1. Lighting up the Higgs sector with photons at CDF Baylor HEP Seminar 1 Karen Bland November 12, 2011

2. Outline Introduction Tevatron and CDF Detector Photon ID and Efficiency SM H  γγ Search Fermiophobic h  γγ Search Summary and Conclusions 2

3. Outline Introduction – Theoretical Overview – Motivation Tevatron and CDF Detector Photon ID and Efficiency SM H  γγ Search Fermiophobic h  γγ Search Summary and Conclusions 3

4. The Standard Model Higgs boson is only SM particle that hasn’t been observed! Through the Higgs mechanism: (1) Electroweak symmetry is broken (2) Other SM particles acquire mass But mass of Higgs boson a free parameter… Has to be determined experimentally if exists What we know so far: Lower mass SM Higgs boson mass preferred by electroweak constraints Search focus: m H =114 – 145 GeV SM Higgs boson not to be discovered at Tevatron However through spring 2012 Tevatron, more sensitive in much of search region than LHC experiments 4

5. The Standard Model Higgs boson is only SM particle that hasn’t been observed! Through the Higgs mechanism: (1) Electroweak symmetry is broken (2) Other SM particles acquire mass But mass of Higgs boson a free parameter… Has to be determined experimentally if exists What we know so far: Lower mass SM Higgs boson mass preferred by electroweak constraints Search focus: m H =114 – 145 GeV SM Higgs boson not to be discovered at Tevatron However through spring 2012 Tevatron, more sensitive in much of search region than LHC experiments 5 Tevatron experiments are still working very hard to improve analysis techniques and add final datasets to fill in as much of the remaining gaps as possible. The H  γγ analysis contributes to this effort Tevatron experiments are still working very hard to improve analysis techniques and add final datasets to fill in as much of the remaining gaps as possible. The H  γγ analysis contributes to this effort H  γγ contributes sensitivity here

6. SM Higgs Production at the Tevatron 6 Gluon Fusion Associated Production Vector Boson Fusion gg  H is largest cross section Excluded from channels where Higgs decays to quarks due to multijet backgrounds (like H  bb) gg  H is largest cross section Excluded from channels where Higgs decays to quarks due to multijet backgrounds (like H  bb)

7. SM Higgs Production at the Tevatron Produced only rarely: ◦ One out of every 10 12 collisions ◦ That’s about 2 Higgs bosons produced each week 7 Gluon Fusion ~ 1000 fb @ 120 GeV Associated Production ~ 225 fb @ 120 GeV Vector Boson Fusion ~ 70 fb @ 120 GeV gg  H is largest cross section Excluded from channels where Higgs decays to quarks due to multijet backgrounds (like H  bb) H  γγ gains by using all three production methods (~1300 fb) gg  H is largest cross section Excluded from channels where Higgs decays to quarks due to multijet backgrounds (like H  bb) H  γγ gains by using all three production methods (~1300 fb)

8. SM H  γγ Decay Dominant low mass decay mode is H  bb H  γγ Br < 0.25% 8 Signal expectation @ 120 GeV: N = σ × L × Br = 1300fb × 7.0fb -1 × 0.002 ~ 18 H  γγ events produced (~ 6 reconstructed)

9. Small Br, however contributes sensitivity to Tevatron search in difficult region ~125 GeV: Many beyond SM scenarios include a larger Br(H  γγ) New results for one such scenario shown later in the talk 9 Is a H  γγ search interesting at the Tevatron?

10. Clean signature compared to H  bb – Photons (or electrons from photon conversions) easier to identify/reconstruct than b-jets – Larger fraction of H  γγ events accepted in comparison – Total acceptance: ~35% accepted for gg  H ~30% accepted for VH and VBF – Largest efficiency losses from fiducial requirements and ID efficiency – Also improves reconstructed mass resolution… 10 Is a H  γγ search interesting at the Tevatron?

11. Great mass resolution: – Mass resolution limited only by electromagnetic (EM) calorimeter and ability to select correct vertex of event (natural width negligible) – 1σ width ~3 GeV or less (M jj width is ~16 GeV) – Resolution ~5x better than best jet algorithms for H  bb – Great background discrimination using M γγ alone – Search for narrow resonance on smoothly falling background – Fits to non-signal region of mass spectrum can be used to estimate background 11 Is a H  γγ search interesting at the Tevatron?

12. Outline Introduction Tevatron and CDF Detector Photon ID and Efficiency SM H  γγ Search Fermiophobic h  γγ Search Summary and Conclusions 12

13. Tevatron 13 pp collisions at √s = 1.96 TeV Peak luminosity 414 × 10 32 cm -2 s -1 Shut down on Sept. 30 th, 2011 11.9 fb -1 delivered 9.9 fb -1 stored on tape at CDF pp collisions at √s = 1.96 TeV Peak luminosity 414 × 10 32 cm -2 s -1 Shut down on Sept. 30 th, 2011 11.9 fb -1 delivered 9.9 fb -1 stored on tape at CDF Results shown here use 7.0 fb -1

14. CDF Detector and Particle Identification 14 Want a detector that can differentiate between different types of final state particles “Jets” come from quarks or gluons fragmenting Muons long-lived and won’t interact in calorimetry; leave track in tracking detector and muon chambers Hadrons interact in calorimetry via cascades of nuclear interactions (much more complex than EM cascades) e’s and γ’s interact in calorimetry via electromagnetic cascades (i.e. ionization and bremmstrahlung for e’s and photoelectric effect, Compton scattering, and pair production for γ’s) Charged particles leave a “track” in the tracking chambers

15. pp Silicon Vertex Detector Central Tracker Muon Chambers Electromagnetic Calorimeter CDF Detector Solenoid Hadronic Calorimeter

16. Outline Introduction Tevatron and CDF Detector Photon ID and Efficiency – Introduction – Central Photons – Forward Photons – Conversion Photons SM H  γγ Search Fermiophobic h  γγ Search Summary and Conclusions 16

17. Photon Identification “Central” – |η|<1.1 “Plug” – 1.2<|η|<2.8 – Tracking efficiency lower than in central region – Easier to miss a track and reconstruct fake object as a photon – Higher backgrounds then for plug photons 17 Central Plug Cross sectional view of one detector quadrant

18. Photon Identification Basic Photon Signature: – Compact EM cluster – Isolated – No high momentum track associated with cluster – Profile (lateral shower shape) consistent with that of a prompt photon Unlike that of π 0 /η  γγ decays (the largest background for prompt photons) Hard to do this with calorimeters alone 18 Inside jets Background Signal

19. Photon Identification ΕΜ calorimeter segmentation: – Δη×Δ ϕ ~ 0.1×15° (|η|<1) – Not fine enough to fully reject π 0 /η jets 19 Hadronic Calorimeter Electromagnetic Calorimeter Shower maximum detector Signal Background Shower max detector – ~6 radiation lengths into EM calorimeter – Finely segmented – Gives resolution to better reject π 0 /η  γγ – Αlso refines EM cluster position measurement to better match associated tracks

20. Central Photon Identification Three level selection (1) Loose requirements – Fiducial in shower max detector – Ratio of hadronic to electromagnetic transverse energy (Had/EM) < 12.5% – Calorimeter isolation. Cut slides with – Track isolation < 5 GeV (2) Track veto – Number tracks ≤ 1 – If 1, then p T trk1 < 1 GeV (3) Cut on NN Output – More details on next slides 20

21. Central Electron Identification Three level selection (1) Loose requirements – Fiducial in shower max detector – Ratio of hadronic to electromagnetic transverse energy (Had/EM) < 12.5% – Calorimeter isolation. Cut slides with – Track isolation – p T trk1 < 5 GeV (2) Track veto – Number tracks ≤ 2 – If 2, then p T trk2 < 1 GeV (3) Cut on NN Output – More details on next slides No pure high statistics data sample of photons to validate ID efficiency Selection chosen so can be modified for electrons Then use Z  e + e – decays (more detail later) 21

22. Central Photon Identification Relative to standard photon selection, increases signal efficiency by 5% and background rejection by 12% NN discriminant constructed from seven well understood variables: – Ratio of hadronic to EM transverse energy – Shape in shower max compared to expectation – Calorimeter Isolation – Track isolation – Ratio of energy at shower max to total EM energy – Lateral sharing of energy between towers compared to expectation 22 Trained using inclusive photon MC and jet MC (with ISR photons removed and energy reweighting) s/sqrt(b) for H  γγ vs NN cut gives optimum cut of 0.74

23. Central Photon ID Efficiency ID efficiency checked in data and MC from Z  e + e – decays Z mass constraint applied to get a pure sample of electrons to probe Effect of pile-up seen through N vtx dependence Net efficiencies obtained by folding ε vtx into N vtx distribution of diphoton data and signal MC (a weighted average) Net photon ID efficiency: Data: 83.2% MC: 87.8% MC scale factor of 94.8% applied Total systematic uncertainty of 2% applied from: – Differences between electron vs photon response (checked in MC) – Data taking period dependence – Fits made to Z mass distribution Small uncertainties using this method! 23

24. Plug Photon ID and Efficiency Standard CDF Cut-Based ID Fiducial in shower max detector Ratio of hadronic to EM transverse energy * < 5% Calorimeter isolation * < 2 GeV Track isolation * < 2 GeV Shape in shower max compared to expectation Same Efficiency Technique as for Central Photons Net photon ID efficiency: – Data: 73.2% – MC: 80.6% MC scale factor of 90.7% applied Total systematic uncertainty of 4.5% 24 * Slides with EM energy or E T

25. Photon Conversions γ  e + e – Colinear tracks moving in approximately same direction Occurs in presence of detector material More material, higher the probability of converting 25 COT inner cylinder port cards, cables ISL outer screen L7 L6 L00, L0-L4

26. Photon Conversions Conversion probability at CMS substantially higher * … ~70% of H  γγ events have at least one photon that converts!! Similarly for ATLAS Much more important at LHC experiments! * J. Nysten, Nuclear Instruments and Methods in Physics Research A 534 (2004) 194-198 26 Use central only Then for two photons, % of events lost from a single central photon converting is: – 26% for CC channel – 15% for CP channel CDF had only one Run I measurement using converted photons: γ cross section  PRD,70, 074008 (2004) H  γγ is the second analysis to use it for Run II p ≈ 15% for central γ

27. Conversion ID 27 r- ϕ separation (cm) cotθ = p z /p T cut ~94% efficient cut ~95% efficient Δcotθ Example trident

28. Outline Introduction Photon ID and Efficiency SM H  γγ Search – Event Selection – Background Modeling – Results – Tevatron Combination Fermiophobic h  γγ Search Summary and Conclusions 28

29. Event Selection Inclusive photon trigger – Single photon E T > 25 GeV – Trigger efficiency after offline selection obtained from trigger simulation assuming z vtx = 0 and trigger tower clustering Use photon ID as previously described Photon p T > 15 GeV Four orthogonal diphoton categories: – Central-central photons (CC) – Central-plug photons (CP) – Central-central conversion photons (CC Conv) where one converts – Central-plug conversion photons (CP Conv) where central converts 29

30. Primary Background Composition Real SM photons – Irreducible background – Via QCD processes from hard interaction Fake backgrounds – Reducible backgrounds – Electrons from Z/γ *  e + e – – Jets fragmenting to neutral mesons (π 0 /η  γγ) which then decay to pairs of colinear photons and are reconstructed as a single photons 30

31. Data-Driven Background Model Fit to non-signal (“sideband”) regions of M γγ distribution We use a 6 parameter polynomial times exponential to model smooth portion of the data Fit is then interpolated into the 12 GeV signal region to estimate background expectation for Higgs mass hypothesis Example shown here for a test mass at 115 GeV for CC channel 31

32. Data-Driven Background Model Channels with a plug photon have a non-neglible contaminated from Z background Breit-Wigner function added to smooth distribution to model this, where mean and width are bounded in fit Example shown here for a test mass at 115 GeV for CP channel 32

33. Background Model Vertical red lines show window excluded from fit for Higgs mass hypothesis being tested Interpolated fit used to obtain data-fit residuals Used to inspect for signs of a resonance for each mass and channel No significant resonance observed 33 CC Channel CC Conversion Channel

34. Background Model 34 CP Channel CP Conversion Channel Vertical red lines show window excluded from fit for Higgs mass hypothesis being tested Interpolated fit used to obtain data-fit residuals Used to inspect for signs of a resonance for each mass and channel No significant resonance observed

35. Background Rate Uncertainty Parameters of fit function varied within uncertainties to obtain a new test fit Integral in 12 GeV signal region calculated for test fit Repeated many times Largest upper and lower differences from standard fit stored Then symmetrized to obtain rate uncertainty for each test mass and channel Model dependence checked by testing alternate fit functions Variation in normalization as compared to standard found to be within uncertainties already obtained 35 Approximate Systematic Errors on Background (%) CC4 CP1 CC Conv8 CP Conv4

36. CC Channel Discriminant and some Math Checks We show these distributions so you can check our results with some simple math… Use 12 GeV signal region, so as an example, here it would be 120 +- 6 GeV Real limits are much more complicated, but this is a rough check that results are in the right ball park N Signal (S) = 2.2, N Bkg (B) = 271 +- 16.5 16.5  1σ statistical error on the background expectation  68% of the time For S = 2.2, obviously we’re not sensitive to a SM Higgs observation How many signal events would we be sensitive to at say a 95% C.L? 33  2σ  about 95% data fluctuates between 271 +- 33 The number 33 is simple 95% upper C.L. limit on the amount of reconstructed signal we’re sensitive to based on bkg expectation alone We can excludes models with predict > 33 γγ decays at this mass @ 95% C.L. How much is this relative to the SM prediction?: 33/2.2 = 15  This is a simple 95% C.L. upper limit relative to the SM prediction Including systematic uncertainties degrades limits by about 10-15% 36

37. Final Discriminants 37 Limits add as ~ 1/L i 2 similar to a || resister, where L i is limit for an individual channel

38. SM Limits Shown are 95% upper C.L. limits on σ×Br relative to SM prediction using a Bayesian method Most sensitive expected limit is for 120 GeV where limit is ~13.0×SM An improvement of ~33% on last result presented! Improvements from better central photon ID, including forward photons, and reconstructing photon conversions Observed limit at ~28×SM above 2σ but we didn’t consider “look- elsewhere effect” With this affect considered, has less than 2σ significance This result included in Tevatron Higgs combination First SM Higgs result from CDF Run II to be published Has also been combined with D0 results to give a Tevatron SM H  γγ result… 38 arXiv:1109.4427 To be published in PRL

39. DØ’s SM H→γγ Search Uses a boosted decision tree as final H  γγ discriminant Βased on five kinematic inputs: M γγ, p T γγ, E T 1, E T 2, Δφ γγ Example output shown for mass of 115 GeV From March 2011 Using 8.2fb -1 Observed @ 115: 12.5xSM Expected @ 115: 15.8xSM PRL 107, 151801 (2011) 39

40. Tevatron 95% C.L. Limits on H  γγ For M H of 115 GeVLuminosityExpected/SMObserved/SM Tevatron H  γγ Combo ≤ 8.2 fb -1 8.510.5 40 arXiv:1107.4960

41. Tevatron vs LHC Due to higher jet backgrounds, the LHC is betting on the H  γγ channel rather than H  bb for a low mass Higgs discovery… Also gain from higher cross sections, calorimeter resolution, and mass resolution Limited by Br (as at Tevatron) and multiple interactions (pileup) As of Sept, 2011 both CMS (CMS-PAS-HIG-11-021) and ATLAS (arXiv:1108.5895) have results in H  γγ of about 3-4xSM expectation: Tevatron is clearly not competitive with LHC in this channel… how about in BSM models? 41

42. Outline Introduction Tevatron and CDF Detector Photon ID and Efficiency SM H  γγ Search Fermiophobic h  γγ Search – Theory Motivation – Differences in search from SM – Results – Tevatron Combination Summary and Conclusions 42

43. Fermiophobic Higgs (h f ) It’s likely nature doesn’t follow the SM Higgs mechanism… We also consider a “benchmark” fermiophobic model A two-Higgs doublet model extension to the SM Spontaneous symmetry breaking mechanism different for fermions and bosons  5 Higgs We search for one in which: – No Higgs coupling to fermions – SM Higgs coupling to bosons – SM production cross sections assumed 43

44. Gluon Fusion ~ 1000 fb @ 120 GeV Fermiophobic Higgs (h f ) Production No gg  h f σ ~ 300 fb @ 120 GeV 44 Associated Production ~ 225 fb @ 120 GeV Vector Boson Fusion ~ 70 fb @ 120 GeV

45. Fermiophobic Higgs (h f ) Decay h  bb no longer dominant 45 Suppressed by m 2 b /m 2 W Dominant low mass decay mode is now h  γγ Signal expectation @ 120 GeV: N = σ × L × Br = 300fb × 7.0fb -1 × 0.03 ~63 (22) h f  γγ events produced (reconstructed) ~4x higher than SM expectation Br ~ 13x higher than SM @ 120 GeV

46. Fermiophobic Higgs (h f ) Decay h  bb no longer dominant 46 Signal expectation @ 100 GeV: N = σ × L × Br = 560fb × 7.0fb -1 × 0.18 ~700 (245) h f  γγ events produced (reconstructed) ~30x higher than SM expectation Suppressed by m 2 b /m 2 W Dominant low mass decay mode is now h  γγ Br ~ 120x higher than SM @ 100 GeV

47. Event Selection Inclusive photon trigger – Single photon E T > 25 GeV – Trigger efficiency after offline selection obtained from trigger simulation assuming z vtx = 0 and trigger tower clustering Use photon ID as previously described Photon p T > 15 GeV Four orthogonal diphoton categories: – Central-central photons (CC) – Central-plug photons (CP) – Central-central conversion photons (CC conv) where one converts – Central-plug conversion photons (CP conv) where central converts gg  h f suppressed Optimize for VH/VBF Split into three diphoton pt bins: – High: p T > 75 GeV – Medium: 35 < p T < 75 GeV – Low: p T < 35 GeV 4 diphoton categories x 3 p T bins = 12 total channels 47 Greatest sensitivity! Same as SM SearchDifferent for h f search

48. Background Model Example fits for CC for each p T γγ bin 48 Same approach for background model as done for SM High p T γγ Bin At 120 GeV: N signal = 2.9 s/sqrt(b) = 0.66 Medium p T γγ Bin At 120 GeV: N signal = 2.5 s/sqrt(b) = 0.37 Low p T γγ Bin At 120 GeV: N signal = 1.3 s/sqrt(b) = 0.09

49. Results At 95% C.L., observed (expected) on B(h f  γγ) exclude a fermiophobic Higgs boson with a mass < 114 GeV (111 GeV) A limit of 114 GeV is currently the world’s best limit on a h f Higgs from a single experiment To be published in PRL with SM result Has also been combined with D0 results to give a Tevatron h f  γγ result… 49 arXiv:1109.4427 To be published in PRL Experimentm hf limit (GeV) LEP109.7 Previous CDF PRL (3.0 fb-1)106 D0’s PRL result (8.2 fb-1)112.9 CMS Prelim result (1.7 fb-1)112 CDF new result (7.0 fb-1)114

50. DØ’s fermiophobic h f →γγ Search Same as SM, but no gg  H and higher Br Uses a boosted decision tree as final h f  γγ discriminant Βased on five kinematic inputs: M γγ, p T γγ, E T 1, E T 2, Δφ γγ Example output shown for mass of 115 GeV From March 2011 Using 8.2fb -1 Observed @ 115: 12.5xSM Expected @ 115: 15.8xSM PRL 107, 151801 (2011) 50

51. Tevatron 95% C.L. Limits on h f This is what exclusion looks like Tevatron results exclude fermiophobic Higgs bosons with masses below 119 GeV at 95% C.L. This is the world’s best limit (arXiv:1109.0576) 51

52. Summary and Conclusions CDF SM H and h f  γγ soon to be published (we just got the bill) Improvements at CDF upon previous analyses: – Inclusion of forward and conversion photons – Better central ID from a NN – Including lower p T regions to h f search SM H  γγ search: – Not competitive with LHC, but contributes to final Higgs search at Tevatron – Tevatron Combination ~ 13xSM (observed) 9xSM (expected) Fermiophobic h f  γγ: – Tevatron setting world’s best limit on this model @ m hf > 119 GeV – CDF limit of 114 GeV most stringent limit by a single experiment Both CDF analyses to be updated with full dataset and a few improvements, so stay tuned… 52

53. Backup 53

54. Indirect constraints Precision electroweak observables are sensitive to the Higgs boson mass via quantum corrections. m H < 161 GeV (95% CL) Stalking the Higgs Boson Direct searches at LEP Tantalizing hints (~1.7  ) of a SM-like Higgs boson with m H ~115 GeV: m H > 114.4 GeV (95% CL) 114.4 < m H < 161 GeV (95% CL) Combining indirect and direct constraints

55. Signal Shapes Widths ~3 GeV (or less) for each channel Use 2σ width to determine signal window  12 GeV Shapes used to fit for signal in the data when setting limits 55

56. Systematic Uncertainties on H  γγ Signal 56

57. Used two central photons from cut-based ID 12 GeV/c 2 signal region for each test mass used to set upper limits set on σ  Br relative to SM prediction Expected and observed limits in good agreement Expected limits of 19.4xSM @ 120 GeV Most sensitive for range 110 – 130 GeV/c 2 Added to SM Higgs Tevatron combination this past summer 57 Previous Limits on H  γγ at CDF using 5.4/fb

58. Fermiophobic Higgs (h f ) It’s likely nature doesn’t follow the SM Higgs mechanism… We also consider a “benchmark” fermiophobic model A Two Higgs Doublet Model (2HDM) extension to SM: With vacuum expectation values: 5 Physical Higgs Particles: h 0, H 0, A 0, H +, and H – 2HDM type-I – Scalar field mixing angle α can lead to different couplings to fermions for h 0 and H 0 – sin(α) for H 0 and cos(α) for h 0 – Limit of α  π/2 yields a Higgs with enhanced coupling to bosons: h 0  h f Standard model cross sections assumed Not present in MSSM 1 1 Akeroyd, hep-ph/9511347, 1995Introduction58

59. Tight Conversion ID Primary electron: – Fiducial – Had/EM < ~5.5% Secondary electron: – Fiducial – p T > 1.0 GeV Conversion photon – pT > 15 GeV – E/P ratio (optimized for Η  γγ) – Calorimeter isolation (optimized for H  γγ) – Rconv > 2.0 cm Photon ID and Efficiency – Conversion Photons59 ~24% of events ~72% of events

60. Conversion ID Efficiency Search for “tridents” where one electron leg brems a photon which then converts Probed conversions of lower momentum range than those from H  γγ Obtain an uncertainty on conversion ID rather than MC scale factor In data and MC calculate ratio: R data /R MC  7% uncertainty Photon ID and Efficiency – Conversion Photons60