Asteroseismic Analysis of δ Scuti Components of Binary Systems: The Case of KIC 8504570

The present work concerns the Asteroseismology of the Kepler-detached eclipsing binary KIC 8504570. Particularly, it focuses on the pulsational behaviour of the oscillating component of this system and the estimation of its physical parameters in order to enrich the so far poor sample of systems of this kind. Using spectroscopic observations, the spectral type of the primary component was determined and used to create accurate light curve models and estimate its absolute parameters. The light curve residuals were subsequently analysed using Fourier transformation techniques to obtain the pulsation models. Theoretical models of δ Scuti stars were employed to identify the oscillation modes of the six detected independent frequencies of the pulsator. In addition, more than 385 combination frequencies were also detected. The absolute and the pulsational properties of the δ Scuti star of this system are discussed and compared with all the currently known similar cases. Moreover, using a recent(empirical) luminosity–pulsation period relationship for δ Scuti stars, the distance of the system was estimated.


Introduction
The δ Scuti stars are short-period and multiperiodic pulsating variables. In general, they oscillate in radial and low-order non-radial pulsations due to κ-mechanism [1,2]. However, recently, it has been proposed that the turbulent pressure in the hydrogen convective zone may explain the observed high-order non-radial modes [3,4]. Their masses typically range between 1.4 and 2.5 M [1], their spectral types between AIII/V and FIII/V, and they are located inside the classical instability strip. Thanks to the Kepler [5,6], the K2 [7], the GAIA [8] and the Transiting Exoplanet Survey Satellite (TESS) [9,10] missions, as well as the All-Sky Automated Survey for Supernovae (ASAS-SN) [11] project, many studies, e.g., [12][13][14][15][16], based on large data sets of δ Scuti stars, have been published, providing new, tremendous knowledge for pulsators of this kind.
The eclipsing binaries (EBs) can be considered as the utmost tools for the calculation of stellar absolute parameters (e.g., masses, radii, luminosities) and the evolutionary stages of their components, particularly in cases when spectroscopy and photometry are combined. However, it should be noted that for systems with large luminosity differences between their components, the radial velocity measurements of the less luminous component are, in general, very difficult to get, because the light of the more luminous component dominates the spectrum.
In addition, given that the phase parts along the quadratures are the most important for the calculation of the amplitudes of the radial velocity curves, the systems with orbital periods longer than

Spectroscopy
The purpose of the spectroscopic observations was the estimation of the spectral type of the primary component of the system. The spectra of the target were obtained with the 2.3 m Ritchey-Cretien "Aristarchos" telescope at Helmos Observatory in Greece on 6 October 2016. The Aristarchos Transient Spectrometer 3 (ATS) instrument [50] using the low resolution grating (600 lines mm −1 ) was employed for the observations. This set-up provided a resolution of ∼3.2 Å pixel −1 and a spectral coverage between approximately 4000 and 7260 Å. Three successive spectra with 10 min exposures were acquired for KIC 8504570 during the orbital phase 0.81 and added together in order to achieve a better signal-to-noise ratio (S/N). The mean S/N of the individual spectra was ∼13, while that of the final integrated spectrum was ∼18. For the spectral classification, a spectral line correlation technique for the spectra of the variable and standard stars was applied. The selected standard stars, suggested by the Gemini Observatory 4 , ranged between A0 and K8 spectral classes (one standard star per subclass) and were observed with the same set-up during August-October 2016. All spectra were calibrated (bias, dark, flat-field corrections) using the MAXIM DL software. The data reduction (wavelength calibration, cosmic rays removal, spectra normalisation, sky background removal) was done with the RAVERE v.2.2c software [51].
The applied correlation method has been described in detail in Liakos [39], but is briefly presented here too. For the comparison between the spectrum of KIC 8504570 and the standard stars, the Balmer and the strong metallic lines between 4000 and 6800 Å were used. The differences of spectral line depths between each standard star and the target star were compared via sums of squared residuals in each case, with the least squares sums indicating the best fit. This method is quite efficient in cases of EBs with large luminosity differences between their components, because the total spectrum is practically dominated by the light of the primary star. In our study we did not use any synthetic spectrum approximation (c.f. [52]) in order to avoid any instrumental effects (e.g., distortion) that cannot be taken into account in a synthetic model. Therefore, using the direct comparison method, and given that all spectra were acquired with the same set-up, any systematic effects were directly removed.
On the other hand, in cases with small luminosity differences between the components, this method does not provide accurate results, and more specifically, it might lead to an underestimation of the spectral type of the primary. Therefore, in order to avoid this, the following method described in [32] was applied. Using the spectra of the standard stars, all the possible combinations were calculated by simply adding and normalising the spectra. Furthermore, for every spectra combination, the spectrum of each component was given a weight between 0 and 1 denoting its light contribution to the combined spectrum. The starting value for the contribution of the primary component was 0.5 and the step was 0.05. Finally, for each spectral combination, ten sub-combinations with different light contributions of the components were derived. Similarly to the previous method, the same spectral lines were used for the comparison of the combined spectra with that of KIC 8504570, again via deriving sums of squared residuals. Hence, the smallest value of these residuals indicated again the best match.
The spectrum of KIC 8504570 was found to be dominated (at least 95%) by the light of the primary component. Therefore, its spectrum was directly compared with those of the standards. The sum of squared residuals against the spectral type for this system is plotted in Figure 1, which shows that the best fit was found with the spectrum of an A9V standard star. The spectrum of the system along with that of best-match standard star are illustrated in Figure 2. It should be noted that for the spectra continuum normalisation, polynomials of various orders were used according to the spectral types of the stars, since each spectral type has a different peak wavelength. However, for the continuum normalisation of the spectra of the standard stars with spectral types close to those of the targets (e.g., between A5-F5), the same polynomials were used. Therefore, since our method is based on direct comparison (i.e., subtraction of spectra), the non-perfect continuum normalisation does not affect the results. The present spectral classification, with an error assumption of one sub-class, corresponds to a temperature T eff = 7450 ± 150 K for the primary, based on the relations between T eff and spectral types of Cox [53]. The present result comes in relatively good agreement with those given in previous studies (see Section 1).  Figure 2. Comparison spectra of KIC 8504570 (black line) and the standard star (red line) with the closest spectral types. The Balmer and some strong metallic lines are also indicated.

Light Curve Modelling and Absolute Parameter Calculations
The system was observed in long-and short-cadence modes by the Kepler mission during various quarters. However, since the primary goal of this study concerns the asteroseismic analysis of the pulsating star of KIC 8504570 (i.e., pulsation modelling and mode identification), only the short-cadence data downloaded from the KEBC [35] were used for the frequency analysis. However, it should be noted that the data obtained for this system during non successive quarters of the Kepler mission provide significant time gaps, something that is crucial for the frequency analysis alias effect [34]. Furthermore, time gaps exist also within the data of a single quarter. Therefore, the selection of data for this system was made according to their continuity and total amount in time in order to include the most compact data sample possible. More specifically, the data of Q13 and a part of Q14 were selected for analysis. In total, 150,458 available points were used. These data were obtained during 106.9 consecutive days and provide 27 full LCs. The level of light contamination for this system is zero (as listed in the Mikulski Archive for Space Telescopes; MAST). The total covering and continuous time of observations is more than three months (with negligible time gaps), which is sufficient for the study of short-period pulsations and for LC modelling. The short-cadence Kepler LCs of the first 40 days of observations for KIC 8504570 are illustrated in Figure 3. The orbital phases and the flux to magnitude conversions for this system were derived using the ephemeris (T 0 = 2,454,955.78(3) BJD, P orb = 4.007705(8) d) and the Kepler magnitude K p = 13.25 mag, respectively, as listed in KEBC.
The LC analyses were done with the PHOEBE v.0.29d software [54] that is based on the 2003 version of the Wilson-Devinney code [55][56][57]. The temperature (T eff, 1 ) of the primary component was given a value as yielded from the spectral classification (see Section 2), and it was kept fixed during the analysis. On the other hand, the temperature of the secondary component (T eff, 2 ) was adjusted. The albedos (A) and the gravity darkening coefficients (g) were assigned values according to the spectral types of the components [58][59][60]. The (linear) limb darkening coefficients (x) were taken from the lists of van Hamme [61]. The synchronicity parameters (F) were initially adjusted, but due to the absence of significant changes during the iterations, the system was assumed to be tidally locked (i.e., F 1 = F 2 = 1) following the preliminary findings of Lurie et al. [62]. The dimensionless potentials (Ω), the fractional luminosity of the primary component (L 1 ) and the inclination of the system (i) were set as adjustable parameters. Since there is no supporting evidence for the existence of a tertiary component, and additionally, since the light contamination was zero, the third light parameter (l 3 ) was not taken into account. At this point, it should be noted that the R filter (Bessell photometric system-range between 550 and 870 nm and with a transmittance peak at 597 nm) simulated the best spectral response of the CCD sensors of Kepler (410-910 nm with a peak at ∼588 nm). Therefore, it was used for the calculation of the filter depended parameters (i.e., x and L) in PHOEBE.
In the absence of spectroscopic mass ratio (q) for KIC 8504570, the q-search method (for details, see e.g., [63]) was applied. For this, a mean LC exempted from the presence of pulsations was needed. Moreover, in this system, except the short-period pulsations, brightness variations due to magnetic activity (e.g., spots), occurring mostly in the out-of-eclipse phase parts, were also found. Therefore, the mean LC (folded into the orbital period) was calculated from two to four successive LCs; there were no major brightness changes between them. It should be noted that a complete LC of KIC 8504570 contains approximately 5500 data points. The mean LC, using averaged points per phase, contained approximately 300 normal points, and the variations of both the pulsations and the spots almost vanished. The q-search was applied in modes 2 (detached system), 4 (semi-detached system with the primary component filling its Roche lobe) and 5 (conventional semi-detached binary) to find feasible ("photometric") estimates of the mass ratio. The step of q change during the search was 0.1 starting from q = 0.1. The sums of the squared residuals were systematically lower for all q values in mode 2; therefore, this system can be plausibly considered as a detached EB.  According to the q-search method, the minimum sum of squared residuals was found for q = 0.5 ( Figure 4). This value was initially assigned to q, but later on it was adjusted. This system presents remarkable brightness changes from cycle to cycle after the 10 day of observations. It was found that for 40 continuous days after the 10 day, a hot spot on the surface of the secondary component describes the individual LCs very well. The selection of the hot spot was based on the results of Davenport [49] regarding possible flare activity in the system and fits well to a profile of a star with temperature of 5300 K (secondary component). Between 52 and 75 days of observations, no spots were required for the LC model, in contrast with the time range between 76 and 104 days, for which a cool spot was adopted on the surface of the same component. The spot parameters (colatitude Colat, longitude long, radius and temperature factor T f ) were adjusted in the individual LC models. Finally, for this system, one model per LC was obtained; thus, 27 models were totally derived and combined for the final average model.
The analyses of Kepler LCs for EBs require special handling due to light variations caused by magnetic spots between successive LCs; c.f. [32,39,40]. That justifies our choice not to model all the available points folded into the P orb , but to model each LC separately. This method provides more realistic errors for the final model results, since its single parameter (except from those of the spots) is the average from those of the individual models, while its error is the standard deviation of them. Moreover, using this method, the brightness changes due to the spots and other proximity effects are well modelled; hence, the final LC residuals can be considered as free as possible of the binarity, something that is extremely crucial for the subsequent pulsation analysis (Section 4). The LCs' modelling results for KIC 8504570 are listed in Table 1. Examples of LC modelling and Roche geometry representation are plotted in Figure 5. The LC residuals after the subtraction of the individual models are illustrated below the observed LCs in Figure 3. Moreover, the parameters of the spots for each LC (cycle) are given in Table A1 in Appendix A. Figure A1 includes the immigration plots of the spots and their locations on the surface of the secondary component for two different dates of observations.
Although no RV curves exist for this system, the absolute parameters of its components can be estimated making plausible assumptions. The adopted mass (1.67 M ) of the primary was based on its spectral type according to the spectral type-mass correlations of Cox [53] for main-sequence stars. A fair mass error value of 10% was also adopted. The mass of the secondary component can be directly derived from the calculated (photometric) mass ratio. The semi-major axes a can then be derived from Kepler's third law. The luminosities (L), gravity's acceleration (log g) and the bolometric magnitude values (M bol ) were calculated using the standard definitions. The calculations of the absolute parameters were done with the software ABSPAREB [64], and they are listed in Table 1.

Pulsation Modelling
The search for pulsation frequencies was done with the software PERIOD04 v.1.2 [65] that is based on classical Fourier analysis. Although the typical frequency range of δ Scuti stars is 4-80 d −1 [34], the present analysis included the regime 0-4 d −1 too. This selection was based on the fact that it has been noticed (e.g., [32,39]) that these stars may also exhibit longer-period oscillations due either to tidal effects, which are connected to their P orb , or even to the intrinsic hybrid behaviour of γ-Doradus-δ Scuti type. Therefore, the present pulsation analysis was done in the range 0-80 d −1 on the LC residuals of the system ( Figure 3). Moreover, since the eclipses affect the amplitudes of the pulsations (i.e., variations of the total light) and in order to keep the data sample homogeneous, only the out-of-eclipse data were used. The ranges of orbital phases (Φ orb ) of the excluded data were 0.97-0.03 and 0.47-0.53. For the signal-to-noise ratio (S/N) calculation of the frequencies, the method for the background noise estimation, as described in detail in Liakos [39], was applied. Particularly, the background noise of the data set was calculated as 7.51 µmag in regimes with absence of frequencies, with a spacing of 2 d −1 , and a box size of 2. A 4σ limit (i.e.,S/N= 4) [65] regarding the reliability of the detected frequencies was adopted (0.03 mmag). Hence, after the first frequency computation the residuals were subsequently pre-whitened for the next one until the detected frequency had S/N∼4. The Nyquist frequency and the frequency resolution according to the Rayleigh criterion (i.e., 1/T, where T is the observation time range in days; c.f. Aerts et al. [1], Schwarzenberg-Czerny [66]) for the present data set were 239.5 d −1 and 0.009 d −1 , respectively. According to the present spectroscopic and LC modelling results (Sections 2 and 3), only the primary component of KIC 8504570 adequately simulates the properties of δ Scuti-type stars (i.e., mass and temperature); hence, it can be plausibly concluded that this star is the pulsator of this system.
After the frequency search, the pulsation constant for each independent frequency ( f ) was calculated based on the relation of Breger [34]: Moreover, the following pulsation constant-density relation was used for the calculation of the density of the pulsators: where f dom is the frequency of the dominant pulsation mode (i.e., that with the largest amplitude). At this point it should be noted that the f dom of the multiperiodic δ Scuti stars varies over time. Therefore, for a more realistic estimation of the density of this pulsator, the average value of Q of the independent frequencies was used.
The identification of the oscillating modes (i.e., l-degrees and type) employed the theoretical MAD models for δ Scuti stars [67] in the FAMIAS software v.1.01 [68]. The l-degrees from the closest MAD models (i.e., f , log g, M and T eff ) to the detected independent frequencies were adopted as the most possible pulsation modes. Moreover, the ratio P pul /P orb of all independent frequencies was calculated in order to check whether it is less than 0.07, which is the upper value, according to Zhang et al. [22], for the discrimination of p-type modes. Table 2 includes the pulsation modelling results regarding the independent frequencies for KIC 8504570 as well as their respective mode identification. Particularly, this table lists: The frequency value f i , the amplitude A, the phase Φ, the S/N, the Q, the P pul /P orb , the l-degrees and the mode of each detected independent frequency. The rest of the detected frequencies (i.e., dependent/combination frequencies) are given in Appendix B (Table A2). Figure 6 shows the periodogram of the pulsating star of KIC 8504570 and the distribution of its oscillation frequencies. Representative Fourier fittings on the LC residuals are plotted in Figure 7.    The pulsator of KIC 8504570 oscillates in a total of 393 frequencies. Six of them are independent and were detected in the regime 11.8-26.2 d −1 . Among the other 387 depended frequencies, 309 were spread almost uniformly in the range 10-43.4 d −1 ; 72 had values less than 4.4 d −1 ; five were found between 5.5-9 d −1 ; and only one, namely, f 282 , exceeded 50 d −1 . As can be seen in Figure 6, one main concentration of frequencies is between 12 and 17 d −1 , while a slightly more spread out one is between 23 and 30 d −1 . The results based on MAD models show that all oscillations are probably non-radial pressure modes. Although the ratio f 4 / f 10 has value ∼0.78, f 4 was not identified as a radial mode by the MAD models. Finally, a value of ρ pul = 0.215(4) ρ was derived.

Summary, Discussion and Conclusions
In the present work, detailed LC and pulsation modellings for KIC 8504570, a neglected Kepler-detached EB with an oscillating component, are presented. The spectral classification of its primary component, based on our spectroscopic observations, provided the means for accurate LC analyses, and for the estimation of the absolute parameters and evolutionary stages of both the components of the EB. The primary component was also identified as a δ Scuti star and its pulsational characteristics (pulsation frequencies model and mode identification) were accurately determined.
The primary component of KIC 8504570 was classified as an A9-type star and pulsates in six independent frequencies in the regime 11.89-26.2 d −1 with the dominant part at 14.37 d −1 . These frequencies were identified as non-radial (pressure) modes according to the MAD models. Moreover, this star oscillates in another 387 combination frequencies. During the LC modelling, initially a hot and subsequently a cool spot on the surface of the secondary component were used to overcome the brightness asymmetries in the quadratures. This selection can be justified from the fact that this EB was listed as a possible flare system [49].
For the estimation of the evolutionary stages of the components of KIC 8504570, the locations of its members on the mass-radius (M − R) and Hertzsprung-Russell (HR) diagrams are illustrated in Figures 8 and 9, respectively. Both components are located inside the main-sequence and follow the theoretical evolutionary tracks of Girardi et al. [69] (see Figure 9) very well according to their derived masses and the corresponding error ranges (see Table 1). Therefore, it seems that they have been evolving without any significant interactions so far. In terms of evolution, the δ Scuti component of KIC 8504570 has similar absolute properties to other δ Scuti stars in detached binary systems. It is among the eight less massive and less luminous stars of this sample and it is located closer to red edge of the classical instability strip.  Figure 8. Locations of the primary (filled symbol) and secondary (empty symbol) components of KIC 8504570 (diamonds) within the mass-radius diagram. The stars and the "×" symbols denote the δ Scuti components of other detached systems with P orb shorter and longer than 13 d, respectively (taken from Liakos and Niarchos [26] and Liakos [32]). The black solid lines represent the main-sequence edges. In order to check the accordance of the pulsational properties of the δ Scuti star of KIC 8504570 with others that belong in similar systems, it was placed on the P pul − P orb and log g − P pul diagrams ( Figures 10 and 11, respectively) along with the well established empirical relations of Liakos [32] for δ Scuti stars in detached binaries with P orb < 13 d. The studied star in these plots follows very well both the distributions of the sample stars and the empirical relations.   Figure 11. Location of the δ Scuti (primary) star of KIC 8504570 within the log g − P pul diagram. Symbols and lines have the same meanings as in Figure 10.
Using the current dominant oscillation frequency of the pulsator and the pulsation period-luminosity relation for δ Scuti stars of Ziaali et al. [14], it is feasible to calculate its absolute magnitude (M V = 2.06(13) mag). Hence, using the apparent magnitude (m V ) and the distance modulus, its distance can be calculated. The m V of KIC 8504570 is 13.28 mag according to the NOMAD-1 catalogue [71] and the extinction in V band is A V = 0.336 mag [47]; thus, its distance is determined as 1502 +93 −87 pc. This value is in very good agreement with the value 1488 ± 41 pc. as derived by Berger et al. [47] and Bailer-Jones et al. [72], and in slight disagreement with the value 1305 pc of Queiroz et al. [73]. The latter discrepancy is attributed to the different extinction value (A V = 0.474 mag) used by Queiroz et al. [73]. It should be noted that the aforementioned M V is in very good agreement with the M bol,1 = 2.17(6) mag, which was calculated based on the LC model;ing (Table 1).
For the future, radial velocity measurements are welcome to validate the present results for the LC model, although the ∼95% light domination of the primary component makes the acquisition of the radial velocities of the secondary an extremely difficult task. At best, we anticipate that only the radial velocities of the primary can be measured, which will only constrain the mass of the primary component, and hence the mass ratio of the system. However, these potential future measurements cannot significantly change the present pulsations models, especially the results for the dominant and the independent frequencies, which were the main goals of the present study. The asteroseismic modelling of other similar systems, especially of those observed by satellite missions, is highly encouraged and recommended because the sample of δ Scuti stars in binary systems is still small and we lack of enough information. Moreover, systems with P orb between 10 and 20 d should be prioritised for detailed analysis in order to check the reasons for the existence of the boundary of P orb ∼ 13 d.

Acknowledgments:
The authors wish to thank Mrs Maria Pizga for proofreading the text and the three anonymous reviewers for their fruitful comments. The "Aristarchos" telescope is operated on Helmos Observatory by the Institute for Astronomy, Astrophysics, Space Applications and Remote Sensing of the National Observatory of Athens. This research has made use of NASA's Astrophysics Data System Bibliographic Services, the SIMBAD, the Mikulski Archive for Space Telescopes (MAST) and the Kepler Eclipsing Binary Catalog data bases.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:

Appendix A. Spot Migration
This appendix includes information for the variation of the spots locations in time, which were assumed to be located on the surface of the secondary part of the system (see also Section 3). It should be noted that the present solution is just a suggestion for describing the LC asymmetries in the quadratures, and more solutions (e.g., more spots with different sizes and temperatures) may result in the same LC behaviour (degeneracy of solutions). The average BJD values of the points included in the models from which the respective parameters (colatitude Colat., longitude long., radius and temperature factor T f ) were calculated, were set as corresponding timings for each cycle in Table A1. The upper part of Figure A1 shows the changes of the parameters of all spots over time for the system, and the lower parts show the spot(s) on the secondary part's surface during different days of observations.   Table A2 contains the values of the depended frequencies f i (where i is an increasing number), semi-amplitudes A, phases Φ and S/N for KIC 8504570. Moreover, in the last column of this table, the most likely combination for each frequency is also given. The combinations were calculated only for the first 255 frequencies because this is the maximum number of frequencies that the software can detect during one run (i.e., Fourier model). However, it should be noted that in order to continue the search, the residuals from these Fourier models were given as new data sets to the same software, but no combinations could be calculated using the first 255 frequencies.   Table A2. Cont.  Table A2. Cont.    (4) 163.6(6.0) 5. 3  Table A2. Cont.