A Link Between Rocky Planet Density and Host Star Chemistry

TL;DR

The study investigates whether rocky planet interior densities correlate with host-star chemistry, testing the idea that planetary bulk composition tracks Galactic chemical evolution. Using FGK and M-dwarf hosts from SDSS-V/MWM, the authors assemble 22 rocky planets with robust density measurements and re-derive homogeneous planetary properties. They find a strong inverse correlation between planet bulk density and host-star [Mg/Fe], which strengthens when including M-dwarf hosts, suggesting denser, iron-rich interiors in Mg-poor environments and more mantle-dominated interiors in alpha-enhanced regions. This link between Galactic chemical evolution and rocky planet formation implies Earth-like interiors may vary systematically across the Galaxy, with thick-disk, alpha-enhanced environments potentially yielding different interior partitions than thin-disk regions.

Abstract

Planets and their host stars form from the same cloud of gas and dust, so we assume that their chemical compositions are linked. However, a clear correlation between rocky planet interior properties and host star chemistry remains elusive for planets around FGK dwarfs, and non-existent for planets around M dwarfs because cool stars frequently lack detailed chemical information. Here, we investigate the relationship between small (R$_{P}$ $\leq$ 1.8 R$_{\oplus}$) planet densities and host star elemental abundances. We use the Sloan Digital Sky Survey-V/Milky Way Mapper and an accompanying data-driven framework to obtain abundances for FGK and M dwarf hosts of 22 rocky planets. We find that planet densities exhibit a strong, inverse relationship to [Mg/Fe] abundances of FGK hosts (p = 0.001). This correlation becomes more significant with the addition of M dwarf hosts (p = 0.0005). If we assume that rocky planets have terrestrial-like compositions, this suggests that low [Mg/Fe] environments form planets with larger Fe-rich cores and thus higher densities. The thick disk planets in our sample help anchor this trend, illustrating the importance of sampling exoplanet properties across a range of host star populations. This finding highlights the connection between Galactic chemical evolution and rocky planet formation, and indicates that Earth-like planet compositions may vary significantly across different regions of the Galaxy.

Figures (6)

• Figure 1: All small ($R_{P}$$\leq$ 1.8 $R_{\oplus}$), rocky planets from the Exoplanet Archive christiansen2025 with $<$30% fractional radius and mass uncertainties (gray points). Planets with masses and radii consistent with a rocky composition (pure iron to pure MgSiO$_{3}$ rock) are contained within the brown shaded region zeng2019. Our rocky planet sample with host star abundances from SDSS-V are shown in color, which corresponds to host star [Fe/H]. Planets hosted by FGK dwarfs are marked as circles, while planets hosted by M dwarfs are marked as triangles. The one planet not contained within the zeng2019 region is TOI-561 b, the only high-confidence thick disk planet in our sample. We also label Kepler-10 b, another possible thick disk planet.
• Figure 2: [Mg/Fe] vs. [Fe/H] for stars that host our rocky planet sample, as well as all main-sequence dwarfs in SDSS-V with reliable abundances. The SDSS-V dwarfs are plotted in the background as transparent blue dots, while the FGK and M dwarfs that host our rocky planet sample are marked as circles and triangles, respectively, and are colored by the densities of their rocky planets. The blue contours trace the background star 10% density distribution levels, and the dashed black line delineates the thin-thick disk separation franchini2020. TOI-561 and Kepler-10 are highlighted as high-confidence and possible thick disk stars, respectively.
• Figure 3: Planet bulk density vs. all rocky elements considered in our analysis in [X/H] form. Rocky solar system bodies (Moon, Mercury, Venus, Earth, and Mars) are plotted for comparison as orange diamonds. Rocky planets hosted by FGK and M dwarfs are denoted as circles and triangles, respectively, and are colored by planet radii values. The best-fit linear trends from our MCMC fitting routine are plotted as blue lines along with their 1$\sigma$ confidence intervals. We provide the Pearson correlation statistics in the bottom left corners of each panel considering just the FGK-hosted rocky planets, and the entire FGKM sample. TOI-1347 b is not present in the Si, O, and Ni panels because it has unreliable host star abundances for these elements.
• Figure 4: Planet bulk density vs. all rocky elements considered in our analysis in [X/Fe] form. As in Figure \ref{['fig:figure3']}, rocky solar system bodies (Moon, Mercury, Venus, Earth, and Mars) are plotted for comparison as orange diamonds. Rocky planets hosted by FGK and M dwarfs are denoted as circles and triangles, respectively, and are colored by planet radii values. The best-fit linear trends from our MCMC fitting routine are plotted as blue lines along with their 1$\sigma$ confidence intervals. We provide the Pearson correlation statistics in the bottom left corners of each panel considering just the FGK-hosted rocky planets, and the entire FGKM sample. TOI-1347 b is not present in the Si, O, and Ni panels because it has unreliable host star abundances for these elements.
• Figure 5: Rocky planet densities vs. host star [Mg/Fe] for planets hosted by both FGK (circles) and M dwarfs (triangles), colored by planet radii values. Rocky solar system bodies (Moon, Mercury, Venus, Earth, and Mars) are plotted for comparison as orange diamonds. Our best fit linear model from our MCMC fitting routine is plotted as a blue line along with the associated 1$\sigma$ confidence intervals. Pearson correlation statistics are provided in the bottom left corner considering just the FGK-hosted rocky planets, and the entire FGKM sample. We label the high-confidence and possible thick disk planets TOI-561 b and Kepler-10 b, respectively.