1. Underlying event at the LHC Multi-parton scattering Tuning of PYTHIA LHC predictions comparison with PHOJET and JIMMY UE in physics analyses Summary Craig Buttar University of Glasgow

2. YETI 06 Durham 2 What is the underlying event High P T scatter Beam remnants ISR Hard and soft physics ISR beam remnants central particle production colour flow/string drawing

3. YETI 06 Durham 3 How to describe low-pt behaviour ?  2→2 >  pp at p T ~5GeV Different approaches but PYTHIA+HERWIG(JIMMY) based on multi-parton scattering (simple scenario with sharp cut-off) Multi-parton interactions

4. YETI 06 Durham 4 Evidence for multi-parton interactions D0 multijet analysis HERA photoproduction indirect Direct CDF underlying event UA5 KNO distributions CDF

5. YETI 06 Durham 5 PYTHIA model Multiple interactions solve total xsect problem Need to tame the PT divergence over QCD cross-section Parameters of the model: p T -min Impact parameter energy dependence Abrupt vs smooth cut- off Matter distribution Number of interactions Number of interactions and fluctuations Parameters not looked at: string drawing, effect of ISR (CDF)

6. YETI 06 Durham 6 How to tune UE – Data Very little UE data –Extensive studies by CDF in ppbar collisions at 1800GeV and some studies at 630GeV Need to cover a bigger energy range to be able to extrapolate to LHC  Use minimum bias data  This assumes that minimum bias is similar to UE of a hard scatter!  Can be used to model the multi-parton interactions  Insensitive to radiation

7. YETI 06 Durham 7 Minimum bias data: UA5 – SPS CDF - Tevatron E735 - Tevatron CERN – ISR Experiment Z. Phys. C 43, 357 (1989) Phys. Rep. 154(5,6) 247 (1987) Phys. Rev. D 41 2330 (1990) Phys. Lett. B 435 453 (1998) Z. Phys. C 37, 191 (1988) Phys. Rev. D 30 528 (1984) References pp at √s = 1.8TeV pp at √s = 200, 546 and 900GeV pp at √s = 30.4, 44.5, 52.6 and 62.2 GeV Colliding beams - - Multiplicity information: ‹n ch ›, dN/d η, KNO, FB, etc. Set π 0, K 0 s and Λ 0 stable Can cover 53-1800GeV but need to be careful when comparing data! Triggers can include diffractive events Some data has reconstructed particles such as  o s (5% effect)

8. YETI 06 Durham 8 Pt-min can be tuned to the give average charged multiplicity as a function of eta Pt-min is ~1.9GeV default value Use MB central rapidity distributions to tune pt-min

9. YETI 06 Durham 9 Compare abrupt and smooth pt-cut-off: Abrupt cut-off generates a Poisson distribution with too few multi-parton interactions in a single event Compare matter distributions: uniform, gaussian, double gaussian  Use double gaussian Use MB multiplicity distributions to tune fluctuations in number of events abrupt smooth Uniform gaussian d-gaussian

10. YETI 06 Durham 10 The underlying event requires less activity => higher pt Lose ‘unification’ of min-bias and underlying event Increasing P t -min Double gaussian with default Core size = 0.2 Core x2 default Default P t -min=1.9 Increasing core size Default Alternatively increase the core size This reduces the core density-reducing activity CDF Run 1 underlying event analysis Phys. Rev. D, 65 092002 (2002) Transverse vs jet p T in UE data from CDF Charged Particles |  | 0.5 GeV/c R<0.7

11. YETI 06 Durham 11 The rapidity distributions are insensitive to the matter distribution Agreement with KNO improves As it reduces the large fluctuations in multiplicity Effect on high multiplicity events.  Change core size rather than pt-min

12. YETI 06 Durham 12 CommentsPYTHIA6.2 - DefaultCDF – Tune A (PYTHIA6.206) PYTHIA6.214 - Tuned Generated processes (QCD + low-pT) Non-diffractive inelastic (MSEL=1) Non-diffractive inelastic + double diffraction (MSEL=0, ISUB 94 and 95) Non-diffractive + double diffraction (MSEL=0, ISUB 94 and 95) p.d.f. CTEQ 5L (MSTP(51)=7) CTEQ 5L (MSTP(51)=7) CTEQ 5L (MSTP(51)=7) Multiple interactions models MSTP(81) = 1 MSTP(82) = 1 MSTP(81) = 1 MSTP(82) = 4 MSTP(81) = 1 MSTP(82) = 4 pT min PARP(82) = 1.9 PARP(89) = 1 TeV PARP(90) = 0.16 PARP(82) = 2.0 PARP(89) = 1.8 TeV PARP(90) = 0.25 PARP(82) = 1.8 PARP(89) = 1 TeV PARP(90) = 0.16 Core radius 20% of the hadron radius (PARP(84) = 0.2) 40% of the hadron radius (PARP(84) = 0.4) 50% of the hadron radius (PARP(84) = 0.5) Gluon production mechanism PARP(85) = 0.33 PARP(86) = 0.66 PARP(85) = 0.9 PARP(86) = 0.95 PARP(85) = 0.33 PARP(86) = 0.66 α s and K-factors MSTP(2) = 1 MSTP(33) = 0 MSTP(2) = 1 MSTP(33) = 0 MSTP(2) = 1 MSTP(33) = 0 Regulating initial state radiation PARP(67) = 1PARP(67) = 4 PARP(67) = 1 ATLAS tuning based on MB and CDF ‘average’ UE data Run-I data CDF based on CDF data including min- max  difference in ISR  difference in gluons

13. YETI 06 Durham 13 PARP(67): Q 2 scale of the hard scattering is multiplied by PARP(67) to define the maximum parton virtuality allowed in showers PARP(67)=4 harder p T spectrum PARP(67)=1 p T ljet > 30 GeV

14. YETI 06 Durham 14 New data from CDF “Leading Jet” “Back-to-Back” TransMAX and transMIN sensitive to radiation and soft UE respectively Back-to-back sample suppresses radiation difference between tranMAX region and transMIN in leading jet and b-2-b jet sample UE depends on topology Tune-A still seems to work

15. YETI 06 Durham 15 Difference is sensitive to ‘hard’ part of UE – radiation

16. YETI 06 Durham 16 Trans-Diff seems ok – Hard part of UE Trans-max and trans-min not so good –Particularly ETsum density

17. YETI 06 Durham 17 UE tuning Parameters Pt-min Matter distribution ISR/FSR String Energy dependence Data – MB dN/deta N ch FB correlation Data – UE Transverse N ch and P t TransMAX, TransMIN, TransDIFF dN/dP t in UE (n ch ) Describing UE is a complicated multiparameter problem covering soft and hard physics

18. YETI 06 Durham 18 LHC predictions Tevatron ● CDF 1.8 TeV PYTHIA6.214 - tuned dN/dη (η=0) N ch jet- p t =20GeV 1.8TeV (pp)4.12.3 14TeV (pp)7.0 increase~x1.8~x3 ~80% ~200% LHC prediction Tevatron PYTHIA6.214 - tuned ● CDF 1.8 TeV LHC MB only UE includes radiation and small impact parameter

19. YETI 06 Durham 19 Other models: PHOJET Developed mainly for soft and semi-hard particle production. Implements ideas of Dual Parton Model for low-p T processes. Multiple Pomeron exchanges (sea- quark multi-chains) enhances the event activity. Uses JETSET (PYTHIA) for high-p t processes Useful comparison for MB and UE studies

20. YETI 06 Durham 20 PYTHIA vs PHOJET: Minimum bias PYTHIA6.214 - tuned PHOJET1.12 pp at √s = 200 GeV pp at √s = 1.8 TeV - - No tuning of PHOJET, defaults used

21. YETI 06 Durham 21 PYTHIA vs PHOJET: Underlying event Transverse PYTHIA6.214 - tuned PHOJET1.12 PYTHIA6.214 - tuned PHOJET1.12 P t (leading jet in GeV) Transverse (GeV) CDF Run 1 underlying event analysis Phys. Rev. D, 65 092002 (2002)

22. YETI 06 Durham 22 Extrapolation to the LHC- Comparison with PHOJET PYTHIA6.2 PHOJET1.12 ● UA5 53, 200, 546 and 900 GeV ▲ CDF 630 and 1800 GeV 0.023ln 2 (s) – 0.025ln(s) + 2.5 √s (GeV) dN/d η at η=0 LHC 0.27ln(s) –3.2 PYTHIA exceeds exp. extrapolation. Transverse PYTHIA6.214 - tuned PHOJET1.12 P t (leading jet in GeV) x 3 LHC x1.5

23. YETI 06 Durham 23 Herwig+Jimmy Jimmy is multi-parton interaction model similar to PYTHIA Main parameter is pt-min Only for hard UE cannot model low-pt ie MB  difficult to get energy dependence Matter distribution is determined from em form factor Gives a good description of CDF data with increased pt-min 2.5  3.25GeV

24. YETI 06 Durham 24 Herwig+Jimmy Extrapolation to LHC predicts much larger UE c.f. underlying event No energy dependent pt-min Which is correct? Transverse P t (leading jet in GeV) Tevatron LHC x3 x5 x4

25. YETI 06 Durham 25 VBF Signal (H  WW  l l ) forward tagging jets correlated isolated leptons low hadronic activity in central region central Higgs production Tagging jet H W W Z/W Important discovery channel For Higgs in mass range 120-200GeV

26. YETI 06 Durham 26 ModelParameter Simple MSTP(82)=1 PARP(81)=1.9 Complex MSTP(82)=4 PARP(82)=1.9 Tuned MSTP(82)=4 PARP(82)=1.8 PARP(84)=0.5 ModelCJV (eff)LEPACC (eff)All Vetoes (eff) Simple0.943 ± 0.0030.610 ±0.0010.084 ±0.001 Complex0.885 ± 0.0030.575 ±0.0010.076 ±0.001 Tuned0.915 ± 0.0030.589 ±0.0010.080 ±0.001 Uncertainty at the level of ~6% on CJV Pt>20GeV and ~3% on lepton Giving a total uncertaintly in the range ~8%

27. YETI 06 Durham 27 Effect of UE on lepton efficiency Vary pt-min by 3  Determine from data Good muons –Barrel  7GeV –Endcap 1.1 9GeV Isolation –  Pt for charged tracks excluding  s with Pt>0.8GeV and  R<0.3 around  in  -  space S.Abdullin et al (CMS) Les Houches 05 Lepton isolation in H->4  ProcessEvent eff Default Event eff –3  Event eff +3  H  4  MH=150GeV 0.775 ± 0.0040.707 ± 0.0050.812 ± 0.004 ZZ background0.780 ± 0.0040.721 ± 0.0050.838 ± 0.004 4 random  Z-inclusive 0.762 ± 0.0070.706 ± 0.0070.821 ±0.006

28. YETI 06 Durham 28 Extract effect of UE from data Use inclusive Z-sample, high statistics Similar dependence to ZZ sample but small systematic shift     random UE

29. YETI 06 Durham 29 First measurements at the LHC ? Charged particle density at  = 0 (Only need central inner tracker and a few thousand pp events) LHC? Multiple interaction model in PHOJET predicts a ln(s) rise in energy dependence. PYTHIA suggests a rise dominated by the ln 2 (s) term.

30. YETI 06 Durham 30 Measuring the minimum bias events at ATLAS  Minimum-bias trigger (commissioning & low-luminosity runs) Minimum bias trigger scintilator

31. YETI 06 Durham 31  Simulated events: Track p T > 400 MeV Track p T > 500 MeV Track p T > 1 GeV

32. YETI 06 Durham 32 Measuring the minimum bias events at ATLAS dN ch /d  dN ch /dp T Black = Generated (Pythia6.2) Blue = TrkTrack: iPatRec Red = TrkTrack: xKalman Only a fraction of tracks reconstructed, because: 1)limited rapidity coverage 2)can only reconstruct track p T with good efficiency down to ~500MeV, and most particles in min-bias events have p T < 500MeV. Previous dN ch /d  measurements published for p T > 0, so need to apply correction factor. Biggest systematic uncertainty? –Run with zero or reduced B-field Reconstruct tracks with: 1) pT>500MeV 1) pT>500MeV 2) |d 0 | < 1mm 2) |d 0 | < 1mm 3) # B-layer hits >= 1 3) # B-layer hits >= 1 4) # precision hits >= 8 4) # precision hits >= 8 p T (MeV) 

33. YETI 06 Durham 33 Ratio / Leading jet E T (GeV) Reconstructing the underlying event CDF Run 1 underlying event analysis Phys. Rev. D, 65 092002 (2002) Njets > 1, | ηjet | < 2.5, ETjet >10 GeV, | ηtrack | < 2.5, pTtrack > 1.0 GeV/c ATLAS DC2 Simulated data How to trigger?

34. YETI 06 Durham 34 UE in  -jet Look a  -jet balance to calibrate jet energies with much better known EM- energy UE event (amongst other effects) will bias the energy balance S.Jorgensen Preliminary ATLAS study

35. YETI 06 Durham 35 UE for different processes R.Field PY Tune A

36. YETI 06 Durham 36 pT balance for different algorithms Parton level Particle level Kt Reconstruction level Kt Parton level Particle level Cone 0.7 Reconstruction level Cone 0.7 Parton level Particle level Cone 0.4 Reconstruction level Cone 0.4 AlgorithmsCone 0.7Cone 0.4Kt Parton level1 - 0% Recon level2 - 0%15 - 7%10 - 2% Biases on pT for the different jet algorithms To understand differences Cone7 and kT, study underlying event, min.bias noise, etc.

37. YETI 06 Durham 37 UE in ep collisions Multi-jet production in photoproduction at HERA requires UE n.b. PYTHIA UE model does not describe  p – always read the manual!

38. YETI 06 Durham 38 Summary and conclusions Tuning the UE is complex as there are both soft physics processes and hard physics processes (ISR/FSR) involved –Beam remnants –Colour flow –Central particle production –Radiation –Topology dependent – leading jet vs back-to-back jets UE is not the same at MB but are related Extrapolation to LHC energies is very uncertain –Uncertain at the level of ~500% (PHOJET  Herwig+Jimmy) –Data will be the ultimate arbiter  how to trigger? –Until then analyses should test how robust they are against variations in UE UE turns up in a number of LHC analyses –Minijets  central jet vetp in VBF studies –Lepton isolation –Energy scale from  -jet

39. YETI 06 Durham 39 Extra slides

40. YETI 06 Durham 40

41. YETI 06 Durham 41

42. YETI 06 Durham 42 Define Charged Particle “Jets” as Circular Regions in  -  Space Order Charged Particles in P T (|  | 0.5 GeV/c). Start with highest P T particle and include in the “jet” all particles (|  | 0.5 GeV/c) within radius R = 0.7 (considering each particle in decreasing P T and recalculating the centroid of the jet after each new particle is added to the jet). Go to next highest P T particle (not already included in a previous jet) and include in the “jet” all particles (|  | 0.5 GeV/c) within radius R = 0.7 (not already included in a previous jet). Continue until all particles are in a “jet”. Maximum number of “jets” is about 2(2)(2  )/(  0.7) 2 ) or 16. “Jets” contain particles from the underlying event in addition to particles from the outgoing partons. 6 particles 5 “jets”