A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host

TL;DR

This paper reports the discovery and analysis of KELT-9b, a highly irradiated giant exoplanet orbiting a very hot star, with extreme ultraviolet exposure potentially causing significant atmospheric mass loss.

Contribution

It presents the first detailed characterization of KELT-9b, the hottest known transiting exoplanet, and discusses its atmospheric properties and potential for ablation due to intense UV irradiation.

Findings

KELT-9b's day-side temperature is approximately 4600K.

The host star's temperature is about 10,170K, near the boundary between A and B types.

KELT-9b receives roughly 700 times more extreme UV radiation than WASP-33b.

Abstract

The amount of ultraviolet irradiation and ablation experienced by a planet depends strongly on the temperature of its host star. Of the thousands of extra-solar planets now known, only four giant planets have been found that transit hot, A-type stars (temperatures of 7300-10,000K), and none are known to transit even hotter B-type stars. WASP-33 is an A-type star with a temperature of ~7430K, which hosts the hottest known transiting planet; the planet is itself as hot as a red dwarf star of type M. The planet displays a large heat differential between its day-side and night-side, and is highly inflated, traits that have been linked to high insolation. However, even at the temperature of WASP-33b's day-side, its atmosphere likely resembles the molecule-dominated atmospheres of other planets, and at the level of ultraviolet irradiation it experiences, its atmosphere is unlikely to be significantly ablated over the lifetime of its star. Here we report observations of the bright star HD 195689, which reveal a close-in (orbital period ~1.48 days) transiting giant planet, KELT-9b. At ~10,170K, the host star is at the dividing line between stars of type A and B, and we measure the KELT-9b's day-side temperature to be ~4600K. This is as hot as stars of stellar type K4. The molecules in K stars are entirely dissociated, and thus the primary sources of opacity in the day-side atmosphere of KELT-9b are likely atomic metals. Furthermore, KELT-9b receives ~700 times more extreme ultraviolet radiation (wavelengths shorter than 91.2 nanometers) than WASP-33b, leading to a predicted range of mass-loss rates that could leave the planet largely stripped of its envelope during the main-sequence lifetime of the host star.

Tables

Table 1. Table 1: Median values and 68% confidence intervals for the physical properties of the KELT-9 system from a global fit to the light curves, radial velocities, and Doppler tomography data, using a custom version of the EXOFAST transit fitting code [ 23 ] (Methods). We use the Yonsei-Yale (YY) isochrones [ 24 ] as well as empirically-calibrated stellar relations [ 25 ] as constraints. In addition, KELT-9 has a measured parallax from Hipparcos [ 26 ] and Gaia [ 27 ] , thereby allowing us to estimate the stellar radius empirically [ 28 ] from its distance, effective temperature, bolometric (wavelength-integrated) flux, and interstellar extinction (Methods). We adopt the values derived using constraints from the YY isochrones and Hipparcos-derived radius as our fiducial system parameters.

Stellar Parameters:
$M_{\ast }$	Mass ($M_{\odot }$)	${2.52}_{-0.20}^{+0.25}$
$R_{\ast }$	Radius ($R_{\odot }$)	${2.362}_{-0.063}^{+0.075}$
$L_{\ast }$	Luminosity ($L_{\odot }$)	${53}_{-10}^{+13}$
${\rho }_{\ast }$	Density (cgs)	$0.2702\pm 0.0029$
$\log g_{\ast }$	Surface gravity (cgs)	$4.093\pm 0.014$
$T_{\mathrm{eff}}$	Effective temperature (K)	$10170\pm 450$
$\left[\mathrm{Fe}/H\right]$	Metallicity	$-0.03\pm 0.20$
$v\sin I_{\ast }$	Rotational velocity (km/s)	$111.4\pm 1.3$
$\lambda$	Spin-orbit alignment (degrees)	$-84.8\pm 1.4$
Planetary Parameters:
$P$	Period (days)	$1.4811235\pm 0.0000011$
$a$	Semi-major axis (AU)	${0.03462}_{-0.00093}^{+0.00110}$
$M_P$	Mass ($M_J$)	$2.88\pm 0.84$
$R_P$	Radius ($R_J$)	${1.891}_{-0.053}^{+0.061}$
${\rho }_P$	Density (cgs)	$0.53\pm 0.15$
$\log g_P$	Surface gravity	${3.30}_{-0.15}^{+0.11}$
$T_{eq}$	Equilibrium temperature (K)	$4050\pm 180$
$⟨F⟩$	Incident flux (109 erg s-1 cm-2)	${61.1}_{-9.8}^{+11.0}$

Full text

B. Scott Gaudi1, Keivan G. Stassun2,3, Karen A. Collins2, Thomas G. Beatty4,5, George Zhou6, David W. Latham6, Allyson Bieryla6, Jason D. Eastman6, Robert J. Siverd7, Justin R. Crepp8, Erica J. Gonzales8, Daniel J. Stevens1, Lars A. Buchhave9,10, Joshua Pepper11, Marshall C. Johnson1, Knicole D. Colon12,13, Eric L. N. Jensen14, Joseph E. Rodriguez6, Valerio Bozza15,16, Sebastiano Calchi Novati15,17, Giuseppe D‘Ago18,19, Mary T. Dumont20,21, Tyler Ellis22,23, Clement Gaillard20, Hannah Jang-Condell22, David H. Kasper22, Akihiko Fukui24, Joao Gregorio25, Ayaka Ito26,27, John F. Kielkopf28, Mark Manner29, Kyle Matt20, Norio Narita26,30,31, Thomas E. Oberst32, Phillip A. Reed33, Gaetano Scarpetta15,17, Denice C. Stephens20, Rex R. Yeigh22, Roberto Zambelli34, B.J. Fulton35,36, Andrew W. Howard35, David J. James37, Matthew Penny1,38, Daniel Bayliss39, Ivan A. Curtis40, D.L. DePoy41, Gilbert A. Esquerdo6, Andrew Gould1,42, Michael D. Joner20, Rudolf B. Kuhn43, Jonathan Labadie-Bartz11, Michael B. Lund2, Jennifer L. Marshall41, Kim K. McLeod44, Richard W. Pogge1, Howard Relles6, Chistopher Stockdale45, T.G. Tan46, Mark Trueblood47, Patricia Trueblood47

{affiliations}

Department of Astronomy, The Ohio State University, Columbus, OH, 43210, USA

Department of Physics and Astronomy, Vanderbilt University, 6301 Stevenson Center, Nashville, TN 37235, USA

Department of Physics, Fisk University, 1000 17th Avenue North, Nashville, TN 37208, USA

Center for Exoplanets and Habitable Worlds, The Pennsylvania State University, 525 Davey Lab, University Park, PA 16802, USA

Department of Astronomy & Astrophysics, The Pennsylvania State University, 525 Davey Lab, University Park, PA 16802, USA

Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA

Las Cumbres Observatory Global Telescope Network, 6740 Cortona Dr., Suite 102, Santa Barbara, CA 93117, USA

Department of Physics, University of Notre Dame, 225 Nieuwland Science Hall, Notre Dame, IN 46556, USA

Niels Bohr Institute, University of Copenhagen, Juliane Maries vej 30, 21S00 Copenhagen, Denmark

Centre for Star and Planet Formation, Geological Museum, Øster Voldgade 5, 1350 Copenhagen, Denmark

Department of Physics, Lehigh University, 16 Memorial Drive East, Bethlehem, PA 18015, USA

NASA Ames Research Center, M/S 244-30, Moffett Field, CA 94035, USA

Bay Area Environmental Research Institute, 625 2nd St. Ste 209 Petaluma, CA 94952, USA

Department of Physics and Astronomy, Swarthmore College, Swarthmore, PA 19081, USA

Dipartimento di Fisica “E. R. Caianiello”, Università di Salerno, Via Giovanni Paolo II 132, 84084 Fisciano (SA), Italy

Istituto Nazionale di Fisica Nucleare, Sezione di Napoli, 80126 Napoli, Italy

IPAC, Mail Code 100-22, Caltech, 1200 E. California Blvd., Pasadena, CA 91125

Istituto Internazionale per gli Alti Studi Scientifici (IIASS), Via G. Pellegrino 19, 84019 Vietri sul Mare (SA), Italy

INAF-Observatory of Capodimonte, Salita Moiariello, 16, 80131, Naples, Italy

Department of Physics and Astronomy, Brigham Young University, Provo, UT 84602, USA

Department of Astronomy and Astrophysics, University of California Santa Cruz, Santa Cruz, CA, 95064, USA

Department of Physics and Astronomy, University of Wyoming, 1000 E. University, Laramie, WY 82071, USA

Department of Physics & Astronomy, Louisiana state University, 202 Nicholson Hall, Baton Rouge, LA 70803, USA

Okayama Astrophysical Observatory, National Astronomical Observatory of Japan, NINS, Asakuchi, Okayama 719-0232, Japan

Atalaia Group & Crow-Observatory, Portalegre, Portugal

National Astronomical Observatory of Japan, NINS, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Kanto, Japan

Graduate School of Science and Engineering, Hosei University, 3-7-2 Kajino-cho, Koganeishi, Tokyo 184-8584, Japan

Department of Physics and Astronomy, University of Louisville, Louisville, KY 40292, USA

Spot Observatory, Nashville, TN 37206 USA

Department of Astronomy, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan

Astrobiology Center, NINS, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan

Department of Physics, Westminster College, New Wilmington, PA, 16172, USA

Department of Physical Sciences, Kutztown University, Kutztown, PA 19530, USA

Società Astronomica Lunae, Castelnuovo Magra 19030, Italy

Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, HI 96822-1839, USA

NSF Graduate Research Fellow

Astronomy Department, University of Washington, Box 351580, Seattle, WA 98195, USA

Sagan Fellow

Observatoire Astronomique de l’Université de Genève, 51 Chemin des Maillettes, 1290 Versoix, Switzerland

ICO, Adelaide, Australia

George P. and Cynthia Woods Mitchell Institute for Fundamental Physics and Astronomy, and Department of Physics and Astronomy, Texas A & M University, College Station, TX 77843-4242, USA

Max Planck Institute for Astronomy, Königstuhl 17, D-69117 Heidelberg, Germany

South African Astronomical Observatory, PO Box 9, Observatory 7935, South Africa

Wellesley College, 106 Central St, Wellesley, MA 02481, USA

Hazelwood Observatory, Victoria, Australia

Perth Exoplanet Survey Telescope, Perth, Australia

Winer Observatory, Sonoita, AZ 85637, USA

The first transiting planets were discovered around cool, solar-type stars[7, 8], primarily because hot stars have few spectral lines and rotate rapidly, making Doppler confirmation of planets more difficult. Only in the past few years have transiting planets been confirmed around hot stars of types early-F and A[9, 10], inspired by the discovery of WASP-33b[1]. That discovery demonstrated that it is possible to confirm transiting planets around rapidly rotating hot stars via a combination of relatively low-precision radial-velocity measurements and Doppler tomography. However, even the hottest of these few A-type transiting-planet host stars only reach temperatures of $\sim$7500 K. Thus, while transit surveys, in particular Kepler[11], have extended the census of planets around low-mass stars, our understanding of planets around massive, hot stars remains poor.

Massive stars cool and spin down as they evolve, enabling precise Doppler measurements. Thus the primary strategy to search for planets around high-mass stars has been surveys of “retired A-stars”[12], high-mass stars that have already evolved into subgiant and giant stars. These stars have revealed a paucity of short-period giant planets relative to sun-like main-sequence stars[12]. One interpretation is that the initial planet population of high-mass stars is similar to that seen in unevolved sun-like stars, but that the short-period planets are subsequently engulfed during the evolution of their parent stars or ablated by the intense irradiation of their host stars while they are still hot[6]. Another interpretation is that these stars actually have masses similar to the Sun[13], implying that the paucity of short-period planets among the retired A-stars is indeed a signature of planet engulfment[14], as sun-like stars do not emit strong ultraviolet radiation with which to ablate their planets[6].

It is therefore critical to assay the population of short-period planets around bona fide high-mass stars while they are still on the main sequence, and then to map the evolution of these planets through to their later evolutionary phases. Although there have been radial-velocity surveys targeting unevolved high-mass stars[15, 16], there are still no known transiting planets around unevolved stars more massive than 2 $\,{M}_{\odot}$ that produce high levels of extreme ultraviolet irradiation.

The Kilodegree Extremely Little Telescope (KELT) is an all-sky survey for planets transiting bright (visual magnitude 8–11) stars[17, 18]. HD 195689 (hereafter KELT-9), exhibited repeating transit-like events of $\sim$0.6% depth with a period of $P\sim$1.48 d (Figure 1), and was selected as a candidate transiting planet (see the Methods). KELT-9’s basic properties (Table 1) include a high effective temperature and rapid rotation. Following the approach that led to the discovery of WASP-33b, we obtained follow-up observations (see Figure 1, Figure 2, and the Methods) that ultimately confirmed KELT-9b as a transiting planet.

KELT-9 is a hot ($T_{\rm eff}\simeq 10,170$ K), massive ($\,{M}_{*}\simeq 2.5~{}\,{M}_{\odot}$) star of spectral type B9.5–A0, with a relatively young age of 300 million years (Methods), comfortably in its 500-million-year main-sequence phase of evolution; it has evidently not yet begun its evolution toward becoming a “retired A star”. Indeed, from comparison with other known planet-hosting stars (Fig. 3), it is clear that KELT-9 is a likely progenitor of at least a subset of the putative “retired A star” hosts of planets detected by radial-velocity surveys. KELT-9, and other A-star transiting planet hosts, thereby provides an important missing link between these samples of planets, and planets detected in more traditional radial-velocity surveys of sun-like stars.

KELT-9 is only the fifth A-type star known to host a transiting giant companion, and is by a significant margin the hottest, most massive, and most luminous known transiting giant planet host. The host star also has the brightest $V$-band magnitude of any transiting hot Jupiter host, being slightly brighter in $V$ than HD 209458b[7, 8].

Given the high stellar luminosity and close orbit, the planet receives a large stellar insolation flux (Table 1). As a result, it has an extremely high equilibrium temperature, assuming zero albedo and perfect heat redistribution, of $\sim$4050 K. This is as hot as a late K-type star[5], and thus we expected a large thermal emission signal, which we easily confirmed with our $z^{\prime}$-band detection of the secondary eclipse with a depth of $\sim 0.1\%$ (Figure 1). This measurement implies an even hotter day-side temperature of $\sim 4600\pm 150$ K, likely indicating poor redistribution of energy to the night side of the planet and a temperature closer to that of a mid-K star. The planet is also extremely inflated relative to theoretical models, with a radius of $\sim 1.9~{}R_{\rm J}$. Poor redistribution of heat and radius inflation (both of which are also observed in WASP-33b[1, 2]), have been linked to high stellar insolation[3, 4], although the exact physical mechanisms remain uncertain.

Thus, although other transiting planets have been found around A-type stars, and indeed some (such as WASP-33b) have been discovered with the temperatures of low-mass stars, the KELT-9 planet and host star are hotter by $\sim 1000$ K and $\sim 2500$ K, respectively, than any other known transiting gas-giant system. Consequently, one expects all of the opacity sources on the day-side to be atomic, as in a K-type star. In contrast, all other known transiting planets, which have day-side temperatures of $<3850$ K[5], are cool enough to contain molecular species. Furthermore, the very high flux of extreme ultraviolet radiation (wavelengths shorter than 91.2 nanometers) from KELT-9, $\sim 700$ times higher than WASP-33, may lead to unique photochemistry in the planet atmosphere[19]. This places KELT-9b in a qualitatively new regime of planetary atmospheres, and makes characterization of the atmosphere of KELT-9b particularly compelling.

Fortunately, the brightness of KELT-9 and the extreme properties of its transiting planet make the prospects for detailed characterization of this system promising. Observations using ground-based facilities, Spitzer, the Hubble Space Telescope (HST), and ultimately the James Webb Space Telescope, will allow for the measurement of the phase-resolved spectrum of its thermal emission from the far-optical through the infrared ($\sim$30 $\mu$m). The low surface gravity of KELT-9b combined with the high temperature lead to one of the largest atmospheric scale heights of any known transiting planet (Figure 3). The expected few-percent variations in the transmission spectrum during the primary transit should be easily detectable.

The future evolution of the KELT-9 system is uncertain but certainly interesting. The high ultraviolet flux impinging on KELT-9b likely means that its atmosphere is being significantly ablated, with an estimated mass loss rate[6] of $\sim 10^{10}$–$10^{13}~{}{\rm g~{}s^{-1}}$. At the upper end of these rates, the planet may be completely stripped of its outer envelope in $<600$ Myr, roughly the time scale for the host to evolve from the main-sequence to the base of the red giant branch (see Methods). These estimates are so uncertain because the physics of planet evaporation is extraordinarily complex, particularly due to the unknown magnitude of the host star activity, and thus very high-energy (extreme ultraviolet and X-ray) non-thermal radiation and stellar wind. In any event, even the lower end of mass-loss rates we estimate should be easily detectable and measurable with HST observations.

On the other hand, as KELT-9 eventually exhausts its core hydrogen supply in $\sim$200 million years, it will grow from its current radius of $\sim$2.4 $R_{\odot}$ to a radius of $\sim 5~{}R_{\odot}$, while simultaneously cooling to $T_{\rm eff}\sim 8000$ K. Soon after, it will rapidly traverse the ‘subgiant branch’ whereby it will cool to $\sim 5000$ K and expand to $\sim 8~{}R_{\odot}$. As it reaches the base of the red giant branch, the stellar surface of KELT-9 will encroach upon the orbit of KELT-9b. Exactly what will happen to the star and planet at this point is far from clear. If the mass loss from ablation is lower than estimated above, the planet may remain intact as a gas giant, and it will be engulfed by its host star, perhaps leading to a bright transient event[20], and an unusually rapidly rotating red giant star with enriched lithium provided by the dissolved planetary companion[21]. On the other hand, if the planet possesses a rocky core and is fully ablated before this point, this could imply the existence a population of close-in, super-Earth remnant cores orbiting subgiant stars, a prediction that could be testable with the upcoming Transiting Exoplanet Survey Satellite (TESS) mission.

More detailed theoretical studies are needed to provide a clearer picture of the future evolution of the KELT-9 system and its analogs, and provide testable predictions. Further follow-up observations using ground and space-based telescopes will test models of heat redistribution, radius inflation, unusual photochemistry, and rapid ablation of planetary atmospheres. The leap from WASP-33b to KELT-9b should invigorate further exploration of the planet population of even higher-mass host stars, complementing efforts to discover planets orbiting ever lower-mass host stars[22]. The KELT-9 system provides an important benchmark system for understanding the nature of planetary systems around massive stars, from birth to death.

References

KELT-North Observations and Photometry

KELT-North field 11 is centered on $\alpha=$ 19h 27m 00s, $\delta=$ 31∘ 39′ 56$.\!\!^{\prime\prime}$16 (J2000) and was observed 6001 times from UT 2007 May 30 to UT 2013 June 14. Following the standard KELT candidate selection strategy[29], we reduced the data and extracted $\sim 116,000$ light curves from the east orientation and $\sim 138,000$ light curves from the west. The combined east and west light curves that passed the reduced proper motion cut[1] were searched for transiting exoplanet candidates. One bright ($V\sim 7.6$) candidate, KC11C043952 (HD 195689, TYC 3157-638-1, 2MASS J20312634+3956196) located at $\alpha=$ 20h 31m 26$.\!\!^{\mathrm{s}}$35401, $\delta=$ 39∘ 56′ 19$.\!\!^{\prime\prime}$7744 (J2000), robustly passed our selection criteria[29] making it a top candidate. Hereafter, we refer to the candidate host star as “KELT-9” and the candidate planet as “KELT-9b”. KELT-9 is also located in KELT-North field 12, which is centered on $\alpha=$ 21h 22m 52$.\!\!^{\mathrm{s}}$8, $\delta=$ 31∘ 39′ 56$.\!\!^{\prime\prime}$16 (J2000). The field was observed 5,700 times from UT 2007 June 08 until UT 2013 June 14. Although not originally used to select KELT-9 as a candidate, the signal in the light curve from field 12 bolstered KELT-9 as a strong candidate.

The KELT-North field 11 phased KELT-9 discovery light curve is shown in the top panel of Fig. 1 in the main manuscript. The field 11 light curve shows apparent out-of-transit variations (OOTVs), while these are absent in the field 12 light curve. We find that the source of the field 11 OOTVs is due to variability, possibly caused by saturation, of a star a few arc-minutes southeast of KELT-9 that is blended in the field 11 KELT-9 aperture. Given the magnitude of the variability, we suspect the contaminating source is the bright object HD 195728, rather than one of the fainter neighbors. The variation is long-term and not periodic, so it tends to inject power at and near aliases of our sampling rate. The variability from the neighbor does not affect the field 12 light curve because the point spread function of KELT-9 in field 12 is smaller and is elongated in a different direction than in field 11. As a result, KELT-9 is minimally blended with the bright neighbor HD 195728, as well as the other faint neighbors near HD 195728.

In short, the apparent out-of-transit variations (OOTVs) seen in the field 11 light curve are not due to intrinsic variability of KELT-9 itself, and thus do not lead us to question its reality as a bona fide planet-host candidate.

Following the discovery of the primary transit signal, we carried out an intensive photometric followup campaign through which we obtained a total of 17 primary and 7 secondary transit light curves. These will be described in a forthcoming paper (Collins et al. 2017, in preparation), and are used in the global fit of the system presented below. The combined, binned follow-up light curve is shown in Figure 1 of the main manuscript in order to highlight the statistical power of this follow-up dataset. We only fit to the individual data; the binned data are shown only for illustration.

Spectroscopic Follow-up

To constrain the mass and enable eventual Doppler tomographic (DT) detection of KELT-9b, we obtained a total of 115 spectroscopic observations of the host star with the Tillinghast Reflector Echelle Spectrograph (TRES) on the 1.5 m telescope at the Fred Lawrence Whipple Observatory, Arizona, USA. Each spectrum delivered by TRES has a spectroscopic resolution of $\lambda/\Delta\lambda=41000$ over the wavelength range of $3900-9100$ Å over 51 echelle orders. This includes 40 observations covering the entire orbital phase to constrain the mass of the planet, and 75 observations made in-transit over three epochs to perform the tomographic line profile analysis. We measured the relative radial velocity from 104 of the observations (see Extended Data Table \spacing1A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host) and used a total of 43 out-of-transit RVs (40 plus one out-of-transit RV from each of the spectroscopic transit observations) to constrain the planet’s orbit and mass. The phased radial velocities are displayed in Figure 1 of the main article.

Three spectroscopic transits of KELT-9 were observed with TRES on 2014-11-15, 2015-11-06, and 2016-06-12. The line broadening kernel is derived from each spectrum via a least-squares deconvolution[2, 1, 3]. The planetary shadow is seen crossing the stellar surface on all three nights, as shown in Extended Data Figure 3. In addition, the rotational profiles allowed us to accurately determine a rotational velocity of $\,{v}\sin{I_{*}}=111.4\pm 1.3\,\mathrm{km\,s}^{-1}$.

Host Star Properties

Extended Data Table \spacing1A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host lists various properties and measurements of KELT-9 collected from the literature and derived in this work. The data from the literature include the four monochromatic near-UV fluxes from the Catalog of Stellar UV Fluxes[4], $UBV$ photometry[5], optical fluxes in the $B_{\rm T}$ and $V_{\rm T}$ passbands from the Tycho-2 catalog[6], $I_{\rm C}$ from the TASS catalog[7], near-infrared (IR) fluxes in the $J$, $H$ and $K_{\rm S}$ passbands from the 2MASS Point Source Catalog[8, 9], near- and mid-IR fluxes in four WISE passbands[10, 11], distances from Hipparcos[12] and Gaia[27], and proper motions from the NOMAD catalog[13, 14].

SED Analysis

We construct an empirical, broad-band spectral energy distribution (SED) of KELT-9, shown in Extended Data Figure 1. We use the 17 photometric measurements from the literature discussed in Section Host Star Properties and shown in Extended Data Table \spacing1A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host. In total, the observed SED spans the wavelength range 0.16–22 $\mu$m. We fit this observed SED to Kurucz stellar atmosphere models[15]. For simplicity we adopted a fixed $\log{g_{\star}}=4.0$, based on the light curve transit analysis. The fit parameters were thus the effective temperature ($T_{\rm eff}$), the metallicity ($\left[{\rm Fe}/{\rm H}\right]$), the extinction ($A_{V}$), and the overall flux normalization. The maximum permitted extinction was set to $A_{V}=0.30$ based on the total line-of-sight extinction in the direction of KELT-9 from Galactic dust maps[16].

Stellar Parameters from SED

The best fit model has a reduced $\chi^{2}$ of 2.56 for 13 degrees of freedom (17 flux measurements, 4 fit parameters). We find $A_{\rm V}$ = $0.09\pm 0.05$, $\left[{\rm Fe}/{\rm H}\right]=0.0\pm 0.2$, and $T_{\rm eff}=9560\pm 550$ K. We note that the quoted statistical uncertainties on $A_{\rm V}$ and $T_{\rm eff}$ are likely to be slightly underestimated because we have not accounted for the uncertainty in the value of $\log{g_{\star}}$ used to derive the model SED, although this parameter generally does not strongly affect the overall shape of the SED, and moreover $\log{g_{\star}}$ is strongly constrained from the light curve transit analysis.

We can integrate the best-fit SED to obtain the (unextincted) bolometric flux at Earth, $F_{\rm bol}=3.18(\pm 0.09)\times 10^{-8}$ erg s*-1* cm*-2*. Together with the best-fit $T_{\rm eff}$ and the distance newly provided by the Gaia parallax, we obtain a direct constraint on the stellar radius of R\textsubscript{\star}=2.37\pm 0.35 $\,{R}_{\odot}$. From the distance provided by the Hipparcos parallax, we obtain a direct constraint[28] on the stellar radius of R\textsubscript{\star}=2.17\pm 0.33 $\,{R}_{\odot}$. As noted in the footnote of Extended Data Table \spacing1A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host, we generally believe the Hipparcos-derived distance and stellar radius to currently be more reliable, and use this for most of the analysis in this paper.

Stellar Models and Age

With $T_{\rm eff}$ and $\log{g_{\star}}$, and an estimated stellar mass from the global analysis (see below), we can place the KELT-9 system in the Hertzsprung-Russell diagram for comparison with theoretical stellar evolutionary models (Extended Data Figure 2). From the values obtained with the initial SED fit, we infer a system age of $\approx$0.4 Gyr; the final age estimate using the final global fit parameters is $\approx$0.3 Gyr. Thus, it is clear that the KELT-9 system is nearly unevolved from the zero-age main-sequence, and in any event is at an early stage of evolution well before the “blue hook” transition to the subgiant and eventual red giant evolutionary phases.

Global System Fit

We determined the physical and orbital parameters of the KELT-9 system by jointly fitting 17 primary and 7 secondary light curves, 43 TRES out-of-transit RVs, and 3 Doppler tomographic data sets (see the section Doppler tomographic model below). To perform the global fit, we used MULTI-EXOFAST (MULTIFAST hereafter), which is a custom version of the public software package EXOFAST[23]. MULTIFAST first performs an AMOEBA[17] best fit to each of the RV and light curve data sets individually to determine uncertainty scaling factors. The uncertainties are scaled such that the probability that the $\chi^{\,2}$ for a data set is larger than the value we achieved, $P\left(>\chi^{\,2}\right)$, is $0.5$, to ensure the resulting parameter uncertainties are roughly accurate. The resulting RV uncertainty scaling factor is 1.23. The DT uncertainties were scaled based on the $\chi^{\,2}$ of the out-of-transit data relative to the median value. The uncertainties of the UT 2014-11-15, UT 2015-11-06, and UT 2016-06-12 DT observations were scaled by 0.79, 0.78, and 0.79, respectively. Finally, MULTIFAST performs a joint AMOEBA model fit to all of the datasets and executes Markov Chain Monte Carlo (MCMC), starting at the global best fit values, to determine the median and 68% confidence intervals for each of the physical and orbital parameters. Siverd et al.[29] provide a more detailed description of MULTIFAST, except the Doppler tomographic model implementation, which was newly implemented as part of this work.

Doppler Tomographic Model

To model the Doppler tomographic signal, we construct and integrate our own models into the MULTIFAST fitting process[18]. We treat the Doppler shadow of the planet as a combination of three different broadening mechanisms which we account for consecutively. Our shadow model begins as a Gaussian profile, with a standard deviation equal to the mean inherent spectral line width (in velocity space) of an equivalent non-rotating star. Second, we convolve this base Gaussian profile with a second Gaussian of width $\sigma=c/R$ to account for the finite spectral resolution of the TRES spectrograph ($R\approx 44,000$). Third, and finally, we convolve the resulting shadow model with a normalized rotational broadening kernel.

For our rotational broadening kernel we use the kernel given by Equation 18.14 in Gray[19]. For simplicity we assume that the linear limb-darkening coefficient ($\epsilon$) for the kernel is equal to zero. We set the velocity width of the kernel equal to $(R_{p}/R_{*})\,\,{v}\sin{I_{*}}$, to represent the fraction of the stellar rotational surface actually obscured by the planet.

We normalize the resulting shadow model using the depth of a $V$-band transit light curve for the system at each observation time, so that the integrated light obscured by the shadow matches the actual depth of the transit. To calculate the trajectory of the shadow in time-velocity space, we follow the method given by Equations 7, 8, and 10 in Collier Cameron et al.[20]

Besides the physical parameters already included in the transit model, our Doppler tomography model therefore has three free parameters: the rotation velocity of the host star ($\,{v}\sin{I_{*}}$), the inherent line width of the non-rotating stellar spectrum ($\sigma_{0}$), and the spin-orbit angle ($\lambda$). Since KELT-9 has $\,{v}\sin{I_{*}}=111.4\pm 1.3\,\mathrm{km\,s}^{-1}$, $\sigma_{0}$ is essentially negligible, so we use a prior on $\sigma_{0}=10.0\pm 1.0\,\mathrm{m\,s}^{-1}$ (i.e. a value near zero that avoids significant numbers of negative $\sigma_{0}$ MCMC trials) to improve convergence of the MCMC chains. To evaluate the goodness-of-fit for a particular set of model parameters, we use the $\Delta\chi^{2}$ between the predicted shadow model and the observed Doppler tomographic observations. Since the observations themselves are super-sampled below the instrumental resolution of the spectrograph during the data reduction process, we divide the resulting $\Delta\chi^{2}$ value by a factor of $(c/R)/\Delta v$, where $c$ is the speed of light, $R$ is the instrumental spectral resolution, and $\Delta v$ is the velocity interval between the individual points in the super-sampled spectra. This reduction of the $\Delta\chi^{2}$ accounts for the fact that due to the instrumental spectral resolution, adjacent points in the super-sampled spectra were not truly independent.

Global Model Results

We adopt a fiducial model (Model 1) with YY constraints, a fixed circular orbit, a fixed RV slope $\dot{\gamma}=0$, and a Hipparcos-based R$\star$ prior, and compare the results to those from ten other global models that systematically explore the results of differing constraints. The posterior median parameter values and 68% confidence intervals are shown in Extended Data Table \spacing1A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host for the initial TTV and final ten global model fits. The KELT-9 fiducial model indicates the system has a host star with mass M\textsubscript{\star}=2.52\,{M}_{\odot}, radius R\textsubscript{\star}=2.362\,{R}_{\odot}, and effective temperature $T_{\rm eff}=10,170$ K, and an extremely hot planet with $T_{\rm eq}=4050$ K, mass $M\textsubscript{P}=2.88M\textsubscript{J}$, and radius $R\textsubscript{P}=1.891\,M_{\rm J}$.

The initial TTV model (Model 0) parameter median values are well within $1\sigma$ of the fiducial model, and the uncertainties are essentially identical to those of the fiducial model results, except the ephemeris parameter uncertainties ($T_{C}$ and $P$) are $\sim$ 25% larger due to the additional seven secondary transits in the fiducial model and the linear ephemeris constraint.

We also explored several other models involving different choices of observational constraints and free-fit parameters, which will be described in a forthcoming paper (Collins et al. 2017, in preparation) but are summarized in Extended Data Table \spacing1A giant planet undergoing extreme ultraviolet irradiation by its hot massive-star host.

In summary, we find that all combinations of stellar constraints result in system parameter values that are within $1\sigma$, and in almost all cases, well within $1\sigma$ of the fiducial model.

False Positive Analysis

KELT-9 was a sufficiently unusual system that we were especially vigilant about ruling out false positive scenarios. We have many lines of evidence that rule out essentially all viable false positive scenarios. First, false positives around rapidly rotating stars can be ruled out via relatively imprecise RV measurements, high precision light curves, and a positive Rossiter-McLaughlin[21, 22] (RM) or DT detection[1, 23]. An upper limit on the Doppler signal with relatively imprecise RV precisions of a few hundred ${\rm m~{}s^{-1}}$ can rule out low-mass stars and brown dwarfs as the occulter of the primary star. Precise follow-up light curves, along with the $\,{v}\sin{I_{*}}$ measured from the spectra, can be used to predict the magnitude, impact parameter, and duration of the DT signal (although not the direction $\lambda$). Thus a measurement of the DT signal that is consistent with the light curve effectively confirms the planet interpretation.

In the case of KELT-9, our first spectroscopic measurement during transit yielded a very weak RM signal, which was sufficiently noisy that we considered it could simply be due to the relatively large uncertainties of the RVs. This lack of a RM signal would be surprising for a transiting planet given the transit depth and large $\,{v}\sin{I_{*}}$ of the star. We thus originally concluded that the system was likely a false positive. However, we decided to proceed with a full DT analysis, which showed tentative evidence for a planet shadow that was nearly coincident with the projected stellar rotation axis, thus possibly explaining the small RM signal. Improved reduction methods and two additional DT measurements definitively showed that the planet does indeed transit nearly along the projected stellar rotation axis, as shown in Extended Data Figure 3.

Second, we were also concerned about the lack of a definitive measurement of the Doppler reflex signal. Our initial fits to a subset of the data presented here provided only upper limits to $K$, and thus we were originally not able to measure the planet mass. However, after acquiring additional data, we were ultimately able to measure $K$ to roughly $\sim 30\%$ precision. Indeed, a Lomb-Scargle[24, 25] periodogram of the out-of-transit RV data yielded several significant signals. The two signals with the highest power are at long periods of $\sim 100$ and $\sim 200$ days (depending on the exact choice of datasets that are used), which are likely aliases of each other, and may be due to another planet, intrinsic long-term stellar activity, or may simply be due to systematic errors in the RVs. However, the signal with the next highest power has a ephemeris of $T_{\rm 0}=2456969.7901$ and $P=1.4811228$. This period is consistent with the final period we derive from our follow-up light curves to within $\sim 10^{-5}$ d, and projecting the ephemeris from our follow-up light curves forward yields a value of $T_{\rm 0}$ that is consistent with that from the RV periodogram to within $\sim 10^{-4}$ d. We are therefore confident that the reflex RV signal is real and due to the transiting planet.

Finally, there are several other pieces of evidence that support the planet hypothesis. (1) The stellar density we infer from the global fit to the light curve (which is essentially a direct observable) is consistent with what one would expect for an unevolved A0 star with the spectroscopically-measured temperature. (2) The limb darkening in the redder bandpasses are noticeably smaller by eye than planet transits of a cooler star, as one would expect for a hot star and bandpasses near the Rayleigh-Jeans part of the SED. Although we do not fit for the limb darkening, the fact that the light curves spanning from $U$ to $z$ are well-fit by the smaller limb darkening coefficients that are predicted for a star of this temperature and surface gravity (as compared to, e.g., those expected for a solar-type star) provides quantitative support for this qualitative conclusion. (3) Adaptive optics observations (Collins et al. 2017, in preparation) do not reveal a blended stellar companion that could cause a false positive, down to the mass of $\sim 0.14~{}\,{M}_{\odot}$ with a projected separation of $\gtrsim 200$ AU. (4) We detect a secondary eclipse in the $z$ band with a depth that is consistent with what one would expect given the amount of stellar irradiation that the planet receives. (5) The basic consistency between the model fits using various constraints (YY, Torres, and Hipparcos and Gaia-inferred radii) also provides support for our interpretation.

We therefore conclude that, despite the very unusual and extreme nature of the system, all available data are consistent with the interpretation that KELT-9 is being transited by an extremely irradiated, highly inflated planet on a near-polar orbit of only $\sim 1.5$ d.

References