Quality: Stellar libraries tests

Elodie library

Date of this test: 2021/06/23

To assess the performance of VOSA to estimate effective temperatures of FGK stars, we have made use of the Elodie library (v3.1) (1389 objects).

First, we kept only entries in Elodie having a quality flag=4 ("excellent") in effective temperatures (159 objects)

Then, we cross-matched with the Gaia EDR3 catalogue looking for counterparts in 5arcsec. In order to avoid extinction - effective temperature degeneracies, we kept objects with parallaxes > 10 mas and relative errors < 20% (parallax_error/parallax < 0.2). After this, we kept 127 objects.

The observational SED of these objects were built gathering photometry from the following catalogues: SLOAN DR12, APASS9, Gaia EDR3, Pan-Starrs DR2, and 2MASS. Also the following grid of models were used: Kurucz ODFNEW/NOVER, alpha:0.0; Kurucz ODFNEW/NOVER, alpha0.4; BT-Settl, BT-Settl (CIFITS). Only objects with good SED fitting (vgfb < 12) were considered for the comparion.

Kurucz model, alpha: 0.0.

Teff_VOSA - Teff_Elodie: 6.43 ± 102.77 K (42 objects)

Kurucz model, alpha: 0.4.

Teff_VOSA - Teff_Elodie: 18.33 ± 96.57 K (42 objects)

BT-Settl.

Teff_VOSA - Teff_Elodie: 50.47 ± 113.42 K (42 objects)

BT-Settl-CIFIST.

Teff_VOSA - Teff_Elodie: 15.68 ± 99.32 K (42 objects)

We can see how, for the four collections of models, VOSA estimates effective temperatures close to those given in Elodie. Only to mention that temperatures estimated using BT-Settl are slightly lower, in particular at Teff < 5200K.

The position in the H-R diagram of the 42 objects used in the comparison is given below.

The forty-two objects used in the comparison are the following:

VOSA and hot stars.

To assess the performance of VOSA at high temperatures we have used the compilation of sdO stars made by Stroeer et al. (2007 A&A, 462, 269)

For our analysis we have selected only those sdO targets not flagged as "outliers" in effective temperature (Table1 of the paper). Then, for these targets, the observational SED has been built using photometry (GALEX, CMC-14, 2MASS) retrieved from VO services using VOSA. The following criteria were adopted:

• J(2MASS) < 17
• H(2MASS) < 16.2
• K(2MASS)< 15
   
• r (CMC-14) < 17
   
• FUV (GALEX) > 12
• NUV(GALEX)>11

An average value of T_eff (VOSA)-T_eff (Stroeer)=2800 ± 6700K is found for a sample of 14 objects.

Comparison with Yee et al. (2017ApJ...836...77Y)

Date of this test: 2017/05/16

We compare the results in Yee et al. with the fit results obtained by VOSA.

• Yee et al. (2017ApJ...836...77Y)
  • Library of optical spectra of 404 touchstone stars observed with Keck/HIRES. High-resolution (R~60000), high signal-to-noise (SNR~150/pixel).
  • Properties (M, R, Teff, [Fe/H]) derived from interferometry, asteroseismology, LTE spectral synthesis, and spectrophotometry.
  • Spectral types ~ M5-F1 (Teff ~ 3000-7000K, Rstar ~ 0.1-1.6 Rsun).
• SED building using VOSA
  • Photometric SED built using Galex, Gaia, APASS, 2MASS and WISE data.
  • Model fit using Kurucz (logg: 2.5 - 5.0; [M/H]:1-5-0.5)

  Effective temperatures

  Only objects with good fit (vgfb<=12) and sigma<200K in the Bayesian fitting are considered (155 objects).

  Using Kurucz model we find:

  Teff (Yee) - Teff(VOSA)

  • Mean: -4.91K
  • Std: 208.84K
  • Median: -34.77K
  
  

  Surface gravities

  Only objects with good fit (vgfb<=12) and sigma<0.3dex in the Bayesian fitting are considered (38 objects).

  Using Kurucz model we find:

  logg (Yee) - logg (VOSA)

  • Mean: 1.14dex
  • Std: 0.64dex
  • Median: 1.24dex
  
  

  But if we use BT-Settl instead of Kurucz, the situation is the reverse, with the gravity values computed by VOSA systematically higher than those given in the paper (28 objects have been used this time).
  

  Metallicities

  Only objects with good fit (vgfb<=12) and sigma<0.3dex in the Bayesian fitting are considered (141 objects).

  Using Kurucz model we find:

  [M/H] (Yee) - [M/H] (VOSA)

  • Mean: 0.16dex
  • Std: 0.58dex
  • Median: -0.02dex
  
  

  A similar result is obtained is BT-Settl is used:
  

  Radius

  Only objects with good fit (vgfb<=12) and errors in Parallaxes (TGAS) < 10% (190 objects).

  Excellent agreement between the distances used in the paper and those used in VOSA (from TGAS).
  

  Radius1 (VOSA); defined by: Md = (R/D)^2

  Using Kurucz model we find:

  Radius1 (Yee) - Radius1 (VOSA)

  • Mean: -0.23 Rsun
  • Std: 0.47 Rsun
  • Median: -0.06 Rsun
  
  

  Radius2 (VOSA); defined by: Lbol = 4 * pi * R^2 * σ * Teff^4

  Using Kurucz model we find:

  Radius2 (Yee) - Radius2 (VOSA)

  • Mean: -0.24 Rsun
  • Std: 0.50 Rsun
  • Median: -0.06 Rsun
  
  

  Similar plots are obtained if BT-Settl models are used instead.
  

  Masses

  Only objects with good fit (vgfb<=12). 54 objects (restricted to masses below 1.4 Msun)

  BTSettl isochrones and tracks.

  Excellent agreement for subsolar masses. Masses over 1Msun are overestimated in VOSA-BTSettl.

  Mass(Yee) - Mass (VOSA_BTSettl)

  • Mean: -0.11Msun
  • Std: 0.09Msun
  • Median: -0.13Msun
  
  

  Similar results are obtained if the BTSettl-CFITS isochrones and tracks are used:
  

  Comparison with Lindgren & Heiter 2017 (arXiv170508785L)

  Date of this test: 2017/07/18 (by Miriam Cortés Contreras)

  (Download this test in pdf)

  Summary

  We compare the results in Lindgren & Heiter 2017 (here-after LH17) with the fit results obtained with VOSA.

  • Effective temperatures

    Efective temperatures computed by VOSA are in agreement with those given in LH17. On average, LH17 temperatures are systematically higher by less than 100K both for BT-Settl and CIFITS. Standard deviations are below 150 K in both cases.

    Below 3400 K, LH17 effective temperatures are larger (250 K and 450 K) than those provided by BT-Settl. This trend does not appear if CIFITS models are used. Anyway, a larger number of objects would be neces- sary to confirm this result.

  • Surface gravities, metallicities

    As expected from the minor contribution of these parameters to the SED shape, the values obtained from VOSA are affected by large uncertainties and, thus, are not reliable.

  • Stellar radii

    There are not significant differences between the radii derived using BT-Settl or BT-Settl CIFIST models and both are in very good agreement with the values derived by LH17.

  • Stellar masses

    While masses directly derived from M = gR 2 /G are not reliable due to the large uncertainties associated to the surface gravities estimated with VOSA, those obtained using the BT-Settl and BHAC isochrones are in reasonable agreement with the ones obtained in LH17. The agreement is slightly worse if the BHAC isochrones are used.

  Sample and input parameters

  • Lindgren & Heiter 2017.
    • Parameter determination for sixteen cool dwarfs using high-resolution spectra taken with CRIRES at VLT:
      • J band (1100-1400 nm)
      • R = 50 000
      • SNR: 55-205
    • Stellar properties:
      • Temperatures determined from FeH lines for M dwarfs cooler than 3575 K, and from photometric calibration for warmer stars. 3350 < T_eff[K] < 4550 (±100 K)
      • Metallicities determined using synthetic spectra fitting. -0.50 < [M/H] < +0.40 (±0.05 dex)
      • Spectral types: K4/K5 - M3.5 V
      • Masses derived from the mass-magnitude empirical relation by Benedict et al. (2016). 0.178 < M [M ] < 0.524
      • Radii derived from the mass-magnitude empirical relation by Mann et al. (2015). 0.214 < R [R ] < 0.698
      • Surface gravity (g = GM/R 2 ): 4.56 < log g [cm s-2 ] < 5.03
  • SED building using VOSA.
    • Photometric SED built using photometry from GALEX, Johnson, SDSS, TYCHO, APASS, GAIA, DENIS, 2MASS, WISE, AKARI and IRAS, retrieved from VO services.
    • Model fit using BT-Settl (log g : 4-6; [M/H]: -0.5-0.5, T_eff: 3000 - 5500 K)
    • Model fit using BT-Settl CIFIST (log g : 4 - 6; [M/H] = 0, T_eff: 3000 - 5500 K)

  Parameters determination

  For comparison and to assess whether the parameters obtained with VOSA are model-dependent, we performed this analysis using two models: BT-Settl and BT-Settl CIFIST. One of the sixteen stars has not enough photometric data. Thus, this analysis was carried out for the fifteen remaining stars.

  Effective Temperatures

  • BT-Settl (Fig. 1)

    Mean(T_eff(LH17) - T_eff(VOSA)) = 92.9 K; std = 132.4 K

  • BT-Settl CIFIST (Fig. 2)

    Mean(T_eff(LH17) - T_eff(VOSA)) = 86.3 K; std = 117.2 K

  Both models give consistent values for the effective temperature.
  

  

  
  

  

  Metallicity

  • BT-Settl (Fig. 3)

    Mean(Metallicity(LH17) - Metallicity(VOSA)) = 0.18; std= 0.38

  BT-Settl does not provide good results for the metallicities.
  

  

  Surface gravity

  • BT-Settl (Fig. 4)

    Mean(log g(LH17) - log g(VOSA)) = 0.05; std= 0.61

  • BT-Settl CIFIST (Fig. 5)

    Mean(log g(LH17) - log g(VOSA)) = -0.48; std = 0.35

  Surface gravities provided by VOSA are not consistent with the values given in the paper. Using BT-Settl we obtain higher values for stars with the lowest gravities in LH17 and lower values for the stars with highest gravities (see Fig. 4). On the other hand, this does not happen using BT-Settl CIFIST but we obtain significantly higher values.
  

  

  
  

  

  Radii and masses

  VOSA computes two stellar radii from two different equations: $$ M_d = (R_1 /D)$$ $$ L_{\rm bol} = 4\pi R_2^2 \ \sigma \ T_{\rm eff}^4$$

  where M d is the proportionality factor used to fit the model to the observations, D is the distance and $\sigma$ is the Stephan-Boltzmann constant.

  From $R_1$ and $R_2$ , VOSA provides also stellar masses by applying: $$ g = \frac{GM}{R^2}$$

  Since the surface gravities provided by VOSA do not agree with those given in the paper, we do not expect consistent masses either. In any case, we performed for the masses the same analysis as for the radii and will derive proper masses from the HR diagram.

  • BT-Settl (Figs. 6 and 7)
    • Mean(Radius(LH17) - Radius1(VOSA)) = -0.007; std = 0.033
    • Mean(Radius(LH17) - Radius2(VOSA)) = -0.003; std = 0.032
    • Mean(Mass(LH17) - Mass1(VOSA)) = -0.76; std= 1.59
    • Mean(Mass(LH17) - Mass2(VOSA) = -0.74; std = 1.56
  • BT-Settl CIFIST (Figs. 8 and 9)
    • Mean(Radius(LH17) - Radius1(VOSA)) = -0.003; std = 0.035
    • Mean(Radius(LH17) - Radius2(VOSA)) = -0.002; std = 0.036
    • Mean(Mass(LH17) - Mass1(VOSA)) = -1.69; std = 1.56
    • Mean(Mass(LH17) - Mass2(VOSA) = -1.68; std = 1.56

  There are not significant differences between the radii derived using BT- Settl or BT-Settl CIFIST models. Similar radii are obtained from Eqs. 1 and 2 and both are in very good agreement with the values derived by LH17.

  On the contrary, masses are not consistent with the masses expected for cool dwarfs and, hence, do not agree with those given in the paper, as expected from the log g values obtained with VOSA.
  

  

  
  

  

  
  

  

  
  

  

  Masses from HRD

  • BT-Settl (Fig. 10)

    Mean(Mass(LH17) - Mass(VOSA)) = 0.07; std = 0.06

    Two K dwarfs lie outside the area covered by the isochrone. With a few exceptions, we found good agreement between values for the thirteen remaining dwarfs.

  • BT-Settl CIFIST (Fig. 11)

    Mean(Mass(LH17) - Mass(VOSA)) = 0.08; std = 0.08

    In this case, only one K dwarf lies outside the area covered by the isochrone. The agreement with the masses in LH17 is worse using BHAC isochrones.

  
  

  

  
  

  

  Comparison with 48-Carmencita stars

  Date of this test: 2017/09/22 (by Miriam Cortés Contreras)

  (Download this test in pdf)

  Summary

  We compare the effective temperatures and luminosities derived by Carlos Cifuentes San Román (Master thesis, Sept. 2017, Universidad Complutense de Madrid; hereafter CCSR), and the effective temperatures from Passeger et al. in prep. (hereafter Pass17) with the fit results obtained with VOSA.

  • Effective temperatures

    VOSA provides effective temperatures using BT-Settl models in agreement with the estimated values of CCSR within 200 K. The comparison with the effective temperatures computed by Pass17 results in a higher dispersion. This differences are explained by the differences among CCSR's and Pass17's temperatures (the relation between them gives a correlation coefficient of r=0.88).

  • Luminosities

    Excellent agreement between the bolometric luminosities provided by VOSA and CCSR's.

  Sample and input parameters

  • CCSR
    • Effective temperatures estimation for 48 M dwarfs from their spectral types and low-resolution model spectra.
    • Luminosity determination for 48 M dwarfs given the magnitudes (u) B g V R r i J H K W1 W2 W3 W4 (u only used when available) performing numerical integration via Simpson's rule and Trapezoidal rule.
      • Up to 16 photometric passbands in the range 154 to 22088 nm.W.
      • Spectral types of the sample: M0 V -- M7.0 V.
      • 2600 < T_eff < 4100 K.
      • 0.0007 < L < 0.1162 L_sun.
  • Pass17
    • Effective temperatures for 30 M dwarfs of the previous sample derived using high-resolution spectra taken with FEROS at the 2.2 m of the European Southern Observatory (La Silla, Chile), CAFE and CARMENES at the 2.2 m and 3.5 m telescopes in Calar Alto (Almería, Spain), and HRS at the 9.2 m HET (Texas).
    • 230 < T_eff < 4169 K.
  • SED building using VOSA
    • Photometric SED built using photometry from GALEX, Stromgren, Johnson, SDSS, TYCHO, APASS, Gaia, DENIS, 2MASS, UKIDSS, VISTA, WISE, MSX, IRC and IRAS retrieved from VO services.
    • Model fit using BT-Settl (log{g}: 4.0 - 6.0; [M/H]: -0.5 - 0.5, T_eff: 2300 - 5200 K)

  Parameters determination

  Of the 48 stars in this study, five have not enough photometric points retrieved by VOSA for the fit.

  • Effective temperatures

    To give an idea of the temperatures used for the analysis, the difference between them has a mean value of 58 K and a standard deviation of 111 K.

    • CCSR:

      Mean(T_eff (CCSR) - T_eff (VOSA)) = -7 K; std = 210 K

      Effective temperatures provided by VOSA are in agreement with those derived by CCSR with one exception which effective temperature is 1000 K higher than estimated by CCSR. Fig. 1.

      

    • Pass17

      Mean(T_eff (Pass17) - T_eff (VOSA)) = -21 K; std = 334 K

      In this case, the concordance between temperatures is slightly worse, but also consistent. On average, VOSA provides higher values. Fig. 2.

      

  • Luminosity

    In CCSR, luminosities were derived from two different approaches: via Simpson's rule and Trapezoidal rule. The difference between them has a mean value of 0.00008 L_sun and, therefore, the comparison will be carried out using the luminosities obtained via Trapezoidal rule. The comparison with those obtained via Simpson's rule would be analogue.

    Mean(L (CCSR) - L(VOSA)) = -0.002 L_sun, std= 0.004L_sun.

    The estimated luminosities are in very good agreement. Fig. 3.

    

    Comparison with Rajpurohit et al. 2017

    Date of this test: 2017/09/21 (by Miriam Cortés Contreras)

    (Download this test in pdf)

    Summary

    We compare the results in Rajpurohit et al. 2017, arXiv170806211R (hereafter Ra17) with the fit results obtained with VOSA.

    • Effective temperatures

      Efective temperatures computed by VOSA are in agreement with those given by Rajpurohit et al. (2017) in the studied range from 3000 to 4000 K, with some dispersion towards higher values between 3100 and 3300 K. On average, temperatures provided by VOSA are systematically lower by less than 100 K and standard deviations are below 150 K for both BT-Settl and CIFIST models.

    • Surface gravities, metallicities

      Metallicities and surface gravities provided by VOSA are not reliable due to the minor contribution of these parameters to the SED shape.

    Sample and input parameters

    • Rajpurohit et al. 2017
      • Parameter determination for 45 M dwarfs using spectral synthesis employing BT-Settl models and high-resolution spectra taken with APOGEE on the Sloan 2.5 m Telescope at Apache Point Observatory in the H band:
        • H band (1.51 - 1.7 μm)
        • R ~ 22 500
      • Stellar properties:
        • Spectral types: M1.0 - M8.0 V
        • 3100 < T_eff [K] < 3900 (± 100 K)
        • -0.50 < [Fe/H] < +0.50 with errors between 0.03 and 0.11.
        • 4 < logg [cm s²] < 5.5 with errors between 0.2 and 0.5.
    • SED building using VOSA
      • Photometric SED built using photometry from GALEX, Johnson, SDSS, APASS, Gaia, IPHAS, DENIS, UKIDSS, 2MASS, WISE and AKARI, retrieved from VO services.
      • Model fit using BT-Settl (logg: 4 - 5.5, [Fe/H]: -0.5 - 0.5, T_eff:3000 - 4000 K).
      • Model fit using BT-Settl CIFIST (logg: 4 - 5.5, [M/H] = 0, T_eff:3000-4000 K).

    Parameters determination

    We performed this analysis using BT-Settl models and used the more recent BT-Settl CIFIST models for comparison. Of the 45 M dwarfs of the analysis, only four had parallactic distances retrieved from VO services and another six had not enough photometric data for the fit.

    • Effective Temperatures
      • BT-Settl (Fig. 1)

        Mean(T_eff (Ra17) - T_eff (VOSA)) = 20.5 K; std = 111.4 K

      • BT-Settl CIFIST (Fig. 2)

        Mean(T_eff (Ra17) - T_eff (VOSA)) = 18.0 K; std = 110.6 K

      Both models provide quite similar values for the effective temperatures and are overall consistent within the errorbars with those given in Rajpurojit et al. (2017).

      

      

    • Metallicity
      • BT-Settl (Fig. 3)

        Mean(Metallicity(Ra17) - Metallicity(VOSA)) = -0.19; std= 0.41

      No good determination of the metallicities using VOSA. On average, metallicities obtained using BT-Settl models differ with the values given by Rajpurohit et al. (2017) by more than 7σ.

      

    • Surface gravity
      • BT-Settl (Fig. 4)

        Mean(logg (Ra17) - logg (VOSA)) = 0.36; std= 0.65

      • BT-Settl CIFIST (Fig. 5)

        Mean(logg (Ra17) - logg (VOSA)) = 0.29; std = 0.56

      The surface gravities given by VOSA strongly differ with those given by Rajpurohit et al. (2017) for near half of the analyzed sample. Hence, these values are not trustworthy.

      

      

    Use Case

    VO photometry

    (Click here for a short link to this documentation section)