Data Formats in GeneratorInterface

Event content definition

RECO Data Formats

GeneratorInterface collections (in RECOSIM and AODSIM)
generator	LHEEventProduct	General characteristics of a generated event (only present if the event starts from LHE events)
generator	GenEventInfoProduct	General characteristics of a generated event.
generator	GenRunInfoProduct	Run-specific parameters that define event generation, such as cross-sections, etc.

GeneratorInterface collections (in RECOSIM only)
generator	edm::HepMCProduct	A tree of final-state particles that form a generated event.
generator	LHEEventProduct	General characteristics of a generated event (only present if the event starts from LHE events)
generator	GenEventInfoProduct	General characteristics of a generated event.
generator	GenRunInfoProduct	Run-specific parameters that define event generation, such as cross-sections, etc.

How to use the table

In the header of your analyzer you should have the header(s) of the data format(s) you will access. Generator data formats belong to the in the SimDataFormats/GeneratorProducts; the headers are located in the /interface subdirectory. Thus, your EDAnalyzer should contain all or one of the following:

#include "SimDataFormats/GeneratorProducts/interface/HepMCProduct.h"
#include "SimDataFormats/GeneratorProducts/interface/GenEventInfoProduct.h"
#include "SimDataFormats/GeneratorProducts/interface/GenRunInfoProduct.h"

Using the HepMCProduct directly (only works with GEN/GENSIM)

Note that edm::HepMCProduct is a wrapper on top of HepMC::GenEvent class of the standard LCG package HepMC.
For details about features and access methods of the HepMC::GenEvent class please refer to the HepMC documentation for details.

Here we would like to show how to access the generated event particle tree and to use 2 of the GenEvent::HepMC access methods.
In the analyze(const Event& e, const EventSetup& ) method of your EDAnalyzer you can do something like the following:

edm::Handle<edm:: HepMCProduct > genEvtHandle;
e.getByLabel( "generator", genEvtHandle) ;
const HepMC::GenEvent* Evt = genEvtHandle->GetEvent() ;
//
// this is an example loop over the hierarchy of vertices
//
for ( HepMC::GenEvent::vertex_const_iterator
          itVtx=Evt->vertices_begin(); itVtx!=Evt->vertices_end(); ++itVtx )
{
      //
      // this is an example loop over particles coming out of each vertex in the loop
      //
      for ( HepMC::GenVertex::particles_out_const_iterator
              itPartOut=(*itVtx)->particles_out_const_begin();
              itPartOut!=(*itVtx)->particles_out_const_end(); ++itPartOut )
      {
         // and more of your code...
      }
}

Out of the event size consideration, in the AOD event content the edm::HepMCProduct format of the is generated event is replaced by the "lighter" record called reco::GenParticleCollection. Formally, it is outside the GeneratorInterface domain, and belongs to the CMS.PhysicsTools of CMSSW. For more details, please visit WorkBookGenParticleCandidate and related materials.

Using other products (AODSIM)

If you wish to analyze general characteristics of an event rather than specific generated particles, in the analyze(const Event& e, const EventSetup& ) of your EDAnalyzer do the following:

edm::Handle<GenEventInfoProduct> genEvtInfo;
e.getByLabel( "generator", genEvtInfo );
double qScale = genEvtInfo->qScale();  // in case of Pythia6, this will be pypars/pari(23)
const std::vector<double>& binningValues = genEvtInfo->binningValues(); // in case of Pythia6, this will be pypars/pari(17)

Particularly important is the use of weights, which must always be activated otherwise events with generators like Madgraph5_aMCatNLO will give wrong results.

 
std::vector<double>& evtWeights = genEvtInfo->weights();
double theWeight = genEvtInfo->weight();
// and some more of your code again...

Please note that the weights() return the container of the associated event weights, where there may be 1 or more elements. For example, in the case of running Pythia6 generators there'll be 2 elements; please see SWGuidePythia6Interface#Example_8_CSA_mode_with_Event_Re for more details.
The weight() method returns the value which is a result of multiplication of all elements in the weights container and is normally the one to be used.

If you wish to access run-specific information about event generation, you can do so via GenRunInfoProduct. Please be advised that this product

• belongs to edm::Run rather than edm::Event
• is not finalized until the end of run

edm::Handle<GenRunInfoProduct> genRunInfo;
r.getByLabel( "generator", genRunInfo );

Retrieving information on LHE event (AODSIM)

Only in case the generation started from LHE events the GenEventInfo will contain information on matching like the number of original partons before hadronization:

 
int nAllPartons = genEvtInfo->nMEPartons();
int nPartonsEnteringMatching = genEvtInfo->nMEPartonsFiltered();
// and some more of your code again...

Also, if the LHE was produced by Madgraph5_aMCatNLO or POWHEG in official production weights from scale variations, PDFs etc. are stored in the relative product. Notice that to be used they need to be renormalized to the central event weight at LHE level which may be different from genEvtInfo->weight() .

edm::Handle<LHEEventProduct> EvtHandle ;
e.getByLabel( theSrc , EvtHandle ) ;

int whichWeight = XXX;
theWeight *= EvtHandle->weights()[whichWeight].wgt/EvtHandle->originalXWGTUP(); 

Some commonly used weights for LO samples:

• XXX = 0,...,8 : QCD scales variations (all combinations of mu_R x2 - x0.5 and muF x2 - x0.5)
• XXX = 9,...,109: Members of NNPDF3.0LO (default PDF, central + 100 variations, alpha_s = 0.130). These numbers are instead 110,...,210 if the sample uses the 4-flavor version of PDFs.
• XXX = 211: NNPDF3.0LO with alpha_s varied to 0.118
• XXX = 313: NNPDF3.0LO with alpha_s varied to 0.119
• XXX = 314: CTEQ6L1 (alternative PDF)
• XXX = 315,...,365: Members of MMHT14lo68cl (alternative PDF, central + 50 variations)
• XXX = 369,...,389: Members of HERAPDF15LO (alternative PDF, central + 20 variations)

Some commonly used weights for NLO samples:

Review status

Responsible:
Last reviewed by: