All the definitions are cited from $ALICE_ROOT/TPC/doc/Definitions/Definitions.pdf
# of TPC clusters N_cls: A charged particle traversing the TPC induces a signal on a given pad-row. If the charge exceeds threshold and fulfills all necessary quality criteria, it is called a cluster. Therefore the maximum number of clusters per track is 159 which corresponds to the total number of pad rows in a given TPC sector.
# of TPC clusters in first iteration N_cls_iter1: the number of TPC clusters assigned during the first (inward) tracking iteration.
# of missing clusters N_miss: Some TPC clusters along the track trajectory can be missing, because their charge is below threshold(e.g. due to baseline shifts etc.). These missing clusters can be identified;ed by looking into the neighboring pad-rows, e.g. if there is no reconstructed cluster on pad row i, but clusters are found on the pad rows i-1 and i+1 (or i-r and i+r in general).
# of crossed rows N_eff: N_eff = N_cls + N_miss. This variable can be viewed proportional to the effectively sampled track length of a particle in the TPC. It is the relevant quantity for pt resolution of a track, because the pt resolution scales with 1/sqrt(N_cls) (statistics) and (N_eff)^2 (level arm).
Findable clusters N_find: The number of findable clusters is the number of geometrically possible clusters which can be assigned to a track.
Remarks
TPC clusters can be lost due to unknown reasons (e.g. dead zones, missing partitions and decays, etc) or because their charge is below threshold (caused by the front-end electronics; these clusters are called missing clusters). So the variable (N_cls)/(N_find) depends on the digital threshold, on multiplicity, energy loss, drift length and track angle. On the other hand (N_eff)/(N_find) is almost independent from energy loss, multiplicity and η. Supporting plots on 2010 PbPb data can be found in $ALICE_ROOT/TPC/doc/Definitions/Definitions.pdf
2012-01-05: Marco: The number nCrossedRows is based on the clustermap, which is a bitmap which has a flag for every padrow telling whether there was a hit found along the track trajectory or not. We then loop over this bitmap and count every hit and also counting neighboring empty rows, up to 2 away from a found hit. So, the number of crossed rows is at most (last row - first row) + a few. This measure is less sensitive to cluster finding efficiency than the number of clusters, because if you lose a cluster somewhere in the middle of the track; it is counted, as long as it has a found cluster in the neighboring row or the one next to it. The crucial point according to the TPC experts is that nCrossedRows 'fills in the gaps' if there is an occasional cluster missing and is therefore less sensitive to the details of the gas amplification and cluster finding efficiency. It does not extend the track beyond the first and last measured point (or at least not more than 2 points on each side). It does not 'correct' for large gaps in the track.
2012-01-19: Why would the ratio nCrossedRows/nFindableClusters be larger than 1? Marian: nFindableClusters can be under/overestimated by a large number of clusters for track near the sector boundaries. The ExB distortions are of the order of 2 mm at the inner radius of the TCP and 1 cm at the outer radius and are not fully treated in the calculation of nFindableClusters.
Track cuts
Definition
Hybrid tracks
Use good global tracks when they are available, otherwise use global constrained tracks
Cuts for good global tracks:
pT dependent cut on # of TPC clusters in the first iteration: if pt<20, N_max = 70 + 30/20 * pt. If pt>20, N_max = 100.
chi2 per TPC cluster in first iteration < 4
No kink daughters
Require TPC refit
Fraction of shared TPC clusters < 0.4
Require ITS refit
Chi2 per ITS cluster < 36
DCA_xy < 2.4cm
DCA_z < 3.2cm
At least one hit on SPD
Chi2 between TPC constrained and global < 36
Cuts for global constrained tracks
Same cuts as for the good global tracks shown above except no ITS refit or SPD hits requirement
Constrained to primary vertex to improve pT resolution
Good track pT resolution, uniform phi distribution, but track quality is mixed.
TPCOnly tracks
TPCOnly track cuts + constrain to SPD vertex
# of clusters in TPC > 70
Chi2 per cluster in TPC < 4
No kink daughters
DCA_xy < 2.4cm
DCA_z < 3.2cm
Constrained to SPD vertex
Uniform phi distribution, but bad pT resolution
Global/Gloden/Primary tracks
Global/gloden/primary tracks are selected by using standard track cuts: AliESDtrackCuts::GetStandardITSTPCTrackCuts2010(kTRUE,1)
# of crossed rows in TPC > 70
Ratio of crossed rows over findable clusters in TPC > 0.8
Chi2 per cluster in TPC < 4
No kink daughters
Require TPC refit
Require ITS refit
At least one hit on SPD
pT dependent DCA in x-y plane
DCA < 2cm in z
Good pT resolution, but non-uniform phi distribution
In the hybrid track cuts, there are two ways to cut on track quality: NTPCclsIter1 (cut on # of TPC clusters in the first iteration) vs NCrossedRow (cut on # of crossed pad rows).
Pass2
This data set suffers from the bug of bookkeeping the TPC cluster map. So all the tracks with more # of TPC clusters than # of crossed pad rows are removed.
Before cut
All the following plots are produced with base cuts: meaning standard hybrid track cuts without cut on NTPCclsIter1
Correlation of NTPCclsIter1 and NCrossedRow in different track pt bins
Distribution
Distribution of NTPCclsIter1 in different track pt bins. Official cut is: if pt<20, N_max = 70 + 30/20 * pt. If pt>20, N_max = 100.
Distribution of NCrossedRow in different track pt bins. Official cut is: N_max = 70.
Distribution of NCrossedRow/NFindableClusters in different track pt bins. Official cut is ratio > 0.8
Correlate with track phi
Correlation of NTPCclsIter1 with track phi in different track pt bins
Correlation of NCrossedRow with track phi in different track pt bins
Correlate with track momentum resolution
Correlation of NTPCclsIter1 with track momentum resolution in different track pt bins. Compared to the plots below, it looks like the NTPCclsIter1 cut cuts away more track with bad resolution.
Correlation of NCrossedRow with track momentum resolution in different track pt bins
Effects of the two different cuts
The official cuts:
Cut on NTPCclsIter1: if pt<20, N_max = 70 + 30/20 * pt. If pt>20, N_max = 100
Cut on NCrossedRow: N_max = 70; ratio of crossed rows over findable clusters > 0.8
Effects on track pt distribution: within statistical uncertainties, the difference is < 5% and cut on NCrossedRow includes more tracks.
Effects on track momentum resolution: cut on NCrossedRow clearly selects more tracks with bad resolution
Effects on raw jet yield: within statistical uncertainties, the difference is within a few percentage.
Summary
Given the fact that NCrossedRow cut includes more tracks with bad resolution, I think it is better to use the cut on NTPCclsIter1.
Hybrid vs TPCOnly
Pass2
Track pT resolution
Left: hybrid track: three band shows up. From the best to worst:
Global tracks with ITSrefit and a hit in one of the two or both SPD layers
Global tracks with ITSrefit but no hits in the 2 SPD layers
Global tracks without ITSrefit (note that these tracks can have a hits in the SPD layers but it is not required)
Track pT distribution in full phase space
TPCOnly vs global tracks
Track pT:
Reconstructed track pT: from AliESDs.root
Constrained track pT: after updated by constraining to SPD vertex
Delta pT vs DCA_xy: the peak of DCA distribution at 0 is not artificial, meaning these tracks do have very small DCA
Delta pT vs DCA_z
Delta pT vs # of TPC clusters:
Delta pT vs track eta
Pass2
MB: Constrained pT vs reconstructed pT
EMC-trigger: Constrained pT vs reconstructed pT
Pass1
Number of tracks per event in triggered and MB events.
pT distribution of tracks in triggered and MB events.
Phi distribution of tracks in triggered and MB events.
Eta distribution of tracks in triggered and MB events.
To have a closer look into the bump at eta=0, I show eta distribution of TPCOnly tracks in MB events with pT>0,1,2 [GeV/c]. As shown in the plot, this bump exists for higher pT tracks
Pass2
Hybrid tracks
Track pT distribution
Track phi and eta distribution
TPCOnly tracks
Track pT distribution
pT distribution of all the tracks in both HT and MB events.
pT distribution of tracks in EMCal acceptance both in HT and MB events (left) and their ratio (right).
Track phi-eta distribution
Track phi-eta distribution
Top row -> HT events; bottom row -> MB events
From left to right: different pT cuts -> 0, 0.2, 1, 5 GeV/c
Pass3
TPCOnly tracks
Ratio of pos to neg
Ration of positive tracks to negative ones:
Pass1
Global tracks
pT distribution
pT distribution of tracks:
phi-eta distribution
This plot shows the phi-eta distribution of global tracks pointing to EMCal in triggered and MB events with different pT cut.
Upper: triggered event, pT>0, 0.2, 1, 5
Lower: MB event, pT>0, 0.2, 1, 5
More activity shows up in area corresponds to EMCal SM 5
MC study
TPCOnly tracks
Tracking efficiency
Cut on |eta| < 0.9 both for generated and reconstructed tracks
Use primary charged pion sample
Pure tracking efficiency with (left) and without (right) upper kinematic cut, namely 40GeV/c, on the reconstructed tracks.
Track pT resolution
This plot is created by matching reconstructed tracks to the generated primary tracks.