1. Fast Simulation and the Higgs: Parameterisations of photon reconstruction efficiency in H  events Fast Simulation and the Higgs: Parameterisations of photon reconstruction efficiency in H  events Neil Cooper-Smith (RHUL), Pedro Teixeira-Dias(RHUL) IOP Physics Meeting 6 th Apr 2009

2. 2Overview Simulation Options available in ATLAS. Fast Simulation - Atlfast I Brief outline of how it works. AtlfastC & Photon parameterisations Why are they needed? How are the parameterisations created? What can be done with them? Higgs Physics with Atlfast I Benefits of using the parameterisations in Atlfast I Higgs events.

3. 3 Simulation in ATLAS Geant4 simulation –1990s per ttbar event, full detector description. Fast Geant4 –Frozen showers replace low energy EM particles in calorimeter. –757s per ttbar event. Fast calorimeter simulation (Atlfast-2A) –Parameterised shower for each particle hitting the calorimeter face. –101s per ttbar event. Fast calorimeter + inner detector simulation (Atlfast-2F) –Track hit distribution parameterised. –7.41s per ttbar event. Fully parameterized fast simulation (Atlfast-1) –Particle kinematics modelled, most detector effects not modelled. –0.097s per ttbar event (includes reco). Timings courtesy of Simon Dean (UCL)

4. 4 Atlfast I Atlfast I is a very fast, simplified simulation tool for describing the effect of the ATLAS detector on reconstructed particle observables. The 4-vectors of particles are smeared according to the following measured detector resolutions: –Electrons (P T ) –Muons (P T,  + efficiency (if selected) ) –Photons (P T,  ) –Jets (P T ) –b/c-jets and  -jets tagged statistically based on efficiency (P T,  ) By default Atlfast I does not account for losses due to the reconstruction process. No Geant4 layer  No photon conversion information.

5. 5 Motivation for  reconstruction efficiency parameterisations Default Atlfast I does not account for particle reconstruction efficiencies. Standard solution: apply a flat 80% reconstruction efficiency to Atlfast photons. However, lower efficiencies in crack and higher  regions not accounted for. For multi-photon final states, these effects add up. Creation of large statistics background samples with accurate reconstruction efficiencies.

6. 6 Solution: AtlfastCorrectors Developed by the Glasgow and Tokyo groups. Including particle identification/reconstruction efficiencies and rejection rates,  ij (P T,  ) and C ij (P T,  ), in AtlfastI via an ID and contamination matrix. I will detail how the photon ID parameterisations (   ) have been developed, and will demonstrate them in action using H  events in the next part of the talk. Reconstructed candidates  γγ  ee  jj CγeCγe CeγCeγ CγjCγj CjγCjγ C je C ej e   j j   e Generated particles   CeCe CjCj CγCγ CeCe CjCj CγCγ

7. 7 Creation Of Parameterisations Extract parameterisations from full simulation/reco photon rich samples. gg  H(120)  - 50,000 Events gg  H(140)  - 10,000 Events gg  H(200)  - 10,000 Events Choose definition of isolated 1 reconstructed photon: H  WG definition: cuts on photon shower shapes derived by G.Unal. Valid for photons with: 0<|  |<1.37 & 1.52<|  |<2.37 pT>4GeV Truth photons matched (  R < 0.1) to H  reconstructed photons  reconstruction efficiency. Reconstruction efficiency can then be parameterised with just pT and  1 sum pT tracks in a cone (  R=0.4) < 4 GeV

8. 8 Converted Photons To accurately describe the reconstruction efficiency, converted and unconverted photons must be treated separately. For example in a ggH  sample approximately 1/3 of all truth photons are converted photons. Converted (truth level) photon reconstruction efficiency. Truth converted photons can be matched to unconverted photons, single and double track converted photons Unconverted (truth level) photon reconstruction efficiency in a ggH  sample. Now 2 separate sets of params, 1 for converted & 1 for unconverted photons.

9. Converted Photons in ATLFAST 9 ATLFAST simulation does not use GEANT  No conversion information in ATLFAST. Therefore created a parameterisation of the probability of a truth photon converting: Now can distinguish between converted and unconverted photons in ATLFAST and then apply appropriate reconstruction efficiency parameterisation. No dependence on , pT. Just parameterise with .

10. 10 Parameterisations with gg  h(120)  Distributions here show reconstructed photons in a ggH  sample. Reconstructed Photons: Full Sim (black), ATLFAST (red) and AtlfastC (blue).

11. 11 Full Sim (black), ATLFAST (red) and AtlfastC (blue). Efficiencies are well matched. Reconstruction Efficiencies with gg  h(120) 

12. 12 Parameterisations tested on H  background Di-Photon sample. Reconstructed Photons: Full Sim (black), ATLFAST (red) and AtlfastC (blue). Parameterisations with  background

13. 13 Reconstruction Efficiencies with  background Full Sim (black), ATLFAST (red) and AtlfastC (blue). Again everything looks good.

14. 14 Parameterisations tested on ttH(120)  sample. Reconstructed Photons: Full Sim (black), ATLFAST (red) and AtlfastC (blue). Parameterisations with tt  (h(120)  )

15. 15 Reconstruction Efficiencies with tt  (h(120)  ) Full Sim (black), ATLFAST (red) and AtlfastC (blue). Efficiencies appear to be 3-4% out. Due to differences in the photon isolation method used in full simulation and Atlfast.

16. 16 Isolation Problem Photon reconstruction efficiency parameterisations are taken from full simulation and then applied to Atlfast. This method assumes that the isolation algorithms used in full simulation and Atlfast are the same, which they are not. This is where the difference in the ttH  sample comes from. Unfortunately the two algorithms cannot be compared directly as the full simulation algorithm uses variables that are not available within the Atlfast framework. Therefore, a common variable (between full & fast simulation) is needed to describe isolation in full simulation. Possible variable -  R to nearest jet (pT > 10 GeV): Turn of photon isolation algorithm in Atlfast, then apply reconstruction & isolation efficiency parameterisations at the same time.

17. 17 Isolation Solution Photon reconstruction efficiency vs.  R to nearest jet ( pT >10 GeV ), for ggH(120)  (red) and ttH(120)  (black) full simulation samples. Define photons with  R to nearest jet >= 0.6 to be “isolated”.

18. 18 “Isolated” Photons Reconstruction efficiencies (where  R to nearest jet >= 0.6, i.e. “isolated”): ggH(120)  (red) and ttH(120)  (black). Conclusion: - For photons where the  R to nearest jet >= 0.6, the standard parameterisations can be used.

19. 19 “Non-isolated” Photons Reconstruction efficiencies (where  R to nearest jet < 0.6, i.e. “non-isolated”): ggH(120)  (red) and ttH(120)  (black). Conclusion: For photons where the  R to nearest jet < 0.6, parameterise with pT, |  | and energy of nearest jet.

20. 20Summary Presented a working set of physics independent parameterisations for the isolated photon reconstruction efficiency for use in Atlfast I. Parameterisations valid for isolated photons (converted & unconverted) with pT>10 GeV, |  |<2.37 Work is ongoing towards a set of parameterisations for isolated and “non-isolated” photons. –Challenge is to replicate the effect of full simulation photon isolation in ATLFAST. –Initial study suggests that it is possible to parameterise non-isolated photons with pT, ,  R to nearest jet (pT >10 GeV ) & the Energy of the nearest jet. –Turn off Atlfast default photon isolation and apply the parameterisations of reconstruction efficiency & isolation to photons in ATLFAST. –Next release of ATLFAST I will include a full simulation like isolation algorithm, making comparisons between ATLFAST and full simulation easier. Final goal is to have a set of parameterisations for the photon reconstruction efficiency that can be used regardless of the physics of the event.