1. Lily Asquith (ANL) on behalf of ATLAS Boost 2012, Valencia jet shapes 0

2. Outline What are jet shapes, and why are we measuring them? Experimental challenges. The measurements. arxiv:1206.5369arxiv:1206.5369 What’s new? 1

3. What are jet shapes? η φ All of these observables are constructed using the angular separation and energy of the jet constituents. e.g. mass: A jet. A constituent. Traditionally jet shapes are differential and integrated. arxiv:1101.0070, arxiv:1204.3170 arxiv:1101.0070arxiv:1204.3170 These ‘shapes’ are different measures of energy flow: mass, width, planar flow, eccentricity and angularity. 2

4. Core-heavy jet: width  0 η φ Width 3

5. Broad jet: width  1 η φ 4

6. Quark/ gluon? Quark/gluon jets: width (or girth); gluon jets are broader than quark jets, with more tracks. arXiv:1106.3076v2 5

7. Planar flow η φ Two-body jet: Linear energy deposition: Planar flow  0 6

8. Planar flow η φ Three-body jet: Planar energy deposition: Planar flow  1 7

9. Eccentricity Isotropic energy deposition: eccentricity  0 η φ 8

10. Eccentricity η φ Elongated energy deposition: eccentricity  1 9

11. Two- and higher-body decays Planar flow can distinguish between three-body (top) jets and two- body (light quark/ gluon) jets. arXiv:0807.0234 10

12. Angularity τ -2 η φ Asymmetric energy deposition: τ -2  maximum 11

13. Angularity τ -2 η φ Symmetric energy deposition: τ -2  0 12

14. Different two-body decays Angularities can distinguish between two-body (W/Z/H) jets with different polarisation and two-body (light quark/gluon) jets. arXiv:0807.0234 Longitudinal Z/ QCD Transverse Z/ QCD z=m/pT Longitudinal Z jets QCD (light quark, gluon) jets 13

15. Correlations between observables High pT, central, Pythia6 dijets. Mass and width are strongly correlated. Planar flow and eccentricity are strongly anti-correlated. 14

16. Correlations between observables At high mass, the correlations change. These are for QCD. Mass > 100 GeVNo mass cut 15

17. The experimental challenges: aka Pileup 16

18. Why pileup is such a problem for jet shapes and substructure 1: These jets are big. These sorts of observables generally change under pileup like R 2 or more… 17

19. Why pileup is such a problem for jet shapes and substructure 2: We want to be able to distinguish A from B… AB 18

20. Why pileup is such a problem for jet shapes and substructure 2: We want to be able to distinguish A from B AB … in these conditions. 19

21. Pileup 2010: ~2 (28% of events NPV=1)  special dataset The Number of reconstructed Primary Vertices - NPV – can tell us how much additional radiation we are dealing with. 2011: ~ 10 2012 2012*: ~ 25+. 20

22. Controlling pileup Complementary cone technique (CDF) looks in region transverse (in azimuth) to the jet. Energy deposits in this region are attributed to pileup and underlying event (UE): soft radiation that is always present. arxiv:1101.3002arxiv:1101.3002, 1106.5952v21106.5952v2 21

23. Controlling pileup Single vertex events contain only the UE contribution  characterise pileup by comparing events with single and multiple vertices. expected measured arxiv:1206.5369 Can then find the scaling of e.g. ΔM with R  obtain subtractions for R=1 jets. 22

24. Controlling pileup Complementary cone technique restores distributions to shape seen in single vertex events. 23

25. The measurements 24

26. Details Events are selected based on run conditions, data quality and detector conditions. The anti-kT algorithm is used with locally calibrated topological clusters as input. The highest pT jet in each event is measured, must have pT>300 GeV. ObservableRMass rangePileup correction Mass M0.6,1.0All ✔ Width W0.6,1.0All ✔ Planar flow P1.0130-210NPV=1 Eccentricity ε0.6,1.0>100 ✔ Angularity τ -2 0.6100-130Not needed 25

27. Jet mass PYTHIA8, PYTHIA6 HERWIG++ 2.4.2, 2.5.1POWHEG, PYTHIA6 R=0.6 R=1.0 26

28. Jet mass Herwig++ 2.5.1 jet mass prediction is greatly improved w.r.t 2.4.2 27

29. Jet mass Eikonal approx of QCD for gluons and quarks is compatible with our expectation that the data is a mixture of quark and gluon initiated jets. 28

30. Jet mass Dominant contributions to the systematic uncertainty are the cluster energy scale and Monte Carlo predictions. These show ΔC on the y-axis: C is the correction factor in bin i when going from detector-level to particle-level jets in the “baseline” Pythia6 (AMBT1) MC sample. ΔC is the difference when we vary the sample w.r.t this baseline. Shading is statistical uncertainty. 29

31. Jet width Width is well-modeled by all MCs beyond the first bin. 30

32. Details ObservableRMass rangePileup correction Mass M0.6,1.0All ✔ Width W0.6,1.0All ✔ Planar flow P1.0130-210NPV=1 Eccentricity ε0.6,1.0>100 ✔ Angularity τ -2 0.6100-130Not needed Planar flow is measured for jets with mass in a window around the top mass. Not many R=0.6 jets have such a high mass: Only measure P for R=1.0 jets. Only measure P in pileup-free (NPV=1) events. 31

33. Planar flow Again we see Herwig++ 2.5.1 providing a superior description of the energy flow wrt 2.4.2. Note: this is not the same mass range as the eccentricity measurements. 32

34. Details Eccentricity is measured in the general “region of interest” for boosted particle searches: M>100 GeV. ObservableRMass rangePileup correction Mass M0.6,1.0All ✔ Width W0.6,1.0All ✔ Planar flow P1.0130-210NPV=1 Eccentricity ε0.6,1.0>100 ✔ Angularity τ -2 0.6100-130Not needed 33

35. Eccentricity Eccentricity is a magnifying glass for differences in the distributions of constituents on the “local” angular scale: 34

36. Eccentricity Eccentricity is a magnifying glass for differences in the distributions of constituents on the “local” angular scale: This piece varies significantly between MCs, but (mostly) washes away with energy weight (soft particles). Highly anti-correlated with planar flow (-90% for jets in same high mass range) 35

37. Details QCD small-angle approximation gives a prediction for the peak and maximum values of the τ -2 distribution: Valid for “fixed” high mass and pT (we choose 100<M<130) Meaningful for “smaller” jets only Corrections in 2010 pileup conditions are negligible, so none applied. ObservableRMass rangePileup correction Mass M0.6,1.0All ✔ Width W0.6,1.0All ✔ Planar flow P1.0130-210NPV=1 Eccentricity ε0.6,1.0>100 ✔ Angularity τ -2 0.6100-130Not needed 36

38. Angularity Nice agreement between data and MC and with QCD small angle approx. 37

39. What’s new? 38

40. Jet mass and 2011 pileup The jet mass versus the number of reconstructed primary vertices per event (NPV) in 2011 data for five different jet algorithm/pruning configurations. From left to right these are [1] Anti-kt, [2] Pruned anti-kt, [3] Trimmed anti-kt, [4] Cambridge-Aachen and [5] Filtered Cambridge-Aachen. As the animation plays, the distance parameter (R) of the jet increases from 0.4 to 1.6. The mean mass in each bin of NPV is indicated by the black markers 39

41. Angularity and 2011 pileup The jet angularity versus the number of reconstructed primary vertices per event (NPV) in 2011 data for five different jet algorithm/pruning configurations. From left to right these are [1] Anti-kt, [2] Pruned anti-kt, [3] Trimmed anti-kt, [4] Cambridge-Aachen and [5] Filtered Cambridge-Aachen. The mean mass in each bin of NPV is indicated by the black markers. 40

42. In summary Our current MC generators are correctly describing the jet substructure we see in data, in some detail. 2011 and 2012 data: – More data! – More opportunity to explore methods for dealing with pileup! – More opportunity to ask questions about how the characteristics of a jet vary according to its parenthood! 41

43. 42

44. Details of the grooming configurations Pruned: During the reclustering of the jet we look at the pT fraction and angle of the cluster we are seeking to combine into our protojet. If the cluster is soft, i.e. carries less than 6% of the protojet pT, or is "wide angle" dR>0.3, then we chuck it. Then we rebuild the jet from remaining clusters. Trimmed: The jet constituents are reclustered with a small distance parameter R=0.3 into subjets. Any subjet with pT<5% of the jet pT is chucked. Then we rebuild the jet from remaining clusters. Filtered: The jet constituents are reclustered with a small distance parameter R=0.3 into subjets. Anything outside the three hardest subjets is chucked. Then we rebuild the jet from remaining clusters.