Search for Lepton Flavor Violating Decay τ⁻ → ℓ⁻η at the Belle II Experiment = Belle II 실험에서의 렙톤 맛깔 위반 붕괴 τ⁻ → ℓ⁻η 탐색

국문 초록 (Abstract)

렙톤 맛깔 위반의 관측은 표준모형을 넘어선 새로운 물리현상의 존재를 암시한다. 이 논문에서는 렙톤 맛깔을 위반하는 붕괴인 τ⁻ → ℓ⁻η (여기서 ℓ⁻은 e⁻ 또는 μ⁻) 의 탐색을 다룬다. 이 분석은 비대칭적 에너지 양전자–전자 충돌기 SuperKEKB에 위치한 Belle II 검출기에서 모은 427.9 fb⁻¹의 데이터 샘플을 이용하였다. 이는 약 3.9×10⁸ 타우–반타우 쌍에 해당한다. η 메존은 η → γγ 와 η → π⁺π⁻π⁰로 붕괴하는 채널을 이용하여 재구성되었다. 잠재적인 편향을 피하기위해, 모든 탐색 과정은 신호영역이 가려진 채로 진행되었다. 커트 조건을 적용한 이벤트 선택은 η 의 두 채널에 모두 적용되었고, 다변수선택 기술을 통한 선택은 η→γγ에 추가적으로 적용되어 신호의 효율은 10%로 유지하면서 배경사건들은 거의 0에 가깝게 배제되었다. 사이드밴드 영역의 데이터를 이용한 데이터 기반 방법이 신호영역의 배경사건의 숫자를 추정하는데 사용되었다. 먼저, 탐색의 예상되는 감도를 계산하기 위해 갈래비에 대한 상한을 배경사건만 존재한다는 영가설 하에서 90%의 신뢰도로 추정되었다. 추정된 상한은 다음과 같다. ℬ(τ⁻ → e⁻η) < 4.62 × 10⁻⁸, ℬ(τ⁻ → μ⁻η) < 4.44 × 10⁻⁸. 신호영역을 확인한 결과, 영가설과 비교했을 때 데이터에서 통계적으로 유의한 초과는 관측되지 않았다. 이에 따라 관측된 데이터를 이용하여 90% 신뢰수준에서 갈래비의 상한을 결정하였다. 그 결과 얻어진 상한은 ℬ(τ⁻ → e⁻η) < 9.21×10⁻⁸, ℬ(τ⁻ → μ⁻η) < 4.23×10⁻⁸이다. 전자 채널의 결과는 이전 실험과 일치하였으며, 뮤온 채널에서는 현재까지 가장 엄격한 상한을 달성하였다.

다국어 초록 (Multilingual Abstract)

Observation of lepton flavor violation (LFV) would indicate the presence of new physics beyond the Standard Model. This thesis presents a search for the LFV decay τ⁻ → ℓ⁻η, where ℓ⁻ denotes either e⁻ or μ⁻. The analysis uses 427.9 fb⁻¹ of data collected with the Belle II detector at the SuperKEKB asymmetric-energy e⁺e⁻ collider, corresponding to approximately 3.9 × 10⁸ τ⁺τ⁻ pairs. The η mesons are reconstructed via the decay modes η → γγ and η → π⁺π⁻π⁰. To avoid potential bias, all search procedures are performed with signal regions blinded. A cut-based selection is applied to both η decay channels, and a multivariate analysis technique is additionally employed for the η → γγ channel to suppress backgrounds to nearly zero, while retaining signal efficiencies of about 10%. A data-driven method using sideband data is adopted to estimate the background yields in the signal regions. First, the upper limits on the branching fractions are estimated at the 90% confidence level under the background-only hypothesis to evaluate the expected sensitivity of the searches. The expected upper limits are ℬ(τ⁻ → e⁻η) < 4.62 × 10⁻⁸ and ℬ(τ⁻ → μ⁻η) < 4.44 × 10⁻⁸. After unblinding the signal regions, no significant excess is observed in the data compared to the null hypothesis. Therefore, using the observed data, the upper limits on the branching fractions are determined at the 90% confidence level. The observed upper limits are ℬ(τ⁻ → e⁻η) < 9.21 × 10⁻⁸ and ℬ(τ⁻ → μ⁻η) < 4.23 × 10⁻⁸. For the electron channel, the result is consistent with the previous experimental search. For the muon channel, it represents the most stringent limit to date.

목차 (Table of Contents)

• Abstract i
• 1 Introduction 1
• 1.1 The Standard Model 1
• 1.1.1 Elementary Particles and the Standard Model 1
• 1.1.2 The η Meson and Flavor Symmetry 5
• 1.1.3 Lepton Flavor and the Tau Lepton 5
• 1.2 Limitations of the Standard Model and Physics Beyond the Standard Model 7
• 1.3 Lepton Flavor Violation (LFV) 9
• 1.3.1 Neutrino Oscillations 10
• 1.3.2 BSM Theories 11
• 1.3.3 Current Status of Experimental Searches for LFV in τ Decays 16
• 1.4 Overview 17
• 2 The Belle II Experiment 19
• 2.1 The SuperKEKB Accelerator 19
• 2.2 The Belle II Detector 21
• 2.2.1 Pixel Vertex Detector (PXD) 22
• 2.2.2 Silicon Vertex Detector (SVD) 24
• 2.2.3 Central Drift Chamber (CDC) 25
• 2.2.4 Aerogel Ring-Imaging Cherenkov Detector (ARICH) 27
• 2.2.5 Time-of-Propagation Counter (TOP) 28
• 2.2.6 Electromagnetic Calorimeter (ECL) 30
• 2.2.7 K⁰_L and Muon Detector (KLM) 31
• 2.2.8 Trigger and Data Acquisition 31
• 2.2.9 Software and Computing 34
• 3 Data and Monte Carlo Samples 37
• 3.1 Real Data 37
• 3.2 Monte Carlo Simulation 38
• 3.2.1 Signal Monte Carlo 39
• 3.2.2 Background Monte Carlo 39
• 4 Event Reconstruction 41
• 4.1 Track Selection 41
• 4.2 Neutral (Photon) Selection 42
• 4.3 Skimming 42
• 4.4 Signal Reconstruction 43
• 4.5 Bremsstrahlung Correction 46
• 4.6 MC Yield Correction 47
• 4.7 Data Correction 48
• 4.8 Trigger Selection 48
• 4.9 Fake Signal and Cross-feed 49
• 4.10 Reconstruction Results 52
• 5 Background Suppression 55
• 5.1 Bhabha Suppression 55
• 5.2 η → γγ Decay Mode 56
• 5.2.1 Analysis Region Definition 57
• 5.2.2 Cut-based Preselection 62
• 5.2.3 Gradient Boosted Decision Tree (GBDT) Application 67
• 5.2.4 Data/MC Comparison 79
• 5.2.5 Background Estimation in the Signal Region 85
• 5.2.6 Result 91
• 5.3 η → π⁺π⁻π⁰ Decay Mode 91
• 5.3.1 Analysis Region Definition 92
• 5.3.2 Cut-based Selection 95
• 5.3.3 Signal Region Optimization 99
• 5.3.4 Data/MC Comparison 100
• 5.3.5 Background Estimation in the Signal Region 102
• 5.3.6 Result 107
• 6 Systematic Uncertainty 109
• 6.1 η → γγ Decay Mode 109
• 6.1.1 Particle Identification 109
• 6.1.2 Trigger Efficiency 110
• 6.1.3 Tracking Efficiency 112
• 6.1.4 η Reconstruction 112
• 6.1.5 η Branching Fraction 115
• 6.1.6 BDT Efficiency 115
• 6.1.7 Tau-pair Production Cross-section 117
• 6.1.8 Luminosity 117
• 6.1.9 MC Statistics 117
• 6.1.10 Track Momentum Scale and Photon Energy Bias 117
• 6.1.11 Fake Signal Contribution 118
• 6.2 η → π⁺π⁻π⁰ Decay Mode 118
• 6.2.1 Particle Identification 118
• 6.2.2 Trigger Efficiency 119
• 6.2.3 Tracking Efficiency 120
• 6.2.4 η Reconstruction 120
• 6.2.5 η Branching Fraction 122
• 6.2.6 Tau-pair Production Cross-section 122
• 6.2.7 Luminosity 123
• 6.2.8 MC Statistics 123
• 6.2.9 Track Momentum Scale and Photon Energy Bias 123
• 6.2.10 Fake Signal Contribution 123
• 6.3 Summary of Systematic Uncertainty 123
• 7 Upper Limit Estimation 127
• 7.1 η → γγ Decay Mode 128
• 7.2 η → π⁺π⁻π⁰ Decay Mode 129
• 7.3 Combined Result 129
• 7.4 Summary on the Expectation 131
• 8 Unblinded Results 133
• 9 Conclusion and Prospects 137
• A Additional Materials 141
• A.1 η → γγ Decay Mode 146
• A.1.1 Background Distribution after the Reconstruction 146
• A.1.2 η Mass Window Determination 146
• A.1.3 Details on Cut-based Selection 146
• A.1.4 GBDT 150
• A.1.5 Distributions of Input Variables after BDT Selection 152
• A.1.6 Data/MC Comparison in Sidebands 154
• A.1.7 BDT Bias Check in the Electron Channel 161
• A.1.8 BDT Bias Check in the Muon Channel 162
• A.1.9 ABCD Method 163
• A.2 η → π⁺π⁻π⁰ Decay Mode 165
• A.2.1 Background Distribution after the Reconstruction 165
• A.2.2 Details of Selection 165
• A.2.3 Signal Region Optimization 168
• A.2.4 Data/MC Comparison in Sidebands 169
• A.2.5 Fit Toys 171
• A.3 Systematic Uncertainties 173
• A.3.1 GBDT 173
• Bibliography 175
• Abstract in Korean 185