There is a gap in the size distribution of exoplanets — the radius valley — where planets between about 1.5 and 2 Earth radii are mysteriously scarce. Two explanations compete. Either all these planets start rocky and the smaller ones lose their hydrogen envelopes (a single population shaped by atmospheric escape), or some are rocky and some are water-rich from the start (two populations with genuinely different compositions). The data on radii alone can't distinguish between them.
Shibata and Izidoro (arXiv 2602.23250) found that orbital eccentricity breaks the tie. In N-body simulations where both rocky and icy protoplanets form, energy equipartition during gravitational scattering transfers kinetic energy from the more massive icy bodies to the less massive rocky ones. The result is an eccentricity contrast: near the radius valley, planets on slightly more eccentric orbits are statistically more likely to be water-rich, while those on more circular orbits are rocky. This pattern vanishes entirely when all planets share the same composition.
The mechanism is simple physics. In a system of mixed-mass bodies interacting gravitationally, the heavier bodies slow down and the lighter ones speed up — the same energy redistribution that thermalizes a gas of mixed-mass molecules. But here the mass difference has a specific origin: icy protoplanets form beyond the snow line where water ice adds to their bulk, while rocky ones form closer in where it's too hot for ice. The mass difference isn't a coincidence. It's a compositional fingerprint written in momentum.
What makes this result powerful is what it doesn't require. You don't need to measure a planet's density (which demands both radius and mass, and mass is hard to get for small planets). You don't need atmospheric spectroscopy (which requires the right geometry and enormous telescope time). You need the orbital eccentricity distribution of a population — something that transit timing and radial velocity surveys already provide. The evidence was in existing data, visible only once you knew what pattern to expect.
This connects to a broader principle about how confirmation works in science. Neither the radius distribution alone nor the eccentricity pattern alone proves water-rich mini-Neptunes exist. But any single formation mechanism must explain both simultaneously. A model that reproduces the radius valley but predicts no eccentricity contrast — or vice versa — is falsified by the joint constraint. The confirmation power lives in the intersection, not in either observable alone. When two independent measurements point the same direction, and their agreement would be a coincidence under the alternative hypothesis, that coincidence becomes the evidence.