UNIVERSITY OF CALIFORNIA, IRVINE
A search for proton decay via p → μ+K0 in Super Kamiokande I
DISSERTATION
submitted in partial satisfaction of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in Physics
by
Christopher Regis
Dissertation Committee: Professor David Casper, Chair Professor Hank Sobel Professor Jonas Schultz
2011 c 2011 Christopher Regis TABLE OF CONTENTS
Page
List Of Figures vi
List Of Tables xii
Acknowledgments xiv
Curriculum Vitae xv
Abstract Of The Dissertation xvi
1 Introduction 1 1.1TheoreticalMotivation...... 1 1.1.1 ConservationandSymmetries...... 2 1.1.2 StandardModel...... 3 1.1.3 GrandUnifiedTheories...... 5 1.2 Past Experimental Measurements of p → μ+K0 ...... 9 1.2.1 Soudan2 ...... 9 1.2.2 IrvineMichiganBrookhaven(IMB)...... 9 1.2.3 Kamiokande...... 10 1.2.4 SuperKamiokande...... 10 1.3OverviewofExperimentalMethod...... 10
2 Detector 14 2.1CherenkovRadiation...... 14 2.2SuperKamiokandeDetector...... 16 2.3PhotomultiplierTube...... 18 2.4Electronics&DataAcquisition...... 18 2.4.1 InnerDetector...... 18 2.4.2 OuterDetector...... 21 2.4.3 TriggerSystem...... 21 2.5WaterPurificationSystem...... 23 2.6Calibration...... 24 2.6.1 TimingCalibration...... 24 2.6.2 RelativeGainCalibration...... 25 2.6.3 AbsoluteGainCalibration...... 27 2.6.4 WaterTransparency...... 28
ii 2.6.5 AbsoluteEnergyCalibration...... 30
3 Simulation 38 3.1ProtonDecay...... 38 3.1.1 FermiMomentumandNuclearBindingEnergy...... 39 3.1.2 Kaon-NucleonInteractionintheOxygenNucleus...... 40 0 → 0 3.1.3 KL KS RegenerationintheOxygenNucleus...... 41 3.2AtmosphericNeutrinoFlux...... 42 3.3NeutrinoInteraction...... 44 3.3.1 Elastic&Quasi-ElasticScattering...... 44 3.3.2 SingleMesonProduction...... 46 3.3.3 DeepInelasticScattering(DIS)...... 48 3.3.4 CoherentPionProduction...... 50 3.3.5 NuclearEffects...... 52 3.4DetectorSimulation...... 54 3.4.1 Photon Propagation ...... 54 3.4.2 Hadron Propagation ...... 54 0 → 0 3.4.3 KL KS RegenerationinWater...... 54
4 Fully Contained Event Reduction 56 4.1Overview...... 56 4.2FirstReduction...... 56 4.3SecondReduction...... 57 4.4ThirdReduction...... 58 4.4.1 Through-goingMuonCut...... 58 4.4.2 StoppingMuonCut...... 59 4.4.3 CablePortMuons...... 59 4.4.4 FlasherEventCut...... 60 4.4.5 AccidentalCoincidenceCut...... 60 4.4.6 LowEnergyEventsCut...... 61 4.5FourthReduction...... 61 4.6FifthReduction...... 63 4.6.1 StoppingMuonCut...... 63 4.6.2 InvisibleMuonCut...... 63 4.6.3 CoincidenceMuonCut...... 64 4.6.4 LongTailFlasherCut...... 64 4.7FinalFullyContainedFiducialVolumeSample...... 64
5 Event Reconstruction 66 5.1StandardEventReconstruction...... 66 5.1.1 Vertex...... 67 5.1.2 RingCounting...... 71 5.1.3 ParticleIdentification...... 73 5.1.4 MSVertexFit...... 77 5.1.5 MomentumDetermination...... 77
iii 5.1.6 DecayElectronSearch...... 82 5.1.7 RingNumberCorrection...... 83 5.2MultipleVertexEventReconstruction...... 84 5.2.1 StandardReconstructionWithoutTimingCuts...... 85 5.2.2 PMTMaskingRegion...... 86 5.2.3 Primary μ-candidateReconstruction...... 86 5.2.4 RemainingParticleReconstruction...... 92
6 Overview of the p → μ+K0 search 95 6.1DataSet...... 95 6.2 Strategy of the Combined p → μ+K0 Search...... 96
0 → 0 0 7Searchfor“KS π π ”99 7.1EventSelection...... 99 7.2SK-IDataSideband...... 106 7.3SK-IDataEventRatevs.EventSelection...... 111 7.4BreakdownofRemainingBackground...... 118 7.5SK-IDataCandidates...... 120 0 → 0 0 7.6 Total Systematic Error of “KS π π ”Search...... 122
0 → + − 8Searchfor“KS π π Method #1” 124 8.1EventSelection...... 124 8.2SK-IDataSideband...... 130 8.3SK-IDataEventRatevs.EventSelection...... 133 8.4BreakdownofRemainingBackground...... 139 8.5SK-IDataCandidates...... 140 0 → + − 8.6 Total Systematic Error of “KS π π Method#1”Search..... 148 0 → + − 9Searchfor“KS π π Method #2” 150 9.1EventSelection...... 150 9.2SK-IDataSideband...... 154 9.3SK-IDataEventRatevs.EventSelection...... 157 9.4BreakdownofRemainingBackground...... 162 0 → + − 9.5 Total Systematic Error of “KS π π Method#2”Search..... 163
0 10 Search for “KL” 165 10.1EventSelection...... 165 10.2SK-IDataSideband...... 175 10.3SK-IDataEventRatevs.EventSelection...... 180 10.4BreakdownofRemainingBackground...... 188 10.5SK-IDataCandidates...... 189 0 10.6 Total Systematic Error of “KL”Search...... 191
11 Results 193
12 Comparison With Previous SK p → μ+K0 Search 196
iv 13 Future Prospects 202
14 Conclusion 204
Appendices 205 A SystematicErrors...... 205 A.1 DataSet...... 205 A.2 SourcesofError...... 206
Bibliography 217
v LIST OF FIGURES
Page
1.1 FeynmandiagramsofSU(5)GUTgaugebosons...... 7 1.2 Feynman diagram of Higgsino exchange from d=5 operator in the SUSYGUTLagrangian...... 9 1.3 Feynman diagram of p → νK¯ + decay...... 9
2.1 RelativeCherenkovlightspectruminpurewater...... 15 2.2 TheSuperKamiokandedetector...... 17 2.3 The50cmPMT...... 19 2.4 Thequantumefficiencyofthe50cmPMT...... 19 2.5 Thetransittimespreadofthe50cmPMT...... 20 2.6 SchematicoftheSKIDDAQsystem...... 22 2.7 SchematicoftheSKwaterpurificationsystem...... 24 2.8 Timingcalibrationsystem...... 25 2.9 RelativegaincalibrationsystemusingXelamp...... 26 2.10 AbsolutegaincalibrationusingaNi+Cfsource...... 27 2.11 Waterattenuationlengthmeasurementsystem...... 28 2.12 Distributionofcorrectedp.e.vs.photontravellength...... 30 2.13 Momentumdistributionfordecayelectrons...... 31 2.14 Invariant mass distribution of neutrino induced π0 events...... 32 2.15 Distribution of reconstructed momentum from observed charge vs. reconstructed Cherenkov opening angle for cosmic ray stopping μ.. 33 2.16 Ratio of the muon momentum from observed p.e. to momentum from openingangleasafunctionofthemomentum...... 34 2.17 The ratio of momentum/range as a function of range for cosmic ray stopping μ...... 35 2.18 Timevariationofenergyscalecalibration...... 36 2.19 The non-uniformity of the detector gain as a function of zenith angle. 37 2.20 Summaryoftheabsoluteenergyscalecalibration...... 37
3.1 Nucleon momentum distributions of 12C...... 39 3.2 Invariant mass of μ+K0 in the p → μ+K0 MC simulation in water. . 40 3.3 Primary cosmic ray flux measurements compared with the model used intheHondafluxcalculation...... 43 3.4 ThepredictedatmosphericneutrinofluxesattheSKsite...... 43 3.5 Quasi-elastic scattering cross-sections of ν andν ¯...... 46
vi 3.6 Cross-sections for charged current single pion productions of νμ... 48 3.7 Cross-sections for charged current νμ andν ¯μ interactions...... 50 3.8 The cross-sections of coherent pion productions off the carbon nucleus forCCandNCinteractions...... 52 3.9 Cross-sections of π+-16O scattering as a function of π+ momentum. 53 3.10 Wavelength dependence of photon attenuation coefficients used in MCsimulation...... 55
5.1 The reconstructed-true muon momentum of three control samples describedinthetext...... 81 5.2 Performance of the initial vertex fitting of the μ-candidate for p → + 0 μ KL MC...... 88 5.3 The ΔV distribution of the initial μ-candidate vertex for atmospheric neutrinoMC...... 88 → + 0 5.4 The PID likelihood of the true primary muon for p μ KL MC. . 89 5.5 Performance of the MS vertex fitting of the μ-candidate for p → + 0 μ KL MC...... 90 5.6 The ΔV distribution of the μ-candidate MS vertex for atmospheric neutrinoMC...... 91 5.7 The reconstructed-true momentum of the true primary muon for p → + 0 μ KL MC...... 91 5.8 Performance of the remaining particle vertex fitting of the μ-candidate → + 0 for p μ KL MC...... 93 5.9 The ΔV− distribution for the remaining particle vertex of the atmo- sphericneutrinoMC...... 94
6.1 The p → μ+K0 combinedsearchalgorithm...... 97
7.1 The nring distribution for p → μ+K0 and atmospheric neutrino MC 0 → 0 0 along the KS π π searcheventselection...... 102 7.2 The PID likelihood distribution for p → μ+K0 and atmospheric neu- 0 → 0 0 trino MC along the KS π π searcheventselection...... 103 7.3 The number of decay electron distribution for p → μ+K0 and atmo- 0 → 0 0 spheric neutrino MC along the KS π π search event selection. . 103 → + 0 7.4 The MKS distribution for p μ K and atmospheric neutrino MC 0 → 0 0 along the KS π π searcheventselection...... 104 + 0 7.5 The pμ distribution for p → μ K and atmospheric neutrino MC 0 → 0 0 along the KS π π searcheventselection...... 104 + 0 7.6 The Ptot distribution for p → μ K and atmospheric neutrino MC 0 → 0 0 along the KS π π searcheventselection...... 105 + 0 7.7 The Mtot distribution for p → μ K and atmospheric neutrino MC 0 → 0 0 along the KS π π searcheventselection...... 105 7.8 The number of decay electron distribution for data and atmospheric 0 → 0 0 neutrino MC using the KS π π searchsidebandsample...... 108
vii 7.9 The MKS distribution for data and atmospheric neutrino MC using 0 → 0 0 the KS π π searchsidebandsample...... 108 7.10 The pμ distribution for data and atmospheric neutrino MC using the 0 → 0 0 KS π π searchsidebandsample...... 109 7.11 The Ptot distribution for data and atmospheric neutrino MC using 0 → 0 0 the KS π π searchsidebandsample...... 109 7.12 The Mtot distribution for data and atmospheric neutrino MC using 0 → 0 0 the KS π π searchsidebandsample...... 110 0 → 0 0 7.13 The event rate along the KS π π searcheventselection...... 113 7.14 The nring distribution for data and atmospheric neutrino MC along 0 → 0 0 the KS π π searcheventselection...... 114 7.15 The PID likelihood distribution for data and atmospheric neutrino 0 → 0 0 MC along the KS π π searcheventselection...... 114 7.16 The number of decay electron distribution for data and atmospheric 0 → 0 0 neutrino MC along the KS π π searcheventselection...... 115
7.17 The MKS distribution for data and atmospheric neutrino MC along 0 → 0 0 the KS π π searcheventselection...... 115 7.18 The pμ distribution for data and atmospheric neutrino MC along the 0 → 0 0 KS π π searcheventselection...... 116 7.19 The Ptot distribution for data and atmospheric neutrino MC along 0 → 0 0 the KS π π searcheventselection...... 116 7.20 The Mtot distribution for data and atmospheric neutrino MC along 0 → 0 0 the KS π π searcheventselection...... 117 7.21 The Mtot vs. Ptot scatter plot for data, atmospheric neutrino MC and → + 0 0 → 0 0 p μ K MC along the KS π π searcheventselection..... 117 0 → 0 0 7.22 Event display of the candidate for the KS π π search...... 121 8.1 The nring distribution for p → μ+K0 and atmospheric neutrino MC 0 → + − along the KS π π Method#1searcheventselection...... 126 8.2 The PID likelihood distribution for p → μ+K0 and atmospheric neu- 0 → + − trino MC along the KS π π Method #1 search event selection. 127 8.3 The number of decay electron distribution for p → μ+K0 and atmo- 0 → + − spheric neutrino MC along the KS π π Method #1 search event selection...... 128 + 0 8.4 The pμ distribution for p → μ K and atmospheric neutrino MC 0 → + − along the KS π π Method#1searcheventselection...... 128 + 0 8.5 The Ptot distribution for p → μ K and atmospheric neutrino MC 0 → + − along the KS π π Method#1searcheventselection...... 129 8.6 The number of decay electron distribution for data and atmospheric 0 → + − neutrino MC using the KS π π Method #1 search sideband sample...... 131 8.7 The pμ distribution for data and atmospheric neutrino MC using the 0 → + − KS π π Method#1searchsidebandsample...... 131 8.8 The Ptot distribution for data and atmospheric neutrino MC using 0 → + − the KS π π Method#1searchsidebandsample...... 132
viii 0 → + − 8.9 The event rate along the KS π π Method #1 search event selec- tion...... 135 8.10 The nring distribution for data and atmospheric neutrino MC along 0 → + − the KS π π Method#1searcheventselection...... 135 8.11 The PID likelihood distribution for data and atmospheric neutrino 0 → + − MC along the KS π π Method #1 search event selection. . . . 136 8.12 The number of decay electron distribution for data and atmospheric 0 → + − neutrino MC along the KS π π Method #1 search event selection. 136 8.13 The pμ distribution for data and atmospheric neutrino MC along the 0 → + − KS π π Method#1searcheventselection...... 137 8.14 The Ptot distribution for data and atmospheric neutrino MC along 0 → + − the KS π π Method#1searcheventselection...... 137 8.15 The pμ vs. Ptot scatter plot for data, atmospheric neutrino MC and → + 0 0 → + − p μ K MC along the KS π π Method #1 search event selection...... 138 0 → + − 8.16 Event display of the first candidate for the KS π π Method #1 search...... 142 0 → + − 8.17 Event display of the second candidate for the KS π π Method #1search...... 143 0 → + − 8.18 Event display of the third candidate for the KS π π Method #1 search...... 144 0 → + − 8.19 Event display of the fourth candidate for the KS π π Method #1search...... 145 0 → + − 8.20 Event display of the fifth candidate for the KS π π Method #1 search...... 146 0 → + − 8.21 Event display of the sixth candidate for the KS π π Method #1 search...... 147
9.1 The number of decay electron distribution for p → μ+K0 and atmo- 0 → + − spheric neutrino MC along the KS π π Method #2 search event selection...... 152 → + 0 9.2 The MKS distribution for p μ K and atmospheric neutrino MC 0 → + − along the KS π π Method#2searcheventselection...... 152 + 0 9.3 The Ptot distribution for p → μ K and atmospheric neutrino MC 0 → + − along the KS π π Method#2searcheventselection...... 153 + 0 9.4 The Mtot distribution for p → μ K and atmospheric neutrino MC 0 → + − along the KS π π Method#2searcheventselection...... 153 9.5 The number of decay electron distribution for data and atmospheric 0 → + − neutrino MC using the KS π π Method #2 search sideband sample...... 155
9.6 The MKS distribution for data and atmospheric neutrino MC using 0 → + − the KS π π Method#2searchsidebandsample...... 155 9.7 The Ptot distribution for data and atmospheric neutrino MC using 0 → + − the KS π π Method#2searchsidebandsample...... 156
ix 9.8 The Mtot distribution for data and atmospheric neutrino MC using 0 → + − the KS π π Method#2searchsidebandsample...... 156 0 → + − 9.9 The event rate along the KS π π Method #2 search event selec- tion...... 159 9.10 The number of decay electron distribution for data and atmospheric 0 → + − neutrino MC along the KS π π Method #2 search event selection. 159
9.11 The MKS distribution for data and atmospheric neutrino MC along 0 → + − the KS π π Method#2searcheventselection...... 160 9.12 The Ptot distribution for data and atmospheric neutrino MC along 0 → + − the KS π π Method#2searcheventselection...... 160 9.13 The Mtot distribution for data and atmospheric neutrino MC along 0 → + − the KS π π Method#2searcheventselection...... 161 9.14 The Mtot vs. Ptot scatter plot for data, atmospheric neutrino MC → + 0 0 → + − and p μ K MC along the KS π π Method #2 search event selection...... 161
10.1 The potot distribution for p → μ+K0 and atmospheric neutrino MC 0 along the KL searcheventselection...... 169 → + 0 10.2 The nringt distribution for p μ K and atmospheric neutrino MC 0 along the KL searcheventselection...... 170 10.3 The PID likelihood distribution for p → μ+K0 and atmospheric neu- 0 trino MC along the KL searcheventselection...... 171 10.4 The number of decay electron distribution for p → μ+K0 and atmo- 0 spheric neutrino MC along the KL searcheventselection...... 172 + 0 10.5 The pμ−cand. distribution for p → μ K and atmospheric neutrino 0 MC along the KL searcheventselection...... 172 + 0 10.6 The vsep. distribution for p → μ K and atmospheric neutrino MC 0 along the KL searcheventselection...... 173 10.7 The pattern PID likelihood distribution for p → μ+K0 and atmo- 0 spheric neutrino MC along the KL searcheventselection...... 173 10.8 The μ-candidate Cherenkov angle distribution for p → μ+K0 and 0 atmospheric neutrino MC along the KL search event selection. . . . 174 10.9 The proton ID likelihood distribution for p → μ+K0 and atmospheric 0 neutrino MC along the KL searcheventselection...... 174 10.10 The number of decay electron distribution for data and atmospheric 0 neutrino MC in the KL searchsidebandsample...... 176 10.11 The pμ−cand. distribution for data and atmospheric neutrino MC in 0 the KL searchsidebandsample...... 177 10.12 The vsep. distribution for data and atmospheric neutrino MC in the 0 KL searchsidebandsample...... 177 10.13 The pattern PID likelihood distribution for data and atmospheric 0 neutrino MC in the KL searchsidebandsample...... 178 10.14 The μ-candidate Cherenkov angle distribution for data and atmo- 0 spheric neutrino MC in the KL searchsidebandsample...... 178
x 10.15 The proton ID likelihood distribution for data and atmospheric neu- 0 trino MC in the KL searchsidebandsample...... 179 0 10.16 The event rate along the KL searcheventselection...... 182 10.17 The potot distribution for data and atmospheric neutrino MC along 0 the KL searcheventselection...... 182 10.18 The nringt distribution for data and atmospheric neutrino MC along 0 the KL searcheventselection...... 183 10.19 The PID likelihood distribution for data and atmospheric neutrino 0 MC along the KL searcheventselection...... 184 10.20 The number of decay electron distribution for data and atmospheric 0 neutrino MC along the KL searcheventselection...... 185 10.21 The pμ−cand. distribution for data and atmospheric neutrino MC along 0 the KL searcheventselection...... 185 10.22 The vsep. distribution for data and atmospheric neutrino MC along 0 the KL searcheventselection...... 186 10.23 The pattern PID likelihood distribution for data and atmospheric 0 neutrino MC along the KL searcheventselection...... 186 10.24 The μ-candidate Cherenkov angle distribution for data and atmo- 0 spheric neutrino MC along the KL searcheventselection...... 187 10.25 The proton ID likelihood distribution for data and atmospheric neu- 0 trino MC along the KL searcheventselection...... 187 0 10.26 Event display of the candidate for the KL search...... 190 12.1 The total momentum vs. μ+ momentumscatterplot...... 199 12.2 Event display of first event selected by this analysis and not the referencedata...... 200 12.3 Event display of second event selected by this analysis and not the referencedata...... 200 12.4 The reconstructed-true μ-candidatemomentumdistribution..... 201
xi LIST OF TABLES
Page
0 1.1 KS decaybranchingratios...... 12 0 1.2 KL kinematicsinwater...... 13 7.1 Data events, signal detection efficiency and background rate along the 0 → 0 0 KS π π searcheventselection...... 113 7.2 Breakdown of remaining atmospheric neutrino MC background for the 0 → 0 0 KS π π search...... 119 7.3 Summary of reconstructed information for the data candidate of the 0 → 0 0 KS π π search...... 120 0 → 0 0 7.4 Systematic errors on signal detection efficiency for the KS π π search...... 122 0 → 0 0 7.5 Systematic errors on background rate for the KS π π search. . . . 123 8.1 Data events, signal detection efficiency and background rate along the 0 → + − KS π π Method#1searcheventselection...... 134 8.2 Breakdown of remaining atmospheric neutrino MC background for the 0 → + − KS π π Method#1search...... 139 8.3 Summary of reconstructed information for the data candidates of the 0 → + − KS π π Method#1search...... 141 0 → + − 8.4 Systematic errors on signal detection efficiency for the KS π π Method#1search...... 148 0 → + − 8.5 Systematic errors on background rate for the KS π π Method #1 search...... 149
9.1 Data events, signal detection efficiency and background rate along the 0 → + − KS π π Method#2searcheventselection...... 158 9.2 Breakdown of remaining atmospheric neutrino MC background for 0 → + − KS π π Method#2search...... 162 0 → + − 9.3 Systematic errors on signal detection efficiency for the KS π π Method#2search...... 163 0 → + − 9.4 Systematic errors on background rate for the KS π π Method #2 search...... 164
10.1 Data events, signal detection efficiency and background rate along the 0 KL searcheventselection...... 181
xii 10.2 Breakdown of remaining atmospheric neutrino MC background for the 0 KL search...... 188 10.3 Summary of reconstructed information for the data candidate of the 0 KL search...... 189 0 10.4 Systematic errors on signal detection efficiency for the KL search. . . 191 0 10.5 Systematic errors on background rate for the KL search...... 192 11.1Resultofthelowerlifetimelimitcalculation...... 195
→ + 0 12.1 Summary of the p μ KS search results from Kobayashi [1] and this analysis...... 198
xiii ACKNOWLEDGMENTS
First, I would like to express my deepest gratitude to Dr. Shun’ichi Mine. Without his dedication and very supportive guidance this thesis would not be possible. I also would like to thank my advisor Prof. Dave Casper for providing many suggestions for this analysis as well as Prof. Hank Sobel for his unwavering support of my work. I want to thank Prof. Yoichiro Suzuki, the spokesman of the Super Kamiokande experiment and I would like to thank especially Prof. Masato Shiozawa, Prof. Ed Kearns, Prof. Kenji Kaneyuki, Prof. Takaaki Kajita and Prof. Chris Walter for all of their generous suggestions. I am also grateful to all members of the atmospheric and proton decay analysis group at ICRR including Y. Hayato, C. Ishihara, J. Kameda, M. Miura, S. Moriyama, S. Nakayama, H. Nishino, Y. Obayashi, K. Okumura, N. Tanimoto. Much of the analysis within this thesis would not have been possible without the hard work and dedication of the Super Kamiokande collaboration to whom I am very thankful. I also want to thank Parker Cravens, Mike Litos, Jen Raaf and Roger Wendell for their assistance as well as the many great adventures we had in Japan. I would like to thank my fellow students that I studied with during my graduate career: Shane Curry, Mike Hood, Shiu Liu, Parker Lund, Matt Teig and Erik Trask. I always enjoyed talking over the latest homework problems with you all and enjoying our time together. This work would not be possible without the funding made possible by the US Department of Energy and the cooperation of the Kamioka Mining and Smelting Company. Finally, I’d like to thank my family for always being supportive of me during this long difficult path.
xiv CURRICULUM VITAE
Christopher Regis
Education:
2001 B.S., Physics, University of California, Santa Barbara
2004 M.S., Physics, University of California, Irvine
2011 Ph.D., Physics, University of California, Irvine
Awards and Honors:
2004-2005 Outstanding Contributions to the Department Award
xv ABSTRACT OF THE DISSERTATION
A search for proton decay via p → μ+K0 in Super Kamiokande I
By
Christopher Regis
Doctor of Philosophy in Physics
University of California, Irvine, 2011
Professor David Casper, Chair
Asearchforp → μ+K0 was performed using the Super Kamiokande I (SK-I) data set with a live-time of 1489.2 days corresponding to a total exposure of 91.2ktyears.
→ + 0 In addition to searching this decay mode via the p μ KS channel, previously performed by Kobayashi using the SK-I data set [1], new event reconstruction was → + 0 developed specifically to search for separated vertices characteristic of the p μ KL
0 channel. This new reconstruction takes advantage of the long lifetime of the KL which results in a unique event signature in which there are two distinct vertices; one
0 corresponding to the proton decay point and another at the KL disappearance point. These are typically separated by about 2 m in space and 20 ns in time.
We did not observe statistically significant evidence for p → μ+K0 decay. There-
33 fore, a lower limit on the proton partial lifetime, τ/Bp→μ+K0 ,of1.1 × 10 years was obtained at 90% confidence level.
xvi Chapter 1
Introduction
1.1 Theoretical Motivation
The Standard Model (SM) of particle physics is a theoretical construct that attempts to explain the interactions between the fundamental particles that make up the uni- verse. It generates electromagnetic, weak and strong interactions via corresponding
gauge symmetries mediated by the corresponding gauge fields, W ±, Z0 and photon, A. Although this model has been very successful in accurately describing results from a multitude of experiments, there are some limitations to the theory. As one example, the SM does not explain the observation of non-zero neutrino mass.
In an attempt to rectify this situation, as well as motivated by the observation of the apparent merging of the electromagnetic, strong and weak coupling constants at higher energies, Grand Unified Theories (GUTs) have been developed which attempt to unify these three fundamental forces under a larger gauge symmetry. One typical feature that arises during the construction of these GUTs that is not present in the SM is proton decay. Grand unification is not without fault however, particularly the so-called “hierarchy problem”, the inelegant necessity to introduce careful tuning of parameters in the calculation of the Higgs mass. One method to solve this problem
1 2 is known as super-symmetry (SUSY) in which a super-symmetric partner for every particle in the SM is introduced. In some of these SUSY GUT models p → μ+K0 is one of the dominant expected proton decay modes [2]. Details of the SM, GUTs and SUSY GUTs are described sections 1.1.2, 1.1.3 and 1.1.3.2 respectively.
1.1.1 Conservation and Symmetries
In theoretical physics, one of the most useful tools is known as Noether’s theorem which states that there exists a conservation law for every continuous symmetry of the
Lagrangian that describes a physical system [3]. Some examples of this are momentum and energy conservation which are due to the invariance of the Lagrangian under translations in space and time. However, there are other quantities such as baryon number which seem to be conserved but have no underlying symmetry. As baryon
number is not a conserved quantity in proton decay, a search for proton decay is a experimental test of the validity of baryon number conservation. To describe these symmetries in a mathematical formalism, group theory is used.
1 An example of this is the SU(2) group used to describe spin- 2 systems in quantum mechanics. The commutation relation between the generators completely describes the group.
[Li,Lj]=iijkLk (1.1)
where ijk is the completely anti-symmetric tensor and (i, j, k) run over indices 1, 2,
3. The generators, Li, can be expressed in terms of the traceless, unitary Pauli spin
σi matrices, σi,asLi = 2 . These matrices span the SU(2) space. Any arbitrary unitary
matrix M with determinant = 1 can be expressed in terms of Li as:
M = eiαi·L i (1.2) 3 where α is an arbitrary scale vector. The two physical states of this system are the orthogonal eigenstates up |+ and down |− which form a basis of the SU(2) space. In SU(2) there are two important operators:
L± =(L1 ∓ iL2) (1.3)
these represent the mixing between the states L±|∓ = |±. The complete system can now be described in terms of the generator matrix, L, and corresponding doublet,
ψ: ⎡ ⎤ ⎛ ⎞
1 ⎢Lz L+ ⎥ ⎜|+⎟ L = ⎣ ⎦ ψ → ⎝ ⎠ (1.4) 2 L− −Lz |−
The diagonal terms of L represent the projection of ψ and the off-diagonal terms represent transformations between eigenstates. Now that we have this formalism, we can make an analogy to particle physics. For instance, the doublet can represent a doublet of fermions such as an electron and the νe while the generators, Lz and
0 ± L±, correspond to the gauge bosons (Z and W for example). This gives rise to interactions which allow the states to mix; for instance changing the electron to a νe via the exchange of a W−.
1.1.2 Standard Model
In the SM of particle physics the fundamental particles and interactions are described
by the SU(3)c ⊗ SU(2)L ⊗ U(1)Y gauge symmetry. The SU(3)c gauge symmetry describes strong interactions between quarks. In this group, the quarks have color charge and form a color triplet, ψc, of three states: red, green and blue. In this case the mediators of the strong force are called gluons and are represented as the 4
generator matrix, Ggluon: ⎡ ⎤ ⎛ ⎞
⎢grr¯ grg¯ gr¯b⎥ ⎜ red ⎟ ⎢ ⎥ ⎜ ⎟ ⎢ ⎥ → ⎜ ⎟ Ggluon = ⎢g g g ¯⎥ ψc ⎜green⎟ (1.5) ⎣ gr¯ gg¯ gb⎦ ⎝ ⎠
gbr¯ gbg¯ gb¯b blue
− where gb¯b = grr¯ ggg¯.
The SU(2)L ⊗ U(1)Y gauge symmetry describes the unification of the electromag- netic and weak interaction into electro-weak theory. Under this symmetry, left handed fermions form SU(2)L doublets with weak isospin, T3, and right-handed fermions are singlets. Each fermion has a weak hypercharge, Y ,fromtheU(1)Y symmetry and is assigned as:
Y =2(Q − T3) (1.6) where Q is the electrical charge of the fermion.
The SU(2)L ⊗ U(1)Y gauge symmetry has four generators expressed as the mass-
1 2 3 less gauge bosons, W ,W ,W from SU(2)L and B from U(1)Y . However this is inconsistent with the experimental observation of three massive gauge bosons, W ±
and Z0 that mediate the weak interaction and the massless photon, A, that mediates the electromagnetic force. One method to resolve this issue is to spontaneously break the SU(2)L ⊗ U(1)Y gauge symmetry via the Higgs mechanism by introduction of a scalar particle known as the Higgs boson. This generates mass for the W ± and
Z0 bosons which are formed from the massless gauge bosons parameterized by the
Weinberg mixing angle θw as:
0 3 Z = W cos θw − B sin θw (1.7)
3 A = W sin θw + B cos θw (1.8) 5
Although the SM is a very successful theory, it does not represent a complete theory of particle physics. Some reasons for this include: the arbitrary assignment of the weak hypercharge to each fermion, the lack of an explanation for the factor of three relationship in charge quantization between quark and leptons, the observation of non-zero neutrino mass and lack of any description of gravity.
1.1.3 Grand Unified Theories
Grand unification is the attempt to unify the strong, weak and electromagnetic in-
teractions under a larger GUT gauge symmetry. One of the earliest indications of this unification is the apparent convergence of the running coupling constants that represent the strength of each of these interactions at an energy scale known as the GUT scale. Furthermore, since the SM has been so successful, it suggests that the
SM is a result of spontaneous symmetry breaking of the GUT gauge symmetry at the GUT energy scale.
1.1.3.1 Minimal SU(5)
One of the earliest attempts at a GUT was proposed by Georgi and Glashow using the
SU(5) group, the smallest gauge group that can contain the SU(3)c ⊗SU(2)L ⊗U(1)Y group of the SM [4]. The generators of this group are written as:
⎡ ⎤ 2 g − √ Bg g ¯ X Y ⎢ rr¯ 30 rg¯ rb 1 1 ⎥ ⎢ ⎥ ⎢ − √2 ⎥ ⎢ ggr¯ ggg¯ Bgg¯b X2 Y2 ⎥ ⎢ 30 ⎥ ⎢ 2 ⎥ VSU(5) = ⎢ g g g ¯ − √ BX Y ⎥ ⎢ br¯ bg¯ bb 30 3 3 ⎥ ⎢ ⎥ ⎢ X¯ X¯ X¯ √1 W 3 + √3 BW+ ⎥ ⎣ 1 2 3 2 30 ⎦ Y¯ Y¯ Y¯ W − − √1 W 3 + √3 B 1 2 3 2 30 (1.9) 6 with the fermions contained in 5¯ and 10 representations as:
⎛ ⎞ ⎡ ⎤ ¯ dr ⎢ 0¯ub −u¯g −ur −dr ⎥ ⎜ ⎟ ⎢ ⎥ ⎜ ⎟ ⎢ ⎥ ⎜ ¯ ⎟ ⎢ −u¯ 0¯u −u −d ⎥ ⎜ dg ⎟ ⎢ b r g g ⎥ ⎜ ⎟ ⎢ ⎥ 5¯ = ⎜ ¯ ⎟ 10 = ⎢ − − − ⎥ (1.10) ⎜ db ⎟ ⎢ u¯g u¯r 0 ub db ⎥ ⎜ ⎟ ⎢ ⎥ ⎜ − ⎟ ⎢ ⎥ ⎜ e ⎟ ⎢ u u u 0 e+ ⎥ ⎝ ⎠ ⎣ r g b ⎦ + −νe dr dg db −e 0 L L
± 3 The gcc¯ are the gluons of SU(3)c,theW , W and B are the gauge bosons of the electroweak theory of the SM. In this GUT there are twelve new bosons X and Y
15 with mass MX,Y ∼ 10 GeV, the GUT energy scale, which do not appear in the standard model. These bosons have both color and flavor and mediate interactions between the leptons and quarks. These interactions, whose Feynman diagrams are illustrated in Figure 1.1, do not conserve baryon number (B) or lepton number (L) but do conserve (B-L) and lead to nucleon decay via the exchange of one of these new bosons with a lifetime proportional to:
4 ∼ 1 MX τ 2 5 (1.11) αGUT mp
where αGUT is the value of the coupling constant at the GUT scale. The favored decay mode for SU(5) is p → e+π0 with a calculated partial lifetime
29±0.7 of τ/Bp→eπ0 =3.7 × 10 years [5]. However, this GUT model has been ruled out with the most recent lower lifetime limit set for this decay mode at 8.2×1033 years [6]. An additional difficulty with the SU(5) model is the value of the weak mixing angle
2 measured by experiment of sin θW =0.23108 ± 0.00005 [7] is inconsistent with the
2 0.0037 predicted value of sin θW =0.2102−0.0031 [8]. 7
dα e+ uα e+ dα ν¯
Xα Y α Y α
uβ u¯γ dβ u¯γ
Xα Y α
Figure 1.1: Feynman diagrams representing the interactions mediated by the Xα & Y α gauge bosons of the SU(5) GUT.
1.1.3.2 Supersymmetric GUTS
Assuming the existence of a GUT, when calculating the mass of the electro-weak
2 ∼ 2 Higgs boson of mH (250 GeV) , the radiative corrections introduce an enormous correction ∼ 1030 GeV2 and so extremely fine tuning of this correction is necessary. This is known as the “hierarchy problem”. One method to solve this is introduction of a symmetry between fermions and bosons known as SUSY. In SUSY every fermion of the SM gets a bosonic partner with the same mass but spin 1/2 different and vice versa. In this way, the radiative corrections are canceled by the new SUSY particles.
In addition to solving the hierarchy problem, in a SUSY GUT MX,Y is predicted to be 1016 GeV instead of 1015 GeV in the non-SUSY case thus increasing the proton decay lifetime by four orders of magnitude from (1.11) which is consistent with the experimental lifetime limits on p → e+π0. Furthermore, in SUSY GUTs the pre-
2 +0.0031 diction of the weak mixing angle is sin θW =0.2334−0.0027 which is consistent with experimental data [8]. Although the p → e+π0 decay mode may be suppressed in SUSY GUT models, 8 there are other interactions available in the SUSY GUT Lagrangian which mediate nucleon decay. The most significant contributor is the dimension 5 (d=5) operator via the exchange of a SUSY Higgsino illustrated in the Feynman diagram shown in Figure 1.2. Interactions via this operator require the family of the final state be different than the family of the initial state. Since a proton is composed of quarks from the first family this means that interactions involving this operator require quarks from the second or third families. As the strange quark is the only one that is kinematically allowed, an anti-strange quark must be in the final state. This will bind to a spectator u or d quark forming a K meson. In minimal SU(5) SUSY, the favored decay modes are therefore p → νK¯ + shown in Figure 1.3 and n → νK¯ 0 and the lifetime of these decay modes is calculated to be ∼ 1030±2 years [9].
In addition to SUSY GUT models based on SU(5), others have been proposed basedontheSO(10) group [2]. One feature of these models is the neutrino can obtain mass provided a new set of color triplet fields is introduced. These fields generate a new set of d=5 operators which predict comparable rates for p → νK¯ + and p → μ+K0
of [2]:
Γ(p → μ+K0) ≈ (20% to 50%) Γ(p → νK¯ +) (1.12)
Given (1.12) and the predicted lifetime of p → νK¯ + of ∼ 1032∼34 years the expected lifetime of p → μ+K0 is:
33∼35 τ/Bp→μ+K0 ∼ 10 years (1.13) which is within reach of the SK sensitivity of ∼ 1033 years. 9
q˜ q l
W˜ H˜ q q q˜ Figure 1.2: Feynman diagram representing the exchange of a Higgsino from the d=5 operator in the SUSY GUT Lagrangian. d˜ u ν¯μ W˜ H˜ p d s¯ u˜ K+ u u
Figure 1.3: Feynman diagram representing the p → νK¯ + decay mode.
1.2 Past Experimental Measurements of p → μ+K0
1.2.1 Soudan 2
The Soudan 2 experiment was a 963 t (770 t fiducial) iron calorimeter experiment
located in the Soudan mine in Minnesota at a depth of 2100 meters water equivalent (m.w.e.). The active portion of the detector consisted of 1 m long 1.5 cm diameter drift tubes encased in a honeycomb matrix of 1.6 mm steel plates. This experiment took data from 1989-2001 corresponding to a fiducial mass exposure of 5.9kt-yr. This → + 0 → + 0 experiment searched both p μ KS and p μ KL modes and set a lower lifetime limit on each of 1.5 × 1032 years and 8.3 × 1031 years respectively giving a combined lower lifetime limit on p → μ+K0 of 1.2 × 1032 years [10].
1.2.2 Irvine Michigan Brookhaven (IMB)
The IMB experiment was an 8 kt (3.3 kt fiducial), rectangular, water Cherenkov de-
tector located at a depth of 1580 m.w.e. in the Fairpoint Mine in Ohio. It took data from 1982-1991. This experiment set a lower lifetime limit on p → μ+K0 of 10
× 32 → + 0 1.2 10 years [11] by searching only the p μ KS mode.
1.2.3 Kamiokande
The Kamiokande experiment was the precursor to the Super Kamiokande (SK) ex-
periment and also was located in the Mozumi Mine. The experiment was a water Cherenkov detector with a total mass of 3 kt surrounded by 948 50 cm photo-multiplier tubes (PMT) resulting in 20% photocathode coverage. It took data from 1986-1996. This experiment set a lower lifetime limit on p → μ+K0 of 1.2 × 1032 years [12] by → + 0 searching only the p μ KS mode.
1.2.4 Super Kamiokande
The current best lower lifetime limit on p → μ+K0 is 1.3 × 1033 years by Kobayashi
→ + 0 set using the SK-I data set searching the p μ KS mode only [1]. Currently, there → + 0 is no published result using p μ KL with SK data.
1.3 Overview of Experimental Method
The SK detector is the largest water Cherenkov detector in the world with a fiducial mass of 22.5 kt. Assuming the ideal but unrealistic case of 100% signal detection efficiency with 0 background and an exposure of 91.7 kt-yr from SK-I, upper limits on the proton decay lifetime of ∼ 1.3 × 1034 years are possible for decay modes that include final state particles above the Cherenkov threshold.
0 0 0 → + 0 The K is a combination of 50% KS and 50% KL so searches for p μ K → + 0 → + 0 typically involve searching for p μ KS and p μ KL separately. For this thesis, we search for p → μ+K0 in SK-I data by updating the p →
+ 0 μ KS mode search performed by Kobayashi [1]. We use the latest improved detector calibration and event reconstruction, use a new combined search analysis strategy 11
→ + 0 described in section 6.2 and introduce a new search of p μ KL described in Chapter 10.
0 0 The lifetime of the KS is 0.9 ns; kinematically in water the KS nearly always
0 decays. The KS decay modes along with their respective branching ratios are sum- marized in Table 1.1 [7].
→ + 0 For the search of p μ KS in this thesis, an exclusive search is performed for the 0 → 0 0 0 → + − KS π π and KS π π decay modes only; the remaining fraction is neglected. → + 0 The decay of the proton via p μ KL provides a unique event signature due
0 to the much longer lifetime of the KL of 51 ns. Because of this, events from this proton decay mode are characterized by two vertices that are separated in time and
∼ ∼ 0 space within the detector; typically 20 ns and 2 m respectively for KL decay. The primary vertex is located at the proton decay point the remaining vertex is the
0 location of the KL decay point. For this thesis a new event reconstruction, described in section 5.2, was developed to look for these separated vertices and use this information to help separate the proton decay signal from the atmospheric neutrino background.
0 In addition to decaying, the KL can hadronically interact in the water in flight. 0 → Additionally, the KL can undergo KL KS regeneration a process not previously included in the event simulation and was introduced for this analysis as described in sections 3.1.3 and 3.4.3. Using the p → μ+K0 monte-carlo (MC) and SK detector simulation described in sections 3.1 and 3.4 respectively, the branching fraction was determined for these kinematic processes and summarized in Table 1.2. We perform
→ + 0 0 an inclusive search of p μ KL which is independent of the KL kinematic process. In this thesis we present the results of a proton decay search via p → μ+K0 in
→ + 0 → + 0 the SK detector using the combined results of p μ KS and p μ KL searches. The presentation begins in Chapter 2 with a description of the SK detector, data acquisition and calibration. A description of the simulation of the proton decay, atmospheric neutrino interaction and the detector is given in Chapter 3. The data 12 reduction steps for the fully contained sample is described in Chapter 4. Chapter 5 details the event reconstruction. In Chapter 6 an overview of the proton decay search is presented; the results are shown in Chapters 7-11. A discussion of these results, including a comparison with the previous SK result, future prospects and conclusion are found in Chapters 12, 13 and 14 respectively. Finally, details of systematic error estimation is described in Appendix A.
0 KS decay mode Branching ratio Γ (%) 0 → + − KS π π 69.2 0 → 0 0 KS π π 30.7 0 → KS other 0.1
0 Table 1.1: KS decay branching ratios. 13
0 KL kinematic process in water Branching ratio Γ (%) 0 → 2 3 KL N N,α, H, H 16.5 0 → ± KL N NK 15.7 0 → + 0 KL p π Λ 10.4 K0 p → π0Σ+ 5.8 0 → L KL interaction 0 → + 0 KL p π Σ 5.8 72.2% 0 → − + KL n π Σ 5.5 0 → + − KL n π Σ 5.1 0 → 0 0 KL n π Σ 3.6 0 → 0 0 KL n π Λ 2.2 0 → KL N other 1.6 0 → ± ∓ KL π e ν 11.1 0 → 0 → ± ∓ KL decay KL π μ ν 7.0 0 → 0 27.5% KL 3π 5.7 0 → + − 0 KL π π π 3.7
KL → KS regeneration 0.1 0 → KL other 0.2
0 Table 1.2: KL kinematics in water. Chapter 2
Detector
2.1 Cherenkov Radiation
When a charged particle of mass, m, and momentum, p, travels faster than light in a medium it produces a coherent electromagnetic wavefront known as Cherenkov radiation or Cherenkov light. This wavefront moves outward at an angle with respect
to the particle direction known as the Cherenkov angle, θc. The Cherenkov angle, the index of refraction of the medium, n, and velocity of the particle, β = v/c,are related by the following: 1 1 m2 cos θ = = 1+ (2.1) c nβ n p2
The requirement that the charged particle travel faster than light in the medium for
Cherenkov light to be emitted leads to a minimum momentum threshold, pthresh:
m pthresh = √ (2.2) n2 − 1
14 15
Figure 2.1: Relative Cherenkov light spectrum in pure water. For comparison the quantum efficiency of the 50 cm PMT is also shown.
As the particle moves through the medium a unit length, dx, the unit number, dN, of Cherenkov photons emitted with unit wavelength, dλ,is:
d2N 1 1 =2πα 1 − (2.3) dxdλ (nβ)2 λ2 where α ≈ 1/137 is the fine structure constant. As shown in Figure 2.1 the SK detector is sensitive to wavelengths ∼ 300 nm to 600 nm. The number of photons within this wavelength range emitted along the track of a charged particle traveling close to the speed of light (β 1) in water (n 1.34) is
∼ 340 cm−1. The Cherenkov light that is emitted by the particle is projected onto the wall of the SK detector as a ring pattern. From this ring pattern the kinematics of the particle that generated the Cherenkov light are reconstructed. 16
2.2 Super Kamiokande Detector
The SK detector is a water Cherenkov detector located at 2700 m.w.e. in the Mozumi mine beneath Mt. Ikenoyama in Kamioka, Gifu Prefecture, Japan. The detector, diagrammed in Figure 2.2, consists of a stainless steel right cylinder tank 41.4m in height and 39.3 m in diameter filled with 50 kt of ultra-purified water. Within the tank, a stainless steel cylindrical superstructure divides the tank volume into three concentric cylindrical volumes. The inner-most volume is known as the Inner Detector (ID); it is 36.2 m in height and 33.8 m in diameter. Mounted to the inner wall of the superstructure are 11146 50 cm diameter PMTs which face the ID corresponding to 40% photocathode coverage. Lining the inner wall of the superstructure, filling the gaps between the ID PMTs is an opaque black plastic material that serves as an optical barrier between the ID and remaining tank volume. The outer-most volume is known as the Outer Detector (OD); it completely sur- rounds the ID and has a thickness of 1.95 m to 2.2 m. Mounted to the outer wall of the superstructure are 1885 20 cm diameter PMTs with 60 cm square wave-shifter plates that face the OD. This wall is lined with Tyvek bonded to black low density polyethylene which serves as an optical barrier to the ID. The outer-most wall of the OD is lined with DuPont Tyvek; a white reflective material which helps to increase the number of detected photons in the OD. The third volume is the 0.55 m thick dead space region between the ID and OD which is occupied by the superstructure that supports the PMTs as well as the coaxial cables that carry the PMT signals outside the water tank. The detector began taking data in April 1996. In July, 2001 after 5 years of observation, a period known as SK-I, the data taking was stopped for maintenance of the detector. In November, 2001 during the refilling of the detector after the maintenance work, an accident occurredinwhichanIDPMTatthebottomregion 17
Figure 2.2: The Super Kamiokande detector. of the tank imploded. This generated a shockwave in the water that caused a chain reaction which destroyed more than half of the PMTs in the detector. The detector was rebuilt with half the ID PMT density of SK-I and began to take data in October,
2002. This second phase of the detector is known as SK-II. SK-II continued until October, 2005 when work began to replace the PMTs that had been destroyed in the accident and bring the ID PMT density back to the SK-I level. This work completed in June, 2006 and the detector took data in a period known as SK-III. SK-III ended in
September, 2008 with the upgrade of the electronics and online systems in preparation of the T2K experiment. From the completion of the upgrade in October, 2008 through the time of this thesis writing the detector has continued taking data in a period known as SK-IV. In this thesis we present the results of the proton decay search using the SK-I data set. 18
2.3 Photomultiplier Tube
Installed in the ID are Hamamatsu R3600 PMTs with a photocathode 50 cm in diam- eter as shown in Figure 2.3. The bulbs of the PMTs are hand blown from borosilicate glass to a thickness of about 5 mm. The inner surface of the glass is coated with
Bialkali (Sb-K-Cs) photocathode designed so its quantum efficiency (QE) coincides with the peak of the Cherenkov light spectrum as shown in Figure 2.4. Within the PMT, an 11 stage Venetian blind style dynode multiplies the ejected photoelectron (p.e.) from the photocathode across an applied voltage of 1700 V to 2000 V creating a gain of 107. The OD PMTs are Hamamatsu R1408 PMTs with a 20 cm diameter photocath- ode. An acrylic wavelength shifting plate of 60 cm × 60cmdopedwith50mgl−1 of bis-MSB scintillator is attached to the bulbs of the OD PMTs. The scintillator within the plate absorbs ultraviolet light (UV) then emits scintillation light at a wavelength which matches the peak sensitivity of the PMTs. This improves the collection effi- ciency of the OD PMT by a factor of 1.5. The signals for both ID and OD PMTs are read out via a 70 m coaxial cable that passes through the detector dead space region to exit the top of the detector into one of four electronic huts where they are recorded by the electronics and data acquisition system.
2.4 Electronics & Data Acquisition
2.4.1 Inner Detector
A schematic of the data acquisition system (DAQ) for the ID PMT is shown in Figure 2.6. The ID PMT signals are collected by 12 channel analog timing modules (ATM) [13] that are housed in Tristan KEK Online (TKO) crates [14]. The ATM per- 19
Figure 2.3: The 50 cm diameter PMT Hamamatsu R3600 PMT.
0.2
Quantum efficiency 0.1
0 300 400 500 600 700 Wave length (nm) Figure 2.4: The quantum efficiency of the 50 cm PMT. 20
Figure 2.5: The transit time spread of the 50 cm PMT for single p.e. equivalent signals. This spread corresponds to the timing resolution of the PMT. forms an analog to digital conversion of the charge and timing of the PMT signal. The dynamic range for charge and timing for each channel is ∼ 450 pC (picoCoulumbs) with 0.2 pC resolution and ∼ 1300 ns with a resolution of 0.4 ns respectively. Within each channel of the ATM, the analog signal from the PMT is amplified 100 times then split into four separate signals. The first of these split signals is added to the PMTSUM which represents the analog sum of the 12 channels of the ATM. PMT- SUM is used for the FLASH ADC DAQ described elsewhere [15]. The second of the split signals is sent to a discriminator with a threshold of −1 mV corresponding to 0.25 p.e. When this threshold is reached, a 200 ns width 15 mV/channel HITSUM signal is generated which is used to determine the global triggering. The remaining two split signals are sent to a pair of time to analog converters (TAC) and charge to analog converters (QAC). If a global trigger is issued, then the information in the TAC and QAC is digitized. The purpose for two banks of TAC/QAC is to enable the recording of two successive events without dead time introduced by the digitiza- tion process. Events like this occur, for example, by the electron generated from a 21 decaying muon. In total there are 946 ATMs that are read out by 8 online computers via 48 Versa
Module Europe (VME) memory modules called Super-Memory Partner (SMP). The data collected by each SMP is sent to a online host computer where is assembled and merged with the other SMP data to make complete events.
2.4.2 Outer Detector
The electronics system for the OD consists of a charge to timing converter (QTC)
module which takes the OD PMT signal as input and generates a rectangular signal whose width is proportional to the total charge of the input signal. If the signal exceeds the threshold of 0.25 p.e., the QTC also generates a rectangular OD HITSUM signal and the PMT signal is digitized by a Lecroy 1877 multi-hit timing to digital
conversion (TDC) module. This TDC module has a dynamic range of 16 ñswitha resolution of 0.5 ns. This TDC data is then read out by a separate online computer system, transferred to the on-line host computer where it is merged with the ID data.
2.4.3 Trigger System
The HITSUM that is generated by each ATM and QTC is collected and added to
produce a total HITSUM for the ID and OD respectively. For the ID there are three triggers: High Energy (HE), Low Energy (LE) and Super Low Energy (SLE) that are activated if the total ID HITSUM exceeds a specified threshold. In SK-I the HE trigger threshold is set at −340 mV corresponding to 31 hits within a 200 ns time window. The LE threshold is −320 mV corresponding to 29 hits which is expected from a 5.7 MeV electron assuming 50% trigger efficiency. The SLE trigger threshold, implemented in May 1997 is set to 4.7 MeV equivalent. SLE trigger events are not used in this analysis. The OD has only one trigger that is generated when the total OD HITSUM reaches a threshold corresponding to 19 hit PMTs in the OD. 22
20-inch PMT
SCH ATM
x 20 x 240 ATM GONG 20-inch PMT
SCH ATM interface Ultra Sparc x 20 x 240 SMP ATM GONG SMP online CPU(slave) SMP SMP SMP SMP Super Memory Partner x 6
VME 20-inch PMT
SCH Ultra Sparc ATM online CPU(slave) x 20 x 240 ATM Ultra Sparc GONG Analog Timing Module online CPU(host) online CPU(slave) TKO Ultra Sparc FDDI Ultra sparc online CPU(slave)
Ultra Sparc
online CPU(slave)
Ultra Sparc FDDI online CPU(slave)
Ultra Sparc
online CPU(slave)
interface online CPU Ultra Sparc (slave) SMP Ultra Sparc SMP online CPU(slave) SMP SMP SMP SMP Super Memory Partner interface VME VME interrupt reg. 20-inch PMT TRG
SCH ATM TRIGGER
TRIGGER x 20 PROCESSOR x 240 HIT INFORMATION ATM GONG Analog Timing Module
TKO
PMT x 11200 ATM x ~1000 SMP x 48 online CPU(slave) x 9
Figure 2.6: Schematic of the SK ID DAQ system. 23
These four trigger types are sent to the TRG module where the trigger type and event number are recorded and trigger timing is determined with 20 ns accuracy. This information is then sent to the online-host computer where it is merged with the PMT data.
2.5 Water Purification System
The origin of the water that fills the SK detector is spring water from the Mozumi mine. Prior to entering the tank, the water is ultra purified using a sophisticated
water system shown schematically in Figure 2.7. The SK water system consists of
eight stages in the order as follows: ñ 1. 1 ñm filter: This rejects relatively larger particles of size > 1 m.
2. Heat exchanger: This maintains the water temperature near 14 ◦C to inhibit growth of bacteria within the water.
3. Ion Exchanger: This removes metal ions from the water.
4. Ultraviolet sterilizer: This kills any live bacteria in the water
5. Vacuum degasifier: This removes any gasses such as oxygen and radon.
6. Cartridge polisher: High performance ion exchanger.
7. Ultra filter: Removes smaller particles > 10 nm.
8. Reverse Osmosis: Removes organisms up to 100 molecular weight.
To maintain the water quality of the detector, the water in the tank is continuously recirculated at a rate of about 35 t/hour. After the purification process, the number
3 of particles larger than 0.2 ñm is reduced to 6 particles/cm , the light attenuation length is ∼ 100 m and the resistivity of the water is increased from about 11 MΩ cm to an average of 18.2MΩcm. 24
VACUUM DEGASIFIER PUMP CARTRIDGE POLISHER ULTRA MEMBRANE FILTER DEGASIFIER UV STERILIZER
HEAT FILTER HEAT PUMP EXCHANGER (1μm Nom.) EXCHANGER
PUMP
BUFFER REVERSE TANK OSMOSIS PUMP RN-LESS-AIR
RN-LESS-AIR REVERSE DISSOLVE TANK OSMOSIS SK TANK SUPER-KAMIOKANDE WATER PURIFICATION SYSTEM Figure 2.7: Schematic of the SK water purification system.
2.6 Calibration
2.6.1 Timing Calibration
Timing calibration of the PMTs is crucial to achieve the precision timing that is necessary for accurate vertex reconstruction. There are a two important quantities that must be measured for each PMT to ensure this precision timing is obtained. First, the transit time of the PMT and its ∼ 70 m signal cable. Second, the effect
known as time walk in which a larger signal will trigger the discriminator threshold earlier than a smaller signal. The system used to calibrate the PMT timing is diagrammed in Figure 2.8a. A diffuser ball is placed in the water tank. The diffuser ball is composed of a silica gel material called LUDOX made from 20 nm glass fragments. Inserted into the ball is
one end of an optical fiber with a TiO2 tip. The optical fiber connects the diffuser ball to a nitrogen dye laser that sends a ∼ 3 ns pulse of light at a wavelength of 384 nm. This wavelength corresponds to the peak quantum efficiency of the ID PMT. The intensity of the light is adjusted using neutral density (ND) optical filters. The TiO2 25
980 variable attenuation filter 970 Linear Scale Log Scale optical fiber N laser 2 generator 960 λ=384nm 950 monitor Super Kamiokande PMT 940 inner tank
optical fiber sig. 930 20’PMT T (nsec) diffuser 920 ball trigger elec. 910
sig. 900
sig. 890 diffuser tip(TiO ) DAQ 2 elec. LUDOX 20’PMT sig. 880 1 5 10 100 Q (p.e.) (a) (b)
Figure 2.8: (a) Schematic of the timing calibration system. (b) TQ-map, a timing vs. charge distribution for a single PMT. tip and LUDOX diffuse the light within the ball to create an isotropic light source. By changing the optical filter, data is taken at a range of intensities and the PMT timing is measured. A distribution of the measured timing as a function of the detected p.e. is made for each PMT. This distribution is known as TQ-map. Figure
2.8b shows an example for one PMT. Each point is data. The open circles represent the average timing for each p.e. bin. The timing of each PMT is then corrected using the TQ-map.
2.6.2 Relative Gain Calibration
Uniform response of the detector is necessary to determine the momentum of a particle without a systematic difference that is dependent on the vertex or direction within the tank. To achieve this, calibration data is taken and the PMT high voltage is adjusted to obtain a uniform gain for all PMTs. A diagram of the relative gain calibration system is shown in Figure 2.9a. Light from a Xe lamp passes through UV and ND optical filters before being injected into a scintillator ball located within the detector via optical fiber. The scintillator ball is 26
UV filter ND filter Xe Flash Lamp 1400 Optical fiber 1200 σ gain = 7.0% 1000
SK TANK 800 Photo Diode Photo Diode
Number of PMTs in Each Bin 600
Scintilator Ball ADC 400
PMT 200
Scintilator 0 Monitor 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 20inch Trigger Relative PMT Gain PMT (a) (b)
Figure 2.9: (a) Schematic of the relative gain calibration system using a Xe lamp. (b) Relative gain distribution of all ID PMTs as measured in 1996.
composed of an acrylic resin doped with MgO powder diffuser and BBOT scintillator. The scintillator absorbs the input UV light from the fiber and emits light with a peak
wavelength of 440 nm which corresponds to typical Cherenkov light. The intensity of the Xe source is monitored using two photodiodes and a PMT attached to scintillator material.
The relative gain, Gi,ofthei-th PMT is given by: Qi 2 li Gi = li exp (2.4) Q0f(θ) L
where Qi is the charge detected at the i-th PMT, f(θ) is the angular acceptance of the PMT, li is the distance from the scintillator ball to the i-th PMT, L is the water attenuation length and Q0 is a normalization factor. The high voltage of the PMTs are adjusted to give a minimum spread to the relative gain distribution shown in
Figure 2.9b. 27
mean 2.055pC
3 Ni wire +water 200mm
Cf 2
1
0 0510pC 190mm (a) (b)
Figure 2.10: (a) Diagram of the Ni+Cf gamma-ray calibration source. (b) Charge distribution of a typical ID PMT from the Ni+Cf gamma-ray source.
2.6.3 Absolute Gain Calibration
Absolute gain calibration is done to determine the conversion between charge mea- sured in pC by the PMT and the number of p.e. To do this we use the low energy gamma ray generated from neutron capture on Nickel nucleus to measure the single p.e. distribution. A diagram of the Nickel calibration source is shown in Figure 2.10a. Neutrons
that are produced from the fission of the 252Cf source thermalize in the water and are captured by Ni nuclei in the surrounding wire generating a 6 MeV to 9 MeV gamma. An example of the single photo electron distribution from this source is shown in Figure 2.10b. The measured conversion is:
2.055 pC = 1 p.e. 28
Figure 2.11: Schematic of the attenuation length measurement system using laser.
2.6.4 Water Transparency
The water transparency measurement is necessary to determine the water attenuation length. This parameter represents the effect of scattering and absorption of light by the water within the tank. There are two methods used to measure the attenuation length, the first uses a nitrogen dye laser and the second uses cosmic ray muons.
2.6.4.1 Laser Injection
The system diagrammed in Figure 2.11 shows the the configuration of the laser and CCD camera used to measure water attenuation length. Using a nitrogen laser in
conjunction with a dye, monochromatic light with wavelengths of 337 nm, 365 nm, 400 nm, 420 nm, 460 nm, 500 nm and 580 nm are generated. The light passes via optical fiber to a diffuser ball that is located within the SK tank. The intensity of the light from the diffuser ball as a function of distance is then measured by the CCD
camera located at the top of the detector. The intensity of the light from the laser is also monitored by the PMT. The attenuation length is then determined using the intensity from the diffuser ball normalized by the intensity measured at the PMT vs. the separation distance between the diffuser ball and the CCD camera. 29
2.6.4.2 Cosmic Ray Muon
Another method to measure the water attenuation length is to use vertical through- going cosmic ray muons which deposit energy at a nearly constant rate along the track. One advantage to this method is cosmic ray muons make up a part of the normal data taking so the water attenuation length can be measured continuously.
To select through-going muons, the following event selection is applied using the fitted entrance point (xin,yin,zin) and exit point (xout,yout,zout)ofthemuon:
(A) 50000 < total number of p.e. < 125000 ≡ 2 2 (B) rin xin + yin < 15.9m,zin > 18.1m ≡ 2 2 − (C) rout xout + yout < 15.9m,zout < 18.1m 2 2 (D) (xin − xout) +(yin − yout) < 5m
Criterion (A) roughly corresponds to the muon track length of 25 m to 63 m. The distance between the top and bottom wall is 39.2 m. This criterion also rejects en- ergetic cosmic ray muons causing hadronic interactions in the detector. By criteria (B) and (C), the entrance and exit points are required on the top and bottom wall, respectively. Criterion (D) selects vertically going muons. For events which pass this event selection, the corrected p.e., qcorr, at each PMT is calculated as:
1 q = Q × l × (2.5) corr f(θ)
Where Q is the detected p.e. at the PMT, l is the travel length of the Cherenkov pho- ton, f(θ) is the PMT acceptance as a function of incident angle θ. This distribution is fitted using the following exponential function:
1 q (l)=G × exp − (2.6) corr L 30 )) θ RUN 3106 P1 5.987 0.4609E-03 500 P2 -0.9487E-04 0.4444E-06
log(Ql/f( 400
300
200
0 1000 2000 3000 4000 5000
Figure 2.12: Distribution of corrected p.e. vs. photon travel length. where G is proportional to the PMT gain and L is the measured attenuation length. Figure 2.12 shows a typical distribution of the corrected p.e. as a function of photon travel length.
2.6.5 Absolute Energy Calibration
The momentum of a particle is determined using the charge information from the PMTs. Since the systematic error in the particle momentum can affect the proton decay search, it is essential to have an accurate measurement of the absolute energy scale. To calibrate the absolute energy, four methods using independent calibration sources with momentum ranges between a few 10 MeV and about 10 GeV are used.
2.6.5.1 Decay Electrons
Many electrons produced by the decay of stopping cosmic ray muons are detected in SK. These electrons have a well understood energy spectrum up to ∼ 53 MeV. The absolute energy scale is determined by comparing the observed data and MC simulation. To select the decay electron sample the following selection criteria is
used: ñ (A) The time interval from the stopping muon event is 1.5 ñsto8.0 s.
(B) The number of hit PMT in a 50 ns time window is larger than 60. 31
225
Events 200
175
150
125
100
75
50
25
0 0 1020304050607080 MOMENTUM (MeV/c)
Figure 2.13: Momentum distribution for decay electrons for data (point) and MC (line).
(C) The goodness of the vertex fit is greater than 0.5.
(D) The vertex position is reconstructed more than 2 m away from the ID wall.
Figure 2.13 shows the momentum spectra of decay electrons as compared to the prediction from the MC. In the MC the measured μ+/μ− ratio of 1.37 [16] is used and μ− capture in the oxygen nucleus is also taken into account. The mean values of the data agree with the MC within 0.6%.
2.6.5.2 Neutrino Induced π0 Events
Atmospheric neutrino events that generate π0 are used to calibrate the energy scale in the few hundred MeV range. A π0 will immediately decay into two gammas from
0 which the invariant π mass, Mπ0 , can be determined via:
Mπ0 = 2Pγ1Pγ2(1 − cos θ) (2.7)
Where Pγ1, Pγ2 and θ are the reconstructed momentum and opening angle between the gammas respectively. To select π0 events, the following selection criteria is applied: 32
140
s 120 t n ve 100 E f o
r
e 80 b m u N 60
40
20
0 0 50 100 150 200 250 300 π0 invariant mass (MeV/c2 ) Figure 2.14: Invariant mass distribution of neutrino induced π0 events for data (point) and MC (box).
(A) Two electron-like rings.
(B) The number of decay electron = 0.
(C) The vertex position is reconstructed more than 2 m away from the ID wall.
Criterion (B) rejects contamination to the sample from π±π0 or μ±π0 events. Figure
2.14 shows the Mπ0 distribution for the data and MC simulation. There is a clear peak near 135 MeV/c2, the true mass of the π0, the position of the peak in the data is 0.7% higher than the MC.
2.6.5.3 Cosmic Ray Stopping μ Cherenkov Angle
Another calibration method using stopping cosmic ray muons is to utilize the half opening angle of the Cherenkov ring. Equation (2.1) describes the dependence of the
Cherenkov ring opening angle, θ, on the muon momentum, pμ. Because |dθ/dpμ| is large for small pμ this calibration method is possible in the low momentum region of pμ 500 MeV/c. The criteria for selecting cosmic ray muons for this calibration are:
(A) Total p.e. < 1500. 33
DATA Monte Carlo 500 500 ) ) c c V/ V/
e 450 e 450 M M ( (
m 400 m 400 u u t t m m e 350 e 350 m m o o m m
. 300 . 300 t t s s on on 250 250 ec ec r r
200 200 30 32 34 36 38 40 42 44 30 32 34 36 38 40 42 44 opening angle (deg.) opening angle (deg.) (a) Data (b) MC
Figure 2.15: Scatter plots of the reconstructed momentum from observed charge vs. reconstructed Cherenkov opening angle for (a) data and (b) MC.
(B) One cluster of hit PMTs in the OD.
(C) The entrance point is on the top wall.
(D) The direction is downward (cos θzenith > 0.9).
(E) There is one decay electron.
Criterion (A) roughly corresponds to the muon momentum of < 380 MeV/c. Crite- rion (B) requires the entrance point of the muon in the OD. Criterion (C) and (D) require the muon is downward going. Figure 2.15 shows the scatter plots of the re- constructed Cherenkov opening angle and the momentum of the observed p.e. from the data and MC. The energy scale of the data is compared to the MC using the ratio
Pp.e./Pθ,wherePp.e. is the momentum estimated from the observed charge and Pθ is the momentum estimated from the Cherenkov angle. Figure 2.16 shows the aver-
aged Pp.e./Pθ for data and MC and the MC/data ratio as a function of the expected momentum Pθ. They agree within 1.1%. 34
) 1.4 θ
1.3
1.2 P(p.e.)/P(
1.1
1
0.9
0.8
0.7
0.6 200 250 300 350 400 450 500 P(θ) (MeV/c)
Figure 2.16: Ratio of the muon momentum from observed p.e. to momentum from opening angle (Pp.e./Pθ) as a function of the momentum (Pθ) for the data (solid line) and MC (dotted line).
2.6.5.4 Cosmic Ray Stopping μ Track Length
The momentum of high energy muons can be estimated using the track length since the range is approximately proportional to the momentum. We can estimate the range independent of the observed p.e. so we use the measurement of the momentum from the range for checking the energy scale from 1 GeV/c to 10 GeV/c. Higher momentum muons typically do not stop within the detector. The criteria for selecting cosmic ray muons for the calibration are:
(B) The entrance point is on the top wall.
(C) The direction is downward (cos θzenith > 0.94).
(D) One decay electron.
(E) Reconstructed muon range > 7m.
We determine the muon range as the distance from the entrance point to the recon- structed vertex of the decay electron. Figure 2.17 shows the averaged value of the ratio of momentum/range as a function of the range for data and MC. The energy 35
2.6
2.5
2.4
2.3 Momentum(MeV/c)/Range(cm)
2.2
2.1
2 0 500 1000 1500 2000 2500 3000 3500 Range (cm) Figure 2.17: The ratio of momentum/range as a function of range for data (solid line) and MC (dotted line) loss per cm is about 2.3 MeV/c. The energy scale of the data and MC agree within 0.7%.
2.6.5.5 Time Variation of Energy Scale
Using the stopping muon and decay electron events we can monitor the time variation of the energy scale. Figures 2.18a and 2.18b show the time variation of the decay elec- tron and momentum/range ratio respectively. The RMS of the energy scale variation is 0.83%.
2.6.5.6 Uniformity of Energy Scale
Non-uniformity in the energy scale can lead to inaccuracy in the total momentum reconstruction which is one of the important quantities used for proton decay searches. Decay electrons from stopping cosmic ray muons are a good calibration source of detector uniformity since their vertex and direction are nearly uniform in location and direction within the fiducial volume (FV). To avoid bias from muon polarization, only electrons whose direction is perpendicular to the parent muon are used. This is achieved by requiring −0.25 < cos Θeμ < 0.25, where cos Θeμ is the opening angle 36
1.05 1.04 1.03 1.02 1.01 1 0.99 0.98 0.97 0.96 0.95 0 500 1000 1500 2000 2500 3000 3500 elapsed days from 1996/Apr./1 (a)
1.05 1.04 1.03 1.02 1.01 1 0.99 0.98 0.97 0.96 0.95 0 500 1000 1500 2000 2500 3000 3500 elapsed days from 1996/Apr./1 (b)
Figure 2.18: The time variation of (a) decay electron momentum and (b) ratio of momentum/range of stopping muons. The SK-I run period consists of the block of data points between day 0 and ∼ 2000. between the electron and muon directions. Figure 2.19 shows the momentum of decay electrons for the MC normalized by the observed data as a function of the zenith angle of the electrons. The detector gain is uniform within ±0.6%.
2.6.5.7 Summary of Energy Scale Calibration
Figure 2.20 summarizes the absolute energy scale calibration. The energy scale uncer- tainly is estimated to be less than 0.74% over the momentum range of a few 10 MeV/c to about 10 GeV/c. Combining this result in quadrature with the RMS of the time variation, the systematic error of the energy scale is estimated to be 1.1%. 37
Figure 2.19: The non-uniformity of the detector gain as a function of zenith angle. The vertical axis is the momentum of the decay electron events for the MC normalized by the observed data.
SK-I 8 6 (%) 4 2
/DATA 0 -2 μ range -4 μ angle π0 mass -6 decaye (MC - DATA) - (MC -8 2 3 4 5 10 10 10 10 10 momentum range (MeV/c)
Figure 2.20: Summary of the absolute energy scale calibration. Chapter 3
Simulation
To estimate the detection efficiency of proton decays and the number of background events a detailed MC simulation program was developed. Details of the simulation of the proton decay, atmospheric neutrino flux, atmospheric neutrino interactions and
detector simulation are described below in sections 3.1, 3.2, 3.3, 3.4 respectively.
3.1 Proton Decay
The source of protons for the proton decay search is the water (H2O) in the SK detector. A water molecule consists of two free protons and eight protons bound in the oxygen nucleus. We assume equal probability of decay for any of these protons.
For the free protons, the kinematics of the p → μ+K0 decay products can be uniquely determined. The direction of the μ+ and K0 is opposite each other each with a momentum of 326.5 MeV/c. In the case of proton decay in oxygen, Fermi momentum, correlation with other nucleons, nuclear binding energy, kaon-nucleon interaction and 0 → 0 KL KS regeneration must be considered.
38 39
Figure 3.1: Nucleon momentum distributions. The left and right figures show mo- menta of 1s state and 1p state in 12C nucleus, respectively. Solid lines show theoretical calculations used in our simulation [17].
3.1.1 Fermi Momentum and Nuclear Binding Energy
We use the Fermi momentum and nuclear binding energy measured by electron-12C scattering [17]. Figure 3.1 shows the Fermi momentum distribution of the 1s state and 1p state for the experimental data. The theoretical calculation shown in the figure is used in our simulation.
Nuclear binding energy is taken into account by modifying the proton mass. The − modified proton mass, mp, is calculated by mp = mp Eb where mp is the proton rest mass and Eb is the nuclear binding energy. The measured nuclear binding energy is 39.0MeV/c2 and 15.5MeV/c2 for the 1s and 2p states respectively.
Ten percent of decaying protons have wavefunctions that are correlated with other nucleons within the nucleus [18]. These correlated decays cause the total invariant mass of the decay products to be smaller than the proton mass because of the momen- tum carried by the correlated nucleons. Figure 3.2 shows the invariant proton mass distribution in 16Ousedinthep → μ+K0 simulation. Correlated decays produce the broad spectrum below ∼ 850 MeV/c2 shown by the hatched region. 40
10 4 number of events
10 3
10 2
10
0 200 400 600 800 1000 MeV/c2 μ+K0 invariant mass in 16O Figure 3.2: Invariant mass of μ+K0 in the p → μ+K0 MC simulation in water (solid line). The hatched region shows the contribution from nucleon correlated decay. The two peaks near the proton mass in the distribution correspond to the 1s and 1p state which are separated by the difference in nuclear binding energy.
3.1.2 Kaon-Nucleon Interaction in the Oxygen Nucleus
The position of the decaying proton in 16O is calculated according to the Wood-saxon nuclear density model, ρp(r), which is expressed as [19]:
Z ρ0 ρp(r)= r−c (3.1) A 1+exp( a )
3 where r is the distance from the center of the nucleus, ρ0 =0.48mπ is the average nuclear density, 2a =0.82 fm is the thickness of the nuclear surface and c =2.69 fm is the average nuclear radius for 16O. The kaon nucleon interactions that are considered include elastic scattering and inelastic scattering via charge exchange. The type of interaction is determined using the calculated mean free path. The mean free path is calculated from the differential
0 ¯0 cross-sections using the results of KS and K scattering experiments [20, 21]. If there is an interaction, the Pauli blocking effect is taken into account by requiring the nucleon momentum after the interaction to be larger than the Fermi surface 41
momentum, pF , defined as: 3π2 p (r)= ρ(r) (3.2) F 2 where ρ(r) is the same as (3.1)
0 0 3.1.3 KL→KS Regeneration in the Oxygen Nucleus
Neutral kaons consist of a down quark and strange quark. There are four types of neutral kaons K0 (S=+1), K¯0 (S=-1) which are the eigenstates of strangeness and
0 0 KL, KS which are the weak eigenstates. They are related by the following:
K0 − K¯0 K0 + K¯0 K0 = √ K0 = √ (3.3) L 2 S 2
Since strong interactions conserve strangeness, there are more strong interactions
¯0 0 0 available for K than for K . Therefore, when a KL propagates through a material, the K0 and K¯0 components interact at different rates with the material which in-
0 0 troduces a KS component in the original KL. It is therefore possible to observe a
0 0 KS after the original KL has propagated some distance through a material. This is 0 → 0 known as KL KS regeneration. 0 → 0 KL KS regeneration in the oxygen nucleus is modeled in the simulation as a
0 decay channel of the KL that is parameterized by a regeneration lifetime in the oxygen nucleus, τRnucl. ; a tunable parameter that affects the probability of regeneration. We assume τ is related to τ (see section 3.4.3) by: Rnucl. RH2O
ρ τ = τ nucl. (3.4) Rnucl. RH2O ρH2O
× 17 −3 Where ρnucl. =3.7 10 kg m is the density of the oxygen nucleus and ρH2O = 1000 kg m−3 is the density of water. 42
3.2 Atmospheric Neutrino Flux
Interactions of atmospheric neutrinos in the SK detector are the primary source of irreducible background for a search of proton decay. To estimate this background the atmospheric neutrino flux and interactions are simulated. In our simulation of the
atmospheric neutrino flux we use the calculation of Honda et. al. [22] (Honda Flux) made at the SK site. The BESS [23] and AMS [24] experiments have precisely measured the primary cosmic ray flux up to 100 GeV. In the Honda flux calculation, the model of the
primary flux is a parametrization and fit of experimental data as shown in Figure 3.3. The effect of the solar wind and geomagnetic field of the Earth on the primary flux are considered in the Honda flux calculation. The difference of the flux at solar maximum and solar minimum is more than a factor of two for 1 GeV cosmic rays, while it decreases to ∼ 10% for 10 GeV. The simulation of hadronic interactions in the atmosphere are treated using two theoretical models depending on the primary cosmic ray energy. NUCRIN [25] is used for energies less than 5 GeV and DPMJET-III [26] is used for energies greater than
5 GeV. The neutrino flux is obtained from decays of secondary particles, primarily pions, that are produced in these interactions. The calculated energy spectrum of atmospheric neutrinos at the SK site is shown in Figure 3.4 with the predictions of G. Battistoni et. al. [27] (Fluka flux) and G. Barr et. al. [28] (Bartol flux). Neutrinos with energy on the order of 1 GeV contribute to the background for proton decay searches. The systematic error of the absolute flux was estimated to be 10% [29] stemming from the uncertainties of the absolute primary cosmic ray flux and cosmic ray interactions. 43
Figure 3.3: Primary cosmic ray flux measurements compared with the model used in the Honda flux calculation. The data are taken from Webber [30] (crosses), LEAP [31] (upward triangles), MASS1 [32] (open circles), CAPRICE [33] (vertical diamonds), IMAX [34] (downward triangles), BESS98 [23] (circles), AMS [24](squares), Ryan [35] (horizontal diamonds), JACEE [36] (downward open triangles), Ivanenko [37] (upward open triangles), Kawamura [38] (open squares) and Runjob [39] (open diamonds). GeV)
−1 ν μ + ν μ sr −1
2 10sec ν ν −2 e + e (m 2 ν Honda flux Fluka flux Flux*E 10 Bartol flux
1 10−1 1 10 102 Eν (GeV) Figure 3.4: The predicted atmospheric neutrino fluxes at the SK site. Solid lines show the Honda flux, dashed lines show the Fluka flux [27], and dotted lines show the Bartol flux [28]. 44
3.3 Neutrino Interaction
The atmospheric neutrinos interact with the nucleons and electrons in the SK detector which we simulate using NEUT [40, 41, 42]. The following types of charged current (CC) and neutral current (NC) neutrino interactions are considered in NEUT:
CC/NC (quasi-)elastic scattering: ν + N → l + N
CC/NC single meson production: ν + N → l + N +meson
CC/NC deep inelastic interaction: ν + N → l + N hadrons
CC/NC coherent π production: ν +16 O → l +16 O+π where ν is a neutrino or anti-neutrino, N is a nucleon and l is a lepton. Systematic uncertainties of neutrino interactions are also considered, the details of which are described elsewhere [29].
3.3.1 Elastic & Quasi-Elastic Scattering
The differential cross-section of the charged current quasi-elastic scattering for free protons is given by [43, 44]:
dσν(¯ν) M 2G2 cos2 θ s − u (s − u)2 F C 2 ∓ 2 2 2 = 2 A(q ) B(q ) 2 + C(q ) 4 (3.5) dq 8πEν M M
where Eν is the neutrino energy, M is the mass of the target nucleon, GF is the Fermi coupling constant, θC is the Cabbibo angle, q is the four-momentum transferred to the lepton and s and u are Mandelstam variables [43]. The factors A, B and C are 45 given by:
m2 − q2 q2 q2 A q2 − |F |2 − |F 1 |2 ( )= 2 4 2 A 4+ 2 V 4M M M q2 q2 4q2F 1 ξF2 − |ξF2 |2 − V V 2 V 1+ 2 2 M 4M M m2 − (F 1 + ξF2 )2 + |F |2 (3.6) M 2 V V A 2 2 q 1 2 B(q )= FA(FV + ξFV ) (3.7) M2 1 q2 C(q2)= |F |2 + |F 1 |2 − |ξF2 |2 (3.8) 4 A V 4M 2 V
where m is the charged lepton mass and ξ ≡ μp −μn =3.71. The vector form factors,
1 2 2 2 2 FV (q )andFV (q ), and the axial form factor, FA(q ) are determined experimentally and given by:
− q2 1 q2 F 1 (q2)= 1 − G (q2) − G (q2) (3.9) V 4M 2 E 4M 2 M − q2 1 ξF2 (q2)= 1 − G (q2) − G (q2) (3.10) V 4M 2 E M − q2 2 F (q2)=− 1.23 1 − (3.11) A M 2 A 2 −2 2 −1 2 q GE(q )=(1+ξ) GM (q )= 2 (3.12) MV
where GE and GM are the electric and magnetic form factor, the vector mass MV is
2 2 set to be 0.84 GeV/c and the axial vector mass MA is set to be 1.21 GeV/c ,which is a common parameter for the single meson production model. The Fermi motion of the nucleons and Pauli exclusion principle must be considered for scattering off nucleons in 16O [45]. Since nucleons are fermions, the outgoing momentum of the nucleons in the interactions is required to be greater than the Fermi surface momentum to allow quasi-elastic scattering to occur. In NEUT, the Fermi surface momentum is set to be 225 MeV/c. 46
2 2 ANL Serpukhov – GGM (1977) 1.8 (a) ν n→μ-p 1.8 (b) ν p→μ+n μ BNL GGM (1977) μ GGM (1979) 1.6 SKAT GGM (1979) 1.6 Serpukhov
) 1.4 ) 1.4 SKAT 2 2 1.2 1.2 cm cm 1 1 -38 -38 0.8 0.8 (10 (10
σ 0.6 σ 0.6 0.4 0.4 0.2 0.2 0 0 02468101214 02468101214
Eν (GeV) Eν (GeV) (a) ν (b)ν ¯
Figure 3.5: Crosssections of (a) ν and (b)ν ¯ with the experimental data from ANL [48], Gargamelle [49, 50, 51], BNL [52], Serpukhov [53] and SKAT [54].
The cross-section for neutral current elastic scattering are estimated from the following relations [46, 47]:
σ(νp → νp)=0.153 × σ(νn → e−p) (3.13)
σ(¯νp → νp¯ )=0.218 × σ(¯νn → e+n) (3.14)
σ(νn → νn)=1.5 × σ(νp → νp) (3.15)
σ(¯νn → νn¯ )=1.0 × σ(¯νp → νp¯ ) (3.16)
Figure 3.5 shows the cross-section of the quasi-elastic scattering for the experimental data and the calculation by NEUT.
3.3.2 Single Meson Production
The resonant single meson production of π, K,andη was simulated based on the model of Rein & Sehgal [55]. An intermediate baryon resonance is assumed in this method:
ν + N → l + N ∗ (3.17)
N ∗ → N + meson (3.18) 47 where N and N are nucleons, N ∗ is a baryon resonance, and l is a lepton. The differential cross-section of single meson production is a product of the amplitude of each resonance production and the probability of their baryon resonance decay to the meson. For a baryon resonance with a negligible decay width, the differential cross-section is:
d2σ 1 1 = × |T (νN → lN∗)|2δ(W 2 − M 2) (3.19) dq2dE 32πME2 2 j j ν ν j,spin
where M is the mass of the target nucleon, Eν is neutrino energy, W is the invariant mass of the hadronic system (or the mass of the intermediate baryon resonance), Mj is the mass of the baryon resonance, and T (νN → lN∗) is the amplitude of resonance production, which is calculated using the Feynman-Kislinger-Ravndal model [56]. The invariant mass, W , is restricted to be less than 2 GeV/c2.ForW>2GeV/c2,the interactions are simulated as deep inelastic scattering as described in section 3.3.3. For a baryon resonance with a finite decay width, the differential cross-section can be derived by replacing the δ-function with a Breit-Wigner factor:
1 Γ δ W 2 − M 2 → × ( j ) 2 2 (3.20) 2π (W − MJ ) +Γ /4
2 For a single meson production, the axial vector mass MA is also set to be 1.21 GeV/c . A total of 28 resonances were simulated in our simulation. The effect of the Pauli exclusion principle in the decay of the baryon resonance was considered by requiring the momentum of the scattered nucleon to be greater than the Fermi surface mo- mentum. The pion-less decay of Δ resonance in 16O nuclei, where about 20% of the events do not have a pion in the decay, was also simulated [57]. Figure 3.6 shows the cross-sections of charged current resonant single meson productions for our calcula- tions and the experimental data. Most of the baryon resonances decay to the final states including π. The production cross-section for the η meson and K meson are 48 ) ) 2 − 2 1 ν → μ π+ − 1.2 μ p p ν → μ π+ cm cm 0.9 μ n n −38 −38 0.8 )(10 )(10
μ 1 μ ν ν ( (
σ 0.7 σ
0.8 0.6
0.5 0.6 0.4
0.4 0.3
0.2 0.2 0.1
0 0 1 10 1 10 Eν(GeV) Eν(GeV) − + − + (a) νµp → μ pπ (b) νµn → μ nπ )
2 1 ν → μ− π0 cm 0.9 μ n p ANL Radecky, Phys.Rev.D 25, 1161 (1982) −38 0.8 ANL Campbell, Phys.Rev.Lett. 30, 225 (1973) )(10 μ ν ( 0.7 ANL Barish, Phys.Rev.D 19, 2521 (1979) σ
0.6 BEBC Allen Nucl.Phys.B 264, 221 (1986) 0.5 BEBC Allen Nucl.Phys.B 176, 269 (1980) 0.4 BEBC Allasia Nucl.Phys.B 343, 285 (1990) 0.3 BNL Kitagaki Phys.Rev.D 34, 2554 (1986) 0.2 FNAL Bell Phys.Rev.Lett. 41, 1008 (1978) 0.1
0 1 10 Eν(GeV) (d) List of experiments − 0 (c) νµn → μ pπ
Figure 3.6: Cross-sections for charged current single pion productions of νμ. Solid lines indicate our calculations. Experimental data are summarized in the panel (d). much less than that for π. The fractions for the single η meson and the K meson production in the total single meson production events in the atmospheric neutrino MC were approximately 4% and 0.6%, respectively.
3.3.3 Deep Inelastic Scattering (DIS)
The cross-section of charged current deep inelastic scattering is calculated by integrat-
ing the following equation in the range of the invariant mass W>1.3GeV/c2 [58]:
d2σν,ν¯ G2 M E y2 y = F N ν (1 − y + + C )F (x, q2) ± (1 − + C )xF (x, q2) (3.21) dxdy π 2 1 2 2 2 3 yM2 xyM m2 m2 C l − N − l − l 1 = 2 (3.22) 4MN Eν x 2En 4Eν 2MN Eνx 2 ml C2 = − (3.23) 4MN Eνx 49
2 where x(= −q /(2M(Eν − El))) and y(= (Eν − El)/Eν) are Bjorken scaling parame- ters, MN is the nucleon mass, ml is the outgoing lepton mass, Eν and El are the energy of the incoming neutrino and outgoing lepton in the laboratory frame respectively.
We use the nucleon structure functions F2 and xF3 taken from GRV98 [59]. To obtain the cross-sections for deep inelastic scattering induced by the neutral current, we use the following relations which are estimated from experimental re- sults [60]: ⎧ ⎪ ⎪ ⎪0.26 (Eν < 3GeV) ⎨⎪ σ(νN → νX) = − ≤ (3.24) → − ⎪0.26 + 0.04(Eν/3 1) (3 GeV Eν < 6GeV) σ(νN μ X) ⎪ ⎪ ⎩⎪ 0.30 (Eν ≥ 6GeV) ⎧ ⎪ ⎪ ⎪0.39 (Eν < 3GeV) ⎨⎪ σ(¯νN → νX¯ ) = − − ≤ (3.25) → + ⎪0.39 0.02(Eν/3 1) (3 GeV Eν < 6GeV) σ(¯νN μ X) ⎪ ⎪ ⎩⎪ 0.37 (Eν ≥ 6GeV)
To generate events, a combination of PYTHIA/JETSET [61] and custom software is used. Since PYTHIA/JETSET is intended for simulation at higher energies it is used only for events with W>2.0GeV/c2. This package treats not only π but other mesons such as K, η and ρ. In the region of 1.3GeV/c2 2 nπ =0.09 + 1.83 ln W (3.26) The number of pions in each event is determined by using KNO (Koba-Nielsen-Olsen) scaling. Since the range of W overlaps with that in single pion production, n ≥ 2 is required in this W region. The forward-backward asymmetry of pion multiplicity 50 ) 1 -1 GRV94LO 0.9 GRV94LO (w/ B-Y corr.) GRV98LO 0.8 GRV98LO (w/ B-Y corr.) GeV 0.7 2 0.6 cm 0.5 -38 0.4 0.3 (10 0.2 ν CCFR 90 CHARM 88 BEBC-WBB 79 CCFRR 84 0.1 CDHSW 87 BNL 80 IHEP-JINR 96 SKAT /E GGM-PS 79 CRS 80 IHEP-ITEP 79 σ 0 010203050 100 150 200 250 Eν (GeV) Figure 3.7: cross-sections for charged current νμ andν ¯μ interactions. Upper(lower) lines are νμ(¯νμ) in the hadronic center of mass system is included using the results from the BEBC experiment [63]: F 2 nπ 0.35 + 0.41 ln(W ) B = 2 (3.27) nπ 0.5+0.09 ln(W ) The cross-section of the CC νμ andν ¯μ interactions are shown in Figure 3.7. 3.3.4 Coherent Pion Production Coherent pion production occurs when a neutrino interacts with an oxygen nucleus, which remains intact, and one pion is produced with the same charge as the incoming weak current. Since very little momentum is transferred to the oxygen nucleus, the angular distributions of the outgoing leptons and pions are peaked in the forward direction. The formalism developed by Rein and Sehgal [64] was used to simulate the 51 interactions, and the differential cross-section is given by: d3σ G2M 1 =β × N f 2A2E (1 − y) (σπN )2 dQ2dydt 2π2 π ν 16π total M 2 2 × r2 A e−b|t|F (1 + ) 2 2 abs (3.28) MA + Q r =Re(fπN(0))/Im(fπN(0)) (3.29) where Q2 is the square of the four-momentum transfer of the lepton, β is the axial vector coupling constant with β = 1 for neutral current and β = 2 for charged current, G is the weak coupling constant, MN is the nucleon mass, fπ =0.93mπ is the pion decay constant, A is the atomic number (=16 for oxygen), Eν is the neutrino energy, − πN y(= (Eν El)/Eν) is the lepton fractional energy loss, σtotal is the averaged total −2 pion-nucleon cross-section, MA is the axial-vector mass, b =80GeV is in the order of the nucleus transverse dimension, t is the square of the four-momentum transfer to the nucleus, fπN is the pion-nucleon scattering amplitude, Fabs accounts for the absorption of pions in the nucleus and is expressed as: − x σπN ρ Fabs = e inel (3.30) where x is the average path length of the pion in the oxygen, ρ is the nuclear πN density and σinel is the averaged total inelastic pion-nucleon cross-section. However, the K2K experiment set an upper limit on the cross-section of coherent CC pion production [65] that is significantly lower than the predicted cross-section by Rein and Sehgal. Therefore, some modification is necessary to the coherent pion production cross-sections used in the MC. The calculated cross-sections of the coherent pion production for the CC and NC interactions by Kartavtsev and Paschos [66] agree better with the several experimental data than that of Rein and Sehgal as shown in Figure 3.8. The difference between the experimental data and the model proposed by 52 250 140 Rein and Sehgal 225 CHARM(93) Kartavtsev et al. MiniBooNE(05) 120 200 Aachen-Padova(83) Gargamelle(84) 175 100 Rein and Sehgal 150 Kartavtsev et al. /Carbon Nucleous) /Carbon Nucleous) 2 2 80 cm 125 cm -40 -40 60 100 (10 (10 σ σ 75 40 50 20 25 0 0 -1 2 12345678910 10 1 10 10 Eν(GeV) Eν(GeV) (a) CC (b) NC Figure 3.8: The cross-sections of coherent pion productions off the carbon nucleus for CC and NC interactions by two models with the experimental data. The solid lines are calculation by Rein and Sehgal, while the dashed lines are by Kartavtsev and Paschos. On the left figure, the arrow shows the experimental upper limit by K2K [65], and on the right figure. The experimental data are from CHARM [67], MiniBooNE [68], Aachen-Padove [69] and Gargamelle [70]. Kartavtsev and Paschos was taken into account as a systematic uncertainty of 100%. 3.3.5 Nuclear Effects The secondary mesons produced by the neutrino interactions with the 16Onucleiare tracked from their production points until they exit or are absorbed. This is done for π, K and η by using the same cascade model as that used for kaons in the proton decay simulation described in section 3.1. The interactions of pions are especially im- portant since the cross-section for pion production is large for Eν > 1GeV,andthe pion-nucleon crosssection is also large. The pion-nucleon interactions that are con- sidered in the simulation are: inelastic scattering, charge exchange, and absorption. The pion production point where the neutrino-nucleon interaction occurred is deter- mined by the Woods-Saxon density distribution (3.1). The type of pion interaction is 53 600 inela+abs+cex inelastic 500 absorption charge exchange O (mb) 16 400 + + π inela+abs+cex 300 200 inelastic absorption cross section for 100 charge exchange 0 0 100 200 300 400 500 600 momentum of π+ (MeV/c) Figure 3.9: cross-sections of π+-16O scattering as a function of π+ momentum. Solid lines show the calculation from our simulation and points show the experimental data [73]. determined by the calculated mean free path of each interaction based on the model of Oset [71]. In the calculation, Fermi motion and Pauli blocking are considered. In the case of inelastic scattering or charge exchange the direction and momentum of the pion are determined using the results of a phase shift analysis obtained from pion-nucleon scattering experiments [72]. The pion interaction model is tested using the experimental results from the following interactions: π−12C scattering, π−16O scattering, and pion photo-production (γ +12 C → π− + X) [73, 74] as shown in Fig- ure 3.9. Absorption for η mesons (ηN → N ∗ → π(π)N) is also considered [75] with the resulting pions tracked using the method described above. Nucleon re-scattering is also considered using the cascade model. The considered interactions are elastic scattering, a single or two delta(s) production. The differential cross-sections are obtained from nucleon-nucleon scattering experiments [76]. For delta production, the isobar production model is used [77]. 54 3.4 Detector Simulation Simulated kinematics of proton decays and neutrino interactions are passed through a detector simulation program. The detector simulation simulates the propagation of particles, Cherenkov radiation, propagation of Cherenkov photons in the detector water, and the response of the PMTs and electronics. Based on the GEANT pack- age [78], the custom detector simulation program was developed for the SK detector. 3.4.1 Photon Propagation Absorption, Mie and Rayleigh scattering processes are taken into account in the sim- ulation of photon propagation in the detector. At short wavelengths, Rayleigh scat- tering is dominant due to the λ−4 dependence. Photons are scattered symmetrically in the forward and backward directions in this process. Absorption is a dominant pro- cess for longer wavelengths ( 450 nm). Coefficients for these processes were tuned to reproduce the calibration data as described in section 2.6.4. Figure 3.10 shows the wavelength dependence of these coefficients and calibration data. 3.4.2 Hadron Propagation In order to reproduce low energy pion interactions, hadronic interactions are treated by a custom simulation program [79] based on experimental data of π-16O scatter- ing [80] and π-p scattering [81] for charged pions of pπ > 500 MeV/c. The CALOR package [82] is used for propagation of nucleons and charged pions of pπ > 500 MeV/c. 0 0 3.4.3 KL→KS Regeneration in Water 0 → 0 0 KL KS regeneration is modeled in the simulation as a decay channel of the KL that is parameterized by a regeneration lifetime in water, τ ; a tunable parameter that RH2O affects the probability of regeneration. The probability of regeneration in carbon is 55 ) -1 Dec.-3,4-96 Dec.-14-96 Dec.-17-96 Dec.-27-96 -1 Jan.-18-97 Mar.-6-97 10 total absorption -2 10 Mie scattering Rayleigh scattering attenuation coefficient (m -3 10 200 250 300 350 400 450 500 550 600 650 700 wavelength (nm) Figure 3.10: Wavelength dependence of photonattenuationcoefficientsusedinMC simulation (solid lines). Measured results from calibration data described in section 2.6.4 is shown (points). measured to be 0.1% for kaons with momentum of 400 MeV/c [83]. We assume the probability of regeneration in water is the same as in carbon. We generate 400 MeV/c pure K0 events in the detector simulation and tune τ such that the fraction L RH2O with regeneration in water is ∼ 0.1% of total generated events. Regeneration is not simulated in water for the atmospheric neutrino MC. Chapter 4 Fully Contained Event Reduction 4.1 Overview Each day, about 106 events are collected by the SK detector not including SLE trigger events. Many of these are due to cosmic ray muons and low energy events from radioactive material such as radon near the detector wall. Also, a non-negligible fraction are PMT flashers from discharges near the dynodes inside the PMT. All of these sources are background to the proton decay search and so we reject these using the fully contained (FC) reduction steps described in this chapter. An FC event by definition only has activity within the ID volume. There are five steps to the FC reduction all of which are automatically applied. 4.2 First Reduction The selection criteria that comprise the first reduction step of the FC event sample are the following: (A) PE300 ≥ 200 p.e. PE300 is defined as the maximum number of p.e.s observed by ID PMTs within 56 57 in a sliding 300 ns window. AND (B) NHITA800 ≤ 50 or OD trigger is off NHITA800 is defined as the total number of hit OD PMTs in a fixed 800 ns win- dow that extends from −400 ns to +400 ns before and after the trigger timing. AND (C) TDIFF> 100 ñs TDIFF is the time interval from the previous event. Criteria (A) is designed to reject low energy events from radioactive isotopes, (B) rejects cosmic ray muons and (C) rejects decay electrons from cosmic ray muons which stop in the ID. Events within 30 ñs after another event selected by the above criteria are retained to keep the decay electrons. These retained events are attached to the fully contained candidates as sub-events and not counted as a primary events. The data rate is reduced from 106 events/day to 3000 events/day after the first reduction step. 4.3 Second Reduction The selection criteria that define the second reduction step of the FC event sample are the following: (A) NHITA800 < 25 if PEtot < 100.000 p.e. or OD trigger is off. PEtot is the total observed p.e.s in the ID. AND (B) PEmax/PE300 < 0.5 PEmax is the maximum number of p.e.s observed by an ID PMT. 58 Criteria (A) is designed to further eliminate cosmic ray muons, (B) rejects low energy and electrical noise events, which have one large hit signal from a single PMT. The data rate is 200 events/day after the second reduction. 4.4 Third Reduction After the first two reduction steps the remaining background consists of noise and cosmic ray muons with only a few hits in the OD. 4.4.1 Through-going Muon Cut The following criteria are used to reject through-going muons: (A) PEmax > 230 p.e. AND (B) goodness of through-going muon fit > 0.75 AND (C) NHITAin ≥ 10 or NHITAout ≥ 10 NHITAin (NHITAout) is the number of hit OD PMTs located within 8 m from the entrance (exit) point in a fixed 800 ns time window. If criteria (A) is satisfied, a muon fitter is applied. The output of the muon fitter is used for criteria (B) and (C). The muon fitter selects an entrance point based on the earliest hit PMT with respect to neighboring PMTs and an exit point defined as the center of the saturated ID PMTs. The fitter then calculates a goodness which is 59 defined as: 1 1 (t − T )2 goodness = × exp − i i (4.1) 1 σ2 2(1.5 × σ )2 i i i σ2 i i where ti and σi are the observed hit time and resolution of the i-th PMT respectively and Ti is the hit time expected from the entering time of the muon and its track. 4.4.2 Stopping Muon Cut To reject stopping muons, a stopping muon fitter is applied which fits the entrance point in a similar fashion as the through-going muon fitter. The following selection criteria are the used to reject stopping muons: (A) NHITAin ≥ 10 or NHITAin ≥ 5 if goodness of stopping muon fit > 0.5 The direction of the muon is reconstructed to maximize the total number of p.e.s inside the cone with a half opening angle of 42◦. The goodness definition is the same as (4.1). 4.4.3 Cable Port Muons Located at the top of the detector tank are 12 cable ports that allow the high voltage and signal cables access from the PMTs to the electronics huts. Four of these ports sit above the ID and block the OD. There is a possibility that cosmic ray muons may go through the port and enter the ID without triggering the OD. To eliminate this occurrence, veto counters (2 m × 2.5 m scintillation counters) were installed above 60 the cable ports in April 1997. To eliminate such muons entering the cable ports the following criteria is applied: (A) One veto counter hit AND (B) Lveto < 4m Lveto is the distance from the cable hole to the the reconstructed vertex. 4.4.4 Flasher Event Cut Flasher events occur when electrical discharges in a PMT cause it to act as a light source. Flashers are characterized by a hit timing distribution that is more broad than neutrino events. To eliminate these, the following criteria is applied: (A) NMIN100 ≥ 14 OR (B) NMIN100 ≥ 10 if the number of hit ID PMTs < 800 NMIN100 is the minimum number of hit ID PMTs in a sliding 100 ns time window from +300 ns to +800 ns after the trigger. 4.4.5 Accidental Coincidence Cut The accidental coincidence occurs when a low energy event starts the trigger and a cosmic ray muon follows within a single trigger gate. These are not rejected by earlier reduction steps because of the absence of the OD activity on the trigger timing and the large number of total p.e.s in the ID due to the muons. The accidental coincidence events are removed by the following criteria: 61 (A) NHITAoff ≥ 20 NHITAoff is the number of hit OD PMTs in a fixed 500 ns off-timing window from +400 ns to +900 ns after the trigger timing. AND (B) PEoff > 5000 p.e. PEoff is the number of p.e.s observed by ID PMTs in a fixed 500 ns off-timing window from +400 ns to +900 ns. 4.4.6 Low Energy Events Cut The remaining events are low energy events from the decay of radio isotopes and electrical noise. The following criteria is applied to remove these background: (A) NHIT50 < 50 NHIT50 is the number of hit ID PMTs in a sliding 50 ns time window. Where NHIT50 is counted after subtracting the time of flight of each observed photon assuming all photons are generated at a point. The vertex is determined as the position at which the timing residual distribution is peaked. NHIT50 = 50 corresponds to visible energy of 9 MeV/c and is low enough not to lose efficiency for contained neutrino events with Evis > 30 MeV/c where Evis is the total visible energy defined as the sum of the energy of all rings assuming they are produced by electrons. The data rate is 45 events/day after the third reduction. 4.5 Fourth Reduction This reduction step is designed to remove additional flasher events that are missed by earlier reduction steps. Because flasher events are caused by electrical discharges in a PMT the event has a characteristic pattern that may repeat over the course of a 62 few hours or days that the flasher is active. A pattern matching algorithm is used to look for these events with a similar pattern and remove them. The algorithm is the following: (A) Divide the ID wall into 1450 patches of 2 m × 2m square. (B) Compute the correlation factor, r, by comparing the total charge in each patch of two events, A and B. The correlation is defined as: 1 (QA −QA) × (QB −QB) r = i i (4.2) N σ × σ i A B A(B) where N is the number of the patches, and Q and σA(B) are the averaged charge and its standard deviation respectively for event A(B). (C) Calculate the distance, DISTmax, between the PMTs with the maximum pulse heights in the two compared events. (D) If DISTmax < 75 cm, an offset value is added to r: r = r +0.15. (E) If r exceeds the threshold, rth, events A and B are recognized as matched events. rth is defined as: × A B rth =0.168 log10((PEtot +PEtot)/2) + 0.130 (4.3) where PEtot is the total number of p.e.s observed in the ID. (F) Repeat the above calculation over 10,000 events around the target event and count the number of matched events. (G) Remove the events with large correlation factor r, or large number of matched events. The data rate is 18 events/day after the fourth reduction. 63 4.6 Fifth Reduction 4.6.1 Stopping Muon Cut The remaining stopping muons are rejected by tighter criteria than those in the third reduction stage. Events satisfying the following criteria are rejected: (A) NHITAin ≥ 5 NHITAin is the number of hit OD PMTs located within 8 m from the entrance point in a sliding 200 ns time window from −400 ns to +400 ns. The entrance position in the OD is estimated by a backward extrapolation from the reconstructed vertex determined by TDC fit (see section 5.1.1.3). 4.6.2 Invisible Muon Cut Invisible muons are caused by cosmic ray muons with momenta less than the Cherenkov threshold and the subsequent decay electrons being observed. These are characterized by a low energy signal from decay electron and a signal in the OD before the trigger timing. Events which satisfy the following cut criteria are rejected as invisible muons: (A) PEtot < 1000 p.e. PEtot is the total number of p.e.s observed in the ID. AND (B) NHITACearly > 4 NHITACearly is the maximum number of hit PMTs in the OD hit cluster in a sliding 200 ns width time window from −800 ns to −100 ns. AND (C) NHITACearly + NHITAC500 > 9ifDISTclust < 500 cm NHITACearly > 9otherwise 64 NHITAC500 is the number of hit PMTs in the OD hit cluster in a fixed 500 ns time window from −100 ns to +400 ns. DISTclust is the distance between the OD hit clusters which are used for NHITACearly and NHITAC500. 4.6.3 Coincidence Muon Cut The remaining accidental coincidence muons are removed by: (A) PE500 < 300 p.e. PE500 is the total number of p.e.s observed in the ID in a fixed 500 ns time window from −100 ns to +400 ns. AND (B) PElate ≥ 20 p.e. PElate is the maximum number of hit OD PMTs in a 200 ns sliding time window from +400 ns to +1600 ns. 4.6.4 Long Tail Flasher Cut This is a more strict version of the flasher cut in the third FC reduction stage. Events satisfying the following criterion are removed: (A) NMIN100 > 5 if the goodness of point fit < 0.4 NMIN100 is the minimum number of the hit ID PMTs in a sliding 100 ns time window from +300 ns to +800 ns. Point fit is explained in section 5.1.1.1. The data rate is 16 events/day after the fifth reduction. 4.7 Final Fully Contained Fiducial Volume Sample To select the final fully contained fiducial volume (FCFV) sample the following event selection is applied: 65 (A) NHITAC < 10 NHITAC is the number of hit PMTs in the highest charge OD cluster. (B) Evis > 30 MeV/c Evis is the total visible energy defined as the sum of the energy of all rings assuming they are produced by electrons. (C) Dwall > 200 cm where Dwall is the distance of the reconstructed vertex from the ID PMT surface. This defines the FV. This event selection defines a FV of 22.5 kt. The detection efficiency is estimated to be 97.2% with systematic error of 0.2%. The data rate is 8.18 ± 0.7events/day. Chapter 5 Event Reconstruction After the event reduction steps described in Chapter 4, the event reconstruction is applied to the FC sample. There are two event reconstructions used for this analysis. First, is the standard event reconstruction which is described below in section 5.1. This is the standard SK reconstruction algorithm that is used for many of the other analysis by the SK collaboration. The performance of the standard reconstruction is summarized in other documents [84, 85]. Second, is the multi-vertex reconstruction 0 algorithm which was developed specifically for the KL search of this analysis. The algorithm and performance of the multi-vertex reconstruction is described in section 5.2. 5.1 Standard Event Reconstruction The standard event reconstruction algorithms consist of the following: (A) Vertex fit Locates the vertex (origin) of the interaction generating the event as well as the direction of the most dominant (energetic) ring. (B) Ring counting 66 67 Uses the vertex and direction of the dominant ring found in (A) to search for additional rings. (C) Particle Identification (PID) Determines a particle type for each ring; showering type (from e±,γ) and non- showering type (from μ±,π±, p). These showering and non-showering types are also known as e-like and μ-like respectively. (D) MS Vertex Fit For single ring events the MS Vertex fit is applied. The PID information from (C) is used to reconstruct the vertex position. (E) Momentum Reconstruction The momentum of each ring found in (B) is determined. (F) Decay electron finding Searches the event for decay electrons. (G) Ring number correction Rejects some low energy rings and overlapping rings found in (B) that are considered as fake rings. 5.1.1 Vertex There are three steps to the vertex reconstruction algorithm: point fit, ring edge fitting and TDC fit. 5.1.1.1 Point Fit The first step of vertex reconstruction is appropriately called point fit as it is based on the assumption that the all Cherenkov photons are emitted from a single point. Using this assumption, a rough estimate of the vertex is calculated by searching for 68 the point within the ID volume in which the timing residual (PMT hit time with time of flight subtracted) distribution is most peaked. The timing residual, ti, is calculated as: n(q ,l) t = t0 − i i ×|P − O | (5.1) i i c i 0 where ti is the timing of the i-th PMT, Pi and O are the position of the i-th PMT and estimated vertex respectively. The index of refraction n(qi,li) is a function of the detected p.e., qi, and photon travel length, li = |Pi − O |. The goodness of the point-fit is defined as: 1 (t − t )2 G = exp − i 0 (5.2) p N 2(1.5 × σ)2 i where N is the number of hit PMTs, t0 is a free parameter chosen to maximize Gp, and σ is the PMT timing resolution taken to be 2.5 ns. The numerical factor 1.5 is chosen to optimize the fitter performance. The fitter searches for the vertex point which gives the maximum value of Gp. Using the estimated vertex from the point fit algorithm a rough estimate of the particle direction is determined as the charge weighted average of all PMTs: P − O d = q × i 0 (5.3) 0 i | − | i Pi O0 where d0 is the direction of the particle, O0 is the vertex position determined by point fit, qi and Pi are the charge and position of the i-th PMT respectively. 5.1.1.2 Ring Edge Fitting The Cherenkov ring edge and direction of the most dominant ring is fit in this step. To determine the Cherenkov ring edge, the p.e. for all PMTs are filled into a 69 histogram as a function of the Cherenkov opening angle, θ. Here, the p.e. are corrected for attenuation length and PMT acceptance. Using this PE(θ) histogram the criteria to determine the Cherenkov ring edge θedge are: (A) θedge >θpeak θpeak is the opening angle where PE(θ) is maximum. 2 d PE(θ) (B) = 0 the second derivative at θ = θedge should be zero. d2θ θedge If multiple θedge satisfy (A) and (B), then the θedge nearest θpeak is selected. The ring direction is determined using the following estimator: θedge 2 PE(θ)dθ − 2 0 dPE(θ) (θedge θexp) Q(θedge)= × × exp − (5.4) sin θ dθ 2σ2 edge θ=θedge θ where θexp and σθ are the Cherenkov opening angle expected from the charge within the cone and its resolution, respectively. The estimator Q(θedge) is calculated changing the ring direction about the ring direction d0 found in point fit. The ring direction and edge position are determined which maximize Q(θedge). 5.1.1.3 TDC Fit In this final step the vertex position is more precisely determined by now taking into account scattered Cherenkov light and finite track length. The track of the particle is considered in the time residual, ti,ofthei-th PMT as: ⎧ ⎪ ⎨ 0 − 1 ×| − |− n ×| − | ti c Xi O c Pi Xi for PMTs inside the Cherenkov ring t = i ⎪ ⎩ 0 − n ×| − | ti c Pi O for PMTs outside the Cherenkov ring (5.5) where O is the vertex position, X i is the position at which Cherenkov photons are emittedtowardthei-th PMT, n is the refractive index of water, Pi is the position 0 and ti is the hit timing of the i-th PMT. 70 There are three components to the estimator used for TDC fit which are calculated based on PMT location and hit timing. For PMTs inside the Cherenkov ring, the estimator GI is defined: 1 (t − t )2 G = exp − i 0 (5.6) I σ2 2(1.5 × σ)2 i i where σi is the timing resolution of the i-th PMT depending on the detected p.e.s, σ is the timing resolution for the average p.e.s of all hit PMTs, ti is the time residual of the i-th PMT as calculated by (5.5) and t0 is a free parameter used to maximize GI . For PMTs outside the Cherenkov ring there are two estimators GO1 and GO2,if the hit timing is later than t0 the contribution from scattered photons is considered. The estimators are defined as: 1 (t − t )2 G = exp − i 0 × 2 − 1 (for t ≤ t ) O1 σ2 2(1.5 × σ)2 i 0 i i (5.7) 1 (t − t )2 G = max exp − i 0 ,G (t ,t ) × 2 − 1 (for t > t ) O2 σ2 2(1.5 × σ)2 scatt i 0 i 0 i i (5.8) where R (t − t )2 R t − t G (t ,t )= q × exp − i 0 + 1 − q × exp − i 0 (5.9) scatt i 0 1.52 2(1.5 × σ)2 1.52 60 ns Rq is the fraction of the charge detected inside the Cherenkov ring. The numerical fac- tors in the above equations are determined empirically to give the best reconstruction performance. 71 The total estimator of TDC fit is: G +(G or G ) G = I O1 O2 (5.10) T 1 σ2 i i The vertex position and ring direction are selected which maximize GT . 5.1.2 Ring Counting The ring counting algorithm is used to search for additional Cherenkov rings in the event. 5.1.2.1 Ring Candidate Selection Ring candidate selection searches for additional Cherenkov rings using an algorithm know as a Hough transform [86] which is commonly used in pattern recognition. However, before beginning the Hough transform, p.e. contributions from rings that are already regarded as true rings are subtracted to enhance the ring finding ability. The Hough space is defined as the (Θ, Φ) plane where Θ is the polar angle and Φ is the azimuthal angle as measured from the reconstructed vertex. The Hough transform is performed by mapping the detected p.e. corrected for PMT acceptance and attenuation length from the i-th hit PMT into the (Θi, Φi) pixel in the Hough space for which the opening angle toward the hit PMT is 42◦. As a result of this mapping, the center of ring candidates are observed as peaks in the (Θ, Φ) plane. 5.1.2.2 Ring Candidate Test In this step, a likelihood method is used to determine whether the candidate ring found by the Hough transform is a true ring or not. When N rings have been found, the likelihood used to check if the (N + 1)-th candidate is probable or not is defined 72 as: N+1 obs · exp LN+1 = log prob qi , αn qi,n (5.11) i n=1 obs where the sum extends over hit PMTs inside N + 1 Cherenkov rings, qi is the · exp observed p.e.s in the i-th PMT and αn qi,n is the expected p.e.s in the i-th PMT from the n-th ring. The LN+1 is maximized by changing the αn scale factors with a lower momentum limit constraint. The probability prob is defined as: ⎧ ⎪ (qobs,qexp)2 ⎪√ 1 − i i exp ⎪ exp (for qi > 20 p.e.) ⎪ 2πσ 2σ2 ⎪ ⎨⎪ obs exp Probability obtained by the probability density prob(qi ,qi ) ⎪ ⎪ exp ⎪distribution function based on the convolution of a (for qi < 20 p.e.) ⎪ ⎪ ⎩⎪single p.e. distribution and a Poisson distribution. (5.12) where σ is the resolution for qexp. If no candidate satisfies LN+1 ≥ LN ,thenumber ofringsisdeterminedtobeN and the ring counting procedure is finished. For ring candidates which satisfy LN+1 ≥ LN , the following evaluation functions are calculated: F1: The difference LN+1 − LN corrected for the total p.e.s. F2: The average of the expected p.e.s from the (N + 1)-th ring, near the edge of the Cherenkov ring. F3: The average of the expected p.e.s outside the (N + 1)-th ring. F4: The residual p.e.s from the expectation with N rings. 73 The evaluation functions are used to calculate the final evaluation function as: 4 FSK−I = [αiFi] (5.13) i=1 where αi are optimized weighting parameters. If FSK−I is positive, then the (N +1)-th ring candidate is considered a true ring. During this process, the effect of muon decay electrons is eliminated by restricting the ring search to PMTs hit within a timing window of width tw starting from the lower edge of the peak of the TOF subtracted hit timing distribution. This is done using the ritofcut routine. tw is given by: tw =(qtot × 65 + 1400)/(6.4 × 40) + 5 (5.14) where qtot is the total p.e.s in the ID. 5.1.3 Particle Identification The PID algorithm classifies each Cherenkov ring into two types: showering (e- like) and non-showering (μ-like). Electrons and gamma-rays undergo electromagnetic showers and multiple scattering which results in Cherenkov rings with diffuse edges. The Cherenkov ring edges from muons and charged pions are more sharp. Another distinction is the Cherenkov opening angle. The Cherenkov opening angle for elec- trons and gamma-rays is 42◦ because they are highly relativistic (β ∼ 1). However, muons and pions can have smaller opening angles if they are not highly relativistic. The PID algorithm uses these differences between the Cherenkov ring pattern and opening angle to make its determination of ring type. 74 5.1.3.1 Expected Charge Distribution The first necessary step in particle identification is the calculation of the expected p.e. distribution for electrons and muons. These distributions are calculated using the following for electrons and muons respectively: 1.5 exp × exp × R × 1 × scatt qi (e)=αe Q (pe,θi) f(Θi)+qi (5.15) ⎛ ri exp(ri/L) ⎞ sin2 θ 1 exp ⎝ × xi knock⎠ × × scatt qi (μ)= αμ + qi f(Θi)+qi r sin θ + r · dθ exp(ri/L) i xi i dx x=xi (5.16) where αe,αμ : a normalization factor for electrons and muons respectively ri : is the distance from the vertex to the i-th PMT θi : is the opening angle between the i-th PMT direction and the ring direction L : the attenuation length of light in water f(Θi) : a correction of acceptance to the i-th PMT as a function of the photon incidence angle Θi R : radius of the virtual sphere (16.9m) exp Q (pe,θi) : expected p.e. distribution from an electron as a function of the open- ing angle and electron momentum. This is obtained by MC simula- tion. scatt qi : expected p.e.s for the i-th PMT from scattered photons x : track length of the muon 75 xi : track length of the muon at which Cherenkov photons are emitted toward the i-th PMT scatt knock qi ,qi : expected p.e.s for the i-th PMT from scattered photons and knock-on electrons respectively. θ (θxi ) : Cherenkov opening angle of the muon traversing at x (xi). The energy loss of the muon is taken into account. 1.5 In the electron case, the (R/ri) term corrects for the distance dependence of the Cherenkov intensity. For the muon, the sin2 θ term corrects for the angular depen- dence of the Cherenkov intensity and the r(sin θ + r(dθ/dx)) term corresponds to the area where Cherenkov photons are emitted when the muon travels a distance dx. 5.1.3.2 Determination of Particle Type The likelihood used to estimate particle type of the n-th ring is given by: obs exp exp Ln(e or μ)= prob qi ,qi,n (e or μ)+ qi,n (5.17) θi<(1.5×θc) n = n obs where the product extends over PMTs inside the n-th ring. qi is the observed exp p.e.s in the i-th PMT, qi,n (e or μ) is the expected p.e.s from the n-th ring with the exp assumption of an electron (5.15) or muon (5.16) and qi,n are the expected p.e.s from the n-th ring without any assumption of particle type. The function prob returns the obs exp probability to detect qi in the i-th PMT when qi is expected (5.12). By changing exp exp the direction and opening angle of the n-th ring, qi,n (e)andqi,n (μ) are optimized to give the maximum likelihood value. To combine this likelihood with another that uses the Cherenkov opening angle, 76 the likelihood is converted into the the following χ2 distribution: 2 − χn(e or μ)= 2logLn(e or μ) + constant (5.18) The probabilities from the ring pattern are: (χ2 (e or μ) − min[χ2 (e),χ2 (μ)])2 P pattern(e or μ)=exp − n n n (5.19) n 2σ2 χn2 √ 2 where σ 2 = 2N and N is the number of PMTs used in the calculation. χn The probabilities from Cherenkov opening angle are: θobs − θexp(e or μ) 2 angle − n n Pn (e or μ)=exp 2 (5.20) 2(δθn) obs where θn and δθn are the reconstructed opening angle of the n-th ring and the fitting exp error respectively. θn (e or μ) is the expected opening angle of the n-th ring, which is estimated from the reconstructed momentum assuming an electron or muon. For single and multi ring events the probability functions of PID are respectively given by: pattern × angle Psingle(e, μ)=Psingle (e, μ) Psingle(e, μ) (5.21) pattern Pmulti(e, μ)=Pmulti (e, μ) (5.22) With the above probabilities we construct the PID likelihood, PPID ≡ − log P (μ)− − log P (e), used to determine ring type. If PPID < 0 the ring is considered e-like and if PPID ≥ 0 the ring is considered μ-like. This algorithm was tested by a beam test at KEK [87]. 77 5.1.4 MS Vertex Fit The vertex fitter described in section 5.1.1 mainly uses timing to determine the vertex. However, using only timing for single ring events results in poor resolution of the vertex along the ring direction. The MS vertex fitter attempts to improve the vertex reconstruction by using the Cherenkov ring pattern information to re-fit the vertex along the ring direction. The fitter acts iteratively, changing both the vertex and ring direction using a likelihood based on the expected p.e. distributions similar to the PID algorithm until changes in vertex position and ring direction are less than 5 cm and 0.5◦ respectively. During the fit the Cherenkov opening angle is fixed. 5.1.5 Momentum Determination The momentum is estimated from the total number of p.e.s detected within a 70◦ half opening angle towards the reconstructed ring direction. To determine the momentum for each ring, the p.e.s in hit PMTs must be separated based on the contribution from each ring. This separation is carried out based on the expected p.e. distribution which is a function of the opening angle θ and assumes uniformity in azimuthal angle φ. The observed p.e.s of the i-th PMT are then separated as: qexp obs obs × i,n qi,n = qi exp (5.23) n qi,n obs obs where qi,n is the fractional p.e.s assigned to the n-th ring, qi is the observed p.e.s exp in the i-th PMT and qi,n is the expected p.e.s. The observed p.e.s for each PMT is corrected for light attenuation in water and PMT acceptance and summed to obtain the integrated total charge within the 70◦ 78 cone of the n-th ring, RTOTn,as: ⎡ ⎤ G ⎢ r cos Θ ⎥ MC ⎢ × obs × i × i − ⎥ RTOTn = ⎣α qi,n exp Si⎦ Gdata ◦ L f(Θi) ◦ θi,n<70 θi,n<70 −50ns α : a normalization factor Gdata,GMC : the relative PMT gain for the data and MC respectively θi,n : is the opening angle between the i-th PMT direction and the n-th ring direction ti : the TOF subtracted hit timing of the i-th PMT position L : the attenuation length of light in water f(Θi) : a correction of acceptance to the i-th PMT as a function of the photon incidence angle Θi ri : the distance from the vertex to the i-th PMT Si : expected p.e.s for the i-th PMT from scattered photons To eliminate the effect from muon decay electrons, the summation is restricted to PMTs hit within a time window from −50 ns to +250 ns about the peak of the TOF subtracted hit timing distribution. This is done by the sptofcut routine. 5.1.5.1 Ring Separation For multi-ring events, it is necessary to separate the observed p.e.s in each PMT to the contribution of each ring so the momentum of each ring can be determined. The 79 expected charge for the i-th PMT used for charge separation is calculated as: R 1 qexp e, μ Qexp p, θ × × × f qscatt i ( )= e,μ ( i) exp MC (Θi)+ i (5.25) L exp(ri/L) where ri : the distance from the vertex to the i-th PMT θi,n : is the opening angle between the i-th PMT direction and the n-th ring direction L : the attenuation length of light in water LMC : the attenuation length of light in water in the MC f(Θi) : a correction of acceptance to the i-th PMT as a function of the photon incidence angle Θi R : the radius of the virtual sphere (16.9m) exp Qe,μ (p, θi) : expected p.e. distribution for e and μ on the virutal sphere as a function of the momentum, p, and opening angle to the i-th PMT, θi scatt qi : expected charge for the i-th PMT from scattered photons For all e-like rings except for the most energetic, the γ conversion length is taken into account in the distance calculation. Reflection on the PMT surface is considered as: exp → × exp exp qi (e, μ) α qi (e, μ)+ βijqj (e, μ) (5.26) j= i 80 where α is a normalization factor and βij is the PMT reflection factor from the j-th to i-th PMT calculated as: