Study of Z decays to τ pairs with CMS detector at √s = 14 TeV Michail Bachtis CMS Group University of Wisconsin - Madison test

Outline Physics of Z ττ The CMS experiment Z ττ study in CMS The Standard Model Z production at the LHC τ phenomenology and identification principles New Physics with τ Background processes to Z ττ The CMS experiment Design and sub-detectors The CMS Trigger system Z ττ study in CMS Generator studies Detector and Trigger Performance studies Zττ Analysis Summary/Next plans

The Standard Model 12 Elementary Particles (fermions) Three generations of quarks. Three charged leptons and corresponding neutrinos 4 Force Carriers (bosons) Gluon (Strong) Photon (EM) W,Z (Weak) Not a complete theory Higgs boson to be discovered

Importance of Z boson studies Test of the Standard model in the new energy domain. Detector performance studies Optimization of τ trigger and offline reconstruction Background for new physics Higgs, Z’ Z production via Drell-Yan in proton collisions

τ Decays τ lepton Decays to lighter particles mass = 1.8 GeV Leptonic Decay Hadronic Decay τ lepton mass = 1.8 GeV mean lifetime ~10-13s Decays to lighter particles Leptonic decays (~35%) Electron/Muon + 2 Neutrinos Lepton+ Missing Et signature Hadronic decays (~65%) Mostly one or three charged particles (prongs) +neutrals+ neutrino “Narrow” jet signature in the detector Most Relevant τ decays and BRs Decay Mode BR τeνeνт 17.8% τμνμνт 17.4% τh +neutrals (one prong) 46.8% τ3h+neutrals (three prongs) 14.0% τ5h+neutrals (five prongs) 0.1% τK±Xνт 3.7% Others 0.03%

Z  ττ Three decay modes τe+τμ Both τ decay leptonically BR = 6.2% Three decay modes Both τ decay leptonically electron + muon electron +electron muon+muon One τ decays leptonically and one hadronically One τ can give one/three prongs Most favored mode Both τ leptons decay hadronically Large Jet Background τh+τl BR = 45.3% τh+τh BR = 41.4%

Background processes QCD Jets Drell-Yan W+Jets Top quark pairs Extremely large cross- section (order of mb) Narrow jets fake hadronic τ Drell-Yan Background for e,μ from τ Leptons fake one prong τ too! W+Jets Wlν Jets fake hadronic τ Top quark pairs Contain τ,W,Jets

New Physics with τ Standard Model Higgs searches (Hττ) MSSM Higgs Higgs couples to mass ττ BR=~10% for low Higgs mass Example:Vector boson fusion Two τ + forward Jet Signature Higgs mass limits >114 GeV (LEP) <160 GeV from indirect searches MSSM Higgs Both charged and neutral Higgs possible Large Branching ratio to τ Heavy neutral Higgs to τ pair Charged Hτν

The CMS Experiment Calorimeters Iron Yoke Muon Endcap Inner Detector ECAL HCAL 76k scintillating PbWO4 crystals Plastic scintillator/brass sandwich Iron Yoke Muon Endcap Cathode Strip Chambers (CSC) Resistive Plate Chambers (RPC) Inner Detector Pixel Silicon Tracker 210 m2 of silicon sensors 9.6M channels Weight: 12,500 T Diameter: 15.0 m Length: 21.5 m Solenoid Magnet 4 T Magnetic Field Muon Barrel Drift Tube Resistive Plate Chambers Chambers

The CMS Experiment Today Lowering the last heavy element Tracker in position Solenoid

Silicon Tracker Tracker installation Silicon Technology Performance Pixel Detector near the interaction point Strips in surrounding area (Barrel, Endcap) Performance High tracker granularity, large size + strong B-field make the tracker efficient for a broad Pt spectrum. Resolution :

Electromagnetic Calorimeter Crystal Technology Lead Tungstate Crystals (~76000) High density (8.2 g/cm3) Short radiation length (8.9 mm) Small Moliere radius (22 mm) High segmentation for precise position measurement Acceptance to |η|<3.0 Resolution:

Hadronic Calorimeter Barrel and Endcap part (|η|<3) Brass / Scintillation layers Resolution: Forward Region (3<|η|<5) Steel plates / Quartz fibers Resolution: Absorber geometry 7 Interaction lengths at η = 0 11 Interaction lengths at η = 1.3

Muon System Operation Principles Muons are identified in Muon System For low Pt muons, Pt is assigned by the tracker For high Pt muons, Muon system contributes to the measurement All muon sub-detectors contribute to the trigger Layout Barrel Drift Tube chambers (DT) |η|<1.3 Resistive Plates (RPC) |η|<1.3 Endcap Cathode Strip Chambers (CSC) 0.9<|η|<2.4 Resistive Plate Chambers (RPC) |η|<2.1 Endcap Disc made in UW

CMS Trigger Overview 2-Level Trigger Design Level 1 Trigger Hardware 100 kHz output (50kHz at first runs) Latency = 3 μs High Level Trigger Software running on Processor Farm Algorithms similar to offline reconstruction ~100Hz output Crossing rate =40 MHz Trigger Rejection ~ 4x105

L1 Trigger Design Calorimeter Trigger Muon Trigger Global Trigger Regional Calorimeter Trigger (RCT) Finds e/γ,regional energy deposits Forwards RCT objects to GCT Global Calorimeter Trigger (GCT) Finds jets,τ Sorts RCT Objects, Calculates Missing Et Forwards Calorimeter quiet regions to Muon Trigger Muon Trigger Regional Triggers Find Segments on chambers Tracks are created in DT,CSC Global Muon trigger Sorts muons Checks Muon Isolation Global Trigger Applies selection criteria Communicates L1 decision

Ratio of Produced Signal Analysis Outline Zττ predicted cross section is 530 pb Main background for τ hadronic decays: QCD Jets QCD cross section = ~108 pb!! Leptonic τ decays faked by Electroweak Processes Z,W, Drell-Yan Optimization of Trigger and τ-ID Important for suppressing backgrounds Zττ analysis procedure Trigger and detector performance studies Zττ analysis Monte Carlo Samples (Pythia) Zττ (500K events) QCD 1B events σ=1.8x108 pb Electroweak (EWK) 50M events W+Jets, Z (excluding τ)+Jets, Drell-Yan σ=2.1x105 pb Expected Events @ ∫ L=100pb-1 Ratio of Produced Signal to Background Events (Before Trigger) 1:350000 Events! Zττ : 53000 QCD : 18B EWK : 21M

Generator Level Cuts Require visible τ Pt>10 GeV (73% accepted) Visible Pt much smaller 73% of the generated τ accepted Require visible τ |η|<2.5 τ must be in tracker acceptance 65% of the generated τ accepted 48% of generated events accepted MC Fiducial Cut (Accept) Tracker Acceptance

Generated Z invariant Mass Broad mass distribution Mass peak shifted 15-25 GeV in leptonic τ decays Neutrinos from e,μ Mass window expected in 20-100 GeV Zττ decays to μμ,ee Drell Yan μμ,ee is irreducible background S:B ~ 50:3000 events! Pythia, Zττ hh hμ/he μe μμ/ee Pythia, Drell-Yan

Calorimeter Geometry Crack [Tracker Cabling/ Services] η η

L1 e/γ Trigger Algorithm e/γ Triggers Large energy deposit in 2 adjacent towers. Shower profile Fine Grain spread in central cell of 3x3 Longitudinal Profile Ratio of HCAL-ECAL energies Isolation on nearest neighbors for isolated object triggers Efficiency per electron candidate = 98% for Pt>10 GeV Pythia, Zττ

L1 τ Trigger Algorithm L1 τ algorithm Uses towers in 12x12 region Specific isolated energy patterns allowed in 4x4 region Non isolated patterns set a veto τ accepted if all vetos are off. Additional Isolation Requires Et<2 GeV in 7 of 8 neighboring 4x4 regions Efficiency per τ candidate = 78% for Et>10 GeV Pythia, Zττ τ Candidate

Muon Geometry Full coverage to |η|<2.4 Three main coverage regions Overlaps with Tracker Coverage. Three main coverage regions |η|<0.8: Barrel only 0.8<|η|<1.3: Barrel and endcap 1.3<|η|<2.4: Endcap only.

L1 μ Trigger Algorithms Local Tracking on Chambers Track Finders Segment Reconstruction Track Finders Cathode Strip Chamber and Drift Tubes Segments combined to global tracks Momentum assigned to the tracks Efficiency per Muon Candidate = 99% for Pt>10 GeV CSCs , 0.9<|η|<2.4 Pythia, Zττ DTs , |η|<1.3

L1 Global Trigger Paths Zτh+τh Zτl+τh Zτμ+τe Double τ Trigger Requires 2 τ with Et>20 GeV Zτl+τh muon + hadronic τ Requires an isolated muon with Pt>5 GeV and a hadronic τ with Et>10GeV electron + hadronic τ Requires an isolated e/γ object with Et>10GeV and a hadronic τ with Et>10GeV Zτμ+τe An “electron + muon” L1 trigger is required Requires an isolated e/γ object with Et>10GeV and a muon with Pt> 5GeV

τ High Level Trigger Those algorithms won’t be used with first data L1 Seeding L2 Calorimeter Isolation L2.5 Pixel Isolation L3 Tracker Isolation Regional Jet Reconstruction around L1 tagged τ Isolation using Calorimeter only Isolation using Pixel Tracks Isolation using Silicon Tracks Those algorithms won’t be used with first data

L2 τ High Level Trigger Three Trigger Algorithms My work.. Start with L1 seeded Jet and a cone around jet axis ECAL Isolation Sum of Crystal Et in isolation annulus Tower Isolation using CaloTowers Sum of Tower Et of isolation annulus “Fast” ECAL Clustering Clustering using ECAL crystals Number of Clusters Cluster spreading around jet center η φ Signal Cone Isolation Cone ECAL Clusters Isolation annulus η φ This is a Jet η φ This is a τ

Cone Isolation Hadronic τ : narrower than QCD jets ECAL Isolation Algorithm Measures total ECAL Crystal Et in isolation annulus Require ECAL Et<3 GeV QCD Jets Removal of 40% τ Efficiency = 98% Tower Isolation Algorithm Measures tower Et (ECAL+HCAL) in the isolation annulus Require Tower Et<5 GeV QCD Jets Removal of 50% τ Efficiency = 97% Important cut for candidates without ECAL contribution Zтт QCD Reject Zтт QCD Reject

ECAL Clustering For further background removal, a Clustering algorithm on ECAL Crystals is applied Clusters are created by ECAL crystals Cuts Number of clusters<7 QCD Rejection =55% τ Efficiency =96% η RMS<0.04 QCD Rejection 60% τ Efficiency =93% Zττ QCD Reject Zττ QCD Reject

HLT Performance Results after applying all the previous algorithms τ Efficiency = 90% QCD Rejection = 75% Cuts can be tuned to provide tighter (looser) configurations Maximum QCD Rejection by a factor of 10 with 83% of τ preserved Maximum τ Efficiency of 99% with 40% of QCD rejection Pythia, Zττ

Inclusive Trigger Performance Trigger Acceptance (Events at ∫ L=100pb-1) Mode τ+τ e+τ μ+τ μ+e Trigger Thresholds τ Et>20 GeV τ Et>10 GeV e Pt>10 GeV μ Pt>5 GeV Z  тт 2026 489 940 324 QCD Jets 118192 1407980 38626 32040 EWK (W+Jets, Z+Jets,Drell-Yan) 176 420025 10899 840 Signal:Background 1:60 1:3738 1:50 1:101 Triggers reduce background rates but we can do even better Recall: We started with 1:350000 events

e/μ Offline Reconstruction Electrons Calorimeter Reconstruction Create “super-clusters” of clusters to include radiated photons Apply Et thresholds Tracker Reconstruction Electron is matched to a track. Cuts are applied on e/p and HCAL energy deposits Muons Standalone Reconstruction Muon tracks reconstructed from the muon system Combined Reconstruction Muon Tracks are matched to tracker tracks and combined muons are created Isolation can be applied in both cases High Level trigger algorithms are similar. ET/pT cut ET γ e- Tracker Strips pT • Pixels Inner Detector Track Standalone Muon Track

Lepton Offline Reconstruction Efficiency Muons (from τ decays) Muon efficiency = ~95% for Pt > 5 GeV Coverage of the Muon Detector up to η = 2.4 for Pt > 5 GeV Electrons (from τ decays) Electron efficiency = ~88% for Pt>15 GeV Future improvement: Optimizing Electron Offline Performance Geometrical acceptance “Crack” reduces efficiency in the transition region (Barrel – Endcap) Reconstruction harder near the tracker boundary (η=2.5) Muons Electrons Pythia, Zττ Muons Electrons Pythia, Zττ

e,μ Resolution Muons Electrons Curvature resolution provided by Tracker Curvature Resolution = 1.9% Electrons Bremsstrahlung blurs Resolution for electrons Peak shifted by ~0.04 Curvature Resolution = 4.9% Pythia,Zττ Pythia,Zττ

τ Identification with Cone Isolation Leading Track axis Two algorithms CaloTau Algorithm Associates tracks to jets Identifies τ by track isolation Particle Flow The algorithm Reconstructs particles Applies Pt corrections in particle level Forms jets from particles γ Jet Axis Signal Cone ΔR=0.15 Isolation Cone ΔR=0.5 Jet cone Require no charged,γ candidates in isolation annulus

Hadronic τ Performance Particle Flow improves Resolution Distribution is better centered. Peak is sharper CaloTau Efficiency is Higher Particle flow can miss a High Pt candidate Particle Flow τ-ID still under basic development Tail under investigation to raise efficiency Pythia, Zττ PFTau CaloTau Pythia, Zττ PFTau CaloTau

Calorimeter Missing Et Estimates Et of particles undetected in Calorimeter Muons Neutrinos Particles outside geometrical acceptance It is defined as: where the sum is evaluated over all the Calorimeter towers Critical for many physics studies Top Studies SUSY Studies τ Studies Pythia, Zττ Poor Resolution MET Higher Underestimation of jet energies Need to be adjusted by calibration

Zτμ+τh studies MC Samples (PYTHIA) Zττ 500K events (Full Simulation) QCD Jets (Pt >30 GeV) Muon preselection in MC Level Require 1 final state muon Generated: 1 Billion events (Fast Simulation) Electroweak W+Jets Z+Jets (excluding ττ) Drell-Yan 50 Million Events (Fast Simulation)

μ,τ and MET spectra EWK & QCD :dominant backgrounds Requirements Muon Pt and MET larger for EWK sample W decays Requirements Event has passed μ+τ HLT Path τ tagged by PF-TauID Only one isolated μ, Pt > 10 GeV Only one τ, Et > 20 GeV Zττ Accept EWK QCD EWK X 10 Signal X 20 Zττ Accept EWK QCD EWK X 10 Signal X 20 Zττ Process Events(100pb-1) S:B EWK Zττ 827 QCD EWK 9046 1:11 EWK X 10 Signal X 20 QCD 20 Not enough QCD statistics to populate spectrum!

Opposite direction and sign τ expected back to back in r,φ Neutrinos blur τ alignment Require |Δφ|>2.5 Require opposite sign between τ,μ 83% of Electroweak Background rejected Zττ EWK Reject QCD Process Events(100pb-1) S:B Zττ 637 EWK 1809 1:2.8 QCD <20

Rejection of W decays Apply (μ,MET) transverse mass cut Expected to be larger for W MET larger in W decays W mass larger Require Mt < 30 GeV Zтт Reject EWK Process Events (100pb-1) S:B Zττ 495 EWK 251 2:1 QCD <20

μ+τ Invariant Mass Z peak visible in mass spectrum 495 events at ∫L=100pb-1 Mass Window 32-104 GeV For M(μ,τ) >110GeV Background < 2 events @100pb-1 Expect good results for Hττ with similar analysis Zττ EWK Signal Efficiency = 53% EWK Rejection = 97.7%

Conclusion Summary Conclusions Improved τ Trigger Achieved maximum QCD Suppression by a factor of 10 using Calorimeter cuts Implemented Zττ Analysis Achieved a S/B ratio of 2 Conclusions Zττ is detectable at ∫L=100pb-1 If a SUSY Higgs appears in low luminosity (large (tanβ)2 ), it possibly can be observed This is the first step for a SM Hττ study

Next plans Next plans Work on L1 and High Level trigger Improve Trigger and Reconstruction performance for leptons and hadronic τ Optimize Zττ analysis and measure σ(Zττ) Optimization with Linear Fisher Discriminant slightly improves performance (S/B = 2.8) Search for the Higgs

Backup Slides

Linear Fisher Discriminant Take two sets of points x in a N dimensional space (one for signal, one for background). Define a linear transformation y=wt·x : RNR We need the transformation w such that the clusters will be best separated in the 1D space. Best separation |μ1-μ2|2 One idea is to maximize J= where μ,σ σ12+σ22 is the mean and variance in 1D space. (maximum distance and minimum spreading in the final space) wt M w J can be written as : J= with M=(m1-m2)(m1-m2)t S=Σ(x-m1)(x-m2)t wt S w in ND space Setting J = λ gives: wtMw = λ wtSw  Mw = λSw S-1Mw = λw So we have an eigenvalue equation for S-1M. The maximum eigenvalue gives maximum separation and the corresponding eigenvector gives the linear transformation.

Further Optimization Using Linear Fisher Disciminant Fisher Discriminant projects the variable space in one dimension Projection with maximum separation Input Variables [4 Dimensions] Δφ (μ,τ) Mt(μ,MET) Δφ (μ,METu) MET Cut value computed by maximizing: Sig = S / (S+(Ls/Lb)B)1/2 where: S: # of signal events that satisfy cut B: # of background events that satisfy cut Li:integrated Luminosity of sample Optimized value: 4.1 Zтт EWK Reject

Cut on Fisher Discriminant Optimization easier in one dimension Discriminant provides one dimensional variable Similar results with cut based analysis Signal efficiency increased to 61% (+10%) Background Acceptance decreased to 1.9% (-1.3%) Process Events (100pb-1) S:B Zττ 574 EWK 207 2.8:1 QCD <20

Muon Isolation QCD Jets often contain leptonic quark decays μ+narrow Jet can fake Zτμτh Apply Muon Isolation Sum of the ECAL Et< 3 GeV in Cone of ΔR = 0.3 Sum of Track Pt< 3 GeV in cone of ΔR = 0.3 Zττ Reject EWK QCD Zττ Reject EWK QCD