Muon Reconstruction and Commissioning with Early Data

1
Muon Reconstruction and Commissioning with Early
Data

• Kevin Black
• Harvard University

2
Outline

• Overview of Muon Reconstruction Software
• General Reconstruction
• Low PT Muon Identification
• Calorimeter Information
• Comparison of Algorithms (see talk by David
  Adams)
• Commissioning with early data
• Cosmics
• Beam Gas Events from single beam running
• Collision Data (_at_900 GeV and then _at_14 TeV)
• Will not discuss test-beam, alignment and
  calibration, or electronic calibration (see talk
  by Ed Diehl)

3
Active Developers (and many others in the past)

• Moore MuID
• BNL - David Adams, Ketevi Assamagan CERN -
  Alan Poppleton, Harvard KB, Steve Cavanaugh,
  Ben Smith, Srivas Prisad INFN Napoli Michela
  Biglietti, INFN Leece Gabriella Cataldi Michigan
  Dan Levin NIKHEF Niels Van Eldik SUNY
  Albany Vivek Jain, Victoria Rojo U Mass
  Amherst Ed Moyse, Thomas Moore, Stephane
  Willocq
• MuonBoy STACO MuTag
• Saclay Florian Baur, Laurent Chevalier, Jean
  Ernwein, Andrea Formica, Pierre-Francios Giraud,
  Claude Guyot, Samira Hassani, Eric Lancon,
  Jean-Francois Laporte, Rosy Nicolaidou, Amimed
  OuraouSaMuSog
• MuGirl
• CERN Zvi Tarem Technion - Natalia
  Panikashvili, Shlomit Tarem, Tel Aviv Orfirt
  Belkind, David Primor
• Muon Identification with Calorimeter
• NIKHEF Peter Kluit, G Ordonez, Wisconsin - L.
  R. Flores-Castillo, B. Mellado, Sau Lan Wu
• dE/dX and Energy Loss in the Calorimeter
• Athens - C. Kourkoumelis, D. Fassouliotis, K.
  Nikolopoulos Saclay SaMuSog, CERN A.
  Poppleton
• Material Treatment
• CERN A. Poppleton UMass Amherst Thomas Moore,
  Saclay SaMuSog Tufts S. Todorova

4
Identify and Measure track parameters
For B 0.5 T, L 5 m p 5 GeV/c R 33 m
s 0.1 m p 1 TeV/c R 6700 m s
500 µm
? need 50 µm resolution to achieve 10 momentum
resolution at 1 TeV
5
Moore PatternRecognition
rpc
barrel ? projection
rpc
rpc
Search for region of activity in the ?
projection and RZ projection
barrel RZ projection
MDT
6
Segment Finding,Track Fitting

• Pattern recognition in individual MDT multilayer
• the drift distance is calculated from the drift
  time, by applying various corrections on it (TOF,
  second coordinate, propagation along the wire,
  Lorenz effect). Among the 4 tangential lines the
  best one is found.
• In the CSC, cluster the strips and form segments
  using a histogram method
• Track segment combination.

MDT pattern recognition
MDT multilayer

• Track fit
• track parameters (a0, z0, ?, cot??, 1/pT ) are
  expressed at the first measured point

7

Package Muid

• Two steps
• 1. Track extrapolation at the I.P.
• Multiple scattering parameterized by means of
  scattering planes in the calorimeters
• Energy loss in calorimeters parameterized in
  function of (??, ?) or measured from calorimeter
  reconstruction
• Re-fit track parameters expressed at vertex
• 2. Tracks from the muon spectrometer and from the
  inner detector are combined with a ?2 lt cut-off
• - ?2 calculated from differences of track
  parameters and from covariance matrix
• Final fit of the successfully
  combined track

8
Moore and MuId Performance and Development
Rome Era (10.0.1)
Recent Release (11.0.3)

• Albany, BNL, CERN, Harvard,
• UMass Amherst
• EDM Migration
• Improved CSC treatment
• Improve pattern recognition
• Improve material treatment
• Develop tools to enhance efficiency and reduce
  fake rate

9
Muonboy and STACO

• Muonboy is another muon spectrometer track
  reconstruction program
• Similar in strategy
• Main Differences 3D pattern recognition,
  extrapolates from inner layer outward adding hits
  as it goes along, material description
• STACO No Refit
• Compare muon spectrometer track and inner
  detector tracks
• Use parameters and covariance matrices to match
  and find the parameters of the combined track
• See backup for mathematical details

Saclay
10
MuonBoy and Staco Performance
Saclay

• Developments
• Staco try a full refit at Calorimeter surface
• Further Development of track extrapolation
• Further Development of their low PT algorithm
  (MuTag)

?
11
Energy loss in the Calorimeter and dE/dx

• Energy loss in Calorimeter
• Either use measured or parameterized energy loss
• Correct for long Landau tail
• See backup for details
• Material Description
• Moore Geantino Map to determine the amount of
  material and then add special scattering hits
  onto the tracks
• MuonBoy collect material during extrapolation
  and track finding and add scattering centers

Athens, CERN
UMass Amherst, Saclay
12
Low Pt Muons
Low PT muons often do not reach the outer
stations, Start with inner detector tracks and
extrapolate out to the muon spectrometer MuTag
and MuGirl
Extrapolation Region
Extrapolation Region
13
MuGirl Performance and Development

• In AODs in release 12
• Working on rejection of muons from K/? decays
• Including segments from TGCs

Bs ?J/?(µ6µ3) bb?µ6X W?b?µ6X Higgs?4µ
CERN, Tel Aviv, Technion
4 Working points evaluated on 4 samples
14
Using the Calorimeter to tag muons
Wisconsin
StacoMuTag StacoMuTagCaloLR
H-gt4 ?µ

• Hardware-related inefficiencies for turn-on
• Tracking efficiency 100
• 4-2-0 topo clusters
• 100 efficient for muons
• Many samplings available
• Longitudinal and transverse shape information

• From single muons and pions select 11 variables
  for Likelihood Ratio ? rejection1000 at 90
  efficiency
• For now, only for ?lt1.4
• Efficiency 94.2 ? 97.7
• Will make use of NN, study more complicated
  background events with pileup

15
Commissioning with Cosmics

• Sector 13 without trigger chambers for about 9
  months
• Sector 13 with trigger chambers starting a few
  weeks ago
• Apply Algorithms to real data
• Beginning of Alignment and Calibration with
  Cosmics
• Analysis of Cosmics
• CERN, Harvard, NIKHEV, Saclay, U Mass Amherst

16
Challenges with Early Data

• Very non-uniform magnetic field (introduces
  complications to calibration, track extrapolation
  and hence alignment, and resolutions)
• Very complicated large area detector (alignment,
  calibration) From S. Goldfarb
• 105 pT gt 2 GeV tracks for alignment with cosmics
• 106 pT gt 3 GeV for alignment with straight
  tracks
• 106 pT gt 6 GeV for alignment with inner detector
• 106 pT gt 5 GeV for alignment between small and
  large sectors
• What is the real inert material?
• How will the real cavern background and pile up
  affect muon reconstruction and identification?
  How realistic is the simulation?
• How severe will punch-through be? Will it be as
  expected from simulation?
• Dedicated triggers and streams for calibration,
  alignment, and tests of the reconstruction (see
  talk by Ed Diehl)
• What is the Standard Model at 14 TeV?

17
Cosmic Ray Commissioning
Rates are substantial 2.3 KHz for a hit
anywhere in detector 0.5 Hz for Z lt 60
cm,R lt 20 cm Trigger in barrel or end-cap 40 Day
Atlas Global Run before beam
From R.McPherson and J. Pilcher Talks
18
Collider Data at 900 GeV
B physics group Heavy Quarkonium 2006 Talk

• Initial Commissioning Run
• C.M. energy 900 GeV (injection energy to LHC
  ring)
• Luminosity (P. Jenni) 1029 cm-2
• Only a handful of Z events ?Br .2 nb after 60
  days running may get 1-3 events (P. Jenni)
• Best focus on J/? and other resonances for
  resolution and scale
• Use all muons for alignment and calibration

Mass resolution s(M2µ)43MeV
Stable beams
Preparation
First collisions.
Shutdown 3 to 4 months?
July
Aug.
Sept.
Oct.
Nov.
Dec.
Jan.
Feb.
Mar
19
Collider data at 14 TeV
_at_1033 cm-2 s-1
Channel Recorded 10fb-1
W ? mn 7 x 107
Z ? mm 1 x 107
tt ? m X 0.1 x 107
Jets pTgt150GeV (if 10 bandwidth) ?107
Min Bias (10 bandwidth) ?107 (can be larger)
gg (M1 TeV) 103-104
With 100 pb-1 0.1 Million Z ? µµ, recall 10
Million events recorded at Tevatron after 20
years
20
Simplistic View of Strategy

• 1st understand and calibrate detector and trigger
  in situ
• Use all muons
• Calibration and alignment
• Alignment with inner detector
• Use well known standard candles (dimuon
  resonances)
• Study Resolution, efficiency, material
  description, realism of monte carlo simulations
• Understand the SM at 14 TeV (first measurements
  of cross-section, rapidity distributions, etc)
• 2nd
• Prepare for the road for discovery
• Measure backgrounds and obtain control samples
• 3rd
• Real excitement phase, first time to
  substantially probe physics at the few TeV scale

21
Summary and Conclusions

• Muon Reconstruction Algorithms performing well
  with similar performance (see D. Adams on
  performance)
• Lots of active interest in North America, many
  recent and continuing developments
• Several Stages to commissioning
• Muon standalone cosmics
• Global Atlas Cosmic runs
• Beam Gas and Beam Halo events for single beam
  runs
• Early focus for 900 GeV Run
• Multistage process with 14 TeV collisions
• Much accomplished, but still much to do
• Primarily, shift focus to preparing for first
  data (more cosmic data)
• Develop tools to evaluate performance without the
  help of truth information (e.g. tag and probe)
• Difficult challenge for alignment and calibration
  because of the size and complexity of the Muon
  Spectrometer and Toroid magnets (See E. Diehls
  talk on calibration and alignment)

22
Back up Slides
23
Big Wheel
RIBS
24
Muonboy Strategy

• Similar Strategy search for Regions of Interest
• Form Segments
• Combined Segments
• Track Fit
• Main Differences
• 3d pattern recognition from the start
• Start with segments in inner station and
  extrapolate the position out to next layer
• Handles inert material differently
• Core program written in early 90s with F90,
  wrapped in C for use with athena

25
Principle of STACO

• For two tracks on some reference location defined
  by their
• Parameters vectors P1 and P2
• Covariance matrices C1 and C2
• P parameters vectors of combined track is the
• solution of the equation
• C covariance matrix of combined track is
  given by
• The corresponding ?2
• Track combination is tried only for pairs of
  tracks that show a reasonable matching in the
  (?,?) plane
• Track combination is accepted only if the global
  ?2 is below a maximal value
• If different combinations are possible, the pair
  given the best ?2 is retained
• Uses Parameterization of Energy Loss in Inert
  Material

26
Strategy for µ energy loss reconstruction
Athens, CERN
27
Long Landau Tail..
A correction was found for this effect. Instead
of using the MPV as is, use a weighted mean in
the region
where
Athens, CERN
28
Different Approaches

• All use inner detector extrapolated track
• MuTag is run after Muonboy and extrapolates ID
  track and tries to match with unused segments
• MuGirl extrapolates ID track and forms new
  segments using ID seeding Extension and
  replacement of older package MuIdLowPt
• Uses ANN to discriminate between true particles
  and fakes
• Uses vertex constraint to discriminate between
  muons from IP and from in flight decay of ?/K
• As of release 12, MuIdLowPt will be deprecated,
  MuTag and MuGirl available at AOD level
• MuGirl is the newest package on the market, but
  is already showing great progress

29
Calorimetric Muon ID
Wisconsin
etileb0/EHad
eemb2/EEMHad
EtopoMax/Etopo
etileb1/EEMHad

• Rejection of single ? 1000 at 90 efficiency
• Plans
• Neural Network for µ/?? separation
• Make use of track to calorimeter extrapolation
  tools
• Study fakes, efficiency, effects on significance
  with background and piluep

30
Plans with first Data

• Cosmics
• Analysis of sector 13 data with trigger chambers,
  as the detector is assembled take more cosmic
  runs
• Initial Calibration, alignment with tracks,
  debug/improve software with the clean and simple
  tracks, alignment with inner detector
• Initial Run with only injection energy, 900 GeV
  low luminosity (although not as exciting,
  benchmark point for understanding)
• Alignment and Calibration with any muon tracks
• Handful of Z events, focus on J/? (and other
  resonances) for studies of efficiency,
  resolution, fake rates with data tag and probe,
  comparison with inner detector
• Alignment with inner detector, study material
  effects, energy loss
• Initial run at full energy
• Focus on Z events tag and probe to study eff,
  res, fake, comparison with inner detector

31
Beam Gas and Beam Halo Events

• ?? pT gt 1 GeV inside ? 3m
• 1.0x103 Hz
• 1.5x109 events in 2 months assuming 30
  efficiency
• Beam Halo events
• Especially usefull for end cap

Beam-gas collisions are essentially boosted
minimum-bias events ? low-pT particles
Rate 2500 interactions/m/s
From J. Pilcher