Lepton Flavor Violation at LHCb The τ → μμμ decay Matt Jaffe Physics Department, Penn State University

The LHCb Experiment Matt Jaffe Penn State University 2 Indirect search for new physics – CP Violation – Measurement of rare decays 700 scientists from 52 institutions Lepton Flavor Violating Decays

Outline The τ → μμμ Decay – Why study it? – How to study it? Data Analysis Results Matt Jaffe Penn State University 3 {

Motivation: Why τ → μμμ ? Study Lepton Flavor Violation Standard Model forbidden process Test and constrain New Physics models Matt Jaffe Penn State University 4 SM + Heavy Majorana ν r SUSY SO(10) mSUGRA + seesaw OR (more likely)

Methods: How τ → μμμ ? Select signal, reject background – Draft sets of cuts – Study cut effects on signal and background distributions Matt Jaffe Penn State University 5 Example of a variable distribution which might be cut on.

Control Channels Matt Jaffe Penn State University 6 D s → π π π D s → φ(μμ) π – Distributions as similar as possible to signal – Want cuts with similar effects across all three channels Example of a variable distribution demonstrating good agreement between signal and control channels.

Data Analysis Matt Jaffe Penn State University 7 40 MHz L0 Trigger 1 MHz High Level Trigger 3 kHz “Online” Analysis “Offline” Analysis Stripping Lines Stored Data Selection Cuts PHYSICS! My Decay Channel Other Decay Channels

Step 1: Stripping Matt Jaffe Penn State University 8 Data τ -> μ μ μ Normalization Channels < 0.05 % Selectionε r (τ -> μμμ)ε r (Ds -> πππ)ε r (Ds ->φπ) Proposed Previous Selectionε r (τ -> μμμ)ε r (Ds -> πππ)ε r (Ds ->φπ) Proposed 0.04% 0.21% (Prescale) 0.044% Signal Monte Carlo Efficiency Background Elimination – Remaining Events

Step 2: Selection Matt Jaffe Penn State University events Selection Cuts Selection Cuts Efficiencies ε r (τ -> μμμ) * (Monte Carlo) ε r (Ds -> πππ)* (Monte Carlo) ε r (Ds ->φπ)* (Monte Carlo) LHCb Real Data (Background) Currently: 8.6 %17.7%19.0 % % Efficiency = ( # candidates after stripping + selection ) / (# candidates after stripping ) D s mass peak reconstructed by our selections from LHCb data *(Not yet signal specific selection) LHCb Internal

Control Channel Mass Peaks Matt Jaffe Penn State University 10 D s peak split into the 2 channels Signal MCs on top Extracted peaks below LHCb Internal D s -> 3 Pi Signal Monte CarloD s -> Phi Pi Signal Monte Carlo D s -> Phi Pi from dataD s -> 3 Pi from data

Future Work Matt Jaffe Penn State University 11 Re-optimize D s →π π π stripping to eliminate need for pre-scaling Further study of daughter PID cuts to split the three channels Analyze selection of τ→μ μ μ. We imagine starting with an earlier, looser version of the D s cuts, then optimizing to the channel. Perform rigorous statistical analysis of cut effects at all stages Look at a lot of data!

Thank You! I cannot thank the following people enough: CERN Summer Student Program – Dr. Johannes Albrecht – Supervisor – Dr. Marco Gersabeck – Substitute Supervisor The University of Michigan REU program – Dr. Homer Neal – Dr. Jean Krisch – Dr. Junjie Zhu – Dr. Steve Goldfarb – Lauren Rigani Matt Jaffe Penn State University 12

Thank you! Matt Jaffe Penn State University 13

Matt Jaffe Penn State University 14

BACKUP SLIDES Matt Jaffe Penn State University 15

Backup Slides - τ -> μ μ μ Stripping Line Previous Cuts Mother Cuts Mass Window = ± 400 MeV Vchi2 < 25 c * τ > 70 μm Child with P t > 1 GeV Daughter Cuts (for all 3) P t > 300 MeV mu track χ 2 / DoF < 5 (Mu IPS) 2 > Matt Jaffe Penn State University 16 Proposed Cuts Mother Cuts Mass Window = ± 400 MeV Vchi2 < 20 c * τ > 100 μm τ IPS < 15 Daughter Cuts (for all 3) P t > 300 MeV mu track χ 2 / DoF < 5 (Mu IPS) 2 > 9

Backup Slides - Control Channel Stripping Line Matt Jaffe Penn State University 17 Previous Cuts Mother Cuts Mass Window = ± 250 MeV Vchi2 < 25 c * τ > 200 μm D s IPS < 3 P t > 2 GeV Daughter Cuts (for all 3) P t > 300 MeV mu track χ 2 / DoF < 5 (Mu IPS) 2 > 12 Proposed Cuts Mother Cuts Mass Window = ± 80 MeV Vchi2 < 20 c * τ > 200 μm D s IPS < 15 Ds->3π required to be TIS Daughter Cuts (for all 3) P t > 300 MeV mu track χ 2 / DoF < 5 (Mu IPS) 2 > 9

Backup Slides - Current Selection Cuts (e) Matt Jaffe Penn State University 18 Current Working Version of Selection Cuts 1)Mother Impact Parameter Significance ( = √(χ 2 ) ) < 1.8 2)Mother Vertex χ 2 < 6 3)max ( Delta Log Likelihood( Pion - Kaon ) ) < 0 4)max ( Daughter Track χ 2 ) < 1.4 5)Cos ( angle ) > )max ( Daughter P tot ) < 40 GeV 7)max ( Delta Log Likelihood( Pion - Proton ) ) < 5

Stripping Cuts Tuning: Step Matt Jaffe Penn State University 19 Selection #ε r (Tau -> 3 Mu)ε r (Ds -> 3 Pi)ε r (Ds -> Phi Pi ) 2_ _ _ ε r := ( # candidates after new stripping lines) / ( # candidates after truth matching )

Matt Jaffe Penn State University 20 Selection #Ds -> 3 PiDs -> Phi PiSignal Blinded Data (a) (b) (c) (d) (e) Selection Cuts - efficiencies Efficiency = ( # candidates after stripping + selection ) / (# candidates after stripping )

More Detailed Efficiency Matt Jaffe Penn State University 21 CutDs -> 3 PiDs -> Phi PiBackgroundDs -> 3 PiDs -> Phi PiBackground (e) (N-1) ε for this cutε for this cut only

Mass Peak Fit Parameters Matt Jaffe Penn State University 22 (a)(b)(c)(d)(e)S / B - fits integrated on (1900,2050) Ds -> 3 Pi (1) (a):0.380 (2) (b):0.374 (3) (c):0.680 (4) (d):0.734 (5) (e):0.719 Ds -> Phi Pi (1) (a):0.340 (2) (b):0.375 (3) (c):0.591 (4) (d):0.664 (5) (e):0.791 #Parameter (1)Linear fit constant (2)Linear fit slope (3)Gaussian amplitude (4)Gaussian Center (5)Gaussian Sigma

Backup Slides - Kinematics Plots Matt Jaffe Penn State University 23

Backup Slides - More Plots Matt Jaffe Penn State University 24

Backup Slides - Flight/IPS Plots Matt Jaffe Penn State University 25

Backup Slides - More Plots Matt Jaffe Penn State University 26