Hunting for Dusty Trails: Ten New Exocomet Transits Discovered in Kepler Data

The Kepler Space Telescope, which observed over 200,000 stars between 2009 and 2013, continues to reveal new celestial secrets even more than a decade after its mission ended. In a new study, P. Dumond and collaborators from the Institut d’Astrophysique de Paris used machine learning to search for exocomet transits, dips in starlight caused when comets from other planetary systems pass in front of their host stars. These transits are challenging to detect because they are faint, irregular, and often occur only once, but their discovery helps astronomers understand how comets behave around other stars and how planetary systems evolve over time.

Exocomets and Their Unusual Signatures

Exocomets, like comets in our Solar System, are icy or rocky bodies that produce long, bright tails when heated by their stars. When they transit, or pass in front of, their host star, they cause a characteristic asymmetric dip in brightness: the starlight drops sharply as the comet’s nucleus passes by, then recovers slowly as its dusty tail trails behind. This distinct shape was first predicted in the late 1990s and confirmed in systems such as Beta Pictoris and 49 Ceti. Detecting these faint events requires precise photometry at levels of one part in ten thousand, a capability first achieved by space missions like Kepler and TESS.

Teaching a Neural Network to Find Comets

To locate potential exocomet transits within Kepler’s massive archive, Dumond’s team trained a neural network, a type of artificial intelligence designed to recognize patterns. The model was taught using thousands of synthetic light curves, some containing simulated comet transits and others showing normal stellar brightness variations. By learning from these examples, the network could later distinguish comet-like dips from random noise. The researchers further improved the model with special computational layers that help detect single, non-repeating transits, since comets rarely pass in front of their stars more than once within Kepler’s observation window.

Testing and Filtering the Results

After training, the neural network examined light curves for over 200,000 stars, slicing each into overlapping 10-day windows. The initial search produced tens of thousands of possible detections, most of them false alarms caused by instrumental noise or stellar variability. The team used mathematical filters and physical models to weed out these false positives, keeping only those events with the right asymmetrical shapes and stable baselines. Additional checks removed cases caused by spacecraft glitches or known exoplanets. The final candidate list was then visually inspected by the authors to confirm the most promising signals.

Building a New Catalog of Exocomets

From this exhaustive process, the researchers produced three catalogs of decreasing confidence. The first tier lists 17 high-confidence exocomet transits, seven already known from earlier work and ten entirely new detections. Each of these comes from a different star. A second tier includes 30 lower-confidence events, and a third lists 49 symmetrical transits that might instead be caused by long-period exoplanets or binary stars. Interestingly, Dumond’s team found that exocomet activity is not limited to young stars as once thought; several detections were made around older and even red-giant stars, suggesting that cometary activity may persist longer in a star’s life than previously believed.

What These Discoveries Mean

The study demonstrates that combining physics-based simulations with artificial intelligence can uncover subtle astrophysical signals hidden in archival data. The detection of ten new exocomet transits expands the known population and challenges the long-standing view that comet activity fades quickly as stars age. While the faintness of the Kepler targets prevents follow-up spectroscopy, similar methods can now be applied to newer missions like TESS and PLATO, which observe brighter stars. As Dumond and collaborators note, these future searches could reveal more about the dusty debris and icy bodies orbiting distant suns, shedding light on how planetary systems like our own evolve over billions of years.

Source: Dumond