Molecules with ALMA at Planet-forming Scales (MAPS). A Circumplanetary Disk Candidate in Molecular Line Emission in the AS 209 Disk

We report the discovery of a circumplanetary disk (CPD) candidate embedded in the circumstellar disk of the T Tauri star AS 209 at a radial distance of about 200 au (on-sky separation of 1."4 from the star at a position angle of $161^\circ$), isolated via $^{13}$CO $J=2-1$ emission. This is the first instance of CPD detection via gaseous emission capable of tracing the overall CPD mass. The CPD is spatially unresolved with a $117\times82$ mas beam and manifests as a point source in $^{13}$CO, indicating that its diameter is $\lesssim14$ au. The CPD is embedded within an annular gap in the circumstellar disk previously identified using $^{12}$CO and near-infrared scattered light observations, and is associated with localized velocity perturbations in $^{12}$CO. The coincidence of these features suggests that they have a common origin: an embedded giant planet. We use the $^{13}$CO intensity to constrain the CPD gas temperature and mass. We find that the CPD temperature is $\gtrsim35$ K, higher than the circumstellar disk temperature at the radial location of the CPD, 22 K, suggesting that heating sources localized to the CPD must be present. The CPD gas mass is $\gtrsim 0.095 M_{\rm Jup} \simeq 30 M_{\rm Earth}$ adopting a standard $^{13}$CO abundance. From the non-detection of millimeter continuum emission at the location of the CPD ($3\sigma$ flux density $\lesssim26.4~\mu$Jy), we infer that the CPD dust mass is $\lesssim 0.027 M_{\rm Earth} \simeq 2.2$ lunar masses, indicating a low dust-to-gas mass ratio of $\lesssim9\times10^{-4}$. We discuss the formation mechanism of the CPD-hosting giant planet on a wide orbit in the framework of gravitational instability and pebble accretion.

origin: an embedded giant planet. We use the 13 CO intensity to constrain the CPD gas temperature and mass. We find that the CPD temperature is 35 K, higher than the circumstellar disk temperature at the radial location of the CPD, 22 K, suggesting that heating sources localized to the CPD must be present. The CPD gas mass is 0.095M Jup 30M Earth adopting a standard 13 CO abundance. From the non-detection of millimeter continuum emission at the location of the CPD (3σ flux density 26.4 µJy), we infer that the CPD dust mass is 0.027M Earth 2.2 lunar masses, indicating a low dust-to-gas mass ratio of 9 × 10 −4 . We discuss the formation mechanism of the CPD-hosting giant planet on a wide orbit in the framework of gravitational instability and pebble accretion.

INTRODUCTION
One of the best and most direct ways to study planet formation processes is to observe young planets while they are forming. Although this task has long been challenging, the situation is rapidly changing. In particular, recent observations with the Atacama Large Millimeter/submillimeter Array (ALMA) have demonstrated the possibility of detecting young planets by spatially and kinematically resolving the characteristic gas flows around forming planets (see review by Pinte et al. 2022), including localized velocity perturbations associated with planet-driven spirals (Pinte et al. 2018;) and large-scale meridional flows falling on to planet-hosting gaps (Teague et al. 2019a;Yu et al. 2021).
As a giant planet forms, it opens a radial gap in its natal circumstellar disk (Lin & Papaloizou 1986). Within the Hill (or Bondi) sphere of the planet, a circumplanetary disk (CPD) forms in order to conserve the angular momentum of the material accreted from the circumstellar disk (Korycansky et al. 1991;Ward & Canup 2010). The growth of a planet after gap opening is controlled by gas accretion through its CPD (Lubow et al. 1999). CPDs also play a crucial role in satellite formation. Indeed, the prograde, nearly circular and co-planar orbits of regular satellites of Jupiter and Saturn suggest that their formation must have happened within CPDs. Despite their importance, however, many attempts to search for CPDs have yielded non-detections (Isella et al. 2014;Pérez et al. 2019;Pineda et al. 2019;Andrews et al. 2021), and the detections made to date are limited in millimeter continuum observations (Isella et al. 2019;Benisty et al. 2021;Wu et al. 2022). Consequently, fundamental properties of CPDs, such as their sizes (see discussion in Paardekooper et al. 2022) and whether they are indeed rotationally-supported "disks" or instead pressure-supported envelopes (Szulágyi et al. 2016;Fung et al. 2019) are not fully understood.
In this Letter, we report the discovery of a CPD candidate embedded in the circumstellar disk of AS 209, * NASA Hubble Fellowship Program Sagan Fellow identified using 13 CO J = 2 − 1 emission. AS 209 is a T Tauri star located in the outskirts of the Ophiuchus star-forming region at a distance of 121 pc (Gaia Collaboration et al. 2020), with a stellar mass of 1.2 M and an age of 1-2 Myr (Andrews et al. 2009;Öberg et al. 2021, and references therein). In (sub-)millimeter continuum emission, the AS 209 disk is 140 au in size and has pronounced multiple concentric rings and gaps (Fedele et al. 2018;Huang et al. 2018;Guzmán et al. 2018;Sierra et al. 2021). Hydrodynamic planet-disk interaction simulations showed that the continuum gaps can be explained by one or more planets embedded in the disk (Fedele et al. 2018;Zhang et al. 2018). The gas disk probed by 12 CO emission extends to 300 au, much further out than the continuum disk (Guzmán et al. 2018). In 12 CO an annular gap is seen centered at a radius of 200 au, sandwiched by two bright rings whose intensities peak at 168 and 245 au (Guzmán et al. 2018;Law et al. 2021a). Near-infrared scattered light observations also revealed this gap around 200 au (Avenhaus et al. 2018). This 12 CO and near-infrared gap at 200 au is within which the CPD candidate is found in this study.
This Letter is organized as follows. In Section 2, we summarize the observations. In Section 3, we present the 12 CO data, which reveals an annular gap and localized kinematic perturbations, and the 13 CO data, which shows a CPD candidate. We also present a null detection of C 18 O and continuum emission at the location of the CPD candidate. In Section 4, we place constraints on the size, temperature, and mass of the CPD, and discuss the potential formation pathway of the planet. We summarize our findings and state our conclusions in Section 5.

OBSERVATIONS
The 12 CO, 13 CO, and C 18 O data used in this study were obtained as part of the ALMA Large Program Molecules with ALMA at Planet-forming Scales (MAPS;2018.1.01055.L;Öberg et al. 2021). We refer the reader toÖberg et al. (2021) for the observational setup and calibration process, and Czekala et al. (2021) for the imaging process including the Jorsater & van Moorsel (1995, hereafter JvM) correction, which is used to prop-erly scale the CLEAN residual map into consistent units of Janskys per CLEAN beam (instead of Janskys per dirty beam) and is crucial for correctly estimating the flux density of faint extended emission. Throughout this paper, we use the robust = 0.5 weighted, JvMcorrected images 1 unless stated otherwise. The 12 CO, 13 CO, and C 18 O images have synthesized beam sizes of 134 × 100 mas, 140 × 104 mas, and 141 × 105 mas with position angles of −89.2 • , −89.1 • , and −88.6 • , respectively. The rms noise measured in line-free channels is 0.605 mJy beam −1 , 0.447 mJy beam −1 , and 0.337 mJy beam −1 , respectively.
While the JvM correction allows the CLEAN model and residual image to have consistent units, Casassus & Carcamo (2022) cautioned that it may exaggerate the peak signal-to-noise ratio of restored images. As such, we carry out our analyses using both JvMcorrected images (presented in the main Sections) and JvM-uncorrected images (presented in the Appendix) in order to ensure that the CPD detection is statistically significant regardless of which image we use.
For the analysis of continuum data, we combined the continuum data from ALMA Large Programs MAPS and Disk Substructures at High Angular Resolution Project 2 (DSHARP; 2016.1.00484.L; Andrews et al. 2018), using CASA version 6.4. We refer the reader to Sierra et al. (2021) for the imaging of the MAPS continuum data and Andrews et al. (2018) for the observational setup and calibration process of the DSHARP continuum data. To account for any source proper motions or atmospheric/instrumental effects between observations, we aligned each dataset to a common phase center using the fixvis and fixplanets tasks to apply the necessary phase shifts. We then made a composite image at 240 GHz, the mean frequency of the DSHARP and MAPS continuum data. The continuum image was generated with a Briggs weighting scheme with robust = 0.5 and a 65 × 20 mas FWHM Gaussian taper (at PA = 0 • ) in order to enhance the point source sensitivity near the location of the CPD candidate. The synthesized beam has FWHM of 106 × 94 mas at a position angle of −69.2 • degrees. We find that the PSF of the combined data is complex due to the combined u, v coverage of the datasets. For that reason, we do not perform the JvM correction on the continuum data. The rms noise within an elliptical annulus ±0. 1 around the 1 Publicly available from https://alma-maps.info/data.html 2 In Andrews et al. (2018), the continuum data obtained as part of the DSHARP program was combined with archival data from programs 2013.1.00226.S (Huang et al. 2016) and 2015.1.00486.S (Fedele et al. 2018). radial location of the CPD candidate (i.e., between 1. 6 and 1. 8 in the deprojected plane) is 8.8 µJy beam −1 . The DSHARP and MAPS observations were taken about two years apart. Between the two observations, the CPD candidate is expected to have moved about 8 mas on the sky, less than 10% of the synthesized beam of the combined continuum data. We thus opt not to account for the orbital motion of the CPD candidate.

An Annular Gap and Localized Kinematic
Perturbations in 12 CO We present selected 12 CO channel maps in Figure  1 (see Figure 7 in Appendix A for additional channel maps). The 12 CO emission reveals a 78 au-wide gap (determined by the distance between the adjacent peaks in the radial intensity profile; Law et al. 2021a) around 200 au, which is most clearly seen along the semimajor axis of the disk (see 5.6 − 6.0 km s −1 in Figure  1). The radial location of the gap is consistent with what is previously reported in Guzmán et al. (2018) and  using independent 12 CO datasets, and in Avenhaus et al. (2018) using near-infrared scattered light images.  showed that the 12 CO gap is associated with deviations in the rotational motion of the gas with variations of up to 5% relative to the background Keplerian rotation, which arises from the steep radial gas pressure gradient at the edges of the gap. While we defer more comprehensive kinematic studies to a forthcoming paper (Galloway-Sprietsma et al., in prep.), comparable rotational velocity deviations are found around the 12 CO gap using the MAPS 12 CO data.
In addition to the annular gap, 12 CO reveals a strong localized (both in space and velocity) perturbation on the southern side (see 4.6 − 5.0 km s −1 in Figure 1), reminiscent of a so-called velocity kink induced by planets (Pinte et al. 2018). Similar velocity perturbations are found at the same location and velocity in independent archival 12 CO datasets presented in Guzmán et al. (2018) and . The velocity perturbation is seen near the semi-minor axis of the disk, suggesting that the perturbation is associated with radial and/or vertical motions rather than rotational motions (Teague et al. 2019b). We note that a similar velocity perturbation is not seen on the northern side of the disk, highlighting the localized nature of the perturbation. tern arising from Keplerian rotation of the AS 209 disk, we find a point source emission at a projected distance of 1. 4 and a position angle of 161 • (east from north). The offset from the star (assumed to be at the center of the concentric continuum rings) is (∆RA, ∆Dec) = (0. 450±0.004, −1. 332±0.003). Assuming that the point source is embedded in the midplane of the AS 209 disk, the deprojected distance between the point source and the central star is 1. 7 ( 206 au). It is worth pointing out that the 13 CO point source is located at the center of the 12 CO annular gap and is coincident in both radius and azimuth with the localized velocity perturbations seen in 12 CO. In Section 4.1, we will discuss the potential (common) origin of these features. The 13 CO data is limited by the 200 m s −1 velocity resolution set by the MAPS program. In the fiducial image cube, the point source is found in two adjacent channels, 4.7 km s −1 (Figure 2 left) and 4.9 km s −1 (Figure 2 right). The peak intensities at the two channels are 5.75 and 5.85 mJy beam −1 , respectively. While we cannot image at a finer channel spacing than the ve-locity resolution, we produce an additional image cube with the same channel spacing of 200 m s −1 , but with the starting velocity offset by half of the velocity resolution (100 m s −1 ). The middle panel in Figure 2 shows the channel map at 4.8 km s −1 from the half-channelshifted image cube. The point source is more clearly seen, with the peak intensity of 9.62 mJy beam −1 (rms noise = 0.543 mJy beam −1 ). Simply comparing the peak intensity with the rms noise, the point source in the three velocity channels is detected above 12, 18, and 13σ levels, respectively.

Point Source Emission in 13 CO
Although the 13 CO emission from the AS 209 disk is less extended than the 12 CO emission, 13 CO emission is detected out to 2 (in deprojected distance), beyond the radial location of the point source. If the point source is embedded in the midplane within the 12 CO gap, it is thus possible that some of the emission toward the point source originates from the residual gas within the gap. To assess the potential contamination from the gas in the gap, in Figure 3 we present a radial profile of the integrated intensity averaged over  Azimuthally-averaged radial profile of the 13 CO integrated intensity excluding the point source. The gray and red shaded regions show the standard deviation of the integrated intensity in each 0. 05 radial bin. The point source is seen well above the azimuthally-averaged integrated intensity. The gap at 0. 5 − 0. 6 is consistent with the 13 CO gap previously reported by Law et al. (2021a). The size of the major axis of the synthesized beam is shown with the black horizontal bar.
the azimuthal region ±2 • around the point source. We compare that to an azimuthally-averaged, radial integrated intensity profile excluding the ±2 • region around the point source. At the deprojected radial distance of 1. 7, the integrated intensity of the point source is 2.54 mJy beam −1 km s −1 , while the integrated intensity excluding the point source is 1.15 mJy beam −1 km s −1 with a standard deviation of 0.33 mJy beam −1 km s −1 .
The brightness temperature of the 13 CO emission within the gap (excluding the point source) is 8 K. For a reasonable gas temperature of 20 K (so that CO does not freeze), the gap must be optically thin (τ gap 0.5) because the brightness temperature is lower than the gas temperature. The fact that the integrated intensity toward the point source does not increase with the beam size (see Appendix B) further supports that the gap is optically thin in 13 CO and the point source dominates the emission. If we assume that a similar amount of gas is present around the point source to elsewhere within the gap, the contamination from the gas in the gap is thus expected to be 1.15/2 = 0.575 mJy beam −1 km s −1 (because only half of the gas in front of the point source would contribute). Subtracting this from the integrated intensity toward the point source, we find that the point source is detected at a 6.0σ significance. We carry out the same analysis using the JvM-uncorrected image and find that the point source is detected at a 4.4σ significance (see Appendix C).
In order to determine whether or not the point source is spatially resolved, we follow Isella et al. (2019) and Benisty et al. (2021) and image the data using various robust parameters ranging from −1 to 2. The synthesized beam size, rms noise, Gaussian fit to the point source emission are summarized in Appendix B and Tables 1 and 2. We find that the peak and integrated intensities are constant within uncertainties when the semi-minor axis of the beam is 120 mas, indicating that the point source is spatially unresolved at those scales. For larger beam sizes, the peak and integrated intensities increase since the point source is no longer spatially separated from the AS 209 disk.

Null Detection in C 18 O and Continuum
In Figure 4, we present C 18 O J = 2 − 1 emission at 4.8 km s −1 and the continuum image. No C 18 O emission at a > 3σ significance is detected at the location of the 13 CO point source. Similarly, no emission coincident with the 13 CO point source is detected at a > 3σ significance in the continuum image. We further inspected the continuum data by modeling the continuum emission in the visibility plane following Andrews et al. (2021), but no point source emission coincident with the 13 CO point source (and elsewhere) is found. In Section 4.2, we place upper limits on the CPD gas and dust masses based on the null detection of C 18 O and continuum emission. We find that the 13 CO point source lies at the center of the 12 CO gap and that it is colocated with the 12 CO velocity kink ( Figure 5). The coincidence suggests that the most likely explanation is that we are witnessing a planet and its CPD embedded in the AS 209 disk. In fact, the compact nature of the 13 CO emission separated from the circumstellar disk bears a striking resemblance to what is predicted in numerical simulations of a CPD embedded in a circumstellar disk presented in Perez et al. (2015). There is no obvious counterpart of the 13 CO point source in 12 CO, but this is likely because the annular gap is optically thick in 12 CO. Similarly, we do not find a counterpart of the 12 CO velocity kink in 13 CO or C 18 O due to their low optical depth within and beyond the circumstellar disk gap (see e.g., Figure 5d).
Recent numerical simulations have shown that circumplanetary materials form a rotationally-supported disk when the cooling is efficient, but when cooling is inefficient, a pressure-supported envelope is instead formed (Szulágyi et al. 2016;Fung et al. 2019). The point source is spectrally unresolved in the existing dataset, and whether the emission is associated with a disk or an envelope has to be confirmed in the future using high velocity resolution molecular line observations that can probe the kinematics of the point source. For the rest of this paper, we opt not to distinguish between circumplanetary disk and envelope, but simply use the term CPD to refer to the two collectively.
What is the mass of the planet candidate? We can estimate the planet mass using the empirical relation between the gap width and planet mass derived from numerical simulations. The width of the gap, determined by the distance between the adjacent peaks in the 12 CO radial intensity profile, is 78 au (Law et al. 2021a). The scaling relation between the gap width normalized by the planet's orbital semi-major axis ∆ gap /R p , disk aspect ratio (h/r) p , disk viscosity α, and planet mass (Kanagawa et al. 2016, see also similar relations in e.g., Zhang et al. 2018;Yun et al. 2019). Adopting the disk aspect ratio at the midplane (h/r) p = 0.118 from the circumstellar disk temperature constrained using CO isotopologues (Law et al. 2021b), ∆ gap /R p = 0.38 converts to a planet mass of M p = 1.3 M Jup · (α/10 −3 ) 1/2 . The inferred mass is comparable to those suggested to ex-  (c), showing the velocity kink in 12 CO (background color map) within the 12 CO gap (black dashed curves). The blue contours show 13 CO emission at 10 and 15σ levels. Note that the CPD candidate locates at the center of the 12 CO gap and coincides with the velocity kink in 12 CO, highly suggestive of the presence of an embedded planet. The synthesized beams for 12 CO and 13 CO data are shown in the lower-left and lower-right corner of the panel, respectively.

CPD size and temperature
As shown in Appendix B, the CPD is spatially unresolved. When the CPD is optically thick and the disk gas is in local thermodynamic equilibrium (LTE), the actual CPD gas temperature, T CPD (assumed to be uniform across the CPD for simplicity), and the observed brightness temperature of the CPD, T obs , are related as T obs = T CPD where is the beam dilution factor 3 . Adopting a thin disk geometry, the beam dilution factor is given by where f is a dimensionless parameter introduced to account for the fraction of the emitting area in a given velocity channel compared to the full geometric area of the CPD (see below), D CPD is the diameter of the CPD, i is the inclination of the CPD, θ maj and θ min are the semi-major and semi-minor axis of the synthesized beam, respectively. We can then relate the temperature and diameter of the CPD as Without the knowledge of the emitting area we opt to adopt f = 1, although it is reasonable to expect f < 1 because the channel spacing of the data (200 m s −1 ) is smaller than the FWHM of the line emission arising from the CPD (300 − 400 m s −1 ; see Tables 1 and 2). With f = 1, our analysis thus offers a lower limit to the CPD temperature. We use the brightness temperature in the 4.8 km s −1 channel ( 19 K) since it is likely that the channel contains the largest emitting area. We adopt i = 35 • (Öberg et al. 2021) assuming that the CPD and AS 209's circumstellar disk are co-planar. Figure 6a shows the CPD temperature required to explain the observed 13 CO emission as a function of the CPD diameter, calculated with Equation (2). We note again that the CPD temperature in Figure 6a presents a lower limit because a higher CPD temperature is needed to explain the observed 13 CO emission (1) when attenuation by the gas in the circumstellar disk gap is not negligible, (2) when the emitting area in the central channel is smaller than the full geometric area of the CPD (f < 1), and/or (3) when the CPD is optically thin. We find that the CPD temperature is 35 K, higher than the circumstellar disk gas temperature at the radial loca-tion of the CPD, 22 K (Law et al. 2021b). This suggests that additional heating sources localized to the CPD are required to explain the observed 13 CO emission, such as planet's thermal/accretion heating, the CPD's internal viscous/turbulent heating, or shock/compressional heating by infalling circumstellar disk material (Szulágyi et al. 2016;Szulágyi & Mordasini 2017).
In the discussion above (and also below in Section 4.2.2), we assume that the CPD and AS 209's circumstellar disk are co-planar. This is a reasonable assumption to start, but numerical simulations show that it is possible that CPDs and their parent circumstellar disk can be misaligned if the planet is formed in a turbulent environment (Jennings & Chiang 2021). In fact, based on the observationally constrained large obliquity of a planetary mass (M p = 12 − 27 M Jup ) companion 2M0122b, 48 +28 −21 degrees, Bryan et al. (2020) proposed that the CPD of 2M0122b would have to be tilted out of the circumstellar disk plane (see also Bryan et al. 2021). For our analysis, the CPD temperature would be minimized when the CPD is face-on (i = 0 • ; see Equation 2). However, even when the CPD is face-on we infer the CPD temperature is 29 K, higher than the circumstellar disk temperature, suggesting that additional heating sources localized to the CPD are still required. Future high spatial/velocity resolution observations would constrain the CPD geometry from its rotation, allowing us to place more accurate constraints on the CPD temperature.

CPD mass
For a given CPD size and the corresponding lower limit to the CPD temperature (black curve in Figure  6a), we can place a lower limit to the 13 CO mass. This is possible because the upper state energy of the 13 CO J = 2 − 1 transition (15.8 K) is below the lower limit of the CPD temperature; the population in the J = 2 rotational level would always drop if the CPD temperature is higher than the lower limit, requiring a larger CPD mass than for cooler temperatures.
To obtain the total 13 CO number density in the CPD, we first calculate the number density of 13 CO molecules in the J = 2 level as where h is the Planck constant, A 21 = 6.04 × 10 −7 s −1 is the spontaneous emission coefficient from molecular data from the LAMDA database (Schöier et al. 2005), Idv = 2.50 mJy beam −1 km s −1 is the integrated line intensity, d is the distance to AS 209, 121 pc (Gaia Collaboration et al. 2020), and Ω is the solid angle of the beam. We then calculate the total 13 CO number density under the assumption that all energy levels are populated under LTE: where Z = ∞ J=0 (2J + 1) exp[−hB e J(J + 1)/kT ] and B e = 55.10101 GHz is the rotational constant (Müller et al. 2005). We convert the 13 CO number density to the total gas mass enclosed in the CPD adopting interstellar CO abundance [ 12 CO]/[H 2 ] = 10 −4 and 13 CO-to-12 CO abundance ratio [ 13 CO]/[ 12 CO] = 1/60 (Wilson & Rood 1994). The resulting CPD mass is presented in Figure  6b. The CPD gas mass is 0.095 M Jup ( 30 M ⊕ ) when D CPD = 14 au, and increases with decreasing CPD size because the population in the J = 2 level decreases with higher CPD temperature. The CPD-to-planet mass ratio is M CPD /M p 0.02.
In the discussion above, we assumed a CO-to-H 2 ratio inferred from the local interstellar medium (ISM). However, near the radial location of the CPD candidate, CO in the AS 209's circumstellar disk is inferred to be depleted in the gas phase by about an order of magnitude compared with the local ISM ). If the CO in the CPD is depleted at a comparable level the CPD-to-planet mass ratio would be M CPD /M p 0.2, a level at which the CPD may be gravitationally unstable.
We can place upper limits to the CPD gas mass from the null detection of C 18 O emission (3σ 1.011 mJy beam −1 ) following the same approach to 13 CO. To do so, we assume that 13 CO and C 18 O share the same line width ( 400 m s −1 ) and temperature (presented in Figure 6a). We adopt A 21 = 6.01 × 10 −7 s −1 (Schöier et al. 2005), B e = 54.89142 GHz (Müller et al. 2005), and [C 18 O]/[ 12 CO] = 1/560 (Wilson & Rood 1994). The resulting upper limit to the CPD gas mass is presented in Figure 6b. Interestingly, the null detection of C 18 O places a tight upper limit that is only about 50% larger than the lower limit obtained with 13 CO. When the CPD temperature is larger than the lower limit presented in Figure 6a, both 13 CO lower limit and C 18 O upper limit would increase. However, because the rotational constant of the two lines are comparable, the ratio between the upper limit inferred by C 18 O and the lower limit inferred by 13 CO would remain nearly constant (see Equation 4) and the null detection of C 18 O would still provide tight constraints.
We can also place constraints on the CPD dust mass from the non-detection of continuum emission (3σ 26.4 µJy beam −1 ). Adopting the DSHARP dust opacity ) with a maximum grain size of 1 mm (κ ν = 2.0 cm 2 g −1 at 240 GHz) and the CPD temperature constrained as in Figure 6a . The gray shaded region shows the planet mass constrained by width of the 12 CO gap, 0.42 − 4.2 MJup, adopting a range of disk viscosity α = 10 −4 − 10 −2 (see Section 4.1). (b) A lower limit to the CPD gas mass inferred from the 13 CO emission (blue) and an upper limit to the CPD gas mass inferred from the null detection of C 18 O (red). The dashed curves show MCPD = 1.0, 0.3, 0.1, and 0.03 Mp (from left to right). (c) The CPD dust mass inferred from the non-detection of the continuum emission. The solid curve assumes a maximum grain size of 1 mm, while the dashed curve assumes a maximum grain size of 1 µm. (d) The CPD dust-to-gas mass ratio obtained from the inferred CPD gas and dust masses. In this figure, the black solid/dashed curves show lower (temperature in panel a) or upper (dust mass in panel c) limits, while the gray hatched regions present the allowed region in the parameter space bounded by limits from the 13 CO (blue) and C 18 O (red) data. mass in the optically thin regime is 0.006 M ⊕ 0.47 lunar masses. If instead the CPD has only small, µm-sized grains, κ ν = 0.42 cm 2 g −1 ) and the CPD dust mass is 0.027 M ⊕ 2.2 lunar masses. These results are presented in Figure  6c.
Using the constraints on the CPD gas and dust mass, we infer the dust-to-gas mass ratio to be 2 × 10 −4 when the maximum grain size in the CPD is 1 mm, and 9 × 10 −4 when the maximum grain size in the CPD is 1 µm (Figure 6d). This suggests that the CPD is likely lacking dust at a level below that in the typical ISM environment, presumably due to a limited dust supply (recall that AS 209's continuum disk is confined within the inner 140 au) and/or rapid radial drift of dust within the CPD.

How Did the Planet Form?
Based on the width of the 12 CO gap, we infer that the planet mass is around a Jupiter-mass. How did a giant planet form at an orbital radius of 200 au?
One possibility is that the AS 209 disk was gravitationally unstable in the past and the planet formed via gravitational instability (GI; Boss 1997). Limitations of the GI scenario often mentioned in the literature are that the masses of GI-induced fragments are generally large, often in the regime of brown dwarfs, and that GIinduced fragments suffer from tidal disruption and/or rapid radial migration (Baruteau et al. 2011;Zhu et al. 2012). However, recent magnetohydrodynamic simulations including the disk's self-gravity have shown that magnetic fields can limit the mass of the GI-induced fragments to a fraction to a few Jupiter masses, and also can prevent fragments from being tidally disrupted (Deng et al. 2021). It is also shown that orbital migration can stall when the planet actively accretes and carves a deep gap (Fletcher et al. 2019).
Despite recent theoretical developments, one challenge to the GI scenario is that the AS 209 disk has a very small mass of 0.003−0.0045 M 3.1−4.7 M Jup (Favre et al. 2019;Zhang et al. 2021). According to the inferred gas surface density profile from Zhang et al. (2021) and the midplane temperature constrained by CO isotopologues from Law et al. (2021b), the Toomre Q parameter (Toomre & Toomre 1972) is > 100 at all radii in the AS 209 disk with Q 300 at 200 au, indicating that the disk must be gravitationally stable presently. While it is not impossible that the disk was sufficiently massive in the past, the small present-day disk mass implies that the disk should have lost its mass very efficiently since then.
An alternative to the GI scenario is pebble accretion (Johansen & Lacerda 2010;Ormel & Klahr 2010). In order for the pebble accretion scenario to work, a few things have to be reconciled. First, the continuum emission of the AS 209 disk is confined within the inner 140 au presently. How did the core grow and how did the planet not trap millimeter grains beyond its orbit? We propose that in this scenario, the core of the planet never reached the pebble isolation mass (Morbidelli & Nesvorny 2012) and pebbles could freely drift inward crossing the planet's orbit. Based on scaling relations from hydrodynamic simulations (e.g., Bitsch et al. 2018), the pebble isolation mass at 200 au in the AS 209 disk (h/r = 0.118; Law et al. 2021b) is expected to be 100 M ⊕ , much more massive than the critical core mass of M crit = 10 − 20 M ⊕ at which point runaway gas accretion can start (Mizuno 1980). We can thus envision a scenario in which the pebbles that existed beyond the orbit of the core migrated inward, leaving the core behind.
Second, we need to check if there existed sufficient pebbles to form a core of the giant planet. To do so, we first estimate the total mass of pebbles necessary to drift toward the core to grow to 10 M ⊕ following . We obtain M peb ≈ 130 M ⊕ · (M c /20M ⊕ ) 1/3 (r/5 au) 1/2 (St/0.05) 1/3 ≈ 650 M ⊕ assuming that the dominant pebble size has a Stokes number of 0.05. The total dust mass in the AS 209 disk is currently about 300 M ⊕ (Sierra et al. 2021), but given that it is very unlikely that all the millimeter grains currently inward of the planet's orbit were once beyond 200 au, more efficient pebble accretion than the standard pebble accretion model of  is likely required. A few possibilities include the presence of pressure bumps (Pinilla et al. 2012) or a vortex that might have formed early on from the infalling flows from the protostellar envelope (Bae et al. 2015), and changes of sticking properties of grains that lead to the traffic jam effect around snow lines (Drkazkowska & Alibert 2017), which can slow down or even halt the radial drift of grains. In fact, it is interesting to note that the midplane gas temperature at the current radial location of the planet (T 22 K; Law et al. 2021b) is close to the expected CO and N 2 freezeout temperature (T frz,CO = 19 − 24 K, T frz,N2 = 17 − 21 K; Huang et al. 2018).
Third, we need to make sure that there had been sufficient time for a core to form via pebble accretion. The timescale for pebble accretion to form a core is given by t PA ≈ 4 × 10 4 years ·(M crit /10M ⊕ ) 1/3 (r/5 au) (Lambrechts & Johansen 2012). Adopting a critical core mass of 10 M ⊕ , at 200 au the pebble accretion time scale is t PA ≈ 1.6 Myr although, as mentioned earlier, slower radial drift or particle trapping can enhance the pebble accretion efficiency and shorten the required time to form the core. Given that the estimated age of AS 209 is 1 − 2 Myr (Andrews et al. 2009;Öberg et al. 2021), the long core accretion timescale might indicate that the planet could have entered the runaway accretion phase in the last 1 Myr. If the planet has been accreting 1 M Jup over the last million year or so, the average accretion rate would be 10 −6 M Jup yr −1 , orders of magnitude larger than the accretion rate of PDS 70b and c measured from Hα line emission (10 −8±1 M Jup yr −1 ; Wagner et al. 2018;Haffert et al. 2019). Observations of Hα line can confirm if the planet is indeed undergoing a runaway accretion, although long-term monitoring observations are likely required to distinguish steadily high accretion from episodic burst-type accretion (Lubow & Martin 2012).

Future Observational Direction
We conclude the discussion by listing a few future directions from the observational perspective. While we could only place upper/lower limits from the existing data, future higher-level transition observations of 13 CO (together with the existing 13 CO J = 2 − 1 data) will allow us to determine whether 13 CO is optically thick or thin, and thus to place stronger constraints on the CPD temperature. As shown in Figure  6b, the null detection of C 18 O already placed a tight upper limit to the CPD gas mass that is only about 50 % larger than the lower limit inferred by 13 CO. Deeper C 18 O observations (by a factor of 2 in sensitivity assuming that isotope selective photodissociation is not important) should thus detect the CPD and place stringent constraints on the CPD gas mass. The existing data is limited by a 200 m s −1 velocity resolution and we presently cannot conclude whether the point source emission is consistent with a rotationallysupported disk or a pressure-supported envelope that is potentially heated by an embedded planet (e.g., Alves et al. 2020). Future observations with higher velocity resolution will enable us to distinguish the two scenarios and to dynamically constrain the planet mass from the rotation of the CPD. High spatial/velocity resolution observations will also enable us to constrain the geometry of the CPD (e.g., position angle, inclination) from the disk rotation. Knowing the CPD geometry will allow us to place more accurate constraints on the CPD temperature and mass. In addition, estimating the obliquity of the CPD can help us to better understand the formation mechanism and environment of the CPD-hosting planet, in particular whether the planet had formed in a turbulent environment that would induce large obliquities (see e.g., Bryan et al. 2020;Jennings & Chiang 2021). The inferred CPD temperature suggests that heating sources localized to the CPD must be present. Such planetary/CPD heating can produce chemical asymmetries in the circumstellar disk, which are shown to be detectable with high-sensitivity molecular line observations using ALMA (Cleeves et al. 2015). Besides additional ALMA observations, future observations in the infrared wavelengths using the James Webb Space Telescope and ground-based telescopes would independently confirm the planet/CPD and allow us to probe the thermal emission from the planet which can help to constrain the mass of the planet. In addition, searching for accretion signatures (e.g., Hα line emission) would yield invaluable constraints on the current position of the planet in its evolutionary stages.

CONCLUSION
We report the discovery of a CPD candidate in the 13 CO J = 2 − 1 emission, embedded in the AS 209 disk at a radial distance of about 200 au. This is the first instance of CPD detection via gaseous emission, allowing us to probe the overall CPD mass. The CPD candidate is located in the middle of an annular gap identified in 12 CO and near-infrared scattered light observations, and is associated with localized velocity perturbations in 12 CO (see Figure 5). The coincidence of these features strongly suggest that we are witnessing the signature of a giant planet and its CPD embedded in the AS 209 disk.
The CPD is spatially unresolved with a 117 × 82 mas beam, indicating that its diameter is 14 au. Based on the 13 CO intensity, we were able to constrain the CPD temperature and gas mass. The CPD temperature is 35 K, greater than the circumstellar disk temperature at the midplane of the radial location of the CPD, 22 K (Law et al. 2021b). This suggests that there must be heating sources localized to the CPD in order to maintain the high temperature. Potential sources include planet's thermal/accretion heating, CPD's internal viscous/turbulent heating, and shock/compressional heating by infalling circumstellar disk material. The CPD gas mass is 0.095 M Jup 30 M ⊕ , adopting a standard 13 CO abundance. Based on the non-detection of continuum emission toward the CPD location, we found that the CPD has 0.027 M ⊕ 2.2 lunar masses of dust. Together with the inferred gas mass, this indicates a low dust-to-gas mass ratio of 9 × 10 −4 within the CPD, presumably due to a limited dust supply to the CPD and/or rapid radial drift of dust within the CPD.
The estimated age of the system is only 1 − 2 Myr (Andrews et al. 2009;Öberg et al. 2021). If confirmed, this CPD-hosting planet would be one of the youngest exoplanets detected to date. Observing planets at this young age allows us to place strong constraints on the mechanism and timescale of planet formation, crucial to gaining new insights into the formation and evolution of giant planets.

ACKNOWLEDGMENTS
We thank the anonymous referee for providing us with a helpful report that improved the initial manuscript.    Placing the cursor in the image will display positional offset and intensity.

B. 13 CO EMISSION WITH VARIOUS IMAGING PARAMETERS
Tables 1 and 2 present a summary of the 13 CO point source in the images adopting various robust parameters from −1 to 2. The rms noise is measured within the synthesized beam around the CPD over the first 20 line-free channels of the cube (v LSR = 14.7 − 10.9 km s −1 ). The central velocity, FWHM, peak intensity, and integrated intensity are obtained from Gaussian fitting of the emission within the synthesized beam around the CPD, using the CASA task specfit.
In Figure 10, we present the peak and integrated intensities as a function of the beam minor axis. The peak and integrated intensities remain constant within uncertainties when the beam minor axis is 120 mas, indicating that the point source is spatially unresolved at those scales. The peak and integrated intensities increase for larger beam sizes because the beam contains emission from the AS 209 disk.  Figure 11 presents 13 CO channel maps without JvM correction. Note that the CPD is clearly seen in all three channels even before the JvM correction is applied, with the peak intensity of 8.80, 13.5, and 8.47 mJy beam −1 , respectively. The peak intensity-to-noise ratio is about 6, 9, and 6 in the three channels, indicating that the detection is statistically significant in the JvM-uncorrected image.
In Figure 12, we present radial integrated intensity profiles.
At 1. 7, the integrated intensity of the point source is 4.37 mJy beam −1 km s −1 , while the integrated intensity excluding the point source is 2.02 mJy beam −1 km s −1 with a standard deviation of σ = 0.77 mJy beam −1 km s −1 . After subtracting the contribution from the gap, the point source is detected at a 4.4σ significance. fiducial channel with half-channel shift Figure 10. (Left) Peak intensity as a function of the beam minor axis. (Right) Integrated intensity as a function of the beam minor axis. Note that both peak intensity and integrated intensity stay constant when the beam is 120 mas, suggesting that the CPD is spatially unresolved. When the beam is > 120 mas the peak intensity and integrated intensity increase due to the contamination from the circumstellar disk. In both panels, blue points present intensities in the fiducial image cube whereas red points present intensities in the image cube where half-channel shift is applied.