Università degli Studi di Napoli Federico II
DOTTORATO DI RICERCA IN FISICA
Ciclo XXXIII Coordinatore: prof. Salvatore Capozziello
Pre-Supernova Alert System for Super-Kamiokande with Gadolinium
Settore Scientifico Disciplinare _____/______
Dottorando Tutor Lucas NASCIMENTO MACHADO Prof.ssa Gianfranca DE ROSA Prof. Vittorio PALLADINO
Anni 2018/2021 Declaration of Authorship
I, Lucas Nascimento Machado, declare that this thesis titled, ‘Pre-Supernova Alert System for Super-Kamiokande with Gadolinium’ and the work presented in it are my own. I confirm that:
⌅ This work was done wholly or mainly while in candidature for a research degree at this University.
⌅ Where any part of this thesis has previously been submitted for a degree or any other qualification at this University or any other institution, this has been clearly stated.
⌅ Where I have consulted the published work of others, this is always clearly attributed.
⌅ Where I have quoted from the work of others, the source is always given. With the exception of such quotations, this thesis is entirely my own work.
⌅ Ihaveacknowledgedallmainsourcesofhelp.
⌅ Where the thesis is based on work done by myself jointly with others, I have made clear exactly what was done by others and what I have contributed myself.
Signed:
Date:
1 UNIVERSITA` DEGLI STUDI DI NAPOLI FEDERICO II
Abstract (English)
”Ettore Pancini” Physics Department
Doctor of Philosophy
by Lucas Nascimento Machado
The current phase of the Super-Kamiokande experiment, SK-Gd, is character- ized by the addition of gadolinium sulfate to the water Cherenkov detector, which improves the detection capability of thermal neutrons. For low energy events, the main detection channel for electron anti-neutrinos is the Inverse Beta Decay interaction, which has, in its final state, a positron and a neutron. The neutron thermal capture by gadolinium emits an 8 MeV gamma-ray cascade, improving the identification of the products of this process. This improved identification reduces the background for low energy events, allowing the analysis of neutrinos with en- ergies below the usual Super-Kamiokande thresholds. One possible detection by SK-Gd is the neutrinos coming from pre-Supernova stars, which correspond to the last evolutionary state of massive stars before core-collapse Supernova. Dur- ing this stage, pair annihilation and beta decay processes are the main cooling mechanisms of the massive stars, emitting high fluxes of electron anti-neutrinos. Their detection could provide an early warning for core-collapse Supernovae. The techniques for the development of the pre-Supernova alert system for SK-Gd are presented in this thesis. UNIVERSITA` DEGLI STUDI DI NAPOLI FEDERICO II
Abstract (Italian)
Dipartimento di Fisica ”Ettore Pancini”
Dottore in Filosofia
di Lucas Nascimento Machado
L’esperimento Super-Kamiokande ha recentemente iniziato una nuova fase sper- imentale, SK-Gd, caratterizzata dall’aggiunta di solfato di gadolinio al rilevatore Cherenkov, migliorando cos`ıla capacit`adi rivelazione dei neutroni termici. Per eventi a bassa energia, il canale di rivelazione principale per gli anti-neutrini elet- tronici `eil processo di decadimento beta inverso, che ha, nel suo stato finale, un positrone e un neutrone. La cattura termica dei neutroni in gadolinio emette una cascata di raggi gamma di 8 MeV, migliorando l’identificazione dei prodotti di questo processo. Questa migliore identificazione riduce il background per eventi di bassa energia, consentendo l’analisi di neutrini con energie al di sotto dei liveli fino ad oggi accessibili in Super-Kamiokande. In SK-Gd `epossibilie la rivelazione di neutrini da pre-Supernova, che si prevede siano emessi nell’ultimo stato evolutivo di stelle di massa maggiore di otto masse solari, prima del collasso. Durante questa fase, la produzione di coppie neutrini e antineutrini e i procesi di decadimento beta sono il principale meccanismo di ra↵reddamento e ci si aspetta un flusso intenso di anti-neutrini elettronici. La rivelazione di questi anti-neutrini elettronici potrebbe fornire un preallarme per collassi di Supernovae. In questa tesi vengono presentati i sistemi di allerta pre-Supernova per SK-Gd. UNIVERSITA` DEGLI STUDI DI NAPOLI FEDERICO II
Resumo (Portuguese)
Departamento de Fsica ”Ettore Pancini”
Doutor em Filosofia
por Lucas Nascimento Machado
AfaseatualdoexperimentoSuper-Kamiokande,SK-Gd,´ecaracterizadapela adi¸c˜aode sulfato de gadol´ınio`a´aguado detector Cherenkov, o que melhora a capacidade de detec¸c˜aode nˆeutrons t´ermicos. Para eventos de baixa energia, o principal canal de detec¸c˜aode anti-neutrinos de el´etrons ´ea intera¸c˜aode Decai- mento Beta Inverso, que tem, em seu estado final, um p´ositron e um nˆeutron. A captura t´ermica de nˆeutrons pelo gadol´ınio emite uma cascata de raios gama de 8 MeV, melhorando a identifica¸c˜ao dos produtos deste processo. Esta identifica¸c˜ao aprimorada reduz o background para eventos de baixa energia, permitindo at´e mesmo a an´alise de neutrinos em energias abaixo dos limites usuais do Super- Kamiokande. Uma poss´ıvel detec¸c˜ao pelo SK-Gd ´ea de neutrinos vindos de es- trelas pr´e-Supernova, que correspondem ao ´ultimoestado evolutivo de estrelas massivas, logo antes do colapso do n´ucleo da estrela. Durante este est´agio,os processos de aniquila¸c˜aode pares e decaimento beta s˜aoos principais mecanismos de resfriamento da estrela, emitindo altos fluxos de anti-neutrinos de el´etrons, cuja detec¸c˜ao poderia fornecer um alerta precoce de explos˜oes de Supernovas. As t´ecnicas para o desenvolvimento de um sistema de alerta de pr´e-Supernova para o SK-Gd s˜aoapresentadas nesta tese. Acknowledgements
First and foremost, I would like to thank my supervisor, Gianfranca De Rosa. From the very first day she has supported me in every single aspect during the PhD period. She helped me establish my research goals and assisted me in every step of the way until they were achieved. In these three years of PhD her professional and emotional support kept me going. I could not have imagined having a better advisor and mentor for my studies. Also a special gratitude to Vittorio Palladino for all the help and incentive and to all my colleagues in Naples.
IwouldliketoexpressgratitudetotheSuper-Kamiokandecollaborationfor accepting my participation and contribution to the SK-Gd project. I have learned a lot of fundamental physics and data analysis. Each meeting, discussion, workshop and email exchange was valuable. My sincere thanks to Mark Vagins, LLuis Marti- Magro, Alex Goldsack, and all the Gadolinium group members for the treasured advices that were very significant in shaping my research methods and critiquing my results. Also to Charles Simpson for introducing me to the pre-Supernova neutrino analysis and supporting me to continue his incredible work.
I also appreciate the encouragement from my friend Guido Celentano. Not only he made my life easier, resolving all the bureaucracies I had for living abroad, but he always had a word of advice and care.
My appreciation goes out to my family in Brazil. To my mother, Maria Ol´ımpia, and my father, Marcelo, for the moral and emotional foundation in my life. To my siblings Hugo, Guilherme, Mariana and Gabriel for their friendship and caring and my sisters in law, Gabriela e Poliana. To my grandparents and my aunts Rosi and Magda, for always keeping me in their heart and prayers. And to my niece Manuela that has already filled our lives with purpose and hope.
The realization of this PhD would have been impossible without the support from my friends. Even living in di↵erent continents they were always there for me, giving me incentive and optimism along the way. My special gratitude goes to B´arbara, Luana, Pedro, Matheus, Rafaela, Let´ıcia, Gabriel, Allan, Andressa 5 6 and Paulo. Also to my friends Rodrigo, Rachel, Murilo, Carla, Bruno, Eduardo, Guilherme, Juvenal, Matheus, Paulo, Pedro, Yuji, Beatriz, Isabela and Patr´ıcia.
Finally, I have been beyond fortunate to meet amazing people after moving to Italy. Living abroad is never easy, but their company and a↵ection gave me strength and courage to continue my journey. I am deeply grateful to Daisy, who has followed closely and daily every single step of the way, giving me advice and motivation to overcome obstacles and never give up. Many thanks to Brurya, for her constant caring and presence. I also wish to thank Mirko, Andy, Carol, Ally, Kelly, and so many others for believing in me more than I believe in myself. Contents
Declaration of Authorship 1
Abstract (English) 2
Abstract (Italian) 3
Resumo (Portuguese) 4
Acknowledgements 5
List of Figures 11
List of Tables 17
Abbreviations 19
Introduction (English) 24
Introduzione (Italian) 27
Introdu¸c˜ao (Portuguese) 30
1 Neutrinos 33 1.1 Neutrino Oscillations in Vacuum ...... 34 1.2 The MSW E↵ect ...... 38 1.3 Neutrino Sources ...... 39 1.3.1 Atmospheric Neutrinos ...... 39 7 Contents 8
1.3.2 Solar Neutrinos ...... 43 1.3.3 Reactor Neutrinos ...... 45 1.3.4 Accelerator Neutrinos ...... 47 1.3.5 Astrophysical Neutrinos ...... 48 1.4 Neutrino Interactions ...... 50
2 Supernova and Pre-Supernova 54 2.1 Supernova Neutrinos ...... 54 2.1.1 Supernova ...... 54 2.1.2 Supernova Burst and Neutrino Emission ...... 57 2.1.3 SN1987A ...... 62 2.2 Di↵use Supernova Neutrino Background ...... 63 2.3 Pre-Supernova Neutrinos ...... 66 2.3.1 Pre-Supernova Models ...... 70
3 Super-Kamiokande with Gadolinium 76 3.1 Detector Overview ...... 77 3.2 Super-Kamiokande Phases ...... 81 3.3 Cherenkov Radiation ...... 82 3.4 Water Purification System ...... 86 3.4.1 Preparation for SK-Gd ...... 87 3.5 PhotoMultiplier Tubes ...... 91 3.6 Electronics and DAQ ...... 96 3.7 Trigger Conditions and WIT ...... 97 3.7.1 Event Reconstruction ...... 99 3.7.2 Isotropy Parameters ...... 103 3.8 Calibrations in Super-Kamiokande ...... 104 3.9 Super-Kamiokande with Gadolinium ...... 105 3.9.1 EGADS ...... 107 3.9.2 Radiopurity ...... 110 3.9.3 First Gadolinium Loading and Future ...... 114 3.9.4 Simulation of TNC on gadolinium ...... 115
4 Low Energy Inverse Beta Decay Detection in Super-Kamiokande120 4.1 Inverse Beta Decay Identification ...... 121 4.2 Backgrounds ...... 124 4.2.1 Fake Neutrons and Accidental Coincidences ...... 125 4.2.2 Radioactive Contamination ...... 127 4.2.3 Reactor Neutrinos ...... 130 4.2.4 Spallation ...... 132 Contents 9
4.3 Event Selection ...... 134 4.3.1 Selection of Coincidence Events ...... 136 4.4 Multivariate Selection methods ...... 140 4.4.1 Boosted Decision Tree ...... 142 4.4.2 Previous Estimations ...... 146
4.4.3 BDTonline ...... 150
5 Pre-Supernova Neutrino Sensitivity at SK-Gd 155 5.1 Previous Estimations ...... 156 5.2 Results for First Gadolinium Loading ...... 161 5.2.1 Final Selection ...... 162 5.2.2 Statistical Parameters ...... 164 5.2.3 Results ...... 167 5.3 Future Gadolinium Loading Phases ...... 172
6 Pre-Supernova Alarm 177 6.1 Preliminary Considerations ...... 178 6.2 Alarm Decision ...... 180 6.2.1 False Positive Rates ...... 181 6.3 System design ...... 182 6.4 Prospects ...... 188
7 Prospects for Low Energy detection in Hyper-Kamiokande exper- iment 190 7.1 The mPMT concept ...... 192 7.1.1 mPMT Geometry for Hyper-K ...... 193 7.1.2 Acrylic Vessel ...... 196 7.2 mPMT Tests at Memphyno ...... 197 7.3 Reduction of mPMT dark rates with BDT ...... 201 7.4 Detection of Supernova and Pre-Supernova Neutrinos in Hyper-K . 206
8 Conclusion (English) 210
9 Conclusioni (Italian) 213
10 Conclus˜ao (Portuguese) 216
A Nearby pre-Supernova Candidates 219 Contents 10
Bibliography 222 List of Figures
1.1 Illustration of atmospheric neutrinos production from cosmic-rays interactions in the upper atmosphere. Reproduced directly from [17]. 40 1.2 Latest atmospheric neutrino analysis fit results in Super-Kamiokande. These parameters will be explained in Section 1.1.From[20]. .... 42 1.3 Energy spectrum of Solar Neutrinos for their four most import sources: pp, 8B, 7Be and pep.Thefigureshowsalsotheenergy thresholds for di↵erent experiments that reported solar neutrino measurements related to each cycle. Ga: Gallium experiment, Cl: Chlorine experiment, SK: Super-Kamiokande, and SNO: Sudbury Neutrino Observatory. From [12]...... 43 1.4 Schematic view of the pp-chains and CNO nuclear fusion sequences. From [23]...... 44 1.5 Scheme of T2K experiment showing the neutrino beam produced at J-PARC and being detected by near detectors a few meters away from production and later by Super-Kamiokande, 295 km away [6]. 48 1.6 Cross sections for low energy neutrino interactions. Reproduced from [43]...... 50 1.7 Feynman diagram representing the Inverse Beta Decay interaction. . 52
2.1 Supernova Classification. Reproduced from [48]...... 55 2.2 Nearby core collapse Supernova candidates within 1 kpc. Colored labels show the star’s spectral Type. Masses and distances of the stars are shown in parenthesis. From [53]...... 58 2.3 Comparison of the time integrated e↵ective anti-neutrino spectrum from SN1987A between data from Kamiokande and IMB. Repro- duced from [65]...... 63 2.4 Predicted DSNB spectrum for di↵erent models and showing the three main backgrounds for this measurement and the energy region where they apply. Reproduced from [65]...... 65 2.5 Characteristic onion-shape of a massive star just before the core collapse, where the shells show the abundance of each element. Re- produced from [16]...... 67 11 List of Figures 12
2.6 Normalized spectrum (normalized energy distribution function) of the pair-annihilation neutrinos emitted during C (solid), Ne (dashed), O (dotted) and Si (dot-dashed) burning stage and the solar pp neu- trinos (thin line), from [60]. The spectrum was calculated based on the Monte Carlo method of [70]...... 69 2.7 For a 20 solar masses star, the emitted anti-neutrino spectrum for di↵erent stages. Dashed line: carbon burning, dotted line: neon burning, short-dashed line: oxygen burning, and solid line: silicon burning. Reproduced from [71]...... 72 2.8 IBD rate in Super-Kamiokande for the 12 hours before the core- collapse for di↵erent star evolution models. It is assumed a distance of 200 parsecs. Reproduced from [16]...... 73
2.9 Comparison of the considered models for the⌫ ¯e luminosity and mean energy prior to the core-collapse. Reproduced from [16]. ... 74
3.1 Schematic overview of the Super-Kamiokande Detector ...... 77 3.2 Arrangement of PMTs in modules inside the Super-Kamiokande Detector. From [5]...... 79 3.3 Picture taken during the refurbishment work in 2018 of the Super- Kamiokande inner detector. Copyright: Kamioka Observatory, ICRR (Institute for Cosmic Ray Research), The University of Tokyo. ... 80 3.4 Spectrum of the Cherenkov radiation in water...... 84 3.5 Scheme of the Super-Kamiokande water system. Reproduced from [5]...... 85 3.6 Scheme of the EGADS water system to simulate the gadolinium sulfate injection in Super-Kamiokande. AE = anion exchange resin, DI = dionisation, TOC = total organic coumpound lamp, RO = reserve osmosis. Reproduced from [79]...... 88 3.7 Water transparency measurements of EGADS compared to SK- III/IV phases. Reproduced from [79]...... 89 3.8 Most recent water transparency measurements in EGADS, that from March 31st, 2018 started to simulate the first phase of SK- Gd...... 90 3.9 General scheme of the 20-inch Photo-Multiplier. From [5]...... 92 3.10 Quantum E ciency for 20-inch PMT. From [5]...... 92 3.11 Picture taken during the full reconstruction of Super-Kamiokande in 2005, showing the PMT with the Fiber Reinforced Case (FRC). Copyright: Kamioka Observatory, ICRR (Institute for Cosmic Ray Research), The University of Tokyo...... 94 List of Figures 13
3.12 Picture taken during the refurbishment work in 2018 of a new PMT being inspected installing in Super-Kamiokande. Copyright: Kamioka Observatory, ICRR (Institute for Cosmic Ray Research), The University of Tokyo...... 95 3.13 Likelihood values as a function of the total electron energy and opening angle between the reconstructed direction and the direction from vertex to each PMT. From [85]...... 100 3.14 Energy resolution as a function of energy found using Monte Carlo events in [77]...... 102 3.15 Scheme of LINAC calibration at the SK detector. The dashed line represents the fiducial volume. The black dots indicate where the calibration data is usually taken. From [88]...... 105 3.16 Neutron capture e ciency for di↵erent gadolinium sulfate concen- trations in Super-Kamiokande. Edited from [92]...... 108 3.17 Scheme of EGADS detector. On the left a side view of the tank and on the right the floor and the PMT support frame. Reproduced from [93]...... 109 3.18 Uranium 238 decay chain, from [95]...... 111 3.19 Thorium 232 decay chain, from [95]...... 112 3.20 Actinium 227/Uranium 235 decay chain, from [96]...... 113 3.21 Scheme of the gadolinium sulfate dissolving plan to the Super- Kamiokande detector. From [92]...... 114 3.22 Evolution of the water transparency in the Super-Kamiokande de- tector and of the PMTs dark rates for the end of SK-V and first months of SK-VI (after gadolinium loading)...... 116 3.23 Emission energies in the two considered models for TNC on gadolin- ium. From [16]...... 117 3.24 Multiplicity of -rays in the two considered models for TNC on gadolinium. From [16] ...... 118 3.25 Comparison of di↵erent variables for the two models of TNC on gadolinium in consideration. From [16]...... 119
4.1 WIT e ciency compared to electron true total energy for events in the whole ID and only inside the FV. The Cherenkov threshold and SK-IV solar neutrino analysis threshold are showed for comparison. 123 4.2 Comparison of expected accidental coincidences events over time for SK-IV and SK-V pure water data. For SK-V, results with improved selection techniques are shown. These techniques will be discussed in Section 4.4.3...... 126 4.3 Comparison between reactor fluxes for the di↵erent years considered to estimate the background due to reactor neutrinos...... 131 List of Figures 14
4.4 Background rates for the di↵erent years in consideration. Results with improved techniques are shown. These techniques will be dis- cussed in Section 4.4.3...... 131 4.5 Distribution of capture times for the fast neutrons and for neutrons from spallation daughters. From [16]...... 133 4.6 Distribution of the energies of spallation daughters. From [16]. ..134 4.7 Variables dR and dT for simulated signal events (black) and acci- dental background (red). From [16]...... 138 4.8 Variables from BONSAI Online reconstruction (bx, by, bz, bt, bgood- ness) and n18. The black solid line shows the variables before any cut, the red dashed line shows the variables for coincidence positron events and blue for coincidence neutron events after the Initial Search.139 4.9 Variables dR and dT calculated after the pre-selection...... 141 4.10 Output from the TMVA software [108]showingarandomdecision tree that was used for the training for signal and background sepa- ration in the pre-Supernova neutrino analysis. Further discussions regarding the discriminating variables used in this scheme will be held in next sections...... 144 4.11 ROC curves example for di↵erent classifiers. For this example, in comparison to the Boosted Decision Tree (BDT) ROC curve, are the Fisher Linear Discriminant (Fisher) and the rectangular cuts (Cuts)...... 146 4.12 ROC curves from reproduced BDT used in [16](black),thenew BDT with o✏ine variables used for Final Selection (blue) and two di↵erent trials to get the best performance out of a training using only online variables (green and yellow)...... 151 4.13 Background and Signal separation for the BDT with online variables.152
5.1 Distribution of the variables from the DC channel in the final se- lection for [16]results...... 157 5.2 Distribution of the BDT score for the single channel in the final selection for [16]results...... 157 5.3 Distribution of variables in the coincidence channel. Solid lines show before the final selection and dashed lines after...... 163 5.4 Evolution of significance level for the normal ordering neutrino mass models for stars at 200 pc in the last hours before the core-collapse at SK-Gd with 0.02% Gd (SO ) 8H O...... 170 2 4 3 · 2 List of Figures 15
5.5 Expected early warning above 3 against the distance for 15 M and 25 M stars for the considered pre-Supernova models, evaluated in steps of 0.01 kparsec. Dashed line shows predictions using the GLG4SIM -ray model and solid line for the ANNRI -ray model evaluated over a 7-hour window...... 171 5.6 Expected early warning considering a rate of 3.2 Supernovae per century against the distance for 15 M and 25 M stars for the considered pre-Supernova models, evaluated in steps of 0.01 kpar- sec. Dashed line shows predictions using the GLG4SIM -ray model and solid line for the ANNRI -ray model evaluated over a 7-hour window...... 172
6.1 Edited from [83]. Illustration of the acquisition system in Super- Kamiokande and where the pre-Supernova alarm is placed...... 179 6.2 Screenshot of the main menu of the alarm system after initialization.183 6.3 Screenshot of the alarm view when data is being processed. Infor- mation regarding last evaluations are also showed in the screen. ..183 6.4 Main web page of the alarm system used to verify latest evaluations and the alarm status...... 185 6.5 Screenshot of the secondary web page showing plots for the evo- lution of evaluated significance levels and selected IBD pairs for di↵erent time intervals...... 187 6.6 Screenshot of the secondary web page showing plots of di↵erent variable distributions over the last 12 hours for the events selected by the system...... 188
7.1 Illustrated scheme of the Hyper-K Detector...... 191 7.2 Illustration of the mPMT module...... 194 7.3 The two initial prototypes for Hyper-K from INFN (left) and TRI- UMF (right)...... 194 7.4 Assembling of the mPMT prototype in INFN Naples...... 195 7.5 Acrylic samples transmittance ...... 197 7.6 Picture of the MEMPHYno test bench at APC in Paris, where the mPMT prototype developed by INFN Naples is being tested. ....198 7.7 Results of dark rate evaluation of di↵erent PMTs in the mPMT module prototype tests at MEMPHYno...... 200 7.8 Vertex resolution (top) and electron/muon separation (bottom) im- proved by the use of 3-inch PMTs (red) in comparison with 20-inch (blue) and 8-inch (black). From [122]...... 202 7.9 ROC curves for di↵erent trials of BDT trainings to be used for the software dark rate reduction of mPMTs...... 204 List of Figures 16
7.10 Signal and background separation for the most e cient BDT of Figure 7.9...... 206 7.11 Predicted Inverse Beta Decay reactions due to Supernova Relic Neu- trinos in function of years for current and planned experiments, from [128]...... 207 List of Tables
2.1 Approximate duration of burning stages for a 20 M star and the fraction and average energy of electron neutrinos emitted by pair- annihilation [69]...... 68
3.1 Trigger schemes in Super-Kamiokande...... 97 3.2 Properties of Gd isotopes, in comparison with Hydrogen. From the information in [79]...... 106
4.1 Radioisotopes activity level required for the Supernova Relic Neu- trino and Solar Neutrino analyses...... 128 4.2 Rates for backgrounds per hour due to radioisotopes contamination on gadolinium sulfate. Requirements for the current analysis are shown in Table 4.1.Previousrequirementsforcontaminationdi↵er from current ones for 235U < 3mBq/kg...... 130 4.3 Ranked importance of the used online variables...... 152 4.4 Approximate time spent to process data for previous approach and optimized (with BDTonline)...... 153
5.1 Number of expected signal events from pre-Supernova star at 200 parsecs in the final 12 hours before core-collapse at SK-Gd with 0.2% Gd (SO ) 8H O,frompreviousestimations[16]...... 158 2 4 3 · 2 5.2 Maximum range of detection of pre-Supernova stars by SK-Gd with 0.2% Gd (SO ) 8H O,frompreviousestimations[16]...... 159 2 4 3 · 2 5.3 Expected early warning in SK-Gd with 0.2% Gd2(SO4)3 8H2O, from previous estimations [16]...... · 160 5.4 Remaining background events per 7 hour window for the DC events after final selection...... 164 5.5 Number of expected signal events from pre-Supernova star at 200 parsecs in the final 7 hours before core-collapse at SK-Gd with 0.02% Gd2(SO4)3 8H2O.Resultsareshownfortwo -ray emis- sion models...... · 168
17 List of Tables 18
5.6 Maximum range of detection of pre-Supernova stars by SK-Gd with 0.02% Gd2(SO4)3 8H2O,fornormalorderingneutrinomassmodels. Results are shown· for two -ray emission models evaluated over a 7-hour window...... 169
5.7 Expected early warning in SK-Gd with 0.02% Gd2(SO4)3 8H2O,for Betelgeuse-like models with normal ordering neutrino mass· models. Results are shown for two -ray emission models evaluated over a 7-hour window...... 171 5.8 Expected rates for the Spontaneous Fission and Pairs of Neutrons backgrounds for the three phases of SK-Gd...... 173 5.9 Maximum range of detection of pre-Supernova stars by SK-Gd with 0.06% Gd2(SO4)3 8H2O,fornormalorderingneutrinomassmodels. Results are shown· for two -ray emission models evaluated over a 7-hour window...... 174 5.10 Maximum range of detection of pre-Supernova stars by SK-Gd with 0.2% Gd2(SO4)3 8H2O,fornormalorderingneutrinomassmodels. Results are shown· for two -ray emission models evaluated over a 7-hour window...... 175
5.11 Expected early warning in SK-Gd with 0.06% Gd2(SO4)3 8H2O,for Betelgeuse-like models with normal ordering neutrino mass· models. Results are shown for two -ray emission models evaluated over a 7-hour window...... 175
5.12 Expected early warning in SK-Gd with 0.2% Gd2(SO4)3 8H2O,for Betelgeuse-like models with normal ordering neutrino mass· models. Results are shown for two -ray emission models evaluated over a 7-hour window...... 176
6.1 Estimated early warning lead time of the pre-Supernova alert system.186
7.1 Variables in order of importance of the BDT for the software Dark Rate reduction in Hyper-Kamiokande...... 205 7.2 Expected maximum range of detection of pre-Supernova stars in Hyper-K loaded with 0.02% Gd2(SO4)3 8H2O,fornormalordering neutrino mass models...... · 209 7.3 Expected early warning in Hyper-K loaded with 0.02% Gd (SO ) 2 4 3 · 8H2O, for Betelgeuse-like models with normal ordering neutrino mass models...... 209
A.1 List of pre-Supernova candidates with estimated masses and dis- tances. Reproduced from [53]...... 221 Abbreviations
M Solar Masses Ac Actinium Adaboost Adaptive boost AmBe Americium-Beryllium BDT Boosted Decision Tree
BDTonline Boosted Decision Tree with online variables BONSAI Branch Optimization Navigating Successive Annealing Iterations CCarbon CC Charged Current CCSN Core-Collapse SuperNova DAQ Data AcQuisition DC Delayed Coincidence DSNB Di↵use Supernova Neutrino Background EGADS Evaluating Gadolinium’s Action on Detector Systems Fe Iron FPR False Positive Rate FRP Fibre Reinforced Plastic FV Fiducial Volume HHydrogen HK Hyper-Kamiokande
19 Abbreviations 20
HPG Hyper-Pure Germanium IBD Inverse Beta Decay ID Inner Detector IWCD Intermediate Water Cherenkov Detector JUNO Jiangmen Underground Neutrino Observatory LD Fisher’s Linear Discriminant LEAF Low Energy Algorithm Framework LMC Large Magellanic Cloud MO Mass Ordering mPMT multi-PMT MSW Mikheyev-Smirnov-Wolfenstein NNitrogen NC Neutral Current Ne Neon NSE Nuclear Statistical Equilibrium OOxygen OD Outer Detector PE PhotoElectron PMT PhotoMultiplier Tube QE Quantum E ciency ROC Response Operator Characteristic RSG Red SuperGiant SF Spontaneous Fission Si Silicon SK Super-Kamiokande SM Standard Model Physics SN SuperNova SNe SuperNovae Abbreviations 21
SRF Supernova Relic Flux Th Thorium TNC Thermal Neutron Capture UUranium WC Water-Cherenkov WCSim Water Cherenkov Simulator WIT Wideband Intelligent Trigger ZAMS Zero Age Main Sequence To my parents. . .
22
Introduction (English)
In 2020, the Super-Kamiokande experiment has moved to its next stage, SK-
Gd, in which gadolinium sulfate was added to the water in the detector. This improves the capability of neutron identification, which has great implications for the detection of low energy neutrinos. Approximately 13 tons of gadolinium sulfate was dissolved to achieve the initial phase of 0.02% concentration by mass.
The observation of astronomical objects beyond our solar system by neutrino detection is called Neutrino Astronomy and it was initiated with the first observa- tion of neutrinos from Supernova in 1987 by the Kamiokande experiment. SK-Gd has the potential to contribute significantly for Neutrino Astronomy, observing neutrinos from di↵erent astronomical sources such as pre-Supernova stars.
During the last stages of the evolution of massive stars, which have mass greater than eight solar masses, the production of neutrino and anti-neutrino pairs become main stellar cooling mechanism. This characteristic behaviour goes over for a few hours preceding the core-collapse supernova. The emitted electron anti-neutrino 24 25 exceeds the threshold for inverse beta decay in hydrogen, so that detection would be possible in Super-Kamiokande. Neutrons, which result from the inverse beta decay, are subjected to capture by Hydrogen nuclei, emitting gamma-rays with an energy of 2.2 MeV. However, for gadolinium, the neutron capture would emit gamma ray cascades with approximately 8 MeV. SK-Gd is characterized for having reduced backgrounds and enhanced e ciency for neutron tagging. The experiment is now capable of detecting neutrinos emitted in the final hours of massive star lives, known as pre-Supernova Neutrinos.
The detection of pre-Supernova neutrinos could provide an early warning for
Supernova events, sending alerts to the community in case of a potential core- collapse Supernova within a few hours. The goal of this project is to create an alarm system for Supernovae based on the detection of pre-Supernova neutrinos.
This system performs a real time search for inverse beta decay candidates on low energy events in SK-Gd and apply the necessary statistical tests, saving long term information to make alarm decisions.
In this thesis the latest results of detection capabilities of pre-Supernova Neutri- nos at Super-Kamiokande with Gadolinium will be presented. Chapters 1 and 2 in- troduce physical concepts of Neutrino physics and Supernova and Pre-Supernova.
Chapter 3 discusses about the Super-Kamiokande detector and the characteristics of its new phase, SK-Gd. In Chapter 4, the backgrounds for the pre-Supernova 26
Neutrino detection will be shown, as well as the multivariate techniques used for event selection.
Chapters 5 and 6 will present the improved models and sensitivities for pre-
Supernova Neutrino detection and the application of this study, which is the pre-
Supernova alert system.
At last, Chapter 7 will show prospects of Supernova and pre-Supernova neutrino detection in next generation experiment Hyper-Kamiokande. Introduzione (Italian)
Nel 2020, l’esperimento Super-Kamiokande ha iniziato una nuova fase sperimen- tale, SK-Gd, aggiungendo polvere di gadolinio alle 50.000 tonnellate di acqua ultrapura del rivelatore. Ci`omigliora la capacit`adi identificazione dei neutroni, con importanti implicazioni per la rilevazione di neutrini di bassa energia. Circa
13 tonnellate di solfato di gadodinio sono state utilizzate per ottenere, in questa prima fase, la concentrazione percentuale in massa dello 0,02%.
L’osservazione dei neutrini da sorgenti astrofisiche, chiamata Astronomia Neu- trinica, `enata con la prima osservazione di neutrini da Supernova nel 1987 dall’esperimento
Kamiokande, estendendo l’astronomia classica oltre la radiazione elettromagnet- ica. Il progetto SK-Gd ha il potenziale per contribuire in modo significativo all’Astronomia Neutrinica, osservando i neutrini provenienti da diverse sorgenti astrofisiche.
Durante le ultime fasi dell’evoluzione delle stelle massicce, che hanno cio`emassa maggiore di otto masse solari, la produzione di neutrini e antineutrini diventa il 27 28 principale meccanismo di ra↵reddamento. Questa emissione caratteristica si pro- trae per alcune ore prima del collasso del nucleo della supernova. Gli antineutrini elettronici emessi in questa fase hanno energia su ciente per il processo di decadi- mento beta inverso in idrogeno, e la loro rivelazione `equindi possibile in Super-
Kamiokande. I neutroni prodotti nel decadimento beta inverso, sono catturati dai nuclei di idrogeno, emettendo raggi gamma con un’energia di 2,2 MeV. Il processo di cattura neutronica con il gadolinio produce una cascata di raggi gamma con energia di circa 8 MeV, migliorando le possibilit`adi rivelazione.
L’attuale fase sperimentale SK-Gd `equindi caratterizzata da una maggiore ef-
ficienza per l’identificazione dei neutroni permettendo la rivelazione di neutrini emessi nelle ultime ore di vita di stelle massicce, noti come neutrini da pre-
Supernova. La rivelazione di neutrini da pre-Supernova potrebbe fornire alla co- munit`aastrofisica un allarme immediato e a dabile al verificarsi di una Supernova nella Galassia.
L’obiettivo di questo lavoro di tesi `eappunto la realizzazione di un sistema di allarme per Supernovae basato sulla rilevazione di neutrini da pre-Supernova. Per sviluppare questo sistema, `erealizzata una ricerca in tempo reale di eventi di bassa energia di decadimento beta inverso in SK-Gd e il sistema di allarme `ebasato sullo studio della significativit`astatistica.
In questa tesi sono presentati gli ultimi risultati sulla capacit`adi rivelazione di 29 neutrini da pre-Supernova in Super-Kamiokande con Gadolinio. I capitoli 1 e 2 introducono i concetti principali della fisica dei neutrini, delle Supernova e pre-
Supernova. Il capitolo 3 descrive il rilevatore Super-Kamiokande e le caratteristiche dell’attuale fase sperimentale SK-Gd. Nel Capitolo 4 sono discussi le sorgenti di background per la rivelazione di neutrini da pre-Supernova e gli algoritmi di analisi multivariata utilizzati per la selezione degli eventi. Nei capitoli 5 e 6 sono presentati i modelli e le sensibilit`asperimentali per la rivelazione di neutrini da pre-Supernova e lo sviluppo del sistema di allarme.
Infine, il Capitolo 7 presenta le prospettive di rivelazione di neutrini da Super- nova e pre-Supernova nel futuro esperimento Hyper-Kamiokande. Introdu¸c˜ao (Portuguese)
Em 2020, o experimento Super-Kamiokande passou para sua pr´oxima fase, SK-
Gd, no qual o sulfato de gadol´ınio foi adicionado `a´agua no detector, melhorando acapacidadedeidentifica¸c˜aodenˆeutrons,oquetemgrandesimplica¸c˜oesparaa detec¸c˜ao de neutrinos de baixa energia. Aproximadamente 13 toneladas de sulfato de gadol´ınioforam dissolvidas para atingir a fase inicial de 0,02% de concentra¸c˜ao em massa.
Aobserva¸c˜aodeobjetosastronˆomicos,paraal´emdonossosistemasolar,por detec¸c˜ao de neutrinos ´econhecida como ”Astronomia de Neutrinos” e foi iniciada com a primeira observa¸c˜aode neutrinos de Supernova em 1987, pelo experimento
Kamiokande. SK-Gd tem o potencial de contribuir significativamente para a As- tronomia de Neutrinos, observando neutrinos de diferentes fontes astronˆomicas, como estrelas pr´e-Supernova.
Durante os ´ultimosest´agiosde evolu¸c˜aodas estrelas massivas, que s˜aoestrelas com massas superiores a oito massas solares, a produ¸c˜aode pares de neutrinos e 30 31 anti-neutrinos torna-se o principal mecanismo de resfriamento estelar. Este com- portamento caracter´ıstico continua por algumas horas antes do colapso do n´ucleo da estrela. Os anti-neutrinos de el´etrons emitidos excedem o limite para o De- caimento Beta Inverso no hidrogˆenio, de modo que tal detec¸c˜ao seria poss´ıvel no Super-Kamiokande. Os nˆeutrons, que resultam do Decaimento Beta Inverso, s˜aosubmetidos `acaptura por n´ucleos de hidrogˆenio, emitindo raios gama com energia de 2,2 MeV. No entanto, para o gadol´ınio, a captura de nˆeutrons emite cascatas de raios gama com aproximadamente 8 MeV. SK-Gd ´ecaracterizado por ter backgrounds reduzidos e eficiˆencia aprimorada para a identifica¸c˜ao de nˆeutrons.
Atualmente, o experimento ´ecapaz de detectar neutrinos emitidos nas horas finais da vida de estrelas massivas, conhecidos como neutrinos pr´e-Supernova.
A detec¸c˜ao de neutrinos pr´e-Supernova pode fornecer um alerta precoce para fenˆomenos de Supernova, enviando alertas para a comunidade no caso de um poss´ıvel colapso do n´ucleo de uma estrela dentro de algumas horas. O objetivo deste projeto ´ecriar um sistema de alarme para Supernovas baseado na detec¸c˜ao de neutrinos pr´e-Supernova. Este sistema realiza uma busca em tempo real por candidatos de Decaimento Beta Inverso em eventos de baixa energia no SK-Gd e aplica os testes estat´ısticosnecess´arios,salvando informa¸c˜oesde longo prazo para tomar as decis˜oes de envio de alertas.
Nesta tese, ser˜ao apresentados os resultados mais recentes das capacidades de 32 detec¸c˜ao de Neutrinos pr´e-Supernova no Super-Kamiokande com Gadol´ınio. Os cap´ıtulos 1 e 2 introduzem conceitos f´ısicos de F´ısica de Neutrino e Supernova e
Pr´e-Supernova. O Cap´ıtulo 3 discute sobre o experimento Super-Kamiokande e as caracter´ısticas de sua nova fase, SK-Gd. No Cap´ıtulo 4, os backgrounds para a detec¸c˜ao pr´e-Supernova Neutrino ser˜ao mostrados, bem como as t´ecnicas de estat´ıstica multivari´avel usadas para a sele¸c˜aode eventos.
Os cap´ıtulos 5 e 6 apresentar˜ao os modelos aprimorados e as sensibilidades para adetec¸c˜aodeneutrinospr´e-Supernovaeaaplica¸c˜aodesteestudo,que´eosistema de alerta de pr´e-Supernova.
Por fim, o Cap´ıtulo 7 mostrar´aas perspectivas de detec¸c˜ao de neutrinos de Su- pernova e pr´e-Supernova no experimento de pr´oxima gera¸c˜ao Hyper-Kamiokande. Chapter 1
Neutrinos
The main candidates to reveal new physics, Neutrinos, are currently the focus of several experiments worldwide dedicated to study these mysterious particles.
The initial concept of neutrinos was introduced in 1930 by W. Pauli as being a neutral particle to explain missing energy in beta decays [1]. Later in 1933 they were given the name of neutrinos by E. Fermi [2]. Their first observation was in
1956 by C. Cowan and F. Reines [3]. Since then, many experiments were design to study their properties (see, e.g., [4]andreferencestherein).Currently,Super-
Kamiokande [5], T2K [6], SNO [7], IceCube [8], and many other experiments are in the search of Neutrino’s oscillation parameters and masses. The understanding of their physical properties might prove the matter-antimatter asymmetry as given by the CP violation [9; 10; 11].
33 Neutrinos 34
Some examples of neutrinos coming from natural sources include: the ones from the decay of the products of cosmic ray interactions with the atmosphere
(Atmospheric Neutrinos), from nuclear reactions in the Sun (Solar Neutrinos), or, the one with most importance to this thesis, generated from astrophysical sources.
More specifically, neutrinos from the di↵erent stages before and after the explosion of massive stars, known as Supernova, are the object of this study.
Neutrinos can also be produced artificially, like in the T2K experiment, in which the production of neutrinos are of big importance since it is known the production properties such as direction, energy, flavor, etc.
The following sections will introduce di↵erent sources of neutrinos, current experiments and physical properties. The discussion of Supernovae and pre-
Supernova Neutrinos will be the focus of next chapter.
1.1 Neutrino Oscillations in Vacuum
Neutrinos exist in three flavours in the Standard Model, in which each one of them forms a doublet with the associated charged lepton. They are represented by the weak-eigenstates ⌫e, ⌫µ and ⌫⌧ .
These eigenstates are usually written as linear superposition of three mass eigen- states ⌫1, ⌫2 and ⌫3,withrespectivemassesm1, m2 and m3.Therelationbetween Neutrinos 35 mass and weak eigenstates is given by Equation (1.1), in which the unitary ma- trix U corresponds to the Pontecorvo-Maki-Nakagawa-Sakata (PMNS) matrix, or neutrino mixing matrix [12].
⌫ = U ⇤ ⌫ , (1.1) | ↵i ↵j| ji j X where flavour eigenstates are represented by greek letters and mass eigenstates by latin letters.
The neutrino mixing matrix can be parametrized by three mixing angles ✓12,
✓13 and ✓23,andaCPviolatingphase .
i 10 0 c13 0 s13e c12 s12 0 0 1 0 1 0 1 U = B0 c23 s23C B 010C B s12 c12 0C B C B C B C B C B C B C B C B i C B C B0 s23 c23C B s13e 0 c13 C B 001C B C B C B C @ A @ A @ A (1.2)
i c12c13 s12c13 s13e 0 1 = i i , B s12c23 c12s13s23e c12c23 s12s13s23e c13s23 C B C B C B i i C B s12s23 c12s13s23e c12c23 s12s13c23e c13c23 C B C @ A where cab = cos(✓ab)andsab = sin(✓ab), which (a, b)=(1, 2, 3), corresponding to the mass eigenstates. This parametrization is written in a way that, without the Neutrinos 36
CP violating phase, it is equivalent to a rotation in three dimensions.
If Neutrinos are Majorana particles, two new phases need to be added to (1.2).
However, the Majorana phases have no e↵ect when looking into neutrino oscilla- tions, since they cancel out when evaluating the flavour evolution [12].
i 10 0 c13 0 s13e c12 s12 0 0 1 0 1 0 1 U = B0 c23 s23C B 010C B s12 c12 0C ⇥ B C B C B C B C B C B C B C B i C B C B0 s23 c23C B s13e 0 c13 C B 001C B C B C B C @ A @ A @ A (1.3)
ei↵1/2 00 0 1 i↵2/2 ⇥ B 0 e 0C B C B C B C B 001C B C @ A
When studying neutrino oscillations in vacuum, the probability of a neutrino with flavor ”↵”tobecomeaneutrinowithflavor” ”whentravellingadistance Neutrinos 37
”L”, is given by the oscillation probability (1.4)[12].
2 2 L imj 2E P (⌫↵ ⌫ )= U jU↵⇤je ! j X = U 4 U 4+ | j| | ↵j| j=1,3 X (1.4) 2 mjkL + 2Re[U U ⇤ U ⇤ U ]cos + j k ↵j ↵k 2E Xj An important definition arising from equation (1.4) is of the squared mass dif- ference m2 = m2 m2. A still open question in Neutrino Physics is related to jk k j whether the eigenstate m3 is lighter or heavier than m1 and m2. So, measurements of oscillation parameters by experiments with known length ”L” and energy ”E” are very important to answer questions about the Neutrino Mass Hierarchy [13]. 2 Usually, models with the mass hierarchy m1 2 called Normal Ordered models, and for m1 When applying CP transformation to equation (1.4), in order to obtain the correspondent probability for anti-neutrino oscillation, we see that the oscillation Neutrinos 38 probability will only violate CP if the phase is di↵erent from 0 or ⇡.So,again, measurements of the phase are important for the studies of CP violation in the lepton sector and understanding the matter-antimatter asymmetry in the Universe. More details can be found in the review work [14] and references therein. 1.2 The MSW E↵ect The Neutrino oscillations introduced in last section happen in vacuum. However, when neutrinos propagate through matter, they will be a↵ected by interactions, changing the pattern of neutrino oscillations. This change occurs because electron neutrinos can interact with electrons in matter through both neutral current (NC) and charged current (CC), while the other neutrino flavors, muon and tau, can only interact through NC. This e↵ect of oscillations in matter is called Mikheyev- Smirnov-Wolfenstein (MSW) e↵ect [15]. For Neutrinos coming from stars, the path from the core to the surface will lead to resonant conversion between neutrino flavours [16]. The Hamiltonian describing a charged current interaction, mediated by the W boson, will have an additional term, leading to the e↵ective mass term: A =2p2(GF Ye/mn)⇢E, (1.5) Neutrinos 39 where Ye is the electron fraction of the matter, mn is the nucleon mass, ⇢ is the density and GF is the Fermi’s constant. For the two flavours oscillation case, the e↵ective mass splitting will be (( m2)cos(2✓) A)2 +(( m2)sin(2✓))2 (1.6) p The resonance occurs at the density: ( m2)cos(2✓) ⇢ = (1.7) 2Ep2GF 2 The MSW resonance in the Sun depends on m21 and it is called low resonance and doesn’t apply to anti-neutrinos, which is di↵erent than the high resonance 2 by m32,whichappliestoanti-neutrinossoitisrelevantforthepre-Supernova studies [16]. 1.3 Neutrino Sources 1.3.1 Atmospheric Neutrinos Protons and heavy nuclei arriving from anywhere in the Universe to the Earth are known as Cosmic Rays. These particles come in an isotropic and constant flux. Neutrinos 40 Figure 1.1: Illustration of atmospheric neutrinos production from cosmic-rays interactions in the upper atmosphere. Reproduced directly from [17]. When they interact with the upper atmosphere, they typically produce unstable mesons called Pions ⇡ and Kaons K, which decay into other particles, including neutrinos. We call them Atmospheric Neutrinos. Neutrinos 41 The possible decays are: K+ µ+ + ⌫ e+ +¯⌫ + ⌫ ! µ ! e µ K µ +¯⌫ e + ⌫ +¯⌫ ! µ ! e µ K+ ⇡+ + ⇡0 ! + 0 K ⇡ + ⇡ ! + + + K ⇡ + ⇡ + ⇡ ! (1.8) + + K ⇡ + ⇡ + ⇡ ! K+ ⇡0 + e+ + ⌫ ! e + 0 K ⇡ + e +¯⌫ ! e ⇡+ µ+ + ⌫ e+ +¯⌫ + ⌫ ! µ ! e µ ⇡ µ +¯⌫ e + ⌫ +¯⌫ ! µ ! e µ The computation of these neutrino fluxes is very di cult, once it needs to take into account, among other factors, the di↵erent composition of Earth’s atmosphere depending on the location, the e↵ects from the Sun on cosmic rays and the dif- ferent channels of production. Also, the energy of the neutrinos depends on the original cosmic ray, which increases significantly their energy range. See the review work [18] and references within for more information regarding the calculation of neutrino fluxes. Neutrinos 42 Figure 1.2: Latest atmospheric neutrino analysis fit results in Super- Kamiokande. These parameters will be explained in Section 1.1. From [20]. The observation of atmospheric neutrinos was responsible for the first indication of neutrino oscillations [19]. Figure 1.2 shows the latest fit results for oscillation parameters in Super-Kamiokande [20]. Experiments studying atmospheric neutrinos are: Super-Kamiokande [5], SNO [7], MINOS [21], and IceCube [8]. The energy range in which they are observed is from 100 MeV to higher than 10 TeV [17]. Neutrinos 43 Figure 1.3: Energy spectrum of Solar Neutrinos for their four most import sources: pp, 8B, 7Be and pep. The figure shows also the energy thresholds for di↵erent experiments that reported solar neutrino measurements related to each cycle. Ga: Gallium experiment, Cl: Chlorine experiment, SK: Super- Kamiokande, and SNO: Sudbury Neutrino Observatory. From [12]. 1.3.2 Solar Neutrinos Neutrinos are also produced in thermonuclear reactions inside the stars. For the Sun, di↵erent spectra of neutrinos are emitted depending which ”cycle” of reaction is burning Hydrogen. Solar neutrinos fluxes are predicted by the Solar Standard Model (SSM) devel- oped by J. Bahcall [22]. Electron neutrinos are products of thermonuclear fusions in the solar core, which is how our Sun obtains its thermal energy. Figure 1.4 shows a schematic view of the pp-chains and CNO cycles of the Sun. In summary, Neutrinos 44 Figure 1.4: Schematic view of the pp-chains and CNO nuclear fusion se- quences. From [23]. the e↵ective fusion reaction, which proceeds through all these chains and cycles, is given by: 4 4p +2e He +2⌫ +26.73MeV (1.9) ! e Estimations to the time the heat takes to travel from the center to the surface of the Sun are uncertain, between 100,000 to 1,000,000 years. However, the generated neutrinos reach the Earth in only 8 minutes approximately. 10 2 1 The flux of electron neutrinos that reaches the Earth is of 6 10 (cm s) .The ⇥ energy produced by the solar nuclear reactions will be shared among the particles in the final state. Neutrinos are capable of escaping from the Sun, carrying part of this energy. Since they are produced through di↵erent cycles and chains, it Neutrinos 45 implies di↵erent neutrino energy distributions. Figure (1.3)showsdi↵erentenergy spectrum related to those cycles and chains, as well as which experiments reported the measurements related to each one of them. The combined observations from Super-Kamiokande and SNO in 2001 gave evi- dence for the neutrino oscillations. Super-Kamiokande used neutrino-electron elas- tic scattering, ⌫ +e ⌫ +e ,tomeasuretherecoilofelectronsandzenithangle e ! e dependence of solar neutrinos, while SNO measured neutrino-deuteron charged- current reaction rate [24], ⌫ + d ⌫ + p + p,andexploititsuniquecapabilityto e ! e observe neutral current interactions, evaluating the flux of all neutrino types [25]. These measurements combined provided evidence of electron neutrino oscillation into other flavors of neutrinos. 1.3.3 Reactor Neutrinos Reactors are currently one of the main artificial sources of neutrinos. It is a controllable source of electron anti-neutrino with low energy and it has a great importance for the topic of this thesis, since it is the main background for the detection of pre-Supernova neutrinos. The fission of Uranium and Plutonium isotopes 235U, 239Pu, 238Uand241Pu is what give power to Reactors. It produces lighter elements, which will also decay Neutrinos 46 and contribute to the energy. The beta decay of products is the source of electron anti-neutrinos. The energy range of reactor neutrinos is 1 to 10 MeV, with mean energy peak around 4 MeV. Early experiments to study this kind of neutrinos, as IIL [26], Bugey [27], Bugey-3 [28], Palo Verde [29], and Chooz [30], performed important measurements of reactor neutrino spectrum and oscillation parameters. Recent experiments were designed to achieve even more precise measurements: Daya Bay [31], Double Chooz [32], RENO [33], among others. One of the next generation ex- periments, JUNO [34], will further improve the detection capabilities to use reactor neutrinos to determine neutrino mass ordering and more precise measurements of oscillation parameters [35]. The detection of reactor neutrinos is of great impact for oscillation parameters estimations. More details about Reactor Neutrino and current reactors in Japan will be discussed later when reviewing backgrounds for the low energy detection in Super- Kamiokande with Gadolinium. Neutrinos 47 1.3.4 Accelerator Neutrinos Another main source of artificial neutrinos is the accelerator neutrinos. Neutrino beams are created to travel a fixed distance. For that, protons are accelerated and hit a target, producing new particles which will decay into neutrinos. The experiments have the tuning power to create both neutrinos and anti-neutrinos, using magnets to select positive or negative pions created by the interaction of protons with the target. Depending on the distance between the source of the beam and the detector, experiments are defined as Short-baseline or Long-baseline. Short-baseline experiments include MicroBooNE [36], MINERvA [37], etc. They have detectors close to the neutrino beam source and study anomalous neutrino ap- pearance and disappearance in regions where the oscillation probability is expected to be small. They often serve as near detectors for long-baseline experiments. Long-baseline experiments include T2K [6] and NOvA [38], which measure the neutrinos after travelling a long distance, understanding neutrino mixing in regions where large oscillation probability. The T2K (Tokai-to-Kamioka) experiment is a long-baseline neutrino experiment located in Japan, illustrated by Figure (1.5). Protons are accelerated at the accel- erator complex J-PARC (Japan Proton Accelerator Research Complex), hitting a graphite target to produce the neutrino or anti-neutrino beam. T2K has near Neutrinos 48 Figure 1.5: Scheme of T2K experiment showing the neutrino beam produced at J-PARC and being detected by near detectors a few meters away from pro- duction and later by Super-Kamiokande, 295 km away [6]. detectors 280 meters from the beam production, in order to study neutrinos before the oscillation. The near facility contains an on-axis detector, the Interactive Neu- trino GRID (INGRID) and an o↵-axis detector, ND280, to measure the neutrino beam director and flux, characterizing the beam before any oscillation occurred. The beam then travels for 295 km to reach the far detector, Super-Kamiokande. The T2K experiment, using a combined analysis of appearance and disappear- ance of both neutrino and anti-neutrino channels, has established a constraint on the value of CP with 99.73% confidence level [39]. 1.3.5 Astrophysical Neutrinos Since neutrinos are charge-less particles, they can travel without any major signif- icant deflections in the Cosmos. It means that they can be used as an important Neutrinos 49 source for studying our Universe. The observation of astrophysical neutrinos by the Kamiokande experiment in 1987 introduced the Neutrino Astronomy [40]. Astrophysical Neutrinos can come from Supernovae, Active Galactic Nuclei, Neutron Stars, Black Holes, etc. In 1987, the first detection of extra-solar neutrino was made with the Super- nova SN1987A in the Large Magellanic Cloud. In 2013, the experiment IceCube detected very high energy neutrinos with energies between 10 TeV and 2 PeV [41; 42]. The Neutrino Astronomy also contributes to the cross-checks among di↵erent sources of particles coming from the Cosmos, such as Cosmic Rays, Gravitational Waves, and Photons, bringing us to a multi-messenger era in astrophysics. Neutrinos coming from Supernova are of main importance for this thesis. More specifically, the detection of neutrinos emitted in the few latest hours of the stars lives before the explosion, will allow to create a Supernova alert, allowing di↵erent experiments to be prepared for its detection. The discussion of astrophysical neutrinos from Supernova and pre-Supernova will be continued in the next chapter. Neutrinos 50 Figure 1.6: Cross sections for low energy neutrino interactions. Reproduced from [43]. 1.4 Neutrino Interactions When considering the neutrino detection, one important characteristic to be taken into account is the interaction cross-section. Depending on which energy range neutrinos are being studied, di↵erent interactions will be dominant, so particular cross-sections need to be studied for each case. For ultrahigh energies (above 100 GeV), the main interaction is the deep inelastic scattering and the cross sections are well known. For intermediate energies (around 1 GeV), dominant processes + are: quasi-elastic scattering ⌫ + n l + p or⌫ ¯ + p l + n,wherel is the ! ! produced charged lepton, and single pion production, which is given by di↵erent Neutrinos 51 processes: + ⌫ + p l + p + ⇡ l ! 0 ⌫ + n l + p + ⇡ l ! + ⌫l + n l + n + ⇡ ! (1.10) + ⌫¯ + p l + p + ⇡ l ! ⌫¯ + p l+ + n + ⇡0 l ! + ⌫¯ + n l + n + ⇡ l ! For the lower energy region, below 100 MeV, the dominant interaction processes are elastic scattering, charged current ⌫ +(N,Z) e +(N 1,Z +1),¯⌫ + e ! e + (N,Z) e +(N +1,Z 1) and neutral current ⌫ + A ⌫ + A⇤ interactions ! ! with nuclei, and the Inverse Beta Decay (IBD). The latter is the most relevant for the pre-Supernova neutrino detection. The IBD is characterized by the reaction: ⌫¯ + p e+ + n (1.11) e ! Neutrinos 52 Figure 1.7: Feynman diagram representing the Inverse Beta Decay interaction. Equation (1.11)meansthatananti-neutrinointeractswithaproton,resultingin apositronandaneutron.TheenergythresholdoftheinteractionEthr is: 2 2 (me + mn) mp Ethr = =1.806MeV, (1.12) 2mp where me, mn and mp are the electron, neutron and proton masses, respectively. Agreatpartofthepre-Supernovafluxisofelectronanti-neutrino,soIBD becomes the main detection channel for its detection. The positron energy is related to the incoming anti-neutrino by: + E⌫¯e = Ee +1.293MeV, (1.13) neglecting the recoil of the neutron. The direction of the incoming anti-neutrino Neutrinos 53 can’t be determined from the outgoing direction of the positron at low energies, as the positrons carry most of the kinetic energy of the anti-neutrino, but the neutron carries most of its momentum. The overall IBD cross-section for energies less than 10 MeV is given by [44]: = 0(1 + rad)(1 + rad)(1 + rec+WM)(1 + thr), (1.14) where rad and rad are the ”Outer” and ”Inner” radiative corrections. The outer correction is dependent on the positron spectrum and the inner correction is de- pendent on the hadronic structure. rec+WM represents the nucleon recoil and weak magnetism and thr the threshold correction. All these corrections have been calculated using di↵erent methods over the years, considering experimental data [44; 45; 46] The inverse beta decay is of extreme importance for the topic of this thesis. Pre-Supernova electron anti-neutrinos interacting via IBD in Super-Kamiokande will produce observable products by the detector. Descriptions of techniques for identifying these products from equation (1.11)willbeshownlater. Chapter 2 Supernova and Pre-Supernova 2.1 Supernova Neutrinos 2.1.1 Supernova A core-collapse Supernova occurs at the end of the life of a massive star, which is a star with at least eight solar masses, when the core of the star goes through a gravitational collapse. This process releases an energy of 3 1053 ergs. About ⇥ 99% of this explosion energy is released by neutrinos of all flavors with energies of tens of MeVs. Based on [47] the total rate of Galactic Supernovae to account the +7.4 observations in the last millennium is of approximately 4.6 2.7 per century. 54 Supernova and Pre-Supernova 55 Figure 2.1: Supernova Classification. Reproduced from [48]. After the explosion, this energy spreads through the universe, which is funda- mental for the evolution of galaxies, planets, and other stars, making the Super- nova phenomena a object of huge interest to cosmology. Supernovas are classified in Types I and II with subcategories, as described in Figure 2.1.Thecategorizationismadebytheabundanceofspecificspectrallines and light curve. Type II Supernovae have H lines in their spectra while Type I lost its H envelope [49]. Based on [47], which considered modeled observability and +1.4 historical records to estimate the Supernovae rates, Types Ia has a rate of 1.4 0.8 per century, while the other Types, which are core collapse Supernovae, have a +7.3 rate of 3.2 2.6 per century. There are many other estimations for Supernovae rates. See e.g. [50], in which the rates were estimated by combining independent results from the literature. Supernova and Pre-Supernova 56 Current models describe Type Ia as a Supernova due to thermonuclear disrup- tion of white dwarfs. For stars with mass greater than 8M ,whereM is the Solar Mass, the cause of the Supernova is the core collapse. They are known as Types Ib, Ic, and II in the classification, and they are more interested in terms of neutrino physics because it releases all kinds of neutrinos, while Type Ia only produce electron neutrinos by electron captures and much lower neutrino fluxes. After the explosion, Supernovae can produce a compact remnant, a Neutron Star or a Black-Hole, depending on their original mass, composition and history of mass loss. Usually, if the star had a mass between 8-20M ,averydenseneutron core is left, which is called Neutron Star. If the mass was greater than 30M ,the star collapse may go straight into a Black Hole, as failed Supernovae [51]. For a Neutron Star, approximated by a homogeneous sphere with Newtonian gravity, the binding energy, which is carried away by neutrinos, can be estimated as [52]: 2 1 3 GM 2 M R E 3.6 1053 [erg], (2.1) ⇡ 5 R ⇡ ⇥ 1.5M 10km ✓ ◆ ✓ ◆ where M and R are the mass and radius of the star. This equation is consistent with the observed for SN1987A, described in the following sections. Next observation of a Supernova will provide much more knowledge about as- trophysics, as well as Neutrino Physics and how great is the neutrino’s role for Supernova and Pre-Supernova 57 the explosion. Figure 2.2 shows core collapse Supernova candidates in a 1 kpc distance. 2.1.2 Supernova Burst and Neutrino Emission After all the burning stages of a star, i.e. nuclear fusion of hydrogen, carbon, oxygen, neon and silicon, the star is left with an iron core of about 1 solar mass, surrounded by shells of these elements with decreasing atomic mass. The central density is about 1010g/cm3 and temperature of about 1 MeV. This structure is sus- tained by the pressure of degenerated electrons. Processes of photo-disintegration of iron 2.2 described in: + 56Fe 13↵ +4n 124.4MeV, (2.2) ! will start to take place due to the high temperature and pressure, reducing the kinetic energy of the electrons. Other processes will also reduce the pressure of electrons, for instance electron capture of nuclei 2.3: e + N(Z, A) N(Z 1,A)+⌫ (2.3) ! e The produced electron neutrinos leave the center carrying most of the kinetic energy. These capture processes will last only a few microseconds until the pressure Supernova and Pre-Supernova 58 ]. 53 Masses and distances of the stars are shown in parenthesis. From [ Nearby core collapse Supernova candidates within 1 kpc. Colored labels show the star’s spectral Type. Figure 2.2: Supernova and Pre-Supernova 59 of the electrons can no longer sustain the structure and the collapse starts. When the core reaches a super density of about 1011-1012 g/cm3,itcantrap neutrinos in the stellar matter. The inner part of the core at this stage collapses with subsonic velocity proportional to the radius, while the outer part has su- personic free-fall velocity [48]. After one second of this instability, the inner core stops the collapse and finds equilibrium. A supersonic shock wave forms at its surface, propagating to the outer core, which is still collapsing. The shock wave will disassociate nuclei in the outer core, which will cause electron capture on free protons, creating a large neutrino flux. However the flux will be trapped behind the shock wave for its density. The shock breakout, when the neutrinos are re- leased, happens when the shock wave reaches a zone with density of 1011g/cm3. The correspondent neutrino emission is known as neutronization burst. It occurs in the first 25 milliseconds after the core bounce when the star emits with very ⇠ high luminosity mainly ⌫e neutrinos. At this stage the inner core is still with high density and opaque to neutrinos. The density needed for neutrino emission will be di↵erent for each flavor due to di↵erence in matter interactions. Neutrino-spheres will be created for di↵erent neu- trino flavors, as they interact di↵erently with matter. The production of neutrinos in the inner core occurs because of di↵erent processes, such as electron-positron Supernova and Pre-Supernova 60 annihilation eq. (2.4): + e + e ⌫ +¯⌫ (2.4) ! x x electron–nucleon eq. (2.6)andnucleon-nucleonbremsstrahlungeq.(2.5): N + N N + N + ⌫ +¯⌫ , (2.5) ! x x e± + N e± + N + ⌫ +¯⌫ , (2.6) ! x x plasmon decay eq. (2.7): ⌫ +¯⌫ , (2.7) ! x x and photoannihilation eq. (2.8): + e± e± + N + ⌫ +¯⌫ , (2.8) ! x x where x = e, µ, ⌧. These interactions in the dense core produces pairs of neutrino and anti-neutrinos from all three flavours. It is produced around 3 1053 ergs of gravitational energy, ⇥ in which 99% goes into neutrinos, 0.01% into photons, and 1% into kinetic energy ejecting material. Due to the di↵erent interactions of the associated leptons, the evolution of the Supernova and Pre-Supernova 61 di↵erent flavours of neutrinos will be di↵erent, giving the following energy hierar- chy: E < E < E , (2.9) h ⌫e i h ⌫¯e i h ⌫x i where x = µ, ⌧, and the energies range between 10 MeV and 30 MeV. The increasing density of the core due to the presence of heavy elements will make it reach a limit of 1.4M ,knownastheChandrasekharmasslimit,starting the core collapse Supernova within 10 seconds. Thanks to optical measurements of living stars, we can have a very good esti- mation of the Supernova rate, e.g. [54; 55; 56; 57; 58]. However, many questions about neutrino emission rates by Supernovae, energy spectrum and luminosity behavior of each of the neutrino flavors are still to be answered (see review work [59] and references within). It is clear the important role neutrinos have during the star explosion, from burning stages before the Supernova to all of the di↵erent stages of accretion, cooling, collapse and formation of compact remnants. The neutrino luminosity from core-collapse Supernovae has, for a few milliseconds, a peak of 1020 L ⇠ 2.5 1053 erg/sec, where L is the solar luminosity, and average of 1017 L ⇠ ⇥ ⇠ ⇠ 1051 erg/sec during the star cooling process [60]. The understanding of Supernova explosions is directly correlated to the understanding of neutrino physics. Supernova and Pre-Supernova 62 2.1.3 SN1987A The first observation of neutrinos produced outside of our solar system was thanks to a Supernova explosion known as SN1987A in the Large Magellanic Cloud. It was detected on February 23rd of 1987 due to an extremely strong neutrino burst. It occurred 51.4 kpc from Earth, and the detection of the neutrino burst, with energy of 1053 ergs, occurred three hours earlier than the optical telescopes and lasted about one minute. SN1987A became our best laboratory to study Supernova neutrinos, but it is the only Supernova in the history detected by a neutrino burst. It was observed by di↵erent experiments at the time: Kamiokande II (11 events) [61], IMB (8 events) [62], Baksan (5 events) [63], and Mont Blanc Liquid Scintillator (5 events) [64]. A comparison of data from Kamiokande and IMB for electron anti neutrino is shown in Figure 2.3. The progenitor of SN1987A was a blue supergiant, contradicting old models which predicted the Type II Supernovae to be produced only by red supergiants. The remnant is believed to be a Neutron Star [49]. Neutrino observations from SN1987A helped to study properties of neutrinos, axions, majorons, supersymmetric particles and unparticles. It also proved pre- dictions regarding neutrino fluence from core collapse Supernovae, spectra, time Supernova and Pre-Supernova 63 Figure 2.3: Comparison of the time integrated e↵ective anti-neutrino spec- trum from SN1987A between data from Kamiokande and IMB. Reproduced from [65]. profile, etc. [49], [66]. These features are very important for many predictions regarding the Di↵use Supernova Neutrino Background, which is discussed in next. 2.2 Di↵use Supernova Neutrino Background The Di↵use Supernova Neutrino Background (DSNB) is the overall neutrino flux of all core collapse Supernovae that happened in the history of our universe. Supernova and Pre-Supernova 64 These neutrinos are also known as Supernova Relic Neutrinos (SRN). In Super- Kamiokande, thy have an expected event rate of 1.3-6.7 events/year/22.5kton, with energy between 10-30 MeV [67]. Equation 2.10 gives the di↵erential Supernova Relic Neutrinos flux in Earth, taking into account also redshift e↵ects. d 1 dN [E (1 + z)] dt = c R (z)(1 + z) ⌫ ⌫ dz, (2.10) dE SN dE dz ⌫ Z0 ⌫ where RSN(z)istheSupernovarateatredshiftz,E⌫ is the neutrino energy re- ceived, dN⌫/dE⌫ is the number spectrum emitted by the explosion where the neu- trino was emitted at a higher energy E⌫(1 + z). The relation between cosmic time t and redshift z is given by dz = H (1 + z) ⌦ +⌦ (1 + z)3,whichdepends dt 0 ⇤ m p on the Hubble constant H0 and cosmological parameters ⌦. Regarding the Supernova rate RSN(z), there are several models to predict its value. After the information acquired from SN1987A discussed in the previous section, it was possible get better estimations for it. Figure 2.4 shows the DSNB spectrum for di↵erent models. Recently, these models have been based on the history of cosmic star formation, where the Supernova rate is proportional to the Star Formation Rate (SFR). The SFR is a function of the star mass, which thanks to current measurements of Supernova and Pre-Supernova 65 Figure 2.4: Predicted DSNB spectrum for di↵erent models and showing the three main backgrounds for this measurement and the energy region where they apply. Reproduced from [65]. emissions of living massive stars, gives a very reliable core collapse Supernova rate function. These neutrinos have not been detected yet. However, detection capabilities have been improved significantly in the last few years. It is predicted that the new phase of Super-Kamiokande, with gadolinium, will be able to detect a few of these events per year [65]. Supernova and Pre-Supernova 66 2.3 Pre-Supernova Neutrinos The amount of neutrino emission from a star increases significantly in early stages of a Supernova. The Zero Age Main Sequence (ZAMS) of a star is when it stars to burn hydrogen in its core through fusion reaction. After stars with ZAMS mass > 8M burn most of the Hydrogen, producing Helium, the density of its core starts to increase, which makes the fusion of the available Helium into carbon. After the ignition of the carbon burning, the star is classified as neutrino-cooled star. The terminology of neutrino-cooled star was introduced by David Arnett [68], and it is important since it distinguishes this very special stage of the star evolu- tion, in which the main cooling mechanism of the star is the neutrino emission. For instance, while the lifetime of stars can vary from million to billion years, the neutrino-cooled star lasts only hundreds of years. A neutrino-cooled star can reach neutrino luminosity of about 1012L ,whileitsphotonluminosityisonly105L [60]. When the neutrino-cooled stage begins, massive stars will usually proceed with nuclear fusion of helium, carbon, oxygen, neon and silicon, getting a onion shape with an iron core, characteristic structure of the star just before the core collapse, represented in Figure 2.5.Theburningoftheseelementsispossiblebecauseof Supernova and Pre-Supernova 67 Figure 2.5: Characteristic onion-shape of a massive star just before the core collapse, where the shells show the abundance of each element. Reproduced from [16]. very high temperatures of the star before the start of density limitation due to electron degeneracy [16]. In temperatures above 5 108 K, the electron-positron annihilation process ⇠ ⇥ generating thermal neutrinos (eq. 2.11)becomesthestar’sdominantformof cooling. Other processes such as plasmon decay, neutrino brehmsstrahlung and pair production also generates thermal neutrinos at this stage. The cooling rate increases with temperature, leading to fusion of even heavier nuclei [16]. Supernova and Pre-Supernova 68 + e + e ⌫ +¯⌫ , (2.11) ! x x where x = e, µ, ⌧. The last stage of these stars before the core-collapse is the Si-burning, which is commonly known as pre-Supernova star. In other words, the pre-Supernova star is the massive star at the onset of the collapse. At this last stage, the star forms an iron core and burns Silicon at a temperature of approximately 3.4 109K. ⇥ Neutrinos emitted at the Si-burning stage have an average energy of 1.85 MeV. Usually the definitions of neutrino-cooled star and pre-Supernova star are ap- proximated into one, since the evolution after the carbon ignition is very fast compared to the life of a star. Table 2.1 shows approximate duration of burning stages for a M star and the average energy of the emitted neutrinos for each phase. Table 2.1: Approximate duration of burning stages for a 20 M star and the fraction and average energy of electron neutrinos emitted by pair-annihilation [69]. Burning Stage Duration ⌫e (¯⌫e) fraction Average ⌫ energy C300years42.5%0.71MeV Ne 140days 39.8% 0.99MeV O180days38.9%1.13MeV Si 2 days 36.3% 1.85 MeV Supernova and Pre-Supernova 69 Figure 2.6: Normalized spectrum (normalized energy distribution function) of the pair-annihilation neutrinos emitted during C (solid), Ne (dashed), O (dotted) and Si (dot-dashed) burning stage and the solar pp neutrinos (thin line), from [60]. The spectrum was calculated based on the Monte Carlo method of [70]. Solar neutrinos and pre-Supernova neutrinos have similar energies. However, pre-Supernova stars emits anti-neutrinos of every flavor, which is not true for the Sun. The neutrino luminosity during Si burning is 1013 times greater than the solar neutrino luminosity. Due to their di↵erent spectrum and the presence of anti neutrinos, the pre-Supernova stars can be detected from kiloparsec distances [71]. The spectrum of the pair-annihilation neutrinos and comparison with pp neutrinos from the Sun is shown in Figure 2.6,butitdoesnotincludetheproducedanti- neutrinos, which is the main di↵erence between the two sources. Supernova and Pre-Supernova 70 About 1/3 of the produced neutrino flux in pre-Supernova stars is of electron anti-neutrino. This is of extremely importance for detection purposes, since its detection can be done though the inverse beta-decay⌫ ¯ + p n + e+,already e ! introduced in the previous chapter. This interaction for the pre-Supernova⌫ ¯e 43 2 has a large averaged cross section ¯ 0.7 10 cm ,whichisaboutthree Si ' ⇥ times more than the neutrino-electron elastic scattering [60]. Although the pre- Supernova neutrinos have lower energies than Supernova neutrinos, their emission is over a very long timescale compared to the Supernova [16]. Stars with ZAMS mass < 8M do not have the Si-burning stage and have a core collapse made of O/Ne/Mg burning instead of an iron core collapse [48]. Very massive stars > 30M may collapse directly to a black hole with no Supernova explosion [51]. 2.3.1 Pre-Supernova Models As explained in the previous section, the main cooling mechanism of massive stars at late stages of their evolution (C, Ne, O and Si burning) before core collapse + is by pair annihilation e e ⌫⌫¯ ,producingallflavoursofneutrinoandanti- ! neutrino pairs. Beta processes will also contribute significantly right before the core collapse, with an average more energetic⌫ ¯e [72]. Only electron anti-neutrinos from the Silicon burning phase can be detected in Super-Kamiokande due to a Supernova and Pre-Supernova 71 larger cross section for detection of⌫ ¯e through Inverse Beta Decay at their energy range. The rate in which pre-Supernova would happen is the same as the rate for Supernova for the progenitor star with ZAMS mass > 8M , if we consider that all these massive stars will end as core-collapse Supernovae. There are many methods that estimates this rate. In this work, the rate considered is of 3.2 per century [47]. In order to perform these estimations of the expected signal from pre-Supernovae neutrinos in Super-Kamiokande, two models of the thermodynamics of stellar mod- els were analyzed: Odrzywolek and Heger [71], [69]andPattonetal.,[72]. Both models provide online data sets for calculation of anti-neutrino emission during the pre-Supernova phase of a star. Odrzywolek provides for stars with 15 M and 25 M [73] and Patton for 15 M and 30 M [72], which are used to estimate the pre-Supernova sensitivity in Super-Kamiokande. The model from Odrzywolek and Heger assumes that the entire neutrino flux comes from pair annihilation. The normalized spectrum for the emitted anti- neutrinos from Odrzywolek and Heger model is shown in Figure 2.7.Forthe nuclear isotopic composition of the star, the model assumes a nuclear statistical equilibrium (NSE), which is a treatment only dependent on the temperature, den- sity and electron fraction, making it a simple flux estimate by only post-processing Supernova and Pre-Supernova 72 Figure 2.7: For a 20 solar masses star, the emitted anti-neutrino spectrum for di↵erent stages. Dashed line: carbon burning, dotted line: neon burning, short-dashed line: oxygen burning, and solid line: silicon burning. Reproduced from [71]. an already existing stellar model. The model from Patton et al., [72]includesamorecompleteevaluationofthe neutrino flux from the pre-Supernova star, including contributions not only from pair annihilation, but also from plasmon decay, photoneutrino process, -decay, and electron capture. Patton et al. also uses a modern approach to the star evolution code, which couples the isotopic evolution to the stellar evolution, by using the software MESA (Modules for Experiments in Stellar Astrophysics) [74]. Supernova and Pre-Supernova 73 Figure 2.8: IBD rate in Super-Kamiokande for the 12 hours before the core- collapse for di↵erent star evolution models. It is assumed a distance of 200 parsecs. Reproduced from [16]. The Inverse Beta Decay rate evolution in time prior to the core collapse in Super- Kamiokande is illustrated in Figure 2.8 for the considered models, in a distance of 200 parsecs. Figure 2.9 shows the evolution in time of neutrino luminosity and mean energy of emitted⌫ ¯e. Following [16], in order to calculate the expected signal from the considered models, an adiabatic transition is assumed and the ratio of electron anti-neutrinos is changed at high MSW resonance and depends on the mass ordering of neutrinos. The assumed transition probability is P (¯⌫ ⌫¯ )=0.675 for normal mass order- e ! e ing and P (¯⌫ ⌫¯ )=0.024 for inverted mass ordering. It accounts the change in e ! e Supernova and Pre-Supernova 74 Figure 2.9: Comparison of the considered models for the⌫ ¯e luminosity and mean energy prior to the core-collapse. Reproduced from [16]. Supernova and Pre-Supernova 75 ratio of electron flavour neutrinos due to the dense stellar medium and the e↵ects of oscillations in vacuum. Recently, Super-Kamiokande started a new period of observations with the ad- dition of gadolinium sulfate to the water in the detector, which has good prospects for detecting pre-Supernova neutrinos. As it will be explained in the next chap- ter, the enhanced neutron tagging by a more e cient capture of gamma-rays will reduce backgrounds and allow the detection of⌫ ¯e emitted during the Si-burning phase. Although their mean energy is below the Inverse Beta Decay threshold, as we can see in Figure 2.7,thetailcouldpossiblybedetected. Figure 2.2 shows nearby stars within 1 kpc from Earth (which are listed also in Appendix A.1). The best candidates for observation currently are ↵-Ori (Betel- geuse) and ↵-Sco (Antares) for their proximity. The expected signal from di↵erent pre-Supernova models and characteristics of the star in Super-Kamiokande will be shown in Chapter 5. Chapter 3 Super-Kamiokande with Gadolinium Super-Kamiokande is a 50 kton water Cherenkov detector located at the Kamioka Observatory of the Institute for Cosmic Ray Research, University of Tokyo. It was designed to study neutrino oscillations and carry out searches for the decay of the nucleon operating since 1996, detecting neutrinos from natural and artificial sources. It is a big collaboration between 40 institutions around the world. In 2020, Super-Kamiokande moved to its next phase, in which Gadolinium Sulfate was dissolved to the water in the detector. 76 Super-Kamiokande with Gadolinium 77 Figure 3.1: Schematic overview of the Super-Kamiokande Detector 3.1 Detector Overview With a diameter of 39.3 m and height of 41.4 m, Super-Kamiokande is cylindrical stainless-steel tank filled with 50 ktons of water. It is located in the Kamioka mine in Japan, overburden with 1000 m of rock (equivalent to 2700 meters water) to reduce cosmic ray muon backgrounds. The reduced muon flux is by a factor of 105. Figure 3.1 shows a schematic view of the Super-Kamiokande detector. The detector is divided into an Inner Detector (ID) and an Outer Detector (OD). This division is made by a cylindrical structure of 55 cm thickness, supporting Super-Kamiokande with Gadolinium 78 PhotoMultipliers Tubes (PMTs) to face both the ID and the OD. The inner part is left with dimensions 33.8m diameter and 36.2m height and, around the inner surface, 26 sets of Helmholtz coils are placed to compensate the magnetic field from the Earth [5]. The OD has a thickness of about 2m and and it is composed of 1,185 8-inch PMTs, facing the outside of the detector. It has a function of veto for charged cosmic rays that generate background and shield for neutrons and gamma rays emitted by radiation of the surrounding rocks. The ID had initially 11,146 20-inch PMTs facing the inner part of the detector. The number of PMTs changed in other phases of the detector, as it will be explained in section 3.2.ThePMTs in the ID are responsible for the overall detection. The water mass in the ID is of 32 kilotons. We also define the Fiducial Volume (FV) 2m inside the inner detector, in order to reduce backgrounds, corresponding to 22.5 ktons of water, which contains approximately 7.5 1033 protons and 6.0 1033 neutrons. These ⇥ ⇥ backgrounds come from radioactive impurities in the PMT glasses, as it will be explained in Section 3.9. Figure 3.2 shows a illustration of the support structure for the inner detector. Each one of the modules showed in Figure 3.2 contains twelve 20-inch PMTs for the ID and two 8-inch PMTs for the OD. The photo-coverage, which is coverage of the surface by the PMTs is about 42%. The number of active PMTs inside ⇠ Super-Kamiokande with Gadolinium 79 Figure 3.2: Arrangement of PMTs in modules inside the Super-Kamiokande Detector. From [5]. the detector varies with respect to the phase in which the experiment is running, which will be discussed in Section 3.2. In order to enhance the ability to observe neutrinos and prepare the detector for gadolinium sulfate loading, in 2018, it was conducted in Super-Kamiokande a refurbishment work to the detector, which included water containment reinforce- ment, replacement of defective PMTs, and improvement of the piping system in the tank. Figure 3.3 shows a picture taken inside the tank during the refurbish- ment work in 2018. Super-Kamiokande with Gadolinium 80 Picture taken during the refurbishment work in 2018 of the Super-Kamiokande inner detector. Copyright: Kamioka Observatory, ICRR (Institute for Cosmic Ray Research), The University of Tokyo. Figure 3.3: Super-Kamiokande with Gadolinium 81 A Monte Carlo simulation built on the GEANT3 framework called SKDETSIM models the detector. 3.2 Super-Kamiokande Phases The Super-Kamiokande experiment began to acquire data in April, 1996. Since then, its operation has been divided into di↵erent phases: SK-I : First phase of the experiment when it started taking data in April 1996 until July 2001, where it had to be shut down for maintenance. It was operating with a photon coverage of 40% by 11,146 ID PMTs and 1,885 OD PMTs. SK-II : During the refilling of the tank in November 2001, an implosion of one PMT destroyed more than half of the PMTs inside the tank. Then it started acquiring data in October 2002 with only 5,182 ID PMTs, with a photon coverage of 19%. Also to each PMT was attached a fiber reinforced plastic, or FRP case to prevent accidents like before. It ran until October 2005, when it stopped to be fully rebuild. SK-III : After being rebuild, the experiment restarted acquiring data in July 2006 until September 2008 with now 11,129 PMTs in the ID and again a photon coverage of 40%. Super-Kamiokande with Gadolinium 82 SK-IV :Thefourthphaseoftheexperiment,whichmaintainedthesamenumber of PMTs and photon coverage as before, but this time with the upgraded Front-End Electronics. It operated between September 2008 until May 2018, when a refurbishment started in the detector towards the new project of the experiment known as SK-Gd. SK-V : Started on January 29th, 2019. Phase in which the experiment started the preparation for gadolinium sulfate loading. Some new PMTs that will be used in the Hyper-Kamiokande experiment were installed for testing. It ran until dissolving Gadolinium Sulfate in July, 2020. SK-VI : The current phase of the experiment, started in July 14th,2020.0.02% gadolinium sulfate was dissolved to the water in the detector to improve the neutron tagging e ciency. 3.3 Cherenkov Radiation Since neutrinos do not have electric charge, they are very hard to detect. However, when they occasionally interacts with the water in Super-Kamiokande, charged particles are generated and they can be detected through the Cherenkov radia- tion. When a charged particle moves faster than the phase velocity of light a certain medium, vparticle >c/n,wherec is the speed of light in vacuum and n is Super-Kamiokande with Gadolinium 83 the refraction index of the medium, it emits a Cherenkov light [75]. The thresh- old energy, which minimum energy to generate the Cherenkov light is given by equation 3.1. m Ethr = , (3.1) 1 (1/n( ))2 p where m is the mass of the charged particle, n is the refraction index of the medium, and is the wavelength of the Cherenkov light. The threshold kinetic energy for an electron or positron to emit Cherenkov light in pure water is 0.257 MeV. The Cherenkov photons are emitted in a cone with an open angle: 1 cos✓ = (3.2) C n( ) The number of photons emitted per unit length per unit wavelength is given by equation 3.3. d2N 2⇡z2↵ 1 2⇡z2↵ = 1 = sin2✓ , (3.3) dxd 2 2n2( ) 2 C ✓ ◆ where z is the particle charge in unit and ↵ is the fine structure constant. Super-Kamiokande with Gadolinium 84 Figure 3.4: Spectrum of the Cherenkov radiation in water. The Cherenkov light is emitted mostly in the UV range, but for Super-Kamiokande, the typical range of photon wavelength detection by the PMTs is between 300nm and 600nm. Figure 3.4 shows the Cherenkov radiation spectrum in water. Inte- grating over this range of wavelength: dN =764z2sin2✓ photons/cm (3.4) dx C Super-Kamiokande with Gadolinium 85 Figure 3.5: Scheme of the Super-Kamiokande water system. Reproduced from [5]. Also, the refraction index for pure water is approximately n 1.34 and if we con- ⇡ sider that we are dealing with ultra relativistic particles, 1, we have from equa- ⇡ tion 3.2 that ✓ 42 , so it is reasonable to assume that inside Super-Kamiokande, C ⇡ 342 Cherenkov photons are produced per centimeter. The Cherenkov photons ⇠ are detected by the PMTs, which are discussed next. Super-Kamiokande with Gadolinium 86 3.4 Water Purification System The water transparency a↵ects significantly the quality of data acquired by the detector. Di↵erent impurities in the water need to be eliminated to maintain high transparency. UV lamps are used to kill di↵erent bacteria, reverse osmosis and filters for the removal of small particulates. Heavy ions are removed with an ion exchange resin called Polisher Cartridge. Radon contamination is reduced with vacuum and membrane degasifiers. Figure 3.5 shows a scheme of the SK water system that was used for pure water. The water needs to be recirculated to maintain the transparency. In the ID, the water injection is at the bottom and the extraction from the top. Also, the water temperature needs to be low for better e ciency of PMTs and suppress growth of bacteria. Small increases in temperature at the top half of the ID causes a di↵erence in the water transparency between top and bottom of the detector. To remove the e↵ects of the di↵erence in transparency, convections need to be done during calibration works. Injected water is cooled below the tank temperature for the convection. Super-Kamiokande with Gadolinium 87 3.4.1 Preparation for SK-Gd The current phase of Super-Kamiokande, SK-Gd (SK-VI), has gadolinium sulfate loaded to the water in the detector. The compound was chosen for not absorb- ing much light in the range of Cherenkov photons and for being highly soluble, inexpensive, and transparent in solution [78; 79]. Changes to the water system had to be performed so the gadolinium sulfate would not be removed with other impurities from the water in Super-Kamiokande. To study the impacts of gadolinium loading in Super-Kamiokande, a small de- tector called Evaluating Gadolinium’s Action on Detector System (EGADS) was realized to simulate the injection of gadolinium sulfate to the water in Super- Kamiokande and the water system. Figure 3.6 shows the apparatus of the water system used in EGADS. The water removed to the top of the tank goes through compatible with gadolin- ium sulfate UV lamps and filters to eliminate particulates and bacteria. To control temperature, and consequently the solubility of gadolinium sulfate, chillers and heating exchangers are used. Fast-recirculation can also be applicable to proce- dures compatible with the gadolinium sulfate. Anion exchange resins used are not the same as the ones used for the pure water in Super-Kamiokande, but compatible with the gadolinium sulfate. Total organic Super-Kamiokande with Gadolinium 88 ]. 79 duced from [ Scheme of the EGADS water system to simulate the gadolinium sulfate injection in Super-Kamiokande. Figure 3.6: AE = anion exchange resin, DI = dionisation, TOC = total organic coumpound lamp, RO = reserve osmosis. Repro- Super-Kamiokande with Gadolinium 89 ]. 79 Water transparency measurements of EGADS compared to SK-III/IV phases. Reproduced from [ Figure 3.7: Super-Kamiokande with Gadolinium 90 the first phase of SK-Gd. Most recent water transparency measurements in EGADS, that from March 31st, 2018 started to simulate Figure 3.8: Super-Kamiokande with Gadolinium 91 compound lamps are used to create free radicals with far-UV. A nanofilter can be used to separate the ions of gadolinium sulfate, so to simulate the removal of the compound. EGADS was able to run for years maintaining the water quality. Figure 3.7 shows the water quality for the loading of 0.2% gadolinium sulfate to EGADS and Figure 3.8 shows it for the 0.02% loading, started in March 31st, 2018. More discussions regarding the EGADS detector will be held later in the chapter. 3.5 PhotoMultiplier Tubes The PhotoMultiplier Tubes (PMTs) were developed by Hamamatsu Photonics in collaboration with the Kamiokande collaboration for use in the Kamiokande experiment. They were further improved for Super-Kamiokande for a better timing resolution and discrepancy from dark current for a single photo-electron. Figure 3.9 shows a scheme of the 20-inch model. The material used for the photocathode is bi-alkali (Sb-K-Cs), which is sensitive between 300-600nm and has quantum e ciency peak around 390nm (Figure 3.10). The PMTs used in the ID have a peak quantum e ciency of 21% at wavelengths of 360-400 nm. The photoelectron collection e ciency is above 70%, with transit time spread of 2.2 ns. Super-Kamiokande with Gadolinium 92 Figure 3.9: General scheme of the 20-inch Photo-Multiplier. From [5]. Figure 3.10: Quantum E ciency for 20-inch PMT. From [5]. Super-Kamiokande with Gadolinium 93 After the accident in 2001, all the PMTs were covered with FRP (fiber reinforced cases) and acrylic covers, as shown in Figure 3.11,whichunfortunatelyincreases the dark noise. Until SK-IV, the model of all PMTs used was the Hamamatsu R3600. It has adynodestructureof11stageVenetianblindtype.FromtheSK-Vphase,it is used both the PMTs model R3600 and also 140 Hamamatsu R12860 PMTs, which were installed during the refurbishment work in 2018. These new models will be installed on the future experiment Hyper-Kamiokande. They have a dynode configuration of 10 stage box and line and, compared to the R3600 model, they have twice the detection e ciency, twice as good charge resolution and more than twice as good timing resolution [76]. Figure 3.12 shows one of the new PMT being inspected before installation in Super-Kamiokande. Each pulse created by each PMT in Super-Kamiokande will record information regarding the pulse arrival time and total charge. If the collection of information inside a time window by a number of PMTs is above some threshold, the system is triggered and the data of hit time and charge are recorded. The total charged collected is correlated with the energy of the charged particle that produced the light. The time window is set to be the maximum time the light would take to travel the maximum diagonal distance inside the ID, which is approximately 200 ns [77]. Discussions of triggering conditions will be held in Section 3.7. Super-Kamiokande with Gadolinium 94 University of Tokyo. Picture taken during the full reconstruction of Super-Kamiokande in 2005, showing the PMT with Figure 3.11: the Fiber Reinforced Case (FRC). Copyright: Kamioka Observatory, ICRR (Institute for Cosmic Ray Research), The Super-Kamiokande with Gadolinium 95 Tokyo. Picture taken during the refurbishment work in 2018 of a new PMT being inspected installing in Figure 3.12: Super-Kamiokande. Copyright: Kamioka Observatory, ICRR (Institute for Cosmic Ray Research), The University of Super-Kamiokande with Gadolinium 96 3.6 Electronics and DAQ The DAQ electronics used in Super-Kamiokande changed over the years. From SK- I to SK-III, the electronics were Analogue Timing Modules (ATM), in which time- to-analogue and charge-to-analogue converters were used [80]. Starting from SK- IV, the ATM electronics were replaced with QTC-based Electronics with Ethernet (QBEE), with better charge and time resolution [81]. High speed charge-to-time converters (QTC) and time-to-digital (TDC) are used. Each board has eight QTC chips and each one of them takes the pulse produced by each PMT. The gaining of the QTC is optimized to give improved dynamic ranges and resolutions. The TDCs tracks the QTC output signal and sends the starting time and signal width to the data sort mapping FPGA, which saves into 6 bytes the source channel, charge and start time. This data is stored by another FPGA to the first-in-first- out (FIFO) memory. Data from the FIFO memory to the front-end computers are transferred by an Ethernet interface card as TCP packets. Data will be then sorted in 20 ms blocks. Parallel computers merge the data blocks into time order and build the hit data into events. Then an organizer computer sorts the data blocks, resulting in a single data stream of events sorted in time, which are then saved to the disk [82]. Triggering is handled by software, which allows changes to trigger conditions. Super-Kamiokande with Gadolinium 97 3.7 Trigger Conditions and WIT AsignificantproportionofhitsinsidetheDAQtimewindowisduetodarknoise, which is caused by thermionic emission of electrons from the photocathode and due to radioactive decays in the PMT material [83]. Events with very low numbers of hits are dominated by these backgrounds, so applying thresholds is a way to reduce them. Di↵erent triggers are applied in Super-Kamiokande, which are related to the energy region of neutrino being analyzed. Table 3.1 shows the di↵erent triggers and their conditions. Table 3.1: Trigger schemes in Super-Kamiokande. Trigger Condition Rate Wide-band Intelligent WIT - kHz ⇠ Super Low Energy SLE > 34/31 hits kHz ⇠ Low Energy LE > 47 hits 40 kHz ⇠ High Energy HE > 50 hits 10 kHz ⇠ Super High Energy SHE > 70/58 hits After AFT (SHE).(OD) Outer Detector OD > 22 hits in OD The Wide-band Intelligent Trigger (WIT), as seen in table 3.1,doesnotapply any threshold to the minimum number of hits recorded by the PMTs. The WIT Super-Kamiokande with Gadolinium 98 system is installed and integrated in the online DAQ system in Super-Kamiokande, and it was design to process all the data, extracting and reconstructing very low energy electrons with e ciency close to 100% in real time, with the use a great amount of parallel computing [83]. The system discards events with reconstructed vertices close to the PMTs or not well reconstructed, in this way eliminating many background events due to radioisotope decays in the PMT glass and FRP cases, for their closeness to the PMTs. The WIT system is composed of seven computers that receive, triggers and re- constructs raw data blocks. These data blocks contain time and charge information of 1344 consecutive data reads (which is about 23 ms of data). In a 220 ns time window, which is the maximum time Cherenkov photons can take to transverse the detector as explained in Section 3.5, coincidence signals of 11 hits (around 2 MeV) above the dark noise (usually 12 hits) are searched inside the data blocks. Hits passing this criterion are called ”pre-triggered events” and to them is applied ver- tex reconstruction, described in the next Section 3.7.1.Thereconstructedevents are sent to an eighth computer, which sorts the events by time. The sorted events are then transferred to the o✏ine disk for physical analyses [83]. Super-Kamiokande with Gadolinium 99 3.7.1 Event Reconstruction The first filter applied to pre-triggered events is the Software Trigger Online Re- construction of Events (STORE). The filter looks for PMT hits that might be originating from a single vertex, since low energy electrons can only travel a few centimeters in water, so their Cherenkov light is approximately a point source [83]. The filter calculates the temporal and spatial separations among hits. The largest set of hits that are consistent with being from a point source is selected. The vertex fit ClusFit is applied to the hits selected by STORE, in which isolated hits are eliminated to reduce e↵ects of reflected and scattered light and dark noise. If the vertex reconstructed by ClusFit is inside the Fiducial Volume, the Branch Optimization Navigating Successive Annealing Iterations (BONSAI) fitter will be applied. BONSAI performs a maximum likelihood fit to timing residuals of the PMT hits, testing each vertex hypothesis. Timing residual is defined as: T = T T T , (3.5) residual hit flight vertex where Thit is the PMT hit time, Tflight is the time of flight from the testing vertex to the hit PMT, and Tvertex is the time of the testing vertex. The hypothesis with maximized likelihood is chosen as reconstructed vertex. Super-Kamiokande with Gadolinium 100 Figure 3.13: Likelihood values as a function of the total electron energy and opening angle between the reconstructed direction and the direction from vertex to each PMT. From [85]. BONSAI is a much more accurate fitter, but slower. For this matter, applying STORE and ClusFit, which are fast filters, is important for CPU time consider- ations [84]. BONSAI returns also a goodness value to evaluate the quality of the applied fit. For the direction reconstruction, an energy dependent likelihood is used to com- pare Cherenkov ring patterns between data and MC simulation. These patterns and their energy dependences were simulated for di↵erent energy ranges [85]. Fig- ure 3.13 shows the likelihood as a function of electron energy and open angle between the reconstructed direction and the direction from vertex to each PMT. The likelihood of hits in an event is maximised to get the most likely direction. Super-Kamiokande with Gadolinium 101 Then it is applied, around the reconstructed direction, a Kolmogorov-Smirnov (KS) circular test on the azimuthal symmetry of the Cherenkov cone, resulting in a ”dirks” score, which lower values can indicate a mis-reconstructed vertex. Many backgrounds in Super-Kamiokande come from radioactive decays in the PMT glasses. These events can be rejected for having low dirks score, since they are not well fit due to their proximity to the walls. For the energy reconstruction, the calculation is done using the number of PMT hits with a residual time within 50 ns window, called N50. However, before con- verting this number of PMT hits to energy, many corrections are applied to take into account e↵ects of scattering, reflection, water transparency and dark noise, so an e↵ective number of hits can be obtained, which is called Neff ,givenbyequation 3.6. N50 Nall Rcover ri N = (X ✏ ✏ ) e (run) G (t) , (3.6) eff i tail dark ⇥ N ⇥ S(✓ , ) ⇥ ⇥ i i=1 norm i i X ⇢ where the factors of these equations correspond to: occupancy (Xi), late hits (✏tail), dark noise (✏dark), bad PMTs (Nall and Nnorm), photo-cathode coverage (Rcover and S(✓i, i)), water transparency (ri and (run)) and PMT gain (Gi(t)). The total energy is calculated as a function of Neff .Theenergyconversionis precisely calibrated using LINAC data, to be discussed in Section 3.8,andthe Super-Kamiokande with Gadolinium 102 Figure 3.14: Energy resolution as a function of energy found using Monte Carlo events in [77]. accuracy the the absolute energy scale is better than 1% [86]. Figure 3.14 shows the energy resolution as a function of energy. Other important variables calculated by the event reconstruction are: n18, which is the number of hits in the interval 6 ns < T < 18 ns at the residual reconstructed vertex, fromwall,whichisthedistancefromthereconstructedver- tex to the wall of the detector by the shortest vector, and effwall,whichisthe same distance as fromwall,butalongavectordefinedbythereconstructeddirec- tion. All these variables are important later when describing the event selection to study the sensitivity for pre-Supernova at Super-Kamiokande. Super-Kamiokande with Gadolinium 103 3.7.2 Isotropy Parameters Other variables can be calculated after the vertex reconstruction. For instance, the angular distribution of PMT hits can characterize an event by its isotropy. The parameters l,givenbyequation3.7,representsagoodwaytoquantifythe isotropy of an event [87]. N 1 N 2 = P (cos✓ )(3.7) l N(N 1) l ij i=1 j=i+1 X X This equation considers that a vector is traced from the reconstructed vertex to each PMT hit, and the angle between pairs of vectors i and j is ✓ij.Theisotropic parameter is the average of the Legendre polynomial of cos(✓ij). Higher values of l indicates less isotropy [16]. In Thermal Neutron Capture (TNC) on gadolinium, multiple -rays with ran- dom directions relative to each other will be created, but the pattern of PMT hits in Super-Kamiokande should be isotropic. So, discriminating events by their isotropy can be contribute to reduce backgrounds. Super-Kamiokande with Gadolinium 104 3.8 Calibrations in Super-Kamiokande The calibration procedures in Super-Kamiokande have great importance for the physics analyses and event reconstruction done in Super-Kamiokande. The PMT and electronics responses, water properties and other items are calibrated to be used as parameters for the Monte Carlo simulation of the detector, SKDETSIM, so it can be used with confidence in the data analyses. For the calibration of the Super-Kamiokande for low energy neutrino measure- ments, a linear accelerator (LINAC) for electrons installed in the detector is used. The electrons are injected into the ID with a known energy. LINAC data is taken at various positions in the detector, getting detector responses to the variables relevant to low energy analyses. Electrons can be selected between 5-16 MeV/c [88]. LINAC is located near the top platform of the tank and the beam goes through an evacuated pipe and guided by magnets along the platform. The pipe is extended to di↵erent positions in the tank, as illustrated in Figure 3.15. The calibration with LINAC is used to set absolute energy scale for low energy analyses and also for determining energy, vertex, and angular resolutions. Other calibration techniques are used to calibrate absolute energy scale, for example Super-Kamiokande with Gadolinium 105 Figure 3.15: Scheme of LINAC calibration at the SK detector. The dashed line represents the fiducial volume. The black dots indicate where the calibration data is usually taken. From [88]. the DT (deuterium-tritium) neutron generator [89]. More information regarding calibrations in Super-Kamiokande can be found in [90]and[88]. 3.9 Super-Kamiokande with Gadolinium From July 14th, 2020, Super-Kamiokande o cially started its new phase, SK-VI, in which 0.02% of gadolinium sulfate, Gd (SO ) 8H O,bymasswasaddedtothe 2 4 3 · 2 water in the detector. It is usually referred as Super-Kamiokande with Gadolinium, Super-Kamiokande with Gadolinium 106 SK-Gd. The project was proposed in 2003 [78], with the main motivation is the detection of DSNB. The final concentration goal of Gd (SO ) 8H O is 0.2% by 2 4 3 · 2 mass. The detection of DSNB is through Inverse Beta Decay (IBD) in Super-Kamiokande. The products of IBD are a positron and a neutron. The Thermal Neutron Capture (TNC) on Hydrogen emits 2.2 MeV -rays, which has low detection e ciency for being close or below the Cherenkov threshold. The estimated detection e ciency for neutron captures on Hydrogen is approximately 20%, assuming that the -rays are uniformly produced in the detector [91]. With small amounts of Gadolinium (Gd) dissolved in the water of the detector, the majority of TNC will be on it, due to its high neutron capture cross section [16]. While the TNC cross section in hydrogen is only 0.3 barns, the average e↵ective cross section in gadolinium is high as 49,000 barns. The largest contributions for the neutron capture come from the isotopes 157Gd and 155Gd. Table 3.2 shows the properties of these isotopes, comparing to the capture on Hydrogen. Table 3.2: Properties of Gd isotopes, in comparison with Hydrogen. From the information in [79]. Isotope NaturalAbundance TNCcrosssection Energy 157Gd 15.65% 255,000 barns 7.9 MeV from -ray cascade 155Gd 14.80% 61,000 barns 8.5 MeV from -ray cascade H0.3barn2.2MeVfromsingle -ray Super-Kamiokande with Gadolinium 107 The -ray cascade produced by the TNC on Gd has a higher energy than on Hydrogen, so many more photons of Cherenkov photons are detected. The energy of the cascade is around 8 MeV. This will allow the detection of DSNB within afewyearsofSK-Gdoperation[78]. Other analyses such as galactic supernova neutrinos, atmospheric neutrinos, proton decay, reactor neutrinos will benefit from the Gd loading. The current phase of SK-Gd corresponds to the loading of 0.02% gadolinium sulfate by mass (or 0.01% Gd), causing approximately 50% of TNC to be on Gd. The final goal is to have Super-Kamiokande loaded with 0.2% gadolinium sulfate by mass (or 0.1% Gd), which will cause approximately 90% of TNC on Gd. The -rays from TNC produce signals in Super-Kamiokande mainly by Compton scattering electrons, which produce Cherenkov photons. Figure 3.16 shows the neutron capture e ciency for three planned phases of SK-Gd, with concentrations of gadolinium sulfate of 0.02%, 0.06% and 0.2%. 3.9.1 EGADS To simulate the gadolinium sulfate loading in Super-Kamiokande, the detector EGADS (Evaluating Gadolinium’s Action on Detector System) has been used. Features regarding the its water system were introduced in section 3.4. EGADS was designed to be a small prototype of SK-Gd. It is a water Cherenkov detector Super-Kamiokande with Gadolinium 108 Figure 3.16: Neutron capture e ciency for di↵erent gadolinium sulfate con- centrations in Super-Kamiokande. Edited from [92]. with 20-inch PMTs, but with only 200 tons of water and without an outer detector. Figure 3.17 shows a schematic view of the detector. EGADS has 227 PMTs installed, in which 148 have no cover, 16 have an FRP housing only, and 60 have both an FRP and acrylic cover like in Super- Kamiokande. It has the same photocathode coverage of 40% and a similar coil to compensate the Earth’s magnetic field. It uses the same QBEEs for data acquisi- tion [79]. EGADS was able to maintain the water quality without loss of gadolinium. The initial gadolinium sulfate loading reached the concentration of 0.2% by mass and ran for a few years. Tests for the 0.02% gadolinium sulfate loading and operation ran from March 31st, 2018 until November 18th,2020,whenitstartedthetestsfor Super-Kamiokande with Gadolinium 109 ]. 93 support frame. Reproduced from [ Scheme of EGADS detector. On the left a side view of the tank and on the right the floor and the PMT Figure 3.17: Super-Kamiokande with Gadolinium 110 the next phase of SK-Gd, in which the concentration of gadolinium sulfate will be 0.06% by mass. Besides being a prototype for SK-Gd, EGADS is used also as a neutrino detector. It has its own galactic core-collape Supernova alert system called HEIMDALL (High E ciency IBD Monitoring Detector and Automated caLL) [94]. 3.9.2 Radiopurity Many of the backgrounds in Super-Kamiokande come from radioactive decays from the impurities distributed in the water of the detector or in the PMT glass and covers. For that reason, it is important to consider a fiducial volume to eliminate events due to radioactive decays near the walls of the detector. Radon contamination, as already explained in previous sections, are partially reduced by temperature control and fiducial volume cuts. The addition of gadolinium sulfate to the water in Super-Kamiokande will bring contaminations that are uniformly distributed through the detector. Backgrounds come from the production of -rays or -rays from uranium U (Figure 3.18), tho- rium Th (Figure 3.19), and actinium Ac (Figure 3.20) chain isotopes. Acceptable Super-Kamiokande with Gadolinium 111 Figure 3.18: Uranium 238 decay chain, from [95]. Super-Kamiokande with Gadolinium 112 Figure 3.19: Thorium 232 decay chain, from [95]. Super-Kamiokande with Gadolinium 113 Figure 3.20: Actinium 227/Uranium 235 decay chain, from [96]. Super-Kamiokande with Gadolinium 114 Figure 3.21: Scheme of the gadolinium sulfate dissolving plan to the Super- Kamiokande detector. From [92]. levels of contamination from these radioisotopes were determined by the low en- ergy neutrino analyses, which established upper limits that would be tolerable for the DSNB and solar neutrino physics goals [79]. 3.9.3 First Gadolinium Loading and Future Approximately 13 tons of gadolinium sulfate Gd (SO ) 8H O were dissolved 2 4 3 · 2 between July 14th and August 18th in 2020. This loading was to achieve a concen- tration of 0.02% by mass. Figure 3.21 shows a scheme of the plan for dissolving: gadolinium loaded water is inject to the bottom of the tank. After the start of injection of the compound to the tank, samples of water were taken in four di↵erent positions in Super-Kamiokande: two in the ID and two in the OD. Temperature and conductivity were measured to trace the spread of gadolinium in the detector. With the water samples it was possible to measure the concentrations at each step of the loading program. Super-Kamiokande with Gadolinium 115 After the loading was finished, the circulation flow of the water system was set to 120 m3/hour,startingtheconvectionmodetogetauniformconcentrationof gadolinium in the tank. Water quality improved gradually, but the dark rate of all PMTs increased significantly during this phase. Figure 3.22 shows the evolution of the dark rate. The main hypothesis for the increased dark rate is the presence of bacteria in the ID and studies about the phenomenon is being currently held. Dark rates decreased after changing the water flow and dropping the temperature by 0.05 C. At the time of writing, dark rates are still decreasing and the water quality is now similar to previous Super-Kamiokande phases. EGADS is currently simulating the next phase of SK-Gd, in which the concentration of 0.06% Gd (SO ) 8H O by 2 4 3 · 2 mass will be reached in the detector. The next loading is planned for 2022. The final goal of SK-Gd is to have the concentration of 0.2% Gd (SO ) 8H O by 2 4 3 · 2 mass. 3.9.4 Simulation of TNC on gadolinium The evaluation of the pre-Supernova sensitivity at SK-Gd uses simulation of TNC on gadolinium in Super-Kamiokande. The current detector simulation, SKDET- SIM does not include the TNC on gadolinium. To better simulate SK-Gd, a new Super-Kamiokande with Gadolinium 116 the end of SK-V and first months of SK-VI (after gadolinium loading). Evolution of the water transparency in the Super-Kamiokande detector and of the PMTs dark rates for Figure 3.22: Super-Kamiokande with Gadolinium 117 Figure 3.23: Emission energies in the two considered models for TNC on gadolinium. From [16]. detector simulator based on GEANT4 [97], which will include the neutron capture on gadolinium, is in development. Following what it was done by [16], -rays vectors are generated externally to SKDETSIM. Two di↵erent -ray emission models are used for that, the generic liquid scintillator simulator GLG4SIM, developed by the Chooz experiment [98] and a second model created at the Japan Proton Accelerator Research Complex (J-PARC) after measurements of -ray properties with an array of germanium de- tectors in the Accurate Neutron-Nucleus Reaction Measurement Instrument (AN- NRI) at the facility [99]. Super-Kamiokande with Gadolinium 118 Figure 3.24: Multiplicity of -rays in the two considered models for TNC on gadolinium. From [16] The spectrum of -rays produced by TNC on gadolinium consists of a few known energy -rays close to the Q value, and a continuum at lower energies where the lev- els are so densely populated that they are indistinguishable [100]. The GLG4SIM model combines a parametric model of the continuum with the known physics of high energy -rays. The model provided by J-PARC includes its measurements with ANNRI to account -ray combinations that come together, and also a dif- ferent continuum distribution. Both models assume a isotropic distribution from the angles of -ray emission. The energy distribution and multiplicity of both methods are compared in Fig- ures 3.23 and 3.24. Some reconstructed variables in both models are shown in Figure 3.25. The two models are included in this work to provide systematic uncertainties for the pre-SN neutrino sensitivity results in Super-Kamiokande. Super-Kamiokande with Gadolinium 119 Figure 3.25: Comparison of di↵erent variables for the two models of TNC on gadolinium in consideration. From [16]. Chapter 4 Low Energy Inverse Beta Decay Detection in Super-Kamiokande As explained before, the detection of low energy electron anti-neutrinos from pre- Supernova stars in Super-Kamiokande is through the Inverse Beta Decay (IBD) interaction: ⌫¯ + p e+ + n (4.1) e ! Positrons are detected by their produced Cherenkov light, explained in Section 3.3. Neutrons are thermalised and captured by free protons and in SK-Gd also by gadolinium. They travel a short distance before being captured, meaning that an IBD event is represented by a prompt part, which corresponds to the reconstructed 120 Low Energy Inverse Beta Decay Detection in Super-Kamiokande 121 positron, and a delayed part, which is the reconstructed neutron. The time be- tween these two parts of the IBD event is called coincidence time and the distance between them is called coincidence distance. The IBD events are also called Co- incidence Events. This signature is used to distinguish background events, since the probability of uncorrelated events to produce this signature is low. The goal of this work is the implementation of a Supernova alarm based on the detection of pre-Supernova neutrinos starting from the first phase of Gd loading in Super-Kamiokande, which is 0.02% Gd (SO ) 8H O by mass. For that, methods 2 4 3 · 2 are optimized for the current phase and they will be eventually improved for next phases of SK-Gd. In this chapter details on reconstruction of IBD events is described, as well as discussing the main backgrounds for the pre-Supernova neutrinos detection and the techniques used to reduce them. 4.1 Inverse Beta Decay Identification One of the challenges for the detection of pre-Supernova⌫ ¯e in Super-Kamiokande is that the positron produced by the IBD interaction has a very low energy, since the expected⌫ ¯e energy from pre-Supernova is also low. The total energy of the Low Energy Inverse Beta Decay Detection in Super-Kamiokande 122 IBD positron is related to the⌫ ¯e energy: E + = E , (4.2) e ⌫ where = m m =1.293 MeV [100]. If not enough photons are detected, these n p events might not get reconstructed. Figure 4.1 shows the trigger e ciency of WIT, which is 100% for electrons with total energy over 4 MeV, but drops significantly below this energy. This trigger e ciency was evaluated with simulation and the figure shows true energy, not reconstructed energy [100]. The simulations were performed with SK-IV conditions and the same trigger conditions are expected for SK-Gd. The detection of the neutron produced in IBD interaction, equation 4.1,occurs by the thermal capturing of the particle by free protons or gadolinium. The TNC on gadolinium emits an approximately 8 MeV -ray cascade, while for hydrogen this cascade has only 2.2 MeV. Neutrons travel a shorter distance before being cap- tured by gadolinium compared to Hydrogen, and the shorter coincidence distance improve the vertex reconstruction resolutions and detection e ciency, lowering backgrounds and increasing tagging e ciency. The addition of gadolinium sulfate lowers the coincidence time from 220 µstoabout120µsforthe0.02%gadolinium sulfate loading. Using the coincidence detection of neutron and positron to identify the IBD Low Energy Inverse Beta Decay Detection in Super-Kamiokande 123 Figure 4.1: WIT e ciency compared to electron true total energy for events in the whole ID and only inside the FV. The Cherenkov threshold and SK-IV solar neutrino analysis threshold are showed for comparison. interaction, increases the tagging e ciency, improving the capability in distin- guishing signal from background. These two parts of the event are handled inde- pendently by WIT, as a coincidence trigger, which will characterize the detection of both positron and neutron as the same event, is still in development for the WIT system. Two detection channels are considered for the pre-Supernova detection. If both products of the IBD reaction are detected, with prompt and delayed parts, i.e. observation of coincidence events, it is called Delayed Current (DC) channel. If only TNC events are observed, it is called Single Neutron channel. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 124 The detection of the electron anti-neutrinos from a pre-Supernova will, apart from find properties of the massive stars in the stages preceding a Supernova and initial conditions of the core-collapse, give warnings to di↵erent experiments in order to prepare for the observation of the subsequent Supernova a few hours in advance. 4.2 Backgrounds The sources of background for the pre-Supernova neutrino are di↵erent for the two considered channels for the analysis, single neutrons events and delayed current (DC) events. The evaluation of backgrounds was done in two parts: initially it was reproduced the rates previously obtained in [16]andlaterare-evaluationof the background sources with new considerations needed for current status of the detector. The background studies have been also performed with an optimized selection, which will be described in Section 4.3. In this section, it will be shown the sources of background and the di↵erences between the previous and current estimations. The main backgrounds for the single neutron channel include: Fake neutrons, from radioactive decays and dark noise. • Low Energy Inverse Beta Decay Detection in Super-Kamiokande 125 Real neutrons that are not from pre-Supernova neutrinos. • Reactor neutrinos and geoneutrinos. • And for the DC channel: Accidental coincidences, from radioactive decays and dark noise. • Spontaneous Fission of 238U, due to radioisotope contamination in the de- • tector. Pairs of neutrons, also due to radioisotope contaminations. • Spallation, due to cosmic rays muons. • Reactor neutrinos and geoneutrinos. • The evaluated backgrounds for the pre-Supernova analysis will be shown in Chapter 5.Amoredetaileddescriptionofthesesourcesisgivennext. 4.2.1 Fake Neutrons and Accidental Coincidences Fake neutrons and accidental coincidences are intrinsic backgrounds for the single neutron and DC channels respectively. They come from radioactive decays and dark noise. Previous estimations of these backgrounds were done in [16]byusing Low Energy Inverse Beta Decay Detection in Super-Kamiokande 126 Figure 4.2: Comparison of expected accidental coincidences events over time for SK-IV and SK-V pure water data. For SK-V, results with improved selection techniques are shown. These techniques will be discussed in Section 4.4.3. 6000 hours of SK-IV real pure water data. The same background rates were obtained in this work for the same conditions. Since these backgrounds come mainly from dark rates, which have changed since SK-IV, a re-evaluation of these background rates were performed using approxi- mately 1000 hours of SK-V pure water data. Figure 4.2 shows the comparison of accidental coincidences rates for SK-IV and SK-V. At the time of writing, SK-VI is still in commissioning phase, so evaluations with gadolinium data will be done in the future. Other intrinsic backgrounds due to physical characteristics of the detector are Low Energy Inverse Beta Decay Detection in Super-Kamiokande 127 expected. The materials inside the detector can emanate radon and also have radioactive decays. These e↵ects happen close to the detector walls, so FV cuts generally reduce them. 4.2.2 Radioactive Contamination With the loading of gadolinium sulfate to the water in Super-Kamiokande, radioac- tive impurities are distributed in the detector. In order to avoid undesirable ef- fects in the physics analyses of Super-Kamiokande, these contamination have been carefully quantified and studied continuously before and after the first loading, by measuring them with HPGe detectors and collaborating with manufacturers to match acceptable levels of contamination [79]. Requirements for maximum levels of contamination of some radioisotopes have been defined by the Solar Neutrino and the Supernova Relic Neutrino (SRN) analyses and are shown in table 4.1. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 128 Table 4.1: Radioisotopes activity level required for the Supernova Relic Neu- trino and Solar Neutrino analyses. Requirements (mBq/kg) Chain Part of the chain SRN Solar ⌫ 238U 238U < 5- 226Ra - < 0.5 232Th 228Ra - < 0.05 228Th - < 0.05 235U 235U-< 30.0 227Ac/227Th - < 30.0 The presence of these radioisotopes in the gadolinium sulfate will be responsible for di↵erent sources of background for the pre-Supernova analysis: real neutrons, pairs of neutrons and spontaneous fission. 238Uwillcontributeprimarilytothespontaneousfissionbackground,processin which one or more neutrons are produced, leading to real neutron background to single neutron events, as well as backgrounds for DC events with the coincidence of a -ray with one of these neutrons or two neutron detection as delayed event. Di↵erent multiplicities need to be accounted when evaluating these background rates [101], depending on which event type, single or coincidence, is being analyzed. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 129 235Uontheotherhandwillcontributemainlytothepairsofneutronsback- 18 21 ground. This isotope and its chain are ↵ emitters, increasing the rates of O(↵,n) Ne⇤ 21 processes. Each reaction of Ne⇤ decay produces two neutrons, so one of the neu- trons could be mistaken for a positron, becoming a background for coincidence events. For the background rates estimations it was considered the usual isotope con- centrations in Nature of 0.20% for 18Oand0.038%for17O. Neutrons will come mainly from 18Oandthemain↵ sources are 211Po and 215Po, coming from the Ac/235Udecaychain[16]. Using the software SOURCES-4C [102], the rates of these backgrounds for both 0.2% and 0.02% gadolinium sulfate concentration were calculated for the current activity level requirements and are shown in Table (4.2). These results are also compared previous background predictions made by [100] based on di↵erent re- quirements for the activity levels from the radioisotope contaminations. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 130 Table 4.2: Rates for backgrounds per hour due to radioisotopes contamination on gadolinium sulfate. Requirements for the current analysis are shown in Table 4.1. Previous requirements for contamination di↵er from current ones for 235U < 3mBq/kg. Pairs of Neutrons Spontaneous Fission Reported in [100], for 0.2% Gd (SO ) 8H O 0.28 0.62 2 4 3 · 2 Previous requirements, for 0.2% Gd (SO ) 8H O 0.275 0.624 2 4 3 · 2 Current requirements, for 0.2% Gd (SO ) 8H O 2.298 0.624 2 4 3 · 2 Current requirements, for 0.02% Gd (SO ) 8H O 0.231 0.062 2 4 3 · 2 4.2.3 Reactor Neutrinos One of the main sources of backgrounds for pre-Supernova neutrino detection is Reactor Neutrinos due to active Japanese nuclear reactors. They are in the same energy range and are detected on Super-Kamiokande through the same interaction, so it is an irreducible background. The website geoneutrinos.org is used to evaluate reactor neutrinos and geoneu- trinos fluxes and the rate and energy spectrum of anti-neutrino interactions in Super-Kamiokande [103]. The accounting of reactor activity, the mean values given by IAEA’s PRIS for specific years of activity are used [104]. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 131 Figure 4.3: Comparison between reactor fluxes for the di↵erent years consid- ered to estimate the background due to reactor neutrinos. Figure 4.4: Background rates for the di↵erent years in consideration. Results with improved techniques are shown. These techniques will be discussed in Section 4.4.3. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 132 In [100]thefluxesofyears2010and2017areconsideredtodefinethehighest and lowest expected backgrounds. In 2010 most of the Japanese nuclear reactors were active, while 2017 represented the period with less reactor activity. For the application in the pre-Supernova alarm, as well as to give more current estimations for the reactor and geoneutrinos background, the most up-to-date information from geoneutrinos.org was used in this work. So far, the most recent values provided are for the year 2019. Figure 4.3 shows a comparison of reactor fluxes for the years of 2010, 2017, 2018, and 2019. Figure 4.4 shows the evolution of expected reactor neutrinos and geoneutrinos background rates. 4.2.4 Spallation One of the main sources of background for low energy analysis in Super-Kamiokande, such as Solar neutrino and DSNB, is the nuclear spallation reactions, which are initiated by beta decays of isotopes produced by cosmic ray muons and their pro- duced daughter particles [105]. Both the cosmic ray muons and the daughter particles pass through the detector. Neutrons produced by the cosmic rays muons are not a concern. If these fast neutrons are produced before the cosmic ray muon reaches the detector, they would need to pass over 4.5 m of water to be captured, so they mostly do not penetrate the FV. With correlations between low energy events and muon tracking, the products Low Energy Inverse Beta Decay Detection in Super-Kamiokande 133 Figure 4.5: Distribution of capture times for the fast neutrons and for neutrons from spallation daughters. From [16]. of spallation are well rejected by Super-Kamiokande by removing events associated within a time and space windows with the muon track [106]. The background from fast neutrons are controlled by vetoing for 1 ms after each muon, as they are thermalised and captured very quickly [100]. Figure 4.5 shows the distribution of capture times for the fast neutrons and for neutrons from spallation daughters. The identification of the unstable daughter particles produced by the cosmic ray muons is through their light emission profiles [105], [107]. They can produce - delayed neutrons, which are neutrons associated with the beta decay of the fission products. Figure 4.6 shows the distribution of energies of spallation products. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 134 Figure 4.6: Distribution of the energies of spallation daughters. From [16]. In particular, -ray energy from the decay of nitrogen-17 is in the energy range of interest for the pre-Supernova analysis. These isotopes should be e ciently identified by spallation reduction methods in Super-Kamiokande [106], which introduces 10% dead-time with 95% reduction e ciency, making Spallation backgrounds small compared to other background sources. 4.3 Event Selection Previous studies [16] for the sensitivity of pre-Supernova neutrinos in Super- Kamiokande with Gadolinium included the analysis of over 6000 hours of WIT Low Energy Inverse Beta Decay Detection in Super-Kamiokande 135 data for the intrinsic background estimations, accidental coincidences and fake neutrons. Also, simulations of TNC on gadolinium were generated for estimations of the other background sources described in section 4.2 and to calculate expected signal of pre-Supernova neutrinos. As already mentioned in this chapter, changes had to be applied to the previ- ous analysis in order to improve the detection capability and update background models. Since [100]waspublished,newrequirementsofactivitylevelsofradioiso- topes were defined, more currently information regarding reactor activity in Japan became available, and most importantly, gadolinium sulfate was loaded in Super- Kamiokande. The goal of this work is the creation of an online alert system, so that the event selection had to be built around the best possible system e ciency and processing speed. At the time of writing, the software SKDETSIM for SK-Gd was not ready, so the simulations were handled as described in Section 3.9.4: for TNC on Gadolinium it is used both the -ray models GLG4SIM from Chooz experiment and the one created by ANNRI at J-PARC. For coincidence events, the simulated TNC on Gadolinium events were combined with simulated positrons with SKDETSIM. These coincidence events were simulated in di↵erent positions in the inner detector, with positrons with energies between 0.8 MeV and 7 MeV. The pre-Supernova alert system will work with an online search of coincidence Low Energy Inverse Beta Decay Detection in Super-Kamiokande 136 events throughout WIT data. Single neutron events take more time for processing, so they will be only used as cross-check for the alert system. For this reason, improved techniques for event selection are for the DC channel and the updated results of sensitivity for pre-Supernova neutrino detection in SK-Gd will be given for only coincidence events. Multivariate methods are used for selection. In particular Boosted Decision Tree (BDT) methods are used at two di↵erent levels of the data processing for both speed and e ciency reasons. Apre-selectionisappliedtocandidateevents.Forsingleneutronevents,besides being inside the FV, a minimum number of hits is required by applying cuts to the variable n18 and also volume cut around calibration sources in the detector is applied. Tests from [16]showedthateventswithn18 < 26 and effwall < 800 were almost never selected, so these cuts are also included in the pre-selection. The final selection for single neutron events is based on the BDT, as discussed in further sections. 4.3.1 Selection of Coincidence Events The alert system will search for coincidence candidate events in WIT data. The prompt and delayed parts of the event are independently reconstructed. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 137 These events are characterized by two variables, dR and dT, which are the di↵erence in reconstructed position and time between the prompt and delayed parts of a coincidence event. When comparing accidental coincidences with signal, these variables have very di↵erent distributions, as shown in Figure 4.7. Candidates are initially chosen using the number of hits n18 and the BONSAI variables reconstructed position (bx, by and bz) and reconstructed time (bt). Fig- ure 4.8 shows the variables used for the search of pair candidates. These variables are available online after the events are saved by the WIT system, as described in Section 3.7.1. Not only they are used to the search for coincidence candidates, but also for training a BDT, which is called BDTonline,whichwillbedescribedin Section 4.4.3. The search for potential coincidences is performed by looping over all events. For each prompt event candidate, a delayed candidate event is searched in a time window from -17 to 290 µs relative to the prompt candidate. If the delayed candidate has n18 > 13 and the events are inside the FV, the variables dR and dT for the pair candidate is calculated. The pair candidates need to be inside the ranges 0 Gd (SO ) 8H O loading. 2 4 3 · 2 Low Energy Inverse Beta Decay Detection in Super-Kamiokande 138 Figure 4.7: Variables dR and dT for simulated signal events (black) and accidental background (red). From [16]. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 139 bx Initial Search (coincidences) by Initial Search (coincidences) Primary file Primary file neutron cut neutron cut positron cut positron cut 104 104 103 103 102 102 −1500 −1000 −500 0 500 1000 1500 −1500 −1000 −500 0 500 1000 1500 bz Initial Search (coincidences) bt Initial Search (coincidences) Primary file Primary file neutron cut neutron cut positron cut 4 positron cut 104 10 3 103 10 2 102 10 −1500 −1000 −500 0 500 1000 1500 0 5000 10000 15000 20000 bgoodness Initial Search (coincidences) n18 Initial Search (coincidences) 5 10 Primary file Primary file neutron cut neutron cut 105 positron cut positron cut 104 104 103 103 102 102 10 10 1 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 10 15 20 25 30 Figure 4.8: Variables from BONSAI Online reconstruction (bx, by, bz, bt, bgoodness) and n18. The black solid line shows the variables before any cut, the red dashed line shows the variables for coincidence positron events and blue for coincidence neutron events after the Initial Search. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 140 Many optimizations were made to the search for pair candidates. The introduc- tion of the BDTonline made the process much faster and increased the expected sensitivity, as will be shown in the next chapter. Also more strict cuts were applied for a better choice of candidates. In the analysis, neutrons are assumed to be captured at the same point of production. The assumption is reasonable as the distance travelled by the neutron is much smaller than the position reconstruction resolution [16]. The pre-selection described for single neutron events is also applied to the neutron part of coincidence events. Figure 4.9 shows the variables dR and dT before and after the pre-selection for a random data sample. The cuts applied during the search for pair candidates and the pre-selection are not enough to reduce backgrounds. For that reason Multivariate Selection methods are implemented in the analysis. 4.4 Multivariate Selection methods The reduction of backgrounds for the pre-Supernova analysis is performed by ap- plying both rectangular cuts and a particular multivariate selection method, the Boosted Decision Tree (BDT). BDTs are machine learning techniques consisting of tree structured classifiers used for discriminating events as signal or background. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 141 dR Initial Search 106 dR after initial search dR after pre-selection 105 104 103 102 10 1 10−1 0 200 400 600 800 1000 dT Initial Search 105 dT after initial search dT after pre-selection 104 103 102 10 1 3 10−1 ×10 0 20 40 60 80 100 120 140 160 180 200 Figure 4.9: Variables dR and dT calculated after the pre-selection. BDTs have many benefits when compared to other multivariate selection meth- ods. An artificial neural network is expected to perform better than BDTs, but they need large CPU times. Also, they require large data samples. BDTs on the other hand are fast, easy to tune, and have very reasonable performances. The trainings of BDTs are done using the Toolkit for Multivariate Analysis (TMVA) [108]integratedintheROOTsoftware.Simulatedsignalwasproducedusingthe GLG4SIM -ray model, already discussed in the previous chapter. For background samples, data from SK-IV was used. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 142 Rectangular cuts assume an important role during event selection for the initial searches and the pre-selection applied, as already discussed in Section 4.3. How- ever, even with optimized rectangular cuts, they are not enough the reduce the backgrounds. A more sophisticated and e cient method was developed in [16] using BDT for further reductions, which was used to estimate the sensitivity of pre-Supernova neutrinos in SK-Gd. For the development of the pre-Supernova alert system, another BDT was trained using only online variables, in order to improve the data processing time. Discussions and results will be shown in future sections. First a more explained introduction to BDT is given. 4.4.1 Boosted Decision Tree Adecisiontreeisapatternrecognitiontechniquethatclassifieseventsassignal or background. The concept of a decision tree is to not reject right away events that fail a specific criterion, searching instead whether other criteria may be more appropriate to classify these events properly. In this way, a decision tree creates a model to predict the target value of a variable by learning decision rules from the observation of this variable. Decision trees are rooted binary trees, with only two classes, signal and background. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 143 The classifier begins with an initial note, the root node, which represents the whole sample to get divided in homogeneous sets, and splits it into a tree structure, as illustrated in Figure 4.10,inwhichadecisionismadeineachnodeforadi↵erent discriminating variable, aiming to classify each event as signal or background. Each node can be recursively split into two daughters or branches, until some stopping condition is reached. For each split, the discriminating variable is the one that gives the best separation between signal and background. At each node, all variables can be considered, even if they have been used in the previous iteration. Thus, same variables can be used more than once, while others might not even be used. The split of nodes are created recursively until it reaches a stop condition, inserted by the user, declaring such node as terminal leaves. For decision trees, the CPU consumption time scales as nNlog(N), for n variables and N training events. In each node, a cut will be applied to these variables, resulting in the split of the samples into two leaves. Defining the Gini coe cient G as: G = p(1 p), (4.3) where p is the signal purity p = s/(s+b), namely the probability of an event being classified as signal, and (1 p) is the background purity. The cut in each node is chosen in a way the maximizes the di↵erence between the Gini coe cient, which Low Energy Inverse Beta Decay Detection in Super-Kamiokande 144 Figure 4.10: Output from the TMVA software [108] showing a random deci- sion tree that was used for the training for signal and background separation in the pre-Supernova neutrino analysis. Further discussions regarding the dis- criminating variables used in this scheme will be held in next sections. gives the probability of misclassification, of the original leaf and the split leaves (weighted by the proportion of events). In a BDT, the concept of decision tree is extended from one to several trees. They are derived from the same training ensemble by reweighting events, resulting in a final classifier that considers the average of each individual decision tree. This is very relevant for fluctuations in the training sample, given a better performance than a single tree [108]. There are di↵erent methods used for boosting, for example AdaBoost [109], Gradient Boosting [110] and NGBoost [111]. The one chosen for this study is AdaBoost (adaptative boost), in which, for every new decision tree, data that were misclassified in the previous one are given an adjusted weight for Low Energy Inverse Beta Decay Detection in Super-Kamiokande 145 the following tree. The weight of each event in the new decision tree is multiplied by the boost weight ↵boost, given by equation 4.4. 1 ✏ boost ↵ = , (4.4) boost ✏ ✓ ◆ where ✏ is the misclassification rate of the tree and boost is the boost rate related to the learning rate of the classifier, inserted by the user. The event classification yboost(x)ofN classifiers combined using AdaBoost is given by equation 4.5. 1 N y (x)= ln(↵ )h (x), (4.5) boost N i i i X th where ↵i is the weight of the i ensemble and hi(x)istheclassificationforthetree ifortheinputvariablex.Smallvaluesofyboost(x)meanstheeventisbackground- like, while large values means signal-like. To compare the e ciency between di↵erent BDT trainings, and also other clas- sifiers, a classic method is to get the area under the plot of background rejection against signal e ciency, known as the Response Operator Characteristic (ROC) curve. Figure 4.11 shows an example of these curves for di↵erent classifiers. In the figure, BDT is compared to rectangular cuts and Fisher Linear Discriminant, Low Energy Inverse Beta Decay Detection in Super-Kamiokande 146 Figure 4.11: ROC curves example for di↵erent classifiers. For this example, in comparison to the Boosted Decision Tree (BDT) ROC curve, are the Fisher Linear Discriminant (Fisher) and the rectangular cuts (Cuts). which is a method to perform event selection by distinguishing mean values of variable distributions for signal and background [108]. The Fisher Linear Discrim- inant is an optimal classifier that creates an axis in each input variable hyperspace, pushing background and signal events in this axis as far away from each other as possible. Figure 4.11 shows that the Fisher Linear Discriminant has a lower e - ciency when compared to the BDT for the pre-Supernova analysis. 4.4.2 Previous Estimations As stated before, in the previous study of pre-Supernova neutrinos [100], a BDT was used to predict the sensitivity of these neutrinos in Super-Kamiokande. For the Low Energy Inverse Beta Decay Detection in Super-Kamiokande 147 pre-Supernova alert system creation, some optimizations were performed, as well as the development of a new BDT using only online variables. Before discussing the currently used BDTs, it will be summarized in this section the main features and variables considered for the BDT used in [100]. For the training and testing of the BDT, representative SK-IV real pure water data were used, extracted as random sub-samples of many di↵erent runs. The training on broader samples had poor performances. As for the resources available, the memory was limited to 4096 MB, and be- cause the TMVA software loads entire event tree into memory while running, the numbers of used variables and events were limited (107 events per 10-20 variables). Di↵erent values for the parameters to train the BDT were tested to find the best overall performance, in particular NTree and boost,whicharethenumberof trees in the BDT structure and the learning rate of the AdaBoost algorithm [108]. The number of trained events for both signal and background were varied, so as the variables used for training. Details regarding di↵erent trainings and tests can be found in [16]. The variables considered, in order of importance to the classifier, for the BDT used in [16]are: Low Energy Inverse Beta Decay Detection in Super-Kamiokande 148 fromwall :distancefromthereconstructedvertextothedetectorwall.Radioactivity close to the wall generates background events; e↵wall : distance in reconstructed direction from reconstructed vertex to the detec- tor wall. Again background events from radioactivity with direction pointing towards the nearest PMT; MSG : Multiple Scattering Goodness (MSG), measure of hit pattern. Events from asinglemultiplescatteredelectron/positronhavehighscore,whilesignal has low score (multiple gamma rays); cgoodness : ClusFit goodness of fit. Some background events do not fit well due to being comprised of a high proportion of dark noise; bz :reconstructedvertexverticallocation.Backgroundismorelikelytobe found in the bottom of the detector due to the pattern of contamination; pclose :proportionofhitpairswithcloseangle,backgroundshavehitsthatare close in angle to each other; bsdirks : BONSAI reconstructed direction KS test; Beta[1] :Isotropyparameter.Isotropicdistributionsofhitshavescorescloserto zero. The neutron capture on Gadolinium produces more isotropic hit pat- terns. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 149 mode Ch3hit :binnedmodelopeningangleof3hitcombinations.Backgroundeventsare more likely to contain opening angles close to ⇡/2; mean Ch3hit :Meanthreehitconecosine.Eventsinwhichthehitswereproducedby Cherenkov emisson will have cosine angles close to the Cherenkov angle; Beta[3] :Isotropyparameter,similartoBeta[1]; cldirks : ClusFit reconstructed direction KS. Test of the hit pattern against expected for single electron in reconstructed direction. Lower score for capture on gadolinium; bgoodness : BONSAI goodness of fit, similar to cgoodness; n18 :numberofselectedonlinehits. Regarding the performance of the BDT, it was much better than rectangular cuts and linear discriminant methods, as showed already in Figure 4.11. The strategy in [16]hasbeenusedasapreliminarystepanditwasfollowed by many attempts to get a new classifier with improved e ciencies. A BDT with o✏ine variables with improved e ciency was achieved and, among all attempts, some trainings were performed using only online variables, which are variables available after the triggering. They are further referenced as BDTonline. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 150 4.4.3 BDTonline As it will be shown in Chapter 6, the pre-Supernova alert system is an online system installed in Super-Kamiokande that needs to process many data sets look- ing for IBD events candidates and run evaluations to make alarm decisions and send alerts. For that reason, the system needs to process data as fast as possible, but without losing detection e ciency. With that in mind, the BDTonline was developed. Due to the complexity of the BDT trained with o✏ine variables, it is natural to expect that it would be more e cient than a BDT trained with only online variables. Figure (4.12)showstwoBDTonline’s ROC curves, which di↵er from each other by the choice of parameters NTrees and boost and used variables, with the second BDTonline having a better performance. They were trained to get e ciencies close to the one with o✏ine variables. The performance of the BDTonline reaches a limit of e ciency at certain point. However, both of the BDTs, online and o✏ine, can be used during di↵erent stages of the data processing. The importance that each variable used by the chosen BDTonline has for the classification is listed in Table (4.3). Also in this case, the boosting method chosen was AdaBoost, but the choice of parameters was di↵erent to the BDT with o✏ine variables. In particular, the most e ciency BDT had the parameters: NTree = 1000, boost =1.0. To have a less stricted BDT, cuts for training and testing were Low Energy Inverse Beta Decay Detection in Super-Kamiokande 151 ROC curves 1 0.95 0.9 0.85 Backgr rejection (1-eff) 0.8 Reproduced 0.75 BDT with offline variables 0.7 BDTonline #1 BDTonline #2 0.65 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 Signal eff Figure 4.12: ROC curves from reproduced BDT used in [16] (black), the new BDT with o✏ine variables used for Final Selection (blue) and two di↵erent trials to get the best performance out of a training using only online variables (green and yellow). chosen for a broader event selection than what is done by the BDT with o✏ine variables. For that, it was extended the range of allowed hits, which characterized by the variable n18. Low Energy Inverse Beta Decay Detection in Super-Kamiokande 152 Online variables BDT 10−1 10−2 Signal 10−3 Background 10−4 10−5 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 BDT score Figure 4.13: Background and Signal separation for the BDT with online vari- ables. Table 4.3: Ranked importance of the used online variables. Rank Variable Importance 1n183.949e-01 2bx*bx+by*by2.768e-01 3bt1.260-01 4bgoodness8.760e-02 5by7.627e-02 6bx6.308e-02 7bz5.055e-02 Low Energy Inverse Beta Decay Detection in Super-Kamiokande 153 Figure (4.13) shows the separation between background signal for coincidence events. For this figure, random WIT data corresponding to approximately 200 hours were used for background sample and for the signal sample simulated coin- cidence events using the GLG4SIM -ray model. Initially it was verified that the BDTonline improved the speed of the data pro- cessing. A cut was chosen by means of selecting the same number of events as if the BDT was not applied. Reductions were applied to the same samples with and without the BDTonline to compare its performance. A random cut on BDT score > 0.6resultedinselectingthesamenumberofeventsforall online samples that were processed without the BDTonline. However, the process with the BDTonline was much faster. For di↵erent memory sizes of data being analyzed, the comparison between the time spent by the system to process data is shown in Table 4.4. Since the data evaluation for the online alert system needs to be fast, these results are very important. The BDTonline allows the system to achieve optimal processing speed. Table 4.4: Approximate time spent to process data for previous approach and optimized (with BDTonline). Number of WIT data subsets Time spend with previous approach Optimized 13minutes40seconds 10 18 minutes 1.5 minutes 100 3 hours 4 minutes Low Energy Inverse Beta Decay Detection in Super-Kamiokande 154 The previous cut on BDT score > 0.6, to evaluate its speed performance, online is not the best cut choice for the signal background separation, which means that the BDTonline can also be used to improve the sensitivity of pre-Supernova neutrinos in Super-Kamiokande. Optimized results, consequent of the application of di↵erent cuts on the BDTonline score, will be shown in the next chapter. Chapter 5 Pre-Supernova Neutrino Sensitivity at SK-Gd Previous sensitivity results for detection of pre-Supernova neutrinos at Super- Kamiokande with 0.2% Gd (SO ) 8H O by mass were given in [100]. However, 2 4 3 · 2 the first phase of SK-Gd has 0.02% Gd (SO ) 8H O by mass and the develop- 2 4 3 · 2 ment of a Supernova alert system based on the detection of pre-Supernova neu- trinos required a dedicated response model for the current phase. In this chapter, sensitivity results to pre-Supernova neutrinos and estimations for the alert system in the first phase of SK-Gd will be shown. General details about the system itself will be covered in next chapter. 155 Pre-Supernova Neutrino Sensitivity at SK-Gd 156 5.1 Previous Estimations In [100], sensitivity results were given for the loading in SK of 0.2% Gd (SO ) 2 4 3 · 8H2O by mass, which is the goal for the final concentration of gadolinium in Super-Kamiokande. Cuts were chosen to give the greatest range of detection of a 25 M pre-Supernova start for the Odrzywolek model [71], 0.1 hour before the core-collapse. These choices were based on the facts that real coincidences would be close together in time and space and that events with higher BDT score are more like the model of TNC on gadolinium. The cuts were: BDTscore > 0, coincidence distance dR < 300cm,andcoincidencetimedT < 100µs. Figure 5.2 shows the distributions of these variables before and after cuts for signal and main backgrounds. As for the single neutrons channel, the final selection was made only with the BDT, which score had to be greater than 0.56. Figure 5.1 show the distribution of the variable for this channel. Trigger and selection e ciency were found to be 4.3-6.7% for coincidence events and 9.5-10% for single neutron events for the last 12 hours before core-collapse. The uncertainty was given by di↵erent neutrino spectrum models. The results were also a combination of the single neutron and DC channels, as they are highly correlated. The signal window chosen to perform statistical evaluations was 12 hours. Results are dependent on neutrino mass ordering, ZAMS Pre-Supernova Neutrino Sensitivity at SK-Gd 157 Figure 5.1: Distribution of the variables from the DC channel in the final selection for [16]results. Figure 5.2: Distribution of the BDT score for the single channel in the final selection for [16]results. Pre-Supernova Neutrino Sensitivity at SK-Gd 158 mass of the star, distance to the star, and background level, and were given based on the red supergiant star ↵ Ori (Betelgeuse) as it is currently the best candidate for detecting pre-Supernova neutrinos due to its mass and distance from the Earth. Table 5.1: Number of expected signal events from pre-Supernova star at 200 parsecs in the final 12 hours before core-collapse at SK-Gd with 0.2% Gd (SO ) 8H O, from previous estimations [16]. 2 4 3 · 2 Model Mass (M ) MassOrdering Singleneutron DC Odrzywolek 15 Normal 55-71 33-36 Odrzywolek 15 Inverted 16-20 9-10 Odrzywolek 25 Normal 120-160 59-65 Odrzywolek 25 Inverted 34-44 17-18 Patton 15 Normal 65-84 45-50 Patton 15 Inverted 18-23 13-15 Patton 30 Normal 130-170 100-110 Patton 30 Inverted 40-52 34-37 Table 5.1 summarizes the expected number of signal events from pre-supernova stars at 200 parsecs away in the last 12 hours before core-collapse and the estimated background for a 12-hour window in SK-Gd. Tables 5.2 and 5.3 show the most optimistic range of detection of pre-Supernova stars and the early warning in Super-Kamiokande, with models following predictions of mass and distance for Pre-Supernova Neutrino Sensitivity at SK-Gd 159 Betelgeuse. Scenario uncertainties are used, for di↵erent TNC on gadolinium models and di↵erent background rate estimations. Table 5.2: Maximum range of detection of pre-Supernova stars by SK-Gd with 0.2% Gd (SO ) 8H O, from previous estimations [16]. 2 4 3 · 2 Maximum Range for Model Mass (M ) FPR = 1/year FPR = 1/cy (pc) (pc) (pc) (pc) Odrzywolek 15 Normal 300-400 250-300 Odrzywolek 15 Inverted 160-200 130-200 Odrzywolek 25 Normal 330-400 280-400 Odrzywolek 25 Inverted 180-200 150-200 Patton 15 Normal 420-600 360-500 Patton 15 Inverted 220-300 190-200 Patton 30 Normal 480-600 410-500 Patton 30 Inverted 270-400 230-300 Pre-Supernova Neutrino Sensitivity at SK-Gd 160 Table 5.3: Expected early warning in SK-Gd with 0.2% Gd (SO ) 8H O, 2 4 3 · 2 from previous estimations [16]. Warning (hours) for Model Mass/Distance MassOrdering FPR=1/year FPR=1/cy Odrzywolek 15M /150pc Normal 5.3-8.4 3.4-6.3 Odrzywolek 15M /150pc Inverted 0.1-2.0 0.0-0.8 Odrzywolek 25M /250pc Normal 4.7-7.4 3.3-5.7 Odrzywolek 25M /250pc Inverted 0.0-0.6 0.0-0.0 Patton 15M /150pc Normal 7.1-14.1 5.1-9.8 Patton 15M /150pc Inverted 0.3-4.1 0.0-2.2 Patton 30M /250pc Normal 1.0-1.6 0.7-1.1 Patton 30M /250pc Inverted 0.1-0.4 0.0-0.1 The plan for the pre-Supernova alert system is to implement it starting from the first phase of loading, which is 0.02% Gd (SO ) 8H O by mass. For that 2 4 3 · 2 reason all the methods were optimized for the lower loading and will be eventually improved for next phases of SK-Gd. A few estimations for the first loading were also given in [16], but for the same background rates as the 0.2% loading. A full evaluation of the sensitivity for the first phase of SK-Gd is the subject of this work and it will be shown next. Pre-Supernova Neutrino Sensitivity at SK-Gd 161 5.2 Results for First Gadolinium Loading As described in Chapter 3, it was loaded to the water in the Super-Kamiokande detector 0.02% Gd (SO ) 8H O by mass. [100]predictedthat,forSK-Gdwitha 2 4 3 · 2 concentration of 0.2% Gd (SO ) 8H O by mass it would be possible the detection 2 4 3 · 2 of pre-Supernova neutrinos, even when considering higher background rates from di↵erent sources. However, the first phase of SK-Gd has a lower concentration of gadolinium sulfate, and in order to understand the viability of developing an alert system for Supernova based on the detection of pre-Supernova neutrinos, their sensitivity for the first phase of SK-Gd is evaluated in this section. Response models were created for the 0.02% loading. The TNC on gadolinium at this concentration has an e ciency of 50%, instead of 90% aimed for the 0.2% loading [92]. Even though the signal e ciency is lower, background rates due to radioisotope contamination are also reduced. For instance, one of the main backgrounds sources of the coincidence channel are pairs of neutrons, which are significantly less due to a lower contamination when compared to higher loadings of gadolinium sulfate. Also, all the backgrounds are reduced after introducing the BDTonline.Thelowerconcentrationofgadoliniumsulfatehasalsoalonger capture time, of about 120 µs, while for the 0.2% loading this time is expected to be around 30 µs. Pre-Supernova Neutrino Sensitivity at SK-Gd 162 Only the DC channel was used to estimate the sensitivity. As it was explained in the last chapter, the alert system will look for coincidence events in the WIT data, so the sensitivity evaluations need to be consistent to the methods applied in the system. Single neutron events might still be used for cross-check in case of potential alerts. Uncertainties using di↵erent models for TNC in gadolinium are calculated, even though they are not useful for statistical evaluation. Discussions regarding uncertainties will be held in the next chapter. The final selection used for the sensitivity studies, and also applied by the alert system, will be discussed next. 5.2.1 Final Selection With the implementation of the BDTonline described in Chapter 4, it was possible to improve the sensitivity estimations. Many cuts were tested to find the optimal cut to the BDT score. The best was BDT score > 0.2. Another online online important feature is that the higher the cut in the BDTonline,thefasterthedata processing will be, as less events will be carried through the reduction after the initial search for pair candidates. Other optimization cuts to the variables used for final selection were also per- formed in order to get the longest early warnings possible for the alarm. For a lower gadolinium loading, it is expected a longer coincidence time. Many cuts on Pre-Supernova Neutrino Sensitivity at SK-Gd 163 7 Accidental 7 Accidental 10 10 Two Neutrons Two Neutrons 6 106 Signal 10 Signal 5 5 10 10 Events per 4e-03 Events per 0.1 MeV 104 104 103 103 102 102 10 10 1 1 2 3 4 5 6 7 8 9 10 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 Prompt Reconstructed Total Energy (MeV) BDT Score of Delayed Events 107 Accidental Accidental 106 Two Neutrons Two Neutrons 106 Signal Signal 105 105 Events per 10 cm 104 Events per 1.7e+03 ns 4 10 3 10 3 10 102 10 102 3 1 ×10 0 100 200 300 400 500 600 700 800 900 1000 0 20 40 60 80 100 120 140 160 180 200 Coincidence Distance (cm) Coincidence Time (ns) Figure 5.3: Distribution of variables in the coincidence channel. Solid lines show before the final selection and dashed lines after. the BDT with o✏ine variables, used for final selection, were tested to get the best e ciency possible. The optimized cuts used for final selection are: dR < 300cm, dT < 150µs, and BDT score > 0.3. All of these cuts, including the one in BDTonline,are di↵erent for other concentrations of gadolinium. Figure 5.3 shows the distribution of variables before and after applying the final selection. The remaining backgrounds for the DC events per 7 hour window are shown in Table 5.4. Pre-Supernova Neutrino Sensitivity at SK-Gd 164 Table 5.4: Remaining background events per 7 hour window for the DC events after final selection. Source NumberofEvents(/7hours) Accidentalcoincidences 1.44 Pairs of neutrons 1.25 Spontaneousfissioncoincidences 0.02 Reactorneutrinosandgeoneutrinos 0.23 5.2.2 Statistical Parameters Before moving to the results for the pre-Supernova neutrino sensitivity at the first phase of SK-Gd, a few statistical considerations need to be discussed. Some important statistical parameters need to be taken into account, such as signal and background time-windows, hypothesis tests and false positive rates. The statistical evaluation that will be done in the alarm system is also used to provide the early warnings and estimate the range of detection of pre-Supernova stars. The chosen approach is to look at this as a Poisson Counting Experiment [112], calculating the p-value for hypothesis testing. False Positive Rate (FPR) also needs to be considered in order to take into account the rate in which Supernovae happen at a detectable distance. Choices of FPR for the alert system will be discussed in the next chapter. Pre-Supernova Neutrino Sensitivity at SK-Gd 165 Consider the hypothesis in which only background is present in a sample, called null hypothesis H0.Agoodsignalmeasurementwouldconsisthavingthesample inconsistent with this hypothesis. The probability of a test statistic to have a value consistent with the null hypothesis is called p-value. The p-value will have very small values in the presence of signal, which leads to believe that the sample actually belongs to an alternative hypothesis H1. For a counting experiment, the p-value is the probability of counting the number of events greater or equal to the case where we have no signal, only background. One way to calculate the p-value of the total number of selected events inside a pre-defined time window Nevents,foranexpectednumberofbackgroundevents NBG is: 1 p = P (N N )= P ois(n; N )(5.1) events BG BG n=N Xevents where P ois(n; NBG)isthePoissondistributionofthevariablen with average NBG. The evidence of a signal is if the signal has a significance of at least ”3 ”anda discovery is for ”5 ”. The significance level Z can be evaluated from the p-value from: 1 1 x2/2 p = e dx =1 (Z)(5.2) p2⇡ ZZ Pre-Supernova Neutrino Sensitivity at SK-Gd 166 For Poisson counting experiments in which the number of events Nevents has the form: Nevents = s + b (5.3) were s is the expected number of signal events and b is the expected number of background events, the significance level can be approximated to: Z = 2 s (5.4) (s + b)log(1 + b ) s s The significances are then evaluated using Equation 5.4.Fordi↵erentBetel- geuse models, it is found which is the best time window to perform the statistical evaluations. Single neutron events can be used as a cross-check in case of positive alerts from the alarm system. In this case, the Stou↵er’s method [113]canbeusedtocombine the significance for both channels. The results for the updated sensitivity for pre-Supernova stars at 0.02% loading SK-Gd will be shown next. Pre-Supernova Neutrino Sensitivity at SK-Gd 167 5.2.3 Results The results are dependent on the models for pre-Supernova being considered, which change in neutrino mass ordering, ZAMS mass of the star, distance to the star, and star evolution model, which was discussed in Chapter 2. The models for TNC on gadolinium being used are the -ray models GLG4SIM from Chooz experiment and the model from ANNRI at J-PARC. The background rate used for the evaluation of the sensitivity at the 0.02% load- ing SK-Gd was predicted using data for the accidental coincidences backgrounds and simulations for other sources like reactor and geo neutrinos and pairs of neu- trons. As already discussed in last section, with the introduction of the BDTonline, optimizations of other cuts and the updates performed to the response models, the irreducible background rate is 0.306 events/hour. Further discussions regarding uncertainties and updates will be held in next chapter. ↵-Ori (Betelgeuse) is currently the best candidate for detecting pre-Supernova neutrinos. In current estimations, its mass and distance from the Earth are cor- related, and for that matter, the results are given in combinations of: 150 pc for 15M and 220 pc for 25M [114; 115]. The signal window chosen to perform statistical evaluations was based on which the best significances for detection of pre-Supernova neutrinos from Betelgeuse would be achieved, which is 7 hours. Pre-Supernova Neutrino Sensitivity at SK-Gd 168 Table 5.5 shows the expected number of events of DC events at 200 pc away in the final 7 hours before the core-collapse. Results are shown for the GLG4SIM and ANNRI -ray emission models, as discussed in Section 3.9.4.Invertedorder neutrino mass models have very low number of expected events and for that rea- son, significance levels evaluated for these models are almost always below any acceptable False Positive Rate, apart from the model Patton with 30M ,whichis unrealistic for this distance. This was already predicted by [16]forthefirstphase of SK-Gd, but it won’t be the case for the next phases. Table 5.5: Number of expected signal events from pre-Supernova star at 200 parsecs in the final 7 hours before core-collapse at SK-Gd with 0.02% Gd (SO ) 8H O. Results are shown for two -ray emission models. 2 4 3 · 2 Model Mass (M )MassOrdering DC DetectionSignificance( ) Odrzywolek 15 Normal 13.83-15.42 5.54-6.04 Odrzywolek 15 Inverted 3.92 - 4.38 1.95 - 2.14 Odrzywolek 25 Normal 24.40-27.19 8.55-9.27 Odrzywolek 25 Inverted 6.92 - 7.72 3.17 - 3.47 Patton 15 Normal 19.14-21.33 7.12-7.73 Patton 15 Inverted 5.57-6.21 2.64-2.89 Patton 30 Normal 47.46-52.96 13.84-14-64 Patton 30 Inverted 15.50-17.30 6.06-6.60 Pre-Supernova Neutrino Sensitivity at SK-Gd 169 Table 5.6: Maximum range of detection of pre-Supernova stars by SK-Gd with 0.02% Gd (SO ) 8H O, for normal ordering neutrino mass models. Results 2 4 3 · 2 are shown for two -ray emission models evaluated over a 7-hour window. Maximum Range for Model Mass (M ) > 3 FPR = 3.2/cy FPR = 1/cy > 5 (pc) (pc) (pc) (pc) Odrzywolek 15 310-320 270-280 260-270 230-240 Odrzywolek 25 400-420 350-370 340-360 300-310 Patton 15 360-380 320-330 300-320 270-280 Patton 30 560-590 490-510 470-490 410-430 Figure 5.4 shows the evolution in significance level prior to the core-collapse for normal ordering neutrino mass models. As expected, the event rates increases significantly closer to the explosion, leading to greater significances right before the Supernova. The maximum range of detection of pre-Supernova stars are summarized in Table 5.6. Expected early warnings for the pre-Supernova alarm are shown in Figures 5.5 and 5.6 and 5.7 shows the early warning for Betelgeuse-like models. Di↵erent false positive rates are being considered in these results, as they will be important for further discussions about alarm decision. Pre-Supernova Neutrino Sensitivity at SK-Gd 170 Figure 5.4: Evolution of significance level for the normal ordering neutrino mass models for stars at 200 pc in the last hours before the core-collapse at SK-Gd with 0.02% Gd (SO ) 8H O. 2 4 3 · 2 From these results we see that, for neutrino normal mass ordering, signifi- cantly high significance levels are expected in the first phase of SK-Gd with 0.02% Gd (SO ) 8H O by mass, allowing to claim discoveries and it is also an evidence 2 4 3 · 2 for the neutrino mass ordering. Also, the results in Table 5.5 and Figures 5.5 and 5.6 prove the viability of developing an alert system for Supernova based on the detection of pre-Supernova neutrinos, with potential early warning of five hours. Pre-Supernova Neutrino Sensitivity at SK-Gd 171 Figure 5.5: Expected early warning above 3 against the distance for 15 M and 25 M stars for the considered pre-Supernova models, evaluated in steps of 0.01 kparsec. Dashed line shows predictions using the GLG4SIM -ray model and solid line for the ANNRI -ray model evaluated over a 7-hour window. Table 5.7: Expected early warning in SK-Gd with 0.02% Gd (SO ) 8H O, 2 4 3 · 2 for Betelgeuse-like models with normal ordering neutrino mass models. Results are shown for two -ray emission models evaluated over a 7-hour window. Warning (hours) for Model Mass/Distance > 3 FPR = 3.2/cy FPR = 1/cy Odrzywolek 15M /150pc 3.1-3.2 2.3-2.4 2.2-2.3 Odrzywolek 25M /220pc 3.4-4.1 3.2-3.3 3.1-3.3 Patton 15M /150pc 5.0-5.2 4.1-4.2 3.8-4.1 Pre-Supernova Neutrino Sensitivity at SK-Gd 172 Figure 5.6: Expected early warning considering a rate of 3.2 Supernovae per century against the distance for 15 M and 25 M stars for the considered pre-Supernova models, evaluated in steps of 0.01 kparsec. Dashed line shows predictions using the GLG4SIM -ray model and solid line for the ANNRI -ray model evaluated over a 7-hour window. Discussions about the development of the pre-Supernova alarm will be held in Chapter 7. Next phases of SK-Gd will have greater concentrations of gadolinium sulfate and a predictions for their pre-Supernova sensitivity will be shown next. 5.3 Future Gadolinium Loading Phases As already shown in Chapter 4, SK-Gd will have a few phases corresponding to the concentration of gadolinium sulfate being loaded to the detector. The Pre-Supernova Neutrino Sensitivity at SK-Gd 173 first phase, with 0.02% Gd (SO ) 8H O initiated in 2020 and its sensitivity for 2 4 3 · 2 pre-Supernova neutrinos was shown in this chapter. The next proposed phase is for a concentration of 0.06% Gd (SO ) 8H O and the final phase will have 2 4 3 · 2 0.2% Gd (SO ) 8H O. The EGADS detector is already loaded with 0.06% 2 4 3 · 2 Gd (SO ) 8H O for simulating the future loading in Super-Kamiokande. 2 4 3 · 2 It is expected that for higher concentrations of gadolinium sulfate in Super- Kamiokande, the TNC will be more e cient, which will improve the capabilities for pre-Supernova neutrino detection. However, the higher concentrations carry also more backgrounds due to radioisotope activities. The change in rates due to this contamination is shown in Table 5.8. Table 5.8: Expected rates for the Spontaneous Fission and Pairs of Neutrons backgrounds for the three phases of SK-Gd. Rate (/hour FV) ⇥ Gd (SO ) 8H O concentration by mass Spontaneous Fission Pairs of Neutrons 2 4 3 · 2 0.02% 0.06 0.23 0.06% 0.38 1.37 0.20% 0.62 2.28 These next phases of SK-Gd will benefit of many studies using data from the current stage. For example, models for background evolution in time can be Pre-Supernova Neutrino Sensitivity at SK-Gd 174 created and multivariate methods will be trained using gadolinium data, giving more reliable a selection. For now, keeping the same cuts and BDT used for the 0.02% Gd (SO ) 8H O 2 4 3 · 2 loading, as described in Section 5.2.1,afewpredictionsaremadeforthenexttwo phases of SK-Gd. Tables 5.9 and 5.10 show the expected range of detection and tables 5.11 and 5.12 show the early warnings for Betelgeuse-like models. Table 5.9: Maximum range of detection of pre-Supernova stars by SK-Gd with 0.06% Gd (SO ) 8H O, for normal ordering neutrino mass models. Results 2 4 3 · 2 are shown for two -ray emission models evaluated over a 7-hour window. Maximum Range for Model Mass (M ) > 3 FPR = 3.2/cy FPR = 1/cy > 5 (pc) (pc) (pc) (pc) Odrzywolek 15 340-360 300-320 290-300 250-270 Odrzywolek 25 450-470 400-420 380-400 330-350 Patton 15 400-420 350-370 330-350 300-310 Patton 30 610-640 540-560 510-540 450-480 Pre-Supernova Neutrino Sensitivity at SK-Gd 175 Table 5.10: Maximum range of detection of pre-Supernova stars by SK-Gd with 0.2% Gd (SO ) 8H O, for normal ordering neutrino mass models. Results 2 4 3· 2 are shown for two -ray emission models evaluated over a 7-hour window. Maximum Range for Model Mass (M ) > 3 FPR = 3.2/cy FPR = 1/cy > 5 (pc) (pc) (pc) (pc) Odrzywolek 15 350-370 310-330 300-310 260-280 Odrzywolek 25 470-490 410-430 400-420 350-370 Patton 15 410-430 360-380 350-370 310-320 Patton 30 630-670 560-590 530-560 470-500 Table 5.11: Expected early warning in SK-Gd with 0.06% Gd (SO ) 8H O, 2 4 3 · 2 for Betelgeuse-like models with normal ordering neutrino mass models. Results are shown for two -ray emission models evaluated over a 7-hour window. Warning (hours) for Model Mass/Distance > 3 FPR = 3.2/cy FPR = 1/cy Odrzywolek 15M /150pc 4.1-4.2 3.1-3.2 2.4-3.1 Odrzywolek 25M /220pc 4.2-4.3 3.6-4.1 3.3-3.6 Patton 15M /150pc 5.2-6.0 4.3-5.1 4.2-5.0 Pre-Supernova Neutrino Sensitivity at SK-Gd 176 Table 5.12: Expected early warning in SK-Gd with 0.2% Gd (SO ) 8H O, 2 4 3 · 2 for Betelgeuse-like models with normal ordering neutrino mass models. Results are shown for two -ray emission models evaluated over a 7-hour window. Warning (hours) for Model Mass/Distance > 3 FPR = 3.2/cy FPR = 1/cy Odrzywolek 15M /150pc 4.2-5.1 3.2-4.0 3.1-3.2 Odrzywolek 25M /220pc 4.3-5.1 4.0-4.1 3.3-4.0 Patton 15M /150pc 6.0-6.1 5.0-5.2 4.3-5.1 In figure 3.16 of Section 3.9, it was shown the improvement of neutron cap- ture e ciency with the increasement of the concentration of gadolinium sulfate in Super-Kamiokande. Next phase with 0.06% concentration by mass is expected for 2022. Many studies are still to be done to improve the sensitivities for these two next phases. Chapter 6 Pre-Supernova Alarm Having determined the expected early warnings, expected detection range, and statistical tools, the creation of an alert system based on the detection of pre- Supernova neutrinos was possible. The goal is to have an alarm in Super-Kamiokande that will send alerts to the community in case of a potential core-collapse Super- nova, which is called pre-Supernova alarm. It was shown in the last chapter the viability to design the system for the first phase of SK-Gd with 0.02% Gd (SO ) 8H O.Detailsaboutdataacquisition, 2 4 3 · 2 event selection, alarm decision, and the design of the system will be shown in this chapter. 177 Pre-Supernova Alarm 178 6.1 Preliminary Considerations The first version of the pre-Supernova alarm is a quasi-online system. The o✏ine reconstruction is still being applied to improve the selection e ciency, as the some of the background reductions in Super-Kamiokande are applied o✏ine. Discussions regarding future plans for online background reductions will be held in Section 6.4. Because of the impossibility of creating a complete online system for the pre- Supernova alarm currently, optimizations were performed in order to reach a faster o✏ine reduction for every new data being analyzed by the system. The creation and implementation of the BDTonline was the main accomplishment, followed by optimization in cuts of other variables and studies of best statistical parameters for the alarm decisions. A smart and e cient system had also to be developed for storing only necessary information and using less memory as possible. Many tests were performed for di↵erent SK data periods and stages of data acquisition in the WIT trigger. Figure 6.1 shows the data flow for the WIT system. The alarm is currently installed in the Organizer PC, processing data right after their acquisition and before being sorted by the Organizer processes, which assemble all individual data blocks collected into sequential sub-runs (87.381 seconds of data). The organizer takes a significantly time for its processes and for this reason, the alarm system was best placed before it, having access to new data instantly. Pre-Supernova Alarm 179 Figure 6.1: Edited from [83]. Illustration of the acquisition system in Super- Kamiokande and where the pre-Supernova alarm is placed. The alarm system receives constantly a sub-set of events, which then goes through the reduction and the number of candidate events selected from the sub- set is saved. All the selected events inside a specific time-window is accounted and the system performs a statistical evaluation to give an alarm decision. False alerts are also taken into consideration. For calibration sources inside the detector it is applied a volume cut around them. Also, when a calibration work is being performed in Super-Kamiokande, the system will ignore data from that period. False alerts from a statistical point of view will be discussed in the next section. Pre-Supernova Alarm 180 6.2 Alarm Decision The alarm decision is whether the software should send an alert or not. Many considerations have to be taken into account: Background levels; • False Positive Rates; • Alarm e ciency; • Current detector status; • Concentration of loading; • Expected early warning; • Expected range of detection. • After processing new data, the number of selected events needs to go through a statistical evaluation in order to understand what is the significance of the detec- tion. To do it so, we look at it as a Poissonian Counting Experiment, as explained in Chapter 5. It is important to notice that the background window used is of 7 hours. The number of events used for the statistical evaluation also have to account for the Pre-Supernova Alarm 181 same time window. The system accounts the number of events in the last 7 hours to perform all the evaluations and make the appropriate alarm decision. So, using equations (5.1)and(5.4), the system calculates the significance every time it receives a new sub-set of data, always being consistent with the chosen time window. The significance level found is then compared to a False Positve Rate (FPR), and in case it is above its value, an alert is sent. 6.2.1 False Positive Rates When testing a hypothesis, one of the outcome may be given by a false positive, which is the probability of wrongly reject one correct hypothesis. For a pre- Supernova neutrino detection, this is related to the rate in which a Supernova may occur within the Milky Way. To introduce this factor in the statistical evaluation, a cut-o↵in the p-value is introduced. For example, if it is expected one Supernova event per year, the false positive rate (FPR) would be given by: 1 3 FPR = =2.74 10 (6.1) 365 ⇥ This would lead to a minimum significance level of 2.78 when applied to equation 5.2. Pre-Supernova Alarm 182 However, the literature usually considers this rate if Supernova in our galaxy to be one per century, which gives: 1 1 5 FPR = =2.74 10 , (6.2) 100 365 ⇥ ✓ ◆ which leads to a minimum significance level of 4.03. The choice of the false positive rate a very important for the alarm e ciency and also guarantees a high confidence level. Based on [116], the rate of core-collapse Supernova given is 3.2 per century, in order to account all the Supernovae events observed over the last millennium, which is a very reasonable rate to be used to apply the FPR rate in the alarm system. The FPR in this case is: 3.2 1 5 FPR = =8.78 10 , (6.3) 100 365 ⇥ ✓ ◆ which leads to a minimum significance level of 3.75. 6.3 System design When initiated, the system gives the options to change the values of background rate (default: 0.306 event/hour) and time window (default: 7 hours) for the sta- tistical evaluation. Figure 6.2 shows the options given to the user after the system Pre-Supernova Alarm 183 Figure 6.2: Screenshot of the main menu of the alarm system after initializa- tion. Figure 6.3: Screenshot of the alarm view when data is being processed. In- formation regarding last evaluations are also showed in the screen. initialization. If the user’s choice is to run the system, it will first create a list of all the available data. A “to do” queue is created for the data that have not been processed yet, which includes new files in the data directory and also files that were modified. Files in the queue will go through the data processing. Figure 6.3 shows a screenshot of when the system is processing data. The Pre-Supernova Alarm 184 events passing the selection will have their information saved: absolute time in which these events happened, and information regarding di↵erent variables, such as the coincidence time and distance. When the queue is empty, the system will look again for new data. If they are not available yet, it keeps checking for new or altered sub-sets. Every time a new sub-set of data is processed, the system calls a function to perform the statistical evaluation, which accounts the number of events that were selected in the last 7 hours. If the significance level is above the False Positive Rate, it sends an email alerting the possibility of a Supernova. The estimated early warning lead time, which is the interval between the time when an event really occurs and it is triggered by the WIT system and the time the alert system makes an alarm decision, is summarized in Table 6.1.Atthetimeof writing, the WIT system, where the alarm is installed, is currently being upgraded with new machines, which will reduce the lead time in about 10 minutes. ⇠ Pre-Supernova Alarm 185 Figure 6.4: Main web page of the alarm system used to verify latest evalua- tions and the alarm status. Pre-Supernova Alarm 186 Table 6.1: Estimated early warning lead time of the pre-Supernova alert sys- tem. Process Estimated Time Data Acquirement (WIT system) 15 minutes Data Reconstruction 50 seconds Data Processing 40 seconds Statistical Evaluation/Alarm Decision 10 seconds Total 16 minutes ⇠ The results of the evaluation and other information of the alert system is printed to a web page. Figure 6.4 shows the web page available for the user to follow the alarm status. The system also monitors the rates of IBD pairs and the evolution of the significance levels found in the last hours. They can be checked at the main web page and also in more details in a secondary page. Figures 6.5 and 6.6 show the plots for random values used for testing. They can be used for monitoring random fluctuations, increments in IBD pairs and significances, which can become potential alerts. Pre-Supernova Alarm 187 Figure 6.5: Screenshot of the secondary web page showing plots for the evo- lution of evaluated significance levels and selected IBD pairs for di↵erent time intervals. Pre-Supernova Alarm 188 Figure 6.6: Screenshot of the secondary web page showing plots of di↵erent variable distributions over the last 12 hours for the events selected by the system. 6.4 Prospects By the time of writing, the gadolinium loading in Super-Kamiokande is still in the commissioning phase, which prevents the update of backgrounds estimations using gadolinium data. After a large sample of SK-Gd data is collected, models Pre-Supernova Alarm 189 of background varying in time will have to be developed. Background rates from Spallation and Reactor neutrinos ideally need to be tracked and reduced online. Simulations also need to be updated when the software SKDETSIM for SK-Gd becomes available, which takes into account current dark rates in the detector and optimized models for TNC in gadolinium for Super-Kamiokande. Another main update to the currently used models is to account the systematic uncertainties. The data files come from the WIT trigger, which has no information yet regarding the uncertainties related to event reconstruction. Also, some of the background sources, like Reactor neutrinos fluxes, don’t account uncertainties for their fluxes calculations. With all of these improvements in the future, the alert system can be moved from quasi-online to completely online. Chapter 7 Prospects for Low Energy detection in Hyper-Kamiokande experiment Hyper-Kamiokande (Hyper-K) is the next generation Water-Cherenkov detector with multi-purpose scientific goals, as the investigation on CP-violation in lep- tonic sector, determination of the neutrino mass ordering, observation of Cosmic Neutrinos and Proton Decay. For its high statistics data within a huge fiducial volume and longevity, Hyper-K has the potential to many discoveries in Neutrino Astronomy. 190 Prospects for Low Energy detection in Hyper-Kamiokande experiment 191 Figure 7.1: Illustrated scheme of the Hyper-K Detector. The detector will be a cylindrical tank with 71 meter height and 68 meters diameter. The total and fiducial masses will be respectively 260 ktons and 187 ktons, which are about ten times larger than Super-Kamiokande. Hyper-K is currently under construction and the experiment will be ready for data taking in 2027. There are proposals for the construction of a second tank in Korea. Accelerator neutrinos measurement from the complex JPARC will be also im- proved with a higher intensity beam. A new water Cherenkov detector of 1 kton at about 1 km from the source, the Intermediate Water Cherenkov Detector (IWCD), will provide flux predictions in Hyper-K and reduce systematic errors related to neutrino cross section uncertainty. Hyper-K will also have enhanced PMTs with improved performances as new Prospects for Low Energy detection in Hyper-Kamiokande experiment 192 geometry, better timing resolution and detector e ciency [117; 118]. The baseline design foresees 40,000 20-inch photomultiplier tubes (PMTs) observing the Inner Detector, resulting in a photo-cathode coverage of 40%, and 10,000 3-inch PMTs in the Outer Detector, which acts as veto for the entering particles. Following the requirements for the Hyper-K detector, alternative designs have been studied in order to improve the experiment physics capabilities. One of them, the multi-PMT (mPMT) concept, is presented next. 7.1 The mPMT concept The mPMT module is an array of small PMTs and their electronics arranged inside a pressure resistant vessel. It was first implemented in the KM3NeT experiment [119], which showed advantages like: Improved segmentation: Single-photon hits can be distinguished from multi- • photon hits analysing the signal in neighboring PMTs. Superior photon counting: Total number of deposited photoelectrons can be • directly obtained from hits on PMTs. Local coincidences: When considering coincidences between individual PMT • inside the modules, suppression of uncorrelated single-hit noise can be im- proved. Prospects for Low Energy detection in Hyper-Kamiokande experiment 193 Directional sensitivity: With the individual direction of each single PMT, • the mPMT gives information on the direction of detected photons, improving signal-to-background separation. Extended dynamic range: When multiple photons arrive at the same time, • it will be more likely to hit di↵erent PMTs. The alternative option for photodetection studied for Hyper-K is the combina- tion of the 20-inch PMTs and the mPMTs modules in the far detector, facing the Inner Detector. For the IWCD it is already decided that the detector will use the mPMT photodetection system. 7.1.1 mPMT Geometry for Hyper-K As illustrated in Figure 7.2, the basic geometry of the mPMT module for Hyper-K is a cylindrical vessel, housing 3-inch PMTs. The PMTs are supported by a 3D printed structure and optically coupled by Silicon Gel. The vessel has an acrylic cover and the cylinder is in high density polyethilen and the base is in stainless steel. The 3-inch PMT used in the modules will be the Hamamatsu R14374. The module has a front-end board integrated inside the vessel, as well as HV suppliers for the PMTs. To each small photodetector is added a reflector cone, in order to increase the e↵ectiveness of the photocathode area. Prospects for Low Energy detection in Hyper-Kamiokande experiment 194 Figure 7.2: Illustration of the mPMT module. Figure 7.3: The two initial prototypes for Hyper-K from INFN (left) and TRIUMF (right). Designs of mPMT modules for the Hyper-K experiment have been in develop- ment, taking into account di↵erent configurations for the PMTs, vessel materials, PMT read-outs, assembling processes, etc. The initial prototypes, which are under test, are: INFN Prototype: based on KM3NeT experiment, a mPMT module with a • 17-inch vessel and 3-inch PMTs inside to test the acrylic vessel and electron- ics. Figure 7.4 shows the assembling of a prototype in INFN Naples. TRIUMF Prototype: optimized design to test of new mechanics and assem- • bling procedure. Prospects for Low Energy detection in Hyper-Kamiokande experiment 195 Assembling of the mPMT prototype in INFN Naples. Figure 7.4: Prospects for Low Energy detection in Hyper-Kamiokande experiment 196 Preliminary simulation studies have already showed improved vertex resolution and increased PID separation compared to 20-inch PMTs [120], which will be shown in Section 7.3. 7.1.2 Acrylic Vessel Km3NeT experience demonstrated that glass vessels are characterized by high 40K and other radioactive contamination. Tests on acrylic samples from several companies have been performed to identify the best solution for Hyper-K. Tests of optical and mechanical performances, absorption, compatibility with optical gel, radioactive contamination, and pressure resistance, showed that the ’PLEXIGLAS GS UV transmitting’ by Evonik is the best solution for the mPMT module. Results of optical transmittance test in di↵erent acrylic materials are in Figure 7.5. A hydro-static pressure test has been realized for 15mm and 20mm- thick spherical vessels to verify the safe constraint to resist to 1.26 MPa. No damage has been registered up to 1.8 MPa for both the vessels, much higher than the Hyper-K constraints. Nuclear tests at the National Gran Sasso Laboratories (LNGS) have been per- formed to investigate radioactive contaminations by measuring gamma emissions, showing radioactivity levels comparable to the ’KURARAY PARAGLAS-UV00’ acrylic currently used in Super-Kamiokande [121]. Prospects for Low Energy detection in Hyper-Kamiokande experiment 197 Figure 7.5: Acrylic samples transmittance 7.2 mPMT Tests at Memphyno At the time of writing, the initial mPMT prototype developed in INFN Naples is under test at the MEMPHYno test bench at the AstroParticule et Cosmologie (APC) in Paris. It consists of a 8m3 filtered water tank with four identical planes of scintillators, two at the top and two at the bottom, to allow the reconstruction of crossing muons. Figure 7.6 shows a picture of the tank. The prototype is being tested in water since October 18th, 2019. The 17-inch mPMT module prototype contains old models of 3-inch PMTs, Hamamatsu R12199, in two di↵erent hemispheres: the ID with 13 PMTs and the OD with 4 PMTs. Prospects for Low Energy detection in Hyper-Kamiokande experiment 198 Figure 7.6: Picture of the MEMPHYno test bench at APC in Paris, where the mPMT prototype developed by INFN Naples is being tested. Following the requirements for Hyper-K, the the electronics for the mPMT mod- ules needs to take into account low power consumption and good charge and time resolution. The prototype has as HV supplier a basic Cockcroft-Walton voltage multiplier circuit, used to generate multiple voltages for the PMT. It raws less than 1.5 mA of supply current at 3.3 V and can provide up to -1500 V cathode voltage. Prospects for Low Energy detection in Hyper-Kamiokande experiment 199 Two di↵erent designs for the mPMT digitization are in development: a Q/T digitization based on discrete components and one based on Flash Analog to Dig- ital Converters (FADC) digitization, with on-board signal processing. The INFN Naples prototype current has the digitizer unit based on Q/T digitization, with a time-to-digital converter (TDC) and a 2-step integrator coupled to a sample-and- hold analog-to-digital converter (ADC), which showed good performance. The achieved timing resolution is 100 ps, the charge resolution is 0.1% and good lin- ⇠ earity of the circuit response to charge was observed. The Q/T digitization based on discrete components will be used for the Hyper-K far detector. The system has very low power consumption, reaching 12.5 mW for the HV circuit and 40.5 mW for the digitizer per-channel, with a total power consumption of 4W per mPMT module [121]. ⇠ Recent updates to the main board firmware have been performed in order to improve measurements and monitoring of the mPMT module. Tests of the mPMT prototype include: directionality, time response, charge resolution, environmental measurements, etc. Dark rates are being carefully eval- uated, as it a↵ects directly the mPMT reconstruction capability, especially at low energies. Figure 7.7 shows results from dark rates of di↵erent PMTs inside the module in respect to threshold. Prospects for Low Energy detection in Hyper-Kamiokande experiment 200 Figure 7.7: Results of dark rate evaluation of di↵erent PMTs in the mPMT module prototype tests at MEMPHYno. Strategies are being studied to improve dark counts such as adding coating material to insulate the PMT tube, reduction of radioactive materials in the PMT glass, and perform software reductions to dark rates, which will be showed next. Prospects for Low Energy detection in Hyper-Kamiokande experiment 201 7.3 Reduction of mPMT dark rates with BDT As showed in section 7.1,theimplementationofmPMTscanbringmanyadvan- tages for neutrino detection.1 Improved vertex resolution will give a higher fiducial volume, improved energy resolution, etc., which will make the low energy neutrino detection more e cient. Preliminary results on vertex resolution already showed improvements with the use of mPMTs. Figure 7.8 shows the improved vertex resolution and the particle identification for the use of 3-inch PMTs compared to 8-inch or 20-inch PMTs. Simulations studies are currently being implemented to evaluate the detector performance. The software system for Hyper-K is being developed to be a reliable and adaptive tool. The simulation flow consists in generate event topologies by Neutrino interaction packages (GENIE [123] and NEUT [124]) and modeling them with a Monte Carlo response code called WCSim. To reconstruct low energy events it is used BONSAI, same as explained in section 3.7.1, or LEAF. For high energy events, it is used fiTQun [125]. The Water Cherenkov Simulator, WCSim, is a GEANT4 [97]basedcodeusedto simulate the geometry and physics response of a water Cherenkov detector [126]. When simulating the Hyper-K detector, WCSim reads the type and number of Prospects for Low Energy detection in Hyper-Kamiokande experiment 202 Figure 7.8: Vertex resolution (top) and electron/muon separation (bottom) improved by the use of 3-inch PMTs (red) in comparison with 20-inch (blue) and 8-inch (black). From [122]. Prospects for Low Energy detection in Hyper-Kamiokande experiment 203 photodetectors being used (20-inch PMTs, mPMTs or both), the detector diam- eter and radius and dark rate information, among other configurations. WCSim outputs both raw hits (which and how many times the PMTs were hit) and dig- itized information (number of hits in a trigger window, time and charge). The events generated by WCSim are then reconstructed. For low energy events it is used BONSAI and, as an alternative option, the Low Energy Algorithm Frame- work (LEAF), which is is a simple low energy fitter [127]. Multivariate selection techniques such as Boosted Decision Tree (BDT) are cur- rently being applied to simulated events for a detector configuration with mPMTs, in order to achieve a software reduction of dark counts. Following the description introduced in Section 4.4, di↵erent samples of background and signal are used in order to create a BDT classificator. Figure 7.9 shows the ROC curves of preliminary trials of BDT trainings. For sig- nal events it was used 1 MeV electrons generated with WCSim and reconstructed with LEAF, while for background events it was used 500 keV electrons, which are much below the expected energy threshold in Hyper-Kamiokande, so they can be approximated as Dark Rate events. This approximation was done due to software limitations and further implications will be studied. Figure 7.10 shows the sig- nal and background separation to the most e cient one, which had as choice of Prospects for Low Energy detection in Hyper-Kamiokande experiment 204 Figure 7.9: ROC curves for di↵erent trials of BDT trainings to be used for the software dark rate reduction of mPMTs. parameters boost =0.8andNTrees =1000.Table7.1 shows the ranked impor- tance of each variable outputted from LEAF for the BDT classification and a brief description of them. Variable Importance Definition mPMT hits 7.76E-02 NumberofhitsinthemPMTs. ID hits 400 7.22E-02 NumberofhitsintheIDwithin400nswindow. mPMT hits 400 6.99E-02 Number of hits in the mPMTs within 400 ns window. mPMT hits 200 6.71E-02 Number of hits in the mPMTs within 200 ns window. ID hits 6.64E-02 NumberofhitsintheID. lf vertex[1] 6.35E-02 Reconstructed vertex, Y. Prospects for Low Energy detection in Hyper-Kamiokande experiment 205 lf vertex[0] 6.30E-02 Reconstructed vertex, X. lf vertex[2] 6.22E-02 Reconstructedvertex,Z. ID hits 200 6.22E-02 NumberofhitsintheIDwithin200nswindow. lf NLL 6.13E-02 NegativeLogLikelihoodforbestvertexcandidate. rawhit num 6.09E-02 Collection of hits, including dark noise. lf vertex[3] 5.69E-02 Reconstructedvertex,T. mPMT hits 50 5.39E-02 Number of hits in th emPMTs within 50 ns window. ID hits 50 5.04E-02 NumberofhitsintheIDwithin50nswindow. lf intime 4.66E-02 NumberofhitswithinatimewindowusedbyLEAF. digithit num 4.24E-02 Numberofdigitizedhits. lf ctime 2.36E-02 CPUtimeforeventreconstruction. Table 7.1: Variables in order of importance of the BDT for the software Dark Rate reduction in Hyper-Kamiokande. The results from figures 7.9 and 7.10 demonstrate the possibility of reducing the dark rate of mPMTs even further. At the time of writing, studies are still being performed to get the most e cient BDT and cut on the BDT score for the software reduction. Other multivariate methods will also be tested. Prospects for Low Energy detection in Hyper-Kamiokande experiment 206 Figure 7.10: Signal and background separation for the most e cient BDT of Figure 7.9. 7.4 Detection of Supernova and Pre-Supernova Neutrinos in Hyper-K Hyper-K has several advantages compared to the current or planned experiments, starting from its large volume, which provides high statistics for neutrino events. Enhanced energy and direction sensitivities, as well as low detection threshold energy and longevity will allow a comprehensive study of Supernova neutrinos. The design of the DAQ and bu↵ering system are being developed to accept a broad range of rates for galactic Supernovae within 10 kpc. Hyper-K can observe Prospects for Low Energy detection in Hyper-Kamiokande experiment 207 Figure 7.11: Predicted Inverse Beta Decay reactions due to Supernova Relic Neutrinos in function of years for current and planned experiments, from [128]. 50,000 to 75,000 inverse beta decay events, 3,400 to 3,600 ⌫e-scattering events, 80 16 16 to 7,900 ⌫e + O CC events, and 660 to 5,900⌫ ¯e + O CC events for a 10 kpc Supernova, a total of 54,000 to 90,000 events. The study of Supernova Relic Neutrinos could also be improved with Hyper-K. The first observation of SRN is expected in SK-Gd, in an energy range of 10-20 MeV. In [128], a few estimations for SRN detection in Hyper-K was performed, in which the experiment could measure these neutrinos with energies of 16-30 MeV. It is expected the selecting of 70 SRN events after 10 years of observations in Hyper-K. The expected number of IBD reactions due to SRN are shown in Figure 7.11 in several experiments along the years. These are very preliminary results Prospects for Low Energy detection in Hyper-Kamiokande experiment 208 and a full Monte Carlo analysis for SRN predictions in Hyper-K is still to be performed. If Hyper-K will be gadolinium-doped, the lower energy end for the detection of SRN could be lower as 10 MeV, which is very important to explore the history of Supernova bursts back to the epoch of red shift (z) 1. In addition, a Gd- ⇠ doped Hyper-K will also be able to detect pre-Supernova neutrinos, with great sensitivity due to a much bigger FV. Also a powerful pre-Supernova alarm could be developed, as many more events are expected, increasing early warnings and ranges of detection compared to SK-Gd. Assuming the same backgrounds for SK-Gd, simple estimations of a pre-Supernova alarm for Hyper-K were evaluated, adjusting the target mass. Results for early warning and range of detection for a 0.02% Gd (SO ) 8H O loading in Hyper-K are shown in Tables 7.2 and 7.3. 2 4 3 · 2 Prospects for Low Energy detection in Hyper-Kamiokande experiment 209 Table 7.2: Expected maximum range of detection of pre-Supernova stars in Hyper-K loaded with 0.02% Gd (SO ) 8H O, for normal ordering neutrino 2 4 3 · 2 mass models. Maximum Range for Model Mass (M ) > 3 FPR = 3.2/cy FPR = 1/cy > 5 (pc) (pc) (pc) (pc) Odrzywolek 15 850 750 710 630 Odrzywolek 25 1120 980 940 830 Patton 15 990 870 840 740 Patton 30 1560 1370 1310 1150 Table 7.3: Expected early warning in Hyper-K loaded with 0.02% Gd (SO ) 2 4 3 · 8H2O, for Betelgeuse-like models with normal ordering neutrino mass models. Warning (hours) for Model Mass/Distance > 3 FPR = 3.2/cy FPR = 1/cy Odrzywolek 15M /150pc 11.0 10.0 9.2 Odrzywolek 25M /220pc 10.0 9.3 8.2 Patton 15M /150pc 13.0 12.1 12.0 By the time of writing, it is still unclear if Hyper-K will be gadolinium doped someday. Although many studies are yet to be performed for its pre-Supernova neutrino sensitivity, these preliminary results shows a very great potential for a pre-Supernova alarm in Hyper-K. Chapter 8 Conclusion (English) The new phase of Super-Kamiokande detector, SK-Gd, is improving the capabil- ities of the experiment to identify neutrons, which enable the reduction of the energy threshold for the detection of neutrinos. In the context of this work, the never-before-detected⌫ ¯e from pre-Supernova stars is now possible to be observed and its detection can provide an early warning for potential core collapse Super- novae. Previous work [16]evaluatedthesensitivityforthedetectionofpre-Supernova neutrinos in Super-Kamiokande with 0.2% Gd (SO ) 8H O by mass, prior to the 2 4 3 · 2 loading of a lower concentration the compound in the detector in 2020. Since then, new requirements for radioisotope contamination of gadolinium sulfate were estab- lished, as well as updated information regarding reactor fluxes, which are two of 210 Conclusion (English) 211 the main backgrounds for the pre-Supernova neutrino detection. All backgrounds sources need to be carefully evaluated as the detection of pre-Supernova neutrinos is below usual energy thresholds of Super-Kamiokande. At the time of writing, the first stage of SK-Gd is still in the commissioning phase and after acquiring a good amount of data, backgrounds will be better determined and evaluated, as well as improved techniques for their reduction. ApossibleapplicationofthisstudyisthedevelopmentofaSupernovaalarm for Super-Kamiokande with Gadolinium. The system was designed and optimized for the first phase of SK-Gd with 0.02% Gd (SO ) 8H O by mass, looking for 2 4 3 · 2 Inverse Beta Decay candidates and provide warnings in case of increasement in event rate. The pre-Supernova alert system can give early warnings of 2.4 to 4.2 hours assuming a false positive rate of 3.2 per century for ↵ Ori, which is the nearest red giant star to Earth. For detection above 3 , the early warnings is 3.2 to 5.2. Maximum detection range for the alarm is 400 pc. Next generation experiment, Hyper-Kamiokande, has a great potential for pre- Supernova detection if it will be gadolinium doped. Detection ranges are estimated to be greater than 1 kpc for a false positive rate of 3.2 per century and an alert system could give warnings of over 10 hours for ↵ Ori. Alternative designs for photodetector using a hybrid combination of 20-inch PMTs and mPMT modules Conclusion (English) 212 are being studied for Hyper-K. This configuration could improve low energy recon- struction capabilities for its enhanced vertex and energy resolution, among other benefits. Chapter 9 Conclusioni (Italian) L’attuale fase sperimentale del rivelatore Super-Kamiokande SK-Gd permette di migliorare la capacit`adell’esperimento nell’identificazione di neutroni, riducendo la soglia di energia per la rivelazione dei neutrini. E’ cos`ıora possibile osservare neutrini⌫ ¯e da pre-Supernova, mai rilevati prima, e realizzare un sistema di allarme per i collassi di Supernovae. In [16]`estatavalutatalasensibilit`aperlarivelazionedineutrinidapre- Supernova in Super-Kamiokande con una concentrazione percentuale in massa dello 0,2% Gd (SO ) 8H O. 2 4 3 · 2 I risultati riportati in questo lavoro di tesi sono basati su una dettagliata stima della contaminazione da radioisotopi del solfato di gadolinio e una aggiornata stima 213 Conclusioni (Italian) 214 dei flussi di neutrini prodotti dai reattori nelle vicinanze del rivelatore, che sono due delle principali sorgenti di contaminazione nella rivelazione dei neutrini pre- Supernova. Tutte le altre sorgenti di background sono state attentamente valutate. Questo lavoro `epresentato mentre la prima fase sperimentale di SK-Gd `ein fase di commissioning. Gli eventi di background saranno meglio determinati e valutati non appena un campione statisticamente significativo dei dati sperimentali sar`a disponibile, cos`ıcome saranno migliorate le strategie per la loro riduzione. I risultati di questo studio costituiscono la base per lo sviluppo di un sistema di allarme per la rivelazione multimessanger di Supernova. Il sistema di allarme sviluppato in questo lavoro di tesi `estato progettato e ottimizzato per individuare gli eventi di decadimento beta inverso di neutrino nella prima fase di SK-Gd, con concentrazione percentuale in massa di 0,02% Gd (SO ) 8H O.Ilsistema 2 4 3 · 2 proposto per la rivelazione di neutrini da pre-Supernova pu`ofornire un’allerta (early warning) stimata tra le 2,4 e le 4,2 ore precedenti l’esplosione di Supernova, assumendo un FPR (False Positive Rate) di 3,2 per secolo per la sorgente ↵ Ori (Betelgeuse), che `ela stella gigante rossa pi`uvicina alla Terra. Per sensitivit`a maggiori di 3 ,l’allerta`etrale3,2ele5,2oreprecedentiilcollasso,entrouna distanza massima `edi 400 pc. Il futuro esperimento Hyper-Kamiokande, attualmente in fase di costruzione, Conclusioni (Italian) 215 potr`amigliorare ulteriormente la sensibilit`anella ricostruzione di neutrini da pre- Supernova se sar`adrogato con gadolinio. Si stima che si possano rivelare eventi di Supernova a distanze maggiori di 1 kpc. Considerando un tasso di falsi positivi di 3,2 per secolo, un sistema di allerta potrebbe fornire l’allarme con oltre 10 ore di anticipo per il collasso di ↵-Ori. Per Hyper-K sono allo studio progetti alternativi che propongono una combinazione ibrida di PMT da 20 pollici e moduli mPMT. Questa configurazione potrebbe migliorare le capacit`adi ricostruzione di eventi di neutrino di bassa energia. Chapter 10 Conclus˜ao (Portuguese) AnovafasedodetectorSuper-Kamiokande,SK-Gd,est´aaprimorandoascapaci- dades do experimento de identificar nˆeutrons, o que possibilita a redu¸c˜aodo limiar de energia para a detec¸c˜ao de neutrinos. No contexto deste trabalho,⌫ ¯e de estrelas pr´e-Supernova, que nunca foram detectados, s˜aoposs´ıveis de serem observados e sua detec¸c˜aopode fornecer um alerta precoce para um potencial colapso do n´ucleo de Supernovas. Em um trabalho anterior [16]foiavaliadaasensibilidadeparaadetec¸c˜aode neutrinos pr´e-Supernova no Super-Kamiokande com 0,2% Gd (SO ) 8H O.Em 2 4 3 · 2 2020, uma concentra¸c˜aomais baixa do composto foi dilu´ıdano detector. Desde ent˜ao,novos requisitos para contamina¸c˜aopor radiois´otopos do sulfato de gadol´ınio foram estabelecidos, bem como informa¸c˜oesatualizadas sobre fluxos de reatores, 216 Conclus˜ao(Portuguese) 217 que juntos s˜aodois dos principais backgrounds para a detec¸c˜aode neutrinos pr´e- Supernova. Todas as fontes de background precisam ser avaliadas cuidadosamente, pois a detec¸c˜ao de neutrinos pr´e-Supernova est´aabaixo dos limites de energia usuais de Super-Kamiokande. No momento da reda¸c˜ao desta tese, o primeiro est´agio do SK-Gd ainda est´aem fase de comissionamento e, ap´osa aquisi¸c˜ao de uma boa quantidade de dados, os backgrounds ser˜aomelhor determinados e avaliados, bem como t´ecnicas aprimoradas para sua redu¸c˜ao. Uma poss´ıvel aplica¸c˜ao deste estudo ´eo desenvolvimento de um alarme para Supernovas para o Super-Kamiokande com Gadol´ınio. O sistema de alerta foi projetado e otimizado para a primeira fase do SK-Gd com 0,02% Gd (SO ) 8H O, 2 4 3· 2 procurando candidatos de Decaimento Beta Inverso e fornecendo avisos em caso de um aumento na taxa de eventos esperada. O sistema de alerta de pr´e-Supernova pode dar avisos antecipados de 2,4 a 4,2 horas, assumindo uma taxa de falsos positivos de 3,2 por s´eculo para ↵ Ori, que ´ea estrela gigante vermelha mais pr´oximada Terra. Para detec¸c˜oes acima de 3 ,osavisoss˜aode3,2a5,2.O intervalo m´aximode detec¸c˜aodo alarme ´e400 pc. O futuro experimento, Hyper-Kamiokande, tem um grande potencial para de- tec¸c˜aopr´e-Supernova se a ´aguado detector tamb´em for for dopada com gadol´ınio. Os intervalos de detec¸c˜ao s˜ao estimados em mais de 1 kpc, para uma taxa de fal- sos positivos de 3,2 por s´eculo, e um sistema de alerta poderia dar avisos de mais Appendices 218 de 10 horas para ↵ Ori. Projetos alternativos para fotodetectores usando uma combina¸c˜ao h´ıbrida de PMTs de 20 polegadas e m´odulos mPMT est˜ao sendo es- tudados para Hyper-K. Esta configura¸c˜ao h´ıbrida pode aprimorar as capacidades de reconstru¸c˜ao de eventos de baixa energia, devido `as melhorias nas resolu¸co˜oes de v´ertice e de energia, entre outros benef´ıcios. Appendix A Nearby pre-Supernova Candidates Catalog Name Common Name Constellation Distance (kpc) Mass (M ) +1.15 HD 116658 Spica/↵ Virginis Virgo 0.077 0.004 11.43 1.15 ± HD 149757 ⇣ Ophiuchi Ophiuchus 0.112 0.002 20.0 ± +1.0 HD 129056 ↵ Lupi Lupus 0.143 0.003 10.1 1.0 ± +1.5 HD 78647 Velorum Vela 0.167 0.003 7.0 1.0 ± HD 148478 Antares/↵ Scorpii Scorpius 0.169 0.030 11.0 14.3 ± +0.8 HD 206778 ✏ Pegasi Pegasus 0.211 0.006 11.7 0.8 ± +5.0 HD 39801 Betelgeuse/↵ Orionis Orion 0.222 0.040 11.6 3.9 ± +0.6 HD 89388 q Car/V337 Car Carina 0.230 0.020 6.9 0.6 ± 219 Nearby pre-Supernova Candidates 220 +0.1 HD 210745 ⇣ Cephei Cepheus 0.256 0.006 10.1 0.1 ± +3.0 HD 34085 Rigel/ Orion Orion 0.264 0.024 21.0 3.0 ± HD 200905 xi Cygni Cygnus 0.278 0.029 8.0 ± HD47839 SMonocerotisA Monoceros 0.282 0.040 29.1 ± HD47839 SMonocerotisB Monoceros 0.282 0.040 21.3 ± +0.1 HD 93070 w Car/V520 Car Carina 0.294 0.023 7.9 0.1 ± HD 68553 NS Puppis Puppis 0.321 0.032 9.7 ± +2.00 HD36389 CETauri/119Tauri Taurus 0.326 0.070 14.37 2.77 ± 2 +0.6 HD 68273 Velorum Vela 0.342 0.035 9.0 0.6 ± 1 +2.0 HD 50877 ø Canis Majoris Canis Major 0.394 0.052 7.83 2.0 ± +0.7 HD207089 12Pegasi Pegasus 0.415 0.031 6.3 0.7 ± +0.18 HD213310 5Lacertae Lacerta 0.505 0.046 5.11 0.18 ± +0.1 HD 52877 Canis Majoris Canis Major 0.513 0.108 12.3 0.1 ± +1.0 HD 208816 VV Cephei Cepheus 0.599 0.083 10.6 1.0 ± +3.0 HD 196725 ✓ Delphini Delphinus 0.629 0.029 5.60 3.0 ± HD 203338 V381 Cephei Cepheus 0.631 0.086 12.0 ± +1.0 HD 216946 V424 Lacertae Lacerta 0.634 0.075 6.8 1.0 ± +0.5 HD 17958 HR 861 Cassiopeia 0.639 0.039 9.2 0.5 ± +0.2 HD 80108 HR 3692 Vela 0.650 0.061 12.1 0.2 ± +0.5 HD56577 145CanisMajor CanisMajor 0.697 0.078 7.8 0.5 ± +0.5 HD219978 V809Cassiopeia Cassiopeia 0.730 0.074 8.3 0.5 ± Nearby pre-Supernova Candidates 221 +0.7 HD 205349 HR 8248 Cygnus 0.746 0.039 6.3 0.7 ± +4.0 HD 102098 Deneb/↵ Cygni Cygnus 0.802 0.066 19.0 4.0 ± Table A.1: List of pre-Supernova candidates with estimated masses and dis- tances. Reproduced from [53]. Bibliography [1] Pauli’s letter of the 4th of December 1930, [Web: https://cds.cern.ch/record/83282/] [2] E. Fermi, An attempt of a theory of beta radiation. 1.,Z.Phys.88,1934, [DOI: 10.1007/BF01351864] [3] C. L. Cowan Jr., F. Reines, F. B. Harrison, H. W. Kruse, A. D. McGuire, Detection of the Free Neutrino: a Confirmation,ScienceVol.124,Issue3212, pp. 103-104, 1956, [DOI: 10.1126/science.124.3212.103]. [4] P.A. Zyla et al. (Particle Data Group), The Review of Particle Physics ,Prog. Theor. Exp. Phys. 2020, 083C01 (2020). [5] S.Fukuda, et al. (Super-Kamiokande Collaboration), The Super- Kamiokande detector, Nucl.Instrum.Meth.A 501 (2003) 418-462, [DOI:10.1016/S0168-9002(03)00425-X]. 222 BIBLIOGRAPHY 223 [6] K. Abe, et al. (The T2K Collaboration), The T2K experiment, Nucl.Instrum.Meth.A 659 (2011) 106-135, [DOI:10.1016/j.nima.2011.06.067]. [7] J. Boger, et al. (The SNO Collaboration), The Sudbury Neu- trino Observatory, Nucl.Instrum.Meth.A 449 (2000) 172-207, [DOI:10.1016/S0168-9002(99)01469-2]. [8] M. Walter (for the IceCube Collaboration), The IceCube Neutrino Observa- tory - Design and Performance, Nucl.Phys.B Proc.Suppl. 172 (2007) 13-16, [DOI:10.1016/j.nuclphysbps.2007.07.031]. [9] A.D. Sakharov, Violation of CP Invariance, C asymmetry, and baryon asym- metry of the universe, Pisma Zh. Eksp. Teor. Fiz. 5 (1967) 32-35, JETP Lett. 5(1967)24-27,[DOI: 10.1070/PU1991v034n05ABEH002497]. [10] M. Fukugita, Baryogenesis Without Grand Unification,Phys.Lett.B174 (1986), [DOI: 10.1016/0370-2693(86)91126-3]. [11] R. Gonzalez Felipe, Neutrinos and the matter-antimatter asymmetry in the Universe, International Journal of Modern Physics EVol. 20, No. supp01, pp. 56-64 (2011), [DOI: 10.1142/S0218301311040074]. BIBLIOGRAPHY 224 [12] Paolo Lipari, Introduction to Neutrino Physics,Contributionto: 1st CERN-CLAF School of High-Energy Physics, 115-199, 2003, [DOI:10.5170/CERN-2003-003.115]. [13] X. Qian, P. Vogel, Neutrino Mass Hierarchy, PPNP, 83, 1-30 (2015), [DOI: 10.1016/j.ppnp.2015.05.002]. [14] H. Nunokawa, S. Parke, J. W. F. Valle, CP Violation and Neu- trino Oscillations, Prog.Part.Nucl.Phys. 60 (2008) 338-402, [DOI: 10.1016/j.ppnp.2007.10.001]. [15] A. Yu. Smirnov, The MSW e↵ect and Solar Neutrinos,Invitedtalkgivenat the 11th workshop on Neutrino Telescopes, Venice, March 11-14, 2003 [16] C. Simpson, Pre-supernova Neutrinos at Super-Kamiokande with Gadolinium,[Web https://ora.ox.ac.uk/objects/ uuid:c90b5754-e52e-4d84-a247-b9eb27580a55]. [17] Takaaki Ka jita, Atmospheric Neutrinos, Advances in High En- ergy Physics, vol. 2012, Article ID 504715, 24 pages, 2012. https://doi.org/10.1155/2012/504715 [18] T. Kajita, Atmospheric Neutrinos, Advances in High Energy Physics, Volume 2012, Article ID 504715, [DOI: 10.1155/2012/504715]. BIBLIOGRAPHY 225 [19] Y. Fukuda, et al. (The Super-Kamiokande Collaboration), Evidence for os- cillation of atmospheric neutrinos,Phys.Rev.Lett.81:1562-1567,1998,[DOI: 10.1103/PhysRevLett.81.1562]. [20] V. Takhistov, Review of Atmospheric Neutrino Results from Super- Kamiokande, proceedings of 40th International Conference on High Energy Physics (ICHEP-2020), [e-Print:2012.06864] [21] P. A. Symes, MINOS: Design and detectors, Prog.Part.Nucl.Phys. 57 (2006) 144-146, [DOI:10.1016/j.ppnp.2005.11.001]. [22] J. N. Bahcall, Neutrino Astrophysics, CAMBRIDGE, UK: UNIV. PR. (1989) 567p, 1989, [ISBN: 9780521379755] [23] The Borexino Collaboration, Comprehensive measurement of pp-chain solar neutrinos, Nature, volume 562, pages 505–510 (2018), [DOI: 10.1038/s41586-018-0624-y]. [24] M. B. Smy, Implications for Solar Neutrino Oscillations from Super- Kamiokande and SNO Data, Proceedings, NO-VE International Workshop on Neutrino Oscillations in Venice, 35-42, [e-Print:hep-ex/0108053] [25] A. W. P. Poon, Neutrino Observations from the Sudbury Neutrino Observa- tory, AIP Conf. Proc. 610:218-230, 2002, [DOI: 10.1063/1.1469931]. BIBLIOGRAPHY 226 [26] F. Boehm, et al. Experimental study of neutrino oscillations at a fission re- actor,Phys.Lett.B97,1980,[DOI:10.1016/0370-2693(80)90608-5]. [27] J. F. Cavaignac, et al. Indication for Neutrino Oscillation From a High Statistics Experiment at the Bugey Reactor,Phys.Lett.B148,1984, [DOI:10.1016/0370-2693(84)90109-6]. [28] M. Abbes, et al. The Bugey-3 neutrino detector, Nucl. Instrum. Meth. A 374, 1996, [DOI:10.1016/0168-9002(96)00220-3]. [29] F. Boehm, et al. The Palo Verde reactor neutrino oscilla- tion experiment, Nucl.Phys.B Proc.Suppl. 77 (1999) 166-170, [DOI:10.1016/S0920-5632(99)00413-2]. [30] M. Apollonio, et al. (Chooz Collaboration), Initial results from the CHOOZ long baseline reactor neutrino oscillation experiment,Phys.Lett.B420,1998, [DOI:10.1016/S0370-2693(97)01476-7]. [31] K. V. Tsang, et al. Results from the Daya Bay Reactor Neu- trino Experiment, Nucl.Phys.B Proc.Suppl. 246-247 (2014) 18-22, [DOI:10.1016/j.nuclphysbps.2013.10.059]. [32] A. Cabrera (for the Double Chooz collaboration), The Double Chooz Experiment, Nucl.Phys.B Proc.Suppl. 229-232 (2012) 87-91, [DOI:10.1016/j.nuclphysbps.2012.09.014]. BIBLIOGRAPHY 227 [33] Y. Oh (for the RENO collaboration), Current status of RENO, Nucl.Phys.B Proc.Suppl. 188 (2009) 109-111, [DOI:10.1016/j.nuclphysbps.2009.02.024]. [34] C. Cerna (on behalf of the JUNO collaboration), The Jiangmen Underground Neutrino Observatory (JUNO), Nucl.Instrum.Meth.A 958 (2020) 162183, [DOI:10.1016/j.nima.2019.05.024]. [35] Liang-Jian Wen, Jun Cao, Yi-Fang Wang, Reactor Neutrino Experiments: Present and Future, Ann.Rev.Nucl.Part.Sci. 67 (2017) 183-211. [36] L. Camilleri, MicroBooNE, Nucl.Phys.B Proc.Suppl. 237-238 (2013) 181-183, [DOI:10.1016/j.nuclphysbps.2013.04.084]. [37] L. Fields, Recent Results from MINERvA, J.Phys.Conf.Ser. 888 (2017) 1, 012015, [DOI:10.1088/1742-6596/888/1/012015]. [38] R. B. Patterson, The NOvA experiment: status and out- look, Nucl.Phys.B Proc.Suppl. 235-236 (2013) 151-157, [DOI:10.1016/j.nuclphysbps.2013.04.005]. [39] The T2K Collaboration, Constraint on the Matter-Antimatter Symmetry- Violating Phase in Neutrino Oscillations, Nature 580, 339–344 (2020), [DOI: 10.1038/s41586-020-2177-0]. BIBLIOGRAPHY 228 [40] M. Koshiba, Observation of Neutrinos from Supernova 1987A by Kamiokande-II,EuropeanSouthernObservatoryConferenceandWorkshop Proceedings, 1987. [41] Ofelia Pisanti, Astrophysical neutrinos: theory, Journal of Physics: Conference Series, Volume 1263, Number 1 (2019), [DOI: 10.1088/1742-6596/1263/1/012004]. [42] M. G. Aartsen, et al. (IceCube Collaboration), Observation of High-Energy Astrophysical Neutrinos in Three Years of IceCube Data,Phys.Rev.Lett. 113, 101101 (2014), [DOI: 10.1103/PhysRevLett.113.101101]. [43] G. Fogli, E. Lisi, A. Mirizzi, D. Montanino, Probing supernova shock waves and neutrino flavor transitions in next-generation water Cerenkov detec- tors, Journal of Cosmology and Astroparticle Physics. 2005. 10.1088/1475- 7516/2005/04/002. [44] AOralbaevetal,The inverse beta decay: a study of cross section, 2016 J. Phys.: Conf. Ser. 675 012003. [45] U. Raha, F. Myhrer, K.Kubodera, Radiative corrections to anti-neutrino pro- ton scattering at low energies,[DOI: 10.1103/PhysRevC.85.045502]. [46] A. N. Ivanov, R. Hollwieser, N. I. Troitskaya, M. Wellenzohn, O. M. Zherebtsov, A. P. Serebrov, Deficit of reactor antineutrinos at distances BIBLIOGRAPHY 229 smaller than 100 m and inverse beta-decay,Phys.Rev.C88,055501,2013, [DOI: 10.1103/PhysRevC.88.055501]. [47] S. M. Adams, C. S. Kochanek, J. F. Beacom, M. R. Vagins, K. Z. Stanek, Observing the Next Galactic Supernova, Astrophys.J. 778 (2013) 164, [DOI: 10.1088/0004-637X/778/2/164]. [48] C. Giunti, C. W. Kim, Fundamentals of Neutrino Physics and Astrophysics, Oxford, UK: Univ. Pr. (2007), [ISBN: 9780198508717]. [49] C. Volpe, Neutrino Astrophysics, Lectures presented at ”52th Winter School of Theoretical Physics”, Ladek Zdroj, (2016). [50] K. Rozwadowska, F. Vissani, E. Cappellaro, On the rate of core collapse supernovae in the Milky Way, New Astronomy, Volume 83, February 2021, 101498, [DOI:10.1016/j.newast.2020.101498]. [51] K. M. Patton, C. Lunardini, R. J. Farmer, Presupernova neutrinos: re- alistic emissivities from stellar evolution, Astrophys.J. 840 (2017) 1, 2, [e-Print:1511.02820] [52] H.-Th. Janka, Neutrino Emission from Supernovae, In: Alsabti A., Murdin P. (eds) Handbook of Supernovae. Springer, Cham, [DOI: 10.1007/978-3-319-21846-5-4]. BIBLIOGRAPHY 230 [53] M. Mukhopadhyay, C. Lunardini, F. X. Timmes, K. Zuber, Presupernova neutrinos: directional sensitivity and prospects for progenitor identification, Astrophys.J. 899 (2020) 2, [DOI: 10.3847/1538-4357/ab99a6] [54] E. A. Magnier, S. Prins, J. van Paradijs, W. H. G. Lewin, R. Supper, G. Hasinger, New supernova remnant candidates in M 31., Astronomy and As- trophysics Supplement, v.114, p.215. [55] T. G. Pannuti, N. Duric, C. K. Lacey, A. M. N. Ferguson, M. A. Magnor, C. Mendelowitz, An X-Ray, Optical, and Radio Search for Supernova Rem- nants in the Nearby Sculptor Group Sd Galaxy NGC 7793, The Astrophysical Journal, 565:966-981, 2002 February 1, [DOI: 10.1086/337918]. [56] E. Sonbas, A. Akyuz, S. Balman, M. E. Ozel, A Search For Supernova Rem- nants in The Nearby Spiral Galaxy M74 (NGC 628), Astron.Astrophys. 517 (2010) A91, [DOI: 10.1051/0004-6361/200913858]. [57] I. Leonidaki, P. Boumis, A. Zezas, A multiwavelength study of supernova rem- nants in six nearby galaxies - II. New optically selected supernova remnants, Mon.Not.Roy.Astron.Soc. 429 (2013) 189, [DOI: 10.1093/mnras/sts324]. [58] J. Lee, M. Lee, Properties of Optically Selected Supernova Remnant Candi- dates in M33, 2014 ApJ 793 134, [DOI: 10.1088/0004-637X/793/2/134] BIBLIOGRAPHY 231 [59] S. Horiuchi, J. P. Kneller, What can be learned from a future supernova neu- trino detection?, Journal of Physics G: Nuclear and Particle Physics, Volume 45, Number 4, 2018, [DOI: 10.1088/1361-6471/aaa90a]. [60] A. Odrzywolek, M. Misiaszek, M. Kutschera, Neutrinos from pre-supernova star, Acta Phys.Polon. B35 (2004) 1981 [61] K.S. Hirata, et al., Observation in the Kamiokande-II Detector of the Neu- trino Burst from Supernova SN 1987a,Phys.Rev.D38(1988)448-458,[DOI: 10.1103/PhysRevD.38.448] [62] T. Haine (for the IMB Collaboration), Neutrinos from SN1987a in the IMB detector, Nucl.Instrum.Meth.A 264 (1988) 28-31, [DOI: 10.1016/0168-9002(88)91097-2] [63] E. N. Alexeyev, L. N. Alexeyeva, I. V. Krivosheina, V. I. Volchenko, Detection of the neutrino signal from SN 1987A in the LMC using the INR Baksan underground scintillation telescope,Phys.Lett.B205(1988)209-214,[DOI: 10.1016/0370-2693(88)91651-6] [64] M. Aglietta et al., On the Event Observed in the Mont Blanc Underground Neutrino Observatory during the Occurrence of Supernova 1987a,1987,EPL (Europhysics Letters), Volume 3, Number 12. BIBLIOGRAPHY 232 [65] J. F. Beacom, The Di↵use Supernova Neutrino Background, Annual Review of Nuclear and Particle Science, Volume 60 (2010), [DOI: 10.1146/annurev.nucl.010909.083331] [66] F. Vissani, Comparative analysis of SN1987A antineutrino fluence, J.Phys.G 42 (2015) 013001, [DOI: 10.1088/0954-3899/42/1/013001]. [67] C. Xu, Super-Kamiokande Collaboration, Current status of SK-Gd project and EGADS, 2016, J. Phys.: Conf. Ser. 718 062070. [68] D. Arnett, Supernovae and Nucleosynthesis: An Investigation of the History of Matter, from the Big Bang to the Present, Princeton Series in Astrophysics, 1996, [ISBN: 9780691011479]. [69] A. Odrzywolek, Marcin Misiaszek, M. Kutschera, Detection possibility of the pair-annihilation neutrinos from the neutrino-cooled pre-supernova star, As- tropart.Phys. 21 (2004) 303-313. [70] X. Shi, G. M. Fuller, Neutrinos and Supermassive Stars: Prospects for Neutrino Emission and Detection, Astrophys. J. 503 (1998) 307, [DOI: 10.1086/305992]. [71] A. Odrzywolek, A. Heger, Neutrino signatures of dying massive stars: From main sequence to the neutron star, Acta Phys.Polon.B 41 (2010) 1611-1628. BIBLIOGRAPHY 233 [72] K. M. Patton, et al. Neutrinos from beta processes in a presupernova: prob- ing the isotopic evolution of a massive star, Astrophys.J. 851 (2017) 1, 6, [DOI:10.3847/1538-4357/aa95c4]. [73] A. Odrzywolek, Pre Supernova Neutrino Spectrum,Web, [http://web.archive.org/web/20181109002141/http: //th.if.uj.edu.pl/ odrzywolek/psns/]. [74] B. Paxton, L. Bildsten, A. Dotter, F. Herwig, P. Lesa↵re and F. Timmes, MODULES FOR EXPERIMENTS IN STELLAR ASTRO- PHYSICS (MESA), The Astrophysical Journal Supplement Series, Volume 192, Number 1, [DOI: 10.1088/0067-0049/192/1/3]. [75] P. A. Cerenkov,ˇ Visible Radiation Produced by Electrons Moving in a Medium with Velocities Exceeding that of Light,1937,Phys.Rev.52,378,[DOI: 10.1103/PhysRev.52.378] [76] C. Bronner, Y. Nishimura, J. Xia, T. Mochizuki and M. Chabera, Per- formances of the Hyper-Kamiokande 20-inch PMT,Proceedingsof5thInt. Workshop New Photon-Detectors (PD18), JPS Conf. Proc. 27, 011013, 2019, [DOI: 10.7566/JPSCP.27.011013] [77] Super-Kamiokande Collaboration, Solar neutrino measurements in Super-Kamiokande-I,Phys.Rev.D73:112001,2006,[DOI: 10.1103/PhysRevD.73.112001]. BIBLIOGRAPHY 234 [78] J. F. Beacom, M. R. Vagins, GADZOOKS! Antineutrino Spectroscopy with Large Water Cerenkov Detectors,inProceedings,30thPhys.Rev.Lett.93 (2004) 171101, [DOI:10.1103/PhysRevLett.93.171101]. [79] Ll. Marti, et al., Evaluation of Gadolinium’s Action on Water Cherenkov Detector Systems with EGADS, Nucl.Instrum.Meth.A 959 (2020) 163549, [e-Print:1908.11532]. [80] T. Tanimori, et al., Design and performance of semi-custom analog IC in- cluding two TACs and two current integrators for ’Super-Kamiokande’,IEEE Trans.Nucl.Sci. 36 (1989) 497-501, [DOI:10.1109/23.34489]. [81] H.Nishino, et al., High-speed charge-to-time converter ASIC for the Super-Kamiokande detector, Nucl.Instrum.Meth.A 610 (2009) 710-717, [DOI:10.1016/j.nima.2009.09.026]. [82] S. Yamada, Y. Hayato; Y. Obayashi; M. Shiozawa, New online system with- out hardware trigger for the Super-Kamiokande experiment,Contributionto: NSS/MIC 2007, 111-114, [DOI:10.1109/NSSMIC.2007.4436298]. [83] G. Carminati for the Super-Kamiokande Collaboration The new Wide-band Solar Neutrino Trigger for Super-Kamiokande,Phys.Procedia61(2015)666- 672, [DOI:10.1016/j.phpro.2014.12.068]. BIBLIOGRAPHY 235 [84] M. Smy, for the Super-Kamiokande Collaboration, Low Energy Event Reconstruction and Selection in Super–Kamiokande–III,in Proceedings, 30th International Cosmic Ray Conference (ICRC 2007). [http://indico.nucleares.unam.mx/contributionDisplay. py?contribId=213confId=4]. [85] K. Abe, et al. (Super-Kamiokande Collaboration), Solar neutrino results in Super-Kamiokande-III,Phys.Rev.D83:052010,2011, [DOI:10.1103/PhysRevD.83.052010]. [86] Z. Haibing Study of Low Energy Electron Anti-neutrinos at Super- Kamiokande IV,2012,PhDthesis. [87] J. A. Dunmore, The Separation of CC and NC Events in the Sudbury Neutrino Observatory,2004,PhDthesis. [88] M. Nakahata, et al., Calibration of Super-Kamiokande using an electron LINAC: The Super-Kamiokande Collaboration, Nucl.Instrum.Meth.A 421 (1999) 113-129, [DOI:10.1016/S0168-9002(98)01200-5]. [89] The Super-Kamiokande Collaboration, 16N as a calibration source for Super-Kamiokande, Nucl.Instrum.Meth.A 458 (2001) 638-649, [DOI:10.1016/S0168-9002(00)00900-1]. BIBLIOGRAPHY 236 [90] K. Abe, et al., Calibration of the Super-Kamiokande detector, Nucl.Instrum.Meth.A 737 (2014) 253-272, [DOI:10.1016/j.nima.2013.11.081]. [91] H. Watanabe, H. Zhang, et al, (Super-Kamiokande Collabora- tion), First Study of Neutron Tagging with a Water Cherenkov Detector, Astroparticle Physics, Volume 31, Issue 4, 2009, [DOI:10.1016/j.astropartphys.2009.03.002]. [92] L. Marti-Magro, On behalf of the Super-Kamiokande Collaboration, Status of the SK-Gd project, 40th International Conference on High Energy physics (ICHEP2020), [DOI: 10.22323/1.390.0164] [93] T. Mori, Development of A Gadolinium-doped Water Cherenkov Detector for The Observation of Supernova Relic Neutrinos,Thesis,2015. [94] Institute for Cosmic Ray Research, The University of Tokyo, Annual Report, April 2018 - March 2019,Specialeditionextractingtheactivityoftheneu- trino and astroparticle division. [95] http://commons.wikimedia.org/wiki/User:ThaLibster. CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0). [96] W. Wolszczak, P. Dorenbos, Shape of intrinsic alpha pulse height spec- tra in lanthanide halide scintillators, Nuclear Instruments and Methods in BIBLIOGRAPHY 237 Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment Volume 857, 11 June 2017, Pages 66-74, [DOI: 10.1016/j.nima.2017.02.041]. [97] S.Agostinelli, et al. Geant4 — a simulation toolkit, Nucl.Instrum.Meth.A 506 (2003) 250-303, [DOI:10.1016/S0168-9002(03)01368-8]. [98] K. Zbiri, Additional gadolinium support for GLG4sim, https://web.archive.org/web/20060907034310/neutrino.phys. ksu.edu/ GLG4sim/Gd.html, 2006. Archived 2006-09-07, original un- available. [99] I. Ou, T. Yano, Y. Yamada, T. Mori, T. Kayano, M. Sakuda, A. Kimura, H. Harada, Measurement of the Energy, Multiplicity and Angular Correlation of -rays from the Thermal Neutron Capture Reaction Gd(n, ), JPS Conf.Proc. 1(2014)013053,[DOI: 10.7566/JPSCP.1.013053]. [100] C. Simpson, et al. (Super-Kamiokande Collaboration), Sensitiv- ity of Super-Kamiokande with Gadolinium to Low Energy Anti- neutrinos from Pre-supernova Emission, Astrophys.J. 885 (2019) 133, [DOI:10.3847/1538-4357/ab4883]. [101] Holden, N.E., Zucker, M.S., Reevaluation of the average prompt neutron emission multiplicity (nubar) values from fission of uranium and transura- nium nuclides, (BNL-NCS–35513). BIBLIOGRAPHY 238 [102] W. B. Wilson, R. T. Perry, E. F. Shores, et al., SOURCES 4C: a code for calculating (↵,n), spontaneous fission, and delayed neutron sources and spectra, LA-UR–02-1839 (Jan, 2002). [103] A. Barna, S. Dye, Global Antineutrino Modeling: A Web Application, [e-Print:1510.05633]. [104] IAEA, Power Reactor Information System (PRIS), [https://pris.iaea.org/PRIS/About.aspx]. [105] S. W. Li, J. F. Beacom, Spallation Backgrounds in Super-Kamiokande Are Made in Muon-Induced Showers,Phys.Rev.D91(2015)10,105005, [e-Print:1503.04823] [106] Y. Zhang, et al. (Super-Kamiokande Collaboration), First measure- ment of radioactive isotope production through cosmic-ray muon spalla- tion in Super-Kamiokande IV,Phys.Rev.D93,012004,2016,[DOI: 10.1103/PhysRevD.93.012004]. [107] S. W. Li, J. F. Beacom, Tagging spallation backgrounds with showers in water Cherenkov detectors,2015,Phys.Rev.D92,105033,[DOI: 10.1103/PhysRevD.92.105033] [108] A. Hoecker, et al., TMVA - Toolkit for Multivariate Data Analysis, (CERN- OPEN-2007-007). BIBLIOGRAPHY 239 [109] A. Lykov, S.Muzychka, K. Vaninsky, The AdaBoost Flow, [e-Print:1110.6228 ] [110] G. Biau , B. Cadre , Optimization by gradient boosting, [e-Print:1707.05023] [111] T. Duan, A. Avati, D. Y. Ding, K. K. Thai, S. Basu, A. Y. Ng, A. Schuler, NGBoost: Natural Gradient Boosting for Probabilistic Prediction, 2015, [e-Print:1910.03225] [112] L. Lista, Statistical Methods for Data Analysis in Particle Physics, Lect.Notes Phys. 941 (2017) 1-257, [DOI:10.1007/978-3-319-62840-0] [113] S. Kim, S. Lee, W. Lee, Y. Yum, J. Kim, S. Sohn, J. Park, J. Lee, J. Lim , S. Kwon, Stou↵er’s Test in a Large Scale Simultaneous Hypothesis Testing, [DOI: 10.1371/journal.pone.0063290]. [114] M. Dolan, G. Mathews, D. Lam, N. Lan, G. Herczeg, D. Dearborn, Evolu- tionary tracks for Betelgeuse, The Astrophysical Journal, Volume 819, Num- ber 1, 2016, [DOI: 10.3847/0004-637X/819/1/7]. [115] G. M. Harper, A. Brown, E. F. Guinan, E. O’Gorman, A. M. S. Richards, P. Kervella, L. Decin, An Updated 2017 Astrometric Solution for Betel- geuse, The Astronomical Journal, Volume 154, Number 1, 2017, [DOI: 10.3847/1538-3881/aa6ff9]. BIBLIOGRAPHY 240 [116] S. M. Adams, C. S. Kochanek, J. F. Beacom, M. R. Vagins, K. Z. Stanek Observing the Next Galactic Supernova, Astrophys.J. 778 (2013) 164, [DOI:10.1088/0004-637X/778/2/164] [117] Hyper-Kamiokande Proto-Collaboration, Hyper-Kamiokande Design Report, [physics.ins-det/1805.04163]. [118] Y. Nishimura, on behalf of the Hyper-Kamiokande Proto-collaboration, New 50-cm Photo-Detectors for Hyper-Kamiokande in proceedings of International Conference on High Energy Physics, PoS(ICHEP2016) (2017). [119] Adrian-Martinez, et al. (KM3NeT Collaboration), Deep sea tests of a pro- totype of the KM3NeT digital optical module, Eur. Phys. J. C74 (9) (2014) 3056 [astro-ph/1405.0839]. [120] B. Quilain, Multi-PMT modules for Hyper-Kamiokande, Proceedings in preparation, [https://www-conf.kek.jp/PD18]. [121] Gianfranca De Rosa, on behalf of Hyper-Kamiokande Proto-Collaboration, A multi-PMT photodetector system for the Hyper-Kamiokande experiment, NIMA PROCEEDINGS-D-19-00151, [DOI:10.1016/j.nima.2019.163033]. [122] B. Quilain, M. Hartz, Y. Nishimura, M. Kuze, H. Morikawa, M. Inomoto, M. Ishitsuka, N. Izumi, Multi-PMT Modules for Hyper-Kamiokande, JPS Conf.Proc. 27 (2019) 011017, [DOI: 10.7566/JPSCP.27.011017]. BIBLIOGRAPHY 241 [123] C. Andreopoulos, et al. The GENIE Neutrino Monte Carlo Generator, Nucl.Instrum.Meth.A614:87-104, 2010, [DOI:10.1016/j.nima.2009.12.009] [124] Y. Hayato, Neutrino interactions: From theory to Monte Carlo simulations, Proceedings, 45th Karpacz Winter School in Theoretical Physics, Acta Phys. Polon. B40, 2477 (2009). [125] S. Tobayama, An analysis of the oscillation of atmospheric neutrinos,PhD thesis, University of British Columbia (2016). [126] Web, https://github.com/WCSim/WCSim [127] Web, https://github.com/bquilain/LEAF [128] Hyper-Kamiokande Proto-Collaboration, Hyper-Kamiokande Design Report, [ePrint: e-Print:1805.04163].