New jet tools for discovering New Physics at the LHC. Lian-Tao Wang Princeton University

Transcription

1 New jet tools for discovering New Physics at the LHC Lian-Tao Wang Princeton University

2 Outline Brief overview. Improving jet algorithms. 2 examples. Jet substructure and new physics searches. Boosted tops. Higgs search. New physics in WW scattering. Heavy squark. Outlook.

3 LHC, the energy frontier. Many fundamental questions to be answered. Electroweak symmetry breaking, origin of mass. Dark matter. Supersymmetry, extra-dimension...

4 Major challenge: Tackling hadronic final states

5 The importance of hadronic final state: Everywhere at hadron colliders. Present in (almost) all new physics signals. Many of them only have hadronic channels. TeV new physics states can decay to SM heavy particles, e.g. t, W, Z, often look like a cluster of hadrons. Understanding of basic structure of QCD and the properties of new physics has lead to the development of a set of modern tools which significantly enhanced the discovery potential.

6 Why is it hard?

7 Why is it hard?

8 Why is it hard?

9 Why is it hard? jet jet Sterman & Weinberg, PRL,1977. jet jet We would like to preserve.

10 Why is it hard? jet jet ISR jet jet

11 Why is it hard? jet jet ISR beam Multiple interaction, underlying events, pile-up jet jet

12 Why is it hard? jet jet ISR beam Multiple interaction, underlying events, pile-up jet jet Overlapping jets.

13 Why is it hard? jet jet ISR beam Multiple interaction, underlying events, pile-up jet jet Overlapping jets. ISR (beam) clustered

14 Why is it hard? beam jet ISR jet Part of the beam? Multiple interaction, underlying events, pile-up jet jet Overlapping jets. ISR (beam) clustered

15 Why is it hard? Proper size of jets. jet ISR jet Part of the beam? beam Multiple interaction, underlying events, pile-up jet jet Overlapping jets. ISR (beam) clustered

16 Why is it hard? Proper size of jets. jet ISR jet Part of the beam? beam Multiple interaction, underlying events, pile-up jet To best preserve jet Overlapping jets. ISR (beam) clustered we would like to: Use smart jet shapes. Control contamination.

17 Two main characters of better jets Light parton initiated jets No loss of FSR. No contamination. boosted t, W Z initiated jets How to distinguish them from the light parton QCD jets.

18 New improved jet algorithms

19 Begin with jet algorithm An algorithm of clustering together close by objects. Basic ingredients of a sequential jet algorithm. Two types of distances Jet-jet distance: Jet-beam distance: d ij d ib when to cluster Pair wise comparison of all distances when to stop clustering If smallest distance at any stage in clustering is jet-jet, add together corresponding four-momenta, else take jet with smallest jet-beam distance and set it aside. Repeat till all jets are set aside.

20 η d 12 d 23 d 13 φ d 12 <d 13 <d 23 <d (1,2,3)B <d i4 4

21 η φ 4

22 η d 12 φ d 12 <d (1,2)B <d i4 4

23 η φ 4

24 η φ 4 Done!

25 Standard Recombination Algorithms k T algorithm d ij = min(p 2 Ti,p 2 Tj) ( R R 0 ) 2,d ib = p 2 Ti C/A algorithm d ij = ( ) 2 R,d ib =1 R 0 anti-k T algorithm d ij = min(p 2 Ti,p 2 Tj ) ( R R 0 ) 2,d ib = p 2 Ti

26 Standard Recombination Algorithms k T algorithm d ij = min(p 2 Ti,p 2 Tj) ( R R 0 ) 2,d ib = p 2 Ti A B C/A algorithm d ij = ( ) 2 R,d ib =1 R 0 A k T B anti-k T algorithm C/A d ij = min(p 2 Ti,p 2 Tj ) ( R R 0 ) 2,d ib = p 2 Ti A B anti k T

27 Boost invariant dynamical jet shapes. Jets with fixed cone size in Jets from new physics signal have different shapes (see later). Risking either missing too much in the forward region or taking in too much extra radiation in the center.

28 Smart jet shapes. D. Krohn, J. Thaler, LTW, arxiv: Jets from new physics resonances decay are likely to be isotropical in the resonance center of mass frame. Boost isotropical jets to the lab frame: Center of Mass frame S Lab frame R boost S R For small opening angle: Therefore, we propose varying the jet size as VR algorithm Dynamical jet shape, boost invariant, infrared and collinear safe.

29 Effect of VR jets Jets from the anti-kt Algorithm Jets from the anti-kt Algorithm Jets from the AKT Algorithm Jets from the AKT-VR AKTVR anti-kt(vr) Jets from Algorithm the anti-kt(vr) Algorithm Phi Phi Eta Eta Figure 1: The same event reconstructed by anti-k T (left) and its VR modification (right). Note that in going to the Fixed VR algorithm, size the high-p T jets (dark blue, green) VR have been reduced in size while softer jets (yellow, purple, light blue) have grown. In this example, only the two harder jets are expected to exhibit VR-symmetry, and the softer jets are saturating the R m a x =1.0 constraint. In VR algorithm, jet size is properly scaled, and appropriate for the underlying process. Our signal and background samples have both been generated in Pythia [10], with parton-level signal events generated in MadGraph [11]. We use nominal LHC beam parameters (14 TeV proton-proton collisions). Final state hadrons are grouped into

30 Jet trimming Effect of the contamination. Initial state radiation (ISR), multiple interaction (MI), underlying events (UE), pile-up (PU). Cross Section [A.U.] R=0.9 R=1.1 R=1.3 R=1.5 Cross Section [A.U.] R=0.9 R=1.1 R=1.3 R= Mass [GeV] FSR only No contamination Mass [GeV] Including ISR, MI, UE, pile-up Room for improvement!

31 A closer look at the soft radiations ISR scale with the hard collision p ISR q 2 Λ 2 hard MI, UE, and pileup incoherent, independent of the hard collision scale. A universal soft background.

32 Jet trimming. D. Krohn, J. Thaler, LTW, arxiv: Introducing a cut on soft radiation. Discard stuff below the cut after jet clustering. Our implementation. Cluster all calorimeter data using any algorithm Take the constituents of each jet and recluster with smaller radius Rsub (Rsub = 0.2 seems to work well). Discard the subjet i if p Ti <f cut Λ hard Best choice of the hard scattering scale and fcut. Process dependent. Can be optimized experimentally. Related but different approaches: Filtering: J. Butterworth, A. Davison, M. Rubin, G. Salam, arxiv: Pruning: S. Ellis, C. Vermilion, J. Walsh, arxiv: ISR argument.

33 Start "# "! 1 2 "# "! Cluster into subjets "# "# 1.5 Discard soft subjets "! "! Reassemble

34 Reduced jet area Cross Section [A.U.] anti-k T anti-k T trimmed Jet Area

35 Simple test case: di-jet resonance Cross Section [A.U.] anti-k 0.25 T trimmed anti-k T Cross Section [A.U.] 0.25 VR VR trimmed Mass [GeV] Mass [GeV] Improvement f cut, N cut R sub R 0, ρ Γ [GeV] M [GeV] anti-k T anti-k T (N) 40% anti-k T (f, p T ) 59% anti-k T (f, H) 61% VR 30% GeV VR (N) 53% GeV VR (f, p T ) 68% GeV VR (f, H) 73% GeV Filtering 27% 2 R 0 / We provide plugins fully compatible with Fastjet.

36 Jet substructure, and applications in new physics searches.

37 Boosted tops. Tops are interesting! Top plays an important role in electroweak symmetry breaking. Top generically couples to heavy new resonances which is an important part of TeV new physics. Examples. Composite top couples strongly to other composite resonances. Many examples. K. Agashe, A. Delgado, M. May, R. Sundrum, hep-ph/ M. Carena, B. Panes, A. Medina, N. Shah, C. Wagner, arxiv: , New heavy scalars couple like Higgs. For example: A. Manohar and M. Wise, hep-ph/ A good example of subjet techniques.

38 Boosted top is also hard to identify. Heavy resonance decay. t X t b e +, u,... boost No isolated objects t W + ν, d,... For m t t > 3 TeV, > 90% events with at least one top fully collimated. Large fraction of events 2-object -like. QCD b b, jj background. A few % with lepton isolation B. Lillie, L. Randall, and LTW, hep-ph/ L. Almeida, S. Lee, G. Perez, I. Sung, J. Virzi, arxiv:

39 (hadronic) Top tagging at the LHC Fully collimated tops look like QCD jets. Basic intuition Top decay. QCD: radiation. 3 hard objects. Energetic tops should lead to massive jets with some substructures. How well can this be distinguished from (massive) QCD jets?

40 QCD jets: parton shower A QCD jet is built up by many radiations (branching). A process is approximated by parton shower.

41 QCD vs top jets: QCD jet: No obvious feature in jet mass Sudakov factor: Radiate more. Top jet, first branching (decay):

42 Top tagging: jet mass QCD jets also have mass. Jet mass. Useful. Additional variable? Using jet mass only.

43 Additional help from new jet algorithm FSR only anti-k T Without contamination Cross Section [A.U.] ISR/MI/pileup anti-k T Trimming FSR only Trimming ISR/MI/pileup With contamination With trimming Jet Mass [GeV] More faithful (smaller) jet mass for the background. Effect of radiation contamination on the jet mass Trimming gives large improvement by reducing effective jet size significantly.

44 QCD vs top jets: QCD jet: Prefers soft radiation Sudakov factor: Radiate more. Top jet, first branching (decay): z at the first branching can distinguish top from QCD jet.

45 Other choices of z-variables? Many possible choices 1. z cell = min(e A,E B ) E A +E B, E X i X E i, 2. z cut d cut d cut +Q 2 M min(e A,E B ) E A +E B where d cut = min(p 2 TA,p2 TB ) R2 AB, R2 AB (φ A φ B ) 2 +(η A η B ) 2 3. z LI = min(p ref p A,p ref p B ), with any p p ref (p A +p B ) ref Preserve IR singularity and approximation factorization at leading log as long as z min(e A, E B )/E M in collinear on-shell limit. Similar performance.

46 Substructure, z-finding!1'2-/.-)*#3*+,-*4-+*5$6.+-/&'(*,&.+#/07 Jet clustering history is approximately the inverse of parton shower.

47 Top jets vs QCD jets Rough approximation of finite calorimetry: δη δφ = QCD soft singularity is in effect regulated. J. Thaler, LW, arxiv: S. Catani, Y. L. Dokshitzer, M. H. Seymour and B. R. Webber, Nucl. Phys. B 406, 187 (1993).

48 Top jets vs QCD jets Combined cuts on jet mass and z can enhance further Rough approximation of finite calorimetry: δη δφ = the signal with respect to the background. QCD soft singularity is in effect regulated. J. Thaler, LW, arxiv: S. Catani, Y. L. Dokshitzer, M. H. Seymour and B. R. Webber, Nucl. Phys. B 406, 187 (1993).

49 Top tagging efficiency J. Thaler and LTW, arxiv: Performance of different z variables. Combined cuts z-variable gives an additional about factor of 2 enhancement in performance. Together with jet mass, an enhancement of 100 of S/B is possible. Related studies: D. Kaplan, K. Reherman, M. Schwartz, B. Tweedie, arxiv: L. Almeida, S. Lee, G. Perez, G. Sterman, I. Sung, J. Virzi, arxiv: Gustaaf H. Brooijmans, arxiv: ; CMS, CMS PAS JME

50 More jet shape variables. Top decay is more like 3-body. Span a plane perpendicular to the jet axis. Transverse sphericity, or planar flow S ij = p i α p j α p α p α α jet α jet. J. Thaler and LTW, arxiv:

51 Using planar flow to identify top jets. is not very well modeled by parton shower. Also affected by contamination from underlying events.

52 Better reconstruction of the jet shape anti-k T FSR only Planar flow Defined in L. Almeida, S. Lee, G. Perez, G. Sterman, I. Sung, J Virzi, arxiv: Cross Section [A.U.] With contamination anti-k T ISR/MI/pileup Trimming FSR only Trimming ISR/MI/pileup 200 With no contamination 100 With trimming Planar Flow Can be used to further improve top tagging. An additional factor of several possible. Interesting to compare with improved QCD calculation, using modern technologies such as SCET.

53 Slow tops, Standard Model top pair. For non-boosted tops, 6 objects in the final state. Very crowed event. Fully hadronic, 6+ jets, very hard. New VR algorithm can help since it has a dynamically adjustable size. We see at least improvement. D. Krohn, C. Popa, and LTW, in progress

54 Additional applications of top reconstruction in new physics signal. Top partner decay. Gluino decay. P. Meade, M. Reece, hep-ph/ T. Han, R. Mahbubani, D. Walker, and LTW, arxiv: B. Acharya, P. Grajek, G. Kane, E. Kuflik, K. Suruliz, and LTW, arxiv: We expect new jet algorithms described here to help in both cases.

55 Hiding Higgs. Alternative decay channels can dramatically change Higgs search strategy. Why can new jet technology help? For example: P. Graham, A. Pierce, J. Wacker, hep-ph/ M. Carena, T. Han, G. Huang, C. Wagner, arxiv: For example: B. Bellazzini, C. Csaki, A. Falkowski, A. Weiler, arxiv: , arxiv: Less radiation outside this cone Jet substructure h Higgs Jet Boosted Higgs, studied in the context of SM-like Higgs by J. Butterworth, A. Davidson, M. Rubin, G. Salam, arxiv:

56 Some preliminary results. Min/Max ratio of eta inv. masses(r=0.3 subjets,pt_min=10 GeV) hist Entries 4171 Mean RMS Min/Max ratio of subjet inv. masses(r=0.3 subjets,pt_min=10 GeV) hist Entries 1134 Mean RMS Min/Max ratio of eta inv. masses (R=0.3) Min/Max ratio of subjet inv. masses (R=0.3) Rad. Outside two subjets/fat-jet energy for Higgs jet 0.12 hist Entries 4171 Mean RMS Rad. Outside two subjets/fat-jet energy for Z+j jet hist Entries 1134 Mean RMS Ratio Rad. Outside subjets Ratio Rad. Outside subjets Higgs + Z signal Z+jet background A. Falkowski, D. Krohn, J. Shelton, A. Thalapillil, and LTW, in progress.

57 Encouraging results. (rates in fb) jet mass planar flow radiation pattern Z + h Z + j m h = 80 m h = 100 m h = 120 m h = 80 m h = 100 m h = 120 Start m j α β Cross Section [fb/5-gev] GeV Higgs 100 GeV Higgs 120 GeV Higgs Background Jet Mass [GeV]

58 New physics in WW scattering Direct probe of electroweak symmetry breaking. Typical search strategy involves using leptonic mode, tagging the forward jets, and the so-called central jet veto. No color flow. Low hadronic activity. ν l W W l ν tagging jets

59 Problem with the traditional strategy. Initial state radiation can affect both jet tagging and central jet veto. Very sensitive to factorization scale. ν l ν l W W W W SM background l ν Ambiguity can only be resolved with proper NLO (matrix element+matched parton shower). Use W polarization as a tool. Requires using the boosted hadronic W, and reconstruction based on the 2 subjets of the W-jet. l ν

60 New strategy Use jet substructure to help identify W boson. See also J. Butterworth, B. Cox, J. Forshaw, hep-ph/ T. Han, D. Krohn, LTW and W. Zhu, arxiv: W ν l Reconstruct the W rest frame, and measure W polarization. New physics generically predicts different longitudinal fraction. More robust, less sensitive to QCD corrections. W q q W jet with sub-structure

61 Effectiveness of W-polarization. Example: new physics parameterized by [A.U.] SM prediction c H " c H " c H " c H " c H " c H " c H " P ± (cos θ )= 3 8 (1 ± cos θ ) 2 P L (cos θ )= 3 4 (1 cos2 θ ) * cos! T. Han, D. Krohn, LTW and W. Zhu, arxiv: Can certainly be useful for looking for new resonances in this channel as well.

62 Heavy squark. Scenarios with heavy squark, ~ several TeV, and light gluinos are appealing. Flavor and CP friendly. A feature of many scenarios. Why can jet substructure help as well? hard jet q Gluino jet, with substructure! g LSP q q D. Krohn, P. Mosteiro, and LTW, in progress.

63 Conclusions Better handles on the hadronic final states are instrumental for discovery at the LHC. Based on consideration of QCD radiation, we proposed a set of carefully constructed new jet algorithms and substructure variables. Much improved performance, jet mass, jet shape, etc. We also demonstrate they can significantly enhance new physics signals in many important new physics channels. Boosted or slow hadronic tops, WW scattering, Higgs search, heavy squark... Similar technique can be applied to Tevatron data. A promising direction. Stay tuned.

64 Extras

65 Infrared and collinear safety Infrared safety. No soft radiation can change the number and the directions of the hard jets. VR is IR safe just like other sequential algorithms. Soft radiation clustered either near the end or at the beginning, not affecting hard dynamics. Collinear safety, jets are robust against colinear (within resolution) splittings. Satisfied by VR with

66 Implementation of the algorithm Distance measures. D. Krohn, J. Thaler, LTW, arxiv: The VR algorithm Parameter VR works best if can be optimized. Infrared and collinear safe.

67 Slow tops, SM ttbar. For non-boosted tops, 6 objects in the final state. Very crowed event. Fully hadronic, 6 jets, very hard. New VR algorithm should help since it has a dynamically adjustable size. VR R: jet size parameter C/A Anti-KT D. Krohn, C. Popa, and LTW, in progress

68 Cross Section [A.U.] anti-k T Pruning Trimming Mass [GeV]

69 Why is it possible to gain? MI, UE, and pile-up are incoherent soft background. They can be effectively removed with a cut on soft radiation. Both FSR (want to keep) and ISR (want to discard) have soft radiation, but ISR: FSR is controlled by both collinear and soft singularities: Therefore, a soft cut relative to the jet energy flow could enhance FSR relative to ISR.

70 A model with warped extra-dimension warped space Top is composite, localized towards the IR brane. Top couples strongly to other composite states, KKgluons,... K. Agashe, A. Delgado, M. May, R. Sundrum, hep-ph/

71 Planar Flow [ I kl w = i w i p i,k w i p i,l w i Pf = 4λ 1λ 2 (λ 1 + λ 2 ) 2 of the matrix

72 WW reach Events/100 fb c H " -0.4 c H " * cos! Figure 4: Projected distribution and associated statistical uncertainties of cos θ for the leptonically decaying vector using 100 fb 1 of luminosity.

73 Here are some example cross sections for a particular set of VBF cuts and for different anomalous couplings (labeled c H ξ, which is 0 for the SM). Parton Level [fb] Jet Level [fb] c H ξ β =0.5 β =1.0 β =2.0 β =0.5 β =1.0 β = Stable before parton shower Sensitive afterward Basically, the central jet veto meant to reduce QCD backgrounds makes the analysis very sensitive to the treatment of the forward jets.

74 Event picture from