Gabriele Cosmo
168 lines
8.2 KiB
Plaintext
168 lines
8.2 KiB
Plaintext
$Id: README,v 1.1 2006/11/22 14:51:26 gcosmo Exp $
|
|
-------------------------------------------------------------------
|
|
|
|
=========================================================
|
|
Geant4 - an Object-Oriented Toolkit for Simulation in HEP
|
|
=========================================================
|
|
|
|
MCTRUTH using HepMC
|
|
-------------------
|
|
|
|
This example demonstrates a mechanism for Monte Carlo truth handling
|
|
using HepMC as the event record. The user does not interact directly
|
|
with the HepMC classes but with the MCTruthManager class which takes
|
|
care with storing all the necessary information about particles,
|
|
vertices and relations between them. A specialized tracking action is
|
|
used to test whether given particle is to be stored or not. The
|
|
decision criteria for storing particle are configurable via the
|
|
MCTruthConfig class.
|
|
|
|
HOW TO BUILD THE EXAMPLE ?
|
|
|
|
- install HepMC event record (version 1.27)
|
|
|
|
- set HEPMC_DIR variable to point to the directory where HepMC is installed;
|
|
add the path "$HEPMC_DIR/lib" to your LD_LIBRARY_PATH variable.
|
|
|
|
- compile and link to generate the executable:
|
|
% gmake
|
|
|
|
- execute the application:
|
|
% mctruthex
|
|
|
|
DESCRIPTION OF THE MCTRUTH HANDLING MECHANISM
|
|
|
|
The main element of the MC truth handling machinery is the
|
|
MCTruthManager class. This class is responsible for all the
|
|
interaction with the HepMC event and does not depend on Geant4. It is
|
|
a singleton, therefore it is guaranteed to be instanciated only once
|
|
and the static 'GetInstance' method allows to access it from anywhere
|
|
in the code. It contains methods like 'NewEvent' to start a new event,
|
|
'AddParticle' to add particle to the current event, as well as
|
|
'PrintEvent' for the purpose of the debugging. The core of the
|
|
algorithm which deals with building up the MC truth event tree within
|
|
the HepMC event is implemented in AddParticle method.
|
|
|
|
The AddParticle method is called with the following arguments:
|
|
four-momentum, production position and 'end' position of the particle,
|
|
PDG code of the particle, as well as the particle ID (unique identifier,
|
|
as we will see later, corresponding to Geant4 TrackID) and the ID of
|
|
the mother. Finally, there is a boolean flag specifying whether the
|
|
direct mother of the given particle has been stored, or not.
|
|
|
|
The first step, which always takes place, is to instanciate a new
|
|
HepMC::GenParticle with the barcode corresponding to particle ID, as
|
|
well as to instanciate a new HepMC::GenVertex which will represent the
|
|
'end' vertex of the particle. The barcode of the 'end vertex' is equal
|
|
to minus the barcode of the particle.
|
|
|
|
We can now distinguish several cases:
|
|
|
|
1) the particle is a primary in the Geant4 language, i.e. its
|
|
mother ID is 0
|
|
|
|
This is the simplest case, we just instanciate a new 'primary'
|
|
(without any incoming particles) GenVertex, we add to it the
|
|
particle and we put it all in the event. Additionally we store the
|
|
ID of the particle in a special vector, where all the IDs of
|
|
primary particles will be stored, allowing quick access to each of
|
|
the main 'branches' of the event. We return from the method.
|
|
|
|
2) the particle is not a primary
|
|
|
|
We use the 'event->barcode_to_particle(motherID)' method to get the
|
|
pointer to its mother.
|
|
|
|
We check if the 'end vertex' of the mother corresponds to the
|
|
'production vertex' of the particle in question.
|
|
|
|
2.1) If the two vertices do match, we attach the new particle to
|
|
the 'end vertex' of the mother. We return from the method.
|
|
|
|
2.2) If the two vertices do not match, i.e. the new particle is not
|
|
a product of the 'end vertex' of the mother particle, we can
|
|
have two cases:
|
|
|
|
2.2.1) The boolean flag says that the direct mother of the
|
|
particle has _not_ been stored. This means that the
|
|
particle has been 'adopted' by one of its ancestors, or
|
|
in other words, the mother ID of the particle does not
|
|
correspond to its direct mother (so clearly the
|
|
vertices cannot match). This for instance could happen
|
|
if we decided not to store gamma coming from pi0 decay
|
|
but did decide to store e+/- coming from the gamma
|
|
conversion (so the gamma between pi0 and e+/- was
|
|
missing). In such a case we instanciate (or use one of
|
|
the existing ones, if vertices match) a 'dummy'
|
|
particle (with pdg = -999999) which then acts as the
|
|
link between the 'adopted' particle and the
|
|
(non-direct) mother. In such a way, the navigability up
|
|
in the event is still possible, but in the same time,
|
|
we can clearly see that the link is not a direct
|
|
one. We return from the method.
|
|
|
|
2.2.2) The boolean flag says that direct mother of the
|
|
particle _has_ been stored. Taking into account that
|
|
the vertices do not match, it can mean only one
|
|
thing. The new particle has been produced 'on the
|
|
flight', i.e. somewhere 'before' the 'end vertex' of
|
|
the mother. This can be the case, for instace, for
|
|
delta electrons, bremsstrahlung gammas, etc. In such a
|
|
situation, we 'split' the mother particle in two
|
|
particles and create a new vertex from which the
|
|
secondary will be going out. The complication, however,
|
|
arises when we have more than one generated 'on the
|
|
flight' particle attached to the same mother. In such a
|
|
case, for each secondary we need to locate the right
|
|
'segment' of the mother particle (i.e. we need to find
|
|
between which two vertices we need to add a new
|
|
one). To keep track of those segmentations we introduce
|
|
a map where each particle ID we map into the number of
|
|
existing segments (in the normal case one). Each new
|
|
'segment' gets barcode equal to the barcode of the
|
|
original particle + N*10000000, where N is the segment
|
|
number. In such a way, one can easily follow the
|
|
'segmentation' (if any) of each particle. We return
|
|
from the method.
|
|
|
|
This concludes the description of MCTruthManager. The MCTruthConfig
|
|
class is a collection of criteria (minimal energy, PDG, creator
|
|
process, etc) that we want to apply when deciding whether to store or
|
|
not given particle. These values are used by the
|
|
'MCTruthTrackingAction' which we describe below. This class can
|
|
certainly be extended with other members.
|
|
|
|
The actual Geant4-dependent part of the MCTruth handling machinery
|
|
consists of a few 'G4 user actions' as well as an implementation of
|
|
G4VUserTrackInformation. The later one is, for the moment, used only
|
|
to store one boolean flag indicating whether the direct mother of the
|
|
given track has been stored or not.
|
|
|
|
The first user action is MCTruthEventAction which is only reponsible
|
|
for calling MCTruthManager::GetInstance()->NewEvent() at the beginning
|
|
of each event. It can also be used for printing out events for the
|
|
purpose of debugging.
|
|
|
|
The actual 'decision making' concerning which particle to store is
|
|
done in MCTruthTrackingAction. At the end of each track the method
|
|
trackToBeStored(track) is called to check for various characteristics
|
|
of the particle. These, for instance can be energy, particle ID,
|
|
creator process, etc.
|
|
|
|
If the particle satisfies the conditions the
|
|
MCTruthManager::GetInstance()->AddParticle is called and all the
|
|
procedure described above is performed. The important element here is
|
|
that the Geant4 TrackID is used as the unique particle ID in
|
|
MCTruthManager and eventually as the barcode of the
|
|
HepMC::GenParticle.
|
|
|
|
If the particle does not qualify to be stored, there are two actions
|
|
performed. First the 'ParentID' of the _daughters_ is set to the
|
|
'ParentID' of the currenly processed particle. In other words, the
|
|
'ParentID' of the daughters is set to the ID of the last stored
|
|
particle. Second, the 'directParent' flag from MCTruthTrackInformation
|
|
of the __daughters__ is set to FALSE. In such a way, one is still able
|
|
to navigate up in the event (to get the ancestors of the particle),
|
|
but in the same time, the particle is flagged as 'not having direct
|
|
parent'.
|