Spectroscopic follow-up of candidate runaway stars

ABSTRACT

Runaway stars are characterized by higher space velocities than typical field stars. They are presumed to have been ejected from their birth places by one or more energetic mechanisms, including supernova explosions. Accurate radial velocities are essential for investigating their origin, by tracing back their Galactic orbits to look for close encounters in space and in time with neutron stars and young associations. While most studies of runaways have focused on OB stars, later-type stars have also been considered on occasion. Here, we report the results of a long-term high-resolution spectroscopic monitoring program with the goal of providing accurate radial velocities for 188 runaway candidates of spectral type A and later, proposed by Tetzlaff, Neuhäuser, and Hohle. We obtained multiple measurements over a period of about 13 yr to guard against the possibility that some may be members of binary or multiple systems, adding archival observations going back another 25 yr in some cases. We report new spectroscopic orbital solutions for more than three dozen systems. Many more are also found to be binaries based on available astrometric information. A small-scale study carried out here to trace back the paths of our targets together with those of four well-studied, optically visible neutron stars among the so-called Magnificent Seven, resulted in no credible encounters.

1 INTRODUCTION

The peculiar velocities of stars represent their motion with respect to their surroundings or to their birth cluster. While the dispersion of the peculiar 3D velocities of most stars is only a few |${\rm km\, s}^{-1}$|⁠, some show much higher velocities, presumably because they have been ejected from their birthplaces. They are referred to as ‘runaway’ stars. They occupy a Maxwellian tail in the distribution of peculiar velocities, which is distinguishable from the larger and relatively narrow peak populated by normal field stars (see e.g. Tetzlaff et al. 2011; their fig. 1). The dividing line, as proposed in that study and others, is usually taken to be around 30 |${\rm km\, s}^{-1}$|⁠. Numerical simulations (e.g. Renzo et al. 2019) show that, depending on initial conditions, stars can also be ejected with lower velocities, such as from a binary with a wide orbit. Such objects have been called ‘walkaways’.

Several mechanisms have been suggested for a star to attain higher than normal velocities: (i) dynamical ejection from the birth cluster or association within the first few Myr, through interactions among three or more stars in the core of a dense cluster (Poveda, Ruiz & Allen 1967); (ii) binary supernova ejection, in which the former companion of a massive star exploding as a supernova can become unbound, if the ejected mass and/or the kick of the newborn neutron star is sufficiently high (Blaauw 1961). Ejection velocities from this and the previous mechanism can be as high as 100 |${\rm km\, s}^{-1}$|⁠; (iii) black hole ejection, in which stars can be accelerated to even higher (possibly escape) velocities by close encounters with the supermassive black hole at the centre of our Galaxy (Hills 1988). These ejected objects are referred to as hypervelocity stars (Brown 2015).

Aside from the interesting question about their origin, stars ejected from their birthplaces are also an important component to include in the calculation of present-day mass functions, when comparing them to initial mass functions (IMFs).

Runaway stars are usually investigated by tracing back their 3D motion through the Galactic potential, to find instances in which, e.g. an OB star and a neutron star may have been at the same location in space at the same time. This could be evidence for a supernova occurring in a binary, ejecting both the pulsar and the runaway star (see e.g. Neuhäuser, Gießler & Hambaryan 2020). In practice, the validity of traceback calculations is typically limited to a few tens of Myr due to various observational uncertainties. Thus, the ejection is required to have happened within that time frame. For this reason, OB stars are most favourable for this type of analysis, because their short main-sequence lifetimes imply that they must be young (less than |$\sim$|30 Myr for spectral types earlier than B3). Nevertheless, later-type stars can also be ejected as runaway stars by any of the scenarios described above. Unfortunately, however, it may be more difficult to precisely determine their ages, by methods such as the use of the Li |$\lambda$|6708 absorption line as a youth diagnostic, or by other means. For example, lithium in stars is destroyed at temperatures above about |$2.5 \times 10^6$| K as it gets mixed into the interior (e.g. Bodenheimer 1965; Neuhäuser 1997), but the rate of Li depletion for mid-F to late-M dwarfs is influenced by the extent of the convection zone. It has a complex and not yet fully understood dependence on temperature, mass, and metallicity, and may also depend on rotation and stellar activity (e.g. Soderblom et al. 1993). On the other hand, age determinations using stellar evolution models are limited in their sensitivity, because observables such as luminosity, temperature, and surface gravity change very slowly with time upon arrival of the star on the main sequence. Other techniques such as asteroseismology are not widely applicable for most field stars, due to a lack of suitable observations. Late-type runaway stars have therefore received less attention than early-type stars, but are no less important. They are the subject of this paper.

An extensive catalogue of candidate runaway stars of all spectral types within 3 kpc of the Sun was published some years ago by Tetzlaff et al. (2011). Those objects were presumed to be young based on the application of several criteria, but they were considered merely as candidates because their radial velocities (RVs) were largely unknown at the time. Their proper motions and parallaxes relied on the Hipparcos catalogue (van Leeuwen 2007). Particularly with its most recent data release (DR3), the Gaia mission (Gaia Collaboration 2023) has now provided greatly improved parallaxes and proper motions for large numbers of stars, as well as average radial velocities in many cases. However, the cadence of the spectroscopic observations is such that spectroscopic binaries or higher order multiples among the runaway candidates may be missed, or their orbits may not be solvable with the Gaia data alone. This can lead to mean velocities that are biased. While it may be true that most runaway stars are expected to be single on account of the violence of the ejection mechanism, close binaries are possible as well, and in fact appear to be quite common among walkaway stars (e.g. Bhat, Irrgang & Heber 2022). Therefore, repeated ground-based RV measurements remain important for the study of runaway stars, in order to characterize them fully and accurately.

With that in mind, soon after the Tetzlaff et al. (2011) list appeared, we embarked on a long-term project to monitor the RVs of a subset of about 190 of them with spectral types A and later. The main goal was to provide precise and accurate RVs, accounting for binarity for objects found to be in multiple systems. Here, we report on the full observational results of that survey, including the discovery of many binaries among our targets, of which Gaia has detected only a small fraction. We also carry out a pilot study to explore the possible origin of these stars, in the context of the supernova ejection scenario described earlier.

Our paper begins with an explanation of how the list of targets we followed up was assembled (Section 2), and continues in Section 3 with a description of our spectroscopic material and the derivation of radial velocities for single- and double-lined objects. We then comment on how we identified binaries among our targets (Section 4), based both on the RVs and on astrometric information from various sources. This section also presents the spectroscopic orbits we have determined. The distribution of peculiar velocities for our sample is reported in Section 5. Then, as a preview of a fuller study left for a future paper, Section 6 presents a small-scale analysis in which we trace back the 3D motion of our targets together with a handful of well-studied neutron stars, to explore a possible connection consistent with the supernova ejection scenario. We end with final remarks in Section 7. An Appendix is included in which we collect the orbital determinations from Gaia for comparison with ours, and provide details of interest for selected targets.

2 THE SAMPLE

The target list for our program, provided by Tetzlaff (2013, private communication), was selected mostly from the catalogue of candidate runaway stars of Tetzlaff et al. (2011), subject to the condition that the stars be located north of about declination |$-35^\circ$| in order to be observable with the telescope facilities at our disposal. They were chosen for spectroscopic follow-up because, at the time of that study, they lacked a radial-velocity measurement, and only the tangential velocities could be computed based on the Hipparcos proper motions and parallaxes. A total of 159 objects were taken from those authors’ Table 4, which lists candidates in which the runaway probability was estimated to be higher than 50 per cent in at least one of the peculiar tangential velocity components. An additional 15 stars come from table 1 of Tetzlaff et al. (2011), which includes objects with poorly determined kinematics that they deemed to be young, a condition required for runaway status. The remaining 14 stars were added based on information gathered after the 2011 publication.

The complete list of 188 targets, ranging from main-sequence stars to supergiants and with spectral types A through M, is given in Table 1, with information collected from the Gaia DR3 catalogue. In addition to the parallaxes, proper motions, and G-band magnitudes, we include the zero-point corrections to the parallaxes (⁠|$\Delta \pi$|⁠) advocated by Lindegren et al. (2021), as well as the Renormalised Unit Weight Error (RUWE), which is an indicator of the quality of the astrometric fit, and possibly an indicator of binarity (see below). Spectral types were taken from the catalogue of Tetzlaff et al. (2011), or from SIMBAD.

3 SPECTROSCOPIC OBSERVATIONS

Observations of our targets were conducted at the Center for Astrophysics (CfA) using several telescope/instrument combinations. Most of the spectra, beginning in September of 2009, were collected with the Tillinghast Reflector Echelle Spectrograph (TRES; Fűrész 2008; Szentgyorgyi A. & Fűrész 2007) on the 1.5-m Tillinghast reflector at the Fred L. Whipple Observatory, on Mount Hopkins (AZ, USA). This is a bench-mounted, fibre-fed instrument delivering a resolving power of |$R \approx 44\, 000$| in 51 orders, over a wavelength range of 3800–9100 Å. The signal-to-noise ratios of the 1898 spectra we obtained depend strongly on the brightness and sky conditions, and range from about 10 to 400 per resolution element of 6.8 |${\rm km\, s}^{-1}$|⁠. Reductions were performed with a dedicated pipeline (see Buchhave et al. 2012). The velocity zero-point was monitored with observations of IAU standards each run, and appropriate corrections were applied to remove any drifts with time.

Earlier archival observations of a few stars were gathered with two nearly identical spectrographs, attached to the Tillinghast reflector and to the 1.5-m Wyeth reflector at the Oak Ridge Observatory (now closed), in the town of Harvard (MA, USA). These instruments (Digital Speedometers; Latham 1992) were equipped with intensified photon-counting Reticon detectors that limited the recorded output to a single echelle order about 45 Å wide, centred on the Mg i b triplet near 5187 Å. They delivered a resolving power of |$R \approx 35\, 000$|⁠. Signal-to-noise ratios for these observations ranged between 10 and 60 per resolution element of 8.5 |${\rm km\, s}^{-1}$|⁠. Reductions were carried out with a dedicated pipeline (Latham 1985, 1992), and the velocity zero-point was monitored with observations of the twilight sky at dusk and dawn. As with TRES, run-to-run corrections were applied to maintain a stable velocity system. A total of 181 spectra were obtained with these instruments.

The native velocity system of the Digital Speedometers is slightly offset from the IAU system by 0.14 |${\rm km\, s}^{-1}$| (Stefanik, Latham & Torres 1999), as determined from observations of minor planets in the Solar system. In order to remove this shift, we added a correction of +0.14 |${\rm km\, s}^{-1}$| to all our raw velocities from these instruments. Similarly, observations of asteroids were used to place the measurements from TRES also on the IAU system.

Radial velocities for single-lined objects were determined by cross-correlation, using templates from a large pre-computed library of synthetic spectra based on model atmospheres by R. L. Kurucz (see Nordström et al. 1994; Latham et al. 2002), coupled with a line list tuned to better match real stars. The templates are restricted to a region near the Mg i b triplet, which captures most of the velocity information. The template parameters were chosen as follows. For simplicity, we adopted the working assumption of solar composition for all our targets, as this has little effect on the velocities. Surface gravity (⁠|$\log g$|⁠) is a subtle effect that is challenging to determine accurately from our spectra, in part because it is strongly correlated with the effective temperature (⁠|$T_{\rm eff}$|⁠). As |$\log g$| also has little impact on the velocities, we set it by locating each target in the colour-magnitude diagram, based on their |$G_{\rm BP}-G_{\rm RP}$| colour index from Gaia and their absolute G-band magnitude |$M_G$|⁠. We used stellar evolution models from the PARSEC v1.2S series (Chen et al. 2014) to infer a rough surface gravity, and then rounded off these |$\log g$| values to the nearest step in our grid of templates (step size = 0.5 dex). Effective temperatures and suitable rotational broadenings¹ for the templates (⁠|$v \sin i$|⁠) were chosen following the procedure of Torres, Neuhäuser & Guenther (2002), in which grids of correlations were run over wide ranges in those parameters, to identify the combination giving the highest correlation value averaged over all exposures.

In a few cases the model-inferred |$\log g$| values were beyond the range of our template library, or the temperatures reached the upper or lower limits available. For those targets, we manually adjusted either |$\log g$| or |$T_{\rm eff}$| to bring them within range, and reran the procedure of Torres et al. (2002) to maximize the average correlation, thereby minimizing any possible bias in the velocities.

For objects whose spectra are double-lined, RVs were measured using TODCOR (Zucker & Mazeh 1994), a 2D cross-correlation algorithm that uses two templates matched to the components. To set the parameters of the synthetic spectra for these cases, we applied a similar methodology as explained above, successively optimizing the primary and secondary templates by iterations. We also measured the flux ratio between the components at the mean wavelength of our observations.

The individual RVs with their formal (internal) uncertainties for the single-lined objects are listed in Table 2. Those for the double-lined targets are given in Table 3.

4 BINARIES IN THE SAMPLE

A summary of the velocity information for each target based on the measurements obtained here is presented in Table 4, and includes the time span and number of observations, as well as the weighted mean velocity and corresponding uncertainty. These mean velocities are used below for computing the Galactic orbits of the stars, in combination with the proper motions and parallaxes listed earlier. A large number of objects in our sample are binary or multiple systems of one type or another. In these cases, the mean velocities may not be representative of the true line-of-sight motion. We discuss these situations below.

More than three dozen of our targets are found to be obvious spectroscopic binaries, and have sufficient RV measurements and phase coverage for orbits to be derived. Eight of them are double-lined. These cases are flagged with an ‘SB1’ or ‘SB2’ note in Table 4, for single- and double-lined systems, respectively. We show the orbits for the single-lined systems in Fig. 1, along with the observations. The corresponding orbital elements are listed in Table 5. For each one, we computed also the mass function |$f(M)$|⁠, the coefficient of the minimum secondary mass²  |$M_2 \sin i$|⁠, and the projected semimajor axis of the primary orbit, |$a_1 \sin i$|⁠. Also included is the root-mean-square (rms) residual |$\sigma$| from the orbital solution, the number of observations, and a multiplicative scale factor F, which was applied to the internal velocity errors presented in Table 2 in order to achieve reduced |$\chi ^2$| values near unity.

Fig. 2 displays the orbits for the double-lined binaries in the sample. The corresponding orbital elements are collected in Table 6. Derived properties including the minimum masses, mass ratios (⁠|$q \equiv M_2/M_1$|⁠), and projected semimajor axes are presented separately in Table 7, along with the rms residuals for the primary and secondary, the number of observations for each, and the scale factors for the internal errors. The spectroscopic flux ratios we derived using TODCOR are included there as well.

For binaries with orbits, the mean velocities listed in Table 4 are the centre-of-mass velocities (⁠|$\gamma$|⁠) with their associated uncertainty, taken from Tables 5 or 6. These are the proper values to use for calculating Galactic orbits. One of our targets (HIP017878) is an overcontact binary of the W UMa class, for which our velocities are not reliable. In that case, the |$\gamma$| velocity listed was taken from the work of Rucinski et al. (2008).

A few other objects are also obvious binaries that show trends in their RVs, but their periods are longer than the timespan of the observations. The clearest examples are shown graphically in Fig. 3, and are flagged as ‘VAR’ in Table 4. Many others have been identified as binaries or possible binaries by visual inspection of our RVs, or by comparing our RVs with the median RVs listed in the Gaia DR3 catalogue.³ These cases are discussed individually in Appendix A, and are also flagged in the notes to Table 4 as ‘VAR’ or ‘VAR?’, depending on our degree of confidence in their RV variability. It goes without saying that the mean velocities listed in Table 4 for the ‘VAR’ and ‘VAR?’ objects are not necessarily representative of their centre-of-mass velocities. Furthermore, in some of them the variability may be due to pulsation rather than binarity.

For 21 of our targets, the Gaia DR3 catalogue reports ‘non-single star solutions’, representing the detection of binary motion, either on the plane of the sky (from astrometry only) or perpendicular to it (RVs), or both. In all but three of these cases, our own RVs also reveal the binary nature of the object. The three that we missed are HIP009008, HIP025386, and HIP033263. Full single-lined spectroscopic orbits are presented by Gaia for the last two of these (flagged as ‘SB1[GAIA]’ in Table 4), while the first was identified by Gaia as a binary from the curvature of its motion on the plane of the sky. We indicate this astrometric detection in the table with ‘AST[GAIA]’. For the last two cases, as well as for HIP062810, the table gives the |$\gamma$| velocity as determined by Gaia, instead of the mean RV from our own measurements. We also adopted the Gaia  |$\gamma$| velocity for HIP005680, which we detected as variable from our own observations, but for which we have insufficient data to determine an orbit. All Gaia non-single star solutions for the objects in our sample are gathered in a table in Appendix A, with an explanation of the different classes.

In many other cases, the Gaia DR3 catalogue or other literature sources, such as the Washington Double Star catalogue (WDS; Worley & Douglass 1997; Mason et al. 2001), identify nearby visual companions that are considered to be physically associated with the target. We have flagged these cases in Table 4 as ‘VIS’, but only if they have not already been identified as RV variables. They are mentioned as well in the Appendix. For these kinds of visual binaries, the orbital periods are presumed to be long, so the mean velocities in the table are not expected to be biased as much as the ‘VAR’ or ‘VAR?’ systems might be.

Other long-period binaries can also be identified from the difference |$\Delta \mu$| between their Hipparcos and Gaia proper motions (see e.g. Kervella, Arenou & Thévenin 2022), which are separated by an interval of 24.75 yr between the mean catalogue epochs. Any such differences are suggestive of orbital motion. To this end, we have consulted the catalogue of astrometric accelerations of Brandt (2021), and identified about a dozen of our targets with |$\Delta \mu$| values larger than 3|$\sigma$| in either coordinate. All except for two of these cases are already identified as binaries by one of the other methods above. The two that are not, HIP037104 and HIP095138, are flagged as ‘ACC’ in Table 4. All of these instances are mentioned in the individual notes in the Appendix. Their mean RVs may be expected to be little affected by their binarity, as in the visual binaries mentioned above.

Finally, other long-period binaries are revealed by the quality of the astrometric solutions from Gaia DR3, as quantified by the RUWE value reported in the catalogue. This metric is expected to be near 1.0 for sources with well-behaved solutions, while values larger than about 1.4 are often indicative of unmodelled binary motion, or some other problem (see Lindegren 2018; Belokurov et al. 2020). Even sources with RUWE in the range 1.0–1.4 may also be binaries in some cases (e.g. Stassun & Torres 2021). Of all the objects in our sample, 41 have |${\rm RUWE} > 1.4$|⁠, but almost all of these are already recognized as binaries in one or more of the ways explained above. The six targets that are not, are flagged in Table 4 with the code ‘RUWE’, and are mentioned in the Appendix.

All in all, there are 39 new spectroscopic orbital solutions derived from our measurements in this work (31 SB1s and 8 SB2s), 4 from Gaia, and one that was previously published (the W UMa system HIP017878), for a total of 44.

5 PECULIAR VELOCITIES AND RUNAWAY STATUS

With the mean RVs in Table 4, along with the parallaxes and proper motions from the Gaia DR3 catalogue (Table 1), we proceeded to compute the peculiar velocities of our targets relative to the Local Standard of Rest (LSR). They are listed in the last column of Table 1, with uncertainties that account for contributions from all observational errors, as well as from the solar motion itself. The components of the solar motion adopted here are |$(UVW)_{\odot } = (+11.10, +12.24, +7.25)~{\rm km\, s}^{-1}$| (Schönrich et al. 2010), with formal uncertainties of (0.75, 0.47, 0.37) |${\rm km\, s}^{-1}$|⁠, as reported by those authors.

Fig. 4 shows the histogram of the LSR velocities for our sample, with the conventional 30 |${\rm km\, s}^{-1}$| threshold for runaway stars indicated with a dashed line. We find that the majority of our stars (80 per cent) have peculiar velocities higher than this, indicating that the original selection of runaway candidates by Tetzlaff et al. (2011), carried out in the absence of RV information at the time, was quite effective. The remaining 20 per cent of our targets have lower velocities, and are normal stars or potential walkaway stars.

Six of our stars have LSR velocities in excess of 100 |${\rm km\, s}^{-1}$|⁠, of which four are included in the figure. The other two are HIP084038 (157 |${\rm km\, s}^{-1}$|⁠) and HIP004106 (314 |${\rm km\, s}^{-1}$|⁠). Interestingly, our RV monitoring has revealed that both of these objects happen to be single-lined spectroscopic binaries, with relatively long orbital periods of 349 and 383 days. They are also both among the more distant stars in our sample, at about 670 and 880 pc, respectively.

6 TRACEBACK OF RUNAWAY STARS: A PILOT STUDY

The results from the preceding sections bring a meaningful improvement in the reliability of the RVs for our runaway targets, and provide the missing velocities for the ones lacking the corresponding entries in the Gaia DR3 catalogue. The impact of binarity is fully accounted for through the determination of spectroscopic orbital solutions, where possible, or it is at least alleviated by identifying wider binaries detectable by astrometric means, which will presumably have less of an effect on the mean velocities. We now turn to the use of this information and the improved space velocities it enables.

Tracing back the 3D motion of a runaway star in the Galactic potential can determine whether, at some time in the past, it was sufficiently close in space to the location of the agent that may have caused its ejection at that time, such as a young association (dynamical ejection mechanism) or a neutron star (via a supernova explosion). As core-collapse supernovae are expected to occur within OB associations, in addition to tracing back a runaway star and a neutron star as a pair, one must generally also consider simultaneously tracing back triplets consisting of the runaway, a neutron star, and an association. Any encounters placing the supernova ejection outside of the association would not be credible.

The 3D velocities of our runaway stars are now well known from the astrometric information in the Gaia catalogue and our extensive RV monitoring. For the associations, their 3D motions are in general sufficiently well established based on either Hipparcos or Gaia data (see e.g. Tetzlaff et al. 2010; Neuhäuser et al. 2020). Neutron star motions, on the other hand, are more problematic: their distances are often poorly determined, and except for a few cases with indirect determinations, their RVs are essentially unknown, and must be drawn from statistical distributions inferred from their 2D motions under certain assumptions (e.g. Hobbs et al. 2005).

Here, we have performed Monte Carlo simulations to account for all observational uncertainties involved in the traceback, particularly the poorly established distances and the unknown RVs of the neutron stars (see also Hoogerwerf, de Bruijne & de Zeeuw 2000, 2001). The latter two factors are the dominant contributors to the uncertainty. Integrating the Galactic orbits to compare our targets against all known neutron stars and all young associations represents a computationally expensive effort of a scale that is beyond the scope of this paper. Instead, in this section we exemplify this procedure by performing a more limited exercise, considering four of the nearest, well known, optically visible neutron stars, among the group known as the Magnificent Seven (Treves et al. 2001; Zampieri et al. 2001). Their probable site of origin has been studied by Tetzlaff et al. (2010, 2011, 2012) based on Hipparcos information for runaway stars and associations, and their relevant properties are collected in Table 8.

The traceback calculations were performed using the algorithm and software developed for this purpose by Neuhäuser et al. (2020), which adopts a three-component model for the Galactic potential. We ran |$3\times 10^6$| Monte Carlo simulations for each pair consisting of a neutron star and a runaway from our sample. For pairs coming closer than 1 pc of each other, the results were analysed in more detail (see Hoogerwerf, de Bruijne & de Zeeuw 2001).

A careful review of the results, considering the four Magnificent Seven neutron stars in Table 8 along with young (⁠|$< 50$| Myr) nearby OB associations within 300 pc, revealed no credible close encounters with any of our runaway stars, with one possible exception. The analysis for HIP076768 and the pulsar RX J0720.4−3125 appeared initially promising, suggesting they could have been located at the same place at the same time within the |$\beta$| Pic/Capricorn group. This would work for either of the two published parallaxes for RX J0720.4−3125 (⁠|$3.6 \pm 1.6$| mas or |$2.8 \pm 0.9$| mas; see Table 8 for references). The encounter, as indicated in about 8000 out of |$3 \times 10^6$| Monte Carlo runs, would have taken place ca. 0.35 Myr ago within a few parsecs of the centre of the |$\beta$| Pic group. This flight time is shorter than the neutron star spin-down upper age limit. The runaway star HIP076768 is a known member of the young Upper Scorpius association (e.g. Miret-Roig et al. 2022)), but from its |$XYZ$| position, it could also be a member of the much larger |$\beta$| Pic/Capricorn group. While its kinematics are not quite typical for |$\beta$| Pic/Capricorn, that is not unexpected for an ejected runaway star.

However, an assessment of the credibility (or probability) of this encounter indicated that it is not high. Specifically, tracing back the neutron star and this young group with their actual properties, but the runaway star HIP076768 with a random direction of motion, we found encounters with the same neutron star within the same group in a similar number of simulations as above (⁠|$\sim$|10 000 out of |$3 \times 10^6$| runs). Furthermore, HIP076768 is a known visual binary, as mentioned in the Appendix, and wide binaries such as this are less likely to be ejected as runaway stars and to remain bound. Therefore, we consider the likelihood of a true encounter between HIP076768 and RX J0720.4−3125 within the |$\beta$| Pic/Capricorn group to be low.

While the null result from this limited test may be disappointing, a more comprehensive analysis is planned in which our list of targets will be paired against several hundred neutron stars with sufficiently precise input data. The outcome will be reported in a follow-up publication.

7 CONCLUDING REMARKS

Runaway stars have received a good deal of attention in the last decade or two, and the question of which of the formation mechanisms dominates the creation of these objects has been a subject of much debate (see e.g. Hoogerwerf et al. 2000; Fujii & Portegies Zwart 2011; McEvoy et al. 2017; Evans, Renzo & Rossi 2020; Bhat et al. 2022). Runaway stars are of interest for a number of reasons. For example, if tracing back the Galactic paths of a runaway star and a neutron star reveals a close encounter in space and in time, this could be evidence that the supernova may have exploded then and there. In that case, the flight time of the neutron star (i.e. its kinematic age), along with the age of the parent association (on the assumption of contemporaneous star formation), can provide an estimate of the progenitor’s mass, and other properties, with the aid of stellar evolution models. Runaway stars have also been applied to investigate the source of the radioactive |$^{60}$|Fe deposited in the Earth’s crust and ocean sediments by recent, nearby, core-collapse supernovae (see e.g. Fry, Fields & Ellis 2015; Hyde & Pecaut 2018). Tantalizing evidence of the location and time of at least one of perhaps several such events was reported by Neuhäuser et al. (2020). It involved the O-type runaway star |$\zeta$| Oph and the radio pulsar PSR 1706−16, split apart and ejected at high velocities some 1.78 Myr ago, from distance of 107 pc.

Much of the work on runaways has focused on OB stars. There are good reasons for this: they are brighter, a significant fraction of them show high space velocities, their youth makes them more favourable candidates, and they appear more frequently as former binary companions of the even more massive supernova progenitors. Many lists of potential OB runaway stars have been followed up over the years (e.g. Hoogerwerf et al. 2001; Mdzinarishvili 2004; Dinçel et al. 2015; Maíz Apellániz et al. 2018; Carretero-Castrillo, Ribó & Paredes 2023; Guo et al. 2024). Later-type examples have been considered as well, including the works of Tetzlaff et al. (2011), Lux et al. (2021), Teklehaimanot & Gebrehiwot (2024), and others. The present project draws from the candidate list released in the first of these later-type studies.

Our more than decade-long spectroscopic monitoring program, augmented with earlier archival observations, was designed to determine accurate mean RVs for nearly 190 runaway (or walkaway) candidates of spectral types A and later, accounting for their binary nature where needed. For the most part, these objects had no RV measurements at the time the Tetzlaff et al. (2011) list was released. As a result of this work, fully characterized orbits are now available for 44 spectroscopic binaries among our targets (39 of them new), both single- and double-lined, with orbital periods ranging between 0.3 and 5300 d. Only a fraction of them were reported in the Gaia DR3 list of non-single-star solutions. Many other targets have been identified here as being binary or multiple systems, based on astrometric information from the WDS, Gaia, or other sources. In total, about 52 per cent of the stars show either spectroscopic and/or astrometric evidence of multiplicity. In another 10 per cent of the cases, there are hints of velocity variability that need to be confirmed. For the remaining 38 per cent, no evidence of binarity is apparent.

The combination of the highly precise Gaia DR3 parallaxes and proper motions for these objects, along with the new RV information we have obtained, allows for much more accurate space velocities to be calculated for this sample than previously possible. This is an essential ingredient for investigating the time and place of their origin in the Galaxy. Most of our targets (80 per cent) have velocities relative to the LSR that are above 30 |${\rm km\, s}^{-1}$|⁠, the classical threshold for considering an object a runaway star. The most extreme is HIP004106, moving at |$V_{\rm LSR} = 314 \pm 6$| |${\rm km\, s}^{-1}$|⁠.

As a prelude to a larger investigation, here we have carried out a limited modelling exercise in which we traced back the orbits of our targets along with those of four well-known neutron stars from the Magnificent Seven group. We found no credible cases in which a star from our sample and a neutron star were sufficiently near each other in the recent past.

ACKNOWLEDGEMENTS

A large number of observers and colleagues at the CfA have contributed to gathering the spectroscopic observations for this project over the years. We thank them all. We are also grateful to Robert Davis and Jessica Mink for maintaining the Digital Speedometer and TRES databases at the CfA. We thank Frank Gießler (U Jena), whose software was used in this paper for tracing back neutron stars, runaway stars, and associations. RN and SAH would also like to thank Baha Dinçel, Surodeep Sheth, and Luca Cortese for support in the traceback calculations. We additionally acknowledge helpful comments from the anonymous referee. VVH and RN would like to thank Deutsche Forschungsgemeinschaft (DFG) for financial support through grant numbers NE 515/61-1 and 65-1. VVH acknowledges support from the Yerevan State University in the framework of an internal grant. This research has made extensive use of the SIMBAD and VizieR databases, operated at the CDS, Strasbourg, France, and of NASA’s Astrophysics Data System Abstract Service. We also acknowledge the use of the Washington Double Star Catalogue, maintained at the U.S. Naval Observatory. The work has used 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.

DATA AVAILABILITY

The data underlying this article are available in the article and in its online supplementary material.

Footnotes

REFERENCES

APPENDIX A

We include below notes for many of our targets, pertaining to their binarity or other details of interest. In 21 cases, the Gaia DR3 catalogue has reported non-single star models of various kinds. They are defined as follows:

• ‘Orbital’ = Orbital model for an astrometric binary;

• ‘SB1’ = Single-lined spectroscopic binary model;

• ‘AstroSpectroSB1’ = Combined astrometric + single-lined spectroscopic orbital model;

• ‘Acceleration7’ = Acceleration model with 7 parameters (astrometry only);

• ‘Acceleration9’ = Acceleration model with 9 parameters (astrometry only);

• ‘SecondDegreeTrendSB1’ = Second degree polynomial fit to the RV trend.

The parameters of these Gaia solutions for our targets are collected in Table A1, for the benefit of the reader. In another dozen cases, the catalogue of astrometric accelerations of Brandt (2021) indicates they are long-period binaries.

HIP001008: SB1 with |$P = 700.8$| d. Gaia DR3 ‘AstroSpectroSB1’ solution.

HIP001602: SB1 with |$P = 1084$| d. Gaia DR3 ‘Orbital’ solution.

HIP001733: Two of our velocities are considerably lower than the rest, but we have not been able to determine a period. Possible binary.

HIP002838: Gaia DR3 lists a common proper motion companion with the same parallax, at a separation of 40|${_{.}^{\prime\prime}}$|9 and 2.3 mag fainter in G. The Gaia RV for this companion is the same as that of the primary, confirming the association.

HIP004106: Tentative SB1 with |$P = 384$| d. This is a long-period photometric variable (CO Cet), classified by Hipparcos as semi-regular, and claimed to have a possible photometric period of 73.5 d. The RV variations have a low amplitude, and seem less regular after mid-2018. They may be caused by pulsations rather than orbital motion.

HIP005680: The RVs clearly indicate this is a binary, possibly with a rather eccentric orbit. Gaia DR3 reports an ‘AstroSpectroSB1’ solution with a very long but uncertain period, and a high eccentricity. We adopt the |$\gamma$| velocity from Gaia as the best representation of its mean velocity. There is a much wider companion (⁠|$\rho = 33{_{.}^{\prime\prime}} 6$|⁠, |$\Delta G = 4.2$| mag) listed by Gaia that shares the same parallax and proper motion. It has a median Gaia RV of |$-4.04 \pm 21.89~{\rm km\, s}^{-1}$|⁠, where the large formal error is a reflection of the scatter, and indicates that it too may be variable. This would make it a quadruple system. The target displays astrometric acceleration (Brandt 2021).

HIP009008: There is only a hint of variability in our RVs. Gaia DR3 ‘Acceleration7’ solution.

HIP009355: SB1 with |$P =1002$| d. Gaia DR3 ‘Orbital’ solution. Long-period photometric variable (DE Psc). There is a claim of a (noisy) detection of a binary companion in a lunar occultation observation on Christmas day, 1982, by Beskin et al. (1987) (projected separation |$8.3 \pm 0.7$| mas, vector angle 34|$^\circ$|⁠, |$\Delta R = 1.3 \pm 0.1$| mag). However, it was not seen in a subsequent lunar occultation event by Dyachenko et al. (2018), who only reported a measure of the angular diameter of the star.

HIP009470: SB1 with |$P = 555$| d. No companions were detected in a lunar occultation observation by Eitter & Beavers (1979).

HIP011242: Although our three RV measurements over a two month period show no change, this appears to be a binary. A lunar occultation observation by Evans et al. (1985) detected a companion with a projected separation of |$23.0 \pm 1.0$| mas, vector angle 191|${_{.}^{\circ}}$|6, and |$\Delta m = 1.99 \pm 0.19$| mag in a red filter. It was unresolved in a speckle observation by Mason (1996).

HIP011339: Clearly a binary, based on the large difference between the median Gaia RV and ours. Our RVs alone show possible signs of variability, with a hint of asymmetry in the CCFs due perhaps to blended lines from a companion. The target displays astrometric acceleration. Gaia classifies it as a photometric variable with short-period oscillations similar to the |$\delta$| Sct stars.

HIP011663: The median Gaia RV is slightly lower than our three measurements obtained about two years earlier, suggesting this may be a binary.

HIP012297: This is a binary with a 232 d period, for which we present an SB2 orbit. The detection of the faint M-dwarf secondary at a flux ratio |$\ell _2/\ell _1$| of just 0.5 per cent is somewhat tentative.

HIP013284: The median Gaia RV is marginally lower than ours. Possible binary.

HIP015373: Candidate massive member of the Psc-Eri stream (Curtis et al. 2019). Probable binary based on a median Gaia RV about 2 |${\rm km\, s}^{-1}$| higher than ours.

HIP016608: The observations indicate this is a binary. One of our RVs is 10 |${\rm km\, s}^{-1}$| lower than the rest, while the median Gaia RV is 10 |${\rm km\, s}^{-1}$| higher. Gaia classifies it as a photometric variable with short-period oscillations similar to the |$\delta$| Sct stars.

HIP016615: The velocity is variable, indicating it is a binary.

HIP017635: The median Gaia RV is 2 |${\rm km\, s}^{-1}$| higher than ours, which were obtained two years earlier. The target displays astrometric acceleration, revealing its binary nature.

HIP017878: This is a W UMa overcontact eclipsing binary (V1128 Tau) with a period of 0.305371 d (e.g. Çalışkan et al. 2014). Our RVs are not reliable because of the rapid rotation and resulting severe line blending, so we do not report them. However, Rucinski et al. (2008) has measured velocities and reported a double-lined spectroscopic orbit with a centre-of-mass velocity of |$-12.27 \pm 0.76~{\rm km\, s}^{-1}$|⁠. We list this value in Table 4. There is a 12|${_{.}^{\prime\prime}}$|1 companion (HIP017876) about 1.1 mag fainter in G that shares the parallax and proper motion of the primary, and has a median Gaia RV of |$-10.72 \pm 0.32~{\rm km\, s}^{-1}$|⁠. This is therefore a hierarchical triple system.

HIP019972: SB1 with a period of 73 d. Gaia DR3 also reports an ‘SB1’ solution with the same period.

HIP020513: This is a long-period photometric variable known as V1142 Tau. There appears to be a roughly 2 yr periodicity in our RVs, which may well be due to semiregular pulsation rather than orbital motion.

HIP021560: SB1 with a period of 1147 d. Gaia DR3 reports an ‘SB1’ solution with a similar period.

HIP022104: The Gaia DR3 catalogue lists a companion with the same parallax and proper motion at a separation of 31|${_{.}^{\prime\prime}}$|5 and 4.1 mag fainter in G. Its median RV as reported by Gaia (⁠|$-14.03 \pm 1.10~{\rm km\, s}^{-1}$|⁠) agrees well with our average primary velocity, supporting the physical association.

HIP023933: Probable binary: the median Gaia RV is some 2 |${\rm km\, s}^{-1}$| higher than ours.

HIP024478: The Gaia catalogue lists a companion 4|${_{.}^{\prime\prime}}$|4 away that is 8.2 mag fainter in G, and shares the parallax and proper motion of the primary.

HIP024780: The Gaia catalogue lists a companion 20|${_{.}^{\prime\prime}}$|7 away that is 8.1 mag fainter in G, and shares the parallax and proper motion of the primary. The primary itself appears to have a constant RV, within the uncertainties. The target displays astrometric acceleration.

HIP025386: Gaia DR3 reports an ‘SB1’ solution with a 555 d period and a small velocity semi-amplitude. The handful of RVs we obtained are not inconsistent with that orbit. We adopt the |$\gamma$| velocity from Gaia as the best representation of its mean velocity.

HIP025793: Herbig Ae/Be star (e.g. Thomas et al. 2023). There is a hint of a long period in our RV measurements. Gaia DR3 does not report any RV information.

HIP027380: The median Gaia RV is about 1 |${\rm km\, s}^{-1}$| larger than ours, possibly indicating binarity. The WDS lists a visual companion at 7|${_{.}^{\prime\prime}}$|3, some 5 or 6 mag fainter than the primary. Gaia confirms the physical association from the similar proper motions and parallaxes. The RV changes may be due to this companion.

HIP027634: Visual binary, according to the WDS, with a 3|${_{.}^{\prime\prime}}$|8 companion about 1.3 mag fainter. Gaia confirms the physical association of the two stars. The primary has a constant RV.

HIP028539: The Gaia DR3 catalogue reports a large RUWE value of 2.020, suggesting it may be a binary.

HIP029639: The median Gaia RV is 1.5 |${\rm km\, s}^{-1}$| lower than ours, suggesting it may be a binary.

HIP031498: Very rapid rotator. This makes our RVs very poor. Nevertheless, there is a hint of variability based on the pattern of the measurements. Gaia DR3 reports a median RV of |$-11.58 \pm 2.11~{\rm km\, s}^{-1}$| based on 11 measurements. This may be a binary.

HIP031613: Clearly a binary. Our SB1 solution with a period of roughly 5300 d is preliminary, as the observations do not yet cover a full cycle.

HIP031807: The median Gaia RV is 1 |${\rm km\, s}^{-1}$| lower than ours, suggesting it may be a binary.

HIP033263: Gaia DR3 reports an ‘SB1’ solution with a very small semi-amplitude of only |$K_1 = 0.410 \pm 0.054~{\rm km\, s}^{-1}$|⁠. Our 4 RVs show no significant change, but are not inconsistent with the Gaia orbit. We adopt the |$\gamma$| velocity from Gaia as the best representation of its mean velocity.

HIP034729: Other RVs from the literature, as well as Gaia, agree with ours and indicate a constant RV. The WDS lists a visual companion at 10|${_{.}^{\prime\prime}}$|8, more than 7 mag fainter than the primary. However, the information from Gaia DR3 indicates it is unrelated.

HIP036158: The RVs show an upward drift, and the median Gaia RV at an intermediate epoch is consistent with the trend. This is a binary.

HIP037104: Visual binary, according to the WDS, with a 12|${_{.}^{\prime\prime}}$|2 companion about 5 mag fainter. While the parallaxes are somewhat similar, the proper motions are not, according to Gaia, so the two stars do not appear to be associated. However, the target displays astrometric acceleration, suggesting there may be another companion.

HIP038062: SB1 system with a 372 d period. The WDS reports a companion at 9|${_{.}^{\prime\prime}}$|3 about 4.4 mag fainter, although a note says there is doubt as to its identification. The Gaia DR3 catalogue shows no companions at this distance.

HIP038923: Long-period photometric variable known as V407 Pup. The RVs may be variable as well, possibly due to semiregular pulsations.

HIP039121: We report this as an SB1 system with a period of 1010 d. The scatter from the orbital fit is unexpectedly small for an object with a rotational broadening as large as we estimate (⁠|$\sim$|70 |${\rm km\, s}^{-1}$|⁠). We suspect it to be double-lined, consistent with its sizeable minimum secondary mass, but with a line blending too severe to allow us to separate the components. In that case, the significant line broadening is the result of the blending, and the formal individual RV uncertainties listed in Table 2 may be overestimated. Gaia DR3 reports a visual companion at 34|${_{.}^{\prime\prime}}$|5 and about 10.3 mag fainter in G, which shares the same parallax and proper motion. Interestingly, it is bluer than the target, indicating it is likely a white dwarf.

HIP042331: Our RVs indicate this is a binary. Gaia DR3 reports an ‘Acceleration7’ solution, as well as a ‘SecondDegreeTrendSB1’ solution describing a non-linear RV variation. Astrometric acceleration is also reported by Brandt (2021). While our observations alone are insufficient to determine the period, it is possible to derive a very tentative orbital solution by making use of the median Gaia RV. The period of such a solution is about 3300 d, and the centre-of-mass velocity is 4.5 |${\rm km\, s}^{-1}$|⁠. The latter is similar to the value reported by Gaia (Table A1).

HIP044580: Our 3 RV measurements show no change, but only span a few weeks. The median Gaia RV over more than 2 yr is consistent with our average. The Gaia catalogue reports that about a quarter of the scans display more than one peak, suggesting a partially resolved (sub-arcsecond) binary. A speckle observation by Mason et al. (1999) revealed no companions.

HIP046130: SB2 binary with |$P = 464$| d. We find a significant primary/secondary velocity offset that is likely due to template mismatch. The primary is a rapid rotator.

HIP047155: The Gaia DR3 catalogue reports a large RUWE value of 3.268, suggesting it may be a binary.

HIP050417: Our RVs show no significant change within the errors, but the median Gaia RV is about 2.5 |${\rm km\, s}^{-1}$| higher than ours, possibly due to orbital motion.

HIP050719: No significant RV change in our 3 measurements. The median Gaia RV is 1 |${\rm km\, s}^{-1}$| higher than ours, possibly due to a companion.

HIP050999: The median Gaia RV is nearly 1 |${\rm km\, s}^{-1}$| lower than our 3 measurements.

HIP053557: This is clearly a binary with a period of the order of 4000 or 5000 d. The spectrum at the maximum velocity is double-lined, and some of the others may also be, but we are unable to measure separate velocities for the two components in most of the spectra.

HIP053759: SB1 binary with a period of 46 d. Gaia lists a companion 43|${_{.}^{\prime\prime}}$|8 away and 7.3 mag fainter in G, with the same parallax and proper motion. The median RV of this companion is listed by Gaia as |$14.92 \pm 5.06~{\rm km\, s}^{-1}$|⁠. It is somewhat different from the centre-of-mass velocity we derive for the primary (⁠|$\gamma = 8.05 \pm 0.48~{\rm km\, s}^{-1}$|⁠), but the large formal error for the companion suggests it too may be a binary. If so, this could be a hierarchical quadruple system.

HIP056703: This SB1 with a period of 152 d is the fainter secondary component of the visual binary ADS 8243 (WDS J11376−1656), separation 13|${_{.}^{\prime\prime}}$|7, |$\Delta V = 0.8$| mag. There is no parallax or proper motion listed for the primary in the Gaia DR3 catalogue. However, the very small relative motion between the stars recorded in the WDS over nearly 200 yr strongly suggests they are physically associated. This would then be a triple system. The Gaia catalogue reports that the SB1 is partially resolved in about half of the observations.

HIP056777: The Gaia DR3 catalogue indicates there is a faint companion (⁠|$\Delta G = 8.0$| mag) at a separation of 2|${_{.}^{\prime\prime}}$|5, which has the same parallax and proper motion as the primary. Our RVs for the primary show no change.

HIP057241: The median Gaia RV is formally about 2 |${\rm km\, s}^{-1}$| lower than ours, although this is only marginally significant given the uncertainties. Gaia classifies it as a photometric variable.

HIP058028: The WDS lists a companion at 15|${_{.}^{\prime\prime}}$|2 nearly 10 mag fainter. Gaia DR3 confirms it is physically associated, and that it is a very red star.

HIP058179: The median Gaia RV is 12 |${\rm km\, s}^{-1}$| lower than ours, indicating this is a binary.

HIP058861: This is the primary of a 30 arcsec visual binary, with |$\Delta V \sim 2$| mag. Gaia DR3 confirms the two objects are physically associated. It is a very rapid rotator. The median Gaia RV for the primary is more than 10 |${\rm km\, s}^{-1}$| higher than ours, indicating it is a binary. This therefore appears to be a hierarchical triple system.

HIP061018: A planet with a 20.8 d period has been identified orbiting this star (Livingston et al. 2018). The Gaia DR3 catalogue reports a large RUWE value of 1.474, suggesting it may have a stellar companion as well.

HIP062810: Gaia DR3 reports an ‘SB1’ solution with a period of 812 d. Our RVs show changes, and are not inconsistent with that orbit. We adopt the |$\gamma$| velocity from Gaia as the best representation of its mean velocity. Gaia also indicates there is a common proper motion companion 4.4 mag fainter in G, which is 13|${_{.}^{\prime\prime}}$|4 away and has the same parallax. Its median RV, according to Gaia, is |$-35.63 \pm 0.88~{\rm km\, s}^{-1}$|⁠. This value is not far from the centre-of-mass velocity that Gaia reports for the primary: |$-33.99 \pm 0.41~{\rm km\, s}^{-1}$|⁠. We conclude this is a triple system.

HIP066045: Primary of a wide (48|${_{.}^{\prime\prime}}$|1) visual binary, with the companion being 1 mag fainter, according to the WDS. Gaia DR3 confirms the physical association. The median Gaia RV is more than 1 |${\rm km\, s}^{-1}$| higher than ours, hinting at binarity. If this is confirmed, it would constitute a triple system. The median Gaia RV for the companion, according to Gaia, is |$-2.21 \pm 0.19~{\rm km\, s}^{-1}$|⁠, not far from the average for the primary.

HIP069848: This is a |$\delta$| Sct star (MX Vir), with a measured photometric frequency of 6.49 d|$^{-1}$| (⁠|$P = 0.154$| d; Barac et al. 2022). Although there is no evidence of binarity, we performed an ‘orbital’ fit simply as a means of deriving a representative average velocity. We obtained a period virtually the same as the photometric one, and a RV amplitude of 3.66 |${\rm km\, s}^{-1}$|⁠.

HIP071712: Long-period photometric variable known as EG Boo.

HIP073216: The RV appears to be variable, but we are unable to identify a plausible orbital period.

HIP074425: The Gaia DR3 catalogue reports a large RUWE value of 3.250, suggesting this may be a binary.

HIP076426: SB2 binary with a period of 1.34 d. The secondary contributes only 0.6 per cent of the flux. This is a known RS CVn variable (BI CrB; Norton et al. 2007) with a similar photometric period as the spectroscopic orbit. Gaia DR3 has an ‘SB1’ solution with the same period as ours.

HIP076768: Our RVs show no change. Gaia DR3 gives no information on the RV. The WDS has this as a visual binary with an astrometric orbit (WDS J15405−1842AB, |$P = 79.5$| yr, |$a = 0{_{.}^{\prime\prime}} 6$|⁠, magnitude difference |$\sim$| 1 mag; Horch et al. 2021).⁴  Gaia resolves the two components, but only gives the parallax and proper motion for the primary. The target displays astrometric acceleration, most likely from the same companion. The All Sky Automatic Survey (ASAS; Kiraga 2012) lists it as a photometric (rotational) variable, with a period of 3.546 d and an amplitude of 0.068 mag in V. See Section 6.

HIP078171: Visual binary with a separation of about 0|${_{.}^{\prime\prime}}$|7, according to the WDS, partially resolved by Gaia. The companion is 3 mag fainter. Our RVs show no change.

HIP078846: The median Gaia RV is several |${\rm km\, s}^{-1}$| larger than our average, suggesting it may be a binary. This is a very rapid rotator.

HIP079687: The median Gaia RV is about 5 |${\rm km\, s}^{-1}$| higher than our average, suggesting binarity.

HIP080988: SB1 binary with a period of 517 d. Gaia DR3 reports an ‘Orbital’ solution with a similar period. The secondary is invisible in our spectra, and is expected to be a low-mass star based on the Gaia inclination angle of about 97|$^\circ$|⁠.

HIP084038: SB1 binary with a period of 349 d. Two RVs from Sperauskas & Bartkevicius (2002) agree with our orbit. This is listed as a long-period photometric variable (V940 Her). Hipparcos classifies it as a semiregular variable, and gives a photometric period of |$313 \pm 2$| d (somewhat similar to the orbital period), and an epoch of maximum brightness of JD 2,448,576, which coincides with the phase of maximum RV in the orbit. Gaia DR3 reports a photometric period of about 350 d (though with a large uncertainty), again near the orbital period.

HIP084385: Our SB1 orbit with a period of 529 d has a low RV semi-amplitude, and is somewhat tentative. SIMBAD indicates it is a long-period photometric variable (V942 Her). This could be the reason for the velocity variations.

HIP084680: Long-period photometric variable (V818 Her) classified as a semiregular (SRb), according to Jerzykiewicz (1984), with a cycle length of tens of days. Its brightness also appears to vary on a shorter time-scale of 6.75 d (Koen & Eyer 2002). The median Gaia RV is slightly different from our average, but this may be related to the oscillations rather than to binarity.

HIP085015: SB1 with a period of 340 d. Gaia DR3 reports an ‘SB1’ solution with the same period, but less precise orbital parameters.

HIP088984: The median Gaia RV is about 1.5 |${\rm km\, s}^{-1}$| lower than ours. This could be a binary.

HIP089397: The period of this SB1 is 2580 d. There is a suggestive periodicity of about 580 d in the residuals, which could be due to photometric variability, although we find no record of that in the literature.

HIP089553: Photometric variable (NSV 24367) with a period of 20.4 d and a 0.1 mag amplitude (Burggraaff et al. 2018).

HIP090886: SB1 binary with a period of 1250 d. Gaia DR3 reports an ‘Acceleration7’ solution.

HIP091444: SB1 binary with a period of 1171 d. Gaia DR3 reports an ‘Acceleration7’ solution.

HIP091828: This is the secondary in an 11|${_{.}^{\prime\prime}}$|5 visual binary (WDS J18434+3533), according to the WDS, in which the primary (HIP091829, not in our sample) is 1 mag brighter. Gaia DR3 confirms the physical association. Our RVs show no change, with a mean velocity that agrees with the median value reported for the primary by Gaia (⁠|$-8.56 \pm 0.77~{\rm km\, s}^{-1}$|⁠).

HIP092787: One of our RVs is almost 2 |${\rm km\, s}^{-1}$| higher than the rest, suggesting it may be a binary. The Gaia catalogue does not report a non-single star solution, but indicates that the object was partially resolved 16 per cent of the time, supporting our suspicions. This is a Mira variable (V913 Aql).

HIP093510: Our SB1 orbit with a period of 731 d has a very small RV amplitude, and consequently a small minimum secondary mass of |$M_2 \sin i = 0.00838 \pm 0.00071 (M_1+M_2)^{2/3} M_{\odot }$|⁠. It is possible this variability is due to pulsation, rather than binarity. Gaia DR3 lists a red companion about 4|${_{.}^{\prime\prime}}$|5 away and 9.6 mag fainter in G, which appears to be physically associated, based on the very similar parallaxes and proper motions. This could then be a triple system.

HIP095138: This target displays astrometric acceleration, suggesting binarity.

HIP095537: The WDS lists this object as a close visual binary (⁠|$\rho = 0{_{.}^{\prime\prime}} 1$|⁠, |$\Delta m = 0.3$| mag). Our RVs show a slight downward drift, possibly caused by the companion. This is also a long-period semiregular variable (V557 Lyr, SRd).

HIP097678: Long-period photometric variable (⁠|$\Pi$| Dra, SRd:). A period of 21.7 d has been reported (Dubath et al. 2011).

HIP097957: Our last RV is almost 5 |${\rm km\, s}^{-1}$| lower than the rest, gathered 3 yr earlier. The median Gaia RV agrees with the trend. Gaia DR3 reports an ‘Acceleration9’ solution. This is clearly a long-period binary. The target displays astrometric acceleration.

HIP098286: The cross-correlation functions (CCFs) give the impression of having multiple peaks (at least three), suggesting a composite spectrum. As we do not have enough spectra to confirm this, the RVs reported here were derived under the assumption that it is a single rapidly rotating star (with all CCF peaks blended together). The median Gaia RV agrees with the average from our 7 observations. The Gaia DR3 catalogue lists a very red companion at 13|${_{.}^{\prime\prime}}$|5 with the same parallax and proper motion, which is 9 mag fainter in the G band. This companion may itself be a binary, as the Gaia catalogue indicates it was partially resolved in most scans. In that case, the system would be at least triple, or possibly more complex. HIP098286 is also listed as a photometric variable (NSV 24947), with a possible period of 0.17328 d (Samus’ et al. 2017), perhaps a |$\delta$| Sct star. An occultation of the primary star by asteroid 248 Lameia was observed in June of 1998, and used to estimate the size and shape of the asteroid (Fraser & Overbeek 1998).

HIP098443: This is an SB2 (BD+17 4185) with a period of 2297 d. Gaia DR3 reports an ‘Acceleration7’ solution. The WDS lists several very wide companions, of which the brightest is BD+17 4187. A WDS note indicates our target is variable, and the AAVSO lists it as type ‘LB’ (Watson, Henden & Price 2006), which is a slow irregular variable (NSV 12668).

HIP098762: Our last RV is about 1 |${\rm km\, s}^{-1}$| lower than the previous ones, obtained 3 yr earlier. The median Gaia RV seems to follow the trend. We consider this to be a likely binary.

HIP099070: The Gaia DR3 catalogue reports a large RUWE value of 1.666, suggesting this may be a binary.

HIP100180: The WDS reports this is a visual binary with a 0|${_{.}^{\prime\prime}}$|12 companion and a magnitude difference of 0.9 mag (Hipparcos detection). It is also photometrically variable (NSV 25104; Samus’ et al. 2017). Our RVs show no significant change.

HIP100376: Our observations do not cover the phase of maximum RV, but the SB1 orbit with |$P = 379$| d is still well determined.

HIP100534: The variable RV indicates this is a binary with a long period. The median Gaia RV agrees with the trend.

HIP100845: The median Gaia RV is about 3 |${\rm km\, s}^{-1}$| higher than ours, suggesting binarity. This is a very rapid rotator.

HIP101219: While we report an SB1 orbit here (⁠|$P = 566$| d), the low-amplitude RV variability could also be due to pulsation rather than orbital motion.

HIP101796: Very poor RVs. The Gaia DR3 catalogue reports a large RUWE value of 2.420, suggesting this may be a binary.

HIP102641: Visual binary, according to the WDS. The companion is currently at 17|${_{.}^{\prime\prime}}$|7, according to Gaia, and is about 3.5 mag fainter. The similar parallaxes and proper motions from Gaia DR3 confirm the physical association. This is also supported by the median RV reported by Gaia for the companion (⁠|$-16.62 \pm 1.02~{\rm km\, s}^{-1}$|⁠), which agrees with the primary velocity. Our 3 RVs of the primary show no change, but span less than a year. The companion may itself be a binary, as it was reported by Gaia to be partially resolved in about a quarter of the scans. In that case, it would be a triple system.

HIP102804: Photometric variable (NSV 25356) with |$P = 64.9$| d and a 0.15 mag amplitude (Burggraaff et al. 2018). The median Gaia RV is about 1.5 |${\rm km\, s}^{-1}$| lower than ours, suggesting this could be a binary. Gaia reports the star was partially resolved in 20 per cent of the scans, suggesting it is a binary. Several wider companions listed in the WDS are not physically associated, according to Gaia.

HIP104444: A lunar occultation observation by Eitter & Beavers (1974) did not detect any companions.

HIP107325: Our last RV is marginally lower than the rest, from 3 yr earlier. The median Gaia RV is intermediate. Gaia DR3 lists a 3|${_{.}^{\prime\prime}}$|8 companion 7.5 mag fainter in G, which has a similar parallax and somewhat similar (if uncertain) proper motion.

HIP107588: Although we see only one set of lines in our spectra, the CCF peak has a triangular shape different from that of other stars with the same rotational broadening. According to the WDS, this is a visual binary with a separation of 0|${_{.}^{\prime\prime}}$|6 and a magnitude difference of 1.3 mag. It was also partially resolved by Gaia. We suspect, then, that the CCF shape is the result of a blend between a rapidly rotating star, presumably the primary, and a more slowly rotating secondary, with a very small velocity separation that we cannot resolve. Gaia classifies it as a photometric variable.

HIP108552: The secondary in this SB2 system (⁠|$P = 105$| d) contributes only about 0.5 per cent as much flux as the primary.

HIP109339: SB1 binary with a period of 21 d. The WDS lists a wide companion at 22|${_{.}^{\prime\prime}}$|9 and about 1.3 mag fainter. Gaia confirms the physical association from the similar parallaxes and proper motions, so this is at least a triple system.

HIP110590: This is a binary. We see an almost linear increase in the RV over 9 yr. The median Gaia RV agrees with this trend.

HIP110993: The median Gaia RV is 8 |${\rm km\, s}^{-1}$| higher than ours, suggesting binarity. This is a rapid rotator.

HIP111003: Gaia lists a 4|${_{.}^{\prime\prime}}$|2 common proper motion companion 8.8 mag fainter in G, which has the same parallax as the primary. Our RVs for the primary span 3 yr, and are constant.

HIP111522: The RV is variable, indicating binarity. The orbital period is likely long. The median Gaia RV agrees with the trend.

HIP112250: Long-period irregular variable (QV Peg). The median Gaia RV is 1.5 |${\rm km\, s}^{-1}$| lower than ours, suggesting binarity.

HIP112272: The RVs are clearly variable, likely with a long orbital period. Gaia DR3 reports an ‘Acceleration7’ astrometric solution, as well as a non-linear RV solution of type ‘SecondDegreeTrendSB1’. The median Gaia RV agrees with the trend. Brandt (2021) also reported that the target displays significant astrometric acceleration.

HIP112931: SB1 binary with a period of 488 d. Gaia DR3 also reports an ‘SB1’ solution, with the same period.

HIP113787: SB1 binary with a 469 d period. Gaia DR3 reports an ‘Orbital’ solution with a similar period.

HIP116987: This is a very rapid rotator. The median Gaia RV is about 3 |${\rm km\, s}^{-1}$| higher than ours, hinting that it may be a binary.