The STAR Collaboration soon plans to implement a tracking algorithm with the goal to provide better versitility and efficiency of TPC tracks. However, in an effort trouble-shoot problems prio to a full application, a charge was given to the PWGs to review the new tracking algorithms. Therefore, a jet analysis has been perfomed on all 4 tracking alogrithms (Sti, StiCA, Stv, and StvCA) to compare the effects they have on physics observables. This blog provides a decription of the data produced, the jet alogrithm applied, a comparison of the different tracking alogrithms, the possible effects any differences have on physics, and then plans for further endeavors.

Data Produced:

 A total of 1.2M events were produced of the year 2009 pp500 st_physics_* data stream for each tracking algorithm. This amount proved to provide sufficient statistics to examine jet distributions. Also, 50K pp 500GeV simulated events were thrown using pythia 6.423 with pileup 1MHz. However, this data has not yet been examined with a jet analysis, but plans to be done soon (see Further Endeavors).

The Jet Algorithm:

The following cuts are applied to the tracks and towers which are used to create a list of proto-jets:

  anapars12->setTowerEnergyCorrection(new StjTowerEnergyCorrectionForTracksFraction(1.00));

  // TPC cuts
  anapars12->addTpcCut(new StjTrackCutFlag(0));
  anapars12->addTpcCut(new StjTrackCutNHits(12));
  anapars12->addTpcCut(new StjTrackCutPossibleHitRatio(0.51));
  anapars12->addTpcCut(new StjTrackCutDca(3));
  anapars12->addTpcCut(new StjTrackCutDcaPtDependent); // dca <=2 for track pT< 0.5 & dca <3. -2.*pT for track pT > 0.5 and dca < 1 for track pT >=1

  anapars12->addTpcCut(new StjTrackCutChi2(0,4));
  anapars12->addTpcCut(new StjTrackCutPt(0.2,200));
  anapars12->addTpcCut(new StjTrackCutEta(-2.5,2.5));
  anapars12->addTpcCut(new StjTrackCutLastPoint(125));

  // BEMC cuts
  anapars12->addBemcCut(new StjTowerEnergyCutBemcStatus(1));
  anapars12->addBemcCut(new StjTowerEnergyCutAdc(4,3)); // ADC-ped>4 AND ADC-ped>3*RMS
  anapars12->addBemcCut(new StjTowerEnergyCutEt(0.2));

  // EEMC cuts
  anapars12->addEemcCut(new StjTowerEnergyCutBemcStatus(1));
  anapars12->addEemcCut(new StjTowerEnergyCutAdc(4,3)); // ADC-ped>4 AND ADC-ped>3*RMS
  anapars12->addEemcCut(new StjTowerEnergyCutEt(0.2));

  // Jet cuts
  anapars12->addJetCut(new StProtoJetCutPt(5,200));
  anapars12->addJetCut(new StProtoJetCutEta(-100,100));
 

Then the Mid-point cone algorithm, a cone having a radius R=0.7) is applied to the proto-jet list to form the jets.

 // Set cone jet finder parameters
  StConePars* conepars = new StConePars;
  conepars->setGridSpacing(105,-3.0,3.0,120,-TMath::Pi(),TMath::Pi());
  conepars->setSeedEtMin(0.5);
  conepars->setAssocEtMin(0.1);
  conepars->setSplitFraction(0.5);
  conepars->setPerformMinimization(true);
  conepars->setAddMidpoints(true);
  conepars->setRequireStableMidpoints(true);
  conepars->setDoSplitMerge(true);
  conepars->setDebug(false);
 

The JP1, JP2, and BHT3 triggers were examined, but for conciseness only the JP1 trigger is shown in this blog. If you want to examine all the distributions of the other applied triggers feel free to look inspect this link. After the Jet Finder was applied to the produced data, a comparsion of the tracking alogrithms could be performed as seen in the following section.

Results:

In this comparision, the different tracking algorithm distributions, Sti (black), StiCA (red), Stv (blue) andStvCA (greem), have been overlayed. In total there are 15 histogram plots followed by an informative statistics table, which effectively demonstrates any differences the tracking algorithms has on the jet produced.

Figure 1 : Jet pT Spectrum

Figure 2: Jet eta spectrum

 

Figure 3:Jet phi spectrum

 

Figure 4: Jet Rt Spectrum

 

Firgure 5: Number of Jets per Event

 

 

Figure 6: Raw Track pT distribution

 

Figure 7: Track pT distribution normalized by the total number of tracks.

 

Fig 8: Track pT spectrum normalized by the total number of jets.

Figure 9: Track eta Distributions

Figure 10: Track phi Distributions

Figure 11:Track chi^2 distribtution

Figure 12:  Number of tracks per Event

Figure 13: Tower Energy Distribution

 

Figure 14: Jet vertex x-position

Figure 15:  Jet vertex y-position

Figure 16: Jet vertex z-position

 

Trigger JP1:	Sti Tracks	StiCA Tracks	StvTracks	StvCATracks
# of jets	55499	63336	51650	53035
<jet pT>	11.6989	11.7802	11.3521	11.2933
<jet eta>	0.00309	0.00339	0.00074	0.000542
<jet phi>	0.01276	0.01142	0.00708	0.015006
<jet Rt>	0.50335	0.48093	0.58184	0.59648
<jet number>	1.84812	1.89487	1.70445	1.67581
# of tracks	272470	330133	189510	272470
<track pT>	1.17585	1.1632	1.29561	1.3292
<track eta>	0.000671	0.00543	-0.00576	-0.00508
<track phi>	-0.01348	-0.02449	-0.04421	-0.04312
<track chiSq>	1.23854	1.31078	1.68514	1.78493
<track number>	5.06704	5.35027	3.90019	3.69263
# of towers	444969	493260	453416	444969
<tower E>	0.86552	0.863732	0.87695	0.865523
# of vertices	31798	35242	32181	33100
<vertex x-pos>	0.41192	0.41321	0.41269	0.41255
<vertex y-pos>	0.01383	0.01408	0.014102	0.01406
<vertex z-pos>	-2.6488	-2.2612	-2.7225	-2.6812

 Table 1: Statisitics for the histograms shown above (Fig 1-15)

 

Average Values	Sti Tracks	StiCA Track	Stv Track	StvCA Tracks
# of jets	1489	1550	811	787
<jet pT>	6.4645	6.4229	6.41545	6.5312
<jet eta>	-0.017217	-0.013134	-0.01864	0.000114
<jet phi>	0.012259	-0.03887	0.100873	0.045450
<jet number>	3.17663	3.1800	2.27123	2.31893
# of tracks	6127	6703	1900	1876
<track pT>	6.4645	6.4229	6.41545	6.5312
<track eta>	-0.012393	-0.002930	-0.02791	0.004524
<track phi>	0.0039206	-0.063453	0.064331	-0.01378
<track chi2>	1.42963	1.59322	2.08618	2.06762
# of vertices	18705	18827	17876	17937
<vertex z-position>	1.4715	0.990976	0.39557	0.19136

 

Observations and Analysis:

First of all, focusing only on the jet spectra (Fig 1, 2, 3, 4, 5 and Rows 2-7 of Table 1), one clearly sees that the StiCA algorithm produces more jets than the others. Also, it appears that there is very little difference between the Stv (blue) and the StvCA (green) distributions. The jet eta and phi distributions appear reasonable. However, the neutral energy ratio (Rt = ratio betwee the total jet tower energy and the total jet energy) plot show significant differences between the Sti and Stv tracking. The Stv distriribution are clearly shifted higher indicating that more heavy in neutral engery (NE) jets are produced. These NE jets are typically not used in a cross-section analysis, which means that the Stv Tracking algorithms will reduce valuable statistics. 

However, before we thrown the Stv algorithms under the bus, lets examine (Fig 6 7 8 9 10 11 12 and Rows 8-13 of Table 1) the track spectra. For clarity purposes these are the tracks within jets and not all the tracks produced by the TPC. A vital observable for the tracking review is the track pT distribution (Fig 6 7 8), which indicates that Sti and StiCA algorithms produce significantly more low pT tracks than the Stv algorithms.  This explains why we see the different distributions in the jet Rt. In the Stv algos there are fewer number of tracks (granted they are low pT), causing more NE jets to be formed.

The contaminations of pile-up of the Run 9 pp 500GeV data is a major concern. Pile-up in the TPC is created in two ways: recontruction of out-of-time tracks and multiple hard scattering within a single event. The former suspected to be larger contributing factor. Nonetheless, these out-of-time pile-up tracks tend to be reconstructed as low pT tracks. Therefore, it would be a benefit to a cross-section analysis if the Stv algorithms are less sensitive to pile-up tracks. 

Another interesting observation are the track eta and phi distrubtions shown Fig. 9 and 10 respectively. The Stv algorithm appears to have trouble reconstructing tracks at an eta ~ 0, which leads us to conclude that the alogrithms may have difficulty with tracks near or crossing the CM. Also, the phi distributions are extremely interesting, due to the region around -1 to 0. The infamous TPC sector 20 has troublesome anode voltages near the sector boudary of the outer sector which makes it near impossible to properly calibrate. One can see that the there is a large dip in the number of Sti tracks in this region. However, the other 3 alogrithms appear to have a better and uniform phi distribution. If the other alogrithms can bring sector 20 back into the fold, then this would greatly improve our acceptance. 

Continuing our examination to the track chi² distributions we see bit of an oddity. The slopes between the Sti and Stv algorithms are vastly different. The Sti tracks clearly have a steeper slope than the Stv tracks, which means that the jet algorithm in place could be cutting out more of the low pT tracks just due to the chi2 < 4 requirement. This applied that was base upon analysis using the Sti tracking and may not be appropriate for the Stv. 

Furthermore, the distributions of the tower energy (you can see other tower observables in the link above) was shown (Fig 13)not to greatly vary due to the different algorithm as expected. Also, the vertex position distributions (Fig 14, 15, 16)apear to be close to one another. However there are more jet vertices in the StiCA than any of the other algorithms and there are more jet vertices at high vertex z-postion |verZ > 40cm|.  

Further Endeavors:

 1) Apply the DiJet Code to determine how the tracking algorithm effects dijet ditributions.

 2) Re-run the Jet Analysis and Dijet Code over the MC data-sample produced.

 3) Then look into the chi2 cut effects. Propose removing the cut completely to see how it effects all track distributions and determine if a better cut should be used for the Stv tracking

 4) Suggestions Welcome.