Sallyportal: Madly Blogging Reed

Physicists use AI to Probe Mystery of Subatomic Jets

Following a collision of two protons, a Higgs boson is produced which decays into two jets of hadrons and two electrons. The lines represent the possible paths of particles produced by the proton-proton collision in the detector while the energy these particles deposit is shown in blue. [Description from CERN.] Lucas Taylor / CERN

Several thousand miles away, on the Franco-Swiss border, scientists at the Large Hadron Collider smash protons together at near-light speed, producing streams of highly energetic particles known as jets. Much closer to home, Reed’s own newly minted alumnus Kaustuv Datta ’17 studies data from these jets to figure out new ways of extracting information from them. In fact, Kaustuv and his advisor Professor Andrew Larkoski had so much success in analyzing these subatomic streams that results from his thesis were recently published in the Journal of High Energy Physics in an article titled “How much information is in a jet?”

Kaustuv offered this analogy to help you understand his research: imagine trying to tell a computer the difference between an object that is a cat and one that is not a cat. You would begin by telling the computer features of the cats (whiskers, paws, tails, etc.) until it knew enough about cats in general to identify any cat. These distinctive features are known in the physics world as “observables.” Kaustuv’s research revealed that a computer only needs a finite set of these observables to identify a jet. In other words, there comes a certain point when any more descriptors of the cat become redundant—and only slow the analysis down.

Kaustuv Datta ’17 published a paper in the Journal of High Energy Physics on extracting information from subatomic jets.

Jets are important to theoretical physicists because they contain information about the bizarre subatomic particles that are briefly liberated by the collisions. Although they only persist for a few microseconds, these particles represent elemental states of matter. By studying them, physicists hope to learn more about the structure of the universe.

The Reed physics department recently demonstrated its interest in this line of research--and its confidence in Kaustuv--by buying an NVIDIA DevBox, a supercomputer with deep learning capabilities that will doubtless prove useful to students in physics, math, and computer science.

Kaustuv’s career in physics is off to an incredible start, including a stint on location at CERN and a published paper. He is extremely grateful to all of his professors and advisors in the physics department for all the support they have provided for him. Without them, Kaustuv insists, none of this would have been possible.