Monday, February 07, 2022

Gigapixel radio image of the Universe using Europe as a radio telescope

An illustration of the difference in resolving power between using only the Dutch LOFAR stations and all of the international stations throughout Europe, arranged in the formation of an HBA core stations. The animation fades from an angular resolution of 6" to a resolution of 0.3". Credit: Frits Sweijen

An international team of astronomers has created one of the largest and most detailed radio maps at megahertz frequencies thanks to Dutch supercomputers. This research, led by Frits Sweijen at Leiden University, has been published in Nature Astronomy on Thursday. Using the International LOFAR Telescope they have, in the wake of tremendous progress late last year, mapped an area of the sky the size of 25 full moons in great detail, with a resolving power comparable to optical telescopes on Earth. The resulting radio image contains nearly seven billion pixels and contains just shy of 2500 radio galaxies.

LOFAR

The sky is filled with radiation invisible to the naked eye, including radio waves at frequencies ten million times lower than red light. With tens of thousands of antennas across Europe, the LOFAR telescope listens to those cosmic radio waves at a frequency of 144 MHz, just above the FM radio band. Through these antennas the European continent transforms into an almost 2000 km big radio telescope. This tremendous size means that LOFAR can see exquisite and unprecedented detail at such low radio frequencies, with a resolving power high enough to make out the Great Pyramid if it were on the Moon. The combined area of all the antennas make it sensitive enough to detect a mobile phone ringing all the way out on Mars.

This resolving power is challenged by Earth's atmosphere however. Ultraviolet radiation from the Sun creates a layer of charged particles in the upper atmosphere. This so-called ``ionosphere'' distorts radio waves from space before they reach the telescope. For LOFAR it therefore is like looking at the sky from the bottom of the ocean. With advanced techniques these distortions can be corrected, focusing the telescope across its entire field of view, thus allowing it to be mapped.

Super computers

Determining these corrections and subsequently converting radio waves into an image, requires modern algorithms and a lot of compute power. Thanks to super computers this was not an issue. Locally, the Academic Leiden Interdisciplinary Cluster Environment (ALICE) lent its power to the scientists. Nationally, SURF in the Netherlands, provided early access to the new platform data processing platform named Spider. This platform is specifically designed for data-intensive projects like this. Lengthy calculations could be run in a massively parallel fashion thanks to these supercomputers. The final image was too large to be made in one go. To image the full field of view, it was processed in 25 smaller chunks each covering an area the size of a full moon. Each of these chunks was turned into an image over seven days, using software recently developed at ASTRON. Piece by piece, on a single computer, this process would have taken more than 175 days to complete. Thanks to the large scale compute infrastructure at Leiden and SURF, however, it only took seven days effectively.


The sharp eyes of the International LOFAR Telescope allow scientists to study the evolution of black holes and their host galaxies in more detail than before. Galaxies in the early Universe, for example, that would otherwise be too small to resolve due to their distance or young age, can have their spatial structure studied. The published results enable this for thousands of sources at once. With its near seven billion pixels this single image contains almost as many pixels as radio surveys from the past did covering the entire sky. These new results explore a tip of the ice berg with a detailed map of the entire Northern sky as the future goal.

This work made use of the Dutch national e-infrastructure with the support of the SURF Cooperative using grant no. EINF-251; the ERC Starting Grant ClusterWeb 804208; the Medical Research Council grant MR/T042842/1; the UK STFC ST/R000972/1 and ST/V000594/1 grants and the Academic Leiden Interdisciplinary Cluster Environment (ALICE) provided by Leiden University.