Identification of additional young nearby runaway stars based on Gaia data release 2 observations and the lithium test

Runaway stars are characterised by their remarkably high space velocities, and the study of their formation mechanisms has attracted considerable interest. Young, nearby runaway stars are the most favorable for identifying their place of origin, and for searching for possible associated objects such as neutron stars. Usually the research field of runaway stars focuses on O- and B-type stars, because these objects are better detectable at larger distances than late-type stars. Early-type runaway stars have the advantage, that they evolve faster and can therefore better be confirmed to be young. In contrast to this, the catalogue of young runaway stars within 3 kpc by Tetzlaff, Neuh\"auser,&Hohle (2011) contains also stars of spectral type A and later. The objects in this catalogue were originally classified as young ($\le 50$ Myr) runaway stars by using Hipparcos data to estimate the ages from their location in the Hertzsprung-Russell diagram and evolutionary models. In this article, we redetermine and/or constrain their ages not only by using the more precise second data release of the Gaia mission, but also by measuring the equivalent width of the lithium (6708 $\unicode{xC5}$) line, which is a youth indicator. Therefore, we searched for lithium absorption in the spectra of 51 target stars, taken at the University Observatory Jena between March and September 2020 with the \'Echelle spectrograph FLECHAS, and within additional TRES-spectra from the Fred L. Whipple Observatory. The main part of this campaign with its 308 reduced spectra, accessible at VizieR, was already published. In this work, which is the continuation and completion of the in 2015 initiated observing campaign, we found three additional young runaway star candidates.


INTRODUCTION
Runaway stars can result from gravitational interactions in dense stellar clusters (Poveda, Ruiz, & Allen, 1967) or from a supernova explosion in a binary system (Blaauw, 1961).
In both cases the ejected stars move on with higher velocities compared to typical field stars. Runaway stars can be traced back to their birth place by considering the influence of the Galactic potential. The correctness of those calculations depends mainly on the time since the ejection and will be even more challenging after several Myr. Credible runaway stars, that originated e.g. from the explosion in a binary system, should not be older than about 50 Myr (including the lifetime of the progenitor star until the supernova).
In 2011 the catalogue of Young runaway stars within 3 kpc 1 was published by Tetzlaff et al., which contains not only O-and B-type stars, but also every type of star with a peculiar space velocity pec > 28 km/s. These objects were identified within a combined analysis of spatial, tangential and radial velocities, measured by the Hipparcos satellite (Perryman et al., 1997). Furthermore, their ages were derived by comparing luminosity and effective temperature to different evolutionary models.
However, these age estimations can be improved and/or constrained with the more accurate second data release (Gaia Collaboration et al., 2018) of the Gaia mission (Gaia DR2) of the European Space Agency. Additionally, we searched for another youth indicator, namely the absorption line of the lithium doublet at 6708 Å based on Neuhäuser (1997), within our spectroscopic observing program for selected stars from the catalogue by Tetzlaff et al.. This observing program started in 2015 and the first results were already published in Bischoff, Mugrauer, Torres, et al. (2020). The remaining targets of this project are presented and discussed in this article.
In section 2, we describe the sample selection, spectroscopic observations and data reduction. In section 3, we characterise the physical properties of our targets based on their Gaia DR2 data and in section 4 we explain the measurements of the Li (6708 Å) line in all taken spectra. Section 5 contains the age estimation of the dwarf stars. Finally, all results are discussed in section 6 and we draw conclusions in section 7.

SAMPLE SELECTION, OBSERVATIONS AND DATA REDUCTION
Our sample was selected from the catalogue by Tetzlaff et al. (2011). These targets had to be brighter than ≤ 8.5 mag in order to record spectra with sufficiently high signal-to-noiseratio (SNR) > 50 within integration times of a few minutes. Furthermore, their declination angle had to be > −14 • so that they can be observed at air masses < 2.4 from Jena. As mentioned in Bischoff, Mugrauer, Torres, et al. (2020), in total 460 stars were identified that fulfill these conditions and 308 of them were already observed and published. However, the remaining list of 152 stars could be further shortened 1 http://cdsarc.u-strasbg.fr/viz-bin/cat/J/MNRAS/410/190 with a rough analysis by using data from Gaia DR2 and the catalogue of Bailer-Jones, Rybizki, Fouesneau, Mantelet, & Andrae (2018) to rule out most of the giants. This results in a target list that contains 51 stars, which were observed with the fibre-linked Échelle spectrograph FLECHAS (Mugrauer, Avila, & Guirao, 2014) and processed in the following.
Our spectra were taken with FLECHAS, operated at the Nasmyth-focus of the 90 cm-telescope ( ∕ = 15) of the University Observatory Jena (Pfau, 1984). The observations, that include 153 spectra with a total integration time of 25.75 h, were carried out between March and September 2020.
All spectra were recorded with the 1x1 binning mode of the spectrograph FLECHAS, using individual detector integration times in the range between 150 s and 1200 s dependent on the target brightness. The instrument has a resolving power of ≈ 9, 300 and covers a spectral range from 3900 Å to 8100 Å within 29 orders (Mugrauer et al., 2014). Three spectra per target were always taken to remove cosmics and to reach a sufficiently high SNR, which was measured in all fully reduced spectra at = 6700 Å, which is the centre of the spectral order with the Li (6708 Å) line. On average SNR = 101 is reached in the FLECHAS spectra of our targets, with range from 50 for HIP 30030 to 201 for HIP 19587. Further details are given in the observation log in Table A1.
Three flat-field frames of a tungsten lamp and three spectra of a thorium-argon (ThAr) lamp are recorded immediately before the observation of each target for calibration purposes. Each calibration file has an individual integration time of 5 s. About 700 detected emission lines are available in the ThAr spectra for wavelength calibration. The long-term stability of the wavelength calibration of FLECHAS was confirmed in studies by Irrgang, Desphande, Moehler, Mugrauer, & Janousch (2016), Bischoff et al. (2017), Heyne et al. (2020) and Bischoff, Mugrauer, Lux, et al. (2020). Additionally, for the dark subtraction, three dark frames for all used integration times were taken in every observing night. An overscan region is always read out to measure and later correct the bias level. The FLECHAS detector has a typical read-noise of about 11 − and the gain is 1.3 − /ADU. The FLECHAS CCD-sensor and the whole instrument is described in detail in Mugrauer et al. (2014).
The observations were reduced with a dedicated pipeline for FLECHAS, developed at the Astrophysical Institute Jena, which includes dark and bias subtraction, flat-fielding, extraction and wavelength calibration of the individual spectral orders. Furthermore, including the final averaging and normalisation of the spectra (Mugrauer et al., 2014).
As part of a separate long-term spectroscopic monitoring program to measure radial velocities and discover binary systems in another sample of runaway stars from Tetzlaff et al. (2011), HIP 2710 and HIP 12297 were also observed between September 2013 and March 2017 with the Tillinghast Reflector Echelle Spectrograph TRES (Fűrész, 2008;Szentgyorgyi & Furész, 2007). The spectrograph is attached to the 1.5 m Tillinghast reflector at the Fred L. Whipple Observatory on Mount Hopkins (Arizona, USA). This bench-mounted, fiberfed instrument generates spectra at a resolving power of ≈ 44, 000 that cover the wavelength region between 3800 Å and 9100 Å in 51 orders. Exposure times ranged from 60 s to 250 s, depending on brightness and weather conditions. Exposures of a ThAr lamp were taken before and after each science frame, and the observations were also reduced with a dedicated pipeline, which follows the procedure described in Fűrész (2008).

TARGET CHARACTERISATION WITH GAIA DR2 DATA
The detailed characterisation of our targets in this article focusses mainly on data from the Gaia DR2. We considered only gold flag photometry (as described by Andrae et al. 2018) entries from the Gaia DR2. The apparent brightness in theband for each target was corrected according to the brightness relations in Maíz Apellániz & Weiler (2018) and taking into account the new defined transmission profiles for the Gaia filters from Weiler (2018). We did not use the estimates for the -band extinction in Gaia DR2, because they were not available for 16 targets of our sample and sometimes they were significantly overestimated, e.g. in the case of HIP 56770, an extinction of G = 0.995 +0.188 −0.314 mag seems unrealistic, given its distance of 47.  Table A3. Additionally, we list effective temperatures eff , stellar radii and luminosities of our targets, if available and the target was not a spectroscopic binary.
We show the distance distribution of our targets in Figure 1  the Hipparcos parallax (van Leeuwen, 2007) instead. For the same four targets, we converted the apparent magnitude of the Hipparcos system p and the − magnitudes, provided by the Hipparcos catalogue (Perryman et al., 1997), into theband with the relations by Evans et al. (2018). Their eff were derived from their Hipparcos spectral type (SpT) with the corresponding SpT-log( eff )-relations by Damiani et al. (2016). HIP 113811 had also no entry in the extinction catalogue of Gontcharov & Mosenkov (2017). Therefore, we used the reddening ( − ) from Green, Schlafly, Zucker, Speagle, & Finkbeiner (2019), which was transformed into r with the relation given there and afterwards into G (Wang & Chen, 2019).
We searched for multiplicity within our sample in the 9th Catalogue of Spectroscopic Binary Orbits (Pourbaix et al., 2004) to correct their absolute brightness. In the case of a double-lined spectroscopic binary, we took the mass ratio of the secondary and the primary component and converted it into a luminosity ratio via ∝ 4.5 , which is applicable for stellar masses between 2 M ⊙ and 0.5 M ⊙ (Salaris & Cassisi, 2005), suitable for our sample of spectral types ranging between A7 and K2. The luminosity ratios were then used to determine how much brighter is the binary in comparison to a single source. We assumed that the brightness difference between the secondary and the primary is at least 1 mag for the single-lined binaries, if no further information about mass ratios or the systems were available. It follows that those systems could be up to ∼ 0.364 mag brighter than a corresponding single star. The identified spectroscopic binaries and their mass ratios are listed in Table 1.
Our targets are illustrated in a Hertzsprung-Russell-Diagram (HRD) in Figure 2. For example, HIP 113811 ( G = −1.891 +1.275 −3.070 mag) and  (2017) HIP 115906 ( G = −1.754 +0.459 −0.531 mag) can be excluded as possible young runaway stars, because they are far too bright for their given eff to be dwarf stars and are clearly located on the giant branch. Typical Geff -relations for dwarf stars are given in Pecaut & Mamajek (2013) 2 and Baraffe, Homeier, Allard, & Chabrier (2015) 3 . However, it is not always easy to decide, based on their location in the HRD alone, whether an object is either a pre-main-sequence star or it has already left the main-sequence. Therefore, we studied their listed surface gravities in the StarHorse catalogue (Anders et al., 2019) and if the given range of log( [cm/s 2 ]) ≳ 3.8, the target was classified as dwarf star.
To identify the young stars among the dwarfs, further analysis is needed.

LI (6708 Å) EQUIVALENT WIDTH MEASUREMENTS AND ABUNDANCES
The Ca (6718 Å) line was used to correct the doppler shift in every spectrum, adopting 0 = 6717.685 Å as laboratory wavelength as listed in the ILLSS catalogue (Coluzzi, 1993), because it is the most prominent spectral line nearby Li (6708 Å), and is also detected in the same spectral order.
We measured the equivalent width via a direct integration of the line profiles in the reduced spectra by using the IRAF _colors_Teff.txt 3 http://perso.ens-lyon.fr/isabelle.baraffe/BHAC15dir/ (Tody, 1993) task splot as explained in Bischoff, Mugrauer, Torres, et al. (2020). The equivalent widths of the Li (6708 Å) line of all targets are given in Table A2. Furthermore, in appendix Dwarf stars with significant lithium detection and Sub-giant/giant stars with significant lithium detection, we illustrate all FLECHAS spectra that show a significant detection, that means Li ≥ 3 ⋅ Li . All spectra are sorted by their spectral type according to SpT-log( eff )-relations from Damiani et al. (2016). The uncertainty of the spectral classification is about two sub-classes. In additional TRES-spectra of HIP 2710 and HIP 12297, we searched for lithium as described above. The measured average equivalent width from the four spectra of HIP 2710 is Li = (43 ± 7) mÅ . In the case of HIP 12297, none of the 24 spectra showed a significant Li (6708 Å) line. These results are consistent with the measurements from FLECHAS.
Our equivalent widths measurements of the identified dwarf stars with significant lithium detection were converted into abundances by using curves of growth from Soderblom et al. (1993), that are based on an abundance scale of log 10 ( ) H = 12 and we assigned the best matching values of log 10 ( ) and eff ( Table 2 in Soderblom et al. 1993) to those of our sample. Indeed, some of our eff values were outside the covered range of Soderblom et al. (1993). Therefore, we fit quadratic polynomials as a function of eff for constant values of log 10 ( ). The results of this conversion are presented in Table 2.

AGE ESTIMATION
We can constrain the ages of our identified dwarf stars with further isochrones. The isochrones in Figure 3 were calculated with models of Bressan et al. (2012) for metallicity = 0.0152. Assuming solar metallicity is justified, because all dwarfs exhibit an average metallicity of [M∕H] = 0.09 with a standard deviation of 0.13 dex. This estimate is based on a compilation of metallicities for our stars from the VizieR database (Ochsenbein, Bauer, & Marcout, 2000), taken from the catalogues by Brewer, Fischer, Valenti, & Piskunov (2016) Valenti & Fischer (2005). The influence of the metallicity scatter is shown in Figure 4. Here, we show as an example the 50 Myr isochrone. The differences between different metallicities are smaller or  (Bressan et al., 2012) for 50 Myr and 5 Gyr with solar metallicity = 0.0152 in both distributions.

TABLE 2
The classified dwarf stars with their identification numbers as shown in Figure 2, with their effective temperatures eff from Gaia DR2 and the measured equivalent widths of the Li (6708 Å) line Li . We list only targets with significant lithium detection. The abundances log 10 ( Li ) based on Soderblom et al. (1993)  However, we have to consider that many of our dwarf stars are consistent with more than one isochrone within their uncertainties in Figure 3, especially if they are matching one of the Gyr isochrones. For that reason, the estimated ages based on the location in the HRD in Table 3, are sometimes only listed with lower limits. Additional information, as explained in the following, are necessary to identify and/or further constrain the young stars among our targets. We compared all dwarf stars with significant lithium detection in their spectra to distributions of stellar clusters with known ages, as illustrated in Figure 5. The curves are polynomial fits for observed average equivalent width measurements dependent on eff (cluster data and fits from E. Mamajek, priv. communication). E. Mamajek's original plot is available online 4 .
The age errors for this method appear to be 10% − 20%, as stated by Soderblom, Hillenbrand, Jeffries, Mamajek, & Naylor (2014), and the detection of lithium in a low-mass star with known effective temperature can give an upper limit to its age. HIP 22524 (# 5 in Table 3 and Figure 5) is only consistent with the 50 Myr curve and within its uncertainties it reaches clearly the area for stars that are younger than 50 Myr. Hence, its assigned age is ≤ 50 Myr. HIP 51386 (# 7) and HIP 71631 (# 14) fit with more than one age curve and are also consistent with ages below 50 Myr within their uncertainties. Therefore, they were considered to be ≤ 50 ... 120 Myr and ≤ 50 ... 90 Myr, respectively.
HIP 30030 (# 9) and HIP 16563 (# 13) should be handled with care, because as stated by Soderblom et al. (2014) the lithium method does not give reliable age estimations below 20 Myr. For that reason, these stars were classified to have an age of < 50 Myr. Furthermore, for eff > 6300 K the < 5 Myr age curve in Figure 5 is an extrapolation, because in this range no stars with lithium and the corresponding age were observed.
The remaining dwarfs with a significant detected Li (6708 Å) line cross more than one age curve in Figure 5. Their age estimation was encircled by the youngest age curve and the oldest age curve, that were hit within their uncertainties. For example, HIP 19855 (# 12) matches the curves for 500 Myr and 625 Myr and therefore, it was estimated to have an age of 500 to 625 Myr. The age estimations for the others stars are given in the corresponding column of Table 3. HIP 40774 (# 16) is about 175 Myr old, because it only fit with those age curve.  Table 3.

DISCUSSION
The aim of our project was to identify and/or confirm young mid-and late-type runaway stars from the catalogue by Tetzlaff et al. (2011) based on their location in the HRD with the more accurate Gaia DR2 data. In addition, we took spectra and searched for the absorption of the Li (6708 Å) line, which is a youth indicator. Our sample consists of 2 A-type, 38 F-type, 7 G-type and 4 K-type stars. Their SpT was assigned based on their eff with the SpT-log( eff )-relation from Damiani et al. (2016).
We studied the surface gravity of our targets within the StarHorse catalogue to rule out possible sub-giants, that have typically log( [cm/s 2 ]) < 3.8. As a result of this, 23 targets could be excluded as already evolved stars.
The main goal of this study was to find stars that are younger or about 50 Myr. Therefore, we used isochrones, as illustrated in Fig. 3, to give an age limit of our identified dwarf stars. Due to their derived scatter of metallicity as explained above, assuming solar metallicity for our sample and using it TABLE 3 Our dwarf stars with their identification numbers as shown in Figure 3, Figure 4 and Figure 5, listed with their SpT derived from their eff using the SpT-log( eff )-relation from Damiani et al. (2016), their distances according to Bailer-Jones et al. (2018), the measured equivalent width of the Hydrogen Balmer line H (all in absorption) and the measured equivalent width of Li (6708 Å) Li . Furthermore, we list the estimated age derived from the position in the HRD, as well as the age according to the lithium test from this work and the age from Tetzlaff et al. (2011 54.9 ± 10.4 * this work † both H -lines of this spectroscopic binary could be measured for isochrone fitting seems reasonable. We considered possible multiplicity within our sample and checked the catalogue of Pourbaix et al. (2004) for spectroscopic binaries, to correct their position in the HRD. We list all identified binaries in Table 4 with their measured radial velocity, which was determined from the Ca (6718 Å) line. The secondary component of the double-lined binaries HIP 5081 and HIP 114379 could also be measured.
Nearly all dwarf stars are consistent within their uncertainties with isochrones in the range of a few Gyr. Therefore, another indicator is needed to confirm the youth of the targets. For this, we measured the equivalent width of the Li (6708 Å) line. HIP 16563 has the strongest lithium line with (254 ± 14) mÅ . In contrast to this, 30 stars of the sample showed no significant Li (6708 Å) line within their spectra.
Equivalent width measurements of the dwarfs were then converted into abundances using the curves of growth from Soderblom et al. (1993). HIP 44212 is also listed in the catalogues of (Ramírez, Fish, Lambert, & Allende Prieto, 2012) and (Lambert & Reddy, 2004) and their measurements (log( Li ) = 2.65 ± 0.03 and log( Li ) = 2.55 ± 0.10, respectively) are consistent with our determined lithium abundance.
We found two young targets within our sample, namely HIP 30030 (# 9) and HIP 16563 (# 13), that show a relatively large amount of lithium in comparison to their expected lower age limit from their location in the HRD. Their spectra are given in Figure 6. These objects were assigned from HRD isochrone fitting to be ≥ 20 Myr and > 20 Myr, respectively. Their position in Figure 5 could suggest an age of ∼ 5 Myr. However, these two stars should be handled with care, because as mentioned above the lithium method does not give very reliable age estimations below 20 Myr (Soderblom et al., 2014). Therefore, in combination with isochrone fitting and the lithium test, HIP 30030 (# 9) is more likely older or equal than 20 Myr and younger than 50 Myr, while HIP 16563 (# 13) is older than 20 Myr and younger than 50 Myr.
GJ 182 (HIP 23200) is one of the youngest stars (e.g. Bischoff, Mugrauer, Torres, et al., 2020) and, given its distance of 24.38 ± 0.02 pc (Bailer-Jones et al., 2018), there is no star known that is both younger and more nearby. Its age was estimated by Bischoff, Mugrauer, Torres, et al. (2020) to be ranging between 20 Myr and 50 Myr, which is consistent with ages from Brandt et al. (2017), Binks & Jeffries (2014) and Bell, Mamajek, & Naylor (2015) -and also with membership to the Pic moving group (Lee & Song, 2018). HIP 113174 (# 2) and HIP 51386 (# 7) have comparable ages to GJ 182 and are not much further away than GJ 182 ( ≤ 2 ⋅ GJ 182 , as listed in Table 3). Therefore, these two young runaway star candidates, which were listed as field stars in David & Hillenbrand (2015) and/or Pace (2013), are the best targets for follow-up investigations of their origin from either dynamical or supernova ejection based on their young age and proximity to the Earth. Even if HIP 26690 (# 3) is further out, given its distance of 167 ± 2 pc (Bailer-Jones et al., 2018), it has a comparable age to HIP 113174 (# 2) and HIP 51386 (# 7) and is therefore also a good candidate. In contrast to the mentioned three targets, HIP 28469 (#1), HIP 22524 (# 5), HIP 30030 (# 9), HIP 16563 (# 13) and HIP 71631 (# 14) are rather no runaway stars, because they are associated with young nearby stellar clusters (Gagné et al., 2018;López-Santiago, Montes, Crespo-Chacón, & Fernández-Figueroa, 2006;Montes et al., 2001). HIP 28469 (#1) and HIP 22524 (# 5) were assigned to be members of the Hyades cluster and HIP 30030 (# 9) belongs to the Columba association (Gagné et al., 2018). Furthermore, HIP 16563 (# 13) is part of the AB Doradus moving group (Gagné et al., 2018) and HIP 71631 (# 14) is listed as member of the Local Association subgroup B4 in López-Santiago et al. (2006). As presented in Table 3, our derived age for HIP 30030 (# 9) is consistent with 42 +6 −4 Myr, the age of its associated cluster in Gagné et al. (2018). We derived an age of 20 ... 50 Myr for HIP 16563 (# 13), which younger than 149 +51 −19 Myr of its moving group as given in Gagné et al. (2018). However, López-Santiago et al. (2006) list ages for the AB Doradus moving group ranging between 30 Myr and 150 Myr, which are consistent with our derived age for its possible member star HIP 16563 (# 13). The Local Association subgroup B4 has an average age of ∼ 150 Myr but also contains stars which are consistent with 80 Myr (López-Santiago et al., 2006). That fits with our derived age of 20 ... 90 Myr for HIP 71631 (# 14). The determined age of the stars in the Hyades cluster is 750 ± 100 Myr according to Gagné et al. (2018). However, the basis for the 625 Myr isochrone in Figure 5 is also the distribution of the Hyades cluster. Those stars scatter around the 625 Myr isochrone in the original plot 5 , that was done by Eric Mamajek, and the location of the lithium richest members are consistent with the location of HIP 22524 (# 5). HIP 28469 (#1) is also close to this area within its uncertainties. If HIP 28469 (#1) and HIP 22524 (# 5) are actually members of the Hyades cluster, their upper age limit would be then the cluster age.

CONCLUSIONS
We carried out spectroscopic follow-up observations for 51 targets from the catalogue by Tetzlaff et al. (2011) to search for the Li (6708 Å) absorption line, which is a youth indicator. 21 stars have a significantly detected lithium line within their spectra. In combination with isochrones based on the Gaia DR2, we classified 8 objects as young with ages ≤ 50 Myr. Some of these targets are already associated with young nearby stellar clusters. HIP 113174 (# 2), HIP 26690 (# 3) and HIP 51386 (# 7) are the remaining young runaway star candidates, which are outside of known clusters. They are suitable for further follow-up observations to identify their place of origin and/or to search for possible companions.
As it is the standard in our survey the fully reduced FLECHAS spectra as well as the measured equivalent widths of the Li (6708 Å) line will be made available in VizieR after publication.