The universe’s transition from darkness to light was a pivotal moment in its development, known as the Cosmic Dawn. However, despite the most powerful telescopes, we cannot directly observe these earliest stars. This makes determining their properties one of the biggest challenges in astronomy.

An international team of astronomers led by the University of Cambridge has made significant progress in understanding how the first stars and their remnants affected a specific radio signal – the 21-centimeter signal – created by hydrogen atoms filling the gaps between star-forming regions, just a hundred million years after the Big Bang.

By studying this signal, researchers have shown that future radio telescopes will be able to learn about the masses of the earliest stars. Their results were reported in the journal Nature Astronomy.

“This is a unique opportunity to learn how the universe’s first light emerged from the darkness,” said Professor Anastasia Fialkov from Cambridge’s Institute of Astronomy. “The transition from a cold, dark universe to one filled with stars is a story we’re only beginning to understand.”

The 21-centimeter signal provides a rare window into the universe’s infancy. It is influenced by the radiation from early stars and black holes. Researchers have found that this signal is sensitive to the masses of first stars.

Fialkov leads the theory group of REACH (the Radio Experiment for the Analysis of Cosmic Hydrogen). REACH is a radio antenna still in its calibration stage but promises to reveal data about the early universe. The Square Kilometre Array (SKA), under construction, will map fluctuations in cosmic signals across vast regions of the sky.

Both projects are vital in probing the masses, luminosities, and distribution of the universe’s earliest stars. In this study, Fialkov and her collaborators developed a model that makes predictions for the 21-centimeter signal for both REACH and SKA. They found that the signal is sensitive to the masses of first stars.

“We are the first group to consistently model the dependence of the 21-centimeter signal on the masses of the first stars, including the impact of ultraviolet starlight and X-ray emissions from X-ray binaries produced when the first stars die,” said Fialkov. “These insights are derived from simulations that integrate the primordial conditions of the universe, such as the hydrogen-helium composition produced by the Big Bang.”

Radio astronomy relies on statistical analysis of faint signals, unlike optical telescopes like the James Webb Space Telescope, which capture vivid images. REACH and SKA will not be able to image individual stars but will provide information about entire populations of stars, X-ray binary systems, and galaxies.

“It takes a bit of imagination to connect radio data to the story of the first stars, but the implications are profound,” said Fialkov.

The predictions made in this study have huge implications for understanding the nature of the very first stars in the universe. Researchers show evidence that their radio telescopes can tell us details about the mass of those first stars and how these early lights may have been very different from today’s stars.

Radio telescopes like REACH are promising to unlock the mysteries of the infant Universe, and these predictions are essential to guide the radio observations being done from the Karoo in South Africa. The research was supported in part by the Science and Technology Facilities Council (STFC), part of UK Research and Innovation (UKRI).

The Milky Way’s central region has long been a subject of fascination for astronomers, but recent research led by Dr. James De Buizer at the SETI Institute and Dr. Wanggi Lim at IPAC at Caltech has revealed a surprising finding: massive star formation is occurring in this area at a lower rate than expected. The study primarily relied on observations from NASA’s retired SOFIA airborne observatory, focusing on three star-forming regions – Sgr B1, Sgr B2, and Sgr C – located at the heart of the Galaxy.

Contrary to previous assumptions that star formation is likely depressed near the Galactic Center, these areas have been found to produce stars with relatively low masses. Despite their dense clouds of gas and dust, conditions typically conducive to forming massive stars, these regions struggle to create such high-mass stars. Furthermore, they appear to lack sufficient material for continued star formation, suggesting that only one generation of stars is produced.

The researchers discovered over 60 presently-forming massive stars within the Galactic Center regions, but found that these areas formed fewer stars and topped out at lower stellar masses than similar-sized regions elsewhere in the Galaxy. The team’s study also suggested that extreme conditions in the Galactic Center, such as its rapid rotation and interaction with older stars and material falling towards the black hole, might be inhibiting gas clouds from forming stars.

However, Sgr B2 was found to be an exception among the studied areas, maintaining a reservoir of dense gas and dust despite having an unusually low rate of present massive star formation. The researchers proposed that this region may represent a new category of stellar nursery or challenge traditional assumptions about giant H II regions hosting massive star clusters.

The study’s findings have significant implications for our understanding of star formation in the Milky Way, highlighting the importance of continued research into the complex dynamics at play within the Galactic Center.

The research team leveraged high-throughput computing capabilities provided by the Center for High Throughput Computing (CHTC) to automate computing tasks across a network of thousands of computers. This innovation allowed them to analyze millions of simulations, making it possible to extract new insights from the data behind the Event Horizon Telescope images of black holes.

The neural network was trained on synthetic data files generated by CHTC, enabling the researchers to make a better comparison between the EHT data and models. The analysis revealed that the emission near the black hole is mainly caused by extremely hot electrons in the surrounding accretion disk, rather than a jet. Additionally, the magnetic fields in the accretion disk appear to behave differently from usual theories of such disks.

Lead researcher Michael Janssen stated that defying prevailing theory is exciting but sees their AI and machine learning approach as a first step towards further improvement and extension of associated models and simulations. The research has significant implications for our understanding of black holes and the cosmos, and it will be interesting to see how this knowledge evolves in the future.

Content:

An international team of astronomers has made groundbreaking discoveries about the black hole at the center of our Milky Way using a neural network. By analyzing millions of synthetic simulations generated by the Center for High Throughput Computing (CHTC), they found that the black hole is spinning at nearly top speed, with its rotation axis pointing towards Earth.

The research team published their findings in three papers in Astronomy & Astrophysics, providing new insights into the behavior of black holes. The neural network was trained on synthetic data files generated by CHTC, enabling the researchers to make a better comparison between the Event Horizon Telescope (EHT) data and models.

Previous studies by the EHT Collaboration used only a handful of realistic synthetic data files, but the Madison-based CHTC enabled the astronomers to feed millions of such data files into a so-called Bayesian neural network. This allowed them to extract as much information as possible from the data and make a more accurate comparison with the models.

The researchers found that the emission near the black hole is mainly caused by extremely hot electrons in the surrounding accretion disk, rather than a jet. Additionally, the magnetic fields in the accretion disk appear to behave differently from usual theories of such disks.

Lead researcher Michael Janssen stated that defying prevailing theory is exciting but sees their AI and machine learning approach as a first step towards further improvement and extension of associated models and simulations. The research has significant implications for our understanding of black holes and the cosmos, and it will be interesting to see how this knowledge evolves in the future.

The Event Horizon Telescope project performed more than 12 million computing jobs in the past three years, using the Open Science Pool operated by PATh. This pool offers computing capacity contributed by more than 80 institutions across the United States, making it an ideal platform for large-scale simulations like those used in this research.

Scientific papers referenced

* Deep learning inference with the Event Horizon Telescope I: Calibration improvements and a comprehensive synthetic data library. By: M. Janssen et al. In: Astronomy & Astrophysics, 6 June 2025.
* Deep learning inference with the Event Horizon Telescope II: The Zingularity framework for Bayesian artificial neural networks. By: M. Janssen et al. In: Astronomy & Astrophysics, 6 June 2025.
* Deep learning inference with the Event Horizon Telescope III: Zingularity results from the 2017 observations and predictions for future array expansions. By: M. Janssen et al. In: Astronomy & Astrophysics, 6 June 2025.