Machine learning learns to listen to the sun’s acoustic heartbeat

For decades, solar physicists have tracked pressure waves that travel through the sun’s interior, bounce off its core, and return to the surface carrying acoustic imprints of what lies beneath. These waves, called p-modes, behave somewhat like the seismic pulses that geologists use to map the interior of the Earth. They carry structural information that no telescope could ever observe directly. Now, a research team led by Dr. Rekha Jain has demonstrated that machine learning can decode those waves with a precision and speed no previous method achieved. The study, published in Solar Physics, represents a meaningful expansion in how scientists track the solar cycle and anticipate its most energetic phases.

What p-modes reveal

P-modes are not silent. They produce measurable oscillations on the sun’s surface, and their frequency shifts in a recognizable pattern that follows the 11-year solar cycle. When the sun is quiet, the pitch of these waves behaves one way. When magnetic activity climbs toward solar maximum, with sunspots multiplying and coronal mass ejections becoming more frequent, that acoustic signature changes. Scientists have accumulated roughly thirty years of p-mode observations, a dataset rich enough to train an artificial intelligence system to find patterns that human analysis would struggle to isolate.

The AI model examined those three decades of frequency shift data and learned to connect what happens acoustically at the surface to what is occurring magnetically in the sun’s deeper layers. What makes this significant is that those magnetic changes begin building up in the interior long before they become visible in surface phenomena like flares or ejections. If the acoustic signal can be read accurately enough, it offers an early warning that precedes the disruptions themselves.

The connection to space weather

Solar storms are not purely astronomical concerns. When a coronal mass ejection reaches Earth, it interacts with the magnetosphere and can induce currents capable of damaging power infrastructure, interrupt satellite communications, and affect the positioning systems that aviation, shipping, and financial networks depend on. The more advance notice operators of those systems receive, the more they can do to reduce exposure.

Current space weather forecasting relies on a combination of direct solar observations, magnetic field measurements, and models of how eruptions propagate through the inner heliosphere. Helioseismology has historically been a tool for solar interior science rather than an operational forecasting resource. This study argues that it can become both. By establishing that AI can reliably characterize p-mode frequency shifts and project when the current cycle will enter its quieter phase, the team is proposing that acoustic data become an independent forecasting indicator alongside the traditional observational tools.

Solar Cycle 25 in focus

The research team applied their model specifically to Solar Cycle 25, the cycle that began in December 2019 and has been tracked closely by the international solar physics community. Cycle 25 has proven more active than many forecasters initially expected, with elevated sunspot counts and several significant geomagnetic storms already on record. Understanding when its peak will pass and how its acoustic signature will evolve toward solar minimum is now a question with direct operational relevance.

The ability to predict the quieter phase from acoustic data alone, without relying exclusively on surface magnetic observations, adds a valuable independent data stream to that effort. In heliophysics, where forecasting depends on the convergence of multiple independent signals, a new kind of evidence is always worth having.

Listening as a scientific method

Dr. Jain described the approach in terms that connect the technical to the immediate: by using machine learning to follow the acoustic heartbeat of the sun, the team is attempting to trace the energy that moves from the deep interior toward the surface and then out into space, where it eventually affects the technology and communications infrastructure that modern society depends on.

That framing matters. Helioseismology began as a discipline concerned with understanding the sun for its own sake, mapping the speed of sound at different depths, inferring rotation rates in layers no instrument could ever reach directly. This study does not abandon that scientific ambition, but it extends the method into territory with applied consequences. The acoustic interior of the sun and the geomagnetically active near-Earth environment are now, through machine learning, a little more directly connected than they were before.

The publication details: Rekha Jain et al., Machine Learning–Based Characterization of Solar p-Mode Frequency Shifts During Solar Cycle 25, Solar Physics (2026). DOI: 10.1007/s11207-026-02660-y

© 2026 SKYCR.ORG | Homer Dávila Gutiérrez, FRAS. All rights reserved. Total or partial reproduction is prohibited without express authorization. Original source: Solar Physics, DOI: 10.1007/s11207-026-02660-y