A search for runaway stars in twelve Galactic supernova remnants
RReceived 09 November 2020; Revised –; Accepted 07 December 2020DOI: xxx/xxxx
ORIGINAL ARTICLE
A search for runaway stars in twelve Galactic supernova remnants † Oliver Lux* | Ralph Neuhäuser | Markus Mugrauer | Richard Bischoff
Astrophysical Institute and UniversityObservatory, Friedrich Schiller UniversityJena, Thuringia, Germany
Correspondence *Oliver Lux, Schillergäßchen 2, 07745 Jena.Email: [email protected]
Runaway stars can result from core-collapse supernovae in multiple stellar systems.If the supernova disrupts the system, the companion gets ejected with its formerorbital velocity. A clear identification of a runaway star can yield the time and placeof the explosion as well as orbital parameters of the pre-supernova binary system.Previous searches have mostly considered O- and B-type stars as runaway starsbecause they are always young in absolute terms (not much older than the lifetime ofthe progenitor) and can be detected up to larger distances. We present here a searchfor runaway stars of all spectral types. For late-type stars, a young age can be inferredfrom the lithium test. We used
Gaia data to identify and characterise runaway starcandidates in nearby supernova remnants, obtained spectra of 39 stars with UVES atthe VLT and HDS at the Subaru telescope and found a significant amount of lithiumin the spectra of six dwarf stars. We present the spectral analysis, including measure-ments of radial velocities, atmospheric parameters and lithium abundances. Thenwe estimate the ages of our targets from the Hertzsprung-Russell diagram and withthe lithium test, present a selection of promising runaway star candidates and drawconstraints on the number of ejected runaway stars compared to model expectations.
KEYWORDS: supernova remnants – stars: fundamental parameters – stars: kinematics
Runaway stars are characterised by having higher peculiarvelocities than typical field stars, 𝑣 pec ≳
20 − 30 km s −1 (e.g.Renzo et al. 2019; Tetzlaff 2013). There are two suggestedejection mechanisms: (i) In the binary supernova ejectionscenario (BES), the secondary component from a binary ormultiple system is ejected after the primary explodes in asupernova (SN). A binary becomes unbound if more than halfof the total system mass gets suddenly ejected during the SN( Blaauw kick , Blaauw 1961). Systems that do not fulfil thiscriterion can still get unbound if the newborn neutron star getsa sufficiently high kick from an asymmetric SN (Renzo etal. 2019, hereafter Rz19). The ejected companion flies awaywith its former orbital velocity. The O9.5 star 𝜁 Oph and PSR † Based on service mode observations obtained from ESO-VLT in project0100.D − B1929+10 were the first pair of a runaway star and a neu-tron star suggested to have been together as a massive binary(Hoogerwerf, de Bruijne, & de Zeeuw, 2000, 2001). How-ever, later works (Chatterjee et al., 2004; Tetzlaff, Neuhäuser,Hohle, & Maciejewski, 2010; Zehe et al., 2018) show thatthe association is very unlikely. Recently, Neuhäuser, Gießler,& Hambaryan (2020) connected 𝜁 Oph to PSR B1706 − .
78 ±0 . megayears (Myr) ago. The moderate distance 𝑑 = 107 ±4 pc of the corresponding SN makes it likely that this eventcontributed to the Fe content that was detected in the Earthcrust and ocean sediments, dated to an arrival time at the Earth ∼ a r X i v : . [ a s t r o - ph . S R ] J a n LUX
ET AL . OB associations (Poveda, Ruiz, & Allen, 1967). Examplesare AE Aur and 𝜇 Col, which were ejected from the Trapez-ium Cluster in the Orion Nebula, possibly by an encounterwith the massive binary 𝜄 Ori (Hoogerwerf et al., 2001). Thismechanism generally produces higher velocities than the BES(Rz19), whereas the highest stellar velocities can be explainedby the Hills mechanism: The tidal encounter with the super-massive black hole in the Galactic centre (Hills, 1988) caneject a hypervelocity star with up to 𝑣 pec ≳ km s −1 .The limit 𝑣 pec ≳
20 − 30 km s −1 is based on statistical rea-sons. Tetzlaff (2013) fitted the velocity distribution of nearbystars with two Maxwellians; one for the regular populationand one for the high-velocity population. Both intersect at 𝑣 pec ≈ 25 km s −1 . However, Rz19 performed an extensive pop-ulation synthesis of massive binaries and emphasised that upto 95 % of SN-ejected main-sequence (MS) companions areexpected to have 𝑣 pec ≤ km s −1 and should be called walk-away stars . In this work, we will not divide between runaway and walkaway stars. All stars with a possible BES origin willbe classified as runaway star candidates, regardless of theirvelocity.While both ejection scenarios are verified by observations,it is still unclear whether one of them is dominant. Therefore,it is important to observe a large number of runaway stars tofind whether they are produced by the BES or the DES. Onepossibility to confirm a BES origin is to look for SN debrisin the stellar atmosphere, which can principally be done byhigh-resolution spectroscopy (e.g. Przybilla, Fernanda Nieva,Heber, & Butler 2008). Another possibility is to identify run-away stars inside SN remnants (SNRs) or in the vicinity ofneutron stars (e.g. Dinçel et al. 2015). SNRs are only visiblefor up to ∼ ∼
820 pc for a B9.5 star travelling with ∼
25 km s −1 ). OB runaway stars were searched for both out-side of SNRs (e.g. Hoogerwerf et al. 2000, 2001; Tetzlaff et al. 2010; Tetzlaff, Eisenbeiss, Neuhäuser, & Hohle 2011; Tetzlaff2013; Tetzlaff, Dinçel, Neuhäuser, & Kovtyukh 2014), andinside SNRs (e.g. Dinçel et al. 2015; Boubert, Fraser, Evans,Green, & Izzard 2017, hereafter Bo17).Dinçel et al. (2015) found the B0.5 V runaway starHD 37424 ( 𝑣 pec = 74 ± 8 km s −1 ) in the SNR S147 andPSR J0538+2817 to have been close to the geometric centre ofthe SNR
30 ± 4 kyr ago, confirming that the BES does happen.Bo17 confirmed HD 37424 and suggested three further likelycandidates, located in the SNRs Cygnus Loop, HB21 andMonoceros Loop. They did not exclude late-type runaways,but they considered early-type stars to be more probable.For our search in nearby SNRs, we know that an ejectedcompanion has to be young. A SN progenitor has a MS life-time of up to 32 Myr (Ekström et al. 2012, Table 3, in the caseof an 𝑀 ⊙ star). The timespans of the subsequent burningcycles and the lifetime of the SNR ( ∼ Gaia mission (Gaia Collaboration, Prusti, etal., 2016). In Section 2 we explain how the SNRs and corre-sponding runaway star candidates were selected. In Section 3we describe the observations. In Section 4 we show thespectral analysis and estimate the ages of our candidates. InSection 5 we discuss the results and possible implications forthe frequency of runaway stars from the BES and, finally, inSection 6 we conclude our work and give an outlook on futurework.
Two main resources were used to find suitable SNRs, namely(i) the Green catalogue of SNRs (Green, 2014) for posi-tions and sizes and (ii) the catalogue from the University ofManitoba (Ferrand & Safi-Harb 2012 and references therein), UX ET AL . compiled mainly from high-energy observations, for distancesand ages.The project was started in 2016 and the first target selectionwas done before Gaia data release (DR) 2 (Gaia Collaborationet al., 2018) was available. Therefore,
Gaia
DR1 (Gaia Col-laboration, Brown, et al., 2016) was used for these targets. Themuch larger number of stars in
Gaia
DR2 had consequencesfor the SNR selection, which are described in the following.For observations of late-type stars, it was necessary to seta magnitude limit and, hence, a distance limit for the SNRs.While the limiting magnitude of a spectrograph at an 8 m-classtelescope is 𝑚 𝑉 , max ≈ 19 . mag (see UVES/VLT manual), forour selection we use 𝑚 𝐺 ≤ . mag in order to reach a suf-ficient signal-to-noise ratio, 𝑆 ∕ 𝑁 ≳ , in the spectra (seeSection 3). Using this limit, mass-brightness relations for themain-sequence from Henry & McCarthy (1993), and choos-ing a limiting distance of 𝑑 < pc, where four SNRs areavailable, we get a mass limit of 𝑀 = 0 . 𝑀 ⊙ , correspond-ing to spectral type K8. These limits were applied for starsselected from Gaia
DR2. For stars selected earlier from
Gaia
DR1,
Tycho-Gaia Astrometric Solution (TGAS), we chose 𝑑 < pc, including up to 25 SNRs. With this limit we getdown to 𝑀 = 0 . 𝑀 ⊙ , corresponding to spectral type K0.5.The limits were chosen as a compromise between investigat-ing a high number of SNRs and getting down to the latestpossible spectral types. Due to the higher number of stars in Gaia
DR2, we limited the distance to 𝑑 < pc in order tocreate feasible projects for follow-up observations.Analysing historical SNRs is particularly interesting,because their age is accurately known, so we can directly studythe SNR expansion and if we find a runaway star, the deter-mination of the distance and pre-SN binary properties can bedone much more accurately. Therefore, we also added fourhistorical SNRs to our sample, despite their larger distances.For the most distant one, Cassiopeia A at 𝑑 ≈ 3 . kpc, withthe limits described above we could detect stars down to spec-tral type ∼ F9. The limiting spectral types of the other SNRs,as well as further properties, are given in Table 1, where welist our selection of the four SNRs with 𝑑 ≤ pc, four fur-ther SNRs with pc ≤ 𝑑 ≤ pc and the four historicalSNRs Cassiopeia A, 3C58 (SN 1181), Crab Nebula (SN 1054)and SNR G347.3 − In order to define an area for the search for runaway star can-didates, it is important to know the position of the SN, whichshould be close to the geometric centre (GC) of the SNR.In most cases, we chose the coordinates from Green (2014),which are mainly based on radio observations, where the SNRmorphology usually is best visible. In the case of the Vela
FIGURE 1
The central ′ ×70 ′ of the Vela SNR. The crossesmark the three geometric centres from X-rays (magenta,Sushch et al. 2011), radio (blue, Bock et al. 1998) and theGreen catalogue (yellow, Green 2014). Their average andstandard deviation are shown by a red cross and ellipse,respectively. Background image from ESO DSS-2-red.SNR, we combined the centre from Green (2014) with twoimages from different wavelengths, namely the 843 MHz radioimage from Bock, Turtle, & Green (1998) and the X-ray imagefrom Sushch, Hnatyk, & Neronov (2011). The average coor-dinates of the three solutions were taken as the actual GC (seeFig. 1) .Around the centre, we defined a search radius, depending onthe maximum possible velocity of runaway stars and the ageof the SNR. Tauris (2015) gives a maximum ejection velocityof 𝑣 max = 1050 km s −1 for stars with . 𝑀 ⊙ , correspondingto extreme cases which are very rare (Rz19). We use 𝑣 max =1000 km s −1 , which yields a search radius of 𝑟 search = 3 . 𝑡 [yr] 𝑑 [pc] arcmin , (1)with the age 𝑡 and the distance 𝑑 of the SNR. As SNRexpansion velocities are well above typical runaway starvelocities during the Sedov-Taylor phase, which is the currentstate of most of our SNRs, we expect potential runaway starsto still be located inside the SNR.For each SNR, we first selected all
Gaia stars in the givensearch radius with 𝑚 𝐺 = 𝐺 ≤ mag. We then traced back the Note that in the case of the Vela SNR we did not trace back the stellartrajectories to the GC, but to the past position of the Vela pulsar.
LUX
ET AL . TABLE 1
Properties of the SNRs studied in this work. The equatorial coordinates RA/DEC and angular diameters Θ are takenfrom Green (2014), the distances 𝑑 and ages 𝑡 from Ferrand & Safi-Harb (2012) and references therein, where the errors comefrom the range between the lowest and highest values in the given literature. For 3C58, Crab Nebula and SNR G347.3 − lim that could be observed at the corresponding nominal distance and a limiting magnitude of 𝑚 𝐺 = 17 . mag, as descibed above.SNR Name Alternative Name RA [h:m:s] DEC [d:m] Θ [arcmin] 𝑑 [kpc] 𝑡 [kyr] SpT lim G074.0 − +
230 × 160 0 .
79 ± 0 .
21 15 ± 5
K5G160.9 + +
140 × 120 0 .
80 ± 0 .
40 5 . . K5G180.0 − +
180 × 180 1 .
30 ± 0 .
20 30 ± 4
K1.5G205.5 + +
220 × 220 1 .
44 ± 0 .
54 90 ± 60
K1.5G260.4 − −
60 × 50 1 . . .
08 ± 0 . K1.5G263.9 − −
255 × 255 0 .
275 ± 0 .
025 18 ± 9
M0.5G266.2 − −
120 × 120 0 .
75 ± 0 .
25 3 . . K5G330.0 + −
180 × 180 0 .
33 ± 0 .
18 23 ± 8
M0.5G111.7 − ∗ Cassiopeia A 23:23:26 + .
50 ± 0 .
20 0 .
334 ± 0 . F9G130.7 + ∗ + .
60 ± 0 . − ∗ Crab Nebula 05:34:31 + .
85 ± 0 . − ∗ SN 393 17:13:50 −
65 × 55 1 . . ∗ Historical SNR.projected trajectories of the stars, using their proper motionsand the age of the SNR to obtain their coordinates at the timeof the SN. We neglected the Galactic gravitational potentialbecause the lifetime of a SNR is so short that the potentialwill not have a significant effect. Gaussian error propagationwas used to calculate the uncertainties of the past positions.We selected the stars that, at the time of the SN, were locatedinside the error ellipse of the GC. For stars from
Gaia
DR1,we used the standard deviations in right ascension and declina-tion from the determination of the Vela GC to define the errorellipse, which was then also used for the other SNRs, scaled tothe corresponding diameter as given in Green (2014). For starsfrom
Gaia
DR2, we used the error estimate given by Green(2009), which we translate here to Δ GC = 0 . ° × 0 . ° , if ′ ≤ Θ ≤ ′ Δ GC = 0 . ° × 0 . ° , if ′ ≤ Θ ≤ ′ (2) Δ GC = 0 . ° × 0 . ° , if ′ ≤ Θ where Θ is the angular diameter of the SNR. However, if aneutron star is associated with the SNR, we selected only thestars that could be traced back to a possible common originwith the neutron star. This was the case for S147, Vela, 3C58and the Crab Nebula. Questionable cases or neutron stars withunknown proper motions were not considered.Note that the strict limitation of Eqn. 2 means that we looseup to ∼
33 % of runaway star candidates that were locatedmore than 𝜎 from the nominal GC. We decided to make this limitation in order to create feasible observing projects,concentrating on the most promising candidates.We also checked if the Gaia parallaxes of the stars are con-sistent with the range of possible distances of the SNR. Wecompared these to distances derived by Bailer-Jones, Rybizki,Fouesneau, Mantelet, & Andrae (2018) and found no signifi-cant deviations for the parallax range considered in our work.All candidates are still within the correct range when usingtheir distances from Bailer-Jones et al. (2018). For stars from
Gaia
DR2, we also checked the given values for radius, lumi-nosity, effective temperature and surface gravity, so that giantscould be excluded beforehand. The selection could be madeby plotting the candidates into a Hertzsprung-Russell diagram(HRD), shown in Fig. 2, where we compare our candidates toPARSEC isochrones (Bressan et al., 2012). Reddish coloursindicate if a star was classified as a giant. If a star has anincreased luminosity compared to the zero-age main-sequence(ZAMS), it can be either a pre-MS star or an evolved star(i.e. terminal-age MS or giant). As we are looking for youngstars, it might be better to leave a few evolved stars in thesample than excluding too many pre-MS stars. The divisionwas roughly made at the isochrone for 1 Myr. For questionablecases and because luminosities were not given for all possi-ble targets in Gaia
DR2, we also checked the surface gravities log( 𝑔 ) of the targets in the StarHorse catalogue (Anders et al., PAdova and TRieste Stellar Evolution Code, see http://stev.oapd.inaf.it/cgi-bin/cmd_3.3 UX ET AL . FIGURE 2
HRD of the runaway star candidates with effective temperatures and luminosities taken from
Gaia
DR2. In the key,for stars that were observed by us, we give the names of the corresponding spectrographs, UVES and HDS. For comparison,we show isochrones calculated from PARSEC models (Bressan et al., 2012). The stars marked in red, dark-red and orange wereconsidered to be post-MS giants and ruled out from the investigation.2019). Four targets with log( 𝑔 ) < , where no luminosity wasgiven, were excluded.Stars that fulfill all criteria were considered as good run-away star candidates and selected as targets for the spectro-scopic follow-up observations. In Fig. 3 the selection is shownfor the Monoceros Loop. Among the 21 stars found here from Gaia
DR1, we only show the remaining runaway star candi-dates and two stars that show Li in their spectra. No neutronstar is known in the Monoceros Loop, so the GC was used asa reference position.After the detection and characterisation of runaway starcandidates, the goal was to determine the radial velocity (RV)and the atmospheric parameters and to search for the Li6708 Å line as youth indicator in the spectra of late-type stars.
For southern targets, we have used the
UV-Visual EchelleSpectrograph (UVES), which is mounted at the
Nasmyth focusof
Kueyen (UT2) at the
Very Large Telescope (VLT), run bythe
European Southern Observatory (ESO) and located on the
Cerro Paranal in Chile. We observed 33 runaway starcandidates, selected from
Gaia
DR1 TGAS, in five SNRson the southern sky, namely (i) G180.0 − − − − Å and
Å. Using binning,which is adequate for the relatively bright targets observedin P100, a slit width of . ′′ yields a resolving power of 𝑅 = 32250 .For the used magnitude limit of 𝐺 = 17 mag, correspondingto 𝑉 = 17 . mag , with the given setup we reach 𝑆 ∕ 𝑁 = 33 at 6711 Å in one hour integration time, derived with the UVESexposure time calculator (ETC), version P106.2 , for a G0 Vstar at airmass 1.1, with a seeing of ′′ and with binning. For the calculation, 𝐺 = 17 . mag was converted to 𝑉 = 17 . mag byusing the polynomial relations by Evans et al. (2018) with 𝐵𝑃 − 𝑅𝑃 = 0 . givenin Pecaut & Mamajek (2013, Table 5) for a G0 V star. = UVES + INS.MODE = spectro LUX
ET AL . FIGURE 3
The runaway star candidates in the MonocerosLoop. The image shows the central ′ × 70 ′ of the SNR. Thegeometric centre of the SNR is marked by a red cross with cor-responding 𝜎 error ellipse. The current positions of runawaystar candidates with and without detected lithium are markedby yellow and blue crosses, respectively. They are connectedwith the positions at the time of the SN (surrounded by a cor-responding error ellipse) by arrows in the direction of theirproper motion. The star marked in black is HD 261393, whichwas suggested as a runaway star by Bo17 and also observedby us. The background image was taken by us in H 𝛼 withthe Schmidt-Teleskop-Kamera (Mugrauer & Berthold, 2010)at the University Observatory Jena.A G0 V star was chosen for this example calculation becauseit represents a typical target of our selection.The VLT observations were executed between October andDecember 2017 in service mode. In the appendix, Table A1 ,we list the 33 runaway star candidates observed with UVES.The individual exposure times varied between 10 s and 500 s,depending on the brightness of the star. Two exposures weretaken for each star, where an average 𝑆 ∕ 𝑁 of 64.7 wasachieved for the single exposures. The total on-source integra-tion time was 5 hours and 4 minutes.Standard calibrations were used, i.e. Bias frames, Flatfieldimages of a Halogen lamp and wavelength calibration imagesof a Thorium-Argon (ThAr) lamp which are taken regularlyfor each standard setup as described in the UVES calibrationplan. Data reduction was done with the EsoReflex pipeline forUVES. The individual reduction steps are bias subtraction, order detection, flatfielding, wavelength calibration and spec-trum extraction. We obtained four spectra for each star; foreach of the two exposures, we got one spectrum for the lowerand one for the upper wavelength regime.The two exposures for each star were averaged to onespectrum with
IRAF . Normalisation was done with iSpec (Blanco-Cuaresma, Soubiran, Heiter, & Jofré, 2014).
Six northern targets, all located in SNR G160.9+02.6 (HB9),were observed with the High Dispersion Spectrograph (HDS),mounted at the
Nasmyth focus of the Subaru telescope, whichis run by the
National Astronomical Observatory of Japan (NAOJ) and located on the
Mauna Kea , Hawaii.HDS uses two CCD detectors, where with the Ra setup thefirst one covers the wavelength range
Å and thesecond one
Å. We used binning and a slitwidth of . ′′ , yielding a resolving power of 𝑅 = 36000 .For this setup, with the HDS ETC , we obtain 𝑆 ∕ 𝑁 ≈ 24 at 6697 Å for a G0 V star with 𝑉 = 17 . mag at airmass 1.1,with a seeing of ′′ and an exposure time of one hour.At the bottom of Table A1 we list the six runaway starcandidates that were observed with Subaru in period S18B.All observations were done on 2018 Oct 26. The individualexposure times varied between 40 s and 300 s. Two exposureswere taken for each target and 𝑆 ∕ 𝑁 = 105 . was reached onaverage for the single exposures.Standard calibrations were used and data reduction wasdone manually with IRAF . We corrected for the bias leveland bad pixels and applied order tracing, flatfielding, spectrumextraction and wavelength calibration, following the proce-dure described in Aoki & Helminiak (2014). As described forUVES/VLT, also here we obtained four spectra for each star.The last step of the data reduction already included the nor-malisation of the spectra by fitting a spline to the continuum,before the spectral orders were merged.
In the averaged spectra, we searched for the Li 6708 Å absorp-tion line to measure the equivalent width (see Section 4.4).Li was found in ten of our VLT spectra. Four of these starswere found to be giants (see Fig. 2) and are therefore tooold to be runaway stars. Fig. 4 shows seven of our VLTspectra around the Li line. While TYC159-1896-1 is onlyshown for comparison, six of them show Li absorption in theirspectra, namely (1) TYC159-251-1 and (2) TYC159-343-1,located in the Monoceros Loop, (3) TYC8150-2802-1 and UX ET AL . FIGURE 4
Normalised, RV corrected and shifted UVES/VLT spectra between ∼ iSpec (Blanco-Cuaresma et al., 2014) todetermine their RVs and atmospheric parameters. The spectrum of TYC159-251-1 shows additional absorp-tion lines, each redshifted by ∼ . The continuum levels of the tem-plates were normalised before to have the flux level of thecombined spectrum of TYC159-251-1, multiplied by the fac-tors 0.661 and 0.339, respectively, for the two components,corresponding to the luminosities of the used spectral typesaccording to Pecaut & Mamajek (2013, Table 5, hereafterPM13). Adding up these modified templates yielded the bestrepresentation of the observed spectrum. Fig. 5 shows theobserved spectrum (blue) as well as the individual components(magenta, green) after subtracting the template spectra (black,yellow). The residuals (light-blue) between the observed spec-trum and the added templates (red) only show a significantdeviation from zero at the Li lines, which is due to the manualremoval of Li in the templates (to retain it in the componentspectra). In the following subsections, potential peculiaritiesin the analysis of this SB2 will be described in additionalparagraphs. The spectral types F6 and G0 were found to yield the best representationof the observed spectrum after iterating with different combinations of templatespectra from Bagnulo et al. (2003) between F6 and G0 (F6+F9, F6+G0, F7+F9,F8+F8, F8+G0 and F9+F9. Other combinations were excluded beforehand).
LUX
ET AL . FIGURE 5
The spectrum of TYC159-251-1 (blue, 1) and its two components TYC159-251-1 A (purple, 1A) and TYC159-251-1 B (green, 1B) between ∼ The barycentric correction was done with iSpec before com-paring the spectral lines to an atomic line mask (cross-correlation) to determine the RVs and shift the spectra cor-respondingly. For the cross-correlation, a line list from a
Narval solar spectrum (
370 − 1048 nm) was used. For theearly-type stars, the RV determination with iSpec was notpossible due to the fewer and broader lines. We determinedtheir RVs manually with
IRAF from the positions of thefollowing lines, if available: Fe II 5018.4 Å, Mg I 5172.7 Å,Mg I 5183.6 Å, He I 5875.6 Å, Si II 6347.1 Å, Si II 6371.4 Å,Fe II 6456.4 Å, H 𝛼 𝑣 𝑟 for each line were calculated.The 𝑣 𝑟 in Table 2 correspond to average and standard devi-ation of these measurements. By combining the RVs withproper motions and parallaxes from Gaia
DR2, we calcu-lated the heliocentric space velocities of the stars. For thecalculation of the peculiar velocity with respect to the localstandard of rest, we first calculated the Galactic space veloc-ity components ( 𝑈 , 𝑉 , 𝑊 ) , following the equations given inJohnson & Soderblom (1987). Then we corrected for the solarmotion with respect to the local standard of rest by adding ( 𝑈 ⊙ , 𝑉 ⊙ , 𝑊 ⊙ ) = (8 . , . , . km s −1 (Cos , kunoˇglu et al.,2011). The peculiar velocity is the absolute value of the result-ing vector. The kinematic data of the six Li-rich stars and thebest early-type candidates are given in Table 2. TYC8150-2802-1, TYC8152-1456-1, TYC8152-550-1, TYC3344-235-1, TYC159-2771-1 and HD 76060 have 𝑣 pec > km s −1 andcould therefore be classified as runaway stars according toTetzlaff (2013).For the SB2 system TYC159-251-1, we determined themomentary RVs of the two components with IRAF , measuringthe positions of 38 absorption lines of Fe I, Fe II, Ca I, Ni I andSi II. As line list we chose the
Identification List of Lines inStellar Spectra (ILLSS, Coluzzi 1999). To obtain the systemicRV 𝛾 , we need the mass ratio 𝑞 = 𝑀 ∕ 𝑀 < . In Section 4.3we will find the components to have spectral types F6 − F7and F9 − G0, respectively, corresponding to 𝑞 = 0 . +0 . . . Thesystemic RV can be calculated as 𝛾 = 𝑣 𝑟 + 𝑞 𝑞 × ( 𝑣 𝑟 − 𝑣 𝑟 ) (3) UX ET AL . The following results were obtained: 𝑣 𝑟 = −10 . . km s −1 𝑣 𝑟 = 48 . . km s −1 𝛾 = 17 . +0 . . km s −1 (4)The systemic RV is then used to determine the space motion(see 𝑣 pec in Table 2). Extensive follow-up RV monitoring isrequired to fit a RV curve and determine the orbital parame-ters. Only then the systemic and space velocity of the systemcan be given reliably. With the synthetic spectral fitting technique provided by iSpec ,we determined the atmospheric parameters of the six Li-richstars. The routine computes synthetic spectra and comparesthem to the observed spectra by nonlinear least-squares fit-ting, minimising the 𝜒 . We used the continuum-normalisedand RV-corrected spectra. The routine considers the follow-ing parameters: (i) Effective temperature 𝑇 eff , (ii) Surfacegravity log( 𝑔 ) , (iii) Metallicity [M/H], (iv) Microturbulencevelocity 𝑣 mic 𝑡 , (v) Macroturbulence velocity 𝑣 mac 𝑡 , (vi) Rota-tion 𝑣 rot sin( 𝑖 ) and (vii) Limb darkening coefficient, taking intoaccount the resolving power 𝑅 = 32250 .To create the synthetic spectra, we used the radiative trans-fer code SPECTRUM with the
ATLAS9.Kurucz model atmo-sphere. Solar abundances were taken from
Kurucz and as linelist we chose
GESv5_atom_hfs_iso.420_92 . For the fitting, wechose the red part of the spectrum (
Å), becauseit yields a higher 𝑆 ∕ 𝑁 . We chose a large number (> 100) oftemperature-sensitive Fe I, Fe II, Ca I, Ca II, Ni I, Si I and Si IIlines as well as the wings of the H 𝛼 absorption line, which hasa big effect on the results because it spans a large wavelengthrange. After the automatic selection of the lines it had to bechecked if they were suitable for fitting, i.e. they should reachback to the continuum within the segments that were drawnaround each line. In some cases the segments were modifiedin order to reach that. In Fig. 6, we show examples of selectedlines for TYC8152-550-1. The synthetic spectra are created inthe segments (grey) while the differences to the observed spec-trum are only computed within the line regions (yellow). If nocontinuum regions were available around a line, or if strongdeviations between spectrum and fit were recognised withinthe line mask, the corresponding line was rejected.As initial parameters for the fits, we used solar parametersas given in Blanco-Cuaresma (2019) and the 𝑇 eff of each starfrom Gaia
DR2. In each run, consisting of up to twelve iter-ations, it could be decided about which parameters should befixed or free. As all parameters were unknown, we left all ofthem free, except either macroturbulence or 𝑣 rot sin( 𝑖 ) , which are strongly correlated with each other. By running the fit sev-eral times with different combinations, we guaranteed that thefixed parameter among 𝑣 mac 𝑡 and 𝑣 rot sin( 𝑖 ) was set to a reason-able value, while the other was left free. The limb darkeningcoefficient was fixed to 0.6, because changing it did not showany effect.The resulting atmospheric parameters of the six Li-richstars are displayed in Table 3. The given intrinsic errors arecomputed from the covariance matrix which connects theerrors of the individual parameters after the least-squares fit-ting (Blanco-Cuaresma et al., 2014). The flux errors for ourspectra were calculated from gain and read noise of UVESwith the given setup. For comparison, we also show 𝑇 eff from Gaia
DR2. We give the results for both components ofTYC159-251-1, where the fits were done individually afterdisentangling the spectra. Note that the disentangled spectrahave lower 𝑆 ∕ 𝑁 . The resulting 𝑇 eff for TYC159-251-1 B cor-responds to SpT F9, but is also consistent with G0, which wasused to disentangle the original spectrum. The derived 𝑇 eff forTYC159-251-1 A is fully consistent with the assumption ofan F6 template spectrum. 𝑣 mac 𝑡 = 0 km s −1 and 𝑣 rot sin( 𝑖 ) =0 km s −1 were obtained for both components in all iterations,even when one or both of these parameters were left free. The Li equivalent widths 𝐸𝑊 Li of the six Li-rich stars weremeasured with IRAF by integrating over the area of the Li6708 Å line with respect to the local continuum. The errorswere calculated by adding the uncertainty in fitting the con-tinuum and the error related to read-noise ( . e − ) and gain( . e − ∕ ADU) of the instrument. The continuum error wasderived by varying the flux level of the local continuum for theintegration.If we do not disentangle the spectrum of a spectroscopicbinary, the equivalent widths will be underestimated due tothe additional continuum flux from the other component.Therefore, we measured the 𝐸𝑊 Li for both components ofTYC159-251-1 individually after disentangling. The absorp-tion line is even stronger for the secondary component. Asthe F6 and G0 template spectra also show Li absorption, weremoved the lines manually with IRAF before subtracting thetemplates from the original spectrum. So it was guaranteedthat the Li lines of both components remained in the disentan-gled spectra, where we could then measure the corresponding 𝐸𝑊 Li .The 𝐸𝑊 Li were converted to abundances by using thecurves of growth given by Soderblom et al. (1993, Table2, hereafter So93). The abundance scale is based on log( 𝑁 H ) = 12 . The curves of growth are calculated for LUX
ET AL . TABLE 2
Gaia
DR2 𝐺 magnitudes, parallaxes 𝜋 and kinematic parameters of the six Li-rich stars as well as the remainingearly-type candidates in HB9, Monoceros Loop and Vela Jr. Parallaxes and proper motions are from Gaia
DR2, the barycentriccorrections ( 𝐵𝐶 ) were determined with iSpec . The radial velocities 𝑣 𝑟 of the Li-rich stars were determined with iSpec , the 𝑣 𝑟 of the early-type stars were determined with IRAF . For TYC159-251-1, we give the systemic RV 𝛾 of the binary, approximatedusing a mass ratio of 𝑞 = 0 . +0 . . . The individual RVs of the binary components were derived with IRAF . For the spectraltypes, see Table 4.Name
𝐺 𝜋 𝜇 𝑅𝐴 𝜇 𝐷𝐸𝐶
𝐵𝐶 𝑣 𝑟 𝑣 pec [mag] [mas] [mas yr −1 ] [mas yr −1 ] [km s −1 ] [km s −1 ] [km s −1 ]Li-rich starsTYC159-251-1 11.51 .
774 ± 0 .
041 +3 .
63 ± 0 .
07 −9 .
44 ± 0 . +17 . +0 . . . +1 . . TYC159-343-1 10.87 .
383 ± 0 .
042 +2 .
51 ± 0 .
09 −2 .
76 ± 0 . +22 .
27 ± 0 .
41 13 . . TYC8150-2802-1 12.03 .
226 ± 0 .
028 −16 .
94 ± 0 .
05 +29 .
17 ± 0 . +53 .
84 ± 0 .
07 58 . . TYC8150-3105-1 11.94 .
351 ± 0 .
028 −16 .
00 ± 0 .
05 +10 .
05 ± 0 . +20 .
37 ± 0 .
29 19 . . TYC8152-1456-1 12.47 .
661 ± 0 .
037 +5 .
17 ± 0 .
06 −11 .
56 ± 0 . −10 .
59 ± 0 .
08 49 . . TYC8152-550-1 11.92 .
435 ± 0 .
026 +4 .
95 ± 0 .
04 −10 .
71 ± 0 . +36 .
12 ± 0 .
10 37 . . HB9TYC3344-235-1 11.07 .
91 ± 0 .
04 −0 .
23 ± 0 .
09 +1 .
37 ± 0 . −31 . . . . TYC3344-679-1 11.40 .
01 ± 0 .
06 +0 .
74 ± 0 .
18 −4 .
76 ± 0 . +8 . . . . TYC3344-683-1 12.33 .
21 ± 0 .
04 −1 .
09 ± 0 .
07 −7 .
02 ± 0 . +9 . . . . TYC3344-553-1 10.87 .
53 ± 0 .
04 −1 .
98 ± 0 .
08 −7 .
22 ± 0 . −14 . . . . Monoceros LoopTYC159-2771-1 12.01 .
75 ± 0 .
04 +1 .
46 ± 0 .
07 −3 .
78 ± 0 . +31 . . . . HD261359 11.79 .
65 ± 0 .
05 −0 .
48 ± 0 .
08 −3 .
15 ± 0 . +0 . . . . HD261393 10.05 .
79 ± 0 .
05 −0 .
20 ± 0 .
09 −0 .
97 ± 0 . +31 . . . . TYC159-2337-1 11.80 .
49 ± 0 .
05 −1 .
70 ± 0 .
09 −0 .
45 ± 0 . +22 . . . . TYC159-2671-1 12.25 .
69 ± 0 .
04 +0 .
08 ± 0 .
07 −0 .
78 ± 0 . +23 . . . . Vela Jr.HD 76060 7.85 .
06 ± 0 .
04 −12 .
42 ± 0 .
08 +10 .
31 ± 0 . +25 . . . . FIGURE 6
Different example lines of TYC8152-550-1 (blue) that were used to fit the atmospheric parameters (red). Theyellow areas mark the line regions used for the fit, the grey areas show segments of 0.5 Å on each side of each line used forsynthesising the spectra. In particular we show Fe I 6065, Fe II 6084, Ni I 6086, Fe I 6344, Si II 6347, Ca I 6472 and Si I 6527 Å. UX ET AL . TABLE 3
The atmospheric parameters of the six Li-rich stars determined with iSpec . The effective temperatures from
Gaia
DR2 are also shown for comparison. A missing error indicates that the parameter was kept constant for all iterations. The resultsfor TYC159-251-1 are given for both components individually.Name 𝑇 eff ,𝐺𝑎𝑖𝑎 𝑇 eff log( 𝑔 ∕ cm s −2 ) [M/H] 𝑣 mic 𝑡 𝑣 mac 𝑡 𝑣 rot sin( 𝑖 ) [K] [K] [km s −1 ] [km s −1 ] [km s −1 ]TYC159-251-1 A ∗ .
16 ± 0 .
22 −0 .
19 ± 0 .
06 1 .
49 ± 0 . ∗ .
94 ± 0 .
27 −0 .
16 ± 0 .
07 1 .
86 ± 0 . .
77 ± 0 .
35 −0 .
57 ± 0 .
08 4 . . . . .
99 ± 0 .
20 −0 .
08 ± 0 .
06 1 .
28 ± 0 . .
78 ± 0 .
21 −0 .
19 ± 0 .
09 1 . . . . . TYC8152-1456-1 .
14 ± 0 .
23 −0 .
09 ± 0 .
07 1 .
26 ± 0 .
26 0 . . .
26 ± 0 .
24 −0 .
27 ± 0 .
06 1 .
50 ± 0 . ∗ Gaia
DR2 value for the unresolved system 𝑇 eff = − log( 𝑁 Li ) by choosingthe values from So93 that best represent the given temper-ature and logarithmic equivalent width (for 𝐸𝑊 Li in mÅ)of the Li-rich stars, taking into account the errors. With 𝑇 eff = 6749 ± 69 K, TYC159-343-1 is not covered by the 𝑇 eff range in So93. Therefore, here we extrapolated the givenabundances with a quadratic function and determined theLi abundance for 𝑇 eff = 6750 K and log( 𝐸𝑊 Li ) = 1 . , log( 𝐸𝑊 Li ) = 1 . and log( 𝐸𝑊 Li ) = 1 . . The resultingabundance error, Δ log( 𝑁 Li ) = +0 . . , contains the error fromvarying 𝐸𝑊 Li ( Δ var = +0 . . ) and the root mean square (RMS)of the fit ( Δ RMS = 0 . in all three cases). For the other Li-rich stars, the errors only come from the variation of 𝐸𝑊 Li and/or 𝑇 eff , according to the errors. The Li equivalent widthsand abundances, together with temperatures, spectral typesand magnitudes are listed in Table 4. The abundances can becompared to typical initial values, which are expected to be log( 𝑁 Li ) = 3 . . for population I stars (Sestito & Randich,2005).From D’Antona & Mazzitelli (1984, Table 7) it can beseen that there is no significant Li depletion in stars with 𝑀 ≳ . 𝑀 ⊙ if no additional mixing mechanisms occur. Thiscorresponds to spectral type F7 (PM13). Therefore, the Lisignal detected in TYC159-343-1 might not be conclusive toestimate its age and also TYC159-251-1 and TYC8152-550-1 have to be taken with care. Nevertheless, we will make anattempt to obtain rough age estimates for the Li-rich stars inthe next subsection. In the following, we will compare two methods to estimatethe ages of our targets. Firstly, we will use the positions of thestars in the HRD, where we can compare them to isochrones (Figs. 7 and 8). Then we perform the Li test by using the deter-mined 𝑇 eff and 𝐸𝑊 Li (or log( 𝑁 Li ) ) to compare them to starsof clusters with known ages (Figs. 9 and 10). In the HRD (Fig. 7, for a zoom-in of the Li-rich stars seeFig. 8), we show the stars observed with UVES/VLT andHDS/Subaru, using 𝑇 eff either from Gaia
DR2, from theirspectral type as given in the
Skiff catalogue (Skiff, 2009),from Bai et al. (2019) or from our parameter fits. Using
Gaia
DR2 parallaxes and
StarHorse extinctions (Anders etal., 2019), we converted the apparent magnitudes 𝐺 to abso-lute magnitudes 𝑀 𝐺 . For the 𝐺 -band extinctions, no errors aregiven in the StarHorse catalogue. We converted the 𝑉 bandextinction errors to 𝐺 with the relation Δ 𝐴 𝐺 = 0 . 𝐴 𝑉 (Mugrauer, 2019). Δ 𝑀 𝐺 was then determined with Gaussianerror propagation.We compare our stars to isochrones which were calculatedfrom PARSEC models (Bressan et al., 2012), reflecting thewhole stellar evolution until the first thermal pulse. We usedsolar metallicity, 𝑍 = 0 . , which corresponds to [ M/H ] =0 . . This is close to [ M/H ] = −0 .
20 ± 0 . , the averagemetallicity of our targets as given in the StarHorse catalogue.The corresponding shift of isochrone positions is rather smallcompared to the temperature and magnitude errors for early-and late-type stars, respectively. However, TYC159-343-1 hasa significantly lower metallicity of [ M/H ] = −0 .
57 ± 0 . .Therefore, for this target we plotted seperate isochrones todetermine its age more precisely (see Fig. 8 and Table 5).We directly see that the red giants are located on thegiant branch of the isochrones in Fig. 7, consistent with agesbetween 0.2 Gyr and >
10 Gyr. The Li-rich stars are mostlylocated relatively close to the ZAMS, only TYC159-343-1is located close to the isochrone for 5 Myr due to its highbrightness. LUX
ET AL . TABLE 4
Temperatures, spectral types and magnitudes of the six Li-rich stars and the remaining observed early-type runawaystar candidates. Effective Temperatures 𝑇 eff of the Li stars and the stars in HB9 are from our parameter fits and from Bai et al.(2019), respectively. The spectral types of the stars in Monoceros and Vela Jr. were taken from the Skiff catalogue (Vo85, Houk1978). Conversion between 𝑇 eff and SpT was done with PM13. G, BP and RP magnitudes are from Gaia
DR2, absolute 𝐺 magnitudes 𝑀 𝐺 were calculated from Gaia G , extinction and parallax and 𝐺 band extinctions are from the StarHorse catalogue(Anders et al., 2019). The last two columns show Li equivalent widths 𝐸𝑊 Li and abundances log( 𝑁 Li ) for the six Li-richstars. The abundance errors correspond to the minimum and maximum possible abundance from So93, taking into account theuncertainties in 𝐸𝑊 Li and 𝑇 eff . For TYC159-251-1, we give the values for the unresolved system as well as for the individualcomponents, assuming that TYC159-251-1 A contributes 66.1 % of the flux and TYC159-251-1 B 33.9 %.Name 𝑇 eff SpT
𝐺 𝐵𝑃 𝑅𝑃 𝑀 𝐺 𝐴 𝐺 𝐸𝑊 Li log( 𝑁 Li ) [K] [mag] [mag] [mag] [mag] [mag] [mÅ]Li-rich starsTYC159-251-1 F5–7 11.51 11.78 11.09 +2 . +0 . . . +0 . .
17 ± 2 2 . +0 . . TYC159-251-1 A
F6–7 11.96 12.23 11.54 +3 . +0 . . . +0 . .
22 ± 3 2 . +0 . . TYC159-251-1 B
F9–G0 12.68 12.96 12.26 +3 . +0 . . . +0 . .
97 ± 7 2 . +0−0 . TYC159-343-1
F2–4 10.87 11.08 10.54 +1 . +0 . . . +0 . .
35 ± 5 2 . +0 . . TYC8150-2802-1
G0–G2 12.03 12.36 11.55 +4 . +0 . . . +0 . .
50 ± 3 2 . +0 . . TYC8150-3105-1
F9.5–G1 11.94 12.26 11.42 +4 . +0 . . . +0 . .
177 ± 7 3 .
37 ± 0
TYC8152-1456-1
F9.5–G1 12.47 12.81 11.97 +3 . +0 . . . +0 . .
66 ± 4 2 . +0 . TYC8152-550-1
F7–F9 11.92 12.21 11.48 +3 . +0 . . . +0 . .
33 ± 4 2 . +0 . . HB9TYC3344-235-1
A5–F0 11.07 11.38 10.61 −0 . +0 . . . +0 . . – –TYC3344-679-1 A6–F0 11.40 11.72 10.92 −0 . +0 . . . +0 . . – –TYC3344-683-1 A7–F0 12.33 12.58 11.94 +2 . +0 . . . +0 . . – –TYC3344-553-1 A7–A9 10.87 11.08 10.53 +1 . +0 . . . +0 . . – –Monoceros LoopTYC159-2771-1 +2550−1250 B8–A0 12.01 12.13 11.78 +1 . +0 . . . +0 . . – –HD 261359 +2550−1250 B8–A0 11.79 11.87 11.62 +0 . +0 . . −0 . +0 . . – –HD 261393 +1150−1450 B4–B6 10.05 10.07 10.02 −0 . +0 . . . +0 . . – –TYC159-2337-1 +2550−1250 B8–A0 11.80 11.88 11.61 −0 . +0 . . . +0 . . – –TYC159-2671-1 +1750−1950 B7–B9 12.25 12.31 12.08 +1 . +0 . . . +0 . . – –Vela Jr.HD 76060 +1750−950 B7–B9 7.85 7.83 7.92 −0 . +0 . . . +0 . . – –TYC159-251-1 was discovered by us to be a SB2. Pho-tometrically unresolved binarity changes the positions in theHRD. The brightness is then overestimated, so the star islocated significantly above the ZAMS. We correct for thiseffect by assuming that the primary component contributes66.1 % and the secondary 33.9 % of the flux. This adds0.45 mag and 1.17 mag, respectively, to their 𝑀 𝐺 magni-tude, shifting them back towards the ZAMS. Furthermore, therevised atmospheric parameter fits for the individual compo-nents yield lower effective temperatures. The two individualcomponents differ by 276 K and 0.72 mag.The positions of the early-type stars in the HRD have tobe taken with care because Gaia temperature estimates aretrained only up to 𝑇 eff = 10000 K and tend to be under-estimated when approaching this limit. Therefore, for the early-type stars observed with UVES, we adopted the spec-tral types from the
Skiff catalogue (Skiff 2009). Only onereference is given for each star (Houk 1978 for HD76060,Voroshilov et al. 1985, hereafter Vo85, for the ten other early-type stars) and no error ranges are given there, so we assumethem to be ±1 subclass. The corresponding temperatures wereinferred from PM13. Although these stars were also observedby us, the spectra were not suited for determining the spec-tral type. Our spectral range of Å is quite narrowand misses most spectral lines that are usually taken for theanalysis of early-type stars. Additionally, the few lines wefound are often affected by strong rotational broadening. Nev-ertheless, we checked the
Skiff spectral types for consistencyby comparing our spectra to standard star spectra observedwith the
Fibre Linked ECHelle Astronomical Spectrograph UX ET AL . FIGURE 7
HRD with absolute G -band magnitudes of stars observed with UVES/VLT and HDS/Subaru. For an age estimate,we show PARSEC isochrones (Bressan et al., 2012) with metallicity 𝑍 = 0 . , corresponding to solar metallicity. For theearly-type stars (blue), we used spectral types from the Skiff catalogue (Skiff 2009 and references therein) and corresponding 𝑇 eff ranges from PM13. The remaining candidates in the Monoceros Loop (M) and Vela Jr. (VJ) are marked by grey squares andlabeled as (M1) TYC159-2771-1, (M2) HD 261359, (M3) HD 261393, (M4) TYC159-2337-1, (M5) TYC159-2671-1, (VJ1)HD 76060. For stars located in HB9 (cyan), labeled as (H1) TYC3344-235-1, (H2) TYC3344-679-1, (H3) TYC3344-683-1 and(H4) TYC3344-553-1, we checked the 𝑇 eff from Bai et al. (2019), which are used here for the stars with spectral type late-A.For stars with 𝑇 eff ≲ K (SpT FG) and no Li (purple), we use 𝑇 eff from Gaia
DR2. For the Li-rich stars, marked withcoloured squares, we show the 𝑇 eff from our parameter fits. They are labeled as (1) TYC159-251-1, (2) TYC159-343-1, (3)TYC8150-2802-1, (4) TYC8150-3105-1, (5) TYC8152-1456-1 and (6) TYC8152-550-1. TYC159-251-1 is a binary and theHRD positions of the individual components correspond to the black triangles.(FLECHAS, Mugrauer, Avila, & Guirao 2014), operated atthe 0.9 m telescope of the University Observatory Jena. Wemainly used the He I line at 5015.7 Å (visible up to B9) andthe Fe lines at . . Å (visible from A0 on). Ourresults are largely consistent with the spectral types given byVo85, with deviations of up to four subclasses.For the stars observed with HDS/Subaru, from compari-son with the spectra we found that the 𝑇 eff in Gaia
DR2are underestimated. For these targets, which are not listed inthe
Skiff catalogue, instead we used the values given in Baiet al. (2019), which fit best to our spectra. Therefore, theyprobably have spectral types A5 − F0. However, their absolute magnitudes differ a lot from each other, indicating differentages.HD 261393, suggested by Bo17 as the best candidate in theMonoceros Loop, has 𝑇 eff = 14250 − 16850 K (Vo85) and 𝑀 𝐺 ≈ −0 . mag. Therefore, it is probably an evolved butstill young star of ≲
105 Myr, consistent with being the ejectedcompanion of a SN-progenitor ( 𝑡 ≲ Myr). LUX
ET AL . FIGURE 8
Left: Zoom-in of Fig. 7, centered on the Li-rich stars. Right: HRD with absolute G -band magnitudes of TYC159-343-1. For an age estimate, we show PARSEC isochrones (Bressan et al., 2012) with metallicity [M/H] = −0 . . This valueand the 𝑇 eff were determined from our parameter fits. The axis scales are similar in both panels. For the key, see Fig. 7. FIGURE 9 𝑇 eff versus 𝐸𝑊 Li for the six Li-rich stars,shown with coloured error boxes and labeled as (1) A TYC159-251-1 A, (1) B ∼ emamajek/images/li.jpg. Fig. 9 shows 𝑇 eff and 𝐸𝑊 Li for stars in several open clus-ters of different ages . Eric Mamajek fitted polynomials to thedata of cluster stars, allowing a rough localisation of ages inthe 𝑇 eff − 𝐸𝑊 Li space. By comparing the values of the Li-rich stars, with 𝑇 eff from our parameter fits, to these curves,we obtained the age ranges displayed in Table 6. Furtheruncertainty comes from the unknown initial Li abundances ofour targets, depending on the metallicity (Lambert & Reddy,2004). If the initial abundance was lower than in the case ofthe cluster stars, less time would have been necessary to reachthe current abundance. Also, the data that were used to fit thepolynomials show a large scatter, e.g. due to different metal-licities and rotational velocities, as well as possible systematiceffects like unresolved binarity or starspots. Therefore, quan-tifying ages for individual stars is unreliable and the valuesgiven in Table 6 should only be seen as a rough estimate. We summarise the ages inferred from the two diagrams inTables 5 and 6. TYC8150-2802-1 and TYC8150-3105-1 areconsistent with the ZAMS in Fig. 7. For TYC8150-2802-1,the low Li content indicates that it is an evolved star, at least afew hundred Myr old. TYC8150-3105-1, however, shows verystrong Li absorption indicating that it is young. From combin-ing both results, we obtain an age of
Myr, so we suggest ∼ emamajek/images/li.jpg UX ET AL . TABLE 5
Estimated HRD ages (see Figs. 7 and 8) for the six Li-rich stars, including both components ofthe SB2 TYC159-251-1, as well as for the remainingobserved runaway star candidates in HB9, MonocerosLoop and Vela Jr. If a star is consistent with the zero-age main-sequence, only a lower limit is given incolumn 2. Otherwise, the values in column 2 representthe case that the star is on its pre-MS and the valuesin column 3 represent the case that the star is on itsterminal-age- or post-MS.Name 𝑡 HRD [Myr] 𝑡 HRD [Myr]if (pre-)MS if post-MSLi-rich starsTYC159-251-1 A 11 – 16 3200 – 4500TYC159-251-1 B 14 – 30 4800 – 7500TYC159-343-1* 2.1 – 3.8 1600 – 2300TYC8150-2802-1 > –TYC8150-3105-1 > –TYC8152-1456-1 8 – 16 5000 – 8000TYC8152-550-1 13 – 20 3800 – 6000HB9TYC3344-235-1 < .
380 – 650TYC3344-679-1 .
490 – 800TYC3344-683-1 > –TYC3344-553-1 3.7 – 7 850 – 1300Monoceros LoopTYC159-2771-1 > . –HD 261359 > . –HD 261393 < –TYC159-2337-1 1.6 – 3.3 50 – 490TYC159-2671-1 > –Vela Jr.HD 76060 1.1 – 1.8 90 – 350* HRD ages determined with isochrones for [ M/H ] = −0 . .that it is on its early MS. In the SB2 system TYC159-251-1,the primary is slightly above the ZAMS while the secondaryis consistent with it. Still, their ages are consistent with eachother. A young age of ∼
15 Myr can not be excluded fromthe HRD, but we consider it more likely that the binary is anevolved system having 𝑡 HRD ≈ 4 . . Gyr because their Licontent is too small for a young age.TYC159-343-1 is a difficult case: From the parameter fitswe obtained a very low metallicity of [M/H] = −0 . . , sowe inferred its HRD-age from comparison to low-metallicityisochrones (see Fig. 8). Its high brightness of 𝑀 𝐺 = 1 .
24 ±0 . mag indicates an age of only . . Myr, if it is pre-MS. However, it is more likely an evolved star with an age of
TABLE 6
Ages of the six Li-rich stars, including both com-ponents of the SB2 TYC159-251-1, estimated from Fig. 9. Name 𝑡 Li [Myr]TYC159-251-1 A > TYC159-251-1 B 90 – 4000TYC159-343-1 90 – 625TYC8150-2802-1 625 – 4000TYC8150-3105-1 5 – 50TYC8152-1456-1 250 – 4000TYC8152-550-1 > . . Gyr. The Li abundance, log( 𝑁 Li ) = 2 .
91 ± 0 . , isnot a good age tracer in this case, because of two caveats: (i)The low metallicity also means that the star probably had alow initial Li abundance, e.g. log( 𝑁 Li ) = 2 .
64 ± 0 . accord-ing to Lambert & Reddy (2004, Table 2), which is even belowthe measured abundance. A lower initial abundance reducesthe time which is necessary to reach the current abundance andtherefore the age. (ii) TYC159-343-1 has spectral type F2 − 𝑡 Li = 90 − 625 Myr. However, dueto the caveats stated above, we rather rely on the age rangesestimated from Fig. 8.The HRD positions of TYC8152-1456-1 and TYC8152-550-1 also allow both possibilities, i.e. they could be eitherpre- or post-ZAMS. Applying the Li test to them excludesthe possibility that they are very young. So, they are probablyevolved stars, unrelated to the birth association of the Vela Jr.progenitor, whereas TYC8150-3105-1 could be from the samestellar group that gave birth to the Vela progenitor.
The Hyades, at an age of ∼
625 Myr, show an interestingfeature that contradicts standard mixing mechanisms: Starsof 𝑇 eff ≈ 6600 ± 300 K show Li depletions which areenhanced by a factor ∼
100 compared to neighbouring hot-ter and cooler stars (Boesgaard & Tripicco, 1986; Deliyannis,Anthony-Twarog, Lee-Brown, & Twarog, 2019; Thorburn,Hobbs, Deliyannis, & Pinsonneault, 1993). In a weaker extent,this feature can also be seen in younger clusters like M35( ∼
175 Myr), where the gap is just forming (Steinhauer, 2003;Steinhauer & Deliyannis, 2004). The gap of enhanced Lidepletion is caused by non-standard mixing on the MS, actingpredominantly in the above stated 𝑇 eff regime. Slow mixing,e.g. by rotational and/or gravitational instabilities, and diffu-sion can play a role in the formation of the gap (Steinhauer LUX
ET AL . & Deliyannis, 2004). Recent works (Steinhauer & Deliyannis,2021) show the effect of rotational mixing for the formationof the gap. The authors find that cluster stars with a high 𝑣 rot sin( 𝑖 ) have a more strongly decreasing Li abundance withage. Therefore, observations of stars within the gap and com-paring their Li abundances and 𝑣 rot sin( 𝑖 ) to clusters of knownages can give a further estimate of their ages. Unfortunately,the method is not applicable for us due to the insignificantmeasurements of 𝑣 rot sin( 𝑖 ) and the unknown initial Li abun-dances. Additionally, the method only works for stars withinthe Li depletion gap and only for ages of ∼ −
650 Myr.However, by comparing the measured abundances to clusterdata we can still obtain a consistency check of the ages givenin Tables 5 and 6. This comparison also works for cooler stars,while it also suffers from the same caveats as described for theLi test above, e.g. the large scatter of the log( 𝑁 Li ) .In Fig. 10 we compare our Li targets to data of theclusters Pleiades (Butler, Cohen, Duncan, & Marcy 1987;Pilachowski, Booth, & Hobbs 1987; Boesgaard, Budge, &Ramsay 1988; So93) and Hyades (Boesgaard & Tripicco,1986; Thorburn et al., 1993). The Li depletion gap of theHyades is indicated by the brown arrow and the attached hor-izontal bar at 𝑇 eff = 6300 − 6900 K. Note that the hot edgeof the gap is steeper than the cold edge, where a larger scatteris observed. The 𝑇 eff of TYC159-251-1 A and TYC159-343-1 are consistent with the Li gap. TYC159-251-1 A lies at thecold edge of the gap and its Li abundance is lower than ofHyades stars with the same 𝑇 eff . Therefore, the Hyades agecould be seen as a lower limit for the SB2 TYC159-251-1.TYC159-343-1 has one of the highest Li abundances amongour sample, despite its low metallicity. Its Li abundance iscomparable to Pleiades stars in this 𝑇 eff -region, so the Pleiadesage ( ∼
125 Myr) can be seen as an upper limit. Possibly the staris younger than inferred from Fig. 9 (
90 − 625
Myr), where wedid not consider the metallicity.The other four Li-rich targets have lower 𝑇 eff and do not fallin the Li depletion gap. The abundances of TYC8150-2802-1, TYC8152-1456-1 and TYC8152-550-1 are consistent withthe Hyades or older, whereas TYC8150-3105-1 has more Lithan comparable Pleiades stars. Therefore, it is very young,consistent with what we found from Fig. 9. We list in Table 7 the SNRs with their corresponding num-bers of runaway star candidates from the different selectionsteps and if they were already observed, as well as additionalinformation (e.g. the best candidates). A list of the individualcandidate stars can be found in Table A2.
FIGURE 10
Lithium abundances of Pleiades and Hyadesstars. The data for the Pleiades (green) were taken fromSoderblom et al. (1993, Table 1), including data from Boes-gaard et al. (1988); Butler et al. (1987) and Pilachowski etal. (1987). The data for the Hyades (brown) were taken fromThorburn et al. (1993) and Boesgaard & Tripicco (1986).Inverted triangles show upper limits in the case of non-detections. Our Li targets, shown with coloured squares repre-senting the error boxes, are labeled as in Fig. 9. The stars (1) A TYC159-251-1 A and (2) TYC159-343-1 are consistent withthe Li depletion gap, which is marked by the brown arrow andthe attached horizontal bar.Note that for the candidates selected from
Gaia
DR2 theallowed locations around the SNR centres at the time of theSN were reduced from an error estimate based on the determi-nation of the Vela SNR centre to a smaller error based on thedescription in Green (2009) (see Section 2.2). This was neces-sary to limit the high number of stars from DR2 more strictly,in order to create feasible projects for follow-up observations.In both cases the errors scale with the size of the SNR. Tra-jectories of associated pulsars were considered for the DR2selection.In S147, we can confirm the B0.5 V star HD 37424 basedon DR2 data. Its position at the time of the SN is well withinthe 1 𝜎 error ellipse of the GC used for the DR1 selection andonly slightly outside (by 8.5 %) of the more strict error ellipseused for the DR2 selection. We also checked for the possibil-ity that more than one star might be ejected from a SN in amultiple system. But the three other runaway star candidatesbesides HD 37424, which projected trajectories could origi-nate from the GC and which were observed with UVES, areneither consistent with PSR J0538+2817 nor with HD 37424. UX ET AL . TABLE 7
Numbers of runaway star candidates with
𝐺 < mag in the SNRs covered in this work. Columns 2 and 3 give thenumbers of candidates identified in Gaia
DR1 TGAS and DR2, respectively. Column 4 gives the number of observed stars andif they were observed with UVES/VLT (U) or with HDS/Subaru (H). These observations are all based on the DR1 selection,whereupon Gaia DR2 was used for the further characterisation of the targets. Column 5 gives the names of associated neutronstars and column 6 the numbers of remaining candidates that could not be excluded during the analysis.SNR Name DR1 DR2 Obs. PSR Rem. cands. Additional infoG074.0 − − G205.5+00.5 21 116 ∗∗
21 (U) – 5 incl. HD 261393 G260.4 − − − − − − − ∗ ∗ − ∗ − ∗ ∗ Historical SNR; ∗∗ not suggested for observations, more precise constraints required; Dinçel et al. (2015); Boubertet al. (2017)Two of the DR2 candidates in the Lupus Loop are par-ticularly promising. Their positions at the time of the SNwere . ′ ± 6 . ′ and . ′ ± 6 . ′ off the GC, respectively(see Table A2) and both have very high proper motionsclearly directing away from the centre. Spectroscopic follow-up observations are highly suggested for them.The historical SNRs generally have very small diametersdue to their low age. This reduces the number of stars inthe cone search. Just as Fraser & Boubert (2019), for CasA and the Crab Nebula we did not find any candidates. ForSNR G130.7+03.1 (from SN 1181) we found one star with 𝐺 < . mag to be consistent with the GC, but it was ruledout due to the missing kinematical consistency with PSRJ0205+6449. SNR G347.3 − Θ = 65 ′ × 55 ′ ). Here we found 18 candidates to be consistentwith the GC. The other SNRs are described in more detail inthe following subsection. HB9 is a ′ × 120 ′ SNR on the northern hemisphere,where we found six runaway star candidates around the GCfrom
Gaia
DR1, which were observed with HDS/Subaru. Theadopted SNR distance and age are 𝑑 = 0 . . kpc and 𝑡 = 5 . . kyr, respectively (Leahy & Tian, 2007). Among the observed candidates, we found two giants. The param-eters of the other four are given in Tables 2 and 4. Theyhave a large range of effective temperatures in the literature,where the Gaia
DR2 values are probably underestimated, ascan be seen from comparison with the spectra. Unfortunately,the chosen spectral range was not convenient for spectral typedetermination of early-type stars. Therefore, we rely on the 𝑇 eff given by Bai et al. (2019), which correspond to spectraltype late-A. The RVs were calculated from the positions ofindividual absorption lines (see Section 4.2), where the uncer-tainties are large due to the low number of available lines. Themost precise value is given for TYC3344-235-1, which alsoshows the most promising kinematics. Although the propermotion is moderate, its high radial velocity (pointing towardsus) gives it the highest peculiar velocity among all observedearly-type stars ( 𝑣 pec = 36 . . km s −1 ). Its high distanceof 𝑑 = 1103 +48−45 pc would then require that the SNR lies atthe upper edge of its distance range. TYC3344-235-1 andTYC3344-679-1 have high luminosities of 𝐿 ⊙ and 𝐿 ⊙ ,respectively. From their spectra it is clear that they are notgiants. They could be very young A-type stars but it is morelikely that they are evolved, with ages between ∼ 𝜏 𝑐 = 1 . Myr(Manchester, Hobbs, Teoh, & Hobbs, 2005) and its propermotion indicate that it is unrelated. The magnetar SGR LUX
ET AL . ∼1 . ° to the south of the centre. If thelocation of the GC is close to the explosion site, an associa-tion would mean that SGR 0501+4516 travels with more thantwo times the largest 2D velocity of any other known pulsar(Hobbs, Lorimer, Lyne, & Kramer, 2005), which makes theassociation very unlikely.From Gaia
DR2, we found one fainter, yet unobservedobject consistent with the GC. As both neutron stars are prob-ably unrelated, we suggest that this is the most probablecandidate. It has 𝐺 = 16 . mag and was located . ′ ± 3 . ′ from the GC at the time of the SN. Its spectral type isK5.5 − K7 according to its
Gaia
DR2 𝑇 eff = 4172 +184−67 K. Fromits position in Fig. 2 ( 𝐿 = 0 .
199 ± 0 . 𝐿 ⊙ ), it should have ayoung age of ∼ −
50 Myr, so it could indeed be the ejectedcompanion of the HB9 progenitor.
The Monoceros Loop is one of the largest (
Θ = 220 ′ ) andprobably the oldest SNR of our sample, although the agedetermination is very uncertain (
30 − 150 kyr, Welsh, Sfeir,Sallmen, & Lallement 2001). From Ferrand & Safi-Harb(2012), we adopted 𝑑 = 1 .
44 ± 0 . kpc, as recent works (Yu,Chen, Jiang, & Zijlstra, 2019; Zhao, Jiang, Gao, Li, & Sun,2018) place it at 𝑑 = 0 . +0 . . kpc or 𝑑 = 1 . +0 . . kpcand 𝑑 = 1 . kpc, respectively. The discrepancy connects tothe question if the Monoceros Loop interacts with the RosetteNebula (3C 163), an HII region bordering the SNR at thesouth-western edge.We observed 21 stars with UVES/VLT in the MonocerosLoop, where no associated neutron star is known. Eightof those were ruled out because they were found to begiants. Among the others, the most interesting cases are (i)HD 261393, a B5 V star according to Vo85, which was pro-posed as the most likely runaway companion by Bo17, (ii)TYC159-251-1, which was newly discovered by us to be adouble-lined spectroscopic binary and (iii) TYC159-343-1, anF2–F4 star. TYC159-343-1 and both components of TYC159-251-1 show some Li absorption. However, the ages inferredin Section 4.5 are higher than what we expect for a BES run-away star. Also, from the updated distance to the MonocerosLoop and from using proper motions and parallaxes from Gaia
DR2, TYC159-343-1 ( 𝑑 = 0 . +0 . . kpc) and TYC159-251-1 ( 𝑑 = 0 . +0 . . kpc) turn out to be foreground stars, evenwhen the lower limit of the SNR distance given by Yu et al.(2019) is extended to 𝜎 ( . kpc − 0 . kpc = 0 . kpc).Bo17 did not have Gaia
DR2, so we can still ask the ques-tion why HD 261393 gained a higher probability in their workthan TYC159-343-1. There are two reasons: (1) It is/waslocated closer to the GC. Bo17 used a prior to describe thelocation of the centre as a 2D-Gaussian with a full width at halfmaximum (FWHM) of ̃ Θ = max (5 ′ , .
05 Θ) . This corresponds to ̃ Θ = 11 ′ in the case of the Monoceros Loop. The angulardistance between TYC159-343-1 and the Green centre at thetime of the SN was ′ ± 22 ′ , almost twice the FWHM. Thiscompares to ′ ± 25 ′ for HD 261393. Note that the errors alsotake into account the uncertainty of the GC. (2) It is more mas-sive. Multiplicity studies (e.g. Pinsonneault & Stanek 2006)have shown that massive stars tend to form binaries withsimilar masses. Therefore, Bo17 assign higher runaway prob-abilities to more massive, early-type stars. However, in widesystems the companions can form more independently and thedistribution of mass ratios 𝑞 becomes almost consistent withrandom pairings from the initial mass function (IMF) (Moe& Di Stefano, 2017). Sana & Evans (2011) and Sana et al.(2012) present relatively flat mass ratio distributions down to 𝑞 = 0 . . Unfortunately, the regime of the most extreme massratios, 𝑞 < . , is usually not covered by multiplicity stud-ies, because the extreme flux ratios as well as decreasing RVprecision in the case of early-type stars make low-mass com-panions hard to detect. However, there is no reason to assumethat the mass ratio distributions should change for 𝑞 < . .Among the other observed stars in the Monoceros Loop,there is TYC159-1896-1 which has spectral type F3 − 𝑇 eff as listed in Gaia
DR2) and which does notshow Li, hence is too old, and nine stars of spectral types mid-B to mid-A. Five of those turn out to be spatially inconsistentwith the GC after reevaluating them with DR2 data.So in total, we have five good runaway star candidates in theMonoceros Loop, including HD 261393 and four stars of spec-tral types B7 − A1, which trajectories are displayed in Fig. 3and which kinematic data are noted in Table 2. Their RVs weredetermined from the positions of individual absorption lines(see Section 4.2). For TYC159-2671-1, H 𝛼 was the only vis-ible absorption line, so here the error was adopted from thehighest error among the other stars. Their spectral types wereadopted from Vo85, where the given range comes from ourerror estimate of ±1 subclass.A search for further targets in the DR2 catalogue yieldsmore than 100 candidates with 𝐺 < mag, due to the largeuncertainties of distance and age of the SNR. We abstainfrom listing them all in Table A2 because it is not realistic toobserve them all in the near future. It appears more fruitfulto first concentrate on the brighter candidates observed withUVES/VLT.Considering the large discrepancy of possible distances andages, the identification of a runaway star would be particu-larly important for the Monoceros Loop to constrain theseparameters significantly. UX ET AL . The Vela region, located on the southern sky in the Vela con-stellation, is a very extended and complex region with severalstellar clusters, star-forming regions and three SNRs.The most distant SNR in this region is Puppis A. Here, tworunaway star candidates were observed with UVES, namelyTYC7669-1336-1 and TYC7669-1414-1, which could bothoriginate from the GC in projection. According to their 𝑇 eff from Gaia
DR2, their spectral types are F3 − F6 and F1 − F2,respectively. No Li was found in the spectra. After reevalu-ating them with the DR2 parallax, it was found that they aretoo close to be consistent with the distance given for Puppis A( 𝑑 = 1 . . kpc, Reynoso, Cichowolski, & Walsh 2017).A search in Gaia
DR2 yields nine candidates with
𝐺 < mag, where eight of them have proper motions pointing tonorth-west, similar to most other stars in this region of the sky.Only one candidate, a K1.5 − K3 star with 𝐺 = 13 . mag, hasa higher proper motion and a slightly different direction com-pared to the other stars, pointing to west-north-west. It might,therefore, be the best candidate. It was located . ′ ± 2 . ′ fromthe GC at the time of the SN. With 𝑇 eff = 5042 +115−180 K and 𝐿 = 4 .
64 ± 0 . 𝐿 ⊙ , from Fig. 2 it could be consistent with ayoung age of ∼ − . Gyr.
With 𝑑 = 0 .
275 ± 0 . kpc, Vela is the closest SNR, there-fore having a large diameter of Θ = 255 ′ . Its age is onlypoorly constrained to 𝑡 = 18 ± 9 kyr (Ferrand & Safi-Harb,2012). However, the characteristic age of the Vela pulsar, 𝜏 𝑐 = 11 . kyr, indicates that the SNR age could be close to thelower limit.Two stars were observed, TYC8150-2802-1 and TYC8150-3105-1, both showing Li in their spectra. Their spectral typesare G0 − G2 and F9.5 − G1, respectively. From the Li contentand the position in the HRD, we can conclude that TYC8150-3105-1 has a relatively young age of
25 − 50
Myr, whileTYC8150-2802-1 is probably an evolved MS star. Anyway,they cannot be associated runaway stars, because their tra-jectories are not consistent with the past position of the Velapulsar (J0835 − Gaia
DR2, which are consistent with theVela pulsar: The star marked in blue was identified by us andcould be observed in the near future. The star marked in blackwas identified by Fraser & Boubert (2019) (hereafter FB19),
FIGURE 11
The central ′ × 70 ′ of the Vela SNR. Thered labels mark the geometric centre and its error ellipse,the yellow labels the Li-rich stars observed in ESO P100,namely (3) TYC8150-2802-1 and (4) TYC8150-3105-1. Thecrosses mark the current positions, connected to the posi-tions at the time of the SN (with error ellipse) by an arrowin proper motion direction. The motion of the Vela PSR(J0835 − Gaia
DR2 in blue and the candidate identified by FB19in black. Background image from ESO DSS-2-red.therein denoted as
Star A . It is closer to the nominal past posi-tion of the Vela pulsar, but with 𝐺 = 20 . mag it is muchtoo faint to obtain a spectrum with sufficient 𝑆 ∕ 𝑁 . FB19 notethat it is unlikely that the ejected companion of the Vela SNprogenitor would be such faint (absolute magnitude 𝑀 𝐺 =12 . mag), so probably Star A is an unrelated background star.FB19 used the distribution of runaway star velocities given byRz19 to constrain their search radius, which limits their selec-tion more strictly than the 𝑣 max = 1000 km s −1 used by us.Therefore, the star marked in blue in Fig. 11 was not found bythem. Vela Jr. is located about ° east of Vela. With an age between2.4 kyr and 5.1 kyr it is younger than Vela and its distance isbetween 0.5 kpc and 1.0 kpc (Allen et al., 2015).Five stars were observed with UVES: TYC8152-120-1 is amid-F star, where no Li was discovered. TYC8152-104-1 isa K giant, and TYC8152-1456-1 (F9.5 − G1) and TYC8152-550-1 (F7 − F9) show Li absorption in their spectra. However, LUX
ET AL . the measured abundances indicate that they are much too oldto be runaway stars associated with the Vela Jr. progenitor(see Figs. 9, 10, Table 6). HD 76060 is the brightest star ofour observed sample with 𝐺 = 7 . mag. Its spectral type isB8 IV–V, according to Houk (1978). From its Gaia
DR2 par-allax, we infer a distance of 𝑑 = 0 . +0 . . kpc, which couldstill be consistent with the lower distance limit of Vela Jr. Itsposition at the time of the SN is consistent with the GC accord-ing to Eqn. 2 and its peculiar velocity of . . km s −1 (see Table 2 for further parameters) is relatively high. Itsage, according to Fig. 7, is either . . Myr (if pre-MS)or
90 − 350
Myr (if post-MS). Although the post-MS age,which would be too high, is more likely, the pre-MS cannotbe excluded. Furthermore, with 𝜎 error bars it could also beyoung enough if it is post-MS. Therefore, we still considerHD 76060 as a promising candidate.As the association to PSR J0855 − 𝐺 < mag which are considered good runaway candi-dates together with HD 76060 and are suggested for follow-upobservations. Here, we want to give an estimate how the findings of thiswork fit model populations of runaway stars in the literature(Bo17, Rz19). We investigated twelve SNRs and can excluderunaway stars with
𝐺 < mag in three of them. For SNRS147, we can confirm HD 37424 with Gaia
DR2 data. Basedon this association, we find no further runaway stars withinS147. In each of the remaining eight SNRs, we find one ormore candidates, but none of those can be confirmed yet.So the minimum number of BES runaway stars within thetwelve investigated SNRs is one, while the maximum numberis nine, corresponding to a fraction of ejected runaway starsper SNR of %. The upper limit could be even higherif we also count the cases where more than one runaway starcould have been ejected. In three of the SNRs (Cygnus Loop,Vela Jr., Lupus Loop), two of the remaining candidates couldhave a common origin. Although many more error ellipses ofthe runaway star candidates are overlapping, it was consid-ered that for a certain age of the SNR the error regimes wouldshrink correspondingly.This compares to theoretical values for runaway starsejected from core-collapse SNe between 32.5 % (Bo17) and68 % (Rz19). The latter value comes from +9−22 % of binarysystems that do not merge before the first SN, multiplied by +10−22 % that get disrupted during the SN, which gives +17−26 %of high-mass binary systems that eject a runaway star. Further-more, Bo17 state that ∼ +13−21 % of core-collapseSNe that eject a runaway star.Both values are consistent with our findings. We needfollow-up observations of our best candidates to search for SNdebris as well as more precise estimates for the ages, distancesand explosion sites of the SNRs to further constrain the num-bers of ejected runaway stars in SNRs. From the theoreticalvalues, we expect to verify of our runaway star candidateswithin the eleven SNRs besides S147 in the future. Our search in the
Gaia
DR2 catalogue for runaway starsin twelve Galactic SNRs yielded no further certain runawaystars besides HD 37424, but 73 new promising candidates,which are listed together with HD 37424 in Table A2. Amongthese, five stars in the Monoceros Loop and one in Vela Jr.were observed with UVES/VLT and four stars in HB9 withSubaru/HDS.In total, we observed 39 stars, where 29 were ruled out,e.g. after revisiting their motion with
Gaia
DR2 data or theyturned out to be giants. We found six dwarf stars with lithium;two in Monoceros, two in Vela and two in Vela Jr. TYC159-343-1, located in the Monoceros Loop in projection, mightbe a young star. We obtained
90 − 625
Myr from Fig 9but its Li signal is hard to interpret due to the early spec-tral type (F ) (D’Antona & Mazzitelli, 1984), and itslow metallicity ([M/H] = −0 .
57 ± 0 . ) (Lambert & Reddy,2004). Even . . Myr, the pre-MS solution from Fig. 7,is possible. However, using DR2 data and an updated distanceof the Monoceros Loop, TYC159-343-1 had to be ruled outbecause it is too close. The same applies to TYC159-251-1,which was discovered by us to be a double-lined spectro-scopic binary. TYC8150-3105-1 in Vela shows the strongestLi absorption among all observed stars, so it was found to beyounger than ∼
50 Myr. The Li-rich stars in Vela (TYC8150-3105-1 and TYC8150-2802-1) were ruled out because theyare not consistent with the motion of the PSR. They could beinterlopers near the volume of the SNR while TYC8150-3105-1 could also be a member of the same OB association fromwhich the SNR formed. The Li-rich stars in Vela Jr. were ruledout because they are too old, which can be seen from theirrelatively small Li abundances.The B-type star HD 261393 is the most likely candidate inthe Monoceros Loop. In Vela, we point out two stars which areconsistent with the motion of the Vela PSR. The star markedin blue in Fig. 11 is suggested for observations, while
Star A (FB19) is too faint to be investigated spectroscopically.In total, we found 74 good runaway star candidates with
𝐺 < mag, mostly with spectral type K. We suggest to UX ET AL . take high-resolution spectra to search for the Li 6708 Å line.Among the twelve investigated SNRs, nine SNe could haveejected a runaway star, while only HD 37424 in S147 is con-firmed. Three of these SNe could possibly have ejected tworunaway stars.Note that we might miss some stars with 𝐺 > mag whichwould mainly be spectral type M. Due to missing observationsof such late-type stars around high-mass primaries, we do notknow how many M-type runaway stars we could expect. Fur-thermore, we might miss some Gaia
DR2 candidates due toour strict selection limit from the allowed angular distance tothe geometric centre (GC) at the time of the SN, which wasnecessary to create feasible observing projects.In order to finally proof a BES origin, we need to observethe best candidates with very high resolution and 𝑆 ∕ 𝑁 to beable to detect SN debris, e.g. heavy- and 𝛼 -elements, in thestellar atmospheres. We also emphasise that it is important toprecisely know the explosion site. The case of HD 37424 inSNR S147 is rather exceptional, because both the pulsar andthe runaway star trace back to the Green centre and meet itwithin just a few arcminutes. We did not find other cases wherea pair of a pulsar and a runaway star candidate was located soclose to the GC. In general, the explosion site will not coincidewith the GC, due to the motion of the local standard of restand due to asymmetries of the SNR expansion (e.g. Meyer,Langer, Mackey, Velázquez, & Gusdorf 2015; Meyer, Petrov,& Pohl 2020). A careful analysis of the time-dependent SNRmorphology is needed to locate the explosion site. ACKNOWLEDGEMENTS
We want to thank our present and former colleagues for manyhints and fruitful discussions, in particular Baha Dinçel, AnnaPannicke and Christian Adam. We acknowledge Sergi Blanco-Cuaresma for many useful hints for the usage of iSpec and theanalysis of stellar spectra in general; Eric Mamajek for pro-viding the data and fits for 𝑇 eff versus 𝐸𝑊 Li for clusters ofdifferent ages; and Aaron Steinhauer for providing the datathat relate log( 𝑁 Li ) , 𝑣 rot sin( 𝑖 ) and age. This work made useof ESO VLT data from run 0100.D-0314 and of NAOJ Sub-aru data from run S18B0195S. Therefore, we would like tothank the ESO and NAOJ staff, in particular the local supportastronomers as well as John Pritchard (ESO), Chie Yoshida(NAOJ) and Akito Tajitsu (NAOJ), for the observations andmany useful hints and comments about the data reduction.The NAOJ Subaru telescope is located on the Mauna Keaand we would like to acknowledge the important culturaland spiritual role, which the mountain has for the indigenousHawaiian community. This work made extensive use of data from the ESA-mission Gaia , which were processed and pro-vided by the Gaia Data Processing and Analysis Consortium (DPAC). We further used
VizieR, Simbad and
Aladin , providedby the
Centre de Donnée astronomiques de Strasbourg (CDS),and the
Australia Telescope National Facility (ATNF) PulsarCatalogue (Manchester et al., 2005).Funding was provided by the
Deutsche Forschungsge-meinschaft , project
NE 515/57-1 . REFERENCES
Allen, G. E., Chow, K., DeLaney, T., Filipovi´c, M. D., Houck, J. C.,Pannuti, T. G., & Stage, M. D. 2015, Jan,
ApJ , (2), 82.Anders, F., Khalatyan, A., Chiappini, C. et al. 2019, Aug, A&A , ,A94.Aoki, W., & Helminiak, K. 2014, Data reduction of echelle spectrawith IRAF [Computer software manual]. Tokyo, Japan.Bagnulo, S., Jehin, E., Ledoux, C., Cabanac, R., Melo, C., Gilmozzi,R., & ESO Paranal Science Operations Team. 2003, Dec, TheMessenger , , 10-14.Bai, Y., Liu, J., Bai, Z., Wang, S., & Fan, D. 2019, Aug, AJ , (2),93.Bailer-Jones, C. A. L., Rybizki, J., Fouesneau, M., Mantelet, G., &Andrae, R. 2018, Aug, AJ , (2), 58.Blaauw, A. 1961, May, Bull. Astron. Inst. Netherlands , , 265.Blanco-Cuaresma, S. 2019, Jun, MNRAS , (2), 2075-2101.Blanco-Cuaresma, S., Soubiran, C., Heiter, U., & Jofré, P. 2014, Sep, A&A , , A111.Bock, D. C. J., Turtle, A. J., & Green, A. J. 1998, Oct, AJ , (4),1886-1896.Bodenheimer, P. 1965, Aug, ApJ , , 451.Boesgaard, A. M., Budge, K. G., & Ramsay, M. E. 1988, Apr, ApJ , , 389.Boesgaard, A. M., & Tripicco, M. J. 1986, Mar, ApJ , , L49.Boubert, D., Fraser, M., Evans, N. W., Green, D. A., & Izzard, R. G.2017, Sep, A&A , , A14.Breitschwerdt, D., Feige, J., Schulreich, M. M., Avillez, M. A. D.,Dettbarn, C., & Fuchs, B. 2016, Apr, Nature , (7597), 73-76.Bressan, A., Marigo, P., Girardi, L., Salasnich, B., Dal Cero, C.,Rubele, S., & Nanni, A. 2012, Nov, MNRAS , (1), 127-145.Butler, R. P., Cohen, R. D., Duncan, D. K., & Marcy, G. W. 1987,Aug, ApJ , , L19.Chatterjee, S., Cordes, J. M., Vlemmings, W. H. T., Arzoumanian,Z., Goss, W. M., & Lazio, T. J. W. 2004, Mar, ApJ , (1),339-345.Coluzzi, R. 1999, Sep, VizieR Online Data Catalog , VI/71A.Cos , kunoˇglu, B., Ak, S., Bilir, S. et al. 2011, Apr, MNRAS , (2),1237-1245.Covino, E., Alcala, J. M., Allain, S., Bouvier, J., Terranegra, L., &Krautter, J. 1997, Dec, A&A , , 187-202.D’Antona, F., & Mazzitelli, I. 1984, Sep, A&A , , 431-442.Deliyannis, C. P., Anthony-Twarog, B. J., Lee-Brown, D. B., &Twarog, B. A. 2019, Oct, AJ , (4), 163.Dinçel, B., Neuhäuser, R., Yerli, S. K., Ankay, A., Tetzlaff, N.,Torres, G., & Mugrauer, M. 2015, Apr, MNRAS , (4),3196-3205.Eastman, J., Siverd, R., & Gaudi, B. S. 2010, Aug, PASP , (894),935. LUX
ET AL . Ekström, S., Georgy, C., Eggenberger, P. et al. 2012, Jan,
A&A , ,A146.Evans, D. W., Riello, M., De Angeli, F. et al. 2018, Aug, A&A , ,A4.Ferrand, G., & Safi-Harb, S. 2012, May, Advances in SpaceResearch , (9), 1313-1319.Fraser, M., & Boubert, D. 2019, Jan, ApJ , (1), 92.Gaia Collaboration, Brown, A. G. A., Vallenari, A. et al. 2018, Aug, A&A , , A1.Gaia Collaboration, Brown, A. G. A., Vallenari, A. et al. 2016, Nov, A&A , , A2.Gaia Collaboration, Prusti, T., de Bruijne, J. H. J. et al. 2016, Nov, A&A , , A1.Green, D. A. 2009, Mar, Bulletin of the Astronomical Society ofIndia , (1), 45-61.Green, D. A. 2014, Jun, Bulletin of the Astronomical Society of India , (2), 47-58.Henry, T. J., & McCarthy, J., Donald W. 1993, Aug, AJ , , 773.Hills, J. G. 1988, Feb, Nature , (6158), 687-689.Hobbs, G., Lorimer, D. R., Lyne, A. G., & Kramer, M. 2005, July, MNRAS , (3), 974-992.Hoogerwerf, R., de Bruijne, J. H. J., & de Zeeuw, P. T. 2000, Dec, ApJ , (2), L133-L136.Hoogerwerf, R., de Bruijne, J. H. J., & de Zeeuw, P. T. 2001, Jan, A&A , , 49-77.Houk, N. 1978, Michigan catalogue of two-dimensional spectraltypes for the HD stars.Johnson, D. R. H., & Soderblom, D. R. 1987, Apr, AJ , , 864.Knie, K., Korschinek, G., Faestermann, T., Wallner, C., Scholten, J.,& Hillebrandt, W. 1999, Jul, Phys. Rev. Lett. , (1), 18-21.Lambert, D. L., & Reddy, B. E. 2004, Apr, MNRAS , (2), 757-767.Leahy, D. A., & Tian, W. W. 2007, Jan, A&A , (3), 1013-1018.Manchester, R. N., Hobbs, G. B., Teoh, A., & Hobbs, M. 2005, Apr, AJ , (4), 1993-2006.Meyer, D. M. A., Langer, N., Mackey, J., Velázquez, P. F., &Gusdorf, A. 2015, Jul, MNRAS , (3), 3080-3100.Meyer, D. M. A., Petrov, M., & Pohl, M. 2020, Apr, MNRAS , (3),3548-3564.Moe, M., & Di Stefano, R. 2017, Jun, ApJS , (2), 15.Mugrauer, M. 2019, Dec, MNRAS , (4), 5088-5102.Mugrauer, M., Avila, G., & Guirao, C. 2014, Jan, AstronomischeNachrichten , (4), 417.Mugrauer, M., & Berthold, T. 2010, Apr, AstronomischeNachrichten , (4), 449.Neuhäuser, R. 1997, Jan, Science , , 1363-1370.Neuhäuser, R., Gießler, F., & Hambaryan, V. V. 2020, Oct, MNRAS , (1), 899-917. doi:Pecaut, M. J., & Mamajek, E. E. 2013, Sep, ApJS , (1), 9.Pilachowski, C. A., Booth, J., & Hobbs, L. M. 1987, Dec, PASP , ,1288-1291.Pinsonneault, M. H., & Stanek, K. Z. 2006, Mar, ApJ , (2), L67-L70.Pourbaix, D., Tokovinin, A. A., Batten, A. H. et al. 2009, Aug, VizieROnline Data Catalog , B/sb9.Poveda, A., Ruiz, J., & Allen, C. 1967, Apr,
Boletin de losObservatorios Tonantzintla y Tacubaya , , 86-90.Przybilla, N., Fernanda Nieva, M., Heber, U., & Butler, K. 2008,Sep, ApJ , (2), L103.Renzo, M., Zapartas, E., de Mink, S. E. et al. 2019, Apr, A&A , ,A66.Reynoso, E. M., Cichowolski, S., & Walsh, A. J. 2017, Jan, MNRAS , (3), 3029-3039.Sana, H., de Mink, S. E., de Koter, A. et al. 2012, Jul, Science , (6093), 444.Sana, H., & Evans, C. J. 2011, Jul, The multiplicity of massivestars. C. Neiner, G. Wade, G. Meynet, et al. (Eds.), Active OBStars: Structure, Evolution, Mass Loss, and Critical Limits. Proc.International Astronomical Union Vol. 272, p. 474-485.Sestito, P., & Randich, S. 2005, Nov, AAP , (2), 615-627.Skiff, B. A. 2009, Feb, VizieR Online Data Catalog , B/mk.Soderblom, D. R. 2010a, Sep,
ARA&A , , 581-629.Soderblom, D. R. 2010b, Apr, Using lithium to estimate ages forsolar-type stars. C. Charbonnel, M. Tosi, F. Primas, et al. (Eds.),Light Elements in the Universe. Proc. International AstronomicalUnion Vol. 268, p. 359-360.Soderblom, D. R., Jones, B. F., Balachand ran, S., Stauffer, J. R.,Duncan, D. K., Fedele, S. B., & Hudon, J. D. 1993, Sep, AJ , ,1059.Steinhauer, A. 2003. Formation and evolution of the open clus-ter lithium gap (Unpublished doctoral dissertation). IndianaUniversity.Steinhauer, A., & Deliyannis, C. P. 2004, Oct, ApJ , (1), L65-L68.Steinhauer, A., & Deliyannis, C. P. 2021, submitted .Sushch, I., Hnatyk, B., & Neronov, A. 2011, Jan, A&A , , A154.Tauris, T. M. 2015, Mar, MNRAS , , L6-L10.Tetzlaff, N. 2013. Identifying birth places of young neutron starsto determine their kinematic age (doctoral thesis). Friedrich-Schiller-Universität Jena (FSU).Tetzlaff, N., Dinçel, B., Neuhäuser, R., & Kovtyukh, V. V. 2014,Mar, MNRAS , (4), 3587-3593.Tetzlaff, N., Eisenbeiss, T., Neuhäuser, R., & Hohle, M. M. 2011,Oct, MNRAS , (1), 617-626.Tetzlaff, N., Neuhäuser, R., Hohle, M. M., & Maciejewski, G. 2010,Mar, MNRAS , (4), 2369-2387.Thorburn, J. A., Hobbs, L. M., Deliyannis, C. P., & Pinsonneault,M. H. 1993, Sep, ApJ , , 150.Voroshilov, V. I., Guseva, N. G., Kalandadze, N. B., Kolesnik, L. N.,Kuznetsov, V. I., Metreveli, M. D., & Shapovalov, A. N. 1985,Catalogue of BV magnitudes and spectral classes for 6000 stars.Ukrainian Acad. Nauk, Kiev, 1-140.Wallner, A., Feige, J., Kinoshita, N. et al. 2016, Apr, Nature , (7597), 69-72.Welsh, B. Y., Sfeir, D. M., Sallmen, S., & Lallement, R. 2001, Jun, A&A , , 516-526.Yu, B., Chen, B. Q., Jiang, B. W., & Zijlstra, A. 2019, Sep, MNRAS , (3), 3129-3142.Zehe, T., Mugrauer, M., Neuhäuser, R. et al. 2018, Jan, Astronomis-che Nachrichten , (1), 46-52.Zhao, H., Jiang, B., Gao, S., Li, J., & Sun, M. 2018, Mar, ApJ , (1), 12. APPENDIX A: OVERVIEW OF THERUNAWAY CANDIDATES
Table A1 shows the individual stars that were observed withUVES/VLT in P100 and with HDS/Subaru in S18B, respec-tively. All observations were executed in service mode. For theHDS spectra, the 𝑆 ∕ 𝑁 was measured in continuum regions UX ET AL . between 6620 Å and 6760 Å in the combined spectra with IRAF , while for the UVES spectra we obtained the 𝑆 ∕ 𝑁 fromthe ESO Archive Science Portal . Julian date and barycen-tric Julian date were calculated with the online tools fromAAVSO and Eastman, Siverd, & Gaudi (2010) , respec-tively.Table A2 gives an overview of the 74 possible runawaystars that were identified during this analysis. The spectraltypes SpT were estimated from the Gaia
DR2 𝑇 eff if avail-able, with the following exceptions: HD 37424 was observed,described and confirmed as a runaway star by Dinçel et al.(2015). The stars in SNR G205.5+00.5 as well as HD76060in G266.2 − Skiff catalogue (Skiff2009; Houk 1978; Vo85). The stars in G160.9+02.6 (exceptthe faintest one) were observed by us with HDS and theirspectral types were adopted from the 𝑇 eff given in Bai et al.(2019). AUTHOR BIOGRAPHY
Oliver Lux.
The author obtained a Bachelor of Sciencein Physics and a Master of Science in Astrophysics atthe Universities of Bochum and Bonn, respectively, gain-ing knowledge about neutron stars. In his current workas a PhD student at the University of Jena, he works onrunaway stars and supernova remnants. Further researchinterests are variable stars and exoplanets. http://archive.eso.org/scienceportal/home http://astroutils.astronomy.ohio-state.edu/time/utc2bjd.html LUX
ET AL . TABLE A1
List of observed stars. The columns give target name, instrument (U for UVES/VLT; H for HDS/Subaru), equa-torial coordinates (J2000) calculated with
VizieR from
Gaia
DR2 data, the SNR where the target is located,
Gaia
DR2 G magnitude, barycentric Julian date BJD, total on source integration time 𝑡 exp and the 𝑆 ∕ 𝑁 reached in the fully reduced andaveraged spectra.Name Instr. RA DEC SNR 𝐺 BJD −2450000 𝑡 exp 𝑆 ∕ 𝑁 [h:m:s] [d:m:s] [mag] [d] [s]TYC1869-1435-1 U 05:38:15.15 + − + − + − + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + − − − − − − − − − − − − − − − − − − + + + + + + + + + + + + UX ET AL . TABLE A2
List of identified runaway star candidates. Equatorial coordinates (J2000) were calculated with
VizieR from
Gaia
DR2 data. The fifth column shows the angular distance 𝜌 to the geometric centre at the time of the SN, except for the candidatein Vela (G263.9 − 𝐺 magnitudes, taken from Gaia
DR2, and the last column shows the spectral types SpT.Gaia DR2 RA [h:m:s] DEC [d:m:s] SNR 𝜌 [arcmin] 𝐺 [mag] SpT1859462217726265728 20:50:22.21 + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + − . . + + + +
10 ± 17 + + . . + +
12 ± 17 + + + − . . + +
12 ± 26 + + + +
10 ± 25 + +
11 ± 22 + + − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . TYC3344-235-1; TYC3344-679-1; TYC3344-683-1; TYC3344-553-1; HD 37424; TYC159-2771-1; HD 261359; HD 261393; TYC159-2337-1; TYC159-2671-1 LUX
ET AL . TABLE A2
ContinuedGaia DR2 RA [h:m:s] DEC [d:m:s] SNR 𝜌 [arcmin] 𝐺 [mag] SpT5329655845686016256 08:51:50.00 − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − + . . − + . . − + . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . . − − . .11