Using the Monte Carlo Information
Update on MonteCarloInfo Status
Feb 3, 2004
Motivation:
- Note: The following is really only interesting
to people working at the reconstruction or below level. Current Monte Carlo
information output to ntuples is more than sufficient for higher level
analyses. If you are a user of recon output it is safe to sleep for the next
few minutes...
- Current Monte Carlo information includes
(created in the G4Generator package):
- McParticle - Basic information about each
particle created in the event
- Particle properties
- Starting and ending position
- Starting and ending four momentum
- Except: gammas have zero four momentum at
conversion points
- etc.
- McPositionHit - Created for "sensitive"
volumes in Tracker (and ACD), one for each McParticle traversing the
sensitive volume
- Entry and exit positions
- Energy deposited
- etc.
- Really designed for the hit digitization
algorithms
- McIntegratingHit - Created for "sensitive"
volumes in Calorimeter, except "integrates" over shower particles
- The current information is more than
satisfactory for:
- The digitization algorithms in order to
provide "hits" to feed to the recon algorithms
- Obtaining global Monte Carlo truth for the
event
- Incident gamma direction
- Incident gamma energy
- Type of event
- However, current set up insufficient for
detailed studies:
- Detailed interactions in the LAT
- Example plots of Gammas which Compton
scatter before converting
- Detailed studies of the current pattern
recognition
- Is the pattern recognition finding all the
tracks?
- Is the pattern recognition finding the
right tracks?
- Is the pattern recognition using all the
hits?
- etc.
- Studying alternate track finding algorithms
- For example, use MC to train a neural net
algorithm
- Detailed studies of the track fitting
- Feed fitter tracks from MC and check output
track parameters
- Detailed studies of the vertexing
- etc.
- One could imagine the same sorts of studies for
the Calorimeter...
- To do more detailed studies requires augmenting
the currently available Monte Carlo information.
- Mostly this involves associating the existing
base Monte Carlo information to various levels of the reconstruction (digis,
hits/clusters, etc.)
- Make heavy use of Relational Tables for this
The Pieces:
- Current implementation creates three new
classes:
- McEventStructure
- Overall description of the event
- gamma event
- charged event
- summarize interactions encountered
- List of direct daughters which leave
McPositionHits (ie hits in tracker/acd)
- List of "associated" grandchildren
(daughters of daughters which also leave hits)
- etc.
- McSiLayerHit
- Ties together McPositionHits with Clusters.
A Cluster can be made of several McPositionHits
- McCandTrack
- Ties together McSiLayerHits with their
association tracks (McParticles)
- Lots and Lots of Relational tables
- McPositionHits to TkrDigis
- TkrDigis to TkrClusters
- McPositionHits to TkrClusters
- McSiLayerHits to MC Tracks
- MC Tracks to TkrCandidates
- etc., etc., etc.
- Then provide Tools to extract information from
the relational tables
- Relational tables are great for associating
information but definitely need tools for getting out what you want...
- And a tool (McAnalValsTool) in AnalysisNtuple
for putting interesting information into the output ntuple for analysis (this
assumed to be done only for people doing detailed studies of recon, it is NOT
intended to be part of the "standard" output)
Current Status:
- First implementation (just to get it all
working) was "hidden" into TkrRecon
- Convenient (for me) just to get started
- Not desirable for general release (or
expansion if that is desired)
- Also had "hidden" above output classes in a
TkrRecon private TDS
- For more general release:
- Move code to create new objects and tables,
creation and information tools, etc. to a new package: MonteCarloInfo
- Move new objects into Event/MonteCarlo
-