Search for New Phenomena in Two-Body Invariant Mass Distributions Using Unsupervised Machine Learning for Anomaly...

Area of Science:

• High Energy Physics
• Particle Physics
• Collider Physics

Background:

• The Standard Model of particle physics has been highly successful but does not explain phenomena such as dark matter and neutrino masses.
• Searches for new physics beyond the Standard Model are crucial for understanding fundamental particles and forces.
• The ATLAS experiment at the Large Hadron Collider (LHC) collects vast amounts of data from high-energy particle collisions.

Purpose of the Study:

• To search for new, unexpected resonances in particle collision data.
• To apply an unsupervised anomaly-detection technique to identify potential new physics signals.
• To set limits on the properties of hypothetical new resonances in specific decay channels.

Main Methods:

• Utilized an unsupervised anomaly-detection technique, specifically an autoencoder, to identify unusual events.
• Analyzed 140 inverse femtobarns of proton-proton collision data at 13 TeV recorded by the ATLAS detector.
• Focused on invariant mass spectra of object pairs including leptons, photons, and jets in regions identified as anomalous.

Main Results:

• No significant deviations from the expected background processes were observed in the analyzed data.
• The anomaly-detection method successfully identified regions for targeted investigation.
• Exclusion limits were placed on the contributions of generic Gaussian signals across nine invariant mass spectra.

Conclusions:

• The current dataset, analyzed with this unsupervised method, does not show evidence for new resonances.
• The study demonstrates the utility of anomaly detection for exploring new physics in collider data.
• Further investigation with larger datasets or refined anomaly-detection techniques may be warranted.