Layout Performance Analysis with Particle Gun & Pileup Studies
The study focuses on the performance analysis of layout variations using a particle gun and pileup technique. The layouts ExtBrl4, IExtBrl4, and InclBrl4 were compared for resolution, efficiency, fake probability, and hits per track. Details on Soshi's technique, truth/track ratios, total hits distribution, and setup information for the project are presented.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
Step1 layouts performance analysis with particle gun + pileup Federico Massa (Pisa) Speaker: Giorgio Chiarelli (Pisa) Supervisors: Claudia Gemme (Genova) Chiara Roda (Pisa) 15th of March, 2016 Pixel R&D Projects Meeting Bologna 1
Layouts We studied three layouts with: same range different pixel barrel geometry ExtBrl4 IExtBrl4 InclBrl4 Unless specified, plots shown will always follow this layout ordering 2
Tool: Particle Gun + Pileup - Generated files used for testing: Sample generated with release 17.2.0.2 using Soshi Tsuno s technique (see next slide) 50K events of single pion with added pileup 10, 50, 140, 200 fixed pt: 15 GeV/c flat eta: [-4.5, 4.5] flat phi: [-pi, pi] Layouts: ExtBrl4, IExtBrl4, InclBrl4 3
Soshis technique in brief - Cut at generation level of pileup truth outside cone in R = 0.1 (see Soshi: https://indico.cern.ch/event/387170) - In the past months we tested this technique on LoI samples with performances studies - We have now repeated the study with Step1 samples - We also studied stability against choice of R = 0.1 to see if the effect of particle loss due to curvature is negligible no significant change within statistics from 0.1 to 0.2 (see https://indico.cern.ch/event/464857) In this presentation: Resolution, efficiency, fake probability, number of hits/track for three layouts: ExtBrl4, IExtBrl4, InclBrl4 Ideas for physics studies 4
Number of total hits InclBrl4 ExtBrl4 IExtBrl4 <ExtBrl4> = 13.6 <IExtBrl4> = 15.2 <InclBrl4> = 15.4 Cut on total hits > 5 at reconstruction level 6
Setup Release setup: asetup 20.20.0.1, AtlasProduction, here - - conditionsTag: OFLCOND-MC12-ITK-27-00 In this reconstruction, ExtBrl4 layout doesn t fully use long clusters Its performances are affected by it so it always shows worse performances than the others For all layouts: - Since our sample production there has been a major improvement in the passive material description, so there could be some change in fakes and efficiencies Step 1 is a somewhat rough description of the detector: the results we obatined should be considered only qualitatively - 7
Resolution and bias: definitions Hard scatter track is defined as the track of minimum R from hard scatter truth For every hard scatter track: 1. We take the histogram corresponding to the interval the track belongs to 2. We fill it with the difference between the track and the truth variable being examined (pt, D0, ...) NB: For resolution and bias calculation we take into account ALL reconstructed matched tracks (fakes included) However, we do NOT expect this to affect the resolution in a significant way as the fake probability is pretty low At the end of the run: 1. We fit the residuals with a Gaussian function following an iterative method 2. Resolution( ) = Sigma taken from fit 3. Bias( ) = Mean taken from fit 8
Resolution: Pt - Fixed pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 50 pu 200 9
Resolution: Pt - Fixed layout, different pileup Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 ExtBrl4 IExtBrl4 InclBrl4 10
Bias: Pt - Fixed Pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 200 pu 50 11
Resolution: D0 - Fixed Pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 200 pu 50 12
Resolution: D0 - Fixed layout, different pileup Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 InclBrl4 ExtBrl4 IExtBrl4 13
Bias: D0 - Fixed pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 50 pu 200 14
Fake definition We use Soshi s fake definition Hard scatter track is defined as the track of minimum R from hard scatter truth Mismatched tracks Fake Track with R from truth > 0.02 Plot of minimum R between track and truth (log scale) 15
Fake probability - Fixed pileup, different layout Fake Probability ( ) = P(track is a fake | hard scatter track is in a certain bin) With 50K events the error is still large and the layouts are compatible pu 200 Integrating in : Pileup/Layout ExtBrl4 IExtBrl4 InclBrl4 0.22 +- 0.02 % 0.15 +- 0.02 % 0.15 +- 0.02 % 10 0.20 +- 0.02 % 0.14 +- 0.02 % 0.15 +- 0.02 % 50 0.24 +- 0.02 % 0.17 +- 0.02 % 0.17 +- 0.02 % 140 0.23 +- 0.02 % 0.17 +- 0.02 % 0.14 +- 0.02 % 200 Significantly worse than the others *** Similar performances 16 *** Suffers from lack of efficient use of long clusters
Efficiency - Fixed pileup, different layout Efficiency ( ) = P(track is not a fake | hard scatter truth is in a certain bin) pu 50 pu 200 17
Efficiency - Fixed layout, different pileup Efficiency ( ) = P(track is not a fake | hard scatter truth is in a certain bin) IExtBrl4 ExtBrl4 18
Efficiency - Fixed pileup, different layout Efficiency ( ) = P(track is not a fake | hard scatter truth is in a certain bin) pu 50 pu 200 19
Efficiency - Fixed layout, different pileup Efficiency ( ) = P(track is not a fake | hard scatter truth is in a certain bin) ExtBrl4 IExtBrl4 20
Future studies: H->4 The idea is to generate a sample of H->4 We can add pileup to this sample following Soshi s technique, but with four cones instead of one By using the same Higgs sample for different pileup we can eliminate generation statistical fluctuations Could the cones overlap? At 7 TeV the isolation requirement for muons is R > 0.1, so if we keep using the same value for the generation cone, they should not overlap (see https://cds.cern.ch/record/1408784?, p.19) Background? We mean to simulate with the same technique background events like ZZ->4 This way, we could undertake physics studies without a high CPU requirement or having to wait for official Full simulation samples Particularly useful during early layout optimization stages 21
Conclusions The performances of IExtBrl4 and InclBrl4 are very similar, whereas the ExtBrl4 seems to be significantly worse. With the reconstruction tag used (OFLCOND-MC12-ITK-27-00) the long clusters in ExtBrl4 are not used yet, so that is one reason. Since the production of our samples there has been a major correction to the passive material description, that could also explain the mild dependency of the fake probability on pileup conditions. To-Do list: Consider possibility to use already existent samples (they are alternated in pt and charge!) Share our samples with the community Simulate physics processes with Soshi s technique: H->4 22
Number of hits: ExtBrl4 Cut on total hits > 5 at reconstruction level 24
Number of hits: IExtBrl4 Cut on total hits > 5 at reconstruction level 25
Number of hits: InclBrl4 Cut on total hits > 5 at reconstruction level 26
Resolution: Eta - Fixed pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 50 pu 200 29
Resolution: Eta - Fixed layout, different pileup Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 ExtBrl4 InclBrl4 IExtBrl4 30
Bias: Eta - Fixed pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 50 pu 200 31
Resolution: Phi - Fixed pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 50 pu 200 32
Resolution: Phi - Fixed layout, different pileup Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 ExtBrl4 IExtBrl4 InclBrl4 33
Bias: Phi - Fixed pileup, different layout Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 pu 200 pu 50 34
Bias: Pt Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 35
Bias: D0 Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 36
Bias: Eta Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 37
Bias: Phi Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 38
Fake probability Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 39
Efficiency vs eta Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 40
Efficiency vs phi Cuts: NB: Only considering hard scatter track , not pileup tracks s pt > 3 GeV/c Total Hits > 5 41
