First Observation of the Bottomonium Ground State

1. First Observation of the Bottomonium Ground State Chris West SLAC National Accelerator Laboratory Fermilab HEP SeminarApril 27, 2010

2. Outline Introduction Previous searches for the b Υ(3S) → b : first observation of b Υ(2S) → b : confirmation of b Combination of results Conclusions

3. Introduction

4. Hyperfine Splitting in Hydrogen Atom • Hyperfine splitting from Zeeman effect due to magnetic field of nucleus Very small effect, proportional to a4(m/mp); Responsible for 21 cm line in microwave astronomy

5. Bottomonium Bound state of b quark and b antiquark The b is the ground state Last ground state of a quark-antiquark system to be observed Large mb → nonrelativistic system, small as Many transitions between states; allowed transitions restricted by symmetries Of interest in this study are the magnetic dipole transitions to the b and the Υ(1S)-b hyperfine mass splitting

6. Theoretical Tools Lattice QCD Effective field theories (EFTs) Non-relativistic QCD (NRQCD) Potential NRQCD (pNRQCD) Potential models Simulation of action density of QCD vacuum in lattice QCD

7. Hyperfine Splitting in Bottomonium Hyperfine splitting Mass difference between triplet and singlet states, m-mb QCD analog of hyperfine splitting in hydrogen Lattice (NRQCD): 61(4)(12)(6) MeV Gray, et al., PRD 72, 094507 (HPQCD and UKQCD Collaborations) Errors due to Statistical/fitting/discretization Radiative corrections Relativistic corrections Perturbative QCD 44 ± 11 MeV (static QCD potential) S. Recksiegel and Y. Sumino, PLB 578, 369 (2004) 41 ± 11+8-9 (s) MeV (pNRQCD) Kniehl, et al. PRL 24, 242001 Potential models 35 – 100 MeV Expected expt. precision NLL NLO LO LL Hyperfine splitting assuming s± 0.003 Kniehl, et al. PRL 24, 242001

8. Hyperfine Splitting from Kniehl, et al. Leading log (LL) Leading order (LO) Next to leading log (NLL) Dependent only on fundamental parameters as and mb possibly useful for extracting as

9. Branching Fraction Predictions Primarily calculated in potential models Often neglecting relativistic corrections Including relativistic corrections plagued with technical ambiguities Range of theoretical predictions: (1-15)x10-4 for Υ(2S) → b (1-20)x10-4for Υ(3S) → b Other methods: Radiative transition rates calculated in lattice QCD only for charmonium No similar study done for bottomonium No EFT-based calculations for transitions from excited states

10. Width of b q~ s2|q(0)|2mq ~s5mq (q=c, b) b width smaller than c width of 26.5 MeV due to smaller s() at =mb versus mc Predictions range from 4-20 MeV

11. Previous Searches for the b

12. Previous Knowledge of b Entry in PDG from 2002 to 2008 12

13. Previous Searches for the b In two-photon events at ALEPH, L3, and DELPHI, b reconstructed in set of exclusive modes Best limit on xBF from ALEPH (95% CL): < 48 eV (4 charged tracks), <132 eV (six charged tracks) Assumes b) = 557 ± 85 eV CLEO III limit: BF[(3S)b] < 4.3x10-4, BF[(2S)b] < 5.1x10-4 @ 90% CL Unpublished CDF limit (at 95% CL): b(|y|<0.6)∙BF(bJ/J/) ∙ BF(J/)2< 2.6 pb

14. Search for Υ(3S) → b at BaBar

16. BaBar Calorimeter Used in this analysis for measurement of photon energies Composed of 6580 CsI(Tl) crystals

18. Simulated event hb decays (through two gluons) have high track and cluster multiplicity Cluster in calorimeter consistent with EM shower, isolated from charged tracks and rest of event, inconsistent with being a p0 daughter, away from edges of calorimeter

19. Analysis Overview Decay modes of b not known or predicted; analysis must remain as inclusive as possible Two body decay: look for a bump in Edistribution around Reduce continuum/0background with photon isolation cuts and 0veto Accurately model peaking background Huge background! Blind analysis Expected signal position

20. Signal PDF Photon peaks normally fit with Crystal Ball function: a Gaussian with a power law tail to model energy leakage Signal probability density function (PDF) modeled with a single Crystal Ball function convolved with a non-relativistic Breit-Wigner of width 10 MeV Fit signal MC with all selection criteria imposed to determine signal PDF and efficiency of = (35.8 ± 0.2) % Crystal Ball Function

21. Background Sources Non-peaking backgrounds udsc production Generic ISR Bottomonium decays Peaking backgrounds Υ(3S)→ bJ(2P), bJ(2P)→ Υ(1S) (J=0, 1, 2) e+e-→ ISRΥ(1S) b ISR(1S) b ?

22. Background: e+e- → ISRΥ(1S) Photon from ISR production of Υ(1S) peaks at 856 MeV Close to signal. Very important to model correctly! Yield fixed from off-resonance Υ(4S) data, adjusted for luminosity, cross-section and efficiency Fitted yield: 35800±1600 Yield extrapolated to Υ(3S): 25200±1700 Yield could also be fixed using Υ(3S) off-resonance data Extrapolated yield consistent Lower statistical precision After Bkg Subtraction Off-resonance Υ(4S) data before Bkg. Subtraction

23. Background: b(2P)→ Υ(1S) Second transition in Υ(3S) → b(2P), b(2P)→ Υ(1S) Three overlapping peaks: b0(2P) E = 743 MeV b1(2P) E = 764 MeV b2(2P) E = 777 MeV Model each as a Crystal Ball function Transition point and power law tail parameter fixed to same value for each peak Means fixed to PDG values minus a common offset Ratio of yields taken from PDG Offset of 3.8 MeV observed in data used to correct energy scale of other peaks. Shape fixed from full dataset with signal region excluded Bkgd subtracted distribution bJ(2P)->(1S) J=0,1,2 ISR(1S) PDF Signal region excluded

24. Fit Strategy b peak shape fixed, yield allowed to float ISR peak position and lineshape fixed; yield fixed from Υ(4S) off-resonance data Zero-width b shape fixed from MC, convolved with Breit-Wigner shape Non-peaking background modeled by empirical function:

25. Fit Results b peaks ISR(1S) b ?

26. First Observation of the b 19200±2000 events Bkg. subtracted Cont. bkg. subtracted b ISR b 10 significance!

27. Observation of b in (3S) Sample b mass Hyperfine splitting Branching fraction The implications of these values will be discussed later in the talk

28. Search for Υ(2S) → b at BaBar

29. Confirmation of hb in different dataset with signal peak at a different energy Improved absolute energy resolution at lower signal photon energies → better separation between peaks Ratio of branching fraction to hb at Υ(2S) and Υ(3S) resonances a probe of nature of peak seen in Υ(3S) sample Motivation for Υ(2S) Analysis

30. Event Selection Use same selection as Υ(3S) analysis with re-optimized |cosT| and E2 selections |cosT|<0.8 Was 0.7 in Υ(3S) analysis Due to lower continuum background fraction in Υ(2S) analysis E2 > 40 MeV and |m-m|<15 MeV Was E2 > 50 MeV in Υ(3S) analysis More combinatorial p0 background at Eg=614 MeV versus 921 MeV Hadronic event and photon selection criteria identical

31. Sources of Background

32. Sources of Background Non-peaking background udsc production – mainly 0decays Generic ISR Bottomonium decays Modeled by exponential of 4th order polynomial Υ(2S)→ bJ(1P), bJ(1P)→  Υ(1S) (J =0, 1, 2) e+e- →  Υ(1S) Other Υ(2S) backgrounds ΥS ΥS absorbed into non-peaking component ΥS ΥS ΥS ΥS The presence of these backgrounds is considered as a (small) systematic error

33. Background: Υ(2S) → b, b → Υ(1S) Second transition in Υ(2S)→ b, b→  Υ(1S) Three overlapping peaks: b0 E = 391.1 MeV b1 E = 423.0 MeV b2 E = 441.6 MeV Improved energy resolution → some technical issues become important Doppler broadening due to b CMmomentum non-negligible compared to Gaussian width: ~5 MeV compared to ~10 MeV Scaling widths from c states show that the width of the b peaks is negligible Relative rates fixed from control sample Υ(2S)→b, b Υ(1S) , Υ(1S)→

34. Background: e+e- → Υ(1S) Decided to float ISR yield in fit Compared to Υ(3S) analysis, peaks better separated, toy studies show that it is not necessary to fix ISR yield Error on extrapolated ISR yield comparable to fitted ISR yield Estimated ISR yield used as consistency check Use ISR yield from Υ(4S) off-peak data to estimate yield in on-Υ(2S) data Estimated yield from Υ(4S) sample consistent with off-Υ(3S) and off-Υ(2S) yields Bkg. Subtracted (4S) off-resonance

35. Tests of Fit Procedure

36. Tests of Fit Procedure Fit to optimization sample Fit of full data sample with signal region excluded Toy studies using simulated datasets

37. Fit to Optimization Sample Test fit procedure on 1/10 optimization sample 2/ndof=94/93 ISR yield consistent with expectation of 1423 b b ISR 38

38. Fit of Blinded Sample Fitted ISR yield of 15200+4200-4000 consistent with expected yield of 16700 A check of the fitted background yield near the signal region Residuals show no unexpected features in signal region 2/ndof=116.2/93 Blinded region 39

39. Fit Results

40. Fitted Spectrum and Residuals b peaks ISR(1S) b ?

41. Background-subtracted Spectrum b b ISR

42. Zoomed Spectrum b b ISR

43. Comparison with Υ(3S) Spectrum Υ(2S) spectrum Υ(3S) spectrum

44. Fit Results b yield: Corrected b peak position: 2/ndof=115.1/93

45. Width Variations We decided before unblinding to use an b width of 10 MeV. Theoretical predictions vary between 4 and 20 MeV. Other widths not significantly favored by the data

46. Yield and Peak Position Systematics

47. Branching Fraction Systematics Selection efficiency Branching fraction

48. Summary of Υ(2S) Results Branching fraction: b mass: Hyperfine splitting: Hyperfine splitting consistent with result from Υ(3S) analysis:

49. Combination of Results Combined hb mass Ratio of branching fractions Consistent with lattice QCD calculation of HPQCD and UKQCD collaborations but 2s higher than pNRQCD calculation making extraction of as problematic Consistent with (large!) range of predictions from potential models ~ 0.3 – 0.7

50. Updated CLEO Analysis After the BaBar b publications, CLEO updated their b analysis, including |cosT| information and ISR background. They now find 4s evidence for the b; their results are consistent with those of BaBar • |cosT|<0.3 • 0.3<|cosT|<0.7 • |cosT|>0.7