The heliosphere’s boundary: a neural network provides a more accurate reconstruction of the maps of energetic neutral atoms from the future HERMA imager

On simulated data, convolutional network inversion outperforms the conventional algorithm, particularly when the signal is weak and the exposure time is short, paving the way for rapid mapping of the entire sky.

The energetic neutral atoms (ENA) that come to us from the heliosphere boundary provide information on how the solar wind collides with the interstellar medium. The HERMA (Helios ENA Rapid Mapper) project is preparing a new type of instrument covering a wide range of energies with high spatial resolution, based on a modulation grid and a 256-pixel silicon linear detector.

To test its ability to quickly produce images of the entire sky, the team modelled the detector's response with the Geant4 simulation tool and then compared the traditional inversion algorithm to a convolutional neural network, using data from the Cassini mission's INCA instrument as a reference. Regardless of the sampling time, the distribution reconstructed by the array sticks more closely to the expected source, and the gap widens in its favour when the signal-to-noise ratio is low and the observation time is short.

The authors see it as a technical building block for the validation of the instrument and its operation in orbit.