Studies of Higgs hadronic decay channels at CLIC. IMPRS Young Scientist Workshop at Ringberg Castle 18 th July 2014 Marco Szalay

Transcription

1 Studies of Higgs hadronic decay channels at CLIC IMPRS Young Scientist Workshop at Ringberg Castle 8 th July 24

2 Outline Physics at e + e - colliders CLIC - collider and detectors BDT intermezzo Higgs branching ratio measurements Conclusions 2

3 Physics at e + e - colliders precision measurement of: - Higgs sector (couplings, mass, potential) - model independent H to Z coupling - top quark (mass, width) - EW sector Search for new physics: - direct reach up to - model-dependent indirect reach far beyond Cross-section [fb] 3 t t H+X s [GeV] 3

4 Higgs production in e + e - collisions Main H production channels at 35 GeV:! - Higgs strahlung (allows model-independ Z to H coupling measurement) e + e Z - Vector Boson Fusion Z H HX) [fb] - e + σ(e 2 t t Hν e ν e H + - H e e HH ν e ν e H Z W + - H H Z W source: Frank Simon - arxiv: s [GeV] 4

5 Linear Colliders - CLIC A lepton collider allows for precision measurements: - leptons clean events, well defined initial state - linear no synchrotron radiation energy losses - complementary to LHC CLIC (Compact LInear Collider) 2-beam acceleration scheme 3 stages: 35 GeV,.4 & 3 TeV source: linearcollider.org 5

6 Linear Colliders - CLIC II 89 klystrons 5 MW, 42 µs drive beam accelerator 2.4 GeV,. GHz circumferences delay loop 73 m CR 293 m CR2 439 m 89 klystrons 5 MW, 42 µs drive beam accelerator 2.4 GeV,. GHz (c)ft Drive Beam 2.5 km delay loop CR CR2 CR2 CR delay loop 2.5 km decelerator, 24 sectors of 878 m TA BC2 e main linac, 2 GHz, MV/m, 2 km BDS 2.75 km IP BDS 2.75 km e + main linac BC2 TA CR combiner ring TA turnaround DR damping ring PDR predamping ring BC bunch compressor BDS beam delivery system IP interaction point dump 48.3 km e injector, 2.86 GeV e PDR 389 m Almost 5 km long! e DR 427 m BC booster linac 2.86 to 9 GeV e + DR 427 m e + PDR 389 m Main Beam e + injector, 2.86 GeV 2 beams: high intensity (~A) low energy driver beam produces the 2 GHz RF power for a low intensity, high energy beam Goal is to achieve MV/m 6

7 Detectors GOALS! High precision Z l + l - reconstruction Separation of W and Z hadronic decays Measure Higgs couplings to fermions source: REQUIREMENTS! Excellent tracking Highly granular calorimeters Very efficient b and c jet tagging Precise time stamping 7

8 Physics Benchmarks Simulation of a set of physics processes to quantify the response of a particular detector design Evaluate the physics potential of new accelerators Crucial to compare detectors before building them (or even prototypes) Can help to tune hardware parameters 8

9 Higgs production at CLIC - 35 GeV 9

10 Higgs event selection MC events with Particle Flow info search for isolated leptons found e + /e - found μ + /μ - remove e + /e - remove μ + /μ - 2 jets 4 jets Durham algorithm + NN flavor tagging 2 jets 2 jets find best jet match for Z and H Multiclass TMVA classifier (BDTG) E, pt, thrust, Cos(θH), Ynn, MZ, MH, Cos(θZ), PtH, PtZ, # charged particles

11 Boosted Decision Trees Intermezzo Start with a MC sample of signal and background events Split the sample in 2 (train and test) Find the variable and the value that gives the best signal/background separation Repeat on the children nodes BOOSTING: Make new decision tree, previously misclassified events are weighted more source: TMVA user guide

12 Higgs event selection - II Entries Signal events Mistagged signal events Background.4.3 WORK IN PROGRESS reconstructed H mass (GeV) The BDT effectively suppresses the non Higgs background and properly separate the various signal final states 2

13 Higgs Branching Ratios Higgs BR + Total Uncert cc µµ bb gg Z WW ZZ M H [GeV] LHC HIGGS XS WG 23 For 26 GeV Higgs boson H 56% H 23.3% H 8.5% H 6.% H 2.9% H 2.8% H.23% H.2% Separate H bb, H cc and H gg based on flavor tagging information 3

14 Flavor Tagging Precise vertex reconstruction allow to identify the interaction point and secondary vertices Jet flavor tagging information come from these vertices (e.g. M and p for the leading hadron in the decay) 4

15 Higgs decay separation Btag jet + Btag jet WORK IN PROGRESS H->bb 4 3 Btag jet + Btag jet WORK IN PROGRESS H->cc Ctag jet + Ctag jet Ctag jet + Ctag jet2 Btag jet + Btag jet WORK IN PROGRESS H->gg 3 Btag jet + Btag jet2 2.2 H gg and other Higgs decays (WW,ττ, ZZ) can t be easily separated Additional.8 filtering needed! WORK IN PROGRESS H->other Ctag jet + Ctag jet Ctag jet + Ctag jet2 5

16 Higgs decay separation - II dn/.84 #.6.4 Additional TMVA.2 (BDTG). filter on:! Flavor tagging.2 info H jet2 - N particles [#] number of particles in jet (discriminates 25 tau jets) dn/.254 F 2 Higgs invariant 5 mass (discriminates W leptonic decay - missing 5 E) H jet2 - ctag [F] Ynn (discriminates H WW qqqq - more jets than what is clustered) dn/.84 # dn/.948 dn/2.59 # dn/.254 F dn/.52 dn/.246 F WORK IN PROGRESS H jet2 Jet - N Clustering particles [#] - Y2 H jet2 - N charged parts [#] Input variable: m_ynn_nomumu_4_ WORK IN PROGRESS H jet2 Y45 - ctag - No MuMu [F] H - btag + btag2 [F] Hbb U/O-flow:. /. %Hothers U/O-flow:. /.6 % dn/.84 dn/.948 # # dn/.256 F dn/.254 dn/.52 F F dn/.488 F Input variable: m_ynn_nomumu_2_ H jet2 - N charged parts [#] H jet2 - N H particles jet2 - btag [#] [F] WORK IN PROGRESS WORK IN PROGRESS H - btag + btag2 [F] H - jet2 ctag - ctag + ctag2 [F] [F] 3 binary classifier give the probability to be an H bb, H cc or H gg event Hbb U/O-flow:. /. %Hgg U/O-flow:. /.3 % dn/.948 dn/.256 # F dn/.52 dn/.488 F F 6

17 Templates BB template CC template GG template c tag b tag c tag b tag c tag b tag D projection of 3D templates H->others template SM BKG template c tag b tag c tag b tag

18 Fitting the templates Binned maximum likelihood fit Assume Poissonian fluctuation for each data bin: with n = number of data entries in bin ijk and μ = w m T m for the same bin Then the Likelihood is the product of P ijk in all bins Find the w m that maximize this value 8

19 Putting all together To compare experimental results to theory, we want to obtain the Higgs couplings starting from the measurement of σ and σ x BR Global fit including all the available measurement χ 2 = (Ci - ) 2 where C i depend on the couplings ( e.g. C ZH,H bb = ghzz 2 ghbb 2 ) Γ 2 ΔFi 2 and ΔF i is the experimental precision 9

20 Global Fit 2 possible fit: model-independent and model-dependent coupling relative to SM. ± 5% ± % CLIC preliminary model independent 35 GeV +.4 TeV + 3 TeV coupling relative to SM.5 ± 2.5% ±.5% CLIC preliminary model dependent 35 GeV +.4 TeV + 3 TeV c τ b t W Z g γ Γ H c τ b W Z g.9 Γ H µ.95 µ t γ Model-independent: total width Γ is a free parameter of the fit CAN NOT BE DONE AT LHC! Model-dependent: - assume SM only - Total width described by few parameters that account for deviation with SM 2

21 Conclusions e + e - linear colliders are a key tool to explore the Higgs sector with precision measurements (complementary to LHC) A new generation of detectors with excellent jet energy resolution and tracking is being developed Model independent H to Z couplings can be measured with such machines Unique possibility in e + e - due to clean environment and precise vertexing Higgs hadronic decay is very interesting since H cc and H gg can t be measured at LHC (and H bb is very difficult too) 2

22 BACKUP 22

23 Detectors - Particle Flow algorithms Remaining Track finder PFA PFA tracks algorithm assign removes are Raw ECAL identified assigned Overall data reconstruct from and HCAL as result hits photons the detector charged hits from to or the particles tracks neutral list hadrons tracks 23

24 Linear Collider Design Concepts A lepton collider allows for precision measurements (clean events, well defined initial state) - in the TeV range to complement LHC - linear to prevent synchrotron radiation energy losses Two machine concepts: - ILC: superconductive accelerator technology - ready to build - CLIC: two-beam accelerator for higher energies - still in development source: linearcollider.org source: clic-study.org 24