In the ATLAS detector at the Large Hadron Collider, particles leave tracks like skiers in the snow. Some tracks break off after a couple of centimeters: that's what a particle would look like if it were born, flew a little way, and then decayed into something invisible. Physicists were looking for exactly such 'vanishing tracks' in the data from 13 TeV proton collisions.

The search relied on distinctive signatures: a track of three or four points in the first layers of the detector, a powerful jet of particles nearby, and a large missing transverse momentum (carried away by something invisible). To distinguish a real signal from random coincidences, machine learning and specialized background estimation methods were used.

This result is like combing a forest: no tracks of the sought-after beast, but the map of the area is now more accurate. Future searches will know where not to look and can focus on heavier or longer-lived particles.