Transcript Document
Performance of the ATLAS
ID Reconstruction
Nectarios Ch. Benekos
CERN/ATLAS
EESFYE - HEP 2003 Workshop, NTUA, April 17-20, 2003
OUTLINE
ATLAS Inner Detector
Pattern Recognition Programs
xKalman
iPatRec
Fitting Method in iPatRec
Material Tuning
Performance studies
Conclusions
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
1
Diameter
25 m
Barrel toroid length
26 m
Endcap end-wall chamber span
46 m
Overall weight
7000 Tons
ATLAS Coordinates
XYZ right handed coordinate system with
Z in beam direction
Barrel + end-cap
inner detector
Nectarios Ch. Benekos
Radius [m]
1.15
Length [m]
6.8
h-coverage
|h|<2.5
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
2
The ATLAS ID
ATLAS Tracker
A side view ID layout
Requirements of the ID Reconstruction:
to reconstruct efficiently the tracks and vertices in an event
to perform, together with the calorimeter and muon systems,
electron,pion and muon identification
to find short lived particle decay vertices.
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
3
The updated ATLAS ID layout
Updated ID Layout:
main change is insertable pixel layout:
to accommodate construction delayed1 year later installation
consequences:
increased structural material (> 6m long cylinders)
>double material at low radius (insertable + realism)
b-layer: same modules as outer layers
pixel size increased from 50x300 mm2(TDR)50x400 mm2
change of the b-layer radial position 4350.5 mm
(due to the change in outer diameter beam pipe 5069.2 mm)
SCT small changes to forward layout
to increase inner radius in order to allow insertable pixels
TRT reduced straw length(occupancy) in endcaps
the continuous tracking of the TRT is approximated using 4 discrete layers
The updated initial layout (low lumi) has:
only 2 pixel layers +
2(+/-) pixel wheels instead of
3 pixel layers + 3(+/-) pixel wheels
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
4
Requirements of any track reconstruction algorithm
Find the tracks of particles in the detector
Introducing the minimum number of fake tracks
pattern
recognition
Give best estimation of the tracks’
actual momenta
direction, slope (cot (q)) of the track
Vertex finding
impact parameter estimation
track
fitting
Track fitting to minimize c2
measures how close the measured parameters are to what they are assumed
to be from a particular fit hypothesis (e.g., helical trajectory)
Track fitting would be trivial if it was not for complications arising because:
of multiple scattering
energy loss
non-uniform magnetic filed, ….and of course
IF we understood our detectors PERFECTLY.
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
5
ATLAS ID Pattern recognition algorithms
Two inner detector pattern recognition and track
reconstruction packages based on two different
techniques are existing in ATLAS:
o xKalman is a pattern recognition package
based upon a Kalman –filter smoother
formalism for finding and fitting tracks in the
inner detector.
o iPatRec uses a helix fitting method.
Its basic strategy is to initiate track finding from
space-points and fit these tracks using an iterative
method based on Newton-Raphson technique
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
6
xKalman
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
7
iPatRec
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
8
iPatRec
Searches for tracks using SP formed
in Pixel and SCT
Reconstruction is performed within a “narrow canonical
raod” joins Vxregion to a Sdregion on the outer surface of ID
Seeds can be:
o e/g candidates from EM calo,
o jets from HAD and,
o muon tracks found in the external muon detectors.
Tracks extension into TRT
detector after passing quality cuts
Track fitting using c2 minimization fit
also TRT hits are included by a histogramming
method in a narrow road around the reconstructed
helix of the track
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
1
8
iPatRec: stand alone pattern recognition (cont.)
o form space points from matching f and z hits :
o find up to 7 space-points on different layers that might form a track
The points are required:
• to be close enough azimuthally
• to lie in a “conical narrow road” defined as a+b/pT
(multiple scattering term)
• tracks extension into TRT detector after passing
quality cuts
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
9
Introduction to Track Fitting
The trajectory of a particle moving in a uniform magnetic field
with no multiple scattering and negligible bremsstrahlung radiation
is described by a helix.
Basically a helix can be decoupled into:
o moving along a circle in the xy-plane
(3 points needed to define it) and
1
r a0 0 r
r2
2 Rcurv
o in the rz plane by a straight line:
(2 points needed to define it)
z z0 cot( )r
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
10
Fitting a model to data using c2 minimization
In order to start fitting a track, one needs two things:
o a model which approximates the trajectory of the tracks
o an understanding of the detector accuracy(resolution)
Track fitting : is a procedure to determine the helix parameters by fitting a set of
coordinates(measurements) measured in a tracking detector to a helix.
We want to fit a model :
o with M parameters aj
o to a set of N uncorrelated measurements yi with error si.
o fi(a) is the expected i-th coordinate when the helix parameter vector is
a[q/pT,tanq…] for yi
2
yi f i (a )
2
Minimizing the c2 to determine the values of aj
c
si
i 1
N
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
11
Fitting a model to data using c2 minimization (cont.)
N
yi f i (a ) f i (a )
c 2
0
2
2
a
si
a
i 1
for a linear model :
o the solution is independent of the starting estimator a0 and
o NO iteration is needed
for a non-linear model (helix) one needs to iterate.
o it gives the correct answer
o i.e. converges to the global minimum, if a0 is
sufficiently close to a l
a0
al
so called Newton-Raphson method
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
12
Generalization
This method is global in the sense that it fits all the measurements at the same time
IF all measurements are independent of each other, the execution
time is ~ number of measurements (n)
BUT
IF we have correlations between measurements the covariance matrix
will contain non-diagonal terms
and inverting it becomes VERY time consuming for large n
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
13
Particle Interactions with matter - Energy Loss
The trajectory of a charged particle is affected by any material
several types of secondary interactions between particles and
material may occur.
Therefore energy loss and multiple scattering have to be applied to the track fitting.
at low energies ionization (described by Bethe-Bloch formula) dominates:
dE
1 1 2me c 2g 2 2 max
2 Z
2
kq 2 ln
Ekin
dx
A 2
I2
2
at high energies, bremsstrahlung dominates
Radiation length:
716.4 gcm2 A
X0
287
Z Z 1 ln
Z
o Mean distance over which
a high energy e- loses all but
1/e of its energy
by bremsstrahlung.
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
14
Multiple Scattering in iPatRec c2-fit
Mostly due to Coulomb scattering from nuclei
x
13.6 MeV
x
0
z
1 0.038ln
cp
X0
X 0
For small angles roughly Gaussian distribution
Thickness of the
scattering material
in radiation lengths
Multiple Scattering(MS) in Track Fitting
multiple
scattering
angles pMSadditional parameters pMS,
MSThe
at the
detector
planes introduces
Full description of the path of a
+
o i.e. the two (fitted) deflections particle
(Df,Dcotq)
at each
through
thedetection
detector plane:
1, Df2,Dcotq2,…,Dfn,Dcotqn)
Helix opareameters
p
pMS=(Df1,Dcotq
Scattering centres are expensive
typically # parameters = 2N+5 (5 track params + 2 x N scat. angles/scattering centre)
o (instead of 5 params ,ignoring material effect)
The scattering processes in the different planes(centres) are independent
from each other
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
15
Tuning Multiple Scattering in iPatRec
Method :
pulls on 5 perigee parameters
residual for a track parameter a:
r ameas atrack
where atrack is the result of the fit
pull for a track parameter a is defined as:
pulla
ameas atrack
sa
• tune material to give :
mean=0 (dE/dx)
sigma=1 (X0)
Nectarios Ch. Benekos
IF the fit is reasonable and errors
are correctly described
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
16
Tuning Multiple Scattering in iPatRec
Procedure :
• need lowest Etrack material effects dominate
• high statistics (to cut on limited region with uniform material)
• start with tuning inner layers then work outwards
• reduce # of layers lower PT for material to dominate
•start with barrel as already ~ 1/3 of phase-space (uniform material)
|h|<0.8 , total acceptancy to 2.5)
Plots in the following using first 7 layers (Pixels + SCT) only
•1/PT
so plotting pulls can see IF errors are correct
•1/PT pull
or over/under estimated !
•a0 (impact parameter d0)
•a0 pull
Increase material - tuned to give all 5 parameters fitting correctly in barrel
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
16
D1/pT)/(1/pT)
|h|<0.8
Well centered
1.6<|h|<2.5
• single muons tracks pT =200 GeV/c
• Pixel + SCT using iPatRec
D1/pT)/(1/pT) ~ 9% (~7% in TDR) in barrel
~ 20% (~15% in TDR) in endcap
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
17
D1/pT)/(1/pT)
|h|<0.8
1.6<|h|<2.5
•single muons tracks pT =1 GeV/c
• Pixel + SCT using iPatRec
D1/pT)/(1/pT) ~ 1.8% in barrel
~ 2.7% (~3% in TDR) in endcap
Increased material thickness !
Systematic shifts on mean dE/dX underestimated
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
18
Impact parameter resolution
N
N
|h|<0.8
1.6<|h|<2.5
DRf
• single muons tracks pT =200 GeV/c
• Pixel + SCT using iPatRec
Nectarios Ch. Benekos
DRf
Impact parameter ~ 13-15 mm
(TDR 11 mm)
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
19
Impact parameter resolution
N
N
|h|<0.8
1.6<|h|<2.5
DRf
DRf
• single muons tracks pT =1 GeV/c
• Pixel + SCT using iPatRec
Nectarios Ch. Benekos
Impact parameter ~ 100 mm / √(sinθ)
(TDR 73 mm / √(sinθ)
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
20
Tuning of pull distributions (plot before corrections)
N
N
|h|<0.8
1.6<|h|<2.5
DRf
DRf
N
0.8<|h|<1.6
• single muons tracks pT =1 GeV/c
• Pixel + SCT using iPatRec
Pull
~ .87 in barrel
~ .91 in endcap
Overestimated X0 in b-layer
guessed 3% X0 corrected
Nectarios Ch. Benekos
DRf2003 Workshop, NTUA, April 17-20, 2003
EESFYE – HEP
21
Tuning (cont.)
N
N
|h|<0.8
1.6<|h|<2.5
DRf
DRf
200 GeV muons using Pixel+SCT
Rel. 6.0.1. using iPatRec
N
Pull
0.8<|h|<1.6
~ 1.0 in barrel
~ .91 in endcap
Errors slighlty over-estimated at
higher h
Nectarios Ch. Benekos
EESFYE – HEP
DRf2003 Workshop, NTUA, April 17-20, 2003
22
Momentum resolution vs eta
In the absence of multiple scattering:
1
s 2 AN
pT L
In the presence of multiple
scattering:
1
BPT
s APT
pT sin q
pT
o reducing further the pT, the effect of multiple scattering
is starting to dominate and
o at pT=1 GeV/c multiple scattering is dominating at all
|h| with a marked degradation in resolution and
with degrading resolution with increasing |h|.
o non-uniform magnetic field correction in forward region
(higher h )
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
23
Eta dependency on impact parameter resolution
s d 0 APT
APT 14 mm
BPT 100mm
Nectarios Ch. Benekos
BPT
pT sin q
(TDR 11 mm)
(TDR 73 mm / √(sinθ))
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
24
Conclusions
The single track reconstruction performance of the ATLAS ID has been
investigated using the simulation of single muons.
Material tuning in iPatRec
resolution studied of the impact parameters, over the complete studied |h|
and pT-range
Measurement errors understood and correctly accounted
Due to the updated ID layout (more realistic material) the
impact parameter resolution was found to be:
o ~ 100 mm (as a function of sinq) for pT=1 GeV/c
(multiple scattering effect is dominated)
o and ~14 mm for pT=200 GeV/c
Nectarios Ch. Benekos
EESFYE – HEP 2003 Workshop, NTUA, April 17-20, 2003
25