K2-19: A Compact Resonant Exoplanet System

• K2-19 is a compact exoplanetary system with three confirmed transiting planets (and one candidate) arranged in a tightly packed configuration.
• Transit timing variations (TTVs) in the resonant b-c pair enable precise determination of masses, eccentricities, and orbital dynamics.
• Decade-long, multi-method observations have established K2-19 as a benchmark for resonance theory and interior structure inference.

K2-19 is a compact exoplanetary system centered on the star EPIC 201505350. It was first identified in K2 Campaign 1 as a two-planet transiting system with orbital periods extremely close to a 3:2 commensurability, and it subsequently became a benchmark for transit-timing variation (TTV) analysis, photo-dynamical modeling, and resonance theory. The system is now treated as a three-planet system with confirmed transiting planets d, b, and c, and a possible outer candidate e; its scientific importance derives mainly from the strongly interacting b-c pair, whose large, structured TTVs have enabled repeated revisions of the masses, eccentricities, and resonant state over a decade of observations (Armstrong et al., 2015, Almenara et al., 2 Oct 2025).

1. Discovery and observational development

K2-19 entered the literature as one of the earliest dynamically interesting K2 multiplanet systems. The original K2 analysis used Campaign 1 photometry over an $\sim 80$-day time span and recovered two transiting planets, K2-19b and K2-19c, with radii $7.23^{+0.56}_{-0.51}\,R_\oplus$ and $4.21 \pm 0.31\,R_\oplus$, respectively. Their period ratio was measured as $P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$, corresponding to $\Delta = 0.00234 \pm 0.00002$ relative to the 3:2 resonance, and a NITES transit of planet b demonstrated a large TTV signal. The two transiting objects were statistically validated with the PASTIS tool, with posterior probability $99.2^{+0.4}_{-0.8}\%$ for the scenario in which both planets transit the target star (Armstrong et al., 2015).

Follow-up observations quickly broadened the empirical basis for the system. Subaru/HDS spectroscopy and Subaru/HiCIAO adaptive-optics imaging refined the stellar parameters and excluded contaminating nearby sources bright enough to mimic or significantly dilute the transits. Ground-based transit follow-up of b with KeplerCam, TRAPPISTCAM, NITES, and MuSCAT confirmed TTVs and enabled an early analytic “synodic chopping” analysis (Narita et al., 2015).

Subsequent characterization introduced additional datasets and methods. A five-additional-RV plus photo-dynamical analysis combined K2 photometry, ground-based transits, and SOPHIE radial velocities, even though the stellar reflex motion was not significantly detected (Barros et al., 2015). Later studies added FIES@NOT, HARPS-N@TNG, and [email protected] radial velocities and performed joint RV+TTV fitting with TRADES (Nespral et al., 2016). A more extensive photodynamical model combined K2, Spitzer, ground-based telescopes, and Keck/HIRES, and also incorporated the inner planet d (Petigura et al., 2019). The most recent revisions extend the time baseline to roughly a decade using K2, Spitzer, TESS, CHEOPS, and multiple ground-based facilities, and apply either gradient-aware TTV modeling with jnkepler or full photodynamical modeling of the transit photometry (Entrican et al., 22 Sep 2025, Almenara et al., 2 Oct 2025).

2. Host star

Published stellar characterizations place K2-19 among slightly sub-solar late G-type or early K-type dwarfs. Early work reported $T_{\rm eff} = 5230 \pm 417~\text{K}$, $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$, $M_\star = 0.89 \pm 0.06~M_\odot$, and $R_\star = 0.88 \pm 0.06~R_\odot$ from transit-based stellar parameters, with a rotation period $7.23^{+0.56}_{-0.51}\,R_\oplus$0 inferred from the K2 light curve (Armstrong et al., 2015). Subaru/HDS spectroscopy later yielded substantially tighter atmospheric parameters, including $7.23^{+0.56}_{-0.51}\,R_\oplus$1, $7.23^{+0.56}_{-0.51}\,R_\oplus$2, $7.23^{+0.56}_{-0.51}\,R_\oplus$3, $7.23^{+0.56}_{-0.51}\,R_\oplus$4, $7.23^{+0.56}_{-0.51}\,R_\oplus$5, and $7.23^{+0.56}_{-0.51}\,R_\oplus$6, together with an age constraint $7.23^{+0.56}_{-0.51}\,R_\oplus$7 (Narita et al., 2015).

Later analyses using spectroscopy, Gaia-based stellar information, and rotation diagnostics converged on somewhat smaller radii and younger ages. Representative values include $7.23^{+0.56}_{-0.51}\,R_\oplus$8, $7.23^{+0.56}_{-0.51}\,R_\oplus$9, $4.21 \pm 0.31\,R_\oplus$0, and $4.21 \pm 0.31\,R_\oplus$1, with gyrochronological ages of roughly $4.21 \pm 0.31\,R_\oplus$2 Gyr and $4.21 \pm 0.31\,R_\oplus$3 Gyr reported in the decade-baseline studies (Entrican et al., 22 Sep 2025, Almenara et al., 2 Oct 2025). A 2016 RV paper, by contrast, adopted $4.21 \pm 0.31\,R_\oplus$4, $4.21 \pm 0.31\,R_\oplus$5, $4.21 \pm 0.31\,R_\oplus$6, $4.21 \pm 0.31\,R_\oplus$7, and $4.21 \pm 0.31\,R_\oplus$8 (Nespral et al., 2016).

These differences are method-dependent rather than merely cosmetic. Planetary radii, densities, and equilibrium temperatures scale directly with the adopted stellar radius and mass, and age estimates strongly affect the interpretation of atmospheric loss, tidal damping, and interior structure. A plausible implication is that K2-19 has served not only as a laboratory for planet dynamics but also as a case study in how stellar characterization propagates into exoplanet inference.

3. Planetary architecture

The current working architecture contains an inner Earth-size planet d, the resonant pair b and c, and a possible outer candidate e. The resonant pair remains the defining feature of the system, but the addition of d and the proposal of e changed the dynamical context from a two-planet subsystem to a compact chain-like architecture (Nespral et al., 2016, Almenara et al., 2 Oct 2025).

Body	Orbital period	Size and mass summary
K2-19d	$4.21 \pm 0.31\,R_\oplus$9 d	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$0, $P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$1 (99%)
K2-19b	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$2 d	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$3, $P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$4
K2-19c	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$5 d	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$6, $P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$7
K2-19e (candidate)	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$8 d	$P_c/P_b = 1.503514^{+0.000052}_{-0.000057}$9, $\Delta = 0.00234 \pm 0.00002$0 (99%)

The b-c pair sits close to the exact 3:2 ratio, with $\Delta = 0.00234 \pm 0.00002$1 in the revised decade-long photodynamical solution, while earlier analyses reported values such as $\Delta = 0.00234 \pm 0.00002$2, $\Delta = 0.00234 \pm 0.00002$3, and $\Delta = 0.00234 \pm 0.00002$4 depending on whether “instantaneous,” photodynamical, or revised long-baseline periods were used (Armstrong et al., 2015, Petigura et al., 2019, Almenara et al., 2 Oct 2025). The planets are tightly packed inside $\Delta = 0.00234 \pm 0.00002$5 AU in the early two-planet description, and in the revised four-body architecture the semimajor axes are $\Delta = 0.00234 \pm 0.00002$6 au, $\Delta = 0.00234 \pm 0.00002$7 au, $\Delta = 0.00234 \pm 0.00002$8 au, and $\Delta = 0.00234 \pm 0.00002$9 au (Almenara et al., 2 Oct 2025).

The physical hierarchy is also unusual. In K2-19 the inner resonant planet is larger than the outer resonant planet. Earlier analyses emphasized $99.2^{+0.4}_{-0.8}\%$0, with b at roughly $99.2^{+0.4}_{-0.8}\%$1 and c at roughly $99.2^{+0.4}_{-0.8}\%$2, a configuration described as unusual among first-order resonant systems (Narita et al., 2015). Later masses reinforced the asymmetry: b is consistently more massive than c in all mature analyses, even though the absolute values vary by method and epoch. This mass hierarchy has been repeatedly invoked in discussions of resonant capture and long-term stability (Nespral et al., 2016, Entrican et al., 22 Sep 2025).

4. Transit timing variations and resonance

K2-19 is one of the clearest TTV systems among compact transiting multiplanet systems. Early K2 plus ground-based follow-up already showed that the transits of b deviated from a linear ephemeris by more than $99.2^{+0.4}_{-0.8}\%$3 hour, and Narita et al. modeled those deviations with the synodic chopping formalism of Deck & Agol, truncating the harmonic series at $99.2^{+0.4}_{-0.8}\%$4 and finding two nearly symmetric solutions slightly above and below exact 3:2 commensurability (Narita et al., 2015). In that framework, the relevant short-timescale signal is tied to the synodic period,

$99.2^{+0.4}_{-0.8}\%$5

for which values $99.2^{+0.4}_{-0.8}\%$6 d and $99.2^{+0.4}_{-0.8}\%$7 d were reported for the two degenerate solutions (Narita et al., 2015).

The resonance question became the central interpretive issue. Classical first-order 3:2 angles are usually written as

$99.2^{+0.4}_{-0.8}\%$8

A 2019 photodynamical analysis found that the classical resonant angles circulate rather than librate and concluded that K2-19b and c are in a “strictly non-resonant configuration,” despite the extremely small offset from exact commensurability (Petigura et al., 2019). A 2020 reanalysis argued that this conclusion was too restrictive: in the proper canonical variables for a first-order two-planet resonance, the true resonant coordinate is $99.2^{+0.4}_{-0.8}\%$9, and the relevant resonant angle librates even when the conventional angles circulate. In that formulation, the resonant motion is diagnosed by $T_{\rm eff} = 5230 \pm 417~\text{K}$0, and K2-19b and c were shown to lie deep inside the resonant island (Petit et al., 2020).

The decade-long analyses resolved the dispute empirically. One study, using eight new transit times from TESS and a gradient-aware TTV code, reported that the system resides in resonance with a small libration amplitude and that $T_{\rm eff} = 5230 \pm 417~\text{K}$1 librates with an amplitude of about $T_{\rm eff} = 5230 \pm 417~\text{K}$2 (Entrican et al., 22 Sep 2025). Another, using a 10-year photodynamical dataset, described the pair as “deep inside” the 3:2 resonance and identified two incommensurate TTV frequencies: a superperiod $T_{\rm eff} = 5230 \pm 417~\text{K}$3 days and a resonant period $T_{\rm eff} = 5230 \pm 417~\text{K}$4 days (Almenara et al., 2 Oct 2025). The same work fit the TTVs with a two-frequency Fourier decomposition and reported amplitudes $T_{\rm eff} = 5230 \pm 417~\text{K}$5 min, $T_{\rm eff} = 5230 \pm 417~\text{K}$6 min, $T_{\rm eff} = 5230 \pm 417~\text{K}$7 min, and $T_{\rm eff} = 5230 \pm 417~\text{K}$8 min (Almenara et al., 2 Oct 2025). In parallel, the jnkepler analysis described the TTV amplitudes more directly as $T_{\rm eff} = 5230 \pm 417~\text{K}$9 hour for b and $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$0 hours for c, with dominant components near $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$1 days and $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$2 days (Entrican et al., 22 Sep 2025).

The common misconception that K2-19 is merely “near 3:2” now requires qualification. Historically that statement was reasonable, because early analytic fits and circulation of $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$3 and $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$4 admitted a non-resonant interpretation. The later canonical and decade-baseline analyses show that the physically relevant statement is different: K2-19b and c are in 3:2 resonance, but the resonance is not faithfully diagnosed by the conventional angles alone (Petit et al., 2020, Entrican et al., 22 Sep 2025).

5. Masses, densities, and internal structure

Mass determinations for K2-19 evolved substantially as the baseline lengthened and the modeling changed. The 2015 photo-dynamical analysis reported $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$5 and $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$6, leading to $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$7 and $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$8 (Barros et al., 2015). A 2016 RV-only plus TRADES analysis found a larger b mass and a smaller c mass, with combined RV+TTV values $[\mathrm{Fe/H}] = 0.38 \pm 0.23~\text{dex}$9 and $M_\star = 0.89 \pm 0.06~M_\odot$0, and corresponding densities $M_\star = 0.89 \pm 0.06~M_\odot$1 and $M_\star = 0.89 \pm 0.06~M_\odot$2 (Nespral et al., 2016).

A later, more precise photodynamical model with Keck/HIRES and extended transit data reported $M_\star = 0.89 \pm 0.06~M_\odot$3, $M_\star = 0.89 \pm 0.06~M_\odot$4, $M_\star = 0.89 \pm 0.06~M_\odot$5, and $M_\star = 0.89 \pm 0.06~M_\odot$6, implying bulk densities of about $M_\star = 0.89 \pm 0.06~M_\odot$7 for b and $M_\star = 0.89 \pm 0.06~M_\odot$8 for c (Petigura et al., 2019). The decade-baseline papers broadly preserved this lower-mass scale for b and a roughly $M_\star = 0.89 \pm 0.06~M_\odot$9–$R_\star = 0.88 \pm 0.06~R_\odot$0 scale for c, although they did not agree perfectly. The jnkepler TTV analysis reported $R_\star = 0.88 \pm 0.06~R_\odot$1 and $R_\star = 0.88 \pm 0.06~R_\odot$2 (Entrican et al., 22 Sep 2025), while the revised 10-year photodynamical solution gave $R_\star = 0.88 \pm 0.06~R_\odot$3, $R_\star = 0.88 \pm 0.06~R_\odot$4, $R_\star = 0.88 \pm 0.06~R_\odot$5, and $R_\star = 0.88 \pm 0.06~R_\odot$6 (Almenara et al., 2 Oct 2025).

Despite the quantitative revisions, the qualitative physical picture has been stable. K2-19b is a low-density sub-Saturn or envelope-rich sub-Neptune/sub-Saturn transition object, while K2-19c is a low-density mini-Neptune or Neptune-like planet. Earlier Lopez-model structure fits gave $R_\star = 0.88 \pm 0.06~R_\odot$7, $R_\star = 0.88 \pm 0.06~R_\odot$8, $R_\star = 0.88 \pm 0.06~R_\odot$9, and $7.23^{+0.56}_{-0.51}\,R_\oplus$00, which made b appear close to the canonical crossover regime for runaway accretion (Petigura et al., 2019). By contrast, the 2025 revised architecture used the GASTLI interior-atmosphere model and obtained for b a core mass fraction $7.23^{+0.56}_{-0.51}\,R_\oplus$01, envelope metal fraction $7.23^{+0.56}_{-0.51}\,R_\oplus$02, total planet metallicity $7.23^{+0.56}_{-0.51}\,R_\oplus$03, and envelope mass fraction $7.23^{+0.56}_{-0.51}\,R_\oplus$04, concluding that the planet is consistent with the mass-metallicity relation of Thorngren et al. and with core accretion (Almenara et al., 2 Oct 2025).

This revision matters because K2-19b had been singled out as a challenge to standard core-nucleated accretion. The newer structure modeling suggests that the earlier tension was at least partly contingent on the older radius, mass, and eccentricity solution rather than on an intrinsically anomalous composition.

6. Dynamical evolution, controversies, and current synthesis

The principal controversy in the literature concerned the conjunction of three claims: that b and c were very close to 3:2, that the classical resonant angles circulated, and that both planets had moderate eccentricities near $7.23^{+0.56}_{-0.51}\,R_\oplus$05. In 2019, that combination was interpreted as a challenge to standard convergent migration and resonance capture, since the system appeared near-commensurate but non-resonant, with apsidal alignment $7.23^{+0.56}_{-0.51}\,R_\oplus$06 and eccentricities $7.23^{+0.56}_{-0.51}\,R_\oplus$07, $7.23^{+0.56}_{-0.51}\,R_\oplus$08 (Petigura et al., 2019). The 2020 canonical analysis partly resolved the first part of the problem by showing that resonance can persist even if the classical angles circulate (Petit et al., 2020).

The second part of the problem concerned tides and long-term survival of planet d. With $7.23^{+0.56}_{-0.51}\,R_\oplus$09 and $7.23^{+0.56}_{-0.51}\,R_\oplus$10 near $7.23^{+0.56}_{-0.51}\,R_\oplus$11, secular coupling can force nonzero eccentricity in the innermost planet, and the 2020 study found that the configuration could not be maintained for more than a few hundred million years. It reported a median half-decay time $7.23^{+0.56}_{-0.51}\,R_\oplus$12 Myr and found that $7.23^{+0.56}_{-0.51}\,R_\oplus$13 of initial conditions had $7.23^{+0.56}_{-0.51}\,R_\oplus$14 Gyr, apparently inconsistent with a multi-Gyr system age unless the orbital solution was incomplete or a fourth planet altered the TTVs (Petit et al., 2020).

The decade-long analyses substantially alleviated this difficulty. The gradient-aware TTV study found broader, lower eccentricity constraints, with $7.23^{+0.56}_{-0.51}\,R_\oplus$15 and $7.23^{+0.56}_{-0.51}\,R_\oplus$16, and argued that the broader posterior reflects the true resonant degeneracies more faithfully than the earlier DE-MCMC exploration (Entrican et al., 22 Sep 2025). The revised 10-year photodynamical analysis went further and lowered the median eccentricities to $7.23^{+0.56}_{-0.51}\,R_\oplus$17 and $7.23^{+0.56}_{-0.51}\,R_\oplus$18, attributing the previously reported high eccentricities largely to a single LCOGT transit for which twilight data mimicked ingress and produced a $7.23^{+0.56}_{-0.51}\,R_\oplus$19-minute error in the midtransit time (Almenara et al., 2 Oct 2025). In the jnkepler study, the residual tidal constraint on b was expressed as $7.23^{+0.56}_{-0.51}\,R_\oplus$20, consistent with gas/ice giant values and therefore compatible with persistence of the resonance over the system age (Entrican et al., 22 Sep 2025).

The current synthesis is accordingly more coherent than the intermediate picture. K2-19 is best understood as a compact system with an inner Earth-size planet d and a b-c pair genuinely locked in the 3:2 mean-motion resonance, exhibiting large, multi-harmonic TTVs over a decade-long baseline. The resonant pair has modest but nonzero free eccentricity, small libration amplitude, and a clear mass hierarchy in which the inner resonant planet is the more massive and larger body. A plausible implication is that K2-19 is no longer a primary challenge to resonance capture itself; instead, it is a high-value empirical testbed for how resonant dynamics, TTV inversion, and interior-structure inference evolve when the observing baseline expands from months to a decade (Entrican et al., 22 Sep 2025, Almenara et al., 2 Oct 2025).