Testing the Bell inequality for Higgs boson decay With supervised learning Multi-Variate Analysis classifiers

Abstract

The Higgs boson was discovered in the ATLAS and CMS detectors at LHC in 2012. Since then, more and more have been learned about its properties including its mass. Bell introduced a test to verify if particles behaved quantum mechanically or classically. This was later modified to be able to test a three-level system such as the Higgs boson decay to W bosons. The data from ATLAS open data was used to train neural networks to classify these events and filter out the background on simulated data and then use these to filter out the actual data. This was then tested in the generalized Bell inequality for a three-level system. It was found that the actual data filtered out by the network broke the inequality with over 4σ but was far away from the expected value. This can be due to precision lower than for the training and validation data because of differences between simulations and actual data but also because the assumptions of the Higgs boson being at rest in the lab frame were insufficient.