Molecules with ALMA at Planet-forming Scales (MAPS). VII. Substellar O/H and C/H and Superstellar C/O in Planet-feeding Gas

We will focus on three of the five disks targeted by MAPS: the disks around AS 209, HD 163296, and MWC 480. All three of these disks have deep gaps in the millimeter emission indicative of ongoing planet formation (see Figure 1; Andrews et al. 2018; Guzmán et al. 2018; Huang et al. 2018; Long et al. 2018; Zhang et al. 2018; Liu et al. 2019), as well as gas velocity structures thought to be perturbations induced by unseen planets in these gaps (Pinte et al. 2018; Teague et al. 2018, 2019; Teague et al. 2021). We leave out the disks around IM Lup and GM Aur. The composition of the IM Lup disk has already been discussed in detail in Cleeves et al. (2018). The GM Aur disk is accreting new material, influencing the chemical composition of the disk (Huang et al. 2021; Schwarz et al. 2021), complicating modeling of the disk.

Zoom In Zoom Out Reset image size

Furthermore, in these disks, we will focus on the gas on the scale of the millimeter disk, that is, the size of the disk in the millimeter continuum, using these large grains as a signpost of active planet formation. Figure 1 shows millimeter continuum radial profiles, rich in structure, for the three disks that are studied together with the planet masses that are inferred to cause these structures (Zhang et al. 2018; Liu et al. 2019; Sierra et al. 2021). ²⁶ All three disks show potentially planet-induced dust structure that is clearly seen at the resolution of the MAPS continuum data (∼01 × 01; Sierra et al. 2021). The MAPS data are thus able to distinguish the gap region from the surrounding disk material.

Our focus is on understanding the chemical composition of localized regions associated with incipient planet formation. In the AS 209 disk, the structures of interest are the gaps at 61 and 105 au, straddling a bright ring at 74 au (Guzmán et al. 2018). These are all well outside of the CO ice line, which is located around 15 au (Alarcón et al. 2021; Zhang et al. 2021). In the HD 163296 disk, we will focus on the dust gaps at 48 and 86 au, with one gap inside and one gap outside the ice line at 70 au (Isella et al. 2018; Zhang et al. 2020a; Calahan et al. 2021; Zhang et al. 2021). Finally, for the MWC 480 disk, we focus on the 74 au gap, which is inside the CO ice line at 100 au (Liu et al. 2019; Zhang et al. 2021). We disregard the inner ∼10 au, as an accurate CO column density has not been derived for this region (Zhang et al. 2021).

To constrain the oxygen and carbon elemental abundances in the three selected disks, we base ourselves on results and models developed within the MAPS ALMA large program (Öberg et al. 2021). For C₂H, the column density derived in Guzmán et al. (2021) is used. These column densities are derived from the image cube by fitting the different C₂H hyperfine components. The C₂H column density derived from the images is uncertain within 30 au. As such, we derive a separate constraint on the C₂H column within 30 au using the kinematic information in the line profile. This is discussed in Appendix A. The shown upper limits in Figure 1 assume a column of C₂H that is constant in the inner 20 (AS 209) or 30 (HD 163296 and MWC 480) au.

We adopt the CO-depletion factor Zhang et al. (2021) derived from the observed CO isotopologue radial profiles (Law et al. 2021a). Zhang et al. (2021) built disk-specific models based on high-resolution continuum and CO isotopologue (¹³CO and C¹⁸O) line data. A total gas-to-dust mass ratio of 100 is assumed, with the gas disk size set by the extent of CO isotopologue emission. Gas and small dust (0.005–1 μm) are well coupled and have the standard tapered power-law surface density distribution and a vertical Gaussian density distribution (see Zhang et al. 2021, for the disk parameters). The large dust is settled at 0.2 times the gas scale height and structured to match the observed millimeter structure (see Tables 1 and 2 and Figure 5 in Zhang et al. 2021). The CO abundance in the disk is then varied until the model CO isotopologue line emission matches the observed line emission for ¹³CO and C¹⁸O for both the J = 2–1 and J = 1–0 transitions. The ratio between the abundance necessary to reproduce that data and the ISM CO abundance is called the CO-depletion factor. The model CO columns are consistent with the CO column derived from the hyperfine lines of C¹⁷O in the regions where the latter is detected. Both the C₂H column and the CO-depletion factor are shown in Figure 1.

The chemistry of C₂H and its use as a tracer of the elemental C/O ratio in protoplanetary disks has been discussed for a variety of sources and conditions (e.g., Jansen et al. 1995; Nagy et al. 2015; Bergin et al. 2016; Cleeves et al. 2018; Bosman et al. 2021). As a radical, it is most abundant in regions with an active photochemistry, where it is continuously produced and destroyed. The chemistry of C₂H is tightly linked to that of CH₄ and C₂H₂, which is the main reservoir of carbon that is not locked in CO. The C₂H can be directly formed by photodissociation of C₂H₂, or it is formed by dissociative recombination of the ${{\rm{C}}}_{2}{{\rm{H}}}_{2}^{+}$ and ${{\rm{C}}}_{2}{{\rm{H}}}_{3}^{+}$ ions, which are part of a gas-phase hydrocarbon radical and ion cycle (see, for example, Figure 1 of Bosman et al. 2021).

The C₂H abundance is very dependent on the C/O ratio in the gas, with reactions such as C₂H + O → CO + CH and CH + O → CO + H directly driving down the C₂H abundance, as well as removing carbon from the hydrocarbon reservoir. Reactions generally result in the formation of CO, which is stable against photodissociation (van Dishoeck & Black 1988). As such, this carbon is no longer available for hydrocarbon chemistry. Depending on the physical structure of the disk and the stellar radiation field, the C₂H column density in a disk model can vary by 4 orders of magnitude just by variations of the C/O ratio between ∼0.5 and 2.0 (Bergin et al. 2016; Cleeves et al. 2018). This makes C₂H a very good tracer of the C/O ratio.

As a radical, C₂H traces the photodissociation region (PDR) layers of the disk. At low C/O ratios, all carbon for C₂H comes from the dissociation of CO, which happens only in the top layer of the disk PDR, where CO self-shielding is not yet active. The mass in this thin layer is relatively constant in a disk, as well as disk to disk, as it depends on a fundamental property of the CO molecule, which starts to self-shield at a column of 10¹⁵ cm⁻² (van Dishoeck & Black 1988). At C/O ratios above 1, the carbon for the formation of C₂H can be extracted from hydrocarbons such as CH₄ and C₂H₂. The thickness of the layer in which these hydrocarbons can be dissociated depends on UV attenuation by dust. The small-dust content of the disk atmosphere thus has an effect on the measured C₂H column. Initial calculations in Bosman et al. (2021) predict that the small-dust content will impact the C₂H abundance but is relatively weak. An order-of-magnitude drop in the small-dust abundance should only lead to an increase in the C₂H abundance by a factor of 2.

In our modeling, we start with the model structure and CO-depletion factor as derived by Zhang et al. (2021). We vary the C/O ratio and the amount of small dust in the surface layers, as these parameters have the most impact of the C₂H abundance. The small-dust content of the disks is varied by multiplying the small-dust abundances by a constant factor of 0.001, 0.01, 0.1, or 1 from the base models of Zhang et al. (2021). The last of these are the fiducial dust structures and are identical to those used in Zhang et al. (2021). ²⁷ Each model has a constant C/O ratio of either 0.47, 1.0, or 2.0. These correspond to the volatile ISM level (0.47), CO-dominated gas (1.0), and a high C/O value (2.0). Increasing the C/O ratio above 2.0 no longer has an impact on the C₂H abundance (e.g., Cleeves et al. 2018). Combining the four different small-dust abundances and three different C/O ratios, this results in 12 models per disk studied.

The model calculation is done in two steps. In the first step, the dust and gas temperatures are calculated. We start from the best-fit structure and CO-depletion factors from Zhang et al. (2021). The structures are here modified to vary the small-dust content. The abundances of H₂O and CH₄ in the initial conditions of the chemical network are also modified at this point, so the gas temperature calculation includes the changes in composition driven by the different C/O ratios. The dust and gas temperatures are then calculated using DALI with the standard reduced chemical network, which correctly calculates the abundances of the major gas coolants (Bruderer et al. 2012; Bruderer 2013). In the second step, a full gas-grain network was used to calculate the 2D chemical composition based on the physical conditions determined in the first step.

This two-step approach was chosen to balance consistency with feasibility. To calculate the gas temperature, the chemical solver has to run multiple times per cell, which is prohibitively expensive to do with the gas-grain chemical network. As mentioned, this approach does capture the effects of the C/O ratio on the gas temperature, and the effect of these changes on the resulting chemistry, while being computationally tractable.

The gas-grain chemical network used for the second step of our calculations is based on the network of Walsh et al. (2015), which is a combination of the UMIST12 ²⁸ (McElroy et al. 2013) for the gas-phase reactions and the Ohio State University network (Garrod et al. 2008) for the grain surface reactions. Molecular binding energies are from McElroy et al. (2013) and Penteado et al. (2017), with small adjustments for NH_x and CH_x radicals (Bosman et al. 2018). These binding energies are tabulated in Bosman et al. (2018, Table B.2). The freeze-out of H₂ is also treated as in Bosman et al. (2018), using the higher (430 K) binding energy for H₂ when there are less than two layers of H₂ ice but lowering the binding energy to 100 K when H₂ covers the ice completely.

Photodesorption yields are taken from experiments where available (Öberg et al. 2009a, 2009b, 2009c) and assumed to be 10⁻³ otherwise. All photodesorption is assumed to yield intact molecules in the gas phase. ²⁹ The grain surface chemistry assumes a diffusion-to-binding energy ratio of 0.3 and a tunneling barrier of 1 Å (as in Bosman et al. 2018). For photodissociation, the wavelength-dependent cross sections of Heays et al. (2017) are used in combination with the wavelength-dependent UV field that DALI calculates. For the few cases in which Heays et al. (2017) did not have a compilation of the wavelength-dependent cross sections, the rate parameterization of McElroy et al. (2013) is used. Self-shielding and mutual shielding is included for C, CO, N₂, and H₂ (Draine & Bertoldi 1996; Kamp & Bertoldi 2000; Visser et al. 2009, 2018).

The initial abundances used in the models with different C/O ratios are tabulated in Table 1. Any molecule, atom, or ion not listed in the table has zero abundance at the beginning of the chemical calculation. The abundances of H₂O, CO, and CH₄, that is, the initial carbon and oxygen carriers, are scaled with the CO-depletion factors derived in Zhang et al. (2021; see Figure 1). Further, the abundances have been chosen such that for the maximum amount of CO that can be formed, that is the lowest of the carbon and oxygen elemental abundances, is the same for all three of the C/O ratios. CH₄ was chosen as the carrier for the excess carbon (instead of elemental C) in the models. The exact initial abundances do not matter greatly for this calculation, as abundances in the C₂H layer will be set by photon-dominated steady-state chemistry (Bosman et al. 2021). The only exceptions to this rule are the large (more than seven carbon atoms) carbon chains, which act as refractories in the intermediate PDR layers. These large carbon chains are only formed in the models when a significant fraction (>10%) of the carbon is in elemental or ionized form. To prevent this, an elemental or ionized form should not be chosen for the initial carbon carrier. In other cases, the exact carrier of the excess carbon does not substantially alter the resulting C₂H abundance (Bosman et al. 2021).

A low cosmic-ray ionization rate of 10⁻¹⁸ s⁻¹ is used for the chemical models, similar to the models in Zhang et al. (2021; see, e.g., Cleeves et al. 2013, for a discussion of cosmic-ray propagation in disks.). A higher cosmic-ray ionization rate, coupled with the already low CO abundances currently assumed, would quickly convert the remaining CO into other species, like CH₄, CO₂, and CH₃OH (e.g., Furuya & Aikawa 2014; Schwarz et al. 2016; Bosman et al. 2018), in the 1 Myr chemical timescale that is assumed. This would create a disagreement between the CO column in the models and the intended, observationally motivated CO column density. This cosmic-ray ionization rate is thus not the rate necessary to lower the CO abundance from the ISM level to the observed levels, but it is the cosmic-ray ionization rate that has to be used to prevent the CO abundance from dropping further. We note that the cosmic-ray ionization rate used does not impact the C₂H abundance in chemical models (Cleeves et al. 2018) and that the cosmic-ray ionization rate used here is consistent with modeling of MAPS sources by Aikawa et al. (2021).

Appendix B shows the CO columns for difference C/O ratios for the three disks compared to the CO columns derived from the data. The difference C/O ratios do not impact the CO column in the model. Furthermore, the agreement between our model CO columns and the results from Zhang et al. (2021) indicate that the amount of volatiles in the surface layer is representative of the disks studied.

The inferred C/O ratios associated with C₂H emission locations in these disks are 2.0 or higher above most of the millimeter disk. This is higher than previously derived for other disks. Miotello et al. (2019) found that a global C/O ratio of 1.5 reproduced the flux for the sample of Lupus disks studied, significantly lower than the C/O ratios found for these disks. This discrepancy is caused by the fact that the MAPS data resolve the disk emission and reveal existing emission structure. In the gas with strong C₂H emission, a C/O ratio ≥2.0 is necessary. However, the C₂H column density is not elevated (>10¹⁴ cm⁻²) over the entire disk. Thus, in models using disk-integrated flux, a lower, global C/O ratio will be estimated. This will be especially prominent for the three disks studied here, as the C₂H column densities drop below 10¹² cm⁻² around 150 au (see Guzmán et al. 2021), which implies C/O ratios <1 for the majority of the disk area.

In the IM Lup disk, the C/O ratio was determined by Cleeves et al. (2018) to be around 0.8, even lower than the C/O ratios found for most of the disk in Miotello et al. (2019). When comparing the C₂H radial profiles in Law et al. (2021a) and C₂H column densities in Guzmán et al. (2021) between the IM Lup disk and the AS 209, MWC 480, and HD 163296 disks, it is clear that the IM Lup disk has less C₂H than the other disks by more than an order of magnitude. The observationally derived C₂H column density for the IM Lup disk is ∼10¹³ cm⁻² between 70 and 230 au, which is of the same order as the AS 209 disk model column densities for a C/O of 0.47 and 1.0. The elemental composition of the IM Lup disk is thus clearly different from the three disks studied here.

The difference between the composition of the IM Lup disk and the other disks studied here could be due to the young age of the IM Lup disk (Mawet et al. 2012). The process responsible for turning the volatile rich and oxygen-dominated ISM composition gas into the volatile poor and carbon-rich gas seen in many other disks could then still be underway (Bergner et al. 2020; Zhang et al. 2020b). This composition transition can be the result of either more efficient trapping of oxygen-carrying species, e.g., water, CO, and CO₂, as ices coating millimeter-sized dust particles that are confined to the midplane (Krijt et al. 2020) or the release of carbon from a refractory source in a uniformly depleted surface layer (Bosman et al. 2021). Alternatively, the IM Lup disk has less pronounced substructures in the millimeter disk. This could simply be a result of the young age, but it can also imply that these substructures are, at least partly, responsible for the high C/O ratios (Bergin et al. 2016; Bosman et al. 2021).

While it is clear that to match the high (>10¹³ cm⁻²) column densities of C₂H in the AS 209, MWC 480, and HD 163296 disks, C/O ≥ 2 is necessary, it is not obvious what causes the radial structures in the C₂H column density. These can be caused by either a variation of C/O ratio when the C/O ratio is below 2.0, a variation in small-dust content when the C/O ratio is larger than unity, or a combination of both. The effects on the C₂H column density due to changes in the C/O ratio are generally larger than those due to changes in the small-dust content (Figure 2).

The regions in which a depletion of small grains must be invoked to match the C₂H column density is wider than the millimeter dust gaps seen in these three disks. As the chemical timescales are short, <1000 yr (Bosman et al. 2021), diffusive spreading of C₂H from the millimeter gap, which happens on timescales >10⁴ yr, can be ruled out. This is thus suggestive of a wider small-grain gap, around the millimeter dust gap, possibly due to the millimeter dust rings acting as a small-grain sink in a manner similar to that predicted from the midplane (Krijt & Ciesla 2016).

In the HD 163296 disk, the C₂H column density outside 70 au is consistent with an elevated C/O ratio of 2.0 and factor of ∼10 of small-dust depletion. At 50 au, the radius of a prominent gap in the millimeter continuum, the C₂H column density is more comparable to the small-grain model with a factor 100 depletion. This indicates that the UV penetrates deeper into the disk around the millimeter dust gap. This is not seen in the second dust gap at 90 au, where there is a local minimum in the C₂H abundance. A possible explanation for the difference in C₂H column density can be found in the scattered-light image. There is a bright ring in scattered light associated with the 66 au submillimeter dust ring (e.g., Garufi et al. 2014; Monnier et al. 2017; Muro-Arena et al. 2018). This ring of abundant small dust could shadow the region of the second dust gap, lowering the available UV flux in the gap and thus lowering the C₂H column density.

The C₂H column density in the MWC 480 disk peaks around 70 au, the location of the millimeter continuum dust gap (Liu et al. 2019; Long et al. 2019; Law et al. 2021a). Similar to the gap at 50 au in the HD 163296 disk, high UV penetration at the gap location—again suggestive of small-grain depletion—is necessary.

The need for additional small dust depletion on top of the fiducial dust models that already have a degree of dust depletion in the surface layers implies gas-to-dust ratios in the surface layers of at least 10,000 and up to 10⁶ at the location of the C₂H emission peak in the AS 209 disk (Zhang et al. 2021). It is thus possible that other processes are increasing the C₂H abundance above the levels predicted by the chemical models. One intriguing possibility is that strong vertical or radial mixing is operating, which could cycle C₂H₂ and other larger carbon species from more shielded regions into the 25–75 au region. This would lead to an out-of-equilibrium chemistry with a higher C₂H abundance. This could be consistent with a large-scale meridional flow into the gaps and would be an interesting subject of further study.

It has been suggested that a low CO abundance and a high C/O ratio have a common dynamical or chemical origin, which should cause a correlation between CO abundance and C/O ratio (Bergin et al. 2016). However, a relation between C₂H flux and CO abundances, as traced by ¹³CO over millimeter dust flux, has not been found (Miotello et al. 2019). The relation has not been studied as a function of radius in a disk, where, for example, the release of CO at the CO ice line is expected to have a strong impact on the gas-phase C/O ratio (Öberg et al. 2011).

For the MWC 480 and HD 163296 disks, the regions within and outside of the CO ice line are resolved by the observations. The emission maps of C¹⁸O (and the derived CO column) do not exhibit a strong response at the CO ice line location predicted by the detailed disk physical models. Most directly, the CO abundance at the CO ice line does not return to the ISM level (e.g., 10⁻⁴ relative to H₂; Zhang et al. 2021, Figure 1). This is consistent with a lack of the expected C₂H column density drop at the CO ice line. Without the return of CO, the C/O ratio does not drop toward unity but stays high. Closer to the star, there is an increase in the CO abundance, reaching greater-than-ISM levels at 50 au in MWC 480 and 30 au in HD 163296.

The increase in CO abundance at these radii is accompanied by a drop in C₂H column density (see Figure 1) consistent with a drop in C/O ratio (Figure 2) due to the return of the oxygen carried by CO into the gas. As mentioned, the location of the CO abundance rises is not where the CO ice line is expected, based on the thermal structure in the models (Calahan et al. 2021; Zhang et al. 2021). Temperatures in the region where the CO abundance rises are between 25 and 30 K in the standard MAPS models (Zhang et al. 2021). These temperatures are in line with modeling by Calahan et al. (2021) for HD 163296 and empirical temperature measurements from the CO isotopologue emission (Law et al. 2021b). Furthermore the location where the C₂H column drops below the C/O = 2 models is at smaller radii than the rise in CO abundance. This implies that when the CO abundance rises to close to the ISM value, the C/O ratio drops.

If the rise of the CO abundance is due to CO coming off the grains, this would imply a higher binding energy than the 850 K (∼22 K ice line temperature; e.g., Harsono et al. 2015) assumed in the models, or could be due to a different physical process. One possibility is large, icy grains drifting across the CO ice line, but only depositing the CO into the gas at a radius well within the CO ice line location. The exact effects are going to be very disk-dependent. A higher CO desorption temperature was also inferred for CO in TW Hya (Zhang et al. 2017). This could be indicative for CO as desorbed on an H₂O or CO₂ mixture (Noble et al. 2012; He & Vidali 2014). In the case of HD 163296 and MWC 480, in which the CO content of the grains is expected to be high, this would indicate that the vast majority of the ∼100 monolayers of CO are all in high binding energy sites. Moving the ice line by ∼40 au requires fast radial drift. Only grains between 1 and 10 cm reach the right drift and desorption speeds (Piso et al. 2015). This is larger than the 3 mm grains that Sierra et al. (2021) needed at these radii to fit the multiband continuum data.

A similar relation between CO and C₂H abundance is seen in the outer disk of AS 209, where outside 100 au, the CO abundance increases and the C₂H column density drops, again implying a low C/O ratio when the CO abundance is high. The HD 163296 and MWC 480 disks also show a drop in C₂H column density and inferred C/O ratio outside ∼120 and ∼150 au, respectively. This is not paired with an increase in CO abundance, however. This implies that a high C/O ratio happens only in the regions with low CO abundance. However, something in addition to low CO abundance is needed to explain a high C/O ratio. This could be related to the presence of an additional carbon source, such as the photoablation of carbonaceous grains (Bosman et al. 2021).

To properly predict the composition of a planet forming at a certain location, it is necessary to know the elemental composition of the building materials. For any planet forming outside of 30 au in the AS 209, HD 163296, and MWC 480 disks, such as those inferred in the various gaps in these systems, it is now possible to make some inferences on the composition of the gas and solids that will go into these putative planets. For this discussion, we will assume that the bulk of the gas mass is accreted late in the disk lifetime, such that the composition of the disk gas inferred here is representative of the gas accreted onto the planet. If planets accrete their gas mass early in the disk lifetime, when accretion onto the disk still delivers significant pristine material, planetary composition will be very similar to that of the host star.

The C₂H emission in these three disks is strong, but an emission surface height cannot be derived unambiguously over the entire disk (Law 2021b). This implies either that the line is optically thin, and thus too weak to measure a height, or that it comes from below the ¹³CO surface, which results in a measured height that is consistent with the disk midplane. The hyperfine ratios, from which the C₂H column densities have been derived, show that the lines are optically thick. This means that the C₂H lines come from deeper than the ¹³CO emission. This is consistent with the disk regions where the C₂H height can be derived (Law 2021b) and in agreement with the models that have most of their column density between one and two scale heights (see Appendix C; the scale heights in these disks at 10–200 au vary between z/r of 0.06 and 0.1, depending on the disk; Zhang et al. 2021). At this height, mixing with the midplane is efficient and happens on 10 kyr timescales for turbulent viscosity coefficients of α ≈ 10⁻³ (Ciesla 2010). As such, it is unlikely that strong vertical elemental abundance gradients exist, so the high C/O abundances and low O/H abundances necessary to explain the C₂H emission likely hold for the disk midplane as well. Furthermore, if accretion onto a planet happens through meridional flows, the gas accretion onto the planet will actually sample the vertical height from where the C₂H emission originates (Tanigawa et al. 2012; Morbidelli et al. 2014); such flows have been observed in the higher layers of the disk with ¹²CO in the HD 163296 disk by Teague et al. (2019). As these are the flows that feed the planet, gas from the C₂H-emitting layer is expected to contribute significantly to the planetary accretion after gap opening (Morbidelli et al. 2014; Cridland et al. 2020).

The number of small grains in the disk atmosphere that can accrete with the gas onto the planet is strongly reduced from the ISM level of 1% of the total mass. This happens both globally, in the disk surface due to grain growth and settling, and locally, as the inferred C₂H column densities necessitate more UV penetration in certain disk locations. The combined low CO abundance and depleted small dust thus implies that most of the ISM carbon and oxygen, both volatile and refractory, is not present in gas accreting onto planets that have reached pebble isolation mass.

Gas giants that accrete most of their atmospheres after reaching the pebble isolation mass thus obtain envelopes, with substellar C/H and O/H and high (≫1.0) C/O ratio compositions, unless efficient release of volatiles from the core or late enrichment by planetesimals takes place. These potential planets are formed at large radii, where CO can be depleted efficiently (>50 au radii for HD 163296 and MWC 480). If these planets do not migrate from their formation location, they are prime targets for composition analysis by direct imaging with the James Webb Space Telescope, the most notable example being the planets in the HR 8799 system. Substellar carbon and oxygen abundances and a high C/O composition would distinguish planets formed through pebble or core accretion from planets formed as a result of a gravitational instability. The latter are expected to have stellar or superstellar C/H and O/H ratios with a stellar or substellar C/O ratio (e.g., Ilee et al. 2017).

Compositions of planets at large radii so far do not show the signature predicted from the disk observations. The planet β Pic b has a low derived C/O ratio, which is expected to be close to or below the stellar C/O ratio (Gravity Collaboration et al. 2020). This rules out a formation scenario in which one of the >40 au putative planets in the disks studied here migrates inward to 9 au (Lagrange et al. 2020), unless this planet accreted >40 M_⊕ of oxygen-rich ice or mixed a similar amount of ice from the core into the atmosphere (Gravity Collaboration et al. 2020). Three other large-radius (>50 au) planets have their water abundance measured (κ And b, HR 8799 b, and HR 8799 c; Todorov et al. 2016; Lavie et al. 2017; Madhusudhan 2019). The water abundances are higher than the disk atmosphere gas, ice, and refractory oxygen abundances combined by a factor of 2–100 for the three disks studied here. If these planets are the successors of the putative planets in these disks, a significant amount of disk water ice needs to make it into the atmosphere, at least 30 M_⊕ for κ And b and HR 8799 b and around 100 M_⊕ for HR 8799 c (Madhusudhan 2019). This would imply very massive cores (60–200 M_⊕) and very efficient core mixing, 6–20 M_⊕ Myr⁻¹ planetesimal accretion rates for a 10 Myr disk age, or a gravitational instability origin from a moderately (factor of 1–3) water ice–enhanced core. The low water abundance measured for many transiting exoplanets is comparable to the low total oxygen abundance in the gas at large (>50 au) radii in the AS 209, HD 163296, and MWC 480 disks (Madhusudhan 2019; Welbanks et al. 2019). This could point to an outer disk origin for these hot Jupiters; however, without constraints on the C/O ratio, this remains speculation. A high C/O ≳ 1.0 would strengthen the case for an outer disk origin for hot Jupiters.