PowerPointPrsentation

1
The new ATLAS Track Extrapolation In a case study
of a future high precision W mass measurement
Andreas Salzburger, CERN PH-ATC University of
Innsbruck, Austria
2
Outline

• Track extrapolation
• - concept / applications
• Design of the new ATLAS Track extrapolatino
  engine
• - realization in the new Event Data Model
• - material integration
• Combined Testbeam 2004
• - first test of new software frame
• Validation procedure
• - from reconstruction to simulation
• Prospect of a future W mass measurement
• - scenario for a praxis test

Andreas Salzburger OEPG, Vienna, 27/09/2005
3
Track extrapolation Introduction
The transport of track parameters and their
associated covariances is an underlying process
of most track fitting algorithms. LHC with its
high event rates and track multiplicity requires
a very stable and reliable framework for these
operations. ATLAS with ist complex geometry
(material distribution) and magnetic field needs
very precise algorithms to match the requirements
of the challenging physics goals.
Global Frame B-Field
x
B
a1
z
p1
y
Andreas Salzburger OEPG, Vienna, 27/09/2005
4
Track extrapolation Definition and Applications

• Can be divided into two basic realms
• geometrical transportation of parameters and
  covariances
• update of parameters and covariance matrix
  taking multiple Coulomb scattering and energy
  loss effects into accout

• Applicaitons
• global and iterative Fitters
• track calorimeter matching
• holes on track search
• vertex fitter
• new InDet MC Simulation

Illustration of a typical step in a Kalman
filter aplication.
Andreas Salzburger OEPG, Vienna, 27/09/2005
5
And this should look like ...
Surfaces
TrackParameters
TrackParameters
ATHENA ALGTOOL
IExtrapolator
StraightLinePropagator HelixPropagator RungeKuttaP
ropagator STEP_Propagator
IPropagator
Andreas Salzburger OEPG, Vienna, 27/09/2005
6
Navigation
The TrackingGeometry enables navigation through
the TrackingVolumes which are taken as an
interface for material and magnetic field
access. If a boundary surface of a
TrackingVolume is hit, a simple projection of the
momentum vector onto the normal vector of the
surface at this point is enough to now the next
TrackingVolume, as the BoundarySurface object
owns TrackingVolume
m_outsideVolume TrackingVolumeArray
m_outsideVolumeArray TrackingVolume
m_insideVolume TrackingVolumeArray
m_insideVolumeArray
n
p
p
n
PixelNegEndcap
PixelPosEndcap
PixelBarrel
owns TrackingGeometry, handles search and
navigation
Andreas Salzburger OEPG, Vienna, 27/09/2005
7
Material Integration

• TrackingGeometry provides mechanism for Layer
  based and Volume based material description.
• Necessarily the TrackingVolume has to be able to
  hold a Layer or a LayerArray.
• Material properties should be retrieved from
  Common Detector
• Description !!
• (if possible)

Layer based material description (point-like
update)
Volume based material description (continuous
update)
steers, decides, configures
IMaterial Effects Updator
STEP_ Propagator
Extrapolator
ATHENA ALGTOOL
ATHENA ALGTOOL
ATHENA ALGTOOL
Andreas Salzburger OEPG, Vienna, 27/09/2005
8
Magnetic Field Access
Inside SCT Barrel Magnetic Field map
parameterized as a Solenoid
Outside SCT Barrel Realistic Field
SCT Barrel
Inner Detector

• The TrackingVolumes have to be identical or
  synchronized with MagneticVolumes
• The CTB 2004 showed be as flexible as possible,
  scaling, shifting, etc. etc.

For the moment, TrackingVolumes have a pointer to
the MagneticFieldTool which holds synchronized
MagneticFieldVolumes.
2nd part Propagation in realistic field
IMagnetic FieldTool
1st part Propagation in solenoidal map
ATHENA ALGTOOL
Andreas Salzburger OEPG, Vienna, 27/09/2005
9
Combined Testbeam 2004 A First Testbed
9 GeV, Pions, B0

• New Extrapolation has been successfully used in
  the new Kalman Fitter, the new Gaussian Sum
  Filter and a new Global Chi2 Fitter.

ATLANTIS Event Display
Residual gained for j
Full simulation geometry
To gain full validation of new software chain, a
controlled testbed was needed.
Andreas Salzburger OEPG, Vienna, 27/09/2005
10
Suddenly ...
... I remembered vaguely that Ive once written a
diploma thesis ...
... and I remembered what I didnt like so much
about the fast simulation ... ATLFAST is in
principle not a fast simulation, it is a fast
simulation and reconstruction! Which had as a
consequence that the fast simulation so far could
not be used for any studies that needed hit
information, or interaction with the offline
reconstruction software.
Andreas Salzburger OEPG, Vienna, 27/09/2005
11
Validation testbed InDetMC_Simulation

• We need a controlled validation testbed
• - full simulation turned out to be too
  complex for understanding single underlying
  processes

• How to create our perfect tracks for
  validation of the TrkTools

Navigator of Extrapolation can in principle
predict the track through the InnerDetector (as
connective TrackingGeometry exists).
trajectory
TrackParameters
ATHENA ALGTOOL
IExtrapolator
InDetMC_ TrackCreator
ATHENA ALGTOOL
Special flavour MC_MaterialEffectsUpdator (enric
hed with RndGenerator)
PRD, RIO, TSOS production
TrkTrack
Andreas Salzburger OEPG, Vienna, 27/09/2005
12
Interaction InDetMC_Simulation

• Allows interaction
• - resolution of detectors
• - efficiency (holes on track creation)
• - misalignment etc.
• - clustering, RIO_OnTrackCreation

InDet MC_Track Creator
ATHENA ALGTOOL

• Allows interaction
• - detector setup
• - material distribution

• Allows interaction
• - particle type
• - different models for E-Loss/MS

• Allows
• - simple switch between used fitting
• technique,
• - cross checks, etc.

IFitter
ATHENA ALGTOOL
Cake - Prolonging the Magic You turn the
screws
Andreas Salzburger OEPG, Vienna, 27/09/2005
13
Application W mass
measured to 0.014 at the Z pole
The W mass in the Standard model
radiative corrections, dominated by mt, mH
known to 0.0009 from muon lifetime
measured precicely at LEP (0.004 )
W mass is sensitive for Higgs mass, but also for
new physics such as SUSY particles due to loop
corrections.
Andreas Salzburger OEPG, Vienna, 27/09/2005
14
Application W mass measurement in ATLAS
In p-p collisions, the W mass can be obtained
from the leptonic decay channels Longitudinal
momentum of neutrino can not be measured,
measurement is done by an extraction from the
transverse mass distribution
where l e, µ
ET
pTe
The transverse neutrino momentum is reconstructed
using teh transverse momenta of the lepton and
the hadronic recoil of the underlying event. Due
to high event rate, the measurement error will be
completely determined by systematical error.
pTW
Neutrino
ET
Hadronic recoil
Andreas Salzburger OEPG, Vienna, 27/09/2005
15
Systematic Errors W mass measurement in ATLAS
Thanks to Esben Klinkby et al.
Andreas Salzburger OEPG, Vienna, 27/09/2005
16
The analysis W mass measurement in ATLAS
The kinematic boundary is smeared by the detector
response, this smearing has to be perfectly
understood (maerial distribution, magnetic field
properties) to guarantee the fit. Generating
large event numbers with full simulation and
different input W mass values and fit the edge in
comparison to taken data would be a strategy, but
this is almost inpossible due to the
time-consuming Geant4 simulation. The idea is to
retry that with the new InDetMC_Simulation, that
is up to two to three magnitudes faster. full
control of the sim/reco material full control
of the sim/reco magnetic field
Andreas Salzburger OEPG, Vienna, 27/09/2005
17
Current Status Simulation and refitting of
single tracks
InDetTrackingGeometry Navigation validation
finished
Material distribution calibration
framework established
Proof of principle with InDetMC_Simulation
and new Kalman Fitter

• redo with adapted material to different ATLAS
  layouts
• interface InDetMC_Simulation with generator
  output
• feed fast simulated and reconstructed files into
  existing W mass measurement
• framework
• chagne sim/reco material and deduce the
  systematics

Todos
Andreas Salzburger OEPG, Vienna, 27/09/2005
18
Conclusion so far

• New Track extrapolation scheme has been
  established
• - used in CTB 2004 reconstruction
• - base of many new fitters and filters in the
  ATLAS offline software
• - also used in physics analysis applications
  already
• Two basic novelties
• - STEP propagator (continous material effect
  update)
• - Full connective TrackingGeometry
  (Navigation)
• Navigation enables new fast MC Simulation for
  ATLAS
• - more than smearing can be done
• - used for various validation studies already,
  but need improvement
• Prospect of a future W mass measurement
• - can we answer the questions of lepton energy
  scale ?

Andreas Salzburger OEPG, Vienna, 27/09/2005
19
Backup slides
Andreas Salzburger OEPG, Vienna, 27/09/2005
20
Material comparison Geant4 / InDetTrackingGeometry
Atlas release 10.2.0
Andreas Salzburger OEPG, Vienna, 27/09/2005
21
THE EDM AFFAIR
working title THE RTF REPORT
The ATLAS offline software has been for a long
time characterized by (mainly two) black-box like
reconstruction programs with no interaction no
possbility to compare internal algorithms and
assumptions. Severall attempts to create a
common ATLAS Event Data Model to guarantee
interchangeability between programs from
different authors failed in the past (RD Event),
etc.
Despite the fact that no-one ever came back who
tried to do this task, a young group of Fellows
and PhD students, inpired by the guideline of the
Final Report of the Reconstruction Task Force
(RTF), tried to do the undoable and ... ...
after a long walk, after barriers and fallbacks
...
Andreas Salzburger - Ibk, 02/08/2005
22
THE EDM AFFAIR
The new Track class was born, a
subdetector-indpendent container for what should
be known as a track. By January this year, with
the introduction of the last converter of a major
reconstruction program to the new track class,
the battle was won (so they say). Today, this
gives me the small overhead of being in charge
of following EDM/Geometry/MagneticField packages
TrkDetDescr
TrkMagneticField
TrkEvent
TrkDetDescrUtils
TrkMagFieldUtils
TrkEventUtils
TrkSurfaces
TrkEventPrimitives
TrkVolumes
TrkMeasurementBase
TrkGeometry
TrkSegment
TrkDetElementBase
TrkParameters
Then ... I just had a little extrapolation
package left ...
Andreas Salzburger - Ibk, 02/08/2005