At least one in a dozen stars exhibits evidence of planetary ingestion

The chemical composition of stars can reveal clues about their formation and evolution, including interactions with their planetary systems. The ingestion of planetary material into a star's atmosphere can alter its elemental abundances, leaving a detectable "planet signature." This signature manifests as correlations between elemental abundance differences and the dust condensation temperature (Tcond). However, detecting these subtle planet signatures is difficult due to several factors: the unknown occurrence rate of planetary ingestion, the small amplitude of abundance changes, and the heterogeneous nature of stellar samples with varying ages and compositions. To overcome these challenges, studying stars born together (co-natal stars) offers significant advantages, as they share the same initial composition, thus minimizing the effects of initial abundance variations. Previous spectroscopic studies have been limited by small sample sizes of binary stars. The launch of the Gaia satellite has provided unprecedented high-precision astrometric data, enabling the identification of a large number of co-moving stellar pairs confirmed to be co-natal. This study leverages this dataset to investigate the prevalence of planetary ingestion signatures in a homogeneous sample of co-natal stars, aiming to improve our understanding of the star-planet-chemistry connection and gain insights into the mechanisms of planet engulfment, formation, and evolution.

Previous studies have explored the link between stellar abundances and planetary systems. Some studies focused on the peculiar solar composition and its possible relation to planet formation, suggesting that the removal of refractory material during planet formation might leave a distinctive imprint on the star's composition. Other research has investigated the influence of giant planets on the composition of solar twins, suggesting that distant giant planets could prevent the accretion of refractory material from outer regions. However, most previous spectroscopic studies investigating planet signatures were limited to small numbers of binary stars, hindering statistically robust conclusions. The reported instances of planetary ingestion in these studies varied, partially due to the limitations of sample size and precision of measurements. The availability of Gaia data offers a significant opportunity to address these limitations by providing a much larger and more homogeneous sample of co-natal stars.

This study utilizes high-precision astrometric data from the Gaia satellite to identify a large sample of co-moving stellar pairs. From this sample, 91 pairs with close spatial separations (Δs < 106 AU) were selected as co-natal pairs, while 34 pairs with larger separations served as a control sample. High-resolution and high signal-to-noise ratio (S/N ≈ 250 per pixel) spectra were obtained from the European Southern Observatory's Very Large Telescope, the Magellan Telescope, and the Keck Telescope. Precise stellar parameters were determined using a line-by-line differential analysis, minimizing systematic uncertainties and achieving extremely high precision (~0.015 dex relative abundance errors for 21 elements). This sample size is significantly larger than in previous studies. A Bayesian analysis was developed and applied to the precise abundance data, employing a model for planetary ingestion. This model quantitatively estimates the mass of bulk Earth material (ME) required to match the observed abundance patterns for each co-moving pair, along with the Bayesian evidence (ln(Z)planet) for the model. The results were then compared against a flat model (null hypothesis) and an atomic diffusion model. The difference in Bayesian evidence (Δln(Z)) between the planetary ingestion and flat models served as a stringent indicator to identify chemical signatures of planetary material ingestion. The analysis also incorporated the effects of atomic diffusion, utilizing the MIST atomic diffusion models to predict abundance changes at different stellar evolutionary phases. The Bayesian evidence for both the planetary ingestion and atomic diffusion models was calculated, and the differences in evidence (Δln(Z)atom) were used to further validate the planetary ingestion findings. In addition to the Bayesian approach, the trend between abundance differences and Tcond was examined using linear least-squares fits, providing an independent assessment of planetary ingestion.

The study identified at least seven new instances of planetary ingestion among the 91 co-natal pairs, corresponding to an occurrence rate of 8% (±3%, assuming Poisson noise). This finding roughly doubles the number of confirmed planetary ingestion events. This occurrence rate is comparable to estimates based on solar twins and theoretical predictions from N-body simulations. The Bayesian modeling was crucial in disentangling planet signatures from other factors, such as random abundance variation and atomic diffusion. The Bayesian evidence comparison provided a more robust and stringent method for identifying planetary ingestion compared to solely relying on Tcond trends. The analysis revealed that the Tcond trends alone are insufficient to effectively distinguish planet signatures from other effects. The study found that 21 pairs fulfilled the Tcond slope criterion, 11 pairs fulfilled the Δln(Z) criterion, and seven pairs fulfilled all three criteria (Tcond slope, Δln(Z), Δln(Z)atom). The seven new instances of planetary ingestion exhibited amplitudes larger than 0.05 dex (up to 0.15 dex), favoring the scenario of planet engulfment over other planet-related scenarios (e.g., accretion of protoplanetary disks or the formation of terrestrial planets). The mass of accreted Earth material for the seven candidates averaged 4.3 ± 0.8 M⊕.

The findings of this study provide valuable insights into the long-term evolution of planetary systems and the frequency of planetary ingestion events. The observed occurrence rate (8%) of planetary ingestion suggests that a significant fraction of Sun-like stars experience this phenomenon. The detection of distinct chemical signatures in co-natal stars strongly supports the interpretation of planet engulfment. The amplitudes of the observed abundance differences favor the scenario of late accretion events, possibly caused by outer perturbers or the slow erosion of inner planets' atmospheres. This challenges the previously perceived stability of super-Earth systems. Considering the estimated occurrence rates of super-Earths, the study suggests that roughly every 4-10 super-Earth systems may experience late ingestion events. The higher occurrence rate compared to some previous studies may be attributed to the more homogeneous sample and the analysis methods utilized in this study, which incorporated a larger set of elements and a more sophisticated Bayesian approach. The results highlight the importance of utilizing co-natal stars in studying planet signatures and the need for advanced statistical techniques to disentangle these signatures from other confounding factors.

This study significantly expands our knowledge of planetary ingestion, doubling the number of confirmed instances and providing a more robust estimate of its occurrence rate. The use of a large homogeneous sample of co-natal stars and a sophisticated Bayesian analysis resulted in a more accurate and reliable detection of planet signatures. Future research could focus on expanding the sample size further, incorporating more sophisticated stellar models (including 3D models), and exploring alternative scenarios to explain the observed abundance differences. Incorporating other data such as asteroseismology could also further constrain stellar models and improve the precision of abundance determination.

The study primarily focused on close, co-natal pairs of stars, potentially missing ingestion events in more widely separated systems. The Bayesian analysis relied on assumptions about the composition of ingested planetary material. The atomic diffusion models used are 1D models; more sophisticated 3D hydrodynamic models could improve the accuracy of accounting for diffusion's effects. The uncertainties associated with the mass fraction of the stellar convection zone could affect the estimates of the mass of accreted planetary material.