KELT-20b: A Giant Planet with a Period of P ∼ 3.5 days Transiting the V ∼ 7.6 Early A Star HD 185603

From a reduction of KELT-North field 11, KELT-20 (HD 185603) was identified as an exoplanet candidate following the same reduction and candidate selection process as described in detail in Siverd et al. (2012). KELT-North field 11 is a 26° × 26° area of the sky centered on $\alpha ={19}^{{\rm{h}}}{27}^{{\rm{m}}}{00}^{{\rm{s}}}$, $\delta =31^\circ$ 39' $56\buildrel{\prime\prime}\over{.} 16$ (J2000) and was observed 6740 times from UT 2007 May 30 to UT 2014 November 25. From our periodicity search using the VARTOOLS (Hartman et al. 2016) implementation of box-least-squares fitting (Kovács et al. 2002), KELT-20b was identified as a candidate with a 3.4739926-day period, 3.06 hr transit duration, and 0.81% transit depth. The phase-folded discovery light curve containing all 6740 points is shown in Figure 1. We note that KELT-20b was first identified as a candidate in a prior reduction of KELT-North field 11 using data that ended in UT 2013 June 14 (∼700 fewer observations than are shown in Figure 1). The BLS results mentioned above are those of the initial discovery parameters. See Table 1 for the photometric and kinematic properties of KELT-20 from the literature and this work.

Zoom In Zoom Out Reset image size

Table 1.  Literature Properties for KELT-20

Other IDs	HD 185603
	TYC 2655-3344-1
	2MASS J19383872+3113091
Parameter	Description	Value	References
${\alpha }_{{\rm{J}}2000}$	Right Ascension (R.A.)	${19}^{{\rm{h}}}{38}^{{\rm{m}}}38\buildrel{\rm{s}}\over{.} 73$	1
${\delta }_{{\rm{J}}2000}$	Declination (decl.)	+31°13'0921	1
156.5 nm	USST (cW/m2/nm/1012)	1.51 ± 0.17	2
196.5 nm	USST (cW/m2/nm/1012)	3.30 ± 0.24	2
236.5 nm	USST (cW/m2/nm/1012)	2.55 ± 0.15	2
274.0 nm	USST (cW/m2/nm/1012)	2.27 ± 0.07	2
${B}_{{\rm{T}}}$	Tycho ${B}_{{\rm{T}}}$ mag	7.697 ± 0.015	3
${V}_{{\rm{T}}}$	Tycho ${V}_{{\rm{T}}}$ mag	7.592 ± 0.010	3
${u}_{\mathrm{Str}}$	${u}_{\mathrm{Str}\ddot{{\rm{o}}}\mathrm{mgren}-\mathrm{Crawford}}$ mag	9.094 ± 0.039	4
${v}_{\mathrm{Str}}$	${v}_{\mathrm{Str}\ddot{{\rm{o}}}\mathrm{mgren}-\mathrm{Crawford}}$ mag	7.874 ± 0.024	4
${b}_{\mathrm{Str}}$	${b}_{\mathrm{Str}\ddot{{\rm{o}}}\mathrm{mgren}-\mathrm{Crawford}}$ mag	7.645 ± 0.014	4
${y}_{\mathrm{Str}}$	${y}_{\mathrm{Str}\ddot{{\rm{o}}}\mathrm{mgren}-\mathrm{Crawford}}$ mag	7.610 ± 0.010	4
J	2MASS J mag	7.424 ± 0.024	5
H	2MASS H mag	7.446 ± 0.018	5
${K}_{{\rm{S}}}$	2MASS ${K}_{{\rm{S}}}$ mag	7.415 ± 0.017	5
WISE1	WISE1 mag	7.394 ± 0.027	6
WISE2	WISE2 mag	7.437 ± 0.020	6
WISE3	WISE3 mag	7.439 ± 0.016	6
WISE4	WISE4 mag	7.350 ± 0.097	6
${\mu }_{\alpha }$	Gaia DR1 proper motion	3.261 ± 0.026	7
	in R.A. (mas yr−1)
${\mu }_{\delta }$	Gaia DR1 proper motion	−6.041 ± 0.032	7
	in decl. (mas yr−1)
RV	Systemic radial	−23.3 ± 0.3	Section 2.3
	velocity ($\mathrm{km}\ {{\rm{s}}}^{-1}$)
$v\sin {I}_{\star }$	Projected stellar rotational	114.0 ± 4.3	Section 4.3
	velocity ($\mathrm{km}\ {{\rm{s}}}^{-1}$)
Spec. type	Spectral type	A2V	Section 3.1
Age	Age (Myr)	$\lesssim 600$	Section 3.3
π	Gaia parallax (mas)	7.41 ± 0.39	5a
d	Gaia-inferred distance (pc)	139.7±6.6	5a
AV	Visual extinction (mag)	0.07 ± 0.07	Section 3.1
Θ	Angular diameter (mas)	0.0555 ± 0.0070	Section 3.1
U*	Space motion ($\mathrm{km}\ {{\rm{s}}}^{-1}$)	1.13 ± 0.17	Section 3.4
V	Space motion ($\mathrm{km}\ {{\rm{s}}}^{-1}$)	−8.98 ± 0.27	Section 3.4
W	Space motion ($\mathrm{km}\ {{\rm{s}}}^{-1}$)	0.75 ± 0.18	Section 3.4

Note.

^aGaia parallax after correcting for the systematic offset of −0.21 mas as described in Stassun & Torres (2016).

References. (1) van Leeuwen 2007; (2) Thompson et al. 1995; (3) Høg et al. 2000; (4) Paunzen 2015; (5) Cutri et al. 2003; (6) Cutri et al. 2012; (7) Gaia Collaboration et al. 2016, Gaia DR1; http://gea.esac.esa.int/archive/.

Download table as:  ASCIITypeset image

We obtained follow-up time-series photometry from the KELT Follow-Up Network (KELT-FUN) to better characterize the transit depth, duration, and shape, as well as to check for potential astrophysical false positives. We used a custom version of the TAPIR software package (Jensen 2013) to predict transits, and we observed 13 transits in a variety of bands between 2014 August and 2017 June, as listed in Table 2. In Figure 2 we display the photometry from all KELT-FUN observations, as well as the transit light curve when all follow-up observations are combined. Unless otherwise stated, all data were calibrated and analyzed using the AstroImageJ package⁴² (Collins & Kielkopf 2013; Collins et al. 2017).

Zoom In Zoom Out Reset image size

Table 2.  Photometric Follow-up Observations of KELT-20b

Observatory	Location	Aperture	Plate Scale	Date	Filter	Exposure	Detrending Parametersa
		(m)	($^{\prime\prime} {\mathrm{pix}}^{-1}$)	(UT)		Time (s)
PvdK	PA, USA	0.6	0.76	2014 Aug 29	${\rm{i}}^{\prime}$	20	airmass, time
GCO	Lucca, Italy	0.2	2.3	2014 Sep 25	V	90	airmass
WCO	PA, USA	0.35	1.45	2015 Oct 06	${\rm{z}}^{\prime}$	12	airmass
DEMONEXT	AZ, USA	0.5	0.90	2016 May 21	${\rm{i}}^{\prime}$	31	None
DEMONEXT	AZ, USA	0.5	0.90	2016 Jun 04	${\rm{i}}^{\prime}$	31	None
DEMONEXT	AZ, USA	0.5	0.90	2016 Jun 11	${\rm{i}}^{\prime}$	31	None
MINERVA	AZ, USA	0.7	0.60	2016 Nov 05	${\rm{g}}^{\prime}$	31	airmass
PvdK	PA, USA	0.6	0.76	2017 May 08	${\rm{i}}^{\prime}$	20	airmass
MORC	KY, USA	0.6	0.39	2017 May 08	${\rm{i}}^{\prime}$	20	airmass
CDK20N	KY, USA	0.5	0.54	2017 May 08	${\rm{z}}^{\prime}$	60, 40, 30	airmass
WCO	PA, USA	0.35	1.45	2017 May 08	${\rm{z}}^{\prime}$	12	airmass
WCO	PA, USA	0.35	1.45	2017 May 15	${\rm{z}}^{\prime}$	12	airmass
CROW	Portalegre, Portugal	0.3	0.84	2017 Jun 11	${\rm{z}}^{\prime}$	150	airmass

Note.

^aPhotometric parameters allowed to vary in global fits as described in the text.

Download table as:  ASCIITypeset image

2.2.1. Peter van de Kamp Observatory (PvdK)

2.2.2. GCO

2.2.3. WCO

2.2.4. Demonext

2.2.5. MINERVA

2.2.6. MORC

2.2.7. CDK20N

2.2.8. CROW

We obtained a series of spectroscopic follow-up observations of KELT-20b with the Tillinghast Reflector Echelle Spectrograph (TRES) on the 1.5 m telescope at the Fred Lawrence Whipple Observatory, Mount Hopkins, Arizona, USA. TRES is a fiber-fed echelle spectrograph, with a spectral resolution of $\lambda /{\rm{\Delta }}\lambda \sim {\rm{44,000}}$ and a wavelength coverage of 3900–9100 Å over the 51 orders. RVs obtained over 11 out-of-transit orbital phases were used to constrain the mass of the planetary companion. Relative RVs were derived by cross-correlating selected orders in each observed spectrum against the strongest observed spectrum of KELT-20, order by order, and this analysis excludes all orders contaminated by telluric lines or with poor signal-to-noise ratio. These "multiorder" velocities are listed in Table 3 and plotted in Figure 3. In addition, 21 in-transit observations were obtained on the night of UT 2017 April 24 to measure the Doppler tomographic transit of the planet. The analysis of these observations is described in Section 4.3.

Zoom In Zoom Out Reset image size

Table 3.  Relative RVs for KELT-20 from TRES

${\mathrm{BJD}}_{\mathrm{TDB}}$	RV	${\sigma }_{\mathrm{RV}}$
	(${\rm{m}}\,{{\rm{s}}}^{-1}$)	(${\rm{m}}\,{{\rm{s}}}^{-1}$)
2,457,885.970564	0	397.81
2,457,890.927060	328.01	313.63
2,457,900.866772	409.74	397.81
2,457,901.852101	230.49	390.78
2,457,902.775118	759.53	355.68
2,457,903.851423	354.69	261.69
2,457,905.775362	418.55	424.11
2,457,906.798723	217.00	377.87
2,457,907.772196	447.92	347.19
2,457,908.828699	−66.87	367.83
2,457,909.823202	−263.25	287.26
2,457,910.774902	257.09	802.67

Note. The TRES RV zero-point is arbitrarily set to the first TRES value.

Download table as:  ASCIITypeset image

We obtained high-resolution imaging for KELT-20 with the infrared camera PHARO behind the adaptive optics (AO) system P3K on the Palomar 200-inch Hale telescope. PHARO has a pixel scale of $0\buildrel{\prime\prime}\over{.} 025$ pixel⁻¹ (Hayward et al. 2001), and the data were obtained in the narrowband filter Br-γ on UT 2017 May 05.

The AO data were obtained in a five-point quincunx dither pattern with each dither position separated by 5''. Each dither position was observed 3 times, each offset from the previous image by $1^{\prime\prime}$ for a total of 15 frames; the integration time per frame was 45 s. We use the dithered images to remove sky background and dark current and then align, flat-field, and stack the individual images. The PHARO AO data have a resolution of $0\buildrel{\prime\prime}\over{.} 09$ (FWHM).

The sensitivity of the AO data was determined by injecting simulated sources into the final combined images with separations from the primary targets in integer multiples of the central source's FWHM (Furlan et al. 2017). The sensitivity curve shown in Figure 4 represents the 5σ limits of the imaging data.

Zoom In Zoom Out Reset image size

For KELT-20, no stellar companions were detected in the infrared adaptive optics, indicating (to the limits of the data) that the star likely has no additional components to either dilute the transit depth or confuse the determination of the origin of the transit signal (e.g., Ciardi et al. 2015).

Note that to exclude a false positive due to an eclipsing binary within the photometric aperture, we have to exclude a companion that is ∼4.7 mag in the visual, given the transit depth of 1.3%. We can exclude companions that are brighter than ∼7.5 mag outside of ∼0.8 arcsec in K using the AO images. Because the contrast ratios of stars are typically maximized in the Rayleigh–Jeans tail, it is unlikely that there is a companion that is less than ∼4.7 mag fainter in the optical in the same aperture. Thus, it is unlikely that an eclipsing binary is causing the photometric signal. Further evidence against a diluted eclipsing binary comes from the achromaticity of the transit signals. Finally, note also that the Doppler tomographic observation further eliminates the possibility of a blended eclipsing binary causing the transit signal. We discuss further evidence against a false-positive scenario in the next section.

Despite the unusual nature of this system and the lack of a definitive measurement of the companion mass, we are confident that this system is truly a hot Jupiter transiting an early A star. The evidence for this comes from several sources that we will briefly review; however, we invite the reader to review Bieryla et al. (2015), Zhou et al. (2016a, 2016b), and Hartman et al. (2015) for a more detailed explanation. Of course, the first system to have been validated in this way was WASP-33b (Collier Cameron et al. 2010).

The Doppler tomographic observation eliminates the possibility of a blended eclipsing binary causing the transit signal. The line profile derived from the least-squares deconvolution shows a lack of spectroscopic companions blended with KELT-20. The spectroscopic transit is seen crossing the entirety of the rapidly rotating target star's line profile, confirming that it is indeed orbiting KELT-20. The summed flux underneath the Doppler tomographic shadow and the distance of closest approach of the shadow from the zero velocity at the center of the predicted transit time are consistent with both the photometric transit depth and impact parameter, suggesting that the photometric transit is not diluted by background stars and is fully consistent with the spectroscopic transit. Adaptive optics observations (Section 2.4) also eliminate blended stars with ${\rm{\Delta }}K\lt 7.5$ and $\gt 0\buildrel{\prime\prime}\over{.} 6$ of KELT-20, consistent with the lack of blending in the spectroscopic analysis.

Finally, the planetary nature of KELT-20b is confirmed by the TRES RV measurements, which constrain the mass of the companion to be $\lesssim 3.5\,{M}_{\mathrm{jup}}$ at $3\sigma$ significance. This eliminates the possibility that the transiting companion is a stellar-mass or brown-dwarf-mass object. As such, KELT-20b is confirmed as a planetary-mass companion transiting the rapidly rotating A star HD 185603.

Thus, we conclude that all the available evidence suggests that the most plausible interpretation is that KELT-20b is a Jupiter-size planet transiting an early A star with a projected spin–orbit alignment that is (perhaps surprisingly) well aligned (see Section 5.2.1).

We assembled the available broadband photometry of KELT-20 (see Table 1) in order to construct a spectral energy distribution (SED) spanning a large range of wavelengths from ∼0.15 to 22 μm (Figure 5). We fit the SED using the model atmospheres of Kurucz (1992), the free parameters being the stellar effective temperature (${T}_{\mathrm{eff}}$), extinction (A_V), and a flux normalization factor (effectively the ratio of the stellar radius to the distance). The stellar surface gravity ($\mathrm{log}{g}_{* }$) and metallicity ($[\mathrm{Fe}/{\rm{H}}]$) have only a minor effect on the SED and are poorly constrained by this type of fit, so we simply adopted a solar metallicity and $\mathrm{log}{g}_{* }$ = 4.3 (corroborated by the final global fit; see Section 4.1 and Table 4). The extinction was limited to the maximum value from the dust maps of Schlegel et al. (1998) for this line of sight, A_V = 1.43 mag.

Zoom In Zoom Out Reset image size

Table 4.  Median Values and 68% Confidence Intervals for the Physical and Orbital Parameters of the KELT-20 System

Parameter	Description (Units)	Adopted Value	Value
		(YY Circular)	(Torres Circular)
Stellar Parameters
M*	Mass ($\,{M}_{\odot }$)	${1.76}_{-0.19}^{+0.14}$	${1.91}_{-0.20}^{+0.22}$
R*	Radius ($\,{R}_{\odot }$)	${1.565}_{-0.064}^{+0.057}$	${1.609}_{-0.064}^{+0.065}$
L*	Luminosity ($\,{L}_{\odot }$)	${12.7}_{-1.9}^{+2.2}$	${13.2}_{-2.1}^{+2.3}$
${\rho }_{* }$	Density (cgs)	${0.641}_{-0.033}^{+0.035}$	${0.646}_{-0.033}^{+0.036}$
$\mathrm{log}{g}_{* }$	Surface gravity (cgs)	${4.290}_{-0.020}^{+0.017}$	4.305 ± 0.022
${T}_{\mathrm{eff}}$	Effective temperature (K)	${8720}_{-260}^{+250}$	${8690}_{-280}^{+260}$
$[\mathrm{Fe}/{\rm{H}}]$	Metallicity	$-{0.29}_{-0.36}^{+0.22}$	$-{0.01}_{-0.49}^{+0.50}$
$v\sin {I}_{* }$	Rotational velocity (m s−1)	117400 ± 2900	117100 ± 2900
λ	Spin–orbit alignment (degrees)	3.1 ± 1.3	3.2 ± 1.3
${NRVel}.W.$	Nonrotating line width (m s−1)	${1460}_{-980}^{+1300}$	${1470}_{-980}^{+1400}$
Planet Parameters
P	Period (days)	3.4741085 ± 0.0000019	3.4741085 ± 0.0000020
a	Semimajor axis (au)	${0.0542}_{-0.0021}^{+0.0014}$	0.0557 ± 0.0020
MP	3σ Mass Limit ($\,{M}_{{\rm{J}}}$)	$\lt 3.382$	$\lt 3.590$
RP	Radius ($\,{R}_{{\rm{J}}}$)	${1.741}_{-0.074}^{+0.069}$	1.789 ± 0.075
${\rho }_{P}$	3σ Limit Density (cgs)	$\lt 0.806$	$\lt 0.781$
$\mathrm{log}{g}_{P}$	3σ Surface gravity	$\lt 3.467$	$\lt 3.611$
Teq	Equilibrium temperature (K)	2262 ± 73	${2252}_{-78}^{+75}$
Θ	Safronov number	${0.0048}_{-0.0040}^{+0.023}$	${0.0047}_{-0.0039}^{+0.022}$
$\langle F\rangle$	Incident flux (${10}^{9}\ \mathrm{erg}\,{{\rm{s}}}^{-1}\,{\mathrm{cm}}^{-2}$)	${5.94}_{-0.73}^{+0.81}$	${5.84}_{-0.77}^{+0.82}$
Radial Velocity Parameters
TC	Time of inferior conjunction (${\mathrm{BJD}}_{\mathrm{TDB}}$)	2,457,485.74965 ± 0.00020	2,457,485.74965 ± 0.00020
K	3σ RV semi-amplitude (m s−1)	$\lt 311.3$	$\lt 306.7$
${M}_{P}\sin i$	3σ Minimum mass ($\,{M}_{{\rm{J}}}$)	$\lt 3.372$	$\lt 3.580$
${M}_{P}/{M}_{* }$	3σ Mass ratio	$\lt 0.001863$	$\lt 0.001788$
u	RM linear limb darkening	${0.532}_{-0.014}^{+0.011}$	${0.533}_{-0.015}^{+0.011}$
${\gamma }_{\mathrm{TRES}}$	m s−1	${248}_{-96}^{+95}$	248 ± 96
Linear Ephemeris
from Follow-up
Transits
PTrans	Period (days)	3.4741070 ± 0.0000019	⋯
T0	Linear ephemeris from transits (${\mathrm{BJD}}_{\mathrm{TDB}}$)	2,457,503.120049 ± 0.000190	⋯

Note. 3σ limits reported for KELT-20b's mass and parameters dependent on mass. The gamma velocity reported here uses an arbitrary zero-point for the multiorder relative velocities. The absolute gamma velocity based on the Mg b order analysis is 23.8 ± 0.3 km s⁻¹.

Download table as:  ASCIITypeset image

The resulting best-fit parameters are ${A}_{V}=0.07\pm 0.07$ mag and ${T}_{\mathrm{eff}}$ = 8800 ± 500 K, with a reduced chi-square of ${\chi }_{\nu }^{2}=3.05$ (Figure 5). By directly integrating the (unextincted) fitted SED model, we obtain a semi-empirical measure of the stellar bolometric flux at Earth, ${F}_{\mathrm{bol}}=(2.46\pm 0.27)\times {10}^{-8}$ erg s⁻¹ cm⁻². From ${F}_{\mathrm{bol}}$ and ${T}_{\mathrm{eff}}$ we obtain a measure of the stellar angular radius, Θ, which in turn provides a constraint on the stellar radius via the distance from the corrected Gaia parallax of ${R}_{\star }=1.61\pm 0.22\ \,{R}_{\odot }$. This estimate of ${R}_{\star }$ is used as a constraint in the global system fit below (Section 4.1). The ${T}_{\mathrm{eff}}$ of 8800 K corresponds to an A2V-type star (Pecaut & Mamajek 2013).

As was originally demonstrated in the context of transiting planets by Seager & Mallén-Ornelas (2003), under the assumption that $k\equiv {R}_{P}/{R}_{* }\ll 1$, it is possible to estimate the density ($\,{\rho }_{* }$) of a host star via a measurement of the FWHM (T_FWHM) of the transit, the period (P), the impact parameter (b), the eccentricity, and the argument of periastron. As these quantities can be measured essentially directly (i.e., without reliance on models), one can obtain an empirical estimate of $\,{\rho }_{* }$. This can then be combined with the essentially direct estimate of R_* as determined from ${T}_{\mathrm{eff}}$, the bolometric flux, and parallax above to estimate the stellar mass (${M}_{* }$), again without reliance on theoretical models (e.g., isochrones) or externally calibrated relations (e.g., Torres et al. 2010). This technique was recently applied to all transiting planets in the first Gaia data release by Stassun et al. (2017).

We do not have a constraint on the eccentricity or argument of periastron, but given the short period, it is reasonable to assume that the orbit has been circularized. In the limit e = 0 and $k\ll 1$,

$$\begin{eqnarray}&&{M}_{* }=\left(\displaystyle \frac{4{{PR}}_{* }^{3}}{\pi {{GT}}_{\mathrm{FWHM}}^{3}}\right){(1-{b}^{2})}^{\tfrac{3}{2}}.\end{eqnarray} \tag{ 1 }$$

We adopt the estimates of P, T_FWHM, and b derived from global modeling (see Section 4.1) using the Yonsei–Yale (YY) isochrone-constrained circular fits given in Tables 4 and 5. We note that while these parameters formally rely on the constraints from the YY isochrones, since they are derived (almost) directly from data, their measurements are not, in fact, affected by these constraints. This can be seen by comparing the values of these parameters measured from the global modeling using the YY isochrones with those from the global modeling using the Torres relations; these parameters differ by $\lt 1 \%$ between these two fits in all cases. Adopting the Gaia-inferred radius of ${R}_{\star }=1.61\pm 0.22\ \,{R}_{\odot }$, we find ${M}_{* }=1.90\pm 0.47\ \,{M}_{\odot }$, with an uncertainty of ∼25%. We note that this uncertainty is dominated by the uncertainty in ${R}_{* }$.

Table 5.  Median Values and 68% Confidence Intervals for the Physical and Orbital Parameters for the KELT-20 System

Parameter	Description (Units)	Adopted Value	Value
		(YY circular)	(Torres circular)
${R}_{P}/{R}_{* }$	Radius of the planet in stellar radii	${0.11440}_{-0.00061}^{+0.00062}$	${0.11431}_{-0.00062}^{+0.00064}$
$a/{R}_{* }$	Semimajor axis in stellar radii	7.42 ± 0.13	7.44 ± 0.13
i	Inclination (degrees)	${86.12}_{-0.27}^{+0.28}$	${86.16}_{-0.27}^{+0.28}$
b	Impact parameter	${0.503}_{-0.028}^{+0.025}$	${0.499}_{-0.028}^{+0.026}$
δ	Transit depth	0.01309 ± 0.00014	${0.01307}_{-0.00014}^{+0.00015}$
TFWHM	FWHM duration (days)	0.12901 ± 0.00048	0.12904 ± 0.00049
τ	Ingress/egress duration (days)	${0.01996}_{-0.00077}^{+0.00080}$	${0.01984}_{-0.00078}^{+0.00081}$
T14	Total duration (days)	${0.14898}_{-0.00088}^{+0.00091}$	${0.14889}_{-0.00088}^{+0.00090}$
PT	A priori nongrazing transit probability	${0.1193}_{-0.0020}^{+0.0021}$	0.1190 ± 0.0021
${P}_{T,G}$	A priori transit probability	0.1502 ± 0.0027	0.1498 ± 0.0027
${u}_{1\mathrm{Sloang}}$	Linear limb darkening	${0.3427}_{-0.017}^{+0.0091}$	${0.3399}_{-0.017}^{+0.0086}$
${u}_{2\mathrm{Sloang}}$	Quadratic limb darkening	${0.3362}_{-0.0038}^{+0.0072}$	${0.3421}_{-0.0083}^{+0.0091}$
${u}_{1\mathrm{Sloani}}$	Linear limb darkening	${0.1924}_{-0.0084}^{+0.011}$	${0.186}_{-0.010}^{+0.012}$
${u}_{2\mathrm{Sloani}}$	Quadratic limb darkening	${0.2442}_{-0.0062}^{+0.010}$	${0.254}_{-0.013}^{+0.026}$
${u}_{1\mathrm{Sloanz}}$	Linear limb darkening	${0.1229}_{-0.0064}^{+0.0099}$	${0.1179}_{-0.0069}^{+0.010}$
${u}_{2\mathrm{Sloanz}}$	Quadratic limb darkening	${0.2393}_{-0.0080}^{+0.0099}$	${0.246}_{-0.012}^{+0.023}$
u1V	Linear limb darkening	${0.300}_{-0.015}^{+0.011}$	${0.295}_{-0.015}^{+0.010}$
u2V	Quadratic limb darkening	${0.3096}_{-0.0035}^{+0.0072}$	${0.3173}_{-0.0100}^{+0.017}$
Secondary Eclipse
TS	Time of eclipse (${\mathrm{BJD}}_{\mathrm{TDB}}$)	2,457,484.01259 ± 0.00020	2,457,484.01260 ± 0.00020

Download table as:  ASCIITypeset image

Interestingly, this inferred mass is nearly identical to the mass inferred from the Torres-constrained global fit, and indeed the radius inferred from this global fit is nearly identical to the Gaia-determined radius. However, in both cases the uncertainties are somewhat smaller. This implies that the mass and radius of the host are largely determined by the direct (model-independent) constraints in the Torres-constrained global fits and completely consistent with the Torres relations. The Torres relations are therefore primarily serving to decrease the uncertainties (slightly).

Importantly, the inferred $\mathrm{log}{g}_{* }\simeq 4.3$ is at the higher end of what is typically expected from A stars of this ${T}_{\mathrm{eff}}$ and solar metallicity (see, e.g., Torres et al. 2010). This implies that the host is exceptionally close to (and perhaps lower than) the ZAMS for solar-metallicity stars in the parameter space of $\mathrm{log}{g}_{* }$ versus ${T}_{\mathrm{eff}}$. This can be explained in several ways. First, the star could indeed have nearly solar metallicity but be very young. Second, the star could be older but have subsolar metallicity, since the ZAMS is at a lower $\mathrm{log}{g}_{* }$ at fixed ${T}_{\mathrm{eff}}$ for stars of lower metallicity. Finally, the measurement of ${R}_{* }$ from the SED and parallax could have a small systematic error.

Since the Torres relations do not encode age, it is possible for this star to have a higher $\mathrm{log}{g}_{* }$ at solar metallicity without resulting in any tension with the empirical parameters using those relations. On the other hand, the YY isochrones do encode age, thus enforcing a maximum $\mathrm{log}{g}_{* }$ for a given metallicity (i.e., that of the ZAMS), and thus the inferred high $\mathrm{log}{g}_{* }$ disfavors this star having solar metallicity. The YY isochrone fits therefore "prefer" lower metallicities for the host star, although we note that a solar metallicity is still allowed within $\sim 1\sigma$. The lower metallicity inferred by the YY fits also results in a somewhat smaller mass and radius than inferred from the empirical methods above and the Torres-constrained global fits.

Overall, we are agnostic about which of these three explanations are correct. Generally, we note that A stars with metallicities of $[\mathrm{Fe}/{\rm{H}}]\sim -0.3$ are not common, and we note that the kinematics of this star (i.e., the low UVW velocities presented in Section 3.4) support the interpretation that the star is young. Of course, we cannot rule out the simpler explanation that there are unrecognized subtle systematics affecting our inference of the radius, mass, and surface gravity of the star.

We note that a Hipparcos parallax also exists for this star and is 8.73 ± 0.50 mas. The radius and mass inferred from the Hipparcos parallax are ${R}_{* }=1.37\pm 0.09\ \,{R}_{\odot }$ and ${M}_{* }=1.17\pm 0.23\,{M}_{\odot }$. These stellar parameters are inconsistent with those inferred from the Gaia parallax of 7.716 ± 0.37 mas. In particular, as can be seen in Figures 5 and 6, these values are completely inconsistent with the SED (${T}_{\mathrm{eff}}$) or even the color of the source. We therefore reject it and adopt the Gaia parallax with the Stassun & Torres (2016) systematic correction. An examination of the reasons for this apparent discrepancy with the Hipparcos parallax is beyond the scope of this paper. Here we simply note the discrepancy and proceed with our analysis utilizing the Gaia parallax as a constraint on the system global solution (Section 4.1).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To put the KELT-20 system in context and to provide an initial estimate of the system age, we show in Figure 6 the KELT-20 host star in the modified Hertzsprung–Russell diagram ($\mathrm{log}{g}_{* }$ vs. ${T}_{\mathrm{eff}}$). Using the YY stellar evolutionary models for a star of mass 1.76 $\,{M}_{\odot }$, we infer an age for KELT-20 of at most ∼600 Myr.

KELT-20 is located at equatorial coordinates $\alpha \,={19}^{{\rm{h}}}{38}^{{\rm{m}}}38\buildrel{\rm{s}}\over{.} 73$ and $\delta =+31^\circ 13^{\prime} 09\buildrel{\prime\prime}\over{.} 21$ (J2000), corresponding to Galactic coordinates of ${\ell }=65\buildrel{\circ}\over{.} 8$ and $b=4\buildrel{\circ}\over{.} 6$. Given the Gaia distance of $139.7\pm 6.6\,\mathrm{pc}$ (Gaia Collaboration et al. 2016), KELT-20 lies at a galactocentric distance of roughly 8.26 kpc, assuming a distance from the Sun to the Galactic center of ${R}_{0}=8.32\,\mathrm{kpc}$ (Gillessen et al. 2017). KELT-20 is located ∼10 pc above the plane, well within the Galactic scale height for A stars of ∼50 pc (Bovy 2017).

Using the Gaia proper motion of $({\mu }_{\alpha },{\mu }_{\delta })=(3.261\,\pm 0.026,-6.041\pm 0.032)\,\mathrm{mas}\ {\mathrm{yr}}^{-1}$, the Gaia parallax, and the absolute RV as determined from the TRES spectroscopy of $-23.8\pm 0.3\,\mathrm{km}\,{{\rm{s}}}^{-1}$, we find that KELT-20 has a three-dimensional Galactic space motion of (U, V, W) = (1.14 ± 0.17, −8.98 ± 0.27, 0.75 ± 0.18) km s⁻¹, where positive U is in the direction of the Galactic center, and we have adopted the Coşkunoǧlu et al. (2011) determination of the solar motion with respect to the local standard of rest. These values yield a 99.5% probability that KELT-20 is a thin-disk star, according to the classification scheme of Bensby et al. (2003), as expected for its young age and early spectral type.

KELT-20 is projected against a supernova remnant, which is also visible in optical and Hα survey data. This is a known supernova remnant, SNR G065.3+05.7, which is about 0.8 kpc away (Boumis et al. 2004). At a distance from Gaia of ∼140 pc, this is evidently a chance projection, with KELT-20 well in front of the supernova remnant.

The line of sight toward KELT-20 in Cygnus is along the so-called Orion Spur or Orion Arm, and thus it would be expected that there would be a large population of young stars in that general direction. Most of the young associations cataloged in that direction (e.g., the Cygnus OB associations, the North America Nebula, the Pelican Nebula, NGC 6914) lie at distances of 1 kpc or more, and we were not able to locate in the literature any evidence of known star-forming regions in the vicinity of the ∼140 pc distance to KELT-20. We also checked KELT-20's Galactic space motion against the known young moving groups, and there is no obvious match. In addition, searching Gaia DR1, there are no sources within 5° of KELT-20 with similar proper motion and distance.

Thus, while we cannot associate KELT-20 with any known star-forming region or known young stellar population in particular, its young age is completely plausible given its location in the Galaxy. We infer that it was likely associated with some earlier episode of star formation in our spiral arm, but its local gas and any associated young stars have since dispersed into the field population.

Using a heavily modified version of EXOFAST (Eastman et al. 2013), an IDL-based exoplanet fitting suite, we perform a series of global fits to determine the system parameters for KELT-20. Within the global fit, all photometric and spectroscopic observations (including the Doppler tomography signal) are simultaneously fit. EXOFAST uses either the YY stellar evolution model tracks (Demarque et al. 2004) or the Torres relations (Torres et al. 2010) to constrain the mass and radius of the host star, KELT-20. See Siverd et al. (2012) for a detailed description of the global modeling routine.

The global fit uses all follow-up raw light curves and their specified detrending parameters (shown in Table 2 as inputs). From the SED analysis (Section 3.1) we impose a prior on ${T}_{\mathrm{eff}}$ of 8800 ± 500 K. As we are unable to suitably constrain the metallicity of KELT-20 from our current observations, we set a prior on $[\mathrm{Fe}/{\rm{H}}]$ of 0.0 ± 0.5 dex. We ran an initial global fit where a prior was set on the period and transit center time from the analysis of the KELT-North light curve. By performing a linear fit to the transit center times, we independently determined an ephemeris for KELT-20b (See Section 4.2). We then reran the Torres and YY circular fits with a prior on the transit center time and period obtained from that analysis. The KELT-North light curve itself is not included in any of the global fits we conducted. Lastly, we use the Gaia parallax shown in Table 1 combined with the bolometric flux estimated from the SED fits to impose a prior on the host star's radius (${R}_{\star }=1.610\pm 0.216$). We performed two separate global fits where we fix the eccentricity of the planet's orbit to zero; one fit uses the YY models to determine the mass and radius of KELT-20, while the other uses the Torres relations for the same purpose. For the discussion and interpretation of the KELT-20 system, we adopt the circular YY fit. The results of both fits are shown in Tables 4 and 5.

For the output parameters shown in this paper that use solar or Jovian units, we adopt the following constants throughout: G $\,{M}_{\odot }$ = 1.3271244 × 10²⁰ m³ s⁻², ${R}_{\odot }$ = 6.9566 × 10⁸ m, ${M}_{{\rm{J}}}$ = 0.000954638698 $\,{M}_{\odot }$, and $\,{R}_{{\rm{J}}}$ = 0.102792236 $\,{R}_{\odot }$(Standish 1995; Torres et al. 2010; Eastman et al. 2013; Prša et al. 2016).

We analyzed the fiducial global model transit center times of all follow-up light curves (see Table 6 and Figure 7) to search for transit timing variations (TTVs) in the KELT-20 system. Before running the global models, we confirm that all photometric time stamps are in ${\mathrm{BJD}}_{\mathrm{TDB}}$ format (Eastman et al. 2010). To ensure the accuracy of the time stamps, follow-up observers provision telescope control computers to synchronize to a standard clock (such as the atomic clock in Boulder, CO). This synchronization is normally done periodically throughout the observing session. To assess the TTV for each light curve, we find the best linear fit to the transit center times. The resulting linear ephemeris has a reference transit center time of ${T}_{0}=2457503.120049\pm 0.000190$ (${\mathrm{BJD}}_{\mathrm{TDB}}$), a period of 3.4741070 ± 0.00000186 days, and a ${\chi }^{2}$ of 60.8 with 11 degrees of freedom. We note that the large ∼9-minute TTV in the GCO data (Table 6) is likely the result of the partial transit coverage and systematics in the light curve (see Figure 2). The largest scatter in the other light curves occurs on epoch 109 (see Table 6), where the transit was simultaneously observed by four telescopes. Using that scatter as the limit of our TTV sensitivity threshold, we find no evidence for astrophysical TTVs in our data. We therefore adopt the linear ephemeris specified above as the best predictor of future transit times from our data.

Table 6.  Transit Times from KELT-20 Photometric Observations

Epoch	${T}_{{\rm{C}}}$	${\sigma }_{{T}_{{\rm{C}}}}$	O–C	O–C	Telescope
	(${\mathrm{BJD}}_{\mathrm{TDB}}$)	(s)	(s)	(${\sigma }_{{T}_{{\rm{C}}}}$)
−174	2,456,898.624275	43	−99.14	−2.27	PvdK
−166	2,456,926.424578	180	544.24	3.02	GCO
−58	2,457,301.621915	74	6.46	0.09	WCO
8	2,457,530.913718	56	70.20	1.23	DEMONEXT
12	2,457,544.810920	44	137.06	3.08	DEMONEXT
14	2,457,551.756911	55	−55.02	−1.00	DEMONEXT
56	2,457,697.671922	62	162.27	2.60	MINERVA
109	2,457,881.799595	48	162.22	3.37	PvdK
109	2,457,881.796557	49	−100.26	−2.01	MORC
109	2,457,881.796903	55	−70.37	−1.28	CDK20N
109	2,457,881.795676	75	−176.38	−2.33	WCO
111	2,457,888.745551	51	−32.88	−0.64	WCO
119	2,457,916.537500	50	−111.28	−2.21	CROW

Note. Epochs are given in orbital periods relative to the value of the inferior conjunction time from the global fit.

Download table as:  ASCIITypeset image

We obtained 21 in-transit spectroscopic observations of KELT-20b with TRES on 2017 April 24. These observations were made and processed as per Zhou et al. (2016a). For each spectrum, we derive a rotational profile via a least-squares deconvolution against a nonrotating template spectrum, as per the techniques described in Donati et al. (1997) and Collier Cameron et al. (2010). We create a median-combined rotational profile that averages out the transit signal. This median-combined rotational profile is then subtracted from each individual exposure, revealing the dark shadow of the planet transiting across the star (Figure 8). These line profile residuals are modeled in the global analysis in Section 4.1 as described in Gaudi et al. (2017). We adopt linear limb-darkening coefficients from Claret (2004) for the V band in the Doppler tomographic modeling. By modeling the rotational broadening profiles, we also measured rotational broadening parameters $v\sin {I}_{* }$ of $114.92\pm 4.24\,\mathrm{km}\ {{\rm{s}}}^{-1}$ and a macroturbulence velocity of ${6.08}_{-2.03}^{+4.44}\,\mathrm{km}\ {{\rm{s}}}^{-1}$. These were adopted as Gaussian priors in the global analysis in Section 4.1. In addition, we also checked the transit Doppler tomography result by deriving multiorder RVs for the same data set. These velocities also clearly show the Rossiter–McLaughlin effect (McLaughlin 1924; Rossiter 1924) consistent with the spin–orbit angle derived from the global analysis (see Figure 9).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In many ways, KELT-20b appears to be quite similar to KELT-9b (Gaudi et al. 2017), albeit orbiting a slightly cooler and less massive star at a somewhat longer (∼2.3 times) period. However, the fact that KELT-20 is nearly as bright as KELT-9 nevertheless makes the prospect for characterization of the system nearly as promising as for KELT-9b.

Figure 10 shows the host star effective temperature versus the V-band magnitude for known transiting planets. Together with 55 Cancri (Demory et al. 2011; Winn et al. 2011), KELT-9b and KELT-20b are the three brightest (in V) transiting planet hosts known, while KELT-9b and KELT-20b are the two brightest hosts of transiting hot Jupiters, which are considerably more amenable to detailed follow-up.

Zoom In Zoom Out Reset image size

Figure 11 shows the primary transit depth, $\delta ={({R}_{P}/{R}_{* })}^{2}$, versus predicted planetary equilibrium temperature ${T}_{\mathrm{eq}}$(assuming zero albedo and complete heat redistribution) for planets with host stars $V\lt 13$, color-coded by the amount of UV flux the planet receives. Although KELT-20b's predicted equilibrium temperature is not nearly as high as KELT-9b, it is nevertheless one of the hottest dozen or so known hot Jupiters. Furthermore, its transit depth is nearly twice that of KELT-9b. Although we only have an upper limit on the mass of KELT-20b, our $3\sigma$ upper limit on the surface gravity $\mathrm{log}{g}_{P}$ is ∼3.5 (cgs). We can therefore predict that the magnitude of the thermal emission spectrum, transmission spectrum, and phase curve should all be easily detectable with Spitzer, the Hubble Space Telescope (HST), and eventually the James Webb Space Telescope. Indeed, the planet is sufficiently hot that secondary eclipse measurements should be possible from ground-based instruments. We also expect that, should the atmosphere be significantly ablated by the high UV flux incident on the planet, this may be detectable via HST.

Zoom In Zoom Out Reset image size

With a sample of six A star hosts of transiting gas giants now known, it starts to become possible to consider and compare the ensemble properties of such systems. Figure 12 shows one such comparison, namely, the location and expected future evolution of these hosts on an ${R}_{* }$ versus ${T}_{\mathrm{eff}}$ (modified Hertzsprung–Russell) diagram. We show the evolutionary tracks based on the YY isochrones for KELT-9 (${M}_{* }\simeq 2.52\,{M}_{\odot }$), KELT-20 (${M}_{* }\simeq 1.76$ $\,{M}_{\odot }$), and KELT-17 (${M}_{* }\simeq 1.63$ $\,{M}_{\odot }$), all assuming solar metallicity. The other three blue circles are (from left to right) Kepler-13 A (${T}_{\mathrm{eff}}\simeq 7650$ K), HAT-P-57 (${T}_{\mathrm{eff}}\simeq 7500$ K), and WASP-33 (${T}_{\mathrm{eff}}\simeq 7430$ K), all of which have quite similar ${T}_{\mathrm{eff}}$ to KELT-17 and radii and masses that differ by only ∼20%.

Zoom In Zoom Out Reset image size

We note that while KELT-9, KELT-17, and Kepler-13 are somewhat evolved from the ZAMS, KELT-20 and to a lesser extent HAT-P-57 and WASP-33 appear to be on (or perhaps even slightly below) the ZAMS, indicating that they are young or (less likely) have subsolar metallicity.

5.2.1. Spin–Orbit Alignment

Doppler tomographic observations allow the measurement of the spin–orbit misalignment (λ). This, however, is merely the sky-projected angle between the stellar spin and planetary orbital angular momentum vectors. Measurement of the full three-dimensional spin–orbit angle (ψ) requires knowledge of the inclination of the stellar rotation axis with respect to the line of sight (I_*), which is typically difficult to measure. We do not have such a measurement of this angle for KELT-20 and so cannot directly calculate ψ.

We can, however, set limits on I_* and thus on ψ. Following Iorio (2011), we can limit I_* by requiring that the star be rotating at less than the breakup velocity. Using our measured stellar and planetary parameters, we obtain a $1\sigma$ limit of $24\buildrel{\circ}\over{.} 4\lt {I}_{* }\lt 155\buildrel{\circ}\over{.} 6$. Together with our measured values of λ and i, this implies $1\buildrel{\circ}\over{.} 3\lt \psi \lt 69\buildrel{\circ}\over{.} 8$ (again at 1σ).

Although the planetary orbit is well aligned if I_* is close to 90^◦ (i.e., the stellar rotation axis is close to perpendicular to the line of sight), in which case $\psi \sim \lambda$, it may still be substantially misaligned if we are viewing the star closer to pole-on. KELT-20 has a projected rotational velocity of $v\sin {I}_{* }=115.9\pm 3.4$ km s⁻¹, which is slightly lower than the median deprojected rotational velocity of 131 km s⁻¹ found by Royer et al. (2007) for A2–A3 main-sequence stars. This suggests that KELT-20 is plausibly close to equator-on and approximately aligned. However, we cannot exclude the possibility that KELT-20 is rotating faster than the median for similar stars and the orbit is misaligned.

A measurement or constraint on I_* may be possible in the future via several methods. First, the detection of rotational modulation would constrain the rotation period and thus I_*; however, this is unlikely and difficult for a hot, likely inactive A star like KELT-20. An asteroseismic measurement of the rotation rate is possible by measuring the rotational splitting of the modes. However, there is no evidence that KELT-20 is pulsating, and thus this would require long-time-baseline, very high precision space-based photometry. It may be possible to measure I_* using very high precision light curves affected by gravity darkening (Barnes 2009), or by measuring the nodal precession of the planet if it is not aligned (Johnson et al. 2015; Iorio 2016). Even in the most optimistic case, however, the precession rate will be $d{\rm{\Omega }}/{dt}\lt 0\buildrel{\circ}\over{.} 03$ yr⁻¹. This is at least an order of magnitude smaller than that measured for WASP-33b by Johnson et al. (2015) and would take several decades to give rise to a detectable change in λ or b.

Because of its larger mass and therefore more rapid evolution, KELT-20 is likely to be quite young ($\lt 600$ Myr) if it has a near-solar metallicity, as expected. This may place interesting constraints on the timescale for its migration to its current orbit. The fact that KELT-20b is one of only two hot Jupiters orbiting A-type stars that could have an aligned orbit,⁴³ as shown in Figure 13, may be particularly interesting in this regard.

Zoom In Zoom Out Reset image size

5.2.2. The Past and Future Evolution of the KELT-20 System

We note that KELT-20 is a somewhat unusual system as compared to many hot Jupiters in that the spin period of the star is shorter than the orbital period of the planet. This implies that tides serve to increase the semimajor axis of the planet, rather than to decrease it. Furthermore, as the star has essentially no convective envelope, one would expect tides to behave quite differently than in stars with convective envelopes. Finally, the expected large oblateness of the host star may affect the efficiency and nature of tidal dissipation.

Nevertheless, we proceed to estimate the past and future orbital evolution of the system under tides. Specifically, we compute the evolution of the semimajor axis in units of the stellar radius and the evolution of the stellar insolation.

The orbital evolution of KELT-20b was calculated under the assumption of a constant phase lag, including the effect of the changing stellar radius due to stellar evolution, following Penev et al. (2014). Due to the poorly constrained efficiency of tidal dissipation in stars, we consider a wide range of dissipation parameters ($Q{{\prime} }_{\star }={10}^{5}$, 10⁶, and 10⁷), where $1/Q{{\prime} }_{\star }$ is the product of the phase lag and the stellar tidal Love number. Given a dissipation parameter, the initial orbital period of the planet was chosen such that the currently observed orbital period is reproduced at an age of 480 Myr. Note that the least dissipative case considered here ($Q{{\prime} }_{\star }={10}^{7}$) was chosen simply because it leads to very little orbital evolution and is in no way physically motivated.

Figure 14 shows the past and future evolution of the orbit of the planet relative to the stellar radius as a function of the age of the system under these assumptions. As mentioned above, unlike the majority of hot Jupiter systems, the measured $v\sin {I}_{* }$ of the host star implies that the stellar spin period is shorter than the orbital period. As a result, the typical picture of a decaying orbit is reversed and the orbit expands over time owing to tidal dissipation. Even under the fairly unrealistic value of $Q{{\prime} }_{\star }\sim {10}^{5}$, the planet will avoid engulfment by the star until well after it begins to extend up the giant branch.

Zoom In Zoom Out Reset image size

Figure 14 also shows the past and future evolution of stellar incident insolation flux received by the planet. The increase in the planet's orbit due to tides is roughly offset by the increase in the radius of the star due to stellar evolution. KELT-20b was likely always above the empirically estimated minimum insolation for inflated giant planets (Demory & Seager 2011), which is not surprising given its inferred radius of ${R}_{P}\sim 1.6\,\,{R}_{{\rm{J}}}$.

Note that at around 1.5 Gyr, the star will cross the Kraft break (Kraft 1967) and begin to develop a deep convective envelope. However, it is unlikely that the planet will have synchronized its period with that of the star, and so we do not expect this system to evolve into an RS CVn system (see Siverd et al. 2012). KELT-20 will eventually engulf its planet, but not until it has ascended the giant branch.