AliRoot Core  ee782a0 (ee782a0)
MUON Simulation

The simulation encompasses the following tasks :

  • Generation of MC particles (the kinematics of the event ends up in the TreeK of Kinematics#.root)
  • Tracking particles through the detector using the Virtual Monte Carlo, producing AliMUONHit objects, that end up in the TreeH of MUON.Hits#.root file(s). This part is steered by AliMUON and its child AliMUONv1 classes.
  • Converting MC hits into AliMUONVDigit, called SDigits, that end up in the TreeS of the MUON.SDigits#.root file(s). A S(ummable)Digit is a pad with its associated charge, but no noise or electronics response function applied. Steered by AliMUONSDigitizerV2 class.
  • Converting SDigits into Digits, by applying electronics calibrations. Actually, we de-calibrate the digits at this stage, by adding a pedestal and dividing by a gain, more or less. Steered by AliMUONDigitizerV3 class. Digits end up in TreeD of MUON.Digits#.root file(s). In addition, for the trigger, we create AliMUONLocalTrigger, AliMUONRegionalTrigger and AliMUONGlobalTrigger objects at this stage, that ends up in TreeD as well.
  • Convert the Digits into RAW data, in a format that should be exactly the same as real data from the DAQ. Performed by AliMUONRawWriter.

From there on, the reconstruction can proceed, in the very same way for real or simulated data, as long as they are in RAW format.

How to run a MUON generation

You only need to run the simulation part of the test script AlirootRun_MUONtest.sh

Tracking parameters, cuts, energy loss and physics processes

Tracking parameters in MUON are automatically defined by GEANT MUON takes the default values of CUTs and physics processes defined by the Config files, except for the gas mixture medium of the tracking chambers. The CUT's and physics processes of the gas mixture medium is then defined in the galice.cuts file in the data directory. In particular ILOSS parameter MUST be equal unity (1) in order simulate a realistic energy loss distribution (mean value and fluctuations) in the active gas.

Tracking of particle in the magnetic field

GEANT has two integration methods for tracking charged particles in the magnetic field: HELIX et RKUTA. HELIX is faster and works well if the gradient of magnetic field is small. For MUON, HELIX is a not a good approximation and we must use RKUTA to get the optimal mass resolution of the spectrometer. The choice of HELIX or RKUTA is done in the config file when the magnetic field is defined:

  TGeoGlobalMagField::Instance()
    ->SetField(new AliMagF("Maps","Maps", INTEG, FACTOR_SOL, FACTOR_DIP, MAXB, AliMagF::k5kG));  

INTEG must be 1 for RKUTA and 2 for HELIX (the default value for aliroot is 2 (HELIX)). FACTOR_SOL, FACTOR_DIP allow you to set the multiplicative factor for solenoid and dipole, respectively; just putting FACTOR_SOL=0 or FACTOR_DIP=0 will set the magnetic field for solenoid or dipole to 0. Default values are 1.0, 1.0. MAXB is the maximum magnetic field which default value is 10.T

Tailing effect

The control to turn on/off the parametrized tailing effect:

AliMUON::SetTailEffect(Bool_t),

The parameter to tune increase/decrease the tailing effect is kept inside, AliMUONResponseV0::DisIntegrate(). This parameter is an integer number (excluding zero and four), the higher the value is the less is the tailing effect:

Int_t para = 5; 

Zero is excluded because it gives straight line transformation, and four is excluded because the AliRoot simulation chain spends VERY VERY long time in AliMUONResponseV0::DisIntegrate method, which reason was not yet understood. The parameter for 1, 2, 3, 5, 6, 7, 8, 9, 10 were checked with no slowing down problem, however parameters greater than 6 give almost no tailing effect since they basically correspond to higher order polynomial transform.

MUON cocktail generator

There is a MUON cocktail generator of the muon sources in the EVGEN directory. This class derives from AliGenCocktail. In the init of this class I have filled the cocktail with the muon sources: J/Psi, Upsilon, Open Charm, Open Beauty, Pion, Kaons. The code needs only the production cross section at 4pi (for the moment this values are in the code since I prefere them do not be modified), and the code calculates the rate of particles in the acceptance, making the scaling based on the number of collisions for the hard probes and on the number of participants for soft sources: Pions and Kaons.

In the Genereate of this class all entries in the cocktail are called and we define a "primordial trigger" with requires a minimum number of muons above a Pt cut in the required acceptance. In order to normalized to the real number of simulated events, there are 2 data members in the class fNsuceeded adn fNGenerate which tell us what is the biais source.

Enclose an example to use this generator:

AliGenMUONCocktail * gener = new AliGenMUONCocktail();
gener->SetPtRange(1.,100.);       // Transverse momentum range  
gener->SetPhiRange(0.,360.);    // Azimuthal angle range 
gener->SetYRange(-4.0,-2.4);
gener->SetMuonPtCut(1.);
gener->SetMuonThetaCut(171.,178.);
gener->SetMuonMultiplicity(2);
gener->SetImpactParameterRange(0.,5.); // 10% most centra PbPb collisions
gener->SetVertexSmear(kPerTrack);  
gener->SetOrigin(0,0,0);        // Vertex position
gener->SetSigma(0,0,0.0);       // Sigma in (X,Y,Z) (cm) on IP position
gener->Init();

How to simulate events with misaligned geometry in local CDB

If you want to use a misaligned geometry to simulate some events you can use a local CDB. For this need to follow the next steps:

  • Generate misaligned data in local CDB. You can use MUONGenerateGeometryData.C as described above in the corresponding section. Let's assume you used the default residual misalignment settings, then you have a local CDB in your working directory called ResMisAlignCDB containing misalignement data (ResMisAlignCDB/MUON/Align).
  • Tell AliSimulation you want to use your local CDB for MUON/Align/Data To do this you need to instantiate the AliCDBManager, set the default storage and set the specific storage for MUON/Align/Data, before instantiating AliSimulation (see for example the commented lines AlirootRun_MUONtest.sh).
aliroot -b  >& testSim.out << EOF
AliCDBManager* man = AliCDBManager::Instance();
man->SetDefaultStorage("local://$ALICE_ROOT/OCDB");
man->SetSpecificStorage("MUON/align/Data","local://ResMisAlignCDB");
AliSimulation MuonSim("$ALICE_ROOT/MUON/Config.C");
MuonSim.SetWriteRawData("MUON");
MuonSim.Run(10);
.q
EOF

How to Merge events

You can merge 2 types of simulated events. For example, you can simulate Hijing events, and then simulate muons merging both.

Merging is done at the sdigits level, so Kinematics files of the merged events will just correspond to the Config.C simulated file).

You must, first, do the Hijing simulation and store it in directory $HIJING_SIM. Note that for merging you won't need Kinematics files of the Hijing simulation...

Hijing simulation

aliroot -b << EOF
AliSimulation HijingSim("$HIJING_SIM/YourConfigForHIJING.C")
HijingSim.Run(5)
.q
EOF

You cand build YourConfigFroHIJING.C File from the ConfigPPR file in AliRoot/macros module.

Then you can do muon simulation and reconstruction merging both simulated events. In next example, we are merging 20 times each Hijing event in order to simulate 100 muons merged with 5 Hijing events.

aliroot -b << EOF
AliSimulation MuonSim("$ALICE_ROOT/MUON/Config.C")
MuonSim.MergeWith("$HIJING_SIM/galice.root",20) //parameters are the Hijing simulation file and the number of times we use each Hijing event
MuonSim.Run(100) // number of muon (Config.C) events
.q
EOF
aliroot -b << EOF
TPluginManager * pluginmanager = gROOT->GetPluginManager()
pluginmanager->AddHandler("AliReconstructor","MUON","AliMUONReconstructor","MUON","AliMUONReconstructor()")
AliReconstruction  MuonRec("galice.root")
MuonRec.SetRunTracking("")
MuonRec.SetRunVertexFinder(kFALSE)
MuonRec.SetRunLocalReconstruction("MUON")
MuonRec.SetFillESD("MUON")
MuonRec.Run()
.q
EOF

On track numbering

All generated particles, including primary and secondary particles are put on the stack. The secondary particles are kept in the stack only if they gave a hit in any of the ALICE detectors The number of all particles placed on the stack for a given event can be obtained with Int_t nPart = AliStack::GetNtrack(); Looping from 0 to nPart via AliStack::Particle(ipart) gives the particle listing as obtained from the particle generator (primaries) and Monte Carlo (secondaries).

The particle response in the detector, a hit, is registered in the hits tree and the hits are filled with each primary track. The total number of "tracks" (fills of the tree) can be obtained with ntracks = AliMUONMCDataInterface::NumberOfTracks(event) and is usually smaller than "nPart". Since particles can also deposit hits in other detectors than the MUON spectrometer, there will be many "tracks" (fills) in the hit-tree without a hit in MUON.

The correspondence between "track ID" in the hits-tree ("itr") and the particle ID for particles on the stack (i.e. generated particles) can be obtained via:

for (Int_t itr = 0; itr < ntracks; itr++) {
    AliMUONVHitStore* hitStore = mcDataInterface.HitStore(event,itr);
    //track "itr" of the hits-tree
    Int_t nhitstot = hitStore->GetSize();
    AliMUONHit* mHit; 
    TIter next(hitStore->CreateIterator());
    while ( ( mHit = static_cast<AliMUONHit*>(next()) ) )
    {   
       Int_t id = mHit->Track(); //gives particle ID on stack
       TParticle* particle = mcDataInterface.Stack(event)->Particle(id);
    }  
}

where mcDataInterface has been obtained by AliMUONMCDataInterface mcDataInterface("galice.root");

During the procedure to go from hits to digits, the hits are summed up such that more than one track can contribute to a given digit. As a consequence the method Int_t AliMUONDigit::Track(Int_t trackID) takes an argument, where "trackID" runs from 0 to AliMUONDigit::Ntracks() to provide the reference to all tracks that contributed to it. The returned track ID is the one referred to in the hit-tree. To know which is the generated particle that deposited a given digit one has to follow the sequence of the kind: (shown here using the simple, but not fast, DataInterface interfaces) :

AliMUONMCDataInterface mcdi("galice.root");
AliMUONDataInterface di("galice.root");
AliMUONVDigitStore* digitStore = di.DigitStore(event);
AliMUONVDigit* mDigit = ... get some digit from the digitStore
for (int tr = 0; tr < mDigit->Ntracks(); tr++)
{
   Int_t hitTrackID = mDigit->Track(tr);
   // get the hits corresponding to this trackID
   AliMUONHitStore* hitStore = mcdi.HitStore(event,hitTrackID);
   // loop over the hits
   TIter hNext(hitStore->CreateIterator());
   AliMUONHit* mHit;
   while ( ( mHit = static_cast<AliMUONHit*>(hNext()) ) )
   {
    Int_t numPart = mHit->Track(); //gives ID of particle on the stack
    Int_t idTrack = mHit->Particle(); //gives flavour code of the particle
   }
}

This chapter is defined in the READMEsim.txt file.