SciPost Submission Page
Tau identification in CMS during LHC Run 2
by Mohammad Hassan Hassanshahi on behalf of the CMS Collaboration
Submission summary
| Authors (as registered SciPost users): | Mohammad Hassan Hassanshahi |
| Submission information | |
|---|---|
| Preprint Link: | scipost_202111_00040v1 (pdf) |
| Date accepted: | 2024-12-09 |
| Date submitted: | 2021-11-22 11:09 |
| Submitted by: | Hassanshahi, Mohammad Hassan |
| Submitted to: | SciPost Physics Proceedings |
| Proceedings issue: | 16th International Workshop on Tau Lepton Physics (TAU2021) |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approach: | Experimental |
Abstract
The LHC Run 2 data-taking period was characterized by an increase in instantaneous luminosity and center-of-mass energy. Several techniques have been deployed in the CMS experiment to reconstruct and identify tau leptons in this environment. The Deep-Tau identification algorithm is used to identify hadronically decaying tau leptons from quark and gluon induced jets, electrons, and muons. Compared to previously used MVA identification algorithms, the use of deep-learning techniques brought a noticeable improvement in the tau identification and rejection of contaminating sources. Low transverse momentum topologies were addressed separately with a dedicated identification algorithm, while machine learning techniques were implemented to improve the identification of the tau hadronic decay channels. These algorithms have been already used for several published physics analyses in CMS. The algorithms are presented together with their measured performances.
Current status:
Editorial decision:
For Journal SciPost Physics Proceedings: Publish
(status: Editorial decision fixed and (if required) accepted by authors)
Reports on this Submission
Report #1 by Swagato Banerjee (Referee 1) on 2024-11-27 (Invited Report)
Report
Three techniques to identify hadronically decaying tau lepton produced in the CMS detector are presented and their performances are studied using Run 2 data.
The first one uses a deep convolutional neural network, the second uses a boosted decision tree, and the third uses an attention-based graph neural network which is suitable to reconstruct three-charged-prong hadronic decays in the low transverse-momentum regime. These provide a great opportunity for measuring the Standard Model parameters and probing beyond the Standard Model physics, in particular in processes with tau leptons in the final state.
Recommendation
Publish (surpasses expectations and criteria for this Journal; among top 10%)