Current Status on ZHH Analysis ’08 5/31 GLD ZHH-group
Analysis menu M H v.s. (ZHH) 500GeV 750GeV 1TeV ZHH study ZHH analysis was started to cover wide Higgs-mass region. E CM =500GeV is the best for M H =120GeV. E CM >750GeV is preferable to study for M H >160GeV. Our analysis menu Light Higgs : M H =120GeV, E CM =500GeV Heavy Higgs : M H =170GeV, E CM =500GeV~1TeV
Study of Light Higgs Case
Introduction Status at TILC08 Analysis was started for HH -mode. Signal significance of 1.4 was obtained against ZZ bbbb. Signal : 12 events ZZ bbbb : 64 events Today’s topic Re-optimization of the selection cut. Analysis with B.G. of ZZ bbbb and tt. (ZZ) = fb (tt) = fb
Selection cuts The new selection cuts are applied to improve B.G. rejection. Primary cut for reconstructed particles Energy cut (E 1,2 < 250GeV) Momentum cut (P 1,2 < 200GeV) cut ( 1,2 < 0.9) Fine selection cut 2 cut ( 2 < 13) Higgs mass cut (95GeV < M H1,2 < 125GeV) Missing mass cut (90GeV < M miss < 200GeV) Angle cut (|cos 1,2 | < 0.9) b-tag cut New selection cuts Example of new cuts (Momentum and cut) is shown.
Momentum cut Momentum of the reconstructed particles was checked. Higgs-momentum of the signal is below 200 GeV/c. The momentum above 200 GeV/c was rejected. P 1 (GeV/c) HH B.G. P 1 (GeV/c) ZZ tt
cut (=p/E) of the reconstructed particles was checked. of the signal is below 0.9. <0.9 was selected. The fine selection cuts were applied after these primary cuts. 1 1 HH B.G. ZZ tt
B-tag cut After the fine selection cuts, b-tagging was applied. N b-tag ≧ 3 was selected. b-tag: 3 tracks with 3 separation from IP events of ZZ bbbb still contaminates for 2 ab -1. All tt events are rejected. N b-tag HH N b-tag B.G. ZZ tt
Signal significance The signal significance was estimated for 2ab -1 Signal : 12.1 events B.G. : 23.5 events Significance : 2.0 Rec. (M H1 +M H2 ) for 2ab -1 TILC08 Signal : 12 events B.G. : 64 1.4 ZZ tt Signal + BG GeV
Reduction summary No cut E 1,2 < 250GeV P 1,2 < 200GeV 1,2 < 0.9 2 < 13 95GeV < M H1,2 <125GeV 90GeV < M miss < 200GeV |cos 1,2 | < 0.9 N b-tag ≧ 3 : 77.6 : 70.8 (x 0.91) : 70.8 (x 1.00) : 60.9 (x 0.86) : 33.8 (x 0.56) : 26.7 (x 0.79) : 25.5 (x 0.96) : 22.2 (x 0.87) : 12.1 (x 0.55) 18,100 14,022 (x 0.77) 4,876 (x 0.35) 2,033 (x 0.42) 686 (x 0.34) 223 (x 0.33) 130 (x 0.58) 41.6 (x 0.32) 23.5 (x 0.56) HH ZZ bbbb The signal efficiency is 16%. 55% of signal is lost at b-tag cut. However, b-tag is necessary to reject tt-B.G.. 1,167,200 35,950 (x 0.031) 29,063 (x 0.81) 25,678 (x 0.88) 6886 (x 0.27) 1517 (x 0.22) 817 (x 0.54) 467 (x 0.57) 0 (x 0.0) tt Next step is to consider tbtb-B.G. (0.7fb).
Summary for HH The selection cut was re-optimized. The signal significance of 2.0 was obtained against the ZZ bbbb and tt. This result is comparable to Djamel’s report at LCWS07, after scaling (ZZ) to the correct value (395.8fb). The other B.G. modes like tbtb should be taken into account.
Study of Heavy Higgs Case
ZHH signal has small xsec 750 GeV, sensitive to HHH coupling many possible final states with different characteristics: from 10 hadronic jets, to 4 charged leptons + 6 neutrinos, and almost everything in between! Many backgrounds with orders of magnitude higher xsec so far, considered only subset with true Z boson ZWW (this has largest xsec), ZZZ, Ztt, ZZWW events generated by MadGraph, QuickSim level Introduction
Fit events to ZHH->ZWWWW hypothesis many possible signal final states need to be considered look for Z->e+e-, mu+mu- if not found, try Z->tau tau, Z->nu nu and Z->jj possibilities look for additional e, mu, assume from W decay identify tau-like jets # charged tracks = 1 or 3 invariant mass < m_tau jet charge = +-1 try two possibilities: true tau jet, or hadronic jet for other Ws, try both tau decay and hadronic decay possibilities for each possible decay topology force non-leptonic part of event into required # jets ask if invariant masses of Z, W, H->jets consistent with expectation Event reconstruction
require that event fits at least one signal final state hypothesis use kinematic variables to decide if event is real ZHH event or background visible energy, momentum vs. # assumed neutrinos Ycut of jet clustering required to force # jets no single variable gives good separation->combine into Neural Network train network to separate signal from dominant ZWW background try at 750 GeV & 1 Tev – slightly different event topologies Signal/background discrimination
Train neural network to separate signal from ZWW background use TMultiLayerPerceptron package (in root) inputs: #jets, #el+#mu, #taus, Z decay type, Ycut of jet forcing, |total momentum|, Z miss. mass 1 hidden layer, with 5 nodes 5k training events (50% ZHH, 50% ZWW) 5k testing events 2k training cycles reached minimum no significant overtraining
Some NN 1TeV centre-of-mass energy ZHH, ZWW Z decay (invis, lept, jet) # taus # hadronic jets Jet forcing ycut NN output
NN 750 GeV NN 1000 GeV reminder: total # signal events ~100 / ab-1 Both look ~impossible...
ZHH production sensitive to gHHH small xsec, large backgrounds for mH = 170 GeV, many possible final states trained NN to separate ZHH & ZWW GeV higgs GeV & 1 TeV centre-of-mass some separation, but signal still swamped by huge backgrounds maybe this measurement is not possible... Summary for heavy Higgs analysis
Supplement
Energy cut Energy of the reconstructed particles was checked. Higgs-energy of the signal is below 250 GeV. The energy above 250 GeV was selected. E 1 (GeV) HH E 1 (GeV) B.G. ZZ tt
2 cut 2 distribution was checked after momentum and selection. The signal has the peak at 2 =0. B.G. has the broader shape than the signal. 2 <13 was selected. 2 HH 2 B.G. ZZ tt
Mass cut The Higgs mass cut is applied to select well-reconstructed events. Selection: 95GeV < M H < 125GeV M H1 (GeV) HH M H1 (GeV) B.G. ZZ tt
Missing mass cut The missing mass was reconstructed. The peak is observed at Z-mass in HH events. Selection: 90GeV < M miss < 200GeV M miss (GeV) HH M miss (GeV) B.G. ZZ tt
Angle cut The angular distributions of the reconstructed particles are checked. Selection: |cos |<0.9 cos HH cos B.G. ZZ tt