[Paper Review] Reconstruction and identification of hadronic decays of tau leptons in ATLAS

Zinonas Zinonos|arXiv (Cornell University)|Sep 1, 2014

Particle physics theoretical and experimental studies2 references3 citations

TL;DR

This paper presents the optimized reconstruction and identification algorithms for hadronic tau decays in the ATLAS experiment during the 2012 LHC run, using a boosted decision tree (BDT) with multi-variate discriminants to achieve high rejection of jets (up to 500) and electrons (over 100), with data-driven validation showing good agreement between simulation and data. The method enables precise tau identification for Higgs and new physics searches at high pile-up conditions.

ABSTRACT

Hadronically decaying tau leptons are of prime importance in numerous physics analyses in ATLAS. The spectrum of the possible applications of hadronically decaying tau leptons reaches from Standard Model measurements, including Higgs searches, to searches for physics beyond the Standard Model. The basic principles behind the sophisticated tau reconstruction and identification techniques, which are specifically designed to identify hadronically decaying taus and reject various background processes, are delineated here along with current data-driven estimates of their respective performance.

Motivation & Objective

• To develop and optimize reconstruction and identification algorithms for hadronically decaying tau leptons in ATLAS under high pile-up conditions during the 2012 LHC run.
• To improve background rejection for jets and electrons misidentified as hadronic taus, especially in high-luminosity environments.
• To validate the performance of identification algorithms using data-driven techniques, ensuring accurate simulation scale factors.
• To provide calibrated efficiency measurements and scale factors for use in physics analyses involving tau decays.
• To enable precise measurements in Standard Model processes and searches for new physics involving hadronic tau final states.

Proposed method

• Reconstruction of hadronic tau candidates using anti-k_t jet clustering with R=0.4, followed by barycenter formation and energy calibration based on clusters within ΔR<0.2.
• Identification using a boosted decision tree (BDT) trained separately on 1-prong and 3-prong decays, combining tracking and calorimeter variables such as energy fraction in ΔR<0.1 cone and track distance from axis.
• Application of three working points (loose, medium, tight) with predefined signal efficiencies (70%, 60%, 40% for 1-prong; 65%, 55%, 35% for multi-prong) to balance signal acceptance and background rejection.
• Electron rejection via a dedicated BDT using transition radiation and shower shape differences, trained on Z→ττ (signal) and Z→ee (background) simulated events.
• Data-driven efficiency measurements using the tag-and-probe method in Z→τℓτhad, W→τν, and t¯t→τhad+jets channels, with extended track multiplicity as the discriminating variable.
• Scale factors derived by comparing data and simulation efficiencies, with uncertainties estimated via systematic template variations.

Experimental results

Research questions

• RQ1 How can hadronic tau reconstruction and identification be optimized for high pile-up conditions in the ATLAS experiment?
• RQ2 What level of jet and electron background rejection can be achieved while maintaining high signal efficiency for hadronic tau decays?
• RQ3 How well do simulation-based identification efficiencies match those measured in actual 2012 LHC data?
• RQ4 What scale factors are required to correct simulation-based efficiencies for use in physics analyses?
• RQ5 What is the performance of the BDT-based identification algorithm across different tau decay modes and transverse momentum regions?

Key findings

• The BDT-based identification algorithm achieves jet background rejection factors of 10–40 at 70% signal efficiency and up to 500 at 35% signal efficiency for 1-prong and multi-prong decays.
• Electron misidentification rejection exceeds 100 with the tight working point, as validated by data-driven measurements.
• Data-driven efficiency measurements in Z→τℓτhad, W→τν, and t¯t→τhad+jets channels show good agreement between data and simulation for all working points.
• Scale factors for tau identification and electron veto efficiencies are provided with associated uncertainties, enabling correction of simulated samples in physics analyses.
• The extended track multiplicity variable in the 0.2<ΔR<0.6 annulus provides effective separation between signal and background, used in fitting templates for efficiency extraction.
• The Winter 2013 BDT version, incorporating π⁰-related variables, improves performance, particularly in jet rejection.