Unmasking the true ages of stars
Modern models of stellar evolution rely on knowledge of helium abundance – a critical, yet hard-to-constrain ingredient. As the second most abundant element in the universe, helium dictates a star’s internal composition, temperature, luminosity and ultimate lifespan. Given that it is notoriously difficult to measure it, traditional models simply guess or assume a star’s helium content rather than determining it from observations. This created systematic uncertainties on our cosmic maps. The central objective of the ERC-funded CartographY project was to move past blind assumptions and make stellar profiles not just more precise but genuinely trustworthy.
Trading guesswork for ‘honest’ data
“A main project activity was to map and measure internal stellar helium using asteroseismology – the study of stellar oscillations,” notes project coordinator Guy Davies. “By combining high-quality asteroseismic data with spectroscopy and astrometry, we moved past the flaw of relying on a single, fixed set of assumptions.” By combining these data, the team inferred stellar properties including mass, radius, age and helium content, while also quantifying the uncertainty in those properties. To turn vast amounts of observational data into reliable science, CartographY combined stellar physics, machine learning and statistical inference. Researchers built large grids of stellar evolution models, mapping out how different masses, helium abundances and internal mixing assumptions affect a star’s observable properties. Because calculating full stellar life cycles is computationally exhausting, the team trained machine-learning emulators to mimic these detailed models much more quickly. Ultimately, instead of treating each star in isolation, researchers introduced Bayesian hierarchical models(opens in new window). This allowed them to study entire star populations simultaneously, which helped reveal broader patterns such as how helium levels change alongside other metals in different star groups. “A key project achievement was the ability to infer stellar ages in a more honest and self-consistent way. Instead of asking what a star’s age is under one rigid, potentially flawed model, CartographY evaluates how robust that age remains across different plausible models of stellar evolution,” outlines Davies. “By incorporating systematic uncertainties directly into the equations, we can now distinguish genuinely well-constrained stellar data from data that only appears precise because model flaws were ignored.” Furthermore, the use of Bayesian model evidence offered the team a principled way to compare competing stellar models. “Rather than picking the model that delivers the smallest formal uncertainty, it identifies which models are actually best supported by the observational data,” explains Davies.
The spin problem and the hunt for exoplanet hosts
To validate the models and truly understand where current stellar models fail, the next phase requires scaling CartographY’s method to larger, more diverse star samples. Another critical frontier is stellar rotation. A star’s spin holds vital clues about its age, yet CartographY revealed that current understanding of rotation evolution is less secure than previously hoped. Applying an incorrect rotation model too aggressively risks biasing the inferred stellar ages. Therefore, rather than rushing to use rotation as a blind metric, the team is focused on using their Bayesian framework to carefully validate and improve rotation theories first. Looking ahead, this new methodology is positioned to play a critical role. Upcoming missions like ESA’s PLATO will soon deliver high-quality data on exoplanet host stars, making it more important than ever to have tools that provide genuinely accurate, rather than just highly precise, stellar profiles.
