An AI-driven tool has the capability to identify the elusive effects of dark matter, potentially bringing us closer to understanding this mysterious component of the universe.
Dark matter is thought to be the unseen force that binds the universe together. It constitutes about 85% of all matter and roughly 27% of everything the universe contains. However, since it is invisible, scientists study the gravitational impacts it has on galaxies and other cosmic structures. Despite years of investigation, the precise nature of dark matter continues to be one of the biggest unanswered questions in science.
A prominent theory suggests that dark matter could be a type of particle that has minimal interactions with other particles, interacting primarily through gravity. Some researchers, however, propose that these particles might occasionally engage with one another, a phenomenon referred to as self-interaction. Detecting such interactions could provide vital insights into the characteristics of dark matter.
Yet, one major obstacle has been differentiating the faint signs of dark matter self-interactions from other cosmic phenomena, such as those caused by active galactic nuclei (AGN)—the supermassive black holes found in the centers of galaxies. The feedback from AGN can influence matter in ways that mimic dark matter effects, complicating the task of distinguishing between the two.
In a notable advancement, astronomer David Harvey from EPFL’s Laboratory of Astrophysics has created a deep-learning algorithm capable of sorting through these intricate signals. The AI-based approach is aimed at separating the impacts of dark matter self-interactions from the effects of AGN feedback by examining images of galaxy clusters—large groups of galaxies held together by gravity. This development is expected to significantly improve the accuracy of dark matter research.
Harvey utilized a Convolutional Neural Network (CNN), which is a specialized AI technique excellent at pattern recognition in images. He trained it using pictures from the BAHAMAS-SIDM project, which simulates galaxy clusters under varying conditions of dark matter and AGN feedback. By analyzing thousands of simulated galaxy cluster images, the CNN learned to differentiate the signals resulting from dark matter self-interactions from those caused by AGN feedback.
Among the different CNN models tested, the most sophisticated one, known as “Inception,” turned out to be the most precise. The AI was trained with two main dark matter scenarios featuring different degrees of self-interaction and was validated on additional models, including a more intricate velocity-dependent dark matter model.
Inception achieved an impressive 80% accuracy in ideal conditions, successfully identifying whether galaxy clusters were affected by self-interacting dark matter or AGN feedback. It retained this high performance even when realistic observational noise was introduced, simulating the data expected from future telescopes like Euclid.
This indicates that Inception—and the AI methods as a whole—could be incredibly beneficial for analyzing the vast volumes of data collected from space. Moreover, its ability to adapt to previously unseen data shows that it is both flexible and dependable, positioning it as a valuable tool for future dark matter studies.
AI techniques like Inception could greatly enhance our comprehension of what dark matter truly is. As new telescopes gather extraordinary amounts of data, these methods will assist scientists in processing that information swiftly and accurately, potentially uncovering the realities of dark matter.
