Tower matching

Reconstruction of Pi0 (or Eta) at low pT from towers only

EEMC response

Pythia event generator, pp events at sqrt(s)=200 GeV

full GSTAR simulation of the EEMC with 2003 geometry was performed

GEANT energy deposit in the tower scintilator is multiplied the sampling fraction

the SF of 0.04 was assumed (note, there is 0.05 in the EEMC CDR) , as 'discovered' from GSTAR simulations half year ago. Look at the second table

Cluster finder is based on the MineSweeper game

seed search:
- each tower has an energy from GSTAR and a tag set to zero
- a tower with the largest energy and tag=0 and energy> seedThres is selected as the cluster seed
- tags in 3x3 patch around the seed are incremented by one
- seed search continue

cluster energy is the sum of 3x3 patch towerEnergy with weights =1/tag

cluster position ( eta and phi) is weighted mean with weights=towerEnergy/tag

cluster shape is characterized by the ratio: seedEnergy/cluserEnergy
- clusters with shape<shapeLimit are discarded

invariant mass

4-momentum is formed from each cluster, assuming E=P and the vertex at 0,0,0

for K clusters in an event there is K*(K-1)/2 combinations

mixing events

to estimate background from accidental cluster combinations the events are processed second time

from every event with at least two clusters one is chosen at random and preserved

for every pair of clusters one chosen at random is replaced by the cluster from previous event

An example of the invariant mass spectrum for PYTHIA+GSTAR high pT pp events (details below) Black line - regular processing of events, green line -mixed events.Cuts: shapeLimit=0.4 and seedThreshold=0.6 GeV.

minb pp events

parameters
There are two parameters in this algo:

seedThres - should be not much lower then MIP response of 0.5 GeV

shapeLimit - should be above 0.5 to keep seed dominant

To optimize the pi0 yield a grid search has been performed. Table below shows the difference between integral of the invM for pi0 region [0.11,0.16] for regular reco and mixed event analysis.

Legend:

Nr - number of real events

Nm - number of mixed events

Npi0 - number of pi0 = Nr-Nm, Err= sqrt(Nr+Nm)

s2b - signal 2 background, s2b=(Nr-Nm)/Nm; Err= s2b *sqrt(1/Nr+1/Nm)

yield- number of pi0 per 10K input events

.ps figure is under each link, black line - regular events, green line -mixed events. Upper panel - raw spectra, middle - the difference, lower - the ratio.

High pT pp events (PYTHIA ckin3=3,4,5 GeV, added 3x10K events), Vertex sigZ=7.5 cm, 30K event total

seedThreshold (GeV)	0.2	0.4	0.6	0.8
	shapeLimit
0.2	Nr=20252 Nm=18500 Npi0=1752 + / - 197 s2b=0.09 + / -0.001 yield=584 + / - 66	Nr=16978 Nm=15430 Npi0=1548 + / - 180 s2b=0.09 + / -0.001 yield=516 + / - 60	Nr=11890 Nm=10791 Npi0=1099 + / - 151 s2b=0.09 + / -0.001 yield=366 + / - 50	Nr=6084 Nm=5288 Npi0=796 + / - 107 s2b=0.13 + / -0.002 yield=265 + / - 36
0.4	Nr=5217 Nm=3875 Npi0=1342 + / - 95 s2b=0.26 + / - 0.005 yield=447 + / - 32	Nr=4741 Nm=3649 Npi0=1092 + / - 92 s2b=0.23 + / - 0.005 yield=364 + / - 31	Nr=3542 Nm=2816 Npi0=726 + / - 80 s2b=0.20 + / - 0.005 yield=242 + / - 27	Nr=1900 Nm=1446 Npi0=454 + / - 58 s2b=0.24 + / - 0.008 yield=151 + / - 19
0.6	Nr=1190 Nm=752 Npi0=438 + / - 44 s2b=0.37 + / - 0.017 yield=146 + / - 15	Nr=1103 Nm=707 Npi0=396 + / - 43 s2b=0.36 + / - 0.017 yield=132 + / - 14	Nr=855 Nm=507 Npi0=348 + / - 37 s2b=0.41 + / - 0.023 yield=116 + / - 12	Nr=407 Nm=226 Npi0=181 + / - 25 s2b=0.44 + / - 0.037 yield=60 + / - 8
0.8	Nr=408 Nm=192 Npi0=216 + / - 24 s2b=0.53 + / - 0.046 yield=72 + / - 8	Nr=384 Nm=202 Npi0=182 + / - 24 s2b=0.47 + / - 0.041 yield=61 + / - 8	Nr=322 Nm=135 Npi0=187 + / - 21 s2b=0.58 + / - 0.060 yield=62 + / - 7	Nr=148 Nm=76 Npi0=72 + / - 15 s2b=0.49 + / - 0.069 yield=24 + / - 5

Conclusion: For the HT trigger data one can reach signal/background of 1:1 if the threshold on the cluster shape and on cluster seed energy is set high enough. Well, based on M-C only.
To get some feeling about topology of the reconstructed pi0 for shapeLimit=0.4 and seedThreshold=0.6 GeV look at the several plots below: energy, pT, pseudorapidity, phi. Mostly pi0 with energy of 2GeV are reconstructed - for higher energy the opening angle is too small, for lower energy of pi0 the gamma energy is too close to MIP.

The plot of opening angle for any pair of clusters (without pi0 cut) is here. Note, for mixed events it goes to zero, since the same tower may fire in two events , this leads to lower values of invariant mass for mixed events - thanks Steve.

Minbias pp events (PYTHIA) Vertex: Gaussian with sigma of 60 cm truncated at +/- 60 cm , 50K event total

seedThreshold (GeV)	0.2	0.4	0.6	0.8
	shapeLimit
0.2	Nr=10922 Nm=10531 Npi0=391 + / - 146 s2b=0.04 + / - 0.000 yield=78 + / - 29	Nr=9533 Nm=8830 Npi0=703 + / - 136 s2b=0.07 + / - 0.001 yield=141 + / - 27	Nr=6850 Nm=6440 Npi0=410 + / - 115 s2b=0.06 + / - 0.001 yield=82 + / - 23	Nr=3582 Nm=3295 Npi0=287 + / - 83 s2b=0.08 + / - 0.002 yield=57 + / - 17
0.4	Nr=2472 Nm=1979 Npi0=493 + / - 67 s2b=0.20 + / - 0.006 yield=99 + / - 13	Nr=2319 Nm=1861 Npi0=458 + / - 65 s2b=0.20 + / - 0.006 yield=92 + / - 13	Nr=1858 Nm=1435 Npi0=423 + / - 57 s2b=0.23 + / - 0.008 yield=85 + / - 11	Nr=1018 Nm=805 Npi0=213 + / - 43 s2b=0.21 + / - 0.010 yield=43 + / - 9
0.6	Nr=484 Nm=315 Npi0=169 + / - 28 s2b=0.35 + / - 0.025 yield=34 + / - 6	Nr=463 Nm=283 Npi0=180 + / - 27 s2b=0.39 + / - 0.029 yield=36 + / - 5	Nr=382 Nm=226 Npi0=156 + / - 25 s2b=0.41 + / - 0.034 yield=31 + / - 5	Nr=192 Nm=93 Npi0=99 + / - 17 s2b=0.52 + / - 0.065 yield=20 + / - 3
0.8	Nr=148 Nm=82 Npi0=66 + / - 15 s2b=0.45 + / - 0.061 yield=13 + / - 3	Nr=140 Nm=87 Npi0=53 + / - 15 s2b=0.38 + / - 0.052 yield=11 + / - 3	Nr=123 Nm=64 Npi0=59 + / - 14 s2b=0.48 + / - 0.074 yield=12 + / - 3	Nr=63 Nm=28 Npi0=35 + / - 10 s2b=0.56 + / - 0.126 yield=7 + / - 2

Conclusion: The pi0 yield per input events is lower for minB events in comparison to high pT ones (less pi0 with E=2 GeV is produced). But the invariant mass peak is sharper since the opening angle is larger.

Script used for generation of plots plMixXXX.C .