A breakthrough machine learning technique has been developed by researchers from Keele University, aiding astronomers in more accurately determining the ages of stars by analyzing the chemicals within their atmospheres. The findings of this innovative research will be presented at the upcoming 2023 National Astronomy Meeting by George Weaver, a Ph.D. student at Keele.
Determining the age of a star has long posed a challenge for scientists. Unlike physical samples such as meteorites or rocks from celestial bodies, stars cannot be directly analyzed for their chemical composition and age using radioactive dating. Instead, astronomers rely on estimating age based on the light emitted by stars. While this is relatively straightforward for groups of stars that evolve together, known as star clusters, it becomes significantly more complex when dealing with individual stars.
In the early stages of a star’s life, the intensifying heat and pressure cause changes in its atmospheric chemical composition. One notable change is the gradual reduction of lithium, referred to as “lithium depletion.” Existing models have not adequately accounted for the intricate nature of this phenomenon.
Fortunately, the Gaia-ESO survey has yielded an extensive collection of high-quality spectra, which provides astronomers with an opportunity to delve deeper into the intricacies of lithium depletion. The newly developed neural network model, an advancement of the previous mathematical model called EAGLES, utilizes data from over 6,000 stars to establish a correlation between a star’s temperature, measured lithium abundance, and age.
This novel method is designed to be scalable, and efforts are already underway to incorporate a larger volume of data into the model. By incorporating as much information as possible, researchers aim to create age estimates that are more comprehensive. Ongoing tests are exploring the inclusion of stellar metallicity, which quantifies the abundance of elements heavier than helium in a star. Furthermore, future expansions of the model may encompass a star’s rotational slowdown over its lifespan and the progressive decrease in its magnetic activity.
George Weaver, the lead author of the forthcoming paper, and a Ph.D. student, elucidates, “While there are multiple independent age estimation methods and models, this artificial neural network offers the opportunity to create a unified approach for estimating a star’s age using spectral measurements.” He further explains, “Not only could this lead to a comprehensive model for determining stellar and cluster ages, but it will also enable us to quantify and establish relationships between these observables and age, potentially revealing previously unknown connections.”
Source: Royal Astronomical Society