1. Searches for the Standard Model Higgs Boson at the Tevatron Gavin Hesketh Northeastern University On behalf of the CDF and D ⊘ Collaborations Hadron Collider Physics Symposium 2007 La Biodola, Isola d'Elba, Italy. 1

2. Overview What we know about the Higgs - direct searches - indirect constraints - search channels The Analyses: - high mass Higgs search - H -> WW - low mass Higgs searches - ZH -> ll bb - ZH -> bb - WH -> l bb Preliminary results with 1 fb -1 - both experiments have now recorded over 2.4 fb -1 Prospects & Conclusions This talk: ~1fb -1 2 G. Hesketh

3. The Standard Model Higgs Boson The Higgs mechanism: - additional scalar field to the SM lagrangian - introduces masses for the W and Z, and leptons - predicts neutral, spin 0 boson - but not its mass Direct searches for ZH production at LEPII: - Higgs mass > 114.4 GeV @95 % CL Indirect constraints: Including the latest Tevatron top and W mass: - see talks by M. Begel, W. Trischuk - best fit Higgs mass = 76 +33 -24 GeV - mass < 182 GeV @ 95 % CL (including LEPII limit) 3 G. Hesketh

4. Excluded m H (GeV) The Higgs at the Tevatron Production cross sections at Tevatron: - 0.1 - 1 pb - compare to: - WZ ~4 pb (5 sigma at CDF, see A. Askew) - single top ~3 pb (3 sigma at DZero, see M. Datta) - ZZ ~2 pb (3 sigma at CDF, see A. Askew) Branching fractions dictate search: - at low mass, H -> bb - gg->H overwhelmed by background - look for associated production - backgrounds: W/Z+jets, top, WW, WZ - at higher mass, H -> WW - gg -> H -> WW - main background: S.M. WW Aim: - isolate signal as much as possible - find variable with good S/B - set limits using this variable 4 G. Hesketh

5. Main search channel for mass > 135 GeV Study ee, mumu and emu channels Signature: - opposite sign leptons (10 - 20 GeV) - missing energy DZero preliminary, 930 pb -1 Main backgrounds are: - Drell-Yan - QCD: semi-leptonic decays jets faking electrons - Diboson (WW, WZ, ZZ) - top gg-> H ->WW* 5 G. Hesketh

6. gg-> H ->WW* DZero preliminary, 930 pb -1 Cut-based analysis: DZero (ICHEP '06), CDF (Moriond EW '07), Drell-Yan background: - low missing energy – require 20-25 GeV - DZero remove Z mass region - CDF require mass – 5 GeV < Higgs mass / 2 QCD background: - isolation, require mass > 16, 15 GeV top background: - cut on number & energy of jets Standard Model WW background dominates - After all cuts, delta phi (l,l) is best discriminant - spin 0 Higgs tends to produce colinear leptons WW 6 G. Hesketh

7. CDF result (Moriond QCD '07) - optimize lepton identification based on WZ analysis - add looser lepton classes - add isolated track (can be e or mu) - gain ~1.6 x data set! - split sample into high and low signal / background Matrix Element approach - take observed leptons and missing energy (x obs ) - integrate over leading order theory predictions for: WW, ZZ, W+gamma, W+jet, 10 Higgs masses - construct discriminant from probabilities: Validate method with LR for background Limit set by fitting LR distribution gg-> H ->WW* 7 G. Hesketh

8. gg-> H ->WW* Limit Setting: - use CLs prescription - set 95 % C.L. exclusion limits For a Higgs mass of 160 GeV: CDF result, Matrix Element: - obs. (exp.): 3.4 (4.8) times SM cut based: - obs. (exp.): 9.2 (6.0) times SM Cut based DZero result, - obs. (exp.): 3.7 (4.2) times SM Fourth generation model excluded in mass range 150-185 GeV Systematics dominated by: - lepton ID and acceptance - theory uncertainty on cross sections 8 G. Hesketh

9. Identifying B Jets CDF: secondary vertex reconstruction Neural net used to improve purity - based on track multiplicity, pT vertex decay length, mass, fit - loose = 50 % efficiency, 1.5 % mistag - tight = 40 % efficiency, 0.5 % mistag DZero uses Neural Net tagger: - combination of secondary vertex & dca - based on same basic variables - high efficiency, purity - 70 %, 4.5 % (loose) - 50 %, 0.3 % (tight) Low mass higgs search: H->bb - Identifying b-jets is essential! - S/B improves ~1:1000 -> 1:100 - Requiring two b-jets: S/B ~ 1:50 9 G. Hesketh

10. Cleanest low mass Higgs signal, but low cross section. - use loose lepton ID (e and mu) - Require two jets: - pT > 15 GeV (DZero); >25,15 GeV (CDF) Main backgrounds: - Z+jets, QCD, top, ZZ & WZ ZH -> llbb Two different approaches: - Require two b-tags, fit the di-jet mass (DZero) - Require 1 tight or two loose b-tags, apply 2-D neural network (CDF) 10 G. Hesketh

11. ZH -> llbb DZero analysis (Nov. '06) Limit extracted by fitting the dijet mass - mass resolution major factor in sensitivity Main systemastics: - b-tagging, jet resolution, lepton ID, b.g. cross sections For Higgs mass = 115 GeV, 95 % CL limit: - expected 22 x SM - observed 23 x SM CDF approach: adjust the jets to balancing the MET Improves the di-jet mass resolution: - 16 % to 10 % 11 G. Hesketh

12. ZH -> llbb CDF analysis (Moriond QCD '07) - split sample into 1 and 2 b-tags, - use optimised mass fit - use 2-D neural network: - separate ZH from tt and ZH from Z+jets - equivalent to ~2.5 times the data set! For Higgs mass = 115 GeV, 95 % CL limit: - expected < 16 x SM - observed < 16 x SM S/B = 1/4 12 G. Hesketh

13. ZH -> bb Compared to Z->ee, mumu: - higher branching fraction - larger acceptance Contribution from WH->lbb - lepton not reconstructed No measured leptons -> more challenging! Trigger on two jets + missing transverse energy. Require:two high energy jets: >20 GeV (DZero); >60, 20 GeV (CDF) high missing energy: >50 GeV (DZero); >75 GeV (CDF) veto on isolated leptons, high jet energy (to reduce top background) one or two b-tags (CDF), one tight + one loose b-tag (DZero) Backgrounds: Z + jets, W + jets, top – measured with Monte Carlo. Challenge is to understand instrumental background: - mis-measured missing energy 13 G. Hesketh

14. ZH -> bb New DZero Result: Define two missing energy variables: - measured with jets (MHT), calorimeter cells (MET) - asymmetry isolates mis-measured jets. Limit set by fitting di-jet mass Jet2 ME T Jet1 14 G. Hesketh

15. ZH -> bb CDF (ICHEP '06): reduce instrumental background using - angle between jets and missing energy - ratio of missing energy to jet energy Extract limit from di-jet mass fit. CDF and Dzero set limits for: - ZH -> bb - WH -> l bb (unreconstructed lepton) CDF: ratio to S.M. cross section DZero: cross section 15 G. Hesketh

16. Higher cross section than ZH -> highest of any low mass Higgs search Use electron and muon W decays Basic event selection: - isolated lepton, pT > 20 GeV - missing energy > 20 GeV - two jets, pT>15 GeV (CDF); > 20 GeV (DZero) Backgrounds: - W+jets, QCD, top, di-boson Both experiments have a cut-based anaysis DZero also has a Matrix Element result. WH -> lbb 16 G. Hesketh

17. WH -> lbb CDF Result (ICHEP '06) - Require 2 b-tags, or 1 + Neural Net - rejects mistags and charm - treat two samples separately, combine results - Extract limit from fit to di-jet mass For Higgs mass = 115 GeV, 95 % CL limit: - 17 x SM (expected) - 26 x SM (observed) Single tagDouble tag 17 G. Hesketh

18. WH -> lbb DZero Result (Moriond EW '07) - OR all triggers for muon mode: - gain ~50 % more muon channel signal! - High efficiency (70 %) using NN b-tagger - Again, treat 1 and 2 b-tags separately For Higgs mass = 115 GeV, 95 % CL limit: - 8.8 x SM (expected) - 11 x SM (observed) 18 G. Hesketh

19. WH -> lbb DZero Result Using Matrix Element (Moriond '07) Derive limit from discriminant fit. Optimised for single top analysis: - currently 10 % worse limit than cut-based - re-optimise for Higgs, use full muon acceptance -> expect 20 - 30 % better limit Double Tag Single Tag 19 G. Hesketh

20. CDF limit (1fb -1 ) 95 % C.L. (x SM) observed (expected) D0 limit (1fb -1 ) 95 % C.L. (x SM) observed (expected) Analysis Z/WH MET +bb @ 115 Technique: M jj 16 (15)14 (9.6) WHlbb @ 115 Technique: M jj Technique: ME 26 (17) - 11 (8.8) 12 (9.5) ZHllbb @ 115 Technique: M jj Technique: NN2D - 16 (16) 23 (22) - HWWll @ 160 Technique: (l,l) Technique: ME 9.2 (6.0) 3.4 (4.8) 3.7 (4.2) - Results Summary 20 G. Hesketh

21. Combinations Tevatron combination in Summer '06. Been working hard since then! - improving sensitivity in all channels - looser lepton ID - trigger redundancy - di-jet mass resolution - Advanced analysis techniques: - neural nets - b-tagging - signal – background separation - matrix element methods - accelerator delivered > 2x luminosity Many of these results are new - combinations ongoing. - DZero alone now has tighter limits: - ratios to SM, obs (exp): - 8.4 (5.9) @ 115 GeV - 3.7 (4.2) @ 160 GeV 21 G. Hesketh

22. Prospects More improvements to come: - updated Tevatron combination - analysing larger datasets - wider use of advanced analysis techniques - more channels: WH -> WWW, H->ZZ, hadronic H->WW, tau modes,... - new Layer 0 at DZero: improves b-tagging Based on DZero's current limits, what could we achieve? -> need ~3 fb -1 to reach 95 % C.L. exclusion 22 G. Hesketh

23. Conclusions Gavin Hesketh, Northeastern University We are passing milestones: - WZ observation (CDF) - single top evidence (DZero) - ZZ evidence (CDF) Is the Higgs next? Analyses presented today all used ~ 1fb -1 - current best expected limits (x S.M. cross section): - 4.2 @ 160 GeV (DZero H->WW) - 5.9 @ 115 GeV (DZero combined) Prospects: - updated CDF + DZero combination soon - accelerator continues to deliver - results are scaling better than luminosity - most sensitive points are 115 and 160 GeV - 95 % C.L. exclusions with ~3 fb -1 - 3 sigma possible with 8 fb -1 - clearer picture after next round of analyses. A very exciting time to be working at the Tevatron! Delivered 23 G. Hesketh

24. Backup slides

25. data deficit data excess

26. Setting Limits The CLs prescription: Use a Log-Likelihood Ratio test statistic: LLR = -2 log Q d is the “data” for model being tested Distributions for backgrounds populated by Poisson trials, mean given by B or S+B hypothesis - systematics folded in Integrate over profile likelihood: - fit to data for each outcome - cross sections allowed to float within uncertainty