K2-18b: Temperate Sub-Neptune Overview

• K2-18b is a transiting temperate sub-Neptune orbiting an M-dwarf, characterized by Earth-like irradiation and ambiguous atmospheric composition.
• Observations from HST and JWST reveal a hydrogen-dominated atmosphere with water, methane, and complex cloud/haze effects that challenge unique molecular interpretations.
• Mass, radius, and density measurements suggest a volatile-rich interior with models supporting either a water ocean layer, mini-Neptune envelope, or magma ocean scenario.

K2-18b is a transiting temperate sub-Neptune orbiting the nearby M-dwarf K2-18 on a 32.94-day orbit and receiving stellar irradiation similar to Earth. It has become a focal point for atmospheric characterization and for assessing the plausibility of “Hycean” atmospheres, because its mass, radius, and transmission spectrum admit multiple physically distinct interpretations: a volatile-rich mini-Neptune, a hydrogen-rich planet overlying a liquid water ocean, or a world whose observable atmosphere is shaped by deeper non-habitable environments such as a magma ocean. Its observational history also intersects debates over cloud formation, methane and carbon-dioxide chemistry, atmospheric escape under M-dwarf irradiation, and the statistical robustness of proposed mid-infrared biosignature-like features (Montet et al., 2015, Wogan et al., 2024).

1. Discovery and system characterization

K2-18b was first identified in K2 Campaign 1 from two transit events separated by about 33 days. Statistical validation using imaging constraints, light-curve morphology, and the vespa false-positive framework yielded $\mathrm{FPP} < 10^{-4}$, establishing the planet as a validated transiting world receiving Earth-like insolation (Montet et al., 2015). A subsequent Spitzer transit at $4.5\,\mu\mathrm{m}$ confirmed that the K2 events were periodic, ruled out the alternative scenario of two long-period planets each transiting once, and repaired a compromised ephemeris after a previously undetected cosmic-ray anomaly in the K2 photometry had shifted the predicted transit time by 1.85 hours (Benneke et al., 2016).

Radial-velocity follow-up with CARMENES, and then joint analysis with HARPS, established the planet’s mass scale and mildly eccentric orbit. Reported solutions include $K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$, $M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$, $\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$, and $e_b = 0.20 \pm 0.08$ (Sarkis et al., 2018). The same study argued that a previously proposed $\sim 9$-day signal was most plausibly stellar activity rather than a second planet, because it was time- and wavelength-dependent and aligned with activity diagnostics rather than an achromatic Keplerian signal (Sarkis et al., 2018).

The host star has been characterized as an M2.5 V, M2.8 dwarf, and M3 dwarf in different analyses, reflecting distinct stellar pipelines and calibrations. A widely used parameter set gives $T_* = 3457 \pm 39\,\mathrm{K}$, $R_* = 0.411 \pm 0.038\,R_\odot$, $M_* = 0.359 \pm 0.047\,M_\odot$, and near-infrared brightness $4.5\,\mu\mathrm{m}$0 mag, $4.5\,\mu\mathrm{m}$1 mag, making the system especially favorable for transmission spectroscopy (Benneke et al., 2016, Sarkis et al., 2018).

2. Bulk properties and internal structure

Published radius and mass estimates place K2-18b firmly in the super-Earth/sub-Neptune or mini-Neptune regime. Representative radius estimates include $4.5\,\mu\mathrm{m}$2 from the initial K2 validation, $4.5\,\mu\mathrm{m}$3 from the HST-era atmospheric analysis, and $4.5\,\mu\mathrm{m}$4 in the Lyman-$4.5\,\mu\mathrm{m}$5 escape study; corresponding mass estimates cluster near $4.5\,\mu\mathrm{m}$6–$4.5\,\mu\mathrm{m}$7 (Montet et al., 2015, Benneke et al., 2019, Santos et al., 2020). Density estimates therefore span substantially different values depending on the adopted stellar radius and mass calibration, including $4.5\,\mu\mathrm{m}$8 and $4.5\,\mu\mathrm{m}$9 (Benneke et al., 2019, Sarkis et al., 2018). This spread does not remove the central inference that low-density volatiles are required.

Interior modeling based on the revised bulk parameters and the transmission spectrum constrains the atmosphere to be H$K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$0-rich with an H$K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$1O volume mixing ratio of $K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$2–$K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$3, while CH$K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$4 and NH$K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$5 are depleted relative to equilibrium expectations and clouds or hazes are not conclusively detected in that framework (Madhusudhan et al., 2020). The same study finds that the H/He envelope mass fraction is $K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$6, spanning $K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$7 for a predominantly water world to $K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$8 for a pure iron interior, and that the thermodynamic conditions at the surface of the H$K_b = 3.55^{+0.57}_{-0.58}\,\mathrm{m\,s^{-1}}$9O layer range from the super-critical to liquid phases (Madhusudhan et al., 2020). In that sense, K2-18b is not constrained to a single internal architecture by current bulk data alone.

A later density reanalysis based on revised stellar parameters derived from HARPS spectra obtained $M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$0, again supporting an H$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$1-dominated mini-Neptune atmosphere rather than a compact rocky planet (Liu et al., 13 Sep 2025). This suggests that the main unresolved question is not whether the planet contains substantial volatiles, but how those volatiles are partitioned among the deep interior, any condensed layers, and the observable atmosphere.

3. Transmission spectroscopy and molecular interpretation

HST/WFC3 observations established the first detailed atmospheric constraints. Using eight spectroscopic transit visits in the final analysis, the HST-era retrieval found a prominent 1.4-$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$2m feature, a Bayes factor of 459:1 relative to a flat spectrum, $M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$3, a cloud-top pressure between 7.74 and 139 mbar, and $M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$4, supporting a hydrogen-dominated atmosphere with water vapor and clouds (Benneke et al., 2019).

That interpretation was immediately qualified by self-consistent forward modeling. For cool H$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$5/He sub-Neptunes, the 1.4-$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$6m band is not uniquely diagnostic of H$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$7O. Exo-REM calculations showed that for K2-18b’s atmospheric conditions CH$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$8 is expected to be abundant, that CH$M_b = 8.92^{+1.70}_{-1.60}\,M_\oplus$9-only spectra are nearly indistinguishable from the full model across WFC3, and that H$\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$0O dominates over CH$\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$1 at 1.4 $\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$2m only at larger temperatures; in self-consistent calculations, water overtakes methane in the 1.335–1.415 $\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$3m band for $\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$4 (Bézard et al., 2020). A parallel 1D Exo-REM study favored atmospheric metallicities between $\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$5 and $\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$6 solar, confirmed that CH$\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$7 absorption features nominally dominate the HST spectral range, and found H$\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$8O-ice clouds but not liquid H$\rho_b = 4.11^{+1.72}_{-1.18}\,\mathrm{g\,cm^{-3}}$9O clouds under favored parameter regimes (Blain et al., 2020).

JWST extended the spectral baseline into the 0.7–12 $e_b = 0.20 \pm 0.08$0m range and shifted the discussion from H$e_b = 0.20 \pm 0.08$1O alone to CH$e_b = 0.20 \pm 0.08$2, CO$e_b = 0.20 \pm 0.08$3, and possible sulfur-bearing molecules. A mini-Neptune interpretation of the JWST data found that a gas-rich atmosphere with $e_b = 0.20 \pm 0.08$4 solar metallicity should have $e_b = 0.20 \pm 0.08$5 CH$e_b = 0.20 \pm 0.08$6 and nearly $e_b = 0.20 \pm 0.08$7 CO$e_b = 0.20 \pm 0.08$8, whereas a lifeless Hycean atmosphere under the same observational constraints supports $e_b = 0.20 \pm 0.08$9 part-per-million CH$\sim 9$0 (Wogan et al., 2024). Independent reanalysis of the full 0.7–12 $\sim 9$1m spectrum confirmed CH$\sim 9$2 and CO$\sim 9$3 and found that the tentative presence of DMS and C$\sim 9$4H$\sim 9$5 is interchangeable in combined-spectrum retrievals, while MIRI-only inferences are highly sensitive to reduction choices (Stevenson et al., 8 Aug 2025).

The proposed mid-infrared DMS/DMDS features remain controversial. A model-agnostic Gaussian-feature analysis of the published MIRI/LRS transmission spectrum found that five of six nested tests preferred a flat spectrum with $\sim 9$6, and that only a two-Gaussian model with centroids fixed at 7 and 8.8 $\sim 9$7m yielded weak evidence over a flat line, with $\sim 9$8 and $\sim 9$9 (Taylor, 22 Apr 2025). A broader independent reduction study concluded that the MIRI transit spectrum is highly susceptible to unresolved instrumental systematics, that 87.5% of retrievals using the favored MIRI binning scheme do not support DMS/DMDS, and that there is no statistically significant evidence for biosignatures in the atmosphere of K2-18b (Stevenson et al., 8 Aug 2025).

4. Clouds, hazes, and atmospheric dynamics

Cloud and haze physics are central to the interpretation of K2-18b’s spectrum. Three-dimensional general-circulation modeling of an H$T_* = 3457 \pm 39\,\mathrm{K}$0-dominated atmosphere showed that, under synchronous rotation, the upper atmosphere is governed by a symmetric day-to-night circulation with cloud formation preferentially at the substellar point or at the terminator. In that framework, water clouds form only for metallicity $T_* = 3457 \pm 39\,\mathrm{K}$1 solar, the cloud fraction at the terminators is small for $T_* = 3457 \pm 39\,\mathrm{K}$2–$T_* = 3457 \pm 39\,\mathrm{K}$3 solar metallicity, and very thick clouds form at the terminator for $T_* = 3457 \pm 39\,\mathrm{K}$4 solar metallicity (Charnay et al., 2020). The same study found that the cloud fraction at the terminator can be highly variable, implying potential variability in transit spectra (Charnay et al., 2020).

One-dimensional Exo-REM work reaches a related but more restrictive conclusion about condensates. H$T_* = 3457 \pm 39\,\mathrm{K}$5O-ice clouds can form for sufficiently high metallicity, but liquid H$T_* = 3457 \pm 39\,\mathrm{K}$6O clouds form only if irradiation drops below about $T_* = 3457 \pm 39\,\mathrm{K}$7 of nominal, slightly below the $T_* = 3457 \pm 39\,\mathrm{K}$8 lower bound considered there, and such cases do not fit the Benneke et al. dataset within $T_* = 3457 \pm 39\,\mathrm{K}$9 (Blain et al., 2020). This makes “water clouds” on K2-18b highly model-dependent: retrieval language based on HST favored water vapor and likely clouds, while self-consistent cloud microphysics in H$R_* = 0.411 \pm 0.038\,R_\odot$0-rich atmospheres tends to place condensates in the ice regime rather than the liquid regime (Benneke et al., 2019, Blain et al., 2020).

Hydrocarbon aerosols provide an alternative continuum source. A joint analysis of NIRISS, NIRSpec, and an independently reduced MIRI/LRS spectrum argued for hydrocarbon hazes across 0.85–12 $R_* = 0.411 \pm 0.038\,R_\odot$1m, an H$R_* = 0.411 \pm 0.038\,R_\odot$2-dominated atmosphere with mean molecular weight $R_* = 0.411 \pm 0.038\,R_\odot$3 Daltons, and no need for instrumental offsets between JWST instruments (Liu et al., 13 Sep 2025). In those haze-inclusive retrievals, CH$R_* = 0.411 \pm 0.038\,R_\odot$4 and CO$R_* = 0.411 \pm 0.038\,R_\odot$5 abundances are systematically lower than in haze-free studies, which suggests that haze can reduce the need for high-$R_* = 0.411 \pm 0.038\,R_\odot$6 solutions and that aerosol opacity is a first-order degeneracy in the interpretation of the planet’s chemistry (Liu et al., 13 Sep 2025).

5. High-energy environment and atmospheric escape

K2-18b is also a benchmark for atmospheric escape in the temperate M-dwarf regime. HST/STIS Lyman-$R_* = 0.411 \pm 0.038\,R_\odot$7 transit spectroscopy found that the average blueshifted stellar emission decreased by $R_* = 0.411 \pm 0.038\,R_\odot$8 during transit relative to the pre-transit level, with the final in-transit orbit reaching $R_* = 0.411 \pm 0.038\,R_\odot$9 absorption, while the red wing changed by only $M_* = 0.359 \pm 0.047\,M_\odot$0 (Santos et al., 2020). Because the line core is absorbed by the interstellar medium, the signal was identified in the wings over $M_* = 0.359 \pm 0.047\,M_\odot$1 and $M_* = 0.359 \pm 0.047\,M_\odot$2, and was interpreted as tentative evidence for neutral hydrogen atoms escaping vigorously and being blown away by radiation pressure (Santos et al., 2020).

Reconstruction of the intrinsic stellar Lyman-$M_* = 0.359 \pm 0.047\,M_\odot$3 profile gave $M_* = 0.359 \pm 0.047\,M_\odot$4–$M_* = 0.359 \pm 0.047\,M_\odot$5 at the planet, a central value of $M_* = 0.359 \pm 0.047\,M_\odot$6, a photoionization rate of $M_* = 0.359 \pm 0.047\,M_\odot$7, and a neutral lifetime of $M_* = 0.359 \pm 0.047\,M_\odot$8 hours (Santos et al., 2020). The same analysis inferred $M_* = 0.359 \pm 0.047\,M_\odot$9 for the ratio of radiation-pressure acceleration to stellar gravitational acceleration and an energy-limited escape estimate of $4.5\,\mu\mathrm{m}$00 at 100% efficiency, implying that the planet would lose less than about $4.5\,\mu\mathrm{m}$01 of its mass over its remaining lifetime (Santos et al., 2020). The authors explicitly emphasized that the detection was tentative because it relied on one partial transit, low S/N, and possible stellar variability (Santos et al., 2020).

Later X-ray observations place the present-day high-energy forcing in a relatively quiet regime. XMM-Newton detected K2-18 as a very faint X-ray source with $4.5\,\mu\mathrm{m}$02, $4.5\,\mu\mathrm{m}$03–$4.5\,\mu\mathrm{m}$04, activity level $4.5\,\mu\mathrm{m}$05, and planetary incident X-ray flux $4.5\,\mu\mathrm{m}$06 (Rukdee et al., 8 Oct 2025). Combining the measured X-ray luminosity with Ly$4.5\,\mu\mathrm{m}$07-inferred EUV gives $4.5\,\mu\mathrm{m}$08 and a present-day energy-limited mass-loss rate of $4.5\,\mu\mathrm{m}$09 under the adopted assumptions (Rukdee et al., 8 Oct 2025). This places K2-18b’s current escape in the weak, atmosphere-retaining regime rather than in catastrophic blow-off.

6. Competing physical interpretations and broader significance

The principal scientific dispute is whether the observed atmosphere overlies a liquid ocean, a deep gas envelope, or a molten surface. In the Hycean framework, K2-18b is treated as a planet with a hydrogen-rich atmosphere overlying a liquid water ocean. Under that assumption, thermodynamic calculations show that the coexistence of abundant H$4.5\,\mu\mathrm{m}$10 with oxidized carbon species creates a strong drive for methanogenesis: more than $4.5\,\mu\mathrm{m}$11 can be released from CO$4.5\,\mu\mathrm{m}$12 hydrogenation across 25–120$4.5\,\mu\mathrm{m}$13C and 1–1000 bar, DMS hydrogenation can yield approximately 62–98 $4.5\,\mu\mathrm{m}$14, and even glycine and alanine synthesis can become energy-releasing or much less costly than in Earth’s ocean (Glein, 2024). These results, however, are explicitly conditional on the existence of a Hycean ocean-atmosphere system.

A contrasting interpretation treats K2-18b as a gas-rich mini-Neptune with no habitable surface. In that model family, a lifeless Hycean atmosphere is hard to reconcile with the JWST data because photochemistry supports $4.5\,\mu\mathrm{m}$15 part-per-million CH$4.5\,\mu\mathrm{m}$16, whereas a $4.5\,\mu\mathrm{m}$17 solar mini-Neptune atmosphere produces about $4.5\,\mu\mathrm{m}$18 CH$4.5\,\mu\mathrm{m}$19 and nearly $4.5\,\mu\mathrm{m}$20 CO$4.5\,\mu\mathrm{m}$21 through deep thermochemistry and vertical mixing, while remaining broadly consistent with the non-detections of H$4.5\,\mu\mathrm{m}$22O, NH$4.5\,\mu\mathrm{m}$23, and CO (Wogan et al., 2024). A different non-Hycean explanation invokes a magma ocean: under reducing conditions, nitrogen dissolves efficiently into silicate melt, so atmospheric NH$4.5\,\mu\mathrm{m}$24 depletion can arise naturally without a liquid water ocean, and the most diagnostic discriminator becomes the CO$4.5\,\mu\mathrm{m}$25/CO ratio in the $4.5\,\mu\mathrm{m}$26m region (Shorttle et al., 2024).

The debate is not resolved by current JWST data. A self-consistent Hycean study coupling photochemistry, radiative–convective equilibrium, and transmission forward modeling found that a $4.5\,\mu\mathrm{m}$27 bar H$4.5\,\mu\mathrm{m}$28 envelope with percent-level CH$4.5\,\mu\mathrm{m}$29 and CO and CO$4.5\,\mu\mathrm{m}$30 buffered at $4.5\,\mu\mathrm{m}$31–$4.5\,\mu\mathrm{m}$32 can reproduce the 0.8–5.2 $4.5\,\mu\mathrm{m}$33m NIRISS+NIRSpec spectrum without invoking DMS, so Hycean and mini-Neptune interpretations both remain viable in that wavelength range (Fujisawa et al., 18 May 2026). This suggests that the decisive observables are likely to be deeper constraints on CO and CO$4.5\,\mu\mathrm{m}$34 between 4 and 5 $4.5\,\mu\mathrm{m}$35m, improved knowledge of stratospheric H$4.5\,\mu\mathrm{m}$36O and OH, and aerosol microphysics rather than additional debate over the same low-S/N mid-infrared bins.

Other lines of inquiry reinforce the picture of a volatile-rich but still ambiguous system. N-body plus CTL tidal simulations indicate that any moons around K2-18b would be extremely unlikely to survive, with lifetimes not exceeding 10 Myr for the adopted Earth-like or Neptune-like tidal parameters, far shorter than the $4.5\,\mu\mathrm{m}$37 Gyr system age (Patel et al., 15 Jul 2025). A coordinated narrowband radio technosignature search with the VLA and MeerKAT found no signals consistent with an astrophysical or artificial origin and placed upper limits of $4.5\,\mu\mathrm{m}$38 to $4.5\,\mu\mathrm{m}$39 on persistent, isotropic narrowband transmitters in the system (Tremblay et al., 10 Feb 2026). K2-18b therefore remains important not because any single interpretation has prevailed, but because it is one of the few temperate sub-Neptunes for which interior structure, atmospheric chemistry, stellar irradiation, and habitability hypotheses can all be confronted directly by data.