What TeV can do for Higgs at LHC?

Transcription

1 A. Nikitenko,, Imperial College What TeV can do for Higgs at LHC? TeV4LHC Workshop, FNAL, 16 Sept Apologize that I will emphasis on CMS Higgs studies. Capabilities of Higgs searches in ATLAS and CMS are very similar

2 The best would be : discover Higgs before LHC! but...

3 Prospects for SM Higgs integrated luminosity (fb -1 ) Tevatron SM Higgs at 120 GeV/c 2 - exclude at 95 % C.L. in σ evidence in 2009 B. Heinemann talk on UK Forum, April 2004 Higgs Sensitivity Study ('03) SUSY/Higgs Workshop ('98-'99) σ discovery 3σ evidence 95% CL exclusion m H (GeV) Statistical Significance LHC; one experiment (CMS example) 10 fb -1 : 5 σ discovery combining all channels for M H > 114 GeV; H γγ, inclusive, NLO H + jet, H γγ, NLO ± ± WH, H γγ, W l + ν l H ZZ* l l l l, NLO H WW* llνν, NLO qqh, H γγ qqh, H τ + τ - lepton + τ jet tth, H bb WH, H bb Total significance -1 CMS, 30 fb -1 5σ at 10 fb -1 5σ at 30 fb -1 5σ at 60 fb m H (GeV/c )

4 The whole mass range for SM Higgs discovery at LHC Statistical Significance CMS, 30 fb qqh, H WW lν l jj qqh, H ZZ llν l ν l H WW*/WW llν l ν l, NLO H ZZ*/ZZ l l l l, NLO qqh, H γγ,τ + τ - H γγ inclusive, NLO tth,wh,h bb Total significance -1 5σ at 2 fb -1 5σ at 10 fb 5-1 5σ at 30 fb m H (GeV/c ) m H > GeV

5 Tevatron data and experience are invaluable for success of Higgs searches at LHC! Physics environment at the LHC is very similar to that at the Tevatron. For LHC Higgs physics it is very important what Tevatron is doing with: Tuning of min-bias and multiple interaction models with TeV data; uncertainties Understanding of reliability and limitations of MC generators; uncertainties Test of theoretical N LO calculations; uncertainties. Experimental methods and techniques for measurement of background from the data measurement of b, τ tagging efficiency from the data measurement of jet -> e, γ, τ, b, c miss id efficiency from the data jets and missing E T measurement; reconstruction, calibration Understanding of the signal and background systematic; theory + experiment I will go through these points giving some examples...

6 Tuning of min-bias and underlying event models; propogate to LHC Outgoing Parton Proton Hard Scattering Outgoing Parton PT(hard) Proton AntiProton Underlying Event Underlying Event Underlying Event Final-State Radiation Outgoing Parton Initial-State Radiation PT(hard) Initial-State Radiation AntiProton Underlying Event R. Field Outgoing Parton Final-State Radiation Pile up and underlying events affect : - isolation of γ, τ, e, µ - jet energy reconstruction ( pedestal ) - jet veto - forward jet tagging in VBF Higgs Very important to understand uncertainties if efficiency can not be evaluated directly from the data

7 Current PYTHIA tunings (used in CMS production) Comments CDF Tune A (PYTHIA6.206) PYTHIA6.214 Tuned (ATLAS) R. Field; CDF UE tuning method Generated processes (QCD + low-pt) 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) Multiple interactions models MSTP(81) = 1 MSTP(82) = 4 MSTP(81) = 1 MSTP(82) = 4 pt min 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 "Transverse" Charged Density "Transverse" Charged Particle Density: dn/dηdφ CDF Preliminary data uncorrected theory corrected PT(charged jet#1) (GeV/c) CDF Run 1 Published CDF Run 2 Preliminary PYTHIA Tune A η <1.0 PT>0.5 GeV/c Core radius Gluon production mechanism α s and K-factors Regulating initial state radiation 40% of the hadron radius (PARP(84) = 0.4) PARP(85) = 0.9 PARP(86) = 0.95 MSTP(2) = 1 MSTP(33) = 0 PARP(67) = 4 50% of the hadron radius (PARP(84) = 0.5) PARP(85) = 0.33 PARP(86) = 0.66 MSTP(2) = 1 MSTP(33) = 0 PARP(67) = 1

8 LHC predictions: PYTHIA6.214 (ATLAS tuning) vs. CDF tuning; different predictions! Transverse < N chg > PYTHIA tuned CDF tuning CDF data dn UE ch /dη ~ 30; min-bias ~7 NEW Multiple Interaction Model In PYTHIA 6.3! T. Sjostrand and P.Z. Skands hep-ph/ dn UE ch /dη ~ 20; min-bias ~ 6 NEW JIMMY JIMMY4.01 & HERWIG dn UE ch /dη ~ 10; min-bias ~ 4 Talk of J. Butterworth & M. Seymor HERA-LHC Workshop TUNNING P t (leading jet in GeV)

9 Effect of underlying event on central jet veto in VBF Higgs H->WW* >WW*->2l>2l in qqh prod. Uncertainty of the central jet veto efficiency due to UE model; ATLAS. Rapidity of the central jet in Higgs events; CMS; full simulation, L=2x10 33 cm -2 s -1 nev/ Pythia ATLFAST η mini j (η j1 +η j2 bkg. like behaviour for soft jets; fake jets: pile up+ue+detector

10 E T (GeV) Can we avoid pile up effect on the jet energy reconstruction? Performance of CMS pile up subtraction algorithm (on event e by event basis) for jet energy reconstruction at high (L=10 34 cm - 2 s - 2 ) luminosity. Reconstructed energy of 50 GeV jets as a function of number of min-bias interactions / 17 A A E-01 E T (GeV) / 17 A A E E Iterative cone algo; R=0.5 before pile up subtraction number of interactions 20 0 after pile up subtraction number of interactions

11 W/Z+nJets is very important background for Higgs at LHC topology W+1j+X W+2j+X W+3j+X W+4j+X Z+1j+X Z+2j+X Z+4j+X Background for Higgs channel (one example) gg->ww* >WW*->2l >2l (?) MSSM gg->bbh bbh,, H->ττH ττ->l+jet (one b-tag) b VBF qq->qqh qqh,, h->ττh ττ->l+jet + 2 tag. jets VBF qq->qqh qqh,, h->wwh >WW->lνjj + 2 tag jets MSSM gg->bbh bbh,, H->ττH ττ->l+jet (one b-tag) b VBF qq->qqh qqh,, h->ττh ττ->l+jet + 2 tag jets VBF qq->qqh qqh,, h->zzh >ZZ->lljj + 2 tag jets Zbb,Zcc, Wbb, Wcc (W/Z+QQ+nj Z+QQ+nj) ) are as important as W/Z+nj

12 W/Z rates at LHC Z, W, tt cross sections and expected number of events after trigger in CMS with 10 fb -1 Very important to understand Z+nj, W+nj, tt~ ~ as background for Higgs (and SUSY) searches J. Campbell, R.K. Ellis, D. Raiwater hep-ph/ W/Z+nJ+X NLO predictions at LHC with cuts : p Tl > 15 GeV ηl < 2.4 p Tj > 20 GeV η j < 4.5 Rlj > 0.4 Rll > 0.2 W/Z bb + X η b < 2.5

13 ME+PS Monte-Carlo verification with Tevatron date is crucial - improve estimates of the background for Higgs studies - test of theoretical calculations and MC generators : ME + PS with HERWIG and PYTHIA (CKKW & MLM approaches) S. Mrenna and P. Richardson hep-ph/ ; LesHouches QCD group report, hep-ph/ ALPGEN + HERWIG works well...

14 Example: qq->qqh, h->ww->l jj+2tag jets CMS; very preliminary, Jacopo Bernardini M higgs = 300 GeV Number of events (30 fb -1 ) Signal = 144 tt = 66 W+4jets ALPGEN = 442; σ=12 (stat) W+4jets PYTHIA = 79; σ=6.4 (stat); with 5% bkg. unc. σ=4.3; σ=n S /sqrt(n B ) S+B B 300 GeV CMS home work (A.N.): W+3jets with PYTHIA MadGrapg ALPGEN p T jet, GeV/c p T jet, GeV

15 MCatNLO S. Frixione and B. Webber JHEP 0206 (2002) 029, hep-ph/ S. Frixione, P. Nason and B. Webber JHEP 0308 (2003) 07 [hep-ph/ ] B. Webber talk on UK HEP Forum. April 2004

16 MCatNLO: is it ready for H->WW->2l analysis? pp->ww->µ + µ - at LHC; φ ll Important to include VV spin correlations in MCatNLO V. Drollinger comparison (CMS)

17 MCatNLO: : pp->tt tt~ ~ at LHC Plots from S. Paganis talk on MC at LHC Workshop, CERN 2003 PYTHIA vs HERWIG vs MCatNLO comparison for LHC Some points on top background for h->wwh >WW->analysis >analysis at LHC : - tt~ spin correlations are not yet in MC@NLO; ~ 7% effect in h->ww->2l analysis - both on-shell and off-shell contributions to top production are important after jet veto - σ NWA (tt~) + σ NWA (Wtb) after cuts leads to large double counting N.Kauer and D. Zeppenfeld arxiv:hep-ph/

18 In the next slides I will go through a few selected analyses and a few selected points (biased view) where TeV can help (or already helped!) - inclusive h->2h >2γ - inclusive h->wwh >WW->2l>2l - tth,, h->bbh - MSSM ggh,, H->bbH - MSSM ggh,, H->ττH

19 H->WW->2l >2l analysis at TeV and LHC (I) Tevatron data and MC (PYTHIA) LHC (CMS) Monte Carlo (PYTHIA) M H =160 GeV M. Zanetti; full simulations, preliminary Very similar event selections: - cuts on lepton p T - cut on miss E T, Z resonance veto - jet veto against tt~ - φ(ll) cut is particularly important; exploit spin correlations

20 Observed Expected Tevatron results Number of events after selections ee 2 2.7±0.4 eµ 2 3.1±0.3 µµ 5 5.3± ev Higgs Expected in SM Dominant bkg. in eµ sample WW W+jets WZ tt 2.51± ± ± ±0.01 LHC results (tab. from old M. Higgs 879 (tab. from old M. Dittmar,, H. Dreiner analysis; 30 fb - 1 ) WW 376 tt~ 64 Wtb 146 From ATLAS analysis of K. From ATLAS analysis of K. Jakobs and T. and T. Trefzger W+jets / WW < 2 %! WZ+ZZ / WW = 2 %

21 W+Jet background in H->WWH >WW->2l>2l Ratio of W+jets and WW backgrounds in Tevatron analysis is much bigger than in LHC analysis (CMS did not take into account W+jet) It can not be explained by difference in cross sections at TeV and LHC : σ, pb W WW WW / W LHC 1.65 x x 10-4 TeV 1.80 x x 10-4 Calculated by E. Boos, CompHEP (LO); Q 2 = M W2, CTEQ6l1 LHC should check W+jets bkg.. with realistic simulation of jet->e miss id.

22 Discovery reaches with H->WWH >WW->2l>2l Excluded cross section times Branching Ratio at 95% C.L. +/- 5 % bkg.. systematic were taken both in ATLAS and CMS; need more justification; learn systematic from Tevatron analysis

23 Possible way to estimate WbWb background for h->wwh >WW->2l >2l at LHC:

24 H->WW->2l: >2l: from discovery to cross section measurement at LHC For cross section measurement signal systematic becomes as important as background one. Jet veto efficiency at generator level Monte Carlo systematic may be significant due to Jet Veto Effect of UE model was shown already. This plot shows efficiency of Jet Veto as a function of Higgs p T for different generators WITHOUT multiple interactions. Uncertainty is ~ 10 % Giovanna Davaz, CMS

25 H->2γ: how TeV can help? CMS plots. K factors included 15 Events/500 MeV for 100 fb Events/500 MeV for 100 fb N S / N B fb 1 (low luminosity) a) m γγ (GeV) b) m γγ (GeV) fb 1 (high luminosity) m γγ (GeV) Background under mass peak will be obtained from the data, but background predictions at LHC can be evaluated today from comparison and tuning of NLO Monte Carlo with TeV data

26 Mγγ at Tevatron: : data comparison with PYTHIA and DIPHOX In CMS we use K factors obtained from comparison of PYTHIA with DIPHOX after experimental selections for different backgrounds: q γ q γ γ q γ g q g γ pp->γ π 0 pp-> π 0 π 0 g Box: K from Dixon et al. γ

27 Composition of the background for LHC. CMS case: full simulation study by S. Shevchenko, T. Lee, V. Litvin, H. Newman (preliminary). PYTHIA K factors from: T. Binoth et al., Les Houches 2001; hep/ph T. Binoth, K. Lassila-Perini (CMS), Les Houches 2003; hep-ph/ Z. Bern, L. Dixon, C. Schmidt, Phys. Rev. D 66 (2002) pp->jj

28 Background from side bands. But how to compare with analytical NLO?

29 NLO γγ+jet background for h->γγ + jet topology using smooth isolation V.Del Duca et al, hep-ph/ Would be problematic to compare directly with the background data, since at Level 1 and High Level trigger for 2γ stream the different isolation criteria has been already optimized in CMS using PYTHIA as bkg. (and signal) generator

30 Photon Fake Rate from data Rate of jets with leading meson (pi0, eta) which cannot be distinguished from prompt photons: Depends on detector capabilities, e.g. granularity of calorimeter Cuts! Systematic error about 30-80% depending on Et Data higher than Pythia and Herwig Pythia describes data better than Herwig B. Heinemann. UK Forume, April 2004 CDF (preliminary result) At TeV Jet->γ miss ID is obtained from γ+jet data. We should evaluate how does it work with LHC detectors

31 Difficult channel : tth, h->bb NLO tth from M. Spira et al., hep-ph/ K=1 σ ttbb [pb] ttbb (and ttjj) predictions at LO has very big scale uncertainties ~ factor 2. V. Drollinger, Les Houches 2003; ALPGEN Q 2 =m t2, CTQ5L, p T (b)>25 GeV, η < 2.4, R(bb) > 0.4 probability / 10 GeV 0.12 Q 2 = ξ 2 *m 2 t 0.1 ξ = 0.5/1.0/ ξ m(b,b) [GeV] Backgrounds: ttbb shape is not affected by scale change, BUT ttb, ttjj, Ztt from LO CompHEP additional jets (at NLO) can give different ttbb is dominant after selections combinatorics which could change the shape NLO predictions for ttbb and ttj(j) ) is very desirable; NLO ttj(j) ) can be verified by Tevatron data

32 tt~ + njet Production Rates at Tevatron (Elizabeth Graves, DO; talk on CMS Higgs meeting) Alpgen production results Cross-section (pb) t tbar + 0 jets 1 jet 2 jets TeV Expected number of events in W lν channel Tevatron: L=5fb -1 Number of produced events t tbar + 0 jets 1 jet 2 jets TeV Take care : qq->tt dominates at Tevatron, gg->tt dominates at LHC same FSR, but different ISR

33 D0: MSSM bbh,, H->bb H at high tanβ Event Selection: At least 3 jets: E T cuts on jets optimized for different Higgs mass values 3 b-tagged jets Look for signal in the invariant mass spectrum from the two leading b- jets Main Background: QCD multi b-production Difficult for LO MC: determined from data and/or ALPGEN 1.2 Signal acceptance about % depending on Mass From Beate Heinemann talk at UK Forum, April 2004 Ldt=131 pb -1

34 CMS: MSSM bbh,, H->bb H at high tanβ Level 1 multi-jet trigger : 1J or 3J or 4J ; thresholds 177, 86, 70 (95% eff) ) => 3 khz HLT single b tagging for next-to-leading jet E T > 160 GeV => 5 Hz Off-line selections are similar to D0 : two hard jets (E T > 220 GeV for M H =600 GeV) two soft jets E T > 20 GeV PYTHIA simulations >= 3 b tagged jets 2 event count / 25 GeV/c CMS 60 fb 1 µ = M 2 = M SUSY = 1 TeV/c2 X t = 6 MSUSY m = 600 GeV/c 2, tan = 50 A β g g b b A/H, A/H b b S/B = 5.6% (in mass window) β tan CMS 60 fb 1 µ = M = M SUSY = 1 TeV/c X = 6 M SUSY t 2 g g b b A/H, A/H b b 2 5σ 4σ 3σ 2σ σ Statistical significance contours without systematical errors M (GeV/c 2 ) bb m (GeV/c ) A Common question : how to evaluate background shape?

35 learning D0 way From the double b-tagged b data to triple b-tagged b data background The shape of the triple b-tagged was estimated from double b-tagged data and extrapolated using a tag-rate-function derived on the multi-jet data sample. This background was then normalized to the triple b-tagged data outside 1 σ signal mass window from the D0 Higgs results page Can it be applied at LHC? Background composition should be different at LHC - triple b-tagged background with two of three real b jets is dominant (~ 72%) - the main contribution come from gg->gg, gb->gb with g->bb~

36 CMS results on bbh,, H->bbH Higgs mass after bkg.. subtraction The 2 σ discovery reach with different with known bkg.. shape assumptions on the level of systematic 2 event count / 25 GeV/c tanβ=50 CMS 60 fb 1 µ = M 2 = M SUSY = 1 TeV/c2 X t = 6 MSUSY m = 600 GeV/c 2, tan β = 50 A g g b b A/H, A/H b b After background subtraction S/B = 5.6% (in mass window) β tan CMS 60 fb 1 µ = M = M = 1 TeV/c 2 SUSY X = 6 M SUSY t 2 2.0% 1.5% 1% 0.5% no systematics g g b b A/H, A/H b b σ significance contours with systematical errors M (GeV/c 2 ) bb m (GeV/c ) A

37 MSSM bbh,, H->2H >2τ.. What we should learn from TeV Cross section exhibits a large sensitivity to tan(β) ) and thus can add a significant observable to a global fit of the SUSY parameters rs R. Kinnunen, S. Lehti, F. Moortgat, A. Nikitenko, M. Spira. hep-ph/ Uncertainty is NLO calculations ~ 20 % for 1b tag [S. Willenbrock, M. Spira et al.] is bigger than stat. uncertainty. tanβ Small error bars: stat errors Large error bars: total uncertainty A t 2 2 = 2450 GeV/c, µ = 300 GeV/c 2 2 M 2 = 200 GeV/c, m SUSY = 1 TeV/c L/L = 5% σ/σ = 20% BR/BR = 3% However systematic due to event selections in this analysis: τ tagging b tagging (1 b tag) jet veto (2 nd b veto) calo scale should be more understood learning how Tevatron estimates uncertainties from the data CMS, 30 fb H SUSY ττ eµ/lj/jj m A (GeV/c )

38 Uncertainties involved in the tan(β) measurement At large tan(β), σ x Br ~ tan 2 (β) eff f(m A ) at fixed µ, M 2, A t, M SUSY N S = tan 2 (β) eff f(m A ) L ε sel tan(β) = tan(β) mes +/- stat +/- syst +/- MCgen syst = 0.5 ( L + σ th + Br th + σ( M H ) + ε sel + B) σ th = 20 % due to NLO scale dependence Br th = 3 % uncertainties of SM input parameters L = 5 % luminosity uncertainty σ( M H ) = % due to mass measurement at 5σ discovery limit B = N B / N S = 10 % at 5σ discovery limit (preliminary) ε sel = ε calo + ε b tag + ε τ tag ε b tag = 2.0 % (prelim.) THIS HAS TO BE MORE ε τ tag = 2.5 % (prelim.) JUSTIFIED EXPLOITING ε calo = 2.9 % (prelim.) TEVATRON EXPERIENCE

39 Exploiting TeV Z τ+τ and W->τν -1 for 30 fb 2 Events/20 GeV/c Signal Background 0 H SUSY CMS ττ l+jet (+X) 2 m A = 200 GeV/c tanβ = 20 Gauss+2 Landau Entries 272 Mean RMS χ / ndf / 17 G:Const ± 16.3 G:Mean ± 3.26 G:Sigma 30.9 ± 3.44 L1:Const ± 14.2 L2:Const 81.9 ± (GeV/c ) m ττ - How TeV evaluates τ id efficiency from the data? - We even did not think about Z->eeZ background. Need to check!

40 Z+b at TeV as benchmark for gb->bh (gg->bbh) Z+b can be used as a benchmark for gb->hb at LHC: test N(N)LO predictions and Monte Carlo. 184 pb -1 for e + e pb -1 for µ + µ - However, be careful: at Teatron both contributions gb->zb and qq~->zbb are important while only gb->zb is dominant at LHC and thus relevant to gb->hb [J. Campbell et all hep-ph/ ] N(N)LO calculations are available for bb->h, gb->hb and gg->bbh and compared in J. Campbell et al, arxiv:hep-ph/ Comparison of p Tb between PYTHIA and NLO gb->hb, gg->bbh was presented in A.N. talk on HERA-LHC Workshop meeting 27 March, 2004

41 VBF Higgs channels (qq->qqh) Signal significance 10 2 L dt = 10 fb -1 (no K-factors) ATLAS qqh qq WW (*) qqh qq ττ VBF, combined VBF, + γγ + tth(bb) + ZZ * Very important for Higgs early discovery at LHC and coupling measurement 10 How TeV can help HLC VBF program, see talk of Dieter Zeppenfeld m H (GeV/c 2 ) dσ/dm T (fb/10 GeV/c 2 ) H->WW->2l ATLAS nev/5 GeV for L=30 fb CMS evts / 5 GeV m H =120 GeV Z jj t t, WW EW ATLAS M T (GeV/c 2 ) M ττ, GeV m ττ (GeV)

42 Higgs cross section: pdf uncertainties see talk of Albert de Roeck Djouadi and Ferrag,, hep-ph/ ph/

43 THE END

44

45 Higgs cross section: dependence on pdfs Djouadi & Ferrag, hep-ph/

46 Djouadi & Ferrag, hep-ph/

47 the differences between pdf sets needs to be better understood! Djouadi & Ferrag, hep-ph/

48 tanβ LEP GeV Preliminary m h -max tanβ m h -max 10 M SUSY =1 TeV M 2 =200 GeV µ=-200 GeV m gluino =800 GeV Stop mix: X t =2M SUSY 10 1 Excluded by LEP 1 Excluded by LEP m A (GeV/c 2 ) m A (GeV/c 2 )

49 MSSM Higgs boson channels important at low tan(β) could be also included : S. Heinemeyer, G. Weiglein, 04 No tan(β) exclusion for m t ->m t +σ mt M SUSY =1 TeV->2 TeV => Low tan(β) not fully excluded by LEP! We should not forget low tan(β) channels : tanβ 10 old LEP upper bound (1999) max m h scen., FH1.0 max m h scen., FH1.3 max m h, FH2.1, mt = 178 GeV max m h, FH2.1, MSUSY = 2 TeV m t = ( ) GeV SM exclusion bound A->Zh (Z->ll, h->bb) A->tt A->2γ H->hh->2γ 2b m h [GeV]

50 MCatNLO B. Webber talk on UK HEP Forum. April 2004 pp->zz, PYTHIA vs MC&NLO at LHC; M. Sani (CMS) p T (ZZ)

51 Current tt~ ~ generation project for H->WWH >WW->2l >2l in CMS Higgs group (I) is coming

52 Current tt~ ~ generation project for H->WWH >WW->2l >2l in CMS Higgs group (II) A.N. Efficiencies of φ(ll) cut for MadGraph llννbb and MadGraph WbWb (W->lν in PYTHIA) are different by ~ 7 %. => effect of the spin correlations