Jet reconstruction with neutrals from full simulation

1
Jet reconstruction with neutralsfrom full
simulation
Magali Estienne IPHC Strasbourg
EMCal Meeting - Frascati March 13. 2007
2
The actual jet finder package JETAN
More information on JETAN see M. L. Noriegas
talk in Bologna June 2006
READER
JETS
JET FINDER

• ESDs (TPC/ITS)
• Pythia events
• New reader to mix HIJING and PYTHIA

2
M.E. 03/13/2007 EMCAL Meeting - Frascati
3
Role of EMCal in jet reconstruction
Case of a monochromatic 100 GeV jet generated by
PYTHIA Cone size for reconstruction Rc 0.4
In the EMCal acceptance

• - Improves the resolution for jet reconstruction
• Trigger jets

Monochromatic 100 GeV jets
3
M.E. 03/13/2007 EMCAL Meeting - Frascati
4
Physical information from EMCal
f
h

• AIM merge TPC/ITS and EMCal information
• In TPC/ITS information from particle tracks
  (px, py, pz, E)
• gt Previously filled in a fMomentumArray
• In EMCAL information from digits or clusters
  (eta, phi, Eintegrated)
• gtTPC and EMCAL information filled in a SAME
  ARRAY fUnitArray

- Limited to EMCal acceptance - Extended to work
on all the TPC acceptance
4
M.E. 03/13/2007 EMCAL Meeting - Frascati
5
The jet finder in practice

• Previously, the philosophy of the code was the
  following
• for each new event in the system (ITS/TPC), a
  fMomentumArray TLorentzVector object was filled
  with px,py,pz and E
• all the track information per event was then
  stored in a ?-? grid which was a Lego plot
  TH2D object
• then, the information was dumped from the histo
  into cells. These cells were then given to the
  UA1 jet-finder

• To merge the (ITS/TPC) and EMCal information,
  creation on a new TObject called fUnitArray which
  is directly filled with all the particle
  information per event. Should avoid
• 1 to use an intermediate
  fMomentumArray
• 2 to use an intermediate histogram
• 3 (maybe) to loose information
  because of possible detector dead-zones
• Given the particle ID in a Grid, ?, Dh, ?, Df, E,
  flags, px, py, pz can be stored for all
  particles in both TPC and EMCal per event

• Main modifications/additions
• In the reader
• 1 new class to set the desired grids only one
  time at the beginning AliJetGrid
• new classes of TTask type to read information
  from tracks or digits or clusters
• AliJetFillUnitArrayTracks, AliJetFillUnitArrayEMC
  ALDigits, AliJetFillUnitArrayEMCALClusters
• In the finder
• Two technics kept
• 1 class added to use Selector structure
  AliJetSelector Analysis framework

5
M.E. 03/13/2007 EMCAL Meeting - Frascati
6
Class relationships
AliJetHadronCorrection
AliJetUnitArray
AliJetGrid
AliJetSelector
6
M.E. 03/13/2007 EMCAL Meeting - Frascati
7
Set an initial grid
New/modified classes (I)
TPC - EMCal

• AliJetGrid
• 2 types of grid can be setted to deal with
  different binnings
• extended to include detector dead-zones
• Functions included

EMCal
// Getter Int_t GetMatrixIndex(Int_t iphi, Int_t
ieta) Int_t GetIndexIFromPhi(Double_t
phi) Int_t GetIndexJFromEta(Double_t eta) Int_t
GetIndex(Double_t phi, Double_t eta) Void
GetIJFromIndex(Int_t index,) Void
GetEtaPhiFromIndex(Int_t index, ) Int_t
GetNEntries() Int_t GetIndexFromEtaPhi() //
Setter
Binning given by the function
TowerIndexFromEtaPhi() or by an other grid
Binning given by the grid initially declared

• In progress fill two full grids and then think
  how to merge them
• Effect of different cell sizes on finder

7
M.E. 03/13/2007 EMCAL Meeting - Frascati
8
A fUnitArray object
New/modified classes (II)

• AliJetUnitArray (AliJetFinderTypes)
• - All the quantities stored from different
  detectors are in the same units
• Its size number of EMCAL towers TPC bins
• (gt Recently extended in order to work on
  all detectors)
• The following quantities can be setted in or
  taken from the Array
• - Digit(/Cluster)/Track h
• - Digit(/Cluster)/Track f
• - Digit(/Cluster)/Track Energy
• - Digit/Cluster/Track ID
• - Detector flag, Cut flag, In/Out jet flag
• - Bin Size Dh, Df
• - Track label
• -
• With the functions
• SetUnitEta(), SetUnitPhi(), SetUnitEnergy(), etc.
• GetUnitEta(), GetUnitPhi(), GetUnitEnergy(), etc.

8
M.E. 03/13/2007 EMCAL Meeting - Frascati
9
For hadron correction
New/modified classes (III)

• In order to avoid double counting between TPC
  tracks selected and EMCAL digits or clusters gt a
  hadron correction has to be applied.
• The following classes have been added
• AliJetHadronCorrection, AliJetHadronCorrectionv0
  and AliJetHadronCorrectionv1

Study with simulations in progress -
Simulation of p/- and p/p in precise (eta,phi)
locations - Eta/Energy mapping and
parametrisation of the detector phase space -
After track matching, extract correction factor
to be applied - Can be improved with PID
- Try other technics
9
M.E. 03/13/2007 EMCAL Meeting - Frascati
10
Fill charged neutral information
New/modified classes (IV)
AliJetESDReader

• Flag added to choose the detector configuration

• 2 alternatives
• switch in the function FillMomentumArray() which
  calls the functions FillUnitArray directly
  included in the class AliJetESDReader
• switch in the function FillMomentumArray() which
  calls objects of TTask type where the object
  fUnitArray is filled

// Fill information from TPC tracks from
ESDs void FillMomentumArray(Int_t event) //
Fill information from TPC and EMCAL from
ESDs void FillUnitArrayFromTPCTracks(Int_t
event) void FillUnitArrayFromEMCALHits(Int_t
event) void FillUnitArrayFromEMCALDigits(Int_t
event) void FillUnitArrayFromEMCALClusters(Int_t
event) void InitUnitArray() // Correction of
hadronic energy virtual void SetHadronCorrection(I
nt_t flag) virtual void SetHadronCorrector(AliEMC
ALHadronCorrectionv1 corr) // For EMCAL digit
calibration Virtual Float_t Calibrate(Int_t amp,
Int_t cellId) Void Exec(Option_t option)
New classes
10
M.E. 13/12/2006
11
Other comments
New/modified classes (V)
AliUA1JetFinder
// Find jets functions void FindJetsTPC() //
Previous FindJets() which is using the
fMomentumArray //
filled to fill then dump a fLego plot.
// It has to be called when
looking for jets in TPC/ITS
// acceptance void FindJets() // New
function which is directly using a fUnitArray to
give // the jet finder
code the desired values
other minor changes in classes

• For the moment, fMomentumArray kept but I will
  not use it anymore to fill the fUnitArray object
  to gain time
• fUnitArray technics takes more time than
  fMomentumArray but in PbPb it should diminish
• fUnitArray more convenient to store information
  flags, eta, phi, P, E, in/out jet
• AliJetSelector structure implemented as well as
  the Andreis Analysis framework.

11
M.E. 03/13/2007 EMCAL Meeting - Frascati
12
2 algo comparison in JETAN
100 GeV Gamma/Jets simulated Jets in EMCal
acceptance Pythia pp evts _at_ 5.5TeV Jet finder
parameters Pt cut 2. GeV R 0.4 MinCellEt
0. Et seed 4. GeV MinJetEt 14 GeV Fake jets
removed
Good agreement
12
M.E. 03/13/2007 EMCAL Meeting - Frascati
13
W/o and W EMCal
- Gamma-jet simulation - Jets in EMCal
acceptance - Fake jets removed
TPC TPCEMCal
Mean energy increases compared to TPC alone BUT
RMS increases !!!
13
M.E. 03/13/2007 EMCAL Meeting - Frascati
14
Resolution still too bad
Entries
Entries
Energy (GeV)
Energy (GeV)
Emean 64.2/-0.6 Resolution 47
Emean 67.8/-1.1 Resolution 42
14
M.E. 03/13/2007 EMCAL Meeting - Frascati
15
Reasons for bad resolution and mean energy
I am trying to identify possible sources of
problem

• Related to my code and the way I fill the
  fUnitArray
• Step by step check of the code from the digit in
  Reco to the ESDs
• Related to the way digits are stored in ESDs
  there was a problem (digit ID higher than the
  number of digits in EMCal). Aleksei has worked on
  it.
• AliRoot HEAD installed and production of 10000
  Gamma-Jets (100GeV) events and 10000 100 GeV jets
  events.
• Bug fixed but worth resolution than before !!!
• Related to hadron correction
• Should have minor effect
• Related to the physics itself ?
• Strategy start to understand what is going on
  on simple cases
• gt Simulation of 100 GeV p0 in EMCal acceptance
• gt Simulation of 100 GeV g in EMCal acceptance

15
M.E. 03/13/2007 EMCAL Meeting - Frascati
16
100 GeV g in EMCal acceptance
Too high energy Counted twice ?
TPCEMCal
Simulation of 2500 events with 1 g emitted inside
the EMCal acceptance Reconstruction of 2370 gs
No more double counts
EMCal only
Tracks responsible for double energy count
TPC only
16
M.E. 03/13/2007 EMCAL Meeting - Frascati
17
TPCEMCal EMCal
Small differences !
Charged electron from double decay or conversion
? Double count of electrons ?
17
M.E. 03/13/2007 EMCAL Meeting - Frascati
18
Remove (true) matches

• True match if track label cluster primary
  index
• Remove matches by looking at true matches
• It removes high pt  jets  reconstructed

All particles
Matches removed

• How to proceed in true life ?
• 1 Do track matching
• Test of Alberto's track matching code
• 2 Identify electron with p/E cut g
  invariant mass ?
• 3 Remove their double energy contribution

18
M.E. 03/13/2007 EMCAL Meeting - Frascati
19
My to Do list !

• Questions to be solved
• Improve the resolution and the mean jet energy
  reconstructed
• - Check the code step by step
• - Identify problems with simple event cases
• - Cut on electrons
• Hadron correction to be correctly done with
  the new EMCAL geometry
• gt in progress but need to obtain first good
  resolution
• Make the code independent on AliEMCALGeometry
• gt class AliJetGeoDummy will be implemented if
  possible using TGeoManager.
• Run on the coming Jet-Jet, Gamma-Jet productions

19
M.E. 03/13/2007 EMCAL Meeting - Frascati
20
Backup
20
21
The cone algorithm
F
1. Estimate ETbkgd per cell from all cells 2.
Sort cells in decreasing ET 3. Until change in
Ebkgd between 2 iterations is smaller than a set
threshold - Clear the jet list and flag
cells outside a jet - jet-finding loop for
each cell, from the highest cell energy. If
Ei-Ebkgd gt Eseed and if the cell is not flagged
as being inside a jet
From Mercedes Lopez-Noriega

• Set the jet-cone centroid to be the center of
  the jet seed cell
• Using cells with ?(hi-h)2(fi-f)2 lt Rc of the
  initial centroid, calculate the energy weighted
  centroid
• Repeat until difference between iterations
  shifts less than one cell.
• Store centroid as jet candidate.
• Recalculate background energy using information
  from cells outside jets.

?

• Parameters
• - Mininum pT for tracks or energy for digits
• - cone radius Rc, Eminseed for the seed, Eminjet
  for a jet

• Input list of cells in an h-f grid sorted in
  decreasing cell energy Ei

22
M.E. 03/13/2007 EMCAL Meeting - Frascati
22
Background estimation methods in the cone
algorithm
F
Statistical method an average background energy
is obtained from N jet-free events Cone method
the energy in all grid cells outside the jet cone
is added and scaled with the inside/outside cone
area ratio Ratio method ratio of the average
energy in the grid with a pT-cut to the one
without a pT-cut. Event by event, the background
energy is estimated by multiplying the average
background energy of the event without pT-cut
with the calculated ratio
Cone area
Background area
?
23
M.E. 03/13/2007 EMCAL Meeting - Frascati
23
Compare with the ALICE PPR II
Fake jets removed
24
M.E. 03/13/2007 EMCAL Meeting - Frascati
24
Simulation in full TPC Acceptance
Simulation in f2p and hlt0.5
TPC TPCEMCal
Simulation in f2p and hlt0.5 BUT jets
reconstruction in the EMCal acceptance
TPC TPCEMCal
25
M.E. 03/13/2007 EMCAL Meeting - Frascati
25
Compare fMomentumArray/fUnitArray with TPC/ITS
only
In order to compare algorithms with a
fMomentumArray filled and a fUnitArray filled, a
cut on h (hlt0.5) and f has been introduced
before the candidate track selection to compare
quantities in the EMCal acceptance.
-gt 400 evts simulated -gt Only TPC -gt 104-125 GeV
Jet multiplicity
No lego
fUnitArray
EMCAL acceptance in h
EMCAL acceptance in f
Jet energy
Need to generate more events for better comparison
26
M.E. 03/13/2007 EMCAL Meeting - Frascati