Tagging $b$ quarks without tracks using an Artificial Neural Network algorithm

TL;DR

This paper introduces an ANN-based method for tagging $b$ quarks in high energy physics experiments, especially effective for highly boosted $B$ hadrons, improving discrimination over traditional algorithms.

Contribution

The study extends previous hit multiplicity methods by incorporating an artificial neural network, significantly enhancing $b$ quark tagging performance in challenging conditions.

Findings

ANN improves rejection of light-quark and charm jets

Performance remains robust with multiple $pp$ interactions

Significantly increases tagging significance

Abstract

Pixel detectors currently in use by high energy physics experiments such as ATLAS, CMS, LHCb, etc., are critical systems for tagging $B$ hadrons within particle jets. However, the performance of standard tagging algorithms begins to fall in the case of highly boosted $B$ hadrons ($\gamma \beta = p/m >200$). This paper builds on the work of our previous study that uses the jump in hit multiplicity among the pixel layers when a $B$ hadron decays within the detector volume. First, multiple $pp$ interactions within a finite luminous region were found to have little effect. Second, the study has been extended to use the multivariant techniques of an artificial neural network (ANN). After training, the ANN shows significant improvements to the ability to reject light-quark and charm jets; thus increasing the expected significance of the technique.