Littérature scientifique sur le sujet « Data analysis, neutrinoless double beta decay »

The experimental evidence of neutrino oscillations clearly showed that the neutrino is a finite-mass particle. Anyway, two big questions concerning the neutrino are still unsolved: its nature (Dirac or Majorana) and the absolute value of its mass. The Double Beta Decay without emission of neutrinos (0nDBD) is at present the most sensitive method to answer the two questions. Bolometers, together with germanium diodes, have provided so far the best results within this kind of research. The choice of the so-called calorimetric approach, where the detector is made of the 0nDBD material, allows the study of large quantities of isotope with an excellent energy resolution (around 0.2-0.3 %). Both features are crucial to disentangle the searched peak from background. The Cuoricino experiment, in which 62 detectors of tellurium dioxide (TeO2) were involved, was carried out in hall A at Laboratori Nazionali del Gran Sasso (LNGS), and represented the bolometric experiment with the best sensitivity concerning the study on the 0nDBD decay. The knowledge achieved in the bolometric technique and its excellent results on 0nDBD led to the development of new experiments to study rare events with bolometric technique. The CUORE (Cryogenic Underground Observatory for Rare Events) experiment, composed by 988 bolometric detectors of TeO2, is under construction and foresees a sensitivity on neutrino mass of the order of about 50 meV. This high sensitivity requires excellent energetic resolutions, low number of spurious counts within the region of interest and high quantity of the 0nDBD active isotope. The reduction of the background in the energetic window where the 0nDBD for the isotope 130Te is expected (Q_bb = 2527 keV) plays a primary role. In this context, the activities I have done are focused primarily on the analysis of the different contributions to the background of CUORE and the optimization of methods for its reduction. This work has been done both through measurements performed in the hall C of LNGS and with Monte Carlo simulations that allowed data interpretation and assessment of the resulting sensitivity. For example, through a series of simulations performed with Geant4 code it is possible to extrapolate the background induced by muons in CUORE. The results of simulations were then validated through the installation of a muon veto above Cuoricino which allowed a direct comparison with experimental measurement. For what concerns the reduction of the bolometers specific background I have also done some measurements for the development, characterization and optimization of some scintillating bolometers. These detectors, thanks to double read-out of thermal and scintillation signals, allow to discriminate different ionizing particles (beta/gamma, alpha and neutrons). This allows a significant reduction of the unwanted background in the region of interest and would therefore allow to approach the condition of `zero background' experiments. Indeed, thanks to the wide choice for the absorber material that allows to study practically all 0nDBD candidate isotopes, it is possible to choose an isotope with a transition energy above 2615 keV and then remove, in practice, all the background due to gamma rays. The 2615 keV gamma line corresponds in fact to the highest energy gamma-ray line from natural radioactivity and is due to 208Tl. Above this energy there are only extremely rare high energy gamma's. Once gamma-rays are no more a worrisome source of background, what is left on the side of radioactivity are alpha emissions. Indeed alpha surface contaminations are already recognized as the most relevant background source in the bolometric experiment Cuoricino. However, thanks to the double read-out of scintillating bolometers, this source of background can be removed too. During my PhD I have tested a number of different crystals and with some of them (CdWO4, ZnSe and some molybdates) I have obtained excellent results both from the point of view of the detector performance and the reduction of the background contributions. Thanks to these results it was possible to evaluate the discrimination power (i.e. the capability to recognize and reject unwanted events) of this technique and extrapolate the sensitivity of a large mass experiment for 0nDBD based on the hybrid heat plus scintillation technique. Moreover, during the analysis of the data collected with the scintillating bolometers, I recognized for the first time, a dependence of the pulse shapes (both on the scintillation and heat channels) on the interacting particle nature. I proposed then to exploit such wonderful feature to discriminate the interacting particles without relying on the much more complicated measurement of both (light and heat) signals. This feature is very promising because it allows to greatly reduce the background in bolometers without getting complicated the assembly of the experiment. In fact in the case of the double read-out of temperature and scintillation both light detectors and reflecting sheet (used to properly collect the scintillation light) are needed in addition to the low temperature calorimeter. Finally, I have proposed a further use of the scintillating bolometer for diagnostic purposes, i.e. the possibility to study surface contaminations with high sensitivity. One of the main limitations in our understanding of the background due to surface contaminations is in fact related to the limited sensitivity of the available standard techniques. Traditionally the devices used in this field are Si surface barrier detectors with an active area of about 10 cm^2, a typical energy resolution of about 25-30 keV FWHM, and counting rates of the order 0.05 count/h/cm^2 between 3 and 8 MeV. A Cuoricino-like bolometer can easily reach a much larger active area (150 cm^2) and, thanks to the absence of a dead layer can reach resolution on surface alpha particles of 10 keV. Moreover a background counting rate in the 3-8 MeV region as low as 0.001 count/h/cm^2 was already reached with this technique. This considerations allow to plan measurements with sensitivities order of magnitude better than standard devices. However, in order to use scintillating bolometer to study surface contaminations, they have not to be surrounded by a reflecting sheet. For these reason it is necessary to use crystals with a very high light yield or crystals that are able to recognize particle from the shape of the thermal pulses (i.e. without any need of collecting the scintillation light). Since this last feature is a very recent discovery and some works have still to be performed before the technique can be considered actually at hand, I have proposed to use a BGO crystal (Bi4Ge3O12), which is characterized by a very high light yield, to study surface contaminations. Preliminary tests with an array of 4 `small' crystals (2x2x2 cm^3) have shown how this possibility could be fulfilled. However this measurement showed a very slow cooling down and an high counting rate due to 207Bi. Therefore, before using the crystal to study surface contamination, a new measurement was carried out with a larger crystal (5x5x5 cm^3). The measurement was successful and showed that the slow cool down and the high counting rate are not intrinsic problems of these crystals. This first test gave excellent results on surface studies and, as supplementary results, a measurement of the rare alpha decay of 209Bi with a high statistical significance was performed.

L'expérience NEMO3 était dédiée à la recherche de la désintégration ββ0ν à l'aide de diverses sources d'isotopes de désintégration double bêta (principalement ¹ººMo, ⁸²Se, ¹¹⁶Cd et ¹³ºTe pour un total d'environ 10 kg). Le détecteur était localisé dans le Laboratoire souterrain de Modane, à mi-parcours du tunnel du Fréjus. Cette expérience a permis de démontrer que la technologie "tracko-calo" est très compétitive et a de plus offert de nouveaux résultats pour la recherche des désintégrations ββ2ν et ββ0ν. Par ailleurs, elle a ouvert la voie pour son successeur SuperNEMO, dont le but est d'atteindre 100 kg de ⁸²Se (pour une sensibilité de 10²⁶ années). Le but principal de cette thèse a été de mesurer le temps de demi-vie des désintégrations ββ2ν et ββ0ν du ¹ººMo vers les états excités 0₁⁺ du ¹ººRu à l'aide des données totales de NEMO3, avec de nouvelles méthodes d'analyse et un développement du programme d'analyse de la collaboration. Les résultats obtenus pour la désintégration ββ2ν du ¹ººMo vers l'état fondamental (gs) et excité (0₁⁺) du ¹ººRu sont T1/2(ββ2ν,gs)=(7,05±0,01(stat)±0,54(syst)).10¹⁸ ans et T1/2(ββ2ν,0₁⁺)=(6,15±1,1(stat)±0,78)).10²º ans. Ces résultats sont compatibles avec les résultats publiés par la collaboration. Quant à la désintégration ββ0ν(0₁⁺), ce travail permet d’obtenir un temps de demi-vie de T1/2(ββ0ν, 0₁⁺)&gt;2,6.10²³ ans, améliorant significativement les derniers résultats publiés. De plus ces méthodes ont aussi permis de présenter un nouveau modèle de bruit de fond de l'expérience, plus exhaustif. Le second but de ce travail a été de mesurer les erreurs systématiques du calorimètre de NEMO3 dues, entre autres, à la longueur d'onde des systèmes d’étalonnage du détecteur. Ce travail a été réalisé notamment à l'aide d'un banc de test basé sur des DEL. Ce banc a aussi permis de contribuer au développement du calorimètre de SuperNEMO, particulièrement au travers de mesures de linéarité et de caractéristiques temporelles des PM destinés au démonstrateur de l'expérience<br>The NEMO3 experiment was researching the ββ0ν decay by using various sources of double beta decay isotopes (mainly ¹ººMo, ⁸²Se, ¹¹⁶Cd and ¹³⁰Te for about 10 kg in total). The detector was located in the “Laboratoire Souterrain de Modane”, in the halfway point of the Frejus tunnel. This experiment demonstrated that the "tracko-calo" technology is really competitive and, in addition, it gives new results for the ββ2ν and the ββ0ν decay research. Moreover it opened a new way for its successor SuperNEMO, which aim is to reach a mass of 100 kg of ⁸²Se (for a sensitivity of 10²⁶ years). The main goal of the thesis is to measure the ββ2ν and ββ0ν decay of the ¹ººMo to the excited state 0₁⁺ of the ¹ººRu thanks to the whole NEMO3 data, with new original methods of analysis and through the development of the collaboration analysis software. The results obtained for the ground states (gs) and excited states ββ2ν of the ¹ººMo are T1/2(ββ2ν,gs)=(7,05±0,01(stat)±0,54(syst)).10¹⁸ years and T1/2(ββ2ν, 0₁⁺)=(6,15±1,1(stat)±0,78)).10²º years. Those results are compatibles with the last published ones by the collaboration. For the ββ0ν(0₁⁺), this work gave a half-life time of T1/2 (ββ0ν, 0₁⁺)&gt;2,6.10²³ years, improving significantly the last published results. Furthermore those methods also allowed to present a new and more exhaustive background noise model for this experiment. The second point of this work was to measure the systematics errors of the NEMO3 calorimeter, among others, due to the wavelength of the NEMO3 calibration systems. This work was done using a new test bench based on LED. This bench also allowed to contribute to the development of the SuperNEMO calorimeter, especially in the time characteristic and the energy linearity measurement of the PMT intended to the demonstrator of the experiments

My PhD thesis project has been carried out within the CUORE (Cryogenic Underground Observatory for Rare Events) experiment. The CUORE experiment is a ton scale cryogenic detector installed in the underground Gran Sasso laboratories of INFN (LNGS) in Italy. The CUORE detector consists of an array of 988 TeO2 bolometric detectors (5 x 5 x 5 cm3 each) arranged into 19 identical structures called 'towers'. Each tower hosts 52 bolometers arranged in 13 floors, each containing 4 crystals. Each crystal has a mass of 750 g, resulting in a total mass of 742 kg of TeO2, or 206 kg of 130Te. The goal of the CUORE experiment is to investigate the 0nuBB decay of 130Te. My main activities have been related to the CUORE detector operation, characterization and optimization. Moreover I developed software tools for the CUORE data processing and analysis. In addition, I have been investigating the CUORE sensitivity to rare processes other than the 0nuBB decay, like the CPT symmetry violation in the 2nuBB decay. In the first part of my PhD thesis, I give a detailed review of the CUORE detector optimization campaigns performed in 2017. It was the first time that such a large number of bolometric detectors were simultaneously operated in a completely new and unique cryogenic system. My main contribution has been related to the analysis of the detectors performance in terms of signal optimization and noise reduction, in order to set both the optimal operating temperature of the cryogenic system and the polarization voltages (working points) for the NTDs coupled with each bolometer. In order to set the best operating conditions of the CUORE detectors, it has been necessary to characterize the response of the CUORE bolometers and NTDs. Utilizing the high number of detectors in CUORE, it was possible to study and model more features of the bolometers response and to try to develop a complete (static and dynamic) thermal model. The development of a thermal model can contribute to the identification of the physical parameters which are affecting the CUORE bolometers energy resolution and which could be possibly better optimized. The second part of the work is focused on the analysis of the CUORE data and on the study of the potential of the CUORE experiment for the search for rare events and/or for physics beyond the Standard Model other than the 0nuBB decay of 130Te. The first CUORE physics data were acquired during the summer of 2017, with a total collected 0nuBB exposure of 86.3 kg.yr of (nat)TeO2. The analysis of these data lead to the first CUORE 0nuBB half-life limit and to a more precise measurement of the 2nuBB decay half-life for 130Te. I have contributed to the analysis by developing and debugging the software for processing and analyzing the CUORE data; moreover I have worked to the evaluation of the selection efficiencies which have been used in the production of the CUORE physics results. Furthermore I studied the sensitivity of the CUORE experiment to the possible CPT symmetry violating terms. Indeed, the violation of this symmetry is described in several Standard Model Extensions (SME); in the case of isotopes which can undergo double-beta decay, the CPT violation would induce a deformation in the spectrum of the total energy of the two electrons emitted in the 2nuBB process. Utilizing the data acquired by CUORE in 2017, it was possible to set the first limit for the CPT violation evaluated from the 130Te isotope.

Neutrinoless Double Beta Decay(0𝜈𝛽𝛽) is one of the major research interests in neutrino physics. The discovery of 0𝜈𝛽𝛽 would answer persistent puzzles in the Standard Model of Elementary Particles. KamLAND-Zen is one of the leading efforts in the search of 0𝛽𝛽 and has acquired data from 745 kg of ^{136}Xe over 224 live-days. This data is analyzed using a Bayesian approach consisting of a Markov Chain Monte Carlo (MCMC) algorithm. The implementation of the Bayesian analysis, which is the focal point of this dissertation, yields a 90\% Credible Interval at T^{0𝜈}_{1/2} = 7.03 × 10^{25} years. Finally, a machine learning event classification algorithm, based on a spherical convolutional neural network (spherical CNN) was developed to increase the T^{0𝜈}_{1/2} sensitivity. The classification power of this algorithm was demonstrated on a Monte Carlo detector simulation, and a data driven classifier was trained to reject crucial backgrounds in the 0𝜈𝛽𝛽 analysis. After implementing the spherical CNN, an increase in T^{0𝜈}_{1/2} sensitivity of 11.0% is predicted. These early studies pave the way for substantial improvements in future 0𝜈𝛽𝛽 analyses.

Tese no âmbito do Doutoramento em Física, Astrofísica, apresentada ao Departamento de Física da Faculdade de Ciências e Tecnologia da Universidade de Coimbra.<br>An elusive form of matter that does not interact via electromagnetic or strong forces permeates the known Universe, and is therefore designated as "dark". This dark matter (DM) is responsible for the evolution of cosmic structures, the cohesion of galaxies and galaxy clusters, and represents around a quarter of the total content of the Universe. Several state-of-the-art experiments are currently searching for dark matter in the form of weakly interacting massive particles (WIMPs), using ultra-low background "observatories" where one of these particles could interact with a material target and produce a readable signature. The LZ experiment is a 10 tonne dark matter detector expected to begin operations in early 2021, that aims to surpass the current world-leading limit on the WIMP-nucleon interaction cross section by more than one order of magnitude. The design of LZ features a dual-phase xenon time projection chamber (TPC) and two additional instrumented veto detectors encompassing the TPC for improved background reduction and active shielding. The projected sensitivity of LZ to the spin-independent WIMP-nucleon scattering cross-section is 1.4×10^{-48} cm^2 for a 40 GeV/c^2 mass WIMP. LZ has the potential to study and discover a wide range of new physics. The inner portions of the TPC of LZ will be one of the most "quiet" environments where rare event searches can be performed. The ultra-low background required for dark matter searches allows LZ to be potentially sensitive to other rare events such as neutrinoless double beta decay of some xenon isotopes, axion interactions or coherent neutrino-nucleus scattering, all of which were not yet observed. The projected sensitivity of LZ to the half-life of the neutrinoless double beta decay of 136Xe is presented in this document. For an exposure of 1360 kg·year, a sensitivity to the half-life of 1.06×10^{26} years with a 90% confidence level is obtained. The projected sensitivity to this same decay from a dedicated run with a 90% 136Xe enriched target and an exposure of 13.8 tonne·year is 1.06×10^{27} years. The development of pulse classification tools for the data processing framework of LZ (LZap) is also presented in this document. These tools represent the groundwork for pulse classification in LZ, both in the form of dedicated heuristics algorithms and machine learning implementations. The Heuristics Algorithm for Discrimination of Event Substructures (HADES) developed in the context of this work is currently the default pulse classification tool in LZap, and provides a measured overall classification accuracy of 98.58% across all pulse topologies in LZ simulated data. The RFClassifier and the TriNet pulse classification tools are two machine learning implementations that use a random forest model and an ensemble of artificial neural networks, respectively, that are aimed at assisting HADES and potentially replacing it in LZap. The RFClassifier algorithm achieved a classification accuracy of 99.37% over LZ simulated data when combined with a powerful clustering analysis using Gaussian mixture models (GMMs). The TriNet algorithm was trained using the results from HADES and achieved a classification accuracy of 95.56% against the GMM clustering results, but demonstrated that it could generalize its results beyond HADES.<br>Uma forma de matéria que não interage através das forças eletromagnética e forte, por isso designada por matéria "escura", permeia o universo visível. Esta matéria escura (ME) representa cerca de um quarto do conteúdo total do universo e é responsável pela evolução das estruturas cósmicas e pela coesão das galáxias e dos aglomerados de galáxias. Várias experiências de ponta procuram pela matéria escura na forma de WIMPs (weakly interacting massive particles na sigla inglesa), usando "observatórios" com fundos radiogénicos e cosmogénicos extremamente reduzidos onde uma destas partículas pode interagir com um material alvo e produzir um sinal mensurável. A experiência LZ é um detetor de matéria escura com 10 toneladas que deverá iniciar operações no início de 2021 e cujo principal objetivo é melhorar o atual limite de exclusão da secção eficaz de interação WIMP-nucleão por mais de uma ordem de grandeza. LZ é composto por uma câmara de projeção temporal (TPC) de duas fases de xénon e por dois detetores adicionais que envolvem a TPC e são usados como vetos, a fim de reduzirem ativamente sinais indesejados (fundos) da experiência. A sensibilidade estimada de LZ à secção eficaz da interação WIMP-nucleão independente de spin é de 1.4×10^{-48} cm^2 para uma WIMP de 40 GeV/c^2 de massa. Para além da matéria escura, LZ tem o potencial de estudar, e talvez descobrir, uma grande variedade de novos processos físicos raros. A região mais interna da TPC de LZ será um dos ambientes mais "calmos" onde o estudo destes processos raros é possível. Os fundos extremamente baixos de LZ permitem-lhe ter uma boa sensibilidade a processos raros nunca observados como o decaimento beta duplo sem emissão de neutrinos de alguns isótopos de xénon, interações de axiões ou dispersão elástica coerente neutrino-núcleo. A sensibilidade de LZ à meia-vida do decaimento beta duplo sem emissão de neutrinos do 136Xe é apresentada neste documento. Para uma exposição de 1360 kg·ano, a sensibilidade estimada é de 1.06×10^{26} anos com um intervalo de confiança de 90%. A sensibilidade estimada para um run dedicado subsequente, com enriquecimento isotópico de 90% de 136Xe e exposição de 13.8 toneladas·ano é de 1.06×10^{27} anos. O desenvolvimento de ferramentas de classificação de sinais para a cadeia de processamento de dados de LZ (LZap) é também apresentado neste documento. Estas ferramentas representam a base para classificação de sinais em LZ, tanto na forma de algoritmos heurísticos dedicados como implementações de Machine Learning. O HADES (Heuristics Algorithm for Discrimination of Event Substructures na sigla inglesa), desenvolvido no contexto deste trabalho, é atualmente a principal ferramenta de classificação de sinais em LZap e consegue uma exatidão global de 98.58% para todas as topologias de sinais presentes nos dados de simulação de LZ. As ferramentas de classificação RFClassifier e TriNet são duas implementações de Machine Learning que usam, respectivamente, um modelo de random forests e um ensemble de redes neuronais para auxiliar o desenvolvimento do HADES e potencialmente substitui-lo na cadeia de LZap. O algoritmo RFClassifier consegue uma exatidão de classificação de 99.37% sobre os dados simulados de LZ quando combinado com Gaussian mixture models (GMMs), uma técnica de clustering poderosa. O algoritmo TriNet foi treinado usando os resultados obtidos pelo HADES e consegue uma exatidão de classificação de 95.56% comparando com resultados do clustering com GMM, mas demonstrou que consegue generalizar os seus resultados para além do HADES.