REVISED STELLAR PROPERTIES OF KEPLER TARGETS FOR THE QUARTER 1–16 TRANSIT DETECTION RUN

The first step in the classification process was to discern dwarfs from giant stars. As described by Batalha et al. (2010) and Huber et al. (2013b), asteroseismology is an efficient tool to identify giant stars using Kepler data since the oscillation amplitudes are large enough to be detectable independent of shot noise, and because oscillation timescales are long enough to be measurable with long-cadence data. We have analyzed Q0–14 long-cadence data of 13,420 unclassified stars using the asteroseismic detection pipelines described by Huber et al. (2009), Hekker et al. (2010), Mosser & Appourchaux (2009) and Mathur et al. (2010). We have used simple-aperture photometry data as opposed to the Pre-search Data Conditioning data (Smith et al. 2012; Stumpe et al. 2012) in order to preserve long-periodic oscillations typical for high-luminosity giant stars. Instrumental trends such as inter-quarter flux discontinuities and pixel-sensitivity dropouts were corrected by fitting linear functions to the start and end of each subset, and by applying a quadratic Savitzky–Golay filter (Savitzky & Golay 1964) with a width of 20 days. In a few stars, we have also applied the procedures described in García et al. (2011) to double check that features in the light curves were not a consequence of the correction procedures and hence verify the reliability of the seismic solutions.

Of the 13420 stars analyzed, 3114 targets showed oscillations that classified them as giants. A subset of 1760 of these targets have been observed for more than nine quarters between Q1–16, with 1176 targets of this subset having been observed continuously for the entire mission. The new detections presented here raise the number of oscillating giant stars detected by Kepler by ∼20% over previous detections (Hekker et al. 2011; Stello et al. 2013) to a total of ∼15, 500 stars. Importantly, the newly detected oscillating giants predominantly have surface gravities well below the red clump (log  g < 2, see Section 5.3). Such stars are underrepresented in previous asteroseismic samples and provide the opportunity to study oscillations in late stages of stellar evolution such as the tip of the red giant branch and the asymptotic giant branch (Bányai et al. 2013; Mosser et al. 2013). Our analysis also yielded 475 classical pulsators, such as γ Doradus and δ Scuti stars, which were identified using the automated classification pipeline by Debosscher et al. (2011).

Figure 2 shows 2MASS J − H versus H − K diagrams for the unclassified sample considered in this study. Note that we have rejected stars without AAA-quality 2MASS photometry from the catalog, reducing the initial sample to 11,191 stars (2762 with asteroseismic detections). Figure 2(a) shows the unclassified sample without asteroseismic detections. The blue lines show near-solar metallicity DSEP isochrones described in Section 3. The observations are offset from the models due to reddening (see red arrow in Figure 2), illustrating the importance of extinction when deriving temperatures for unclassified Kepler stars. Note that dwarfs and giants separate for H − K ≳ 0.1 (corresponding roughly to spectral type K5), with cooler dwarfs retaining a roughly constant J − H color. For hotter stars, however, the colors of dwarfs and giants overlap and an independent luminosity classification, e.g., from asteroseismology or spectroscopy, is required.

Zoom In Zoom Out Reset image size

Figure 2(b) shows the same diagram but for unclassified stars with asteroseismic detections. As expected, asteroseismic detections are mainly found in cool stars in the top right section of the plot. Three typical examples of oscillating giants are marked, and their power spectra showing the presence of convection-driven oscillations are illustrated in Figure 3. Note that as giants become cooler and more luminous (larger J − H and H − K colors), their oscillation periods and amplitudes increase.

Zoom In Zoom Out Reset image size

Notably, there is a significant fraction of stars in Figure 2(a) for which an asteroseismic detection would have been expected based on their 2MASS colors. Reasons for an asteroseismic non-detection in giants include that the star is too cool and too evolved, resulting in pulsation periods that are too long for an unambiguous detection with 14 quarters of long-cadence data. Similarly, many unclassified stars have not been observed for the full mission duration, and hence the data provide lower frequency resolution. This discussion implies that a seismic detection is strong evidence that a star is a giant, but a seismic non-detection does not necessarily imply that a star is a dwarf. We therefore applied an additional color cut of H − K > 0.1 and J − H > 0.75 (see red box in Figure 2(a)) to identify giants based on 2MASS colors only. Using this procedure, we classify a total of 3302 giants (log  g < 3.5) and 7889 subgiant or dwarf stars (log  g > 3.5) in the unclassified sample.

To test our luminosity classifications, we restricted our sample to the same color-cut applied by Mann et al. (2012) (Kp − J > 2, Kp < 14) and found a total fraction of giant stars (selected using the asteroseismic classifications and 2MASS color cuts) of 90% using all stars, and 92% using only stars with a full set of Kepler data. This compares reasonably well to the giant fraction of 96% found by Mann et al. (2012) based on photometric and spectroscopic classifications. The remaining difference could be due to the fact that Mann et al. (2012) considered both classified and unclassified stars, or implies that our crude 2MASS color cut to identify giant stars is too conservative. We also note that, although all seismic detections were checked by eye, it cannot be excluded that the unclassified seismic sample includes a small fraction of outliers which were erroneously classified as giants (for example due to blends). Future work including proper motion measurements will enable an improved giant-dwarf discrimination for Kepler targets.

We determined effective temperatures for unclassified stars by comparing observed colors to theoretical values of the DSEP model grid described in Section 3. As shown in Figure 2, interstellar extinction is significant in this sample. To ensure consistency with the other Kepler targets, we adopted the same reddening model as applied in the KIC (see Section 2). For each star, we calculated the distance corresponding to each model J-band absolute magnitude and the apparent J-band magnitude, and determined the reddening at the distance and galactic latitude of the star. Note that we did not restrict reddening to a maximum value. This process was repeated until convergence in distance was reached. We adopted the reddening law by Cardelli et al. (1989), with A_V/A_J = 0.29. For each star, models were restricted to a given metallicity (see next section).

The best-fitting model was identified by finding the closest matching model color to the observed color. Colors were matched to models depending on the spectral type of the star. Figure 4 shows the relation between J − K, H − K and g − i to effective temperature for DSEP models of dwarfs (black) and giants (red). J − K provides the best thermometer for warmer stars with T_eff > 4500 K. For cooler dwarfs, however, J − K becomes insensitive to T_eff. H − K shows some sensitivity, but only over a color span of ∼0.1 mag, which is relatively small compared to typical errors of 0.03 mag in 2MASS colors. We therefore adopted a third color based on g − i from the Kepler-INT survey (Greiss et al. 2012). Note that the Kepler-INT Sloan photometry is in the Vega system, which we converted into the AB system using the transformations by González-Solares et al. (2011). For stars between 3300–4500 K, temperatures were then derived from g − i when available, and otherwise from H − K. Finally, for the coolest dwarfs (<3300 K), we applied average polynomial relations from BT–Settl models, as described in Section 3.

Zoom In Zoom Out Reset image size

To test the derived effective temperatures, we applied the same procedure to a subset of stars with available temperatures from Dressing & Charbonneau (2013) for dwarfs with T_eff < 4500 K and P12 for dwarfs with T_eff > 4500 K. The result of this comparison is shown in Figure 5(a). For stars with T_eff < 4500 K the agreement between the temperatures is excellent, implying that Kepler-INT and KIC Sloan colors agree well for these stars. For hotter stars the agreement is good for T_eff ≲ 5800 K, but we observe an increasing systematic offset for hotter stars with DSEP temperatures being ∼500 K hotter than P12 temperatures at T_eff ∼ 6500 K. This offset is caused by the fact that a single color (J − K, H − K, or g − i) does not contain independent information on temperature and reddening (or log g), introducing a bias toward hotter, more luminous stars evolving off the main sequence. Since temperature and reddening are degenerate, more luminous and distant stars with greater reddening and lower log g tend to be selected.

Zoom In Zoom Out Reset image size

To correct for this bias and ensure consistency of the unclassified stars with the remaining sample, we apply an ad hoc correction for hot stars with the same form as adopted by P12:

$$\begin{equation} T_{\rm eff,cor} = 5800\,\rm {K}+0.6\; ({T_{\rm eff}}-5800\,\rm {K}) \:, \end{equation} \tag{ 1 }$$

for all stars with T_eff > 5800 K.

Figure 5(b) shows a comparison of DSEP temperatures with temperatures from P12 for a sample of giant stars. We observe an offset with DSEP temperatures being significantly (∼80 K) cooler than the comparison values. This is consistent with the offset between temperatures derived from the J − K infrared flux method and temperatures based on Sloan colors discussed by P12. Since the offset is generally within the adopted uncertainties, we did not apply a systematic correction to temperatures of unclassified giant stars.

The determination of surface gravities and metallicities from broadband colors is a notoriously ill-posed problem due to strong degeneracies between fitted parameters. Our attempts to perform direct fits of observed colors to model colors quickly showed that many stars would be frequently matched with physical parameters that are unlikely to occur, such as massive subgiants in short-lived phases of stellar evolution.

Following B11, we therefore adopted priors on both surface gravity and metallicity. The log g prior was constructed using stars in the Hipparcos catalog (van Leeuwen 2007) with distances <100 pc and fractional parallax uncertainties <10%. We first calculated a grid with a stepsize of 0.05 mag in (B − V) and 0.25 mag in M_V (calculated from the parallax assuming A_V = 0) and counted the number of Hipparcos stars in each grid cell. For each DSEP model at a given metallicity, we then found the grid cell containing the (B − V) color and M_V magnitude of that model and assigned a prior probability corresponding to the number density of Hipparcos stars in that cell. Figure 6(a) shows the resulting log g prior for a typical solar neighborhood metallicity. For the [Fe/H] prior we adopted the same analytic function as used for the KIC, which was constructed from metallicities in the Geneva–Copenhagen survey (Nordström et al. 2004). We note that the priors have been purposefully chosen to be very similar to those adopted by B11 to minimize biases between the unclassified stars and the remaining sample.

Zoom In Zoom Out Reset image size

Incorporating these priors still resulted in distributions of stars that were unrealistically narrow, being confined to the peak of the prior distribution at a fixed temperature. This confirmed that the adopted colors (J − K, H − K, g − i) have little sensitivity to either log g or [Fe/H]. To arrive at more realistic distributions, we calculated for each star a one-dimensional prior probability distribution in bins of 0.01 dex in log g around a slice of 50 K centered on the T_eff determined in the previous section. We then drew a log g value from this distribution with a probability corresponding to the prior value at a given log g. The same procedure was applied to assign metallicities for a given star using the B11 metallicity prior. Note that for giant stars with asteroseismic detections, surface gravities were calculated from the measured frequency of maximum oscillation power, and hence only metallicities were assigned in this manner.

The resulting log g versus T_eff distribution of the unclassified sample is shown in Figure 6(b). By construction, the distribution in log g closely follows the prior distribution shown in Figure 6(a). We stress that the procedure described in the previous paragraph means that log g and/or [Fe/H] for unclassified stars are only statistically accurate, but are drawn from a prior probability on a star-by-star basis. The properties of these stars (except for temperatures) are therefore not suitable for scientific analyses on a star-by-star basis, and we strongly encourage follow-up observations for stars of particular interest (e.g., if planet candidates are detected). Future efforts will improve the properties for these stars by using proper motions or additional colors that contain independent information on surface gravities and metallicities.

Figure 6(b) also shows that there is a significant fraction of ultra cool dwarfs in the unclassified sample, which form a narrow band of stars with T_eff < 3300 K. We recall that this discrete distribution is caused by the fact that we adopted typical stellar properties for ultra cool dwarfs based on polynomial fits to BT–Settl models (see Section 3). We note that ∼100 of these stars were matched to the coolest end of the BT–Settl polynomials due to very red H − K colors (see Figure 2), and were excluded from Figure 6(b) to avoid biasing the color scale. It is likely that a fraction of these ultra-cool dwarfs are giants, and we emphasize that the properties for these stars should be used with caution.

The procedures described in Sections 4 and 5 yielded input values for T_eff, log g and [Fe/H] for a total of 196,468 stars. Prior to fitting these constraints to models, uncertainties need to be specified for each input value. Since uncertainties quoted in the literature are heterogeneous, we adopted typical uncertainties on each parameter depending on the observational method with which the parameter was derived.

Asteroseismic surface gravities have been shown to be accurate to at least 0.03 dex (Creevey et al. 2013; Morel & Miglio 2012; Hekker et al. 2013), which we adopted as a typical uncertainty independent of the evolutionary state of the star. Transit-derived densities have proven to be in good agreement with asteroseismic densities (e.g., Nutzman et al. 2011) but are based on an implicit assumption of circular orbits, and we hence assigned a slightly more conservative uncertainty of 0.05 dex. Typical spectroscopic uncertainties of 2% in T_eff, 0.15 dex in log g, and 0.15 dex in [Fe/H] are based on the comparison of spectroscopic analyses with and without asteroseismic constraints (Huber et al. 2013b).

To estimate typical uncertainties for photometric methods, we have compared published results for a sample with combined spectroscopic and asteroseismic constraints to properties given in the KIC and P12 temperatures. The result is shown in Figure 7. The median and scatter of the residuals are +130 ± 120 K for the P12 temperatures (panel a), −130 ± 140 K for KIC temperatures (panel b), +0.13 ± 0.33 dex for KIC log g, and −0.15 ± 0.31 for KIC [Fe/H]. Based on these residuals and previous estimates of uncertainties of KIC properties (e.g., Bruntt et al. 2011), we assigned uncertainties of 3.5% in T_eff, 0.4 dex in log g, and 0.3 dex in [Fe/H]. We note that the P12 temperatures contain homogeneously derived uncertainties based on the quality of the input photometry. To preserve this information, uncertainties were calculated by adding a 2.5% systematic error in quadrature to the formal uncertainty given by P12. The typical adopted uncertainties are given in Table 2. Table 3 lists the reference key for the literature sources of the input values, and Table 4 lists the input values with adopted uncertainties for the whole catalog (see Section 6.5 for more details on provenances and reference keys).

Zoom In Zoom Out Reset image size

Table 2. Uncertainties Adopted for the Input Parameters

Method	$\sigma _{{T_{\rm eff}}}$	σlog  g	σ[Fe/H]
	(%)	(dex)	(dex)
Asteroseismology	⋅⋅⋅	0.03	⋅⋅⋅
Transits	⋅⋅⋅	0.05	⋅⋅⋅
Spectroscopy	2	0.15	0.15
Photometry	3.5	0.40	0.30
KIC	3.5	0.40	0.30

Note. An error floor of 80 K for spectroscopy and 100 K for photometry has been adopted for effective temperatures.

Download table as:  ASCIITypeset image

Table 4. Consolidated Input Values

KIC	Teff	log g	[Fe/H]>	$P_{{T_{\rm eff}}}$	Plog  g	P[Fe/H]
757076	5164 ± 154	3.601 ± 0.400	−0.083 ± 0.300	PHO1	KIC0	KIC0
757099	5521 ± 168	3.817 ± 0.400	−0.208 ± 0.300	PHO1	KIC0	KIC0
757137	4751 ± 139	2.378 ± 0.030	−0.079 ± 0.300	PHO1	AST9	KIC0
757280	6543 ± 188	4.082 ± 0.400	−0.231 ± 0.300	PHO1	KIC0	KIC0
757450	5330 ± 106	4.500 ± 0.050	−0.070 ± 0.150	SPE51	TRA51	SPE51
891901	6325 ± 186	4.411 ± 0.400	−0.084 ± 0.300	PHO1	KIC0	KIC0
891916	5602 ± 165	4.591 ± 0.400	−0.580 ± 0.300	PHO1	KIC0	KIC0
892010	4834 ± 151	2.163 ± 0.030	0.207 ± 0.300	PHO1	AST9	KIC0
892107	5086 ± 161	3.355 ± 0.400	−0.085 ± 0.300	PHO1	KIC0	KIC0
892195	5521 ± 184	3.972 ± 0.400	−0.054 ± 0.300	PHO1	KIC0	KIC0
⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
1429653	6636 ± 225	4.622 ± 0.400	0.239 ± 0.300	PHO1	KIC0	KIC0
1429729	3903 ± 136	4.735 ± 0.400	−0.200 ± 0.300	PHO2	PHO2	PHO2
1429751	6000 ± 185	4.420 ± 0.400	−0.012 ± 0.300	PHO1	KIC0	KIC0
1429795	5772 ± 164	4.504 ± 0.400	−0.104 ± 0.300	PHO1	KIC0	KIC0
1429893	5068 ± 143	4.583 ± 0.400	−0.071 ± 0.300	PHO1	KIC0	KIC0
1429921	4356 ± 125	4.723 ± 0.400	−0.254 ± 0.300	PHO1	KIC0	KIC0
1429977	5155 ± 180	4.333 ± 0.400	−0.465 ± 0.300	KIC0	KIC0	KIC0
1430118	5070 ± 155	3.124 ± 0.030	−0.137 ± 0.300	PHO1	AST9	KIC0
1430163	6520 ± 130	4.221 ± 0.030	−0.110 ± 0.150	SPE6	AST10	SPE6
1430171	4771 ± 166	4.559 ± 0.400	−0.063 ± 0.300	KIC0	KIC0	KIC0
⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅

Notes. Note that all uncertainties were assigned typical fractional or absolute values for a given method, as listed in Table 2. Provenance abbreviations: KIC = Kepler Input Catalog, PHO = Photometry, SPE = Spectroscopy, AST = Asteroseismology, TRA = Transits. The number at the end of each provenance denotes the reference key, as given in Table 3.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

The input values in Table 4 were fitted to the grid of DSEP isochrones to derive radii, masses, densities, and luminosities. Matching observations to stellar isochrones is a non-trivial task, with important systematics such as the terminal age bias (Pont & Eyer 2004; Jørgensen & Lindegren 2005; Casagrande et al. 2011; Serenelli et al. 2013). We adopted an approach following Kallinger et al. (2010) and calculated for each star:

$$\begin{equation} \mathcal {L}_{X} = \frac{1}{\sqrt{2\pi }\sigma _{X}} \exp {\left(\frac{-(X_{\rm obs}-X_{\rm model})^2}{2\sigma _{X}^2}\right)}, \end{equation} \tag{ 2 }$$

where X = {T_eff, log  g, [Fe/H]} are assumed to be independent Gaussian observables. The combined likelihood is:

$$\begin{equation} \mathcal {L} = \mathcal {L}_{T_{\rm eff}} \mathcal {L}_{{\log\; g}} \mathcal {L}_{\rm {Fe/H}} \:. \end{equation} \tag{ 3 }$$

For each parameter, Equation (3) yields a probability distribution that was used to calculate the best-fit, median and 68% (1σ) intervals.

The reported value for each stellar property in the catalog corresponds to the best-fitting model, which was determined by maximizing Equation (3). To specify an error bar as a single number (as required by the Kepler planet-detection pipeline), we reported the largest distance of the best fit to the upper or lower limit of the 1σ interval around the median of the probability distribution. For highly asymmetric distributions this procedure results in conservative estimates, as further discussed below. We have also derived a second set of uncertainties by calculating the largest difference of the best-fit value to the lower or upper limit of the closest 1σ interval around the best fit.

Figure 8 shows histograms of the derived uncertainties for T_eff, log g, [Fe/H], R, M and ρ. As expected, the uncertainties on T_eff, log g, [Fe/H] largely follow the distribution of uncertainties adopted on the input parameters. Uncertainties based on the 1σ interval around the median (dashed lines) are systematically larger than estimates based on the 1σ interval around the best fit (solid lines) for higher input uncertainties in log g, radius and density. This is due to asymmetric probability distributions for main-sequence stars whose initial log g estimates were based on photometry: assuming a 0.4 dex uncertainty, the finite extent of the isochrone grid causes a sharp cut-off at large log g values. For smaller fractional uncertainties the distributions become more symmetric, and the two uncertainty estimates agree better. Temperature and mass mostly yield symmetric distributions, and so this effect does not arise. Median uncertainties in radius for the full sample over both methods span 40%–55% in radius, and 60%–75% in density. For stellar mass the typical uncertainty is ∼20%, confirming that mass is mainly constrained by T_eff and less affected by large uncertainties in log g. Based on the above discussion, we conclude that the uncertainty estimates based on the central 1σ interval (which were used in the Kepler planet-detection pipeline) are probably conservative, especially for main-sequence stars whose input surface gravities were based on photometry.

Zoom In Zoom Out Reset image size

It is informative to analyze the relative radius uncertainty as a function of effective temperature, as shown in Figure 9(a) (see also Gaidos & Mann 2013). Note that we show the uncertainties based on the 1σ interval around the best fit here, but the comparison is qualitatively similar for the uncertainties based on the 1σ interval around the median. For G-type dwarfs the median radius uncertainty based on photometry is 40%, increasing to higher values for more massive dwarfs. The "band" of points with large uncertainties at T_eff ∼ 5000 K is due to K dwarfs in the "No-Man's-Land" zone (see Section 8), which have highly bimodal radius distributions reaching from the main-sequence to the subgiant branch. For even cooler stars, this subgiant degeneracy disappears, and the much slower evolution of stars constrains the radius to typical uncertainties of 20%. We emphasize that these uncertainties do not include potential systematic errors in the models, which can be significant particularly for cool dwarfs (see, e.g., Boyajian et al. 2012). Dashed–dotted and dashed boxes highlight the smaller subset of stars with constraints from asteroseismology and spectroscopy, respectively. As expected, the relative uncertainties are smaller, making targets with asteroseismic measurements the best characterized stars in the Kepler field.

Zoom In Zoom Out Reset image size

Figure 9(b) shows the relative radius uncertainty distribution for giant stars. Stars with asteroseismic measurements dominate this sample, with typical relative uncertainties in radius of ∼30%. We emphasize that such uncertainties are atypically large since we have only used asteroseismic constraints on log g and ignored any information on the mean density, which is typically much better constrained. Ongoing projects aimed at combining APOGEE H-band spectra (Eisenstein et al. 2011) with seismic constraints for Kepler giants (the APOKASC project, see Mészáros et al. 2013; M. Pinsonneault et al. 2014, in preparation) will soon provide much improved radii, masses, and ages of oscillating giants in the Kepler field.

We have compared our catalog results with published radii and masses derived using different methods and models. Figure 10 shows a comparison for confirmed Kepler planet host stars taken from the NASA exoplanet archive²⁷ (black), as well as the larger sample of planet-candidate host stars by Buchhave et al. (2012) (red) for stars with relative uncertainties better than 20%. In both cases the majority of the radii and masses were derived using Yonsei–Yale (YY) evolutionary tracks (Yi et al. 2001). Overall the residuals show an offset of 1% with a scatter of 7% for radius and an offset of 3% with a scatter of 6% for mass. These offsets, which are more pronounced for the sample by Buchhave et al. (2012), are likely due to differences in the interior models and assumptions of uncertainties on the input values. We also observe systematic differences at the low-mass end (≲ 0.8 M_☉), resulting in a "kink" with higher DSEP masses and radii between ∼0.6–0.8 M_☉, and lower DSEP masses and radii for ≲ 0.6 M_☉. This is consistent with systematic differences between DSEP and YY models due to different equations of state adopted for low-mass stars (Dotter et al. 2008).

Zoom In Zoom Out Reset image size

As an additional test we calculated radii and masses from T_eff, log g and [Fe/H] using the empirical relations by Torres et al. (2010), which were calibrated using a large sample of detached eclipsing binary systems. A comparison with the radii and masses derived in this work is shown in Figure 11 for the mass range in which the Torres et al. (2010) relations are valid. We observe that the catalog radii and masses are systematically smaller by ∼5% than the empirical values. As noted by Torres et al. (2010), a similar offset in mass is found when comparing empirically calculated values to YY isochrone values by Valenti & Fischer (2005), or when using observed values for the Sun. Importantly, the offset to the Q1–Q16 catalog does not vary with stellar mass and radius, and is typically well within the quoted uncertainties.

Zoom In Zoom Out Reset image size

The primary motivation for fitting T_eff, log g and [Fe/H] to a single set of isochrones was to ensure a homogeneous treatment for all stars. However, methods such as asteroseismology provide significantly better constraints on other stellar properties such as the mean stellar density. Hence, omitting such additional information can yield significantly less accurate stellar radii and masses. More importantly, some of the studies used in the consolidation of literature values adopted the same models as in this study, hence removing the need to re-fit the parameters to ensure consistency with the remaining sample. For these reasons, we have adopted literature values for all stars with published masses, radii, and densities that are based on DSEP models. These studies include Dressing & Charbonneau (2013) and Muirhead et al. (2012a) for M dwarfs, and Chaplin et al. (2014) and Huber et al. (2013b) for F-G dwarfs with asteroseismic measurements. Note that the latter two studies used a large variety of models including DSEP, hence providing an even more robust estimate of the uncertainties on stellar properties.

The complete Q1–Q16 star properties catalog is presented in Table 5. For each star we list the best-fitting T_eff, log g, [Fe/H], radius, mass and density, together with the uncertainty based on the 1-σ interval around the best fit, as described in Section 6.2.²⁸ For stars with published masses, radii, and densities based on DSEP models, stellar properties and uncertainties as given in the literature are listed (see Section 6.4). Each entry contains provenance flags specifying the origin of the input T_eff, log g, and [Fe/H]. The provenance consists of a three letter abbreviation of the method used to derive the parameter and a number specifying the reference from which the parameter was adopted. The abbreviations are as follows (see also Section 4):

• 1.  
  AST = Asteroseismology
• 2.  
  TRA = Transits
• 3.  
  SPE = Spectroscopy
• 4.  
  PHO = Photometry
• 5.  
  KIC = Kepler Input Catalog.

Table 5. Q1–Q16 Star Properties Catalog. An interactive version of this table is available at the NASA Exoplanet Archive: http://exoplanetarchive.ipac.caltech.edu/.

KIC	Stellar Properties						Provenances
	Teff	log g	[Fe/H]	R(R☉)	M(M☉)	ρ (g cm−3)	$P_{{T_{\rm eff}}}$	Plog  g	P[Fe/H]	PM, R, ρ
757076	$5160^{+138}_{-163}$	$3.580^{+0.274}_{-0.294}$	$-0.100^{+0.260}_{-0.300}$	$3.13^{+1.41}_{-1.03}$	$1.36^{+0.32}_{-0.40}$	$0.062^{+0.112}_{-0.040}$	PHO1	KIC0	KIC0	DSEP
757099	$5519^{+183}_{-168}$	$3.822^{+0.501}_{-0.276}$	$-0.220^{+0.340}_{-0.280}$	$2.11^{+1.10}_{-1.02}$	$1.08^{+0.33}_{-0.20}$	$0.16^{+0.83}_{-0.11}$	PHO1	KIC0	KIC0	DSEP
757137	$4706^{+81}_{-103}$	$2.374^{+0.029}_{-0.027}$	$-0.100^{+0.280}_{-0.340}$	$15.45^{+3.57}_{-4.60}$	$2.06^{+1.15}_{-1.05}$	$0.00079^{+0.00035}_{-0.00014}$	PHO1	AST9	KIC0	DSEP
757280	$6543^{+155}_{-206}$	$4.082^{+0.228}_{-0.266}$	$-0.240^{+0.240}_{-0.300}$	$1.64^{+0.82}_{-0.46}$	$1.18^{+0.30}_{-0.17}$	$0.38^{+0.50}_{-0.24}$	PHO1	KIC0	KIC0	DSEP
757450	$5332^{+102}_{-98}$	$4.500^{+0.043}_{-0.040}$	$-0.080^{+0.160}_{-0.120}$	$0.843^{+0.051}_{-0.044}$	$0.821^{+0.060}_{-0.040}$	$1.93^{+0.30}_{-0.26}$	SPE51	TRA51	SPE51	DSEP
891901	$6324^{+153}_{-211}$	$4.356^{+0.085}_{-0.327}$	$-0.100^{+0.220}_{-0.300}$	$1.15^{+0.67}_{-0.14}$	$1.08^{+0.28}_{-0.11}$	$1.01^{+0.39}_{-0.71}$	PHO1	KIC0	KIC0	DSEP
891916	$5602^{+183}_{-148}$	$4.587^{+0.039}_{-0.218}$	$-0.580^{+0.360}_{-0.280}$	$0.741^{+0.286}_{-0.056}$	$0.773^{+0.109}_{-0.061}$	$2.68^{+0.48}_{-1.50}$	PHO1	KIC0	KIC0	DSEP
892010	$4729^{+70}_{-182}$	$2.168^{+0.032}_{-0.027}$	$0.070^{+0.140}_{-0.470}$	$26.09^{+0.44}_{-8.70}$	$3.652^{+0.018}_{-2.031}$	$0.000290^{+0.000158}_{-0.000018}$	PHO1	AST9	KIC0	DSEP
892107	$5080^{+114}_{-155}$	$3.354^{+0.261}_{-0.299}$	$-0.080^{+0.220}_{-0.320}$	$4.29^{+2.02}_{-1.47}$	$1.52^{+0.37}_{-0.52}$	$0.027^{+0.047}_{-0.018}$	PHO1	KIC0	KIC0	DSEP
892195	$5522^{+190}_{-157}$	$3.984^{+0.399}_{-0.294}$	$-0.060^{+0.280}_{-0.260}$	$1.67^{+0.97}_{-0.65}$	$0.98^{+0.25}_{-0.10}$	$0.30^{+0.93}_{-0.20}$	PHO1	KIC0	KIC0	DSEP
...	...	...	...	...	...	...	...	...	...	...
1429653	$6622^{+174}_{-314}$	$4.331^{+0.069}_{-0.342}$	$0.210^{+0.150}_{-0.430}$	$1.32^{+0.76}_{-0.21}$	$1.37^{+0.22}_{-0.28}$	$0.83^{+0.32}_{-0.59}$	PHO1	KIC0	KIC0	DSEP
1429729	$3903^{+76}_{-60}$	$4.735^{+0.060}_{-0.070}$	$-0.200^{+0.200}_{-0.100}$	$0.523^{+0.070}_{-0.050}$	$0.541^{+0.070}_{-0.050}$	$5.33^{+2.25}_{-2.25}$	PHO2	PHO2	PHO2	DSEP2
1429751	$6000^{+156}_{-194}$	$4.415^{+0.070}_{-0.287}$	$-0.020^{+0.220}_{-0.300}$	$1.05^{+0.45}_{-0.12}$	$1.04^{+0.20}_{-0.12}$	$1.27^{+0.40}_{-0.81}$	PHO1	KIC0	KIC0	DSEP
1429795	$5768^{+153}_{-150}$	$4.501^{+0.045}_{-0.288}$	$-0.100^{+0.260}_{-0.280}$	$0.906^{+0.377}_{-0.080}$	$0.950^{+0.110}_{-0.098}$	$1.80^{+0.38}_{-1.14}$	PHO1	KIC0	KIC0	DSEP
1429893	$5066^{+158}_{-130}$	$4.578^{+0.035}_{-0.090}$	$-0.080^{+0.320}_{-0.260}$	$0.754^{+0.120}_{-0.059}$	$0.785^{+0.097}_{-0.071}$	$2.58^{+0.45}_{-0.76}$	PHO1	KIC0	KIC0	DSEP
1429921	$4358^{+127}_{-136}$	$4.673^{+0.042}_{-0.051}$	$-0.260^{+0.320}_{-0.320}$	$0.606^{+0.061}_{-0.055}$	$0.629^{+0.056}_{-0.063}$	$3.98^{+0.84}_{-0.75}$	PHO1	KIC0	KIC0	DSEP
1429977	$5166^{+192}_{-158}$	$4.554^{+0.071}_{-0.848}$	$-0.440^{+0.340}_{-0.260}$	$0.728^{+1.287}_{-0.076}$	$0.692^{+0.185}_{-0.047}$	$2.52^{+0.78}_{-2.39}$	KIC0	KIC0	KIC0	DSEP
1430118	$5094^{+82}_{-167}$	$3.126^{+0.029}_{-0.030}$	$-0.100^{+0.180}_{-0.340}$	$6.25^{+0.51}_{-1.61}$	$1.90^{+0.26}_{-0.85}$	$0.0110^{+0.0039}_{-0.0012}$	PHO1	AST9	KIC0	DSEP
1430163	$6520^{+84}_{-84}$	$4.221^{+0.013}_{-0.014}$	$-0.110^{+0.090}_{-0.090}$	$1.480^{+0.030}_{-0.030}$	$1.340^{+0.060}_{-0.060}$	$0.577^{+0.024}_{-0.025}$	SPE6	AST10	SPE6	MULT10
1430171	$4771^{+185}_{-155}$	$4.566^{+0.050}_{-0.051}$	$-0.060^{+0.300}_{-0.280}$	$0.729^{+0.073}_{-0.067}$	$0.714^{+0.092}_{-0.058}$	$2.60^{+0.60}_{-0.46}$	KIC0	KIC0	KIC0	DSEP
...	...	...	...	...	...	...	...	...	...	...

Notes. Reported are for each parameter the best-fitting value and the lower and upper limit of the 68% interval closest to the best fit; Provenance abbreviations: KIC = Kepler Input Catalog, PHO = Photometry, SPE = Spectroscopy, AST = Asteroseismology, TRA = Transits, DSEP = Based on Dartmouth models, MULT = Based on multiple models (including DSEP). Uncertainties set to zero indicate that no uncertainty estimate is available. The number at the end of each provenance denotes the reference key, as given in Table 3.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

In addition to the provenances for T_eff, log g, [Fe/H], Table 5 also lists a provenance for the source of interior parameters (R, M and ρ). The abbreviations are as follows:

• 1.  
  DSEP = Derived from the Dartmouth Stellar Evolution Program Models
• 2.  
  MULT = Derived from multiple evolutionary tracks/isochrones, including DSEP.

Note that entries specifying DSEP without a reference number correspond to values derived with the model grid presented in this work. Interior model flags with reference numbers correspond to entries which were replaced by published solutions (see Section 6.4).

The reference key is provided in Table 3. Using the three letter abbreviations described above and the reference number, each parameter is directly traceable to a single reference and method, and subsets of stars can be filtered according to individual methods or references. For example, restricting the sample to provenances containing SPE+AST+SPE for T_eff+log g+[Fe/H] will extract the "gold-standard" sample of stars with combined asteroseismic and spectroscopic constraints.

The stellar properties presented in this paper have been adopted in the Q1–Q16 transit detection run described in Tenenbaum et al. (2014). We note that the uncertainties adopted in that run (and hence displayed in Kepler pipeline products such as data validation reports at the NASA Exoplanet Archive) differ from those reported in Table 5 for reasons described in Section 6.2. We emphasize, however, that uncertainties on stellar properties are not currently used in the Kepler pipeline, and hence this difference does not affect the results displayed in the data validation reports. We also note that targets that have only been observed in Q0 (commissioning) are not analyzed in the transit detection run, and hence have also not been included in this catalog. Additionally, ∼2000 stars which are unclassified in the KIC and do not have AAA-quality 2MASS photometry remain unclassified in this work, and hence have not been included in the catalog (see Section 5).

The general trends and biases in the sample can be summarized as follows:

• 1.  
  For ∼70% of all stars the input log g and [Fe/H] values are still based on the KIC, and hence any biases in these values (for example potential systematic overestimates of log g for G-type dwarfs, see Verner et al. 2011; Everett et al. 2013) will be included in the Q1–Q16 catalog.
• 2.  
  The catalog is based on literature values from a variety of techniques, and hence includes systematic offsets between these different methods. For example, spectroscopic temperatures are well known to be systematically offset from photometric temperatures (see Figure 7).
• 3.  
  Surface gravities and metallicities for dwarfs and metallicities for giants that are unclassified in the KIC are valid in a statistical sense only, but are not accurate on a star-to-star basis. Follow-up observations of these stars are highly recommended, especially if a planet-candidate is detected. Stellar properties of ultra cool dwarfs that were previously unclassified in the KIC should also be treated with caution.
• 4.  
  The adopted isochrone grid does not include He-core burning models for low-mass stars, and hence radii, masses, and densities for red giant stars are systematically biased as they are more frequently matched to higher-mass models which include He-core burning models. To derive realistic radii and masses for red giants, it is highly recommended to repeat the isochrone or evolutionary track fits using grids which include He-core burning models.
• 5.  
  Uncertainties on stellar properties adopted in the Q1–Q16 transit detection run (Tenenbaum et al. 2014) are conservative estimates and may be overestimated for cases with very asymmetric probability distributions. This particularly applies to F-K dwarfs. The uncertainties presented in this paper provide improved estimates which should be unaffected by these biases. We note, however, that lower or upper 1σ intervals may be significantly underestimated for stars near the edge of the model grid.
• 6.  
  The isochrone fitting method adopted in this study ignores priors on stellar evolution such as an initial mass function or star formation history. Potential biases introduced by e.g., different evolutionary speeds of stars and different densities of models in certain parameter ranges are not yet considered.

In their revision of properties of Kepler planet-candidate host stars, Batalha et al. (2013) identified two groups of stars with KIC surface gravities and temperatures that were incompatible with YY isochrones. These two groups, namely G-type dwarfs with log  g ∼ 5 and K dwarfs with T_eff ∼ 5000 K and log  g ∼ 4.2, were subsequently matched to the closest YY isochrone. Lacking any further observational information, the fitting procedure in this work results in a similar classification of these "No-Man's-Land" stars.

To test the accuracy of this procedure, we selected stars in the KIC that are either cooler than a 14 Gyr isochrone with [Fe/H] = +0.5 dex, or have a log g that is higher than the highest log g of a 1 Gyr isochrone with [Fe/H] = −2.09 dex. Isochrones for this selection were taken from the BaSTI grid (Pietrinferni et al. 2004). While these age and metallicity cuts are somewhat arbitrary, they do not affect the general conclusions presented in this section. The selected sample was then cross-matched to the SEGUE catalog of spectroscopic classifications (Yanny et al. 2009), yielding an overlap of 140 stars (the total overlap between the KIC and SEGUE includes ∼2220 stars, none of which are Kepler targets). The comparison of temperatures and surface gravities for this sample is shown in Figure 14. As expected, stars generally move away from the "No-Man's-Land," with most high-gravity G-type dwarfs moving closer to the main-sequence. For the cool "No-Man's-Land" sample, stars move in roughly equal numbers toward the main-sequence or the subgiant branch. Notably, some of the stars are identified as giants in the SEGUE classification. Additionally, a considerable number of SEGUE classifications remain in the "No-Man's-Land" zone. Such targets may correspond to stars with unusual properties or rare evolutionary stages, such as merger products, unresolved binary stars, pre-main-sequence stars, or cases in which the medium-resolution SEGUE spectra (R ∼ 2000) did not yield a reliable classification.

Zoom In Zoom Out Reset image size

While the comparison shows that moving stars to the nearest isochrone qualitatively yields improved stellar properties, it is clear from Figure 14 that for some stars such a procedure can in fact yield a larger discrepancy to spectroscopic classifications than initially given in the KIC. We stress that stellar radii for some of these targets may be considerably overestimated or underestimated.