240 likes | 258 Views
Tetsuro Sekiguchi, KEK. BNL-E949 Collaboration. BNL, FNAL, UNM, Stony Brook Univ. (USA) Alberta, TRIUMF (Canada) IHEP, INR (Russia) Fukui, KEK, Kyoto, NDA, Osaka, Osaka RCNP (Japan). The E949 experiment The analysis The results Conclusions. Signal = Stopped + nothing.
E N D
Tetsuro Sekiguchi, KEK BNL-E949 Collaboration BNL, FNAL, UNM, Stony Brook Univ. (USA) Alberta, TRIUMF (Canada) IHEP, INR (Russia) Fukui, KEK, Kyoto, NDA, Osaka, Osaka RCNP (Japan) • The E949 experiment • The analysis • The results • Conclusions
Signal = Stopped + nothing • Low energy beam and stop in the target (intensity = E787) • Kinematics measurement momentum, energy and range • in the stopping counter • Photon veto hermetic detectors BNL-E949 = Successor of E787 side view end view
RS Layer 1-5 replacement more light output • RS gain monitor system better energy calibration Improved kinematics measurement Kinamatics measurement is sensitive to signal selection
New detectors in blue. • Rejection to background as a function of acceptance for E787 and E949. • better rejection at 80% of nominal acceptance. Improved photon veto
Data Taking • Physics run in 2002 (12 weeks) • Beam condition was not optimized • Detector worked very well • Smooth data taking
Blind Analysis • Measure Background level with real data • To avoid bias, 1/3 of data cut tuning 2/3 of data background measurement • Characterize backgrounds using back- ground functions • Likelihood Analysis Analysis Signal region“the BOX” Background sources Analysis Strategy
but range is small due to interactions in RS. Changing cut position Acceptance & background levelat each point of parameter Functions Neural net function for and Background characterization Background can be characterized using background functions For muon backgrounds
Branching ratio and Confidence level • Both and are small Poisson statistic • The ratio of two Poisson probabilities • # of observed candidate events in the cell Likelihood Analysis • The signal region is divided into cells. Cell construction by binning the parameter space of each function. • The signal and the background in the cell • The cell is characterized by thesignaltobackgroundratio • Likelihood ratio technique(T. Junk [NIM A434, 435 (1999)]) (BR)( : Acceptance) Likelihood estimator of cells containing candidate events
NK (1012) Total acceptance (%) Sensitivity (10-10) Sensitivity and background Sensitivity Background Note: 10% larger acceptanceby enlarging the signal region, resulting inmore backgrounds For the likelihood analysis, important is the ratio in each cell NOT the total background level in signal region. All cuts are fixed and ready to open the BOX !
Opening the BOX Range (cm) and Energy (MeV) for E949 data after all other cuts applied. Solid line shows signal region. Single candidate found.
Branching ratio & Confidence level • E949 result alone: • Combine E787 and E949 results increase statistics (68% CL) E949(02) = combined E787&E949. E949 projection with full running period. (~60 weeks)
E949 has observed an additional candidate. (68% CL, PNN1 region) from the combined E787 and E949 result. • We need more data. - Further E949 running? - Analysis of “below (PNN2) region” Conclusions • Upgrades of E787 to E949 were successful. • Likelihood analysis was performed to measure
By courtesy of G. Isidori Central value [dashed], 68% interval [dot-dashed], 90% interval [solid] (including theoretical uncertainties)
Events with early decay are selected • signal: blue • background: red • arrow: candidate event
Signal rate and background level for the candidate cell Signal Si Background bi
Acceptance calculation Cross check
Cut R = PV or TD: loosen by Cut K = KIN: loosen by more background event should be observed in loosened BOX. ( ) Verify background prediction by loosening cuts
Branching ratio and confidence limits • 2002 candidate event alone • Branching ratio: • Combined measurement (1995-2002) • Combined BR:
Si (BR) BR : signal • BR: • NK: # of Kaon decay • Ai: Acceptance bi: background di: # of observed events Likelihood Analysis with T. Junk method The Poisson probability to observe di with Si + bi or bi expected
Likelihood Analysis with T. Junk method Summing all the set di to satisfy Then obtain the confidence level