The Astrophysical Journal, 926:157 (14pp), 2022 February 20 https://doi.org/10.

3847/1538-4357/ac3d32
© 2022. The Author(s). Published by the American Astronomical Society.

Spatially Resolved Modeling of Optical Albedos for a Sample of Six Hot Jupiters
Danica J. Adams1 , Tiffany Kataria2 , Natasha E. Batalha3 , Peter Gao4,5 , and Heather A. Knutson1
1
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA; [email protected]
2
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA
3
NASA Ames Research Center, MS 245-3, Moffett Field, CA 94035, USA
4
Department of Astronomy and Astrophysics, University of California, Santa Cruz, Santa Cruz, CA 95064, USA
5
Earth and Planets Laboratory, Carnegie Institution for Science, 5241 Broad Branch Road, NW, Washington, DC 20015, USA
Received 2021 May 12; revised 2021 November 12; accepted 2021 November 23; published 2022 February 21

Abstract
Optical secondary eclipse measurements made by Kepler reveal a diverse set of geometric albedos for hot Jupiters
with equilibrium temperatures between 1550 and 1700 K. The presence or absence of high-altitude condensates,
such as Mg2SiO4, Fe, Al2O3, and TiO2, can significantly alter optical albedos, but these clouds are expected to be
confined to localized regions in the atmospheres of these tidally locked planets. Here, we present 3D general
circulation models and corresponding cloud and albedo maps for six hot Jupiters with measured optical albedos in
this temperature range. We find that the observed optical albedos of K2-31b and K2-107b are best matched by
either cloud-free models or models with relatively compact cloud layers, while Kepler-8b’s and Kepler-17b’s
optical albedos can be matched by moderately extended ( fsed = 0.1) parametric cloud models. HATS-11b has a
high optical albedo, corresponding to models with bright Mg2SiO4 clouds extending to very low pressures
( fsed = 0.03). We are unable to reproduce Kepler-7b’s high albedo, as our models predict that the dayside will be
dominated by dark Al2O3 clouds at most longitudes. We compare our parametric cloud model with a microphysical
cloud model. We find that even after accounting for the 3D thermal structure, no single cloud model can explain
the full range of observed albedos within the sample. We conclude that a better knowledge of the vertical mixing
profiles, cloud radiative feedback, cloud condensate properties, and atmospheric metallicities is needed in order to
explain the unexpected diversity of albedos in this temperature range.
Unified Astronomy Thesaurus concepts: Hot Jupiters (753); Atmospheric clouds (2180); Radiative transfer (1335);
Exoplanet atmospheres (487); Extrasolar gaseous planets (2172)

1. Introduction atmosphere models for hot Jupiters generally predict dayside

albedos of less than 0.1 (Seager et al. 2000), but the observed
Transiting short-period gas giant planets, or “hot Jupiters”, are
optical geometric albedo measurements published to date span a
among the most favorable targets for atmospheric characterization
wide range of values, with the brightest planets exhibiting albedos
studies. By observing the decrease in optical flux when the planet
as high as 0.3 (e.g., Heng & Demory 2013; Niraula et al. 2018). In
passes behind the host star (secondary eclipse), we can measure the
the solar system, the presence of water clouds increases Earth’s
optical dayside albedos for these tidally locked planets. Most of
geometric albedo to approximately 0.37 (e.g., Goode et al. 2001),
these albedo measurements have come from space telescopes
while ammonia clouds in Jupiter’s atmosphere contribute to its
observing in broad optical bandpasses (e.g., Rowe et al. 2008;
geometric albedo of approximately 0.5 (e.g., Marley et al. 1999).
Demory et al. 2011; Kipping & Bakos 2011; Coughlin & Lopez-
For the same reason, models suggest that the presence of reflective
Morales 2012; Demory et al. 2013; Parviainen et al. 2013;
condensates such as Mg2SiO4 or MgSiO3 in hot Jupiter
Angerhausen et al. 2014; von Paris et al. 2016; Niraula et al. 2018).
atmospheres can increase their albedos to values as high as 0.5
The hot Jupiter HD 189733b is currently the only planet with a
(Parmentier et al. 2016, 2021; Roman et al. 2021). Unlike brown
spectroscopically resolved reflected-light measurement (Evans
dwarfs, whose cloud distributions and optical depths correlate
et al. 2013; geometric albedo of 0.40 ± 0.12 at 290–450 nm and
closely with their equilibrium temperatures (e.g., Kirkpatrick 2005;
<0.12 at 450–570 nm). WASP-12b and WASP-43b were also
Marley et al. 2010), hot Jupiter albedos can vary by as much as an
observed with optical spectrographs on the Hubble Space
order of magnitude within a relatively narrow range of equilibrium
Telescope (HST), but both observations resulted in nondetections
temperatures (Figure 1). This suggests that individual planets with
(Bell et al. 2017; Fraine et al. 2021; geometric albedos <0.06).
similar equilibrium temperatures may exhibit diverse cloud
These spectroscopic albedo measurements are invaluable for
properties.
constraining the nature of the scattering particles in the atmo-
Hot Jupiters are expected to be tidally locked as a result of
spheres of these planets (e.g., Barstow et al. 2014).
their short orbital periods, which can lead to significant day–
Theoretical models predict that variations in hot Jupiter optical
night temperature gradients (e.g., Guillot & Showman 2002).
albedos should primarily be driven by the presence or absence of
This inhomogeneous temperature structure affects the cloud
high-altitude aerosols, which are expected to scatter incident
distribution, and 3D atmospheric circulation models predict that
starlight (e.g., Seager et al. 2000; Burrows et al. 2008). Cloud-free
hot Jupiters should host spatially inhomogeneous clouds over a
wide range of equilibrium temperatures (e.g., Parmentier et al.
Original content from this work may be used under the terms
of the Creative Commons Attribution 4.0 licence. Any further 2016, 2021; Roman et al. 2021; also see Figure 1). Furthermore,
distribution of this work must maintain attribution to the author(s) and the title there is direct observational evidence for inhomogeneous cloud
of the work, journal citation and DOI. structures stemming from reflected-light phase curves (e.g.,

1
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

of published studies that use GCMs to predict hot Jupiter dayside

albedos, we find that all of these studies make simplifying
assumptions in order to conserve run time and numerical
complexity. Oreshenko et al. (2016), for example, determined
the locations of clouds by comparing the 3D thermal structure
from a GCM with relevant condensation curves, but assumed a
fixed particle size and local condensation only (i.e., they neglected
vertical mixing) for the cloud layers. In Parmentier et al. (2016),
cloud distributions were determined using the completed temp-
erature structure calculated by a nongray cloud-free GCM, with a
cloud-top minimum pressure fixed at 1 microbar. In Parmentier
et al. (2021), cloud layers were also calculated using the thermal
structure output from a GCM, but the prescribed vertical extent is
limited by temperatures within the extent of 200 mbar to 1 μbar.
Roman & Rauscher (2019) and Roman et al. (2021) differ
significantly from the Parmentier models. In the former, cloud
distributions were determined at each timestep in a double-gray
GCM which included radiative feedback. They included extended
Figure 1. Optical geometric albedo measurements for a sample of hot Jupiters
clouds, which form when the temperature-pressure profile permits
observed in the Kepler bandpass; at these temperatures thermal emission is but are forced to taper off at pressures between 0.3 and
negligible for planets with cloudy skies and the measured secondary eclipse 0.057 mbar. They also consider compact cases with varied optical
depth is dominated by reflected light. The equilibrium temperature is calculated thickness where the clouds are truncated and tapered off after
assuming a Bond albedo of zero and efficient day–night circulation. The planet
colors vary as a function of surface gravity. The dashed box indicates the approximately one scale height regardless of where their base
temperature range considered in this study. Measurements are drawn from forms. In Roman et al. (2021), they also consider a more extensive
Fortney et al. (2011), Desert et al. (2011), Barclay et al. (2012), Esteves et al. grid of planet models, while also varying cloud compositions,
(2013), Shporer & Hu (2015), Niraula et al. (2018), and Heng et al. (2021).
densities, and vertical extents. All of the studies listed above
assumed homogeneous condensation, allowing them to treat each
Demory et al. 2013). To date, three of the four planets with
observed reflected-light phase curves appear to have patchy individual cloud species separately. Although there are studies in
clouds (Desert et al. 2011; Demory et al. 2013; Angerhausen the literature that have combined GCMs with microphysical cloud
et al. 2014; Esteves et al. 2015; Shporer & Hu 2015; von Paris models (e.g., Lee et al. 2015; Lines et al. 2018; Helling 2019;
et al. 2016; Niraula et al. 2018). Kepler-7b, -12b, and -41b all Helling et al. 2019, 2020), these models focused on individual
appear to have patchy clouds that are preferentially concentrated planets and were limited in their albedo predictions in the Kepler
in the western (dawn) hemisphere, causing the peak of the phase bandpass. Although there are studies in the literature that have
curve to occur after the secondary eclipse (Shporer & Hu 2015). combined GCMs with microphysical cloud models (e.g., Lee
Observations of the fourth planet, TrES-2b, indicate that it is et al. 2015; Lines et al. 2018; Helling 2019; Helling et al.
uniformly dark (As < 0.03, which is equivalent to Ag < 0.02 for a 2019, 2020), these models focused on individual planets and were
Lambertian sphere, where As is the spherical albedo and Ag is the limited in their albedo predictions in the Kepler bandpass.
geometric albedo) at all phases (von Paris et al. 2016). These In this study, we utilize a suite of models to investigate the role
studies demonstrate that if we wish to explain the observed of patchy clouds over a sample of six individual hot Jupiters,
planet-to-planet variations in the measured dayside optical chosen due to their diverse observed albedos over a narrow range
albedos of hot Jupiters, we must utilize 3D models capable of of equilibrium temperatures. In Section 2, we describe our sample
capturing the spatially varying cloud structure. selection and summarize our modeling approach. We use 3D
While an increasing number of studies are using general GCMs to derive the thermal structures and eddy diffusion
circulation models (GCMs) to predict cloud patterns, these have coefficients (Kzz). We then use Virga, a phase equilibrium cloud
either focused on individual planets (e.g., Lee et al. 2015; Webber code, to make detailed maps of cloud structure over the dayside of
et al. 2015; Oreshenko et al. 2016; Lines et al. 2018; Helling 2019; each planet. We use Planetary Intensity Code for Atmospheric
Helling et al. 2019; Lines et al. 2019; Roman & Rauscher 2019; Scattering Observations (PICASO), a radiative transfer program, to
Helling et al. 2020) or generic grids of models (e.g., Parmentier calculate the corresponding geometric albedo maps and hemi-
et al. 2016, 2021; Roman et al. 2021). Of these listed studies, only
sphere-integrated dayside albedos. We also consider a micro-
five (Webber et al. 2015; Oreshenko et al. 2016; Parmentier et al.
physical code, Community Aerosol and Radiation Model for
2016, 2021; Roman et al. 2021) use their models to calculate
predicted albedos in the Kepler bandpass. The question of why Atmospheres (CARMA), which enables us to take a closer look at
individual planets with similar equilibrium temperatures would the role of nucleation, condensation, and sedimentation in shaping
exhibit widely varying cloud properties therefore remains largely the distribution of dayside clouds. In Section 3 we compare our
unexplored in the literature. predicted dayside-integrated optical albedos with observations of
Modeling the 3D structure of clouds can be computationally the six planets of interest. We then investigate the relative
demanding, especially given the large number of parameters that importance of various model assumptions, such as equilibrium
must be considered (including but not limited to the particle condensation versus kinetic condensation, by comparing the
number density and size distribution, spatial extent of the clouds, Virga results with the CARMA results. Finally, we discuss the
location of the cloud decks, number and composition of cloud implications of our results in Section 4 and present our conclusions
species, and coupled radiative feedback). In examining the body in Section 5.

2
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Table 1
Properties of our Planet Sample

Planet MP (MJ) RP (RJ) Teq(K )a g (m s−2) a (au) Period (days) Measured Ag Referenceb
K2-31b 1.77 1.06 1550 39.07 0.022 1.26 0.023 ± 0.002 1,9
Kepler-17b 2.45 1.31 1570 35.41 0.026 1.49 0.099 ± 0.017 2,3
HATS-11b 0.85 1.51 1560 7.97 0.051 3.62 0.270 ± 0.052 4,5,9
Kepler-7b 0.45 1.65 1630 4.16 0.062 4.89 0.25 ± 0.01 6,8
K2-107b 0.84 1.44 1650 10.26 0.048 3.31 0.062 ± 0.010 5,7,9
Kepler-8b 0.60 1.41 1680 7.32 0.047 3.85 0.124 ± 0.013 6,10

Notes. (1) Grziwa et al. (2016), (2) Desert et al. (2011), (3) Bonomo et al. (2017), (4) Bayliss et al. (2018), (5) Livingston et al. (2018), (6) Esteves et al. (2015),
(7) Eigmuller et al. (2017), (8) Heng et al. (2021), (9) Niraula et al. (2018), (10) Esteves et al. (2013).
a
Calculated assuming a Bond albedo of zero and efficient day–night recirculation.
b
Reference to geometric albedo measurement is bolded for each row.

2. Methods primitive equations on a staggered Arakawa C grid (Arakawa

et al. 1977) with the finite-volume method. The equations are
2.1. Planet Sample
discretized on a 128 × 64 cubed-sphere grid with 53 vertical
In this study we focus on planets with equilibrium temperatures layers extending from 200 bars at the bottom boundary to 20 μbar
(Teq, calculated assuming an albedo of zero and efficient day– at the top boundary. A horizontal fourth-order Shapiro filter is
night recirculation) between 1550 and 1700 K. For planets in this used to smooth horizontal noise. We let each model run for a
relatively narrow temperature range, we expect that reflective simulated 1000+ Earth days so as to reach quasi-steady-state
silicate clouds should dominate the optical dayside albedos (e.g., equilibrium.
Parmentier et al. 2016; Powell et al. 2018; Gao et al. 2020; The radiative transfer scheme solves the two-stream radiative
Parmentier et al. 2021; Roman et al. 2021). This temperature transfer equations using the correlated-k method (Goody et al.
range contains some of the most reflective hot Jupiters observed to 1989; Marley et al. 1999) over 11 spectral bins (Kataria et al.
date, including Kepler-7b (Demory et al. 2013; Heng et al. 2021) 2013). This coupling allows for the self-consistent calculation of
and HATS-11b (Niraula et al. 2018). Note that two observed the heating and cooling rates of the atmosphere with latitude,
geometric albedos have been reported for Kepler-7b, and from longitude, and pressure. At each grid point, the radiative transfer
here on we consider the most recent value from Heng et al. scheme calculates the upward and downward fluxes at each
(2021). It also includes three moderately reflective hot Jupiters pressure layer, which are used to update the heating/cooling
(Kepler-8b, Kepler-17b, and K2-107b; Desert et al. 2011; Esteves rates. These rates are used by the MITgcm to update the wind
et al. 2013; Niraula et al. 2018) and one very dark hot Jupiter (K2- and temperature fields. Opacities are computed at each pressure-
31b; Niraula et al. 2018). Previous studies (e.g., Demory et al. temperature point assuming chemical and thermodynamic
2013) have concluded that planets with these equilibrium equilibrium, using the solar photospheric elemental abundances
temperatures should have negligible amounts of thermal emission of Lodders (2003). We interpolate across the PHOENIX stellar
in the Kepler bandpass. We checked this by using PICASO to atmosphere models to generate an input spectrum for each host
compute the predicted thermal emission for each of the six planets star. The SPARC/MITgcm has been successfully utilized for a
in our sample and found a contribution of 2 ppm or less. This is series of hot Jupiter studies (e.g., Showman et al. 2009; Kataria
significantly smaller than the uncertainties on the secondary et al. 2013; Parmentier et al. 2013; Showman & Kaspi 2013;
eclipse depths used to calculate the geometric albedos for these Lewis et al. 2014; Kataria et al. 2015; Showman et al. 2015;
planets. Hence, throughout this paper we ignore the thermal Kataria et al. 2016), and we refer the reader to Kataria et al.
contribution to the measured secondary eclipse depth in the (2016) for further details. Figure 2 shows the resulting thermal
Kepler bandpass. structure for each of the six planets in our sample at 1 mbar
In this study we present a GCM tailored to each of the six (approximately the level of unit optical depth for clear skies).
individual planets, which we use to predict planet-to-planet
variations in cloud coverage and dayside albedo in the optical
Kepler band. These planets sample a range of surface gravities, 2.3. Computing Equilibrium Condensate Clouds with Virga
allowing us to investigate its effect on the planet’s thermal We use the thermal structure and vertical mixing rates from the
structure, cloud distributions, and optical albedo. We summar- SPARC/MITgcm model as inputs to Virga (Batalha 2020),6 an
ize the physical and orbital properties of each system in open-source code that calculates phase equilibrium cloud
Table 1. distributions. The cloud parameterization used in this code is
described in Ackerman & Marley (2001) and has been used for
2.2. Modeling Atmospheric Circulation with the SPARC/ 1D model studies across a wide range of exoplanets and brown
MITgcm dwarfs (e.g., Fortney et al. 2006; Marley et al. 2010; Morley
et al. 2015). This parametric approach also allows us to sample
We model each planet’s clear-sky (cloud-free) 3D thermal the 3D cloud structure at a much higher spatial resolution than
structure and atmospheric circulation with the Substellar and for our microphysical models.
Planetary Radiation and Circulation (SPARC) model, which We calculate Kzz profiles from the rms vertical velocities
couples the GCM maintained at the Massachusetts Institute of derived from global horizontal averages at a given pressure
Technology (the MITgcm; Adcrot et al. 2004) with a plane-
parallel, two-stream version of the multistream radiation code, as 6
Code and documentation available at https://natashabatalha.github.io/
described in Marley et al. (1999). The MITgcm solves the 3D virga/.

3
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Figure 2. Map of the dayside temperatures (in Kelvin) of each planet at 1 mbar, roughly the pressure of unit optical depth in a clear atmosphere in the Kepler
bandpass. Each planet is given a unique scale for the color bar to best match the relevant temperature range.

level from the GCMs by assuming Kzz = w(z)L(z), where w(z) tunable parameter and explore a range of possible values. Models
is the horizontally averaged global rms vertical velocity from with larger values of fsed will have high rates of sedimentation,
the GCM simulations, L(z) is approximated as the atmospheric concentrating the condensing species in the lower atmosphere.
pressure scale height H(z) (but could be a fraction of H(z); see Conversely, models with smaller values of fsed will have much
Smith 1998), and z is altitude. Moses et al. (2011) note that this slower sedimentation rates, allowing cloud particles to remain
is only an estimate; a better approach would involve calculating lofted higher in the atmosphere. For each planet, we run a suite of
Kzz from the eddy vertical velocity times the eddy displace- Virga models with fsed values of 0.03, 0.1, 0.3, 1.0, 3.0, and 6.0.
ment, but this information is not readily obtainable from the This range is motivated by comparisons to observational data,
GCMs; this could be resolved by adding passive tracers to which suggest that fsed can be as small as 0.01 for super-Earths
future GCM models. Our treatment may overestimate Kzz in the (Morley et al. 2015) or as large as 2–5 for some gas giants and
∼10–200 bar radiative region, where the vertical motion often brown dwarfs (Saumon & Marley 2008; Skemer et al. 2016;
consists of small-scale wave oscillations. MacDonald et al. 2018). For Jupiter’s ammonia clouds, an fsed of
We reduce the spatial resolution of our longitude and latitude ∼2 appears to provide the best match to observations (Ackerman
grid from 128 × 65 to 10 × 10 by binning the pressure-temp- & Marley 2001). We therefore conclude that our chosen list of
erature profiles and corresponding Kzz profiles prior to running values spans a representative range for this parameter.
Virga. We retain the original 53 pressure levels in the rebinned
grid. This binning has a negligible effect on our calculation of 2.4. Computing Microphysical Clouds with CARMA
the phase-integrated albedo and significantly reduces computa-
In addition to the parametric model described above, we also
tion time. We bin using the area mean with angles from the
utilize CARMA, a more computationally demanding microphysical
Chebyshev–Gauss integration method that vary as a function of
cloud formation model. CARMA calculates the equilibrium cloud
planetary latitude and longitude.
particle-size distribution by solving the 1D discretized continuity
In Virga the molar mixing ratio of the condensed phase, qc,
equation for aerosol particles that experience vertical transport due
is calculated by solving the equation
to sedimentation and eddy diffusion and production and loss due
¶qt to particle nucleation (homogeneous and heterogeneous), con-
- Kzz - fsed w*qc = 0, (1 ) densation, evaporation, and coagulation. CARMA has been prev-
¶z iously used to investigate condensate cloud formation on Earth
where qt is the total mixing ratio (condensed and vapor phases), (e.g., Ackerman et al. 1993; Jensen & Toon 1994; Ackerman et al.
1995), Venus (e.g., James et al. 1997; McGouldrick & Toon 2007;
ω* is the convective velocity scale, and fsed is defined as the
Gao et al. 2014), Mars (e.g., Colaprete et al. 1999), and exoplanets
ratio of the mass-weighted droplet sedimentation velocity to the
(e.g., Gao et al. 2018; Powell et al. 2019; Gao et al. 2020). CARMA
convective velocity, ω*. The product fsedω* describes an has also been used to model photochemical hazes on Titan (Toon
average sedimentation velocity for the condensate, which et al. 1992), Pluto (Gao et al. 2017), ancient Earth (Wolf &
offsets turbulent mixing. We refer the reader to Ackerman & Toon 2010), and warm Jupiters (Adams et al. 2019). In order to
Marley (2001) for more details regarding the equations that make our use of CARMA computationally tractable, we divide each
govern Virga. planet into two zones and calculate averaged temperature and Kzz
In Equation (1), fsed is the only parameter that cannot be profiles for each zone as described in Section 3.2 (see also
calculated directly from the models. We therefore treat it as a Figure 3). We do not consider photochemical hazes here, as the

4
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

particle masses twice that of the previous bin. We use 65 bins in

our model, with the mass in the first bin corresponding to particles
with radii of 0.1 nm for all species.

2.5. Cloud Compositions

For the Virga cloud modeling in this study, we are only
interested in cloud species that are expected to be abundant in
these atmospheres and which condense at relatively low
pressures (approximately 1 bar). We identify three cloud
species that are likely to be important based on comparing the
species’ condensation curves to the planets’ temperature-
pressure profiles (Figure 3; also see Ackerman & Marley 2001;
Morley et al. 2012): Mg2SiO4, Al2O3, and TiO2. This assumes
that all SiO goes into Mg2SiO4 rather than SiO2 or MgSiO3.
The inclusion of SiO2 would require a different modeling
framework including kinetic condensation. We plot the optical
depths in the Kepler bandpass (shown in Figure 6) for these
species as a function of pressure as computed by Virga at a
representative location on the planet K2-107b in Figure 4.
Although some previous studies of spatially inhomogeneous
cloud formation in hot Jupiters also included sulfide clouds
(e.g., Parmentier et al. 2016), recent microphysical cloud
Figure 3. Condensation curves (dashed gray) of Fe, Al2O3, TiO2, and Mg2SiO4 models (Powell et al. 2019; Gao et al. 2020) indicate that the
clouds compared with the two-zone model-averaged temperature profiles (see high surface energy of sulfide condensates makes homoge-
Section 3.2). Dayside (solid) and western limb (dashed) are shown for each neous nucleation unlikely. This conclusion is in good
planet: K2-107b (yellow), Kepler-8b (orange), HATS-11b (blue), K2-31b
(navy), Kepler-17b (purple), and Kepler-7b (green).
agreement with observational data indicating that the observed
cloud opacity for more moderately irradiated hot Jupiters is
well matched by models without sulfide condensates (e.g.,
Chachan et al. 2019; Gao et al. 2020). We therefore exclude
planets in our sample lie above the temperature range where these
sulfide condensates from our Virga models. Fe is also a
hazes are expected to form (Lodders & Fegley 2002; Kawashima
potentially important condensate but, like the sulfides, previous
& Ikoma 2019; Gao et al. 2020). microphysical modeling suggested that Fe cloud formation
In the CARMA model, the formation of condensate clouds proceeds slowly due to Fe’s high surface energy (Gao et al.
begins with either homogeneous or heterogeneous nucleation. 2020). On the other hand, Fe has been considered a major
Cloud particles undergo homogeneous nucleation when stable cloud component in many previous works (Sudarsky et al.
clusters of condensate molecules form and grow directly from the 2003; Fortney 2005; Helling et al. 2008; Marley et al. 2013;
vapor. The rate is controlled by the material properties of the Marley & Robinson 2015). Therefore, while our nominal
condensate, such as its molecular weight and surface energy, and Virga models will not include Fe, we will consider another
the flux of molecules to the cluster, which depends on the set of simulations where Fe is included to explore its impact on
abundance of condensate vapor. Under the same supersaturation our predicted albedos.
and local temperature, high-surface-energy and high-molecular- For the CARMA simulations in this study, we use the same
weight materials tend to nucleate more slowly than low-surface- model setup as in Gao et al. (2020) where clouds of TiO2,
energy and low-molecular-weight materials. Unlike homogeneous Al2O3, Fe, Mg2SiO4, Cr, MnS, Na2S, ZnS, and KCl are
nucleation, heterogeneous nucleation involves the formation of allowed to form. However, as previously discussed, in practice
stable clusters on condensation nuclei, or foreign surfaces, which only Mg2SiO4, Al2O3, and TiO2 form in any significant
are provided by other aerosol particles in the atmosphere. The size abundance. Of these three species, Mg2SiO4 is particularly
and abundance of these particles strongly impact the rate of unlikely to undergo homogeneous nucleation, as it is not
heterogeneous nucleation. The nucleation rate is also dependent abundant in the vapor phase. Instead, it is the product of a
on the interaction between the condensate and the surface, thermochemical reaction between Mg, SiO, and H2O (e.g.,
characterized by the contact angle between the surface and the Visscher et al. 2010). Similarly, Al2O3 does not exist in the
condensate cluster, the energy needed by a condensate molecule gaseous phase as other aluminum oxide species will be more
to desorb from the surface, and the oscillation frequency of the stable (e.g., Patzer et al. 2005). As in Gao et al. (2020), we
condensate molecule on the surface, which is related to the allow these two species to heterogeneously nucleate on
desorption energy (Pruppacher & Klett 1978). homogeneously nucleated TiO2 seeds. Though Fe is permitted
Unlike in Virga, which assumes a log-normal particle-size to both homogeneously and heterogeneously nucleate (on TiO2
distribution, particle-size distributions in CARMA are resolved seeds), the high surface tension of Fe prevents significant Fe
using mass bins and can change over time via condensation, cloud formation.
evaporation, and coagulation. The former two rates are dependent
on the flux of condensate molecules and the rate at which particles
2.6. Computing Optical Albedos with PICASO
may conduct away the latent heat released upon condensation.
Coagulation, or growth via physically sticking upon the collision We use the outputs from the cloud codes Virga and CARMA
of particles, is also considered. Each mass bin corresponds to to calculate the single-scattering albedos, asymmetry parameters,

5
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Figure 4. Nadir optical depths (integrated over the Kepler bandpass) for three condensate species in our Virga model of K2-107b. From left to right: Mg2SiO4, TiO2,
Al2O3. Each panel shows two representative grid points: blue corresponds to a western grid point centered at − 60° W, 8° N and red corresponds to a dayside grid
point centered at 42° E, 8° N. The value of fsed, which ranges from 0.03 to 6.0, is indicated by the shading of each line, where the lightest shades correspond to the
smallest values and darkest shades represent the largest values of fsed.

and optical depths for each pressure layer at each location in the PICASO considers the extinction from three opacity sources
atmosphere assuming Mie scattering. We then convert these in order to calculate the geometric albedo as a function of
quantities into wavelength-dependent albedos using The Planetary wavelength: molecular absorption, Rayleigh scattering, and
Intensity Code for Atmospheric Scattering Observations scattering by clouds. We show a representative calculation of
(PICASO; Batalha et al. 2019). This code is governed by the these three opacity sources as a function of wavelength in
radiative transfer equation Figure 5. To describe the phase-dependence, PICASO
computes the emergent intensity from the disk at multiple
dt
I (ti , m) = I (ti + 1, m) exp ⎛⎜ i ⎟⎞ plane-parallel facets, where each has its own incident and
⎝m⎠ outgoing angles. PICASO uses the Chebyshev–Gauss integra-
dti t tion method to integrate over all emergent intensities. We
- ò0 S (t ¢m) exp ( - ) dt ¢ / m ,
m
(2 ) integrate the wavelength-dependent geometric albedo over the
Kepler response function for each point in our 10 × 10 grid and
where I(τi, μ) is the azimuthally averaged intensity emerging then integrate again over the dayside hemisphere in order to
from the top of a layer, i, with opacity, τi, and outgoing angle, obtain a geometric albedo that we can compare with the Kepler
dt measurements. We have run sensitivity tests that demonstrate
μ. I (ti + 1) exp ( mi ) is the incident intensity at the lower boundary
that a higher resolution grid (20 × 20 grid) yields comparable
of the layer attenuated by the optical depth within the layer, δτ, results.
and S (t ¢ , m ) is the source function integrated over all layers.
The source function has two components: single-scattered and
multiple-scattered radiation integrated over all diffuse angles: 2.7. Effect of Simplifying Model Assumptions
w t¢ In this study we do not consider radiative feedback from
S (t ¢ , m ) = Fo Psingle (m , - mo) exp ( - ) clouds, which might affect our albedo predictions. Clouds can
4p ms
alter the planet’s global thermal structure in several ways. First,
w 1
+
2 -1ò I (t ¢ , m¢) Pmulti (m , m¢) dm¢ , (3 ) they can reduce the amount of heating on the dayside by
increasing the planet’s Bond albedo. Second, they can suppress
where ω is the single-scattering albedo and Fo is the incident cooling on the nightside by preventing the reradiation of
flux. Pmulti and Psingle describe the phase function of the infrared light to space. Parmentier et al. (2021) and Roman
et al. (2021) ran grids of 3D GCM models incorporating
multiple and single scattering, respectively. Psingle is an
radiative feedback from clouds spanning a range of incident
opacity-weighted combination of the Rayleigh phase function
fluxes. They found that the presence of reflective clouds on the
and a two-term Henyey–Greenstein phase function. Pmulti dayside resulted in lower global temperatures, while the
requires integration over all diffuse angles, for which PICASO presence of nightside clouds inhibited cooling, causing a
uses a N = 2 Legendre expansion. This approximation alone is 100–200 K global increase in temperature. With the possible
inadequate to represent cases with high rates of forward exception of Kepler-7b (see Section 4.2.1), we expect that such
scattering, so PICASO implements the delta-Eddington shifts in temperature would not substantially alter the pressures
approximation to scale g, ω, and τ to more accurately capture of the cloud decks or reduce their horizontal extent for the
the forward scattering peak. planets examined here.

6
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Figure 5. Pressure level (bar) of unit optical depth for Rayleigh scattering (pink), cloud opacity (green), and molecular opacity (blue) as a function of wavelength for a
single grid point (−60°W, 8° N) in the Virga model of hot Jupiter K2-107b. fsed of 0.1 is shown on the left and 3.0 on the right. The shaded regions indicate the
dominant opacity source as a function of wavelength. This is the same grid point shown in Figure 4.

We also note that, in this study, our 3D atmospheric of temperature with respect to pressure is related to the inverse
circulation models are decoupled from our cloud models. of the surface gravity (e.g., Gao et al. 2018).
While this does enable more flexibility in exploring different
cloud species and sedimentation parameters in both Virga and
CARMA, cloud formation and transport is ultimately a coupled 3.1. Virga Model Results
process between advection, radiation, and chemistry. Previous In order to determine the effect of the clouds on the albedo,
studies that couple cloud microphysical models and atmo- we must first calculate their vertical extent at each location in
spheric circulation models (e.g., Lee et al. 2015; Lines et al. our model grid. In our Virga equilibrium cloud models, the
2018) suggest that zonal (east/west) and/or meridional (north– vertical extent of the clouds is controlled by our choice of fsed.
south) transport of cloud particles from colder regions of the Taking K2-107b as a representative example, Figure 4
atmosphere could lead to high cloud-particle number densities, indicates that Mg2SiO4 clouds reach a unit optical depth near
even on the hotter dayside, which could in turn enable more 1 mbar for the lowest fsed value, 0.03, and near 50 mbar for
cloud nucleation and growth. These processes will only affect fsed = 0.1. These clouds will only contribute to the albedo at
our dayside albedo predictions if the planet in question has wavelengths where they reach optical depth unity at lower
large dayside temperature (and hence albedo) gradients with pressures than Rayleigh scattering or molecular opacity
longitude or latitude. We revisit both of these assumptions (Figure 5). This means that clouds will have a greater effect
(radiative feedback and cloud microphysics coupled to at wavelengths where the combined molecular and Rayleigh
circulation) in Section 4.2, where we discuss their implications scattering opacity is lower. We show the effect of varying fsed
for individual planets in light of our model results. on the wavelength-dependent albedo of K2-107b in Figure 6.
As fsed decreases and the clouds extend to lower pressures, the
cloud opacity contributes most to the overall albedo for an
3. Results increasingly large fraction of the Kepler bandpass.
We next examine how the contribution of clouds to the band-
The SPARC/MITgcm models indicate that the day–night integrated albedo varies across the dayside atmosphere. In
temperature contrasts for the six planets in our sample vary in Figure 7, we show the 10 × 10 grid of albedos in the Kepler
magnitude (Figure 2). As expected, the hottest region on the bandpass for each planet as a function of fsed. We find that the
dayside in all six models is located to the east of the substellar three planets with the greatest temperature variation as a
point. This shift is caused by eastward equatorial winds, which function of longitude (Kepler-7b, K2-31b, and Kepler-17b)
transport heat to the planet’s nightside (e.g., Showman et al. also have relatively large albedo variations across their dayside
2020, and references therein). atmospheres for low-to-intermediate fsed values. Once fsed
Kepler-7b has the largest thermal gradient of all the planets increases above 0.3, the clouds remain confined below the
in our sample, followed by K2-31b and Kepler-17b; this is optical depth unity level of molecular absorbers, such that the
because the former has a relatively low surface gravity presence or absence of clouds does not affect the observed
(approximately 4 m s−2) and the latter two have the shortest albedo. These three planets appear brighter on their western
orbital periods in the sample. limbs than in the east, in good agreement with the albedo maps
Kepler-7b has the largest thermal gradient of all the planets derived from the Kepler phase curve for Kepler-7b (Demory
in our sample, followed by K2-31b and Kepler-17b; this is et al. 2013) and other planets with comparable equilibrium
because the former has a relatively low surface gravity temperatures (Shporer & Hu 2015). In contrast to these three
(approximately 4 m s−2) and the latter two have the shortest planets, HATS-11b, K2-107b, and Kepler-8b all appear to have
orbital periods in the sample. Our models indicate that Kepler- fairly homogeneous albedos, as expected based on their more
7b is also warmer at depth than the other planets in this sample. homogeneous thermal structures.
This is expected, as Kepler-7b has the lowest surface gravity of Lastly, we compare the hemisphere-averaged dayside albedo
the six planets and under hydrostatic equilibrium the gradient in the Kepler bandpass as a function of fsed to the measured

7
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

greatest difference in albedo at mid-latitudes as determined

from the Virga albedo maps. Occasionally the western zone
includes a limb region in the east; see Table 3 for a list of the
zone definitions for each planet. The resulting zonally averaged
pressure-temperature profiles are plotted in Figure 3. We run a
separate CARMA model for each zone, and also run Virga
models for the same zones in order to facilitate comparisons
between the two models. In the discussion below, we limit our
comparisons to these two-zone Virga models unless other-
wise noted.
When comparing the two models, it is important to note that
they predict distinct particle compositions. While Virga
assumes that particles nucleate homogeneously, CARMA
predicts that the Mg2SiO4 condensates will nucleate hetero-
geneously onto TiO2 cores (see Section 2.5). We find that
Figure 6. Hemisphere-averaged albedo of K2-107b as a function of
wavelength for a clear atmosphere (blue) and cloudy atmospheres of varying accounting for this TiO2 core when we calculate the single-
fsed values (green; 0.03 as a solid line, 0.1 as a dashed line, and 0.3 as a dotted scattering albedos for the Mg2SiO4 particles in our CARMA
line). Cloudy atmospheres including Fe condensates are shown in purple. All models results in a lower single-scattering albedo than when we
cloud distributions are computed using Virga. The Kepler response function repeat the calculation for pure Mg2SiO4 particles, which return
is overplotted in gray.
single-scattering albedo profiles of greater than 0.96. For
comparison, we refer the reader to Figure 9; a single value for
dayside albedo for each planet (Figure 8). We find that K2-31b single-scattering albedo is not representative of heterogeneous
and K2-107b are best described by models with large values of particles, which are sensitive to the core mass fraction of TiO2.
fsed or (equivalently) clear atmospheres, indicating that any However, even after accounting for this effect we find that the
reflective clouds present in these atmospheres do not extend
species-averaged single-scattering albedo shown in Figure 9 for
above the level of unit molecular opacity. Kepler-17b and
our CARMA models with heterogeneous particles is still
Kepler-8b are best matched by models with intermediate fsed
comparable to the Virga models with fsed equal to 0.1
values, implying that their albedos are moderately enhanced by
(Figure 10).
cloud opacity. HATS-11b is best matched by models with low
When we calculate the hemisphere-integrated albedos for
fsed values, corresponding to a high, vertically extended
these planets from the CARMA models we find that they
reflective cloud layer spanning a wide range of longitudes. No
generally lie within the range of Virga predictions for varying
fsed value is able to reproduce the high observed albedo for
Kepler-7b, which our models indicate is too warm for bright fsed (see Figure 8 and Table 3). In Figure 8, we show the
Mg2SiO4 clouds to form over much of the dayside. Instead, our hemisphere-integrated albedos from the full-resolution Virga
models indicate that the dayside of Kepler-7b is dominated by model from Section 2.3 rather than the two-zone model
deep Al2O3 clouds at pressures of around a bar. The addition of discussed in this section.
iron clouds do not greatly change our results except for We find that K2-107b and HATS-11b appear to have the
decreasing the albedos of low fsed models, as iron clouds sink brightest hemisphere-integrated CARMA albedos, with values
below the photosphere for higher fsedʼs. We list the observed consistent with those of the lowest fsed (0.03 − 0.1) Virga
albedos and the predicted albedos as a function of fsed in models. Kepler-8b, Kepler-17b, and K2-31b have somewhat
Table 2. We find that the hemisphere-integrated albedo over the lower predicted CARMA albedos, more comparable to the fsed of
Kepler bandpass is very sensitive to the assumed value of fsed. 0.3 Virga models. Kepler-7b is a notable exception, as our
CARMA model predicts an albedo that is higher than that of the
smallest fsed Virga model.
3.2. Comparison to CARMA Microphysical Models
We find that dayside albedos can vary significantly
depending on the assumed sedimentation efficiency (e.g., 4. Discussion
Figure 7). We therefore use these maps to divide each planet 4.1. Comparison to Previous Modeling Efforts
into two zones, and then utilize the more computationally
demanding CARMA microphysical cloud model to solve for the We can compare our results to generic grids of cloudy GCM
vertical extent of the clouds and corresponding albedo in models in the published literature. First, we note that our GCM
each zone. outputs for Kepler-7b roughly agree with those shown in
Our Virga albedo maps indicate that a subset of the planets Oreshenko et al. (2016), and the general distribution of our
in our sample are likely to have spatially inhomogeneous silicate and corundum clouds agrees with the modeling results
Mg2SiO4 clouds located near their western limbs. Although it of Roman & Rauscher (2019).
would be computationally prohibitive to run a separate CARMA Roman et al. (2021) investigated planetary albedos using a
model for each point in the 10 × 10 grid, we can nonetheless grid of GCMs with varying irradiation temperature and surface
capture this cloud structure using a more computationally gravity. The closest equivalent models in their grid are for a
tractable two-zone model. We define a threshold value in planet with an irradiation temperature of either 2500 K
longitude for each planet based on the albedo distributions (equilibrium temperature of 1500 K) or 2750 K (equilibrium
found by Virga as shown in Figure 7 and in Table 3. We temperature of 1700 K) and a surface gravity of 10 m s−2. We
determine the longitude range defining the two zones (a compare to their nucleation-limited models, which also exclude
western zone and a dayside zone) as the division that yields the iron and sulfide condensates.

8
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Figure 7. Relative dayside albedo contribution at each grid point including appropriate geometric weights. Planets are sorted by increasing equilibrium temperature
from top to bottom, and sorted by increasing fsed (0.03, 0.1, 0.3, and 1.0) from left to right. We omit fsed values of 3.0 and 6.0 from the figure, as the clouds in these
models reside below the level of unit molecular albedo, and thus the albedo remains roughly unchanged compared to the fsed = 1.0 case. The relative albedo
contributions are all smaller than the face-integrated albedo by roughly a factor of π; summing over the points yields the integrated albedo. The models include three
cloud species: Mg2SiO4, Al2O3, and TiO2.

The models presented in Roman et al. (2021) assume a fixed feedback from these clouds when solving for the temperature
pressure-dependent particle size for the clouds, with a size of structure of the atmosphere. These models indicate that compact
0.1 μm at the top of the atmosphere that increases exponentially cloud layers will result in relatively low and uniform dayside
with increasing pressure for pressures greater than 10 mbar. Since albedos, in good agreement with our results (Figure 7). Their
their equilibrium cloud models do not solve for the vertical extent vertically extended cloud models exhibit a range of optical
of the cloud layers, they present two cases corresponding to spherical albedos between 0.2 and 0.3, with lower values for the
compact (cloud tops limited to 1.4 scale heights above the cloud higher temperature model as the reflective silicate clouds become
base) and vertically extended (cloud-top pressure of 0.1 mbar) increasingly confined to the cooler western region of the
cloud layers. Unlike our models, they account for radiative atmosphere. This also agrees with the qualitative picture from

9
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Figure 8. Geometric albedo predictions in the Kepler optical bandpass compared with published albedo measurements as a function of equilibrium temperature.
Albedo predictions from the full-resolution Virga models are shown as circles, where the shading indicates the fsed value, going from 0.03 (light) to 6.0 (dark).
Kepler albedo measurements are shown as stars while the predicted albedos from the two-zone CARMA models are shown as triangles. The left panel excludes Fe in
Virga calculations, while the right panel includes Fe condensates.

Table 2
Hemisphere-averaged Albedos from Virga Models as a Function of fseda

Planet Measured Clear 0.03 0.1 0.3 1.0 3.0 6.0 0.1, with Fe
K2-31b 0.023 ± 0.002 0.015 0.404 0.123 0.023 0.015 0.015 0.015 0.115
Kepler-17b 0.099 ± 0.017 0.017 0.416 0.131 0.023 0.017 0.017 0.017 0.125
HATS-11b 0.270 ± 0.052 0.066 0.301 0.127 0.060 0.066 0.066 0.066 0.127
Kepler-7b 0.194 ± 0.013 0.009 0.064 0.034 0.015 0.010 0.009 0.009 0.030
K2-107b 0.062 ± 0.010 0.065 0.333 0.154 0.063 0.065 0.065 0.065 0.134
Kepler-8b 0.124 ± 0.013 0.069 0.319 0.151 0.063 0.070 0.070 0.070 0.151

Note.
a
Bolded values indicate the simulated Virga albedo that best matches the Kepler eclipse observations.

Table 3
Hemisphere-averaged Dayside Albedos from CARMA Models

Planet Measured Clear West Day Day Zonea Integrated

K2-31b 0.023 ± 0.002 0.015 0.085 0.054 [−17°W, 51°E] 0.073
Kepler-17b 0.099 ± 0.017 0.017 0.074 0.069 [0°E, 90°E] 0.073
HATS-11B 0.270 ± 0.052 0.066 0.169 0.165 [51°E, 90°E] 0.168
Kepler-7b 0.194 ± 0.013 0.009 0.199 0.034 [−38°W, 51°E] 0.128
K2-107b 0.062 ± 0.010 0.065 0.233 0.229 [51°E, 90°E] 0.280
Kepler-8b 0.124 ± 0.013 0.069 0.082 0.109 [51°E, 90°E] 0.092

Note.
a
Eastern/day zone boundaries are listed in the table. The western/limb zone is defined as the region excluded by the dayside definition.

our models, which span an equivalent range of albedos. They dominate the dayside optical albedo for planets with equilibrium
conclude that their optical albedos for the nucleation-limited case temperatures between 1500 and 1700 K. In Parmentier et al.
are dominated by silicate cloud particles, consistent with our (2021) they incorporate cloud radiative feedback for the
conclusions here. condensation of a single cloud species, MnS, but the treatment
Parmentier et al. (2016) and Parmentier et al. (2021) considered of silicate clouds is similar to Parmentier et al. (2016). For that
GCMs spanning a wide range of equilibrium temperatures. In reason, we focus on Parmentier et al. (2016) for our comparison.
Parmentier et al. (2016), the clouds are post-processed (i.e., they In this study they assume a small fixed-particle-size distribution
do not include cloud radiative feedback in the GCM models), and centered at 0.1 μm and a cloud-top pressure of 1 microbar, which
they model the condensation of a wide range of cloud species. yields geometric albedos greater than 0.5 in the Kepler bandpass
These models predict that MgSiO3 and CaTiO3 clouds should for planets of approximately 1500 K equilibrium temperature.

10
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Figure 9. Optical depth (left), single-scattering albedo (center), and asymmetry parameter (right) as a function of pressure for the two-zone Virga (dashed; fsed fixed
to 0.1) and CARMA (solid) models, calculated by dividing the dayside hemisphere into a dayside (red) and western (blue) zone (see Section 3.2). Each row corresponds
to a different planet, sorted left to right by increasing equilibrium temperature.

Figure 10. Dayside and western zone albedos from the CARMA models integrated over the Kepler bandpass. Virga model albedos with fsed equal to 0.1 are listed
below for comparison. Both sets of models are calculated in two zonally averaged regions, defined in Table 3. Planets are sorted from left to right and top to bottom by
increasing equilibrium temperature.

This value is much higher than both our albedos and those 4.2. Comparison to Published Kepler Albedos
reported by Roman et al. (2021), and is most likely due to the very
Our results show that, with the exception of Kepler-7b, it is
low cloud-top pressure assumed in these models. They invoke a
cold trap for silicates to reduce the albedo, while we predict that possible to match the observed optical geometric albedos for all
changing the cloud vertical extent can achieve a similar difference of the planets in our sample using either Virga or CARMA
in observable albedo. models. However, no single model (Virga at a fixed fsed or

11
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

CARMA) can explain the observed albedos of all six planets. brighter cloud particles. We discuss how porous particles may
Although the albedo predictions from the CARMA models are increase HATS-11bs dayside albedo in Section 4.2.3; this same
broadly consistent with those of the Virga models, neither explanation might also apply to Kepler-7b.
model is able to reproduce or explain the observed planet-to-
planet variations in dayside albedo. This suggests that
4.2.2. K2-31b and K2-107b Do Not Host High-altitude Reflective
accounting for differences in equilibrium temperature, host-
Cloud Layers
star spectral type, surface gravity, and rotation rate alone are
not sufficient to capture the observed diversity of hot Jupiter The observed albedos of K2-31b and K2-107b are relatively
albedos in this temperature range. We discuss our results in low, and are well matched by Virga models with clear skies
more detail on a planet-by-planet basis below. and/or deep clouds (i.e., those with opacities dominated by
molecular absorption). For K2-31b, the contribution of clouds to
the albedo is negligible for fsed of 0.3 and larger. For K2-107b, we
4.2.1. Kepler-7b: Spatially Resolved Clouds
obtain a comparable result for fsed of 1.0 and larger. This suggests
Our CARMA and Virga models both indicate that Kepler-7b’s that both of these planets may have relatively efficient
relatively bright dayside albedo is dominated by reflection from sedimentation (deep clouds), or alternatively that they have
the region near the western limb, in good agreement with spatially relatively little condensable material in their atmospheres (perhaps
resolved albedo constraints from phase-curve observations corresponding to a relatively low atmospheric metallicity). If the
(Demory et al. 2013; Heng et al. 2021). This underscores the lack of clouds is due to efficient sedimentation, this would appear
importance of using spatially resolved cloud models for tidally to contradict predictions from our microphysical CARMA models,
locked hot Jupiters. We note that there is some tension between which track the sedimentation of cloud particles explicitly and
our model predictions and the observational data, as fits to Kepler- predict albedos that are a factor of two or more higher than the
7b’s optical phase curve indicate that the bright reflective western observed values for these two planets.
zone extends as far as 10° ± 6° west of the substellar point Our CARMA models utilize vertical mixing rates calculated from
(Muñoz & Isaak 2015; Heng et al. 2021). Our Virga models our GCMs. If these mixing rates are overestimates of the true
predict that the atmosphere will only be cool enough for Mg2SiO4 values, we might expect any clouds near the day–night terminator
clouds to condense in the two westernmost longitude bins on these planets to also be relatively compact. If this is the case,
(extending from the terminator to approximately 38° west of the the transmission spectra of these two planets should show
substellar point; see Figure 7 and Table 3). This is likely why our relatively strong absorption features. Although K2-31b has a high
models underpredict Kepler-7b’s optical geometric albedo. surface gravity and is therefore a more challenging target for
We consider two possible explanations for this discrepancy. transmission spectroscopy, K2-107b might be accessible to future
Zonal transport of cloud particles from the western-limb region space telescopes like the James Webb Space Telescope. More
could increase the albedo in adjacent longitudes where the broadly, the sedimentation rates calculated from GCMs and
atmosphere is otherwise too warm for them to condense (see CARMA could be tested with comparisons to transmission spectra
Section 2.7). However, our models for Kepler-7b prefer small from ongoing surveys (Sing et al. 2016; Crossfield & Kreidberg
cloud particles with a large vertical extent; these small particles 2017; Fu et al. 2017). If CARMA models underestimate the
might have a relatively short lifetime in the hotter substellar sedimentation efficiency for other planets, the disagreement
region of the dayside atmosphere. Instead, perhaps small should be detectable in these data, which are very sensitive to
particles transported meridionally could nucleate and grow in the vertical distribution of cloud particles near the limb. To date,
bands at high latitudes (Lines et al. 2018). Ultimately, this most planets appear to be well matched by CARMA-model
planet would be an interesting test case for microphysical predictions (e.g., Chachan et al. 2020; Gao et al. 2020), indicating
transport models (e.g., Lee et al. 2015; Lines et al. 2018), that such model–data disagreements may be relatively rare.
which can explicitly quantify the timescales of these processes
and predict the resulting horizontal distribution of cloud
4.2.3. HATS-11b, K2-31b, and K2-107b Have Reflective Dayside
particles.
Clouds
Alternatively, if Kepler-7b’s dayside is cooler than predicted
by our GCM, Mg2SiO4 would be able to condense over a wider The measured Kepler albedo for HATS-11b (0.27 ± 0.05) is
range of longitudes. Our GCMs do not account for reflectivity brighter than that of Kepler-7b, in good agreement with our
from clouds when calculating the effect of incident starlight on model predictions. This planet is cooler than Kepler-7b, with a
the dayside atmosphere; this effect might reduce the magnitude smaller day–night temperature gradient. As a result, our models
of dayside heating and result in globally lower temperatures predict a global reflective dayside cloud layer for small fsed
(Lines et al. 2018; Roman & Rauscher 2019; Roman et al. values, and our hemisphere-integrated Virga albedo for the
2021). However, if the clouds extend over a significant fraction fsed of 0.03 model is within 1σ of the measured value. This is an
of the planet’s nightside it could result in net global warming, unusually small value of fsed, compared to the other planets’
as they would act to reduce the amount of energy that can be best-fit value, while more typical values of 0.1 and larger
radiated to space in this region (Roman & Rauscher 2019; underestimate the observation. The CARMA-model albedo is
Parmentier et al. 2021; Roman et al. 2021). For Kepler-7b, somewhat lower, but is still within 2σ of the observed value. If we
whose clouds extend over much of the western hemisphere, it is wish to adapt our models to better match this planet’s high dayside
unclear which of these two competing effects would dominate. albedo, it likely would require increasing the predicted cloud
These explanations assume that the dayside-coverage area is opacity, such as by increasing the porosity of the cloud particles
the most significant limiting factor on the brightness of the (Samra et al. 2020). Although increasing the atmospheric
dayside-integrated albedo. However, it is also important to metallicity might also increase the cloud opacity, published
consider factors that might increase the brightness of the cloudy models for other planets indicate that there is not a simple
region, including a larger vertical extent for the clouds or scaling between these two quantities (e.g., Morley et al. 2013;

12
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Gao et al. 2018), and increasing the metallicity will also affect the indicate that it extends farther east than predicted by our models,
global thermal structure (e.g., Kataria et al. 2015). hinting that a more detailed study of the planet that couples
Our model predictions for Kepler-17b and Kepler-8b are also cloud microphysics and dynamics is required.
in reasonable agreement with the observed albedos. In both We conclude that the sample of optical albedos measured by
cases, an fsed = 0.1 Virga model slightly overestimates the Kepler represents a rich source of information for 3D cloud
albedo while our CARMA model slightly underestimates the models, and that there is no single explanation for the observed
albedo. This may indicate that moderately bright/cloudy diversity of albedos for the planets considered in this study. Future
worlds have moderate sedimentation efficiencies. For Kepler- studies leveraging the large sample of transmission spectra of hot
8b, both CARMA and Virga models predict that the planet will Jupiters could provide complementary constraints on the typical
have relatively uniform cloud coverage in both latitude and sedimentation efficiencies of their atmospheres, while additional
longitude; we therefore do not need to consider further spatial complementary modeling studies exploring the coupled effects of
variations in cloud number density and particle size. For atmospheric dynamics and cloud microphysics, as well as an
Kepler-17b, the fsed = 0.1 Virga model predicts an albedo exploration of the microporosity of cloud particles, would help to
gradient across the dayside atmosphere, but this gradient further illuminate the relative importance of these processes in
appears to be localized near the equatorial (low-latitude) region explaining the high albedos of the brightest planets in our sample.
of the atmosphere. Our CARMA model predicts a relatively
uniform albedo across the two zones, but this may be biased by We thank the anonymous reviewer for an interesting report
our inability to resolve latitudinal gradients in the simplified that greatly helped improve the paper. P. Gao acknowledges
two-zone model. support from NASA through the NASA Hubble Fellowship
grant No. HST-HF2-51456.001-A, awarded by the Space
Telescope Science Institute, which is operated by the Associa-
5. Conclusions
tion of Universities for Research in Astronomy, Inc., for
Optical secondary eclipse measurements made by Kepler NASA, under contract NAS5-26555.
reveal a wide range of geometric albedos for hot Jupiters with Software: numba (Lam et al. 2015), pandas (McKinney
equilibrium temperatures between 1550 and 1700 K. We 2010), bokeh (Bokeh Development Team 2014), NumPy
combine 3D GCMs with both equilibrium (Virga) and (Walt 2011), IPython (Pérez & Granger 2007), Jupyter,
microphysical (CARMA) cloud models to explore whether 3D (Kluyver et al. 2016), Virga (Batalha 2020), PICASO (Batalha
effects can explain these observations. We find that the et al. 2019), SQLite (sqlite3 Development Team 2019).
predicted albedos from our Virga models are very sensitive
to the assumed sedimentation efficiency ( fsed). We can compare ORCID iDs
these albedo predictions to results from our CARMA model,
which use mixing rates calculated from the GCM models to Danica J. Adams https://orcid.org/0000-0001-9897-9680
predict the vertical extent and particle-size distributions of the Tiffany Kataria https://orcid.org/0000-0003-3759-9080
clouds. We find that while the hemisphere-integrated CARMA Natasha E. Batalha https://orcid.org/0000-0003-1240-6844
albedos generally agree with the range of albedos predicted by Peter Gao https://orcid.org/0000-0002-8518-9601
Virga, there is no single fsed value that consistently matches Heather A. Knutson https://orcid.org/0000-0002-5375-4725
the CARMA predictions.
When we compare these model predictions to the measured References
Kepler albedos for each of the six planets in our sample, we
find that the albedos of K2-31b and K2-107b are best matched Ackerman, A. S., Hobbs, P. V., & Toon, O. B. 1995, JAtS, 52, 1204
by models that are either cloud-free or have very deep compact Ackerman, A. S., & Marley, M. S. 2001, ApJ, 556, 872
cloud layers (large fsed values). Kepler-8b and Kepler-17b’s Ackerman, A. S., Toon, O. B., & Hobbs, P. V. 1993, Sci, 262, 226
Adams, D., Gao, P., de Pater, I., & Morley, C. V. 2019, ApJ, 874, 61
optical albedos can be matched by moderately cloudy models Adcrot, A., Campin, J.-M., Hill, C., & Marshall, J. 2004, MWRv, 132, 2845
( fsed greater than 0.3). Both Virga and CARMA tend to Angerhausen, D., Huber, K., Mandell, A., et al. 2014, in Proc. IAU, Formation,
underpredict the dayside albedos of the two most reflective Detection, and Characterization of Extrasolar Habitable Planets
planets in our sample, HATS-11b and Kepler-7b, which are (Cambridge: Cambridge Univ. Press), 293
best matched by Virga models with reflective Mg2SiO4 Arakawa, E. T., Williams, M. W., & Inagaki, T. 1977, JAP, 48, 3176
Barclay, T., Huber, D., Rowe, J. F., et al. 2012, ApJ, 761, 53
clouds extending to very low pressures ( fsed = 0.03); our Barstow, J. K., Aigrain, S., Irwin, P. G. J., et al. 2014, ApJ, 786, 154
CARMA model for HATS-11b predicts a slightly lower albedo Batalha, N. 2020, natashabatalha/virga: Initial Release, v0.0, Zenodo, doi: 10.
value than the brightest Virga model, while our CARMA 5281/zenodo.3759888
model for Kepler-7b predicts a slightly higher albedo value Batalha, N. E., Marley, M. S., Lewis, N. K., & Fortney, J. J. 2019, ApJ, 878, 70
Bayliss, D., Hartman, J. D., Zhou, G., et al. 2018, AJ, 155, 119
than Virga. Although HATS-11b has relatively uniform cloud Bell, T. J., Nikolov, N., Cowan, N. B., et al. 2017, ApJL, 847, L2
coverage across the dayside, it is possible that other factors Bokeh Development Team 2014, Bokeh: Python library for interactive
(such as a low particle porosity) might increase the dayside visualization, http://www.bokeh.pydata.org
cloud opacity beyond the values predicted by our models. Bonomo, A. S., Desidera, S., Benatti, S., et al. 2017, A&A, 602, A107
Our models predict that the observed albedo of Kepler-7b Burrows, A., Ibgui, L., & Hubeny, I. 2008, ApJ, 682, 1277
Chachan, Y., Jontof-Hutter, D., Knutson, H. A., et al. 2020, AJ, 160, 201
should be lower than that of HATS-11b, in good agreement with Chachan, Y., Knutson, H. A., Gao, P., et al. 2019, AJ, 158, 244
the observations. Although a bright reflective cloud layer forms Colaprete, A., Toon, O. B., & Magalhães, J. A. 1999, JGRE, 104, 9043
in the westernmost region of the dayside atmosphere, most Coughlin, J., & Lopez-Morales, M. 2012, ApJ, 143, 39
dayside longitudes in Kepler-7b’s atmosphere are too warm for Crossfield, I. J. M., & Kreidberg, L. 2017, AJ, 154, 261
Demory, B.-O., de Wit, J., Lewis, N., et al. 2013, ApJL, 776, L25
Mg2SiO4 to condense, resulting in a lower hemisphere-averaged Demory, B.-O., Seager, S., Madhusudhan, N., et al. 2011, ApJ, 735, L12
dayside albedo. Empirical constraints on the horizontal extent Desert, J.-M., Charbonneau, D., Fortney, J. J., et al. 2011, ApJS, 197, 11
of the western cloudy region from phase-curve observations Eigmuller, P., Gandolfi, D., Persson, C. M., et al. 2017, AJ, 153, 130

13
The Astrophysical Journal, 926:157 (14pp), 2022 February 20 Adams et al.

Esteves, L. J., De Mooij, E. J. W., & Jayawardhana, R. 2013, ApJ, 772, 51 Marley, M. S., Ackerman, A. S., Cuzzi, J. N., & Kitzmann, D. 2013, in Clouds
Esteves, L. J., De Mooij, E. J. W., & Jayawardhana, R. 2015, ApJ, 804, 150 and Hazes in Exoplanet Atmospheres, ed. S. J. Mackwell et al. (Tucson,
Evans, T. M., Pont, F., Sing, D. K., et al. 2013, ApJL, 772, L16 AZ: Univ. Arizona Press), 367
Fortney, J. J. 2005, MNRAS, 364, 649 Marley, M. S., Gelino, C., Stephens, D., Lunine, J. I., & Freedman, R. 1999,
Fortney, J. J., Cooper, C. S., Showman, A. P., Marley, M. S., & ApJ, 513, 879
Freedman, R. S. 2006, ApJ, 652, 746 Marley, M. S., Saumon, D., & Goldblatt, C. 2010, ApJL, 723, L117
Fortney, J. J., Demory, B.-O., Désert, J.-M., et al. 2011, ApJS, 197, 9 McGouldrick, K., & Toon, O. B. 2007, Icar, 191, 1
Fraine, J., Mayorga, L. C., Stevenson, K. B., et al. 2021, AJ, 161, 269 McKinney, W. 2010, in Proc. of the 9th Python in Science Conf., Data
Fu, G., Deming, D., Knutson, H., et al. 2017, ApJL, 847, L22 Structures for Statistical Computing in Python, ed. S. van der Walt &
Gao, P., Fan, S., Wong, M. L., et al. 2017, Icar, 287, 116 J. Millman, 56
Gao, P., Marley, M. S., & Ackerman, A. S. 2018, ApJ, 855, 86 Morley, C. V., Fortney, J. J., Kempton, E. M.-R., et al. 2013, ApJ, 775, 33
Gao, P., Marley, M. S., Zahnle, K., Robinson, T. D., & Lewis, N. K. 2017, AJ, Morley, C. V., Fortney, J. J., Marley, M. S., et al. 2012, ApJ, 756, 172
153, 139 Morley, C. V., Fortney, J. J., Marley, M. S., et al. 2015, ApJ, 815, 110
Gao, P., Thorngren, D. P., Lee, G. K. H., et al. 2020, NatAs, 4, 951 Moses, J. I., Visscher, C., Fortney, J. J., et al. 2011, ApJ, 737, 15
Gao, P., Zhang, X., Crisp, D., Bardeen, C. G., & Yung, Y. L. 2014, Icar, Muñoz, A. G., & Isaak, K. G. 2015, PNAS, 112, 13461
231, 83 Niraula, P., Redfield, S., de Wit, J., et al. 2018, arXiv:1812.09227
Goode, P. R., Qiu, J., Yurchyshyn, V., et al. 2001, GeoRL, 28, 1671 Oreshenko, M., Heng, K., & Demory, B.-O. 2016, MNRAS, 457, 3420
Goody, R., West, R., Chen, L., & Crisp, D. 1989, JQSRT, 42, 539 Parmentier, V., Fortney, J. J., Showman, A. P., Morley, C., & Marley, M. S.
Grziwa, S., Gandolfi, D., Csizmadia, S., et al. 2016, AJ, 152, 132 2016, ApJ, 828, 22
Guillot, T., & Showman, A. P. 2002, A&A, 385, 156 Parmentier, V., Showman, A. P., & Fortney, J. J. 2021, MNRAS, 501, 78
Helling, C. 2019, AREPS, 47, 583 Parmentier, V., Showman, A. P., & Lian, Y. 2013, A&A, 558, A91
Helling, C., Dehn, M., Woitke, P., & Hauschildt, P. H. 2008, ApJL, 675, L105 Parviainen, H., Deeg, H. J., & Belmonte, J. A. 2013, A&A, 550, A67
Helling, C., Iro, N., Corrales, L., et al. 2019, A&A, 631, A79 Patzer, A., Chang, C., Sedlmayr, E., & Sulzle, D. 2005, EPJD, 32, 329
Helling, C., Kawashima, Y., Graham, V., et al. 2020, A&A, 641, A178 Pérez, F., & Granger, B. E. 2007, CSE, 9, 21
Heng, K., & Demory, B.-O. 2013, ApJ, 777, 100 Powell, D., Louden, T., Kreidberg, L., et al. 2019, ApJ, 887, 170
Heng, K., Morris, B., & Kitzmann, D. 2021, NatAs, 5, 1001 Powell, D., Zhang, X., Gao, P., & Parmentier, V. 2018, ApJ, 860, 18
James, E., Toon, O., & Schubert, G. 1997, Icar, 129, 147 Pruppacher, H. R., & Klett, J. D. 1978, Homogeneous Nucleation (Dordrecht:
Jensen, E. J., & Toon, O. B. 1994, GeoRL, 21, 2019 Springer), 162
Kataria, T., Showman, A. P., Fortney, J. J., et al. 2015, ApJ, 801, 86 Roman, M., & Rauscher, E. 2019, ApJ, 872, 1
Kataria, T., Showman, A. P., Lewis, N. K., et al. 2013, ApJ, 767, 76 Roman, M. T., Kempton, E. M. R., Rauscher, E., et al. 2021, ApJ, 908, 101
Kataria, T., Sing, D. K., Lewis, N. K., et al. 2016, ApJ, 821, 9 Rowe, J., Matthews, J., Seager, S., et al. 2008, ApJ, 689, 1345
Kawashima, Y., & Ikoma, M. 2019, ApJ, 877, 109 Samra, D., Helling, Ch., & Min, M. 2020, A&A, 639, A107
Kipping, D., & Bakos, G. 2011, ApJ, 733, 36 Saumon, D., & Marley, M. S. 2008, ApJ, 689, 1327
Kirkpatrick, J. D. 2005, ARA&A, 43, 195 Seager, S., Whitney, B. A., & Sasselov, D. D. 2000, ApJ, 540, 504
Kluyver, T., Ragan-Kelley, B., Pérez, F., et al. 2016, ELPUB, 87 Showman, A. P., Fortney, J. J., Lian, Y., et al. 2009, ApJ, 699, 564
Lam, S. K., Pitrou, A., & Seibert, S. 2015, in Proc. of the Second Workshop on Showman, A. P., & Kaspi, Y. 2013, ApJ, 776, 85
the LLVM Compiler Infrastructure in HPC. Numba: a LLVM-based Python Showman, A. P., Lewis, N. K., & Fortney, J. J. 2015, ApJ, 801, 95
JIT compiler, 7 (New York, NY, USA: ACM) Showman, A. P., Tan, X., & Parmentier, V. 2020, SSRv, 216, 139
Lee, G., Helling, C., Dobbs-Dixon, I., & Juncher, D. 2015, A&A, Shporer, A., & Hu, R. 2015, AJ, 150, 112
580, A12 Sing, D., Fortney, J., Nikolov, N., et al. 2016, Natur, 529, 59
Lewis, N. K., Showman, A. P., Fortney, J. J., Knutson, H. A., & Marley, M. S. Skemer, A. J., Morley, C. V., Zimmerman, N. T., et al. 2016, ApJ, 817, 166
2014, ApJ, 795, 150 Smith, M. D. 1998, Icar, 132, 176
Lines, S., Mayne, N. J., Boutle, I. A., et al. 2018, A&A, 615, A97 sqlite3 Development Team. 2019, DB-API 2.0 interface for SQLite databases.
Lines, S., Mayne, N. J., Manners, J., et al. 2019, MNRAS, 488, 1332 https://docs.python.org/3.4/library/sqlite3.html
Livingston, J. H., Crossfield, I. J. M., Petigura, E. A., et al. 2018, AJ, 156, 277 Sudarsky, D., Burrows, A., & Hubeny, I. 2003, ApJ, 588, 1121
Lodders, K. 2003, ApJ, 591, 1220 Toon, O., McKay, C., Griffith, C., & Turco, R. 1992, Icar, 95, 24
Lodders, K., & Fegley, B. 2002, Icar, 155, 393 Visscher, C., Lodders, K., & Fegley, B. 2010, ApJ, 716, 1060
MacDonald, R. J., Marley, M. S., Fortney, J. J., & Lewis, N. K. 2018, ApJ, von Paris, P., Gratier, P., Borde, P., & Selsis, F. 2016, A&A, 587, A149
858, 69 Walt, S., Colbert, V. D., Varoquaux, G., et al. 2011, CSE, 13, 22
Marley, M., & Robinson, T. 2015, ARA&A, 53, 279 Webber, M. W., Lewis, N. K., Marley, M., et al. 2015, ApJ, 804, 94
Wolf, E. T., & Toon, O. B. 2010, Sci, 328, 1266

14