The document provides an overview of the SPHENIX Clusterizer algorithm used at CMS. It describes the island algorithm which clusters calorimeter towers starting from the highest energy seed tower and moving in φ and η directions until an energy rise or hole is found. Simple checks are performed using single particle events to evaluate the number and energy of clusters compared to the generated particle properties. Further work is planned to characterize shower shapes and evaluate performance on more complex events.
2. Brandon McKinzie 26 July 2016
The Island Algorithm at CMS
Procedure:
1. Store “seed” towers. Defined by ET
> ET
thresh
2. Remove seeds adjacent to higher energy ones.
3. Starting from highest energy seed:
a. Move both directions in φ until rise in energy or hole.
b. Move one step in η. Repeat φ search.
i. Continue along η until energy rise or hole.
2
3. Brandon McKinzie 26 July 2016
Simple Checks: Number of Clusters Per Event
● Single-particle [(η, φ) = (0, 0)] events.
● Varied (but known) particle pT
3
4. Brandon McKinzie 26 July 2016
Simple Checks: Number of Towers Per Cluster
● Single-particle [(η, φ) = (0, 0)] events.
● Varied (but known) particle pT
4
5. Brandon McKinzie 26 July 2016
Simple Checks: Illustrating Cluster ET
● Single generated particle with pT
= 10 GeV/c.
● Each box is a tower labeled with its collected ET
● The sum of 5x5 tower ET
is the cluster ET
5
Electron Photon
6. Brandon McKinzie 26 July 2016
Clustered ET
vs. Generated pT
For each of e-
, , and 0
:
● Generate one single-particle event
○ (η, φ) = (0, 0)
○ known pT
○ noise included
● Plot the cluster ET
that was found
● Repeat for many (independent)
events with different particle pT
6
7. Brandon McKinzie 26 July 2016
Comparison with Simple 5x5 Clusterizer
For each of e-
, , and 0
:
● Collect seed towers.
● Construct simple 5xt5 clusters
centered on each seed.
● More data needed for “fair”
comparison.
7
8. Brandon McKinzie 26 July 2016
Characterizing Shower Shape with
8
Electron Photon Pi0
9. Brandon McKinzie 26 July 2016
Work in Progress
● Writing an evaluation module similar to CaloEvaluator.
● Exploring effect of superclustering/Bremsstrahlung recovery on performance.
● Inspecting properties of clusters with reconstructed ET
<< generated pT
● Running with PYTHIA events / more complicated events than PHG4SimpleEventGenerator.
9