The Nature and Origins of Sub‐Neptune Size Planets

Abstract Planets intermediate in size between the Earth and Neptune, and orbiting closer to their host stars than Mercury does the Sun, are the most common type of planet revealed by exoplanet surveys over the last quarter century. Results from NASA's Kepler mission have revealed a bimodality in the radius distribution of these objects, with a relative underabundance of planets between 1.5 and 2.0 R⊕. This bimodality suggests that sub‐Neptunes are mostly rocky planets that were born with primary atmospheres a few percent by mass accreted from the protoplanetary nebula. Planets above the radius gap were able to retain their atmospheres (“gas‐rich super‐Earths”), while planets below the radius gap lost their atmospheres and are stripped cores (“true super‐Earths”). The mechanism that drives atmospheric loss for these planets remains an outstanding question, with photoevaporation and core‐powered mass loss being the prime candidates. As with the mass‐loss mechanism, there are two contenders for the origins of the solids in sub‐Neptune planets: the migration model involves the growth and migration of embryos from beyond the ice line, while the drift model involves inward‐drifting pebbles that coagulate to form planets close‐in. Atmospheric studies have the potential to break degeneracies in interior structure models and place additional constraints on the origins of these planets. However, most atmospheric characterization efforts have been confounded by aerosols. Observations with upcoming facilities are expected to finally reveal the atmospheric compositions of these worlds, which are arguably the first fundamentally new type of planetary object identified from the study of exoplanets.

In this article we review the current state of knowledge of close-in, sub-Neptune size planets. We discuss the early history of the discovery and nomenclature of these objects in Section 2. The most important insights Abstract Planets intermediate in size between the Earth and Neptune, and orbiting closer to their host stars than Mercury does the Sun, are the most common type of planet revealed by exoplanet surveys over the last quarter century. Results from NASA's Kepler mission have revealed a bimodality in the radius distribution of these objects, with a relative underabundance of planets between 1.5 and 2.0  R . This bimodality suggests that sub-Neptunes are mostly rocky planets that were born with primary atmospheres a few percent by mass accreted from the protoplanetary nebula. Planets above the radius gap were able to retain their atmospheres ("gas-rich super-Earths"), while planets below the radius gap lost their atmospheres and are stripped cores ("true super-Earths"). The mechanism that drives atmospheric loss for these planets remains an outstanding question, with photoevaporation and core-powered mass loss being the prime candidates. As with the mass-loss mechanism, there are two contenders for the origins of the solids in sub-Neptune planets: the migration model involves the growth and migration of embryos from beyond the ice line, while the drift model involves inward-drifting pebbles that coagulate to form planets close-in. Atmospheric studies have the potential to break degeneracies in interior structure models and place additional constraints on the origins of these planets. However, most atmospheric characterization efforts have been confounded by aerosols. Observations with upcoming facilities are expected to finally reveal the atmospheric compositions of these worlds, which are arguably the first fundamentally new type of planetary object identified from the study of exoplanets.
Plain Language Summary Planets with radii between that of the Earth and Neptune have been found around other stars in large numbers. It wasn't immediately obvious after their initial discovery what the basic characteristics of these planets are and how they formed because there aren't exact analogs of them in the solar system. Scientists have recently concluded that they are most likely Earth-like in composition based on measurements of how common objects of different sizes and densities in this regime are. However, there are two classes of these objects. The class of slightly larger objects harbors moderately thick atmospheres composed primarily of hydrogen gas. The other class of smaller objects are thought to have been born with similar atmospheres, but lost them during their subsequent evolution. Both classes of these planets must have formed very soon after the formation of their host stars in order to have started with hydrogen-dominated atmospheres, but the exact sequence of events leading to the birth of these objects remains uncertain. Efforts to directly study the atmospheres of these objects have been mostly stymied by heavy cloud layers. Observations with new telescopes are expected to yield detailed information on the atmospheres to further our understanding of these objects. BEAN ET AL. Two views of Kepler's planet radius gap for sub-Neptune size planets. In both cases the slope of the radius gap to smaller instellations and larger orbital periods matches the expectations of atmospheric mass loss being the key discriminant between the larger and smaller populations. Left: Planet occurrence as a function of planet size and instellation with host star radii derived from spectroscopy and distances. Two peaks in the distribution centered at 2.4 and 1.3 R ⊕ are visible in the data. The two lines represent expectations for models of the formation and evolution of these objects from Lopez and Fortney (2013). Figure taken from Fulton et al. (2017). Right: Precisely measured radii for individual planets from stellar characterization via asteroseismology. The lines present the best fit to the radius gap. This figure was originally presented as Figure 5 in Van Eylen et al. (2018). using ground-based telescopes (Johnson et al., 2017), distances from European Space Agency (ESA's) Gaia mission , and asteroseismology using Kepler's time-series photometry (Van Eylen et al., 2018) reduced the uncertainties on planetary radii to 5% or less in the last few years.
The more precise measurements of Kepler planetary radii from improved host star characterization revealed a bimodality in the radius distribution of sub-Neptune planets. Importantly, the gap, or valley, between the two peaks in the distribution has a dependency on orbital period, which has been interpreted as a trend in the incident stellar irradiation received by the planets ("instellation"). The slope of the gap with instellation matches the predictions of models where the two populations of planets both originally formed with hydrogen-dominated envelopes, but the more highly irradiated objects subsequently loss their envelopes (see the dashed line in the left panel of Figure 2). For the Sun-like stars that Kepler primarily observed, gas poor formation models are ruled out because they predict the opposite slope in the radius gap from what is observed (see the dotted line in the left panel of Figure 2). Based on these and other similar results, it is widely held that atmospheric mass loss is the key process that sculpts the population of close-in sub-Neptune size planets orbiting Sun-like stars. However, the hypothesis that the 2-4 R ⊕ planets are predominantly water worlds has not been totally abandoned, and we revisit this hypothesis at the end of this section.
Two main drivers for the mass loss that sculpts the sub-Neptune population have been proposed: "photoevaporation" (Owen & Wu, 2013) and "core-powered mass loss" (Ginzburg et al., 2018). These two models share a similar physical basis: heating of the planet's upper atmosphere drives a hydrodynamic outflow, akin to a Parker wind (Parker, 1958), resulting in mass loss. However, the energy source that provides the heating of the upper atmosphere and drives the outflow differs. In the photoevaporation model, high-energy, ionizing, extreme ultraviolet (XUV) photons (hν∼0.01-1 keV) produced in the stellar corona are absorbed by the planet's upper atmosphere. Due to the destruction of molecular coolants by the ionizing photons, the upper atmosphere it is heated to high temperatures (a few thousand to 10 4 K), driving a hyrodynamic outflow (García Muñoz, 2007;Lammer et al., 2003;Murray-Clay et al., 2009;Owen & Jackson, 2012;Yelle, 2004). In the core-powered mass-loss model, heating of the upper atmosphere from infrared (IR) radiation from the cooling planetary interior and bolometric irradiation from the star similarly drives a hydrodynamic outflow, albeit a cooler and slower one.
It is important to emphasize that the physical processes of XUV and IR/bolometric heating of the planet's upper atmosphere will both happen. However, what is not clear at this stage is which heating mechanism (and therefore whether photoevaporation or core-powered mass loss) dominates the mass loss from sub-Neptune sized planets. While both these heating processes are yet to be included self-consistently in a single model, we can hypothesize about the unified picture and speculate on its limits.
The expected structure of the outflow is summarized in a schematic in Figure 3, where due to the high cross-section for the absorption of XUV photons, they only penetrate into the very upper most layers of the atmosphere. Since the outflow upstream of the sonic surface is not in causal contact with the planetary atmosphere, the position of the sonic surface compared to the penetration depth of XUV photons essentially sets whether core-powered mass loss or photoevaporation dominates. We can imagine as the XUV luminosity increases (or the cooling radiation decreases) we transition from a regime where core-powered mass loss dominates, to one where the sonic surface occurs just inside the XUV heated region. At this point the cooling/bolometrically heated outflow is not thin (compared to the planetary radius) and increases the planet's effective cross-sectional area to the absorption of XUV photons. The increase in absorption of XUV photons could lead to mass-loss rates enhanced above the standard expectation for photoevaporation. Eventually, as the XUV luminosity increases, the upper atmosphere will be entirely XUV dominated and we return to the standard picture of photoevaporation.
Both photovepoaration and core-powered mass loss suggests a unified explanation for the close-in sub-Neptune size planets: they are large (i.e., the mean mass is somewhere in the neighborhood of 3-8 M ⊕ ) terrestrial planets born with hydrogen-dominated atmospheres that are a few percent by mass. Planets below the radius gap were fully stripped of these primordial atmospheres, while planets above the gap held on to their hydrogen-dominated atmospheres. Since the radius at which this transition occurs has been observed, and mass loss is sensitive to the planet's mass (with more massive planets better able to hold onto their natal hydrogen atmospheres), then the position of the radius gap is a probe of the core's composition. More BEAN ET AL.

10.1029/2020JE006639
4 of 20 volatile-rich cores are expected to have a radius gap at larger radii (see Figure 4), and this appears to be ruled out by the data.
Numerous works constrain the core compositions to be volatile poor and consistent with an Earth-like silicate-to-iron ratio (Ginzburg et al., 2018;Gupta & Schlichting, 2019;Owen, 2019;Owen & Wu, 2017;Wu, 2019). This conclusion appears to be robust, and is independent of whether the photoevaporation or core-powered mass-loss model is used. Further, the fact that the radius gap is observed to be a relatively sharp feature indicates that there is not a large spread in the core densities. Recent statistical fits of the radius gap within the framework of the photoevaporation model suggest the mean density of an Earth-mass 5.1 0.4g cm , with a variance in the density of Earth-mass cores at the <1g cm −3 level. This mean core density implies, for a typical core mass of 6 M ⊕ , the water content can be no higher than 20% and this is even in the hypothetical and unlikely case that it's composed of iron and water only, with no silicates (J. G. Rogers & Owen, 2020). While photoevaporation and core-powered mass-loss models agree on the core composition, they diverge on other inferred properties. In particular, the photoevaporation model suggests a positive linear correlation between core mass and stellar mass (Wu, 2019), while core-powered mass loss implies the core mass is independent of stellar mass, although some correlation is not ruled out (Gupta & Schlichting, 2020).
A second key observational result comes from the careful analysis of the masses and radii of ultra-short period planets -see Figure 5 ( Dai et al., 2019). These planets are so highly irradiated that it is very unlikely that they harbor hydrogen-dominated atmospheres, thus eliminating a degree of freedom in the interior structure models. Dai et al. (2019) found that most of the small ultra-short period planets were consistent with an Earth-like terrestrial composition, with the exception of two out of the 11 planet sample that were more or less dense. The low density object, 55 Cnc e (see also a discussion of its atmosphere in Section 5), is the most interesting in the context of this review because its large radius suggests it either has a significant BEAN ET AL.  . Schematic cartoon of the expected outflow structure in a unified picture for hydrodynamic mass loss from close in exoplanets (top) and the three (continuously connected) expected mass-loss regimes (bottom). The top panel shows the three layers to the planetary atmosphere. The bound atmosphere (yellow region) is where the hydrodynamic outflow is sub-dominant. The region heated by cooling radiation from the planetary interior (red photons) and the stellar bolometric luminosity (green photons) has an intermediate temperature (blue/green region). Finally, the region heated by stellar XUV irradiation (blue photons) is a few thousand Kelvin or more (orange region). The mass-loss regimes are shown from left to right as a function of increasing XUV luminosity (or decreasing cooling radiation). Corepowered mass loss occurs when the sonic surface sits interior to the penetration of XUV photons, which thus do not affect the outflow (i). When the sonic point occurs in the XUV heated region, but the cooling/bolometric heated region is not thin, photoevaporation is enhanced due to the larger sub-tended absorption area of the planet to XUV photons (ii), and finally when the cooling/bolmetric region is thin mass loss behaves as 'classic' photoevaporation (iii). Only scenarios (i) and (iii) have been calculated, and only for each in isolation. XUV, extreme ultraviolet. component of low-density volatiles, or it has no core and is predominantly made up of Ca and Al minerals that condense at high temperatures (Dorn et al., 2019). The potential presence of a significant amount of volatiles is in tension with the statistics of the radius gap, which can be fully explained with a population model that has essentially no water-rich planets (J. G. Rogers & Owen, 2020). Fifty five Cnc e is also part of a system that is unusually rich in giant planets (Fischer et al., 2008), so it might not be representative of the broader sub-Neptune planet class.
As mentioned above, the mass-loss hypothesis for the nature and origins of sub-Neptune planets currently has the most traction in the field, but alternative hypotheses have not been fully abandoned. In particular, Zeng et al. (2019) and Mousis et al. (2020) have extended earlier work on water worlds and shown that the population of 2-4 R ⊕ planets can be explained by internal structure models with a large fraction of volatiles. They have shown that roughly 50/50 admixtures of rock and water can reproduce the 2.5 R ⊕ peak in the Kepler radius distribution, with variations in water abundance explaining the range of sizes and densities for these planets instead of variations in the hydrogen fraction. The water world model has not gained as wide acceptance as the mass-loss model in the exoplanet community mainly because it doesn't explain the two key observed correlations with instellation that are described above (i.e., the radius distribution with orbital period and the densities of highly irradiated planets). While it remains true that models for the internal structures of individual planets in the 2-4 R ⊕ regime are degenerate, and thus water-world solutions are possible, the mass-loss hypothesis offers a simple, unified explanation for the trends in the entire population of sub-Neptune planets. Nevertheless, work should continue to test and refine all plausible models.

Implied Formation Pathways
Understanding the origins of systems of close-in sub-Neptunes requires a change in our frame of reference. The majority of planet formation studies to date have focused on our solar system (e.g., Safronov, 1972;Wetherill, 1978). This is not surprising given that exoplanets were only discovered in the past few decades, whereas the origins of the solar system planets have been pondered for centuries. Close-in planetary systems represent fundamentally different outcomes of planet formation that must be more common than the one that produced the solar system. While the same physical processes should govern the formation of all planetary systems, the specific sequence of events must play a key role in shaping their orbital architectures (see discussion in Raymond, Izidoro, & Morbidelli, 2018).
Observational constraints on formation models come from the bulk properties of the observed population of close-in sub-Neptunes. Specific quantities include these planets' sizes, densities and orbital distances (including the correlated planet sizes within individual systems; Ciardi et al., 2013;Wu & Lithwick, 2013;Weiss et al., 2018;Weiss & Petigura, 2020). Given that many sub-Neptunes are found in multiple-planet systems (see Figure 6), other constraints come from the multiplicity distribution (Fang & Margot, 2012;Johansen et al., 2012;Tremaine & Dong, 2012;Youdin, 2011)   . The observed radius distribution of Kepler planets with orbital periods <100 days is shown as the gray histogram. The radius distributions predicted by the photoevaporation for different solid core compositions are shown as the colored lines. Lower density cores predict the radius gap to appear at higher radii. The observed radius distribution implies cores have densities consistent with an Earth-like rock-iron mixture (i.e., a model intermediate between the red and yellow models). More sophisticated models tightly constrain the silicate-to-iron ratio to be ∼3:1, that is consistent with Earth's composition (J. G. Rogers & Owen, 2020). Figure from Owen and Wu (2017).

Silicate
Water Iron Figure 5. Radius versus mass from a uniform analysis of small, highly irradiated planets. These planets should not have substantial gaseous envelopes, thus removing a degree of freedom from interior structure models. The data are tightly clustered around the Earth-like composition line, suggesting a common composition for rocky planets. The higher density outlier K2-229b could have a higher iron fraction from collisions (although note that current models struggle to create very iron-rich planets from collisions; Scora et al., 2020), while the low density outlier 55 Cnc e could be the rare small planet with a significant volatile content or no core. how many planets transit each star), as well as the orbital period ratio distribution of adjacent planets (Fabrycky et al., 2014;Lissauer et al., 2011).
At least seven models exist to explain the origins of close-in sub-Neptunes (Raymond et al., 2008), most of which were proposed prior the launch of the Kepler space telescope. Some of these models (e.g., Fogg & Nelson, 2005;Raymond et al., 2006;Zhou et al., 2005) relied on the dynamical influence of giant planets and can be ruled out on the simple grounds that the measured occurrence rate of super-Earths is far higher than that of gas giants (Fressin et al., 2013;Howard et al, 2010Howard et al, , 2012Mayor et al., 2011). The very simplest model-often called "in situ accretion"-proposed that close-in planets grew in the same way as our own terrestrial planets, by successive impacts between ever-larger planetary embryos within disks that were massive enough to have many Earth masses of solids very close to their stars (Chiang & Laughlin, 2013). That model can also be ruled out because it assumes that the planets grew in place, close to their current orbital distances. As such, the model is not self-consistent: any disk massive enough to form such planets in situ would drive orbital migration at such a fast rate as to make migration a central process (Bolmont et al., 2014;Inamdar & Schlichting, 2015;Ogihara et al., 2015).
There are currently two plausible models to explain the origins of close-in small planets (see Figure 7). Both invoke large-scale inward movement of solids within gas-dominated planet-forming disks but at very different size scales. In the drift model the majority of growth takes place close in, from mass that has drifted inward. In the current paradigm of planet formation, dust grains coagulate and grow until they become large enough to partially decouple from the gas and drift inward (see Johansen & Lambrechts, 2017;Ormel et al., 2017). Dust particles that grow big enough to drift rapidly are often referred to as "pebbles." Observations of gas-rich disks around young stars BEAN ET AL.  commonly find evidence for the existence of pebbles (Natta et al., 2007;Pérez et al., 2015), and it is inferred that they drift inward because dust disks are observed to be more compact than gas disks (Andrews et al., 2012;Cleeves et al., 2016;Trapman et al., 2020). The ring-like structures observed in many disks (ALMA Partnership et al., 2015;Andrews et al., 2018) are thought to be produced by growing and drifting dust/ pebbles (e.g., Dullemond et al., 2018). Drifting pebbles may be trapped at a pressure bump in the inner parts of the disk (Boley et al., 2014;Chatterjee & Tan, 2014, 2015X. Hu et al., 2018X. Hu et al., , 2016Jankovic et al., 2019). Indeed, MHD simulations of the inner regions of disks find that pressure bumps should exist close in and are capable of trapping drifting particles (Flock et al., 2017(Flock et al., , 2019. The next phases of growth are thought to involve gravitational instability to form planetary embryos, followed by mutual collisions and orbital migration (e.g., see Dawson et al., 2015;Flock et al., 2019;Hansen & Murray, 2012Moriarty & Ballard, 2016).
In the migration scenario mass is delivered to the inner disk in the form of large planetary cores rather than drifting pebbles. Cores are assumed to form across the disk by planetesimal and pebble accretion (e.g., Johansen & Lambrechts, 2017). Massive cores likely form preferentially at or past the snow line, where pebble accretion is accelerated (Lambrechts & Johansen, 2014;Morbidelli et al., 2015;Ormel et al., 2017). Once they reach a critical mass, cores migrate inward until they reach the inner edge of the disk (e.g., Cossou et al., 2014;Coleman & Nelson, 2016;Ida & Lin, 2010;McNeil & Nelson, 2010;Ogihara & Ida, 2009;L. A. Rogers et al., 2011;Terquem & Papaloizou, 2007). As in the drift model, the final phases of growth involve giant collisions between cores. It is worth noting that our understanding of migration is incomplete, as even the highest resolution simulations to date cannot fully resolve the behavior of low-mass planets in low-viscosity disks (McNally et al., 2019).
In each scenario, the growing planets accrete gaseous envelopes directly from the disk. The structure of these primordial atmospheres is determined by a complex competition between gas flow within the disk, thermal evolution, and loss during impacts (Coleman et al., 2017;Ginzburg et al., 2016;Ikoma & Hori, 2012;Inamdar & Schlichting, 2016;Lambrechts et al., 2019;Lambrechts & Lega, 2017;Lee & Chiang, 2016;Lee et al., 2014;Schlichting, 2014). Once the gas disk has dissipated, these atmospheres are subject to loss processes (see Section 3).
The drift and migration models predict different compositions for close-in planets. Pebbles should lose their volatiles as they drift inward across the snow line (which is itself moving inward as the disk evolves; e.g., Ida et al., 2019;Oka et al., 2011) such that the planets formed in the drift model should be rocky, with little water. In contrast, large migrating cores should retain the bulk of their volatiles. But exactly how water-rich should planets be in the migration scenario? Unfortunately, this is currently unclear. If planetesimals-the seeds of super-Earths-preferentially form just past the snow line (Armitage et al., 2016;Drażkowska & Alibert, 2017; and grow further by pebble accretion (while migrating), then their bulk water contents are often a few to 10% (according to simulations; Bitsch et al., 2019). If, however, planetesimals form quickly across a broad swath of the disk then the final super-Earths are likely to be closer to 50% water by mass ). Yet large migrating cores shepherd material interior to their orbits and catalyze the formation of even closer-in planets, which themselves are often volatile-poor (Izidoro et al., 2014;Raymond, Boulet, et al., 2018). In simple terms, while the drift model predicts high-density volatile-poor planets, the migration model predicts a diversity of volatile contents of such planets, often within the same system. Very high water contents are at odds with the compositional inferences discussed above within the context of evaporative mass loss of the atmospheres of close-in planets (see Section 3). Future high-precision measurements of the bulk densities of close-in planets, coupled with interior structure models, may distinguish between these model predictions.
The late dynamical evolution of the drift and migration models should be similar (see discussion in Raymond, Izidoro, & Morbidelli, 2018). In each model, massive planets form quickly and are close to the inner parts of the disk while the disk is still dense. This implies that migration must be important during the later parts of gaseous disk phase. Migrating cores tend to form configurations in which each pair of neighboring planets is in mean motion resonance (e.g., Cresswell & Nelson, 2008;Terquem & Papaloizou, 2007). In these "resonant chains," the innermost planet is anchored at the inner edge of the gaseous disk, which provides a positive torque to balance the negative ones felt by the other planets (Masset et al., 2006;Romanova et al., 2019). Resonant chains often become dynamically unstable after the gaseous disk dissipates, leading to phase of late giant impacts Ogihara & Ida, 2009;Terquem & Papaloizou, 2007). This is the foundation of the breaking the chains model, which can match the period ratio and multiplicity distributions of Kepler's close-in planets as long as 95% or more of resonant chains become unstable (Izidoro et al., 2017. In this context, exotic multi-resonant systems such as TRAPPIST-1 (Gillon et al., 2017) and Kepler-223 (Mills et al., 2016) represent the rare resonant chains that remained stable after the disk dissipated. While the late stages of growth and migration of the drift model have not yet been modeled, we expect them to follow the breaking the chains pathway.
The distribution of sub-Neptunes' H/He atmospheres place constraints on formation models. For example, envelopes containing a few percent of a planet's mass are needed to explain the radius valley (see Section 3), yet it remains unclear why planets would not accrete substantially more gas from the disk. Several processes have been proposed to explain this, including delayed atmospheric cooling due to high opacity (Lee et al., 2014), dissolution of H 2 in magma oceans (Kite et al., 2019), and rapid disk photoevaporation (Ginzburg et al., 2016;Ogihara et al., 2020;Owen & Wu, 2016). Giant impacts between growing sub-Neptunes are likely to lead to loss of the planets' primordial H/He atmospheres, especially for close-in young planets (Biersteker & Schlichting, 2019). Yet if impacts are generic and systematically remove H/He envelopes, why do sub-Neptunes exist at all? The answer to this apparent contradiction is not immediately clear. Perhaps impacts often occur before the full dissipation of the gaseous disk (Esteves et al., 2020) such that a thin atmosphere can still be accreted . Or perhaps resonant chains of planets are spread out by a different mechanism, such as by the magnetospheric rebound migration torque from the expanding disk cavity . We expect future formation models to take advantage of detailed compositions constraints from studies of sub-Neptune atmospheres (see Section 6).
How does the solar system fit into this picture? Why are there no super-Earths or sub-Neptunes close to the Sun? A number of solutions to this very relevant problem have been proposed (for a detailed discussion, see Raymond, Izidoro, & Morbidelli, 2018). Some models propose that close-in planets did indeed form around the Sun but did not survive. However, these scenarios are hard to reconcile with observations. For example, if our Sun's close-in planets were collisionally ground to dust (Volk & Gladman, 2015), why are such planets so common around other stars and why are their orbital spacings suggestive of a late phase of giant impacts (Izidoro et al., 2017;Pu & Wu, 2015)? Likewise, if planets formed close to the Sun but migrated away, either outward toward the giant planet region (Raymond et al., 2016), or inward onto the Sun (Batygin & Laughlin, 2015), then how can we reconcile this with the abundance of close-in exoplanets?
At present it seems more likely that some mechanism prevented the formation of close-in super-Earths or sub-Neptunes around the Sun. Perhaps Jupiter's growing core reached the pebble isolation mass (Bitsch et al., 2018;Lambrechts & Johansen, 2014) and starved the inner solar system of inward-drifting pebbles, thus preventing the terrestrial planets from growing massive enough to migrate . This mechanism could explain the isotopic dichotomy of chondritic meteorites (Warren, 2011), which appears to require spatial segregation of pebbles in the early solar system due to Jupiter's growing core (Kruijer et al., 2017(Kruijer et al., , 2020 or perhaps pressure bumps in the disk (Brasser & Mojzsis, 2020). However, if the giant planets' cores grew fast enough to block the pebble flux into the inner solar system, why didn't they migrate farther inward themselves? Another possibility is that the full-grown Jupiter and Saturn blocked the inward migration of the progenitors of the ice giants . If that were the case, then we would expect to observe an anti-correlation between close-in planets and outer gas giants; that correlation is not presently observed (Barbato et al., 2018;Bryan et al., 2019;Zhu & Wu, 2018). Thus, while it remains an active area of research, it is unclear exactly which pieces of the puzzle are responsible for the lack of close-in large planets in the solar system.

A Decade of Super-Earth Atmosphere Studies
It has been recognized since the early days of sub-Neptune size exoplanet discovery that understanding their atmospheres holds a key to revealing the nature and origins of these mysterious objects. One reason for this is that determination of the atmospheric composition could help break the degeneracies in interior structure models and uniquely constrain their bulk makeup (E. R. Adams et al., 2008;Miller-Ricci et al., 2009; L. A. Rogers & Seager, 2010a;Valencia et al., 2013). If the atmospheric compositions of these planets could be directly determined that would provide an important boundary condition for the models used to match the planets' masses and/or radii.
Another reason for the importance of these planets' atmosphere is that the composition of the atmosphere itself is also an important record of a planet's formation and evolution (Benneke & Seager, 2013;Miller-Ricci & Fortney, 2010;L. A. Rogers & Seager, 2010b). For example, both primary and secondary type atmospheres are possible for these planets, and these different types of envelopes can be distinguished by their composition. Furthermore, even primary atmospheres could be altered by interaction with the planet's interior, thus offering a potential view to the detailed composition of the bulk (Kite et al., 2019(Kite et al., , 2020. There are still very few direct observational constraints on the compositions of sub-Neptune size exoplanet atmospheres despite considerable effort over the last decade. Most efforts to date have focused on transmission spectroscopy observations because this is in principle the most efficient way to detect the atmospheres of these objects with existing facilities (as opposed to measurements of thermal emission or reflected light via secondary eclipses and phase curves). Transmission spectroscopy is also particularly well suited to addressing the key question of the hydrogen content of these planets' atmospheres because the size of spectral features in transmission spectroscopy measurements is primarily sensitive to the scale height of the atmosphere, which itself is mainly set by the abundance of hydrogen gas through its impact on the mean molecular weight (Miller-Ricci et al., 2009).
Unfortunately, the vast majority of transmission spectroscopy measurements for sub-Neptune sized planets have yielded featureless, or so-called "flat" spectra (e.g., Bean et al., 2011Bean et al., , 2010Berta et al., 2012;Désert et al., 2011;Diamond-Lowe et al., 2018;Fraine et al., 2013;Guo et al., 2020;Knutson et al., 2014;Kreidberg et al., 2014;Libby-Roberts et al., 2020). The featureless spectra for the planets that must have gaseous envelopes (i.e., those with ) can be explained by the presence of thick aerosols at high altitude obscuring our view of the bulk of the atmospheres, and thus they generally can't constrain the compositions of the planets' atmospheres. Aerosols are a particularly pernicious problem for super-Earths because these objects are already hard to observe due to their small size (reminder: signal sizes in transmission spectroscopy scale as 3 p R all else being equal), and because these planets are typically cooler, which enhances aerosol formation (R. Hu & Seager, 2014;Kawashima & Ikoma, 2019;Mbarek & Kempton, 2016;Miller-Ricci Kempton et al., 2012;Morley et al., 2013Morley et al., , 2015). The featureless transmission spectra for planets with are consistent with these planets lacking cloudless, hydrogen-dominated atmospheres, but aren't constraining beyond that.
There have been two notable successes in detecting features in the transmission spectra of sub-Neptune sized planets, both from Hubble Space Telescope Wide Field Camera 3 (WFC3) observations. Tsiaras et al. (2016) presented the detection of relatively large features in the transmission spectrum of 55 Cnc e (R p = 1.9 R ⊕ ), and Benneke et al. (2019) and Tsiaras et al. (2019) present a much more convincing detection of features in the spectrum of K2-18b (R p = 2.6 R ⊕ ). Both of these detections imply hydrogen-dominated atmospheres but with large uncertainties in the overall heavy element abundance (metallicity) due to modeling degeneracies and the limited information content of the data. The presence of a hydrogen-dominated atmosphere on K2-18b (see Figure 8) is consistent with the interpretation of the planet radius gap in the population statistics described above. However, the idea of a similar atmosphere on 55 Cnc e is harder to reconcile given its very high level of irradiation (it has an instellation S ∼ 2500 S ⊕ ). The 55 Cnc e data are also suspect because the host star is right at the brightness limit for WFC3, thus raising the possibility of unmitigated instrument systematics (Hilbert, 2014;Swain et al., 2013;Wilkins et al., 2014). In contrast, Jindal et al. (2020) ruled out the presence of a hydrogen-dominated atmosphere containing water vapor for this planet using groundbased high-resolution spectroscopy.
55 Cnc e is also one of the few sub-Neptune size planets that is amenable to thermal emission measurements. The Spitzer Space Telescope full orbit phase curve for this planet has been shown to have a large amplitude, which is indicative of poor day-night heat redistribution (the planet is likely tidally locked), but also a large hot spot offset, which is indicative of substantial heat transport (B. O. Demory et al., 2012;Demory, Gillon, de Wit, et al., 2016). Surprisingly, the dayside thermal emission of the planet also shows variability, thus making the observations even more difficult to interpret .
Ultimately, 55 Cnc was also at the brightness limit for the now defunct Spitzer, and unrecognized instrument systematics could have impacted these measurements as well.
Beyond the attempts to directly observe the bulk atmospheres of sub-Neptune size planets, there have also been clever attempts to deduce the composition of these planets' atmospheres through observations of their thermospheres, exospheres, and winds. These observations have been obtained for neutral hydrogen at Lyman α (Bourrier et al., 2017;dos Santos et al., 2020;Ehrenreich et al., 2012;García Muñoz et al., 2020;Waalkes et al., 2019) and in the helium IR triplet (Kasper et al., 2020). The detection of neutral hydrogen in an escaping wind would constrain the mass-loss rate of the atmosphere but would have an ambiguous interpretation with regards to the composition because neutral hydrogen could be produced from the photodissociation of H 2 and H 2 O. Helium is a more promising species from this standpoint because it would only exist in large quantities for a primary atmosphere that was accreted from the protoplanetary nebula. However, the most observable helium feature, the IR triplet, arises from a metastable state that requires a finely tuned spectrum of UV irradiation from the host star to be populated (Oklopčić, 2019).
Unfortunately, no clear detections of the upper atmospheres of super-Earths have been made so far. The most promising result is the tentative detection of neutral hydrogen for K2-18b based on a partial transit observed with Hubble (dos Santos et al., 2020). This is consistent with the observation of the bulk atmosphere described above, but more data are needed to confirm the result.

Future Directions
Approximately 15 years since their first discovery, the nature and origins of sub-Neptune size exoplanets have started to come into focus. The global picture that has recently emerged from population level studies is that most close-in, sub-Neptune size planets are actually large terrestrial bodies, with the absence ("true super-Earths") or presence ("gas-rich super-Earths") of hydrogen-dominated atmospheres separating them into two classes. These objects are likely poor in volatiles (≲10% by mass), and their final assembly occurred close to their host stars in the presence of a gas-rich disk. It has been tempting to compare these objects to the solar system ice giants Uranus and Neptune because they have hydrogen-dominated atmospheres as a common factor (e.g., Atreya et al., 2020;Wakeford & Dalba, 2020). However, the more we learn about these objects the less appropriate this comparison is. Uranus and Neptune have roughly 10 times more hydrogen BEAN ET AL.
10.1029/2020JE006639 11 of 20 and helium by mass than the typical gas-rich super-Earth, their bulk and envelopes are likely rich in volatiles, and their formation histories must be quite different to have arrived at very different orbital distances.
The distinct internal structures of gas-rich super-Earths, that is, rock overlaid by thick, hydrogen-dominated atmospheres, leads us to propose that these objects are the first fundamentally new type of planetary object identified from the study of exoplanets. There are a number of observations that can be done to test this hypothesis. One ongoing area of work is the precise measurement of masses and radii for sub-Neptune size planets orbiting stars with a range of masses and ages, and with a wide range of orbital separations. These observations should seek to determine how the planet radius gap varies with stellar mass (Cloutier & Menou, 2020;Hardegree-Ullman et al., 2020) and age (Berger et al., 2020), and to ultimately reveal the statistical distribution of planet densities in the multi-dimensional parameter space. Early results on this topic from further analysis of Kepler/K2 data have yielded tentative evidence that super-Earths form in gas poor disks around low-mass stars (Cloutier & Menou, 2020), and that the mass-loss timescale for these planets around stars with masses ≳1 M ⊙ is approximately a Gyr, which is a potential signpost to the core-powered mechanism (Berger et al., 2020). Continuing work on this topic is currently enabled through the detection of transiting planets around bright stars by NASA's TESS mission (launched 2018; Ricker et al., 2015) and ESA's CHEOPS mission (launched 2019;Benz et al., 2020), and will be furthered by ESA's PLATO mission (scheduled for launch in 2026; Rauer et al., 2014).
Another key observation that can be done for sub-Neptune size planets is precise spectroscopy to reveal their atmospheric compositions. While such observations have been mostly stymied so far, the increased sensitivity and spectral range of the James Webb Space Telescope (Beichman et al., 2014;Greene et al., 2016) and the next generation of ground-based Extremely Large Telescopes (Gandhi et al., 2020;Hood et al., 2020) are expected deliver breakthroughs on this topic. Spectroscopy of gas-rich super-Earths should seek to determine if the metallicities of their atmospheres follow the trend of increasing metallicity with lower planet mass that is expected from extrapolating from giant planet formation . These observations may also reveal atmospheric carbon-to-oxygen abundance ratios, which are a tracer of formation location and migration Öberg et al., 2011). Gas-rich super-Earths are expected to have deep magma oceans in contact with their atmospheres, thus yielding unique chemistry in atmospheric gases that could be detectable (Kite et al., 2020), as well as potentially sculpting the populations statistics at large sizes (Kite et al., 2019).
On the theoretical side, work combining models of photoevaporation and core-powered mass loss into a unified picture of hydrodynamic escape is necessary. This modeling should help identify the regions of parameter space that each mass-loss mechanism dominates. Further, observations of atmospheric escape for the emerging class of very young planets that are the likely antecedents of mature sub-Neptune size planets (David, Cody, et al., 2019;David, Petigura, et al., 2019;David et al., 2016;Newton et al., 2019;Plavchan et al., 2020;Rizzuto et al., 2020) offer the hope of distinguishing between the photoevaporative and core-powered atmospheric loss mechanisms. Ultimately, our quantitative insights into how these planets formed, such as the core-mass function and how much H/He these planets accreted, depend strongly on the assumed mass-loss model.
The main uncertainty in our understanding of the formation of sub-Neptune systems is where large cores (planetary embryos) form. Do they originate past the snow line and undergo large-scale migration, or very close to their stars and only migrate to a limited extent? Future advances will likely be aided by a better understanding of the bulk compositions of close-in planets, in particular their volatile contents (e.g., Gupta & Schlichting, 2019; J. G. Rogers & Owen, 2020). Improved observations and models of the structure and evolution of planet-forming disks will also play a role, as the disk determines how fast pebbles drift, where and when they accumulate to form planetesimals (Drażkowska & Alibert, 2017), and how fast and in what direction growing planets migrate .
Putting our Solar System-and its lack of close-in super-Earths or sub-Neptunes-in the context of extrasolar planets is a challenge (for a discussion, see Raymond, Boulet, et al., 2018). Jupiter is the only Solar System planet that would be detectable if the Sun were observed with present-day technology. Understanding where our system fits within the bigger picture may thus rest on demographic studies that correlate the nature of inner and outer parts of planetary systems, including super-Earths and sub-Neptunes, Jupiter-like gas giants, ice giant analogs, and even debris disks (Barbato et al., 2018;Bryan et al., 2019;Clanton & Gaudi, 2016;Moro-Martín et al., 2015;Raymond et al., 2011;Suzuki et al., 2016;Zhu & Wu, 2018). Fortunately, we are in a golden era of extrasolar planetary astronomy where the observational tools needed for these studies are rapidly advancing. The next 15 years are sure to bring dramatic surprises and insights to match those of the first 15 years of sub-Neptune planet discovery and characterization.

Data Availability Statement
The data underling the previously published figures (Figures 1, 2, 4, 5, and 8) are available in the corresponding publications. A comprehensive database of exoplanet parameters can be found at the NASA Exoplanet Archive (https://exoplanetarchive.ipac.caltech.edu). Data from this archive was used to create Acknowledgments JLB acknowledges generous support over the years from NASA, the NSF, the David and Lucile Packard Foundation, the Heising-Simons Foundation, and the Sloan Foundation. SNR thanks the PNP program of the CNRS as well as the Agence Nationale pour la Recherche for funding of many of the ideas presented here (grant ANR-13-BS05-0003-002) and is grateful to all his colleagues involved in the MOJO project. JEO is supported by a Royal Society University Research Fellowship and a 2019 ERC starting grant (PEVAP).