Chinese researchers develop AI model to process stellar data from different telescopes

BEIJING -- A Chinese research team has developed an artificial intelligence (AI) model called SpecCLIP, which can interpret stellar spectral data from different telescopes, demonstrating the vast potential of AI in processing and integrating massive astronomical datasets, the Science and Technology Daily reported on Wednesday.

Stellar spectra contain unique information about stars, including a star's temperature, chemical composition and surface gravity. By analyzing these spectra, astronomers can trace the evolutionary history of the Milky Way from its beginning to the present.

However, current research faces a significant challenge: different survey projects, such as China's LAMOST and Europe's Gaia satellite, acquire spectral data through varying methods, resolutions and wavelength ranges. These datasets are like stories told in different dialects, making it difficult to combine them directly for large-scale analysis.

To address this data barrier, a research team from the National Astronomical Observatories of the Chinese Academy of Sciences, the University of Chinese Academy of Sciences (UCAS) and other institutions introduced concepts similar to large language models into astronomy and applied a contrastive learning method, creating AI that is capable of learning and establishing intrinsic connections autonomously between spectral data from different sources.

According to Huang Yang from UCAS, SpecCLIP acts as a "translator" that can convert LAMOST's low-resolution spectra and Gaia's high-precision spectra into a "universal language." This allows scientists to perform joint analyses with ease, enabling data alignment and transformation across different instruments and survey projects.

According to the study, which has been published in the Astrophysical Journal, SpecCLIP is not a specialist AI model designed for a single task, but a framework close to a foundational model. It can predict stellar atmospheric parameters and elemental abundances in one go, perform spectral-similarity searches, and even help identify peculiar celestial objects.

These capabilities are particularly crucial in the field of Galactic archaeology, holding the promise of sifting through massive datasets efficiently to find extremely rare, metal-poor ancient stars, which would provide key evidence for the study of the early formation and merger history of the Milky Way.

SpecCLIP has already been applied in multiple cutting-edge exploration missions. On one mission searching for planets similar to Earth, for example it has accurately characterized the features of planet-hosting stars, thereby improving the efficiency of screening for potentially habitable planets.