Neutrinos are one of the most enigmatic elements of the Standard Model, a framework for understanding natural basic forces and particles. While they are among the most numerous known particles in the cosmos, they only interact with matter infrequently, making their discovery a difficult experiment. The Mini Booster Neutrino Experiment (MiniBooNE), which lasted from 2002 to 2017 at the Fermi National Accelerator Laboratory, or Fermilab, in Illinois, solved one of the long-standing riddles in neutrino physics. MiniBooNE discovered substantially more neutrino-electron interactions than one would predict based on our current understanding of the Standard Model, and scientists are trying to figure out why.
In 2007, researchers came up with the idea for MicroBooNE, a follow-up experiment that just concluded collecting data at Fermilab. MicroBooNE is an excellent test of the MiniBooNE excess because it employs an unique detector technology called the liquid argon time projection chamber (LArTPC), which produces high-resolution images of the particles produced by neutrino interactions.
Nicholas Kamp and Lauren Yates, both of the MIT Laboratory for Nuclear Science, as well as Professor Janet Conrad, both performed key roles in MicroBooNE’s deep-learning-based search for an excess of neutrinos in the Fermilab Booster Neutrino Beam. Kamp explores the future of the MiniBooNE anomaly in light of MicroBooNE’s newest results in this interview.
Q: What’s the big deal about the MiniBooNE anomaly?
A: The existence of a hypothetical particle known as the “sterile neutrino” is one of the main unanswered problems in neutrino physics. Finding a new particle would be huge because it may lead to fresh insights into the bigger theory that explains all of the particles we observe. The inclusion of such a sterile neutrino to the Standard Model is the most prevalent explanation for the MiniBooNE excess. Because of the effects of neutrino oscillations, this sterile neutrino would appear in MiniBooNE as an increase in electron neutrinos.
There are a slew of other abnormalities in neutrino physics that point to the existence of this particle. However, explaining these anomalies, as well as MiniBooNE, with a single sterile neutrino is difficult—the whole picture doesn’t match. Our research group at MIT is looking at new physics models that could be able to explain the whole picture.
Q: What do we know about the MiniBooNE surplus right now?
A: Thanks to recent advances in both the experimental and theoretical domains, our knowledge has advanced tremendously.
Our team collaborated with physicists from Harvard, Columbia, and Cambridge universities to investigate novel photon sources that may exist in a theoretical model with a 20% electron signature. We devised a “mixed model” with two sorts of exotic neutrinos: one that transforms into electron flavor and the other that decays into a photon. Physical Review D will publish this research soon.
More recent MicroBooNE studies, which included a deep-learning-based study in which our MIT group played a key role, found no excess of neutrinos that create electrons in the MicroBooNE detector. Considering the accuracy of MicroBooNE’s measurements, this shows that the MiniBooNE excess cannot be wholly explained to additional neutrino interactions. If electrons aren’t the culprit, photons are the only particle capable of producing a comparable signal in MiniBooNE. However, because photons created by interactions are limited to a low level, we are certain it is not photons produced by interactions that we are aware of. As a result, they must be originating from a novel source, such as the exotic neutrino decay in the mixed model. MicroBooNE is now working on a search that will allow them to isolate and identify these extra photons. Keep an eye out!
Q: You said that your group is working on MicroBooNE analysis using deep learning. Why do neutrino physicists utilize deep learning?
A: Humans can discern the difference between species of cats easily when looking at photographs of cats. Similarly, researchers can determine the difference between the particles created in neutrino interactions without problem when looking at photos from a LArTPC. Both objectives, however, turn out to be tough for standard algorithms due to the complexity of the distinctions.
MIT is a crossroads for deep-learning concepts. It is now home to the National Science Foundation AI Institute for Artificial Intelligence and Fundamental Interactions, for example. Building on the significant local experience in the subject made sense for our group. We’ve also had the pleasure of collaborating with outstanding groups at SLAC, Tufts University, Columbia University, and IIT, all of which have extensive experience with the connections between deep learning and neutrino physics.
The concept of a “neutral network,” which is an algorithm that makes judgments (such as detecting particles in a LArTPC) based on past exposure to a set of training data, is one of the most important concepts in deep learning. Our group published the first publication in neutrino physics on particle identification using deep learning, demonstrating that it is a strong approach. This is one of the main reasons why the MicroBooNE deep learning-based analysis results, which were just disclosed, impose severe limitations on an electron neutrino interpretation of the MiniBooNE excess.