Exoplanet census reveals two distinct families of planetary systems

Astronomers surveying the ever-growing catalogue of worlds beyond our Solar System have uncovered a hidden order within the apparent chaos. More than a thousand multi-planet systems are now known, and their architectures span more than four orders of magnitude in mass and five in orbital scale. A new population study published in Nature Astronomy shows that this bewildering diversity is not a continuum of random outcomes but resolves into two clearly separated families, each with its own characteristic architecture, dynamical temperament and breed of host star. The finding suggests that the blueprint of a planetary system is written jointly by the physics of planet formation and by the properties of the star that anchors it, and that these fingerprints survive billions of years of subsequent dynamical evolution.

The research team, led by Diego Turrini of Italy’s National Institute for Astrophysics and Camilla Danielski of the University of Valencia, confronted a problem familiar to anyone who has followed the exoplanet revolution: observational bias. The planets we detect are not necessarily the planets that are most common; they are simply the ones our instruments can reach. Transiting surveys such as NASA’s Kepler mission favour tightly packed, short-period configurations, while radial velocity, imaging and astrometric searches preferentially uncover massive planets at wide separations. Population synthesis models, meanwhile, still struggle to reproduce the Solar System and the observed exoplanet demographics simultaneously. To avoid mistaking a partial view for the whole elephant, the team built four independent samples of multi-planet systems selected by different criteria and analysed them with a common set of architecture-agnostic metrics.

The reference sample, labelled L, contains 298 systems, roughly 28 percent of all known multi-planet systems, for which both the semimajor axis and the mass of every planet are available. A second sample of 189 systems was cross-matched with the SWEET-Cat catalogue to obtain homogeneous stellar parameters, allowing the team to test trends involving host star properties on an even footing. A third sample of 89 systems has measured orbital eccentricities for every planet, and a fourth, strictest sample of 68 systems has absolute planetary masses with well-characterized uncertainties throughout. Where eccentricities or inclinations were missing, the researchers filled the gaps using multiplicity-dependent correlations derived from the Kepler population, assuming the equipartition of dynamical excitation among the secular degrees of freedom of each system.

The heart of the analysis lies in nine normalized parameters that place systems of wildly different scales on a common footing. These include the normalized system mass, the ratio of total planetary mass to stellar mass; the orbital period of the outermost planet; the planetary multiplicity; the fraction of the system’s mass locked in its most massive planet; an orbital spacing metric that captures both spatial and gravitational compactness; and the normalized angular momentum deficit, or NAMD, which the authors describe as a kind of dynamical temperature. Host stars were characterized by metallicity, mass, and a quantity called planet-building material, which estimates the dust content of the now-vanished protoplanetary disk using the empirical correlation between stellar mass and disk dust mass.

When the systems were plotted in the mass-period plane, a striking bimodality emerged, and it held in all four samples. Statistical tests, including Kolmogorov-Smirnov tests corrected for multiple comparisons and Monte Carlo-resampled Anderson-Darling tests with false discovery rate control, confirmed that the split is real and not an artefact of measurement uncertainty. The lower-mass family contains systems whose total planetary mass falls below roughly 80 Earth masses around a solar-mass star. The high-mass family begins above roughly 400 Earth masses. Between the two peaks lies an uncertain membership zone where systems could plausibly belong to either population. Notably, the two families map onto the main groups of exoplanets discovered by the transit and radial velocity methods, and the parameter space where hot Jupiters reside appears conspicuously underpopulated, a puzzle the authors explore in their supplementary analysis.

The two families differ in almost every architectural respect. Lower-mass systems are dominated by two-planet configurations, but up to 45 percent of them host three or more planets, compared with 25 percent of high-mass systems. In most lower-mass systems the heaviest planet carries between one- and two-thirds of the total planetary mass, a distribution consistent with the famous peas-in-a-pod pattern seen among Kepler’s transiting multi-planet systems, where neighbouring planets tend to have similar sizes. High-mass systems instead show a flat distribution of mass concentration: arrangements of similar-mass planets are exactly as probable as systems in which nearly all the mass sits in a single dominant object. The Solar System, with Jupiter holding about 70 percent of the planetary mass, falls within the range shared by both families.

Dynamically, the contrast is even sharper. Both families contain many systems more excited than our own, whose NAMD value of 1.27 times ten to the minus three serves as the reference line between order and chaos. The modal excitation of lower-mass systems is broadly comparable to that of the Solar System, though it shifts upward in the sample with fully measured eccentricities, hinting at the role of instabilities in sculpting even the quiet Kepler multis. High-mass systems, however, run about an order of magnitude hotter than both their lower-mass counterparts and the Solar System, pointing to violent evolutionary histories driven by global instabilities and planet-planet scattering. Their gravitationally packed architectures amplify the perturbations exchanged during close encounters, and their larger share of two-planet systems fits the well-documented anticorrelation between dynamical excitation and multiplicity: chaotic histories tend to thin the planetary herd.

The host stars tell their own story. Lower-mass systems preferentially orbit subsolar-to-solar mass stars with metallicities peaking between minus 0.1 and 0.1 dex, essentially solar values. High-mass systems orbit solar-to-supersolar stars whose metallicity distribution peaks near 0.2 dex, about 1.5 times higher. Individually these differences are modest, but they compound constructively through the star-disk mass relation: the natal disks of high-mass systems contained roughly three times more heavy elements, dust and ices, than those of their lower-mass counterparts. In such gas- and dust-rich environments, planetary growth outpaces orbital migration, allowing massive planets to form at larger distances and to assemble over extended timescales under varying accretion conditions, which naturally produces the wide dispersion of planet masses observed within individual high-mass systems. Lower-mass systems, by contrast, arise from Solar System-like environments where slow growth combined with migration and pebble accretion yields planets of comparatively similar mass packed into a fraction of an astronomical unit.

To probe the dynamical roots of these architectures, the team ran hundreds of instability simulations of synthetic systems starting from four kinds of initial configurations: unstable non-resonant setups, incomplete resonant chains, and stable resonant and non-resonant chains perturbed by passing stars. The verdict was clear. Initially stable configurations, whether resonant or not, failed to reproduce the observed fractions of highly excited systems. The excitations of both real families are best matched by systems that began unstable or in incomplete resonance chains, and most simulations that produced the observed excitation levels ended with at least one planet lost to collision, ejection or stellar engulfment. Combined with the observed scarcity of multi-planet systems locked in resonant chains, this suggests that planet formation is only moderately efficient at producing stable resonant architectures, and that most known systems bear the scars of a turbulent youth.

The broader message is a caution against cosmic parochialism. Both families, despite their differences, are more spatially compact than the inner Solar System and less compact than the outer one, and both contain large fractions of systems more dynamically excited than our own. The authors argue that the Sun and its planets should not be treated as a universal template: the star-planet genetic link must be part of any complete theory of planet formation. The work also offers a natural home for hot Jupiters, which plausibly arrive at their scorching orbits through high-eccentricity migration within the very high-NAMD systems the study identifies. As surveys such as ESA’s PLATO and the ongoing characterization efforts of the Ariel mission expand the census, the two-family picture will face its sternest test, and the boundary region between the families, where uncertain systems currently linger, may yet hide the transitional architectures that reveal how one kind of planetary system transforms into another.

Subject of Research: Architectural and dynamical classification of multi-planet exoplanet systems and their host stars

Article Title: Two families of multi-planet systems with distinct architectures and stellar hosts

Article References: Two families of multi-planet systems with distinct architectures and stellar hosts. (n.d.). https://doi.org/10.1038/s41550-026-02990-8

Image Credits: AI Generated

DOI: 10.1038/s41550-026-02990-8

Keywords: exoplanets, multi-planet systems, planetary architectures, dynamical excitation, planet formation, stellar metallicity, protoplanetary disks, Kepler mission, peas-in-a-pod, planet-planet scattering, hot Jupiters, Nature Astronomy