Observational mapping of the mass discrepancy in eclipsing binaries: Selection of the sample and its photometric and spectroscopic properties

Abridged. Eclipsing spectroscopic double-lined binaries are the prime source of precise and accurate measurements of masses and radii of stars. These measurements provide a stringent test of models of stellar evolution that are persistently reported to contain major shortcomings. The mass discrepancy observed for the eclipsing spectroscopic double-lined binaries is one of the manifestations of shortcomings in stellar evolution models. Our ultimate goal is to provide an observational mapping of the mass discrepancy and propose a recipe for its solution. We initiate a spectroscopic monitoring campaign of 573 candidate eclipsing binaries of which 83 are analysed in this work with the methods of least-squares deconvolution and spectral disentangling. TESS light curves are used to provide photometric classification of the systems according to the type of their intrinsic variability. We confirm 69 systems as either spectroscopic binaries or higher-order multiple systems. Twelve stars are classified as single and two more objects are found at the interface of their line profile variability being interpreted as due to binarity and intrinsic variability of the star. Moreover, 20 eclipsing binaries are found to contain at least one component that exhibits stellar oscillations. The sample presented in this work contains both detached and semi-detached systems and covers a range in the effective temperature and mass of the star of Teff = [7000,30000] K and M = [1.5,15] M_Sun, respectively. We conclude an appreciable capability of the spectral disentangling method to deliver precise and accurate spectroscopic orbital elements from as few as 6-8 orbital phase-resolved spectroscopic observations. Orbital solutions obtained this way are accurate enough to deliver age estimates with accuracy of 10% or better, an important resource for calibration of stellar evolution models.


Introduction 1
Eclipsing spectroscopic double-lined (hereafter eSB2) binaries 2 are the prime source of precise and accurate mass measurements 3 of stars, the parameter which, along with the initial composition Costa et al. (2019) revisited the sample of Claret & Torres (2016, 2017, 2018, 2019) using a combination of the Bayesian statistical framework and parsec stellar evolution models.The authors confirmed the need for a mild convective core overshooting for stars with masses of M ≥ 1.9 M and additionally found a large spread in the inferred α ov values ranging from some 0.3 to 0.8 H p .Costa et al. (2019) arrived at the conclusion that while the low boundary of the inferred α ov values is likely to be explained by an insufficient amount of core-boundary mixing in standard SSE models, the large spread in the inferred values is likely a manifestation of the natural distribution of initial rotation rates and hence indicates active rotational mixing in these intermediate-mass stars.
Finally, Tkachenko et al. (2020) presented a study of eleven high-mass eSB2 systems for which high-precision stellar quantities were derived in a homogeneous way (Pavlovski et al. 2018(Pavlovski et al. , 2023;;Tkachenko et al. 2014b;Johnston et al. 2019a).No dependence of the observed mass discrepancy on the stellar mass was reported, but a strong anti-correlation with the surface gravity of the star was found.The same correlation was found to persist when the stellar mass was replaced with the mass of the convective core of the star, a finding which the authors interpreted as showing the need for larger convective core masses in standard models of SSE.Remarkably, Johnston (2021) arrived at the same conclusion from the study of a combined large sample of binary components and asteroseismically active single stars.Furthermore, Martinet et al. (2021) investigated a large sample of stars from the literature spanning a mass range from 7 to 25 M using a large grid of SSE models computed for different values of the overshooting parameter and initial rotation and assuming a moderate and strong angular momentum transport in the models.The authors confirmed the findings of Tkachenko et al. (2020) and Johnston (2021) and reported the need for larger convective cores at higher stellar masses.
An alternative hypothetical explanation of the mass discrepancy came from the nature of binarity itself and the fact that the majority of high-mass stars are expected to be found in binary systems and/or have experienced binary interactions in the past (e.g.Sana et al. 2012Sana et al. , 2014)).Mahy et al. (2020) investigated 26 eclipsing binary systems in the LMC and reported a good agreement between their dynamical and spectroscopic masses, whereas the evolutionary masses were found to be systematically overestimated.Upon a closer inspection of the obtained results, the authors found the mass discrepancy to be more pronounced in semi-detached binary configurations, a finding that suggests that binary interactions might be (at least partially) responsible for the observed mass discrepancy.
Despite being known for some three decades and many attempts being made to observationally quantify and pinpoint the most likely theoretical explanation for it, the mass discrepancy persists as a phenomenon, and the only definite statement the stellar astrophysics community can make about it is that the problem exists.The majority of previous studies have focussed on a particular mass regime, and there is a large diversity in the employed data analysis methods and input physics used in SSE models, which makes it non-trivial to account for systematic uncertainties when attempting to quantify and ultimately resolve the mass discrepancy problem.Motivated by the current state of the art, we have initiated a systematic search for intermediateto high-mass (spectral types OBAF and masses larger than some 1.2 M ) eclipsing binaries in TESS (Ricker et al. 2015) spacebased photometric data and have organised a complementary ground-based spectroscopic follow-up campaign of the detected candidate eclipsing binaries.Our goal is to build a stellar sample

Spectroscopic observations
From the above-mentioned sample of 3425 eclipsing binaries, we selected all targets with a V magnitude below 11 and that are accessible from the Roque de los Muchachos Observatory (La Palma, Spain) in the course of the year.A total of 545 systems were proposed for observations with the High-Efficiency and high-Resolution Mercator Echelle Spectrograph (hermes; Raskin et al. 2011) instrument attached to the 1.2-m Mercator telescope as part of a dedicated large programme.For each of those systems, we requested eight epochs of spectroscopic observations with uniform sampling of the orbital phase.The latter was computed from the orbital period value derived by IJspeert et al. (2021) and a reference date T 0 taken to be the date of the first spectroscopic observation.A few systems were followed more extensively to resolve the effect of intrinsic stellar oscillations on spectral line profiles, with the time series ranging from a couple of dozen spectra to almost 200 measurements (e.g. the case of 16 Lac in Table A.1; see Southworth & Bowman 2022).
At the time of writing this work, we had completed spectroscopic observations with the hermes instrument for 58 systems from our sample.The criteria of completion are the following: (i) individual epoch observations meet the minimum S/N requirement of 60 per resolution element and (ii) a quasi-uniform orbital phase coverage is achieved to maximise the chances for detecting lines of a companion star in the observed composite spectra.The former requirement guarantees an S/N in excess of 100 in the disentangled spectra of SB2 systems or the combined spectrum for SB1 binaries.Table A.1 gives an overview of the sample of 58 systems, with columns one through seven indicating the star name and a catalogue identifier (col.1-3), orbital period (col.4), number of spectra acquired (col.5), and spectroscopic (col.6) and photometric (col.7) classifications obtained in this work.
A complementary, albeit smaller, sample of 28 southern hemisphere eclipsing binaries was proposed by us for spectroscopic observations with the Fiber-fed Extended Range Optical Spectrograph (feros; Kaufer et al. 1999) instrument attached to the MPG/ESO 2.2-m telescope (under programme 0106.A-0906(A)).Similar to the hermes-based programme, we requested eight epochs of spectroscopic observations and used exactly the same completion criteria for all the observed systems.The criteria were met for a total of 25 systems, while the remaining three stars received only two epochs of spectroscopic observations.Detailed information is provided in Table A.1 for the 25 systems that we consider completed; the columns of the table are the same as for the hermes sample.
The hermes spectra were reduced with version 7.0 of the dedicated hermes pipeline, while the feros spectra were processed with a modification of the ceres pipeline (Brahm et al. 2017) presented in Gebruers et al. (2022).The data reduction steps we used are the standard ones and include bias subtraction, flat fielding, cosmic ray removal, wavelength calibration, barycentric correction, and order merging.Normalisation to the pseudocontinuum was done by selecting knot points and fitting a lowdegree polynomial through them.Special care was taken in the regions of broad Balmer lines to ensure the outer line wings were not altered by the normalisation process.

Spectroscopic analysis
In this section, we present the results of our preliminary spectroscopic analysis of 83 systems, that is, 58 in the hermes sample and 25 in the feros sample.We start with spectroscopic classi-Article number, page 3 of 17 (1) The line mask consists of a set of delta functions that specify 283 wavelength positions and theoretical strengths of spectral lines.

284
This implies that the exact content of a line mask depends on the assumed atmospheric parameters (T eff , log g, ξ, [M/H]) of the star, where the effective temperature has by far the most dominant effect.For the purpose of the analysis in this work and as informed by the spectral classification as OB(A)-type stars in IJspeert et al. ( 2021), we computed line masks for two values of T eff , 8 000 K and 11 000 K, and we used both of them for the calculation of the LSD profiles from every single spectrum of every object in the sample.The two masks are at the same time required and sufficient to account for the variable complexity of the stellar spectrum between these two temperature regimes.
We performed a visual classification of all systems based on the time series of their LSD profiles.Figure 1 shows a time series of the LSD profiles for four objects ranging from a single star to a quadruple system.In the figure, we classify HD 34382 (top left) as a single star exhibiting line profile variations (LPVs).The latter are caused by the intrinsic variability of the star, either due to rotational modulation of inhomogeneities on its surface or stellar pulsations.We found HD 138305 (top right) to be an SB2 system composed of two similar stars.HD 234650 (bottom left) and HD 57158 (bottom right) are higher-order multiple systems, with the former being a spectroscopic triple-lined (SB3) system and the latter being a spectroscopic quadruple-lined (SB4) system.
Article number, page 4 of 17 Fig. 2. Example orbital solutions for SB2 systems with differing numbers of observed spectra and fairly good phase coverage.Black lines depict the orbital solutions, while blue and red dots represent orbital phases of the observed spectra for the primary and secondary components, respectively.V436 Per, α CrB, and NY Cep are highly eccentric binary systems with eccentricities of about 0.37, 0.38, and 0.44, respectively.RX Dra represents the case of a circular-orbit system with the previously undetected secondary component that is revealed by this work.

Spectral disentangling 309
The SPD method was originally introduced and formulated by where the single-column matrices x and c represent a priori unknown disentangled spectra of the components and the time se-324 ries of the observed spectra, respectively.The rectangular matrix

325
M represents a linear transformation of x to c, has a block struc-326 ture, and is constructed using individual Doppler shifts for each 327 of the components at each epoch of the observations.Matrix M also contains information on the light dilution factors for each component of the system and each observed spectrum in the time series.In the case of high-resolution spectra and when the number of observed spectra exceeds the number of individual components, the system of linear equations is overdetermined (i.e.number of equations is greater than the number of unknowns).Mathematically, this is an ill-posed problem, and regularisation conditions are required to solve it.Simon & Sturm (1994) employed an algebraic technique known as a singular value decomposition (SVD).Solving the system is typically a computationally demanding problem due to the need to invert matrix M in Eq. (2).
Independently, a Fourier-based formulation of SPD was developed by Hadrava (1995).With the help of the Fourier transform, a large set of equations can be uncoupled into many small sets of equations for each Fourier mode.Such an approach is significantly less computationally demanding than the original formulation.In either formulation, the SPD method can be exploited in two basic modes: (i) spectrum disentangling, where a priori unknown components' spectra are optimised along with the orbital elements of the system, and (ii) spectrum separation, where RVs or orbital elements are assumed to be known Article number, page 5 of 17 a challenge for the method, the latter is, in principle, a direct vi-365 olation of one of the method's fundamental assumptions.Hence, 366 we comment on both of these challenges in more detail to justify 367 the choice of the SPD method for our purpose.

368
In the method of SPD, the gain in S/N in the final disentangled spectra is proportional to While time-dependent LPVs, due to the effect of non-radial pulsations, represent an extra source of variability that is not recognised by the SPD method, previous studies of binary systems with pulsating component(s) have demonstrated that the method delivers reliable results provided that (i) the LPV amplitude due to the intrinsic variability of the star is much smaller than due to its binary motion and (ii) the orbital period of the system differs from that of the dominant intrinsic variability of the star (e.g.Uytterhoeven et al. 2005a,b;Ausseloos et al. 2006;Tkachenko et al. 2012).For the case of a radially pulsating star where the RV amplitude of the dominant (radial) pulsation mode is comparable to or exceeds that of the orbital motion, a large set of observed spectra with uniform coverage of the pulsation and orbital cycles is required.Indeed, in such a scenario, a time series of the observed spectra can be binned with respect to the orbital phase of the binary system to average out the effect of oscillations on the line profile of a pulsating component.As a result, spectral line profiles in the orbital phase-binned observed spectra contain a minimum amount of distortion due to stellar pulsations, bringing us to the former case of orbital motion being the dominant source of the apparent line profile variability (e.g.Tkachenko et al. 2014aTkachenko et al. , 2016)).
In this work, we employ the Fourier-based method of SPD as implemented in the FDBinary software package5 (Ilijic et al. 2004).We used the disentangling method to verify and, if necessary, refine the LSD-based spectroscopic classification, and more importantly, we derived a preliminary set of orbital elements for case of a refined classification is of an SB1 (LSD-based) to an 406 SB2 (SPD-based) system.This is thanks to the power of the SPD 407 method to detect spectral contributions as faint as a few percent 408 (in the continuum flux units), which may stay unnoticed in the 409 LSD profiles if the most optimal line mask is not chosen for the 410 respective calculations.

411
The results of our combined LSD-and SPD-based spectro-412 scopic classification are provided in Table A.1 (sixth column).

413
For all SB1 and SB2 systems, we employed the FDBinary code the SPD method.The red and blue filled circles respectively indi-464 cate the predicted primary and secondary RVs from the obtained orbital solution and at the epochs of the acquired spectroscopic observations.We note that the method of SPD bypasses the step of RV determination from the observed spectra; hence, the RVs shown in Figure 2 are inferred quantities and used to illustrate the typical phase coverage we achieved for our systems irrespective of their orbital eccentricity value.Examples of the disentangled spectra for the same four systems are shown in Figure 3.In the top row of the figure, we show a system with two similar components (V436 Per, light ratio l B /l A = 0.798 ± 0.021; Southworth & Bowman 2022) and a system with an extreme light ratio between the two stars (α CrB, light ratio l B /l A = 0.0180±0.0002,Pavlovski et al., in preparation).In the latter case, spectral lines of the secondary component have a depth of at most 0.5% in the continuum units, owing to their rotational broadening.In the bottom row of the figure, the disentangled spectra are shown for the NY Cep (left) and RX Dra (right) systems.The primary component in the detached eclipsing binary NY Cep is one of the hottest stars in our sample, which is evidenced by the presence of He ii lines at 4541 Å and 4687 Å.The SPD method also revealed the spectrum of the secondary component in the RX Dra system (right), which makes it an SB2 system with a γ Dor pulsating component.

Photometric analysis
Previous efforts have demonstrated the power of combining high-quality TESS light curves with spectroscopic radial velocities for relatively small numbers of high-mass pulsating eclipsing binaries (see e.g.Bowman et al. 2019b;Southworth et al. 2020Southworth et al. , 2021;;Southworth & Bowman 2022).As discussed in Section 2, our large sample of EBs was originally identified as eclipsing systems by IJspeert et al. ( 2021) using the products of automated light curve extraction tools, such as the MAST SPOC pipeline (Jenkins et al. 2016) and/or the TESS quicklook pipeline (QLP).However, in our current work, we extract a light curve optimised for each pulsating EB using the TESS fullframe images (FFIs), following a similar methodology described in Bowman et al. (2022), which is briefly summarised here for completeness.
TESS pixel cutouts sized 25 × 25 were extracted from the FFIs using astrocut (Brasseur et al. 2019).We used the lightkurve software package (Lightkurve Collaboration et al. 2018) to select aperture masks optimised to maximise the S/N of the eclipses (and pulsations if present) whilst minimising the contribution from nearby contaminating sources based on crosschecking with the location and fluxes of sources in the Gaia catalogue.The background flux was estimated from the median observed flux per frame after excluding pixels that contain flux from the target or other sources.The background flux was subtracted, and the extracted light curves were normalised by dividing through the median flux.Finally, we performed a principal component analysis (PCA) to remove any remaining instrumental systematics in the light curves (Bowman et al. 2022).
Due to the difference in integration times between different TESS data cycles and the large (i.e.> 1 yr) gaps between cycles, we chose to analyse data from each cycle, if available, separately for each star.To classify the dominant variability beyond the eclipses caused by binarity and to ascertain if any additional signal is caused by pulsations or rotational modulation (RotMod) in our sample, we created a multi-frequency co-sinusoid model to remove the orbital harmonics from the extracted TESS light.Our approach is similar to the method used by Bowman et al. (2019b) for the pulsating eclipsing binary system U Gru.After found to contain either a g-(eight binaries) or p-mode (eight binaries) pulsator, and one system (V350 And) contains a g-/pmode hybrid pulsator.Finally, the CD-45 4393 system was found by us to show a steadily decreasing eclipse depth across different TESS sectors such that the signal practically disappears in sectors 61 and 62.Such a phenomenon could be a manifestation of the precession of the binary orbit (e.g.due to the presence of a third body in the system), which causes the orbital inclination angle with respect to the observer's line of sight to change with time.
Below, we discuss systems that have been studied in the literature and, where applicable, compare our photometric and spectroscopic results with those reported by other research groups.We focus primarily on dedicated studies rather than on large catalogues.Hence, the overview provided in this section is probably not inclusive.
BD+13 1880: Šubjak et al. ( 2020) reported that the system is composed of a metallic-line Am star and a brown dwarf companion in a 3.6772±0.0001d orbit.From the analysis of a set of some 50 spectra obtained with multiple high-resolution spectrographs, the authors reported the RV semi-amplitude K of the host star as 4.64±0.03km s −1 , which is in a good agreement with the value of 5.0±0.1 km s −1 derived in this work from a time series of ten hermes spectra.
BD+36 3317: Özdarcan et al. (2012) classified the system as an Algol-type binary based on the analysis of ground-based multi-colour photometry and provided evidence of the system being a member of the δ Lyrae cluster.The system was revisited by Kıran et al. (2016) based on photometric observations from the literature and a newly obtained time series of 20 mediumresolution (R = 11 700) spectra.The authors reinforced the conclusion regarding the cluster membership of the system and reported the RV semi-amplitudes of the primary and secondary component to be K 1 =80.8±1.1 and K 2 =124.1±1.3,respectively.Both values are in good agreement with those derived in this work from eight hermes spectra, namely, K 1 =82.93±0.03and K 2 =127.3±0.1.
65 UMa B: The star was discovered to be magnetic by Bychkov et al. (2003), while Aurière et al. (2007) measured its longitudinal magnetic field and rotation period as B l = −166 ± 20 G and 15.830 d, respectively.Joshi et al. (2010) presented a detailed spectroscopic study of the star based on a time series of high-resolution spectra obtained with the 2.56-m Nordic Optical Telescope (NOT).The authors reported on the absence of RV variations in the spectra and on the abundance pattern characteristic of evolved Ap stars.Zasche et al. (2012, see their fig.7) reported 65 UMa to be a sextuple system consisting of (i) a close eclipsing pair of nearly identical stars (Aa1+Aa2) orbiting each other with a period of ∼1.73 d; (ii) a distant third component (Ab) orbiting the close pair with a period of about 640 d; (iii) a fourth component (65 UMa B), resolved interferometrically, whose period is about 118 yr; and (iv) a fifth (65 UMa C) and a sixth (65 UMa D) component with periods of ∼14 kyr and ∼591 kyr, respectively.Our spectroscopic classification of 65 UMa B as a single star not showing notable RV variability is in full agreement with the literature.The TESS light curve revealed the presence of shallow eclipses that are indicative of an eclipsing binary with a period of about 1.73 d.We concluded that the eclipse signal in the TESS light curve of 65 UMa B is due to a contaminating light from the component A, which has a similar surface brightness.
EL CVn: Maxted et al. (2014) studied the system based on a combined WASP photometry and ses high-resolution (R=60 000) spectroscopy.The authors found the system to be Article number, page 8 of 17 V994 Her: The system was discovered to have a quadruplelined double-eclipsing nature from ground-based multi-colour photometric and high-resolution spectroscopic data by Lee et al. (2008).The authors reported the two eclipsing binaries in the system to have orbital periods of ∼2.08 d and ∼1.42 d, of which the former is in agreement with the period used in this work and reported by IJspeert et al. (2021).Both binaries were found to have slightly elliptic (e<0.1)orbits, with reported masses of 2.83±0.20 M , 2.30±0.16M , 1.87±0.12M , and 1.86±0.12M for the Aa, Ab, Ba, and Bb components, respectively.Zasche & Uhlař (2013) reported V994 Her to be a quintuple system with the two eclipsing binaries orbiting each other with a period of about 6.3 d.These findings were refined in Zasche & Uhlař (2016), where the authors reported an orbital period of ∼2.9 d for the two eclipsing pairs and found evidence for apsidal motion with periods of about 116 and 111 years.The authors also reported updated masses of 3.01±0.06M , 2.58±0.05M , 1.84±0.03M , and 1.93±0.04M for the Aa, Ab, Ba, and Bb components, respectively.Ultimately, Zasche et al. (2023) reported the discovery of a third set of eclipses in the TESS space-based and archival ground-based photometric data, which makes the system a triply eclipsing sextuple star system.The authors refined the orbital period of the A(Aa+Ab)-B(Ba+Bb) core system to 1062±2 d and reported on its close to 3:2 mean motion resonance.The eclipsing pair C(Ca+Cb) was reported to have an elliptic orbit with a period of 1.96 d and component masses of 1.81 +0.17 −0.07 M and 1.08 +0.16 −0.11 M , respectively.Our classification of V994 Her as an SB4 system in this work is in agreement with the previous findings.
V1898 Cyg: The system was classified as a spectroscopic double-lined binary by Abt et al. (1972) and as an eclipsing sys-Article number, page 9 of 17 tem by Halbedel (1985).A detailed analysis of the system based on the newly obtained spectroscopic and archival photometric data was presented in Dervisoglu et al. (2011).The authors reported K 1 and K 2 RV semi-amplitudes of 55.2±0.8 km s −1 and 287.6±2.1 km s −1 , respectively, which are in good agreement with the findings in this work.
GK Cep: The system was classified as an eclipsing spectroscopic double-lined binary with a period of 0.936171 d and a spectroscopic mass ratio of 0.92 by Bartolini et al. (1965).Furthermore, Pribulla et al. (2009)  AH Cep: The system was analysed based on a combined photometric and spectroscopic dataset by Bell et al. (1986).The authors reported K 1 = 249±8 km s −1 and K 2 = 283±8 km s −1 under the assumption of a circular orbit.Holmgren et al. (1990a) and Burkholder et al. (1997) reported K 1 = 237±2 km s −1 and K 2 = 269±2 km s −1 and K 1 = 230.0±3.2 km s −1 and K 2 = 277.6±4.4 km s −1 , respectively, under the same assumption of e = 0.The latter solution is in agreement with the findings in this work.A recent study by Pavlovski et al. (2018) reported a slightly higher value, K 1 = 234.9±1.1 km s −1 while their measure of K 2 = 276.9±1.4 km s −1 remains consistent with our determination of the respective parameter.WW Aur: A detailed analysis of the system Was presented by Southworth et al. (2005).The authors reported K 1 = 116.81±0.23 km s −1 and K 2 = 126.49±0.28km s −1 and M 1 = 1.964±0.007M and M 2 = 1.814±0.007M for the primary and secondary component, respectively.Our determinations of the RV semi-amplitudes are in agreement for the primary component and slightly lower for the secondary star.AW Cam: A single-lined binary solution was presented by Mammano et al. (1967) with K 1 = 112±5 km s −1 under the assumption of a circular orbit.Frey et al. (2010) provided a simultaneous light-and RV-curve solution estimating K 1 to 110 km s −1 .Our estimate of K 1 is in agreement with the previous findings, and we establish the double-lined nature of the system in this work.
V453 Cyg: Popper & Hill (1991), Simon &Sturm (1994), andBurkholder et al. (1997) respectively reported spectroscopic RV semi-amplitudes of K 1 = 171±1.5km s −1 and K 2 = 222±2.5km s −1 ; K 1 = 171.7±2.9 km s −1 and K 2 = 223.1±2.9 km s −1 ; and K 1 = 173.2±1.3 km s −1 and K 2 = 213.6±3.0 km s −1 .Of the more recent studies, Southworth et al. troscopic study of the system based on 26 newly obtained highresolution spectra.Among other things, the authors reported an eccentricity, argument of periastron, and components' RV semi-amplitudes of e = 0.48 ± 0.02, ω = 58 • ± 2, K 1 = 112±3 km s −1 , 782 and K 2 = 158±8 km s −1 , respectively.Albrecht et al. (2011) re-783 visited the system based on 46 newly obtained high-resolution 784 spectra with the sophie spectrograph.The authors reported e = 785 0.443 ± 0.005, ω = 56.3• ± 1, K 1 = 113.8±1.2 km s −1 , and K 2 786 = 139±4 km s −1 assuming P = 15.27566d, as derived in Ahn 787 (1992).Our determinations of the eccentricity and argument of 788 periastron of the system as well as the RV semi-amplitudes of 789 both binary components are in good agreement with the findings 790 by Albrecht et al. (2011).791 16 Lac (= EN Lac): A detailed spectroscopic analysis of 792 the star as an SB1 system based on some 1200 newly obtained 793 and archival spectra was presented by Lehmann et al. (2001).794 The authors reported P = 12.096844 d, e = 0.0392 ± 0.0017, 795 ω = 63.7 • ± 2.1, and K 1 = 23.818±0.033km s −1 .While it has 796 also been known as a single eclipsing system since its discovery 797 by Jerzykiewicz (1980) Harmanec et al. (1997) presented the first detailed 807 spectroscopic study of the system based on a collection of newly 808 obtained and archival data.Using the SPD method, the authors 809 reported the detection of LPVs and determined the orbital eccen-810 tricity, argument of periastron, and RV semi-amplitudes of the 811 components to be e = 0.3882±0.0043,ω = 108.98• ±0.27,K 1 = 812 98.0±1.0 km s −1 , and K 2 = 102.5±1.2 km s −1 , respectively.The 813 system was revisited by Janík et al. (2003) based on a new set of 814 high-resolution spectroscopic observations.The authors did not 815 confirm the previously reported LPVs and presented an updated 816 set of spectroscopic orbital elements: e = 0.3768 ± 0.0014, ω = 817 109.83• ±0.10,K 1 = 97.4±0.1 km s −1 , and K 2 = 91.2±0.1 km s −1 .818 A notable finding in the latter study is that the secondary com-819 ponent has a lower RV semi-amplitude, suggesting it is a more 820 massive star in the system.Southworth & Bowman (2022) anal-821 ysed TESS photometric data of the V346 Per system, reported 822 several local minima in the obtained solution, and emphasised 823 the importance of obtaining an independent (spectroscopic) es-824 timate of the components' light ratio to resolve the degeneracy 825 they encountered.The spectroscopic orbital elements of the sys-826 tem that we derived in this work are in agreement with those 827 reported by Harmanec et al. (1997) except that both of the K-828 values we found are some 0.3% lower.We also note that un-829 like Janík et al. (2003), we found the primary component to be 830 the more massive one in the system.In addition, we found the 831 star to exhibit stochastic low-frequency photometric variability, 832 in a agreement with the classification by Southworth & Bowman 833 (2022).

834
RX Dra: We found the system to contain a γ Dor-type pul-835 sator.This is in agreement with the classification by Southworth 836 & Van Reeth (2022).

837
V1425 Cyg: A detailed study of the system based on a com-838 bined set of spectroscopic and photometric observations was pre-839 sented by Hill & Khalesseh (1993).The authors reported K 1 = 840 142.3±1.8 km s −1 and K 2 = 221.7±2.2 km s −1 under the assump-841 tion of a circular orbit.The respective RV semi-amplitudes de-842 rived by us in this work are in agreement with the findings by 843 Hill & Khalesseh (1993).
Article number, page 10 of 17 Tkachenko et al.: Observational mapping of the mass discrepancy: Selection and properties of the sample V446 Cep: Our photometric classification of the system as a β Cep pulsator is in agreement with the conclusions by Southworth & Bowman (2022).These authors also reported the star to presumably exhibit tidally induced oscillations.
CM Lac: A study of the system based on a combined set of spectroscopic and photometric observations was presented by Liakos & Niarchos (2012).The authors determined K 1 = 119±2 km s −1 and K 2 = 156±2 km s −1 under the assumption of a circular orbit.Southworth & Van Reeth (2022) revisited the system based on TESS photometric data and spectroscopic RV measurements from Liakos & Niarchos (2012).The authors reported K 1 = 120.0±3.4 km s −1 and K 2 = 157.0±3.3 km s −1 , in agreement with the previous findings.Our determinations of the RV semi-amplitudes are in good agreement with both of the abovementioned studies.Southworth & Van Reeth (2022) reported the system to contain a γ Dor pulsator, and we confirm their findings in this work.
V398 Lac: The system was investigated spectroscopically by Çakırlı et al. (2007).The authors reported e = 0.230, K 1 = 110.3±3.7 km s −1 , and K 2 = 128.5±3.8 km s −1 for the eccentricity and RV semi-amplitudes of the components, respectively.A moderate orbital eccentricity of 0.273 and 0.2284 ± 0.0007 was confirmed by Bulut & Demircan (2008) and Wolf et al. (2013), respectively, based on the studies of Hipparcos photometry and apsidal motion of the system.While the orbital eccentricity obtained by us is in agreement with the previous studies, the RV semi-amplitudes of both components are lower by some 15% to 20% than those reported by Çakırlı et al. (2007).
V402 Lac: A detailed analysis of the system based on the newly obtained spectroscopic data and TESS space-based photometry was presented by Baroch et al. (2022).The authors determined the eccentricity and RV semi-amplitudes to be e = 0.376 ± 0.003, K 1 = 128.5±0.8 km s −1 , and K 2 = 129.2±0.8 km s −1 from the combined RV and eclipse timing analysis, suggesting the system is composed of two stars with a similar mass.Our findings are similar but not exactly the same as the determinations by Baroch et al. (2022).In particular, we found a slightly lower eccentricity and a K-value for the primary component, the latter being suggestive of a slightly smaller mass ratio of the system.AE Pic: A spectroscopic orbit of the system was presented by Sahade & Landi Dessy (1950).The authors reported e = 0.1, ω = 39 • , and K 1 = 119 km s −1 .In this work, we classified the star as an SB2 system with a slightly lower eccentricity and RV semi-amplitude of the primary than reported previously.The system was classified as an eclipsing binary containing a pulsator that additionally exhibits signatures of rotational modulation by Barraza et al. (2022).We confirm their findings in this work and classified the star as having a β Cep pulsator with extra signatures of rotational modulation in the light curve.

Conclusions
Of the 573 systems proposed for spectroscopic monitoring with the hermes and feros spectrographs, we report phase-resolved spectroscopic observations for 83 of them (cf.Table A.1).We  B.1.In the case of SB2 systems, parameters of the hotter component are displayed.Solid lines represent mesa evolutionary tracks for the stellar mass values M of (from right to left) 1.5, 3.0, 4.5, 7.0, and 13 M and a minimum amount of overshoot f OV = 0.005 H p .The tracks are from Johnston et al. (2019b).order SB3 or SB4 multiple systems, while we classified the re-906 maining 12 targets as single stars.

907
The log(T eff ) − log(g) Kiel diagram of the above-mentioned 908 65 SB1 and SB2 systems is shown in Figure 5.Where possible, 909 the T eff and log g values have been taken from the literature (cf.910 Section 5); otherwise, the respective values have been estimated 911 from the spectral type and luminosity class of the star by inter-912 polating in the tables of Schmidt-Kaler (1982).We also note that 913 T eff and log g of only the hotter and more massive primary com-914 ponents are displayed in Fig. 5. Therefore, the positions of the 915 majority of stars in the Kiel diagram are not precise but rather 916 indicative and used for the purpose of demonstrating the param-917 eter space covered by the sample presented in this work.With a 918 sub-sample comprising slightly over 10% of the total 573 candi-919 date systems that we are currently monitoring spectroscopically, 920 we achieved the coverage of a large range in the effective tem-921 perature and mass of the star, that is, T eff ∈ [7000, 30000] K 922 and M ∈ [1.5, 15] M .Currently, the lower-mass regime with 923 M 3.5 M is populated more densely, and the sample consists 924 almost exclusively of stars in the core hydrogen burning phase.925 Nevertheless, the present sample allowed us to start look-926 ing into the problem of mass discrepancy as a function of stellar 927 mass, surface properties of the star (i.e.T eff , log g, and v sin i), 928 binary configuration (i.e.detached versus semi-detached sys-929 tems), and intrinsic variability of the star (i.e.pulsating versus 930 non-pulsating components).Indeed, several systems in the sam-931 ple with AF-type primary components are known Algols, and 932 more of those have light curve characteristics of Algol-type bi-933 naries.A mix of well-detached and semi-detached systems in 934 the sample allowed us to assess quantitatively the role of binary 935 interactions in the mass discrepancy problem, as has been sug-936 gested by Mahy et al. (2020).Furthermore, diversity of the sam-937 ple in terms of the effective temperature and mass of the primary 938 component is an asset in the context of the possible connection 939 between the mass discrepancy and amount of the near-core mix-940 ing in the form of the convective core overshoot proposed in 941 the literature (e.g.Guinan et al. 2000;Massey et al. 2012;Mor-942 rell et al. 2014;Claret & Torres 2016, 2017, 2018, 2019)  The above-mentioned accuracy of 3% in stellar mass is sufficiently high to provide a pertinent calibration of SSE models for scientific exploitation of future space-based missions, such as PLATO (Rauer et al. 2014).Indeed, Chaplin et al. (2014) and Silva Aguirre et al. (2017) demonstrated asteroseismically for solar-like pulsators (i.e.main-sequence Sun-like stars and evolved intermediate-mass stars on the red giant brunch) that ∼10% precision in age can be achieved when the mass and radius of the star are measured with the precision of 4% and 2%, respectively.The 10% age accuracy is one of the most fundamental science requirements for the PLATO mission to be able to characterise an Earth-like planet in the habitable zone of a Sun-like star (Rauer et al. 2014).In this work, we demonstrated that our approach to the planning of spectroscopic observations and subsequent analysis of the obtained data precision and accuracy is compliant even with the most stringent requirements of space-based missions such as PLATO.
In the forthcoming papers, we will present detailed analyses of individual systems based on their combined spectroscopic and TESS photometric data.All systems will be studied in the context of the mass discrepancy problem presented in detail in Section 1. Also, the forthcoming papers will include updates on our ongoing spectroscopic monitoring of the entire sample of 573 candidate eclipsing binaries in a format similar to that used in the present study.TYC8514-106-1 TIC 382044531 0.7903919(1) 6 SB2 EB The letter 'N' refers to the number of spectra we have at our disposal.Period uncertainty is provided in parentheses in terms of the last digit. 1Eclipse signal is detected in the TESS data, but contamination as the source cannot be excluded. 2Variable eclipse depth such that eclipses practically disappear in the TESS sectors 61 and 62.

Appendix B: Spectral disentangling-based orbital solutions 1230
In this section we present SPD-based orbital solutions for all stars that are classified as either SB1 or SB2 systems in Table A The terms e, ω, and K i stand for the eccentricity, time of periastron passage, and RV semi-amplitudes, respectively.Cases where eccentricity and argument of periastron were fixed to 0.0 and 90 • , respectively, are indicated with the superscript 'f'.Orbital period values and their uncertainties have been adopted from Table A.1. 1 The star is classified as an SB2 system based on the visual inspection of the LSD profiles and/or original observed spectra; however, the SPD solution could only be obtained for the brighter primary component. 2Same as above but no stable SPD solution could be obtained for either of the components.
3 Either an SB1 system with a low K semi-amplitude of about 1.5 km s −1 or a single star showing LPVs; the methods employed in this work do not favour any of these hypothesis.The terms e, ω, and K i stand for the eccentricity, time of periastron passage, and RV semi-amplitudes, respectively.Cases where eccentricity and argument of periastron were fixed to 0.0 and 90 • , respectively, are indicated with the superscript 'f'.Orbital period values and their uncertainties have been adopted from Table A.1. 1 The star is classified as an SB2 system based on the visual inspection of the LSD profiles and/or original observed spectra; however, the SPD solution could only be obtained for the brighter primary component. 2Same as above but no stable SPD solution could be obtained for either of the components.
3 Either an SB1 system with a low K semi-amplitude of about 1.5 km s −1 or a single star showing LPVs; the methods employed in this work do not favour any of these hypothesis.
Article number, page 16 of 17

Fig. 1 .
Fig.1.Examples of the LSD profiles computed with the T eff =11 000 K mask.From top left to bottom right: HD 34382, a single star showing line profile variability; HD 138305, an SB2 binary system; HD 234650, an SB3 triple system; and HD 57158, an SB4 quadruple system.The individual LSD profiles in each panel have been shifted vertically by a constant factor for clarity.

Fig. 3 .
Fig. 3. Examples of the disentangled spectra for the binary systems shown in Fig 2. In the upper row, V436 Per and α CrB represent systems with an extreme light ratio.In the lower row, binary systems with hot (NY Cep) and cool (RX Dra) components in our sample are shown for comparison.

414 460 Figure 2
Figure 2 shows example orbital solutions for a selection of

Fig. 4 .
Fig. 4. Example summary figure for TIC 150442264.Top panels: FFI-extracted light curve versus time (left) and phase folded on the orbital period for all available cycle 1 sectors (right).Bottom panels: Original and residual amplitude spectra in which vertical red lines denote the location of significant orbital harmonics that comprise the analytic binary model, which is also shown as a red line in the phase-folded light curve panel.
classified the system as a spectroscopic triple, in agreement with the findings in this work.A detailed study of GK Cep was presented in Zhao et al. (2021), which is based on newly obtained photometric data from the lunar-based ultraviolet telescope and the TESS space-based mission.The authors confirmed the presence of a third body in the system and measured the masses of the close binary components to be 1.93 M and 2.11 M .

(
2004) obtained an eccentricity of e = 0.022 ± 0.002 from the apsidal motion solution and K 1 = 173.7±0.8 km s −1 and K 2 = 224.6±2.0 km s −1 from the RV fitting.Consistent with these findings are the solutions obtained by Pavlovski & Southworth (2009); Pavlovski et al. (2018), where the latter study reported e = 0.022 ± 0.002, K 1 = 175.2±0.7 km s −1 , and K 2 = 220.2±1.6 km s −1 .The orbital elements that we determined in this work are in agreement with the previous studies, including the small eccentricity of the system.The star was classified as including a β Cep variable by Southworth et al. (2020), in line with our own classification in this work.NY Cep: Holmgren et al. (1990b) presented a detailed spec-

Fig. 5 .
Fig. 5. Kiel diagram of stars listed in TableB.1.In the case of SB2 systems, parameters of the hotter component are displayed.Solid lines represent mesa evolutionary tracks for the stellar mass values M of (from right to left) 1.5, 3.0, 4.5, 7.0, and 13 M and a minimum amount of overshoot f OV = 0.005 H p .The tracks are fromJohnston et al. (2019b).
. The 943 Article number, page 11 of 17 A&A proofs: manuscript no.output fact that the present sample contains binaries with and without 944 pulsating components will allow us to study the role of pulsa-945 tions in the mass discrepancy problem methodologically (e.g. to 946 what extent the presence of intrinsic variability alters the infer-947 ence of the absolute stellar dimensions from a light curve) and 948 physics-wise (e.g. to what extent a wave-induced mixing alters 949 interior and atmospheric properties of stars in models that are 950 used to quantify the mass discrepancy).Moreover, the detection 951 of stellar oscillations in at least one of the binary components 952 will enable an independent measurement of the amount of the 953 near-core mixing with asteroseismic methods.A further exten-954 sion of the sample towards inclusion of more evolved stars will 955 allow us to investigate a possible connection between the mass 956 discrepancy and evolutionary stage of the star, in particular the 957 role of turbulent and radiative pressure terms in the inference of 958 the effective temperature of the star when its surface gravity is 959 accurately known from a photodynamical model of the system 960 (e.g.Tkachenko et al. 2020).961 The sub-sample representing slightly over 10% of the to-962 tal of 573 candidate eclipsing binary systems that we are cur-963 rently monitoring spectroscopically is one of the largest (if not 964 the largest) eclipsing binary star ensemble being observed in 965 a largely consistent way (i.e. with the same ground-based in-966 struments for spectroscopic data and TESS space-based mission 967 for photometric data sets), and it will be analysed by us with 968 a well-established modelling framework, described in detail in 969 Tkachenko et al. (2020).The high level of consistency in the ob-970 servational strategy and modelling approach is particularly im-971 portant for minimising otherwise numerous and hardly traceable 972 systematic uncertainties in the investigation of the mass discrep-973 ancy problem.Coupled with a homogeneous and, in the near 974 future, complete coverage of the parameter space, we can con-975 clude that our sample of eclipsing binaries has all the potential 976 to become the foundation to quantify and resolve the mass dis-977 crepancy problem in eclipsing binaries.978 Finally, we stress an appreciable capability of the SPD 979 method to infer precise and accurate spectroscopic orbital ele-980 ments and individual components' spectra from as little as six to 981 eight orbital phase-resolved spectroscopic observations.Indeed, 982 by using an estimate of the orbital period of the system as the 983 only input to the planning of our observational campaign, we 984 have demonstrated that a stable SPD solution can be obtained 985 for the majority of binaries in the sample.This is contrary to the 986 established thinking in the community, namely, that an extensive 987 time series typically comprised of a couple of dozen spectro-988 scopic observations is required for the SPD method to deliver 989 meaningful results.In particular, by comparing our orbital so-990 lutions reported in Table B.1 with those obtained by other re-991 search groups based on typically more extended datasets, we 992 found a good agreement for all but a few systems (cf.Sec-993 tion 5) in the sample.Moreover, systems like AE Pic or AW 994 Cam, which are known as SB1s in the literature, were discovered 995 by us to be double-lined binaries and thus became ideal candi-996 dates for detailed modelling with the goal of inferring absolute 997 dimensions of both binary components.Furthermore, the preci-998 sion with which we inferred the RV semi-amplitudes of both bi-999 nary components is often comparable to what is reported in dedi-1000 cated studies in the literature (e.g.Pavlovski et al. 2018; Schmitt 1001 et al. 2016, for the AH Cep and α CrB systems, respectively).1002 This in turn translates into our ability to infer absolute dimen-1003 sions of stars with precision and accuracy better than 3% when a 1004 complementary high-quality ground-or space-based photomet-1005 ric dataset is available (which is the case for all systems in our 1006 sample). 1007 where N is the number of the , a grazing secondary eclipse was for the 798 first time detected in the TESS data by Southworth & Bowman 799 (2022).The spectroscopic orbital elements derived by us in this 800 work are in good agreement with those presented by Lehmann 801 et al. (2001).The star has been reported to be a β Cep variable 802 by Jerzykiewicz (1980); Dziembowski & Jerzykiewicz (1996); 803 Lehmann et al. (2001); Aerts et al. (2003); Jerzykiewicz et al. 804 (2015); Southworth & Bowman (2022), in line with our own 805 findings in this work.

Table A .
1. Spectroscopic and photometric classification for stars studied in this work.N' refers to the number of spectra we have at our disposal.Period uncertainty is provided in parentheses in terms of the last digit.1Eclipsesignalisdetectedin the TESS data, but contamination as the source cannot be excluded.2Variableeclipsedepth such that eclipses practically disappear in the TESS sectors 61 and 62.Article number, page 14 of 17 Tkachenko et al.: Observational mapping of the mass discrepancy: Selection and properties of the sample Table A.1.continued. .1.
Article number, page 15 of 17