<<O>>  Difference Topic ElAlgoritmoParticleFlowEnILD (r1.3 - 23 Dec 2008 - Main.iglesias)

META TOPICPARENT WebHome
Changed:
<
<

4. *Algoritmo Particle Flow *

>
>

4. Algoritmo Particle Flow


4.b. El algoritmo Particle Flow en ILD

Changed:
<
<

4.b.i. Pandora PFA * 4. Algoritmo Particle Flow en la zona hacia delante

>
>

4.b.i. Pandora PFA


Pandora PFA is a C++ implementation of a PFA algorithm running in the MARLIN Framework. It was developed and optimised using events generated with the MOKKA program, which provides a GEANT4 simulation of the detector.

Changed:
<
<

Reconstruction_Framework.jpg

Pandora PFA performs both calorimeter clustering and particle flow in a single stage. using tracking information to help ECAL/HCAL clustering

>
>
Reconstruction_Framework.jpg

Changed:
<
<
It can be applicable to multiple detector concepts
>
>
Pandora PFA performs both calorimeter clustering and particle flow in a single stage. using tracking information to help ECAL/HCAL clustering. Other important advantage of this algorithm is that it have been designed and developped in order to can be applicable to multiple detector concepts.

The algorithm has seven main stages:

1- Tracking:

Changed:
<
<
Charged particle tracks are reconstructed in the tracking detectors. The track parameters are then extracted using the helical fit. For optimal performance it is important to identify :
  • neutral vertices, e.g. from KS +- decays and photon conversions: Neutral particle decays resulting in two charged particle tracks (V0s) are identified by
searching for tracks which do not originate from the interaction point and that are consistent with coming from a single point in space.
  • and kinks from bremstrahlung charged particle decays: Kinked tracks from charged particle decays to a single charged particle and a number of neutrals are also identified. When a kink is identified the “parent” track is effectively removed and is not used in the reconstruction of PFOs.
>
>
Charged particle tracks are reconstructed in the tracking detectors. The track parameters are then extracted using the helical fit. The projections of tracks onto the front face of the ECAL barrel/endcap are calculated using helical fits (without taking into account energy loss along the track) and are stored.

Changed:
<
<
The projections of tracks onto the front face of the ECAL barrel/endcap are calculated using helical fits (without taking into account energy loss along the track) and are stored.
>
>
For optimal performance it is important to identify :
  • neutral vertices, e.g. from KS → π+ π- decays and photon conversions: Neutral particle decays resulting in two charged particle tracks (V0s) are identified by searching for tracks which do not originate from the interaction point and that are consistent with coming from a single point in space.
  • and kinks from bremstrahlung charged particle decays: Kinked tracks from charged particle decays to a single charged particle and a number of neutrals are also identified. When a kink is identified the “parent” track is effectively removed and is not used in the reconstruction of PFOs. kinks.jpg

2- ECAL/HCAL Clustering: The clustering algorithm Start at inner layers and work outward. In this manner, hits are added to clusters or are used to seed new clusters. Throughout the clustering algorithm, clusters are assigned a direction (or directions) in which they are growing. Tracks can be used to “seed”clusters. The initial direction of a track-seeded cluster is obtained from the track direction. A Simple cone based algorithm is used

Added:
>
>

ECAL_HCAL_clustering.jpg

3- Topological Cluster Association:

Added:
>
>
Clusters associated using a number of topological rules.
  • Clear Associations: Join clusters which are clearly associated making use of high granularity + tracking capability: very few mistakes. Clear_associations.jpg
  • Less Clear Associations: Different strategies are followed, for example:
    • Proximity: Use E/p consistency to veto clear mistakes

Less_Clear_associations_PROXIMITY.jpg


4- Associate track to clusters

5- Iterative Reclustering¨: This step using different estrategies:

Added:
>
>
  • Cluster splitting: we reapply entire clustering algorithm to hits in “dubious” cluster. And iteratively we reduce cone angle until cluster splits to give acceptable energy match to track. Therefore, if it will be necessary, plug in alternative clustering algorithm

Changed:
<
<
Cluster splitting we reapply entire clustering algorithm to hits in “dubious” cluster. And iteratively we reduce cone angle until cluster splits to give acceptable energy match to track. Therefore, if it will be necessary, plug in alternative clustering algorithm
>
>
  • Cluster merging with splitting: We look for clusters to add to a track to get sensible energy association. If necessary iteratively we split up clusters to get good match.

Changed:
<
<
Cluster merging with splitting We look for clusters to add to a track to get sensible energy association. If necessary iteratively we split up clusters to get good match.
>
>
  • Track association ambigueties: In dense environment we may have multiple tracks matched to same cluster. We Apply above techniques to get ok energy match.

Changed:
<
<
Track association ambigueties In dense environment we may have multiple tracks matched to same cluster. We Apply above techniques to get ok energy match. Nuclear option If none of above works –we kill track and rely on clusters alone
>
>
  • Nuclear option: If none of above works –we kill track and rely on clusters alone

IterativeReclusteringStrategies.jpg

6- Photon identification/Recovery

Added:
>
>
Previously used simple cut-based photon ID. This is still used in the early (CPU intensive) stages of PandoraPFA?. In the final stages use improved photon ID based on the expected EM longitudinal profile for cluster energy E0
Photon_ID.jpg

Convert cluster into energy depositions per radiation length (use cluster to determine the layer spacing, i.e. geometry indep.)

  • Shower Profile fixed by cluster energy
  • But fit for best shower start, s
  • Normalise areas to unity and calc.:
Normalize_area_energy_photon.jpg
  • Gives a measure of fractional disagreement in obs/exp profiles
  • Use fand sto ID photons
  • Small improvement in PFA perf.

7- Fragment Removal Clustering “fragments”from charged tracks, looking for “evidence” that a cluster is associated with another, as: *Distance of closest approach *Layers in close contact *Distance to track extrapolated *Fraction of energy in cone
Added:
>
>
Fragment_Removal.jpg

-- Main.iglesias - 22 Dec 2008

Line: 63 to 78

-- Main.iglesias - 22 Dec 2008
Added:
>
>


META FILEATTACHMENT Reconstruction_Framework.jpg attr="" comment="" date="1229949205" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Reconstruction_Framework.jpg" size="59274" user="iglesias" version="1.1"
Added:
>
>
META FILEATTACHMENT ECAL_HCAL_clustering.jpg attr="" comment="" date="1230027321" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\ECAL_HCAL_clustering.jpg" size="56043" user="iglesias" version="1.1"
META FILEATTACHMENT IterativeReclusteringStrategies?.jpg attr="" comment="" date="1230028535" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Iterative Reclustering Strategies.jpg" size="37394" user="iglesias" version="1.1"
META FILEATTACHMENT Fragment_Removal.jpg attr="" comment="" date="1230028548" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Fragment_Removal.jpg" size="48941" user="iglesias" version="1.1"
META FILEATTACHMENT Clear_associations.jpg attr="" comment="" date="1230028975" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Clear_associations.jpg" size="21441" user="iglesias" version="1.1"
META FILEATTACHMENT Less_Clear_associations_PROXIMITY.jpg attr="" comment="" date="1230029555" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Less_Clear_associations_PROXIMITY.jpg" size="12367" user="iglesias" version="1.1"
META FILEATTACHMENT kinks.jpg attr="" comment="" date="1230030431" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\kinks.jpg" size="13596" user="iglesias" version="1.1"
META FILEATTACHMENT Photon_ID.jpg attr="" comment="" date="1230031985" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Photon_ID.jpg" size="10661" user="iglesias" version="1.1"
META FILEATTACHMENT Normalize_area_energy_photon.jpg attr="" comment="" date="1230032057" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Normalize_area_energy_photon.jpg" size="2680" user="iglesias" version="1.1"
 <<O>>  Difference Topic ElAlgoritmoParticleFlowEnILD (r1.2 - 22 Dec 2008 - Main.iglesias)

META TOPICPARENT WebHome
Changed:
<
<
  • 4. Algoritmo Particle Flow en la zona hacia delante
    • b. El algoritmo Particle Flow en ILD
      • Pandora PFA
>
>

4. *Algoritmo Particle Flow *

4.b. El algoritmo Particle Flow en ILD

4.b.i. Pandora PFA * 4. Algoritmo Particle Flow en la zona hacia delante


Pandora PFA is a C++ implementation of a PFA algorithm running in the MARLIN Framework. It was developed and optimised using events generated with the MOKKA program, which provides a GEANT4 simulation of the detector.

Added:
>
>

Reconstruction_Framework.jpg

Pandora PFA performs both calorimeter clustering and particle flow in a single stage. using tracking information to help ECAL/HCAL clustering
Line: 56 to 62

-- Main.iglesias - 22 Dec 2008

Added:
>
>

META FILEATTACHMENT Reconstruction_Framework.jpg attr="" comment="" date="1229949205" path="C:\Users\iglesias\Documents\My_Work_USC\My_Work\ILC_my_work\Presentaciones\Plot_Particle_Flow_SiLC_Meeting\Reconstruction_Framework.jpg" size="59274" user="iglesias" version="1.1"
 <<O>>  Difference Topic ElAlgoritmoParticleFlowEnILD (r1.1 - 22 Dec 2008 - Main.iglesias)
Line: 1 to 1
Added:
>
>
META TOPICPARENT WebHome
  • 4. Algoritmo Particle Flow en la zona hacia delante
    • b. El algoritmo Particle Flow en ILD
      • Pandora PFA

Pandora PFA is a C++ implementation of a PFA algorithm running in the MARLIN Framework. It was developed and optimised using events generated with the MOKKA program, which provides a GEANT4 simulation of the detector.

Pandora PFA performs both calorimeter clustering and particle flow in a single stage. using tracking information to help ECAL/HCAL clustering

It can be applicable to multiple detector concepts

The algorithm has seven main stages:

1- Tracking: Charged particle tracks are reconstructed in the tracking detectors. The track parameters are then extracted using the helical fit. For optimal performance it is important to identify :

  • neutral vertices, e.g. from KS +- decays and photon conversions: Neutral particle decays resulting in two charged particle tracks (V0s) are identified by
searching for tracks which do not originate from the interaction point and that are consistent with coming from a single point in space.
  • and kinks from bremstrahlung charged particle decays: Kinked tracks from charged particle decays to a single charged particle and a number of neutrals are also identified. When a kink is identified the “parent” track is effectively removed and is not used in the reconstruction of PFOs.

The projections of tracks onto the front face of the ECAL barrel/endcap are calculated using helical fits (without taking into account energy loss along the track) and are stored.

2- ECAL/HCAL Clustering: The clustering algorithm Start at inner layers and work outward. In this manner, hits are added to clusters or are used to seed new clusters. Throughout the clustering algorithm, clusters are assigned a direction (or directions) in which they are growing. Tracks can be used to “seed”clusters. The initial direction of a track-seeded cluster is obtained from the track direction. A Simple cone based algorithm is used

3- Topological Cluster Association:

4- Associate track to clusters

5- Iterative Reclustering¨: This step using different estrategies:

Cluster splitting we reapply entire clustering algorithm to hits in “dubious” cluster. And iteratively we reduce cone angle until cluster splits to give acceptable energy match to track. Therefore, if it will be necessary, plug in alternative clustering algorithm

Cluster merging with splitting We look for clusters to add to a track to get sensible energy association. If necessary iteratively we split up clusters to get good match.

Track association ambigueties In dense environment we may have multiple tracks matched to same cluster. We Apply above techniques to get ok energy match. Nuclear option If none of above works –we kill track and rely on clusters alone

6- Photon identification/Recovery

7- Fragment Removal Clustering “fragments”from charged tracks, looking for “evidence” that a cluster is associated with another, as: *Distance of closest approach *Layers in close contact *Distance to track extrapolated *Fraction of energy in cone

-- Main.iglesias - 22 Dec 2008

View topic | Diffs | r1.3 | > | r1.2 | > | r1.1 | More
Revision r1.1 - 22 Dec 2008 - 09:54 - Main.iglesias
Revision r1.3 - 23 Dec 2008 - 10:58 - Main.iglesias