We assume planets reflect their stars. A metal-rich star should host metal-rich planets. A carbon-rich star should host carbon-rich planets. The host composition sets the menu; the planet picks from what's available.
In reduced (carbon-rich) stellar systems, thermochemical modeling shows this assumption fails. The planet's composition diverges dramatically from the star's. Carbon-rich disks produce iron-nickel alloys, silicon carbide, and graphite as the primary condensates — materials that bear little resemblance to the solar-composition silicates we're used to. The resulting planets can be “super-Mercuries”: metal-enriched worlds with thin mantles, whose bulk composition is decoupled from the photospheric abundance ratios of their host star.
The decoupling happens because condensation is nonlinear. Small changes in the C/O ratio of the disk gas shift the condensation sequence — which minerals form first, which remain in the gas, which get incorporated into planetary cores. The star's atmosphere, well-mixed and homogeneous, gives no warning of these phase boundaries. Two stars with similar spectra can produce radically different planetary systems.
This breaks one of exoplanet science's favorite shortcuts: inferring planetary composition from stellar spectroscopy. The inference requires a mapping from star to planet, and that mapping is not one-to-one. The same star can produce very different worlds depending on where in the disk the planet forms and what condensation sequence operates there.
Note (2026-11-11): checked against the full text of the paper (Zaveri and co-authors, arXiv 2603.15955, 'A chemical perspective on planet formation in reduced systems'). The core is as the essay says: for Sun-like stars with C/O ratios from 0.65 to 0.95, equilibrium condensation calculations (FactSage, three disk pressures) show that condensation sequences differ from the solar one, carbides, silicides and sulfides appear, and the bulk iron-to-magnesium, iron-to-silicon and iron-to-oxygen ratios of the resulting planetesimals deviate substantially from the star's. Five points differ. (1) The paper computes the compositions of planetesimals, the building blocks aggregated from feeding zones, not of planets; it says these provide 'potential formation pathways' for metal-enriched super-Mercury analogues and chemically distinct carbon- and sulfur-rich rocky planets. The essay says planets 'can be super-Mercuries' with thin mantles as a result. (2) The essay says 'the star's atmosphere ... gives no warning of these phase boundaries' and that two similar stars can yield radically different systems. In the paper the stellar abundances are the input: it uses measured abundances of real stars with different C/O ratios, and the three regimes (solar-like below about 0.7, transitional about 0.7 to 0.91, reduced above about 0.92 in the abstract) are defined by the star's own C/O ratio, so the star does carry the information. What the paper says fails is the use of solar-nebula condensation temperatures to classify elements as refractory or volatile in reduced disks. (3) The deviation is not uniform: in the solar-like regime the ratios track the star, and the paper reports a range from metal-rich, carbon- and sulfur-bearing bodies to more Earth-like ones inside a single disk, depending on temperature and pressure. (4) The essay lists iron-nickel alloy, silicon carbide and graphite as the primary condensates; the paper's abstract names carbides, silicides and sulfides appearing and oxygen-bearing silicates condensing at lower temperatures, and I did not check the alloy and graphite statement. (5) The calculations are equilibrium condensation with a stochastic accretion model; the essay's explanation that condensation is 'nonlinear' and that the mapping is not one-to-one is the essay's wording.