHARA in the K-band.

In the text

	Fig. 4. Same as the left panel of Fig. 3 but for a circular orbit solution for V447 Sct (Table 2), and RVs measured in the Si III λ4553 line.
In the text

	Fig. 5. ARCES spectra of V447 Sct folded with the orbital period for selected lines, as labeled. Similar to V742 Cas shown in Fig. 2, except for the single peaked emission in Si II, which is something also seen in HR 6819 (Rivinius et al. 2020).
In the text

	Fig. 6. Same as Fig. 4, but for a circular orbit solution for V1362 Cyg (Table 3), and RVs measured in the He I λ6678 line.
In the text

	Fig. 7. As Fig. 5, BeSS echelle and ARCES spectra of V1362 Cyg folded with the orbital period for selected lines, as labeled. The infrared region at 7500 Å is only covered by ARCES.
In the text

	Fig. 8. He I λ4471/Mg II λ4481 region of HR 4930, clearly showing the composite nature of the spectrum. In the FEROS spectrum (the upper one here), the secondary component of the He I line is also seen. The difference in strength of the secondary features in the spectra of Chauville et al. (2001) with respect to the FEROS spectrum is present in the original data and remains unexplained here.
In the text

	Fig. 9. Photometric and spectroscopic observations of V1371 Tau. Top: TESS photometry of V1371 Tau normalized and folded with the orbital period. Each phase was observed at least twice. Bottom: Hα profiles available from BeSS.
In the text