AI separates Mars’s oxygen beams from the background

A 22 October preprint uses U-Net to extract energetic O₂⁺ beams from MAVEN’s 2015–2025 observations. These charged oxygen molecules form a distinct population escaping Mars.[1]

MAVEN had already detected escaping ions near the start of its mission. NASA’s December 2014 account described its STATIC instrument measuring relatively cold ionised gas low in the atmosphere and energised particles higher up. Some gain enough energy to overcome the planet’s gravity. The ionosphere is a reservoir of charged particles; understanding escape means following how those particles acquire energy and leave it.[2]

U-Net was originally developed to identify structures within biomedical images. It assigns labels to individual pixels, using a contracting sequence of layers to capture broad context and an expanding sequence to recover detail. Connections between the two preserve fine features. The output is a map of the image’s contents: it can mark a boundary or a region, where a single label for the entire image would discard their locations.[3]

Here the image plots time against ion energy. Human-labelled spectrograms from 200 orbits supplied training and testing data. Requiring reliable field directions retained 14,484 of 36,888 detected beam intervals. Beams favoured the side where the solar wind’s electric field points outwards.[1]

An earlier MAVEN study by Yaxue Dong and colleagues helps separate two physical processes. Solar ultraviolet light creates ions by removing electrons from atmospheric gas. Interactions with the solar wind supply energy that can accelerate those ions. The study found that responses to ultraviolet radiation varied with ion species, energy and escape route. Producing more charged particles and accelerating the existing particles are therefore different parts of the problem.[4]

Report an error or suggest a correction