Background Understanding and Suppression in Very Long Baseline Neutrino Oscillation Experiments with Water Cherenkov Detector Chiaki Yanagisawa Stony Brook Talk at NNN05, Aussois, France April 7-9, 2005
VLBNO Setting the stage e and e +N e + invisible N' + (invisible n s, n 0) e e ' invisible n s, n 0) Look for single electron events Major background e contamination in beam (typically 0.7%) How do we find the signal for v e ~ a half megaton F.V. water Cherenkov detector, for example UNO BNL very long baseline neutrino beam VLB neutrino oscillation experiment See, for example, PRD68 (2003) for physics argument ● Very long baseline neutrino oscillation
Neutrino spectra of on- and off-axis BNL Superbeams on-axis beam 1 o off-axis beam Neutrino energy (GeV) s/GeV/m 2 /POT PRD68 (2003) 12002; private communication w/ M.Diwan VLBNO
How is analysis done ? Use of SK atmospheric neutrino MC Flatten SK atm. spectra and reweight with BNL beam spectra Normalize with QE events: 12,000 events for events for beam e for 0.5 Mt F.V. with 5 years of running, 2,540 km baseline Reweight with oscillation probabilities for and for e Standard SK analysis package + m 2 21 =7.3 x eV 2, m 2 31 =2.5 x eV 2 sin 2 2 ij (12,23,13)=0.86/1.0/0.04, CP =0,+45,+135,-45,-135 o Probability tables from Brett Viren of BNL Oscillation parameters used: distance from BNL to Homestake special 0 finder
VLBNO Selection criteria used to improve Likelihood analysis using the following eight variables: Initial cuts: One and only one electron-like ring with energy and reconstructed neutrino energy more than 100 MeV without any decay electron mass, energy fraction, costh, -likelihood , e-likelihood -likelihood, total charge/electron energy, Cherenkov angle To reduce events with invisible charged pions Traditional SK cuts only With finder
VLBNO QE events only before likelihood cut All CC events before likelihood cut Erec Reconstructed energy E E Erec All CC events that survive the initial cuts are signals What are sources of the signal (QE and nonQE) ? Only single e-like events left after initial cut
0 finder finder reconstruction efficiency with standard SK software measured opening angle vs. mass with finder m (MeV/c 2 ) true opening angle (deg) efficiency opening angle measured(deg) Single e-like events from single int. All single interactions SK atm. neutrino spectra Always looks for an extra ring in a single ring event inefficiency due to overlap inefficiency due to weak 2 nd ring ● Finder
reconstruction efficiency reconstruction efficiency with standard SK + finder efficiency All the single int. True opening angle (deg) finder 0 mass cut:1- and 2-ring events 0 mass cut:2-ring events With atmospheric neutrino spectra with 0 finder without 0 finder with 0 finder w/o 0 finder
Variables GeV GeV GeV GeV GeV GeV GeV GeV background signal 0 mass Useful Variables All the distributions of useful variables are obtained with neutrino oscillation “on” with CPV phase angle 0 mass
Variables GeV GeV GeV GeV GeV GeV GeV GeV background signal Energy fraction of 2 nd ring Fake ring has less energy than real one
Variables Primary electron ring An undetected weak ring initially - One algorithm optimized to find an extra ring near the primary ring (forward region) - Another algorithm optimized to find an extra ring in wider space (wide region) - See the difference ln -likelihood (forward) - ln - likelihood (wide) Difference between ln of two -likelihood (wide vs. forward)
Variables Difference between ln of two -likelihood (wide vs. forward) GeV GeV GeV GeV GeV GeV GeV GeV background signal
Variables GeV GeV GeV GeV GeV GeV GeV GeV background signal costh = cos e
Preliminary Likelihood Cut ln likelihood distributions Trained with e CC events for signal, CC/NC & e NC for bkg GeV GeV GeV GeV GeV3.0- GeV likelihood signal background Difference in ln likelihood between sig and bkg ln likelihood ratio Preliminary
Bkgs 169 (112 from +others) ( 57 from e ) ln likelihood cut (100% signal retained) Background from CP+45 o Signal 700 evBkgs 2005 (1878 from +others) ( 127 from e ) Signal e background o CP+45 ln likelihood cut (~50% signal retained) Effect of cut on ln likelihood E rec Preliminary Signal/Background Signal 321 ev e CC for signal ; all e NC, e beam for background TRADITIONAL ANALYSIS After initial cuts
ln likelihood cut (40% signal retained) Background from CP+45 o Signal 251 evBkgs 118 ( 74 from +others) ( 44 from e ) Signal e background o CP-45 ln likelihood cut (~40% signal retained) Effect of cut on likelihood E rec Preliminary Signal/Background Signal 142 ev e CC for signal ; all e NC, e beam for backgrounds Bkgs 118 ( 75 from +others) ( 43 from e )
Effect of cut on likelihood ln likelihood cut (~40% signal retained) Background from CP+135 o Signal e background o CP-135 ln likelihood cut (~40% signal retained) E rec Preliminary Signal/Background Signal 342 evBkgs 126 ( 81 from +others) ( 45 from e ) Signal 233 evBkgs 122 ( 78 from +others) ( 44 from e ) e CC for signal ; all e NC, e beam for backgrounds
Preliminary Summary of BNL Variable removed SignalBkgSignal Bkg Effic None e CC all, e, NC e CC e CC e CC 50% pi0lh pi0-lh poa e-lh 50% Beam e % e CC Issues ● S/B and variables e CC 50% efrac pi0mass costh Neutrino oscillation was on to define template distributions For analysis CPV=+45 o Some issues all, e, NC ange e CC 50%
Conclusion ● Conclusion Realistic MC simulation studies have been performed for the BNL very long baseline scenario with a water Cherenkov detector. It was found that BNL VLB combined with a UNO type detector seems to DO A GREAT JOB – Very exciting news but needs confirmation. It was demonstrated that there is some room to improve S/B ratio beyond the standard water Cherenkov detector software even with currently available software We may need further improvement of algorithm/software, which is quite possible A larger detector such as UNO has an advantage over a smaller detector such as SK (we learned a lesson from 1kt at K2K) Detailed studies on sensitivity on oscillation parameters needed