Cosmic Simulation Fall 2009 Production

Introduction

• • Calo volume filter will be used for the first time
  • Momentum slicing will be used for some of the samples
  • Additional track records will be in the ESD
  • A new 'perigee' truth record will be included
  • The muon system timing has grown closer to reality (particularly for the MDTs)
  • Additional truth will be saved in the calorimetry

For this simulation has been done with release 15.3.1.5,AtlasProduction. Reconstruction is being done with release 15.3.1.7,AtlasProduction to be consistent with MC09. The transform commands that will be run are similar to those described at CosmicSimHowTo.

As discussed in data preparation the first production will produce the following statistics (all with fields on):

1M Pixel Volume
1.5M ID Volume
1M Calo Volume
0.5M Calo Volume with 100<p<300 GeV
0.5M Calo Volume with 300<p<1000 GeV
0.5M Calo Volume with 1 TeV<p

• 5 Million events in all. Afterwards, we may on-demand increase the statistics considerably (a factor of 10 or more)
• The breakdown in terms of sim and reco jobs should be:
  • Pixel volume (all momenta): 1 M evts (simul tag s598 and task, digit tag d217 and task, reco tag r848 and task)

s: 25000 evts/output, d:  150k events/output; r: 150k events/output.
eff=0.01
MC8.108865.CosSimPixVolSolOnTorOn.py

• • ID volume (all momenta): 1.5 M evts (simul tag s599 and task, digit tag d217 and task, reco tag r848 and task)

s: 50k evts/output: d: 400k events/output r: 400k events/output
eff=0.0036
MC8.108867.CosSimIDVolSolOnTorOn.py

• • Calo volume (all momenta): 1 M evts (simul tag s600 and task, digit tag d217 and task, reco tag r848 and task)

s: 30000 evts/output, d: 60000 evts/output r: 60000 evts/output
MC8.108872.CosSimCaloVolSolOnTorOn.py
eff=1/30

• • Calovolume (100<p<300 GeV): 0.5 M evts (simul tag s601 and task, digit tag d217 and task, reco tag r848 and task)

s: 30000 evts/output, d: 60000 evts/output r: 60000 evts/output
MC8.108873.CosSimCaloVolSolOnTorOn_sl2.py
eff=1/30

• • Calo volume (300<p<1000 GeV): 0.5 M evts (simul tag s602 and task, digit tag d217 and task, reco tag r848 and task)

s: 30000 evts/output, d: 60000 evts/output r: 60000 evts/output
MC8.108874.CosSimCaloVolSolOnTorOn_sl3.py
eff=1/30

• • Calo volume (1TeV<p): 0.5 M evts (simul tag s603 and task, digit tag d217 and task, reco tag r848 and task)

s: 30000 evts/output, d: 60000 evts/output r: 60000 evts/output
MC8.108875.CosSimCaloVolSolOnTorOn_sl4.py
eff=1/30

Actual commands submitted

SIMULATION TASKS:
------------------------------
Pixel sim:
csc_cosmics_sim_trf.py outputHitsFile=???? maxEvents=25000 runNumber=???? firstEvent=???? physicsList=QGSP_BERT randomSeed=???? geometryVersion=ATLAS-GEO-08-00-01 CosmicFilterVolume=TRT_Barrel CosmicFilterVolume2=Pixel jobConfig= CosmicsWithExtraTruth.py DBRelease=/afs/cern.ch/atlas/www/GROUPS/DATABASE/pacman4/DBRelease/DBRelease-7.4.1.tar.gz
where you fill in the ???'s on the per job basis (except runnumber which you set to the same thing for all jobs in this task).
firstEvent should be n*25000 where n=the job number in this task.

number of jobs in this task should be: 4002

ID sim:
csc_cosmics_sim_trf.py outputHitsFile=???? maxEvents=50000 runNumber=???? firstEvent=???? physicsList=QGSP_BERT randomSeed=???? geometryVersion=ATLAS-GEO-08-00-01 CosmicFilterVolume=TRT_Barrel CosmicFilterVolume2=TRT_EC jobConfig=CosmicsWithExtraTruth.py DBRelease=/afs/cern.ch/atlas/www/GROUPS/DATABASE/pacman4/DBRelease/DBRelease-7.4.1.tar.gz
where you fill in the ???'s on the per job basis (except runnumber which you set to the same thing for all jobs in this task).
firstEvent should be n*50000 where n=the job number in this task.

number of jobs in this task should be: 8336

Calo sim:
csc_cosmics_sim_trf.py outputHitsFile=???? maxEvents=30000 runNumber=???? firstEvent=???? physicsList=QGSP_BERT randomSeed=???? geometryVersion=ATLAS-GEO-08-00-01 CosmicFilterVolume=Calo CosmicFilterVolume2=NONE jobConfig= CosmicsWithExtraTruth.py CosmicPtSlice=slice2 DBRelease=/afs/cern.ch/atlas/www/GROUPS/DATABASE/pacman4/DBRelease/DBRelease-7.4.1.tar.gz
where you fill in the ???'s on the per job basis (except runnumber which you set to the same thing for all jobs in this task).
firstEvent should be n*30000 where n=the job number in this task.

The CosmicPtSlice parameter can take the values 'NONE' (all momenta), 'slice1' (p<100 GeV), 'slice2' (100<p<300 GeV), 'slice3' (300<p<1000 GeV), or 'slice4' (1 TeV<p).  No production was requested with 'slice1'.  For slice4, there should be only 15000 events per job.

number of jobs in this task should be: 1000 with all momenta or slice4 and 500 with slice2 or slice3.


DIGITIZATION TASKS
------------------------------
csc_digi_trf.py inputHitsFile=???? outputRDOFile=???? maxEvents=-1 skipEvents=0 geometryVersion=ATLAS-GEO-08-00-01 jobConfig=SimuJobTransforms/CosmicsDigitConfig.py digiSeedOffset1=0 digiSeedOffset2=0 minbiasHitsFile=NONE cavernHitsFile=NONE DBRelease=/afs/cern.ch/atlas/www/GROUPS/DATABASE/pacman4/DBRelease/DBRelease-7.4.1.tar.gz digiRndmSvc=AtRanluxGenSvc samplingFractionDbTag=QGSP_BERT conditionsTag=OFLCOND-SIM-01-00-00

where you fill in the ????.  The input here for each job should be one file output from the corresponding simulation task.

Digitization tasks have been defined with tag d217.

RECONSTRUCTION TASKS
-------------------------------------
Reco_trf.py inputRDOFile=???? preInclude=RecExCommission/CosmicSimulationRecoSetup-HLT.py,RecExCommission/RecExCommission.py beamType=cosmics HIST=myMergedMonitoring09.root outputTAGComm=myTAGCOMM.root outputCBNT=myCommission.ntuple.root outputESDFile=esd09.pool.root outputAODFile=aod09.pool.root geometryVersion=ATLAS-GEO-08-00-00 --ignoreunknown

To fix a bug in monitoring, we added: preExec='DQMonFlags.doHLTMon.set_Value_and_Lock(False)'

To fix a bug in TRT conditions, we added: postExec='if (jobproperties.Rec.doESD): job.InDetCosmicsEventPhase.UseNewEP=False'

where the output files:
myTAGCOMM.root
myMergedMonitoring.root
aod09.pool.root 
esd09.pool.root
should of course have their filename depending on the input data, and should all be kept and registered with DDM (dq2).

We should take 1 input RDO file for 1 job.

Reconstruction tasks have been defined with tag r834. First try had a little problem. Reconstruction has now been re-submitted with tag 837 and tasks have been defined.

Second round of reconstruction is being hit by an additional bug. The tasks have been resubmitted with tag r839. The r837 datasets are fine to use, if they have enough statistics for your study.

Third round of reconstruction had a problem with TRT conditions. The tasks have been resubmitted with reco tag r848. A description of the differences is given in Maria Llacer's talk. The first problem reported there is the relevant one; the second was already repaired in 15.3.1.7.

This newest round is affected by several bugs, but none appear severe enough to force us to stop the reconstruction:

• Return failure in CaloCell2ClusterMap
• Segfault in TrigInDetTrackTruthMaker::PIXspTruth
• Unchecked status code in TrigIDSCAN::hltExecute

Validation of new simulation options

In this production we have included track records as an additional simulation output dataset. These are records of muons as they enter the ATLAS cavern and are only saved for those events passing the filter that has been applied. Two validation sets have been submitted:

• Track record creation and reading validation with sim tag s615, digit tag d217 and task, reco tag r848 and task. If all goes will the final reconstructed dataset can be transparently added to the corresponding 'standard' simulation datasets. In principle, this job only reads in the track records and runs them through the same simulation. Because of random number seed differences, the events will not be identical (but the input flux will look a bit strange).
• Track smearing validation with sim tag s614, digit tag d217 and task, reco tag r848 and task. This does not include the additional calorimeter truth information that most of the sets include, but otherwise should be identical to the 'standard' datasets. This smears the starting position of the input track records a small bit, providing a statistically independent sample without the requirement of starting from scratch (and without the low filter efficiency). The input flux is only smeared a small amount (10 mm and 0.001 radians) in order to properly maintain the same input flux shape (e.g. not smear out the shaft shadows).

Available reconstructed data

Data sets are now all finished.

• mc09_valid.108865.CosSimPixVolSolOnTorOn.recon.ESD.s598_d217_r848/ (and AOD)
• mc09_valid.108867.CosSimIDVolSolOnTorOn.recon.ESD.s599_d217_r848/ (and AOD)
• mc09_valid.108872.CosSimCaloVolSolOnTorOn.recon.ESD.s600_d217_r848/ (and AOD)
• mc09_valid.108873.CosSimCaloVolSolOnTorOn_sl2.recon.ESD.s601_d217_r848/ (and AOD)
• mc09_valid.108874.CosSimCaloVolSolOnTorOn_sl3.recon.ESD.s602_d217_r848/ (and AOD)
• mc09_valid.108875.CosSimCaloVolSolOnTorOn_sl4.recon.ESD.s603_d217_r848/ (and AOD)

10M additional events were produced with the ID filter and with corrected muon digitization configurations. The data is now available:

• mc09_valid.108867.CosSimIDVolSolOnTorOn.recon.ESD.s664_d283_r1076/

Normalization of Calo Slice samples

Several factors change between the calo slice samples above. In order to normalize them properly (even relative to each other) one must take into account:

• Number of events requested for production (note: this is not the output number of events)
• Efficiency on the Grid (i.e. number of jobs that finished successfully)
• Geometric factor (area above the cavern over which cosmics were generated and solid angle of the cavern)
• Energy for generation (flux scaling)

The fortran code used to generate the cosmics resides here. As a note ahead of time, I am neither sure these calculations are correct, nor certain that I didn't miss any factors. You are welcome to double check the answers yourself.

For each of these factors, I find the numbers in the following table. Geometry factors were taken from a log file from a typical run (probably sufficient for an estimate with reasonable errors). Energy factors were calculated by evaluating the code by hand.

-- ZacharyMarshall - 29-Nov-2009